]> code.delx.au - gnu-emacs/blob - src/lread.c
Merge branch 'master' of git.sv.gnu.org:/srv/git/emacs
[gnu-emacs] / src / lread.c
1 /* Lisp parsing and input streams.
2
3 Copyright (C) 1985-1989, 1993-1995, 1997-2015 Free Software Foundation,
4 Inc.
5
6 This file is part of GNU Emacs.
7
8 GNU Emacs is free software: you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation, either version 3 of the License, or
11 (at your option) any later version.
12
13 GNU Emacs is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
17
18 You should have received a copy of the GNU General Public License
19 along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */
20
21 /* Tell globals.h to define tables needed by init_obarray. */
22 #define DEFINE_SYMBOLS
23
24 #include <config.h>
25 #include "sysstdio.h"
26 #include <sys/types.h>
27 #include <sys/stat.h>
28 #include <sys/file.h>
29 #include <errno.h>
30 #include <limits.h> /* For CHAR_BIT. */
31 #include <math.h>
32 #include <stat-time.h>
33 #include "lisp.h"
34 #include "intervals.h"
35 #include "character.h"
36 #include "buffer.h"
37 #include "charset.h"
38 #include "coding.h"
39 #include <epaths.h>
40 #include "commands.h"
41 #include "keyboard.h"
42 #include "frame.h"
43 #include "termhooks.h"
44 #include "blockinput.h"
45
46 #ifdef MSDOS
47 #include "msdos.h"
48 #endif
49
50 #ifdef HAVE_NS
51 #include "nsterm.h"
52 #endif
53
54 #include <unistd.h>
55
56 #ifdef HAVE_SETLOCALE
57 #include <locale.h>
58 #endif /* HAVE_SETLOCALE */
59
60 #include <fcntl.h>
61
62 #ifdef HAVE_FSEEKO
63 #define file_offset off_t
64 #define file_tell ftello
65 #else
66 #define file_offset long
67 #define file_tell ftell
68 #endif
69
70 /* The association list of objects read with the #n=object form.
71 Each member of the list has the form (n . object), and is used to
72 look up the object for the corresponding #n# construct.
73 It must be set to nil before all top-level calls to read0. */
74 static Lisp_Object read_objects;
75
76 /* File for get_file_char to read from. Use by load. */
77 static FILE *instream;
78
79 /* For use within read-from-string (this reader is non-reentrant!!) */
80 static ptrdiff_t read_from_string_index;
81 static ptrdiff_t read_from_string_index_byte;
82 static ptrdiff_t read_from_string_limit;
83
84 /* Number of characters read in the current call to Fread or
85 Fread_from_string. */
86 static EMACS_INT readchar_count;
87
88 /* This contains the last string skipped with #@. */
89 static char *saved_doc_string;
90 /* Length of buffer allocated in saved_doc_string. */
91 static ptrdiff_t saved_doc_string_size;
92 /* Length of actual data in saved_doc_string. */
93 static ptrdiff_t saved_doc_string_length;
94 /* This is the file position that string came from. */
95 static file_offset saved_doc_string_position;
96
97 /* This contains the previous string skipped with #@.
98 We copy it from saved_doc_string when a new string
99 is put in saved_doc_string. */
100 static char *prev_saved_doc_string;
101 /* Length of buffer allocated in prev_saved_doc_string. */
102 static ptrdiff_t prev_saved_doc_string_size;
103 /* Length of actual data in prev_saved_doc_string. */
104 static ptrdiff_t prev_saved_doc_string_length;
105 /* This is the file position that string came from. */
106 static file_offset prev_saved_doc_string_position;
107
108 /* True means inside a new-style backquote
109 with no surrounding parentheses.
110 Fread initializes this to false, so we need not specbind it
111 or worry about what happens to it when there is an error. */
112 static bool new_backquote_flag;
113
114 /* A list of file names for files being loaded in Fload. Used to
115 check for recursive loads. */
116
117 static Lisp_Object Vloads_in_progress;
118
119 static int read_emacs_mule_char (int, int (*) (int, Lisp_Object),
120 Lisp_Object);
121
122 static void readevalloop (Lisp_Object, FILE *, Lisp_Object, bool,
123 Lisp_Object, Lisp_Object,
124 Lisp_Object, Lisp_Object);
125 \f
126 /* Functions that read one byte from the current source READCHARFUN
127 or unreads one byte. If the integer argument C is -1, it returns
128 one read byte, or -1 when there's no more byte in the source. If C
129 is 0 or positive, it unreads C, and the return value is not
130 interesting. */
131
132 static int readbyte_for_lambda (int, Lisp_Object);
133 static int readbyte_from_file (int, Lisp_Object);
134 static int readbyte_from_string (int, Lisp_Object);
135
136 /* Handle unreading and rereading of characters.
137 Write READCHAR to read a character,
138 UNREAD(c) to unread c to be read again.
139
140 These macros correctly read/unread multibyte characters. */
141
142 #define READCHAR readchar (readcharfun, NULL)
143 #define UNREAD(c) unreadchar (readcharfun, c)
144
145 /* Same as READCHAR but set *MULTIBYTE to the multibyteness of the source. */
146 #define READCHAR_REPORT_MULTIBYTE(multibyte) readchar (readcharfun, multibyte)
147
148 /* When READCHARFUN is Qget_file_char, Qget_emacs_mule_file_char,
149 Qlambda, or a cons, we use this to keep an unread character because
150 a file stream can't handle multibyte-char unreading. The value -1
151 means that there's no unread character. */
152 static int unread_char;
153
154 static int
155 readchar (Lisp_Object readcharfun, bool *multibyte)
156 {
157 Lisp_Object tem;
158 register int c;
159 int (*readbyte) (int, Lisp_Object);
160 unsigned char buf[MAX_MULTIBYTE_LENGTH];
161 int i, len;
162 bool emacs_mule_encoding = 0;
163
164 if (multibyte)
165 *multibyte = 0;
166
167 readchar_count++;
168
169 if (BUFFERP (readcharfun))
170 {
171 register struct buffer *inbuffer = XBUFFER (readcharfun);
172
173 ptrdiff_t pt_byte = BUF_PT_BYTE (inbuffer);
174
175 if (! BUFFER_LIVE_P (inbuffer))
176 return -1;
177
178 if (pt_byte >= BUF_ZV_BYTE (inbuffer))
179 return -1;
180
181 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
182 {
183 /* Fetch the character code from the buffer. */
184 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, pt_byte);
185 BUF_INC_POS (inbuffer, pt_byte);
186 c = STRING_CHAR (p);
187 if (multibyte)
188 *multibyte = 1;
189 }
190 else
191 {
192 c = BUF_FETCH_BYTE (inbuffer, pt_byte);
193 if (! ASCII_CHAR_P (c))
194 c = BYTE8_TO_CHAR (c);
195 pt_byte++;
196 }
197 SET_BUF_PT_BOTH (inbuffer, BUF_PT (inbuffer) + 1, pt_byte);
198
199 return c;
200 }
201 if (MARKERP (readcharfun))
202 {
203 register struct buffer *inbuffer = XMARKER (readcharfun)->buffer;
204
205 ptrdiff_t bytepos = marker_byte_position (readcharfun);
206
207 if (bytepos >= BUF_ZV_BYTE (inbuffer))
208 return -1;
209
210 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
211 {
212 /* Fetch the character code from the buffer. */
213 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, bytepos);
214 BUF_INC_POS (inbuffer, bytepos);
215 c = STRING_CHAR (p);
216 if (multibyte)
217 *multibyte = 1;
218 }
219 else
220 {
221 c = BUF_FETCH_BYTE (inbuffer, bytepos);
222 if (! ASCII_CHAR_P (c))
223 c = BYTE8_TO_CHAR (c);
224 bytepos++;
225 }
226
227 XMARKER (readcharfun)->bytepos = bytepos;
228 XMARKER (readcharfun)->charpos++;
229
230 return c;
231 }
232
233 if (EQ (readcharfun, Qlambda))
234 {
235 readbyte = readbyte_for_lambda;
236 goto read_multibyte;
237 }
238
239 if (EQ (readcharfun, Qget_file_char))
240 {
241 readbyte = readbyte_from_file;
242 goto read_multibyte;
243 }
244
245 if (STRINGP (readcharfun))
246 {
247 if (read_from_string_index >= read_from_string_limit)
248 c = -1;
249 else if (STRING_MULTIBYTE (readcharfun))
250 {
251 if (multibyte)
252 *multibyte = 1;
253 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c, readcharfun,
254 read_from_string_index,
255 read_from_string_index_byte);
256 }
257 else
258 {
259 c = SREF (readcharfun, read_from_string_index_byte);
260 read_from_string_index++;
261 read_from_string_index_byte++;
262 }
263 return c;
264 }
265
266 if (CONSP (readcharfun))
267 {
268 /* This is the case that read_vector is reading from a unibyte
269 string that contains a byte sequence previously skipped
270 because of #@NUMBER. The car part of readcharfun is that
271 string, and the cdr part is a value of readcharfun given to
272 read_vector. */
273 readbyte = readbyte_from_string;
274 if (EQ (XCDR (readcharfun), Qget_emacs_mule_file_char))
275 emacs_mule_encoding = 1;
276 goto read_multibyte;
277 }
278
279 if (EQ (readcharfun, Qget_emacs_mule_file_char))
280 {
281 readbyte = readbyte_from_file;
282 emacs_mule_encoding = 1;
283 goto read_multibyte;
284 }
285
286 tem = call0 (readcharfun);
287
288 if (NILP (tem))
289 return -1;
290 return XINT (tem);
291
292 read_multibyte:
293 if (unread_char >= 0)
294 {
295 c = unread_char;
296 unread_char = -1;
297 return c;
298 }
299 c = (*readbyte) (-1, readcharfun);
300 if (c < 0)
301 return c;
302 if (multibyte)
303 *multibyte = 1;
304 if (ASCII_CHAR_P (c))
305 return c;
306 if (emacs_mule_encoding)
307 return read_emacs_mule_char (c, readbyte, readcharfun);
308 i = 0;
309 buf[i++] = c;
310 len = BYTES_BY_CHAR_HEAD (c);
311 while (i < len)
312 {
313 c = (*readbyte) (-1, readcharfun);
314 if (c < 0 || ! TRAILING_CODE_P (c))
315 {
316 while (--i > 1)
317 (*readbyte) (buf[i], readcharfun);
318 return BYTE8_TO_CHAR (buf[0]);
319 }
320 buf[i++] = c;
321 }
322 return STRING_CHAR (buf);
323 }
324
325 #define FROM_FILE_P(readcharfun) \
326 (EQ (readcharfun, Qget_file_char) \
327 || EQ (readcharfun, Qget_emacs_mule_file_char))
328
329 static void
330 skip_dyn_bytes (Lisp_Object readcharfun, ptrdiff_t n)
331 {
332 if (FROM_FILE_P (readcharfun))
333 {
334 block_input (); /* FIXME: Not sure if it's needed. */
335 fseek (instream, n, SEEK_CUR);
336 unblock_input ();
337 }
338 else
339 { /* We're not reading directly from a file. In that case, it's difficult
340 to reliably count bytes, since these are usually meant for the file's
341 encoding, whereas we're now typically in the internal encoding.
342 But luckily, skip_dyn_bytes is used to skip over a single
343 dynamic-docstring (or dynamic byte-code) which is always quoted such
344 that \037 is the final char. */
345 int c;
346 do {
347 c = READCHAR;
348 } while (c >= 0 && c != '\037');
349 }
350 }
351
352 static void
353 skip_dyn_eof (Lisp_Object readcharfun)
354 {
355 if (FROM_FILE_P (readcharfun))
356 {
357 block_input (); /* FIXME: Not sure if it's needed. */
358 fseek (instream, 0, SEEK_END);
359 unblock_input ();
360 }
361 else
362 while (READCHAR >= 0);
363 }
364
365 /* Unread the character C in the way appropriate for the stream READCHARFUN.
366 If the stream is a user function, call it with the char as argument. */
367
368 static void
369 unreadchar (Lisp_Object readcharfun, int c)
370 {
371 readchar_count--;
372 if (c == -1)
373 /* Don't back up the pointer if we're unreading the end-of-input mark,
374 since readchar didn't advance it when we read it. */
375 ;
376 else if (BUFFERP (readcharfun))
377 {
378 struct buffer *b = XBUFFER (readcharfun);
379 ptrdiff_t charpos = BUF_PT (b);
380 ptrdiff_t bytepos = BUF_PT_BYTE (b);
381
382 if (! NILP (BVAR (b, enable_multibyte_characters)))
383 BUF_DEC_POS (b, bytepos);
384 else
385 bytepos--;
386
387 SET_BUF_PT_BOTH (b, charpos - 1, bytepos);
388 }
389 else if (MARKERP (readcharfun))
390 {
391 struct buffer *b = XMARKER (readcharfun)->buffer;
392 ptrdiff_t bytepos = XMARKER (readcharfun)->bytepos;
393
394 XMARKER (readcharfun)->charpos--;
395 if (! NILP (BVAR (b, enable_multibyte_characters)))
396 BUF_DEC_POS (b, bytepos);
397 else
398 bytepos--;
399
400 XMARKER (readcharfun)->bytepos = bytepos;
401 }
402 else if (STRINGP (readcharfun))
403 {
404 read_from_string_index--;
405 read_from_string_index_byte
406 = string_char_to_byte (readcharfun, read_from_string_index);
407 }
408 else if (CONSP (readcharfun))
409 {
410 unread_char = c;
411 }
412 else if (EQ (readcharfun, Qlambda))
413 {
414 unread_char = c;
415 }
416 else if (FROM_FILE_P (readcharfun))
417 {
418 unread_char = c;
419 }
420 else
421 call1 (readcharfun, make_number (c));
422 }
423
424 static int
425 readbyte_for_lambda (int c, Lisp_Object readcharfun)
426 {
427 return read_bytecode_char (c >= 0);
428 }
429
430
431 static int
432 readbyte_from_file (int c, Lisp_Object readcharfun)
433 {
434 if (c >= 0)
435 {
436 block_input ();
437 ungetc (c, instream);
438 unblock_input ();
439 return 0;
440 }
441
442 block_input ();
443 c = getc (instream);
444
445 /* Interrupted reads have been observed while reading over the network. */
446 while (c == EOF && ferror (instream) && errno == EINTR)
447 {
448 unblock_input ();
449 QUIT;
450 block_input ();
451 clearerr (instream);
452 c = getc (instream);
453 }
454
455 unblock_input ();
456
457 return (c == EOF ? -1 : c);
458 }
459
460 static int
461 readbyte_from_string (int c, Lisp_Object readcharfun)
462 {
463 Lisp_Object string = XCAR (readcharfun);
464
465 if (c >= 0)
466 {
467 read_from_string_index--;
468 read_from_string_index_byte
469 = string_char_to_byte (string, read_from_string_index);
470 }
471
472 if (read_from_string_index >= read_from_string_limit)
473 c = -1;
474 else
475 FETCH_STRING_CHAR_ADVANCE (c, string,
476 read_from_string_index,
477 read_from_string_index_byte);
478 return c;
479 }
480
481
482 /* Read one non-ASCII character from INSTREAM. The character is
483 encoded in `emacs-mule' and the first byte is already read in
484 C. */
485
486 static int
487 read_emacs_mule_char (int c, int (*readbyte) (int, Lisp_Object), Lisp_Object readcharfun)
488 {
489 /* Emacs-mule coding uses at most 4-byte for one character. */
490 unsigned char buf[4];
491 int len = emacs_mule_bytes[c];
492 struct charset *charset;
493 int i;
494 unsigned code;
495
496 if (len == 1)
497 /* C is not a valid leading-code of `emacs-mule'. */
498 return BYTE8_TO_CHAR (c);
499
500 i = 0;
501 buf[i++] = c;
502 while (i < len)
503 {
504 c = (*readbyte) (-1, readcharfun);
505 if (c < 0xA0)
506 {
507 while (--i > 1)
508 (*readbyte) (buf[i], readcharfun);
509 return BYTE8_TO_CHAR (buf[0]);
510 }
511 buf[i++] = c;
512 }
513
514 if (len == 2)
515 {
516 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
517 code = buf[1] & 0x7F;
518 }
519 else if (len == 3)
520 {
521 if (buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_11
522 || buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_12)
523 {
524 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
525 code = buf[2] & 0x7F;
526 }
527 else
528 {
529 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
530 code = ((buf[1] << 8) | buf[2]) & 0x7F7F;
531 }
532 }
533 else
534 {
535 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
536 code = ((buf[2] << 8) | buf[3]) & 0x7F7F;
537 }
538 c = DECODE_CHAR (charset, code);
539 if (c < 0)
540 Fsignal (Qinvalid_read_syntax,
541 list1 (build_string ("invalid multibyte form")));
542 return c;
543 }
544
545
546 static Lisp_Object read_internal_start (Lisp_Object, Lisp_Object,
547 Lisp_Object);
548 static Lisp_Object read0 (Lisp_Object);
549 static Lisp_Object read1 (Lisp_Object, int *, bool);
550
551 static Lisp_Object read_list (bool, Lisp_Object);
552 static Lisp_Object read_vector (Lisp_Object, bool);
553
554 static Lisp_Object substitute_object_recurse (Lisp_Object, Lisp_Object,
555 Lisp_Object);
556 static void substitute_object_in_subtree (Lisp_Object,
557 Lisp_Object);
558 static void substitute_in_interval (INTERVAL, Lisp_Object);
559
560 \f
561 /* Get a character from the tty. */
562
563 /* Read input events until we get one that's acceptable for our purposes.
