]> code.delx.au - gnu-emacs/blob - lisp/progmodes/cc-engine.el
Amend CC Mode to handle big C++ raw strings correctly.
[gnu-emacs] / lisp / progmodes / cc-engine.el
1 ;;; cc-engine.el --- core syntax guessing engine for CC mode -*- coding: utf-8 -*-
2
3 ;; Copyright (C) 1985, 1987, 1992-2016 Free Software Foundation, Inc.
4
5 ;; Authors: 2001- Alan Mackenzie
6 ;; 1998- Martin Stjernholm
7 ;; 1992-1999 Barry A. Warsaw
8 ;; 1987 Dave Detlefs
9 ;; 1987 Stewart Clamen
10 ;; 1985 Richard M. Stallman
11 ;; Maintainer: bug-cc-mode@gnu.org
12 ;; Created: 22-Apr-1997 (split from cc-mode.el)
13 ;; Keywords: c languages
14 ;; Package: cc-mode
15
16 ;; This file is part of GNU Emacs.
17
18 ;; GNU Emacs is free software: you can redistribute it and/or modify
19 ;; it under the terms of the GNU General Public License as published by
20 ;; the Free Software Foundation, either version 3 of the License, or
21 ;; (at your option) any later version.
22
23 ;; GNU Emacs is distributed in the hope that it will be useful,
24 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
25 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26 ;; GNU General Public License for more details.
27
28 ;; You should have received a copy of the GNU General Public License
29 ;; along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>.
30
31 ;;; Commentary:
32
33 ;; The functions which have docstring documentation can be considered
34 ;; part of an API which other packages can use in CC Mode buffers.
35 ;; Otoh, undocumented functions and functions with the documentation
36 ;; in comments are considered purely internal and can change semantics
37 ;; or even disappear in the future.
38 ;;
39 ;; (This policy applies to CC Mode as a whole, not just this file. It
40 ;; probably also applies to many other Emacs packages, but here it's
41 ;; clearly spelled out.)
42
43 ;; Hidden buffer changes
44 ;;
45 ;; Various functions in CC Mode use text properties for caching and
46 ;; syntactic markup purposes, and those of them that might modify such
47 ;; properties but still don't modify the buffer in a visible way are
48 ;; said to do "hidden buffer changes". They should be used within
49 ;; `c-save-buffer-state' or a similar function that saves and restores
50 ;; buffer modifiedness, disables buffer change hooks, etc.
51 ;;
52 ;; Interactive functions are assumed to not do hidden buffer changes,
53 ;; except in the specific parts of them that do real changes.
54 ;;
55 ;; Lineup functions are assumed to do hidden buffer changes. They
56 ;; must not do real changes, though.
57 ;;
58 ;; All other functions that do hidden buffer changes have that noted
59 ;; in their doc string or comment.
60 ;;
61 ;; The intention with this system is to avoid wrapping every leaf
62 ;; function that do hidden buffer changes inside
63 ;; `c-save-buffer-state'. It should be used as near the top of the
64 ;; interactive functions as possible.
65 ;;
66 ;; Functions called during font locking are allowed to do hidden
67 ;; buffer changes since the font-lock package run them in a context
68 ;; similar to `c-save-buffer-state' (in fact, that function is heavily
69 ;; inspired by `save-buffer-state' in the font-lock package).
70
71 ;; Use of text properties
72 ;;
73 ;; CC Mode uses several text properties internally to mark up various
74 ;; positions, e.g. to improve speed and to eliminate glitches in
75 ;; interactive refontification.
76 ;;
77 ;; Note: This doc is for internal use only. Other packages should not
78 ;; assume that these text properties are used as described here.
79 ;;
80 ;; 'category
81 ;; Used for "indirection". With its help, some other property can
82 ;; be cheaply and easily switched on or off everywhere it occurs.
83 ;;
84 ;; 'syntax-table
85 ;; Used to modify the syntax of some characters. It is used to
86 ;; mark the "<" and ">" of angle bracket parens with paren syntax, to
87 ;; "hide" obtrusive characters in preprocessor lines, and to mark C++
88 ;; raw strings to enable their fontification.
89 ;;
90 ;; This property is used on single characters and is therefore
91 ;; always treated as front and rear nonsticky (or start and end open
92 ;; in XEmacs vocabulary). It's therefore installed on
93 ;; `text-property-default-nonsticky' if that variable exists (Emacs
94 ;; >= 21).
95 ;;
96 ;; 'c-is-sws and 'c-in-sws
97 ;; Used by `c-forward-syntactic-ws' and `c-backward-syntactic-ws' to
98 ;; speed them up. See the comment blurb before `c-put-is-sws'
99 ;; below for further details.
100 ;;
101 ;; 'c-type
102 ;; This property is used on single characters to mark positions with
103 ;; special syntactic relevance of various sorts. Its primary use is
104 ;; to avoid glitches when multiline constructs are refontified
105 ;; interactively (on font lock decoration level 3). It's cleared in
106 ;; a region before it's fontified and is then put on relevant chars
107 ;; in that region as they are encountered during the fontification.
108 ;; The value specifies the kind of position:
109 ;;
110 ;; 'c-decl-arg-start
111 ;; Put on the last char of the token preceding each declaration
112 ;; inside a declaration style arglist (typically in a function
113 ;; prototype).
114 ;;
115 ;; 'c-decl-end
116 ;; Put on the last char of the token preceding a declaration.
117 ;; This is used in cases where declaration boundaries can't be
118 ;; recognized simply by looking for a token like ";" or "}".
119 ;; `c-type-decl-end-used' must be set if this is used (see also
120 ;; `c-find-decl-spots').
121 ;;
122 ;; 'c-<>-arg-sep
123 ;; Put on the commas that separate arguments in angle bracket
124 ;; arglists like C++ template arglists.
125 ;;
126 ;; 'c-decl-id-start and 'c-decl-type-start
127 ;; Put on the last char of the token preceding each declarator
128 ;; in the declarator list of a declaration. They are also used
129 ;; between the identifiers cases like enum declarations.
130 ;; 'c-decl-type-start is used when the declarators are types,
131 ;; 'c-decl-id-start otherwise.
132 ;;
133 ;; 'c-awk-NL-prop
134 ;; Used in AWK mode to mark the various kinds of newlines. See
135 ;; cc-awk.el.
136
137 ;;; Code:
138
139 (eval-when-compile
140 (let ((load-path
141 (if (and (boundp 'byte-compile-dest-file)
142 (stringp byte-compile-dest-file))
143 (cons (file-name-directory byte-compile-dest-file) load-path)
144 load-path)))
145 (load "cc-bytecomp" nil t)))
146
147 (cc-require 'cc-defs)
148 (cc-require-when-compile 'cc-langs)
149 (cc-require 'cc-vars)
150
151 (eval-when-compile (require 'cl))
152
153 \f
154 ;; Make declarations for all the `c-lang-defvar' variables in cc-langs.
155
156 (defmacro c-declare-lang-variables ()
157 `(progn
158 ,@(c--mapcan (lambda (init)
159 `(,(if (elt init 2)
160 `(defvar ,(car init) nil ,(elt init 2))
161 `(defvar ,(car init) nil))
162 (make-variable-buffer-local ',(car init))))
163 (cdr c-lang-variable-inits))))
164 (c-declare-lang-variables)
165
166 \f
167 ;;; Internal state variables.
168
169 ;; Internal state of hungry delete key feature
170 (defvar c-hungry-delete-key nil)
171 (make-variable-buffer-local 'c-hungry-delete-key)
172
173 ;; The electric flag (toggled by `c-toggle-electric-state').
174 ;; If t, electric actions (like automatic reindentation, and (if
175 ;; c-auto-newline is also set) auto newlining) will happen when an electric
176 ;; key like `{' is pressed (or an electric keyword like `else').
177 (defvar c-electric-flag t)
178 (make-variable-buffer-local 'c-electric-flag)
179
180 ;; Internal state of auto newline feature.
181 (defvar c-auto-newline nil)
182 (make-variable-buffer-local 'c-auto-newline)
183
184 ;; Included in the mode line to indicate the active submodes.
185 ;; (defvar c-submode-indicators nil)
186 ;; (make-variable-buffer-local 'c-submode-indicators)
187
188 (defun c-calculate-state (arg prevstate)
189 ;; Calculate the new state of PREVSTATE, t or nil, based on arg. If
190 ;; arg is nil or zero, toggle the state. If arg is negative, turn
191 ;; the state off, and if arg is positive, turn the state on
192 (if (or (not arg)
193 (zerop (setq arg (prefix-numeric-value arg))))
194 (not prevstate)
195 (> arg 0)))
196
197 \f
198 ;; Basic handling of preprocessor directives.
199
200 ;; This is a dynamically bound cache used together with
201 ;; `c-query-macro-start' and `c-query-and-set-macro-start'. It only
202 ;; works as long as point doesn't cross a macro boundary.
203 (defvar c-macro-start 'unknown)
204
205 (defsubst c-query-and-set-macro-start ()
206 (if (symbolp c-macro-start)
207 (setq c-macro-start (save-excursion
208 (c-save-buffer-state ()
209 (and (c-beginning-of-macro)
210 (point)))))
211 c-macro-start))
212
213 (defsubst c-query-macro-start ()
214 (if (symbolp c-macro-start)
215 (save-excursion
216 (c-save-buffer-state ()
217 (and (c-beginning-of-macro)
218 (point))))
219 c-macro-start))
220
221 ;; One element macro cache to cope with continual movement within very large
222 ;; CPP macros.
223 (defvar c-macro-cache nil)
224 (make-variable-buffer-local 'c-macro-cache)
225 ;; Nil or cons of the bounds of the most recent CPP form probed by
226 ;; `c-beginning-of-macro', `c-end-of-macro' or `c-syntactic-end-of-macro'.
227 ;; The cdr will be nil if we know only the start of the CPP form.
228 (defvar c-macro-cache-start-pos nil)
229 (make-variable-buffer-local 'c-macro-cache-start-pos)
230 ;; The starting position from where we determined `c-macro-cache'.
231 (defvar c-macro-cache-syntactic nil)
232 (make-variable-buffer-local 'c-macro-cache-syntactic)
233 ;; Either nil, or the syntactic end of the macro currently represented by
234 ;; `c-macro-cache'.
235 (defvar c-macro-cache-no-comment nil)
236 (make-variable-buffer-local 'c-macro-cache-no-comment)
237 ;; Either nil, or the last character of the macro currently represented by
238 ;; `c-macro-cache' which isn't in a comment. */
239
240 (defun c-invalidate-macro-cache (beg end)
241 ;; Called from a before-change function. If the change region is before or
242 ;; in the macro characterized by `c-macro-cache' etc., nullify it
243 ;; appropriately. BEG and END are the standard before-change-functions
244 ;; parameters. END isn't used.
245 (cond
246 ((null c-macro-cache))
247 ((< beg (car c-macro-cache))
248 (setq c-macro-cache nil
249 c-macro-cache-start-pos nil
250 c-macro-cache-syntactic nil
251 c-macro-cache-no-comment nil))
252 ((and (cdr c-macro-cache)
253 (< beg (cdr c-macro-cache)))
254 (setcdr c-macro-cache nil)
255 (setq c-macro-cache-start-pos beg
256 c-macro-cache-syntactic nil
257 c-macro-cache-no-comment nil))))
258
259 (defun c-macro-is-genuine-p ()
260 ;; Check that the ostensible CPP construct at point is a real one. In
261 ;; particular, if point is on the first line of a narrowed buffer, make sure
262 ;; that the "#" isn't, say, the second character of a "##" operator. Return
263 ;; t when the macro is real, nil otherwise.
264 (let ((here (point)))
265 (beginning-of-line)
266 (prog1
267 (if (and (eq (point) (point-min))
268 (/= (point) 1))
269 (save-restriction
270 (widen)
271 (beginning-of-line)
272 (and (looking-at c-anchored-cpp-prefix)
273 (eq (match-beginning 1) here)))
274 t)
275 (goto-char here))))
276
277 (defun c-beginning-of-macro (&optional lim)
278 "Go to the beginning of a preprocessor directive.
279 Leave point at the beginning of the directive and return t if in one,
280 otherwise return nil and leave point unchanged.
281
282 Note that this function might do hidden buffer changes. See the
283 comment at the start of cc-engine.el for more info."
284 (let ((here (point)))
285 (when c-opt-cpp-prefix
286 (if (and (car c-macro-cache)
287 (>= (point) (car c-macro-cache))
288 (or (and (cdr c-macro-cache)
289 (<= (point) (cdr c-macro-cache)))
290 (<= (point) c-macro-cache-start-pos)))
291 (unless (< (car c-macro-cache) (or lim (point-min)))
292 (progn (goto-char (max (or lim (point-min)) (car c-macro-cache)))
293 (setq c-macro-cache-start-pos
294 (max c-macro-cache-start-pos here))
295 t))
296 (setq c-macro-cache nil
297 c-macro-cache-start-pos nil
298 c-macro-cache-syntactic nil
299 c-macro-cache-no-comment nil)
300
301 (save-restriction
302 (if lim (narrow-to-region lim (point-max)))
303 (beginning-of-line)
304 (while (eq (char-before (1- (point))) ?\\)
305 (forward-line -1))
306 (back-to-indentation)
307 (if (and (<= (point) here)
308 (save-match-data (looking-at c-opt-cpp-start))
309 (c-macro-is-genuine-p))
310 (progn
311 (setq c-macro-cache (cons (point) nil)
312 c-macro-cache-start-pos here)
313 t)
314 (goto-char here)
315 nil))))))
316
317 (defun c-end-of-macro ()
318 "Go to the end of a preprocessor directive.
319 More accurately, move the point to the end of the closest following
320 line that doesn't end with a line continuation backslash - no check is
321 done that the point is inside a cpp directive to begin with.
322
323 Note that this function might do hidden buffer changes. See the
324 comment at the start of cc-engine.el for more info."
325 (if (and (cdr c-macro-cache)
326 (<= (point) (cdr c-macro-cache))
327 (>= (point) (car c-macro-cache)))
328 (goto-char (cdr c-macro-cache))
329 (unless (and (car c-macro-cache)
330 (<= (point) c-macro-cache-start-pos)
331 (>= (point) (car c-macro-cache)))
332 (setq c-macro-cache nil
333 c-macro-cache-start-pos nil
334 c-macro-cache-syntactic nil
335 c-macro-cache-no-comment nil))
336 (while (progn
337 (end-of-line)
338 (when (and (eq (char-before) ?\\)
339 (not (eobp)))
340 (forward-char)
341 t)))
342 (when (car c-macro-cache)
343 (setcdr c-macro-cache (point)))))
344
345 (defun c-syntactic-end-of-macro ()
346 ;; Go to the end of a CPP directive, or a "safe" pos just before.
347 ;;
348 ;; This is normally the end of the next non-escaped line. A "safe"
349 ;; position is one not within a string or comment. (The EOL on a line
350 ;; comment is NOT "safe").
351 ;;
352 ;; This function must only be called from the beginning of a CPP construct.
353 ;;
354 ;; Note that this function might do hidden buffer changes. See the comment
355 ;; at the start of cc-engine.el for more info.
356 (let* ((here (point))
357 (there (progn (c-end-of-macro) (point)))
358 s)
359 (if c-macro-cache-syntactic
360 (goto-char c-macro-cache-syntactic)
361 (setq s (parse-partial-sexp here there))
362 (while (and (or (nth 3 s) ; in a string
363 (nth 4 s)) ; in a comment (maybe at end of line comment)
364 (> there here)) ; No infinite loops, please.
365 (setq there (1- (nth 8 s)))
366 (setq s (parse-partial-sexp here there)))
367 (setq c-macro-cache-syntactic (point)))
368 (point)))
369
370 (defun c-no-comment-end-of-macro ()
371 ;; Go to the end of a CPP directive, or a pos just before which isn't in a
372 ;; comment. For this purpose, open strings are ignored.
373 ;;
374 ;; This function must only be called from the beginning of a CPP construct.
375 ;;
376 ;; Note that this function might do hidden buffer changes. See the comment
377 ;; at the start of cc-engine.el for more info.
378 (let* ((here (point))
379 (there (progn (c-end-of-macro) (point)))
380 s)
381 (if c-macro-cache-no-comment
382 (goto-char c-macro-cache-no-comment)
383 (setq s (parse-partial-sexp here there))
384 (while (and (nth 3 s) ; in a string
385 (> there here)) ; No infinite loops, please.
386 (setq here (1+ (nth 8 s)))
387 (setq s (parse-partial-sexp here there)))
388 (when (nth 4 s)
389 (goto-char (1- (nth 8 s))))
390 (setq c-macro-cache-no-comment (point)))
391 (point)))
392
393 (defun c-forward-over-cpp-define-id ()
394 ;; Assuming point is at the "#" that introduces a preprocessor
395 ;; directive, it's moved forward to the end of the identifier which is
396 ;; "#define"d (or whatever c-opt-cpp-macro-define specifies). Non-nil
397 ;; is returned in this case, in all other cases nil is returned and
398 ;; point isn't moved.
399 ;;
400 ;; This function might do hidden buffer changes.
401 (when (and c-opt-cpp-macro-define-id
402 (looking-at c-opt-cpp-macro-define-id))
403 (goto-char (match-end 0))))
404
405 (defun c-forward-to-cpp-define-body ()
406 ;; Assuming point is at the "#" that introduces a preprocessor
407 ;; directive, it's moved forward to the start of the definition body
408 ;; if it's a "#define" (or whatever c-opt-cpp-macro-define
409 ;; specifies). Non-nil is returned in this case, in all other cases
410 ;; nil is returned and point isn't moved.
411 ;;
412 ;; This function might do hidden buffer changes.
413 (when (and c-opt-cpp-macro-define-start
414 (looking-at c-opt-cpp-macro-define-start)
415 (not (= (match-end 0) (c-point 'eol))))
416 (goto-char (match-end 0))))
417
418 \f
419 ;;; Basic utility functions.
420
421 (defun c-delq-from-dotted-list (elt dlist)
422 ;; If ELT is a member of the (possibly dotted) list DLIST, remove all
423 ;; occurrences of it (except for any in the last cdr of DLIST).
424 ;;
425 ;; Call this as (setq DLIST (c-delq-from-dotted-list ELT DLIST)), as
426 ;; sometimes the original structure is changed, sometimes it's not.
427 ;;
428 ;; This function is needed in Emacs < 24.5, and possibly XEmacs, because
429 ;; `delq' throws an error in these versions when given a dotted list.
430 (let ((tail dlist) prev)
431 (while (consp tail)
432 (if (eq (car tail) elt)
433 (if prev
434 (setcdr prev (cdr tail))
435 (setq dlist (cdr dlist)))
436 (setq prev tail))
437 (setq tail (cdr tail)))
438 dlist))
439
440 (defun c-syntactic-content (from to paren-level)
441 ;; Return the given region as a string where all syntactic
442 ;; whitespace is removed or, where necessary, replaced with a single
443 ;; space. If PAREN-LEVEL is given then all parens in the region are
444 ;; collapsed to "()", "[]" etc.
445 ;;
446 ;; This function might do hidden buffer changes.
447
448 (save-excursion
449 (save-restriction
450 (narrow-to-region from to)
451 (goto-char from)
452 (let* ((parts (list nil)) (tail parts) pos in-paren)
453
454 (while (re-search-forward c-syntactic-ws-start to t)
455 (goto-char (setq pos (match-beginning 0)))
456 (c-forward-syntactic-ws)
457 (if (= (point) pos)
458 (forward-char)
459
460 (when paren-level
461 (save-excursion
462 (setq in-paren (= (car (parse-partial-sexp from pos 1)) 1)
463 pos (point))))
464
465 (if (and (> pos from)
466 (< (point) to)
467 (looking-at "\\w\\|\\s_")
468 (save-excursion
469 (goto-char (1- pos))
470 (looking-at "\\w\\|\\s_")))
471 (progn
472 (setcdr tail (list (buffer-substring-no-properties from pos)
473 " "))
474 (setq tail (cddr tail)))
475 (setcdr tail (list (buffer-substring-no-properties from pos)))
476 (setq tail (cdr tail)))
477
478 (when in-paren
479 (when (= (car (parse-partial-sexp pos to -1)) -1)
480 (setcdr tail (list (buffer-substring-no-properties
481 (1- (point)) (point))))
482 (setq tail (cdr tail))))
483
484 (setq from (point))))
485
486 (setcdr tail (list (buffer-substring-no-properties from to)))
487 (apply 'concat (cdr parts))))))
488
489 (defun c-shift-line-indentation (shift-amt)
490 ;; Shift the indentation of the current line with the specified
491 ;; amount (positive inwards). The buffer is modified only if
492 ;; SHIFT-AMT isn't equal to zero.
493 (let ((pos (- (point-max) (point)))
494 (c-macro-start c-macro-start)
495 tmp-char-inserted)
496 (if (zerop shift-amt)
497 nil
498 ;; If we're on an empty line inside a macro, we take the point
499 ;; to be at the current indentation and shift it to the
500 ;; appropriate column. This way we don't treat the extra
501 ;; whitespace out to the line continuation as indentation.
502 (when (and (c-query-and-set-macro-start)
503 (looking-at "[ \t]*\\\\$")
504 (save-excursion
505 (skip-chars-backward " \t")
506 (bolp)))
507 (insert ?x)
508 (backward-char)
509 (setq tmp-char-inserted t))
510 (unwind-protect
511 (let ((col (current-indentation)))
512 (delete-region (c-point 'bol) (c-point 'boi))
513 (beginning-of-line)
514 (indent-to (+ col shift-amt)))
515 (when tmp-char-inserted
516 (delete-char 1))))
517 ;; If initial point was within line's indentation and we're not on
518 ;; a line with a line continuation in a macro, position after the
519 ;; indentation. Else stay at same point in text.
520 (if (and (< (point) (c-point 'boi))
521 (not tmp-char-inserted))
522 (back-to-indentation)
523 (if (> (- (point-max) pos) (point))
524 (goto-char (- (point-max) pos))))))
525
526 (defsubst c-keyword-sym (keyword)
527 ;; Return non-nil if the string KEYWORD is a known keyword. More
528 ;; precisely, the value is the symbol for the keyword in
529 ;; `c-keywords-obarray'.
530 (intern-soft keyword c-keywords-obarray))
531
532 (defsubst c-keyword-member (keyword-sym lang-constant)
533 ;; Return non-nil if the symbol KEYWORD-SYM, as returned by
534 ;; `c-keyword-sym', is a member of LANG-CONSTANT, which is the name
535 ;; of a language constant that ends with "-kwds". If KEYWORD-SYM is
536 ;; nil then the result is nil.
537 (get keyword-sym lang-constant))
538
539 ;; String syntax chars, suitable for skip-syntax-(forward|backward).
540 (defconst c-string-syntax (if (memq 'gen-string-delim c-emacs-features)
541 "\"|"
542 "\""))
543
544 ;; Regexp matching string limit syntax.
545 (defconst c-string-limit-regexp (if (memq 'gen-string-delim c-emacs-features)
546 "\\s\"\\|\\s|"
547 "\\s\""))
548
549 ;; Regexp matching WS followed by string limit syntax.
550 (defconst c-ws*-string-limit-regexp
551 (concat "[ \t]*\\(" c-string-limit-regexp "\\)"))
552
553 ;; Holds formatted error strings for the few cases where parse errors
554 ;; are reported.
555 (defvar c-parsing-error nil)
556 (make-variable-buffer-local 'c-parsing-error)
557
558 (defun c-echo-parsing-error (&optional quiet)
559 (when (and c-report-syntactic-errors c-parsing-error (not quiet))
560 (c-benign-error "%s" c-parsing-error))
561 c-parsing-error)
562
563 ;; Faces given to comments and string literals. This is used in some
564 ;; situations to speed up recognition; it isn't mandatory that font
565 ;; locking is in use. This variable is extended with the face in
566 ;; `c-doc-face-name' when fontification is activated in cc-fonts.el.
567 (defvar c-literal-faces
568 (append '(font-lock-comment-face font-lock-string-face)
569 (when (facep 'font-lock-comment-delimiter-face)
570 ;; New in Emacs 22.
571 '(font-lock-comment-delimiter-face))))
572
573 (defsubst c-put-c-type-property (pos value)
574 ;; Put a c-type property with the given value at POS.
575 (c-put-char-property pos 'c-type value))
576
577 (defun c-clear-c-type-property (from to value)
578 ;; Remove all occurrences of the c-type property that has the given
579 ;; value in the region between FROM and TO. VALUE is assumed to not
580 ;; be nil.
581 ;;
582 ;; Note: This assumes that c-type is put on single chars only; it's
583 ;; very inefficient if matching properties cover large regions.
584 (save-excursion
585 (goto-char from)
586 (while (progn
587 (when (eq (get-text-property (point) 'c-type) value)
588 (c-clear-char-property (point) 'c-type))
589 (goto-char (c-next-single-property-change (point) 'c-type nil to))
590 (< (point) to)))))
591
592 \f
593 ;; Some debug tools to visualize various special positions. This
594 ;; debug code isn't as portable as the rest of CC Mode.
595
596 (cc-bytecomp-defun overlays-in)
597 (cc-bytecomp-defun overlay-get)
598 (cc-bytecomp-defun overlay-start)
599 (cc-bytecomp-defun overlay-end)
600 (cc-bytecomp-defun delete-overlay)
601 (cc-bytecomp-defun overlay-put)
602 (cc-bytecomp-defun make-overlay)
603
604 (defun c-debug-add-face (beg end face)
605 (c-save-buffer-state ((overlays (overlays-in beg end)) overlay)
606 (while overlays
607 (setq overlay (car overlays)
608 overlays (cdr overlays))
609 (when (eq (overlay-get overlay 'face) face)
610 (setq beg (min beg (overlay-start overlay))
611 end (max end (overlay-end overlay)))
612 (delete-overlay overlay)))
613 (overlay-put (make-overlay beg end) 'face face)))
614
615 (defun c-debug-remove-face (beg end face)
616 (c-save-buffer-state ((overlays (overlays-in beg end)) overlay
617 (ol-beg beg) (ol-end end))
618 (while overlays
619 (setq overlay (car overlays)
620 overlays (cdr overlays))
621 (when (eq (overlay-get overlay 'face) face)
622 (setq ol-beg (min ol-beg (overlay-start overlay))
623 ol-end (max ol-end (overlay-end overlay)))
624 (delete-overlay overlay)))
625 (when (< ol-beg beg)
626 (overlay-put (make-overlay ol-beg beg) 'face face))
627 (when (> ol-end end)
628 (overlay-put (make-overlay end ol-end) 'face face))))
629
630 \f
631 ;; `c-beginning-of-statement-1' and accompanying stuff.
632
633 ;; KLUDGE ALERT: c-maybe-labelp is used to pass information between
634 ;; c-crosses-statement-barrier-p and c-beginning-of-statement-1. A
635 ;; better way should be implemented, but this will at least shut up
636 ;; the byte compiler.
637 (defvar c-maybe-labelp)
638
639 ;; New awk-compatible version of c-beginning-of-statement-1, ACM 2002/6/22
640
641 ;; Macros used internally in c-beginning-of-statement-1 for the
642 ;; automaton actions.
643 (defmacro c-bos-push-state ()
644 '(setq stack (cons (cons state saved-pos)
645 stack)))
646 (defmacro c-bos-pop-state (&optional do-if-done)
647 `(if (setq state (car (car stack))
648 saved-pos (cdr (car stack))
649 stack (cdr stack))
650 t
651 ,do-if-done
652 (throw 'loop nil)))
653 (defmacro c-bos-pop-state-and-retry ()
654 '(throw 'loop (setq state (car (car stack))
655 saved-pos (cdr (car stack))
656 ;; Throw nil if stack is empty, else throw non-nil.
657 stack (cdr stack))))
658 (defmacro c-bos-save-pos ()
659 '(setq saved-pos (vector pos tok ptok pptok)))
660 (defmacro c-bos-restore-pos ()
661 '(unless (eq (elt saved-pos 0) start)
662 (setq pos (elt saved-pos 0)
663 tok (elt saved-pos 1)
664 ptok (elt saved-pos 2)
665 pptok (elt saved-pos 3))
666 (goto-char pos)
667 (setq sym nil)))
668 (defmacro c-bos-save-error-info (missing got)
669 `(setq saved-pos (vector pos ,missing ,got)))
670 (defmacro c-bos-report-error ()
671 '(unless noerror
672 (setq c-parsing-error
673 (format-message
674 "No matching `%s' found for `%s' on line %d"
675 (elt saved-pos 1)
676 (elt saved-pos 2)
677 (1+ (count-lines (point-min)
678 (c-point 'bol (elt saved-pos 0))))))))
679
680 (defun c-beginning-of-statement-1 (&optional lim ignore-labels
681 noerror comma-delim)
682 "Move to the start of the current statement or declaration, or to
683 the previous one if already at the beginning of one. Only
684 statements/declarations on the same level are considered, i.e. don't
685 move into or out of sexps (not even normal expression parentheses).
686
687 If point is already at the earliest statement within braces or parens,
688 this function doesn't move back into any whitespace preceding it; it
689 returns `same' in this case.
690
691 Stop at statement continuation tokens like \"else\", \"catch\",
692 \"finally\" and the \"while\" in \"do ... while\" if the start point
693 is within the continuation. If starting at such a token, move to the
694 corresponding statement start. If at the beginning of a statement,
695 move to the closest containing statement if there is any. This might
696 also stop at a continuation clause.
697
698 Labels are treated as part of the following statements if
699 IGNORE-LABELS is non-nil. (FIXME: Doesn't work if we stop at a known
700 statement start keyword.) Otherwise, each label is treated as a
701 separate statement.
702
703 Macros are ignored \(i.e. skipped over) unless point is within one, in
704 which case the content of the macro is treated as normal code. Aside
705 from any normal statement starts found in it, stop at the first token
706 of the content in the macro, i.e. the expression of an \"#if\" or the
707 start of the definition in a \"#define\". Also stop at start of
708 macros before leaving them.
709
710 Return:
711 `label' if stopped at a label or \"case...:\" or \"default:\";
712 `same' if stopped at the beginning of the current statement;
713 `up' if stepped to a containing statement;
714 `previous' if stepped to a preceding statement;
715 `beginning' if stepped from a statement continuation clause to
716 its start clause; or
717 `macro' if stepped to a macro start.
718 Note that `same' and not `label' is returned if stopped at the same
719 label without crossing the colon character.
720
721 LIM may be given to limit the search. If the search hits the limit,
722 point will be left at the closest following token, or at the start
723 position if that is less (`same' is returned in this case).
724
725 NOERROR turns off error logging to `c-parsing-error'.
726
727 Normally only `;' and virtual semicolons are considered to delimit
728 statements, but if COMMA-DELIM is non-nil then `,' is treated
729 as a delimiter too.
730
731 Note that this function might do hidden buffer changes. See the
732 comment at the start of cc-engine.el for more info."
733
734 ;; The bulk of this function is a pushdown automaton that looks at statement
735 ;; boundaries and the tokens (such as "while") in c-opt-block-stmt-key. Its
736 ;; purpose is to keep track of nested statements, ensuring that such
737 ;; statements are skipped over in their entirety (somewhat akin to what C-M-p
738 ;; does with nested braces/brackets/parentheses).
739 ;;
740 ;; Note: The position of a boundary is the following token.
741 ;;
742 ;; Beginning with the current token (the one following point), move back one
743 ;; sexp at a time (where a sexp is, more or less, either a token or the
744 ;; entire contents of a brace/bracket/paren pair). Each time a statement
745 ;; boundary is crossed or a "while"-like token is found, update the state of
746 ;; the PDA. Stop at the beginning of a statement when the stack (holding
747 ;; nested statement info) is empty and the position has been moved.
748 ;;
749 ;; The following variables constitute the PDA:
750 ;;
751 ;; sym: This is either the "while"-like token (e.g. 'for) we've just
752 ;; scanned back over, 'boundary if we've just gone back over a
753 ;; statement boundary, or nil otherwise.
754 ;; state: takes one of the values (nil else else-boundary while
755 ;; while-boundary catch catch-boundary).
756 ;; nil means "no "while"-like token yet scanned".
757 ;; 'else, for example, means "just gone back over an else".
758 ;; 'else-boundary means "just gone back over a statement boundary
759 ;; immediately after having gone back over an else".
760 ;; saved-pos: A vector of either saved positions (tok ptok pptok, etc.) or
761 ;; of error reporting information.
762 ;; stack: The stack onto which the PDA pushes its state. Each entry
763 ;; consists of a saved value of state and saved-pos. An entry is
764 ;; pushed when we move back over a "continuation" token (e.g. else)
765 ;; and popped when we encounter the corresponding opening token
766 ;; (e.g. if).
767 ;;
768 ;;
769 ;; The following diagram briefly outlines the PDA.
770 ;;
771 ;; Common state:
772 ;; "else": Push state, goto state `else'.
773 ;; "while": Push state, goto state `while'.
774 ;; "catch" or "finally": Push state, goto state `catch'.
775 ;; boundary: Pop state.
776 ;; other: Do nothing special.
777 ;;
778 ;; State `else':
779 ;; boundary: Goto state `else-boundary'.
780 ;; other: Error, pop state, retry token.
781 ;;
782 ;; State `else-boundary':
783 ;; "if": Pop state.
784 ;; boundary: Error, pop state.
785 ;; other: See common state.
786 ;;
787 ;; State `while':
788 ;; boundary: Save position, goto state `while-boundary'.
789 ;; other: Pop state, retry token.
790 ;;
791 ;; State `while-boundary':
792 ;; "do": Pop state.
793 ;; boundary: Restore position if it's not at start, pop state. [*see below]
794 ;; other: See common state.
795 ;;
796 ;; State `catch':
797 ;; boundary: Goto state `catch-boundary'.
798 ;; other: Error, pop state, retry token.
799 ;;
800 ;; State `catch-boundary':
801 ;; "try": Pop state.
802 ;; "catch": Goto state `catch'.
803 ;; boundary: Error, pop state.
804 ;; other: See common state.
805 ;;
806 ;; [*] In the `while-boundary' state, we had pushed a 'while state, and were
807 ;; searching for a "do" which would have opened a do-while. If we didn't
808 ;; find it, we discard the analysis done since the "while", go back to this
809 ;; token in the buffer and restart the scanning there, this time WITHOUT
810 ;; pushing the 'while state onto the stack.
811 ;;
812 ;; In addition to the above there is some special handling of labels
813 ;; and macros.
814
815 (let ((case-fold-search nil)
816 (start (point))
817 macro-start
818 (delims (if comma-delim '(?\; ?,) '(?\;)))
819 (c-stmt-delim-chars (if comma-delim
820 c-stmt-delim-chars-with-comma
821 c-stmt-delim-chars))
822 c-in-literal-cache c-maybe-labelp after-case:-pos saved
823 ;; Current position.
824 pos
825 ;; Position of last stmt boundary character (e.g. ;).
826 boundary-pos
827 ;; The position of the last sexp or bound that follows the
828 ;; first found colon, i.e. the start of the nonlabel part of
829 ;; the statement. It's `start' if a colon is found just after
830 ;; the start.
831 after-labels-pos
832 ;; Like `after-labels-pos', but the first such position inside
833 ;; a label, i.e. the start of the last label before the start
834 ;; of the nonlabel part of the statement.
835 last-label-pos
836 ;; The last position where a label is possible provided the
837 ;; statement started there. It's nil as long as no invalid
838 ;; label content has been found (according to
839 ;; `c-nonlabel-token-key'). It's `start' if no valid label
840 ;; content was found in the label. Note that we might still
841 ;; regard it a label if it starts with `c-label-kwds'.
842 label-good-pos
843 ;; Putative positions of the components of a bitfield declaration,
844 ;; e.g. "int foo : NUM_FOO_BITS ;"
845 bitfield-type-pos bitfield-id-pos bitfield-size-pos
846 ;; Symbol just scanned back over (e.g. 'while or 'boundary).
847 ;; See above.
848 sym
849 ;; Current state in the automaton. See above.
850 state
851 ;; Current saved positions. See above.
852 saved-pos
853 ;; Stack of conses (state . saved-pos).
854 stack
855 ;; Regexp which matches "for", "if", etc.
856 (cond-key (or c-opt-block-stmt-key
857 "\\<\\>")) ; Matches nothing.
858 ;; Return value.
859 (ret 'same)
860 ;; Positions of the last three sexps or bounds we've stopped at.
861 tok ptok pptok)
862
863 (save-restriction
864 (if lim (narrow-to-region lim (point-max)))
865
866 (if (save-excursion
867 (and (c-beginning-of-macro)
868 (/= (point) start)))
869 (setq macro-start (point)))
870
871 ;; Try to skip back over unary operator characters, to register
872 ;; that we've moved.
873 (while (progn
874 (setq pos (point))
875 (c-backward-syntactic-ws)
876 ;; Protect post-++/-- operators just before a virtual semicolon.
877 (and (not (c-at-vsemi-p))
878 (/= (skip-chars-backward "-+!*&~@`#") 0))))
879
880 ;; Skip back over any semicolon here. If it was a bare semicolon, we're
881 ;; done. Later on we ignore the boundaries for statements that don't
882 ;; contain any sexp. The only thing that is affected is that the error
883 ;; checking is a little less strict, and we really don't bother.
884 (if (and (memq (char-before) delims)
885 (progn (forward-char -1)
886 (setq saved (point))
887 (c-backward-syntactic-ws)
888 (or (memq (char-before) delims)
889 (memq (char-before) '(?: nil))
890 (eq (char-syntax (char-before)) ?\()
891 (c-at-vsemi-p))))
892 (setq ret 'previous
893 pos saved)
894
895 ;; Begin at start and not pos to detect macros if we stand
896 ;; directly after the #.
897 (goto-char start)
898 (if (looking-at "\\<\\|\\W")
899 ;; Record this as the first token if not starting inside it.
900 (setq tok start))
901
902 ;; The following while loop goes back one sexp (balanced parens,
903 ;; etc. with contents, or symbol or suchlike) each iteration. This
904 ;; movement is accomplished with a call to c-backward-sexp approx 170
905 ;; lines below.
906 ;;
907 ;; The loop is exited only by throwing nil to the (catch 'loop ...):
908 ;; 1. On reaching the start of a macro;
909 ;; 2. On having passed a stmt boundary with the PDA stack empty;
910 ;; 3. On reaching the start of an Objective C method def;
911 ;; 4. From macro `c-bos-pop-state'; when the stack is empty;
912 ;; 5. From macro `c-bos-pop-state-and-retry' when the stack is empty.
913 (while
914 (catch 'loop ;; Throw nil to break, non-nil to continue.
915 (cond
916 ;; Are we in a macro, just after the opening #?
917 ((save-excursion
918 (and macro-start ; Always NIL for AWK.
919 (progn (skip-chars-backward " \t")
920 (eq (char-before) ?#))
921 (progn (setq saved (1- (point)))
922 (beginning-of-line)
923 (not (eq (char-before (1- (point))) ?\\)))
924 (looking-at c-opt-cpp-start)
925 (progn (skip-chars-forward " \t")
926 (eq (point) saved))))
927 (goto-char saved)
928 (if (and (c-forward-to-cpp-define-body)
929 (progn (c-forward-syntactic-ws start)
930 (< (point) start)))
931 ;; Stop at the first token in the content of the macro.
932 (setq pos (point)
933 ignore-labels t) ; Avoid the label check on exit.
934 (setq pos saved
935 ret 'macro
936 ignore-labels t))
937 (throw 'loop nil)) ; 1. Start of macro.
938
939 ;; Do a round through the automaton if we've just passed a
940 ;; statement boundary or passed a "while"-like token.
941 ((or sym
942 (and (looking-at cond-key)
943 (setq sym (intern (match-string 1)))))
944
945 (when (and (< pos start) (null stack))
946 (throw 'loop nil)) ; 2. Statement boundary.
947
948 ;; The PDA state handling.
949 ;;
950 ;; Refer to the description of the PDA in the opening
951 ;; comments. In the following OR form, the first leaf
952 ;; attempts to handles one of the specific actions detailed
953 ;; (e.g., finding token "if" whilst in state `else-boundary').
954 ;; We drop through to the second leaf (which handles common
955 ;; state) if no specific handler is found in the first cond.
956 ;; If a parsing error is detected (e.g. an "else" with no
957 ;; preceding "if"), we throw to the enclosing catch.
958 ;;
959 ;; Note that the (eq state 'else) means
960 ;; "we've just passed an else", NOT "we're looking for an
961 ;; else".
962 (or (cond
963 ((eq state 'else)
964 (if (eq sym 'boundary)
965 (setq state 'else-boundary)
966 (c-bos-report-error)
967 (c-bos-pop-state-and-retry)))
968
969 ((eq state 'else-boundary)
970 (cond ((eq sym 'if)
971 (c-bos-pop-state (setq ret 'beginning)))
972 ((eq sym 'boundary)
973 (c-bos-report-error)
974 (c-bos-pop-state))))
975
976 ((eq state 'while)
977 (if (and (eq sym 'boundary)
978 ;; Since this can cause backtracking we do a
979 ;; little more careful analysis to avoid it:
980 ;; If there's a label in front of the while
981 ;; it can't be part of a do-while.
982 (not after-labels-pos))
983 (progn (c-bos-save-pos)
984 (setq state 'while-boundary))
985 (c-bos-pop-state-and-retry))) ; Can't be a do-while
986
987 ((eq state 'while-boundary)
988 (cond ((eq sym 'do)
989 (c-bos-pop-state (setq ret 'beginning)))
990 ((eq sym 'boundary) ; isn't a do-while
991 (c-bos-restore-pos) ; the position of the while
992 (c-bos-pop-state)))) ; no longer searching for do.
993
994 ((eq state 'catch)
995 (if (eq sym 'boundary)
996 (setq state 'catch-boundary)
997 (c-bos-report-error)
998 (c-bos-pop-state-and-retry)))
999
1000 ((eq state 'catch-boundary)
1001 (cond
1002 ((eq sym 'try)
1003 (c-bos-pop-state (setq ret 'beginning)))
1004 ((eq sym 'catch)
1005 (setq state 'catch))
1006 ((eq sym 'boundary)
1007 (c-bos-report-error)
1008 (c-bos-pop-state)))))
1009
1010 ;; This is state common. We get here when the previous
1011 ;; cond statement found no particular state handler.
1012 (cond ((eq sym 'boundary)
1013 ;; If we have a boundary at the start
1014 ;; position we push a frame to go to the
1015 ;; previous statement.
1016 (if (>= pos start)
1017 (c-bos-push-state)
1018 (c-bos-pop-state)))
1019 ((eq sym 'else)
1020 (c-bos-push-state)
1021 (c-bos-save-error-info 'if 'else)
1022 (setq state 'else))
1023 ((eq sym 'while)
1024 ;; Is this a real while, or a do-while?
1025 ;; The next `when' triggers unless we are SURE that
1026 ;; the `while' is not the tail end of a `do-while'.
1027 (when (or (not pptok)
1028 (memq (char-after pptok) delims)
1029 ;; The following kludge is to prevent
1030 ;; infinite recursion when called from
1031 ;; c-awk-after-if-for-while-condition-p,
1032 ;; or the like.
1033 (and (eq (point) start)
1034 (c-vsemi-status-unknown-p))
1035 (c-at-vsemi-p pptok))
1036 ;; Since this can cause backtracking we do a
1037 ;; little more careful analysis to avoid it: If
1038 ;; the while isn't followed by a (possibly
1039 ;; virtual) semicolon it can't be a do-while.
1040 (c-bos-push-state)
1041 (setq state 'while)))
1042 ((memq sym '(catch finally))
1043 (c-bos-push-state)
1044 (c-bos-save-error-info 'try sym)
1045 (setq state 'catch))))
1046
1047 (when c-maybe-labelp
1048 ;; We're either past a statement boundary or at the
1049 ;; start of a statement, so throw away any label data
1050 ;; for the previous one.
1051 (setq after-labels-pos nil
1052 last-label-pos nil
1053 c-maybe-labelp nil))))
1054
1055 ;; Step to the previous sexp, but not if we crossed a
1056 ;; boundary, since that doesn't consume an sexp.
1057 (if (eq sym 'boundary)
1058 (setq ret 'previous)
1059
1060 ;; HERE IS THE SINGLE PLACE INSIDE THE PDA LOOP WHERE WE MOVE
1061 ;; BACKWARDS THROUGH THE SOURCE.
1062
1063 (c-backward-syntactic-ws)
1064 (let ((before-sws-pos (point))
1065 ;; The end position of the area to search for statement
1066 ;; barriers in this round.
1067 (maybe-after-boundary-pos pos))
1068
1069 ;; Go back over exactly one logical sexp, taking proper
1070 ;; account of macros and escaped EOLs.
1071 (while
1072 (progn
1073 (unless (c-safe (c-backward-sexp) t)
1074 ;; Give up if we hit an unbalanced block. Since the
1075 ;; stack won't be empty the code below will report a
1076 ;; suitable error.
1077 (throw 'loop nil))
1078 (cond
1079 ;; Have we moved into a macro?
1080 ((and (not macro-start)
1081 (c-beginning-of-macro))
1082 ;; Have we crossed a statement boundary? If not,
1083 ;; keep going back until we find one or a "real" sexp.
1084 (and
1085 (save-excursion
1086 (c-end-of-macro)
1087 (not (c-crosses-statement-barrier-p
1088 (point) maybe-after-boundary-pos)))
1089 (setq maybe-after-boundary-pos (point))))
1090 ;; Have we just gone back over an escaped NL? This
1091 ;; doesn't count as a sexp.
1092 ((looking-at "\\\\$")))))
1093
1094 ;; Have we crossed a statement boundary?
1095 (setq boundary-pos
1096 (cond
1097 ;; Are we at a macro beginning?
1098 ((and (not macro-start)
1099 c-opt-cpp-prefix
1100 (looking-at c-opt-cpp-prefix))
1101 (save-excursion
1102 (c-end-of-macro)
1103 (c-crosses-statement-barrier-p
1104 (point) maybe-after-boundary-pos)))
1105 ;; Just gone back over a brace block?
1106 ((and
1107 (eq (char-after) ?{)
1108 (not (c-looking-at-inexpr-block lim nil t))
1109 (save-excursion
1110 (c-backward-token-2 1 t nil)
1111 (not (looking-at "=\\([^=]\\|$\\)"))))
1112 (save-excursion
1113 (c-forward-sexp) (point)))
1114 ;; Just gone back over some paren block?
1115 ((looking-at "\\s(")
1116 (save-excursion
1117 (goto-char (1+ (c-down-list-backward
1118 before-sws-pos)))
1119 (c-crosses-statement-barrier-p
1120 (point) maybe-after-boundary-pos)))
1121 ;; Just gone back over an ordinary symbol of some sort?
1122 (t (c-crosses-statement-barrier-p
1123 (point) maybe-after-boundary-pos))))
1124
1125 (when boundary-pos
1126 (setq pptok ptok
1127 ptok tok
1128 tok boundary-pos
1129 sym 'boundary)
1130 ;; Like a C "continue". Analyze the next sexp.
1131 (throw 'loop t))))
1132
1133 ;; ObjC method def?
1134 (when (and c-opt-method-key
1135 (setq saved (c-in-method-def-p)))
1136 (setq pos saved
1137 ignore-labels t) ; Avoid the label check on exit.
1138 (throw 'loop nil)) ; 3. ObjC method def.
1139
1140 ;; Might we have a bitfield declaration, "<type> <id> : <size>"?
1141 (if c-has-bitfields
1142 (cond
1143 ;; The : <size> and <id> fields?
1144 ((and (numberp c-maybe-labelp)
1145 (not bitfield-size-pos)
1146 (save-excursion
1147 (goto-char (or tok start))
1148 (not (looking-at c-keywords-regexp)))
1149 (not (looking-at c-keywords-regexp))
1150 (not (c-punctuation-in (point) c-maybe-labelp)))
1151 (setq bitfield-size-pos (or tok start)
1152 bitfield-id-pos (point)))
1153 ;; The <type> field?
1154 ((and bitfield-id-pos
1155 (not bitfield-type-pos))
1156 (if (and (looking-at c-symbol-key) ; Can only be an integer type. :-)
1157 (not (looking-at c-not-primitive-type-keywords-regexp))
1158 (not (c-punctuation-in (point) tok)))
1159 (setq bitfield-type-pos (point))
1160 (setq bitfield-size-pos nil
1161 bitfield-id-pos nil)))))
1162
1163 ;; Handle labels.
1164 (unless (eq ignore-labels t)
1165 (when (numberp c-maybe-labelp)
1166 ;; `c-crosses-statement-barrier-p' has found a colon, so we
1167 ;; might be in a label now. Have we got a real label
1168 ;; (including a case label) or something like C++'s "public:"?
1169 ;; A case label might use an expression rather than a token.
1170 (setq after-case:-pos (or tok start))
1171 (if (or (looking-at c-nonlabel-token-key) ; e.g. "while" or "'a'"
1172 ;; Catch C++'s inheritance construct "class foo : bar".
1173 (save-excursion
1174 (and
1175 (c-safe (c-backward-sexp) t)
1176 (looking-at c-nonlabel-token-2-key))))
1177 (setq c-maybe-labelp nil)
1178 (if after-labels-pos ; Have we already encountered a label?
1179 (if (not last-label-pos)
1180 (setq last-label-pos (or tok start)))
1181 (setq after-labels-pos (or tok start)))
1182 (setq c-maybe-labelp t
1183 label-good-pos nil))) ; bogus "label"
1184
1185 (when (and (not label-good-pos) ; i.e. no invalid "label"'s yet
1186 ; been found.
1187 (looking-at c-nonlabel-token-key)) ; e.g. "while :"
1188 ;; We're in a potential label and it's the first
1189 ;; time we've found something that isn't allowed in
1190 ;; one.
1191 (setq label-good-pos (or tok start))))
1192
1193 ;; We've moved back by a sexp, so update the token positions.
1194 (setq sym nil
1195 pptok ptok
1196 ptok tok
1197 tok (point)
1198 pos tok) ; always non-nil
1199 ) ; end of (catch loop ....)
1200 ) ; end of sexp-at-a-time (while ....)
1201
1202 ;; If the stack isn't empty there might be errors to report.
1203 (while stack
1204 (if (and (vectorp saved-pos) (eq (length saved-pos) 3))
1205 (c-bos-report-error))
1206 (setq saved-pos (cdr (car stack))
1207 stack (cdr stack)))
1208
1209 (when (and (eq ret 'same)
1210 (not (memq sym '(boundary ignore nil))))
1211 ;; Need to investigate closer whether we've crossed
1212 ;; between a substatement and its containing statement.
1213 (if (setq saved
1214 (cond ((and (looking-at c-block-stmt-1-2-key)
1215 (eq (char-after ptok) ?\())
1216 pptok)
1217 ((looking-at c-block-stmt-1-key)
1218 ptok)
1219 (t pptok)))
1220 (cond ((> start saved) (setq pos saved))
1221 ((= start saved) (setq ret 'up)))))
1222
1223 (when (and (not ignore-labels)
1224 (eq c-maybe-labelp t)
1225 (not (eq ret 'beginning))
1226 after-labels-pos
1227 (not bitfield-type-pos) ; Bitfields take precedence over labels.
1228 (or (not label-good-pos)
1229 (<= label-good-pos pos)
1230 (progn
1231 (goto-char (if (and last-label-pos
1232 (< last-label-pos start))
1233 last-label-pos
1234 pos))
1235 (looking-at c-label-kwds-regexp))))
1236 ;; We're in a label. Maybe we should step to the statement
1237 ;; after it.
1238 (if (< after-labels-pos start)
1239 (setq pos after-labels-pos)
1240 (setq ret 'label)
1241 (if (and last-label-pos (< last-label-pos start))
1242 ;; Might have jumped over several labels. Go to the last one.
1243 (setq pos last-label-pos)))))
1244
1245 ;; Have we got "case <expression>:"?
1246 (goto-char pos)
1247 (when (and after-case:-pos
1248 (not (eq ret 'beginning))
1249 (looking-at c-case-kwds-regexp))
1250 (if (< after-case:-pos start)
1251 (setq pos after-case:-pos))
1252 (if (eq ret 'same)
1253 (setq ret 'label)))
1254
1255 ;; Skip over the unary operators that can start the statement.
1256 (while (progn
1257 (c-backward-syntactic-ws)
1258 ;; protect AWK post-inc/decrement operators, etc.
1259 (and (not (c-at-vsemi-p (point)))
1260 (/= (skip-chars-backward "-+!*&~@`#") 0)))
1261 (setq pos (point)))
1262 (goto-char pos)
1263 ret)))
1264
1265 (defun c-punctuation-in (from to)
1266 "Return non-nil if there is a non-comment non-macro punctuation character
1267 between FROM and TO. FROM must not be in a string or comment. The returned
1268 value is the position of the first such character."
1269 (save-excursion
1270 (goto-char from)
1271 (let ((pos (point)))
1272 (while (progn (skip-chars-forward c-symbol-chars to)
1273 (c-forward-syntactic-ws to)
1274 (> (point) pos))
1275 (setq pos (point))))
1276 (and (< (point) to) (point))))
1277
1278 (defun c-crosses-statement-barrier-p (from to)
1279 "Return non-nil if buffer positions FROM to TO cross one or more
1280 statement or declaration boundaries. The returned value is actually
1281 the position of the earliest boundary char. FROM must not be within
1282 a string or comment.
1283
1284 The variable `c-maybe-labelp' is set to the position of the first `:' that
1285 might start a label (i.e. not part of `::' and not preceded by `?'). If a
1286 single `?' is found, then `c-maybe-labelp' is cleared.
1287
1288 For AWK, a statement which is terminated by an EOL (not a ; or a }) is
1289 regarded as having a \"virtual semicolon\" immediately after the last token on
1290 the line. If this virtual semicolon is _at_ from, the function recognizes it.
1291
1292 Note that this function might do hidden buffer changes. See the
1293 comment at the start of cc-engine.el for more info."
1294 (let* ((skip-chars
1295 ;; If the current language has CPP macros, insert # into skip-chars.
1296 (if c-opt-cpp-symbol
1297 (concat (substring c-stmt-delim-chars 0 1) ; "^"
1298 c-opt-cpp-symbol ; usually "#"
1299 (substring c-stmt-delim-chars 1)) ; e.g. ";{}?:"
1300 c-stmt-delim-chars))
1301 (non-skip-list
1302 (append (substring skip-chars 1) nil)) ; e.g. (?# ?\; ?{ ?} ?? ?:)
1303 lit-range lit-start vsemi-pos)
1304 (save-restriction
1305 (widen)
1306 (save-excursion
1307 (catch 'done
1308 (goto-char from)
1309 (while (progn (skip-chars-forward
1310 skip-chars
1311 (min to (c-point 'bonl)))
1312 (< (point) to))
1313 (cond
1314 ;; Virtual semicolon?
1315 ((and (bolp)
1316 (save-excursion
1317 (progn
1318 (if (setq lit-start (c-literal-start from)) ; Have we landed in a string/comment?
1319 (goto-char lit-start))
1320 (c-backward-syntactic-ws) ; ? put a limit here, maybe?
1321 (setq vsemi-pos (point))
1322 (c-at-vsemi-p))))
1323 (throw 'done vsemi-pos))
1324 ;; In a string/comment?
1325 ((setq lit-range (c-literal-limits from))
1326 (goto-char (cdr lit-range)))
1327 ((eq (char-after) ?:)
1328 (forward-char)
1329 (if (and (eq (char-after) ?:)
1330 (< (point) to))
1331 ;; Ignore scope operators.
1332 (forward-char)
1333 (setq c-maybe-labelp (1- (point)))))
1334 ((eq (char-after) ??)
1335 ;; A question mark. Can't be a label, so stop
1336 ;; looking for more : and ?.
1337 (setq c-maybe-labelp nil
1338 skip-chars (substring c-stmt-delim-chars 0 -2)))
1339 ;; At a CPP construct or a "#" or "##" operator?
1340 ((and c-opt-cpp-symbol (looking-at c-opt-cpp-symbol))
1341 (if (save-excursion
1342 (skip-chars-backward " \t")
1343 (and (bolp)
1344 (or (bobp)
1345 (not (eq (char-before (1- (point))) ?\\)))))
1346 (c-end-of-macro)
1347 (skip-chars-forward c-opt-cpp-symbol)))
1348 ((memq (char-after) non-skip-list)
1349 (throw 'done (point)))))
1350 ;; In trailing space after an as yet undetected virtual semicolon?
1351 (c-backward-syntactic-ws from)
1352 (when (and (bolp) (not (bobp))) ; Can happen in AWK Mode with an
1353 ; unterminated string/regexp.
1354 (backward-char))
1355 (if (and (< (point) to)
1356 (c-at-vsemi-p))
1357 (point)
1358 nil))))))
1359
1360 (defun c-at-statement-start-p ()
1361 "Return non-nil if the point is at the first token in a statement
1362 or somewhere in the syntactic whitespace before it.
1363
1364 A \"statement\" here is not restricted to those inside code blocks.
1365 Any kind of declaration-like construct that occur outside function
1366 bodies is also considered a \"statement\".
1367
1368 Note that this function might do hidden buffer changes. See the
1369 comment at the start of cc-engine.el for more info."
1370
1371 (save-excursion
1372 (let ((end (point))
1373 c-maybe-labelp)
1374 (c-syntactic-skip-backward (substring c-stmt-delim-chars 1) nil t)
1375 (or (bobp)
1376 (eq (char-before) ?})
1377 (and (eq (char-before) ?{)
1378 (not (and c-special-brace-lists
1379 (progn (backward-char)
1380 (c-looking-at-special-brace-list)))))
1381 (c-crosses-statement-barrier-p (point) end)))))
1382
1383 (defun c-at-expression-start-p ()
1384 "Return non-nil if the point is at the first token in an expression or
1385 statement, or somewhere in the syntactic whitespace before it.
1386
1387 An \"expression\" here is a bit different from the normal language
1388 grammar sense: It's any sequence of expression tokens except commas,
1389 unless they are enclosed inside parentheses of some kind. Also, an
1390 expression never continues past an enclosing parenthesis, but it might
1391 contain parenthesis pairs of any sort except braces.
1392
1393 Since expressions never cross statement boundaries, this function also
1394 recognizes statement beginnings, just like `c-at-statement-start-p'.
1395
1396 Note that this function might do hidden buffer changes. See the
1397 comment at the start of cc-engine.el for more info."
1398
1399 (save-excursion
1400 (let ((end (point))
1401 (c-stmt-delim-chars c-stmt-delim-chars-with-comma)
1402 c-maybe-labelp)
1403 (c-syntactic-skip-backward (substring c-stmt-delim-chars 1) nil t)
1404 (or (bobp)
1405 (memq (char-before) '(?{ ?}))
1406 (save-excursion (backward-char)
1407 (looking-at "\\s("))
1408 (c-crosses-statement-barrier-p (point) end)))))
1409
1410 \f
1411 ;; A set of functions that covers various idiosyncrasies in
1412 ;; implementations of `forward-comment'.
1413
1414 ;; Note: Some emacsen considers incorrectly that any line comment
1415 ;; ending with a backslash continues to the next line. I can't think
1416 ;; of any way to work around that in a reliable way without changing
1417 ;; the buffer, though. Suggestions welcome. ;) (No, temporarily
1418 ;; changing the syntax for backslash doesn't work since we must treat
1419 ;; escapes in string literals correctly.)
1420
1421 (defun c-forward-single-comment ()
1422 "Move forward past whitespace and the closest following comment, if any.
1423 Return t if a comment was found, nil otherwise. In either case, the
1424 point is moved past the following whitespace. Line continuations,
1425 i.e. a backslashes followed by line breaks, are treated as whitespace.
1426 The line breaks that end line comments are considered to be the
1427 comment enders, so the point will be put on the beginning of the next
1428 line if it moved past a line comment.
1429
1430 This function does not do any hidden buffer changes."
1431
1432 (let ((start (point)))
1433 (when (looking-at "\\([ \t\n\r\f\v]\\|\\\\[\n\r]\\)+")
1434 (goto-char (match-end 0)))
1435
1436 (when (forward-comment 1)
1437 (if (eobp)
1438 ;; Some emacsen (e.g. XEmacs 21) return t when moving
1439 ;; forwards at eob.
1440 nil
1441
1442 ;; Emacs includes the ending newline in a b-style (c++)
1443 ;; comment, but XEmacs doesn't. We depend on the Emacs
1444 ;; behavior (which also is symmetric).
1445 (if (and (eolp) (elt (parse-partial-sexp start (point)) 7))
1446 (condition-case nil (forward-char 1)))
1447
1448 t))))
1449
1450 (defsubst c-forward-comments ()
1451 "Move forward past all following whitespace and comments.
1452 Line continuations, i.e. a backslashes followed by line breaks, are
1453 treated as whitespace.
1454
1455 Note that this function might do hidden buffer changes. See the
1456 comment at the start of cc-engine.el for more info."
1457
1458 (while (or
1459 ;; If forward-comment in at least XEmacs 21 is given a large
1460 ;; positive value, it'll loop all the way through if it hits
1461 ;; eob.
1462 (and (forward-comment 5)
1463 ;; Some emacsen (e.g. XEmacs 21) return t when moving
1464 ;; forwards at eob.
1465 (not (eobp)))
1466
1467 (when (looking-at "\\\\[\n\r]")
1468 (forward-char 2)
1469 t))))
1470
1471 (defun c-backward-single-comment ()
1472 "Move backward past whitespace and the closest preceding comment, if any.
1473 Return t if a comment was found, nil otherwise. In either case, the
1474 point is moved past the preceding whitespace. Line continuations,
1475 i.e. a backslashes followed by line breaks, are treated as whitespace.
1476 The line breaks that end line comments are considered to be the
1477 comment enders, so the point cannot be at the end of the same line to
1478 move over a line comment.
1479
1480 This function does not do any hidden buffer changes."
1481
1482 (let ((start (point)))
1483 ;; When we got newline terminated comments, forward-comment in all
1484 ;; supported emacsen so far will stop at eol of each line not
1485 ;; ending with a comment when moving backwards. This corrects for
1486 ;; that, and at the same time handles line continuations.
1487 (while (progn
1488 (skip-chars-backward " \t\n\r\f\v")
1489 (and (looking-at "[\n\r]")
1490 (eq (char-before) ?\\)))
1491 (backward-char))
1492
1493 (if (bobp)
1494 ;; Some emacsen (e.g. Emacs 19.34) return t when moving
1495 ;; backwards at bob.
1496 nil
1497
1498 ;; Leave point after the closest following newline if we've
1499 ;; backed up over any above, since forward-comment won't move
1500 ;; backward over a line comment if point is at the end of the
1501 ;; same line.
1502 (re-search-forward "\\=\\s *[\n\r]" start t)
1503
1504 (if (if (forward-comment -1)
1505 (if (eolp)
1506 ;; If forward-comment above succeeded and we're at eol
1507 ;; then the newline we moved over above didn't end a
1508 ;; line comment, so we give it another go.
1509 (forward-comment -1)
1510 t))
1511
1512 ;; Emacs <= 20 and XEmacs move back over the closer of a
1513 ;; block comment that lacks an opener.
1514 (if (looking-at "\\*/")
1515 (progn (forward-char 2) nil)
1516 t)))))
1517
1518 (defsubst c-backward-comments ()
1519 "Move backward past all preceding whitespace and comments.
1520 Line continuations, i.e. a backslashes followed by line breaks, are
1521 treated as whitespace. The line breaks that end line comments are
1522 considered to be the comment enders, so the point cannot be at the end
1523 of the same line to move over a line comment. Unlike
1524 c-backward-syntactic-ws, this function doesn't move back over
1525 preprocessor directives.
1526
1527 Note that this function might do hidden buffer changes. See the
1528 comment at the start of cc-engine.el for more info."
1529
1530 (let ((start (point)))
1531 (while (and
1532 ;; `forward-comment' in some emacsen (e.g. XEmacs 21.4)
1533 ;; return t when moving backwards at bob.
1534 (not (bobp))
1535
1536 (if (let (moved-comment)
1537 (while
1538 (and (not (setq moved-comment (forward-comment -1)))
1539 ;; Cope specifically with ^M^J here -
1540 ;; forward-comment sometimes gets stuck after ^Ms,
1541 ;; sometimes after ^M^J.
1542 (or
1543 (when (eq (char-before) ?\r)
1544 (backward-char)
1545 t)
1546 (when (and (eq (char-before) ?\n)
1547 (eq (char-before (1- (point))) ?\r))
1548 (backward-char 2)
1549 t))))
1550 moved-comment)
1551 (if (looking-at "\\*/")
1552 ;; Emacs <= 20 and XEmacs move back over the
1553 ;; closer of a block comment that lacks an opener.
1554 (progn (forward-char 2) nil)
1555 t)
1556
1557 ;; XEmacs treats line continuations as whitespace but
1558 ;; only in the backward direction, which seems a bit
1559 ;; odd. Anyway, this is necessary for Emacs.
1560 (when (and (looking-at "[\n\r]")
1561 (eq (char-before) ?\\)
1562 (< (point) start))
1563 (backward-char)
1564 t))))))
1565
1566 \f
1567 ;; Tools for skipping over syntactic whitespace.
1568
1569 ;; The following functions use text properties to cache searches over
1570 ;; large regions of syntactic whitespace. It works as follows:
1571 ;;
1572 ;; o If a syntactic whitespace region contains anything but simple
1573 ;; whitespace (i.e. space, tab and line breaks), the text property
1574 ;; `c-in-sws' is put over it. At places where we have stopped
1575 ;; within that region there's also a `c-is-sws' text property.
1576 ;; That since there typically are nested whitespace inside that
1577 ;; must be handled separately, e.g. whitespace inside a comment or
1578 ;; cpp directive. Thus, from one point with `c-is-sws' it's safe
1579 ;; to jump to another point with that property within the same
1580 ;; `c-in-sws' region. It can be likened to a ladder where
1581 ;; `c-in-sws' marks the bars and `c-is-sws' the rungs.
1582 ;;
1583 ;; o The `c-is-sws' property is put on the simple whitespace chars at
1584 ;; a "rung position" and also maybe on the first following char.
1585 ;; As many characters as can be conveniently found in this range
1586 ;; are marked, but no assumption can be made that the whole range
1587 ;; is marked (it could be clobbered by later changes, for
1588 ;; instance).