564
565 If NO_SWITCH_FRAME, switch-frame events are stashed
566 until we get a character we like, and then stuffed into
567 unread_switch_frame.
568
569 If ASCII_REQUIRED, check function key events to see
570 if the unmodified version of the symbol has a Qascii_character
571 property, and use that character, if present.
572
573 If ERROR_NONASCII, signal an error if the input we
574 get isn't an ASCII character with modifiers. If it's false but
575 ASCII_REQUIRED is true, just re-read until we get an ASCII
576 character.
577
578 If INPUT_METHOD, invoke the current input method
579 if the character warrants that.
580
581 If SECONDS is a number, wait that many seconds for input, and
582 return Qnil if no input arrives within that time. */
583
584 static Lisp_Object
585 read_filtered_event (bool no_switch_frame, bool ascii_required,
586 bool error_nonascii, bool input_method, Lisp_Object seconds)
587 {
588 Lisp_Object val, delayed_switch_frame;
589 struct timespec end_time;
590
591 #ifdef HAVE_WINDOW_SYSTEM
592 if (display_hourglass_p)
593 cancel_hourglass ();
594 #endif
595
596 delayed_switch_frame = Qnil;
597
598 /* Compute timeout. */
599 if (NUMBERP (seconds))
600 {
601 double duration = extract_float (seconds);
602 struct timespec wait_time = dtotimespec (duration);
603 end_time = timespec_add (current_timespec (), wait_time);
604 }
605
606 /* Read until we get an acceptable event. */
607 retry:
608 do
609 val = read_char (0, Qnil, (input_method ? Qnil : Qt), 0,
610 NUMBERP (seconds) ? &end_time : NULL);
611 while (INTEGERP (val) && XINT (val) == -2); /* wrong_kboard_jmpbuf */
612
613 if (BUFFERP (val))
614 goto retry;
615
616 /* `switch-frame' events are put off until after the next ASCII
617 character. This is better than signaling an error just because
618 the last characters were typed to a separate minibuffer frame,
619 for example. Eventually, some code which can deal with
620 switch-frame events will read it and process it. */
621 if (no_switch_frame
622 && EVENT_HAS_PARAMETERS (val)
623 && EQ (EVENT_HEAD_KIND (EVENT_HEAD (val)), Qswitch_frame))
624 {
625 delayed_switch_frame = val;
626 goto retry;
627 }
628
629 if (ascii_required && !(NUMBERP (seconds) && NILP (val)))
630 {
631 /* Convert certain symbols to their ASCII equivalents. */
632 if (SYMBOLP (val))
633 {
634 Lisp_Object tem, tem1;
635 tem = Fget (val, Qevent_symbol_element_mask);
636 if (!NILP (tem))
637 {
638 tem1 = Fget (Fcar (tem), Qascii_character);
639 /* Merge this symbol's modifier bits
640 with the ASCII equivalent of its basic code. */
641 if (!NILP (tem1))
642 XSETFASTINT (val, XINT (tem1) | XINT (Fcar (Fcdr (tem))));
643 }
644 }
645
646 /* If we don't have a character now, deal with it appropriately. */
647 if (!INTEGERP (val))
648 {
649 if (error_nonascii)
650 {
651 Vunread_command_events = list1 (val);
652 error ("Non-character input-event");
653 }
654 else
655 goto retry;
656 }
657 }
658
659 if (! NILP (delayed_switch_frame))
660 unread_switch_frame = delayed_switch_frame;
661
662 #if 0
663
664 #ifdef HAVE_WINDOW_SYSTEM
665 if (display_hourglass_p)
666 start_hourglass ();
667 #endif
668
669 #endif
670
671 return val;
672 }
673
674 DEFUN ("read-char", Fread_char, Sread_char, 0, 3, 0,
675 doc: /* Read a character from the command input (keyboard or macro).
676 It is returned as a number.
677 If the character has modifiers, they are resolved and reflected to the
678 character code if possible (e.g. C-SPC -> 0).
679
680 If the user generates an event which is not a character (i.e. a mouse
681 click or function key event), `read-char' signals an error. As an
682 exception, switch-frame events are put off until non-character events
683 can be read.
684 If you want to read non-character events, or ignore them, call
685 `read-event' or `read-char-exclusive' instead.
686
687 If the optional argument PROMPT is non-nil, display that as a prompt.
688 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
689 input method is turned on in the current buffer, that input method
690 is used for reading a character.
691 If the optional argument SECONDS is non-nil, it should be a number
692 specifying the maximum number of seconds to wait for input. If no
693 input arrives in that time, return nil. SECONDS may be a
694 floating-point value. */)
695 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
696 {
697 Lisp_Object val;
698
699 if (! NILP (prompt))
700 message_with_string ("%s", prompt, 0);
701 val = read_filtered_event (1, 1, 1, ! NILP (inherit_input_method), seconds);
702
703 return (NILP (val) ? Qnil
704 : make_number (char_resolve_modifier_mask (XINT (val))));
705 }
706
707 DEFUN ("read-event", Fread_event, Sread_event, 0, 3, 0,
708 doc: /* Read an event object from the input stream.
709 If the optional argument PROMPT is non-nil, display that as a prompt.
710 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
711 input method is turned on in the current buffer, that input method
712 is used for reading a character.
713 If the optional argument SECONDS is non-nil, it should be a number
714 specifying the maximum number of seconds to wait for input. If no
715 input arrives in that time, return nil. SECONDS may be a
716 floating-point value. */)
717 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
718 {
719 if (! NILP (prompt))
720 message_with_string ("%s", prompt, 0);
721 return read_filtered_event (0, 0, 0, ! NILP (inherit_input_method), seconds);
722 }
723
724 DEFUN ("read-char-exclusive", Fread_char_exclusive, Sread_char_exclusive, 0, 3, 0,
725 doc: /* Read a character from the command input (keyboard or macro).
726 It is returned as a number. Non-character events are ignored.
727 If the character has modifiers, they are resolved and reflected to the
728 character code if possible (e.g. C-SPC -> 0).
729
730 If the optional argument PROMPT is non-nil, display that as a prompt.
731 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
732 input method is turned on in the current buffer, that input method
733 is used for reading a character.
734 If the optional argument SECONDS is non-nil, it should be a number
735 specifying the maximum number of seconds to wait for input. If no
736 input arrives in that time, return nil. SECONDS may be a
737 floating-point value. */)
738 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
739 {
740 Lisp_Object val;
741
742 if (! NILP (prompt))
743 message_with_string ("%s", prompt, 0);
744
745 val = read_filtered_event (1, 1, 0, ! NILP (inherit_input_method), seconds);
746
747 return (NILP (val) ? Qnil
748 : make_number (char_resolve_modifier_mask (XINT (val))));
749 }
750
751 DEFUN ("get-file-char", Fget_file_char, Sget_file_char, 0, 0, 0,
752 doc: /* Don't use this yourself. */)
753 (void)
754 {
755 register Lisp_Object val;
756 block_input ();
757 XSETINT (val, getc (instream));
758 unblock_input ();
759 return val;
760 }
761
762
763 \f
764
765 /* Return true if the lisp code read using READCHARFUN defines a non-nil
766 `lexical-binding' file variable. After returning, the stream is
767 positioned following the first line, if it is a comment or #! line,
768 otherwise nothing is read. */
769
770 static bool
771 lisp_file_lexically_bound_p (Lisp_Object readcharfun)
772 {
773 int ch = READCHAR;
774
775 if (ch == '#')
776 {
777 ch = READCHAR;
778 if (ch != '!')
779 {
780 UNREAD (ch);
781 UNREAD ('#');
782 return 0;
783 }
784 while (ch != '\n' && ch != EOF)
785 ch = READCHAR;
786 if (ch == '\n') ch = READCHAR;
787 /* It is OK to leave the position after a #! line, since
788 that is what read1 does. */
789 }
790
791 if (ch != ';')
792 /* The first line isn't a comment, just give up. */
793 {
794 UNREAD (ch);
795 return 0;
796 }
797 else
798 /* Look for an appropriate file-variable in the first line. */
799 {
800 bool rv = 0;
801 enum {
802 NOMINAL, AFTER_FIRST_DASH, AFTER_ASTERIX
803 } beg_end_state = NOMINAL;
804 bool in_file_vars = 0;
805
806 #define UPDATE_BEG_END_STATE(ch) \
807 if (beg_end_state == NOMINAL) \
808 beg_end_state = (ch == '-' ? AFTER_FIRST_DASH : NOMINAL); \
809 else if (beg_end_state == AFTER_FIRST_DASH) \
810 beg_end_state = (ch == '*' ? AFTER_ASTERIX : NOMINAL); \
811 else if (beg_end_state == AFTER_ASTERIX) \
812 { \
813 if (ch == '-') \
814 in_file_vars = !in_file_vars; \
815 beg_end_state = NOMINAL; \
816 }
817
818 /* Skip until we get to the file vars, if any. */
819 do
820 {
821 ch = READCHAR;
822 UPDATE_BEG_END_STATE (ch);
823 }
824 while (!in_file_vars && ch != '\n' && ch != EOF);
825
826 while (in_file_vars)
827 {
828 char var[100], val[100];
829 unsigned i;
830
831 ch = READCHAR;
832
833 /* Read a variable name. */
834 while (ch == ' ' || ch == '\t')
835 ch = READCHAR;
836
837 i = 0;
838 while (ch != ':' && ch != '\n' && ch != EOF && in_file_vars)
839 {
840 if (i < sizeof var - 1)
841 var[i++] = ch;
842 UPDATE_BEG_END_STATE (ch);
843 ch = READCHAR;
844 }
845
846 /* Stop scanning if no colon was found before end marker. */
847 if (!in_file_vars || ch == '\n' || ch == EOF)
848 break;
849
850 while (i > 0 && (var[i - 1] == ' ' || var[i - 1] == '\t'))
851 i--;
852 var[i] = '\0';
853
854 if (ch == ':')
855 {
856 /* Read a variable value. */
857 ch = READCHAR;
858
859 while (ch == ' ' || ch == '\t')
860 ch = READCHAR;
861
862 i = 0;
863 while (ch != ';' && ch != '\n' && ch != EOF && in_file_vars)
864 {
865 if (i < sizeof val - 1)
866 val[i++] = ch;
867 UPDATE_BEG_END_STATE (ch);
868 ch = READCHAR;
869 }
870 if (! in_file_vars)
871 /* The value was terminated by an end-marker, which remove. */
872 i -= 3;
873 while (i > 0 && (val[i - 1] == ' ' || val[i - 1] == '\t'))
874 i--;
875 val[i] = '\0';
876
877 if (strcmp (var, "lexical-binding") == 0)
878 /* This is it... */
879 {
880 rv = (strcmp (val, "nil") != 0);
881 break;
882 }
883 }
884 }
885
886 while (ch != '\n' && ch != EOF)
887 ch = READCHAR;
888
889 return rv;
890 }
891 }
892 \f
893 /* Value is a version number of byte compiled code if the file
894 associated with file descriptor FD is a compiled Lisp file that's
895 safe to load. Only files compiled with Emacs are safe to load.
896 Files compiled with XEmacs can lead to a crash in Fbyte_code
897 because of an incompatible change in the byte compiler. */
898
899 static int
900 safe_to_load_version (int fd)
901 {
902 char buf[512];
903 int nbytes, i;
904 int version = 1;
905
906 /* Read the first few bytes from the file, and look for a line
907 specifying the byte compiler version used. */
908 nbytes = emacs_read (fd, buf, sizeof buf);
909 if (nbytes > 0)
910 {
911 /* Skip to the next newline, skipping over the initial `ELC'
912 with NUL bytes following it, but note the version. */
913 for (i = 0; i < nbytes && buf[i] != '\n'; ++i)
914 if (i == 4)
915 version = buf[i];
916
917 if (i >= nbytes
918 || fast_c_string_match_ignore_case (Vbytecomp_version_regexp,
919 buf + i, nbytes - i) < 0)
920 version = 0;
921 }
922
923 lseek (fd, 0, SEEK_SET);
924 return version;
925 }
926
927
928 /* Callback for record_unwind_protect. Restore the old load list OLD,
929 after loading a file successfully. */
930
931 static void
932 record_load_unwind (Lisp_Object old)
933 {
934 Vloads_in_progress = old;
935 }
936
937 /* This handler function is used via internal_condition_case_1. */
938
939 static Lisp_Object
940 load_error_handler (Lisp_Object data)
941 {
942 return Qnil;
943 }
944
945 static void
946 load_warn_old_style_backquotes (Lisp_Object file)
947 {
948 if (!NILP (Vold_style_backquotes))
949 {
950 AUTO_STRING (format, "Loading `%s': old-style backquotes detected!");
951 CALLN (Fmessage, format, file);
952 }
953 }
954
955 DEFUN ("get-load-suffixes", Fget_load_suffixes, Sget_load_suffixes, 0, 0, 0,
956 doc: /* Return the suffixes that `load' should try if a suffix is \
957 required.
958 This uses the variables `load-suffixes' and `load-file-rep-suffixes'. */)
959 (void)
960 {
961 Lisp_Object lst = Qnil, suffixes = Vload_suffixes, suffix, ext;
962 while (CONSP (suffixes))
963 {
964 Lisp_Object exts = Vload_file_rep_suffixes;
965 suffix = XCAR (suffixes);
966 suffixes = XCDR (suffixes);
967 while (CONSP (exts))
968 {
969 ext = XCAR (exts);
970 exts = XCDR (exts);
971 lst = Fcons (concat2 (suffix, ext), lst);
972 }
973 }
974 return Fnreverse (lst);
975 }
976
977 DEFUN ("load", Fload, Sload, 1, 5, 0,
978 doc: /* Execute a file of Lisp code named FILE.
979 First try FILE with `.elc' appended, then try with `.el',
980 then try FILE unmodified (the exact suffixes in the exact order are
981 determined by `load-suffixes'). Environment variable references in
982 FILE are replaced with their values by calling `substitute-in-file-name'.
983 This function searches the directories in `load-path'.
984
985 If optional second arg NOERROR is non-nil,
986 report no error if FILE doesn't exist.