1589 ;;
1590 ;; Note that some part of the beginning of a sequence of simple
1591 ;; whitespace might be part of the end of a preceding line comment
1592 ;; or cpp directive and must not be considered part of the "rung".
1593 ;; Such whitespace is some amount of horizontal whitespace followed
1594 ;; by a newline. In the case of cpp directives it could also be
1595 ;; two newlines with horizontal whitespace between them.
1596 ;;
1597 ;; The reason to include the first following char is to cope with
1598 ;; "rung positions" that don't have any ordinary whitespace. If
1599 ;; `c-is-sws' is put on a token character it does not have
1600 ;; `c-in-sws' set simultaneously. That's the only case when that
1601 ;; can occur, and the reason for not extending the `c-in-sws'
1602 ;; region to cover it is that the `c-in-sws' region could then be
1603 ;; accidentally merged with a following one if the token is only
1604 ;; one character long.
1605 ;;
1606 ;; o On buffer changes the `c-in-sws' and `c-is-sws' properties are
1607 ;; removed in the changed region. If the change was inside
1608 ;; syntactic whitespace that means that the "ladder" is broken, but
1609 ;; a later call to `c-forward-sws' or `c-backward-sws' will use the
1610 ;; parts on either side and use an ordinary search only to "repair"
1611 ;; the gap.
1612 ;;
1613 ;; Special care needs to be taken if a region is removed: If there
1614 ;; are `c-in-sws' on both sides of it which do not connect inside
1615 ;; the region then they can't be joined. If e.g. a marked macro is
1616 ;; broken, syntactic whitespace inside the new text might be
1617 ;; marked. If those marks would become connected with the old
1618 ;; `c-in-sws' range around the macro then we could get a ladder
1619 ;; with one end outside the macro and the other at some whitespace
1620 ;; within it.
1621 ;;
1622 ;; The main motivation for this system is to increase the speed in
1623 ;; skipping over the large whitespace regions that can occur at the
1624 ;; top level in e.g. header files that contain a lot of comments and
1625 ;; cpp directives. For small comments inside code it's probably
1626 ;; slower than using `forward-comment' straightforwardly, but speed is
1627 ;; not a significant factor there anyway.
1628
1629 ; (defface c-debug-is-sws-face
1630 ; '((t (:background "GreenYellow")))
1631 ; "Debug face to mark the `c-is-sws' property.")
1632 ; (defface c-debug-in-sws-face
1633 ; '((t (:underline t)))
1634 ; "Debug face to mark the `c-in-sws' property.")
1635
1636 ; (defun c-debug-put-sws-faces ()
1637 ; ;; Put the sws debug faces on all the `c-is-sws' and `c-in-sws'
1638 ; ;; properties in the buffer.
1639 ; (interactive)
1640 ; (save-excursion
1641 ; (c-save-buffer-state (in-face)
1642 ; (goto-char (point-min))
1643 ; (setq in-face (if (get-text-property (point) 'c-is-sws)
1644 ; (point)))
1645 ; (while (progn
1646 ; (goto-char (next-single-property-change
1647 ; (point) 'c-is-sws nil (point-max)))
1648 ; (if in-face
1649 ; (progn
1650 ; (c-debug-add-face in-face (point) 'c-debug-is-sws-face)
1651 ; (setq in-face nil))
1652 ; (setq in-face (point)))
1653 ; (not (eobp))))
1654 ; (goto-char (point-min))
1655 ; (setq in-face (if (get-text-property (point) 'c-in-sws)
1656 ; (point)))
1657 ; (while (progn
1658 ; (goto-char (next-single-property-change
1659 ; (point) 'c-in-sws nil (point-max)))
1660 ; (if in-face
1661 ; (progn
1662 ; (c-debug-add-face in-face (point) 'c-debug-in-sws-face)
1663 ; (setq in-face nil))
1664 ; (setq in-face (point)))
1665 ; (not (eobp)))))))
1666
1667 (defmacro c-debug-sws-msg (&rest args)
1668 ;;`(message ,@args)
1669 )
1670
1671 (defmacro c-put-is-sws (beg end)
1672 ;; This macro does a hidden buffer change.
1673 `(let ((beg ,beg) (end ,end))
1674 (put-text-property beg end 'c-is-sws t)
1675 ,@(when (facep 'c-debug-is-sws-face)
1676 `((c-debug-add-face beg end 'c-debug-is-sws-face)))))
1677
1678 (defmacro c-put-in-sws (beg end)
1679 ;; This macro does a hidden buffer change.
1680 `(let ((beg ,beg) (end ,end))
1681 (put-text-property beg end 'c-in-sws t)
1682 ,@(when (facep 'c-debug-is-sws-face)
1683 `((c-debug-add-face beg end 'c-debug-in-sws-face)))))
1684
1685 (defmacro c-remove-is-sws (beg end)
1686 ;; This macro does a hidden buffer change.
1687 `(let ((beg ,beg) (end ,end))
1688 (remove-text-properties beg end '(c-is-sws nil))
1689 ,@(when (facep 'c-debug-is-sws-face)
1690 `((c-debug-remove-face beg end 'c-debug-is-sws-face)))))
1691
1692 (defmacro c-remove-in-sws (beg end)
1693 ;; This macro does a hidden buffer change.
1694 `(let ((beg ,beg) (end ,end))
1695 (remove-text-properties beg end '(c-in-sws nil))
1696 ,@(when (facep 'c-debug-is-sws-face)
1697 `((c-debug-remove-face beg end 'c-debug-in-sws-face)))))
1698
1699 (defmacro c-remove-is-and-in-sws (beg end)
1700 ;; This macro does a hidden buffer change.
1701 `(let ((beg ,beg) (end ,end))
1702 (remove-text-properties beg end '(c-is-sws nil c-in-sws nil))
1703 ,@(when (facep 'c-debug-is-sws-face)
1704 `((c-debug-remove-face beg end 'c-debug-is-sws-face)
1705 (c-debug-remove-face beg end 'c-debug-in-sws-face)))))
1706
1707 (defsubst c-invalidate-sws-region-after (beg end)
1708 ;; Called from `after-change-functions'. Note that if
1709 ;; `c-forward-sws' or `c-backward-sws' are used outside
1710 ;; `c-save-buffer-state' or similar then this will remove the cache
1711 ;; properties right after they're added.
1712 ;;
1713 ;; This function does hidden buffer changes.
1714
1715 (save-excursion
1716 ;; Adjust the end to remove the properties in any following simple
1717 ;; ws up to and including the next line break, if there is any
1718 ;; after the changed region. This is necessary e.g. when a rung
1719 ;; marked empty line is converted to a line comment by inserting
1720 ;; "//" before the line break. In that case the line break would
1721 ;; keep the rung mark which could make a later `c-backward-sws'
1722 ;; move into the line comment instead of over it.
1723 (goto-char end)
1724 (skip-chars-forward " \t\f\v")
1725 (when (and (eolp) (not (eobp)))
1726 (setq end (1+ (point)))))
1727
1728 (when (and (= beg end)
1729 (get-text-property beg 'c-in-sws)
1730 (> beg (point-min))
1731 (get-text-property (1- beg) 'c-in-sws))
1732 ;; Ensure that an `c-in-sws' range gets broken. Note that it isn't
1733 ;; safe to keep a range that was continuous before the change. E.g:
1734 ;;
1735 ;; #define foo
1736 ;; \
1737 ;; bar
1738 ;;
1739 ;; There can be a "ladder" between "#" and "b". Now, if the newline
1740 ;; after "foo" is removed then "bar" will become part of the cpp
1741 ;; directive instead of a syntactically relevant token. In that
1742 ;; case there's no longer syntactic ws from "#" to "b".
1743 (setq beg (1- beg)))
1744
1745 (c-debug-sws-msg "c-invalidate-sws-region-after [%s..%s]" beg end)
1746 (c-remove-is-and-in-sws beg end))
1747
1748 (defun c-forward-sws ()
1749 ;; Used by `c-forward-syntactic-ws' to implement the unbounded search.
1750 ;;
1751 ;; This function might do hidden buffer changes.
1752
1753 (let (;; `rung-pos' is set to a position as early as possible in the
1754 ;; unmarked part of the simple ws region.
1755 (rung-pos (point)) next-rung-pos rung-end-pos last-put-in-sws-pos
1756 rung-is-marked next-rung-is-marked simple-ws-end
1757 ;; `safe-start' is set when it's safe to cache the start position.
1758 ;; It's not set if we've initially skipped over comments and line
1759 ;; continuations since we might have gone out through the end of a
1760 ;; macro then. This provision makes `c-forward-sws' not populate the
1761 ;; cache in the majority of cases, but otoh is `c-backward-sws' by far
1762 ;; more common.
1763 safe-start)
1764
1765 ;; Skip simple ws and do a quick check on the following character to see
1766 ;; if it's anything that can't start syntactic ws, so we can bail out
1767 ;; early in the majority of cases when there just are a few ws chars.
1768 (skip-chars-forward " \t\n\r\f\v")
1769 (when (or (looking-at c-syntactic-ws-start)
1770 (and c-opt-cpp-prefix
1771 (looking-at c-noise-macro-name-re)))
1772
1773 (setq rung-end-pos (min (1+ (point)) (point-max)))
1774 (if (setq rung-is-marked (text-property-any rung-pos rung-end-pos
1775 'c-is-sws t))
1776 ;; Find the last rung position to avoid setting properties in all
1777 ;; the cases when the marked rung is complete.
1778 ;; (`next-single-property-change' is certain to move at least one
1779 ;; step forward.)
1780 (setq rung-pos (1- (c-next-single-property-change
1781 rung-is-marked 'c-is-sws nil rung-end-pos)))
1782 ;; Got no marked rung here. Since the simple ws might have started
1783 ;; inside a line comment or cpp directive we must set `rung-pos' as
1784 ;; high as possible.
1785 (setq rung-pos (point)))
1786
1787 (with-silent-modifications
1788 (while
1789 (progn
1790 ;; In the following while form, we move over a "ladder" and
1791 ;; following simple WS each time round the loop, appending the WS
1792 ;; onto the ladder, joining adjacent ladders, and terminating when
1793 ;; there is no more WS or we reach EOB.
1794 (while
1795 (when (and rung-is-marked
1796 (get-text-property (point) 'c-in-sws))
1797
1798 ;; The following search is the main reason that `c-in-sws'
1799 ;; and `c-is-sws' aren't combined to one property.
1800 (goto-char (c-next-single-property-change
1801 (point) 'c-in-sws nil (point-max)))
1802 (unless (get-text-property (point) 'c-is-sws)
1803 ;; If the `c-in-sws' region extended past the last
1804 ;; `c-is-sws' char we have to go back a bit.
1805 (or (get-text-property (1- (point)) 'c-is-sws)
1806 (goto-char (previous-single-property-change
1807 (point) 'c-is-sws)))
1808 (backward-char))
1809
1810 (c-debug-sws-msg
1811 "c-forward-sws cached move %s -> %s (max %s)"
1812 rung-pos (point) (point-max))
1813
1814 (setq rung-pos (point))
1815 (and (> (skip-chars-forward " \t\n\r\f\v") 0)
1816 (not (eobp))))
1817
1818 ;; We'll loop here if there is simple ws after the last rung.
1819 ;; That means that there's been some change in it and it's
1820 ;; possible that we've stepped into another ladder, so extend
1821 ;; the previous one to join with it if there is one, and try to
1822 ;; use the cache again.
1823 (c-debug-sws-msg
1824 "c-forward-sws extending rung with [%s..%s] (max %s)"
1825 (1+ rung-pos) (1+ (point)) (point-max))
1826 (unless (get-text-property (point) 'c-is-sws)
1827 ;; Remove any `c-in-sws' property from the last char of
1828 ;; the rung before we mark it with `c-is-sws', so that we
1829 ;; won't connect with the remains of a broken "ladder".
1830 (c-remove-in-sws (point) (1+ (point))))
1831 (c-put-is-sws (1+ rung-pos)
1832 (1+ (point)))
1833 (c-put-in-sws rung-pos
1834 (setq rung-pos (point)
1835 last-put-in-sws-pos rung-pos)))
1836
1837 ;; Now move over any comments (x)or a CPP construct.
1838 (setq simple-ws-end (point))
1839 (c-forward-comments)
1840
1841 (cond
1842 ((/= (point) simple-ws-end)
1843 ;; Skipped over comments. Don't cache at eob in case the buffer
1844 ;; is narrowed.
1845 (not (eobp)))
1846
1847 ((save-excursion
1848 (and c-opt-cpp-prefix
1849 (looking-at c-opt-cpp-start)
1850 (progn (skip-chars-backward " \t")
1851 (bolp))
1852 (or (bobp)
1853 (progn (backward-char)
1854 (not (eq (char-before) ?\\))))))
1855 ;; Skip a preprocessor directive.
1856 (end-of-line)
1857 (while (and (eq (char-before) ?\\)
1858 (= (forward-line 1) 0))
1859 (end-of-line))
1860 (forward-line 1)
1861 (setq safe-start t)
1862 ;; Don't cache at eob in case the buffer is narrowed.
1863 (not (eobp)))
1864
1865 ((and c-opt-cpp-prefix
1866 (looking-at c-noise-macro-name-re))
1867 ;; Skip over a noise macro.
1868 (goto-char (match-end 1))
1869 (setq safe-start t)
1870 (not (eobp)))))
1871
1872 ;; We've searched over a piece of non-white syntactic ws. See if this
1873 ;; can be cached.
1874 (setq next-rung-pos (point))
1875 (skip-chars-forward " \t\n\r\f\v")
1876 (setq rung-end-pos (min (1+ (point)) (point-max)))
1877
1878 (if (or
1879 ;; Cache if we haven't skipped comments only, and if we started
1880 ;; either from a marked rung or from a completely uncached
1881 ;; position.
1882 (and safe-start
1883 (or rung-is-marked
1884 (not (get-text-property simple-ws-end 'c-in-sws))))
1885
1886 ;; See if there's a marked rung in the encountered simple ws. If
1887 ;; so then we can cache, unless `safe-start' is nil. Even then
1888 ;; we need to do this to check if the cache can be used for the
1889 ;; next step.
1890 (and (setq next-rung-is-marked
1891 (text-property-any next-rung-pos rung-end-pos
1892 'c-is-sws t))
1893 safe-start))
1894
1895 (progn
1896 (c-debug-sws-msg
1897 "c-forward-sws caching [%s..%s] - [%s..%s] (max %s)"
1898 rung-pos (1+ simple-ws-end) next-rung-pos rung-end-pos
1899 (point-max))
1900
1901 ;; Remove the properties for any nested ws that might be cached.
1902 ;; Only necessary for `c-is-sws' since `c-in-sws' will be set
1903 ;; anyway.
1904 (c-remove-is-sws (1+ simple-ws-end) next-rung-pos)
1905 (unless (and rung-is-marked (= rung-pos simple-ws-end))
1906 (c-put-is-sws rung-pos
1907 (1+ simple-ws-end))
1908 (setq rung-is-marked t))
1909 (c-put-in-sws rung-pos
1910 (setq rung-pos (point)
1911 last-put-in-sws-pos rung-pos))
1912 (unless (get-text-property (1- rung-end-pos) 'c-is-sws)
1913 ;; Remove any `c-in-sws' property from the last char of
1914 ;; the rung before we mark it with `c-is-sws', so that we
1915 ;; won't connect with the remains of a broken "ladder".
1916 (c-remove-in-sws (1- rung-end-pos) rung-end-pos))
1917 (c-put-is-sws next-rung-pos
1918 rung-end-pos))
1919
1920 (c-debug-sws-msg
1921 "c-forward-sws not caching [%s..%s] - [%s..%s] (max %s)"
1922 rung-pos (1+ simple-ws-end) next-rung-pos rung-end-pos
1923 (point-max))
1924
1925 ;; Set `rung-pos' for the next rung. It's the same thing here as
1926 ;; initially, except that the rung position is set as early as
1927 ;; possible since we can't be in the ending ws of a line comment or
1928 ;; cpp directive now.
1929 (if (setq rung-is-marked next-rung-is-marked)
1930 (setq rung-pos (1- (c-next-single-property-change
1931 rung-is-marked 'c-is-sws nil rung-end-pos)))
1932 (setq rung-pos next-rung-pos))
1933 (setq safe-start t)))
1934
1935 ;; Make sure that the newly marked `c-in-sws' region doesn't connect to
1936 ;; another one after the point (which might occur when editing inside a
1937 ;; comment or macro).
1938 (when (eq last-put-in-sws-pos (point))
1939 (cond ((< last-put-in-sws-pos (point-max))
1940 (c-debug-sws-msg
1941 "c-forward-sws clearing at %s for cache separation"
1942 last-put-in-sws-pos)
1943 (c-remove-in-sws last-put-in-sws-pos
1944 (1+ last-put-in-sws-pos)))
1945 (t
1946 ;; If at eob we have to clear the last character before the end
1947 ;; instead since the buffer might be narrowed and there might
1948 ;; be a `c-in-sws' after (point-max). In this case it's
1949 ;; necessary to clear both properties.
1950 (c-debug-sws-msg
1951 "c-forward-sws clearing thoroughly at %s for cache separation"
1952 (1- last-put-in-sws-pos))
1953 (c-remove-is-and-in-sws (1- last-put-in-sws-pos)
1954 last-put-in-sws-pos))))
1955 ))))
1956
1957 (defun c-backward-sws ()
1958 ;; Used by `c-backward-syntactic-ws' to implement the unbounded search.
1959 ;;
1960 ;; This function might do hidden buffer changes.
1961
1962 (let (;; `rung-pos' is set to a position as late as possible in the unmarked
1963 ;; part of the simple ws region.
1964 (rung-pos (point)) next-rung-pos last-put-in-sws-pos
1965 rung-is-marked simple-ws-beg cmt-skip-pos)
1966
1967 ;; Skip simple horizontal ws and do a quick check on the preceding
1968 ;; character to see if it's anything that can't end syntactic ws, so we can
1969 ;; bail out early in the majority of cases when there just are a few ws
1970 ;; chars. Newlines are complicated in the backward direction, so we can't
1971 ;; skip over them.
1972 (skip-chars-backward " \t\f")
1973 (when (and (not (bobp))
1974 (save-excursion
1975 (backward-char)
1976 (or (looking-at c-syntactic-ws-end)
1977 (and c-opt-cpp-prefix
1978 (looking-at c-symbol-char-key)
1979 (progn (c-beginning-of-current-token)
1980 (looking-at c-noise-macro-name-re))))))
1981 ;; Try to find a rung position in the simple ws preceding point, so that
1982 ;; we can get a cache hit even if the last bit of the simple ws has
1983 ;; changed recently.
1984 (setq simple-ws-beg (point))
1985 (skip-chars-backward " \t\n\r\f\v")
1986 (if (setq rung-is-marked (text-property-any
1987 (point) (min (1+ rung-pos) (point-max))
1988 'c-is-sws t))
1989 ;; `rung-pos' will be the earliest marked position, which means that
1990 ;; there might be later unmarked parts in the simple ws region.
1991 ;; It's not worth the effort to fix that; the last part of the
1992 ;; simple ws is also typically edited often, so it could be wasted.
1993 (goto-char (setq rung-pos rung-is-marked))
1994 (goto-char simple-ws-beg))
1995
1996 (with-silent-modifications
1997 (while
1998 (progn
1999 ;; Each time round the next while form, we move back over a ladder
2000 ;; and append any simple WS preceding it, if possible joining with
2001 ;; the previous ladder.
2002 (while
2003 (when (and rung-is-marked
2004 (not (bobp))
2005 (get-text-property (1- (point)) 'c-in-sws))
2006
2007 ;; The following search is the main reason that `c-in-sws'
2008 ;; and `c-is-sws' aren't combined to one property.
2009 (goto-char (previous-single-property-change
2010 (point) 'c-in-sws nil (point-min)))
2011 (unless (get-text-property (point) 'c-is-sws)
2012 ;; If the `c-in-sws' region extended past the first
2013 ;; `c-is-sws' char we have to go forward a bit.
2014 (goto-char (c-next-single-property-change
2015 (point) 'c-is-sws)))
2016
2017 (c-debug-sws-msg
2018 "c-backward-sws cached move %s <- %s (min %s)"
2019 (point) rung-pos (point-min))
2020
2021 (setq rung-pos (point))
2022 (if (and (< (min (skip-chars-backward " \t\f\v")
2023 (progn
2024 (setq simple-ws-beg (point))
2025 (skip-chars-backward " \t\n\r\f\v")))
2026 0)
2027 (setq rung-is-marked
2028 (text-property-any (point) rung-pos
2029 'c-is-sws t)))
2030 t
2031 (goto-char simple-ws-beg)
2032 nil))
2033
2034 ;; We'll loop here if there is simple ws before the first rung.
2035 ;; That means that there's been some change in it and it's
2036 ;; possible that we've stepped into another ladder, so extend
2037 ;; the previous one to join with it if there is one, and try to
2038 ;; use the cache again.
2039 (c-debug-sws-msg
2040 "c-backward-sws extending rung with [%s..%s] (min %s)"
2041 rung-is-marked rung-pos (point-min))
2042 (unless (get-text-property (1- rung-pos) 'c-is-sws)
2043 ;; Remove any `c-in-sws' property from the last char of
2044 ;; the rung before we mark it with `c-is-sws', so that we
2045 ;; won't connect with the remains of a broken "ladder".
2046 (c-remove-in-sws (1- rung-pos) rung-pos))
2047 (c-put-is-sws rung-is-marked
2048 rung-pos)
2049 (c-put-in-sws rung-is-marked
2050 (1- rung-pos))
2051 (setq rung-pos rung-is-marked
2052 last-put-in-sws-pos rung-pos))
2053
2054 (c-backward-comments)
2055 (setq cmt-skip-pos (point))
2056
2057 (cond
2058 ((and c-opt-cpp-prefix
2059 (/= cmt-skip-pos simple-ws-beg)
2060 (c-beginning-of-macro))
2061 ;; Inside a cpp directive. See if it should be skipped over.
2062 (let ((cpp-beg (point)))
2063
2064 ;; Move back over all line continuations in the region skipped
2065 ;; over by `c-backward-comments'. If we go past it then we
2066 ;; started inside the cpp directive.
2067 (goto-char simple-ws-beg)
2068 (beginning-of-line)
2069 (while (and (> (point) cmt-skip-pos)
2070 (progn (backward-char)
2071 (eq (char-before) ?\\)))
2072 (beginning-of-line))
2073
2074 (if (< (point) cmt-skip-pos)
2075 ;; Don't move past the cpp directive if we began inside
2076 ;; it. Note that the position at the end of the last line
2077 ;; of the macro is also considered to be within it.
2078 (progn (goto-char cmt-skip-pos)
2079 nil)
2080
2081 ;; It's worthwhile to spend a little bit of effort on finding
2082 ;; the end of the macro, to get a good `simple-ws-beg'
2083 ;; position for the cache. Note that `c-backward-comments'
2084 ;; could have stepped over some comments before going into
2085 ;; the macro, and then `simple-ws-beg' must be kept on the
2086 ;; same side of those comments.
2087 (goto-char simple-ws-beg)
2088 (skip-chars-backward " \t\n\r\f\v")
2089 (if (eq (char-before) ?\\)
2090 (forward-char))
2091 (forward-line 1)
2092 (if (< (point) simple-ws-beg)
2093 ;; Might happen if comments after the macro were skipped
2094 ;; over.
2095 (setq simple-ws-beg (point)))
2096
2097 (goto-char cpp-beg)
2098 t)))
2099
2100 ((/= (save-excursion
2101 (skip-chars-forward " \t\n\r\f\v" simple-ws-beg)
2102 (setq next-rung-pos (point)))
2103 simple-ws-beg)
2104 ;; Skipped over comments. Must put point at the end of
2105 ;; the simple ws at point since we might be after a line
2106 ;; comment or cpp directive that's been partially
2107 ;; narrowed out, and we can't risk marking the simple ws
2108 ;; at the end of it.
2109 (goto-char next-rung-pos)
2110 t)
2111
2112 ((and c-opt-cpp-prefix
2113 (save-excursion
2114 (and (< (skip-syntax-backward "w_") 0)
2115 (progn (setq next-rung-pos (point))
2116 (looking-at c-noise-macro-name-re)))))
2117 ;; Skipped over a noise macro
2118 (goto-char next-rung-pos)
2119 t)))
2120
2121 ;; We've searched over a piece of non-white syntactic ws. See if this
2122 ;; can be cached.
2123 (setq next-rung-pos (point))
2124 (skip-chars-backward " \t\f\v")
2125
2126 (if (or
2127 ;; Cache if we started either from a marked rung or from a
2128 ;; completely uncached position.
2129 rung-is-marked
2130 (not (get-text-property (1- simple-ws-beg) 'c-in-sws))
2131
2132 ;; Cache if there's a marked rung in the encountered simple ws.
2133 (save-excursion
2134 (skip-chars-backward " \t\n\r\f\v")
2135 (text-property-any (point) (min (1+ next-rung-pos) (point-max))
2136 'c-is-sws t)))
2137
2138 (progn
2139 (c-debug-sws-msg
2140 "c-backward-sws caching [%s..%s] - [%s..%s] (min %s)"
2141 (point) (1+ next-rung-pos)
2142 simple-ws-beg (min (1+ rung-pos) (point-max))
2143 (point-min))
2144
2145 ;; Remove the properties for any nested ws that might be cached.
2146 ;; Only necessary for `c-is-sws' since `c-in-sws' will be set
2147 ;; anyway.
2148 (c-remove-is-sws (1+ next-rung-pos) simple-ws-beg)
2149 (unless (and rung-is-marked (= simple-ws-beg rung-pos))
2150 (let ((rung-end-pos (min (1+ rung-pos) (point-max))))
2151 (unless (get-text-property (1- rung-end-pos) 'c-is-sws)
2152 ;; Remove any `c-in-sws' property from the last char of
2153 ;; the rung before we mark it with `c-is-sws', so that we
2154 ;; won't connect with the remains of a broken "ladder".
2155 (c-remove-in-sws (1- rung-end-pos) rung-end-pos))
2156 (c-put-is-sws simple-ws-beg
2157 rung-end-pos)
2158 (setq rung-is-marked t)))
2159 (c-put-in-sws (setq simple-ws-beg (point)
2160 last-put-in-sws-pos simple-ws-beg)
2161 rung-pos)
2162 (c-put-is-sws (setq rung-pos simple-ws-beg)
2163 (1+ next-rung-pos)))
2164
2165 (c-debug-sws-msg
2166 "c-backward-sws not caching [%s..%s] - [%s..%s] (min %s)"
2167 (point) (1+ next-rung-pos)
2168 simple-ws-beg (min (1+ rung-pos) (point-max))
2169 (point-min))
2170 (setq rung-pos next-rung-pos
2171 simple-ws-beg (point))
2172 ))
2173
2174 ;; Make sure that the newly marked `c-in-sws' region doesn't connect to
2175 ;; another one before the point (which might occur when editing inside a
2176 ;; comment or macro).
2177 (when (eq last-put-in-sws-pos (point))
2178 (cond ((< (point-min) last-put-in-sws-pos)
2179 (c-debug-sws-msg
2180 "c-backward-sws clearing at %s for cache separation"
2181 (1- last-put-in-sws-pos))
2182 (c-remove-in-sws (1- last-put-in-sws-pos)
2183 last-put-in-sws-pos))
2184 ((> (point-min) 1)
2185 ;; If at bob and the buffer is narrowed, we have to clear the
2186 ;; character we're standing on instead since there might be a
2187 ;; `c-in-sws' before (point-min). In this case it's necessary
2188 ;; to clear both properties.
2189 (c-debug-sws-msg
2190 "c-backward-sws clearing thoroughly at %s for cache separation"
2191 last-put-in-sws-pos)
2192 (c-remove-is-and-in-sws last-put-in-sws-pos
2193 (1+ last-put-in-sws-pos)))))
2194 ))))
2195
2196 \f
2197 ;; Other whitespace tools
2198 (defun c-partial-ws-p (beg end)
2199 ;; Is the region (beg end) WS, and is there WS (or BOB/EOB) next to the
2200 ;; region? This is a "heuristic" function. .....
2201 ;;
2202 ;; The motivation for the second bit is to check whether removing this
2203 ;; region would coalesce two symbols.
2204 ;;
2205 ;; FIXME!!! This function doesn't check virtual semicolons in any way. Be
2206 ;; careful about using this function for, e.g. AWK. (2007/3/7)
2207 (save-excursion
2208 (let ((end+1 (min (1+ end) (point-max))))
2209 (or (progn (goto-char (max (point-min) (1- beg)))
2210 (c-skip-ws-forward end)
2211 (eq (point) end))
2212 (progn (goto-char beg)
2213 (c-skip-ws-forward end+1)
2214 (eq (point) end+1))))))
2215 \f
2216 ;; A system for finding noteworthy parens before the point.
2217
2218 (defconst c-state-cache-too-far 5000)
2219 ;; A maximum comfortable scanning distance, e.g. between
2220 ;; `c-state-cache-good-pos' and "HERE" (where we call c-parse-state). When
2221 ;; this distance is exceeded, we take "emergency measures", e.g. by clearing
2222 ;; the cache and starting again from point-min or a beginning of defun. This
2223 ;; value can be tuned for efficiency or set to a lower value for testing.
2224
2225 (defvar c-state-cache nil)
2226 (make-variable-buffer-local 'c-state-cache)
2227 ;; The state cache used by `c-parse-state' to cut down the amount of
2228 ;; searching. It's the result from some earlier `c-parse-state' call. See
2229 ;; `c-parse-state''s doc string for details of its structure.
2230 ;;
2231 ;; The use of the cached info is more effective if the next
2232 ;; `c-parse-state' call is on a line close by the one the cached state
2233 ;; was made at; the cache can actually slow down a little if the
2234 ;; cached state was made very far back in the buffer. The cache is
2235 ;; most effective if `c-parse-state' is used on each line while moving
2236 ;; forward.
2237
2238 (defvar c-state-cache-good-pos 1)
2239 (make-variable-buffer-local 'c-state-cache-good-pos)
2240 ;; This is a position where `c-state-cache' is known to be correct, or
2241 ;; nil (see below). It's a position inside one of the recorded unclosed
2242 ;; parens or the top level, but not further nested inside any literal or
2243 ;; subparen that is closed before the last recorded position.
2244 ;;
2245 ;; The exact position is chosen to try to be close to yet earlier than
2246 ;; the position where `c-state-cache' will be called next. Right now
2247 ;; the heuristic is to set it to the position after the last found
2248 ;; closing paren (of any type) before the line on which
2249 ;; `c-parse-state' was called. That is chosen primarily to work well
2250 ;; with refontification of the current line.
2251 ;;
2252 ;; 2009-07-28: When `c-state-point-min' and the last position where
2253 ;; `c-parse-state' or for which `c-invalidate-state-cache' was called, are
2254 ;; both in the same literal, there is no such "good position", and
2255 ;; c-state-cache-good-pos is then nil. This is the ONLY circumstance in which
2256 ;; it can be nil. In this case, `c-state-point-min-literal' will be non-nil.
2257 ;;
2258 ;; 2009-06-12: In a brace desert, c-state-cache-good-pos may also be in
2259 ;; the middle of the desert, as long as it is not within a brace pair
2260 ;; recorded in `c-state-cache' or a paren/bracket pair.
2261
2262 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2263 ;; We maintain a simple cache of positions which aren't in a literal, so as to
2264 ;; speed up testing for non-literality.
2265 (defconst c-state-nonlit-pos-interval 3000)
2266 ;; The approximate interval between entries in `c-state-nonlit-pos-cache'.
2267
2268 (defvar c-state-nonlit-pos-cache nil)
2269 (make-variable-buffer-local 'c-state-nonlit-pos-cache)
2270 ;; A list of buffer positions which are known not to be in a literal or a cpp
2271 ;; construct. This is ordered with higher positions at the front of the list.
2272 ;; Only those which are less than `c-state-nonlit-pos-cache-limit' are valid.
2273
2274 (defvar c-state-nonlit-pos-cache-limit 1)
2275 (make-variable-buffer-local 'c-state-nonlit-pos-cache-limit)
2276 ;; An upper limit on valid entries in `c-state-nonlit-pos-cache'. This is
2277 ;; reduced by buffer changes, and increased by invocations of
2278 ;; `c-state-literal-at'.
2279
2280 (defvar c-state-semi-nonlit-pos-cache nil)
2281 (make-variable-buffer-local 'c-state-semi-nonlit-pos-cache)
2282 ;; A list of elements which are either buffer positions (when such positions
2283 ;; are not in literals) or lists of the form (POS TYPE START), where POS is
2284 ;; a buffer position inside a literal, TYPE is the type of the literal
2285 ;; ('string, 'c, or 'c++) and START is the start of the literal.
2286
2287 (defvar c-state-semi-nonlit-pos-cache-limit 1)
2288 (make-variable-buffer-local 'c-state-semi-nonlit-pos-cache-limit)
2289 ;; An upper limit on valid entries in `c-state-semi-nonlit-pos-cache'. This
2290 ;; is reduced by buffer changes, and increased by invocations of
2291 ;; `c-parse-ps-state-below'.
2292
2293 (defsubst c-truncate-semi-nonlit-pos-cache (pos)
2294 ;; Truncate the upper bound of the cache `c-state-semi-nonlit-pos-cache' to
2295 ;; POS, if it is higher than that position.
2296 (setq c-state-semi-nonlit-pos-cache-limit
2297 (min c-state-semi-nonlit-pos-cache-limit pos)))
2298
2299 (defun c-state-semi-pp-to-literal (here &optional not-in-delimiter)
2300 ;; Do a parse-partial-sexp from a position in the buffer before HERE which
2301 ;; isn't in a literal, and return information about HERE, either:
2302 ;; (STATE TYPE BEG) if HERE is in a literal; or
2303 ;; (STATE) otherwise,
2304 ;; where STATE is the parsing state at HERE, TYPE is the type of the literal
2305 ;; enclosing HERE, (one of 'string, 'c, 'c++) and BEG is the starting
2306 ;; position of that literal (including the delimiter).
2307 ;;
2308 ;; Unless NOT-IN-DELIMITER is non-nil, when TO is inside a two-character
2309 ;; comment opener, this is recognized as being in a comment literal.
2310 ;;
2311 ;; Only elements 3 (in a string), 4 (in a comment), 5 (following a quote), 7
2312 ;; (comment type), and 8 (start of comment/string), and possibly 10 (in
2313 ;; newer Emacsen only, the syntax of a position after a potential first char
2314 ;; of a two char construct) of STATE are valid.
2315 (save-excursion
2316 (save-match-data
2317 (let* ((base-and-state (c-parse-ps-state-below here))
2318 (base (car base-and-state))
2319 (s (cdr base-and-state))
2320 (s (parse-partial-sexp base here nil nil s))
2321 ty)
2322 (cond
2323 ((or (nth 3 s) (nth 4 s)) ; in a string or comment
2324 (setq ty (cond
2325 ((nth 3 s) 'string)
2326 ((nth 7 s) 'c++)
2327 (t 'c)))
2328 (list s ty (nth 8 s)))
2329
2330 ((and (not not-in-delimiter) ; inside a comment starter
2331 (not (bobp))
2332 (progn (backward-char)
2333 (and (not (and (memq 'category-properties c-emacs-features)
2334 (looking-at "\\s!")))
2335 (looking-at c-comment-start-regexp))))
2336 (setq ty (if (looking-at c-block-comment-start-regexp) 'c 'c++))
2337 (list s ty (point)))
2338
2339 (t (list s)))))))
2340
2341 (defun c-state-full-pp-to-literal (here &optional not-in-delimiter)
2342 ;; This function will supersede c-state-pp-to-literal.
2343 ;;
2344 ;; Do a parse-partial-sexp from a position in the buffer before HERE which
2345 ;; isn't in a literal, and return information about HERE, either:
2346 ;; (STATE TYPE (BEG . END)) if HERE is in a literal; or
2347 ;; (STATE) otherwise,
2348 ;; where STATE is the parsing state at HERE, TYPE is the type of the literal
2349 ;; enclosing HERE, (one of 'string, 'c, 'c++) and (BEG . END) is the
2350 ;; boundaries of that literal (including the delimiters).
2351 ;;
2352 ;; Unless NOT-IN-DELIMITER is non-nil, when TO is inside a two-character
2353 ;; comment opener, this is recognized as being in a comment literal.
2354 ;;
2355 ;; Only elements 3 (in a string), 4 (in a comment), 5 (following a quote), 7
2356 ;; (comment type), and 8 (start of comment/string), and possibly 10 (in
2357 ;; newer Emacsen only, the syntax of a position after a potential first char
2358 ;; of a two char construct) of STATE are valid.
2359 (save-excursion
2360 (save-match-data
2361 (let* ((base-and-state (c-parse-ps-state-below here))
2362 (base (car base-and-state))
2363 (s (cdr base-and-state))
2364 (s (parse-partial-sexp base here nil nil s))
2365 ty start)
2366 (cond
2367 ((or (nth 3 s) (nth 4 s)) ; in a string or comment
2368 (setq ty (cond
2369 ((nth 3 s) 'string)
2370 ((nth 7 s) 'c++)
2371 (t 'c)))
2372 (setq start (nth 8 s))
2373 (parse-partial-sexp here (point-max)
2374 nil ; TARGETDEPTH
2375 nil ; STOPBEFORE
2376 s ; OLDSTATE
2377 'syntax-table) ; stop at end of literal
2378 (list s ty (cons start (point))))
2379
2380 ((and (not not-in-delimiter) ; inside a comment starter
2381 (not (bobp))
2382 (progn (backward-char)
2383 (and (not (and (memq 'category-properties c-emacs-features)
2384 (looking-at "\\s!")))
2385 (looking-at c-comment-start-regexp))))
2386 (setq ty (if (looking-at c-block-comment-start-regexp) 'c 'c++)
2387 start (point))
2388 (forward-comment 1)
2389 (list s ty (cons start (point))))
2390
2391 (t (list s)))))))
2392
2393 (defsubst c-state-pp-to-literal (from to &optional not-in-delimiter)
2394 ;; Do a parse-partial-sexp from FROM to TO, returning either
2395 ;; (STATE TYPE (BEG . END)) if TO is in a literal; or
2396 ;; (STATE) otherwise,
2397 ;; where STATE is the parsing state at TO, TYPE is the type of the literal
2398 ;; (one of 'c, 'c++, 'string) and (BEG . END) is the boundaries of the literal,
2399 ;; including the delimiters.
2400 ;;
2401 ;; Unless NOT-IN-DELIMITER is non-nil, when TO is inside a two-character
2402 ;; comment opener, this is recognized as being in a comment literal.
2403 ;;
2404 ;; Only elements 3 (in a string), 4 (in a comment), 5 (following a quote),
2405 ;; 7 (comment type) and 8 (start of comment/string) (and possibly 9) of
2406 ;; STATE are valid.
2407 (save-excursion
2408 (save-match-data
2409 (let ((s (parse-partial-sexp from to))
2410 ty co-st)
2411 (cond
2412 ((or (nth 3 s) (nth 4 s)) ; in a string or comment
2413 (setq ty (cond
2414 ((nth 3 s) 'string)
2415 ((nth 7 s) 'c++)
2416 (t 'c)))
2417 (parse-partial-sexp (point) (point-max)
2418 nil ; TARGETDEPTH
2419 nil ; STOPBEFORE
2420 s ; OLDSTATE
2421 'syntax-table) ; stop at end of literal
2422 `(,s ,ty (,(nth 8 s) . ,(point))))
2423
2424 ((and (not not-in-delimiter) ; inside a comment starter
2425 (not (bobp))
2426 (progn (backward-char)
2427 (and (not (looking-at "\\s!"))
2428 (looking-at c-comment-start-regexp))))
2429 (setq ty (if (looking-at c-block-comment-start-regexp) 'c 'c++)
2430 co-st (point))
2431 (forward-comment 1)
2432 `(,s ,ty (,co-st . ,(point))))
2433
2434 (t `(,s)))))))
2435
2436 (defun c-cache-to-parse-ps-state (elt)
2437 ;; Create a list suitable to use as the old-state parameter to
2438 ;; `parse-partial-sexp', out of ELT. ELT is either just a number, a buffer
2439 ;; position, or it is a list (POS TYPE STARTING-POS). Here POS is the
2440 ;; buffer position the other elements are pertinent for, TYPE is either 'c
2441 ;; or 'c++ (for a comment) or a character (for a string delimiter) or t
2442 ;; (meaning a string fence opened the string), STARTING-POS is the starting
2443 ;; position of the comment or string.
2444 (if (consp elt)
2445 (let ((depth 0) (containing nil) (last nil)
2446 in-string in-comment (after-quote nil)
2447 (min-depth 0) com-style com-str-start (intermediate nil)
2448 (between-syntax nil)
2449 (type (cadr elt)))
2450 (setq com-str-start (car (cddr elt)))
2451 (cond
2452 ((or (numberp type) (eq type t)) ; A string
2453 (setq in-string type))
2454 ((memq type '(c c++)) ; A comment
2455 (setq in-comment t
2456 com-style (if (eq type 'c++) 1 nil)))
2457 (t (c-benign-error "Invalid type %s in c-cache-to-parse-ps-state"
2458 elt)))
2459 (list depth containing last
2460 in-string in-comment after-quote
2461 min-depth com-style com-str-start
2462 intermediate nil))
2463 (copy-tree '(0 nil nil nil nil nil 0 nil nil nil nil))))
2464
2465 (defun c-parse-ps-state-to-cache (state)
2466 ;; Convert STATE, a `parse-partial-sexp' state valid at POINT, to an element
2467 ;; for the `c-state-semi-nonlit-pos-cache' cache. This is either POINT
2468 ;; (when point is not in a literal) or a list (POINT TYPE STARTING-POS),
2469 ;; where TYPE is the type of the literal, either 'string, 'c, or 'c++, and
2470 ;; STARTING-POS is the starting position of the comment or string.
2471 (cond
2472 ((nth 3 state) ; A string
2473 (list (point) (nth 3 state) (nth 8 state)))
2474 ((nth 4 state) ; A comment
2475 (list (point)
2476 (if (eq (nth 7 state) 1) 'c++ 'c)
2477 (nth 8 state)))
2478 (t ; Neither string nor comment.
2479 (point))))
2480
2481 (defsubst c-ps-state-cache-pos (elt)
2482 ;; Get the buffer position from ELT, an element from the cache
2483 ;; `c-state-semi-nonlit-pos-cache'.
2484 (if (atom elt)
2485 elt
2486 (car elt)))
2487
2488 (defun c-parse-ps-state-below (here)
2489 ;; Given a buffer position HERE, Return a cons (CACHE-POS . STATE), where
2490 ;; CACHE-POS is a position not very far before HERE for which the
2491 ;; parse-partial-sexp STATE is valid. Note that the only valid elements of
2492 ;; STATE are those concerning comments and strings; STATE is the state of a
2493 ;; null `parse-partial-sexp' scan when CACHE-POS is not in a comment or
2494 ;; string.
2495 (save-restriction
2496 (widen)
2497 (save-excursion
2498 (let ((c c-state-semi-nonlit-pos-cache)
2499 elt state pos npos high-elt)
2500 ;; Trim the cache to take account of buffer changes.
2501 (while (and c (> (c-ps-state-cache-pos (car c))
2502 c-state-semi-nonlit-pos-cache-limit))
2503 (setq c (cdr c)))
2504 (setq c-state-semi-nonlit-pos-cache c)
2505
2506 (while (and c (> (c-ps-state-cache-pos (car c)) here))
2507 (setq high-elt (car c))
2508 (setq c (cdr c)))
2509 (setq pos (or (and c (c-ps-state-cache-pos (car c)))
2510 (point-min)))
2511
2512 (if high-elt
2513 (setq state (c-cache-to-parse-ps-state (car c)))
2514 (setq elt (if c (car c) (point-min)))
2515 (setq state
2516 (if c
2517 (c-cache-to-parse-ps-state (car c))
2518 (copy-tree '(0 nil nil nil nil nil 0 nil nil nil nil))))
2519 (while
2520 ;; Add an element to `c-state-semi-nonlit-pos-cache' each iteration.
2521 (<= (setq npos (+ pos c-state-nonlit-pos-interval)) here)
2522 (setq state (parse-partial-sexp pos npos nil nil state))
2523 (setq elt (c-parse-ps-state-to-cache state))
2524 (setq c-state-semi-nonlit-pos-cache
2525 (cons elt c-state-semi-nonlit-pos-cache))
2526 (setq pos npos)))
2527
2528 (if (> pos c-state-semi-nonlit-pos-cache-limit)
2529 (setq c-state-semi-nonlit-pos-cache-limit pos))
2530
2531 (cons pos state)))))
2532
2533 (defun c-state-safe-place (here)
2534 ;; Return a buffer position before HERE which is "safe", i.e. outside any
2535 ;; string, comment, or macro.
2536 ;;
2537 ;; NOTE: This function manipulates `c-state-nonlit-pos-cache'. This cache
2538 ;; MAY NOT contain any positions within macros, since macros are frequently
2539 ;; turned into comments by use of the `c-cpp-delimiter' category properties.
2540 ;; We cannot rely on this mechanism whilst determining a cache pos since
2541 ;; this function is also called from outwith `c-parse-state'.
2542 (save-restriction
2543 (widen)
2544 (save-excursion
2545 (let ((c c-state-nonlit-pos-cache)
2546 pos npos high-pos lit macro-beg macro-end)
2547 ;; Trim the cache to take account of buffer changes.
2548 (while (and c (> (car c) c-state-nonlit-pos-cache-limit))
2549 (setq c (cdr c)))
2550 (setq c-state-nonlit-pos-cache c)
2551
2552 (while (and c (> (car c) here))
2553 (setq high-pos (car c))
2554 (setq c (cdr c)))
2555 (setq pos (or (car c) (point-min)))
2556
2557 (unless high-pos
2558 (while
2559 ;; Add an element to `c-state-nonlit-pos-cache' each iteration.
2560 (and
2561 (setq npos
2562 (when (<= (+ pos c-state-nonlit-pos-interval) here)
2563 (+ pos c-state-nonlit-pos-interval)))
2564
2565 ;; Test for being in a literal. If so, go to after it.
2566 (progn
2567 (setq lit (car (cddr (c-state-pp-to-literal pos npos))))
2568 (or (null lit)
2569 (prog1 (<= (cdr lit) here)
2570 (setq npos (cdr lit)))))
2571
2572 ;; Test for being in a macro. If so, go to after it.
2573 (progn
2574 (goto-char npos)
2575 (setq macro-beg
2576 (and (c-beginning-of-macro) (/= (point) npos) (point)))
2577 (when macro-beg
2578 (c-syntactic-end-of-macro)
2579 (or (eobp) (forward-char))
2580 (setq macro-end (point)))
2581 (or (null macro-beg)
2582 (prog1 (<= macro-end here)
2583 (setq npos macro-end)))))
2584
2585 (setq pos npos)
2586 (setq c-state-nonlit-pos-cache (cons pos c-state-nonlit-pos-cache)))
2587 ;; Add one extra element above HERE so as to to avoid the previous
2588 ;; expensive calculation when the next call is close to the current
2589 ;; one. This is especially useful when inside a large macro.
2590 (when npos
2591 (setq c-state-nonlit-pos-cache
2592 (cons npos c-state-nonlit-pos-cache))))
2593
2594 (if (> pos c-state-nonlit-pos-cache-limit)
2595 (setq c-state-nonlit-pos-cache-limit pos))
2596 pos))))
2597
2598 (defun c-state-literal-at (here)
2599 ;; If position HERE is inside a literal, return (START . END), the
2600 ;; boundaries of the literal (which may be outside the accessible bit of the
2601 ;; buffer). Otherwise, return nil.
2602 ;;
2603 ;; This function is almost the same as `c-literal-limits'. Previously, it
2604 ;; differed in that it was a lower level function, and that it rigorously
2605 ;; followed the syntax from BOB. `c-literal-limits' is now (2011-12)
2606 ;; virtually identical to this function.
2607 (save-restriction
2608 (widen)
2609 (save-excursion
2610 (let ((pos (c-state-safe-place here)))
2611 (car (cddr (c-state-pp-to-literal pos here)))))))
2612
2613 (defsubst c-state-lit-beg (pos)
2614 ;; Return the start of the literal containing POS, or POS itself.
2615 (or (car (c-state-literal-at pos))
2616 pos))
2617
2618 (defsubst c-state-cache-non-literal-place (pos state)
2619 ;; Return a position outside of a string/comment/macro at or before POS.
2620 ;; STATE is the parse-partial-sexp state at POS.
2621 (let ((res (if (or (nth 3 state) ; in a string?
2622 (nth 4 state)) ; in a comment?
2623 (nth 8 state)
2624 pos)))
2625 (save-excursion
2626 (goto-char res)
2627 (if (c-beginning-of-macro)
2628 (point)
2629 res))))
2630
2631 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2632 ;; Stuff to do with point-min, and coping with any literal there.
2633 (defvar c-state-point-min 1)
2634 (make-variable-buffer-local 'c-state-point-min)
2635 ;; This is (point-min) when `c-state-cache' was last calculated. A change of
2636 ;; narrowing is likely to affect the parens that are visible before the point.
2637
2638 (defvar c-state-point-min-lit-type nil)
2639 (make-variable-buffer-local 'c-state-point-min-lit-type)
2640 (defvar c-state-point-min-lit-start nil)
2641 (make-variable-buffer-local 'c-state-point-min-lit-start)
2642 ;; These two variables define the literal, if any, containing point-min.
2643 ;; Their values are, respectively, 'string, c, or c++, and the start of the
2644 ;; literal. If there's no literal there, they're both nil.
2645
2646 (defvar c-state-min-scan-pos 1)
2647 (make-variable-buffer-local 'c-state-min-scan-pos)
2648 ;; This is the earliest buffer-pos from which scanning can be done. It is
2649 ;; either the end of the literal containing point-min, or point-min itself.
2650 ;; It becomes nil if the buffer is changed earlier than this point.
2651 (defun c-state-get-min-scan-pos ()
2652 ;; Return the lowest valid scanning pos. This will be the end of the
2653 ;; literal enclosing point-min, or point-min itself.
2654 (or c-state-min-scan-pos
2655 (save-restriction
2656 (save-excursion
2657 (widen)
2658 (goto-char c-state-point-min-lit-start)
2659 (if (eq c-state-point-min-lit-type 'string)
2660 (forward-sexp)
2661 (forward-comment 1))
2662 (setq c-state-min-scan-pos (point))))))
2663
2664 (defun c-state-mark-point-min-literal ()
2665 ;; Determine the properties of any literal containing POINT-MIN, setting the
2666 ;; variables `c-state-point-min-lit-type', `c-state-point-min-lit-start',
2667 ;; and `c-state-min-scan-pos' accordingly. The return value is meaningless.
2668 (let ((p-min (point-min))
2669 lit)
2670 (save-restriction
2671 (widen)
2672 (setq lit (c-state-literal-at p-min))
2673 (if lit
2674 (setq c-state-point-min-lit-type
2675 (save-excursion
2676 (goto-char (car lit))
2677 (cond
2678 ((looking-at c-block-comment-start-regexp) 'c)
2679 ((looking-at c-line-comment-starter) 'c++)
2680 (t 'string)))
2681 c-state-point-min-lit-start (car lit)
2682 c-state-min-scan-pos (cdr lit))
2683 (setq c-state-point-min-lit-type nil
2684 c-state-point-min-lit-start nil
2685 c-state-min-scan-pos p-min)))))
2686
2687
2688 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2689 ;; A variable which signals a brace dessert - helpful for reducing the number
2690 ;; of fruitless backward scans.
2691 (defvar c-state-brace-pair-desert nil)
2692 (make-variable-buffer-local 'c-state-brace-pair-desert)
2693 ;; Used only in `c-append-lower-brace-pair-to-state-cache'. It is set when
2694 ;; that defun has searched backwards for a brace pair and not found one. Its
2695 ;; value is either nil or a cons (PA . FROM), where PA is the position of the
2696 ;; enclosing opening paren/brace/bracket which bounds the backwards search (or
2697 ;; nil when at top level) and FROM is where the backward search started. It
2698 ;; is reset to nil in `c-invalidate-state-cache'.
2699
2700
2701 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2702 ;; Lowish level functions/macros which work directly on `c-state-cache', or a
2703 ;; list of like structure.
2704 (defmacro c-state-cache-top-lparen (&optional cache)
2705 ;; Return the address of the top left brace/bracket/paren recorded in CACHE
2706 ;; (default `c-state-cache') (or nil).
2707 (let ((cash (or cache 'c-state-cache)))
2708 `(if (consp (car ,cash))
2709 (caar ,cash)
2710 (car ,cash))))
2711
2712 (defmacro c-state-cache-top-paren (&optional cache)
2713 ;; Return the address of the latest brace/bracket/paren (whether left or
2714 ;; right) recorded in CACHE (default `c-state-cache') or nil.
2715 (let ((cash (or cache 'c-state-cache)))
2716 `(if (consp (car ,cash))
2717 (cdar ,cash)
2718 (car ,cash))))
2719
2720 (defmacro c-state-cache-after-top-paren (&optional cache)
2721 ;; Return the position just after the latest brace/bracket/paren (whether
2722 ;; left or right) recorded in CACHE (default `c-state-cache') or nil.
2723 (let ((cash (or cache 'c-state-cache)))
2724 `(if (consp (car ,cash))
2725 (cdar ,cash)
2726 (and (car ,cash)
2727 (1+ (car ,cash))))))
2728
2729 (defun c-get-cache-scan-pos (here)
2730 ;; From the state-cache, determine the buffer position from which we might
2731 ;; scan forward to HERE to update this cache. This position will be just
2732 ;; after a paren/brace/bracket recorded in the cache, if possible, otherwise
2733 ;; return the earliest position in the accessible region which isn't within
2734 ;; a literal. If the visible portion of the buffer is entirely within a
2735 ;; literal, return NIL.
2736 (let ((c c-state-cache) elt)
2737 ;(while (>= (or (c-state-cache-top-lparen c) 1) here)
2738 (while (and c
2739 (>= (c-state-cache-top-lparen c) here))
2740 (setq c (cdr c)))
2741
2742 (setq elt (car c))
2743 (cond
2744 ((consp elt)
2745 (if (> (cdr elt) here)
2746 (1+ (car elt))
2747 (cdr elt)))
2748 (elt (1+ elt))
2749 ((<= (c-state-get-min-scan-pos) here)
2750 (c-state-get-min-scan-pos))
2751 (t nil))))
2752
2753 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2754 ;; Variables which keep track of preprocessor constructs.
2755 (defvar c-state-old-cpp-beg-marker nil)
2756 (make-variable-buffer-local 'c-state-old-cpp-beg-marker)
2757 (defvar c-state-old-cpp-beg nil)
2758 (make-variable-buffer-local 'c-state-old-cpp-beg)
2759 (defvar c-state-old-cpp-end-marker nil)
2760 (make-variable-buffer-local 'c-state-old-cpp-end-marker)
2761 (defvar c-state-old-cpp-end nil)
2762 (make-variable-buffer-local 'c-state-old-cpp-end)
2763 ;; These are the limits of the macro containing point at the previous call of
2764 ;; `c-parse-state', or nil.
2765
2766 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2767 ;; Defuns which analyze the buffer, yet don't change `c-state-cache'.
2768 (defun c-get-fallback-scan-pos (here)
2769 ;; Return a start position for building `c-state-cache' from
2770 ;; scratch. This will be at the top level, 2 defuns back.
2771 (save-excursion
2772 ;; Go back 2 bods, but ignore any bogus positions returned by
2773 ;; beginning-of-defun (i.e. open paren in column zero).
2774 (goto-char here)
2775 (let ((cnt 2))
2776 (while (not (or (bobp) (zerop cnt)))
2777 (c-beginning-of-defun-1) ; Pure elisp BOD.
2778 (if (eq (char-after) ?\{)
2779 (setq cnt (1- cnt)))))
2780 (point)))
2781
2782 (defun c-state-balance-parens-backwards (here- here+ top)
2783 ;; Return the position of the opening paren/brace/bracket before HERE- which
2784 ;; matches the outermost close p/b/b between HERE+ and TOP. Except when
2785 ;; there's a macro, HERE- and HERE+ are the same. Like this:
2786 ;;
2787 ;; ............................................
2788 ;; | |
2789 ;; ( [ ( .........#macro.. ) ( ) ] )
2790 ;; ^ ^ ^ ^
2791 ;; | | | |
2792 ;; return HERE- HERE+ TOP
2793 ;;
2794 ;; If there aren't enough opening paren/brace/brackets, return the position
2795 ;; of the outermost one found, or HERE- if there are none. If there are no
2796 ;; closing p/b/bs between HERE+ and TOP, return HERE-. HERE-/+ and TOP
2797 ;; must not be inside literals. Only the accessible portion of the buffer
2798 ;; will be scanned.
2799
2800 ;; PART 1: scan from `here+' up to `top', accumulating ")"s which enclose
2801 ;; `here'. Go round the next loop each time we pass over such a ")". These
2802 ;; probably match "("s before `here-'.
2803 (let (pos pa ren+1 lonely-rens)
2804 (save-excursion
2805 (save-restriction
2806 (narrow-to-region (point-min) top) ; This can move point, sometimes.
2807 (setq pos here+)
2808 (c-safe
2809 (while
2810 (setq ren+1 (c-sc-scan-lists pos 1 1)) ; might signal
2811 (setq lonely-rens (cons ren+1 lonely-rens)
2812 pos ren+1)))))
2813
2814 ;; PART 2: Scan back before `here-' searching for the "("s
2815 ;; matching/mismatching the ")"s found above. We only need to direct the
2816 ;; caller to scan when we've encountered unmatched right parens.
2817 (setq pos here-)
2818 (when lonely-rens
2819 (c-safe
2820 (while
2821 (and lonely-rens ; actual values aren't used.
2822 (setq pa (c-sc-scan-lists pos -1 1)))
2823 (setq pos pa)
2824 (setq lonely-rens (cdr lonely-rens)))))
2825 pos))
2826
2827 (defun c-parse-state-get-strategy (here good-pos)
2828 ;; Determine the scanning strategy for adjusting `c-parse-state', attempting
2829 ;; to minimize the amount of scanning. HERE is the pertinent position in
2830 ;; the buffer, GOOD-POS is a position where `c-state-cache' (possibly with
2831 ;; its head trimmed) is known to be good, or nil if there is no such
2832 ;; position.
2833 ;;
2834 ;; The return value is a list, one of the following:
2835 ;;
2836 ;; o - ('forward START-POINT) - scan forward from START-POINT,
2837 ;; which is not less than the highest position in `c-state-cache' below HERE,
2838 ;; which is after GOOD-POS.
2839 ;; o - ('backward nil) - scan backwards (from HERE).
2840 ;; o - ('back-and-forward START-POINT) - like 'forward, but when HERE is earlier
2841 ;; than GOOD-POS.
2842 ;; o - ('BOD START-POINT) - scan forwards from START-POINT, which is at the
2843 ;; top level.
2844 ;; o - ('IN-LIT nil) - point is inside the literal containing point-min.
2845 (let ((cache-pos (c-get-cache-scan-pos here)) ; highest position below HERE in cache (or 1)
2846 BOD-pos ; position of 2nd BOD before HERE.
2847 strategy ; 'forward, 'backward, 'BOD, or 'IN-LIT.
2848 start-point
2849 how-far) ; putative scanning distance.
2850 (setq good-pos (or good-pos (c-state-get-min-scan-pos)))
2851 (cond
2852 ((< here (c-state-get-min-scan-pos))
2853 (setq strategy 'IN-LIT
2854 start-point nil
2855 cache-pos nil
2856 how-far 0))
2857 ((<= good-pos here)
2858 (setq strategy 'forward
2859 start-point (max good-pos cache-pos)
2860 how-far (- here start-point)))
2861 ((< (- good-pos here) (- here cache-pos)) ; FIXME!!! ; apply some sort of weighting.
2862 (setq strategy 'backward
2863 how-far (- good-pos here)))
2864 (t
2865 (setq strategy 'back-and-forward
2866 start-point cache-pos
2867 how-far (- here start-point))))
2868
2869 ;; Might we be better off starting from the top level, two defuns back,
2870 ;; instead? This heuristic no longer works well in C++, where
2871 ;; declarations inside namespace brace blocks are frequently placed at
2872 ;; column zero. (2015-11-10): Remove the condition on C++ Mode.
2873 (when (and (or (not (memq 'col-0-paren c-emacs-features))
2874 open-paren-in-column-0-is-defun-start)
2875 ;; (not (c-major-mode-is 'c++-mode))
2876 (> how-far c-state-cache-too-far))
2877 (setq BOD-pos (c-get-fallback-scan-pos here)) ; somewhat EXPENSIVE!!!
2878 (if (< (- here BOD-pos) how-far)
2879 (setq strategy 'BOD
2880 start-point BOD-pos)))
2881
2882 (list strategy start-point)))
2883
2884
2885 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2886 ;; Routines which change `c-state-cache' and associated values.
2887 (defun c-renarrow-state-cache ()
2888 ;; The region (more precisely, point-min) has changed since we
2889 ;; calculated `c-state-cache'. Amend `c-state-cache' accordingly.
2890 (if (< (point-min) c-state-point-min)
2891 ;; If point-min has MOVED BACKWARDS then we drop the state completely.
2892 ;; It would be possible to do a better job here and recalculate the top
2893 ;; only.
2894 (progn
2895 (c-state-mark-point-min-literal)
2896 (setq c-state-cache nil
2897 c-state-cache-good-pos c-state-min-scan-pos
2898 c-state-brace-pair-desert nil))
2899
2900 ;; point-min has MOVED FORWARD.
2901
2902 ;; Is the new point-min inside a (different) literal?
2903 (unless (and c-state-point-min-lit-start ; at prev. point-min
2904 (< (point-min) (c-state-get-min-scan-pos)))
2905 (c-state-mark-point-min-literal))
2906
2907 ;; Cut off a bit of the tail from `c-state-cache'.
2908 (let ((ptr (cons nil c-state-cache))
2909 pa)
2910 (while (and (setq pa (c-state-cache-top-lparen (cdr ptr)))
2911 (>= pa (point-min)))
2912 (setq ptr (cdr ptr)))
2913
2914 (when (consp ptr)
2915 (if (or (eq (cdr ptr) c-state-cache)
2916 (and (consp (cadr ptr))
2917 (> (cdr (cadr ptr)) (point-min)))) ; Our new point-min is
2918 ; inside a recorded
2919 ; brace pair.
2920 (setq c-state-cache nil
2921 c-state-cache-good-pos c-state-min-scan-pos)
2922 (setcdr ptr nil)
2923 (setq c-state-cache-good-pos (1+ (c-state-cache-top-lparen))))
2924 )))
2925
2926 (setq c-state-point-min (point-min)))
2927
2928 (defun c-append-lower-brace-pair-to-state-cache (from here &optional upper-lim)
2929 ;; If there is a brace pair preceding FROM in the buffer, at the same level
2930 ;; of nesting (not necessarily immediately preceding), push a cons onto
2931 ;; `c-state-cache' to represent it. FROM must not be inside a literal. If
2932 ;; UPPER-LIM is non-nil, we append the highest brace pair whose "}" is below
2933 ;; UPPER-LIM.
2934 ;;
2935 ;; Return non-nil when this has been done.
2936 ;;
2937 ;; The situation it copes with is this transformation:
2938 ;;
2939 ;; OLD: { (.) {...........}
2940 ;; ^ ^
2941 ;; FROM HERE
2942 ;;
2943 ;; NEW: { {....} (.) {.........
2944 ;; ^ ^ ^
2945 ;; LOWER BRACE PAIR HERE or HERE
2946 ;;
2947 ;; This routine should be fast. Since it can get called a LOT, we maintain
2948 ;; `c-state-brace-pair-desert', a small cache of "failures", such that we
2949 ;; reduce the time wasted in repeated fruitless searches in brace deserts.
2950 (save-excursion
2951 (save-restriction
2952 (let* (new-cons
2953 (cache-pos (c-state-cache-top-lparen)) ; might be nil.
2954 (macro-start-or-from
2955 (progn (goto-char from)
2956 (c-beginning-of-macro)
2957 (point)))
2958 (bra ; Position of "{".
2959 ;; Don't start scanning in the middle of a CPP construct unless
2960 ;; it contains HERE - these constructs, in Emacs, are "commented
2961 ;; out" with category properties.
2962 (if (eq (c-get-char-property macro-start-or-from 'category)
2963 'c-cpp-delimiter)
2964 macro-start-or-from
2965 from))
2966 ce) ; Position of "}"
2967 (or upper-lim (setq upper-lim from))
2968
2969 ;; If we're essentially repeating a fruitless search, just give up.
2970 (unless (and c-state-brace-pair-desert
2971 (eq cache-pos (car c-state-brace-pair-desert))
2972 (or (null (car c-state-brace-pair-desert))
2973 (> from (car c-state-brace-pair-desert)))
2974 (<= from (cdr c-state-brace-pair-desert)))
2975 ;; DESERT-LIM. Avoid repeated searching through the cached desert.
2976 (let ((desert-lim
2977 (and c-state-brace-pair-desert
2978 (eq cache-pos (car c-state-brace-pair-desert))
2979 (>= from (cdr c-state-brace-pair-desert))
2980 (cdr c-state-brace-pair-desert)))
2981 ;; CACHE-LIM. This limit will be necessary when an opening
2982 ;; paren at `cache-pos' has just had its matching close paren
2983 ;; inserted into the buffer. `cache-pos' continues to be a
2984 ;; search bound, even though the algorithm below would skip
2985 ;; over the new paren pair.
2986 (cache-lim (and cache-pos (< cache-pos from) cache-pos)))
2987 (narrow-to-region
2988 (cond
2989 ((and desert-lim cache-lim)
2990 (max desert-lim cache-lim))
2991 (desert-lim)
2992 (cache-lim)
2993 ((point-min)))
2994 ;; The top limit is EOB to ensure that `bra' is inside the
2995 ;; accessible part of the buffer at the next scan operation.
2996 (1+ (buffer-size))))
2997
2998 ;; In the next pair of nested loops, the inner one moves back past a
2999 ;; pair of (mis-)matching parens or brackets; the outer one moves
3000 ;; back over a sequence of unmatched close brace/paren/bracket each
3001 ;; time round.