987 Print messages at start and end of loading unless
988 optional third arg NOMESSAGE is non-nil (but `force-load-messages'
989 overrides that).
990 If optional fourth arg NOSUFFIX is non-nil, don't try adding
991 suffixes `.elc' or `.el' to the specified name FILE.
992 If optional fifth arg MUST-SUFFIX is non-nil, insist on
993 the suffix `.elc' or `.el'; don't accept just FILE unless
994 it ends in one of those suffixes or includes a directory name.
995
996 If NOSUFFIX is nil, then if a file could not be found, try looking for
997 a different representation of the file by adding non-empty suffixes to
998 its name, before trying another file. Emacs uses this feature to find
999 compressed versions of files when Auto Compression mode is enabled.
1000 If NOSUFFIX is non-nil, disable this feature.
1001
1002 The suffixes that this function tries out, when NOSUFFIX is nil, are
1003 given by the return value of `get-load-suffixes' and the values listed
1004 in `load-file-rep-suffixes'. If MUST-SUFFIX is non-nil, only the
1005 return value of `get-load-suffixes' is used, i.e. the file name is
1006 required to have a non-empty suffix.
1007
1008 When searching suffixes, this function normally stops at the first
1009 one that exists. If the option `load-prefer-newer' is non-nil,
1010 however, it tries all suffixes, and uses whichever file is the newest.
1011
1012 Loading a file records its definitions, and its `provide' and
1013 `require' calls, in an element of `load-history' whose
1014 car is the file name loaded. See `load-history'.
1015
1016 While the file is in the process of being loaded, the variable
1017 `load-in-progress' is non-nil and the variable `load-file-name'
1018 is bound to the file's name.
1019
1020 Return t if the file exists and loads successfully. */)
1021 (Lisp_Object file, Lisp_Object noerror, Lisp_Object nomessage,
1022 Lisp_Object nosuffix, Lisp_Object must_suffix)
1023 {
1024 FILE *stream;
1025 int fd;
1026 int fd_index;
1027 ptrdiff_t count = SPECPDL_INDEX ();
1028 Lisp_Object found, efound, hist_file_name;
1029 /* True means we printed the ".el is newer" message. */
1030 bool newer = 0;
1031 /* True means we are loading a compiled file. */
1032 bool compiled = 0;
1033 Lisp_Object handler;
1034 bool safe_p = 1;
1035 const char *fmode = "r" FOPEN_TEXT;
1036 int version;
1037
1038 CHECK_STRING (file);
1039
1040 /* If file name is magic, call the handler. */
1041 /* This shouldn't be necessary any more now that `openp' handles it right.
1042 handler = Ffind_file_name_handler (file, Qload);
1043 if (!NILP (handler))
1044 return call5 (handler, Qload, file, noerror, nomessage, nosuffix); */
1045
1046 /* The presence of this call is the result of a historical accident:
1047 it used to be in every file-operation and when it got removed
1048 everywhere, it accidentally stayed here. Since then, enough people
1049 supposedly have things like (load "$PROJECT/foo.el") in their .emacs
1050 that it seemed risky to remove. */
1051 if (! NILP (noerror))
1052 {
1053 file = internal_condition_case_1 (Fsubstitute_in_file_name, file,
1054 Qt, load_error_handler);
1055 if (NILP (file))
1056 return Qnil;
1057 }
1058 else
1059 file = Fsubstitute_in_file_name (file);
1060
1061 /* Avoid weird lossage with null string as arg,
1062 since it would try to load a directory as a Lisp file. */
1063 if (SCHARS (file) == 0)
1064 {
1065 fd = -1;
1066 errno = ENOENT;
1067 }
1068 else
1069 {
1070 Lisp_Object suffixes;
1071 found = Qnil;
1072
1073 if (! NILP (must_suffix))
1074 {
1075 /* Don't insist on adding a suffix if FILE already ends with one. */
1076 ptrdiff_t size = SBYTES (file);
1077 if (size > 3
1078 && !strcmp (SSDATA (file) + size - 3, ".el"))
1079 must_suffix = Qnil;
1080 else if (size > 4
1081 && !strcmp (SSDATA (file) + size - 4, ".elc"))
1082 must_suffix = Qnil;
1083 /* Don't insist on adding a suffix
1084 if the argument includes a directory name. */
1085 else if (! NILP (Ffile_name_directory (file)))
1086 must_suffix = Qnil;
1087 }
1088
1089 if (!NILP (nosuffix))
1090 suffixes = Qnil;
1091 else
1092 {
1093 suffixes = Fget_load_suffixes ();
1094 if (NILP (must_suffix))
1095 suffixes = CALLN (Fappend, suffixes, Vload_file_rep_suffixes);
1096 }
1097
1098 fd = openp (Vload_path, file, suffixes, &found, Qnil, load_prefer_newer);
1099 }
1100
1101 if (fd == -1)
1102 {
1103 if (NILP (noerror))
1104 report_file_error ("Cannot open load file", file);
1105 return Qnil;
1106 }
1107
1108 /* Tell startup.el whether or not we found the user's init file. */
1109 if (EQ (Qt, Vuser_init_file))
1110 Vuser_init_file = found;
1111
1112 /* If FD is -2, that means openp found a magic file. */
1113 if (fd == -2)
1114 {
1115 if (NILP (Fequal (found, file)))
1116 /* If FOUND is a different file name from FILE,
1117 find its handler even if we have already inhibited
1118 the `load' operation on FILE. */
1119 handler = Ffind_file_name_handler (found, Qt);
1120 else
1121 handler = Ffind_file_name_handler (found, Qload);
1122 if (! NILP (handler))
1123 return call5 (handler, Qload, found, noerror, nomessage, Qt);
1124 #ifdef DOS_NT
1125 /* Tramp has to deal with semi-broken packages that prepend
1126 drive letters to remote files. For that reason, Tramp
1127 catches file operations that test for file existence, which
1128 makes openp think X:/foo.elc files are remote. However,
1129 Tramp does not catch `load' operations for such files, so we
1130 end up with a nil as the `load' handler above. If we would
1131 continue with fd = -2, we will behave wrongly, and in
1132 particular try reading a .elc file in the "rt" mode instead
1133 of "rb". See bug #9311 for the results. To work around
1134 this, we try to open the file locally, and go with that if it
1135 succeeds. */
1136 fd = emacs_open (SSDATA (ENCODE_FILE (found)), O_RDONLY, 0);
1137 if (fd == -1)
1138 fd = -2;
1139 #endif
1140 }
1141
1142 if (fd < 0)
1143 {
1144 /* Pacify older GCC with --enable-gcc-warnings. */
1145 IF_LINT (fd_index = 0);
1146 }
1147 else
1148 {
1149 fd_index = SPECPDL_INDEX ();
1150 record_unwind_protect_int (close_file_unwind, fd);
1151 }
1152
1153 /* Check if we're stuck in a recursive load cycle.
1154
1155 2000-09-21: It's not possible to just check for the file loaded
1156 being a member of Vloads_in_progress. This fails because of the
1157 way the byte compiler currently works; `provide's are not
1158 evaluated, see font-lock.el/jit-lock.el as an example. This
1159 leads to a certain amount of ``normal'' recursion.
1160
1161 Also, just loading a file recursively is not always an error in
1162 the general case; the second load may do something different. */
1163 {
1164 int load_count = 0;
1165 Lisp_Object tem;
1166 for (tem = Vloads_in_progress; CONSP (tem); tem = XCDR (tem))
1167 if (!NILP (Fequal (found, XCAR (tem))) && (++load_count > 3))
1168 signal_error ("Recursive load", Fcons (found, Vloads_in_progress));
1169 record_unwind_protect (record_load_unwind, Vloads_in_progress);
1170 Vloads_in_progress = Fcons (found, Vloads_in_progress);
1171 }
1172
1173 /* All loads are by default dynamic, unless the file itself specifies
1174 otherwise using a file-variable in the first line. This is bound here
1175 so that it takes effect whether or not we use
1176 Vload_source_file_function. */
1177 specbind (Qlexical_binding, Qnil);
1178
1179 /* Get the name for load-history. */
1180 hist_file_name = (! NILP (Vpurify_flag)
1181 ? concat2 (Ffile_name_directory (file),
1182 Ffile_name_nondirectory (found))
1183 : found) ;
1184
1185 version = -1;
1186
1187 /* Check for the presence of old-style quotes and warn about them. */
1188 specbind (Qold_style_backquotes, Qnil);
1189 record_unwind_protect (load_warn_old_style_backquotes, file);
1190
1191 if (!memcmp (SDATA (found) + SBYTES (found) - 4, ".elc", 4)
1192 || (fd >= 0 && (version = safe_to_load_version (fd)) > 0))
1193 /* Load .elc files directly, but not when they are
1194 remote and have no handler! */
1195 {
1196 if (fd != -2)
1197 {
1198 struct stat s1, s2;
1199 int result;
1200
1201 if (version < 0
1202 && ! (version = safe_to_load_version (fd)))
1203 {
1204 safe_p = 0;
1205 if (!load_dangerous_libraries)
1206 error ("File `%s' was not compiled in Emacs", SDATA (found));
1207 else if (!NILP (nomessage) && !force_load_messages)
1208 message_with_string ("File `%s' not compiled in Emacs", found, 1);
1209 }
1210
1211 compiled = 1;
1212
1213 efound = ENCODE_FILE (found);
1214 fmode = "r" FOPEN_BINARY;
1215
1216 /* openp already checked for newness, no point doing it again.
1217 FIXME would be nice to get a message when openp
1218 ignores suffix order due to load_prefer_newer. */
1219 if (!load_prefer_newer)
1220 {
1221 result = stat (SSDATA (efound), &s1);
1222 if (result == 0)
1223 {
1224 SSET (efound, SBYTES (efound) - 1, 0);
1225 result = stat (SSDATA (efound), &s2);
1226 SSET (efound, SBYTES (efound) - 1, 'c');
1227 }
1228
1229 if (result == 0
1230 && timespec_cmp (get_stat_mtime (&s1), get_stat_mtime (&s2)) < 0)
1231 {
1232 /* Make the progress messages mention that source is newer. */
1233 newer = 1;
1234
1235 /* If we won't print another message, mention this anyway. */
1236 if (!NILP (nomessage) && !force_load_messages)
1237 {
1238 Lisp_Object msg_file;
1239 msg_file = Fsubstring (found, make_number (0), make_number (-1));
1240 message_with_string ("Source file `%s' newer than byte-compiled file",
1241 msg_file, 1);
1242 }
1243 }
1244 } /* !load_prefer_newer */
1245 }
1246 }
1247 else
1248 {
1249 /* We are loading a source file (*.el). */
1250 if (!NILP (Vload_source_file_function))
1251 {
1252 Lisp_Object val;
1253
1254 if (fd >= 0)
1255 {
1256 emacs_close (fd);
1257 clear_unwind_protect (fd_index);
1258 }
1259 val = call4 (Vload_source_file_function, found, hist_file_name,
1260 NILP (noerror) ? Qnil : Qt,
1261 (NILP (nomessage) || force_load_messages) ? Qnil : Qt);
1262 return unbind_to (count, val);
1263 }
1264 }
1265
1266 if (fd < 0)
1267 {
1268 /* We somehow got here with fd == -2, meaning the file is deemed
1269 to be remote. Don't even try to reopen the file locally;
1270 just force a failure. */
1271 stream = NULL;
1272 errno = EINVAL;
1273 }
1274 else
1275 {
1276 #ifdef WINDOWSNT
1277 emacs_close (fd);
1278 clear_unwind_protect (fd_index);
1279 efound = ENCODE_FILE (found);
1280 stream = emacs_fopen (SSDATA (efound), fmode);
1281 #else
1282 stream = fdopen (fd, fmode);
1283 #endif
1284 }
1285 if (! stream)
1286 report_file_error ("Opening stdio stream", file);
1287 set_unwind_protect_ptr (fd_index, fclose_unwind, stream);
1288
1289 if (! NILP (Vpurify_flag))
1290 Vpreloaded_file_list = Fcons (Fpurecopy (file), Vpreloaded_file_list);
1291
1292 if (NILP (nomessage) || force_load_messages)
1293 {
1294 if (!safe_p)
1295 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...",
1296 file, 1);
1297 else if (!compiled)
1298 message_with_string ("Loading %s (source)...", file, 1);
1299 else if (newer)
1300 message_with_string ("Loading %s (compiled; note, source file is newer)...",
1301 file, 1);
1302 else /* The typical case; compiled file newer than source file. */
1303 message_with_string ("Loading %s...", file, 1);
1304 }
1305
1306 specbind (Qload_file_name, found);
1307 specbind (Qinhibit_file_name_operation, Qnil);
1308 specbind (Qload_in_progress, Qt);
1309
1310 instream = stream;
1311 if (lisp_file_lexically_bound_p (Qget_file_char))
1312 Fset (Qlexical_binding, Qt);
1313
1314 if (! version || version >= 22)
1315 readevalloop (Qget_file_char, stream, hist_file_name,
1316 0, Qnil, Qnil, Qnil, Qnil);
1317 else
1318 {
1319 /* We can't handle a file which was compiled with
1320 byte-compile-dynamic by older version of Emacs. */
1321 specbind (Qload_force_doc_strings, Qt);
1322 readevalloop (Qget_emacs_mule_file_char, stream, hist_file_name,
1323 0, Qnil, Qnil, Qnil, Qnil);
1324 }
1325 unbind_to (count, Qnil);
1326
1327 /* Run any eval-after-load forms for this file. */
1328 if (!NILP (Ffboundp (Qdo_after_load_evaluation)))
1329 call1 (Qdo_after_load_evaluation, hist_file_name) ;
1330
1331 xfree (saved_doc_string);
1332 saved_doc_string = 0;
1333 saved_doc_string_size = 0;
1334
1335 xfree (prev_saved_doc_string);
1336 prev_saved_doc_string = 0;
1337 prev_saved_doc_string_size = 0;
1338
1339 if (!noninteractive && (NILP (nomessage) || force_load_messages))
1340 {
1341 if (!safe_p)
1342 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...done",
1343 file, 1);
1344 else if (!compiled)
1345 message_with_string ("Loading %s (source)...done", file, 1);
1346 else if (newer)
1347 message_with_string ("Loading %s (compiled; note, source file is newer)...done",
1348 file, 1);
1349 else /* The typical case; compiled file newer than source file. */
1350 message_with_string ("Loading %s...done", file, 1);
1351 }
1352
1353 return Qt;
1354 }
1355 \f
1356 static bool
1357 complete_filename_p (Lisp_Object pathname)
1358 {
1359 const unsigned char *s = SDATA (pathname);
1360 return (IS_DIRECTORY_SEP (s[0])
1361 || (SCHARS (pathname) > 2
1362 && IS_DEVICE_SEP (s[1]) && IS_DIRECTORY_SEP (s[2])));
1363 }
1364
1365 DEFUN ("locate-file-internal", Flocate_file_internal, Slocate_file_internal, 2, 4, 0,
1366 doc: /* Search for FILENAME through PATH.
1367 Returns the file's name in absolute form, or nil if not found.
1368 If SUFFIXES is non-nil, it should be a list of suffixes to append to
1369 file name when searching.
1370 If non-nil, PREDICATE is used instead of `file-readable-p'.
1371 PREDICATE can also be an integer to pass to the faccessat(2) function,
1372 in which case file-name-handlers are ignored.
1373 This function will normally skip directories, so if you want it to find
1374 directories, make sure the PREDICATE function returns `dir-ok' for them. */)
1375 (Lisp_Object filename, Lisp_Object path, Lisp_Object suffixes, Lisp_Object predicate)
1376 {
1377 Lisp_Object file;
1378 int fd = openp (path, filename, suffixes, &file, predicate, false);
1379 if (NILP (predicate) && fd >= 0)
1380 emacs_close (fd);
1381 return file;
1382 }
1383
1384 /* Search for a file whose name is STR, looking in directories
1385 in the Lisp list PATH, and trying suffixes from SUFFIX.