3002 (while
3003 (progn
3004 (c-safe
3005 (while
3006 (and (setq ce (c-sc-scan-lists bra -1 -1)) ; back past )/]/}; might signal
3007 (setq bra (c-sc-scan-lists ce -1 1)) ; back past (/[/{; might signal
3008 (or (> bra here) ;(> ce here)
3009 (and
3010 (< ce here)
3011 (or (not (eq (char-after bra) ?\{))
3012 (and (goto-char bra)
3013 (c-beginning-of-macro)
3014 (< (point) macro-start-or-from))))))))
3015 (and ce (< ce bra)))
3016 (setq bra ce)) ; If we just backed over an unbalanced closing
3017 ; brace, ignore it.
3018
3019 (if (and ce (< ce here) (< bra ce) (eq (char-after bra) ?\{))
3020 ;; We've found the desired brace-pair.
3021 (progn
3022 (setq new-cons (cons bra (1+ ce)))
3023 (cond
3024 ((consp (car c-state-cache))
3025 (setcar c-state-cache new-cons))
3026 ((and (numberp (car c-state-cache)) ; probably never happens
3027 (< ce (car c-state-cache)))
3028 (setcdr c-state-cache
3029 (cons new-cons (cdr c-state-cache))))
3030 (t (setq c-state-cache (cons new-cons c-state-cache)))))
3031
3032 ;; We haven't found a brace pair. Record this in the cache.
3033 (setq c-state-brace-pair-desert
3034 (cons (if (and ce (< bra ce) (> ce here)) ; {..} straddling HERE?
3035 bra
3036 (point-min))
3037 (min here from)))))))))
3038
3039 (defsubst c-state-push-any-brace-pair (bra+1 macro-start-or-here)
3040 ;; If BRA+1 is nil, do nothing. Otherwise, BRA+1 is the buffer position
3041 ;; following a {, and that brace has a (mis-)matching } (or ]), and we
3042 ;; "push" "a" brace pair onto `c-state-cache'.
3043 ;;
3044 ;; Here "push" means overwrite the top element if it's itself a brace-pair,
3045 ;; otherwise push it normally.
3046 ;;
3047 ;; The brace pair we push is normally the one surrounding BRA+1, but if the
3048 ;; latter is inside a macro, not being a macro containing
3049 ;; MACRO-START-OR-HERE, we scan backwards through the buffer for a non-macro
3050 ;; base pair. This latter case is assumed to be rare.
3051 ;;
3052 ;; Note: POINT is not preserved in this routine.
3053 (if bra+1
3054 (if (or (> bra+1 macro-start-or-here)
3055 (progn (goto-char bra+1)
3056 (not (c-beginning-of-macro))))
3057 (setq c-state-cache
3058 (cons (cons (1- bra+1)
3059 (c-sc-scan-lists bra+1 1 1))
3060 (if (consp (car c-state-cache))
3061 (cdr c-state-cache)
3062 c-state-cache)))
3063 ;; N.B. This defsubst codes one method for the simple, normal case,
3064 ;; and a more sophisticated, slower way for the general case. Don't
3065 ;; eliminate this defsubst - it's a speed optimization.
3066 (c-append-lower-brace-pair-to-state-cache (1- bra+1) (point-max)))))
3067
3068 (defun c-append-to-state-cache (from here)
3069 ;; Scan the buffer from FROM to HERE, adding elements into `c-state-cache'
3070 ;; for braces etc. Return a candidate for `c-state-cache-good-pos'.
3071 ;;
3072 ;; FROM must be after the latest brace/paren/bracket in `c-state-cache', if
3073 ;; any. Typically, it is immediately after it. It must not be inside a
3074 ;; literal.
3075 (let ((here-bol (c-point 'bol here))
3076 (macro-start-or-here
3077 (save-excursion (goto-char here)
3078 (if (c-beginning-of-macro)
3079 (point)
3080 here)))
3081 pa+1 ; pos just after an opening PAren (or brace).
3082 (ren+1 from) ; usually a pos just after an closing paREN etc.
3083 ; Is actually the pos. to scan for a (/{/[ from,
3084 ; which sometimes is after a silly )/}/].
3085 paren+1 ; Pos after some opening or closing paren.
3086 paren+1s ; A list of `paren+1's; used to determine a
3087 ; good-pos.
3088 bra+1 ; just after L bra-ce.
3089 bra+1s ; list of OLD values of bra+1.
3090 mstart) ; start of a macro.
3091
3092 (save-excursion
3093 (save-restriction
3094 (narrow-to-region (point-min) here)
3095 ;; Each time round the following loop, we enter a successively deeper
3096 ;; level of brace/paren nesting. (Except sometimes we "continue at
3097 ;; the existing level".) `pa+1' is a pos inside an opening
3098 ;; brace/paren/bracket, usually just after it.
3099 (while
3100 (progn
3101 ;; Each time round the next loop moves forward over an opening then
3102 ;; a closing brace/bracket/paren. This loop is white hot, so it
3103 ;; plays ugly tricks to go fast. DON'T PUT ANYTHING INTO THIS
3104 ;; LOOP WHICH ISN'T ABSOLUTELY NECESSARY!!! It terminates when a
3105 ;; call of `scan-lists' signals an error, which happens when there
3106 ;; are no more b/b/p's to scan.
3107 (c-safe
3108 (while t
3109 (setq pa+1 (c-sc-scan-lists ren+1 1 -1) ; Into (/{/[; might signal
3110 paren+1s (cons pa+1 paren+1s))
3111 (setq ren+1 (c-sc-scan-lists pa+1 1 1)) ; Out of )/}/]; might signal
3112 (if (and (eq (char-before pa+1) ?{)) ; Check for a macro later.
3113 (setq bra+1 pa+1))
3114 (setcar paren+1s ren+1)))
3115
3116 (if (and pa+1 (> pa+1 ren+1))
3117 ;; We've just entered a deeper nesting level.
3118 (progn
3119 ;; Insert the brace pair (if present) and the single open
3120 ;; paren/brace/bracket into `c-state-cache' It cannot be
3121 ;; inside a macro, except one around point, because of what
3122 ;; `c-neutralize-syntax-in-CPP' has done.
3123 (c-state-push-any-brace-pair bra+1 macro-start-or-here)
3124 ;; Insert the opening brace/bracket/paren position.
3125 (setq c-state-cache (cons (1- pa+1) c-state-cache))
3126 ;; Clear admin stuff for the next more nested part of the scan.
3127 (setq ren+1 pa+1 pa+1 nil bra+1 nil bra+1s nil)
3128 t) ; Carry on the loop
3129
3130 ;; All open p/b/b's at this nesting level, if any, have probably
3131 ;; been closed by matching/mismatching ones. We're probably
3132 ;; finished - we just need to check for having found an
3133 ;; unmatched )/}/], which we ignore. Such a )/}/] can't be in a
3134 ;; macro, due the action of `c-neutralize-syntax-in-CPP'.
3135 (c-safe (setq ren+1 (c-sc-scan-lists ren+1 1 1)))))) ; acts as loop control.
3136
3137 ;; Record the final, innermost, brace-pair if there is one.
3138 (c-state-push-any-brace-pair bra+1 macro-start-or-here)
3139
3140 ;; Determine a good pos
3141 (while (and (setq paren+1 (car paren+1s))
3142 (> (if (> paren+1 macro-start-or-here)
3143 paren+1
3144 (goto-char paren+1)
3145 (setq mstart (and (c-beginning-of-macro)
3146 (point)))
3147 (or mstart paren+1))
3148 here-bol))
3149 (setq paren+1s (cdr paren+1s)))
3150 (cond
3151 ((and paren+1 mstart)
3152 (min paren+1 mstart))
3153 (paren+1)
3154 (t from))))))
3155
3156 (defun c-remove-stale-state-cache (start-point here pps-point)
3157 ;; Remove stale entries from the `c-cache-state', i.e. those which will
3158 ;; not be in it when it is amended for position HERE. This may involve
3159 ;; replacing a CONS element for a brace pair containing HERE with its car.
3160 ;; Additionally, the "outermost" open-brace entry before HERE will be
3161 ;; converted to a cons if the matching close-brace is below HERE.
3162 ;;
3163 ;; START-POINT is a "maximal" "safe position" - there must be no open
3164 ;; parens/braces/brackets between START-POINT and HERE.
3165 ;;
3166 ;; As a second thing, calculate the result of parse-partial-sexp at
3167 ;; PPS-POINT, w.r.t. START-POINT. The motivation here is that
3168 ;; `c-state-cache-good-pos' may become PPS-POINT, but the caller may need to
3169 ;; adjust it to get outside a string/comment. (Sorry about this! The code
3170 ;; needs to be FAST).
3171 ;;
3172 ;; Return a list (GOOD-POS SCAN-BACK-POS CONS-SEPARATED PPS-STATE), where
3173 ;; o - GOOD-POS is a position where the new value `c-state-cache' is known
3174 ;; to be good (we aim for this to be as high as possible);
3175 ;; o - SCAN-BACK-POS, if not nil, indicates there may be a brace pair
3176 ;; preceding POS which needs to be recorded in `c-state-cache'. It is a
3177 ;; position to scan backwards from. It is the position of the "{" of the
3178 ;; last element to be removed from `c-state-cache', when that elt is a
3179 ;; cons, otherwise nil.
3180 ;; o - CONS-SEPARATED is t when a cons element in `c-state-cache' has been
3181 ;; replaced by its car because HERE lies inside the brace pair represented
3182 ;; by the cons.
3183 ;; o - PPS-STATE is the parse-partial-sexp state at PPS-POINT.
3184 (save-excursion
3185 (save-restriction
3186 (narrow-to-region 1 (point-max))
3187 (let* ((in-macro-start ; start of macro containing HERE or nil.
3188 (save-excursion
3189 (goto-char here)
3190 (and (c-beginning-of-macro)
3191 (point))))
3192 (start-point-actual-macro-start ; Start of macro containing
3193 ; start-point or nil
3194 (and (< start-point here)
3195 (save-excursion
3196 (goto-char start-point)
3197 (and (c-beginning-of-macro)
3198 (point)))))
3199 (start-point-actual-macro-end ; End of this macro, (maybe
3200 ; HERE), or nil.
3201 (and start-point-actual-macro-start
3202 (save-excursion
3203 (goto-char start-point-actual-macro-start)
3204 (c-end-of-macro)
3205 (point))))
3206 pps-state ; Will be 9 or 10 elements long.
3207 pos
3208 upper-lim ; ,beyond which `c-state-cache' entries are removed
3209 scan-back-pos
3210 cons-separated
3211 pair-beg pps-point-state target-depth)
3212
3213 ;; Remove entries beyond HERE. Also remove any entries inside
3214 ;; a macro, unless HERE is in the same macro.
3215 (setq upper-lim
3216 (if (or (null c-state-old-cpp-beg)
3217 (and (> here c-state-old-cpp-beg)
3218 (< here c-state-old-cpp-end)))
3219 here
3220 (min here c-state-old-cpp-beg)))
3221 (while (and c-state-cache (>= (c-state-cache-top-lparen) upper-lim))
3222 (setq scan-back-pos (car-safe (car c-state-cache)))
3223 (setq c-state-cache (cdr c-state-cache)))
3224
3225 ;; If `upper-lim' is inside the last recorded brace pair, remove its
3226 ;; RBrace and indicate we'll need to search backwards for a previous
3227 ;; brace pair.
3228 (when (and c-state-cache
3229 (consp (car c-state-cache))
3230 (> (cdar c-state-cache) upper-lim))
3231 (setcar c-state-cache (caar c-state-cache))
3232 (setq scan-back-pos (car c-state-cache)
3233 cons-separated t))
3234
3235 ;; The next loop jumps forward out of a nested level of parens each
3236 ;; time round; the corresponding elements in `c-state-cache' are
3237 ;; removed. `pos' is just after the brace-pair or the open paren at
3238 ;; (car c-state-cache). There can be no open parens/braces/brackets
3239 ;; between `start-point'/`start-point-actual-macro-start' and HERE,
3240 ;; due to the interface spec to this function.
3241 (setq pos (if (and start-point-actual-macro-end
3242 (not (eq start-point-actual-macro-start
3243 in-macro-start)))
3244 (1+ start-point-actual-macro-end) ; get outside the macro as
3245 ; marked by a `category' text property.
3246 start-point))
3247 (goto-char pos)
3248 (while (and c-state-cache
3249 (or (numberp (car c-state-cache)) ; Have we a { at all?
3250 (cdr c-state-cache))
3251 (< (point) here))
3252 (cond
3253 ((null pps-state) ; first time through
3254 (setq target-depth -1))
3255 ((eq (car pps-state) target-depth) ; found closing ),},]
3256 (setq target-depth (1- (car pps-state))))
3257 ;; Do nothing when we've merely reached pps-point.
3258 )
3259
3260 ;; Scan!
3261 (setq pps-state
3262 (c-sc-parse-partial-sexp
3263 (point) (if (< (point) pps-point) pps-point here)
3264 target-depth
3265 nil pps-state))
3266
3267 (if (= (point) pps-point)
3268 (setq pps-point-state pps-state))
3269
3270 (when (eq (car pps-state) target-depth)
3271 (setq pos (point)) ; POS is now just after an R-paren/brace.
3272 (cond
3273 ((and (consp (car c-state-cache))
3274 (eq (point) (cdar c-state-cache)))
3275 ;; We've just moved out of the paren pair containing the brace-pair
3276 ;; at (car c-state-cache). `pair-beg' is where the open paren is,
3277 ;; and is potentially where the open brace of a cons in
3278 ;; c-state-cache will be.
3279 (setq pair-beg (car-safe (cdr c-state-cache))
3280 c-state-cache (cdr-safe (cdr c-state-cache)))) ; remove {}pair + containing Lparen.
3281 ((numberp (car c-state-cache))
3282 (setq pair-beg (car c-state-cache)
3283 c-state-cache (cdr c-state-cache))) ; remove this
3284 ; containing Lparen
3285 ((numberp (cadr c-state-cache))
3286 (setq pair-beg (cadr c-state-cache)
3287 c-state-cache (cddr c-state-cache))) ; Remove a paren pair
3288 ; together with enclosed brace pair.
3289 ;; (t nil) ; Ignore an unmated Rparen.
3290 )))
3291
3292 (if (< (point) pps-point)
3293 (setq pps-state (c-sc-parse-partial-sexp
3294 (point) pps-point
3295 nil nil ; TARGETDEPTH, STOPBEFORE
3296 pps-state)))
3297
3298 ;; If the last paren pair we moved out of was actually a brace pair,
3299 ;; insert it into `c-state-cache'.
3300 (when (and pair-beg (eq (char-after pair-beg) ?{))
3301 (if (consp (car-safe c-state-cache))
3302 (setq c-state-cache (cdr c-state-cache)))
3303 (setq c-state-cache (cons (cons pair-beg pos)
3304 c-state-cache)))
3305
3306 (list pos scan-back-pos cons-separated pps-state)))))
3307
3308 (defun c-remove-stale-state-cache-backwards (here)
3309 ;; Strip stale elements of `c-state-cache' by moving backwards through the
3310 ;; buffer, and inform the caller of the scenario detected.
3311 ;;
3312 ;; HERE is the position we're setting `c-state-cache' for.
3313 ;; CACHE-POS (a locally bound variable) is just after the latest recorded
3314 ;; position in `c-state-cache' before HERE, or a position at or near
3315 ;; point-min which isn't in a literal.
3316 ;;
3317 ;; This function must only be called only when (> `c-state-cache-good-pos'
3318 ;; HERE). Usually the gap between CACHE-POS and HERE is large. It is thus
3319 ;; optimized to eliminate (or minimize) scanning between these two
3320 ;; positions.
3321 ;;
3322 ;; Return a three element list (GOOD-POS SCAN-BACK-POS FWD-FLAG), where:
3323 ;; o - GOOD-POS is a "good position", where `c-state-cache' is valid, or
3324 ;; could become so after missing elements are inserted into
3325 ;; `c-state-cache'. This is JUST AFTER an opening or closing
3326 ;; brace/paren/bracket which is already in `c-state-cache' or just before
3327 ;; one otherwise. exceptionally (when there's no such b/p/b handy) the BOL
3328 ;; before `here''s line, or the start of the literal containing it.
3329 ;; o - SCAN-BACK-POS, if non-nil, indicates there may be a brace pair
3330 ;; preceding POS which isn't recorded in `c-state-cache'. It is a position
3331 ;; to scan backwards from.
3332 ;; o - FWD-FLAG, if non-nil, indicates there may be parens/braces between
3333 ;; POS and HERE which aren't recorded in `c-state-cache'.
3334 ;;
3335 ;; The comments in this defun use "paren" to mean parenthesis or square
3336 ;; bracket (as contrasted with a brace), and "(" and ")" likewise.
3337 ;;
3338 ;; . {..} (..) (..) ( .. { } ) (...) ( .... . ..)
3339 ;; | | | | | |
3340 ;; CP E here D C good
3341 (let ((cache-pos (c-get-cache-scan-pos here)) ; highest position below HERE in cache (or 1)
3342 (pos c-state-cache-good-pos)
3343 pa ren ; positions of "(" and ")"
3344 dropped-cons ; whether the last element dropped from `c-state-cache'
3345 ; was a cons (representing a brace-pair)
3346 good-pos ; see above.
3347 lit ; (START . END) of a literal containing some point.
3348 here-lit-start here-lit-end ; bounds of literal containing `here'
3349 ; or `here' itself.
3350 here- here+ ; start/end of macro around HERE, or HERE
3351 (here-bol (c-point 'bol here))
3352 (too-far-back (max (- here c-state-cache-too-far) (point-min))))
3353
3354 ;; Remove completely irrelevant entries from `c-state-cache'.
3355 (while (and c-state-cache
3356 (>= (setq pa (c-state-cache-top-lparen)) here))
3357 (setq dropped-cons (consp (car c-state-cache)))
3358 (setq c-state-cache (cdr c-state-cache))
3359 (setq pos pa))
3360 ;; At this stage, (>= pos here);
3361 ;; (< (c-state-cache-top-lparen) here) (or is nil).
3362
3363 (cond
3364 ((and (consp (car c-state-cache))
3365 (> (cdar c-state-cache) here))
3366 ;; CASE 1: The top of the cache is a brace pair which now encloses
3367 ;; `here'. As good-pos, return the address. of the "{". Since we've no
3368 ;; knowledge of what's inside these braces, we have no alternative but
3369 ;; to direct the caller to scan the buffer from the opening brace.
3370 (setq pos (caar c-state-cache))
3371 (setcar c-state-cache pos)
3372 (list (1+ pos) pos t)) ; return value. We've just converted a brace pair
3373 ; entry into a { entry, so the caller needs to
3374 ; search for a brace pair before the {.
3375
3376 ;; `here' might be inside a literal. Check for this.
3377 ((progn
3378 (setq lit (c-state-literal-at here)
3379 here-lit-start (or (car lit) here)
3380 here-lit-end (or (cdr lit) here))
3381 ;; Has `here' just "newly entered" a macro?
3382 (save-excursion
3383 (goto-char here-lit-start)
3384 (if (and (c-beginning-of-macro)
3385 (or (null c-state-old-cpp-beg)
3386 (not (= (point) c-state-old-cpp-beg))))
3387 (progn
3388 (setq here- (point))
3389 (c-end-of-macro)
3390 (setq here+ (point)))
3391 (setq here- here-lit-start
3392 here+ here-lit-end)))
3393
3394 ;; `here' might be nested inside any depth of parens (or brackets but
3395 ;; not braces). Scan backwards to find the outermost such opening
3396 ;; paren, if there is one. This will be the scan position to return.
3397 (save-restriction
3398 (narrow-to-region cache-pos (point-max))
3399 (setq pos (c-state-balance-parens-backwards here- here+ pos)))
3400 nil)) ; for the cond
3401
3402 ((< pos here-lit-start)
3403 ;; CASE 2: Address of outermost ( or [ which now encloses `here', but
3404 ;; didn't enclose the (previous) `c-state-cache-good-pos'. If there is
3405 ;; a brace pair preceding this, it will already be in `c-state-cache',
3406 ;; unless there was a brace pair after it, i.e. there'll only be one to
3407 ;; scan for if we've just deleted one.
3408 (list pos (and dropped-cons pos) t)) ; Return value.
3409
3410 ;; `here' isn't enclosed in a (previously unrecorded) bracket/paren.
3411 ;; Further forward scanning isn't needed, but we still need to find a
3412 ;; GOOD-POS. Step out of all enclosing "("s on HERE's line.
3413 ((progn
3414 (save-restriction
3415 (narrow-to-region here-bol (point-max))
3416 (setq pos here-lit-start)
3417 (c-safe (while (setq pa (c-sc-scan-lists pos -1 1))
3418 (setq pos pa)))) ; might signal
3419 nil)) ; for the cond
3420
3421 ((save-restriction
3422 (narrow-to-region too-far-back (point-max))
3423 (setq ren (c-safe (c-sc-scan-lists pos -1 -1))))
3424 ;; CASE 3: After a }/)/] before `here''s BOL.
3425 (list (1+ ren) (and dropped-cons pos) nil)) ; Return value
3426
3427 ((progn (setq good-pos (c-state-lit-beg (c-point 'bopl here-bol)))
3428 (>= cache-pos good-pos))
3429 ;; CASE 3.5: Just after an existing entry in `c-state-cache' on `here''s
3430 ;; line or the previous line.
3431 (list cache-pos nil nil))
3432
3433 (t
3434 ;; CASE 4; Best of a bad job: BOL before `here-bol', or beginning of
3435 ;; literal containing it.
3436 (list good-pos (and dropped-cons good-pos) nil)))))
3437
3438
3439 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
3440 ;; Externally visible routines.
3441
3442 (defun c-state-cache-init ()
3443 (setq c-state-cache nil
3444 c-state-cache-good-pos 1
3445 c-state-nonlit-pos-cache nil
3446 c-state-nonlit-pos-cache-limit 1
3447 c-state-semi-nonlit-pos-cache nil
3448 c-state-semi-nonlit-pos-cache-limit 1
3449 c-state-brace-pair-desert nil
3450 c-state-point-min 1
3451 c-state-point-min-lit-type nil
3452 c-state-point-min-lit-start nil
3453 c-state-min-scan-pos 1
3454 c-state-old-cpp-beg nil
3455 c-state-old-cpp-end nil)
3456 (c-state-mark-point-min-literal))
3457
3458 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
3459 ;; Debugging routines to dump `c-state-cache' in a "replayable" form.
3460 ;; (defmacro c-sc-de (elt) ; "c-state-cache-dump-element"
3461 ;; `(format ,(concat "(setq " (symbol-name elt) " %s) ") ,elt))
3462 ;; (defmacro c-sc-qde (elt) ; "c-state-cache-quote-dump-element"
3463 ;; `(format ,(concat "(setq " (symbol-name elt) " '%s) ") ,elt))
3464 ;; (defun c-state-dump ()
3465 ;; ;; For debugging.
3466 ;; ;(message
3467 ;; (concat
3468 ;; (c-sc-qde c-state-cache)
3469 ;; (c-sc-de c-state-cache-good-pos)
3470 ;; (c-sc-qde c-state-nonlit-pos-cache)
3471 ;; (c-sc-de c-state-nonlit-pos-cache-limit)
3472 ;; (c-sc-qde c-state-brace-pair-desert)
3473 ;; (c-sc-de c-state-point-min)
3474 ;; (c-sc-de c-state-point-min-lit-type)
3475 ;; (c-sc-de c-state-point-min-lit-start)
3476 ;; (c-sc-de c-state-min-scan-pos)
3477 ;; (c-sc-de c-state-old-cpp-beg)
3478 ;; (c-sc-de c-state-old-cpp-end)))
3479 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
3480
3481 (defun c-invalidate-state-cache-1 (here)
3482 ;; Invalidate all info on `c-state-cache' that applies to the buffer at HERE
3483 ;; or higher and set `c-state-cache-good-pos' accordingly. The cache is
3484 ;; left in a consistent state.
3485 ;;
3486 ;; This is much like `c-whack-state-after', but it never changes a paren
3487 ;; pair element into an open paren element. Doing that would mean that the
3488 ;; new open paren wouldn't have the required preceding paren pair element.
3489 ;;
3490 ;; This function is called from c-before-change.
3491
3492 ;; The caches of non-literals:
3493 ;; Note that we use "<=" for the possibility of the second char of a two-char
3494 ;; comment opener being typed; this would invalidate any cache position at
3495 ;; HERE.
3496 (if (<= here c-state-nonlit-pos-cache-limit)
3497 (setq c-state-nonlit-pos-cache-limit (1- here)))
3498 (c-truncate-semi-nonlit-pos-cache here)
3499
3500 ;; `c-state-cache':
3501 ;; Case 1: if `here' is in a literal containing point-min, everything
3502 ;; becomes (or is already) nil.
3503 (if (or (null c-state-cache-good-pos)
3504 (< here (c-state-get-min-scan-pos)))
3505 (setq c-state-cache nil
3506 c-state-cache-good-pos nil
3507 c-state-min-scan-pos nil)
3508
3509 ;; Truncate `c-state-cache' and set `c-state-cache-good-pos' to a value
3510 ;; below `here'. To maintain its consistency, we may need to insert a new
3511 ;; brace pair.
3512 (let ((here-bol (c-point 'bol here))
3513 too-high-pa ; recorded {/(/[ next above or just below here, or nil.
3514 dropped-cons ; was the last removed element a brace pair?
3515 pa)
3516 ;; The easy bit - knock over-the-top bits off `c-state-cache'.
3517 (while (and c-state-cache
3518 (>= (setq pa (c-state-cache-top-paren)) here))
3519 (setq dropped-cons (consp (car c-state-cache))
3520 too-high-pa (c-state-cache-top-lparen)
3521 c-state-cache (cdr c-state-cache)))
3522
3523 ;; Do we need to add in an earlier brace pair, having lopped one off?
3524 (if (and dropped-cons
3525 (<= too-high-pa here))
3526 (c-append-lower-brace-pair-to-state-cache too-high-pa here here-bol))
3527 (setq c-state-cache-good-pos (or (c-state-cache-after-top-paren)
3528 (c-state-get-min-scan-pos)))))
3529
3530 ;; The brace-pair desert marker:
3531 (when (car c-state-brace-pair-desert)
3532 (if (< here (car c-state-brace-pair-desert))
3533 (setq c-state-brace-pair-desert nil)
3534 (if (< here (cdr c-state-brace-pair-desert))
3535 (setcdr c-state-brace-pair-desert here)))))
3536
3537 (defun c-parse-state-1 ()
3538 ;; Find and record all noteworthy parens between some good point earlier in
3539 ;; the file and point. That good point is at least the beginning of the
3540 ;; top-level construct we are in, or the beginning of the preceding
3541 ;; top-level construct if we aren't in one.
3542 ;;
3543 ;; The returned value is a list of the noteworthy parens with the last one
3544 ;; first. If an element in the list is an integer, it's the position of an
3545 ;; open paren (of any type) which has not been closed before the point. If
3546 ;; an element is a cons, it gives the position of a closed BRACE paren
3547 ;; pair[*]; the car is the start brace position and the cdr is the position
3548 ;; following the closing brace. Only the last closed brace paren pair
3549 ;; before each open paren and before the point is recorded, and thus the
3550 ;; state never contains two cons elements in succession. When a close brace
3551 ;; has no matching open brace (e.g., the matching brace is outside the
3552 ;; visible region), it is not represented in the returned value.
3553 ;;
3554 ;; [*] N.B. The close "brace" might be a mismatching close bracket or paren.
3555 ;; This defun explicitly treats mismatching parens/braces/brackets as
3556 ;; matching. It is the open brace which makes it a "brace" pair.
3557 ;;
3558 ;; If POINT is within a macro, open parens and brace pairs within
3559 ;; THIS macro MIGHT be recorded. This depends on whether their
3560 ;; syntactic properties have been suppressed by
3561 ;; `c-neutralize-syntax-in-CPP'. This might need fixing (2008-12-11).
3562 ;;
3563 ;; Currently no characters which are given paren syntax with the
3564 ;; syntax-table property are recorded, i.e. angle bracket arglist
3565 ;; parens are never present here. Note that this might change.
3566 ;;
3567 ;; BUG: This function doesn't cope entirely well with unbalanced
3568 ;; parens in macros. (2008-12-11: this has probably been resolved
3569 ;; by the function `c-neutralize-syntax-in-CPP'.) E.g. in the
3570 ;; following case the brace before the macro isn't balanced with the
3571 ;; one after it:
3572 ;;
3573 ;; {
3574 ;; #define X {
3575 ;; }
3576 ;;
3577 ;; Note to maintainers: this function DOES get called with point
3578 ;; within comments and strings, so don't assume it doesn't!
3579 ;;
3580 ;; This function might do hidden buffer changes.
3581 (let* ((here (point))
3582 (here-bopl (c-point 'bopl))
3583 strategy ; 'forward, 'backward etc..
3584 ;; Candidate positions to start scanning from:
3585 cache-pos ; highest position below HERE already existing in
3586 ; cache (or 1).
3587 good-pos
3588 start-point ; (when scanning forward) a place below HERE where there
3589 ; are no open parens/braces between it and HERE.
3590 bopl-state
3591 res
3592 cons-separated
3593 scan-backward-pos scan-forward-p) ; used for 'backward.
3594 ;; If POINT-MIN has changed, adjust the cache
3595 (unless (= (point-min) c-state-point-min)
3596 (c-renarrow-state-cache))
3597
3598 ;; Strategy?
3599 (setq res (c-parse-state-get-strategy here c-state-cache-good-pos)
3600 strategy (car res)
3601 start-point (cadr res))
3602
3603 (when (eq strategy 'BOD)
3604 (setq c-state-cache nil
3605 c-state-cache-good-pos start-point))
3606
3607 ;; SCAN!
3608 (cond
3609 ((memq strategy '(forward back-and-forward BOD))
3610 (setq res (c-remove-stale-state-cache start-point here here-bopl))
3611 (setq cache-pos (car res)
3612 scan-backward-pos (cadr res)
3613 cons-separated (car (cddr res))
3614 bopl-state (cadr (cddr res))) ; will be nil if (< here-bopl
3615 ; start-point)
3616 (if (and scan-backward-pos
3617 (or cons-separated (eq strategy 'forward))) ;scan-backward-pos
3618 (c-append-lower-brace-pair-to-state-cache scan-backward-pos here))
3619 (setq good-pos
3620 (c-append-to-state-cache cache-pos here))
3621 (setq c-state-cache-good-pos
3622 (if (and bopl-state
3623 (< good-pos (- here c-state-cache-too-far)))
3624 (c-state-cache-non-literal-place here-bopl bopl-state)
3625 good-pos)))
3626
3627 ((eq strategy 'backward)
3628 (setq res (c-remove-stale-state-cache-backwards here)
3629 good-pos (car res)
3630 scan-backward-pos (cadr res)
3631 scan-forward-p (car (cddr res)))
3632 (if scan-backward-pos
3633 (c-append-lower-brace-pair-to-state-cache scan-backward-pos here))
3634 (setq c-state-cache-good-pos
3635 (if scan-forward-p
3636 (c-append-to-state-cache good-pos here)
3637 good-pos)))
3638
3639 (t ; (eq strategy 'IN-LIT)
3640 (setq c-state-cache nil
3641 c-state-cache-good-pos nil))))
3642
3643 c-state-cache)
3644
3645 (defun c-invalidate-state-cache (here)
3646 ;; This is a wrapper over `c-invalidate-state-cache-1'.
3647 ;;
3648 ;; It suppresses the syntactic effect of the < and > (template) brackets and
3649 ;; of all parens in preprocessor constructs, except for any such construct
3650 ;; containing point. We can then call `c-invalidate-state-cache-1' without
3651 ;; worrying further about macros and template delimiters.
3652 (if (eval-when-compile (memq 'category-properties c-emacs-features))
3653 ;; Emacs
3654 (c-with-<->-as-parens-suppressed
3655 (if (and c-state-old-cpp-beg
3656 (< c-state-old-cpp-beg here))
3657 (c-with-all-but-one-cpps-commented-out
3658 c-state-old-cpp-beg
3659 c-state-old-cpp-end
3660 (c-invalidate-state-cache-1 here))
3661 (c-with-cpps-commented-out
3662 (c-invalidate-state-cache-1 here))))
3663 ;; XEmacs
3664 (c-invalidate-state-cache-1 here)))
3665
3666 (defmacro c-state-maybe-marker (place marker)
3667 ;; If PLACE is non-nil, return a marker marking it, otherwise nil.
3668 ;; We (re)use MARKER.
3669 `(and ,place
3670 (or ,marker (setq ,marker (make-marker)))
3671 (set-marker ,marker ,place)))
3672
3673 (defun c-parse-state ()
3674 ;; This is a wrapper over `c-parse-state-1'. See that function for a
3675 ;; description of the functionality and return value.
3676 ;;
3677 ;; It suppresses the syntactic effect of the < and > (template) brackets and
3678 ;; of all parens in preprocessor constructs, except for any such construct
3679 ;; containing point. We can then call `c-parse-state-1' without worrying
3680 ;; further about macros and template delimiters.
3681 (let (here-cpp-beg here-cpp-end)
3682 (save-excursion
3683 (when (c-beginning-of-macro)
3684 (setq here-cpp-beg (point))
3685 (unless
3686 (> (setq here-cpp-end (c-syntactic-end-of-macro))
3687 here-cpp-beg)
3688 (setq here-cpp-beg nil here-cpp-end nil))))
3689 ;; FIXME!!! Put in a `condition-case' here to protect the integrity of the
3690 ;; subsystem.
3691 (prog1
3692 (if (eval-when-compile (memq 'category-properties c-emacs-features))
3693 ;; Emacs
3694 (c-with-<->-as-parens-suppressed
3695 (if (and here-cpp-beg (> here-cpp-end here-cpp-beg))
3696 (c-with-all-but-one-cpps-commented-out
3697 here-cpp-beg here-cpp-end
3698 (c-parse-state-1))
3699 (c-with-cpps-commented-out
3700 (c-parse-state-1))))
3701 ;; XEmacs
3702 (c-parse-state-1))
3703 (setq c-state-old-cpp-beg
3704 (c-state-maybe-marker here-cpp-beg c-state-old-cpp-beg-marker)
3705 c-state-old-cpp-end
3706 (c-state-maybe-marker here-cpp-end c-state-old-cpp-end-marker)))))
3707
3708 ;; Debug tool to catch cache inconsistencies. This is called from
3709 ;; 000tests.el.
3710 (defvar c-debug-parse-state nil)
3711 (unless (fboundp 'c-real-parse-state)
3712 (fset 'c-real-parse-state (symbol-function 'c-parse-state)))
3713 (cc-bytecomp-defun c-real-parse-state)
3714
3715 (defvar c-parse-state-point nil)
3716 (defvar c-parse-state-state nil)
3717 (make-variable-buffer-local 'c-parse-state-state)
3718 (defun c-record-parse-state-state ()
3719 (setq c-parse-state-point (point))
3720 (when (markerp (cdr (assq 'c-state-old-cpp-beg c-parse-state-state)))
3721 (move-marker (cdr (assq 'c-state-old-cpp-beg c-parse-state-state)) nil)
3722 (move-marker (cdr (assq 'c-state-old-cpp-end c-parse-state-state)) nil))
3723 (setq c-parse-state-state
3724 (mapcar
3725 (lambda (arg)
3726 (let ((val (symbol-value arg)))
3727 (cons arg
3728 (cond ((consp val) (copy-tree val))
3729 ((markerp val) (copy-marker val))
3730 (t val)))))
3731 '(c-state-cache
3732 c-state-cache-good-pos
3733 c-state-nonlit-pos-cache
3734 c-state-nonlit-pos-cache-limit
3735 c-state-semi-nonlit-pos-cache
3736 c-state-semi-nonlit-pos-cache-limit
3737 c-state-brace-pair-desert
3738 c-state-point-min
3739 c-state-point-min-lit-type
3740 c-state-point-min-lit-start
3741 c-state-min-scan-pos
3742 c-state-old-cpp-beg
3743 c-state-old-cpp-end
3744 c-parse-state-point))))
3745 (defun c-replay-parse-state-state ()
3746 (message "%s"
3747 (concat "(setq "
3748 (mapconcat
3749 (lambda (arg)
3750 (format "%s %s%s" (car arg)
3751 (if (atom (cdr arg)) "" "'")
3752 (if (markerp (cdr arg))
3753 (format "(copy-marker %s)" (marker-position (cdr arg)))
3754 (cdr arg))))
3755 c-parse-state-state " ")
3756 ")")))
3757
3758 (defun c-debug-parse-state-double-cons (state)
3759 (let (state-car conses-not-ok)
3760 (while state
3761 (setq state-car (car state)
3762 state (cdr state))
3763 (if (and (consp state-car)
3764 (consp (car state)))
3765 (setq conses-not-ok t)))
3766 conses-not-ok))
3767
3768 (defun c-debug-parse-state ()
3769 (let ((here (point)) (min-point (point-min)) (res1 (c-real-parse-state)) res2)
3770 (let ((c-state-cache nil)
3771 (c-state-cache-good-pos 1)
3772 (c-state-nonlit-pos-cache nil)
3773 (c-state-nonlit-pos-cache-limit 1)
3774 (c-state-brace-pair-desert nil)
3775 (c-state-point-min 1)
3776 (c-state-point-min-lit-type nil)
3777 (c-state-point-min-lit-start nil)
3778 (c-state-min-scan-pos 1)
3779 (c-state-old-cpp-beg nil)
3780 (c-state-old-cpp-end nil))
3781 (setq res2 (c-real-parse-state)))
3782 (unless (equal res1 res2)
3783 ;; The cache can actually go further back due to the ad-hoc way
3784 ;; the first paren is found, so try to whack off a bit of its
3785 ;; start before complaining.
3786 ;; (save-excursion
3787 ;; (goto-char (or (c-least-enclosing-brace res2) (point)))
3788 ;; (c-beginning-of-defun-1)
3789 ;; (while (not (or (bobp) (eq (char-after) ?{)))
3790 ;; (c-beginning-of-defun-1))
3791 ;; (unless (equal (c-whack-state-before (point) res1) res2)
3792 ;; (message (concat "c-parse-state inconsistency at %s: "
3793 ;; "using cache: %s, from scratch: %s")
3794 ;; here res1 res2)))
3795 (message (concat "c-parse-state inconsistency at %s: "
3796 "using cache: %s, from scratch: %s. POINT-MIN: %s")
3797 here res1 res2 min-point)
3798 (message "Old state:")
3799 (c-replay-parse-state-state))
3800
3801 (when (c-debug-parse-state-double-cons res1)
3802 (message "c-parse-state INVALIDITY at %s: %s"
3803 here res1)
3804 (message "Old state:")
3805 (c-replay-parse-state-state))
3806
3807 (c-record-parse-state-state)
3808 res2 ; res1 correct a cascading series of errors ASAP
3809 ))
3810
3811 (defun c-toggle-parse-state-debug (&optional arg)
3812 (interactive "P")
3813 (setq c-debug-parse-state (c-calculate-state arg c-debug-parse-state))
3814 (fset 'c-parse-state (symbol-function (if c-debug-parse-state
3815 'c-debug-parse-state
3816 'c-real-parse-state)))
3817 (c-keep-region-active)
3818 (message "c-debug-parse-state %sabled"
3819 (if c-debug-parse-state "en" "dis")))
3820 (when c-debug-parse-state
3821 (c-toggle-parse-state-debug 1))
3822
3823 \f
3824 (defun c-whack-state-before (bufpos paren-state)
3825 ;; Whack off any state information from PAREN-STATE which lies
3826 ;; before BUFPOS. Not destructive on PAREN-STATE.
3827 (let* ((newstate (list nil))
3828 (ptr newstate)
3829 car)
3830 (while paren-state
3831 (setq car (car paren-state)
3832 paren-state (cdr paren-state))
3833 (if (< (if (consp car) (car car) car) bufpos)
3834 (setq paren-state nil)
3835 (setcdr ptr (list car))
3836 (setq ptr (cdr ptr))))
3837 (cdr newstate)))
3838
3839 (defun c-whack-state-after (bufpos paren-state)
3840 ;; Whack off any state information from PAREN-STATE which lies at or
3841 ;; after BUFPOS. Not destructive on PAREN-STATE.
3842 (catch 'done
3843 (while paren-state
3844 (let ((car (car paren-state)))
3845 (if (consp car)
3846 ;; just check the car, because in a balanced brace
3847 ;; expression, it must be impossible for the corresponding
3848 ;; close brace to be before point, but the open brace to
3849 ;; be after.
3850 (if (<= bufpos (car car))
3851 nil ; whack it off
3852 (if (< bufpos (cdr car))
3853 ;; its possible that the open brace is before
3854 ;; bufpos, but the close brace is after. In that
3855 ;; case, convert this to a non-cons element. The
3856 ;; rest of the state is before bufpos, so we're
3857 ;; done.
3858 (throw 'done (cons (car car) (cdr paren-state)))
3859 ;; we know that both the open and close braces are
3860 ;; before bufpos, so we also know that everything else
3861 ;; on state is before bufpos.
3862 (throw 'done paren-state)))
3863 (if (<= bufpos car)
3864 nil ; whack it off
3865 ;; it's before bufpos, so everything else should too.
3866 (throw 'done paren-state)))
3867 (setq paren-state (cdr paren-state)))
3868 nil)))
3869
3870 (defun c-most-enclosing-brace (paren-state &optional bufpos)
3871 ;; Return the bufpos of the innermost enclosing open paren before
3872 ;; bufpos, or nil if none was found.
3873 (let (enclosingp)
3874 (or bufpos (setq bufpos 134217727))
3875 (while paren-state
3876 (setq enclosingp (car paren-state)
3877 paren-state (cdr paren-state))
3878 (if (or (consp enclosingp)
3879 (>= enclosingp bufpos))
3880 (setq enclosingp nil)
3881 (setq paren-state nil)))
3882 enclosingp))
3883
3884 (defun c-least-enclosing-brace (paren-state)
3885 ;; Return the bufpos of the outermost enclosing open paren, or nil
3886 ;; if none was found.
3887 (let (pos elem)
3888 (while paren-state
3889 (setq elem (car paren-state)
3890 paren-state (cdr paren-state))
3891 (if (integerp elem)
3892 (setq pos elem)))
3893 pos))
3894
3895 (defun c-safe-position (bufpos paren-state)
3896 ;; Return the closest "safe" position recorded on PAREN-STATE that
3897 ;; is higher up than BUFPOS. Return nil if PAREN-STATE doesn't
3898 ;; contain any. Return nil if BUFPOS is nil, which is useful to
3899 ;; find the closest limit before a given limit that might be nil.
3900 ;;
3901 ;; A "safe" position is a position at or after a recorded open
3902 ;; paren, or after a recorded close paren. The returned position is
3903 ;; thus either the first position after a close brace, or the first
3904 ;; position after an enclosing paren, or at the enclosing paren in
3905 ;; case BUFPOS is immediately after it.
3906 (when bufpos
3907 (let (elem)
3908 (catch 'done
3909 (while paren-state
3910 (setq elem (car paren-state))
3911 (if (consp elem)
3912 (cond ((< (cdr elem) bufpos)
3913 (throw 'done (cdr elem)))
3914 ((< (car elem) bufpos)
3915 ;; See below.
3916 (throw 'done (min (1+ (car elem)) bufpos))))
3917 (if (< elem bufpos)
3918 ;; elem is the position at and not after the opening paren, so
3919 ;; we can go forward one more step unless it's equal to
3920 ;; bufpos. This is useful in some cases avoid an extra paren
3921 ;; level between the safe position and bufpos.
3922 (throw 'done (min (1+ elem) bufpos))))
3923 (setq paren-state (cdr paren-state)))))))
3924
3925 (defun c-beginning-of-syntax ()
3926 ;; This is used for `font-lock-beginning-of-syntax-function'. It
3927 ;; goes to the closest previous point that is known to be outside
3928 ;; any string literal or comment. `c-state-cache' is used if it has
3929 ;; a position in the vicinity.
3930 (let* ((paren-state c-state-cache)
3931 elem
3932
3933 (pos (catch 'done
3934 ;; Note: Similar code in `c-safe-position'. The
3935 ;; difference is that we accept a safe position at
3936 ;; the point and don't bother to go forward past open
3937 ;; parens.
3938 (while paren-state
3939 (setq elem (car paren-state))
3940 (if (consp elem)
3941 (cond ((<= (cdr elem) (point))
3942 (throw 'done (cdr elem)))
3943 ((<= (car elem) (point))
3944 (throw 'done (car elem))))
3945 (if (<= elem (point))
3946 (throw 'done elem)))
3947 (setq paren-state (cdr paren-state)))
3948 (point-min))))
3949
3950 (if (> pos (- (point) 4000))
3951 (goto-char pos)
3952 ;; The position is far back. Try `c-beginning-of-defun-1'
3953 ;; (although we can't be entirely sure it will go to a position
3954 ;; outside a comment or string in current emacsen). FIXME:
3955 ;; Consult `syntax-ppss' here.
3956 (c-beginning-of-defun-1)
3957 (if (< (point) pos)
3958 (goto-char pos)))))
3959
3960 \f
3961 ;; Tools for scanning identifiers and other tokens.
3962
3963 (defun c-on-identifier ()
3964 "Return non-nil if the point is on or directly after an identifier.
3965 Keywords are recognized and not considered identifiers. If an
3966 identifier is detected, the returned value is its starting position.
3967 If an identifier ends at the point and another begins at it \(can only
3968 happen in Pike) then the point for the preceding one is returned.
3969
3970 Note that this function might do hidden buffer changes. See the
3971 comment at the start of cc-engine.el for more info."
3972
3973 ;; FIXME: Shouldn't this function handle "operator" in C++?
3974
3975 (save-excursion
3976 (skip-syntax-backward "w_")
3977
3978 (or
3979
3980 ;; Check for a normal (non-keyword) identifier.
3981 (and (looking-at c-symbol-start)
3982 (not (looking-at c-keywords-regexp))
3983 (point))
3984
3985 (when (c-major-mode-is 'pike-mode)
3986 ;; Handle the `<operator> syntax in Pike.
3987 (let ((pos (point)))
3988 (skip-chars-backward "-!%&*+/<=>^|~[]()")
3989 (and (if (< (skip-chars-backward "`") 0)
3990 t
3991 (goto-char pos)
3992 (eq (char-after) ?\`))
3993 (looking-at c-symbol-key)
3994 (>= (match-end 0) pos)
3995 (point))))
3996
3997 ;; Handle the "operator +" syntax in C++.
3998 (when (and c-overloadable-operators-regexp
3999 (= (c-backward-token-2 0) 0))
4000
4001 (cond ((and (looking-at c-overloadable-operators-regexp)
4002 (or (not c-opt-op-identifier-prefix)
4003 (and (= (c-backward-token-2 1) 0)
4004 (looking-at c-opt-op-identifier-prefix))))
4005 (point))
4006
4007 ((save-excursion
4008 (and c-opt-op-identifier-prefix
4009 (looking-at c-opt-op-identifier-prefix)
4010 (= (c-forward-token-2 1) 0)
4011 (looking-at c-overloadable-operators-regexp)))
4012 (point))))
4013
4014 )))
4015
4016 (defsubst c-simple-skip-symbol-backward ()
4017 ;; If the point is at the end of a symbol then skip backward to the
4018 ;; beginning of it. Don't move otherwise. Return non-nil if point
4019 ;; moved.
4020 ;;
4021 ;; This function might do hidden buffer changes.
4022 (or (< (skip-syntax-backward "w_") 0)
4023 (and (c-major-mode-is 'pike-mode)
4024 ;; Handle the `<operator> syntax in Pike.
4025 (let ((pos (point)))
4026 (if (and (< (skip-chars-backward "-!%&*+/<=>^|~[]()") 0)
4027 (< (skip-chars-backward "`") 0)
4028 (looking-at c-symbol-key)
4029 (>= (match-end 0) pos))
4030 t
4031 (goto-char pos)
4032 nil)))))
4033
4034 (defun c-beginning-of-current-token (&optional back-limit)
4035 ;; Move to the beginning of the current token. Do not move if not
4036 ;; in the middle of one. BACK-LIMIT may be used to bound the
4037 ;; backward search; if given it's assumed to be at the boundary
4038 ;; between two tokens. Return non-nil if the point is moved, nil
4039 ;; otherwise.
4040 ;;
4041 ;; This function might do hidden buffer changes.
4042 (let ((start (point)))
4043 (if (looking-at "\\w\\|\\s_")
4044 (skip-syntax-backward "w_" back-limit)
4045 (when (< (skip-syntax-backward ".()" back-limit) 0)
4046 (while (let ((pos (or (and (looking-at c-nonsymbol-token-regexp)
4047 (match-end 0))
4048 ;; `c-nonsymbol-token-regexp' should always match
4049 ;; since we've skipped backward over punctuation
4050 ;; or paren syntax, but consume one char in case
4051 ;; it doesn't so that we don't leave point before
4052 ;; some earlier incorrect token.
4053 (1+ (point)))))
4054 (if (<= pos start)
4055 (goto-char pos))))))
4056 (< (point) start)))
4057
4058 (defun c-end-of-current-token (&optional back-limit)
4059 ;; Move to the end of the current token. Do not move if not in the
4060 ;; middle of one. BACK-LIMIT may be used to bound the backward
4061 ;; search; if given it's assumed to be at the boundary between two
4062 ;; tokens. Return non-nil if the point is moved, nil otherwise.
4063 ;;
4064 ;; This function might do hidden buffer changes.
4065 (let ((start (point)))
4066 (cond ((< (skip-syntax-backward "w_" (1- start)) 0)
4067 (skip-syntax-forward "w_"))
4068 ((< (skip-syntax-backward ".()" back-limit) 0)
4069 (while (progn
4070 (if (looking-at c-nonsymbol-token-regexp)
4071 (goto-char (match-end 0))
4072 ;; `c-nonsymbol-token-regexp' should always match since
4073 ;; we've skipped backward over punctuation or paren
4074 ;; syntax, but move forward in case it doesn't so that
4075 ;; we don't leave point earlier than we started with.
4076 (forward-char))
4077 (< (point) start)))))
4078 (> (point) start)))
4079
4080 (defconst c-jump-syntax-balanced
4081 (if (memq 'gen-string-delim c-emacs-features)
4082 "\\w\\|\\s_\\|\\s(\\|\\s)\\|\\s\"\\|\\s|"
4083 "\\w\\|\\s_\\|\\s(\\|\\s)\\|\\s\""))
4084
4085 (defconst c-jump-syntax-unbalanced
4086 (if (memq 'gen-string-delim c-emacs-features)
4087 "\\w\\|\\s_\\|\\s\"\\|\\s|"
4088 "\\w\\|\\s_\\|\\s\""))
4089
4090 (defun c-forward-token-2 (&optional count balanced limit)
4091 "Move forward by tokens.
4092 A token is defined as all symbols and identifiers which aren't
4093 syntactic whitespace \(note that multicharacter tokens like \"==\" are
4094 treated properly). Point is always either left at the beginning of a
4095 token or not moved at all. COUNT specifies the number of tokens to
4096 move; a negative COUNT moves in the opposite direction. A COUNT of 0
4097 moves to the next token beginning only if not already at one. If
4098 BALANCED is true, move over balanced parens, otherwise move into them.
4099 Also, if BALANCED is true, never move out of an enclosing paren.
4100
4101 LIMIT sets the limit for the movement and defaults to the point limit.
4102 The case when LIMIT is set in the middle of a token, comment or macro
4103 is handled correctly, i.e. the point won't be left there.
4104
4105 Return the number of tokens left to move \(positive or negative). If
4106 BALANCED is true, a move over a balanced paren counts as one. Note
4107 that if COUNT is 0 and no appropriate token beginning is found, 1 will
4108 be returned. Thus, a return value of 0 guarantees that point is at
4109 the requested position and a return value less \(without signs) than
4110 COUNT guarantees that point is at the beginning of some token.
4111
4112 Note that this function might do hidden buffer changes. See the
4113 comment at the start of cc-engine.el for more info."
4114
4115 (or count (setq count 1))
4116 (if (< count 0)
4117 (- (c-backward-token-2 (- count) balanced limit))
4118
4119 (let ((jump-syntax (if balanced
4120 c-jump-syntax-balanced
4121 c-jump-syntax-unbalanced))
4122 (last (point))
4123 (prev (point)))
4124
4125 (if (zerop count)
4126 ;; If count is zero we should jump if in the middle of a token.
4127 (c-end-of-current-token))
4128
4129 (save-restriction
4130 (if limit (narrow-to-region (point-min) limit))
4131 (if (/= (point)
4132 (progn (c-forward-syntactic-ws) (point)))
4133 ;; Skip whitespace. Count this as a move if we did in
4134 ;; fact move.
4135 (setq count (max (1- count) 0)))
4136
4137 (if (eobp)
4138 ;; Moved out of bounds. Make sure the returned count isn't zero.
4139 (progn
4140 (if (zerop count) (setq count 1))
4141 (goto-char last))
4142
4143 ;; Use `condition-case' to avoid having the limit tests
4144 ;; inside the loop.
4145 (condition-case nil
4146 (while (and
4147 (> count 0)
4148 (progn
4149 (setq last (point))
4150 (cond ((looking-at jump-syntax)
4151 (goto-char (scan-sexps (point) 1))
4152 t)
4153 ((looking-at c-nonsymbol-token-regexp)
4154 (goto-char (match-end 0))
4155 t)
4156 ;; `c-nonsymbol-token-regexp' above should always
4157 ;; match if there are correct tokens. Try to
4158 ;; widen to see if the limit was set in the
4159 ;; middle of one, else fall back to treating
4160 ;; the offending thing as a one character token.
4161 ((and limit
4162 (save-restriction
4163 (widen)
4164 (looking-at c-nonsymbol-token-regexp)))
4165 nil)
4166 (t
4167 (forward-char)
4168 t))))
4169 (c-forward-syntactic-ws)
4170 (setq prev last
4171 count (1- count)))
4172 (error (goto-char last)))
4173
4174 (when (eobp)
4175 (goto-char prev)
4176 (setq count (1+ count)))))
4177
4178 count)))
4179
4180 (defun c-backward-token-2 (&optional count balanced limit)
4181 "Move backward by tokens.
4182 See `c-forward-token-2' for details."
4183
4184 (or count (setq count 1))
4185 (if (< count 0)
4186 (- (c-forward-token-2 (- count) balanced limit))
4187
4188 (or limit (setq limit (point-min)))
4189 (let ((jump-syntax (if balanced
4190 c-jump-syntax-balanced
4191 c-jump-syntax-unbalanced))
4192 (last (point)))
4193
4194 (if (zerop count)
4195 ;; The count is zero so try to skip to the beginning of the
4196 ;; current token.
4197 (if (> (point)
4198 (progn (c-beginning-of-current-token) (point)))
4199 (if (< (point) limit)
4200 ;; The limit is inside the same token, so return 1.
4201 (setq count 1))
4202
4203 ;; We're not in the middle of a token. If there's
4204 ;; whitespace after the point then we must move backward,
4205 ;; so set count to 1 in that case.
4206 (and (looking-at c-syntactic-ws-start)
4207 ;; If we're looking at a '#' that might start a cpp
4208 ;; directive then we have to do a more elaborate check.
4209 (or (/= (char-after) ?#)
4210 (not c-opt-cpp-prefix)
4211 (save-excursion
4212 (and (= (point)
4213 (progn (beginning-of-line)
4214 (looking-at "[ \t]*")
4215 (match-end 0)))
4216 (or (bobp)
4217 (progn (backward-char)
4218 (not (eq (char-before) ?\\)))))))
4219 (setq count 1))))
4220
4221 ;; Use `condition-case' to avoid having to check for buffer
4222 ;; limits in `backward-char', `scan-sexps' and `goto-char' below.
4223 (condition-case nil
4224 (while (and
4225 (> count 0)
4226 (progn
4227 (c-backward-syntactic-ws)
4228 (backward-char)
4229 (if (looking-at jump-syntax)
4230 (goto-char (scan-sexps (1+ (point)) -1))
4231 ;; This can be very inefficient if there's a long
4232 ;; sequence of operator tokens without any separation.
4233 ;; That doesn't happen in practice, anyway.
4234 (c-beginning-of-current-token))
4235 (>= (point) limit)))
4236 (setq last (point)
4237 count (1- count)))
4238 (error (goto-char last)))
4239
4240 (if (< (point) limit)
4241 (goto-char last))
4242
4243 count)))
4244
4245 (defun c-forward-token-1 (&optional count balanced limit)
4246 "Like `c-forward-token-2' but doesn't treat multicharacter operator
4247 tokens like \"==\" as single tokens, i.e. all sequences of symbol
4248 characters are jumped over character by character. This function is
4249 for compatibility only; it's only a wrapper over `c-forward-token-2'."
4250 (let ((c-nonsymbol-token-regexp "\\s."))
4251 (c-forward-token-2 count balanced limit)))
4252
4253 (defun c-backward-token-1 (&optional count balanced limit)
4254 "Like `c-backward-token-2' but doesn't treat multicharacter operator
4255 tokens like \"==\" as single tokens, i.e. all sequences of symbol
4256 characters are jumped over character by character. This function is
4257 for compatibility only; it's only a wrapper over `c-backward-token-2'."
4258 (let ((c-nonsymbol-token-regexp "\\s."))
4259 (c-backward-token-2 count balanced limit)))
4260
4261 \f
4262 ;; Tools for doing searches restricted to syntactically relevant text.
4263
4264 (defun c-syntactic-re-search-forward (regexp &optional bound noerror
4265 paren-level not-inside-token
4266 lookbehind-submatch)
4267 "Like `re-search-forward', but only report matches that are found
4268 in syntactically significant text. I.e. matches in comments, macros
4269 or string literals are ignored. The start point is assumed to be
4270 outside any comment, macro or string literal, or else the content of
4271 that region is taken as syntactically significant text.
4272
4273 If PAREN-LEVEL is non-nil, an additional restriction is added to
4274 ignore matches in nested paren sexps. The search will also not go
4275 outside the current list sexp, which has the effect that if the point
4276 should be moved to BOUND when no match is found \(i.e. NOERROR is
4277 neither nil nor t), then it will be at the closing paren if the end of
4278 the current list sexp is encountered first.
4279
4280 If NOT-INSIDE-TOKEN is non-nil, matches in the middle of tokens are
4281 ignored. Things like multicharacter operators and special symbols
4282 \(e.g. \"`()\" in Pike) are handled but currently not floating point
4283 constants.
4284
4285 If LOOKBEHIND-SUBMATCH is non-nil, it's taken as a number of a
4286 subexpression in REGEXP. The end of that submatch is used as the
4287 position to check for syntactic significance. If LOOKBEHIND-SUBMATCH
4288 isn't used or if that subexpression didn't match then the start
4289 position of the whole match is used instead. The \"look behind\"
4290 subexpression is never tested before the starting position, so it
4291 might be a good idea to include \\=\\= as a match alternative in it.
4292
4293 Optimization note: Matches might be missed if the \"look behind\"
4294 subexpression can match the end of nonwhite syntactic whitespace,
4295 i.e. the end of comments or cpp directives. This since the function
4296 skips over such things before resuming the search. It's on the other
4297 hand not safe to assume that the \"look behind\" subexpression never
4298 matches syntactic whitespace.
4299
4300 Bug: Unbalanced parens inside cpp directives are currently not handled
4301 correctly \(i.e. they don't get ignored as they should) when
4302 PAREN-LEVEL is set.
4303
4304 Note that this function might do hidden buffer changes. See the
4305 comment at the start of cc-engine.el for more info."
4306
4307 (or bound (setq bound (point-max)))
4308 (if paren-level (setq paren-level -1))
4309
4310 ;;(message "c-syntactic-re-search-forward %s %s %S" (point) bound regexp)
4311
4312 (let ((start (point))
4313 tmp
4314 ;; Start position for the last search.
4315 search-pos
4316 ;; The `parse-partial-sexp' state between the start position
4317 ;; and the point.
4318 state
4319 ;; The current position after the last state update. The next
4320 ;; `parse-partial-sexp' continues from here.
4321 (state-pos (point))
4322 ;; The position at which to check the state and the state
4323 ;; there. This is separate from `state-pos' since we might
4324 ;; need to back up before doing the next search round.
4325 check-pos check-state
4326 ;; Last position known to end a token.
4327 (last-token-end-pos (point-min))
4328 ;; Set when a valid match is found.
4329 found)
4330
4331 (condition-case err
4332 (while
4333 (and
4334 (progn
4335 (setq search-pos (point))
4336 (re-search-forward regexp bound noerror))
4337
4338 (progn
4339 (setq state (parse-partial-sexp
4340 state-pos (match-beginning 0) paren-level nil state)
4341 state-pos (point))
4342 (if (setq check-pos (and lookbehind-submatch
4343 (or (not paren-level)
4344 (>= (car state) 0))
4345 (match-end lookbehind-submatch)))
4346 (setq check-state (parse-partial-sexp
4347 state-pos check-pos paren-level nil state))
4348 (setq check-pos state-pos
4349 check-state state))
4350
4351 ;; NOTE: If we got a look behind subexpression and get
4352 ;; an insignificant match in something that isn't
4353 ;; syntactic whitespace (i.e. strings or in nested
4354 ;; parentheses), then we can never skip more than a
4355 ;; single character from the match start position
4356 ;; (i.e. `state-pos' here) before continuing the
4357 ;; search. That since the look behind subexpression
4358 ;; might match the end of the insignificant region in
4359 ;; the next search.
4360
4361 (cond
4362 ((elt check-state 7)
4363 ;; Match inside a line comment. Skip to eol. Use
4364 ;; `re-search-forward' instead of `skip-chars-forward' to get
4365 ;; the right bound behavior.
4366 (re-search-forward "[\n\r]" bound noerror))
4367
4368 ((elt check-state 4)
4369 ;; Match inside a block comment. Skip to the '*/'.
4370 (search-forward "*/" bound noerror))
4371
4372 ((and (not (elt check-state 5))
4373 (eq (char-before check-pos) ?/)
4374 (not (c-get-char-property (1- check-pos) 'syntax-table))
4375 (memq (char-after check-pos) '(?/ ?*)))
4376 ;; Match in the middle of the opener of a block or line
4377 ;; comment.
4378 (if (= (char-after check-pos) ?/)
4379 (re-search-forward "[\n\r]" bound noerror)
4380 (search-forward "*/" bound noerror)))
4381
4382 ;; The last `parse-partial-sexp' above might have
4383 ;; stopped short of the real check position if the end
4384 ;; of the current sexp was encountered in paren-level
4385 ;; mode. The checks above are always false in that
4386 ;; case, and since they can do better skipping in
4387 ;; lookbehind-submatch mode, we do them before
4388 ;; checking the paren level.
4389
4390 ((and paren-level
4391 (/= (setq tmp (car check-state)) 0))
4392 ;; Check the paren level first since we're short of the
4393 ;; syntactic checking position if the end of the
4394 ;; current sexp was encountered by `parse-partial-sexp'.
4395 (if (> tmp 0)
4396
4397 ;; Inside a nested paren sexp.
4398 (if lookbehind-submatch
4399 ;; See the NOTE above.
4400 (progn (goto-char state-pos) t)
4401 ;; Skip out of the paren quickly.
4402 (setq state (parse-partial-sexp state-pos bound 0 nil state)
4403 state-pos (point)))
4404
4405 ;; Have exited the current paren sexp.
4406 (if noerror
4407 (progn
4408 ;; The last `parse-partial-sexp' call above
4409 ;; has left us just after the closing paren
4410 ;; in this case, so we can modify the bound
4411 ;; to leave the point at the right position
4412 ;; upon return.
4413 (setq bound (1- (point)))
4414 nil)
4415 (signal 'search-failed (list regexp)))))
4416
4417 ((setq tmp (elt check-state 3))
4418 ;; Match inside a string.
4419 (if (or lookbehind-submatch
4420 (not (integerp tmp)))
4421 ;; See the NOTE above.
4422 (progn (goto-char state-pos) t)
4423 ;; Skip to the end of the string before continuing.
4424 (let ((ender (make-string 1 tmp)) (continue t))
4425 (while (if (search-forward ender bound noerror)
4426 (progn
4427 (setq state (parse-partial-sexp
4428 state-pos (point) nil nil state)
4429 state-pos (point))
4430 (elt state 3))
4431 (setq continue nil)))
4432 continue)))
4433
4434 ((save-excursion
4435 (save-match-data
4436 (c-beginning-of-macro start)))
4437 ;; Match inside a macro. Skip to the end of it.
4438 (c-end-of-macro)
4439 (cond ((<= (point) bound) t)
4440 (noerror nil)
4441 (t (signal 'search-failed (list regexp)))))
4442
4443 ((and not-inside-token
4444 (or (< check-pos last-token-end-pos)
4445 (< check-pos
4446 (save-excursion
4447 (goto-char check-pos)
4448 (save-match-data
4449 (c-end-of-current-token last-token-end-pos))
4450 (setq last-token-end-pos (point))))))
4451 ;; Inside a token.
4452 (if lookbehind-submatch
4453 ;; See the NOTE above.
4454 (goto-char state-pos)
4455 (goto-char (min last-token-end-pos bound))))
4456
4457 (t
4458 ;; A real match.
4459 (setq found t)
4460 nil)))
4461
4462 ;; Should loop to search again, but take care to avoid
4463 ;; looping on the same spot.
4464 (or (/= search-pos (point))
4465 (if (= (point) bound)
4466 (if noerror
4467 nil
4468 (signal 'search-failed (list regexp)))
4469 (forward-char)
4470 t))))
4471
4472 (error
4473 (goto-char start)
4474 (signal (car err) (cdr err))))
4475
4476 ;;(message "c-syntactic-re-search-forward done %s" (or (match-end 0) (point)))
4477
4478 (if found
4479 (progn
4480 (goto-char (match-end 0))
4481 (match-end 0))
4482
4483 ;; Search failed. Set point as appropriate.
4484 (if (eq noerror t)
4485 (goto-char start)
4486 (goto-char bound))
4487 nil)))
4488
4489 (defvar safe-pos-list) ; bound in c-syntactic-skip-backward
4490
4491 (defsubst c-ssb-lit-begin ()
4492 ;; Return the start of the literal point is in, or nil.