1386 On success, return a file descriptor (or 1 or -2 as described below).
1387 On failure, return -1 and set errno.
1388
1389 SUFFIXES is a list of strings containing possible suffixes.
1390 The empty suffix is automatically added if the list is empty.
1391
1392 PREDICATE t means the files are binary.
1393 PREDICATE non-nil and non-t means don't open the files,
1394 just look for one that satisfies the predicate. In this case,
1395 return 1 on success. The predicate can be a lisp function or
1396 an integer to pass to `access' (in which case file-name-handlers
1397 are ignored).
1398
1399 If STOREPTR is nonzero, it points to a slot where the name of
1400 the file actually found should be stored as a Lisp string.
1401 nil is stored there on failure.
1402
1403 If the file we find is remote, return -2
1404 but store the found remote file name in *STOREPTR.
1405
1406 If NEWER is true, try all SUFFIXes and return the result for the
1407 newest file that exists. Does not apply to remote files,
1408 or if a non-nil and non-t PREDICATE is specified. */
1409
1410 int
1411 openp (Lisp_Object path, Lisp_Object str, Lisp_Object suffixes,
1412 Lisp_Object *storeptr, Lisp_Object predicate, bool newer)
1413 {
1414 ptrdiff_t fn_size = 100;
1415 char buf[100];
1416 char *fn = buf;
1417 bool absolute;
1418 ptrdiff_t want_length;
1419 Lisp_Object filename;
1420 Lisp_Object string, tail, encoded_fn, save_string;
1421 ptrdiff_t max_suffix_len = 0;
1422 int last_errno = ENOENT;
1423 int save_fd = -1;
1424 USE_SAFE_ALLOCA;
1425
1426 /* The last-modified time of the newest matching file found.
1427 Initialize it to something less than all valid timestamps. */
1428 struct timespec save_mtime = make_timespec (TYPE_MINIMUM (time_t), -1);
1429
1430 CHECK_STRING (str);
1431
1432 for (tail = suffixes; CONSP (tail); tail = XCDR (tail))
1433 {
1434 CHECK_STRING_CAR (tail);
1435 max_suffix_len = max (max_suffix_len,
1436 SBYTES (XCAR (tail)));
1437 }
1438
1439 string = filename = encoded_fn = save_string = Qnil;
1440
1441 if (storeptr)
1442 *storeptr = Qnil;
1443
1444 absolute = complete_filename_p (str);
1445
1446 for (; CONSP (path); path = XCDR (path))
1447 {
1448 filename = Fexpand_file_name (str, XCAR (path));
1449 if (!complete_filename_p (filename))
1450 /* If there are non-absolute elts in PATH (eg "."). */
1451 /* Of course, this could conceivably lose if luser sets
1452 default-directory to be something non-absolute... */
1453 {
1454 filename = Fexpand_file_name (filename, BVAR (current_buffer, directory));
1455 if (!complete_filename_p (filename))
1456 /* Give up on this path element! */
1457 continue;
1458 }
1459
1460 /* Calculate maximum length of any filename made from
1461 this path element/specified file name and any possible suffix. */
1462 want_length = max_suffix_len + SBYTES (filename);
1463 if (fn_size <= want_length)
1464 {
1465 fn_size = 100 + want_length;
1466 fn = SAFE_ALLOCA (fn_size);
1467 }
1468
1469 /* Loop over suffixes. */
1470 for (tail = NILP (suffixes) ? list1 (empty_unibyte_string) : suffixes;
1471 CONSP (tail); tail = XCDR (tail))
1472 {
1473 Lisp_Object suffix = XCAR (tail);
1474 ptrdiff_t fnlen, lsuffix = SBYTES (suffix);
1475 Lisp_Object handler;
1476
1477 /* Concatenate path element/specified name with the suffix.
1478 If the directory starts with /:, remove that. */
1479 int prefixlen = ((SCHARS (filename) > 2
1480 && SREF (filename, 0) == '/'
1481 && SREF (filename, 1) == ':')
1482 ? 2 : 0);
1483 fnlen = SBYTES (filename) - prefixlen;
1484 memcpy (fn, SDATA (filename) + prefixlen, fnlen);
1485 memcpy (fn + fnlen, SDATA (suffix), lsuffix + 1);
1486 fnlen += lsuffix;
1487 /* Check that the file exists and is not a directory. */
1488 /* We used to only check for handlers on non-absolute file names:
1489 if (absolute)
1490 handler = Qnil;
1491 else
1492 handler = Ffind_file_name_handler (filename, Qfile_exists_p);
1493 It's not clear why that was the case and it breaks things like
1494 (load "/bar.el") where the file is actually "/bar.el.gz". */
1495 /* make_string has its own ideas on when to return a unibyte
1496 string and when a multibyte string, but we know better.
1497 We must have a unibyte string when dumping, since
1498 file-name encoding is shaky at best at that time, and in
1499 particular default-file-name-coding-system is reset
1500 several times during loadup. We therefore don't want to
1501 encode the file before passing it to file I/O library
1502 functions. */
1503 if (!STRING_MULTIBYTE (filename) && !STRING_MULTIBYTE (suffix))
1504 string = make_unibyte_string (fn, fnlen);
1505 else
1506 string = make_string (fn, fnlen);
1507 handler = Ffind_file_name_handler (string, Qfile_exists_p);
1508 if ((!NILP (handler) || (!NILP (predicate) && !EQ (predicate, Qt)))
1509 && !NATNUMP (predicate))
1510 {
1511 bool exists;
1512 if (NILP (predicate) || EQ (predicate, Qt))
1513 exists = !NILP (Ffile_readable_p (string));
1514 else
1515 {
1516 Lisp_Object tmp = call1 (predicate, string);
1517 if (NILP (tmp))
1518 exists = false;
1519 else if (EQ (tmp, Qdir_ok)
1520 || NILP (Ffile_directory_p (string)))
1521 exists = true;
1522 else
1523 {
1524 exists = false;
1525 last_errno = EISDIR;
1526 }
1527 }
1528
1529 if (exists)
1530 {
1531 /* We succeeded; return this descriptor and filename. */
1532 if (storeptr)
1533 *storeptr = string;
1534 SAFE_FREE ();
1535 return -2;
1536 }
1537 }
1538 else
1539 {
1540 int fd;
1541 const char *pfn;
1542 struct stat st;
1543
1544 encoded_fn = ENCODE_FILE (string);
1545 pfn = SSDATA (encoded_fn);
1546
1547 /* Check that we can access or open it. */
1548 if (NATNUMP (predicate))
1549 {
1550 fd = -1;
1551 if (INT_MAX < XFASTINT (predicate))
1552 last_errno = EINVAL;
1553 else if (faccessat (AT_FDCWD, pfn, XFASTINT (predicate),
1554 AT_EACCESS)
1555 == 0)
1556 {
1557 if (file_directory_p (pfn))
1558 last_errno = EISDIR;
1559 else
1560 fd = 1;
1561 }
1562 }
1563 else
1564 {
1565 int oflags = O_RDONLY + (NILP (predicate) ? 0 : O_BINARY);
1566 fd = emacs_open (pfn, oflags, 0);
1567 if (fd < 0)
1568 {
1569 if (errno != ENOENT)
1570 last_errno = errno;
1571 }
1572 else
1573 {
1574 int err = (fstat (fd, &st) != 0 ? errno
1575 : S_ISDIR (st.st_mode) ? EISDIR : 0);
1576 if (err)
1577 {
1578 last_errno = err;
1579 emacs_close (fd);
1580 fd = -1;
1581 }
1582 }
1583 }
1584
1585 if (fd >= 0)
1586 {
1587 if (newer && !NATNUMP (predicate))
1588 {
1589 struct timespec mtime = get_stat_mtime (&st);
1590
1591 if (timespec_cmp (mtime, save_mtime) <= 0)
1592 emacs_close (fd);
1593 else
1594 {
1595 if (0 <= save_fd)
1596 emacs_close (save_fd);
1597 save_fd = fd;
1598 save_mtime = mtime;
1599 save_string = string;
1600 }
1601 }
1602 else
1603 {
1604 /* We succeeded; return this descriptor and filename. */
1605 if (storeptr)
1606 *storeptr = string;
1607 SAFE_FREE ();
1608 return fd;
1609 }
1610 }
1611
1612 /* No more suffixes. Return the newest. */
1613 if (0 <= save_fd && ! CONSP (XCDR (tail)))
1614 {
1615 if (storeptr)
1616 *storeptr = save_string;
1617 SAFE_FREE ();
1618 return save_fd;
1619 }
1620 }
1621 }
1622 if (absolute)
1623 break;
1624 }
1625
1626 SAFE_FREE ();
1627 errno = last_errno;
1628 return -1;
1629 }
1630
1631 \f
1632 /* Merge the list we've accumulated of globals from the current input source
1633 into the load_history variable. The details depend on whether
1634 the source has an associated file name or not.
1635
1636 FILENAME is the file name that we are loading from.
1637
1638 ENTIRE is true if loading that entire file, false if evaluating
1639 part of it. */
1640
1641 static void
1642 build_load_history (Lisp_Object filename, bool entire)
1643 {
1644 Lisp_Object tail, prev, newelt;
1645 Lisp_Object tem, tem2;
1646 bool foundit = 0;
1647
1648 tail = Vload_history;
1649 prev = Qnil;
1650
1651 while (CONSP (tail))
1652 {
1653 tem = XCAR (tail);
1654
1655 /* Find the feature's previous assoc list... */
1656 if (!NILP (Fequal (filename, Fcar (tem))))
1657 {
1658 foundit = 1;
1659
1660 /* If we're loading the entire file, remove old data. */
1661 if (entire)
1662 {
1663 if (NILP (prev))
1664 Vload_history = XCDR (tail);
1665 else
1666 Fsetcdr (prev, XCDR (tail));
1667 }
1668
1669 /* Otherwise, cons on new symbols that are not already members. */
1670 else
1671 {
1672 tem2 = Vcurrent_load_list;
1673
1674 while (CONSP (tem2))
1675 {
1676 newelt = XCAR (tem2);
1677
1678 if (NILP (Fmember (newelt, tem)))
1679 Fsetcar (tail, Fcons (XCAR (tem),
1680 Fcons (newelt, XCDR (tem))));
1681
1682 tem2 = XCDR (tem2);
1683 QUIT;
1684 }
1685 }
1686 }
1687 else
1688 prev = tail;
1689 tail = XCDR (tail);
1690 QUIT;
1691 }
1692
1693 /* If we're loading an entire file, cons the new assoc onto the
1694 front of load-history, the most-recently-loaded position. Also
1695 do this if we didn't find an existing member for the file. */
1696 if (entire || !foundit)
1697 Vload_history = Fcons (Fnreverse (Vcurrent_load_list),
1698 Vload_history);
1699 }
1700
1701 static void
1702 readevalloop_1 (int old)
1703 {
1704 load_convert_to_unibyte = old;
1705 }
1706
1707 /* Signal an `end-of-file' error, if possible with file name
1708 information. */
1709
1710 static _Noreturn void
1711 end_of_file_error (void)
1712 {
1713 if (STRINGP (Vload_file_name))
1714 xsignal1 (Qend_of_file, Vload_file_name);
1715
1716 xsignal0 (Qend_of_file);
1717 }
1718
1719 static Lisp_Object
1720 readevalloop_eager_expand_eval (Lisp_Object val, Lisp_Object macroexpand)
1721 {
1722 /* If we macroexpand the toplevel form non-recursively and it ends
1723 up being a `progn' (or if it was a progn to start), treat each
1724 form in the progn as a top-level form. This way, if one form in
1725 the progn defines a macro, that macro is in effect when we expand
1726 the remaining forms. See similar code in bytecomp.el. */
1727 val = call2 (macroexpand, val, Qnil);
1728 if (EQ (CAR_SAFE (val), Qprogn))
1729 {
1730 Lisp_Object subforms = XCDR (val);
1731
1732 for (val = Qnil; CONSP (subforms); subforms = XCDR (subforms))
1733 val = readevalloop_eager_expand_eval (XCAR (subforms),
1734 macroexpand);
1735 }
1736 else
1737 val = eval_sub (call2 (macroexpand, val, Qt));
1738 return val;
1739 }
1740
1741 /* UNIBYTE specifies how to set load_convert_to_unibyte
1742 for this invocation.
1743 READFUN, if non-nil, is used instead of `read'.
1744
1745 START, END specify region to read in current buffer (from eval-region).
1746 If the input is not from a buffer, they must be nil. */
1747
1748 static void
1749 readevalloop (Lisp_Object readcharfun,
1750 FILE *stream,
1751 Lisp_Object sourcename,
1752 bool printflag,
1753 Lisp_Object unibyte, Lisp_Object readfun,
1754 Lisp_Object start, Lisp_Object end)
1755 {
1756 int c;
1757 Lisp_Object val;
1758 ptrdiff_t count = SPECPDL_INDEX ();
1759 struct buffer *b = 0;
1760 bool continue_reading_p;
1761 Lisp_Object lex_bound;
1762 /* True if reading an entire buffer. */
1763 bool whole_buffer = 0;
1764 /* True on the first time around. */
1765 bool first_sexp = 1;
1766 Lisp_Object macroexpand = intern ("internal-macroexpand-for-load");
1767
1768 if (NILP (Ffboundp (macroexpand))
1769 /* Don't macroexpand in .elc files, since it should have been done
1770 already. We actually don't know whether we're in a .elc file or not,
1771 so we use circumstantial evidence: .el files normally go through
1772 Vload_source_file_function -> load-with-code-conversion
1773 -> eval-buffer. */
1774 || EQ (readcharfun, Qget_file_char)
1775 || EQ (readcharfun, Qget_emacs_mule_file_char))
1776 macroexpand = Qnil;
1777
1778 if (MARKERP (readcharfun))
1779 {
1780 if (NILP (start))
1781 start = readcharfun;
1782 }
1783
1784 if (BUFFERP (readcharfun))
1785 b = XBUFFER (readcharfun);
1786 else if (MARKERP (readcharfun))
1787 b = XMARKER (readcharfun)->buffer;
1788
1789 /* We assume START is nil when input is not from a buffer. */
1790 if (! NILP (start) && !b)
1791 emacs_abort ();
1792
1793 specbind (Qstandard_input, readcharfun);
1794 specbind (Qcurrent_load_list, Qnil);
1795 record_unwind_protect_int (readevalloop_1, load_convert_to_unibyte);
1796 load_convert_to_unibyte = !NILP (unibyte);
1797
1798 /* If lexical binding is active (either because it was specified in
1799 the file's header, or via a buffer-local variable), create an empty
1800 lexical environment, otherwise, turn off lexical binding. */
1801 lex_bound = find_symbol_value (Qlexical_binding);
1802 specbind (Qinternal_interpreter_environment,
1803 (NILP (lex_bound) || EQ (lex_bound, Qunbound)
1804 ? Qnil : list1 (Qt)));
1805
1806 /* Try to ensure sourcename is a truename, except whilst preloading. */
1807 if (NILP (Vpurify_flag)
1808 && !NILP (sourcename) && !NILP (Ffile_name_absolute_p (sourcename))
1809 && !NILP (Ffboundp (Qfile_truename)))
1810 sourcename = call1 (Qfile_truename, sourcename) ;
1811
1812 LOADHIST_ATTACH (sourcename);
1813
1814 continue_reading_p = 1;
1815 while (continue_reading_p)
1816 {
1817 ptrdiff_t count1 = SPECPDL_INDEX ();
1818
1819 if (b != 0 && !BUFFER_LIVE_P (b))
1820 error ("Reading from killed buffer");
1821
1822 if (!NILP (start))
1823 {
1824 /* Switch to the buffer we are reading from. */
1825 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1826 set_buffer_internal (b);
1827
1828 /* Save point in it. */
1829 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1830 /* Save ZV in it. */
1831 record_unwind_protect (save_restriction_restore, save_restriction_save ());
1832 /* Those get unbound after we read one expression. */
1833
1834 /* Set point and ZV around stuff to be read. */
1835 Fgoto_char (start);
1836 if (!NILP (end))
1837 Fnarrow_to_region (make_number (BEGV), end);
1838
1839 /* Just for cleanliness, convert END to a marker
1840 if it is an integer. */
1841 if (INTEGERP (end))
1842 end = Fpoint_max_marker ();
1843 }
1844
1845 /* On the first cycle, we can easily test here
1846 whether we are reading the whole buffer. */
1847 if (b && first_sexp)
1848 whole_buffer = (PT == BEG && ZV == Z);
1849
1850 instream = stream;
1851 read_next:
1852 c = READCHAR;
1853 if (c == ';')
1854 {
1855 while ((c = READCHAR) != '\n' && c != -1);
1856 goto read_next;
1857 }
1858 if (c < 0)
1859 {
1860 unbind_to (count1, Qnil);
1861 break;
1862 }
1863
1864 /* Ignore whitespace here, so we can detect eof. */
1865 if (c == ' ' || c == '\t' || c == '\n' || c == '\f' || c == '\r'
1866 || c == NO_BREAK_SPACE)
1867 goto read_next;
1868
1869 if (!NILP (Vpurify_flag) && c == '(')
1870 {
1871 val = read_list (0, readcharfun);
1872 }
1873 else
1874 {
1875 UNREAD (c);
1876 read_objects = Qnil;
1877 if (!NILP (readfun))
1878 {
1879 val = call1 (readfun, readcharfun);
1880
1881 /* If READCHARFUN has set point to ZV, we should
1882 stop reading, even if the form read sets point
1883 to a different value when evaluated. */
1884 if (BUFFERP (readcharfun))
1885 {
1886 struct buffer *buf = XBUFFER (readcharfun);
1887 if (BUF_PT (buf) == BUF_ZV (buf))
1888 continue_reading_p = 0;
1889 }
1890 }
1891 else if (! NILP (Vload_read_function))
1892 val = call1 (Vload_read_function, readcharfun);
1893 else
1894 val = read_internal_start (readcharfun, Qnil, Qnil);
1895 }
1896
1897 if (!NILP (start) && continue_reading_p)
1898 start = Fpoint_marker ();
1899
1900 /* Restore saved point and BEGV. */
1901 unbind_to (count1, Qnil);
1902
1903 /* Now eval what we just read. */
1904 if (!NILP (macroexpand))
1905 val = readevalloop_eager_expand_eval (val, macroexpand);
1906 else
1907 val = eval_sub (val);
1908
1909 if (printflag)
1910 {
1911 Vvalues = Fcons (val, Vvalues);
1912 if (EQ (Vstandard_output, Qt))
1913 Fprin1 (val, Qnil);
1914 else
1915 Fprint (val, Qnil);
1916 }
1917
1918 first_sexp = 0;
1919 }
1920
1921 build_load_history (sourcename,
1922 stream || whole_buffer);
1923
1924 unbind_to (count, Qnil);
1925 }
1926
1927 DEFUN ("eval-buffer", Feval_buffer, Seval_buffer, 0, 5, "",
1928 doc: /* Execute the current buffer as Lisp code.
1929 When called from a Lisp program (i.e., not interactively), this
1930 function accepts up to five optional arguments:
1931 BUFFER is the buffer to evaluate (nil means use current buffer).
1932 PRINTFLAG controls printing of output:
1933 A value of nil means discard it; anything else is stream for print.
1934 FILENAME specifies the file name to use for `load-history'.
1935 UNIBYTE, if non-nil, specifies `load-convert-to-unibyte' for this
1936 invocation.
1937 DO-ALLOW-PRINT, if non-nil, specifies that `print' and related
1938 functions should work normally even if PRINTFLAG is nil.
1939
1940 This function preserves the position of point. */)
1941 (Lisp_Object buffer, Lisp_Object printflag, Lisp_Object filename, Lisp_Object unibyte, Lisp_Object do_allow_print)
1942 {
1943 ptrdiff_t count = SPECPDL_INDEX ();
1944 Lisp_Object tem, buf;
1945
1946 if (NILP (buffer))
1947 buf = Fcurrent_buffer ();
1948 else
1949 buf = Fget_buffer (buffer);
1950 if (NILP (buf))
1951 error ("No such buffer");
1952
1953 if (NILP (printflag) && NILP (do_allow_print))
1954 tem = Qsymbolp;
1955 else
1956 tem = printflag;
1957
1958 if (NILP (filename))
1959 filename = BVAR (XBUFFER (buf), filename);
1960
1961 specbind (Qeval_buffer_list, Fcons (buf, Veval_buffer_list));
1962 specbind (Qstandard_output, tem);
1963 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1964 BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
1965 specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
1966 readevalloop (buf, 0, filename,
1967 !NILP (printflag), unibyte, Qnil, Qnil, Qnil);
1968 unbind_to (count, Qnil);
1969
1970 return Qnil;
1971 }
1972
1973 DEFUN ("eval-region", Feval_region, Seval_region, 2, 4, "r",
1974 doc: /* Execute the region as Lisp code.
1975 When called from programs, expects two arguments,
1976 giving starting and ending indices in the current buffer
1977 of the text to be executed.
1978 Programs can pass third argument PRINTFLAG which controls output:
1979 A value of nil means discard it; anything else is stream for printing it.
1980 Also the fourth argument READ-FUNCTION, if non-nil, is used
1981 instead of `read' to read each expression. It gets one argument
1982 which is the input stream for reading characters.
1983
1984 This function does not move point. */)
1985 (Lisp_Object start, Lisp_Object end, Lisp_Object printflag, Lisp_Object read_function)
1986 {
1987 /* FIXME: Do the eval-sexp-add-defvars dance! */
1988 ptrdiff_t count = SPECPDL_INDEX ();
1989 Lisp_Object tem, cbuf;
1990
1991 cbuf = Fcurrent_buffer ();
1992
1993 if (NILP (printflag))
1994 tem = Qsymbolp;
1995 else
1996 tem = printflag;
1997 specbind (Qstandard_output, tem);
1998 specbind (Qeval_buffer_list, Fcons (cbuf, Veval_buffer_list));
1999
2000 /* `readevalloop' calls functions which check the type of start and end. */
2001 readevalloop (cbuf, 0, BVAR (XBUFFER (cbuf), filename),
2002 !NILP (printflag), Qnil, read_function,
2003 start, end);
2004
2005 return unbind_to (count, Qnil);
2006 }
2007
2008 \f
2009 DEFUN ("read", Fread, Sread, 0, 1, 0,
2010 doc: /* Read one Lisp expression as text from STREAM, return as Lisp object.
2011 If STREAM is nil, use the value of `standard-input' (which see).
2012 STREAM or the value of `standard-input' may be:
2013 a buffer (read from point and advance it)
2014 a marker (read from where it points and advance it)
2015 a function (call it with no arguments for each character,
2016 call it with a char as argument to push a char back)
2017 a string (takes text from string, starting at the beginning)
2018 t (read text line using minibuffer and use it, or read from
2019 standard input in batch mode). */)
2020 (Lisp_Object stream)
2021 {
2022 if (NILP (stream))
2023 stream = Vstandard_input;
2024 if (EQ (stream, Qt))
2025 stream = Qread_char;
2026 if (EQ (stream, Qread_char))
2027 /* FIXME: ?! When is this used !? */
2028 return call1 (intern ("read-minibuffer"),
2029 build_string ("Lisp expression: "));
2030
2031 return read_internal_start (stream, Qnil, Qnil);
2032 }
2033
2034 DEFUN ("read-from-string", Fread_from_string, Sread_from_string, 1, 3, 0,
2035 doc: /* Read one Lisp expression which is represented as text by STRING.
2036 Returns a cons: (OBJECT-READ . FINAL-STRING-INDEX).
2037 FINAL-STRING-INDEX is an integer giving the position of the next
2038 remaining character in STRING. START and END optionally delimit
2039 a substring of STRING from which to read; they default to 0 and
2040 (length STRING) respectively. Negative values are counted from
2041 the end of STRING. */)
2042 (Lisp_Object string, Lisp_Object start, Lisp_Object end)
2043 {
2044 Lisp_Object ret;
2045 CHECK_STRING (string);
2046 /* `read_internal_start' sets `read_from_string_index'. */
2047 ret = read_internal_start (string, start, end);
2048 return Fcons (ret, make_number (read_from_string_index));
2049 }
2050
2051 /* Function to set up the global context we need in toplevel read
2052 calls. START and END only used when STREAM is a string. */
2053 static Lisp_Object
2054 read_internal_start (Lisp_Object stream, Lisp_Object start, Lisp_Object end)
2055 {
2056 Lisp_Object retval;
2057
2058 readchar_count = 0;
2059 new_backquote_flag = 0;
2060 read_objects = Qnil;
2061 if (EQ (Vread_with_symbol_positions, Qt)
2062 || EQ (Vread_with_symbol_positions, stream))
2063 Vread_symbol_positions_list = Qnil;
2064
2065 if (STRINGP (stream)
2066 || ((CONSP (stream) && STRINGP (XCAR (stream)))))
2067 {
2068 ptrdiff_t startval, endval;
2069 Lisp_Object string;
2070
2071 if (STRINGP (stream))
2072 string = stream;
2073 else
2074 string = XCAR (stream);
2075
2076 validate_subarray (string, start, end, SCHARS (string),
2077 &startval, &endval);
2078
2079 read_from_string_index = startval;
2080 read_from_string_index_byte = string_char_to_byte (string, startval);
2081 read_from_string_limit = endval;
2082 }
2083
2084 retval = read0 (stream);
2085 if (EQ (Vread_with_symbol_positions, Qt)
2086 || EQ (Vread_with_symbol_positions, stream))
2087 Vread_symbol_positions_list = Fnreverse (Vread_symbol_positions_list);
2088 return retval;
2089 }
2090 \f
2091
2092 /* Signal Qinvalid_read_syntax error.
2093 S is error string of length N (if > 0) */
2094
2095 static _Noreturn void
2096 invalid_syntax (const char *s)
2097 {
2098 xsignal1 (Qinvalid_read_syntax, build_string (s));
2099 }
2100
2101
2102 /* Use this for recursive reads, in contexts where internal tokens
2103 are not allowed. */
2104
2105 static Lisp_Object
2106 read0 (Lisp_Object readcharfun)
2107 {
2108 register Lisp_Object val;
2109 int c;
2110
2111 val = read1 (readcharfun, &c, 0);
2112 if (!c)
2113 return val;
2114
2115 xsignal1 (Qinvalid_read_syntax,
2116 Fmake_string (make_number (1), make_number (c)));
2117 }
2118 \f
2119 static ptrdiff_t read_buffer_size;
2120 static char *read_buffer;
2121
2122 /* Read a \-escape sequence, assuming we already read the `\'.
2123 If the escape sequence forces unibyte, return eight-bit char. */
2124
2125 static int
2126 read_escape (Lisp_Object readcharfun, bool stringp)
2127 {
2128 int c = READCHAR;
2129 /* \u allows up to four hex digits, \U up to eight. Default to the
2130 behavior for \u, and change this value in the case that \U is seen. */
2131 int unicode_hex_count = 4;
2132
2133 switch (c)
2134 {
2135 case -1:
2136 end_of_file_error ();
2137
2138 case 'a':
2139 return '\007';
2140 case 'b':
2141 return '\b';
2142 case 'd':
2143 return 0177;
2144 case 'e':
2145 return 033;
2146 case 'f':
2147 return '\f';
2148 case 'n':
2149 return '\n';
2150 case 'r':
2151 return '\r';
2152 case 't':
2153 return '\t';
2154 case 'v':
2155 return '\v';
2156 case '\n':
2157 return -1;
2158 case ' ':
2159 if (stringp)
2160 return -1;
2161 return ' ';
2162
2163 case 'M':
2164 c = READCHAR;
2165 if (c != '-')
2166 error ("Invalid escape character syntax");
2167 c = READCHAR;
2168 if (c == '\\')
2169 c = read_escape (readcharfun, 0);
2170 return c | meta_modifier;
2171
2172 case 'S':
2173 c = READCHAR;
2174 if (c != '-')
2175 error ("Invalid escape character syntax");
2176 c = READCHAR;
2177 if (c == '\\')
2178 c = read_escape (readcharfun, 0);
2179 return c | shift_modifier;
2180
2181 case 'H':
2182 c = READCHAR;
2183 if (c != '-')
2184 error ("Invalid escape character syntax");
2185 c = READCHAR;
2186 if (c == '\\')
2187 c = read_escape (readcharfun, 0);
2188 return c | hyper_modifier;
2189
2190 case 'A':
2191 c = READCHAR;
2192 if (c != '-')
2193 error ("Invalid escape character syntax");
2194 c = READCHAR;
2195 if (c == '\\')
2196 c = read_escape (readcharfun, 0);
2197 return c | alt_modifier;
2198
2199 case 's':
2200 c = READCHAR;
2201 if (stringp || c != '-')
2202 {
2203 UNREAD (c);
2204 return ' ';
2205 }
2206 c = READCHAR;
2207 if (c == '\\')
2208 c = read_escape (readcharfun, 0);
2209 return c | super_modifier;
2210
2211 case 'C':
2212 c = READCHAR;
2213 if (c != '-')
2214 error ("Invalid escape character syntax");
2215 case '^':
2216 c = READCHAR;
2217 if (c == '\\')
2218 c = read_escape (readcharfun, 0);
2219 if ((c & ~CHAR_MODIFIER_MASK) == '?')
2220 return 0177 | (c & CHAR_MODIFIER_MASK);
2221 else if (! SINGLE_BYTE_CHAR_P ((c & ~CHAR_MODIFIER_MASK)))
2222 return c | ctrl_modifier;
2223 /* ASCII control chars are made from letters (both cases),
2224 as well as the non-letters within 0100...0137. */
2225 else if ((c & 0137) >= 0101 && (c & 0137) <= 0132)
2226 return (c & (037 | ~0177));
2227 else if ((c & 0177) >= 0100 && (c & 0177) <= 0137)
2228 return (c & (037 | ~0177));
2229 else
2230 return c | ctrl_modifier;
2231
2232 case '0':
2233 case '1':
2234 case '2':
2235 case '3':
2236 case '4':
2237 case '5':
2238 case '6':
2239 case '7':
2240 /* An octal escape, as in ANSI C. */
2241 {
2242 register int i = c - '0';
2243 register int count = 0;
2244 while (++count < 3)
2245 {
2246 if ((c = READCHAR) >= '0' && c <= '7')
2247 {
2248 i *= 8;
2249 i += c - '0';
2250 }
2251 else
2252 {
2253 UNREAD (c);
2254 break;
2255 }
2256 }
2257
2258 if (i >= 0x80 && i < 0x100)
2259 i = BYTE8_TO_CHAR (i);
2260 return i;
2261 }
2262
2263 case 'x':
2264 /* A hex escape, as in ANSI C. */
2265 {
2266 unsigned int i = 0;
2267 int count = 0;
2268 while (1)
2269 {
2270 c = READCHAR;
2271 if (c >= '0' && c <= '9')
2272 {
2273 i *= 16;
2274 i += c - '0';
2275 }
2276 else if ((c >= 'a' && c <= 'f')
2277 || (c >= 'A' && c <= 'F'))
2278 {
2279 i *= 16;
2280 if (c >= 'a' && c <= 'f')
2281 i += c - 'a' + 10;
2282 else
2283 i += c - 'A' + 10;
2284 }
2285 else
2286 {
2287 UNREAD (c);
2288 break;
2289 }
2290 /* Allow hex escapes as large as ?\xfffffff, because some
2291 packages use them to denote characters with modifiers. */
2292 if ((CHAR_META | (CHAR_META - 1)) < i)
2293 error ("Hex character out of range: \\x%x...", i);
2294 count += count < 3;
2295 }
2296
2297 if (count < 3 && i >= 0x80)
2298 return BYTE8_TO_CHAR (i);
2299 return i;
2300 }
2301
2302 case 'U':
2303 /* Post-Unicode-2.0: Up to eight hex chars. */
2304 unicode_hex_count = 8;
2305 case 'u':
2306
2307 /* A Unicode escape. We only permit them in strings and characters,
2308 not arbitrarily in the source code, as in some other languages. */
2309 {
2310 unsigned int i = 0;
2311 int count = 0;
2312
2313 while (++count <= unicode_hex_count)
2314 {
2315 c = READCHAR;
2316 /* `isdigit' and `isalpha' may be locale-specific, which we don't
2317 want. */
2318 if (c >= '0' && c <= '9') i = (i << 4) + (c - '0');
2319 else if (c >= 'a' && c <= 'f') i = (i << 4) + (c - 'a') + 10;
2320 else if (c >= 'A' && c <= 'F') i = (i << 4) + (c - 'A') + 10;
2321 else
2322 error ("Non-hex digit used for Unicode escape");
2323 }
2324 if (i > 0x10FFFF)
2325 error ("Non-Unicode character: 0x%x", i);
2326 return i;
2327 }
2328
2329 default:
2330 return c;
2331 }
2332 }
2333
2334 /* Return the digit that CHARACTER stands for in the given BASE.