4493 ;; We read and write the variables `safe-pos', `safe-pos-list', `state'
4494 ;; bound in the caller.
4495
4496 ;; Use `parse-partial-sexp' from a safe position down to the point to check
4497 ;; if it's outside comments and strings.
4498 (save-excursion
4499 (let ((pos (point)) safe-pos state)
4500 ;; Pick a safe position as close to the point as possible.
4501 ;;
4502 ;; FIXME: Consult `syntax-ppss' here if our cache doesn't give a good
4503 ;; position.
4504
4505 (while (and safe-pos-list
4506 (> (car safe-pos-list) (point)))
4507 (setq safe-pos-list (cdr safe-pos-list)))
4508 (unless (setq safe-pos (car-safe safe-pos-list))
4509 (setq safe-pos (max (or (c-safe-position
4510 (point) (c-parse-state))
4511 0)
4512 (point-min))
4513 safe-pos-list (list safe-pos)))
4514
4515 ;; Cache positions along the way to use if we have to back up more. We
4516 ;; cache every closing paren on the same level. If the paren cache is
4517 ;; relevant in this region then we're typically already on the same
4518 ;; level as the target position. Note that we might cache positions
4519 ;; after opening parens in case safe-pos is in a nested list. That's
4520 ;; both uncommon and harmless.
4521 (while (progn
4522 (setq state (parse-partial-sexp
4523 safe-pos pos 0))
4524 (< (point) pos))
4525 (setq safe-pos (point)
4526 safe-pos-list (cons safe-pos safe-pos-list)))
4527
4528 ;; If the state contains the start of the containing sexp we cache that
4529 ;; position too, so that parse-partial-sexp in the next run has a bigger
4530 ;; chance of starting at the same level as the target position and thus
4531 ;; will get more good safe positions into the list.
4532 (if (elt state 1)
4533 (setq safe-pos (1+ (elt state 1))
4534 safe-pos-list (cons safe-pos safe-pos-list)))
4535
4536 (if (or (elt state 3) (elt state 4))
4537 ;; Inside string or comment. Continue search at the
4538 ;; beginning of it.
4539 (elt state 8)))))
4540
4541 (defun c-syntactic-skip-backward (skip-chars &optional limit paren-level)
4542 "Like `skip-chars-backward' but only look at syntactically relevant chars,
4543 i.e. don't stop at positions inside syntactic whitespace or string
4544 literals. Preprocessor directives are also ignored, with the exception
4545 of the one that the point starts within, if any. If LIMIT is given,
4546 it's assumed to be at a syntactically relevant position.
4547
4548 If PAREN-LEVEL is non-nil, the function won't stop in nested paren
4549 sexps, and the search will also not go outside the current paren sexp.
4550 However, if LIMIT or the buffer limit is reached inside a nested paren
4551 then the point will be left at the limit.
4552
4553 Non-nil is returned if the point moved, nil otherwise.
4554
4555 Note that this function might do hidden buffer changes. See the
4556 comment at the start of cc-engine.el for more info."
4557
4558 (c-self-bind-state-cache
4559 (let ((start (point))
4560 state-2
4561 ;; A list of syntactically relevant positions in descending
4562 ;; order. It's used to avoid scanning repeatedly over
4563 ;; potentially large regions with `parse-partial-sexp' to verify
4564 ;; each position. Used in `c-ssb-lit-begin'
4565 safe-pos-list
4566 ;; The result from `c-beginning-of-macro' at the start position or the
4567 ;; start position itself if it isn't within a macro. Evaluated on
4568 ;; demand.
4569 start-macro-beg
4570 ;; The earliest position after the current one with the same paren
4571 ;; level. Used only when `paren-level' is set.
4572 lit-beg
4573 (paren-level-pos (point)))
4574
4575 (while
4576 (progn
4577 ;; The next loop "tries" to find the end point each time round,
4578 ;; loops when it hasn't succeeded.
4579 (while
4580 (and
4581 (let ((pos (point)))
4582 (while (and
4583 (< (skip-chars-backward skip-chars limit) 0)
4584 ;; Don't stop inside a literal.
4585 (when (setq lit-beg (c-ssb-lit-begin))
4586 (goto-char lit-beg)
4587 t)))
4588 (< (point) pos))
4589
4590 (let ((pos (point)) state-2 pps-end-pos)
4591
4592 (cond
4593 ((and paren-level
4594 (save-excursion
4595 (setq state-2 (parse-partial-sexp
4596 pos paren-level-pos -1)
4597 pps-end-pos (point))
4598 (/= (car state-2) 0)))
4599 ;; Not at the right level.
4600
4601 (if (and (< (car state-2) 0)
4602 ;; We stop above if we go out of a paren.
4603 ;; Now check whether it precedes or is
4604 ;; nested in the starting sexp.
4605 (save-excursion
4606 (setq state-2
4607 (parse-partial-sexp
4608 pps-end-pos paren-level-pos
4609 nil nil state-2))
4610 (< (car state-2) 0)))
4611
4612 ;; We've stopped short of the starting position
4613 ;; so the hit was inside a nested list. Go up
4614 ;; until we are at the right level.
4615 (condition-case nil
4616 (progn
4617 (goto-char (scan-lists pos -1
4618 (- (car state-2))))
4619 (setq paren-level-pos (point))
4620 (if (and limit (>= limit paren-level-pos))
4621 (progn
4622 (goto-char limit)
4623 nil)
4624 t))
4625 (error
4626 (goto-char (or limit (point-min)))
4627 nil))
4628
4629 ;; The hit was outside the list at the start
4630 ;; position. Go to the start of the list and exit.
4631 (goto-char (1+ (elt state-2 1)))
4632 nil))
4633
4634 ((c-beginning-of-macro limit)
4635 ;; Inside a macro.
4636 (if (< (point)
4637 (or start-macro-beg
4638 (setq start-macro-beg
4639 (save-excursion
4640 (goto-char start)
4641 (c-beginning-of-macro limit)
4642 (point)))))
4643 t
4644
4645 ;; It's inside the same macro we started in so it's
4646 ;; a relevant match.
4647 (goto-char pos)
4648 nil))))))
4649
4650 (> (point)
4651 (progn
4652 ;; Skip syntactic ws afterwards so that we don't stop at the
4653 ;; end of a comment if `skip-chars' is something like "^/".
4654 (c-backward-syntactic-ws)
4655 (point)))))
4656
4657 ;; We might want to extend this with more useful return values in
4658 ;; the future.
4659 (/= (point) start))))
4660
4661 ;; The following is an alternative implementation of
4662 ;; `c-syntactic-skip-backward' that uses backward movement to keep
4663 ;; track of the syntactic context. It turned out to be generally
4664 ;; slower than the one above which uses forward checks from earlier
4665 ;; safe positions.
4666 ;;
4667 ;;(defconst c-ssb-stop-re
4668 ;; ;; The regexp matching chars `c-syntactic-skip-backward' needs to
4669 ;; ;; stop at to avoid going into comments and literals.
4670 ;; (concat
4671 ;; ;; Match comment end syntax and string literal syntax. Also match
4672 ;; ;; '/' for block comment endings (not covered by comment end
4673 ;; ;; syntax).
4674 ;; "\\s>\\|/\\|\\s\""
4675 ;; (if (memq 'gen-string-delim c-emacs-features)
4676 ;; "\\|\\s|"
4677 ;; "")
4678 ;; (if (memq 'gen-comment-delim c-emacs-features)
4679 ;; "\\|\\s!"
4680 ;; "")))
4681 ;;
4682 ;;(defconst c-ssb-stop-paren-re
4683 ;; ;; Like `c-ssb-stop-re' but also stops at paren chars.
4684 ;; (concat c-ssb-stop-re "\\|\\s(\\|\\s)"))
4685 ;;
4686 ;;(defconst c-ssb-sexp-end-re
4687 ;; ;; Regexp matching the ending syntax of a complex sexp.
4688 ;; (concat c-string-limit-regexp "\\|\\s)"))
4689 ;;
4690 ;;(defun c-syntactic-skip-backward (skip-chars &optional limit paren-level)
4691 ;; "Like `skip-chars-backward' but only look at syntactically relevant chars,
4692 ;;i.e. don't stop at positions inside syntactic whitespace or string
4693 ;;literals. Preprocessor directives are also ignored. However, if the
4694 ;;point is within a comment, string literal or preprocessor directory to
4695 ;;begin with, its contents is treated as syntactically relevant chars.
4696 ;;If LIMIT is given, it limits the backward search and the point will be
4697 ;;left there if no earlier position is found.
4698 ;;
4699 ;;If PAREN-LEVEL is non-nil, the function won't stop in nested paren
4700 ;;sexps, and the search will also not go outside the current paren sexp.
4701 ;;However, if LIMIT or the buffer limit is reached inside a nested paren
4702 ;;then the point will be left at the limit.
4703 ;;
4704 ;;Non-nil is returned if the point moved, nil otherwise.
4705 ;;
4706 ;;Note that this function might do hidden buffer changes. See the
4707 ;;comment at the start of cc-engine.el for more info."
4708 ;;
4709 ;; (save-restriction
4710 ;; (when limit
4711 ;; (narrow-to-region limit (point-max)))
4712 ;;
4713 ;; (let ((start (point)))
4714 ;; (catch 'done
4715 ;; (while (let ((last-pos (point))
4716 ;; (stop-pos (progn
4717 ;; (skip-chars-backward skip-chars)
4718 ;; (point))))
4719 ;;
4720 ;; ;; Skip back over the same region as
4721 ;; ;; `skip-chars-backward' above, but keep to
4722 ;; ;; syntactically relevant positions.
4723 ;; (goto-char last-pos)
4724 ;; (while (and
4725 ;; ;; `re-search-backward' with a single char regexp
4726 ;; ;; should be fast.
4727 ;; (re-search-backward
4728 ;; (if paren-level c-ssb-stop-paren-re c-ssb-stop-re)
4729 ;; stop-pos 'move)
4730 ;;
4731 ;; (progn
4732 ;; (cond
4733 ;; ((looking-at "\\s(")
4734 ;; ;; `paren-level' is set and we've found the
4735 ;; ;; start of the containing paren.
4736 ;; (forward-char)
4737 ;; (throw 'done t))
4738 ;;
4739 ;; ((looking-at c-ssb-sexp-end-re)
4740 ;; ;; We're at the end of a string literal or paren
4741 ;; ;; sexp (if `paren-level' is set).
4742 ;; (forward-char)
4743 ;; (condition-case nil
4744 ;; (c-backward-sexp)
4745 ;; (error
4746 ;; (goto-char limit)
4747 ;; (throw 'done t))))
4748 ;;
4749 ;; (t
4750 ;; (forward-char)
4751 ;; ;; At the end of some syntactic ws or possibly
4752 ;; ;; after a plain '/' operator.
4753 ;; (let ((pos (point)))
4754 ;; (c-backward-syntactic-ws)
4755 ;; (if (= pos (point))
4756 ;; ;; Was a plain '/' operator. Go past it.
4757 ;; (backward-char)))))
4758 ;;
4759 ;; (> (point) stop-pos))))
4760 ;;
4761 ;; ;; Now the point is either at `stop-pos' or at some
4762 ;; ;; position further back if `stop-pos' was at a
4763 ;; ;; syntactically irrelevant place.
4764 ;;
4765 ;; ;; Skip additional syntactic ws so that we don't stop
4766 ;; ;; at the end of a comment if `skip-chars' is
4767 ;; ;; something like "^/".
4768 ;; (c-backward-syntactic-ws)
4769 ;;
4770 ;; (< (point) stop-pos))))
4771 ;;
4772 ;; ;; We might want to extend this with more useful return values
4773 ;; ;; in the future.
4774 ;; (/= (point) start))))
4775
4776 \f
4777 ;; Tools for handling comments and string literals.
4778
4779 (defun c-in-literal (&optional lim detect-cpp)
4780 "Return the type of literal point is in, if any.
4781 The return value is `c' if in a C-style comment, `c++' if in a C++
4782 style comment, `string' if in a string literal, `pound' if DETECT-CPP
4783 is non-nil and in a preprocessor line, or nil if somewhere else.
4784 Optional LIM is used as the backward limit of the search. If omitted,
4785 or nil, `c-beginning-of-defun' is used.
4786
4787 The last point calculated is cached if the cache is enabled, i.e. if
4788 `c-in-literal-cache' is bound to a two element vector.
4789
4790 Note that this function might do hidden buffer changes. See the
4791 comment at the start of cc-engine.el for more info."
4792 (save-restriction
4793 (widen)
4794 (let ((lit (c-state-semi-pp-to-literal (point))))
4795 (or (cadr lit)
4796 (and detect-cpp
4797 (save-excursion (c-beginning-of-macro))
4798 'pound)))))
4799
4800 (defun c-literal-limits (&optional lim near not-in-delimiter)
4801 "Return a cons of the beginning and end positions of the comment or
4802 string surrounding point (including both delimiters), or nil if point
4803 isn't in one. If LIM is non-nil, it's used as the \"safe\" position
4804 to start parsing from. If NEAR is non-nil, then the limits of any
4805 literal next to point is returned. \"Next to\" means there's only
4806 spaces and tabs between point and the literal. The search for such a
4807 literal is done first in forward direction. If NOT-IN-DELIMITER is
4808 non-nil, the case when point is inside a starting delimiter won't be
4809 recognized. This only has effect for comments which have starting
4810 delimiters with more than one character.
4811
4812 Note that this function might do hidden buffer changes. See the
4813 comment at the start of cc-engine.el for more info."
4814
4815 (save-excursion
4816 (let*
4817 ((pos (point))
4818 (lit-limits
4819 (if lim
4820 (let ((s (parse-partial-sexp lim (point))))
4821 (when (or (nth 3 s) (nth 4 s))
4822 (cons (nth 8 s)
4823 (progn (parse-partial-sexp (point) (point-max)
4824 nil 'syntax-table
4825 s)
4826 (point)))))
4827 (let ((pp-to-lit (c-state-full-pp-to-literal pos not-in-delimiter)))
4828 (car (cddr pp-to-lit))))))
4829 (cond
4830 (lit-limits)
4831
4832 (near
4833 (goto-char pos)
4834 ;; Search forward for a literal.
4835 (skip-chars-forward " \t")
4836 (cond
4837 ((looking-at c-string-limit-regexp) ; String.
4838 (cons (point) (or (c-safe (c-forward-sexp 1) (point))
4839 (point-max))))
4840
4841 ((looking-at c-comment-start-regexp) ; Line or block comment.
4842 (cons (point) (progn (c-forward-single-comment) (point))))
4843
4844 (t
4845 ;; Search backward.
4846 (skip-chars-backward " \t")
4847
4848 (let ((end (point)) beg)
4849 (cond
4850 ((save-excursion
4851 (< (skip-syntax-backward c-string-syntax) 0)) ; String.
4852 (setq beg (c-safe (c-backward-sexp 1) (point))))
4853
4854 ((and (c-safe (forward-char -2) t)
4855 (looking-at "*/"))
4856 ;; Block comment. Due to the nature of line
4857 ;; comments, they will always be covered by the
4858 ;; normal case above.
4859 (goto-char end)
4860 (c-backward-single-comment)
4861 ;; If LIM is bogus, beg will be bogus.
4862 (setq beg (point))))
4863
4864 (if beg (cons beg end))))))
4865 ))))
4866
4867 (defun c-literal-start (&optional safe-pos)
4868 "Return the start of the string or comment surrounding point, or nil if
4869 point isn't in one. SAFE-POS, if non-nil, is a position before point which is
4870 a known \"safe position\", i.e. outside of any string or comment."
4871 (if safe-pos
4872 (let ((s (parse-partial-sexp safe-pos (point))))
4873 (and (or (nth 3 s) (nth 4 s))
4874 (nth 8 s)))
4875 (car (cddr (c-state-semi-pp-to-literal (point))))))
4876
4877 ;; In case external callers use this; it did have a docstring.
4878 (defalias 'c-literal-limits-fast 'c-literal-limits)
4879
4880 (defun c-collect-line-comments (range)
4881 "If the argument is a cons of two buffer positions (such as returned by
4882 `c-literal-limits'), and that range contains a C++ style line comment,
4883 then an extended range is returned that contains all adjacent line
4884 comments (i.e. all comments that starts in the same column with no
4885 empty lines or non-whitespace characters between them). Otherwise the
4886 argument is returned.
4887
4888 Note that this function might do hidden buffer changes. See the
4889 comment at the start of cc-engine.el for more info."
4890
4891 (save-excursion
4892 (condition-case nil
4893 (if (and (consp range) (progn
4894 (goto-char (car range))
4895 (looking-at c-line-comment-starter)))
4896 (let ((col (current-column))
4897 (beg (point))
4898 (bopl (c-point 'bopl))
4899 (end (cdr range)))
4900 ;; Got to take care in the backward direction to handle
4901 ;; comments which are preceded by code.
4902 (while (and (c-backward-single-comment)
4903 (>= (point) bopl)
4904 (looking-at c-line-comment-starter)
4905 (= col (current-column)))
4906 (setq beg (point)
4907 bopl (c-point 'bopl)))
4908 (goto-char end)
4909 (while (and (progn (skip-chars-forward " \t")
4910 (looking-at c-line-comment-starter))
4911 (= col (current-column))
4912 (prog1 (zerop (forward-line 1))
4913 (setq end (point)))))
4914 (cons beg end))
4915 range)
4916 (error range))))
4917
4918 (defun c-literal-type (range)
4919 "Convenience function that given the result of `c-literal-limits',
4920 returns nil or the type of literal that the range surrounds, one
4921 of the symbols `c', `c++' or `string'. It's much faster than using
4922 `c-in-literal' and is intended to be used when you need both the
4923 type of a literal and its limits.
4924
4925 Note that this function might do hidden buffer changes. See the
4926 comment at the start of cc-engine.el for more info."
4927
4928 (if (consp range)
4929 (save-excursion
4930 (goto-char (car range))
4931 (cond ((looking-at c-string-limit-regexp) 'string)
4932 ((or (looking-at "//") ; c++ line comment
4933 (and (looking-at "\\s<") ; comment starter
4934 (looking-at "#"))) ; awk comment.
4935 'c++)
4936 (t 'c))) ; Assuming the range is valid.
4937 range))
4938
4939 (defsubst c-determine-limit-get-base (start try-size)
4940 ;; Get a "safe place" approximately TRY-SIZE characters before START.
4941 ;; This defsubst doesn't preserve point.
4942 (let* ((pos (max (- start try-size) (point-min)))
4943 (s (c-state-semi-pp-to-literal pos)))
4944 (or (car (cddr s)) pos)))
4945
4946 (defun c-determine-limit (how-far-back &optional start try-size)
4947 ;; Return a buffer position HOW-FAR-BACK non-literal characters from START
4948 ;; (default point). This is done by going back further in the buffer then
4949 ;; searching forward for literals. The position found won't be in a
4950 ;; literal. We start searching for the sought position TRY-SIZE (default
4951 ;; twice HOW-FAR-BACK) bytes back from START. This function must be fast.
4952 ;; :-)
4953 (save-excursion
4954 (let* ((start (or start (point)))
4955 (try-size (or try-size (* 2 how-far-back)))
4956 (base (c-determine-limit-get-base start try-size))
4957 (pos base)
4958
4959 (s (parse-partial-sexp pos pos)) ; null state.
4960 stack elt size
4961 (count 0))
4962 (while (< pos start)
4963 ;; Move forward one literal each time round this loop.
4964 ;; Move forward to the start of a comment or string.
4965 (setq s (parse-partial-sexp
4966 pos
4967 start
4968 nil ; target-depth
4969 nil ; stop-before
4970 s ; state
4971 'syntax-table)) ; stop-comment
4972
4973 ;; Gather details of the non-literal-bit - starting pos and size.
4974 (setq size (- (if (or (nth 4 s) (nth 3 s))
4975 (nth 8 s)
4976 (point))
4977 pos))
4978 (if (> size 0)
4979 (setq stack (cons (cons pos size) stack)))
4980
4981 ;; Move forward to the end of the comment/string.
4982 (if (or (nth 4 s) (nth 3 s))
4983 (setq s (parse-partial-sexp
4984 (point)
4985 start
4986 nil ; target-depth
4987 nil ; stop-before
4988 s ; state
4989 'syntax-table))) ; stop-comment
4990 (setq pos (point)))
4991
4992 ;; Now try and find enough non-literal characters recorded on the stack.
4993 ;; Go back one recorded literal each time round this loop.
4994 (while (and (< count how-far-back)
4995 stack)
4996 (setq elt (car stack)
4997 stack (cdr stack))
4998 (setq count (+ count (cdr elt))))
4999
5000 ;; Have we found enough yet?
5001 (cond
5002 ((>= count how-far-back)
5003 (+ (car elt) (- count how-far-back)))
5004 ((eq base (point-min))
5005 (point-min))
5006 (t
5007 (c-determine-limit (- how-far-back count) base try-size))))))
5008
5009 (defun c-determine-+ve-limit (how-far &optional start-pos)
5010 ;; Return a buffer position about HOW-FAR non-literal characters forward
5011 ;; from START-POS (default point), which must not be inside a literal.
5012 (save-excursion
5013 (let ((pos (or start-pos (point)))
5014 (count how-far)
5015 (s (parse-partial-sexp (point) (point)))) ; null state
5016 (while (and (not (eobp))
5017 (> count 0))
5018 ;; Scan over counted characters.
5019 (setq s (parse-partial-sexp
5020 pos
5021 (min (+ pos count) (point-max))
5022 nil ; target-depth
5023 nil ; stop-before
5024 s ; state
5025 'syntax-table)) ; stop-comment
5026 (setq count (- count (- (point) pos) 1)
5027 pos (point))
5028 ;; Scan over literal characters.
5029 (if (nth 8 s)
5030 (setq s (parse-partial-sexp
5031 pos
5032 (point-max)
5033 nil ; target-depth
5034 nil ; stop-before
5035 s ; state
5036 'syntax-table) ; stop-comment
5037 pos (point))))
5038 (point))))
5039
5040 \f
5041 ;; `c-find-decl-spots' and accompanying stuff.
5042
5043 ;; Variables used in `c-find-decl-spots' to cache the search done for
5044 ;; the first declaration in the last call. When that function starts,
5045 ;; it needs to back up over syntactic whitespace to look at the last
5046 ;; token before the region being searched. That can sometimes cause
5047 ;; moves back and forth over a quite large region of comments and
5048 ;; macros, which would be repeated for each changed character when
5049 ;; we're called during fontification, since font-lock refontifies the
5050 ;; current line for each change. Thus it's worthwhile to cache the
5051 ;; first match.
5052 ;;
5053 ;; `c-find-decl-syntactic-pos' is a syntactically relevant position in
5054 ;; the syntactic whitespace less or equal to some start position.
5055 ;; There's no cached value if it's nil.
5056 ;;
5057 ;; `c-find-decl-match-pos' is the match position if
5058 ;; `c-find-decl-prefix-search' matched before the syntactic whitespace
5059 ;; at `c-find-decl-syntactic-pos', or nil if there's no such match.
5060 (defvar c-find-decl-syntactic-pos nil)
5061 (make-variable-buffer-local 'c-find-decl-syntactic-pos)
5062 (defvar c-find-decl-match-pos nil)
5063 (make-variable-buffer-local 'c-find-decl-match-pos)
5064
5065 (defsubst c-invalidate-find-decl-cache (change-min-pos)
5066 (and c-find-decl-syntactic-pos
5067 (< change-min-pos c-find-decl-syntactic-pos)
5068 (setq c-find-decl-syntactic-pos nil)))
5069
5070 ; (defface c-debug-decl-spot-face
5071 ; '((t (:background "Turquoise")))
5072 ; "Debug face to mark the spots where `c-find-decl-spots' stopped.")
5073 ; (defface c-debug-decl-sws-face
5074 ; '((t (:background "Khaki")))
5075 ; "Debug face to mark the syntactic whitespace between the declaration
5076 ; spots and the preceding token end.")
5077
5078 (defmacro c-debug-put-decl-spot-faces (match-pos decl-pos)
5079 (when (facep 'c-debug-decl-spot-face)
5080 `(c-save-buffer-state ((match-pos ,match-pos) (decl-pos ,decl-pos))
5081 (c-debug-add-face (max match-pos (point-min)) decl-pos
5082 'c-debug-decl-sws-face)
5083 (c-debug-add-face decl-pos (min (1+ decl-pos) (point-max))
5084 'c-debug-decl-spot-face))))
5085 (defmacro c-debug-remove-decl-spot-faces (beg end)
5086 (when (facep 'c-debug-decl-spot-face)
5087 `(c-save-buffer-state ()
5088 (c-debug-remove-face ,beg ,end 'c-debug-decl-spot-face)
5089 (c-debug-remove-face ,beg ,end 'c-debug-decl-sws-face))))
5090
5091 (defmacro c-find-decl-prefix-search ()
5092 ;; Macro used inside `c-find-decl-spots'. It ought to be a defun,
5093 ;; but it contains lots of free variables that refer to things
5094 ;; inside `c-find-decl-spots'. The point is left at `cfd-match-pos'
5095 ;; if there is a match, otherwise at `cfd-limit'.
5096 ;;
5097 ;; The macro moves point forward to the next putative start of a declaration
5098 ;; or cfd-limit. This decl start is the next token after a "declaration
5099 ;; prefix". The declaration prefix is the earlier of `cfd-prop-match' and
5100 ;; `cfd-re-match'. `cfd-match-pos' is set to the decl prefix.
5101 ;;
5102 ;; This macro might do hidden buffer changes.
5103
5104 '(progn
5105 ;; Find the next property match position if we haven't got one already.
5106 (unless cfd-prop-match
5107 (save-excursion
5108 (while (progn
5109 (goto-char (c-next-single-property-change
5110 (point) 'c-type nil cfd-limit))
5111 (and (< (point) cfd-limit)
5112 (not (eq (c-get-char-property (1- (point)) 'c-type)
5113 'c-decl-end)))))
5114 (setq cfd-prop-match (point))))
5115
5116 ;; Find the next `c-decl-prefix-or-start-re' match if we haven't
5117 ;; got one already.
5118 (unless cfd-re-match
5119
5120 (if (> cfd-re-match-end (point))
5121 (goto-char cfd-re-match-end))
5122
5123 ;; Each time round, the next `while' moves forward over a pseudo match
5124 ;; of `c-decl-prefix-or-start-re' which is either inside a literal, or
5125 ;; is a ":" not preceded by "public", etc.. `cfd-re-match' and
5126 ;; `cfd-re-match-end' get set.
5127 (while
5128 (progn
5129 (setq cfd-re-match-end (re-search-forward c-decl-prefix-or-start-re
5130 cfd-limit 'move))
5131 (cond
5132 ((null cfd-re-match-end)
5133 ;; No match. Finish up and exit the loop.
5134 (setq cfd-re-match cfd-limit)
5135 nil)
5136 ((c-got-face-at
5137 (if (setq cfd-re-match (match-end 1))
5138 ;; Matched the end of a token preceding a decl spot.
5139 (progn
5140 (goto-char cfd-re-match)
5141 (1- cfd-re-match))
5142 ;; Matched a token that start a decl spot.
5143 (goto-char (match-beginning 0))
5144 (point))
5145 c-literal-faces)
5146 ;; Pseudo match inside a comment or string literal. Skip out
5147 ;; of comments and string literals.
5148 (while (progn
5149 (goto-char (c-next-single-property-change
5150 (point) 'face nil cfd-limit))
5151 (and (< (point) cfd-limit)
5152 (c-got-face-at (point) c-literal-faces))))
5153 t) ; Continue the loop over pseudo matches.
5154 ((and (match-string 1)
5155 (string= (match-string 1) ":")
5156 (save-excursion
5157 (or (/= (c-backward-token-2 2) 0) ; no search limit. :-(
5158 (not (looking-at c-decl-start-colon-kwd-re)))))
5159 ;; Found a ":" which isn't part of "public:", etc.
5160 t)
5161 (t nil)))) ;; Found a real match. Exit the pseudo-match loop.
5162
5163 ;; If our match was at the decl start, we have to back up over the
5164 ;; preceding syntactic ws to set `cfd-match-pos' and to catch
5165 ;; any decl spots in the syntactic ws.
5166 (unless cfd-re-match
5167 (c-backward-syntactic-ws)
5168 (setq cfd-re-match (point))))
5169
5170 ;; Choose whichever match is closer to the start.
5171 (if (< cfd-re-match cfd-prop-match)
5172 (setq cfd-match-pos cfd-re-match
5173 cfd-re-match nil)
5174 (setq cfd-match-pos cfd-prop-match
5175 cfd-prop-match nil))
5176
5177 (goto-char cfd-match-pos)
5178
5179 (when (< cfd-match-pos cfd-limit)
5180 ;; Skip forward past comments only so we don't skip macros.
5181 (c-forward-comments)
5182 ;; Set the position to continue at. We can avoid going over
5183 ;; the comments skipped above a second time, but it's possible
5184 ;; that the comment skipping has taken us past `cfd-prop-match'
5185 ;; since the property might be used inside comments.
5186 (setq cfd-continue-pos (if cfd-prop-match
5187 (min cfd-prop-match (point))
5188 (point))))))
5189
5190 (defun c-find-decl-spots (cfd-limit cfd-decl-re cfd-face-checklist cfd-fun)
5191 ;; Call CFD-FUN for each possible spot for a declaration, cast or
5192 ;; label from the point to CFD-LIMIT.
5193 ;;
5194 ;; CFD-FUN is called with point at the start of the spot. It's passed two
5195 ;; arguments: The first is the end position of the token preceding the spot,
5196 ;; or 0 for the implicit match at bob. The second is a flag that is t when
5197 ;; the match is inside a macro. Point should be moved forward by at least
5198 ;; one token.
5199 ;;
5200 ;; If CFD-FUN adds `c-decl-end' properties somewhere below the current spot,
5201 ;; it should return non-nil to ensure that the next search will find them.
5202 ;;
5203 ;; Such a spot is:
5204 ;; o The first token after bob.
5205 ;; o The first token after the end of submatch 1 in
5206 ;; `c-decl-prefix-or-start-re' when that submatch matches. This
5207 ;; submatch is typically a (L or R) brace or paren, a ;, or a ,.
5208 ;; o The start of each `c-decl-prefix-or-start-re' match when
5209 ;; submatch 1 doesn't match. This is, for example, the keyword
5210 ;; "class" in Pike.
5211 ;; o The start of a previously recognized declaration; "recognized"
5212 ;; means that the last char of the previous token has a `c-type'
5213 ;; text property with the value `c-decl-end'; this only holds
5214 ;; when `c-type-decl-end-used' is set.
5215 ;;
5216 ;; Only a spot that match CFD-DECL-RE and whose face is in the
5217 ;; CFD-FACE-CHECKLIST list causes CFD-FUN to be called. The face
5218 ;; check is disabled if CFD-FACE-CHECKLIST is nil.
5219 ;;
5220 ;; If the match is inside a macro then the buffer is narrowed to the
5221 ;; end of it, so that CFD-FUN can investigate the following tokens
5222 ;; without matching something that begins inside a macro and ends
5223 ;; outside it. It's to avoid this work that the CFD-DECL-RE and
5224 ;; CFD-FACE-CHECKLIST checks exist.
5225 ;;
5226 ;; The spots are visited approximately in order from top to bottom.
5227 ;; It's however the positions where `c-decl-prefix-or-start-re'
5228 ;; matches and where `c-decl-end' properties are found that are in
5229 ;; order. Since the spots often are at the following token, they
5230 ;; might be visited out of order insofar as more spots are reported
5231 ;; later on within the syntactic whitespace between the match
5232 ;; positions and their spots.
5233 ;;
5234 ;; It's assumed that comments and strings are fontified in the
5235 ;; searched range.
5236 ;;
5237 ;; This is mainly used in fontification, and so has an elaborate
5238 ;; cache to handle repeated calls from the same start position; see
5239 ;; the variables above.
5240 ;;
5241 ;; All variables in this function begin with `cfd-' to avoid name
5242 ;; collision with the (dynamically bound) variables used in CFD-FUN.
5243 ;;
5244 ;; This function might do hidden buffer changes.
5245
5246 (let ((cfd-start-pos (point)) ; never changed
5247 (cfd-buffer-end (point-max))
5248 ;; The end of the token preceding the decl spot last found
5249 ;; with `c-decl-prefix-or-start-re'. `cfd-limit' if there's
5250 ;; no match.
5251 cfd-re-match
5252 ;; The end position of the last `c-decl-prefix-or-start-re'
5253 ;; match. If this is greater than `cfd-continue-pos', the
5254 ;; next regexp search is started here instead.
5255 (cfd-re-match-end (point-min))
5256 ;; The end of the last `c-decl-end' found by
5257 ;; `c-find-decl-prefix-search'. `cfd-limit' if there's no
5258 ;; match. If searching for the property isn't needed then we
5259 ;; disable it by setting it to `cfd-limit' directly.
5260 (cfd-prop-match (unless c-type-decl-end-used cfd-limit))
5261 ;; The end of the token preceding the decl spot last found by
5262 ;; `c-find-decl-prefix-search'. 0 for the implicit match at
5263 ;; bob. `cfd-limit' if there's no match. In other words,
5264 ;; this is the minimum of `cfd-re-match' and `cfd-prop-match'.
5265 (cfd-match-pos cfd-limit)
5266 ;; The position to continue searching at.
5267 cfd-continue-pos
5268 ;; The position of the last "real" token we've stopped at.
5269 ;; This can be greater than `cfd-continue-pos' when we get
5270 ;; hits inside macros or at `c-decl-end' positions inside
5271 ;; comments.
5272 (cfd-token-pos 0)
5273 ;; The end position of the last entered macro.
5274 (cfd-macro-end 0))
5275
5276 ;; Initialize by finding a syntactically relevant start position
5277 ;; before the point, and do the first `c-decl-prefix-or-start-re'
5278 ;; search unless we're at bob.
5279
5280 (let (start-in-literal start-in-macro syntactic-pos)
5281 ;; Must back up a bit since we look for the end of the previous
5282 ;; statement or declaration, which is earlier than the first
5283 ;; returned match.
5284
5285 ;; This `cond' moves back over any literals or macros. It has special
5286 ;; handling for when the region being searched is entirely within a
5287 ;; macro. It sets `cfd-continue-pos' (unless we've reached
5288 ;; `cfd-limit').
5289 (cond
5290 ;; First we need to move to a syntactically relevant position.
5291 ;; Begin by backing out of comment or string literals.
5292 ;;
5293 ;; This arm of the cond actually triggers if we're in a literal,
5294 ;; and cfd-limit is at most at BONL.
5295 ((and
5296 ;; This arm of the `and' moves backwards out of a literal when
5297 ;; the face at point is a literal face. In this case, its value
5298 ;; is always non-nil.
5299 (when (c-got-face-at (point) c-literal-faces)
5300 ;; Try to use the faces to back up to the start of the
5301 ;; literal. FIXME: What if the point is on a declaration
5302 ;; inside a comment?
5303 (while (and (not (bobp))
5304 (c-got-face-at (1- (point)) c-literal-faces))
5305 (goto-char (previous-single-property-change
5306 (point) 'face nil (point-min))))
5307
5308 ;; XEmacs doesn't fontify the quotes surrounding string
5309 ;; literals.
5310 (and (featurep 'xemacs)
5311 (eq (get-text-property (point) 'face)
5312 'font-lock-string-face)
5313 (not (bobp))
5314 (progn (backward-char)
5315 (not (looking-at c-string-limit-regexp)))
5316 (forward-char))
5317
5318 ;; Don't trust the literal to contain only literal faces
5319 ;; (the font lock package might not have fontified the
5320 ;; start of it at all, for instance) so check that we have
5321 ;; arrived at something that looks like a start or else
5322 ;; resort to `c-literal-limits'.
5323 (unless (looking-at c-literal-start-regexp)
5324 (let ((lit-start (c-literal-start)))
5325 (if lit-start (goto-char lit-start)))
5326 )
5327
5328 (setq start-in-literal (point))) ; end of `and' arm.
5329
5330 ;; The start is in a literal. If the limit is in the same
5331 ;; one we don't have to find a syntactic position etc. We
5332 ;; only check that if the limit is at or before bonl to save
5333 ;; time; it covers the by far most common case when font-lock
5334 ;; refontifies the current line only.
5335 (<= cfd-limit (c-point 'bonl cfd-start-pos))
5336 (save-excursion
5337 (goto-char cfd-start-pos)
5338 (while (progn
5339 (goto-char (c-next-single-property-change
5340 (point) 'face nil cfd-limit))
5341 (and (< (point) cfd-limit)
5342 (c-got-face-at (point) c-literal-faces))))
5343 (= (point) cfd-limit))) ; end of `cond' arm condition
5344
5345 ;; Completely inside a literal. Set up variables to trig the
5346 ;; (< cfd-continue-pos cfd-start-pos) case below and it'll
5347 ;; find a suitable start position.
5348 (setq cfd-continue-pos start-in-literal)) ; end of `cond' arm
5349
5350 ;; Check if the region might be completely inside a macro, to
5351 ;; optimize that like the completely-inside-literal above.
5352 ((save-excursion
5353 (and (= (forward-line 1) 0)
5354 (bolp) ; forward-line has funny behavior at eob.
5355 (>= (point) cfd-limit)
5356 (progn (backward-char)
5357 (eq (char-before) ?\\))))
5358 ;; (Maybe) completely inside a macro. Only need to trig the
5359 ;; (< cfd-continue-pos cfd-start-pos) case below to make it
5360 ;; set things up.
5361 (setq cfd-continue-pos (1- cfd-start-pos)
5362 start-in-macro t))
5363
5364 ;; The default arm of the `cond' moves back over any macro we're in
5365 ;; and over any syntactic WS. It sets `c-find-decl-syntactic-pos'.
5366 (t
5367 ;; Back out of any macro so we don't miss any declaration
5368 ;; that could follow after it.
5369 (when (c-beginning-of-macro)
5370 (setq start-in-macro t))
5371
5372 ;; Now we're at a proper syntactically relevant position so we
5373 ;; can use the cache. But first clear it if it applied
5374 ;; further down.
5375 (c-invalidate-find-decl-cache cfd-start-pos)
5376
5377 (setq syntactic-pos (point))
5378 (unless (eq syntactic-pos c-find-decl-syntactic-pos)
5379 ;; Don't have to do this if the cache is relevant here,
5380 ;; typically if the same line is refontified again. If
5381 ;; we're just some syntactic whitespace further down we can
5382 ;; still use the cache to limit the skipping.
5383 (c-backward-syntactic-ws c-find-decl-syntactic-pos))
5384
5385 ;; If we hit `c-find-decl-syntactic-pos' and
5386 ;; `c-find-decl-match-pos' is set then we install the cached
5387 ;; values. If we hit `c-find-decl-syntactic-pos' and
5388 ;; `c-find-decl-match-pos' is nil then we know there's no decl
5389 ;; prefix in the whitespace before `c-find-decl-syntactic-pos'
5390 ;; and so we can continue the search from this point. If we
5391 ;; didn't hit `c-find-decl-syntactic-pos' then we're now in
5392 ;; the right spot to begin searching anyway.
5393 (if (and (eq (point) c-find-decl-syntactic-pos)
5394 c-find-decl-match-pos)
5395 (setq cfd-match-pos c-find-decl-match-pos
5396 cfd-continue-pos syntactic-pos)
5397
5398 (setq c-find-decl-syntactic-pos syntactic-pos)
5399
5400 (when (if (bobp)
5401 ;; Always consider bob a match to get the first
5402 ;; declaration in the file. Do this separately instead of
5403 ;; letting `c-decl-prefix-or-start-re' match bob, so that
5404 ;; regexp always can consume at least one character to
5405 ;; ensure that we won't get stuck in an infinite loop.
5406 (setq cfd-re-match 0)
5407 (backward-char)
5408 (c-beginning-of-current-token)
5409 (< (point) cfd-limit))
5410 ;; Do an initial search now. In the bob case above it's
5411 ;; only done to search for a `c-decl-end' spot.
5412 (c-find-decl-prefix-search)) ; sets cfd-continue-pos
5413
5414 (setq c-find-decl-match-pos (and (< cfd-match-pos cfd-start-pos)
5415 cfd-match-pos))))) ; end of `cond'
5416
5417 ;; Advance `cfd-continue-pos' if it's before the start position.
5418 ;; The closest continue position that might have effect at or
5419 ;; after the start depends on what we started in. This also
5420 ;; finds a suitable start position in the special cases when the
5421 ;; region is completely within a literal or macro.
5422 (when (and cfd-continue-pos (< cfd-continue-pos cfd-start-pos))
5423
5424 (cond
5425 (start-in-macro
5426 ;; If we're in a macro then it's the closest preceding token
5427 ;; in the macro. Check this before `start-in-literal',
5428 ;; since if we're inside a literal in a macro, the preceding
5429 ;; token is earlier than any `c-decl-end' spot inside the
5430 ;; literal (comment).
5431 (goto-char (or start-in-literal cfd-start-pos))
5432 ;; The only syntactic ws in macros are comments.
5433 (c-backward-comments)
5434 (backward-char)
5435 (c-beginning-of-current-token))
5436
5437 (start-in-literal
5438 ;; If we're in a comment it can only be the closest
5439 ;; preceding `c-decl-end' position within that comment, if
5440 ;; any. Go back to the beginning of such a property so that
5441 ;; `c-find-decl-prefix-search' will find the end of it.
5442 ;; (Can't stop at the end and install it directly on
5443 ;; `cfd-prop-match' since that variable might be cleared
5444 ;; after `cfd-fun' below.)
5445 ;;
5446 ;; Note that if the literal is a string then the property
5447 ;; search will simply skip to the beginning of it right
5448 ;; away.
5449 (if (not c-type-decl-end-used)
5450 (goto-char start-in-literal)
5451 (goto-char cfd-start-pos)
5452 (while (progn
5453 (goto-char (previous-single-property-change
5454 (point) 'c-type nil start-in-literal))
5455 (and (> (point) start-in-literal)
5456 (not (eq (c-get-char-property (point) 'c-type)
5457 'c-decl-end))))))
5458
5459 (when (= (point) start-in-literal)
5460 ;; Didn't find any property inside the comment, so we can
5461 ;; skip it entirely. (This won't skip past a string, but
5462 ;; that'll be handled quickly by the next
5463 ;; `c-find-decl-prefix-search' anyway.)
5464 (c-forward-single-comment)
5465 (if (> (point) cfd-limit)
5466 (goto-char cfd-limit))))
5467
5468 (t
5469 ;; If we started in normal code, the only match that might
5470 ;; apply before the start is what we already got in
5471 ;; `cfd-match-pos' so we can continue at the start position.
5472 ;; (Note that we don't get here if the first match is below
5473 ;; it.)
5474 (goto-char cfd-start-pos))) ; end of `cond'
5475
5476 ;; Delete found matches if they are before our new continue
5477 ;; position, so that `c-find-decl-prefix-search' won't back up
5478 ;; to them later on.
5479 (setq cfd-continue-pos (point))
5480 (when (and cfd-re-match (< cfd-re-match cfd-continue-pos))
5481 (setq cfd-re-match nil))
5482 (when (and cfd-prop-match (< cfd-prop-match cfd-continue-pos))
5483 (setq cfd-prop-match nil))) ; end of `when'
5484
5485 (if syntactic-pos
5486 ;; This is the normal case and we got a proper syntactic
5487 ;; position. If there's a match then it's always outside
5488 ;; macros and comments, so advance to the next token and set
5489 ;; `cfd-token-pos'. The loop below will later go back using
5490 ;; `cfd-continue-pos' to fix declarations inside the
5491 ;; syntactic ws.
5492 (when (and cfd-match-pos (< cfd-match-pos syntactic-pos))
5493 (goto-char syntactic-pos)
5494 (c-forward-syntactic-ws)
5495 (and cfd-continue-pos
5496 (< cfd-continue-pos (point))
5497 (setq cfd-token-pos (point))))
5498
5499 ;; Have one of the special cases when the region is completely
5500 ;; within a literal or macro. `cfd-continue-pos' is set to a
5501 ;; good start position for the search, so do it.
5502 (c-find-decl-prefix-search)))
5503
5504 ;; Now loop, one decl spot per iteration. We already have the first
5505 ;; match in `cfd-match-pos'.
5506 (while (progn
5507 ;; Go forward over "false matches", one per iteration.
5508 (while (and
5509 (< cfd-match-pos cfd-limit)
5510
5511 (or
5512 ;; Kludge to filter out matches on the "<" that
5513 ;; aren't open parens, for the sake of languages
5514 ;; that got `c-recognize-<>-arglists' set.
5515 (and (eq (char-before cfd-match-pos) ?<)
5516 (not (c-get-char-property (1- cfd-match-pos)
5517 'syntax-table)))
5518
5519 ;; If `cfd-continue-pos' is less or equal to
5520 ;; `cfd-token-pos', we've got a hit inside a macro
5521 ;; that's in the syntactic whitespace before the last
5522 ;; "real" declaration we've checked. If they're equal
5523 ;; we've arrived at the declaration a second time, so
5524 ;; there's nothing to do.
5525 (= cfd-continue-pos cfd-token-pos)
5526
5527 (progn
5528 ;; If `cfd-continue-pos' is less than `cfd-token-pos'
5529 ;; we're still searching for declarations embedded in
5530 ;; the syntactic whitespace. In that case we need
5531 ;; only to skip comments and not macros, since they
5532 ;; can't be nested, and that's already been done in
5533 ;; `c-find-decl-prefix-search'.
5534 (when (> cfd-continue-pos cfd-token-pos)
5535 (c-forward-syntactic-ws)
5536 (setq cfd-token-pos (point)))
5537
5538 ;; Continue if the following token fails the
5539 ;; CFD-DECL-RE and CFD-FACE-CHECKLIST checks.
5540 (when (or (>= (point) cfd-limit)
5541 (not (looking-at cfd-decl-re))
5542 (and cfd-face-checklist
5543 (not (c-got-face-at
5544 (point) cfd-face-checklist))))
5545 (goto-char cfd-continue-pos)
5546 t)))
5547
5548 (< (point) cfd-limit)) ; end of "false matches" condition
5549 (c-find-decl-prefix-search)) ; end of "false matches" loop
5550
5551 (< (point) cfd-limit)) ; end of condition for "decl-spot" while
5552
5553 (when (and
5554 (>= (point) cfd-start-pos)
5555
5556 (progn
5557 ;; Narrow to the end of the macro if we got a hit inside
5558 ;; one, to avoid recognizing things that start inside the
5559 ;; macro and end outside it.
5560 (when (> cfd-match-pos cfd-macro-end)
5561 ;; Not in the same macro as in the previous round.
5562 (save-excursion
5563 (goto-char cfd-match-pos)
5564 (setq cfd-macro-end
5565 (if (save-excursion (and (c-beginning-of-macro)
5566 (< (point) cfd-match-pos)))
5567 (progn (c-end-of-macro)
5568 (point))
5569 0))))
5570
5571 (if (zerop cfd-macro-end)
5572 t
5573 (if (> cfd-macro-end (point))
5574 (progn (narrow-to-region (point-min) cfd-macro-end)
5575 t)
5576 ;; The matched token was the last thing in the macro,
5577 ;; so the whole match is bogus.
5578 (setq cfd-macro-end 0)
5579 nil)))) ; end of when condition
5580
5581 (c-debug-put-decl-spot-faces cfd-match-pos (point))
5582 (if (funcall cfd-fun cfd-match-pos (/= cfd-macro-end 0))
5583 (setq cfd-prop-match nil))
5584
5585 (when (/= cfd-macro-end 0)
5586 ;; Restore limits if we did macro narrowing above.
5587 (narrow-to-region (point-min) cfd-buffer-end)))
5588
5589 (goto-char cfd-continue-pos)
5590 (if (= cfd-continue-pos cfd-limit)
5591 (setq cfd-match-pos cfd-limit)
5592 (c-find-decl-prefix-search))))) ; Moves point, sets cfd-continue-pos,
5593 ; cfd-match-pos, etc.
5594
5595 \f
5596 ;; A cache for found types.
5597
5598 ;; Buffer local variable that contains an obarray with the types we've
5599 ;; found. If a declaration is recognized somewhere we record the
5600 ;; fully qualified identifier in it to recognize it as a type
5601 ;; elsewhere in the file too. This is not accurate since we do not
5602 ;; bother with the scoping rules of the languages, but in practice the
5603 ;; same name is seldom used as both a type and something else in a
5604 ;; file, and we only use this as a last resort in ambiguous cases (see
5605 ;; `c-forward-decl-or-cast-1').
5606 ;;
5607 ;; Not every type need be in this cache. However, things which have
5608 ;; ceased to be types must be removed from it.
5609 ;;
5610 ;; Template types in C++ are added here too but with the template
5611 ;; arglist replaced with "<>" in references or "<" for the one in the
5612 ;; primary type. E.g. the type "Foo<A,B>::Bar<C>" is stored as
5613 ;; "Foo<>::Bar<". This avoids storing very long strings (since C++
5614 ;; template specs can be fairly sized programs in themselves) and
5615 ;; improves the hit ratio (it's a type regardless of the template
5616 ;; args; it's just not the same type, but we're only interested in
5617 ;; recognizing types, not telling distinct types apart). Note that
5618 ;; template types in references are added here too; from the example
5619 ;; above there will also be an entry "Foo<".
5620 (defvar c-found-types nil)
5621 (make-variable-buffer-local 'c-found-types)
5622
5623 (defsubst c-clear-found-types ()
5624 ;; Clears `c-found-types'.
5625 (setq c-found-types (make-vector 53 0)))
5626
5627 (defun c-add-type (from to)
5628 ;; Add the given region as a type in `c-found-types'. If the region
5629 ;; doesn't match an existing type but there is a type which is equal
5630 ;; to the given one except that the last character is missing, then
5631 ;; the shorter type is removed. That's done to avoid adding all
5632 ;; prefixes of a type as it's being entered and font locked. This
5633 ;; doesn't cover cases like when characters are removed from a type
5634 ;; or added in the middle. We'd need the position of point when the
5635 ;; font locking is invoked to solve this well.
5636 ;;
5637 ;; This function might do hidden buffer changes.
5638 (let ((type (c-syntactic-content from to c-recognize-<>-arglists)))
5639 (unless (intern-soft type c-found-types)
5640 (unintern (substring type 0 -1) c-found-types)
5641 (intern type c-found-types))))
5642
5643 (defun c-unfind-type (name)
5644 ;; Remove the "NAME" from c-found-types, if present.
5645 (unintern name c-found-types))
5646
5647 (defsubst c-check-type (from to)
5648 ;; Return non-nil if the given region contains a type in
5649 ;; `c-found-types'.
5650 ;;
5651 ;; This function might do hidden buffer changes.
5652 (intern-soft (c-syntactic-content from to c-recognize-<>-arglists)
5653 c-found-types))
5654
5655 (defun c-list-found-types ()
5656 ;; Return all the types in `c-found-types' as a sorted list of
5657 ;; strings.
5658 (let (type-list)
5659 (mapatoms (lambda (type)
5660 (setq type-list (cons (symbol-name type)
5661 type-list)))
5662 c-found-types)
5663 (sort type-list 'string-lessp)))
5664
5665 ;; Shut up the byte compiler.
5666 (defvar c-maybe-stale-found-type)
5667
5668 (defun c-trim-found-types (beg end old-len)
5669 ;; An after change function which, in conjunction with the info in
5670 ;; c-maybe-stale-found-type (set in c-before-change), removes a type
5671 ;; from `c-found-types', should this type have become stale. For
5672 ;; example, this happens to "foo" when "foo \n bar();" becomes
5673 ;; "foo(); \n bar();". Such stale types, if not removed, foul up
5674 ;; the fontification.
5675 ;;
5676 ;; Have we, perhaps, added non-ws characters to the front/back of a found
5677 ;; type?
5678 (when (> end beg)
5679 (save-excursion
5680 (when (< end (point-max))
5681 (goto-char end)
5682 (if (and (c-beginning-of-current-token) ; only moves when we started in the middle
5683 (progn (goto-char end)
5684 (c-end-of-current-token)))
5685 (c-unfind-type (buffer-substring-no-properties
5686 end (point)))))
5687 (when (> beg (point-min))
5688 (goto-char beg)
5689 (if (and (c-end-of-current-token) ; only moves when we started in the middle
5690 (progn (goto-char beg)
5691 (c-beginning-of-current-token)))
5692 (c-unfind-type (buffer-substring-no-properties
5693 (point) beg))))))
5694
5695 (if c-maybe-stale-found-type ; e.g. (c-decl-id-start "foo" 97 107 " (* ooka) " "o")
5696 (cond
5697 ;; Changing the amount of (already existing) whitespace - don't do anything.
5698 ((and (c-partial-ws-p beg end)
5699 (or (= beg end) ; removal of WS
5700 (string-match "^[ \t\n\r\f\v]*$" (nth 5 c-maybe-stale-found-type)))))
5701
5702 ;; The syntactic relationship which defined a "found type" has been
5703 ;; destroyed.
5704 ((eq (car c-maybe-stale-found-type) 'c-decl-id-start)
5705 (c-unfind-type (cadr c-maybe-stale-found-type)))
5706 ;; ((eq (car c-maybe-stale-found-type) 'c-decl-type-start) FIXME!!!
5707 )))
5708
5709 \f
5710 ;; Setting and removing syntax properties on < and > in languages (C++
5711 ;; and Java) where they can be template/generic delimiters as well as
5712 ;; their normal meaning of "less/greater than".
5713
5714 ;; Normally, < and > have syntax 'punctuation'. When they are found to
5715 ;; be delimiters, they are marked as such with the category properties
5716 ;; c-<-as-paren-syntax, c->-as-paren-syntax respectively.
5717
5718 ;; STRATEGY:
5719 ;;
5720 ;; It is impossible to determine with certainty whether a <..> pair in
5721 ;; C++ is two comparison operators or is template delimiters, unless
5722 ;; one duplicates a lot of a C++ compiler. For example, the following
5723 ;; code fragment:
5724 ;;
5725 ;; foo (a < b, c > d) ;
5726 ;;
5727 ;; could be a function call with two integer parameters (each a
5728 ;; relational expression), or it could be a constructor for class foo
5729 ;; taking one parameter d of templated type "a < b, c >". They are
5730 ;; somewhat easier to distinguish in Java.
5731 ;;
5732 ;; The strategy now (2010-01) adopted is to mark and unmark < and
5733 ;; > IN MATCHING PAIRS ONLY. [Previously, they were marked
5734 ;; individually when their context so indicated. This gave rise to
5735 ;; intractable problems when one of a matching pair was deleted, or
5736 ;; pulled into a literal.]
5737 ;;
5738 ;; At each buffer change, the syntax-table properties are removed in a
5739 ;; before-change function and reapplied, when needed, in an
5740 ;; after-change function. It is far more important that the
5741 ;; properties get removed when they they are spurious than that they
5742 ;; be present when wanted.
5743 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
5744 (defun c-clear-<-pair-props (&optional pos)
5745 ;; POS (default point) is at a < character. If it is marked with
5746 ;; open paren syntax-table text property, remove the property,
5747 ;; together with the close paren property on the matching > (if
5748 ;; any).
5749 (save-excursion
5750 (if pos
5751 (goto-char pos)
5752 (setq pos (point)))
5753 (when (equal (c-get-char-property (point) 'syntax-table)
5754 c-<-as-paren-syntax)
5755 (with-syntax-table c-no-parens-syntax-table ; ignore unbalanced [,{,(,..
5756 (c-go-list-forward))
5757 (when (equal (c-get-char-property (1- (point)) 'syntax-table)
5758 c->-as-paren-syntax) ; should always be true.
5759 (c-unmark-<->-as-paren (1- (point))))
5760 (c-unmark-<->-as-paren pos))))
5761
5762 (defun c-clear->-pair-props (&optional pos)
5763 ;; POS (default point) is at a > character. If it is marked with
5764 ;; close paren syntax-table property, remove the property, together
5765 ;; with the open paren property on the matching < (if any).
5766 (save-excursion
5767 (if pos
5768 (goto-char pos)
5769 (setq pos (point)))
5770 (when (equal (c-get-char-property (point) 'syntax-table)
5771 c->-as-paren-syntax)
5772 (with-syntax-table c-no-parens-syntax-table ; ignore unbalanced [,{,(,..
5773 (c-go-up-list-backward))
5774 (when (equal (c-get-char-property (point) 'syntax-table)
5775 c-<-as-paren-syntax) ; should always be true.
5776 (c-unmark-<->-as-paren (point)))
5777 (c-unmark-<->-as-paren pos))))
5778
5779 (defun c-clear-<>-pair-props (&optional pos)
5780 ;; POS (default point) is at a < or > character. If it has an
5781 ;; open/close paren syntax-table property, remove this property both
5782 ;; from the current character and its partner (which will also be
5783 ;; thusly marked).
5784 (cond
5785 ((eq (char-after) ?\<)
5786 (c-clear-<-pair-props pos))
5787 ((eq (char-after) ?\>)
5788 (c-clear->-pair-props pos))
5789 (t (c-benign-error
5790 "c-clear-<>-pair-props called from wrong position"))))
5791
5792 (defun c-clear-<-pair-props-if-match-after (lim &optional pos)
5793 ;; POS (default point) is at a < character. If it is both marked
5794 ;; with open/close paren syntax-table property, and has a matching >
5795 ;; (also marked) which is after LIM, remove the property both from
5796 ;; the current > and its partner. Return t when this happens, nil
5797 ;; when it doesn't.
5798 (save-excursion
5799 (if pos
5800 (goto-char pos)
5801 (setq pos (point)))
5802 (when (equal (c-get-char-property (point) 'syntax-table)
5803 c-<-as-paren-syntax)
5804 (with-syntax-table c-no-parens-syntax-table ; ignore unbalanced [,{,(,..
5805 (c-go-list-forward))
5806 (when (and (>= (point) lim)
5807 (equal (c-get-char-property (1- (point)) 'syntax-table)
5808 c->-as-paren-syntax)) ; should always be true.
5809 (c-unmark-<->-as-paren (1- (point)))
5810 (c-unmark-<->-as-paren pos))
5811 t)))
5812
5813 (defun c-clear->-pair-props-if-match-before (lim &optional pos)
5814 ;; POS (default point) is at a > character. If it is both marked
5815 ;; with open/close paren syntax-table property, and has a matching <
5816 ;; (also marked) which is before LIM, remove the property both from
5817 ;; the current < and its partner. Return t when this happens, nil
5818 ;; when it doesn't.
5819 (save-excursion
5820 (if pos
5821 (goto-char pos)
5822 (setq pos (point)))
5823 (when (equal (c-get-char-property (point) 'syntax-table)
5824 c->-as-paren-syntax)
5825 (with-syntax-table c-no-parens-syntax-table ; ignore unbalanced [,{,(,..
5826 (c-go-up-list-backward))
5827 (when (and (<= (point) lim)
5828 (equal (c-get-char-property (point) 'syntax-table)
5829 c-<-as-paren-syntax)) ; should always be true.
5830 (c-unmark-<->-as-paren (point))
5831 (c-unmark-<->-as-paren pos))
5832 t)))
5833
5834 ;; Set by c-common-init in cc-mode.el.
5835 (defvar c-new-BEG)
5836 (defvar c-new-END)
5837 ;; Set by c-after-change in cc-mode.el.
5838 (defvar c-old-BEG)
5839 (defvar c-old-END)
5840
5841 (defun c-before-change-check-<>-operators (beg end)
5842 ;; Unmark certain pairs of "< .... >" which are currently marked as
5843 ;; template/generic delimiters. (This marking is via syntax-table text
5844 ;; properties), and expand the (c-new-BEG c-new-END) region to include all
5845 ;; unmarked < and > operators within the certain bounds (see below).
5846 ;;
5847 ;; These pairs are those which are in the current "statement" (i.e.,
5848 ;; the region between the {, }, or ; before BEG and the one after
5849 ;; END), and which enclose any part of the interval (BEG END).
5850 ;;
5851 ;; Note that in C++ (?and Java), template/generic parens cannot
5852 ;; enclose a brace or semicolon, so we use these as bounds on the
5853 ;; region we must work on.
5854 ;;
5855 ;; This function is called from before-change-functions (via
5856 ;; c-get-state-before-change-functions). Thus the buffer is widened,
5857 ;; and point is undefined, both at entry and exit.
5858 ;;
5859 ;; FIXME!!! This routine ignores the possibility of macros entirely.
5860 ;; 2010-01-29.
5861 (save-excursion
5862 (c-save-buffer-state
5863 ((beg-lit-start (progn (goto-char beg) (c-literal-start)))
5864 (end-lit-limits (progn (goto-char end) (c-literal-limits)))
5865 new-beg new-end beg-limit end-limit)
5866 ;; Locate the earliest < after the barrier before the changed region,
5867 ;; which isn't already marked as a paren.
5868 (goto-char (or beg-lit-start beg))
5869 (setq beg-limit (c-determine-limit 512))
5870
5871 ;; Remove the syntax-table/category properties from each pertinent <...>
5872 ;; pair. Firstly, the ones with the < before beg and > after beg....
5873 (while (progn (c-syntactic-skip-backward "^;{}<" beg-limit)
5874 (eq (char-before) ?<))
5875 (c-backward-token-2)
5876 (when (eq (char-after) ?<)
5877 (c-clear-<-pair-props-if-match-after beg)))
5878 (c-forward-syntactic-ws)
5879 (setq new-beg (point))
5880
5881 ;; ...Then the ones with < before end and > after end.
5882 (goto-char (if end-lit-limits (cdr end-lit-limits) end))
5883 (setq end-limit (c-determine-+ve-limit 512))
5884 (while (and (c-syntactic-re-search-forward "[;{}>]" end-limit 'end)
5885 (eq (char-before) ?>))
5886 (c-end-of-current-token)
5887 (when (eq (char-before) ?>)
5888 (c-clear->-pair-props-if-match-before end (1- (point)))))
5889 (c-backward-syntactic-ws)
5890 (setq new-end (point))
5891
5892 ;; Extend the fontification region, if needed.
5893 (and new-beg
5894 (< new-beg c-new-BEG)
5895 (setq c-new-BEG new-beg))
5896 (and new-end
5897 (> new-end c-new-END)
5898 (setq c-new-END new-end)))))
5899
5900 (defun c-after-change-check-<>-operators (beg end)
5901 ;; This is called from `after-change-functions' when
5902 ;; c-recognize-<>-arglists' is set. It ensures that no "<" or ">"
5903 ;; chars with paren syntax become part of another operator like "<<"
5904 ;; or ">=".
5905 ;;
5906 ;; This function might do hidden buffer changes.
5907
5908 (save-excursion
5909 (goto-char beg)
5910 (when (or (looking-at "[<>]")
5911 (< (skip-chars-backward "<>") 0))
5912
5913 (goto-char beg)
5914 (c-beginning-of-current-token)
5915 (when (and (< (point) beg)
5916 (looking-at c-<>-multichar-token-regexp)
5917 (< beg (setq beg (match-end 0))))
5918 (while (progn (skip-chars-forward "^<>" beg)
5919 (< (point) beg))
5920 (c-clear-<>-pair-props)
5921 (forward-char))))
5922
5923 (when (< beg end)
5924 (goto-char end)
5925 (when (or (looking-at "[<>]")
5926 (< (skip-chars-backward "<>") 0))
5927
5928 (goto-char end)
5929 (c-beginning-of-current-token)
5930 (when (and (< (point) end)
5931 (looking-at c-<>-multichar-token-regexp)
5932 (< end (setq end (match-end 0))))
5933 (while (progn (skip-chars-forward "^<>" end)
5934 (< (point) end))
5935 (c-clear-<>-pair-props)
5936 (forward-char)))))))
5937
5938 (defun c-restore-<>-properties (_beg _end _old-len)
5939 ;; This function is called as an after-change function. It restores the
5940 ;; category/syntax-table properties on template/generic <..> pairs between
5941 ;; c-new-BEG and c-new-END. It may do hidden buffer changes.
5942 (c-save-buffer-state ((c-parse-and-markup-<>-arglists t)
5943 c-restricted-<>-arglists lit-limits)
5944 (goto-char c-new-BEG)
5945 (if (setq lit-limits (c-literal-limits))
5946 (goto-char (cdr lit-limits)))
5947 (while (and (< (point) c-new-END)
5948 (c-syntactic-re-search-forward "<" c-new-END 'bound))
5949 (backward-char)
5950 (save-excursion
5951 (c-backward-token-2)
5952 (setq c-restricted-<>-arglists
5953 (and (not (looking-at c-opt-<>-sexp-key))
5954 (progn (c-backward-syntactic-ws) ; to ( or ,
5955 (and (memq (char-before) '(?\( ?,)) ; what about <?
5956 (not (eq (c-get-char-property (point) 'c-type)
5957 'c-decl-arg-start)))))))
5958 (or (c-forward-<>-arglist nil)
5959 (forward-char)))))
5960
5961 \f
5962 ;; Functions to handle C++ raw strings.
5963 ;;
5964 ;; A valid C++ raw string looks like
5965 ;; R"<id>(<contents>)<id>"
5966 ;; , where <id> is an identifier from 0 to 16 characters long, not containing
5967 ;; spaces, control characters, double quote or left/right paren. <contents>
5968 ;; can include anything which isn't the terminating )<id>", including new
5969 ;; lines, "s, parentheses, etc.
5970 ;;
5971 ;; CC Mode handles C++ raw strings by the use of `syntax-table' text
5972 ;; properties as follows:
5973 ;;
5974 ;; (i) On a validly terminated raw string, no `syntax-table' text properties
5975 ;; are applied to the opening and closing delimiters, but any " in the
5976 ;; contents is given the property value "punctuation" (`(1)') to prevent it
5977 ;; interacting with the "s in the delimiters.
5978 ;;
5979 ;; The font locking routine `c-font-lock-c++-raw-strings' (in cc-fonts.el)
5980 ;; recognizes valid raw strings, and fontifies the delimiters (apart from
5981 ;; the parentheses) with the default face and the parentheses and the
5982 ;; <contents> with font-lock-string-face.
5983 ;;
5984 ;; (ii) A valid, but unterminated, raw string opening delimiter gets the
5985 ;; "punctuation" value (`(1)') of the `syntax-table' text property, and the
5986 ;; open parenthesis gets the "string fence" value (`(15)').