2335 Return -1 if CHARACTER is out of range for BASE,
2336 and -2 if CHARACTER is not valid for any supported BASE. */
2337 static int
2338 digit_to_number (int character, int base)
2339 {
2340 int digit;
2341
2342 if ('0' <= character && character <= '9')
2343 digit = character - '0';
2344 else if ('a' <= character && character <= 'z')
2345 digit = character - 'a' + 10;
2346 else if ('A' <= character && character <= 'Z')
2347 digit = character - 'A' + 10;
2348 else
2349 return -2;
2350
2351 return digit < base ? digit : -1;
2352 }
2353
2354 /* Read an integer in radix RADIX using READCHARFUN to read
2355 characters. RADIX must be in the interval [2..36]; if it isn't, a
2356 read error is signaled . Value is the integer read. Signals an
2357 error if encountering invalid read syntax or if RADIX is out of
2358 range. */
2359
2360 static Lisp_Object
2361 read_integer (Lisp_Object readcharfun, EMACS_INT radix)
2362 {
2363 /* Room for sign, leading 0, other digits, trailing null byte.
2364 Also, room for invalid syntax diagnostic. */
2365 char buf[max (1 + 1 + sizeof (uintmax_t) * CHAR_BIT + 1,
2366 sizeof "integer, radix " + INT_STRLEN_BOUND (EMACS_INT))];
2367
2368 int valid = -1; /* 1 if valid, 0 if not, -1 if incomplete. */
2369
2370 if (radix < 2 || radix > 36)
2371 valid = 0;
2372 else
2373 {
2374 char *p = buf;
2375 int c, digit;
2376
2377 c = READCHAR;
2378 if (c == '-' || c == '+')
2379 {
2380 *p++ = c;
2381 c = READCHAR;
2382 }
2383
2384 if (c == '0')
2385 {
2386 *p++ = c;
2387 valid = 1;
2388
2389 /* Ignore redundant leading zeros, so the buffer doesn't
2390 fill up with them. */
2391 do
2392 c = READCHAR;
2393 while (c == '0');
2394 }
2395
2396 while ((digit = digit_to_number (c, radix)) >= -1)
2397 {
2398 if (digit == -1)
2399 valid = 0;
2400 if (valid < 0)
2401 valid = 1;
2402
2403 if (p < buf + sizeof buf - 1)
2404 *p++ = c;
2405 else
2406 valid = 0;
2407
2408 c = READCHAR;
2409 }
2410
2411 UNREAD (c);
2412 *p = '\0';
2413 }
2414
2415 if (! valid)
2416 {
2417 sprintf (buf, "integer, radix %"pI"d", radix);
2418 invalid_syntax (buf);
2419 }
2420
2421 return string_to_number (buf, radix, 0);
2422 }
2423
2424
2425 /* If the next token is ')' or ']' or '.', we store that character
2426 in *PCH and the return value is not interesting. Else, we store
2427 zero in *PCH and we read and return one lisp object.
2428
2429 FIRST_IN_LIST is true if this is the first element of a list. */
2430
2431 static Lisp_Object
2432 read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
2433 {
2434 int c;
2435 bool uninterned_symbol = 0;
2436 bool multibyte;
2437
2438 *pch = 0;
2439
2440 retry:
2441
2442 c = READCHAR_REPORT_MULTIBYTE (&multibyte);
2443 if (c < 0)
2444 end_of_file_error ();
2445
2446 switch (c)
2447 {
2448 case '(':
2449 return read_list (0, readcharfun);
2450
2451 case '[':
2452 return read_vector (readcharfun, 0);
2453
2454 case ')':
2455 case ']':
2456 {
2457 *pch = c;
2458 return Qnil;
2459 }
2460
2461 case '#':
2462 c = READCHAR;
2463 if (c == 's')
2464 {
2465 c = READCHAR;
2466 if (c == '(')
2467 {
2468 /* Accept extended format for hashtables (extensible to
2469 other types), e.g.
2470 #s(hash-table size 2 test equal data (k1 v1 k2 v2)) */
2471 Lisp_Object tmp = read_list (0, readcharfun);
2472 Lisp_Object head = CAR_SAFE (tmp);
2473 Lisp_Object data = Qnil;
2474 Lisp_Object val = Qnil;
2475 /* The size is 2 * number of allowed keywords to
2476 make-hash-table. */
2477 Lisp_Object params[10];
2478 Lisp_Object ht;
2479 Lisp_Object key = Qnil;
2480 int param_count = 0;
2481
2482 if (!EQ (head, Qhash_table))
2483 error ("Invalid extended read marker at head of #s list "
2484 "(only hash-table allowed)");
2485
2486 tmp = CDR_SAFE (tmp);
2487
2488 /* This is repetitive but fast and simple. */
2489 params[param_count] = QCsize;
2490 params[param_count + 1] = Fplist_get (tmp, Qsize);
2491 if (!NILP (params[param_count + 1]))
2492 param_count += 2;
2493
2494 params[param_count] = QCtest;
2495 params[param_count + 1] = Fplist_get (tmp, Qtest);
2496 if (!NILP (params[param_count + 1]))
2497 param_count += 2;
2498
2499 params[param_count] = QCweakness;
2500 params[param_count + 1] = Fplist_get (tmp, Qweakness);
2501 if (!NILP (params[param_count + 1]))
2502 param_count += 2;
2503
2504 params[param_count] = QCrehash_size;
2505 params[param_count + 1] = Fplist_get (tmp, Qrehash_size);
2506 if (!NILP (params[param_count + 1]))
2507 param_count += 2;
2508
2509 params[param_count] = QCrehash_threshold;
2510 params[param_count + 1] = Fplist_get (tmp, Qrehash_threshold);
2511 if (!NILP (params[param_count + 1]))
2512 param_count += 2;
2513
2514 /* This is the hashtable data. */
2515 data = Fplist_get (tmp, Qdata);
2516
2517 /* Now use params to make a new hashtable and fill it. */
2518 ht = Fmake_hash_table (param_count, params);
2519
2520 while (CONSP (data))
2521 {
2522 key = XCAR (data);
2523 data = XCDR (data);
2524 if (!CONSP (data))
2525 error ("Odd number of elements in hashtable data");
2526 val = XCAR (data);
2527 data = XCDR (data);
2528 Fputhash (key, val, ht);
2529 }
2530
2531 return ht;
2532 }
2533 UNREAD (c);
2534 invalid_syntax ("#");
2535 }
2536 if (c == '^')
2537 {
2538 c = READCHAR;
2539 if (c == '[')
2540 {
2541 Lisp_Object tmp;
2542 tmp = read_vector (readcharfun, 0);
2543 if (ASIZE (tmp) < CHAR_TABLE_STANDARD_SLOTS)
2544 error ("Invalid size char-table");
2545 XSETPVECTYPE (XVECTOR (tmp), PVEC_CHAR_TABLE);
2546 return tmp;
2547 }
2548 else if (c == '^')
2549 {
2550 c = READCHAR;
2551 if (c == '[')
2552 {
2553 /* Sub char-table can't be read as a regular
2554 vector because of a two C integer fields. */
2555 Lisp_Object tbl, tmp = read_list (1, readcharfun);
2556 ptrdiff_t size = XINT (Flength (tmp));
2557 int i, depth, min_char;
2558 struct Lisp_Cons *cell;
2559
2560 if (size == 0)
2561 error ("Zero-sized sub char-table");
2562
2563 if (! RANGED_INTEGERP (1, XCAR (tmp), 3))
2564 error ("Invalid depth in sub char-table");
2565 depth = XINT (XCAR (tmp));
2566 if (chartab_size[depth] != size - 2)
2567 error ("Invalid size in sub char-table");
2568 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2569 free_cons (cell);
2570
2571 if (! RANGED_INTEGERP (0, XCAR (tmp), MAX_CHAR))
2572 error ("Invalid minimum character in sub-char-table");
2573 min_char = XINT (XCAR (tmp));
2574 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2575 free_cons (cell);
2576
2577 tbl = make_uninit_sub_char_table (depth, min_char);
2578 for (i = 0; i < size; i++)
2579 {
2580 XSUB_CHAR_TABLE (tbl)->contents[i] = XCAR (tmp);
2581 cell = XCONS (tmp), tmp = XCDR (tmp);
2582 free_cons (cell);
2583 }
2584 return tbl;
2585 }
2586 invalid_syntax ("#^^");
2587 }
2588 invalid_syntax ("#^");
2589 }
2590 if (c == '&')
2591 {
2592 Lisp_Object length;
2593 length = read1 (readcharfun, pch, first_in_list);
2594 c = READCHAR;
2595 if (c == '"')
2596 {
2597 Lisp_Object tmp, val;
2598 EMACS_INT size_in_chars = bool_vector_bytes (XFASTINT (length));
2599 unsigned char *data;
2600
2601 UNREAD (c);
2602 tmp = read1 (readcharfun, pch, first_in_list);
2603 if (STRING_MULTIBYTE (tmp)
2604 || (size_in_chars != SCHARS (tmp)
2605 /* We used to print 1 char too many
2606 when the number of bits was a multiple of 8.
2607 Accept such input in case it came from an old
2608 version. */
2609 && ! (XFASTINT (length)
2610 == (SCHARS (tmp) - 1) * BOOL_VECTOR_BITS_PER_CHAR)))
2611 invalid_syntax ("#&...");
2612
2613 val = make_uninit_bool_vector (XFASTINT (length));
2614 data = bool_vector_uchar_data (val);
2615 memcpy (data, SDATA (tmp), size_in_chars);
2616 /* Clear the extraneous bits in the last byte. */
2617 if (XINT (length) != size_in_chars * BOOL_VECTOR_BITS_PER_CHAR)
2618 data[size_in_chars - 1]
2619 &= (1 << (XINT (length) % BOOL_VECTOR_BITS_PER_CHAR)) - 1;
2620 return val;
2621 }
2622 invalid_syntax ("#&...");
2623 }
2624 if (c == '[')
2625 {
2626 /* Accept compiled functions at read-time so that we don't have to
2627 build them using function calls. */
2628 Lisp_Object tmp;
2629 struct Lisp_Vector *vec;
2630 tmp = read_vector (readcharfun, 1);
2631 vec = XVECTOR (tmp);
2632 if (vec->header.size == 0)
2633 invalid_syntax ("Empty byte-code object");
2634 make_byte_code (vec);
2635 return tmp;
2636 }
2637 if (c == '(')
2638 {
2639 Lisp_Object tmp;
2640 int ch;
2641
2642 /* Read the string itself. */
2643 tmp = read1 (readcharfun, &ch, 0);
2644 if (ch != 0 || !STRINGP (tmp))
2645 invalid_syntax ("#");
2646 /* Read the intervals and their properties. */
2647 while (1)
2648 {
2649 Lisp_Object beg, end, plist;
2650
2651 beg = read1 (readcharfun, &ch, 0);
2652 end = plist = Qnil;
2653 if (ch == ')')
2654 break;
2655 if (ch == 0)
2656 end = read1 (readcharfun, &ch, 0);
2657 if (ch == 0)
2658 plist = read1 (readcharfun, &ch, 0);
2659 if (ch)
2660 invalid_syntax ("Invalid string property list");
2661 Fset_text_properties (beg, end, plist, tmp);
2662 }
2663
2664 return tmp;
2665 }
2666
2667 /* #@NUMBER is used to skip NUMBER following bytes.
2668 That's used in .elc files to skip over doc strings
2669 and function definitions. */
2670 if (c == '@')
2671 {
2672 enum { extra = 100 };
2673 ptrdiff_t i, nskip = 0, digits = 0;
2674
2675 /* Read a decimal integer. */
2676 while ((c = READCHAR) >= 0
2677 && c >= '0' && c <= '9')
2678 {
2679 if ((STRING_BYTES_BOUND - extra) / 10 <= nskip)
2680 string_overflow ();
2681 digits++;
2682 nskip *= 10;
2683 nskip += c - '0';
2684 if (digits == 2 && nskip == 0)
2685 { /* We've just seen #@00, which means "skip to end". */
2686 skip_dyn_eof (readcharfun);
2687 return Qnil;
2688 }
2689 }
2690 if (nskip > 0)
2691 /* We can't use UNREAD here, because in the code below we side-step
2692 READCHAR. Instead, assume the first char after #@NNN occupies
2693 a single byte, which is the case normally since it's just
2694 a space. */
2695 nskip--;
2696 else
2697 UNREAD (c);
2698
2699 if (load_force_doc_strings
2700 && (FROM_FILE_P (readcharfun)))
2701 {
2702 /* If we are supposed to force doc strings into core right now,
2703 record the last string that we skipped,
2704 and record where in the file it comes from. */
2705
2706 /* But first exchange saved_doc_string
2707 with prev_saved_doc_string, so we save two strings. */
2708 {
2709 char *temp = saved_doc_string;
2710 ptrdiff_t temp_size = saved_doc_string_size;
2711 file_offset temp_pos = saved_doc_string_position;
2712 ptrdiff_t temp_len = saved_doc_string_length;
2713
2714 saved_doc_string = prev_saved_doc_string;
2715 saved_doc_string_size = prev_saved_doc_string_size;
2716 saved_doc_string_position = prev_saved_doc_string_position;
2717 saved_doc_string_length = prev_saved_doc_string_length;
2718
2719 prev_saved_doc_string = temp;
2720 prev_saved_doc_string_size = temp_size;
2721 prev_saved_doc_string_position = temp_pos;
2722 prev_saved_doc_string_length = temp_len;
2723 }
2724
2725 if (saved_doc_string_size == 0)
2726 {
2727 saved_doc_string = xmalloc (nskip + extra);
2728 saved_doc_string_size = nskip + extra;
2729 }
2730 if (nskip > saved_doc_string_size)
2731 {
2732 saved_doc_string = xrealloc (saved_doc_string, nskip + extra);
2733 saved_doc_string_size = nskip + extra;
2734 }
2735
2736 saved_doc_string_position = file_tell (instream);
2737
2738 /* Copy that many characters into saved_doc_string. */
2739 block_input ();
2740 for (i = 0; i < nskip && c >= 0; i++)
2741 saved_doc_string[i] = c = getc (instream);
2742 unblock_input ();
2743
2744 saved_doc_string_length = i;
2745 }
2746 else
2747 /* Skip that many bytes. */
2748 skip_dyn_bytes (readcharfun, nskip);
2749
2750 goto retry;
2751 }
2752 if (c == '!')
2753 {
2754 /* #! appears at the beginning of an executable file.