5987 ;;
5988 ;; `c-font-lock-c++-raw-strings' puts c-font-lock-warning-face on the entire
5989 ;; unmatched opening delimiter (from the R up to the open paren), and allows
5990 ;; the rest of the buffer to get font-lock-string-face, caused by the
5991 ;; unmatched "string fence" `syntax-table' text property value.
5992 ;;
5993 ;; (iii) Inside a macro, a valid raw string is handled as in (i). An
5994 ;; unmatched opening delimiter is handled slightly differently. In addition
5995 ;; to the "punctuation" and "string fence" properties on the delimiter,
5996 ;; another "string fence" `syntax-table' property is applied to the last
5997 ;; possible character of the macro before the terminating linefeed (if there
5998 ;; is such a character after the "("). This "last possible" character is
5999 ;; never a backslash escaping the end of line. If the character preceding
6000 ;; this "last possible" character is itself a backslash, this preceding
6001 ;; character gets a "punctuation" `syntax-table' value. If the "(" is
6002 ;; already at the end of the macro, it gets the "punctuation" value, and no
6003 ;; "string fence"s are used.
6004 ;;
6005 ;; The effect on the fontification of either of these tactics is that rest of
6006 ;; the macro (if any) after the "(" gets font-lock-string-face, but the rest
6007 ;; of the file is fontified normally.
6008
6009
6010 (defun c-raw-string-pos ()
6011 ;; Get POINT's relationship to any containing raw string.
6012 ;; If point isn't in a raw string, return nil.
6013 ;; Otherwise, return the following list:
6014 ;;
6015 ;; (POS B\" B\( E\) E\")
6016 ;;
6017 ;; , where POS is the symbol `open-delim' if point is in the opening
6018 ;; delimiter, the symbol `close-delim' if it's in the closing delimiter, and
6019 ;; nil if it's in the string body. B\", B\(, E\), E\" are the positions of
6020 ;; the opening and closing quotes and parentheses of a correctly terminated
6021 ;; raw string. (N.B.: E\) and E\" are NOT on the "outside" of these
6022 ;; characters.) If the raw string is not terminated, E\) and E\" are set to
6023 ;; nil.
6024 ;;
6025 ;; Note: this routine is dependant upon the correct syntax-table text
6026 ;; properties being set.
6027 (let ((state (c-state-semi-pp-to-literal (point)))
6028 open-quote-pos open-paren-pos close-paren-pos close-quote-pos id)
6029 (save-excursion
6030 (when
6031 (and
6032 (cond
6033 ((null (cadr state))
6034 (or (eq (char-after) ?\")
6035 (search-backward "\"" (max (- (point) 17) (point-min)) t)))
6036 ((and (eq (cadr state) 'string)
6037 (goto-char (nth 2 state))
6038 (or (eq (char-after) ?\")
6039 (search-backward "\"" (max (- (point) 17) (point-min)) t))
6040 (not (bobp)))))
6041 (eq (char-before) ?R)
6042 (looking-at "\"\\([^ ()\\\n\r\t]\\{0,16\\}\\)("))
6043 (setq open-quote-pos (point)
6044 open-paren-pos (match-end 1)
6045 id (match-string-no-properties 1))
6046 (goto-char (1+ open-paren-pos))
6047 (when (and (not (c-get-char-property open-paren-pos 'syntax-table))
6048 (search-forward (concat ")" id "\"") nil t))
6049 (setq close-paren-pos (match-beginning 0)
6050 close-quote-pos (1- (point))))))
6051 (and open-quote-pos
6052 (list
6053 (cond
6054 ((<= (point) open-paren-pos)
6055 'open-delim)
6056 ((and close-paren-pos
6057 (> (point) close-paren-pos))
6058 'close-delim)
6059 (t nil))
6060 open-quote-pos open-paren-pos close-paren-pos close-quote-pos))))
6061
6062 (defun c-depropertize-raw-string (id open-quote open-paren bound)
6063 ;; Point is immediately after a raw string opening delimiter. Remove any
6064 ;; `syntax-table' text properties associated with the delimiter (if it's
6065 ;; unmatched) or the raw string.
6066 ;;
6067 ;; ID, a string, is the delimiter's identifier. OPEN-QUOTE and OPEN-PAREN
6068 ;; are the buffer positions of the delimiter's components. BOUND is the
6069 ;; bound for searching for a matching closing delimiter; it is usually nil,
6070 ;; but if we're inside a macro, it's the end of the macro.
6071 ;;
6072 ;; Point is moved to after the (terminated) raw string, or left after the
6073 ;; unmatched opening delimiter, as the case may be. The return value is of
6074 ;; no significance.
6075 (let ((open-paren-prop (c-get-char-property open-paren 'syntax-table)))
6076 (cond
6077 ((null open-paren-prop)
6078 ;; A terminated raw string
6079 (when (search-forward (concat ")" id "\"") nil t)
6080 (let* ((closing-paren (match-beginning 0))
6081 (first-punctuation
6082 (save-match-data
6083 (goto-char (1+ open-paren))
6084 (and (c-search-forward-char-property 'syntax-table '(1)
6085 closing-paren)
6086 (1- (point)))))
6087 )
6088 (when first-punctuation
6089 (c-clear-char-property-with-value
6090 first-punctuation (match-beginning 0) 'syntax-table '(1))
6091 (c-truncate-semi-nonlit-pos-cache first-punctuation)
6092 ))))
6093 ((or (and (equal open-paren-prop '(15)) (null bound))
6094 (equal open-paren-prop '(1)))
6095 ;; An unterminated raw string either not in a macro, or in a macro with
6096 ;; the open parenthesis right up against the end of macro
6097 (c-clear-char-property open-quote 'syntax-table)
6098 (c-truncate-semi-nonlit-pos-cache open-quote)
6099 (c-clear-char-property open-paren 'syntax-table))
6100 (t
6101 ;; An unterminated string in a macro, with at least one char after the
6102 ;; open paren
6103 (c-clear-char-property open-quote 'syntax-table)
6104 (c-truncate-semi-nonlit-pos-cache open-quote)
6105 (c-clear-char-property open-paren 'syntax-table)
6106 (let ((after-string-fence-pos
6107 (save-excursion
6108 (goto-char (1+ open-paren))
6109 (c-search-forward-char-property 'syntax-table '(15) bound))))
6110 (when after-string-fence-pos
6111 (c-clear-char-property (1- after-string-fence-pos) 'syntax-table)))
6112 ))))
6113
6114 (defun c-depropertize-raw-strings-in-region (start finish)
6115 ;; Remove any `syntax-table' text properties associated with C++ raw strings
6116 ;; contained in the region (START FINISH). Point is undefined at entry and
6117 ;; exit, and the return value has no significance.
6118 (goto-char start)
6119 (while (and (< (point) finish)
6120 (re-search-forward
6121 (concat "\\(" ; 1
6122 c-anchored-cpp-prefix ; 2
6123 "\\)\\|\\(" ; 3
6124 "R\"\\([^ ()\\\n\r\t]\\{0,16\\}\\)(" ; 4
6125 "\\)")
6126 finish t))
6127 (when (save-excursion
6128 (goto-char (match-beginning 0)) (not (c-in-literal)))
6129 (if (match-beginning 4) ; the id
6130 ;; We've found a raw string
6131 (c-depropertize-raw-string
6132 (match-string-no-properties 4) ; id
6133 (1+ (match-beginning 3)) ; open quote
6134 (match-end 4) ; open paren
6135 nil) ; bound
6136 ;; We've found a CPP construct. Search for raw strings within it.
6137 (goto-char (match-beginning 2)) ; the "#"
6138 (c-end-of-macro)
6139 (let ((eom (point)))
6140 (goto-char (match-end 2)) ; after the "#".
6141 (while (and (< (point) eom)
6142 (c-syntactic-re-search-forward
6143 "R\"\\([^ ()\\\n\r\t]\\{0,16\\}\\)(" eom t))
6144 (c-depropertize-raw-string
6145 (match-string-no-properties 1) ; id
6146 (1+ (match-beginning 0)) ; open quote
6147 (match-end 1) ; open paren
6148 eom))))))) ; bound.
6149
6150 (defun c-before-change-check-raw-strings (beg end)
6151 ;; This function clears `syntax-table' text properties from C++ raw strings
6152 ;; in the region (c-new-BEG c-new-END). BEG and END are the standard
6153 ;; arguments supplied to any before-change function.
6154 ;;
6155 ;; Point is undefined on both entry and exit, and the return value has no
6156 ;; significance.
6157 ;;
6158 ;; This function is called as a before-change function solely due to its
6159 ;; membership of the C++ value of `c-get-state-before-change-functions'.
6160 (c-save-buffer-state
6161 ((beg-rs (progn (goto-char beg) (c-raw-string-pos)))
6162 (beg-plus (if (null beg-rs)
6163 beg
6164 (max beg
6165 (1+ (or (nth 4 beg-rs) (nth 2 beg-rs))))))
6166 (end-rs (progn (goto-char end) (c-raw-string-pos))) ; FIXME!!!
6167 ; Optimize this so that we don't call
6168 ; `c-raw-string-pos' twice when once
6169 ; will do. (2016-06-02).
6170 (end-minus (if (null end-rs)
6171 end
6172 (min end (cadr end-rs))))
6173 )
6174 (when beg-rs
6175 (setq c-new-BEG (min c-new-BEG (1- (cadr beg-rs)))))
6176 (c-depropertize-raw-strings-in-region c-new-BEG beg-plus)
6177
6178 (when end-rs
6179 (setq c-new-END (max c-new-END
6180 (1+ (or (nth 4 end-rs)
6181 (nth 2 end-rs))))))
6182 (c-depropertize-raw-strings-in-region end-minus c-new-END)))
6183
6184 (defun c-propertize-raw-string-opener (id open-quote open-paren bound)
6185 ;; Point is immediately after a raw string opening delimiter. Apply any
6186 ;; pertinent `syntax-table' text properties to the delimiter and also the
6187 ;; raw string, should there be a valid matching closing delimiter.
6188 ;;
6189 ;; ID, a string, is the delimiter's identifier. OPEN-QUOTE and OPEN-PAREN
6190 ;; are the buffer positions of the delimiter's components. BOUND is the
6191 ;; bound for searching for a matching closing delimiter; it is usually nil,
6192 ;; but if we're inside a macro, it's the end of the macro.
6193 ;;
6194 ;; Point is moved to after the (terminated) raw string, or left after the
6195 ;; unmatched opening delimiter, as the case may be. The return value is of
6196 ;; no significance.
6197 (if (search-forward (concat ")" id "\"") bound t)
6198 (let ((end-string (match-beginning 0))
6199 (after-quote (match-end 0)))
6200 (goto-char open-paren)
6201 (while (progn (skip-syntax-forward "^\"" end-string)
6202 (< (point) end-string))
6203 (c-put-char-property (point) 'syntax-table '(1)) ; punctuation
6204 (c-truncate-semi-nonlit-pos-cache (point))
6205 (forward-char))
6206 (goto-char after-quote))
6207 (c-put-char-property open-quote 'syntax-table '(1)) ; punctuation
6208 (c-truncate-semi-nonlit-pos-cache open-quote)
6209 (c-put-char-property open-paren 'syntax-table '(15)) ; generic string
6210 (when bound
6211 ;; In a CPP construct, we try to apply a generic-string `syntax-table'
6212 ;; text property to the last possible character in the string, so that
6213 ;; only characters within the macro get "stringed out".
6214 (goto-char bound)
6215 (if (save-restriction
6216 (narrow-to-region (1+ open-paren) (point-max))
6217 (re-search-backward
6218 (eval-when-compile
6219 ;; This regular expression matches either an escape pair (which
6220 ;; isn't an escaped NL) (submatch 5) or a non-escaped character
6221 ;; (which isn't itself a backslash) (submatch 10). The long
6222 ;; preambles to these (respectively submatches 2-4 and 6-9)
6223 ;; ensure that we have the correct parity for sequences of
6224 ;; backslashes, etc..
6225 (concat "\\(" ; 1
6226 "\\(\\`[^\\]?\\|[^\\][^\\]\\)\\(\\\\\\(.\\|\n\\)\\)*" ; 2-4
6227 "\\(\\\\.\\)" ; 5
6228 "\\|"
6229 "\\(\\`\\|[^\\]\\|\\(\\`[^\\]?\\|[^\\][^\\]\\)\\(\\\\\\(.\\|\n\\)\\)+\\)" ; 6-9
6230 "\\([^\\]\\)" ; 10
6231 "\\)"
6232 "\\(\\\\\n\\)*\\=")) ; 11
6233 (1+ open-paren) t))
6234 (if (match-beginning 10)
6235 (progn
6236 (c-put-char-property (match-beginning 10) 'syntax-table '(15))
6237 (c-truncate-semi-nonlit-pos-cache (match-beginning 10)))
6238 (c-put-char-property (match-beginning 5) 'syntax-table '(1))
6239 (c-put-char-property (1+ (match-beginning 5)) 'syntax-table '(15))
6240 (c-truncate-semi-nonlit-pos-cache (1+ (match-beginning 5))))
6241 (c-put-char-property open-paren 'syntax-table '(1)))
6242 (goto-char bound))))
6243
6244 (defun c-after-change-re-mark-raw-strings (beg end old-len)
6245 ;; This function applies `syntax-table' text properties to C++ raw strings
6246 ;; beginning in the region (c-new-BEG c-new-END). BEG, END, and OLD-LEN are
6247 ;; the standard arguments supplied to any after-change function.
6248 ;;
6249 ;; Point is undefined on both entry and exit, and the return value has no
6250 ;; significance.
6251 ;;
6252 ;; This function is called as an after-change function solely due to its
6253 ;; membership of the C++ value of `c-before-font-lock-functions'.
6254 (c-save-buffer-state ()
6255 ;; If the region (c-new-BEG c-new-END) has expanded, remove
6256 ;; `syntax-table' text-properties from the new piece(s).
6257 (when (< c-new-BEG c-old-BEG)
6258 (let ((beg-rs (progn (goto-char c-old-BEG) (c-raw-string-pos))))
6259 (c-depropertize-raw-strings-in-region
6260 c-new-BEG
6261 (if beg-rs
6262 (1+ (or (nth 4 beg-rs) (nth 2 beg-rs)))
6263 c-old-BEG))))
6264 (when (> c-new-END c-old-END)
6265 (let ((end-rs (progn (goto-char c-old-END) (c-raw-string-pos))))
6266 (c-depropertize-raw-strings-in-region
6267 (if end-rs
6268 (cadr end-rs)
6269 c-old-END)
6270 c-new-END)))
6271
6272 (goto-char c-new-BEG)
6273 (while (and (< (point) c-new-END)
6274 (re-search-forward
6275 (concat "\\(" ; 1
6276 c-anchored-cpp-prefix ; 2
6277 "\\)\\|\\(" ; 3
6278 "R\"\\([^ ()\\\n\r\t]\\{0,16\\}\\)(" ; 4
6279 "\\)")
6280 c-new-END t))
6281 (when (save-excursion
6282 (goto-char (match-beginning 0)) (not (c-in-literal)))
6283 (if (match-beginning 4) ; the id
6284 ;; We've found a raw string.
6285 (c-propertize-raw-string-opener
6286 (match-string-no-properties 4) ; id
6287 (1+ (match-beginning 3)) ; open quote
6288 (match-end 4) ; open paren
6289 nil) ; bound
6290 ;; We've found a CPP construct. Search for raw strings within it.
6291 (goto-char (match-beginning 2)) ; the "#"
6292 (c-end-of-macro)
6293 (let ((eom (point)))
6294 (goto-char (match-end 2)) ; after the "#".
6295 (while (and (< (point) eom)
6296 (c-syntactic-re-search-forward
6297 "R\"\\([^ ()\\\n\r\t]\\{0,16\\}\\)(" eom t))
6298 (c-propertize-raw-string-opener
6299 (match-string-no-properties 1) ; id
6300 (1+ (match-beginning 0)) ; open quote
6301 (match-end 1) ; open paren
6302 eom)))))))) ; bound
6303
6304 \f
6305 ;; Handling of small scale constructs like types and names.
6306
6307 ;; Dynamically bound variable that instructs `c-forward-type' to also
6308 ;; treat possible types (i.e. those that it normally returns 'maybe or
6309 ;; 'found for) as actual types (and always return 'found for them).
6310 ;; This means that it records them in `c-record-type-identifiers' if
6311 ;; that is set, and that it adds them to `c-found-types'.
6312 (defvar c-promote-possible-types nil)
6313
6314 ;; Dynamically bound variable that instructs `c-forward-<>-arglist' to
6315 ;; mark up successfully parsed arglists with paren syntax properties on
6316 ;; the surrounding angle brackets and with `c-<>-arg-sep' in the
6317 ;; `c-type' property of each argument separating comma.
6318 ;;
6319 ;; Setting this variable also makes `c-forward-<>-arglist' recurse into
6320 ;; all arglists for side effects (i.e. recording types), otherwise it
6321 ;; exploits any existing paren syntax properties to quickly jump to the
6322 ;; end of already parsed arglists.
6323 ;;
6324 ;; Marking up the arglists is not the default since doing that correctly
6325 ;; depends on a proper value for `c-restricted-<>-arglists'.
6326 (defvar c-parse-and-markup-<>-arglists nil)
6327
6328 ;; Dynamically bound variable that instructs `c-forward-<>-arglist' to
6329 ;; not accept arglists that contain binary operators.
6330 ;;
6331 ;; This is primarily used to handle C++ template arglists. C++
6332 ;; disambiguates them by checking whether the preceding name is a
6333 ;; template or not. We can't do that, so we assume it is a template
6334 ;; if it can be parsed as one. That usually works well since
6335 ;; comparison expressions on the forms "a < b > c" or "a < b, c > d"
6336 ;; in almost all cases would be pointless.
6337 ;;
6338 ;; However, in function arglists, e.g. in "foo (a < b, c > d)", we
6339 ;; should let the comma separate the function arguments instead. And
6340 ;; in a context where the value of the expression is taken, e.g. in
6341 ;; "if (a < b || c > d)", it's probably not a template.
6342 (defvar c-restricted-<>-arglists nil)
6343
6344 ;; Dynamically bound variables that instructs
6345 ;; `c-forward-keyword-clause', `c-forward-<>-arglist',
6346 ;; `c-forward-name', `c-forward-type', `c-forward-decl-or-cast-1', and
6347 ;; `c-forward-label' to record the ranges of all the type and
6348 ;; reference identifiers they encounter. They will build lists on
6349 ;; these variables where each element is a cons of the buffer
6350 ;; positions surrounding each identifier. This recording is only
6351 ;; activated when `c-record-type-identifiers' is non-nil.
6352 ;;
6353 ;; All known types that can't be identifiers are recorded, and also
6354 ;; other possible types if `c-promote-possible-types' is set.
6355 ;; Recording is however disabled inside angle bracket arglists that
6356 ;; are encountered inside names and other angle bracket arglists.
6357 ;; Such occurrences are taken care of by `c-font-lock-<>-arglists'
6358 ;; instead.
6359 ;;
6360 ;; Only the names in C++ template style references (e.g. "tmpl" in
6361 ;; "tmpl<a,b>::foo") are recorded as references, other references
6362 ;; aren't handled here.
6363 ;;
6364 ;; `c-forward-label' records the label identifier(s) on
6365 ;; `c-record-ref-identifiers'.
6366 (defvar c-record-type-identifiers nil)
6367 (defvar c-record-ref-identifiers nil)
6368
6369 ;; This variable will receive a cons cell of the range of the last
6370 ;; single identifier symbol stepped over by `c-forward-name' if it's
6371 ;; successful. This is the range that should be put on one of the
6372 ;; record lists above by the caller. It's assigned nil if there's no
6373 ;; such symbol in the name.
6374 (defvar c-last-identifier-range nil)
6375
6376 (defmacro c-record-type-id (range)
6377 (if (eq (car-safe range) 'cons)
6378 ;; Always true.
6379 `(setq c-record-type-identifiers
6380 (cons ,range c-record-type-identifiers))
6381 `(let ((range ,range))
6382 (if range
6383 (setq c-record-type-identifiers
6384 (cons range c-record-type-identifiers))))))
6385
6386 (defmacro c-record-ref-id (range)
6387 (if (eq (car-safe range) 'cons)
6388 ;; Always true.
6389 `(setq c-record-ref-identifiers
6390 (cons ,range c-record-ref-identifiers))
6391 `(let ((range ,range))
6392 (if range
6393 (setq c-record-ref-identifiers
6394 (cons range c-record-ref-identifiers))))))
6395
6396 ;; Dynamically bound variable that instructs `c-forward-type' to
6397 ;; record the ranges of types that only are found. Behaves otherwise
6398 ;; like `c-record-type-identifiers'.
6399 (defvar c-record-found-types nil)
6400
6401 (defmacro c-forward-keyword-prefixed-id (type)
6402 ;; Used internally in `c-forward-keyword-clause' to move forward
6403 ;; over a type (if TYPE is 'type) or a name (otherwise) which
6404 ;; possibly is prefixed by keywords and their associated clauses.
6405 ;; Try with a type/name first to not trip up on those that begin
6406 ;; with a keyword. Return t if a known or found type is moved
6407 ;; over. The point is clobbered if nil is returned. If range
6408 ;; recording is enabled, the identifier is recorded on as a type
6409 ;; if TYPE is 'type or as a reference if TYPE is 'ref.
6410 ;;
6411 ;; This macro might do hidden buffer changes.
6412 `(let (res)
6413 (setq c-last-identifier-range nil)
6414 (while (if (setq res ,(if (eq type 'type)
6415 `(c-forward-type)
6416 `(c-forward-name)))
6417 nil
6418 (cond ((looking-at c-keywords-regexp)
6419 (c-forward-keyword-clause 1))
6420 ((and c-opt-cpp-prefix
6421 (looking-at c-noise-macro-with-parens-name-re))
6422 (c-forward-noise-clause)))))
6423 (when (memq res '(t known found prefix maybe))
6424 (when c-record-type-identifiers
6425 ,(if (eq type 'type)
6426 `(c-record-type-id c-last-identifier-range)
6427 `(c-record-ref-id c-last-identifier-range)))
6428 t)))
6429
6430 (defmacro c-forward-id-comma-list (type update-safe-pos)
6431 ;; Used internally in `c-forward-keyword-clause' to move forward
6432 ;; over a comma separated list of types or names using
6433 ;; `c-forward-keyword-prefixed-id'.
6434 ;;
6435 ;; This macro might do hidden buffer changes.
6436 `(while (and (progn
6437 ,(when update-safe-pos
6438 `(setq safe-pos (point)))
6439 (eq (char-after) ?,))
6440 (progn
6441 (forward-char)
6442 (c-forward-syntactic-ws)
6443 (c-forward-keyword-prefixed-id ,type)))))
6444
6445 (defun c-forward-noise-clause ()
6446 ;; Point is at a c-noise-macro-with-parens-names macro identifier. Go
6447 ;; forward over this name, any parenthesis expression which follows it, and
6448 ;; any syntactic WS, ending up at the next token. If there is an unbalanced
6449 ;; paren expression, leave point at it. Always Return t.
6450 (c-forward-token-2)
6451 (if (and (eq (char-after) ?\()
6452 (c-go-list-forward))
6453 (c-forward-syntactic-ws))
6454 t)
6455
6456 (defun c-forward-keyword-clause (match)
6457 ;; Submatch MATCH in the current match data is assumed to surround a
6458 ;; token. If it's a keyword, move over it and any immediately
6459 ;; following clauses associated with it, stopping at the start of
6460 ;; the next token. t is returned in that case, otherwise the point
6461 ;; stays and nil is returned. The kind of clauses that are
6462 ;; recognized are those specified by `c-type-list-kwds',
6463 ;; `c-ref-list-kwds', `c-colon-type-list-kwds',
6464 ;; `c-paren-nontype-kwds', `c-paren-type-kwds', `c-<>-type-kwds',
6465 ;; and `c-<>-arglist-kwds'.
6466 ;;
6467 ;; This function records identifier ranges on
6468 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
6469 ;; `c-record-type-identifiers' is non-nil.
6470 ;;
6471 ;; Note that for `c-colon-type-list-kwds', which doesn't necessary
6472 ;; apply directly after the keyword, the type list is moved over
6473 ;; only when there is no unaccounted token before it (i.e. a token
6474 ;; that isn't moved over due to some other keyword list). The
6475 ;; identifier ranges in the list are still recorded if that should
6476 ;; be done, though.
6477 ;;
6478 ;; This function might do hidden buffer changes.
6479
6480 (let ((kwd-sym (c-keyword-sym (match-string match))) safe-pos pos
6481 ;; The call to `c-forward-<>-arglist' below is made after
6482 ;; `c-<>-sexp-kwds' keywords, so we're certain they actually
6483 ;; are angle bracket arglists and `c-restricted-<>-arglists'
6484 ;; should therefore be nil.
6485 (c-parse-and-markup-<>-arglists t)
6486 c-restricted-<>-arglists)
6487
6488 (when kwd-sym
6489 (goto-char (match-end match))
6490 (c-forward-syntactic-ws)
6491 (setq safe-pos (point))
6492
6493 (cond
6494 ((and (c-keyword-member kwd-sym 'c-type-list-kwds)
6495 (c-forward-keyword-prefixed-id type))
6496 ;; There's a type directly after a keyword in `c-type-list-kwds'.
6497 (c-forward-id-comma-list type t))
6498
6499 ((and (c-keyword-member kwd-sym 'c-ref-list-kwds)
6500 (c-forward-keyword-prefixed-id ref))
6501 ;; There's a name directly after a keyword in `c-ref-list-kwds'.
6502 (c-forward-id-comma-list ref t))
6503
6504 ((and (c-keyword-member kwd-sym 'c-paren-any-kwds)
6505 (eq (char-after) ?\())
6506 ;; There's an open paren after a keyword in `c-paren-any-kwds'.
6507
6508 (forward-char)
6509 (when (and (setq pos (c-up-list-forward))
6510 (eq (char-before pos) ?\)))
6511 (when (and c-record-type-identifiers
6512 (c-keyword-member kwd-sym 'c-paren-type-kwds))
6513 ;; Use `c-forward-type' on every identifier we can find
6514 ;; inside the paren, to record the types.
6515 (while (c-syntactic-re-search-forward c-symbol-start pos t)
6516 (goto-char (match-beginning 0))
6517 (unless (c-forward-type)
6518 (looking-at c-symbol-key) ; Always matches.
6519 (goto-char (match-end 0)))))
6520
6521 (goto-char pos)
6522 (c-forward-syntactic-ws)
6523 (setq safe-pos (point))))
6524
6525 ((and (c-keyword-member kwd-sym 'c-<>-sexp-kwds)
6526 (eq (char-after) ?<)
6527 (c-forward-<>-arglist (c-keyword-member kwd-sym 'c-<>-type-kwds)))
6528 (c-forward-syntactic-ws)
6529 (setq safe-pos (point)))
6530
6531 ((and (c-keyword-member kwd-sym 'c-nonsymbol-sexp-kwds)
6532 (not (looking-at c-symbol-start))
6533 (c-safe (c-forward-sexp) t))
6534 (c-forward-syntactic-ws)
6535 (setq safe-pos (point))))
6536
6537 (when (c-keyword-member kwd-sym 'c-colon-type-list-kwds)
6538 (if (eq (char-after) ?:)
6539 ;; If we are at the colon already, we move over the type
6540 ;; list after it.
6541 (progn
6542 (forward-char)
6543 (c-forward-syntactic-ws)
6544 (when (c-forward-keyword-prefixed-id type)
6545 (c-forward-id-comma-list type t)))
6546 ;; Not at the colon, so stop here. But the identifier
6547 ;; ranges in the type list later on should still be
6548 ;; recorded.
6549 (and c-record-type-identifiers
6550 (progn
6551 ;; If a keyword matched both one of the types above and
6552 ;; this one, we match `c-colon-type-list-re' after the
6553 ;; clause matched above.
6554 (goto-char safe-pos)
6555 (looking-at c-colon-type-list-re))
6556 (progn
6557 (goto-char (match-end 0))
6558 (c-forward-syntactic-ws)
6559 (c-forward-keyword-prefixed-id type))
6560 ;; There's a type after the `c-colon-type-list-re' match
6561 ;; after a keyword in `c-colon-type-list-kwds'.
6562 (c-forward-id-comma-list type nil))))
6563
6564 (goto-char safe-pos)
6565 t)))
6566
6567 ;; cc-mode requires cc-fonts.
6568 (declare-function c-fontify-recorded-types-and-refs "cc-fonts" ())
6569
6570 (defun c-forward-<>-arglist (all-types)
6571 ;; The point is assumed to be at a "<". Try to treat it as the open
6572 ;; paren of an angle bracket arglist and move forward to the
6573 ;; corresponding ">". If successful, the point is left after the
6574 ;; ">" and t is returned, otherwise the point isn't moved and nil is
6575 ;; returned. If ALL-TYPES is t then all encountered arguments in
6576 ;; the arglist that might be types are treated as found types.
6577 ;;
6578 ;; The variable `c-parse-and-markup-<>-arglists' controls how this
6579 ;; function handles text properties on the angle brackets and argument
6580 ;; separating commas.
6581 ;;
6582 ;; `c-restricted-<>-arglists' controls how lenient the template
6583 ;; arglist recognition should be.
6584 ;;
6585 ;; This function records identifier ranges on
6586 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
6587 ;; `c-record-type-identifiers' is non-nil.
6588 ;;
6589 ;; This function might do hidden buffer changes.
6590
6591 (let ((start (point))
6592 ;; If `c-record-type-identifiers' is set then activate
6593 ;; recording of any found types that constitute an argument in
6594 ;; the arglist.
6595 (c-record-found-types (if c-record-type-identifiers t)))
6596 (if (catch 'angle-bracket-arglist-escape
6597 (setq c-record-found-types
6598 (c-forward-<>-arglist-recur all-types)))
6599 (progn
6600 (when (consp c-record-found-types)
6601 (setq c-record-type-identifiers
6602 ;; `nconc' doesn't mind that the tail of
6603 ;; `c-record-found-types' is t.
6604 (nconc c-record-found-types c-record-type-identifiers)))
6605 t)
6606
6607 (goto-char start)
6608 nil)))
6609
6610 (defun c-forward-<>-arglist-recur (all-types)
6611 ;; Recursive part of `c-forward-<>-arglist'.
6612 ;;
6613 ;; This function might do hidden buffer changes.
6614 (let ((start (point)) res pos
6615 ;; Cover this so that any recorded found type ranges are
6616 ;; automatically lost if it turns out to not be an angle
6617 ;; bracket arglist. It's propagated through the return value
6618 ;; on successful completion.
6619 (c-record-found-types c-record-found-types)
6620 ;; List that collects the positions after the argument
6621 ;; separating ',' in the arglist.
6622 arg-start-pos)
6623 ;; If the '<' has paren open syntax then we've marked it as an angle
6624 ;; bracket arglist before, so skip to the end.
6625 (if (and (not c-parse-and-markup-<>-arglists)
6626 (c-get-char-property (point) 'syntax-table))
6627
6628 (progn
6629 (forward-char)
6630 (if (and (c-go-up-list-forward)
6631 (eq (char-before) ?>))
6632 t
6633 ;; Got unmatched paren angle brackets. We don't clear the paren
6634 ;; syntax properties and retry, on the basis that it's very
6635 ;; unlikely that paren angle brackets become operators by code
6636 ;; manipulation. It's far more likely that it doesn't match due
6637 ;; to narrowing or some temporary change.
6638 (goto-char start)
6639 nil))
6640
6641 (forward-char) ; Forward over the opening '<'.
6642
6643 (unless (looking-at c-<-op-cont-regexp)
6644 ;; go forward one non-alphanumeric character (group) per iteration of
6645 ;; this loop.
6646 (while (and
6647 (progn
6648 (c-forward-syntactic-ws)
6649 (when (or (and c-record-type-identifiers all-types)
6650 (not (equal c-inside-<>-type-key "\\(\\<\\>\\)")))
6651 (c-forward-syntactic-ws)
6652 (cond
6653 ((eq (char-after) ??)
6654 (forward-char))
6655 ((and (looking-at c-identifier-start)
6656 (not (looking-at c-keywords-regexp)))
6657 (if (or (and all-types c-record-type-identifiers)
6658 (c-major-mode-is 'java-mode))
6659 ;; All encountered identifiers are types, so set the
6660 ;; promote flag and parse the type.
6661 (let ((c-promote-possible-types t)
6662 (c-record-found-types t))
6663 (c-forward-type))
6664 (c-forward-token-2))))
6665
6666 (c-forward-syntactic-ws)
6667
6668 (when (looking-at c-inside-<>-type-key)
6669 (goto-char (match-end 1))
6670 (c-forward-syntactic-ws)
6671 (let ((c-promote-possible-types t)
6672 (c-record-found-types t))
6673 (c-forward-type))
6674 (c-forward-syntactic-ws)))
6675
6676 (setq pos (point)) ; e.g. first token inside the '<'
6677
6678 ;; Note: These regexps exploit the match order in \| so
6679 ;; that "<>" is matched by "<" rather than "[^>:-]>".
6680 (c-syntactic-re-search-forward
6681 ;; Stop on ',', '|', '&', '+' and '-' to catch
6682 ;; common binary operators that could be between
6683 ;; two comparison expressions "a<b" and "c>d".
6684 ;; 2016-02-11: C++11 templates can now contain arithmetic
6685 ;; expressions, so template detection in C++ is now less
6686 ;; robust than it was.
6687 c-<>-notable-chars-re
6688 nil t t))
6689
6690 (cond
6691 ((eq (char-before) ?>)
6692 ;; Either an operator starting with '>' or the end of
6693 ;; the angle bracket arglist.
6694
6695 (if (save-excursion
6696 (c-backward-token-2)
6697 (looking-at c-multichar->-op-not->>-regexp))
6698 (progn
6699 (goto-char (match-end 0))
6700 t) ; Continue the loop.
6701
6702 ;; The angle bracket arglist is finished.
6703 (when c-parse-and-markup-<>-arglists
6704 (while arg-start-pos
6705 (c-put-c-type-property (1- (car arg-start-pos))
6706 'c-<>-arg-sep)
6707 (setq arg-start-pos (cdr arg-start-pos)))
6708 (c-mark-<-as-paren start)
6709 (c-mark->-as-paren (1- (point))))
6710 (setq res t)
6711 nil)) ; Exit the loop.
6712
6713 ((eq (char-before) ?<)
6714 ;; Either an operator starting with '<' or a nested arglist.
6715 (setq pos (point))
6716 (let (id-start id-end subres keyword-match)
6717 (cond
6718 ;; The '<' begins a multi-char operator.
6719 ((looking-at c-<-op-cont-regexp)
6720 (goto-char (match-end 0)))
6721 ;; We're at a nested <.....>
6722 ((progn
6723 (backward-char) ; to the '<'
6724 (and
6725 (save-excursion
6726 ;; There's always an identifier before an angle
6727 ;; bracket arglist, or a keyword in `c-<>-type-kwds'
6728 ;; or `c-<>-arglist-kwds'.
6729 (c-backward-syntactic-ws)
6730 (setq id-end (point))
6731 (c-simple-skip-symbol-backward)
6732 (when (or (setq keyword-match
6733 (looking-at c-opt-<>-sexp-key))
6734 (not (looking-at c-keywords-regexp)))
6735 (setq id-start (point))))
6736 (setq subres
6737 (let ((c-promote-possible-types t)
6738 (c-record-found-types t))
6739 (c-forward-<>-arglist-recur
6740 (and keyword-match
6741 (c-keyword-member
6742 (c-keyword-sym (match-string 1))
6743 'c-<>-type-kwds))))))
6744 (or subres (goto-char pos))
6745 subres)
6746 ;; It was an angle bracket arglist.
6747 (setq c-record-found-types subres)
6748
6749 ;; Record the identifier before the template as a type
6750 ;; or reference depending on whether the arglist is last
6751 ;; in a qualified identifier.
6752 (when (and c-record-type-identifiers
6753 (not keyword-match))
6754 (if (and c-opt-identifier-concat-key
6755 (progn
6756 (c-forward-syntactic-ws)
6757 (looking-at c-opt-identifier-concat-key)))
6758 (c-record-ref-id (cons id-start id-end))
6759 (c-record-type-id (cons id-start id-end)))))
6760
6761 ;; At a "less than" operator.
6762 (t
6763 ;; (forward-char) ; NO! We've already gone over the <.
6764 )))
6765 t) ; carry on looping.
6766
6767 ((and
6768 (eq (char-before) ?\()
6769 (c-go-up-list-forward)
6770 (eq (char-before) ?\))))
6771
6772 ((and (not c-restricted-<>-arglists)
6773 (or (and (eq (char-before) ?&)
6774 (not (eq (char-after) ?&)))
6775 (eq (char-before) ?,)))
6776 ;; Just another argument. Record the position. The
6777 ;; type check stuff that made us stop at it is at
6778 ;; the top of the loop.
6779 (setq arg-start-pos (cons (point) arg-start-pos)))
6780
6781 (t
6782 ;; Got a character that can't be in an angle bracket
6783 ;; arglist argument. Abort using `throw', since
6784 ;; it's useless to try to find a surrounding arglist
6785 ;; if we're nested.
6786 (throw 'angle-bracket-arglist-escape nil))))))
6787 (if res
6788 (or c-record-found-types t)))))
6789
6790 (defun c-backward-<>-arglist (all-types &optional limit)
6791 ;; The point is assumed to be directly after a ">". Try to treat it
6792 ;; as the close paren of an angle bracket arglist and move back to
6793 ;; the corresponding "<". If successful, the point is left at
6794 ;; the "<" and t is returned, otherwise the point isn't moved and
6795 ;; nil is returned. ALL-TYPES is passed on to
6796 ;; `c-forward-<>-arglist'.
6797 ;;
6798 ;; If the optional LIMIT is given, it bounds the backward search.
6799 ;; It's then assumed to be at a syntactically relevant position.
6800 ;;
6801 ;; This is a wrapper around `c-forward-<>-arglist'. See that
6802 ;; function for more details.
6803
6804 (let ((start (point)))
6805 (backward-char)
6806 (if (and (not c-parse-and-markup-<>-arglists)
6807 (c-get-char-property (point) 'syntax-table))
6808
6809 (if (and (c-go-up-list-backward)
6810 (eq (char-after) ?<))
6811 t
6812 ;; See corresponding note in `c-forward-<>-arglist'.
6813 (goto-char start)
6814 nil)
6815
6816 (while (progn
6817 (c-syntactic-skip-backward "^<;{}" limit t)
6818
6819 (and
6820 (if (eq (char-before) ?<)
6821 t
6822 ;; Stopped at bob or a char that isn't allowed in an
6823 ;; arglist, so we've failed.
6824 (goto-char start)
6825 nil)
6826
6827 (if (> (point)
6828 (progn (c-beginning-of-current-token)
6829 (point)))
6830 ;; If we moved then the "<" was part of some
6831 ;; multicharacter token.
6832 t
6833
6834 (backward-char)
6835 (let ((beg-pos (point)))
6836 (if (c-forward-<>-arglist all-types)
6837 (cond ((= (point) start)
6838 ;; Matched the arglist. Break the while.
6839 (goto-char beg-pos)
6840 nil)
6841 ((> (point) start)
6842 ;; We started from a non-paren ">" inside an
6843 ;; arglist.
6844 (goto-char start)
6845 nil)
6846 (t
6847 ;; Matched a shorter arglist. Can be a nested
6848 ;; one so continue looking.
6849 (goto-char beg-pos)
6850 t))
6851 t))))))
6852
6853 (/= (point) start))))
6854
6855 (defun c-forward-name ()
6856 ;; Move forward over a complete name if at the beginning of one,
6857 ;; stopping at the next following token. A keyword, as such,
6858 ;; doesn't count as a name. If the point is not at something that
6859 ;; is recognized as a name then it stays put.
6860 ;;
6861 ;; A name could be something as simple as "foo" in C or something as
6862 ;; complex as "X<Y<class A<int>::B, BIT_MAX >> b>, ::operator<> ::
6863 ;; Z<(a>b)> :: operator const X<&foo>::T Q::G<unsigned short
6864 ;; int>::*volatile const" in C++ (this function is actually little
6865 ;; more than a `looking-at' call in all modes except those that,
6866 ;; like C++, have `c-recognize-<>-arglists' set).
6867 ;;
6868 ;; Return
6869 ;; o - nil if no name is found;
6870 ;; o - 'template if it's an identifier ending with an angle bracket
6871 ;; arglist;
6872 ;; o - 'operator of it's an operator identifier;
6873 ;; o - t if it's some other kind of name.
6874 ;;
6875 ;; This function records identifier ranges on
6876 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
6877 ;; `c-record-type-identifiers' is non-nil.
6878 ;;
6879 ;; This function might do hidden buffer changes.
6880
6881 (let ((pos (point)) (start (point)) res id-start id-end
6882 ;; Turn off `c-promote-possible-types' here since we might
6883 ;; call `c-forward-<>-arglist' and we don't want it to promote
6884 ;; every suspect thing in the arglist to a type. We're
6885 ;; typically called from `c-forward-type' in this case, and
6886 ;; the caller only wants the top level type that it finds to
6887 ;; be promoted.
6888 c-promote-possible-types)
6889 (while
6890 (and
6891 (looking-at c-identifier-key)
6892
6893 (progn
6894 ;; Check for keyword. We go to the last symbol in
6895 ;; `c-identifier-key' first.
6896 (goto-char (setq id-end (match-end 0)))
6897 (c-simple-skip-symbol-backward)
6898 (setq id-start (point))
6899
6900 (if (looking-at c-keywords-regexp)
6901 (when (and (c-major-mode-is 'c++-mode)
6902 (looking-at
6903 (cc-eval-when-compile
6904 (concat "\\(operator\\|\\(template\\)\\)"
6905 "\\(" (c-lang-const c-nonsymbol-key c++)
6906 "\\|$\\)")))
6907 (if (match-beginning 2)
6908 ;; "template" is only valid inside an
6909 ;; identifier if preceded by "::".
6910 (save-excursion
6911 (c-backward-syntactic-ws)
6912 (and (c-safe (backward-char 2) t)
6913 (looking-at "::")))
6914 t))
6915
6916 ;; Handle a C++ operator or template identifier.
6917 (goto-char id-end)
6918 (c-forward-syntactic-ws)
6919 (cond ((eq (char-before id-end) ?e)
6920 ;; Got "... ::template".
6921 (let ((subres (c-forward-name)))
6922 (when subres
6923 (setq pos (point)
6924 res subres))))
6925
6926 ((looking-at c-identifier-start)
6927 ;; Got a cast operator.
6928 (when (c-forward-type)
6929 (setq pos (point)
6930 res 'operator)
6931 ;; Now we should match a sequence of either
6932 ;; '*', '&' or a name followed by ":: *",
6933 ;; where each can be followed by a sequence
6934 ;; of `c-opt-type-modifier-key'.
6935 (while (cond ((looking-at "[*&]")
6936 (goto-char (match-end 0))
6937 t)
6938 ((looking-at c-identifier-start)
6939 (and (c-forward-name)
6940 (looking-at "::")
6941 (progn
6942 (goto-char (match-end 0))
6943 (c-forward-syntactic-ws)
6944 (eq (char-after) ?*))
6945 (progn
6946 (forward-char)
6947 t))))
6948 (while (progn
6949 (c-forward-syntactic-ws)
6950 (setq pos (point))
6951 (looking-at c-opt-type-modifier-key))
6952 (goto-char (match-end 1))))))
6953
6954 ((looking-at c-overloadable-operators-regexp)
6955 ;; Got some other operator.
6956 (setq c-last-identifier-range
6957 (cons (point) (match-end 0)))
6958 (goto-char (match-end 0))
6959 (c-forward-syntactic-ws)
6960 (setq pos (point)
6961 res 'operator)))
6962
6963 nil)
6964
6965 ;; `id-start' is equal to `id-end' if we've jumped over
6966 ;; an identifier that doesn't end with a symbol token.
6967 ;; That can occur e.g. for Java import directives on the
6968 ;; form "foo.bar.*".
6969 (when (and id-start (/= id-start id-end))
6970 (setq c-last-identifier-range
6971 (cons id-start id-end)))
6972 (goto-char id-end)
6973 (c-forward-syntactic-ws)
6974 (setq pos (point)
6975 res t)))
6976
6977 (progn
6978 (goto-char pos)
6979 (when (or c-opt-identifier-concat-key
6980 c-recognize-<>-arglists)
6981
6982 (cond
6983 ((and c-opt-identifier-concat-key
6984 (looking-at c-opt-identifier-concat-key))
6985 ;; Got a concatenated identifier. This handles the
6986 ;; cases with tricky syntactic whitespace that aren't
6987 ;; covered in `c-identifier-key'.
6988 (goto-char (match-end 0))
6989 (c-forward-syntactic-ws)
6990 t)
6991
6992 ((and c-recognize-<>-arglists
6993 (eq (char-after) ?<))
6994 ;; Maybe an angle bracket arglist.
6995 (when (let (c-last-identifier-range)
6996 (c-forward-<>-arglist nil))
6997
6998 (c-forward-syntactic-ws)
6999 (unless (eq (char-after) ?\()
7000 (setq c-last-identifier-range nil)
7001 (c-add-type start (1+ pos)))
7002 (setq pos (point))
7003
7004 (if (and c-opt-identifier-concat-key
7005 (looking-at c-opt-identifier-concat-key))
7006
7007 ;; Continue if there's an identifier concatenation
7008 ;; operator after the template argument.
7009 (progn
7010 (when (and c-record-type-identifiers id-start)
7011 (c-record-ref-id (cons id-start id-end)))
7012 (forward-char 2)
7013 (c-forward-syntactic-ws)
7014 t)
7015
7016 (when (and c-record-type-identifiers id-start
7017 (not (eq (char-after) ?\()))
7018 (c-record-type-id (cons id-start id-end)))
7019 (setq res 'template)
7020 nil)))
7021 )))))
7022
7023 (goto-char pos)
7024 res))
7025
7026 (defun c-forward-type (&optional brace-block-too)
7027 ;; Move forward over a type spec if at the beginning of one,
7028 ;; stopping at the next following token. The keyword "typedef"
7029 ;; isn't part of a type spec here.
7030 ;;
7031 ;; BRACE-BLOCK-TOO, when non-nil, means move over the brace block in
7032 ;; constructs like "struct foo {...} bar ;" or "struct {...} bar;".
7033 ;; The current (2009-03-10) intention is to convert all uses of
7034 ;; `c-forward-type' to call with this parameter set, then to
7035 ;; eliminate it.
7036 ;;
7037 ;; Return
7038 ;; o - t if it's a known type that can't be a name or other
7039 ;; expression;
7040 ;; o - 'known if it's an otherwise known type (according to
7041 ;; `*-font-lock-extra-types');
7042 ;; o - 'prefix if it's a known prefix of a type;
7043 ;; o - 'found if it's a type that matches one in `c-found-types';
7044 ;; o - 'maybe if it's an identifier that might be a type;
7045 ;; o - 'decltype if it's a decltype(variable) declaration; - or
7046 ;; o - nil if it can't be a type (the point isn't moved then).
7047 ;;
7048 ;; The point is assumed to be at the beginning of a token.
7049 ;;
7050 ;; Note that this function doesn't skip past the brace definition
7051 ;; that might be considered part of the type, e.g.
7052 ;; "enum {a, b, c} foo".
7053 ;;
7054 ;; This function records identifier ranges on
7055 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
7056 ;; `c-record-type-identifiers' is non-nil.
7057 ;;
7058 ;; This function might do hidden buffer changes.
7059 (when (and c-recognize-<>-arglists
7060 (looking-at "<"))
7061 (c-forward-<>-arglist t)
7062 (c-forward-syntactic-ws))
7063
7064 (let ((start (point)) pos res name-res id-start id-end id-range)
7065
7066 ;; Skip leading type modifiers. If any are found we know it's a
7067 ;; prefix of a type.
7068 (when c-opt-type-modifier-key ; e.g. "const" "volatile", but NOT "typedef"
7069 (while (looking-at c-opt-type-modifier-key)
7070 (goto-char (match-end 1))
7071 (c-forward-syntactic-ws)
7072 (setq res 'prefix)))
7073
7074 (cond
7075 ((looking-at c-typeof-key) ; e.g. C++'s "decltype".
7076 (goto-char (match-end 1))
7077 (c-forward-syntactic-ws)
7078 (setq res (and (eq (char-after) ?\()
7079 (c-safe (c-forward-sexp))
7080 'decltype))
7081 (if res
7082 (c-forward-syntactic-ws)
7083 (goto-char start)))
7084
7085 ((looking-at c-type-prefix-key) ; e.g. "struct", "class", but NOT
7086 ; "typedef".
7087 (goto-char (match-end 1))
7088 (c-forward-syntactic-ws)
7089
7090 (while (cond
7091 ((looking-at c-decl-hangon-key)
7092 (c-forward-keyword-clause 1))
7093 ((and c-opt-cpp-prefix
7094 (looking-at c-noise-macro-with-parens-name-re))
7095 (c-forward-noise-clause))))
7096
7097 (setq pos (point))
7098
7099 (setq name-res (c-forward-name))
7100 (setq res (not (null name-res)))
7101 (when (eq name-res t)
7102 ;; In many languages the name can be used without the
7103 ;; prefix, so we add it to `c-found-types'.
7104 (c-add-type pos (point))
7105 (when (and c-record-type-identifiers
7106 c-last-identifier-range)
7107 (c-record-type-id c-last-identifier-range)))
7108 (when (and brace-block-too
7109 (memq res '(t nil))
7110 (eq (char-after) ?\{)
7111 (save-excursion
7112 (c-safe
7113 (progn (c-forward-sexp)
7114 (c-forward-syntactic-ws)
7115 (setq pos (point))))))
7116 (goto-char pos)
7117 (setq res t))
7118 (unless res (goto-char start))) ; invalid syntax
7119
7120 ((progn
7121 (setq pos nil)
7122 (if (looking-at c-identifier-start)
7123 (save-excursion
7124 (setq id-start (point)
7125 name-res (c-forward-name))
7126 (when name-res
7127 (setq id-end (point)
7128 id-range c-last-identifier-range))))
7129 (and (cond ((looking-at c-primitive-type-key)
7130 (setq res t))
7131 ((c-with-syntax-table c-identifier-syntax-table
7132 (looking-at c-known-type-key))
7133 (setq res 'known)))
7134 (or (not id-end)
7135 (>= (save-excursion
7136 (save-match-data
7137 (goto-char (match-end 1))
7138 (c-forward-syntactic-ws)
7139 (setq pos (point))))
7140 id-end)
7141 (setq res nil))))
7142 ;; Looking at a primitive or known type identifier. We've
7143 ;; checked for a name first so that we don't go here if the
7144 ;; known type match only is a prefix of another name.
7145
7146 (setq id-end (match-end 1))
7147
7148 (when (and c-record-type-identifiers
7149 (or c-promote-possible-types (eq res t)))
7150 (c-record-type-id (cons (match-beginning 1) (match-end 1))))
7151
7152 (if (and c-opt-type-component-key
7153 (save-match-data
7154 (looking-at c-opt-type-component-key)))
7155 ;; There might be more keywords for the type.
7156 (let (safe-pos)
7157 (c-forward-keyword-clause 1)
7158 (while (progn
7159 (setq safe-pos (point))
7160 (looking-at c-opt-type-component-key))
7161 (when (and c-record-type-identifiers
7162 (looking-at c-primitive-type-key))
7163 (c-record-type-id (cons (match-beginning 1)
7164 (match-end 1))))
7165 (c-forward-keyword-clause 1))
7166 (if (looking-at c-primitive-type-key)
7167 (progn
7168 (when c-record-type-identifiers
7169 (c-record-type-id (cons (match-beginning 1)
7170 (match-end 1))))
7171 (c-forward-keyword-clause 1)
7172 (setq res t))
7173 (goto-char safe-pos)
7174 (setq res 'prefix)))
7175 (unless (save-match-data (c-forward-keyword-clause 1))
7176 (if pos
7177 (goto-char pos)
7178 (goto-char (match-end 1))
7179 (c-forward-syntactic-ws)))))
7180
7181 (name-res
7182 (cond ((eq name-res t)
7183 ;; A normal identifier.
7184 (goto-char id-end)
7185 (if (or res c-promote-possible-types)
7186 (progn
7187 (c-add-type id-start id-end)
7188 (when (and c-record-type-identifiers id-range)
7189 (c-record-type-id id-range))
7190 (unless res
7191 (setq res 'found)))
7192 (setq res (if (c-check-type id-start id-end)
7193 ;; It's an identifier that has been used as
7194 ;; a type somewhere else.
7195 'found
7196 ;; It's an identifier that might be a type.
7197 'maybe))))
7198 ((eq name-res 'template)
7199 ;; A template is sometimes a type.
7200 (goto-char id-end)
7201 (c-forward-syntactic-ws)
7202 (setq res
7203 (if (eq (char-after) ?\()
7204 (if (c-check-type id-start id-end)
7205 ;; It's an identifier that has been used as
7206 ;; a type somewhere else.
7207 'found
7208 ;; It's an identifier that might be a type.
7209 'maybe)
7210 t)))
7211 (t
7212 ;; Otherwise it's an operator identifier, which is not a type.
7213 (goto-char start)
7214 (setq res nil)))))
7215
7216 (when res
7217 ;; Skip trailing type modifiers. If any are found we know it's
7218 ;; a type.
7219 (when c-opt-type-modifier-key
7220 (while (looking-at c-opt-type-modifier-key) ; e.g. "const", "volatile"
7221 (goto-char (match-end 1))
7222 (c-forward-syntactic-ws)
7223 (setq res t)))
7224
7225 ;; Step over any type suffix operator. Do not let the existence
7226 ;; of these alter the classification of the found type, since
7227 ;; these operators typically are allowed in normal expressions
7228 ;; too.
7229 (when c-opt-type-suffix-key ; e.g. "..."
7230 (while (looking-at c-opt-type-suffix-key)
7231 (goto-char (match-end 1))
7232 (c-forward-syntactic-ws)))
7233
7234 (when c-opt-type-concat-key ; Only/mainly for pike.
7235 ;; Look for a trailing operator that concatenates the type
7236 ;; with a following one, and if so step past that one through
7237 ;; a recursive call. Note that we don't record concatenated
7238 ;; types in `c-found-types' - it's the component types that
7239 ;; are recorded when appropriate.
7240 (setq pos (point))
7241 (let* ((c-promote-possible-types (or (memq res '(t known))
7242 c-promote-possible-types))
7243 ;; If we can't promote then set `c-record-found-types' so that
7244 ;; we can merge in the types from the second part afterwards if
7245 ;; it turns out to be a known type there.
7246 (c-record-found-types (and c-record-type-identifiers
7247 (not c-promote-possible-types)))
7248 subres)
7249 (if (and (looking-at c-opt-type-concat-key)
7250
7251 (progn
7252 (goto-char (match-end 1))
7253 (c-forward-syntactic-ws)
7254 (setq subres (c-forward-type))))
7255
7256 (progn
7257 ;; If either operand certainly is a type then both are, but we
7258 ;; don't let the existence of the operator itself promote two
7259 ;; uncertain types to a certain one.
7260 (cond ((eq res t))
7261 ((eq subres t)
7262 (unless (eq name-res 'template)
7263 (c-add-type id-start id-end))
7264 (when (and c-record-type-identifiers id-range)
7265 (c-record-type-id id-range))
7266 (setq res t))
7267 ((eq res 'known))
7268 ((eq subres 'known)
7269 (setq res 'known))
7270 ((eq res 'found))
7271 ((eq subres 'found)
7272 (setq res 'found))
7273 (t
7274 (setq res 'maybe)))
7275
7276 (when (and (eq res t)
7277 (consp c-record-found-types))
7278 ;; Merge in the ranges of any types found by the second
7279 ;; `c-forward-type'.
7280 (setq c-record-type-identifiers
7281 ;; `nconc' doesn't mind that the tail of
7282 ;; `c-record-found-types' is t.
7283 (nconc c-record-found-types
7284 c-record-type-identifiers))))
7285
7286 (goto-char pos))))
7287
7288 (when (and c-record-found-types (memq res '(known found)) id-range)
7289 (setq c-record-found-types
7290 (cons id-range c-record-found-types))))
7291
7292 ;;(message "c-forward-type %s -> %s: %s" start (point) res)
7293
7294 res))
7295
7296 (defun c-forward-annotation ()
7297 ;; Used for Java code only at the moment. Assumes point is on the @, moves
7298 ;; forward an annotation and returns t. Leaves point unmoved and returns
7299 ;; nil if there is no annotation at point.
7300 (let ((pos (point)))
7301 (or
7302 (and (looking-at "@")
7303 (not (looking-at c-keywords-regexp))
7304 (progn (forward-char) t)
7305 (looking-at c-symbol-key)
7306 (progn (goto-char (match-end 0))
7307 (c-forward-syntactic-ws)
7308 t)
7309 (if (looking-at "(")
7310 (c-go-list-forward)
7311 t))
7312 (progn (goto-char pos) nil))))
7313
7314 (defmacro c-pull-open-brace (ps)
7315 ;; Pull the next open brace from PS (which has the form of paren-state),
7316 ;; skipping over any brace pairs. Returns NIL when PS is exhausted.
7317 `(progn
7318 (while (consp (car ,ps))
7319 (setq ,ps (cdr ,ps)))
7320 (prog1 (car ,ps)
7321 (setq ,ps (cdr ,ps)))))
7322
7323 (defun c-back-over-compound-identifier ()
7324 ;; Point is putatively just after a "compound identifier", i.e. something
7325 ;; looking (in C++) like this "FQN::of::base::Class". Move to the start of
7326 ;; this construct and return t. If the parsing fails, return nil, leaving
7327 ;; point unchanged.
7328 (let ((here (point))
7329 end
7330 )
7331 (if (not (c-simple-skip-symbol-backward))
7332 nil
7333 (while
7334 (progn
7335 (setq end (point))
7336 (c-backward-syntactic-ws)
7337 (c-backward-token-2)
7338 (and
7339 c-opt-identifier-concat-key
7340 (looking-at c-opt-identifier-concat-key)
7341 (progn
7342 (c-backward-syntactic-ws)
7343 (c-simple-skip-symbol-backward))))
7344 (setq end (point)))
7345 (goto-char end)
7346 t)))
7347
7348 (defun c-back-over-member-initializer-braces ()
7349 ;; Point is just after a closing brace/parenthesis. Try to parse this as a
7350 ;; C++ member initializer list, going back to just after the introducing ":"
7351 ;; and returning t. Otherwise return nil, leaving point unchanged.
7352 (let ((here (point)) res)
7353 (setq res
7354 (catch 'done
7355 (when (not (c-go-list-backward))
7356 (throw 'done nil))
7357 (c-backward-syntactic-ws)
7358 (when (not (c-back-over-compound-identifier))
7359 (throw 'done nil))
7360 (c-backward-syntactic-ws)
7361
7362 (while (eq (char-before) ?,)
7363 (backward-char)
7364 (c-backward-syntactic-ws)
7365 (when (not (memq (char-before) '(?\) ?})))
7366 (throw 'done nil))
7367 (when (not (c-go-list-backward))
7368 (throw 'done nil))
7369 (c-backward-syntactic-ws)
7370 (when (not (c-back-over-compound-identifier))
7371 (throw 'done nil))
7372 (c-backward-syntactic-ws))
7373
7374 (eq (char-before) ?:)))
7375 (or res (goto-char here))
7376 res))
7377
7378 (defmacro c-back-over-list-of-member-inits ()
7379 ;; Go back over a list of elements, each looking like:
7380 ;; <symbol> (<expression>) ,
7381 ;; or <symbol> {<expression>} ,
7382 ;; when we are putatively immediately after a comma. Stop when we don't see
7383 ;; a comma. If either of <symbol> or bracketed <expression> is missing,
7384 ;; throw nil to 'level. If the terminating } or ) is unmatched, throw nil
7385 ;; to 'done. This is not a general purpose macro!
7386 `(while (eq (char-before) ?,)
7387 (backward-char)
7388 (c-backward-syntactic-ws)
7389 (when (not (memq (char-before) '(?\) ?})))
7390 (throw 'level nil))
7391 (when (not (c-go-list-backward))
7392 (throw 'done nil))
7393 (c-backward-syntactic-ws)
7394 (when (not (c-back-over-compound-identifier))
7395 (throw 'level nil))
7396 (c-backward-syntactic-ws)))
7397
7398 (defun c-back-over-member-initializers ()
7399 ;; Test whether we are in a C++ member initializer list, and if so, go back
7400 ;; to the introducing ":", returning the position of the opening paren of
7401 ;; the function's arglist. Otherwise return nil, leaving point unchanged.
7402 (let ((here (point))
7403 (paren-state (c-parse-state))
7404 pos level-plausible at-top-level res)
7405 ;; Assume tentatively that we're at the top level. Try to go back to the
7406 ;; colon we seek.
7407 (setq res
7408 (catch 'done
7409 (setq level-plausible
7410 (catch 'level
7411 (c-backward-syntactic-ws)
7412 (when (memq (char-before) '(?\) ?}))
7413 (when (not (c-go-list-backward))
7414 (throw 'done nil))
7415 (c-backward-syntactic-ws))
7416 (when (c-back-over-compound-identifier)
7417 (c-backward-syntactic-ws))
7418 (c-back-over-list-of-member-inits)
7419 (and (eq (char-before) ?:)
7420 (save-excursion
7421 (c-backward-token-2)
7422 (not (looking-at c-:$-multichar-token-regexp)))
7423 (c-just-after-func-arglist-p))))
7424
7425 (while (and (not (and level-plausible
7426 (setq at-top-level (c-at-toplevel-p))))
7427 (setq pos (c-pull-open-brace paren-state))) ; might be a paren.
7428 (setq level-plausible
7429 (catch 'level
7430 (goto-char pos)
7431 (c-backward-syntactic-ws)
7432 (when (not (c-back-over-compound-identifier))
7433 (throw 'level nil))
7434 (c-backward-syntactic-ws)
7435 (c-back-over-list-of-member-inits)
7436 (and (eq (char-before) ?:)
7437 (save-excursion
7438 (c-backward-token-2)
7439 (not (looking-at c-:$-multichar-token-regexp)))
7440 (c-just-after-func-arglist-p)))))
7441
7442 (and at-top-level level-plausible)))
7443 (or res (goto-char here))
7444 res))
7445
7446 \f
7447 ;; Handling of large scale constructs like statements and declarations.
7448
7449 ;; Macro used inside `c-forward-decl-or-cast-1'. It ought to be a
7450 ;; defsubst or perhaps even a defun, but it contains lots of free
7451 ;; variables that refer to things inside `c-forward-decl-or-cast-1'.
7452 (defmacro c-fdoc-shift-type-backward (&optional short)
7453 ;; `c-forward-decl-or-cast-1' can consume an arbitrary length list
7454 ;; of types when parsing a declaration, which means that it
7455 ;; sometimes consumes the identifier in the declaration as a type.
7456 ;; This is used to "backtrack" and make the last type be treated as
7457 ;; an identifier instead.
7458 `(progn
7459 ,(unless short
7460 ;; These identifiers are bound only in the inner let.
7461 '(setq identifier-type at-type
7462 identifier-start type-start
7463 got-parens nil
7464 got-identifier t
7465 got-suffix t
7466 got-suffix-after-parens id-start
7467 paren-depth 0))
7468
7469 (if (setq at-type (if (eq backup-at-type 'prefix)
7470 t
7471 backup-at-type))
7472 (setq type-start backup-type-start
7473 id-start backup-id-start)
7474 (setq type-start start-pos
7475 id-start start-pos))
7476
7477 ;; When these flags already are set we've found specifiers that
7478 ;; unconditionally signal these attributes - backtracking doesn't
7479 ;; change that. So keep them set in that case.
7480 (or at-type-decl
7481 (setq at-type-decl backup-at-type-decl))
7482 (or maybe-typeless
7483 (setq maybe-typeless backup-maybe-typeless))
7484
7485 ,(unless short
7486 ;; This identifier is bound only in the inner let.
7487 '(setq start id-start))))
7488
7489 (defun c-forward-declarator (&optional limit accept-anon)
7490 ;; Assuming point is at the start of a declarator, move forward over it,
7491 ;; leaving point at the next token after it (e.g. a ) or a ; or a ,).
7492 ;;
7493 ;; Return a list (ID-START ID-END BRACKETS-AFTER-ID GOT-INIT), where ID-START and
7494 ;; ID-END are the bounds of the declarator's identifier, and
7495 ;; BRACKETS-AFTER-ID is non-nil if a [...] pair is present after the id.
7496 ;; GOT-INIT is non-nil when the declarator is followed by "=" or "(".
7497 ;;
7498 ;; If ACCEPT-ANON is non-nil, move forward over any "anonymous declarator",
7499 ;; i.e. something like the (*) in int (*), such as might be found in a
7500 ;; declaration. In such a case ID-START and ID-END in the return value are
7501 ;; both set to nil. A "null" "anonymous declarator" gives a non-nil result.
7502 ;;
7503 ;; If no declarator is found, leave point unmoved and return nil. LIMIT is
7504 ;; an optional limit for forward searching.
7505 ;;
7506 ;; Note that the global variable `c-last-identifier-range' is written to, so
7507 ;; the caller should bind it if necessary.
7508
7509 ;; Inside the following "condition form", we move forward over the
7510 ;; declarator's identifier up as far as any opening bracket (for array
7511 ;; size) or paren (for parameters of function-type) or brace (for
7512 ;; array/struct initialization) or "=" or terminating delimiter
7513 ;; (e.g. "," or ";" or "}").
7514 (let ((here (point))
7515 id-start id-end brackets-after-id paren-depth)
7516 (or limit (setq limit (point-max)))
7517 (if (and
7518 (< (point) limit)
7519
7520 ;; The following form moves forward over the declarator's
7521 ;; identifier (and what precedes it), returning t. If there
7522 ;; wasn't one, it returns nil.
7523 (let (got-identifier)
7524 (setq paren-depth 0)
7525 ;; Skip over type decl prefix operators, one for each iteration
7526 ;; of the while. These are, e.g. "*" in "int *foo" or "(" and
7527 ;; "*" in "int (*foo) (void)" (Note similar code in
7528 ;; `c-forward-decl-or-cast-1'.)
7529 (while
7530 (cond
7531 ((looking-at c-decl-hangon-key)
7532 (c-forward-keyword-clause 1))
7533 ((and c-opt-cpp-prefix
7534 (looking-at c-noise-macro-with-parens-name-re))
7535 (c-forward-noise-clause))
7536 ((and (looking-at c-type-decl-prefix-key)
7537 (if (and (c-major-mode-is 'c++-mode)
7538 (match-beginning 3))
7539 ;; If the third submatch matches in C++ then
7540 ;; we're looking at an identifier that's a
7541 ;; prefix only if it specifies a member pointer.
7542 (progn
7543 (setq id-start (point))
7544 (c-forward-name)
7545 (if (looking-at "\\(::\\)")
7546 ;; We only check for a trailing "::" and
7547 ;; let the "*" that should follow be
7548 ;; matched in the next round.
7549 t
7550 ;; It turned out to be the real identifier,
7551 ;; so flag that and stop.
7552 (setq got-identifier t)
7553 nil))
7554 t))
7555 (if (eq (char-after) ?\()
7556 (progn
7557 (setq paren-depth (1+ paren-depth))
7558 (forward-char))
7559 (goto-char (match-end 1)))
7560 (c-forward-syntactic-ws)
7561 t)))
7562
7563 ;; If we haven't passed the identifier already, do it now.
7564 (unless got-identifier
7565 (setq id-start (point)))
7566 (cond
7567 ((or got-identifier
7568 (c-forward-name))
7569 (save-excursion
7570 (c-backward-syntactic-ws)
7571 (setq id-end (point))))
7572 (accept-anon
7573 (setq id-start nil id-end nil)
7574 t)
7575 (t (/= (point) here))))
7576
7577 ;; Skip out of the parens surrounding the identifier. If closing
7578 ;; parens are missing, this form returns nil.
7579 (or (= paren-depth 0)
7580 (c-safe (goto-char (scan-lists (point) 1 paren-depth))))
7581
7582 (<= (point) limit)
7583
7584 ;; Skip over any trailing bit, such as "__attribute__".
7585 (progn
7586 (while (cond
7587 ((looking-at c-decl-hangon-key)
7588 (c-forward-keyword-clause 1))
7589 ((and c-opt-cpp-prefix
7590 (looking-at c-noise-macro-with-parens-name-re))
7591 (c-forward-noise-clause))))
7592 (<= (point) limit))
7593
7594 ;; Search syntactically to the end of the declarator (";",
7595 ;; ",", a closing paren, eob etc) or to the beginning of an
7596 ;; initializer or function prototype ("=" or "\\s\(").
7597 ;; Note that square brackets are now not also treated as
7598 ;; initializers, since this broke when there were also
7599 ;; initializing brace lists.
7600 (let (found)
7601 (while
7602 (and (setq found (c-syntactic-re-search-forward
7603 "[;,]\\|\\s)\\|\\'\\|\\(=\\|\\s(\\)" limit t t))
7604 (eq (char-before) ?\[)
7605 (c-go-up-list-forward))
7606 (setq brackets-after-id t))
7607 (backward-char)
7608 found))
7609 (list id-start id-end brackets-after-id (match-beginning 1))
7610
7611 (goto-char here)
7612 nil)))
7613
7614 (defun c-forward-decl-or-cast-1 (preceding-token-end context last-cast-end)
7615 ;; Move forward over a declaration or a cast if at the start of one.
7616 ;; The point is assumed to be at the start of some token. Nil is
7617 ;; returned if no declaration or cast is recognized, and the point
7618 ;; is clobbered in that case.
7619 ;;
7620 ;; If a declaration is parsed:
7621 ;;
7622 ;; The point is left at the first token after the first complete
7623 ;; declarator, if there is one. The return value is a list of 4 elements,
7624 ;; where the first is the position of the first token in the declarator.
7625 ;; (See below for the other three.)
7626 ;; Some examples:
7627 ;;
7628 ;; void foo (int a, char *b) stuff ...
7629 ;; car ^ ^ point
7630 ;; float (*a)[], b;
7631 ;; car ^ ^ point
7632 ;; unsigned int a = c_style_initializer, b;
7633 ;; car ^ ^ point
7634 ;; unsigned int a (cplusplus_style_initializer), b;
7635 ;; car ^ ^ point (might change)
7636 ;; class Foo : public Bar {}
7637 ;; car ^ ^ point
7638 ;; class PikeClass (int a, string b) stuff ...
7639 ;; car ^ ^ point
7640 ;; enum bool;
7641 ;; car ^ ^ point
7642 ;; enum bool flag;
7643 ;; car ^ ^ point
7644 ;; void cplusplus_function (int x) throw (Bad);
7645 ;; car ^ ^ point
7646 ;; Foo::Foo (int b) : Base (b) {}
7647 ;; car ^ ^ point
7648 ;;
7649 ;; auto foo = 5;
7650 ;; car ^ ^ point
7651 ;; auto cplusplus_11 (int a, char *b) -> decltype (bar):
7652 ;; car ^ ^ point
7653 ;;
7654 ;;
7655 ;;
7656 ;; The second element of the return value is non-nil when a
7657 ;; `c-typedef-decl-kwds' specifier is found in the declaration.
7658 ;; Specifically it is a dotted pair (A . B) where B is t when a
7659 ;; `c-typedef-kwds' ("typedef") is present, and A is t when some
7660 ;; other `c-typedef-decl-kwds' (e.g. class, struct, enum)
7661 ;; specifier is present. I.e., (some of) the declared
7662 ;; identifier(s) are types.
7663 ;;
7664 ;; The third element of the return value is non-nil when the declaration
7665 ;; parsed might be an expression. The fourth element is the position of
7666 ;; the start of the type identifier.
7667 ;;
7668 ;; If a cast is parsed:
7669 ;;
7670 ;; The point is left at the first token after the closing paren of
7671 ;; the cast. The return value is `cast'. Note that the start
7672 ;; position must be at the first token inside the cast parenthesis
7673 ;; to recognize it.
7674 ;;
7675 ;; PRECEDING-TOKEN-END is the first position after the preceding
7676 ;; token, i.e. on the other side of the syntactic ws from the point.
7677 ;; Use a value less than or equal to (point-min) if the point is at
7678 ;; the first token in (the visible part of) the buffer.
7679 ;;
7680 ;; CONTEXT is a symbol that describes the context at the point:
7681 ;; 'decl In a comma-separated declaration context (typically
7682 ;; inside a function declaration arglist).
7683 ;; '<> In an angle bracket arglist.
7684 ;; 'arglist Some other type of arglist.
7685 ;; nil Some other context or unknown context. Includes
7686 ;; within the parens of an if, for, ... construct.
7687 ;;
7688 ;; LAST-CAST-END is the first token after the closing paren of a
7689 ;; preceding cast, or nil if none is known. If
7690 ;; `c-forward-decl-or-cast-1' is used in succession, it should be
7691 ;; the position after the closest preceding call where a cast was
7692 ;; matched. In that case it's used to discover chains of casts like
7693 ;; "(a) (b) c".
7694 ;;
7695 ;; This function records identifier ranges on
7696 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
7697 ;; `c-record-type-identifiers' is non-nil.
7698 ;;
7699 ;; This function might do hidden buffer changes.
7700
7701 (let (;; `start-pos' is used below to point to the start of the
7702 ;; first type, i.e. after any leading specifiers. It might
7703 ;; also point at the beginning of the preceding syntactic
7704 ;; whitespace.
7705 (start-pos (point))
7706 ;; Set to the result of `c-forward-type'.
7707 at-type
7708 ;; The position of the first token in what we currently
7709 ;; believe is the type in the declaration or cast, after any
7710 ;; specifiers and their associated clauses.
7711 type-start
7712 ;; The position of the first token in what we currently
7713 ;; believe is the declarator for the first identifier. Set
7714 ;; when the type is found, and moved forward over any
7715 ;; `c-decl-hangon-kwds' and their associated clauses that
7716 ;; occurs after the type.
7717 id-start
7718 ;; These store `at-type', `type-start' and `id-start' of the
7719 ;; identifier before the one in those variables. The previous
7720 ;; identifier might turn out to be the real type in a
7721 ;; declaration if the last one has to be the declarator in it.
7722 ;; If `backup-at-type' is nil then the other variables have
7723 ;; undefined values.
7724 backup-at-type backup-type-start backup-id-start
7725 ;; This stores `kwd-sym' of the symbol before the current one.
7726 ;; This is needed to distinguish the C++11 version of "auto" from
7727 ;; the pre C++11 meaning.
7728 backup-kwd-sym
7729 ;; Set if we've found a specifier (apart from "typedef") that makes
7730 ;; the defined identifier(s) types.
7731 at-type-decl
7732 ;; Set if we've a "typedef" keyword.
7733 at-typedef
7734 ;; Set if we've found a specifier that can start a declaration
7735 ;; where there's no type.
7736 maybe-typeless
7737 ;; Save the value of kwd-sym between loops of the "Check for a
7738 ;; type" loop. Needed to distinguish a C++11 "auto" from a pre
7739 ;; C++11 one.
7740 prev-kwd-sym
7741 ;; If a specifier is found that also can be a type prefix,
7742 ;; these flags are set instead of those above. If we need to
7743 ;; back up an identifier, they are copied to the real flag
7744 ;; variables. Thus they only take effect if we fail to
7745 ;; interpret it as a type.
7746 backup-at-type-decl backup-maybe-typeless
7747 ;; Whether we've found a declaration or a cast. We might know
7748 ;; this before we've found the type in it. It's 'ids if we've
7749 ;; found two consecutive identifiers (usually a sure sign, but
7750 ;; we should allow that in labels too), and t if we've found a
7751 ;; specifier keyword (a 100% sure sign).
7752 at-decl-or-cast
7753 ;; Set when we need to back up to parse this as a declaration
7754 ;; but not as a cast.
7755 backup-if-not-cast
7756 ;; For casts, the return position.
7757 cast-end
7758 ;; Have we got a new-style C++11 "auto"?
7759 new-style-auto
7760 ;; Set when the symbol before `preceding-token-end' is known to
7761 ;; terminate the previous construct, or when we're at point-min.
7762 at-decl-start
7763 ;; Save `c-record-type-identifiers' and
7764 ;; `c-record-ref-identifiers' since ranges are recorded
7765 ;; speculatively and should be thrown away if it turns out
7766 ;; that it isn't a declaration or cast.
7767 (save-rec-type-ids c-record-type-identifiers)
7768 (save-rec-ref-ids c-record-ref-identifiers)
7769 ;; Set when we parse a declaration which might also be an expression,
7770 ;; such as "a *b". See CASE 16 and CASE 17.
7771 maybe-expression)
7772
7773 (save-excursion
7774 (goto-char preceding-token-end)
7775 (setq at-decl-start
7776 (or (bobp)
7777 (let ((tok-end (point)))
7778 (c-backward-token-2)
7779 (member (buffer-substring-no-properties (point) tok-end)
7780 c-pre-start-tokens)))))
7781
7782 (while (c-forward-annotation)
7783 (c-forward-syntactic-ws))
7784
7785 ;; Check for a type. Unknown symbols are treated as possible
7786 ;; types, but they could also be specifiers disguised through
7787 ;; macros like __INLINE__, so we recognize both types and known
7788 ;; specifiers after them too.
7789 (while
7790 (let* ((start (point)) kwd-sym kwd-clause-end found-type noise-start)
7791
7792 (cond
7793 ;; Look for a specifier keyword clause.
7794 ((or (looking-at c-prefix-spec-kwds-re)
7795 (and (c-major-mode-is 'java-mode)
7796 (looking-at "@[A-Za-z0-9]+")))
7797 (save-match-data
7798 (if (looking-at c-typedef-key)
7799 (setq at-typedef t)))
7800 (setq kwd-sym (c-keyword-sym (match-string 1)))
7801 (save-excursion
7802 (c-forward-keyword-clause 1)
7803 (setq kwd-clause-end (point))))
7804 ((and c-opt-cpp-prefix
7805 (looking-at c-noise-macro-with-parens-name-re))
7806 (setq noise-start (point))
7807 (c-forward-noise-clause)
7808 (setq kwd-clause-end (point))))
7809
7810 (when (setq found-type (c-forward-type t)) ; brace-block-too
7811 ;; Found a known or possible type or a prefix of a known type.
7812 (when (and (c-major-mode-is 'c++-mode) ; C++11 style "auto"?
7813 (eq prev-kwd-sym (c-keyword-sym "auto"))
7814 (looking-at "[=(]")) ; FIXME!!! proper regexp.
7815 (setq new-style-auto t)
7816 (setq found-type nil)
7817 (goto-char start)) ; position of foo in "auto foo"
7818
7819 (when at-type
7820 ;; Got two identifiers with nothing but whitespace
7821 ;; between them. That can only happen in declarations.
7822 (setq at-decl-or-cast 'ids)
7823
7824 (when (eq at-type 'found)
7825 ;; If the previous identifier is a found type we
7826 ;; record it as a real one; it might be some sort of
7827 ;; alias for a prefix like "unsigned".
7828 (save-excursion
7829 (goto-char type-start)
7830 (let ((c-promote-possible-types t))
7831 (c-forward-type)))))
7832
7833 (setq backup-at-type at-type
7834 backup-type-start type-start
7835 backup-id-start id-start
7836 backup-kwd-sym kwd-sym
7837 at-type found-type
7838 type-start start
7839 id-start (point)
7840 ;; The previous ambiguous specifier/type turned out
7841 ;; to be a type since we've parsed another one after
7842 ;; it, so clear these backup flags.
7843 backup-at-type-decl nil
7844 backup-maybe-typeless nil))
7845
7846 (if (or kwd-sym noise-start)
7847 (progn
7848 ;; Handle known specifier keywords and
7849 ;; `c-decl-hangon-kwds' which can occur after known
7850 ;; types.
7851
7852 (if (or (c-keyword-member kwd-sym 'c-decl-hangon-kwds)
7853 noise-start)
7854 ;; It's a hang-on keyword or noise clause that can occur
7855 ;; anywhere.
7856 (progn
7857 (if at-type
7858 ;; Move the identifier start position if
7859 ;; we've passed a type.
7860 (setq id-start kwd-clause-end)
7861 ;; Otherwise treat this as a specifier and
7862 ;; move the fallback position.
7863 (setq start-pos kwd-clause-end))
7864 (goto-char kwd-clause-end))
7865
7866 ;; It's an ordinary specifier so we know that
7867 ;; anything before this can't be the type.
7868 (setq backup-at-type nil
7869 start-pos kwd-clause-end)
7870
7871 (if found-type
7872 ;; It's ambiguous whether this keyword is a
7873 ;; specifier or a type prefix, so set the backup
7874 ;; flags. (It's assumed that `c-forward-type'
7875 ;; moved further than `c-forward-keyword-clause'.)
7876 (progn
7877 (when (c-keyword-member kwd-sym 'c-typedef-decl-kwds)
7878 (setq backup-at-type-decl t))
7879 (when (c-keyword-member kwd-sym 'c-typeless-decl-kwds)
7880 (setq backup-maybe-typeless t)))
7881
7882 (when (c-keyword-member kwd-sym 'c-typedef-decl-kwds)
7883 ;; This test only happens after we've scanned a type.
7884 ;; So, with valid syntax, kwd-sym can't be 'typedef.
7885 (setq at-type-decl t))
7886 (when (c-keyword-member kwd-sym 'c-typeless-decl-kwds)
7887 (setq maybe-typeless t))
7888
7889 ;; Haven't matched a type so it's an unambiguous
7890 ;; specifier keyword and we know we're in a
7891 ;; declaration.
7892 (setq at-decl-or-cast t)
7893 (setq prev-kwd-sym kwd-sym)
7894
7895 (goto-char kwd-clause-end))))
7896
7897 ;; If the type isn't known we continue so that we'll jump
7898 ;; over all specifiers and type identifiers. The reason
7899 ;; to do this for a known type prefix is to make things
7900 ;; like "unsigned INT16" work.
7901 (and found-type (not (eq found-type t))))))
7902
7903 (cond
7904 ((eq at-type t)
7905 ;; If a known type was found, we still need to skip over any
7906 ;; hangon keyword clauses after it. Otherwise it has already
7907 ;; been done in the loop above.
7908 (while
7909 (cond ((looking-at c-decl-hangon-key)
7910 (c-forward-keyword-clause 1))
7911 ((and c-opt-cpp-prefix
7912 (looking-at c-noise-macro-with-parens-name-re))
7913 (c-forward-noise-clause))))
7914 (setq id-start (point)))
7915
7916 ((eq at-type 'prefix)
7917 ;; A prefix type is itself a primitive type when it's not
7918 ;; followed by another type.
7919 (setq at-type t))
7920
7921 ((not at-type)
7922 ;; Got no type but set things up to continue anyway to handle
7923 ;; the various cases when a declaration doesn't start with a
7924 ;; type.
7925 (setq id-start start-pos))
7926
7927 ((and (eq at-type 'maybe)
7928 (c-major-mode-is 'c++-mode))
7929 ;; If it's C++ then check if the last "type" ends on the form
7930 ;; "foo::foo" or "foo::~foo", i.e. if it's the name of a
7931 ;; (con|de)structor.
7932 (save-excursion
7933 (let (name end-2 end-1)
7934 (goto-char id-start)
7935 (c-backward-syntactic-ws)
7936 (setq end-2 (point))
7937 (when (and
7938 (c-simple-skip-symbol-backward)
7939 (progn
7940 (setq name
7941 (buffer-substring-no-properties (point) end-2))
7942 ;; Cheating in the handling of syntactic ws below.
7943 (< (skip-chars-backward ":~ \t\n\r\v\f") 0))
7944 (progn
7945 (setq end-1 (point))
7946 (c-simple-skip-symbol-backward))
7947 (>= (point) type-start)
7948 (equal (buffer-substring-no-properties (point) end-1)
7949 name))
7950 ;; It is a (con|de)structor name. In that case the
7951 ;; declaration is typeless so zap out any preceding
7952 ;; identifier(s) that we might have taken as types.
7953 (goto-char type-start)
7954 (setq at-type nil
7955 backup-at-type nil
7956 id-start type-start))))))
7957
7958 ;; Check for and step over a type decl expression after the thing
7959 ;; that is or might be a type. This can't be skipped since we
7960 ;; need the correct end position of the declarator for
7961 ;; `max-type-decl-end-*'.
7962 (let ((start (point)) (paren-depth 0) pos
7963 ;; True if there's a non-open-paren match of
7964 ;; `c-type-decl-prefix-key'.
7965 got-prefix
7966 ;; True if the declarator is surrounded by a parenthesis pair.
7967 got-parens
7968 ;; True if there is an identifier in the declarator.
7969 got-identifier
7970 ;; True if there's a non-close-paren match of
7971 ;; `c-type-decl-suffix-key'.
7972 got-suffix
7973 ;; True if there's a prefix match outside the outermost
7974 ;; paren pair that surrounds the declarator.
7975 got-prefix-before-parens
7976 ;; True if there's a suffix match outside the outermost
7977 ;; paren pair that surrounds the declarator. The value is
7978 ;; the position of the first suffix match.
7979 got-suffix-after-parens
7980 ;; True if we've parsed the type decl to a token that is
7981 ;; known to end declarations in this context.
7982 at-decl-end
7983 ;; The earlier values of `at-type' and `type-start' if we've
7984 ;; shifted the type backwards.
7985 identifier-type identifier-start
7986 ;; If `c-parse-and-markup-<>-arglists' is set we need to
7987 ;; turn it off during the name skipping below to avoid
7988 ;; getting `c-type' properties that might be bogus. That
7989 ;; can happen since we don't know if
7990 ;; `c-restricted-<>-arglists' will be correct inside the
7991 ;; arglist paren that gets entered.
7992 c-parse-and-markup-<>-arglists
7993 ;; Start of the identifier for which `got-identifier' was set.
7994 name-start)
7995
7996 (goto-char id-start)
7997
7998 ;; Skip over type decl prefix operators. (Note similar code in
7999 ;; `c-forward-declarator'.)
8000 (if (and c-recognize-typeless-decls
8001 (equal c-type-decl-prefix-key "\\<\\>"))
8002 (when (eq (char-after) ?\()
8003 (progn
8004 (setq paren-depth (1+ paren-depth))
8005 (forward-char)))
8006 (while (and (looking-at c-type-decl-prefix-key)
8007 (if (and (c-major-mode-is 'c++-mode)
8008 (match-beginning 3))
8009 ;; If the third submatch matches in C++ then
8010 ;; we're looking at an identifier that's a
8011 ;; prefix only if it specifies a member pointer.
8012 (when (progn (setq pos (point))
8013 (setq got-identifier (c-forward-name)))
8014 (setq name-start pos)
8015 (if (looking-at "\\(::\\)")
8016 ;; We only check for a trailing "::" and
8017 ;; let the "*" that should follow be
8018 ;; matched in the next round.
8019 (progn (setq got-identifier nil) t)
8020 ;; It turned out to be the real identifier,
8021 ;; so stop.
8022 nil))
8023 t))
8024
8025 (if (eq (char-after) ?\()
8026 (progn
8027 (setq paren-depth (1+ paren-depth))
8028 (forward-char))
8029 (unless got-prefix-before-parens
8030 (setq got-prefix-before-parens (= paren-depth 0)))
8031 (setq got-prefix t)
8032 (goto-char (match-end 1)))
8033 (c-forward-syntactic-ws)))
8034
8035 (setq got-parens (> paren-depth 0))
8036
8037 ;; Skip over an identifier.
8038 (or got-identifier
8039 (and (looking-at c-identifier-start)
8040 (setq pos (point))
8041 (setq got-identifier (c-forward-name))
8042 (setq name-start pos)))
8043
8044 ;; Skip over type decl suffix operators and trailing noise macros.
8045 (while
8046 (cond
8047 ((and c-opt-cpp-prefix
8048 (looking-at c-noise-macro-with-parens-name-re))
8049 (c-forward-noise-clause))
8050
8051 ((looking-at c-type-decl-suffix-key)
8052 (if (eq (char-after) ?\))
8053 (when (> paren-depth 0)
8054 (setq paren-depth (1- paren-depth))
8055 (forward-char)
8056 t)
8057 (when (if (save-match-data (looking-at "\\s("))
8058 (c-safe (c-forward-sexp 1) t)
8059 (goto-char (match-end 1))
8060 t)
8061 (when (and (not got-suffix-after-parens)
8062 (= paren-depth 0))
8063 (setq got-suffix-after-parens (match-beginning 0)))
8064 (setq got-suffix t))))
8065
8066 (t
8067 ;; No suffix matched. We might have matched the
8068 ;; identifier as a type and the open paren of a
8069 ;; function arglist as a type decl prefix. In that
8070 ;; case we should "backtrack": Reinterpret the last
8071 ;; type as the identifier, move out of the arglist and
8072 ;; continue searching for suffix operators.
8073 ;;
8074 ;; Do this even if there's no preceding type, to cope
8075 ;; with old style function declarations in K&R C,
8076 ;; (con|de)structors in C++ and `c-typeless-decl-kwds'
8077 ;; style declarations. That isn't applicable in an
8078 ;; arglist context, though.
8079 (when (and (= paren-depth 1)
8080 (not got-prefix-before-parens)
8081 (not (eq at-type t))
8082 (or backup-at-type
8083 maybe-typeless
8084 backup-maybe-typeless
8085 (when c-recognize-typeless-decls
8086 (not context)))
8087 (setq pos (c-up-list-forward (point)))
8088 (eq (char-before pos) ?\)))
8089 (c-fdoc-shift-type-backward)
8090 (goto-char pos)
8091 t)))
8092
8093 (c-forward-syntactic-ws))
8094
8095 (when (or (and new-style-auto
8096 (looking-at c-auto-ops-re))
8097 (and (or maybe-typeless backup-maybe-typeless)
8098 (not got-identifier)
8099 (not got-prefix)
8100 at-type))
8101 ;; Have found no identifier but `c-typeless-decl-kwds' has
8102 ;; matched so we know we're inside a declaration. The
8103 ;; preceding type must be the identifier instead.
8104 (c-fdoc-shift-type-backward))
8105
8106 ;; Prepare the "-> type;" for fontification later on.
8107 (when (and new-style-auto
8108 (looking-at c-haskell-op-re))
8109 (save-excursion
8110 (goto-char (match-end 0))
8111 (c-forward-syntactic-ws)
8112 (setq type-start (point))
8113 (setq at-type (c-forward-type))))
8114
8115 (setq
8116 at-decl-or-cast
8117 (catch 'at-decl-or-cast
8118
8119 ;; CASE 1
8120 (when (> paren-depth 0)
8121 ;; Encountered something inside parens that isn't matched by
8122 ;; the `c-type-decl-*' regexps, so it's not a type decl
8123 ;; expression. Try to skip out to the same paren depth to
8124 ;; not confuse the cast check below.
8125 (c-safe (goto-char (scan-lists (point) 1 paren-depth)))
8126 ;; If we've found a specifier keyword then it's a
8127 ;; declaration regardless.
8128 (throw 'at-decl-or-cast (eq at-decl-or-cast t)))
8129
8130 (setq at-decl-end
8131 (looking-at (cond ((eq context '<>) "[,>]")
8132 (context "[,)]")
8133 (t "[,;]"))))
8134
8135 ;; Now we've collected info about various characteristics of
8136 ;; the construct we're looking at. Below follows a decision
8137 ;; tree based on that. It's ordered to check more certain
8138 ;; signs before less certain ones.
8139
8140 (if got-identifier
8141 (progn
8142
8143 ;; CASE 2
8144 (when (and (or at-type maybe-typeless)
8145 (not (or got-prefix got-parens)))
8146 ;; Got another identifier directly after the type, so it's a
8147 ;; declaration.
8148 (throw 'at-decl-or-cast t))
8149
8150 (when (and got-parens
8151 (not got-prefix)
8152 ;; (not got-suffix-after-parens)
8153 (or backup-at-type
8154 maybe-typeless
8155 backup-maybe-typeless
8156 (eq at-decl-or-cast t)
8157 ;; Check whether we have "bar (gnu);" where we
8158 ;; are directly inside a class (etc.) called "bar".
8159 (save-excursion
8160 (and
8161 (progn
8162 (goto-char name-start)
8163 (not (memq (c-forward-type) '(nil maybe))))
8164 (progn
8165 (goto-char id-start)
8166 (c-directly-in-class-called-p
8167 (buffer-substring
8168 type-start
8169 (progn
8170 (goto-char type-start)
8171 (c-forward-type)
8172 (c-backward-syntactic-ws)
8173 (point)))))))))
8174 ;; Got a declaration of the form "foo bar (gnu);" or "bar
8175 ;; (gnu);" where we've recognized "bar" as the type and "gnu"
8176 ;; as the declarator, and in the latter case, checked that
8177 ;; "bar (gnu)" appears directly inside the class "bar". In
8178 ;; this case it's however more likely that "bar" is the
8179 ;; declarator and "gnu" a function argument or initializer
8180 ;; (if `c-recognize-paren-inits' is set), since the parens
8181 ;; around "gnu" would be superfluous if it's a declarator.
8182 ;; Shift the type one step backward.
8183 (c-fdoc-shift-type-backward)))
8184
8185 ;; Found no identifier.
8186
8187 (if backup-at-type
8188 (progn
8189
8190 ;; CASE 3
8191 (when (= (point) start)
8192 ;; Got a plain list of identifiers. If a colon follows it's
8193 ;; a valid label, or maybe a bitfield. Otherwise the last
8194 ;; one probably is the declared identifier and we should
8195 ;; back up to the previous type, providing it isn't a cast.
8196 (if (and (eq (char-after) ?:)
8197 (not (c-major-mode-is 'java-mode)))
8198 (cond
8199 ;; If we've found a specifier keyword then it's a
8200 ;; declaration regardless.
8201 ((eq at-decl-or-cast t)
8202 (throw 'at-decl-or-cast t))
8203 ((and c-has-bitfields
8204 (eq at-decl-or-cast 'ids)) ; bitfield.
8205 (setq backup-if-not-cast t)
8206 (throw 'at-decl-or-cast t)))
8207
8208 (setq backup-if-not-cast t)
8209 (throw 'at-decl-or-cast t)))
8210
8211 ;; CASE 4
8212 (when (and got-suffix
8213 (not got-prefix)
8214 (not got-parens))
8215 ;; Got a plain list of identifiers followed by some suffix.
8216 ;; If this isn't a cast then the last identifier probably is
8217 ;; the declared one and we should back up to the previous
8218 ;; type.
8219 (setq backup-if-not-cast t)
8220 (throw 'at-decl-or-cast t)))
8221
8222 ;; CASE 5
8223 (when (eq at-type t)
8224 ;; If the type is known we know that there can't be any
8225 ;; identifier somewhere else, and it's only in declarations in
8226 ;; e.g. function prototypes and in casts that the identifier may
8227 ;; be left out.
8228 (throw 'at-decl-or-cast t))
8229
8230 (when (= (point) start)
8231 ;; Only got a single identifier (parsed as a type so far).
8232 ;; CASE 6
8233 (if (and
8234 ;; Check that the identifier isn't at the start of an
8235 ;; expression.
8236 at-decl-end
8237 (cond
8238 ((eq context 'decl)
8239 ;; Inside an arglist that contains declarations. If K&R
8240 ;; style declarations and parenthesis style initializers
8241 ;; aren't allowed then the single identifier must be a
8242 ;; type, else we require that it's known or found
8243 ;; (primitive types are handled above).
8244 (or (and (not c-recognize-knr-p)
8245 (not c-recognize-paren-inits))
8246 (memq at-type '(known found))))
8247 ((eq context '<>)
8248 ;; Inside a template arglist. Accept known and found
8249 ;; types; other identifiers could just as well be
8250 ;; constants in C++.
8251 (memq at-type '(known found)))))
8252 (throw 'at-decl-or-cast t)
8253 ;; CASE 7
8254 ;; Can't be a valid declaration or cast, but if we've found a
8255 ;; specifier it can't be anything else either, so treat it as
8256 ;; an invalid/unfinished declaration or cast.
8257 (throw 'at-decl-or-cast at-decl-or-cast))))
8258
8259 (if (and got-parens
8260 (not got-prefix)
8261 (not context)
8262 (not (eq at-type t))
8263 (or backup-at-type
8264 maybe-typeless
8265 backup-maybe-typeless
8266 (when c-recognize-typeless-decls
8267 (or (not got-suffix)
8268 (not (looking-at
8269 c-after-suffixed-type-maybe-decl-key))))))
8270 ;; Got an empty paren pair and a preceding type that probably
8271 ;; really is the identifier. Shift the type backwards to make
8272 ;; the last one the identifier. This is analogous to the
8273 ;; "backtracking" done inside the `c-type-decl-suffix-key' loop
8274 ;; above.
8275 ;;
8276 ;; Exception: In addition to the conditions in that
8277 ;; "backtracking" code, do not shift backward if we're not
8278 ;; looking at either `c-after-suffixed-type-decl-key' or "[;,]".
8279 ;; Since there's no preceding type, the shift would mean that
8280 ;; the declaration is typeless. But if the regexp doesn't match
8281 ;; then we will simply fall through in the tests below and not
8282 ;; recognize it at all, so it's better to try it as an abstract
8283 ;; declarator instead.
8284 (c-fdoc-shift-type-backward)
8285
8286 ;; Still no identifier.
8287 ;; CASE 8
8288 (when (and got-prefix (or got-parens got-suffix))
8289 ;; Require `got-prefix' together with either `got-parens' or
8290 ;; `got-suffix' to recognize it as an abstract declarator:
8291 ;; `got-parens' only is probably an empty function call.
8292 ;; `got-suffix' only can build an ordinary expression together
8293 ;; with the preceding identifier which we've taken as a type.
8294 ;; We could actually accept on `got-prefix' only, but that can
8295 ;; easily occur temporarily while writing an expression so we
8296 ;; avoid that case anyway. We could do a better job if we knew
8297 ;; the point when the fontification was invoked.
8298 (throw 'at-decl-or-cast t))
8299
8300 ;; CASE 9
8301 (when (and at-type
8302 (not got-prefix)
8303 (not got-parens)
8304 got-suffix-after-parens
8305 (eq (char-after got-suffix-after-parens) ?\())
8306 ;; Got a type, no declarator but a paren suffix. I.e. it's a
8307 ;; normal function call after all (or perhaps a C++ style object
8308 ;; instantiation expression).
8309 (throw 'at-decl-or-cast nil))))
8310
8311 ;; CASE 10
8312 (when at-decl-or-cast
8313 ;; By now we've located the type in the declaration that we know
8314 ;; we're in.
8315 (throw 'at-decl-or-cast t))
8316
8317 ;; CASE 11
8318 (when (and got-identifier
8319 (not context)
8320 (looking-at c-after-suffixed-type-decl-key)
8321 (if (and got-parens
8322 (not got-prefix)
8323 (not got-suffix)
8324 (not (eq at-type t)))
8325 ;; Shift the type backward in the case that there's a
8326 ;; single identifier inside parens. That can only
8327 ;; occur in K&R style function declarations so it's
8328 ;; more likely that it really is a function call.
8329 ;; Therefore we only do this after
8330 ;; `c-after-suffixed-type-decl-key' has matched.
8331 (progn (c-fdoc-shift-type-backward) t)
8332 got-suffix-after-parens))
8333 ;; A declaration according to `c-after-suffixed-type-decl-key'.
8334 (throw 'at-decl-or-cast t))
8335
8336 ;; CASE 12
8337 (when (and (or got-prefix (not got-parens))
8338 (memq at-type '(t known)))
8339 ;; It's a declaration if a known type precedes it and it can't be a
8340 ;; function call.
8341 (throw 'at-decl-or-cast t))
8342
8343 ;; If we get here we can't tell if this is a type decl or a normal
8344 ;; expression by looking at it alone. (That's under the assumption
8345 ;; that normal expressions always can look like type decl expressions,
8346 ;; which isn't really true but the cases where it doesn't hold are so
8347 ;; uncommon (e.g. some placements of "const" in C++) it's not worth
8348 ;; the effort to look for them.)
8349
8350 ;;; 2008-04-16: commented out the next form, to allow the function to recognize
8351 ;;; "foo (int bar)" in CC (an implicit type (in class foo) without a semicolon)
8352 ;;; as a(n almost complete) declaration, enabling it to be fontified.
8353 ;; CASE 13
8354 ;; (unless (or at-decl-end (looking-at "=[^=]"))
8355 ;; If this is a declaration it should end here or its initializer(*)
8356 ;; should start here, so check for allowed separation tokens. Note
8357 ;; that this rule doesn't work e.g. with a K&R arglist after a
8358 ;; function header.
8359 ;;
8360 ;; *) Don't check for C++ style initializers using parens
8361 ;; since those already have been matched as suffixes.
8362 ;;
8363 ;; If `at-decl-or-cast' is then we've found some other sign that
8364 ;; it's a declaration or cast, so then it's probably an
8365 ;; invalid/unfinished one.
8366 ;; (throw 'at-decl-or-cast at-decl-or-cast))
8367
8368 ;; Below are tests that only should be applied when we're certain to
8369 ;; not have parsed halfway through an expression.
8370
8371 ;; CASE 14
8372 (when (memq at-type '(t known))
8373 ;; The expression starts with a known type so treat it as a
8374 ;; declaration.
8375 (throw 'at-decl-or-cast t))
8376
8377 ;; CASE 15
8378 (when (and (c-major-mode-is 'c++-mode)
8379 ;; In C++ we check if the identifier is a known type, since
8380 ;; (con|de)structors use the class name as identifier.
8381 ;; We've always shifted over the identifier as a type and
8382 ;; then backed up again in this case.
8383 identifier-type
8384 (or (memq identifier-type '(found known))
8385 (and (eq (char-after identifier-start) ?~)
8386 ;; `at-type' probably won't be 'found for
8387 ;; destructors since the "~" is then part of the
8388 ;; type name being checked against the list of
8389 ;; known types, so do a check without that
8390 ;; operator.
8391 (or (save-excursion
8392 (goto-char (1+ identifier-start))
8393 (c-forward-syntactic-ws)
8394 (c-with-syntax-table
8395 c-identifier-syntax-table
8396 (looking-at c-known-type-key)))
8397 (save-excursion
8398 (goto-char (1+ identifier-start))
8399 ;; We have already parsed the type earlier,
8400 ;; so it'd be possible to cache the end
8401 ;; position instead of redoing it here, but
8402 ;; then we'd need to keep track of another
8403 ;; position everywhere.
8404 (c-check-type (point)
8405 (progn (c-forward-type)
8406 (point))))))))
8407 (throw 'at-decl-or-cast t))
8408
8409 (if got-identifier
8410 (progn
8411 ;; CASE 16
8412 (when (and got-prefix-before-parens
8413 at-type
8414 (or at-decl-end (looking-at "=[^=]"))
8415 (not context)
8416 (or (not got-suffix)
8417 at-decl-start))
8418 ;; Got something like "foo * bar;". Since we're not inside
8419 ;; an arglist it would be a meaningless expression because
8420 ;; the result isn't used. We therefore choose to recognize
8421 ;; it as a declaration. We only allow a suffix (which makes
8422 ;; the construct look like a function call) when
8423 ;; `at-decl-start' provides additional evidence that we do
8424 ;; have a declaration.
8425 (setq maybe-expression t)
8426 (throw 'at-decl-or-cast t))
8427
8428 ;; CASE 17
8429 (when (and (or got-suffix-after-parens
8430 (looking-at "=[^=]"))
8431 (eq at-type 'found)
8432 (not (eq context 'arglist)))
8433 ;; Got something like "a (*b) (c);" or "a (b) = c;". It could
8434 ;; be an odd expression or it could be a declaration. Treat
8435 ;; it as a declaration if "a" has been used as a type
8436 ;; somewhere else (if it's a known type we won't get here).
8437 (setq maybe-expression t)
8438 (throw 'at-decl-or-cast t)))
8439
8440 ;; CASE 18
8441 (when (and context
8442 (or got-prefix
8443 (and (eq context 'decl)
8444 (not c-recognize-paren-inits)
8445 (or got-parens got-suffix))))
8446 ;; Got a type followed by an abstract declarator. If `got-prefix'
8447 ;; is set it's something like "a *" without anything after it. If
8448 ;; `got-parens' or `got-suffix' is set it's "a()", "a[]", "a()[]",
8449 ;; or similar, which we accept only if the context rules out
8450 ;; expressions.
8451 (throw 'at-decl-or-cast t)))
8452
8453 ;; If we had a complete symbol table here (which rules out
8454 ;; `c-found-types') we should return t due to the disambiguation rule
8455 ;; (in at least C++) that anything that can be parsed as a declaration
8456 ;; is a declaration. Now we're being more defensive and prefer to
8457 ;; highlight things like "foo (bar);" as a declaration only if we're
8458 ;; inside an arglist that contains declarations.
8459 ;; CASE 19
8460 (eq context 'decl))))
8461
8462 ;; The point is now after the type decl expression.
8463
8464 (cond
8465 ;; Check for a cast.
8466 ((save-excursion
8467 (and
8468 c-cast-parens
8469
8470 ;; Should be the first type/identifier in a cast paren.
8471 (> preceding-token-end (point-min))
8472 (memq (char-before preceding-token-end) c-cast-parens)
8473
8474 ;; The closing paren should follow.
8475 (progn
8476 (c-forward-syntactic-ws)
8477 (looking-at "\\s)"))
8478
8479 ;; There should be a primary expression after it.
8480 (let (pos)
8481 (forward-char)
8482 (c-forward-syntactic-ws)
8483 (setq cast-end (point))
8484 (and (looking-at c-primary-expr-regexp)
8485 (progn
8486 (setq pos (match-end 0))
8487 (or
8488 ;; Check if the expression begins with a prefix keyword.
8489 (match-beginning 2)
8490 (if (match-beginning 1)
8491 ;; Expression begins with an ambiguous operator. Treat
8492 ;; it as a cast if it's a type decl or if we've
8493 ;; recognized the type somewhere else.
8494 (or at-decl-or-cast
8495 (memq at-type '(t known found)))
8496 ;; Unless it's a keyword, it's the beginning of a primary
8497 ;; expression.
8498 (not (looking-at c-keywords-regexp)))))
8499 ;; If `c-primary-expr-regexp' matched a nonsymbol token, check
8500 ;; that it matched a whole one so that we don't e.g. confuse
8501 ;; the operator '-' with '->'. It's ok if it matches further,
8502 ;; though, since it e.g. can match the float '.5' while the
8503 ;; operator regexp only matches '.'.
8504 (or (not (looking-at c-nonsymbol-token-regexp))
8505 (<= (match-end 0) pos))))
8506
8507 ;; There should either be a cast before it or something that isn't an
8508 ;; identifier or close paren.
8509 (> preceding-token-end (point-min))
8510 (progn
8511 (goto-char (1- preceding-token-end))
8512 (or (eq (point) last-cast-end)
8513 (progn
8514 (c-backward-syntactic-ws)
8515 (if (< (skip-syntax-backward "w_") 0)
8516 ;; It's a symbol. Accept it only if it's one of the
8517 ;; keywords that can precede an expression (without
8518 ;; surrounding parens).
8519 (looking-at c-simple-stmt-key)
8520 (and
8521 ;; Check that it isn't a close paren (block close is ok,
8522 ;; though).
8523 (not (memq (char-before) '(?\) ?\])))
8524 ;; Check that it isn't a nonsymbol identifier.
8525 (not (c-on-identifier)))))))))
8526
8527 ;; Handle the cast.
8528 (when (and c-record-type-identifiers at-type (not (eq at-type t)))
8529 (let ((c-promote-possible-types t))
8530 (goto-char type-start)
8531 (c-forward-type)))
8532
8533 (goto-char cast-end)
8534 'cast)
8535
8536 (at-decl-or-cast
8537 ;; We're at a declaration. Highlight the type and the following
8538 ;; declarators.
8539
8540 (when backup-if-not-cast
8541 (c-fdoc-shift-type-backward t))
8542
8543 (when (and (eq context 'decl) (looking-at ","))
8544 ;; Make sure to propagate the `c-decl-arg-start' property to
8545 ;; the next argument if it's set in this one, to cope with
8546 ;; interactive refontification.
8547 (c-put-c-type-property (point) 'c-decl-arg-start))
8548
8549 ;; Record the type's coordinates in `c-record-type-identifiers' for
8550 ;; later fontification.
8551 (when (and c-record-type-identifiers at-type ;; (not (eq at-type t))
8552 ;; There seems no reason to exclude a token from
8553 ;; fontification just because it's "a known type that can't
8554 ;; be a name or other expression". 2013-09-18.
8555 )
8556 (let ((c-promote-possible-types t))
8557 (save-excursion
8558 (goto-char type-start)
8559 (c-forward-type))))
8560
8561 (list id-start
8562 (and (or at-type-decl at-typedef)
8563 (cons at-type-decl at-typedef))
8564 maybe-expression
8565 type-start))
8566
8567 (t
8568 ;; False alarm. Restore the recorded ranges.
8569 (setq c-record-type-identifiers save-rec-type-ids
8570 c-record-ref-identifiers save-rec-ref-ids)
8571 nil))))
8572
8573 (defun c-forward-label (&optional assume-markup preceding-token-end limit)
8574 ;; Assuming that point is at the beginning of a token, check if it starts a
8575 ;; label and if so move over it and return non-nil (t in default situations,
8576 ;; specific symbols (see below) for interesting situations), otherwise don't
8577 ;; move and return nil. "Label" here means "most things with a colon".
8578 ;;
8579 ;; More precisely, a "label" is regarded as one of:
8580 ;; (i) a goto target like "foo:" - returns the symbol `goto-target';
8581 ;; (ii) A case label - either the entire construct "case FOO:", or just the
8582 ;; bare "case", should the colon be missing. We return t;
8583 ;; (iii) a keyword which needs a colon, like "default:" or "private:"; We
8584 ;; return t;
8585 ;; (iv) One of QT's "extended" C++ variants of
8586 ;; "private:"/"protected:"/"public:"/"more:" looking like "public slots:".
8587 ;; Returns the symbol `qt-2kwds-colon'.
8588 ;; (v) QT's construct "signals:". Returns the symbol `qt-1kwd-colon'.
8589 ;; (vi) One of the keywords matched by `c-opt-extra-label-key' (without any
8590 ;; colon). Currently (2006-03), this applies only to Objective C's
8591 ;; keywords "@private", "@protected", and "@public". Returns t.
8592 ;;
8593 ;; One of the things which will NOT be recognized as a label is a bit-field
8594 ;; element of a struct, something like "int foo:5".
8595 ;;
8596 ;; The end of the label is taken to be just after the colon, or the end of
8597 ;; the first submatch in `c-opt-extra-label-key'. The point is directly
8598 ;; after the end on return. The terminating char gets marked with
8599 ;; `c-decl-end' to improve recognition of the following declaration or
8600 ;; statement.
8601 ;;
8602 ;; If ASSUME-MARKUP is non-nil, it's assumed that the preceding
8603 ;; label, if any, has already been marked up like that.
8604 ;;
8605 ;; If PRECEDING-TOKEN-END is given, it should be the first position
8606 ;; after the preceding token, i.e. on the other side of the
8607 ;; syntactic ws from the point. Use a value less than or equal to
8608 ;; (point-min) if the point is at the first token in (the visible
8609 ;; part of) the buffer.
8610 ;;
8611 ;; The optional LIMIT limits the forward scan for the colon.
8612 ;;
8613 ;; This function records the ranges of the label symbols on
8614 ;; `c-record-ref-identifiers' if `c-record-type-identifiers' (!) is
8615 ;; non-nil.
8616 ;;
8617 ;; This function might do hidden buffer changes.
8618
8619 (let ((start (point))
8620 label-end
8621 qt-symbol-idx
8622 macro-start ; if we're in one.
8623 label-type
8624 kwd)
8625 (cond
8626 ;; "case" or "default" (Doesn't apply to AWK).
8627 ((looking-at c-label-kwds-regexp)
8628 (let ((kwd-end (match-end 1)))
8629 ;; Record only the keyword itself for fontification, since in
8630 ;; case labels the following is a constant expression and not
8631 ;; a label.
8632 (when c-record-type-identifiers
8633 (c-record-ref-id (cons (match-beginning 1) kwd-end)))
8634
8635 ;; Find the label end.
8636 (goto-char kwd-end)
8637 (setq label-type
8638 (if (and (c-syntactic-re-search-forward
8639 ;; Stop on chars that aren't allowed in expressions,
8640 ;; and on operator chars that would be meaningless
8641 ;; there. FIXME: This doesn't cope with ?: operators.
8642 "[;{=,@]\\|\\(\\=\\|[^:]\\):\\([^:]\\|\\'\\)"
8643 limit t t nil 1)
8644 (match-beginning 2))
8645
8646 (progn ; there's a proper :
8647 (goto-char (match-beginning 2)) ; just after the :
8648 (c-put-c-type-property (1- (point)) 'c-decl-end)
8649 t)
8650
8651 ;; It's an unfinished label. We consider the keyword enough
8652 ;; to recognize it as a label, so that it gets fontified.
8653 ;; Leave the point at the end of it, but don't put any
8654 ;; `c-decl-end' marker.
8655 (goto-char kwd-end)
8656 t))))
8657
8658 ;; @private, @protected, @public, in Objective C, or similar.
8659 ((and c-opt-extra-label-key
8660 (looking-at c-opt-extra-label-key))
8661 ;; For a `c-opt-extra-label-key' match, we record the whole
8662 ;; thing for fontification. That's to get the leading '@' in
8663 ;; Objective-C protection labels fontified.
8664 (goto-char (match-end 1))
8665 (when c-record-type-identifiers
8666 (c-record-ref-id (cons (match-beginning 1) (point))))
8667 (c-put-c-type-property (1- (point)) 'c-decl-end)
8668 (setq label-type t))
8669
8670 ;; All other cases of labels.
8671 ((and c-recognize-colon-labels ; nil for AWK and IDL, otherwise t.
8672
8673 ;; A colon label must have something before the colon.
8674 (not (eq (char-after) ?:))
8675
8676 ;; Check that we're not after a token that can't precede a label.
8677 (or
8678 ;; Trivially succeeds when there's no preceding token.
8679 ;; Succeeds when we're at a virtual semicolon.
8680 (if preceding-token-end
8681 (<= preceding-token-end (point-min))
8682 (save-excursion
8683 (c-backward-syntactic-ws)
8684 (setq preceding-token-end (point))
8685 (or (bobp)
8686 (c-at-vsemi-p))))
8687
8688 ;; Check if we're after a label, if we're after a closing
8689 ;; paren that belong to statement, and with
8690 ;; `c-label-prefix-re'. It's done in different order
8691 ;; depending on `assume-markup' since the checks have
8692 ;; different expensiveness.
8693 (if assume-markup
8694 (or
8695 (eq (c-get-char-property (1- preceding-token-end) 'c-type)
8696 'c-decl-end)
8697
8698 (save-excursion
8699 (goto-char (1- preceding-token-end))
8700 (c-beginning-of-current-token)
8701 (or (looking-at c-label-prefix-re)
8702 (looking-at c-block-stmt-1-key)))
8703
8704 (and (eq (char-before preceding-token-end) ?\))
8705 (c-after-conditional)))
8706
8707 (or
8708 (save-excursion
8709 (goto-char (1- preceding-token-end))
8710 (c-beginning-of-current-token)
8711 (or (looking-at c-label-prefix-re)
8712 (looking-at c-block-stmt-1-key)))
8713
8714 (cond
8715 ((eq (char-before preceding-token-end) ?\))
8716 (c-after-conditional))
8717
8718 ((eq (char-before preceding-token-end) ?:)
8719 ;; Might be after another label, so check it recursively.
8720 (save-restriction
8721 (save-excursion
8722 (goto-char (1- preceding-token-end))
8723 ;; Essentially the same as the
8724 ;; `c-syntactic-re-search-forward' regexp below.
8725 (setq macro-start
8726 (save-excursion (and (c-beginning-of-macro)
8727 (point))))
8728 (if macro-start (narrow-to-region macro-start (point-max)))
8729 (c-syntactic-skip-backward "^-]:?;}=*/%&|,<>!@+" nil t)
8730 ;; Note: the following should work instead of the
8731 ;; narrow-to-region above. Investigate why not,
8732 ;; sometime. ACM, 2006-03-31.
8733 ;; (c-syntactic-skip-backward "^-]:?;}=*/%&|,<>!@+"
8734 ;; macro-start t)
8735 (let ((pte (point))
8736 ;; If the caller turned on recording for us,
8737 ;; it shouldn't apply when we check the
8738 ;; preceding label.
8739 c-record-type-identifiers)
8740 ;; A label can't start at a cpp directive. Check for
8741 ;; this, since c-forward-syntactic-ws would foul up on it.
8742 (unless (and c-opt-cpp-prefix (looking-at c-opt-cpp-prefix))
8743 (c-forward-syntactic-ws)
8744 (c-forward-label nil pte start))))))))))
8745
8746 ;; Point is still at the beginning of the possible label construct.
8747 ;;
8748 ;; Check that the next nonsymbol token is ":", or that we're in one
8749 ;; of QT's "slots" declarations. Allow '(' for the sake of macro
8750 ;; arguments. FIXME: Should build this regexp from the language
8751 ;; constants.
8752 (cond
8753 ;; public: protected: private:
8754 ((and
8755 (c-major-mode-is 'c++-mode)
8756 (search-forward-regexp
8757 "\\=p\\(r\\(ivate\\|otected\\)\\|ublic\\)\\>[^_]" nil t)
8758 (progn (backward-char)
8759 (c-forward-syntactic-ws limit)
8760 (looking-at ":\\([^:]\\|\\'\\)"))) ; A single colon.
8761 (forward-char)
8762 (setq label-type t))
8763 ;; QT double keyword like "protected slots:" or goto target.
8764 ((progn (goto-char start) nil))
8765 ((when (c-syntactic-re-search-forward
8766 "[ \t\n[:?;{=*/%&|,<>!@+-]" limit t t) ; not at EOB
8767 (backward-char)
8768 (setq label-end (point))
8769 (setq qt-symbol-idx
8770 (and (c-major-mode-is 'c++-mode)
8771 (string-match
8772 "\\(p\\(r\\(ivate\\|otected\\)\\|ublic\\)\\|more\\)\\>"
8773 (buffer-substring start (point)))))
8774 (c-forward-syntactic-ws limit)
8775 (cond
8776 ((looking-at ":\\([^:]\\|\\'\\)") ; A single colon.
8777 (forward-char)
8778 (setq label-type
8779 (if (or (string= "signals" ; Special QT macro
8780 (setq kwd (buffer-substring-no-properties start label-end)))
8781 (string= "Q_SIGNALS" kwd))
8782 'qt-1kwd-colon
8783 'goto-target)))
8784 ((and qt-symbol-idx
8785 (search-forward-regexp "\\=\\(slots\\|Q_SLOTS\\)\\>" limit t)
8786 (progn (c-forward-syntactic-ws limit)
8787 (looking-at ":\\([^:]\\|\\'\\)"))) ; A single colon
8788 (forward-char)
8789 (setq label-type 'qt-2kwds-colon)))))))
8790
8791 (save-restriction
8792 (narrow-to-region start (point))
8793
8794 ;; Check that `c-nonlabel-token-key' doesn't match anywhere.
8795 (catch 'check-label
8796 (goto-char start)
8797 (while (progn
8798 (when (looking-at c-nonlabel-token-key)
8799 (goto-char start)
8800 (setq label-type nil)
8801 (throw 'check-label nil))
8802 (and (c-safe (c-forward-sexp)
8803 (c-forward-syntactic-ws)
8804 t)
8805 (not (eobp)))))
8806
8807 ;; Record the identifiers in the label for fontification, unless
8808 ;; it begins with `c-label-kwds' in which case the following
8809 ;; identifiers are part of a (constant) expression that
8810 ;; shouldn't be fontified.
8811 (when (and c-record-type-identifiers
8812 (progn (goto-char start)
8813 (not (looking-at c-label-kwds-regexp))))
8814 (while (c-syntactic-re-search-forward c-symbol-key nil t)
8815 (c-record-ref-id (cons (match-beginning 0)
8816 (match-end 0)))))
8817
8818 (c-put-c-type-property (1- (point-max)) 'c-decl-end)
8819 (goto-char (point-max)))))
8820
8821 (t
8822 ;; Not a label.
8823 (goto-char start)))
8824 label-type))
8825
8826 (defun c-forward-objc-directive ()
8827 ;; Assuming the point is at the beginning of a token, try to move
8828 ;; forward to the end of the Objective-C directive that starts
8829 ;; there. Return t if a directive was fully recognized, otherwise
8830 ;; the point is moved as far as one could be successfully parsed and
8831 ;; nil is returned.
8832 ;;
8833 ;; This function records identifier ranges on
8834 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
8835 ;; `c-record-type-identifiers' is non-nil.
8836 ;;
8837 ;; This function might do hidden buffer changes.
8838
8839 (let ((start (point))
8840 start-char
8841 (c-promote-possible-types t)
8842 lim
8843 ;; Turn off recognition of angle bracket arglists while parsing
8844 ;; types here since the protocol reference list might then be
8845 ;; considered part of the preceding name or superclass-name.
8846 c-recognize-<>-arglists)
8847
8848 (if (or
8849 (when (looking-at
8850 (eval-when-compile
8851 (c-make-keywords-re t
8852 (append (c-lang-const c-protection-kwds objc)
8853 '("@end"))
8854 'objc-mode)))
8855 (goto-char (match-end 1))
8856 t)
8857
8858 (and
8859 (looking-at
8860 (eval-when-compile
8861 (c-make-keywords-re t
8862 '("@interface" "@implementation" "@protocol")
8863 'objc-mode)))
8864
8865 ;; Handle the name of the class itself.
8866 (progn
8867 ;; (c-forward-token-2) ; 2006/1/13 This doesn't move if the token's
8868 ;; at EOB.
8869 (goto-char (match-end 0))
8870 (setq lim (point))
8871 (c-skip-ws-forward)
8872 (c-forward-type))
8873
8874 (catch 'break
8875 ;; Look for ": superclass-name" or "( category-name )".
8876 (when (looking-at "[:(]")
8877 (setq start-char (char-after))
8878 (forward-char)
8879 (c-forward-syntactic-ws)
8880 (unless (c-forward-type) (throw 'break nil))
8881 (when (eq start-char ?\()
8882 (unless (eq (char-after) ?\)) (throw 'break nil))
8883 (forward-char)
8884 (c-forward-syntactic-ws)))
8885
8886 ;; Look for a protocol reference list.
8887 (if (eq (char-after) ?<)
8888 (let ((c-recognize-<>-arglists t)
8889 (c-parse-and-markup-<>-arglists t)
8890 c-restricted-<>-arglists)
8891 (c-forward-<>-arglist t))
8892 t))))
8893
8894 (progn
8895 (c-backward-syntactic-ws lim)
8896 (c-clear-c-type-property start (1- (point)) 'c-decl-end)
8897 (c-put-c-type-property (1- (point)) 'c-decl-end)
8898 t)
8899
8900 (c-clear-c-type-property start (point) 'c-decl-end)
8901 nil)))
8902
8903 (defun c-beginning-of-inheritance-list (&optional lim)
8904 ;; Go to the first non-whitespace after the colon that starts a
8905 ;; multiple inheritance introduction. Optional LIM is the farthest
8906 ;; back we should search.
8907 ;;
8908 ;; This function might do hidden buffer changes.
8909 (c-with-syntax-table c++-template-syntax-table
8910 (c-backward-token-2 0 t lim)
8911 (while (and (or (looking-at c-symbol-start)
8912 (looking-at "[<,]\\|::"))
8913 (zerop (c-backward-token-2 1 t lim))))))
8914
8915 (defun c-in-method-def-p ()
8916 ;; Return nil if we aren't in a method definition, otherwise the
8917 ;; position of the initial [+-].
8918 ;;
8919 ;; This function might do hidden buffer changes.
8920 (save-excursion
8921 (beginning-of-line)
8922 (and c-opt-method-key
8923 (looking-at c-opt-method-key)
8924 (point))
8925 ))
8926
8927 ;; Contributed by Kevin Ryde <user42@zip.com.au>.
8928 (defun c-in-gcc-asm-p ()
8929 ;; Return non-nil if point is within a gcc \"asm\" block.
8930 ;;
8931 ;; This should be called with point inside an argument list.
8932 ;;
8933 ;; Only one level of enclosing parentheses is considered, so for
8934 ;; instance nil is returned when in a function call within an asm
8935 ;; operand.
8936 ;;
8937 ;; This function might do hidden buffer changes.
8938
8939 (and c-opt-asm-stmt-key
8940 (save-excursion
8941 (beginning-of-line)
8942 (backward-up-list 1)
8943 (c-beginning-of-statement-1 (point-min) nil t)
8944 (looking-at c-opt-asm-stmt-key))))
8945
8946 (defun c-at-toplevel-p ()
8947 "Return a determination as to whether point is \"at the top level\".
8948 Informally, \"at the top level\" is anywhere where you can write
8949 a function.
8950
8951 More precisely, being at the top-level means that point is either
8952 outside any enclosing block (such as a function definition), or
8953 directly inside a class, namespace or other block that contains
8954 another declaration level.
8955
8956 If point is not at the top-level (e.g. it is inside a method
8957 definition), then nil is returned. Otherwise, if point is at a
8958 top-level not enclosed within a class definition, t is returned.
8959 Otherwise, a 2-vector is returned where the zeroth element is the
8960 buffer position of the start of the class declaration, and the first
8961 element is the buffer position of the enclosing class's opening
8962 brace.
8963
8964 Note that this function might do hidden buffer changes. See the
8965 comment at the start of cc-engine.el for more info."
8966 ;; Note to maintainers: this function consumes a great mass of CPU cycles.
8967 ;; Its use should thus be minimized as far as possible.
8968 (let ((paren-state (c-parse-state)))
8969 (or (not (c-most-enclosing-brace paren-state))
8970 (c-search-uplist-for-classkey paren-state))))
8971
8972 (defun c-just-after-func-arglist-p (&optional lim)
8973 ;; Return non-nil if the point is in the region after the argument
8974 ;; list of a function and its opening brace (or semicolon in case it
8975 ;; got no body). If there are K&R style argument declarations in
8976 ;; that region, the point has to be inside the first one for this
8977 ;; function to recognize it.
8978 ;;
8979 ;; If successful, the point is moved to the first token after the
8980 ;; function header (see `c-forward-decl-or-cast-1' for details) and
8981 ;; the position of the opening paren of the function arglist is
8982 ;; returned.
8983 ;;
8984 ;; The point is clobbered if not successful.
8985 ;;
8986 ;; LIM is used as bound for backward buffer searches.
8987 ;;
8988 ;; This function might do hidden buffer changes.
8989
8990 (let ((beg (point)) id-start)
8991 (and
8992 (eq (c-beginning-of-statement-1 lim) 'same)
8993
8994 (not (and (c-major-mode-is 'objc-mode)
8995 (c-forward-objc-directive)))
8996
8997 (setq id-start
8998 (car-safe (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil)))
8999 (< id-start beg)
9000
9001 ;; There should not be a '=' or ',' between beg and the
9002 ;; start of the declaration since that means we were in the
9003 ;; "expression part" of the declaration.
9004 (or (> (point) beg)
9005 (not (looking-at "[=,]")))
9006
9007 (save-excursion
9008 ;; Check that there's an arglist paren in the
9009 ;; declaration.
9010 (goto-char id-start)
9011 (cond ((eq (char-after) ?\()
9012 ;; The declarator is a paren expression, so skip past it
9013 ;; so that we don't get stuck on that instead of the
9014 ;; function arglist.
9015 (c-forward-sexp))
9016 ((and c-opt-op-identifier-prefix
9017 (looking-at c-opt-op-identifier-prefix))
9018 ;; Don't trip up on "operator ()".
9019 (c-forward-token-2 2 t)))
9020 (and (< (point) beg)
9021 (c-syntactic-re-search-forward "(" beg t t)
9022 (1- (point)))))))
9023
9024 (defun c-in-knr-argdecl (&optional lim)
9025 ;; Return the position of the first argument declaration if point is
9026 ;; inside a K&R style argument declaration list, nil otherwise.
9027 ;; `c-recognize-knr-p' is not checked. If LIM is non-nil, it's a
9028 ;; position that bounds the backward search for the argument list. This
9029 ;; function doesn't move point.
9030 ;;
9031 ;; Point must be within a possible K&R region, e.g. just before a top-level
9032 ;; "{". It must be outside of parens and brackets. The test can return
9033 ;; false positives otherwise.
9034 ;;
9035 ;; This function might do hidden buffer changes.
9036 (save-excursion
9037 (save-restriction
9038 ;; If we're in a macro, our search range is restricted to it. Narrow to
9039 ;; the searchable range.
9040 (let* ((macro-start (save-excursion (and (c-beginning-of-macro) (point))))
9041 (macro-end (save-excursion (and macro-start (c-end-of-macro) (point))))
9042 (low-lim (max (or lim (point-min)) (or macro-start (point-min))))
9043 before-lparen after-rparen
9044 (here (point))
9045 (pp-count-out 20) ; Max number of paren/brace constructs before
9046 ; we give up.
9047 ids ; List of identifiers in the parenthesized list.
9048 id-start after-prec-token decl-or-cast decl-res
9049 c-last-identifier-range identifier-ok)
9050 (narrow-to-region low-lim (or macro-end (point-max)))
9051
9052 ;; Search backwards for the defun's argument list. We give up if we
9053 ;; encounter a "}" (end of a previous defun) an "=" (which can't be in
9054 ;; a knr region) or BOB.
9055 ;;
9056 ;; The criterion for a paren structure being the arg list is:
9057 ;; o - there is non-WS stuff after it but before any "{"; AND
9058 ;; o - the token after it isn't a ";" AND
9059 ;; o - it is preceded by either an identifier (the function name) or
9060 ;; a macro expansion like "DEFUN (...)"; AND
9061 ;; o - its content is a non-empty comma-separated list of identifiers
9062 ;; (an empty arg list won't have a knr region).
9063 ;;
9064 ;; The following snippet illustrates these rules:
9065 ;; int foo (bar, baz, yuk)
9066 ;; int bar [] ;
9067 ;; int (*baz) (my_type) ;
9068 ;; int (*(* yuk) (void)) (void) ;
9069 ;; {
9070 ;;
9071 ;; Additionally, for a knr list to be recognized:
9072 ;; o - The identifier of each declarator up to and including the
9073 ;; one "near" point must be contained in the arg list.
9074
9075 (catch 'knr
9076 (while (> pp-count-out 0) ; go back one paren/bracket pair each time.
9077 (setq pp-count-out (1- pp-count-out))
9078 (c-syntactic-skip-backward "^)]}=")
9079 (cond ((eq (char-before) ?\))
9080 (setq after-rparen (point)))
9081 ((eq (char-before) ?\])
9082 (setq after-rparen nil))
9083 (t ; either } (hit previous defun) or = or no more
9084 ; parens/brackets.
9085 (throw 'knr nil)))
9086
9087 (if after-rparen
9088 ;; We're inside a paren. Could it be our argument list....?
9089 (if
9090 (and
9091 (progn
9092 (goto-char after-rparen)
9093 (unless (c-go-list-backward) (throw 'knr nil)) ;
9094 ;; FIXME!!! What about macros between the parens? 2007/01/20
9095 (setq before-lparen (point)))
9096
9097 ;; It can't be the arg list if next token is ; or {
9098 (progn (goto-char after-rparen)
9099 (c-forward-syntactic-ws)
9100 (not (memq (char-after) '(?\; ?\{ ?\=))))
9101
9102 ;; Is the thing preceding the list an identifier (the
9103 ;; function name), or a macro expansion?
9104 (progn
9105 (goto-char before-lparen)
9106 (eq (c-backward-token-2) 0)
9107 (or (eq (c-on-identifier) (point))
9108 (and (eq (char-after) ?\))
9109 (c-go-up-list-backward)
9110 (eq (c-backward-token-2) 0)
9111 (eq (c-on-identifier) (point)))))
9112
9113 ;; Have we got a non-empty list of comma-separated
9114 ;; identifiers?
9115 (progn
9116 (goto-char before-lparen)
9117 (c-forward-token-2) ; to first token inside parens
9118 (and
9119 (setq id-start (c-on-identifier)) ; Must be at least one.