2755 Skip the first line. */
2756 while (c != '\n' && c >= 0)
2757 c = READCHAR;
2758 goto retry;
2759 }
2760 if (c == '$')
2761 return Vload_file_name;
2762 if (c == '\'')
2763 return list2 (Qfunction, read0 (readcharfun));
2764 /* #:foo is the uninterned symbol named foo. */
2765 if (c == ':')
2766 {
2767 uninterned_symbol = 1;
2768 c = READCHAR;
2769 if (!(c > 040
2770 && c != NO_BREAK_SPACE
2771 && (c >= 0200
2772 || strchr ("\"';()[]#`,", c) == NULL)))
2773 {
2774 /* No symbol character follows, this is the empty
2775 symbol. */
2776 UNREAD (c);
2777 return Fmake_symbol (empty_unibyte_string);
2778 }
2779 goto read_symbol;
2780 }
2781 /* ## is the empty symbol. */
2782 if (c == '#')
2783 return Fintern (empty_unibyte_string, Qnil);
2784 /* Reader forms that can reuse previously read objects. */
2785 if (c >= '0' && c <= '9')
2786 {
2787 EMACS_INT n = 0;
2788 Lisp_Object tem;
2789
2790 /* Read a non-negative integer. */
2791 while (c >= '0' && c <= '9')
2792 {
2793 if (MOST_POSITIVE_FIXNUM / 10 < n
2794 || MOST_POSITIVE_FIXNUM < n * 10 + c - '0')
2795 n = MOST_POSITIVE_FIXNUM + 1;
2796 else
2797 n = n * 10 + c - '0';
2798 c = READCHAR;
2799 }
2800
2801 if (n <= MOST_POSITIVE_FIXNUM)
2802 {
2803 if (c == 'r' || c == 'R')
2804 return read_integer (readcharfun, n);
2805
2806 if (! NILP (Vread_circle))
2807 {
2808 /* #n=object returns object, but associates it with
2809 n for #n#. */
2810 if (c == '=')
2811 {
2812 /* Make a placeholder for #n# to use temporarily. */
2813 AUTO_CONS (placeholder, Qnil, Qnil);
2814 Lisp_Object cell = Fcons (make_number (n), placeholder);
2815 read_objects = Fcons (cell, read_objects);
2816
2817 /* Read the object itself. */
2818 tem = read0 (readcharfun);
2819
2820 /* Now put it everywhere the placeholder was... */
2821 substitute_object_in_subtree (tem, placeholder);
2822
2823 /* ...and #n# will use the real value from now on. */
2824 Fsetcdr (cell, tem);
2825
2826 return tem;
2827 }
2828
2829 /* #n# returns a previously read object. */
2830 if (c == '#')
2831 {
2832 tem = Fassq (make_number (n), read_objects);
2833 if (CONSP (tem))
2834 return XCDR (tem);
2835 }
2836 }
2837 }
2838 /* Fall through to error message. */
2839 }
2840 else if (c == 'x' || c == 'X')
2841 return read_integer (readcharfun, 16);
2842 else if (c == 'o' || c == 'O')
2843 return read_integer (readcharfun, 8);
2844 else if (c == 'b' || c == 'B')
2845 return read_integer (readcharfun, 2);
2846
2847 UNREAD (c);
2848 invalid_syntax ("#");
2849
2850 case ';':
2851 while ((c = READCHAR) >= 0 && c != '\n');
2852 goto retry;
2853
2854 case '\'':
2855 return list2 (Qquote, read0 (readcharfun));
2856
2857 case '`':
2858 {
2859 int next_char = READCHAR;
2860 UNREAD (next_char);
2861 /* Transition from old-style to new-style:
2862 If we see "(`" it used to mean old-style, which usually works
2863 fine because ` should almost never appear in such a position
2864 for new-style. But occasionally we need "(`" to mean new
2865 style, so we try to distinguish the two by the fact that we
2866 can either write "( `foo" or "(` foo", where the first
2867 intends to use new-style whereas the second intends to use
2868 old-style. For Emacs-25, we should completely remove this
2869 first_in_list exception (old-style can still be obtained via
2870 "(\`" anyway). */
2871 if (!new_backquote_flag && first_in_list && next_char == ' ')
2872 {
2873 Vold_style_backquotes = Qt;
2874 goto default_label;
2875 }
2876 else
2877 {
2878 Lisp_Object value;
2879 bool saved_new_backquote_flag = new_backquote_flag;
2880
2881 new_backquote_flag = 1;
2882 value = read0 (readcharfun);
2883 new_backquote_flag = saved_new_backquote_flag;
2884
2885 return list2 (Qbackquote, value);
2886 }
2887 }
2888 case ',':
2889 {
2890 int next_char = READCHAR;
2891 UNREAD (next_char);
2892 /* Transition from old-style to new-style:
2893 It used to be impossible to have a new-style , other than within
2894 a new-style `. This is sufficient when ` and , are used in the
2895 normal way, but ` and , can also appear in args to macros that
2896 will not interpret them in the usual way, in which case , may be
2897 used without any ` anywhere near.
2898 So we now use the same heuristic as for backquote: old-style
2899 unquotes are only recognized when first on a list, and when
2900 followed by a space.
2901 Because it's more difficult to peek 2 chars ahead, a new-style
2902 ,@ can still not be used outside of a `, unless it's in the middle
2903 of a list. */
2904 if (new_backquote_flag
2905 || !first_in_list
2906 || (next_char != ' ' && next_char != '@'))
2907 {
2908 Lisp_Object comma_type = Qnil;
2909 Lisp_Object value;
2910 int ch = READCHAR;
2911
2912 if (ch == '@')
2913 comma_type = Qcomma_at;
2914 else if (ch == '.')
2915 comma_type = Qcomma_dot;
2916 else
2917 {
2918 if (ch >= 0) UNREAD (ch);
2919 comma_type = Qcomma;
2920 }
2921
2922 value = read0 (readcharfun);
2923 return list2 (comma_type, value);
2924 }
2925 else
2926 {
2927 Vold_style_backquotes = Qt;
2928 goto default_label;
2929 }
2930 }
2931 case '?':
2932 {
2933 int modifiers;
2934 int next_char;
2935 bool ok;
2936
2937 c = READCHAR;
2938 if (c < 0)
2939 end_of_file_error ();
2940
2941 /* Accept `single space' syntax like (list ? x) where the
2942 whitespace character is SPC or TAB.
2943 Other literal whitespace like NL, CR, and FF are not accepted,
2944 as there are well-established escape sequences for these. */
2945 if (c == ' ' || c == '\t')
2946 return make_number (c);
2947
2948 if (c == '\\')
2949 c = read_escape (readcharfun, 0);
2950 modifiers = c & CHAR_MODIFIER_MASK;
2951 c &= ~CHAR_MODIFIER_MASK;
2952 if (CHAR_BYTE8_P (c))
2953 c = CHAR_TO_BYTE8 (c);
2954 c |= modifiers;
2955
2956 next_char = READCHAR;
2957 ok = (next_char <= 040
2958 || (next_char < 0200
2959 && strchr ("\"';()[]#?`,.", next_char) != NULL));
2960 UNREAD (next_char);
2961 if (ok)
2962 return make_number (c);
2963
2964 invalid_syntax ("?");
2965 }
2966
2967 case '"':
2968 {
2969 char *p = read_buffer;
2970 char *end = read_buffer + read_buffer_size;
2971 int ch;
2972 /* True if we saw an escape sequence specifying
2973 a multibyte character. */
2974 bool force_multibyte = 0;
2975 /* True if we saw an escape sequence specifying
2976 a single-byte character. */
2977 bool force_singlebyte = 0;
2978 bool cancel = 0;
2979 ptrdiff_t nchars = 0;
2980
2981 while ((ch = READCHAR) >= 0
2982 && ch != '\"')
2983 {
2984 if (end - p < MAX_MULTIBYTE_LENGTH)
2985 {
2986 ptrdiff_t offset = p - read_buffer;
2987 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
2988 memory_full (SIZE_MAX);
2989 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
2990 read_buffer_size *= 2;
2991 p = read_buffer + offset;
2992 end = read_buffer + read_buffer_size;
2993 }
2994
2995 if (ch == '\\')
2996 {
2997 int modifiers;
2998
2999 ch = read_escape (readcharfun, 1);
3000
3001 /* CH is -1 if \ newline or \ space has just been seen. */
3002 if (ch == -1)
3003 {
3004 if (p == read_buffer)
3005 cancel = 1;
3006 continue;
3007 }
3008
3009 modifiers = ch & CHAR_MODIFIER_MASK;
3010 ch = ch & ~CHAR_MODIFIER_MASK;
3011
3012 if (CHAR_BYTE8_P (ch))
3013 force_singlebyte = 1;
3014 else if (! ASCII_CHAR_P (ch))
3015 force_multibyte = 1;
3016 else /* I.e. ASCII_CHAR_P (ch). */
3017 {
3018 /* Allow `\C- ' and `\C-?'. */
3019 if (modifiers == CHAR_CTL)
3020 {
3021 if (ch == ' ')
3022 ch = 0, modifiers = 0;
3023 else if (ch == '?')
3024 ch = 127, modifiers = 0;
3025 }
3026 if (modifiers & CHAR_SHIFT)
3027 {
3028 /* Shift modifier is valid only with [A-Za-z]. */
3029 if (ch >= 'A' && ch <= 'Z')
3030 modifiers &= ~CHAR_SHIFT;
3031 else if (ch >= 'a' && ch <= 'z')
3032 ch -= ('a' - 'A'), modifiers &= ~CHAR_SHIFT;
3033 }
3034
3035 if (modifiers & CHAR_META)
3036 {
3037 /* Move the meta bit to the right place for a
3038 string. */
3039 modifiers &= ~CHAR_META;
3040 ch = BYTE8_TO_CHAR (ch | 0x80);
3041 force_singlebyte = 1;
3042 }
3043 }
3044
3045 /* Any modifiers remaining are invalid. */
3046 if (modifiers)
3047 error ("Invalid modifier in string");
3048 p += CHAR_STRING (ch, (unsigned char *) p);
3049 }
3050 else
3051 {
3052 p += CHAR_STRING (ch, (unsigned char *) p);
3053 if (CHAR_BYTE8_P (ch))
3054 force_singlebyte = 1;
3055 else if (! ASCII_CHAR_P (ch))
3056 force_multibyte = 1;
3057 }
3058 nchars++;
3059 }
3060
3061 if (ch < 0)
3062 end_of_file_error ();
3063
3064 /* If purifying, and string starts with \ newline,
3065 return zero instead. This is for doc strings
3066 that we are really going to find in etc/DOC.nn.nn. */
3067 if (!NILP (Vpurify_flag) && NILP (Vdoc_file_name) && cancel)
3068 return make_number (0);
3069
3070 if (! force_multibyte && force_singlebyte)
3071 {
3072 /* READ_BUFFER contains raw 8-bit bytes and no multibyte
3073 forms. Convert it to unibyte. */
3074 nchars = str_as_unibyte ((unsigned char *) read_buffer,
3075 p - read_buffer);
3076 p = read_buffer + nchars;
3077 }
3078
3079 return make_specified_string (read_buffer, nchars, p - read_buffer,
3080 (force_multibyte
3081 || (p - read_buffer != nchars)));
3082 }
3083
3084 case '.':
3085 {
3086 int next_char = READCHAR;
3087 UNREAD (next_char);
3088
3089 if (next_char <= 040
3090 || (next_char < 0200
3091 && strchr ("\"';([#?`,", next_char) != NULL))
3092 {
3093 *pch = c;
3094 return Qnil;
3095 }
3096
3097 /* Otherwise, we fall through! Note that the atom-reading loop
3098 below will now loop at least once, assuring that we will not
3099 try to UNREAD two characters in a row. */
3100 }
3101 default:
3102 default_label:
3103 if (c <= 040) goto retry;
3104 if (c == NO_BREAK_SPACE)
3105 goto retry;
3106
3107 read_symbol:
3108 {
3109 char *p = read_buffer;
3110 bool quoted = 0;
3111 EMACS_INT start_position = readchar_count - 1;
3112
3113 {
3114 char *end = read_buffer + read_buffer_size;
3115
3116 do
3117 {
3118 if (end - p < MAX_MULTIBYTE_LENGTH)
3119 {
3120 ptrdiff_t offset = p - read_buffer;
3121 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3122 memory_full (SIZE_MAX);
3123 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3124 read_buffer_size *= 2;
3125 p = read_buffer + offset;
3126 end = read_buffer + read_buffer_size;
3127 }
3128
3129 if (c == '\\')
3130 {
3131 c = READCHAR;
3132 if (c == -1)
3133 end_of_file_error ();
3134 quoted = 1;
3135 }
3136
3137 if (multibyte)
3138 p += CHAR_STRING (c, (unsigned char *) p);
3139 else
3140 *p++ = c;
3141 c = READCHAR;
3142 }
3143 while (c > 040
3144 && c != NO_BREAK_SPACE
3145 && (c >= 0200
3146 || strchr ("\"';()[]#`,", c) == NULL));
3147
3148 if (p == end)
3149 {
3150 ptrdiff_t offset = p - read_buffer;
3151 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3152 memory_full (SIZE_MAX);
3153 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3154 read_buffer_size *= 2;
3155 p = read_buffer + offset;
3156 end = read_buffer + read_buffer_size;
3157 }
3158 *p = 0;
3159 UNREAD (c);
3160 }
3161
3162 if (!quoted && !uninterned_symbol)
3163 {
3164 Lisp_Object result = string_to_number (read_buffer, 10, 0);
3165 if (! NILP (result))
3166 return result;
3167 }
3168 {
3169 Lisp_Object name, result;
3170 ptrdiff_t nbytes = p - read_buffer;
3171 ptrdiff_t nchars
3172 = (multibyte
3173 ? multibyte_chars_in_text ((unsigned char *) read_buffer,
3174 nbytes)
3175 : nbytes);
3176
3177 name = ((uninterned_symbol && ! NILP (Vpurify_flag)
3178 ? make_pure_string : make_specified_string)
3179 (read_buffer, nchars, nbytes, multibyte));
3180 result = (uninterned_symbol ? Fmake_symbol (name)
3181 : Fintern (name, Qnil));
3182
3183 if (EQ (Vread_with_symbol_positions, Qt)
3184 || EQ (Vread_with_symbol_positions, readcharfun))
3185 Vread_symbol_positions_list
3186 = Fcons (Fcons (result, make_number (start_position)),
3187 Vread_symbol_positions_list);
3188 return result;
3189 }
3190 }
3191 }
3192 }
3193 \f
3194
3195 /* List of nodes we've seen during substitute_object_in_subtree. */
3196 static Lisp_Object seen_list;
3197
3198 static void
3199 substitute_object_in_subtree (Lisp_Object object, Lisp_Object placeholder)
3200 {
3201 Lisp_Object check_object;
3202
3203 /* We haven't seen any objects when we start. */
3204 seen_list = Qnil;
3205
3206 /* Make all the substitutions. */
3207 check_object
3208 = substitute_object_recurse (object, placeholder, object);
3209
3210 /* Clear seen_list because we're done with it. */
3211 seen_list = Qnil;
3212
3213 /* The returned object here is expected to always eq the
3214 original. */
3215 if (!EQ (check_object, object))
3216 error ("Unexpected mutation error in reader");
3217 }
3218
3219 /* Feval doesn't get called from here, so no gc protection is needed. */
3220 #define SUBSTITUTE(get_val, set_val) \
3221 do { \
3222 Lisp_Object old_value = get_val; \
3223 Lisp_Object true_value \
3224 = substitute_object_recurse (object, placeholder, \
3225 old_value); \
3226 \
3227 if (!EQ (old_value, true_value)) \
3228 { \
3229 set_val; \
3230 } \
3231 } while (0)
3232
3233 static Lisp_Object
3234 substitute_object_recurse (Lisp_Object object, Lisp_Object placeholder, Lisp_Object subtree)
3235 {
3236 /* If we find the placeholder, return the target object. */
3237 if (EQ (placeholder, subtree))
3238 return object;
3239
3240 /* If we've been to this node before, don't explore it again. */
3241 if (!EQ (Qnil, Fmemq (subtree, seen_list)))
3242 return subtree;
3243
3244 /* If this node can be the entry point to a cycle, remember that
3245 we've seen it. It can only be such an entry point if it was made
3246 by #n=, which means that we can find it as a value in
3247 read_objects. */
3248 if (!EQ (Qnil, Frassq (subtree, read_objects)))
3249 seen_list = Fcons (subtree, seen_list);
3250
3251 /* Recurse according to subtree's type.