9120 (catch 'id-list
9121 (while
9122 (progn
9123 (forward-char)
9124 (c-end-of-current-token)
9125 (push (buffer-substring-no-properties id-start
9126 (point))
9127 ids)
9128 (c-forward-syntactic-ws)
9129 (eq (char-after) ?\,))
9130 (c-forward-token-2)
9131 (unless (setq id-start (c-on-identifier))
9132 (throw 'id-list nil)))
9133 (eq (char-after) ?\)))))
9134
9135 ;; Are all the identifiers in the k&r list up to the
9136 ;; current one also in the argument list?
9137 (progn
9138 (forward-char) ; over the )
9139 (setq after-prec-token after-rparen)
9140 (c-forward-syntactic-ws)
9141 (while (and
9142 (or (consp (setq decl-or-cast
9143 (c-forward-decl-or-cast-1
9144 after-prec-token
9145 nil ; Or 'arglist ???
9146 nil)))
9147 (progn
9148 (goto-char after-prec-token)
9149 (c-forward-syntactic-ws)
9150 (setq identifier-ok (eq (char-after) ?{))
9151 nil))
9152 (eq (char-after) ?\;)
9153 (setq after-prec-token (1+ (point)))
9154 (goto-char (car decl-or-cast))
9155 (setq decl-res (c-forward-declarator))
9156 (setq identifier-ok
9157 (member (buffer-substring-no-properties
9158 (car decl-res) (cadr decl-res))
9159 ids))
9160 (progn
9161 (goto-char after-prec-token)
9162 (prog1 (< (point) here)
9163 (c-forward-syntactic-ws))))
9164 (setq identifier-ok nil))
9165 identifier-ok))
9166 ;; ...Yes. We've identified the function's argument list.
9167 (throw 'knr
9168 (progn (goto-char after-rparen)
9169 (c-forward-syntactic-ws)
9170 (point)))
9171 ;; ...No. The current parens aren't the function's arg list.
9172 (goto-char before-lparen))
9173
9174 (or (c-go-list-backward) ; backwards over [ .... ]
9175 (throw 'knr nil)))))))))
9176
9177 (defun c-skip-conditional ()
9178 ;; skip forward over conditional at point, including any predicate
9179 ;; statements in parentheses. No error checking is performed.
9180 ;;
9181 ;; This function might do hidden buffer changes.
9182 (c-forward-sexp (cond
9183 ;; else if()
9184 ((looking-at (concat "\\<else"
9185 "\\([ \t\n]\\|\\\\\n\\)+"
9186 "if\\>\\([^_]\\|$\\)"))
9187 3)
9188 ;; do, else, try, finally
9189 ((looking-at (concat "\\<\\("
9190 "do\\|else\\|try\\|finally"
9191 "\\)\\>\\([^_]\\|$\\)"))
9192 1)
9193 ;; for, if, while, switch, catch, synchronized, foreach
9194 (t 2))))
9195
9196 (defun c-after-conditional (&optional lim)
9197 ;; If looking at the token after a conditional then return the
9198 ;; position of its start, otherwise return nil.
9199 ;;
9200 ;; This function might do hidden buffer changes.
9201 (save-excursion
9202 (and (zerop (c-backward-token-2 1 t lim))
9203 (or (looking-at c-block-stmt-1-key)
9204 (and (eq (char-after) ?\()
9205 (zerop (c-backward-token-2 1 t lim))
9206 (or (looking-at c-block-stmt-2-key)
9207 (looking-at c-block-stmt-1-2-key))))
9208 (point))))
9209
9210 (defun c-after-special-operator-id (&optional lim)
9211 ;; If the point is after an operator identifier that isn't handled
9212 ;; like an ordinary symbol (i.e. like "operator =" in C++) then the
9213 ;; position of the start of that identifier is returned. nil is
9214 ;; returned otherwise. The point may be anywhere in the syntactic
9215 ;; whitespace after the last token of the operator identifier.
9216 ;;
9217 ;; This function might do hidden buffer changes.
9218 (save-excursion
9219 (and c-overloadable-operators-regexp
9220 (zerop (c-backward-token-2 1 nil lim))
9221 (looking-at c-overloadable-operators-regexp)
9222 (or (not c-opt-op-identifier-prefix)
9223 (and
9224 (zerop (c-backward-token-2 1 nil lim))
9225 (looking-at c-opt-op-identifier-prefix)))
9226 (point))))
9227
9228 (defsubst c-backward-to-block-anchor (&optional lim)
9229 ;; Assuming point is at a brace that opens a statement block of some
9230 ;; kind, move to the proper anchor point for that block. It might
9231 ;; need to be adjusted further by c-add-stmt-syntax, but the
9232 ;; position at return is suitable as start position for that
9233 ;; function.
9234 ;;
9235 ;; This function might do hidden buffer changes.
9236 (unless (= (point) (c-point 'boi))
9237 (let ((start (c-after-conditional lim)))
9238 (if start
9239 (goto-char start)))))
9240
9241 (defsubst c-backward-to-decl-anchor (&optional lim)
9242 ;; Assuming point is at a brace that opens the block of a top level
9243 ;; declaration of some kind, move to the proper anchor point for
9244 ;; that block.
9245 ;;
9246 ;; This function might do hidden buffer changes.
9247 (unless (= (point) (c-point 'boi))
9248 (c-beginning-of-statement-1 lim)))
9249
9250 (defun c-search-decl-header-end ()
9251 ;; Search forward for the end of the "header" of the current
9252 ;; declaration. That's the position where the definition body
9253 ;; starts, or the first variable initializer, or the ending
9254 ;; semicolon. I.e. search forward for the closest following
9255 ;; (syntactically relevant) '{', '=' or ';' token. Point is left
9256 ;; _after_ the first found token, or at point-max if none is found.
9257 ;;
9258 ;; This function might do hidden buffer changes.
9259
9260 (let ((base (point)))
9261 (if (c-major-mode-is 'c++-mode)
9262
9263 ;; In C++ we need to take special care to handle operator
9264 ;; tokens and those pesky template brackets.
9265 (while (and
9266 (c-syntactic-re-search-forward "[;{<=]" nil 'move t t)
9267 (or
9268 (c-end-of-current-token base)
9269 ;; Handle operator identifiers, i.e. ignore any
9270 ;; operator token preceded by "operator".
9271 (save-excursion
9272 (and (c-safe (c-backward-sexp) t)
9273 (looking-at c-opt-op-identifier-prefix)))
9274 (and (eq (char-before) ?<)
9275 (c-with-syntax-table c++-template-syntax-table
9276 (if (c-safe (goto-char (c-up-list-forward (point))))
9277 t
9278 (goto-char (point-max))
9279 nil)))))
9280 (setq base (point)))
9281
9282 (while (and
9283 (c-syntactic-re-search-forward "[;{=]" nil 'move t t)
9284 (c-end-of-current-token base))
9285 (setq base (point))))))
9286
9287 (defun c-beginning-of-decl-1 (&optional lim)
9288 ;; Go to the beginning of the current declaration, or the beginning
9289 ;; of the previous one if already at the start of it. Point won't
9290 ;; be moved out of any surrounding paren. Return a cons cell of the
9291 ;; form (MOVE . KNR-POS). MOVE is like the return value from
9292 ;; `c-beginning-of-statement-1'. If point skipped over some K&R
9293 ;; style argument declarations (and they are to be recognized) then
9294 ;; KNR-POS is set to the start of the first such argument
9295 ;; declaration, otherwise KNR-POS is nil. If LIM is non-nil, it's a
9296 ;; position that bounds the backward search.
9297 ;;
9298 ;; NB: Cases where the declaration continues after the block, as in
9299 ;; "struct foo { ... } bar;", are currently recognized as two
9300 ;; declarations, e.g. "struct foo { ... }" and "bar;" in this case.
9301 ;;
9302 ;; This function might do hidden buffer changes.
9303 (catch 'return
9304 (let* ((start (point))
9305 (last-stmt-start (point))
9306 (move (c-beginning-of-statement-1 lim nil t)))
9307
9308 ;; `c-beginning-of-statement-1' stops at a block start, but we
9309 ;; want to continue if the block doesn't begin a top level
9310 ;; construct, i.e. if it isn't preceded by ';', '}', ':', bob,
9311 ;; or an open paren.
9312 (let ((beg (point)) tentative-move)
9313 ;; Go back one "statement" each time round the loop until we're just
9314 ;; after a ;, }, or :, or at BOB or the start of a macro or start of
9315 ;; an ObjC method. This will move over a multiple declaration whose
9316 ;; components are comma separated.
9317 (while (and
9318 ;; Must check with c-opt-method-key in ObjC mode.
9319 (not (and c-opt-method-key
9320 (looking-at c-opt-method-key)))
9321 (/= last-stmt-start (point))
9322 (progn
9323 (c-backward-syntactic-ws lim)
9324 (not (or (memq (char-before) '(?\; ?} ?: nil))
9325 (c-at-vsemi-p))))
9326 (save-excursion
9327 (backward-char)
9328 (not (looking-at "\\s(")))
9329 ;; Check that we don't move from the first thing in a
9330 ;; macro to its header.
9331 (not (eq (setq tentative-move
9332 (c-beginning-of-statement-1 lim nil t))
9333 'macro)))
9334 (setq last-stmt-start beg
9335 beg (point)
9336 move tentative-move))
9337 (goto-char beg))
9338
9339 (when c-recognize-knr-p
9340 (let ((fallback-pos (point)) knr-argdecl-start)
9341 ;; Handle K&R argdecls. Back up after the "statement" jumped
9342 ;; over by `c-beginning-of-statement-1', unless it was the
9343 ;; function body, in which case we're sitting on the opening
9344 ;; brace now. Then test if we're in a K&R argdecl region and
9345 ;; that we started at the other side of the first argdecl in
9346 ;; it.
9347 (unless (eq (char-after) ?{)
9348 (goto-char last-stmt-start))
9349 (if (and (setq knr-argdecl-start (c-in-knr-argdecl lim))
9350 (< knr-argdecl-start start)
9351 (progn
9352 (goto-char knr-argdecl-start)
9353 (not (eq (c-beginning-of-statement-1 lim nil t) 'macro))))
9354 (throw 'return
9355 (cons (if (eq (char-after fallback-pos) ?{)
9356 'previous
9357 'same)
9358 knr-argdecl-start))
9359 (goto-char fallback-pos))))
9360
9361 ;; `c-beginning-of-statement-1' counts each brace block as a separate
9362 ;; statement, so the result will be 'previous if we've moved over any.
9363 ;; So change our result back to 'same if necessary.
9364 ;;
9365 ;; If they were brace list initializers we might not have moved over a
9366 ;; declaration boundary though, so change it to 'same if we've moved
9367 ;; past a '=' before '{', but not ';'. (This ought to be integrated
9368 ;; into `c-beginning-of-statement-1', so we avoid this extra pass which
9369 ;; potentially can search over a large amount of text.). Take special
9370 ;; pains not to get mislead by C++'s "operator=", and the like.
9371 (if (and (eq move 'previous)
9372 (c-with-syntax-table (if (c-major-mode-is 'c++-mode)
9373 c++-template-syntax-table
9374 (syntax-table))
9375 (save-excursion
9376 (and
9377 (progn
9378 (while ; keep going back to "[;={"s until we either find
9379 ; no more, or get to one which isn't an "operator ="
9380 (and (c-syntactic-re-search-forward "[;={]" start t t t)
9381 (eq (char-before) ?=)
9382 c-overloadable-operators-regexp
9383 c-opt-op-identifier-prefix
9384 (save-excursion
9385 (eq (c-backward-token-2) 0)
9386 (looking-at c-overloadable-operators-regexp)
9387 (eq (c-backward-token-2) 0)
9388 (looking-at c-opt-op-identifier-prefix))))
9389 (eq (char-before) ?=))
9390 (c-syntactic-re-search-forward "[;{]" start t t)
9391 (eq (char-before) ?{)
9392 (c-safe (goto-char (c-up-list-forward (point))) t)
9393 (not (c-syntactic-re-search-forward ";" start t t))))))
9394 (cons 'same nil)
9395 (cons move nil)))))
9396
9397 (defun c-end-of-decl-1 ()
9398 ;; Assuming point is at the start of a declaration (as detected by
9399 ;; e.g. `c-beginning-of-decl-1'), go to the end of it. Unlike
9400 ;; `c-beginning-of-decl-1', this function handles the case when a
9401 ;; block is followed by identifiers in e.g. struct declarations in C
9402 ;; or C++. If a proper end was found then t is returned, otherwise
9403 ;; point is moved as far as possible within the current sexp and nil
9404 ;; is returned. This function doesn't handle macros; use
9405 ;; `c-end-of-macro' instead in those cases.
9406 ;;
9407 ;; This function might do hidden buffer changes.
9408 (let ((start (point))
9409 (decl-syntax-table (if (c-major-mode-is 'c++-mode)
9410 c++-template-syntax-table
9411 (syntax-table))))
9412 (catch 'return
9413 (c-search-decl-header-end)
9414
9415 (when (and c-recognize-knr-p
9416 (eq (char-before) ?\;)
9417 (c-in-knr-argdecl start))
9418 ;; Stopped at the ';' in a K&R argdecl section which is
9419 ;; detected using the same criteria as in
9420 ;; `c-beginning-of-decl-1'. Move to the following block
9421 ;; start.
9422 (c-syntactic-re-search-forward "{" nil 'move t))
9423
9424 (when (eq (char-before) ?{)
9425 ;; Encountered a block in the declaration. Jump over it.
9426 (condition-case nil
9427 (goto-char (c-up-list-forward (point)))
9428 (error (goto-char (point-max))
9429 (throw 'return nil)))
9430 (if (or (not c-opt-block-decls-with-vars-key)
9431 (save-excursion
9432 (c-with-syntax-table decl-syntax-table
9433 (let ((lim (point)))
9434 (goto-char start)
9435 (not (and
9436 ;; Check for `c-opt-block-decls-with-vars-key'
9437 ;; before the first paren.
9438 (c-syntactic-re-search-forward
9439 (concat "[;=([{]\\|\\("
9440 c-opt-block-decls-with-vars-key
9441 "\\)")
9442 lim t t t)
9443 (match-beginning 1)
9444 (not (eq (char-before) ?_))
9445 ;; Check that the first following paren is
9446 ;; the block.
9447 (c-syntactic-re-search-forward "[;=([{]"
9448 lim t t t)
9449 (eq (char-before) ?{)))))))
9450 ;; The declaration doesn't have any of the
9451 ;; `c-opt-block-decls-with-vars' keywords in the
9452 ;; beginning, so it ends here at the end of the block.
9453 (throw 'return t)))
9454
9455 (c-with-syntax-table decl-syntax-table
9456 (while (progn
9457 (if (eq (char-before) ?\;)
9458 (throw 'return t))
9459 (c-syntactic-re-search-forward ";" nil 'move t))))
9460 nil)))
9461
9462 (defun c-looking-at-decl-block (containing-sexp goto-start &optional limit)
9463 ;; Assuming the point is at an open brace, check if it starts a
9464 ;; block that contains another declaration level, i.e. that isn't a
9465 ;; statement block or a brace list, and if so return non-nil.
9466 ;;
9467 ;; If the check is successful, the return value is the start of the
9468 ;; keyword that tells what kind of construct it is, i.e. typically
9469 ;; what `c-decl-block-key' matched. Also, if GOTO-START is set then
9470 ;; the point will be at the start of the construct, before any
9471 ;; leading specifiers, otherwise it's at the returned position.
9472 ;;
9473 ;; The point is clobbered if the check is unsuccessful.
9474 ;;
9475 ;; CONTAINING-SEXP is the position of the open of the surrounding
9476 ;; paren, or nil if none.
9477 ;;
9478 ;; The optional LIMIT limits the backward search for the start of
9479 ;; the construct. It's assumed to be at a syntactically relevant
9480 ;; position.
9481 ;;
9482 ;; If any template arglists are found in the searched region before
9483 ;; the open brace, they get marked with paren syntax.
9484 ;;
9485 ;; This function might do hidden buffer changes.
9486
9487 (let ((open-brace (point)) kwd-start first-specifier-pos)
9488 (c-syntactic-skip-backward c-block-prefix-charset limit t)
9489
9490 (when (and c-recognize-<>-arglists
9491 (eq (char-before) ?>))
9492 ;; Could be at the end of a template arglist.
9493 (let ((c-parse-and-markup-<>-arglists t))
9494 (while (and
9495 (c-backward-<>-arglist nil limit)
9496 (progn
9497 (c-syntactic-skip-backward c-block-prefix-charset limit t)
9498 (eq (char-before) ?>))))))
9499
9500 ;; Skip back over noise clauses.
9501 (while (and
9502 c-opt-cpp-prefix
9503 (eq (char-before) ?\))
9504 (let ((after-paren (point)))
9505 (if (and (c-go-list-backward)
9506 (progn (c-backward-syntactic-ws)
9507 (c-simple-skip-symbol-backward))
9508 (or (looking-at c-paren-nontype-key)
9509 (looking-at c-noise-macro-with-parens-name-re)))
9510 (progn
9511 (c-syntactic-skip-backward c-block-prefix-charset limit t)
9512 t)
9513 (goto-char after-paren)
9514 nil))))
9515
9516 ;; Note: Can't get bogus hits inside template arglists below since they
9517 ;; have gotten paren syntax above.
9518 (when (and
9519 ;; If `goto-start' is set we begin by searching for the
9520 ;; first possible position of a leading specifier list.
9521 ;; The `c-decl-block-key' search continues from there since
9522 ;; we know it can't match earlier.
9523 (if goto-start
9524 (when (c-syntactic-re-search-forward c-symbol-start
9525 open-brace t t)
9526 (goto-char (setq first-specifier-pos (match-beginning 0)))
9527 t)
9528 t)
9529
9530 (cond
9531 ((c-syntactic-re-search-forward c-decl-block-key open-brace t t t)
9532 (goto-char (setq kwd-start (match-beginning 0)))
9533 (and
9534 ;; Exclude cases where we matched what would ordinarily
9535 ;; be a block declaration keyword, except where it's not
9536 ;; legal because it's part of a "compound keyword" like
9537 ;; "enum class". Of course, if c-after-brace-list-key
9538 ;; is nil, we can skip the test.
9539 (or (equal c-after-brace-list-key "\\<\\>")
9540 (save-match-data
9541 (save-excursion
9542 (not
9543 (and
9544 (looking-at c-after-brace-list-key)
9545 (= (c-backward-token-2 1 t) 0)
9546 (looking-at c-brace-list-key))))))
9547 (or
9548 ;; Found a keyword that can't be a type?
9549 (match-beginning 1)
9550
9551 ;; Can be a type too, in which case it's the return type of a
9552 ;; function (under the assumption that no declaration level
9553 ;; block construct starts with a type).
9554 (not (c-forward-type))
9555
9556 ;; Jumped over a type, but it could be a declaration keyword
9557 ;; followed by the declared identifier that we've jumped over
9558 ;; instead (e.g. in "class Foo {"). If it indeed is a type
9559 ;; then we should be at the declarator now, so check for a
9560 ;; valid declarator start.
9561 ;;
9562 ;; Note: This doesn't cope with the case when a declared
9563 ;; identifier is followed by e.g. '(' in a language where '('
9564 ;; also might be part of a declarator expression. Currently
9565 ;; there's no such language.
9566 (not (or (looking-at c-symbol-start)
9567 (looking-at c-type-decl-prefix-key))))))
9568
9569 ;; In Pike a list of modifiers may be followed by a brace
9570 ;; to make them apply to many identifiers. Note that the
9571 ;; match data will be empty on return in this case.
9572 ((and (c-major-mode-is 'pike-mode)
9573 (progn
9574 (goto-char open-brace)
9575 (= (c-backward-token-2) 0))
9576 (looking-at c-specifier-key)
9577 ;; Use this variant to avoid yet another special regexp.
9578 (c-keyword-member (c-keyword-sym (match-string 1))
9579 'c-modifier-kwds))
9580 (setq kwd-start (point))
9581 t)))
9582
9583 ;; Got a match.
9584
9585 (if goto-start
9586 ;; Back up over any preceding specifiers and their clauses
9587 ;; by going forward from `first-specifier-pos', which is the
9588 ;; earliest possible position where the specifier list can
9589 ;; start.
9590 (progn
9591 (goto-char first-specifier-pos)
9592
9593 (while (< (point) kwd-start)
9594 (if (looking-at c-symbol-key)
9595 ;; Accept any plain symbol token on the ground that
9596 ;; it's a specifier masked through a macro (just
9597 ;; like `c-forward-decl-or-cast-1' skip forward over
9598 ;; such tokens).
9599 ;;
9600 ;; Could be more restrictive wrt invalid keywords,
9601 ;; but that'd only occur in invalid code so there's
9602 ;; no use spending effort on it.
9603 (let ((end (match-end 0)))
9604 (unless (c-forward-keyword-clause 0)
9605 (goto-char end)
9606 (c-forward-syntactic-ws)))
9607
9608 ;; Can't parse a declaration preamble and is still
9609 ;; before `kwd-start'. That means `first-specifier-pos'
9610 ;; was in some earlier construct. Search again.
9611 (if (c-syntactic-re-search-forward c-symbol-start
9612 kwd-start 'move t)
9613 (goto-char (setq first-specifier-pos (match-beginning 0)))
9614 ;; Got no preamble before the block declaration keyword.
9615 (setq first-specifier-pos kwd-start))))
9616
9617 (goto-char first-specifier-pos))
9618 (goto-char kwd-start))
9619
9620 kwd-start)))
9621
9622 (defun c-directly-in-class-called-p (name)
9623 ;; Check whether point is directly inside a brace block which is the brace
9624 ;; block of a class, struct, or union which is called NAME, a string.
9625 (let* ((paren-state (c-parse-state))
9626 (brace-pos (c-pull-open-brace paren-state))
9627 )
9628 (when (eq (char-after brace-pos) ?{)
9629 (goto-char brace-pos)
9630 (save-excursion
9631 ; *c-looking-at-decl-block
9632 ; containing-sexp goto-start &optional
9633 ; limit)
9634 (when (and (c-looking-at-decl-block
9635 (c-pull-open-brace paren-state)
9636 nil)
9637 (looking-at c-class-key))
9638 (goto-char (match-end 1))
9639 (c-forward-syntactic-ws)
9640 (looking-at name))))))
9641
9642 (defun c-search-uplist-for-classkey (paren-state)
9643 ;; Check if the closest containing paren sexp is a declaration
9644 ;; block, returning a 2 element vector in that case. Aref 0
9645 ;; contains the bufpos at boi of the class key line, and aref 1
9646 ;; contains the bufpos of the open brace. This function is an
9647 ;; obsolete wrapper for `c-looking-at-decl-block'.
9648 ;;
9649 ;; This function might do hidden buffer changes.
9650 (let ((open-paren-pos (c-most-enclosing-brace paren-state)))
9651 (when open-paren-pos
9652 (save-excursion
9653 (goto-char open-paren-pos)
9654 (when (and (eq (char-after) ?{)
9655 (c-looking-at-decl-block
9656 (c-safe-position open-paren-pos paren-state)
9657 nil))
9658 (back-to-indentation)
9659 (vector (point) open-paren-pos))))))
9660
9661 (defun c-most-enclosing-decl-block (paren-state)
9662 ;; Return the buffer position of the most enclosing decl-block brace (in the
9663 ;; sense of c-looking-at-decl-block) in the PAREN-STATE structure, or nil if
9664 ;; none was found.
9665 (let* ((open-brace (c-pull-open-brace paren-state))
9666 (next-open-brace (c-pull-open-brace paren-state)))
9667 (while (and open-brace
9668 (save-excursion
9669 (goto-char open-brace)
9670 (not (c-looking-at-decl-block next-open-brace nil))))
9671 (setq open-brace next-open-brace
9672 next-open-brace (c-pull-open-brace paren-state)))
9673 open-brace))
9674
9675 (defun c-cheap-inside-bracelist-p (paren-state)
9676 ;; Return the position of the L-brace if point is inside a brace list
9677 ;; initialization of an array, etc. This is an approximate function,
9678 ;; designed for speed over accuracy. It will not find every bracelist, but
9679 ;; a non-nil result is reliable. We simply search for "= {" (naturally with
9680 ;; syntactic whitespace allowed). PAREN-STATE is the normal thing that it
9681 ;; is everywhere else.
9682 (let (b-pos)
9683 (save-excursion
9684 (while
9685 (and (setq b-pos (c-pull-open-brace paren-state))
9686 (progn (goto-char b-pos)
9687 (c-backward-sws)
9688 (c-backward-token-2)
9689 (not (looking-at "=")))))
9690 b-pos)))
9691
9692 (defun c-backward-typed-enum-colon ()
9693 ;; We're at a "{" which might be the opening brace of a enum which is
9694 ;; strongly typed (by a ":" followed by a type). If this is the case, leave
9695 ;; point before the colon and return t. Otherwise leave point unchanged and return nil.
9696 ;; Match data will be clobbered.
9697 (let ((here (point))
9698 (colon-pos nil))
9699 (save-excursion
9700 (while
9701 (and (eql (c-backward-token-2) 0)
9702 (or (not (looking-at "\\s)"))
9703 (c-go-up-list-backward))
9704 (cond
9705 ((and (eql (char-after) ?:)
9706 (save-excursion
9707 (c-backward-syntactic-ws)
9708 (c-on-identifier)))
9709 (setq colon-pos (point))
9710 (forward-char)
9711 (c-forward-syntactic-ws)
9712 (or (and (c-forward-type)
9713 (progn (c-forward-syntactic-ws)
9714 (eq (point) here)))
9715 (setq colon-pos nil))
9716 nil)
9717 ((eql (char-after) ?\()
9718 t)
9719 ((looking-at c-symbol-key)
9720 t)
9721 (t nil)))))
9722 (when colon-pos
9723 (goto-char colon-pos)
9724 t)))
9725
9726 (defun c-backward-over-enum-header ()
9727 ;; We're at a "{". Move back to the enum-like keyword that starts this
9728 ;; declaration and return t, otherwise don't move and return nil.
9729 (let ((here (point))
9730 up-sexp-pos before-identifier)
9731 (when c-recognize-post-brace-list-type-p
9732 (c-backward-typed-enum-colon))
9733 (while
9734 (and
9735 (eq (c-backward-token-2) 0)
9736 (or (not (looking-at "\\s)"))
9737 (c-go-up-list-backward))
9738 (cond
9739 ((and (looking-at c-symbol-key) (c-on-identifier)
9740 (not before-identifier))
9741 (setq before-identifier t))
9742 ((and before-identifier
9743 (or (eql (char-after) ?,)
9744 (looking-at c-postfix-decl-spec-key)))
9745 (setq before-identifier nil)
9746 t)
9747 ((looking-at c-after-brace-list-key) t)
9748 ((looking-at c-brace-list-key) nil)
9749 ((eq (char-after) ?\()
9750 (and (eq (c-backward-token-2) 0)
9751 (or (looking-at c-decl-hangon-key)
9752 (and c-opt-cpp-prefix
9753 (looking-at c-noise-macro-with-parens-name-re)))))
9754
9755 ((and c-recognize-<>-arglists
9756 (eq (char-after) ?<)
9757 (looking-at "\\s("))
9758 t)
9759 (t nil))))
9760 (or (looking-at c-brace-list-key)
9761 (progn (goto-char here) nil))))
9762
9763 (defun c-inside-bracelist-p (containing-sexp paren-state)
9764 ;; return the buffer position of the beginning of the brace list
9765 ;; statement if we're inside a brace list, otherwise return nil.
9766 ;; CONTAINING-SEXP is the buffer pos of the innermost containing
9767 ;; paren. PAREN-STATE is the remainder of the state of enclosing
9768 ;; braces
9769 ;;
9770 ;; N.B.: This algorithm can potentially get confused by cpp macros
9771 ;; placed in inconvenient locations. It's a trade-off we make for
9772 ;; speed.
9773 ;;
9774 ;; This function might do hidden buffer changes.
9775 (or
9776 ;; This will pick up brace list declarations.
9777 (save-excursion
9778 (goto-char containing-sexp)
9779 (c-backward-over-enum-header))
9780 ;; this will pick up array/aggregate init lists, even if they are nested.
9781 (save-excursion
9782 (let ((class-key
9783 ;; Pike can have class definitions anywhere, so we must
9784 ;; check for the class key here.
9785 (and (c-major-mode-is 'pike-mode)
9786 c-decl-block-key))
9787 bufpos braceassignp lim next-containing macro-start)
9788 (while (and (not bufpos)
9789 containing-sexp)
9790 (when paren-state
9791 (if (consp (car paren-state))
9792 (setq lim (cdr (car paren-state))
9793 paren-state (cdr paren-state))
9794 (setq lim (car paren-state)))
9795 (when paren-state
9796 (setq next-containing (car paren-state)
9797 paren-state (cdr paren-state))))
9798 (goto-char containing-sexp)
9799 (if (c-looking-at-inexpr-block next-containing next-containing)
9800 ;; We're in an in-expression block of some kind. Do not
9801 ;; check nesting. We deliberately set the limit to the
9802 ;; containing sexp, so that c-looking-at-inexpr-block
9803 ;; doesn't check for an identifier before it.
9804 (setq containing-sexp nil)
9805 ;; see if the open brace is preceded by = or [...] in
9806 ;; this statement, but watch out for operator=
9807 (setq braceassignp 'dontknow)
9808 (c-backward-token-2 1 t lim)
9809 ;; Checks to do only on the first sexp before the brace.
9810 (when (and c-opt-inexpr-brace-list-key
9811 (eq (char-after) ?\[))
9812 ;; In Java, an initialization brace list may follow
9813 ;; directly after "new Foo[]", so check for a "new"
9814 ;; earlier.
9815 (while (eq braceassignp 'dontknow)
9816 (setq braceassignp
9817 (cond ((/= (c-backward-token-2 1 t lim) 0) nil)
9818 ((looking-at c-opt-inexpr-brace-list-key) t)
9819 ((looking-at "\\sw\\|\\s_\\|[.[]")
9820 ;; Carry on looking if this is an
9821 ;; identifier (may contain "." in Java)
9822 ;; or another "[]" sexp.
9823 'dontknow)
9824 (t nil)))))
9825 ;; Checks to do on all sexps before the brace, up to the
9826 ;; beginning of the statement.
9827 (while (eq braceassignp 'dontknow)
9828 (cond ((eq (char-after) ?\;)
9829 (setq braceassignp nil))
9830 ((and class-key
9831 (looking-at class-key))
9832 (setq braceassignp nil))
9833 ((eq (char-after) ?=)
9834 ;; We've seen a =, but must check earlier tokens so
9835 ;; that it isn't something that should be ignored.
9836 (setq braceassignp 'maybe)
9837 (while (and (eq braceassignp 'maybe)
9838 (zerop (c-backward-token-2 1 t lim)))
9839 (setq braceassignp
9840 (cond
9841 ;; Check for operator =
9842 ((and c-opt-op-identifier-prefix
9843 (looking-at c-opt-op-identifier-prefix))
9844 nil)
9845 ;; Check for `<opchar>= in Pike.
9846 ((and (c-major-mode-is 'pike-mode)
9847 (or (eq (char-after) ?`)
9848 ;; Special case for Pikes
9849 ;; `[]=, since '[' is not in
9850 ;; the punctuation class.
9851 (and (eq (char-after) ?\[)
9852 (eq (char-before) ?`))))
9853 nil)
9854 ((looking-at "\\s.") 'maybe)
9855 ;; make sure we're not in a C++ template
9856 ;; argument assignment
9857 ((and
9858 (c-major-mode-is 'c++-mode)
9859 (save-excursion
9860 (let ((here (point))
9861 (pos< (progn
9862 (skip-chars-backward "^<>")
9863 (point))))
9864 (and (eq (char-before) ?<)
9865 (not (c-crosses-statement-barrier-p
9866 pos< here))
9867 (not (c-in-literal))
9868 ))))
9869 nil)
9870 (t t))))))
9871 (if (and (eq braceassignp 'dontknow)
9872 (/= (c-backward-token-2 1 t lim) 0))
9873 (setq braceassignp nil)))
9874 (cond
9875 (braceassignp
9876 ;; We've hit the beginning of the aggregate list.
9877 (c-beginning-of-statement-1
9878 (c-most-enclosing-brace paren-state))
9879 (setq bufpos (point)))
9880 ((eq (char-after) ?\;)
9881 ;; Brace lists can't contain a semicolon, so we're done.
9882 (setq containing-sexp nil))
9883 ((and (setq macro-start (point))
9884 (c-forward-to-cpp-define-body)
9885 (eq (point) containing-sexp))
9886 ;; We've a macro whose expansion starts with the '{'.
9887 ;; Heuristically, if we have a ';' in it we've not got a
9888 ;; brace list, otherwise we have.
9889 (let ((macro-end (progn (c-end-of-macro) (point))))
9890 (goto-char containing-sexp)
9891 (forward-char)
9892 (if (and (c-syntactic-re-search-forward "[;,]" macro-end t t)
9893 (eq (char-before) ?\;))
9894 (setq bufpos nil
9895 containing-sexp nil)
9896 (setq bufpos macro-start))))
9897 (t
9898 ;; Go up one level
9899 (setq containing-sexp next-containing
9900 lim nil
9901 next-containing nil)))))
9902
9903 bufpos))
9904 ))
9905
9906 (defun c-looking-at-special-brace-list (&optional lim)
9907 ;; If we're looking at the start of a pike-style list, i.e., `({ })',
9908 ;; `([ ])', `(< >)', etc., a cons of a cons of its starting and ending
9909 ;; positions and its entry in c-special-brace-lists is returned, nil
9910 ;; otherwise. The ending position is nil if the list is still open.
9911 ;; LIM is the limit for forward search. The point may either be at
9912 ;; the `(' or at the following paren character. Tries to check the
9913 ;; matching closer, but assumes it's correct if no balanced paren is
9914 ;; found (i.e. the case `({ ... } ... )' is detected as _not_ being
9915 ;; a special brace list).
9916 ;;
9917 ;; This function might do hidden buffer changes.
9918 (if c-special-brace-lists
9919 (condition-case ()
9920 (save-excursion
9921 (let ((beg (point))
9922 inner-beg end type)
9923 (c-forward-syntactic-ws)
9924 (if (eq (char-after) ?\()
9925 (progn
9926 (forward-char 1)
9927 (c-forward-syntactic-ws)
9928 (setq inner-beg (point))
9929 (setq type (assq (char-after) c-special-brace-lists)))
9930 (if (setq type (assq (char-after) c-special-brace-lists))
9931 (progn
9932 (setq inner-beg (point))
9933 (c-backward-syntactic-ws)
9934 (forward-char -1)
9935 (setq beg (if (eq (char-after) ?\()
9936 (point)
9937 nil)))))
9938 (if (and beg type)
9939 (if (and (c-safe
9940 (goto-char beg)
9941 (c-forward-sexp 1)
9942 (setq end (point))
9943 (= (char-before) ?\)))
9944 (c-safe
9945 (goto-char inner-beg)
9946 (if (looking-at "\\s(")
9947 ;; Check balancing of the inner paren
9948 ;; below.
9949 (progn
9950 (c-forward-sexp 1)
9951 t)
9952 ;; If the inner char isn't a paren then
9953 ;; we can't check balancing, so just
9954 ;; check the char before the outer
9955 ;; closing paren.
9956 (goto-char end)
9957 (backward-char)
9958 (c-backward-syntactic-ws)
9959 (= (char-before) (cdr type)))))
9960 (if (or (/= (char-syntax (char-before)) ?\))
9961 (= (progn
9962 (c-forward-syntactic-ws)
9963 (point))
9964 (1- end)))
9965 (cons (cons beg end) type))
9966 (cons (list beg) type)))))
9967 (error nil))))
9968
9969 (defun c-looking-at-bos (&optional lim)
9970 ;; Return non-nil if between two statements or declarations, assuming
9971 ;; point is not inside a literal or comment.
9972 ;;
9973 ;; Obsolete - `c-at-statement-start-p' or `c-at-expression-start-p'
9974 ;; are recommended instead.
9975 ;;
9976 ;; This function might do hidden buffer changes.
9977 (c-at-statement-start-p))
9978 (make-obsolete 'c-looking-at-bos 'c-at-statement-start-p "22.1")
9979
9980 (defun c-looking-at-inexpr-block (lim containing-sexp &optional check-at-end)
9981 ;; Return non-nil if we're looking at the beginning of a block
9982 ;; inside an expression. The value returned is actually a cons of
9983 ;; either 'inlambda, 'inexpr-statement or 'inexpr-class and the
9984 ;; position of the beginning of the construct.
9985 ;;
9986 ;; LIM limits the backward search. CONTAINING-SEXP is the start
9987 ;; position of the closest containing list. If it's nil, the
9988 ;; containing paren isn't used to decide whether we're inside an
9989 ;; expression or not. If both LIM and CONTAINING-SEXP are used, LIM
9990 ;; needs to be farther back.
9991 ;;
9992 ;; If CHECK-AT-END is non-nil then extra checks at the end of the
9993 ;; brace block might be done. It should only be used when the
9994 ;; construct can be assumed to be complete, i.e. when the original
9995 ;; starting position was further down than that.
9996 ;;
9997 ;; This function might do hidden buffer changes.
9998
9999 (save-excursion
10000 (let ((res 'maybe) passed-paren
10001 (closest-lim (or containing-sexp lim (point-min)))
10002 ;; Look at the character after point only as a last resort
10003 ;; when we can't disambiguate.
10004 (block-follows (and (eq (char-after) ?{) (point))))
10005
10006 (while (and (eq res 'maybe)
10007 (progn (c-backward-syntactic-ws)
10008 (> (point) closest-lim))
10009 (not (bobp))
10010 (progn (backward-char)
10011 (looking-at "[]).]\\|\\w\\|\\s_"))
10012 (c-safe (forward-char)
10013 (goto-char (scan-sexps (point) -1))))
10014
10015 (setq res
10016 (if (looking-at c-keywords-regexp)
10017 (let ((kw-sym (c-keyword-sym (match-string 1))))
10018 (cond
10019 ((and block-follows
10020 (c-keyword-member kw-sym 'c-inexpr-class-kwds))
10021 (and (not (eq passed-paren ?\[))
10022 (or (not (looking-at c-class-key))
10023 ;; If the class definition is at the start of
10024 ;; a statement, we don't consider it an
10025 ;; in-expression class.
10026 (let ((prev (point)))
10027 (while (and
10028 (= (c-backward-token-2 1 nil closest-lim) 0)
10029 (eq (char-syntax (char-after)) ?w))
10030 (setq prev (point)))
10031 (goto-char prev)
10032 (not (c-at-statement-start-p)))
10033 ;; Also, in Pike we treat it as an
10034 ;; in-expression class if it's used in an
10035 ;; object clone expression.
10036 (save-excursion
10037 (and check-at-end
10038 (c-major-mode-is 'pike-mode)
10039 (progn (goto-char block-follows)
10040 (zerop (c-forward-token-2 1 t)))
10041 (eq (char-after) ?\())))
10042 (cons 'inexpr-class (point))))
10043 ((c-keyword-member kw-sym 'c-inexpr-block-kwds)
10044 (when (not passed-paren)
10045 (cons 'inexpr-statement (point))))
10046 ((c-keyword-member kw-sym 'c-lambda-kwds)
10047 (when (or (not passed-paren)
10048 (eq passed-paren ?\())
10049 (cons 'inlambda (point))))
10050 ((c-keyword-member kw-sym 'c-block-stmt-kwds)
10051 nil)
10052 (t
10053 'maybe)))
10054
10055 (if (looking-at "\\s(")
10056 (if passed-paren
10057 (if (and (eq passed-paren ?\[)
10058 (eq (char-after) ?\[))
10059 ;; Accept several square bracket sexps for
10060 ;; Java array initializations.
10061 'maybe)
10062 (setq passed-paren (char-after))
10063 'maybe)
10064 'maybe))))
10065
10066 (if (eq res 'maybe)
10067 (when (and c-recognize-paren-inexpr-blocks
10068 block-follows
10069 containing-sexp
10070 (eq (char-after containing-sexp) ?\())
10071 (goto-char containing-sexp)
10072 (if (or (save-excursion
10073 (c-backward-syntactic-ws lim)
10074 (while (and (eq (char-before) ?>)
10075 (c-get-char-property (1- (point))
10076 'syntax-table)
10077 (c-go-list-backward nil lim))
10078 (c-backward-syntactic-ws lim))
10079 (and (> (point) (or lim (point-min)))
10080 (c-on-identifier)))
10081 (and c-special-brace-lists
10082 (c-looking-at-special-brace-list)))
10083 nil
10084 (cons 'inexpr-statement (point))))
10085
10086 res))))
10087
10088 (defun c-looking-at-inexpr-block-backward (paren-state)
10089 ;; Returns non-nil if we're looking at the end of an in-expression
10090 ;; block, otherwise the same as `c-looking-at-inexpr-block'.
10091 ;; PAREN-STATE is the paren state relevant at the current position.
10092 ;;
10093 ;; This function might do hidden buffer changes.
10094 (save-excursion
10095 ;; We currently only recognize a block.
10096 (let ((here (point))
10097 (elem (car-safe paren-state))
10098 containing-sexp)
10099 (when (and (consp elem)
10100 (progn (goto-char (cdr elem))
10101 (c-forward-syntactic-ws here)
10102 (= (point) here)))
10103 (goto-char (car elem))
10104 (if (setq paren-state (cdr paren-state))
10105 (setq containing-sexp (car-safe paren-state)))
10106 (c-looking-at-inexpr-block (c-safe-position containing-sexp
10107 paren-state)
10108 containing-sexp)))))
10109
10110 (defun c-at-macro-vsemi-p (&optional pos)
10111 ;; Is there a "virtual semicolon" at POS or point?
10112 ;; (See cc-defs.el for full details of "virtual semicolons".)
10113 ;;
10114 ;; This is true when point is at the last non syntactic WS position on the
10115 ;; line, there is a macro call last on the line, and this particular macro's
10116 ;; name is defined by the regexp `c-vs-macro-regexp' as not needing a
10117 ;; semicolon.
10118 (save-excursion
10119 (save-restriction
10120 (widen)
10121 (if pos
10122 (goto-char pos)
10123 (setq pos (point)))
10124 (and
10125 c-macro-with-semi-re
10126 (eq (skip-chars-backward " \t") 0)
10127
10128 ;; Check we've got nothing after this except comments and empty lines
10129 ;; joined by escaped EOLs.
10130 (skip-chars-forward " \t") ; always returns non-nil.
10131 (progn
10132 (while ; go over 1 block comment per iteration.
10133 (and
10134 (looking-at "\\(\\\\[\n\r][ \t]*\\)*")
10135 (goto-char (match-end 0))
10136 (cond
10137 ((looking-at c-block-comment-start-regexp)
10138 (and (forward-comment 1)
10139 (skip-chars-forward " \t"))) ; always returns non-nil
10140 ((looking-at c-line-comment-start-regexp)
10141 (end-of-line)
10142 nil)
10143 (t nil))))
10144 (eolp))
10145
10146 (goto-char pos)
10147 (progn (c-backward-syntactic-ws)
10148 (eq (point) pos))
10149
10150 ;; Check for one of the listed macros being before point.
10151 (or (not (eq (char-before) ?\)))
10152 (when (c-go-list-backward)
10153 (c-backward-syntactic-ws)
10154 t))
10155 (c-simple-skip-symbol-backward)
10156 (looking-at c-macro-with-semi-re)
10157 (goto-char pos)
10158 (not (c-in-literal)))))) ; The most expensive check last.
10159
10160 (defun c-macro-vsemi-status-unknown-p () t) ; See cc-defs.el.
10161
10162 \f
10163 ;; `c-guess-basic-syntax' and the functions that precedes it below
10164 ;; implements the main decision tree for determining the syntactic
10165 ;; analysis of the current line of code.
10166
10167 ;; Dynamically bound to t when `c-guess-basic-syntax' is called during
10168 ;; auto newline analysis.
10169 (defvar c-auto-newline-analysis nil)
10170
10171 (defun c-brace-anchor-point (bracepos)
10172 ;; BRACEPOS is the position of a brace in a construct like "namespace
10173 ;; Bar {". Return the anchor point in this construct; this is the
10174 ;; earliest symbol on the brace's line which isn't earlier than
10175 ;; "namespace".
10176 ;;
10177 ;; Currently (2007-08-17), "like namespace" means "matches
10178 ;; c-other-block-decl-kwds". It doesn't work with "class" or "struct"
10179 ;; or anything like that.
10180 (save-excursion
10181 (let ((boi (c-point 'boi bracepos)))
10182 (goto-char bracepos)
10183 (while (and (> (point) boi)
10184 (not (looking-at c-other-decl-block-key)))
10185 (c-backward-token-2))
10186 (if (> (point) boi) (point) boi))))
10187
10188 (defsubst c-add-syntax (symbol &rest args)
10189 ;; A simple function to prepend a new syntax element to
10190 ;; `c-syntactic-context'. Using `setq' on it is unsafe since it
10191 ;; should always be dynamically bound but since we read it first
10192 ;; we'll fail properly anyway if this function is misused.
10193 (setq c-syntactic-context (cons (cons symbol args)
10194 c-syntactic-context)))
10195
10196 (defsubst c-append-syntax (symbol &rest args)
10197 ;; Like `c-add-syntax' but appends to the end of the syntax list.
10198 ;; (Normally not necessary.)
10199 (setq c-syntactic-context (nconc c-syntactic-context
10200 (list (cons symbol args)))))
10201
10202 (defun c-add-stmt-syntax (syntax-symbol
10203 syntax-extra-args
10204 stop-at-boi-only
10205 containing-sexp
10206 paren-state)
10207 ;; Add the indicated SYNTAX-SYMBOL to `c-syntactic-context', extending it as
10208 ;; needed with further syntax elements of the types `substatement',
10209 ;; `inexpr-statement', `arglist-cont-nonempty', `statement-block-intro', and
10210 ;; `defun-block-intro'.
10211 ;;
10212 ;; Do the generic processing to anchor the given syntax symbol on
10213 ;; the preceding statement: Skip over any labels and containing
10214 ;; statements on the same line, and then search backward until we
10215 ;; find a statement or block start that begins at boi without a
10216 ;; label or comment.
10217 ;;
10218 ;; Point is assumed to be at the prospective anchor point for the
10219 ;; given SYNTAX-SYMBOL. More syntax entries are added if we need to
10220 ;; skip past open parens and containing statements. Most of the added
10221 ;; syntax elements will get the same anchor point - the exception is
10222 ;; for an anchor in a construct like "namespace"[*] - this is as early
10223 ;; as possible in the construct but on the same line as the {.
10224 ;;
10225 ;; [*] i.e. with a keyword matching c-other-block-decl-kwds.
10226 ;;
10227 ;; SYNTAX-EXTRA-ARGS are a list of the extra arguments for the
10228 ;; syntax symbol. They are appended after the anchor point.
10229 ;;
10230 ;; If STOP-AT-BOI-ONLY is nil, we can stop in the middle of the line
10231 ;; if the current statement starts there.
10232 ;;
10233 ;; Note: It's not a problem if PAREN-STATE "overshoots"
10234 ;; CONTAINING-SEXP, i.e. contains info about parens further down.
10235 ;;
10236 ;; This function might do hidden buffer changes.
10237
10238 (if (= (point) (c-point 'boi))
10239 ;; This is by far the most common case, so let's give it special
10240 ;; treatment.
10241 (apply 'c-add-syntax syntax-symbol (point) syntax-extra-args)
10242
10243 (let ((syntax-last c-syntactic-context)
10244 (boi (c-point 'boi))
10245 ;; Set when we're on a label, so that we don't stop there.
10246 ;; FIXME: To be complete we should check if we're on a label
10247 ;; now at the start.
10248 on-label)
10249
10250 ;; Use point as the anchor point for "namespace", "extern", etc.
10251 (apply 'c-add-syntax syntax-symbol
10252 (if (rassq syntax-symbol c-other-decl-block-key-in-symbols-alist)
10253 (point) nil)
10254 syntax-extra-args)
10255
10256 ;; Loop while we have to back out of containing blocks.
10257 (while
10258 (and
10259 (catch 'back-up-block
10260
10261 ;; Loop while we have to back up statements.
10262 (while (or (/= (point) boi)
10263 on-label
10264 (looking-at c-comment-start-regexp))
10265
10266 ;; Skip past any comments that stands between the
10267 ;; statement start and boi.
10268 (let ((savepos (point)))
10269 (while (and (/= savepos boi)
10270 (c-backward-single-comment))
10271 (setq savepos (point)
10272 boi (c-point 'boi)))
10273 (goto-char savepos))
10274
10275 ;; Skip to the beginning of this statement or backward
10276 ;; another one.
10277 (let ((old-pos (point))
10278 (old-boi boi)
10279 (step-type (c-beginning-of-statement-1 containing-sexp)))
10280 (setq boi (c-point 'boi)
10281 on-label (eq step-type 'label))
10282
10283 (cond ((= (point) old-pos)
10284 ;; If we didn't move we're at the start of a block and
10285 ;; have to continue outside it.
10286 (throw 'back-up-block t))
10287
10288 ((and (eq step-type 'up)
10289 (>= (point) old-boi)
10290 (looking-at "else\\>[^_]")
10291 (save-excursion
10292 (goto-char old-pos)
10293 (looking-at "if\\>[^_]")))
10294 ;; Special case to avoid deeper and deeper indentation
10295 ;; of "else if" clauses.
10296 )
10297
10298 ((and (not stop-at-boi-only)
10299 (/= old-pos old-boi)
10300 (memq step-type '(up previous)))
10301 ;; If stop-at-boi-only is nil, we shouldn't back up
10302 ;; over previous or containing statements to try to
10303 ;; reach boi, so go back to the last position and
10304 ;; exit.
10305 (goto-char old-pos)
10306 (throw 'back-up-block nil))
10307
10308 (t
10309 (if (and (not stop-at-boi-only)
10310 (memq step-type '(up previous beginning)))
10311 ;; If we've moved into another statement then we
10312 ;; should no longer try to stop in the middle of a
10313 ;; line.
10314 (setq stop-at-boi-only t))
10315
10316 ;; Record this as a substatement if we skipped up one
10317 ;; level.
10318 (when (eq step-type 'up)
10319 (c-add-syntax 'substatement nil))))
10320 )))
10321
10322 containing-sexp)
10323
10324 ;; Now we have to go out of this block.
10325 (goto-char containing-sexp)
10326
10327 ;; Don't stop in the middle of a special brace list opener
10328 ;; like "({".
10329 (when c-special-brace-lists
10330 (let ((special-list (c-looking-at-special-brace-list)))
10331 (when (and special-list
10332 (< (car (car special-list)) (point)))
10333 (setq containing-sexp (car (car special-list)))
10334 (goto-char containing-sexp))))
10335
10336 (setq paren-state (c-whack-state-after containing-sexp paren-state)
10337 containing-sexp (c-most-enclosing-brace paren-state)
10338 boi (c-point 'boi))
10339
10340 ;; Analyze the construct in front of the block we've stepped out
10341 ;; from and add the right syntactic element for it.
10342 (let ((paren-pos (point))
10343 (paren-char (char-after))
10344 step-type)
10345
10346 (if (eq paren-char ?\()
10347 ;; Stepped out of a parenthesis block, so we're in an
10348 ;; expression now.
10349 (progn
10350 (when (/= paren-pos boi)
10351 (if (and c-recognize-paren-inexpr-blocks
10352 (progn
10353 (c-backward-syntactic-ws containing-sexp)
10354 (or (not (looking-at "\\>"))
10355 (not (c-on-identifier))))
10356 (save-excursion
10357 (goto-char (1+ paren-pos))
10358 (c-forward-syntactic-ws)
10359 (eq (char-after) ?{)))
10360 ;; Stepped out of an in-expression statement. This
10361 ;; syntactic element won't get an anchor pos.
10362 (c-add-syntax 'inexpr-statement)
10363
10364 ;; A parenthesis normally belongs to an arglist.
10365 (c-add-syntax 'arglist-cont-nonempty nil paren-pos)))
10366
10367 (goto-char (max boi
10368 (if containing-sexp
10369 (1+ containing-sexp)
10370 (point-min))))
10371 (setq step-type 'same
10372 on-label nil))
10373
10374 ;; Stepped out of a brace block.
10375 (setq step-type (c-beginning-of-statement-1 containing-sexp)
10376 on-label (eq step-type 'label))
10377
10378 (if (and (eq step-type 'same)
10379 (/= paren-pos (point)))
10380 (let (inexpr)
10381 (cond
10382 ((save-excursion
10383 (goto-char paren-pos)
10384 (setq inexpr (c-looking-at-inexpr-block
10385 (c-safe-position containing-sexp paren-state)
10386 containing-sexp)))
10387 (c-add-syntax (if (eq (car inexpr) 'inlambda)
10388 'defun-block-intro
10389 'statement-block-intro)
10390 nil))
10391 ((looking-at c-other-decl-block-key)
10392 (c-add-syntax
10393 (cdr (assoc (match-string 1)
10394 c-other-decl-block-key-in-symbols-alist))
10395 (max (c-point 'boi paren-pos) (point))))
10396 (t (c-add-syntax 'defun-block-intro nil))))
10397
10398 (c-add-syntax 'statement-block-intro nil)))
10399
10400 (if (= paren-pos boi)
10401 ;; Always done if the open brace was at boi. The
10402 ;; c-beginning-of-statement-1 call above is necessary
10403 ;; anyway, to decide the type of block-intro to add.
10404 (goto-char paren-pos)
10405 (setq boi (c-point 'boi)))
10406 ))
10407
10408 ;; Fill in the current point as the anchor for all the symbols
10409 ;; added above.
10410 (let ((p c-syntactic-context) q)
10411 (while (not (eq p syntax-last))
10412 (setq q (cdr (car p))) ; e.g. (nil 28) [from (arglist-cont-nonempty nil 28)]
10413 (while q
10414 (unless (car q)
10415 (setcar q (point)))
10416 (setq q (cdr q)))
10417 (setq p (cdr p))))
10418 )))
10419
10420 (defun c-add-class-syntax (symbol
10421 containing-decl-open
10422 containing-decl-start
10423 containing-decl-kwd
10424 paren-state)
10425 ;; The inclass and class-close syntactic symbols are added in
10426 ;; several places and some work is needed to fix everything.
10427 ;; Therefore it's collected here.
10428 ;;
10429 ;; This function might do hidden buffer changes.
10430 (goto-char containing-decl-open)
10431 (if (and (eq symbol 'inclass) (= (point) (c-point 'boi)))
10432 (progn
10433 (c-add-syntax symbol containing-decl-open)
10434 containing-decl-open)
10435 (goto-char containing-decl-start)
10436 ;; Ought to use `c-add-stmt-syntax' instead of backing up to boi
10437 ;; here, but we have to do like this for compatibility.
10438 (back-to-indentation)
10439 (c-add-syntax symbol (point))
10440 (if (and (c-keyword-member containing-decl-kwd
10441 'c-inexpr-class-kwds)
10442 (/= containing-decl-start (c-point 'boi containing-decl-start)))
10443 (c-add-syntax 'inexpr-class))
10444 (point)))
10445
10446 (defun c-guess-continued-construct (indent-point
10447 char-after-ip
10448 beg-of-same-or-containing-stmt
10449 containing-sexp
10450 paren-state)
10451 ;; This function contains the decision tree reached through both
10452 ;; cases 18 and 10. It's a continued statement or top level
10453 ;; construct of some kind.
10454 ;;
10455 ;; This function might do hidden buffer changes.
10456
10457 (let (special-brace-list placeholder)
10458 (goto-char indent-point)
10459 (skip-chars-forward " \t")
10460
10461 (cond
10462 ;; (CASE A removed.)
10463 ;; CASE B: open braces for class or brace-lists
10464 ((setq special-brace-list
10465 (or (and c-special-brace-lists
10466 (c-looking-at-special-brace-list))
10467 (eq char-after-ip ?{)))
10468
10469 (cond
10470 ;; CASE B.1: class-open
10471 ((save-excursion
10472 (and (eq (char-after) ?{)
10473 (c-looking-at-decl-block containing-sexp t)
10474 (setq beg-of-same-or-containing-stmt (point))))
10475 (c-add-syntax 'class-open beg-of-same-or-containing-stmt))
10476
10477 ;; CASE B.2: brace-list-open
10478 ((or (consp special-brace-list)
10479 (save-excursion
10480 (goto-char beg-of-same-or-containing-stmt)
10481 (c-syntactic-re-search-forward "=\\([^=]\\|$\\)"
10482 indent-point t t t)))
10483 ;; The most semantically accurate symbol here is
10484 ;; brace-list-open, but we normally report it simply as a
10485 ;; statement-cont. The reason is that one normally adjusts
10486 ;; brace-list-open for brace lists as top-level constructs,
10487 ;; and brace lists inside statements is a completely different
10488 ;; context. C.f. case 5A.3.
10489 (c-beginning-of-statement-1 containing-sexp)
10490 (c-add-stmt-syntax (if c-auto-newline-analysis
10491 ;; Turn off the dwim above when we're
10492 ;; analyzing the nature of the brace
10493 ;; for the auto newline feature.
10494 'brace-list-open
10495 'statement-cont)
10496 nil nil
10497 containing-sexp paren-state))
10498
10499 ;; CASE B.3: The body of a function declared inside a normal
10500 ;; block. Can occur e.g. in Pike and when using gcc
10501 ;; extensions, but watch out for macros followed by blocks.
10502 ;; C.f. cases E, 16F and 17G.
10503 ((and (not (c-at-statement-start-p))
10504 (eq (c-beginning-of-statement-1 containing-sexp nil nil t)
10505 'same)
10506 (save-excursion
10507 (let ((c-recognize-typeless-decls nil))
10508 ;; Turn off recognition of constructs that lacks a
10509 ;; type in this case, since that's more likely to be
10510 ;; a macro followed by a block.
10511 (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil))))
10512 (c-add-stmt-syntax 'defun-open nil t
10513 containing-sexp paren-state))
10514
10515 ;; CASE B.4: Continued statement with block open. The most
10516 ;; accurate analysis is perhaps `statement-cont' together with
10517 ;; `block-open' but we play DWIM and use `substatement-open'
10518 ;; instead. The rationale is that this typically is a macro
10519 ;; followed by a block which makes it very similar to a
10520 ;; statement with a substatement block.
10521 (t
10522 (c-add-stmt-syntax 'substatement-open nil nil
10523 containing-sexp paren-state))
10524 ))
10525
10526 ;; CASE C: iostream insertion or extraction operator
10527 ((and (looking-at "\\(<<\\|>>\\)\\([^=]\\|$\\)")
10528 (save-excursion
10529 (goto-char beg-of-same-or-containing-stmt)
10530 ;; If there is no preceding streamop in the statement
10531 ;; then indent this line as a normal statement-cont.
10532 (when (c-syntactic-re-search-forward
10533 "\\(<<\\|>>\\)\\([^=]\\|$\\)" indent-point 'move t t)
10534 (c-add-syntax 'stream-op (c-point 'boi))
10535 t))))
10536
10537 ;; CASE E: In the "K&R region" of a function declared inside a
10538 ;; normal block. C.f. case B.3.
10539 ((and (save-excursion
10540 ;; Check that the next token is a '{'. This works as
10541 ;; long as no language that allows nested function
10542 ;; definitions allows stuff like member init lists, K&R
10543 ;; declarations or throws clauses there.
10544 ;;
10545 ;; Note that we do a forward search for something ahead
10546 ;; of the indentation line here. That's not good since
10547 ;; the user might not have typed it yet. Unfortunately
10548 ;; it's exceedingly tricky to recognize a function
10549 ;; prototype in a code block without resorting to this.
10550 (c-forward-syntactic-ws)
10551 (eq (char-after) ?{))
10552 (not (c-at-statement-start-p))
10553 (eq (c-beginning-of-statement-1 containing-sexp nil nil t)
10554 'same)
10555 (save-excursion
10556 (let ((c-recognize-typeless-decls nil))
10557 ;; Turn off recognition of constructs that lacks a
10558 ;; type in this case, since that's more likely to be
10559 ;; a macro followed by a block.
10560 (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil))))
10561 (c-add-stmt-syntax 'func-decl-cont nil t
10562 containing-sexp paren-state))
10563
10564 ;;CASE F: continued statement and the only preceding items are
10565 ;;annotations.
10566 ((and (c-major-mode-is 'java-mode)
10567 (setq placeholder (point))
10568 (c-beginning-of-statement-1)
10569 (progn
10570 (while (and (c-forward-annotation)
10571 (< (point) placeholder))
10572 (c-forward-syntactic-ws))
10573 t)
10574 (prog1
10575 (>= (point) placeholder)
10576 (goto-char placeholder)))
10577 (c-beginning-of-statement-1 containing-sexp)
10578 (c-add-syntax 'annotation-var-cont (point)))
10579
10580 ;; CASE G: a template list continuation?
10581 ;; Mostly a duplication of case 5D.3 to fix templates-19:
10582 ((and (c-major-mode-is 'c++-mode)
10583 (save-excursion
10584 (goto-char indent-point)
10585 (c-with-syntax-table c++-template-syntax-table
10586 (setq placeholder (c-up-list-backward)))
10587 (and placeholder
10588 (eq (char-after placeholder) ?<)
10589 (/= (char-before placeholder) ?<)
10590 (progn
10591 (goto-char (1+ placeholder))
10592 (not (looking-at c-<-op-cont-regexp))))))
10593 (c-with-syntax-table c++-template-syntax-table
10594 (goto-char placeholder)
10595 (c-beginning-of-statement-1 containing-sexp t))
10596 (if (save-excursion
10597 (c-backward-syntactic-ws containing-sexp)
10598 (eq (char-before) ?<))
10599 ;; In a nested template arglist.
10600 (progn
10601 (goto-char placeholder)
10602 (c-syntactic-skip-backward "^,;" containing-sexp t)
10603 (c-forward-syntactic-ws))
10604 (back-to-indentation))
10605 ;; FIXME: Should use c-add-stmt-syntax, but it's not yet
10606 ;; template aware.
10607 (c-add-syntax 'template-args-cont (point) placeholder))
10608
10609 ;; CASE D: continued statement.
10610 (t
10611 (c-beginning-of-statement-1 containing-sexp)
10612 (c-add-stmt-syntax 'statement-cont nil nil
10613 containing-sexp paren-state))
10614 )))
10615
10616 ;; The next autoload was added by RMS on 2005/8/9 - don't know why (ACM,
10617 ;; 2005/11/29).
10618 ;;;###autoload
10619 (defun c-guess-basic-syntax ()
10620 "Return the syntactic context of the current line."
10621 (save-excursion
10622 (beginning-of-line)
10623 (c-save-buffer-state
10624 ((indent-point (point))
10625 (case-fold-search nil)
10626 ;; A whole ugly bunch of various temporary variables. Have
10627 ;; to declare them here since it's not possible to declare
10628 ;; a variable with only the scope of a cond test and the
10629 ;; following result clauses, and most of this function is a
10630 ;; single gigantic cond. :P
10631 literal char-before-ip before-ws-ip char-after-ip macro-start
10632 in-macro-expr c-syntactic-context placeholder c-in-literal-cache
10633 step-type tmpsymbol keyword injava-inher special-brace-list tmp-pos
10634 containing-<
10635 ;; The following record some positions for the containing
10636 ;; declaration block if we're directly within one:
10637 ;; `containing-decl-open' is the position of the open
10638 ;; brace. `containing-decl-start' is the start of the
10639 ;; declaration. `containing-decl-kwd' is the keyword
10640 ;; symbol of the keyword that tells what kind of block it
10641 ;; is.
10642 containing-decl-open
10643 containing-decl-start
10644 containing-decl-kwd
10645 ;; The open paren of the closest surrounding sexp or nil if
10646 ;; there is none.
10647 containing-sexp
10648 ;; The position after the closest preceding brace sexp
10649 ;; (nested sexps are ignored), or the position after
10650 ;; `containing-sexp' if there is none, or (point-min) if
10651 ;; `containing-sexp' is nil.
10652 lim
10653 ;; The paren state outside `containing-sexp', or at
10654 ;; `indent-point' if `containing-sexp' is nil.
10655 (paren-state (c-parse-state))
10656 ;; There's always at most one syntactic element which got
10657 ;; an anchor pos. It's stored in syntactic-relpos.
10658 syntactic-relpos
10659 (c-stmt-delim-chars c-stmt-delim-chars))
10660
10661 ;; Check if we're directly inside an enclosing declaration
10662 ;; level block.
10663 (when (and (setq containing-sexp
10664 (c-most-enclosing-brace paren-state))
10665 (progn
10666 (goto-char containing-sexp)
10667 (eq (char-after) ?{))
10668 (setq placeholder
10669 (c-looking-at-decl-block
10670 (c-most-enclosing-brace paren-state
10671 containing-sexp)
10672 t)))
10673 (setq containing-decl-open containing-sexp
10674 containing-decl-start (point)
10675 containing-sexp nil)
10676 (goto-char placeholder)
10677 (setq containing-decl-kwd (and (looking-at c-keywords-regexp)
10678 (c-keyword-sym (match-string 1)))))
10679
10680 ;; Init some position variables.
10681 (if paren-state
10682 (progn
10683 (setq containing-sexp (car paren-state)
10684 paren-state (cdr paren-state))
10685 (if (consp containing-sexp)
10686 (save-excursion
10687 (goto-char (cdr containing-sexp))
10688 (if (and (c-major-mode-is 'c++-mode)
10689 (c-back-over-member-initializer-braces))
10690 (c-syntactic-skip-backward "^}" nil t))
10691 (setq lim (point))
10692 (if paren-state
10693 ;; Ignore balanced paren. The next entry
10694 ;; can't be another one.
10695 (setq containing-sexp (car paren-state)
10696 paren-state (cdr paren-state))
10697 ;; If there is no surrounding open paren then
10698 ;; put the last balanced pair back on paren-state.
10699 (setq paren-state (cons containing-sexp paren-state)
10700 containing-sexp nil)))
10701 (setq lim (1+ containing-sexp))))
10702 (setq lim (point-min)))
10703
10704 ;; If we're in a parenthesis list then ',' delimits the
10705 ;; "statements" rather than being an operator (with the
10706 ;; exception of the "for" clause). This difference is
10707 ;; typically only noticeable when statements are used in macro
10708 ;; arglists.
10709 (when (and containing-sexp
10710 (eq (char-after containing-sexp) ?\())
10711 (setq c-stmt-delim-chars c-stmt-delim-chars-with-comma))
10712 ;; cache char before and after indent point, and move point to
10713 ;; the most likely position to perform the majority of tests
10714 (goto-char indent-point)
10715 (c-backward-syntactic-ws lim)
10716 (setq before-ws-ip (point)
10717 char-before-ip (char-before))
10718 (goto-char indent-point)
10719 (skip-chars-forward " \t")
10720 (setq char-after-ip (char-after))
10721
10722 ;; are we in a literal?
10723 (setq literal (c-in-literal lim))
10724
10725 ;; now figure out syntactic qualities of the current line
10726 (cond
10727
10728 ;; CASE 1: in a string.
10729 ((eq literal 'string)
10730 (c-add-syntax 'string (c-point 'bopl)))
10731
10732 ;; CASE 2: in a C or C++ style comment.
10733 ((and (memq literal '(c c++))
10734 ;; This is a kludge for XEmacs where we use
10735 ;; `buffer-syntactic-context', which doesn't correctly
10736 ;; recognize "\*/" to end a block comment.
10737 ;; `parse-partial-sexp' which is used by
10738 ;; `c-literal-limits' will however do that in most
10739 ;; versions, which results in that we get nil from
10740 ;; `c-literal-limits' even when `c-in-literal' claims
10741 ;; we're inside a comment.
10742 (setq placeholder (c-literal-start lim)))
10743 (c-add-syntax literal placeholder))
10744
10745 ;; CASE 3: in a cpp preprocessor macro continuation.
10746 ((and (save-excursion
10747 (when (c-beginning-of-macro)
10748 (setq macro-start (point))))
10749 (/= macro-start (c-point 'boi))
10750 (progn
10751 (setq tmpsymbol 'cpp-macro-cont)
10752 (or (not c-syntactic-indentation-in-macros)
10753 (save-excursion
10754 (goto-char macro-start)
10755 ;; If at the beginning of the body of a #define
10756 ;; directive then analyze as cpp-define-intro
10757 ;; only. Go on with the syntactic analysis
10758 ;; otherwise. in-macro-expr is set if we're in a
10759 ;; cpp expression, i.e. before the #define body
10760 ;; or anywhere in a non-#define directive.
10761 (if (c-forward-to-cpp-define-body)
10762 (let ((indent-boi (c-point 'boi indent-point)))
10763 (setq in-macro-expr (> (point) indent-boi)
10764 tmpsymbol 'cpp-define-intro)
10765 (= (point) indent-boi))
10766 (setq in-macro-expr t)
10767 nil)))))
10768 (c-add-syntax tmpsymbol macro-start)
10769 (setq macro-start nil))
10770
10771 ;; CASE 11: an else clause?
10772 ((looking-at "else\\>[^_]")
10773 (c-beginning-of-statement-1 containing-sexp)
10774 (c-add-stmt-syntax 'else-clause nil t
10775 containing-sexp paren-state))
10776
10777 ;; CASE 12: while closure of a do/while construct?
10778 ((and (looking-at "while\\>[^_]")
10779 (save-excursion
10780 (prog1 (eq (c-beginning-of-statement-1 containing-sexp)
10781 'beginning)
10782 (setq placeholder (point)))))
10783 (goto-char placeholder)
10784 (c-add-stmt-syntax 'do-while-closure nil t
10785 containing-sexp paren-state))
10786
10787 ;; CASE 13: A catch or finally clause? This case is simpler
10788 ;; than if-else and do-while, because a block is required
10789 ;; after every try, catch and finally.
10790 ((save-excursion
10791 (and (cond ((c-major-mode-is 'c++-mode)
10792 (looking-at "catch\\>[^_]"))
10793 ((c-major-mode-is 'java-mode)
10794 (looking-at "\\(catch\\|finally\\)\\>[^_]")))
10795 (and (c-safe (c-backward-syntactic-ws)
10796 (c-backward-sexp)
10797 t)
10798 (eq (char-after) ?{)
10799 (c-safe (c-backward-syntactic-ws)
10800 (c-backward-sexp)
10801 t)
10802 (if (eq (char-after) ?\()
10803 (c-safe (c-backward-sexp) t)
10804 t))
10805 (looking-at "\\(try\\|catch\\)\\>[^_]")
10806 (setq placeholder (point))))
10807 (goto-char placeholder)
10808 (c-add-stmt-syntax 'catch-clause nil t
10809 containing-sexp paren-state))
10810
10811 ;; CASE 18: A substatement we can recognize by keyword.
10812 ((save-excursion
10813 (and c-opt-block-stmt-key
10814 (not (eq char-before-ip ?\;))
10815 (not (c-at-vsemi-p before-ws-ip))
10816 (not (memq char-after-ip '(?\) ?\] ?,)))
10817 (or (not (eq char-before-ip ?}))
10818 (c-looking-at-inexpr-block-backward c-state-cache))
10819 (> (point)
10820 (progn
10821 ;; Ought to cache the result from the
10822 ;; c-beginning-of-statement-1 calls here.
10823 (setq placeholder (point))
10824 (while (eq (setq step-type
10825 (c-beginning-of-statement-1 lim))
10826 'label))
10827 (if (eq step-type 'previous)
10828 (goto-char placeholder)
10829 (setq placeholder (point))
10830 (if (and (eq step-type 'same)
10831 (not (looking-at c-opt-block-stmt-key)))
10832 ;; Step up to the containing statement if we
10833 ;; stayed in the same one.
10834 (let (step)
10835 (while (eq
10836 (setq step
10837 (c-beginning-of-statement-1 lim))
10838 'label))
10839 (if (eq step 'up)
10840 (setq placeholder (point))
10841 ;; There was no containing statement after all.
10842 (goto-char placeholder)))))
10843 placeholder))
10844 (if (looking-at c-block-stmt-2-key)
10845 ;; Require a parenthesis after these keywords.
10846 ;; Necessary to catch e.g. synchronized in Java,
10847 ;; which can be used both as statement and
10848 ;; modifier.
10849 (and (zerop (c-forward-token-2 1 nil))
10850 (eq (char-after) ?\())
10851 (looking-at c-opt-block-stmt-key))))
10852
10853 (if (eq step-type 'up)
10854 ;; CASE 18A: Simple substatement.
10855 (progn
10856 (goto-char placeholder)
10857 (cond
10858 ((eq char-after-ip ?{)
10859 (c-add-stmt-syntax 'substatement-open nil nil
10860 containing-sexp paren-state))
10861 ((save-excursion
10862 (goto-char indent-point)
10863 (back-to-indentation)
10864 (c-forward-label))
10865 (c-add-stmt-syntax 'substatement-label nil nil
10866 containing-sexp paren-state))
10867 (t
10868 (c-add-stmt-syntax 'substatement nil nil
10869 containing-sexp paren-state))))
10870
10871 ;; CASE 18B: Some other substatement. This is shared
10872 ;; with case 10.
10873 (c-guess-continued-construct indent-point
10874 char-after-ip
10875 placeholder
10876 lim
10877 paren-state)))
10878
10879 ;; CASE 14: A case or default label
10880 ((save-excursion
10881 (and (looking-at c-label-kwds-regexp)
10882 (or (c-major-mode-is 'idl-mode)
10883 (and
10884 containing-sexp
10885 (goto-char containing-sexp)
10886 (eq (char-after) ?{)
10887 (progn (c-backward-syntactic-ws) t)
10888 (eq (char-before) ?\))
10889 (c-go-list-backward)
10890 (progn (c-backward-syntactic-ws) t)
10891 (c-simple-skip-symbol-backward)
10892 (looking-at c-block-stmt-2-key)))))
10893 (if containing-sexp
10894 (progn
10895 (goto-char containing-sexp)
10896 (setq lim (c-most-enclosing-brace c-state-cache
10897 containing-sexp))
10898 (c-backward-to-block-anchor lim)
10899 (c-add-stmt-syntax 'case-label nil t lim paren-state))
10900 ;; Got a bogus label at the top level. In lack of better
10901 ;; alternatives, anchor it on (point-min).
10902 (c-add-syntax 'case-label (point-min))))
10903
10904 ;; CASE 15: any other label
10905 ((save-excursion
10906 (back-to-indentation)
10907 (and (not (looking-at c-syntactic-ws-start))
10908 (not (looking-at c-label-kwds-regexp))
10909 (c-forward-label)))
10910 (cond (containing-decl-open
10911 (setq placeholder (c-add-class-syntax 'inclass
10912 containing-decl-open
10913 containing-decl-start
10914 containing-decl-kwd
10915 paren-state))
10916 ;; Append access-label with the same anchor point as
10917 ;; inclass gets.
10918 (c-append-syntax 'access-label placeholder))
10919
10920 (containing-sexp
10921 (goto-char containing-sexp)
10922 (setq lim (c-most-enclosing-brace c-state-cache
10923 containing-sexp))
10924 (save-excursion
10925 (setq tmpsymbol
10926 (if (and (eq (c-beginning-of-statement-1 lim) 'up)
10927 (looking-at "switch\\>[^_]"))
10928 ;; If the surrounding statement is a switch then
10929 ;; let's analyze all labels as switch labels, so
10930 ;; that they get lined up consistently.
10931 'case-label
10932 'label)))
10933 (c-backward-to-block-anchor lim)
10934 (c-add-stmt-syntax tmpsymbol nil t lim paren-state))
10935
10936 (t
10937 ;; A label on the top level. Treat it as a class
10938 ;; context. (point-min) is the closest we get to the
10939 ;; class open brace.
10940 (c-add-syntax 'access-label (point-min)))))
10941
10942 ;; CASE 4: In-expression statement. C.f. cases 7B, 16A and
10943 ;; 17E.
10944 ((setq placeholder (c-looking-at-inexpr-block
10945 (c-safe-position containing-sexp paren-state)
10946 containing-sexp
10947 ;; Have to turn on the heuristics after
10948 ;; the point even though it doesn't work
10949 ;; very well. C.f. test case class-16.pike.
10950 t))
10951 (setq tmpsymbol (assq (car placeholder)
10952 '((inexpr-class . class-open)
10953 (inexpr-statement . block-open))))
10954 (if tmpsymbol
10955 ;; It's a statement block or an anonymous class.
10956 (setq tmpsymbol (cdr tmpsymbol))
10957 ;; It's a Pike lambda. Check whether we are between the
10958 ;; lambda keyword and the argument list or at the defun
10959 ;; opener.
10960 (setq tmpsymbol (if (eq char-after-ip ?{)
10961 'inline-open
10962 'lambda-intro-cont)))
10963 (goto-char (cdr placeholder))
10964 (back-to-indentation)
10965 (c-add-stmt-syntax tmpsymbol nil t
10966 (c-most-enclosing-brace c-state-cache (point))
10967 paren-state)
10968 (unless (eq (point) (cdr placeholder))
10969 (c-add-syntax (car placeholder))))
10970
10971 ;; CASE 5: Line is inside a declaration level block or at top level.
10972 ((or containing-decl-open (null containing-sexp))
10973 (cond
10974
10975 ;; CASE 5A: we are looking at a defun, brace list, class,
10976 ;; or inline-inclass method opening brace
10977 ((setq special-brace-list
10978 (or (and c-special-brace-lists
10979 (c-looking-at-special-brace-list))
10980 (eq char-after-ip ?{)))
10981 (cond
10982
10983 ;; CASE 5A.1: Non-class declaration block open.
10984 ((save-excursion
10985 (let (tmp)
10986 (and (eq char-after-ip ?{)
10987 (setq tmp (c-looking-at-decl-block containing-sexp t))
10988 (progn
10989 (setq placeholder (point))
10990 (goto-char tmp)
10991 (looking-at c-symbol-key))
10992 (c-keyword-member
10993 (c-keyword-sym (setq keyword (match-string 0)))
10994 'c-other-block-decl-kwds))))
10995 (goto-char placeholder)
10996 (c-add-stmt-syntax
10997 (if (string-equal keyword "extern")
10998 ;; Special case for extern-lang-open.
10999 'extern-lang-open
11000 (intern (concat keyword "-open")))
11001 nil t containing-sexp paren-state))
11002
11003 ;; CASE 5A.2: we are looking at a class opening brace
11004 ((save-excursion
11005 (goto-char indent-point)
11006 (skip-chars-forward " \t")
11007 (and (eq (char-after) ?{)
11008 (c-looking-at-decl-block containing-sexp t)
11009 (setq placeholder (point))))
11010 (c-add-syntax 'class-open placeholder))
11011
11012 ;; CASE 5A.3: brace list open
11013 ((save-excursion
11014 (c-beginning-of-decl-1 lim)
11015 (while (cond
11016 ((looking-at c-specifier-key)
11017 (c-forward-keyword-clause 1))
11018 ((and c-opt-cpp-prefix
11019 (looking-at c-noise-macro-with-parens-name-re))
11020 (c-forward-noise-clause))))
11021 (setq placeholder (c-point 'boi))
11022 (or (consp special-brace-list)
11023 (and (or (save-excursion
11024 (goto-char indent-point)
11025 (setq tmpsymbol nil)
11026 (while (and (> (point) placeholder)
11027 (zerop (c-backward-token-2 1 t))
11028 (not (looking-at "=\\([^=]\\|$\\)")))
11029 (and c-opt-inexpr-brace-list-key
11030 (not tmpsymbol)
11031 (looking-at c-opt-inexpr-brace-list-key)
11032 (setq tmpsymbol 'topmost-intro-cont)))
11033 (looking-at "=\\([^=]\\|$\\)"))
11034 (looking-at c-brace-list-key))
11035 (save-excursion
11036 (while (and (< (point) indent-point)
11037 (zerop (c-forward-token-2 1 t))
11038 (not (memq (char-after) '(?\; ?\()))))
11039 (not (memq (char-after) '(?\; ?\()))
11040 ))))
11041 (if (and (not c-auto-newline-analysis)
11042 (c-major-mode-is 'java-mode)
11043 (eq tmpsymbol 'topmost-intro-cont))
11044 ;; We're in Java and have found that the open brace
11045 ;; belongs to a "new Foo[]" initialization list,
11046 ;; which means the brace list is part of an
11047 ;; expression and not a top level definition. We
11048 ;; therefore treat it as any topmost continuation
11049 ;; even though the semantically correct symbol still
11050 ;; is brace-list-open, on the same grounds as in
11051 ;; case B.2.
11052 (progn
11053 (c-beginning-of-statement-1 lim)
11054 (c-add-syntax 'topmost-intro-cont (c-point 'boi)))
11055 (c-add-syntax 'brace-list-open placeholder)))
11056
11057 ;; CASE 5A.4: inline defun open
11058 ((and containing-decl-open
11059 (not (c-keyword-member containing-decl-kwd
11060 'c-other-block-decl-kwds)))
11061 (c-add-syntax 'inline-open)
11062 (c-add-class-syntax 'inclass
11063 containing-decl-open
11064 containing-decl-start
11065 containing-decl-kwd
11066 paren-state))
11067
11068 ;; CASE 5A.5: ordinary defun open
11069 (t
11070 (save-excursion
11071 (c-beginning-of-decl-1 lim)
11072 (while (cond
11073 ((looking-at c-specifier-key)
11074 (c-forward-keyword-clause 1))
11075 ((and c-opt-cpp-prefix
11076 (looking-at c-noise-macro-with-parens-name-re))
11077 (c-forward-noise-clause))))
11078 (c-add-syntax 'defun-open (c-point 'boi))
11079 ;; Bogus to use bol here, but it's the legacy. (Resolved,
11080 ;; 2007-11-09)
11081 ))))
11082
11083 ;; CASE 5R: Member init list. (Used to be part of CASE 5B.1)
11084 ;; Note there is no limit on the backward search here, since member
11085 ;; init lists can, in practice, be very large.
11086 ((save-excursion
11087 (when (and (c-major-mode-is 'c++-mode)
11088 (setq placeholder (c-back-over-member-initializers)))
11089 (setq tmp-pos (point))))
11090 (if (= (c-point 'bosws) (1+ tmp-pos))
11091 (progn
11092 ;; There is no preceding member init clause.
11093 ;; Indent relative to the beginning of indentation
11094 ;; for the topmost-intro line that contains the
11095 ;; prototype's open paren.
11096 (goto-char placeholder)
11097 (c-add-syntax 'member-init-intro (c-point 'boi)))
11098 ;; Indent relative to the first member init clause.
11099 (goto-char (1+ tmp-pos))
11100 (c-forward-syntactic-ws)
11101 (c-add-syntax 'member-init-cont (point))))
11102
11103 ;; CASE 5B: After a function header but before the body (or
11104 ;; the ending semicolon if there's no body).
11105 ((save-excursion
11106 (when (setq placeholder (c-just-after-func-arglist-p
11107 (max lim (c-determine-limit 500))))
11108 (setq tmp-pos (point))))
11109 (cond
11110
11111 ;; CASE 5B.1: Member init list.
11112 ((eq (char-after tmp-pos) ?:)
11113 ;; There is no preceding member init clause.
11114 ;; Indent relative to the beginning of indentation
11115 ;; for the topmost-intro line that contains the
11116 ;; prototype's open paren.
11117 (goto-char placeholder)
11118 (c-add-syntax 'member-init-intro (c-point 'boi)))
11119
11120 ;; CASE 5B.2: K&R arg decl intro
11121 ((and c-recognize-knr-p
11122 (c-in-knr-argdecl lim))
11123 (c-beginning-of-statement-1 lim)
11124 (c-add-syntax 'knr-argdecl-intro (c-point 'boi))
11125 (if containing-decl-open
11126 (c-add-class-syntax 'inclass
11127 containing-decl-open
11128 containing-decl-start
11129 containing-decl-kwd
11130 paren-state)))
11131
11132 ;; CASE 5B.4: Nether region after a C++ or Java func
11133 ;; decl, which could include a `throws' declaration.
11134 (t
11135 (c-beginning-of-statement-1 lim)
11136 (c-add-syntax 'func-decl-cont (c-point 'boi))
11137 )))
11138
11139 ;; CASE 5C: inheritance line. could be first inheritance
11140 ;; line, or continuation of a multiple inheritance
11141 ((or (and (c-major-mode-is 'c++-mode)
11142 (progn
11143 (when (eq char-after-ip ?,)
11144 (skip-chars-forward " \t")
11145 (forward-char))
11146 (looking-at c-opt-postfix-decl-spec-key)))
11147 (and (or (eq char-before-ip ?:)
11148 ;; watch out for scope operator
11149 (save-excursion
11150 (and (eq char-after-ip ?:)
11151 (c-safe (forward-char 1) t)
11152 (not (eq (char-after) ?:))
11153 )))
11154 (save-excursion
11155 (c-beginning-of-statement-1 lim)
11156 (when (looking-at c-opt-<>-sexp-key)
11157 (goto-char (match-end 1))
11158 (c-forward-syntactic-ws)
11159 (c-forward-<>-arglist nil)
11160 (c-forward-syntactic-ws))
11161 (looking-at c-class-key)))
11162 ;; for Java
11163 (and (c-major-mode-is 'java-mode)
11164 (let ((fence (save-excursion
11165 (c-beginning-of-statement-1 lim)
11166 (point)))
11167 cont done)
11168 (save-excursion
11169 (while (not done)
11170 (cond ((looking-at c-opt-postfix-decl-spec-key)
11171 (setq injava-inher (cons cont (point))
11172 done t))
11173 ((or (not (c-safe (c-forward-sexp -1) t))
11174 (<= (point) fence))
11175 (setq done t))
11176 )
11177 (setq cont t)))
11178 injava-inher)
11179 (not (c-crosses-statement-barrier-p (cdr injava-inher)
11180 (point)))
11181 ))
11182 (cond
11183
11184 ;; CASE 5C.1: non-hanging colon on an inher intro
11185 ((eq char-after-ip ?:)
11186 (c-beginning-of-statement-1 lim)
11187 (c-add-syntax 'inher-intro (c-point 'boi))
11188 ;; don't add inclass symbol since relative point already
11189 ;; contains any class offset
11190 )
11191
11192 ;; CASE 5C.2: hanging colon on an inher intro
11193 ((eq char-before-ip ?:)
11194 (c-beginning-of-statement-1 lim)
11195 (c-add-syntax 'inher-intro (c-point 'boi))
11196 (if containing-decl-open
11197 (c-add-class-syntax 'inclass
11198 containing-decl-open
11199 containing-decl-start
11200 containing-decl-kwd
11201 paren-state)))
11202
11203 ;; CASE 5C.3: in a Java implements/extends
11204 (injava-inher
11205 (let ((where (cdr injava-inher))
11206 (cont (car injava-inher)))
11207 (goto-char where)
11208 (cond ((looking-at "throws\\>[^_]")
11209 (c-add-syntax 'func-decl-cont
11210 (progn (c-beginning-of-statement-1 lim)
11211 (c-point 'boi))))
11212 (cont (c-add-syntax 'inher-cont where))
11213 (t (c-add-syntax 'inher-intro
11214 (progn (goto-char (cdr injava-inher))
11215 (c-beginning-of-statement-1 lim)
11216 (point))))
11217 )))
11218
11219 ;; CASE 5C.4: a continued inheritance line
11220 (t
11221 (c-beginning-of-inheritance-list lim)
11222 (c-add-syntax 'inher-cont (point))
11223 ;; don't add inclass symbol since relative point already
11224 ;; contains any class offset
11225 )))
11226
11227 ;; CASE 5P: AWK pattern or function or continuation
11228 ;; thereof.
11229 ((c-major-mode-is 'awk-mode)
11230 (setq placeholder (point))
11231 (c-add-stmt-syntax
11232 (if (and (eq (c-beginning-of-statement-1) 'same)
11233 (/= (point) placeholder))
11234 'topmost-intro-cont
11235 'topmost-intro)
11236 nil nil
11237 containing-sexp paren-state))
11238
11239 ;; CASE 5D: this could be a top-level initialization, a
11240 ;; member init list continuation, or a template argument
11241 ;; list continuation.
11242 ((save-excursion
11243 ;; Note: We use the fact that lim is always after any
11244 ;; preceding brace sexp.
11245 (if c-recognize-<>-arglists
11246 (while (and
11247 (progn
11248 (c-syntactic-skip-backward "^;,=<>" lim t)
11249 (> (point) lim))
11250 (or
11251 (when c-overloadable-operators-regexp
11252 (when (setq placeholder (c-after-special-operator-id lim))
11253 (goto-char placeholder)
11254 t))
11255 (cond
11256 ((eq (char-before) ?>)
11257 (or (c-backward-<>-arglist nil lim)
11258 (backward-char))
11259 t)
11260 ((eq (char-before) ?<)
11261 (backward-char)
11262 (if (save-excursion
11263 (c-forward-<>-arglist nil))
11264 (progn (forward-char)
11265 nil)
11266 t))
11267 (t nil)))))
11268 ;; NB: No c-after-special-operator-id stuff in this
11269 ;; clause - we assume only C++ needs it.
11270 (c-syntactic-skip-backward "^;,=" lim t))
11271 (memq (char-before) '(?, ?= ?<)))
11272 (cond
11273
11274 ;; CASE 5D.3: perhaps a template list continuation?
11275 ((and (c-major-mode-is 'c++-mode)
11276 (save-excursion
11277 (save-restriction
11278 (c-with-syntax-table c++-template-syntax-table
11279 (goto-char indent-point)
11280 (setq placeholder (c-up-list-backward))
11281 (and placeholder
11282 (eq (char-after placeholder) ?<))))))
11283 (c-with-syntax-table c++-template-syntax-table
11284 (goto-char placeholder)
11285 (c-beginning-of-statement-1 lim t))
11286 (if (save-excursion
11287 (c-backward-syntactic-ws lim)
11288 (eq (char-before) ?<))
11289 ;; In a nested template arglist.
11290 (progn
11291 (goto-char placeholder)
11292 (c-syntactic-skip-backward "^,;" lim t)
11293 (c-forward-syntactic-ws))
11294 (back-to-indentation))
11295 ;; FIXME: Should use c-add-stmt-syntax, but it's not yet
11296 ;; template aware.
11297 (c-add-syntax 'template-args-cont (point) placeholder))
11298
11299 ;; CASE 5D.4: perhaps a multiple inheritance line?
11300 ((and (c-major-mode-is 'c++-mode)
11301 (save-excursion
11302 (c-beginning-of-statement-1 lim)
11303 (setq placeholder (point))
11304 (if (looking-at "static\\>[^_]")
11305 (c-forward-token-2 1 nil indent-point))
11306 (and (looking-at c-class-key)
11307 (zerop (c-forward-token-2 2 nil indent-point))
11308 (if (eq (char-after) ?<)
11309 (c-with-syntax-table c++-template-syntax-table
11310 (zerop (c-forward-token-2 1 t indent-point)))
11311 t)
11312 (eq (char-after) ?:))))
11313 (goto-char placeholder)
11314 (c-add-syntax 'inher-cont (c-point 'boi)))
11315
11316 ;; CASE 5D.5: Continuation of the "expression part" of a
11317 ;; top level construct. Or, perhaps, an unrecognized construct.
11318 (t
11319 (while (and (setq placeholder (point))
11320 (eq (car (c-beginning-of-decl-1 containing-sexp)) ; Can't use `lim' here.
11321 'same)
11322 (save-excursion
11323 (c-backward-syntactic-ws)
11324 (eq (char-before) ?}))
11325 (< (point) placeholder)))
11326 (c-add-stmt-syntax
11327 (cond
11328 ((eq (point) placeholder) 'statement) ; unrecognized construct
11329 ;; A preceding comma at the top level means that a
11330 ;; new variable declaration starts here. Use
11331 ;; topmost-intro-cont for it, for consistency with
11332 ;; the first variable declaration. C.f. case 5N.
11333 ((eq char-before-ip ?,) 'topmost-intro-cont)
11334 (t 'statement-cont))
11335 nil nil containing-sexp paren-state))
11336 ))
11337
11338 ;; CASE 5F: Close of a non-class declaration level block.
11339 ((and (eq char-after-ip ?})
11340 (c-keyword-member containing-decl-kwd
11341 'c-other-block-decl-kwds))
11342 ;; This is inconsistent: Should use `containing-decl-open'
11343 ;; here if it's at boi, like in case 5J.
11344 (goto-char containing-decl-start)
11345 (c-add-stmt-syntax
11346 (if (string-equal (symbol-name containing-decl-kwd) "extern")
11347 ;; Special case for compatibility with the
11348 ;; extern-lang syntactic symbols.
11349 'extern-lang-close
11350 (intern (concat (symbol-name containing-decl-kwd)
11351 "-close")))
11352 nil t
11353 (c-most-enclosing-brace paren-state (point))
11354 paren-state))
11355
11356 ;; CASE 5G: we are looking at the brace which closes the
11357 ;; enclosing nested class decl
11358 ((and containing-sexp
11359 (eq char-after-ip ?})
11360 (eq containing-decl-open containing-sexp))
11361 (c-add-class-syntax 'class-close
11362 containing-decl-open
11363 containing-decl-start
11364 containing-decl-kwd
11365 paren-state))
11366
11367 ;; CASE 5H: we could be looking at subsequent knr-argdecls
11368 ((and c-recognize-knr-p
11369 (not containing-sexp) ; can't be knr inside braces.
11370 (not (eq char-before-ip ?}))
11371 (save-excursion
11372 (setq placeholder (cdr (c-beginning-of-decl-1 lim)))
11373 (and placeholder
11374 ;; Do an extra check to avoid tripping up on
11375 ;; statements that occur in invalid contexts
11376 ;; (e.g. in macro bodies where we don't really
11377 ;; know the context of what we're looking at).
11378 (not (and c-opt-block-stmt-key
11379 (looking-at c-opt-block-stmt-key)))))
11380 (< placeholder indent-point))
11381 (goto-char placeholder)
11382 (c-add-syntax 'knr-argdecl (point)))
11383
11384 ;; CASE 5I: ObjC method definition.
11385 ((and c-opt-method-key
11386 (looking-at c-opt-method-key))
11387 (c-beginning-of-statement-1 nil t)
11388 (if (= (point) indent-point)
11389 ;; Handle the case when it's the first (non-comment)
11390 ;; thing in the buffer. Can't look for a 'same return
11391 ;; value from cbos1 since ObjC directives currently
11392 ;; aren't recognized fully, so that we get 'same
11393 ;; instead of 'previous if it moved over a preceding
11394 ;; directive.
11395 (goto-char (point-min)))
11396 (c-add-syntax 'objc-method-intro (c-point 'boi)))
11397
11398 ;; CASE 5N: At a variable declaration that follows a class
11399 ;; definition or some other block declaration that doesn't
11400 ;; end at the closing '}'. C.f. case 5D.5.
11401 ((progn
11402 (c-backward-syntactic-ws lim)
11403 (and (eq (char-before) ?})
11404 (save-excursion
11405 (let ((start (point)))
11406 (if (and c-state-cache
11407 (consp (car c-state-cache))
11408 (eq (cdar c-state-cache) (point)))
11409 ;; Speed up the backward search a bit.
11410 (goto-char (caar c-state-cache)))
11411 (c-beginning-of-decl-1 containing-sexp) ; Can't use `lim' here.
11412 (setq placeholder (point))
11413 (if (= start (point))
11414 ;; The '}' is unbalanced.
11415 nil
11416 (c-end-of-decl-1)
11417 (>= (point) indent-point))))))
11418 (goto-char placeholder)
11419 (c-add-stmt-syntax 'topmost-intro-cont nil nil
11420 containing-sexp paren-state))
11421
11422 ;; NOTE: The point is at the end of the previous token here.
11423
11424 ;; CASE 5J: we are at the topmost level, make
11425 ;; sure we skip back past any access specifiers
11426 ((and
11427 ;; A macro continuation line is never at top level.
11428 (not (and macro-start
11429 (> indent-point macro-start)))
11430 (save-excursion
11431 (setq placeholder (point))
11432 (or (memq char-before-ip '(?\; ?{ ?} nil))
11433 (c-at-vsemi-p before-ws-ip)
11434 (when (and (eq char-before-ip ?:)
11435 (eq (c-beginning-of-statement-1 lim)
11436 'label))
11437 (c-backward-syntactic-ws lim)
11438 (setq placeholder (point)))
11439 (and (c-major-mode-is 'objc-mode)
11440 (catch 'not-in-directive
11441 (c-beginning-of-statement-1 lim)
11442 (setq placeholder (point))
11443 (while (and (c-forward-objc-directive)
11444 (< (point) indent-point))
11445 (c-forward-syntactic-ws)
11446 (if (>= (point) indent-point)
11447 (throw 'not-in-directive t))
11448 (setq placeholder (point)))
11449 nil)))))
11450 ;; For historic reasons we anchor at bol of the last
11451 ;; line of the previous declaration. That's clearly
11452 ;; highly bogus and useless, and it makes our lives hard
11453 ;; to remain compatible. :P
11454 (goto-char placeholder)
11455 (c-add-syntax 'topmost-intro (c-point 'bol))
11456 (if containing-decl-open
11457 (if (c-keyword-member containing-decl-kwd
11458 'c-other-block-decl-kwds)
11459 (progn
11460 (goto-char (c-brace-anchor-point containing-decl-open))
11461 (c-add-stmt-syntax
11462 (if (string-equal (symbol-name containing-decl-kwd)
11463 "extern")
11464 ;; Special case for compatibility with the
11465 ;; extern-lang syntactic symbols.
11466 'inextern-lang
11467 (intern (concat "in"
11468 (symbol-name containing-decl-kwd))))
11469 nil t
11470 (c-most-enclosing-brace paren-state (point))
11471 paren-state))
11472 (c-add-class-syntax 'inclass
11473 containing-decl-open
11474 containing-decl-start
11475 containing-decl-kwd
11476 paren-state)))
11477 (when (and c-syntactic-indentation-in-macros
11478 macro-start
11479 (/= macro-start (c-point 'boi indent-point)))
11480 (c-add-syntax 'cpp-define-intro)
11481 (setq macro-start nil)))
11482
11483 ;; CASE 5K: we are at an ObjC method definition
11484 ;; continuation line.
11485 ((and c-opt-method-key
11486 (save-excursion
11487 (c-beginning-of-statement-1 lim)
11488 (beginning-of-line)
11489 (when (looking-at c-opt-method-key)
11490 (setq placeholder (point)))))
11491 (c-add-syntax 'objc-method-args-cont placeholder))
11492
11493 ;; CASE 5L: we are at the first argument of a template
11494 ;; arglist that begins on the previous line.
11495 ((and c-recognize-<>-arglists
11496 (eq (char-before) ?<)
11497 (not (and c-overloadable-operators-regexp
11498 (c-after-special-operator-id lim))))
11499 (c-beginning-of-statement-1 (c-safe-position (point) paren-state))
11500 (c-add-syntax 'template-args-cont (c-point 'boi)))
11501
11502 ;; CASE 5Q: we are at a statement within a macro.
11503 (macro-start
11504 (c-beginning-of-statement-1 containing-sexp)
11505 (c-add-stmt-syntax 'statement nil t containing-sexp paren-state))
11506
11507 ;;CASE 5N: We are at a topmost continuation line and the only
11508 ;;preceding items are annotations.
11509 ((and (c-major-mode-is 'java-mode)
11510 (setq placeholder (point))
11511 (c-beginning-of-statement-1)
11512 (progn
11513 (while (and (c-forward-annotation))
11514 (c-forward-syntactic-ws))
11515 t)
11516 (prog1
11517 (>= (point) placeholder)
11518 (goto-char placeholder)))
11519 (c-add-syntax 'annotation-top-cont (c-point 'boi)))
11520
11521 ;; CASE 5M: we are at a topmost continuation line
11522 (t
11523 (c-beginning-of-statement-1 (c-safe-position (point) paren-state))
11524 (when (c-major-mode-is 'objc-mode)
11525 (setq placeholder (point))
11526 (while (and (c-forward-objc-directive)
11527 (< (point) indent-point))
11528 (c-forward-syntactic-ws)
11529 (setq placeholder (point)))
11530 (goto-char placeholder))
11531 (c-add-syntax 'topmost-intro-cont (c-point 'boi)))
11532 ))
11533
11534 ;; (CASE 6 has been removed.)
11535
11536 ;; CASE 7: line is an expression, not a statement. Most
11537 ;; likely we are either in a function prototype or a function
11538 ;; call argument list
11539 ((not (or (and c-special-brace-lists
11540 (save-excursion
11541 (goto-char containing-sexp)
11542 (c-looking-at-special-brace-list)))
11543 (eq (char-after containing-sexp) ?{)))
11544 (cond
11545
11546 ;; CASE 7A: we are looking at the arglist closing paren.
11547 ;; C.f. case 7F.
11548 ((memq char-after-ip '(?\) ?\]))
11549 (goto-char containing-sexp)
11550 (setq placeholder (c-point 'boi))
11551 (if (and (c-safe (backward-up-list 1) t)
11552 (>= (point) placeholder))
11553 (progn
11554 (forward-char)
11555 (skip-chars-forward " \t"))
11556 (goto-char placeholder))
11557 (c-add-stmt-syntax 'arglist-close (list containing-sexp) t
11558 (c-most-enclosing-brace paren-state (point))
11559 paren-state))
11560
11561 ;; CASE 7B: Looking at the opening brace of an
11562 ;; in-expression block or brace list. C.f. cases 4, 16A
11563 ;; and 17E.
11564 ((and (eq char-after-ip ?{)
11565 (progn
11566 (setq placeholder (c-inside-bracelist-p (point)
11567 paren-state))
11568 (if placeholder
11569 (setq tmpsymbol '(brace-list-open . inexpr-class))
11570 (setq tmpsymbol '(block-open . inexpr-statement)
11571 placeholder
11572 (cdr-safe (c-looking-at-inexpr-block
11573 (c-safe-position containing-sexp
11574 paren-state)
11575 containing-sexp)))
11576 ;; placeholder is nil if it's a block directly in
11577 ;; a function arglist. That makes us skip out of
11578 ;; this case.
11579 )))
11580 (goto-char placeholder)
11581 (back-to-indentation)
11582 (c-add-stmt-syntax (car tmpsymbol) nil t
11583 (c-most-enclosing-brace paren-state (point))
11584 paren-state)
11585 (if (/= (point) placeholder)
11586 (c-add-syntax (cdr tmpsymbol))))
11587
11588 ;; CASE 7C: we are looking at the first argument in an empty
11589 ;; argument list. Use arglist-close if we're actually
11590 ;; looking at a close paren or bracket.
11591 ((memq char-before-ip '(?\( ?\[))
11592 (goto-char containing-sexp)
11593 (setq placeholder (c-point 'boi))
11594 (if (and (c-safe (backward-up-list 1) t)
11595 (>= (point) placeholder))
11596 (progn
11597 (forward-char)
11598 (skip-chars-forward " \t"))
11599 (goto-char placeholder))
11600 (c-add-stmt-syntax 'arglist-intro (list containing-sexp) t
11601 (c-most-enclosing-brace paren-state (point))
11602 paren-state))
11603
11604 ;; CASE 7D: we are inside a conditional test clause. treat
11605 ;; these things as statements
11606 ((progn
11607 (goto-char containing-sexp)
11608 (and (c-safe (c-forward-sexp -1) t)
11609 (looking-at "\\<for\\>[^_]")))
11610 (goto-char (1+ containing-sexp))
11611 (c-forward-syntactic-ws indent-point)
11612 (if (eq char-before-ip ?\;)
11613 (c-add-syntax 'statement (point))
11614 (c-add-syntax 'statement-cont (point))
11615 ))
11616
11617 ;; CASE 7E: maybe a continued ObjC method call. This is the
11618 ;; case when we are inside a [] bracketed exp, and what
11619 ;; precede the opening bracket is not an identifier.
11620 ((and c-opt-method-key
11621 (eq (char-after containing-sexp) ?\[)
11622 (progn
11623 (goto-char (1- containing-sexp))
11624 (c-backward-syntactic-ws (c-point 'bod))
11625 (if (not (looking-at c-symbol-key))
11626 (c-add-syntax 'objc-method-call-cont containing-sexp))
11627 )))
11628
11629 ;; CASE 7F: we are looking at an arglist continuation line,
11630 ;; but the preceding argument is on the same line as the
11631 ;; opening paren. This case includes multi-line
11632 ;; mathematical paren groupings, but we could be on a
11633 ;; for-list continuation line. C.f. case 7A.
11634 ((progn
11635 (goto-char (1+ containing-sexp))
11636 (< (save-excursion
11637 (c-forward-syntactic-ws)
11638 (point))
11639 (c-point 'bonl)))
11640 (goto-char containing-sexp) ; paren opening the arglist
11641 (setq placeholder (c-point 'boi))
11642 (if (and (c-safe (backward-up-list 1) t)
11643 (>= (point) placeholder))
11644 (progn
11645 (forward-char)
11646 (skip-chars-forward " \t"))
11647 (goto-char placeholder))
11648 (c-add-stmt-syntax 'arglist-cont-nonempty (list containing-sexp) t
11649 (c-most-enclosing-brace c-state-cache (point))
11650 paren-state))
11651
11652 ;; CASE 7G: we are looking at just a normal arglist
11653 ;; continuation line
11654 (t (c-forward-syntactic-ws indent-point)
11655 (c-add-syntax 'arglist-cont (c-point 'boi)))
11656 ))
11657
11658 ;; CASE 8: func-local multi-inheritance line
11659 ((and (c-major-mode-is 'c++-mode)
11660 (save-excursion
11661 (goto-char indent-point)
11662 (skip-chars-forward " \t")
11663 (looking-at c-opt-postfix-decl-spec-key)))
11664 (goto-char indent-point)
11665 (skip-chars-forward " \t")
11666 (cond
11667
11668 ;; CASE 8A: non-hanging colon on an inher intro
11669 ((eq char-after-ip ?:)
11670 (c-backward-syntactic-ws lim)
11671 (c-add-syntax 'inher-intro (c-point 'boi)))
11672
11673 ;; CASE 8B: hanging colon on an inher intro
11674 ((eq char-before-ip ?:)
11675 (c-add-syntax 'inher-intro (c-point 'boi)))
11676
11677 ;; CASE 8C: a continued inheritance line
11678 (t
11679 (c-beginning-of-inheritance-list lim)
11680 (c-add-syntax 'inher-cont (point))
11681 )))
11682
11683 ;; CASE 9: we are inside a brace-list
11684 ((and (not (c-major-mode-is 'awk-mode)) ; Maybe this isn't needed (ACM, 2002/3/29)
11685 (setq special-brace-list
11686 (or (and c-special-brace-lists ;;;; ALWAYS NIL FOR AWK!!
11687 (save-excursion
11688 (goto-char containing-sexp)
11689 (c-looking-at-special-brace-list)))
11690 (c-inside-bracelist-p containing-sexp paren-state))))
11691 (cond
11692
11693 ;; CASE 9A: In the middle of a special brace list opener.
11694 ((and (consp special-brace-list)
11695 (save-excursion
11696 (goto-char containing-sexp)
11697 (eq (char-after) ?\())
11698 (eq char-after-ip (car (cdr special-brace-list))))
11699 (goto-char (car (car special-brace-list)))
11700 (skip-chars-backward " \t")
11701 (if (and (bolp)
11702 (assoc 'statement-cont
11703 (setq placeholder (c-guess-basic-syntax))))
11704 (setq c-syntactic-context placeholder)
11705 (c-beginning-of-statement-1
11706 (c-safe-position (1- containing-sexp) paren-state))
11707 (c-forward-token-2 0)
11708 (while (cond
11709 ((looking-at c-specifier-key)
11710 (c-forward-keyword-clause 1))
11711 ((and c-opt-cpp-prefix
11712 (looking-at c-noise-macro-with-parens-name-re))
11713 (c-forward-noise-clause))))
11714 (c-add-syntax 'brace-list-open (c-point 'boi))))
11715
11716 ;; CASE 9B: brace-list-close brace
11717 ((if (consp special-brace-list)
11718 ;; Check special brace list closer.
11719 (progn
11720 (goto-char (car (car special-brace-list)))
11721 (save-excursion
11722 (goto-char indent-point)
11723 (back-to-indentation)
11724 (or
11725 ;; We were between the special close char and the `)'.
11726 (and (eq (char-after) ?\))
11727 (eq (1+ (point)) (cdr (car special-brace-list))))
11728 ;; We were before the special close char.
11729 (and (eq (char-after) (cdr (cdr special-brace-list)))
11730 (zerop (c-forward-token-2))
11731 (eq (1+ (point)) (cdr (car special-brace-list)))))))
11732 ;; Normal brace list check.
11733 (and (eq char-after-ip ?})
11734 (c-safe (goto-char (c-up-list-backward (point))) t)
11735 (= (point) containing-sexp)))
11736 (if (eq (point) (c-point 'boi))
11737 (c-add-syntax 'brace-list-close (point))
11738 (setq lim (c-most-enclosing-brace c-state-cache (point)))
11739 (c-beginning-of-statement-1 lim nil nil t)
11740 (c-add-stmt-syntax 'brace-list-close nil t lim paren-state)))
11741
11742 (t
11743 ;; Prepare for the rest of the cases below by going to the
11744 ;; token following the opening brace
11745 (if (consp special-brace-list)
11746 (progn
11747 (goto-char (car (car special-brace-list)))
11748 (c-forward-token-2 1 nil indent-point))
11749 (goto-char containing-sexp))
11750 (forward-char)
11751 (let ((start (point)))
11752 (c-forward-syntactic-ws indent-point)
11753 (goto-char (max start (c-point 'bol))))
11754 (c-skip-ws-forward indent-point)
11755 (cond
11756
11757 ;; CASE 9C: we're looking at the first line in a brace-list
11758 ((= (point) indent-point)
11759 (if (consp special-brace-list)
11760 (goto-char (car (car special-brace-list)))
11761 (goto-char containing-sexp))
11762 (if (eq (point) (c-point 'boi))
11763 (c-add-syntax 'brace-list-intro (point))
11764 (setq lim (c-most-enclosing-brace c-state-cache (point)))
11765 (c-beginning-of-statement-1 lim)
11766 (c-add-stmt-syntax 'brace-list-intro nil t lim paren-state)))
11767
11768 ;; CASE 9D: this is just a later brace-list-entry or
11769 ;; brace-entry-open
11770 (t (if (or (eq char-after-ip ?{)
11771 (and c-special-brace-lists
11772 (save-excursion
11773 (goto-char indent-point)
11774 (c-forward-syntactic-ws (c-point 'eol))
11775 (c-looking-at-special-brace-list (point)))))
11776 (c-add-syntax 'brace-entry-open (point))
11777 (c-add-syntax 'brace-list-entry (point))
11778 ))
11779 ))))
11780
11781 ;; CASE 10: A continued statement or top level construct.
11782 ((and (not (memq char-before-ip '(?\; ?:)))
11783 (not (c-at-vsemi-p before-ws-ip))
11784 (or (not (eq char-before-ip ?}))
11785 (c-looking-at-inexpr-block-backward c-state-cache))
11786 (> (point)
11787 (save-excursion
11788 (c-beginning-of-statement-1 containing-sexp)
11789 (setq placeholder (point))))
11790 (/= placeholder containing-sexp))
11791 ;; This is shared with case 18.
11792 (c-guess-continued-construct indent-point
11793 char-after-ip
11794 placeholder
11795 containing-sexp
11796 paren-state))
11797
11798 ;; CASE 16: block close brace, possibly closing the defun or
11799 ;; the class
11800 ((eq char-after-ip ?})
11801 ;; From here on we have the next containing sexp in lim.
11802 (setq lim (c-most-enclosing-brace paren-state))
11803 (goto-char containing-sexp)
11804 (cond
11805
11806 ;; CASE 16E: Closing a statement block? This catches
11807 ;; cases where it's preceded by a statement keyword,
11808 ;; which works even when used in an "invalid" context,
11809 ;; e.g. a macro argument.
11810 ((c-after-conditional)
11811 (c-backward-to-block-anchor lim)
11812 (c-add-stmt-syntax 'block-close nil t lim paren-state))
11813
11814 ;; CASE 16A: closing a lambda defun or an in-expression
11815 ;; block? C.f. cases 4, 7B and 17E.
11816 ((setq placeholder (c-looking-at-inexpr-block
11817 (c-safe-position containing-sexp paren-state)
11818 nil))
11819 (setq tmpsymbol (if (eq (car placeholder) 'inlambda)
11820 'inline-close
11821 'block-close))
11822 (goto-char containing-sexp)
11823 (back-to-indentation)
11824 (if (= containing-sexp (point))
11825 (c-add-syntax tmpsymbol (point))
11826 (goto-char (cdr placeholder))
11827 (back-to-indentation)
11828 (c-add-stmt-syntax tmpsymbol nil t
11829 (c-most-enclosing-brace paren-state (point))
11830 paren-state)
11831 (if (/= (point) (cdr placeholder))
11832 (c-add-syntax (car placeholder)))))
11833
11834 ;; CASE 16B: does this close an inline or a function in
11835 ;; a non-class declaration level block?
11836 ((save-excursion
11837 (and lim
11838 (progn
11839 (goto-char lim)
11840 (c-looking-at-decl-block
11841 (c-most-enclosing-brace paren-state lim)
11842 nil))
11843 (setq placeholder (point))))
11844 (c-backward-to-decl-anchor lim)
11845 (back-to-indentation)
11846 (if (save-excursion
11847 (goto-char placeholder)
11848 (looking-at c-other-decl-block-key))
11849 (c-add-syntax 'defun-close (point))
11850 (c-add-syntax 'inline-close (point))))
11851
11852 ;; CASE 16F: Can be a defun-close of a function declared
11853 ;; in a statement block, e.g. in Pike or when using gcc
11854 ;; extensions, but watch out for macros followed by
11855 ;; blocks. Let it through to be handled below.
11856 ;; C.f. cases B.3 and 17G.
11857 ((save-excursion
11858 (and (not (c-at-statement-start-p))
11859 (eq (c-beginning-of-statement-1 lim nil nil t) 'same)
11860 (setq placeholder (point))
11861 (let ((c-recognize-typeless-decls nil))
11862 ;; Turn off recognition of constructs that
11863 ;; lacks a type in this case, since that's more
11864 ;; likely to be a macro followed by a block.
11865 (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil))))
11866 (back-to-indentation)
11867 (if (/= (point) containing-sexp)
11868 (goto-char placeholder))
11869 (c-add-stmt-syntax 'defun-close nil t lim paren-state))
11870
11871 ;; CASE 16C: If there is an enclosing brace then this is
11872 ;; a block close since defun closes inside declaration
11873 ;; level blocks have been handled above.
11874 (lim
11875 ;; If the block is preceded by a case/switch label on
11876 ;; the same line, we anchor at the first preceding label
11877 ;; at boi. The default handling in c-add-stmt-syntax
11878 ;; really fixes it better, but we do like this to keep
11879 ;; the indentation compatible with version 5.28 and
11880 ;; earlier. C.f. case 17H.
11881 (while (and (/= (setq placeholder (point)) (c-point 'boi))
11882 (eq (c-beginning-of-statement-1 lim) 'label)))
11883 (goto-char placeholder)
11884 (if (looking-at c-label-kwds-regexp)
11885 (c-add-syntax 'block-close (point))
11886 (goto-char containing-sexp)
11887 ;; c-backward-to-block-anchor not necessary here; those
11888 ;; situations are handled in case 16E above.
11889 (c-add-stmt-syntax 'block-close nil t lim paren-state)))
11890
11891 ;; CASE 16D: Only top level defun close left.
11892 (t
11893 (goto-char containing-sexp)
11894 (c-backward-to-decl-anchor lim)
11895 (c-add-stmt-syntax 'defun-close nil nil
11896 (c-most-enclosing-brace paren-state)
11897 paren-state))
11898 ))
11899
11900 ;; CASE 19: line is an expression, not a statement, and is directly
11901 ;; contained by a template delimiter. Most likely, we are in a
11902 ;; template arglist within a statement. This case is based on CASE
11903 ;; 7. At some point in the future, we may wish to create more
11904 ;; syntactic symbols such as `template-intro',
11905 ;; `template-cont-nonempty', etc., and distinguish between them as we
11906 ;; do for `arglist-intro' etc. (2009-12-07).
11907 ((and c-recognize-<>-arglists
11908 (setq containing-< (c-up-list-backward indent-point containing-sexp))
11909 (eq (char-after containing-<) ?\<))
11910 (setq placeholder (c-point 'boi containing-<))
11911 (goto-char containing-sexp) ; Most nested Lbrace/Lparen (but not
11912 ; '<') before indent-point.
11913 (if (>= (point) placeholder)
11914 (progn
11915 (forward-char)
11916 (skip-chars-forward " \t"))
11917 (goto-char placeholder))
11918 (c-add-stmt-syntax 'template-args-cont (list containing-<) t
11919 (c-most-enclosing-brace c-state-cache (point))
11920 paren-state))
11921
11922 ;; CASE 17: Statement or defun catchall.
11923 (t
11924 (goto-char indent-point)
11925 ;; Back up statements until we find one that starts at boi.
11926 (while (let* ((prev-point (point))
11927 (last-step-type (c-beginning-of-statement-1
11928 containing-sexp)))
11929 (if (= (point) prev-point)
11930 (progn
11931 (setq step-type (or step-type last-step-type))
11932 nil)
11933 (setq step-type last-step-type)
11934 (/= (point) (c-point 'boi)))))
11935 (cond
11936
11937 ;; CASE 17B: continued statement
11938 ((and (eq step-type 'same)
11939 (/= (point) indent-point))
11940 (c-add-stmt-syntax 'statement-cont nil nil
11941 containing-sexp paren-state))
11942
11943 ;; CASE 17A: After a case/default label?
11944 ((progn
11945 (while (and (eq step-type 'label)
11946 (not (looking-at c-label-kwds-regexp)))
11947 (setq step-type
11948 (c-beginning-of-statement-1 containing-sexp)))
11949 (eq step-type 'label))
11950 (c-add-stmt-syntax (if (eq char-after-ip ?{)
11951 'statement-case-open
11952 'statement-case-intro)
11953 nil t containing-sexp paren-state))
11954
11955 ;; CASE 17D: any old statement
11956 ((progn
11957 (while (eq step-type 'label)
11958 (setq step-type
11959 (c-beginning-of-statement-1 containing-sexp)))
11960 (eq step-type 'previous))
11961 (c-add-stmt-syntax 'statement nil t
11962 containing-sexp paren-state)
11963 (if (eq char-after-ip ?{)
11964 (c-add-syntax 'block-open)))
11965
11966 ;; CASE 17I: Inside a substatement block.
11967 ((progn
11968 ;; The following tests are all based on containing-sexp.
11969 (goto-char containing-sexp)
11970 ;; From here on we have the next containing sexp in lim.
11971 (setq lim (c-most-enclosing-brace paren-state containing-sexp))
11972 (c-after-conditional))
11973 (c-backward-to-block-anchor lim)
11974 (c-add-stmt-syntax 'statement-block-intro nil t
11975 lim paren-state)
11976 (if (eq char-after-ip ?{)
11977 (c-add-syntax 'block-open)))
11978
11979 ;; CASE 17E: first statement in an in-expression block.
11980 ;; C.f. cases 4, 7B and 16A.
11981 ((setq placeholder (c-looking-at-inexpr-block
11982 (c-safe-position containing-sexp paren-state)
11983 nil))
11984 (setq tmpsymbol (if (eq (car placeholder) 'inlambda)
11985 'defun-block-intro
11986 'statement-block-intro))
11987 (back-to-indentation)
11988 (if (= containing-sexp (point))
11989 (c-add-syntax tmpsymbol (point))
11990 (goto-char (cdr placeholder))
11991 (back-to-indentation)
11992 (c-add-stmt-syntax tmpsymbol nil t
11993 (c-most-enclosing-brace c-state-cache (point))
11994 paren-state)
11995 (if (/= (point) (cdr placeholder))
11996 (c-add-syntax (car placeholder))))
11997 (if (eq char-after-ip ?{)
11998 (c-add-syntax 'block-open)))
11999
12000 ;; CASE 17F: first statement in an inline, or first
12001 ;; statement in a top-level defun. we can tell this is it
12002 ;; if there are no enclosing braces that haven't been
12003 ;; narrowed out by a class (i.e. don't use bod here).
12004 ((save-excursion
12005 (or (not (setq placeholder (c-most-enclosing-brace
12006 paren-state)))
12007 (and (progn
12008 (goto-char placeholder)
12009 (eq (char-after) ?{))
12010 (c-looking-at-decl-block (c-most-enclosing-brace
12011 paren-state (point))
12012 nil))))
12013 (c-backward-to-decl-anchor lim)
12014 (back-to-indentation)
12015 (c-add-syntax 'defun-block-intro (point)))
12016
12017 ;; CASE 17G: First statement in a function declared inside
12018 ;; a normal block. This can occur in Pike and with
12019 ;; e.g. the gcc extensions, but watch out for macros
12020 ;; followed by blocks. C.f. cases B.3 and 16F.
12021 ((save-excursion
12022 (and (not (c-at-statement-start-p))
12023 (eq (c-beginning-of-statement-1 lim nil nil t) 'same)
12024 (setq placeholder (point))
12025 (let ((c-recognize-typeless-decls nil))
12026 ;; Turn off recognition of constructs that lacks
12027 ;; a type in this case, since that's more likely
12028 ;; to be a macro followed by a block.
12029 (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil))))
12030 (back-to-indentation)
12031 (if (/= (point) containing-sexp)
12032 (goto-char placeholder))
12033 (c-add-stmt-syntax 'defun-block-intro nil t
12034 lim paren-state))
12035
12036 ;; CASE 17H: First statement in a block.
12037 (t
12038 ;; If the block is preceded by a case/switch label on the
12039 ;; same line, we anchor at the first preceding label at
12040 ;; boi. The default handling in c-add-stmt-syntax is
12041 ;; really fixes it better, but we do like this to keep the
12042 ;; indentation compatible with version 5.28 and earlier.
12043 ;; C.f. case 16C.
12044 (while (and (/= (setq placeholder (point)) (c-point 'boi))
12045 (eq (c-beginning-of-statement-1 lim) 'label)))
12046 (goto-char placeholder)
12047 (if (looking-at c-label-kwds-regexp)
12048 (c-add-syntax 'statement-block-intro (point))
12049 (goto-char containing-sexp)
12050 ;; c-backward-to-block-anchor not necessary here; those
12051 ;; situations are handled in case 17I above.
12052 (c-add-stmt-syntax 'statement-block-intro nil t
12053 lim paren-state))
12054 (if (eq char-after-ip ?{)
12055 (c-add-syntax 'block-open)))
12056 ))
12057 )
12058
12059 ;; now we need to look at any modifiers
12060 (goto-char indent-point)
12061 (skip-chars-forward " \t")
12062
12063 ;; are we looking at a comment only line?
12064 (when (and (looking-at c-comment-start-regexp)
12065 (/= (c-forward-token-2 0 nil (c-point 'eol)) 0))
12066 (c-append-syntax 'comment-intro))
12067
12068 ;; we might want to give additional offset to friends (in C++).
12069 (when (and c-opt-friend-key
12070 (looking-at c-opt-friend-key))
12071 (c-append-syntax 'friend))
12072
12073 ;; Set syntactic-relpos.
12074 (let ((p c-syntactic-context))
12075 (while (and p
12076 (if (integerp (c-langelem-pos (car p)))
12077 (progn
12078 (setq syntactic-relpos (c-langelem-pos (car p)))
12079 nil)
12080 t))
12081 (setq p (cdr p))))
12082
12083 ;; Start of or a continuation of a preprocessor directive?
12084 (if (and macro-start
12085 (eq macro-start (c-point 'boi))
12086 (not (and (c-major-mode-is 'pike-mode)
12087 (eq (char-after (1+ macro-start)) ?\"))))
12088 (c-append-syntax 'cpp-macro)
12089 (when (and c-syntactic-indentation-in-macros macro-start)
12090 (if in-macro-expr
12091 (when (or
12092 (< syntactic-relpos macro-start)
12093 (not (or
12094 (assq 'arglist-intro c-syntactic-context)
12095 (assq 'arglist-cont c-syntactic-context)
12096 (assq 'arglist-cont-nonempty c-syntactic-context)
12097 (assq 'arglist-close c-syntactic-context))))
12098 ;; If inside a cpp expression, i.e. anywhere in a
12099 ;; cpp directive except a #define body, we only let
12100 ;; through the syntactic analysis that is internal
12101 ;; in the expression. That means the arglist
12102 ;; elements, if they are anchored inside the cpp
12103 ;; expression.
12104 (setq c-syntactic-context nil)
12105 (c-add-syntax 'cpp-macro-cont macro-start))
12106 (when (and (eq macro-start syntactic-relpos)
12107 (not (assq 'cpp-define-intro c-syntactic-context))
12108 (save-excursion
12109 (goto-char macro-start)
12110 (or (not (c-forward-to-cpp-define-body))
12111 (<= (point) (c-point 'boi indent-point)))))
12112 ;; Inside a #define body and the syntactic analysis is
12113 ;; anchored on the start of the #define. In this case
12114 ;; we add cpp-define-intro to get the extra
12115 ;; indentation of the #define body.
12116 (c-add-syntax 'cpp-define-intro)))))
12117
12118 ;; return the syntax
12119 c-syntactic-context)))
12120
12121 \f
12122 ;; Indentation calculation.
12123
12124 (defun c-evaluate-offset (offset langelem symbol)
12125 ;; offset can be a number, a function, a variable, a list, or one of
12126 ;; the symbols + or -
12127 ;;
12128 ;; This function might do hidden buffer changes.
12129 (let ((res
12130 (cond
12131 ((numberp offset) offset)
12132 ((vectorp offset) offset)
12133 ((null offset) nil)
12134
12135 ((eq offset '+) c-basic-offset)
12136 ((eq offset '-) (- c-basic-offset))
12137 ((eq offset '++) (* 2 c-basic-offset))
12138 ((eq offset '--) (* 2 (- c-basic-offset)))
12139 ((eq offset '*) (/ c-basic-offset 2))
12140 ((eq offset '/) (/ (- c-basic-offset) 2))
12141
12142 ((functionp offset)
12143 (c-evaluate-offset
12144 (funcall offset
12145 (cons (c-langelem-sym langelem)
12146 (c-langelem-pos langelem)))
12147 langelem symbol))
12148
12149 ((listp offset)
12150 (cond
12151 ((eq (car offset) 'quote)
12152 (c-benign-error "The offset %S for %s was mistakenly quoted"
12153 offset symbol)
12154 nil)
12155
12156 ((memq (car offset) '(min max))
12157 (let (res val (method (car offset)))
12158 (setq offset (cdr offset))
12159 (while offset
12160 (setq val (c-evaluate-offset (car offset) langelem symbol))
12161 (cond
12162 ((not val))
12163 ((not res)
12164 (setq res val))
12165 ((integerp val)
12166 (if (vectorp res)
12167 (c-benign-error "\
12168 Error evaluating offset %S for %s: \
12169 Cannot combine absolute offset %S with relative %S in `%s' method"
12170 (car offset) symbol res val method)
12171 (setq res (funcall method res val))))
12172 (t
12173 (if (integerp res)
12174 (c-benign-error "\
12175 Error evaluating offset %S for %s: \
12176 Cannot combine relative offset %S with absolute %S in `%s' method"
12177 (car offset) symbol res val method)
12178 (setq res (vector (funcall method (aref res 0)
12179 (aref val 0)))))))
12180 (setq offset (cdr offset)))
12181 res))
12182
12183 ((eq (car offset) 'add)
12184 (let (res val)
12185 (setq offset (cdr offset))
12186 (while offset
12187 (setq val (c-evaluate-offset (car offset) langelem symbol))
12188 (cond
12189 ((not val))
12190 ((not res)
12191 (setq res val))
12192 ((integerp val)
12193 (if (vectorp res)
12194 (setq res (vector (+ (aref res 0) val)))
12195 (setq res (+ res val))))
12196 (t
12197 (if (vectorp res)
12198 (c-benign-error "\
12199 Error evaluating offset %S for %s: \
12200 Cannot combine absolute offsets %S and %S in `add' method"
12201 (car offset) symbol res val)
12202 (setq res val)))) ; Override.
12203 (setq offset (cdr offset)))
12204 res))
12205
12206 (t
12207 (let (res)
12208 (when (eq (car offset) 'first)
12209 (setq offset (cdr offset)))
12210 (while (and (not res) offset)
12211 (setq res (c-evaluate-offset (car offset) langelem symbol)
12212 offset (cdr offset)))
12213 res))))
12214
12215 ((and (symbolp offset) (boundp offset))
12216 (symbol-value offset))
12217
12218 (t
12219 (c-benign-error "Unknown offset format %S for %s" offset symbol)
12220 nil))))
12221
12222 (if (or (null res) (integerp res)
12223 (and (vectorp res) (= (length res) 1) (integerp (aref res 0))))
12224 res
12225 (c-benign-error "Error evaluating offset %S for %s: Got invalid value %S"
12226 offset symbol res)
12227 nil)))
12228
12229 (defun c-calc-offset (langelem)
12230 ;; Get offset from LANGELEM which is a list beginning with the
12231 ;; syntactic symbol and followed by any analysis data it provides.
12232 ;; That data may be zero or more elements, but if at least one is
12233 ;; given then the first is the anchor position (or nil). The symbol
12234 ;; is matched against `c-offsets-alist' and the offset calculated
12235 ;; from that is returned.
12236 ;;
12237 ;; This function might do hidden buffer changes.
12238 (let* ((symbol (c-langelem-sym langelem))
12239 (match (assq symbol c-offsets-alist))
12240 (offset (cdr-safe match)))
12241 (if match
12242 (setq offset (c-evaluate-offset offset langelem symbol))
12243 (if c-strict-syntax-p
12244 (c-benign-error "No offset found for syntactic symbol %s" symbol))
12245 (setq offset 0))
12246 (if (vectorp offset)
12247 offset
12248 (or (and (numberp offset) offset)
12249 (and (symbolp offset) (symbol-value offset))
12250 0))
12251 ))
12252
12253 (defun c-get-offset (langelem)
12254 ;; This is a compatibility wrapper for `c-calc-offset' in case
12255 ;; someone is calling it directly. It takes an old style syntactic
12256 ;; element on the form (SYMBOL . ANCHOR-POS) and converts it to the
12257 ;; new list form.
12258 ;;
12259 ;; This function might do hidden buffer changes.
12260 (if (c-langelem-pos langelem)
12261 (c-calc-offset (list (c-langelem-sym langelem)
12262 (c-langelem-pos langelem)))
12263 (c-calc-offset langelem)))
12264
12265 (defun c-get-syntactic-indentation (langelems)
12266 ;; Calculate the syntactic indentation from a syntactic description
12267 ;; as returned by `c-guess-syntax'.
12268 ;;
12269 ;; Note that topmost-intro always has an anchor position at bol, for
12270 ;; historical reasons. It's often used together with other symbols
12271 ;; that has more sane positions. Since we always use the first
12272 ;; found anchor position, we rely on that these other symbols always
12273 ;; precede topmost-intro in the LANGELEMS list.
12274 ;;
12275 ;; This function might do hidden buffer changes.
12276 (let ((indent 0) anchor)
12277
12278 (while langelems
12279 (let* ((c-syntactic-element (car langelems))
12280 (res (c-calc-offset c-syntactic-element)))
12281
12282 (if (vectorp res)
12283 ;; Got an absolute column that overrides any indentation
12284 ;; we've collected so far, but not the relative
12285 ;; indentation we might get for the nested structures
12286 ;; further down the langelems list.
12287 (setq indent (elt res 0)
12288 anchor (point-min)) ; A position at column 0.
12289
12290 ;; Got a relative change of the current calculated
12291 ;; indentation.
12292 (setq indent (+ indent res))
12293
12294 ;; Use the anchor position from the first syntactic
12295 ;; element with one.
12296 (unless anchor
12297 (setq anchor (c-langelem-pos (car langelems)))))
12298
12299 (setq langelems (cdr langelems))))
12300
12301 (if anchor
12302 (+ indent (save-excursion
12303 (goto-char anchor)
12304 (current-column)))
12305 indent)))
12306
12307 \f
12308 (cc-provide 'cc-engine)
12309
12310 ;; Local Variables:
12311 ;; indent-tabs-mode: t
12312 ;; tab-width: 8
12313 ;; End:
12314 ;;; cc-engine.el ends here