3252 Every branch must return a Lisp_Object. */
3253 switch (XTYPE (subtree))
3254 {
3255 case Lisp_Vectorlike:
3256 {
3257 ptrdiff_t i = 0, length = 0;
3258 if (BOOL_VECTOR_P (subtree))
3259 return subtree; /* No sub-objects anyway. */
3260 else if (CHAR_TABLE_P (subtree) || SUB_CHAR_TABLE_P (subtree)
3261 || COMPILEDP (subtree) || HASH_TABLE_P (subtree))
3262 length = ASIZE (subtree) & PSEUDOVECTOR_SIZE_MASK;
3263 else if (VECTORP (subtree))
3264 length = ASIZE (subtree);
3265 else
3266 /* An unknown pseudovector may contain non-Lisp fields, so we
3267 can't just blindly traverse all its fields. We used to call
3268 `Flength' which signaled `sequencep', so I just preserved this
3269 behavior. */
3270 wrong_type_argument (Qsequencep, subtree);
3271
3272 if (SUB_CHAR_TABLE_P (subtree))
3273 i = 2;
3274 for ( ; i < length; i++)
3275 SUBSTITUTE (AREF (subtree, i),
3276 ASET (subtree, i, true_value));
3277 return subtree;
3278 }
3279
3280 case Lisp_Cons:
3281 {
3282 SUBSTITUTE (XCAR (subtree),
3283 XSETCAR (subtree, true_value));
3284 SUBSTITUTE (XCDR (subtree),
3285 XSETCDR (subtree, true_value));
3286 return subtree;
3287 }
3288
3289 case Lisp_String:
3290 {
3291 /* Check for text properties in each interval.
3292 substitute_in_interval contains part of the logic. */
3293
3294 INTERVAL root_interval = string_intervals (subtree);
3295 AUTO_CONS (arg, object, placeholder);
3296
3297 traverse_intervals_noorder (root_interval,
3298 &substitute_in_interval, arg);
3299
3300 return subtree;
3301 }
3302
3303 /* Other types don't recurse any further. */
3304 default:
3305 return subtree;
3306 }
3307 }
3308
3309 /* Helper function for substitute_object_recurse. */
3310 static void
3311 substitute_in_interval (INTERVAL interval, Lisp_Object arg)
3312 {
3313 Lisp_Object object = Fcar (arg);
3314 Lisp_Object placeholder = Fcdr (arg);
3315
3316 SUBSTITUTE (interval->plist, set_interval_plist (interval, true_value));
3317 }
3318
3319 \f
3320 #define LEAD_INT 1
3321 #define DOT_CHAR 2
3322 #define TRAIL_INT 4
3323 #define E_EXP 16
3324
3325
3326 /* Convert STRING to a number, assuming base BASE. Return a fixnum if CP has
3327 integer syntax and fits in a fixnum, else return the nearest float if CP has
3328 either floating point or integer syntax and BASE is 10, else return nil. If
3329 IGNORE_TRAILING, consider just the longest prefix of CP that has
3330 valid floating point syntax. Signal an overflow if BASE is not 10 and the
3331 number has integer syntax but does not fit. */
3332
3333 Lisp_Object
3334 string_to_number (char const *string, int base, bool ignore_trailing)
3335 {
3336 int state;
3337 char const *cp = string;
3338 int leading_digit;
3339 bool float_syntax = 0;
3340 double value = 0;
3341
3342 /* Negate the value ourselves. This treats 0, NaNs, and infinity properly on
3343 IEEE floating point hosts, and works around a formerly-common bug where
3344 atof ("-0.0") drops the sign. */
3345 bool negative = *cp == '-';
3346
3347 bool signedp = negative || *cp == '+';
3348 cp += signedp;
3349
3350 state = 0;
3351
3352 leading_digit = digit_to_number (*cp, base);
3353 if (leading_digit >= 0)
3354 {
3355 state |= LEAD_INT;
3356 do
3357 ++cp;
3358 while (digit_to_number (*cp, base) >= 0);
3359 }
3360 if (*cp == '.')
3361 {
3362 state |= DOT_CHAR;
3363 cp++;
3364 }
3365
3366 if (base == 10)
3367 {
3368 if ('0' <= *cp && *cp <= '9')
3369 {
3370 state |= TRAIL_INT;
3371 do
3372 cp++;
3373 while ('0' <= *cp && *cp <= '9');
3374 }
3375 if (*cp == 'e' || *cp == 'E')
3376 {
3377 char const *ecp = cp;
3378 cp++;
3379 if (*cp == '+' || *cp == '-')
3380 cp++;
3381 if ('0' <= *cp && *cp <= '9')
3382 {
3383 state |= E_EXP;
3384 do
3385 cp++;
3386 while ('0' <= *cp && *cp <= '9');
3387 }
3388 else if (cp[-1] == '+'
3389 && cp[0] == 'I' && cp[1] == 'N' && cp[2] == 'F')
3390 {
3391 state |= E_EXP;
3392 cp += 3;
3393 value = INFINITY;
3394 }
3395 else if (cp[-1] == '+'
3396 && cp[0] == 'N' && cp[1] == 'a' && cp[2] == 'N')
3397 {
3398 state |= E_EXP;
3399 cp += 3;
3400 /* NAN is a "positive" NaN on all known Emacs hosts. */
3401 value = NAN;
3402 }
3403 else
3404 cp = ecp;
3405 }
3406
3407 float_syntax = ((state & (DOT_CHAR|TRAIL_INT)) == (DOT_CHAR|TRAIL_INT)
3408 || state == (LEAD_INT|E_EXP));
3409 }
3410
3411 /* Return nil if the number uses invalid syntax. If IGNORE_TRAILING, accept
3412 any prefix that matches. Otherwise, the entire string must match. */
3413 if (! (ignore_trailing
3414 ? ((state & LEAD_INT) != 0 || float_syntax)
3415 : (!*cp && ((state & ~DOT_CHAR) == LEAD_INT || float_syntax))))
3416 return Qnil;
3417
3418 /* If the number uses integer and not float syntax, and is in C-language
3419 range, use its value, preferably as a fixnum. */
3420 if (leading_digit >= 0 && ! float_syntax)
3421 {
3422 uintmax_t n;
3423
3424 /* Fast special case for single-digit integers. This also avoids a
3425 glitch when BASE is 16 and IGNORE_TRAILING, because in that
3426 case some versions of strtoumax accept numbers like "0x1" that Emacs
3427 does not allow. */
3428 if (digit_to_number (string[signedp + 1], base) < 0)
3429 return make_number (negative ? -leading_digit : leading_digit);
3430
3431 errno = 0;
3432 n = strtoumax (string + signedp, NULL, base);
3433 if (errno == ERANGE)
3434 {
3435 /* Unfortunately there's no simple and accurate way to convert
3436 non-base-10 numbers that are out of C-language range. */
3437 if (base != 10)
3438 xsignal1 (Qoverflow_error, build_string (string));
3439 }
3440 else if (n <= (negative ? -MOST_NEGATIVE_FIXNUM : MOST_POSITIVE_FIXNUM))
3441 {
3442 EMACS_INT signed_n = n;
3443 return make_number (negative ? -signed_n : signed_n);
3444 }
3445 else
3446 value = n;
3447 }
3448
3449 /* Either the number uses float syntax, or it does not fit into a fixnum.
3450 Convert it from string to floating point, unless the value is already
3451 known because it is an infinity, a NAN, or its absolute value fits in
3452 uintmax_t. */
3453 if (! value)
3454 value = atof (string + signedp);
3455
3456 return make_float (negative ? -value : value);
3457 }
3458
3459 \f
3460 static Lisp_Object
3461 read_vector (Lisp_Object readcharfun, bool bytecodeflag)
3462 {
3463 ptrdiff_t i, size;
3464 Lisp_Object *ptr;
3465 Lisp_Object tem, item, vector;
3466 struct Lisp_Cons *otem;
3467 Lisp_Object len;
3468
3469 tem = read_list (1, readcharfun);
3470 len = Flength (tem);
3471 vector = Fmake_vector (len, Qnil);
3472
3473 size = ASIZE (vector);
3474 ptr = XVECTOR (vector)->contents;
3475 for (i = 0; i < size; i++)
3476 {
3477 item = Fcar (tem);
3478 /* If `load-force-doc-strings' is t when reading a lazily-loaded
3479 bytecode object, the docstring containing the bytecode and
3480 constants values must be treated as unibyte and passed to
3481 Fread, to get the actual bytecode string and constants vector. */
3482 if (bytecodeflag && load_force_doc_strings)
3483 {
3484 if (i == COMPILED_BYTECODE)
3485 {
3486 if (!STRINGP (item))
3487 error ("Invalid byte code");
3488
3489 /* Delay handling the bytecode slot until we know whether
3490 it is lazily-loaded (we can tell by whether the
3491 constants slot is nil). */
3492 ASET (vector, COMPILED_CONSTANTS, item);
3493 item = Qnil;
3494 }
3495 else if (i == COMPILED_CONSTANTS)
3496 {
3497 Lisp_Object bytestr = ptr[COMPILED_CONSTANTS];
3498
3499 if (NILP (item))
3500 {
3501 /* Coerce string to unibyte (like string-as-unibyte,
3502 but without generating extra garbage and
3503 guaranteeing no change in the contents). */
3504 STRING_SET_CHARS (bytestr, SBYTES (bytestr));
3505 STRING_SET_UNIBYTE (bytestr);
3506
3507 item = Fread (Fcons (bytestr, readcharfun));
3508 if (!CONSP (item))
3509 error ("Invalid byte code");
3510
3511 otem = XCONS (item);
3512 bytestr = XCAR (item);
3513 item = XCDR (item);
3514 free_cons (otem);
3515 }
3516
3517 /* Now handle the bytecode slot. */
3518 ASET (vector, COMPILED_BYTECODE, bytestr);
3519 }
3520 else if (i == COMPILED_DOC_STRING
3521 && STRINGP (item)
3522 && ! STRING_MULTIBYTE (item))
3523 {
3524 if (EQ (readcharfun, Qget_emacs_mule_file_char))
3525 item = Fdecode_coding_string (item, Qemacs_mule, Qnil, Qnil);
3526 else
3527 item = Fstring_as_multibyte (item);
3528 }
3529 }
3530 ASET (vector, i, item);
3531 otem = XCONS (tem);
3532 tem = Fcdr (tem);
3533 free_cons (otem);
3534 }
3535 return vector;
3536 }
3537
3538 /* FLAG means check for ']' to terminate rather than ')' and '.'. */
3539
3540 static Lisp_Object
3541 read_list (bool flag, Lisp_Object readcharfun)
3542 {
3543 Lisp_Object val, tail;
3544 Lisp_Object elt, tem;
3545 /* 0 is the normal case.
3546 1 means this list is a doc reference; replace it with the number 0.
3547 2 means this list is a doc reference; replace it with the doc string. */
3548 int doc_reference = 0;
3549
3550 /* Initialize this to 1 if we are reading a list. */
3551 bool first_in_list = flag <= 0;
3552
3553 val = Qnil;
3554 tail = Qnil;
3555
3556 while (1)
3557 {
3558 int ch;
3559 elt = read1 (readcharfun, &ch, first_in_list);
3560
3561 first_in_list = 0;
3562
3563 /* While building, if the list starts with #$, treat it specially. */
3564 if (EQ (elt, Vload_file_name)
3565 && ! NILP (elt)
3566 && !NILP (Vpurify_flag))
3567 {
3568 if (NILP (Vdoc_file_name))
3569 /* We have not yet called Snarf-documentation, so assume
3570 this file is described in the DOC file
3571 and Snarf-documentation will fill in the right value later.
3572 For now, replace the whole list with 0. */
3573 doc_reference = 1;
3574 else
3575 /* We have already called Snarf-documentation, so make a relative
3576 file name for this file, so it can be found properly
3577 in the installed Lisp directory.
3578 We don't use Fexpand_file_name because that would make
3579 the directory absolute now. */
3580 {
3581 AUTO_STRING (dot_dot_lisp, "../lisp/");
3582 elt = concat2 (dot_dot_lisp, Ffile_name_nondirectory (elt));
3583 }
3584 }
3585 else if (EQ (elt, Vload_file_name)
3586 && ! NILP (elt)
3587 && load_force_doc_strings)
3588 doc_reference = 2;
3589
3590 if (ch)
3591 {
3592 if (flag > 0)
3593 {
3594 if (ch == ']')
3595 return val;
3596 invalid_syntax (") or . in a vector");
3597 }
3598 if (ch == ')')
3599 return val;
3600 if (ch == '.')
3601 {
3602 if (!NILP (tail))
3603 XSETCDR (tail, read0 (readcharfun));
3604 else
3605 val = read0 (readcharfun);
3606 read1 (readcharfun, &ch, 0);
3607
3608 if (ch == ')')
3609 {
3610 if (doc_reference == 1)
3611 return make_number (0);
3612 if (doc_reference == 2 && INTEGERP (XCDR (val)))
3613 {
3614 char *saved = NULL;
3615 file_offset saved_position;
3616 /* Get a doc string from the file we are loading.
3617 If it's in saved_doc_string, get it from there.
3618
3619 Here, we don't know if the string is a
3620 bytecode string or a doc string. As a
3621 bytecode string must be unibyte, we always
3622 return a unibyte string. If it is actually a
3623 doc string, caller must make it
3624 multibyte. */
3625
3626 /* Position is negative for user variables. */
3627 EMACS_INT pos = eabs (XINT (XCDR (val)));
3628 if (pos >= saved_doc_string_position
3629 && pos < (saved_doc_string_position
3630 + saved_doc_string_length))
3631 {
3632 saved = saved_doc_string;
3633 saved_position = saved_doc_string_position;
3634 }
3635 /* Look in prev_saved_doc_string the same way. */
3636 else if (pos >= prev_saved_doc_string_position
3637 && pos < (prev_saved_doc_string_position
3638 + prev_saved_doc_string_length))
3639 {
3640 saved = prev_saved_doc_string;
3641 saved_position = prev_saved_doc_string_position;
3642 }
3643 if (saved)
3644 {
3645 ptrdiff_t start = pos - saved_position;
3646 ptrdiff_t from, to;
3647
3648 /* Process quoting with ^A,
3649 and find the end of the string,
3650 which is marked with ^_ (037). */
3651 for (from = start, to = start;
3652 saved[from] != 037;)
3653 {
3654 int c = saved[from++];
3655 if (c == 1)
3656 {
3657 c = saved[from++];
3658 saved[to++] = (c == 1 ? c
3659 : c == '0' ? 0
3660 : c == '_' ? 037
3661 : c);
3662 }
3663 else
3664 saved[to++] = c;
3665 }
3666
3667 return make_unibyte_string (saved + start,
3668 to - start);
3669 }
3670 else
3671 return get_doc_string (val, 1, 0);
3672 }
3673
3674 return val;
3675 }
3676 invalid_syntax (". in wrong context");
3677 }
3678 invalid_syntax ("] in a list");
3679 }
3680 tem = list1 (elt);
3681 if (!NILP (tail))
3682 XSETCDR (tail, tem);
3683 else
3684 val = tem;
3685 tail = tem;
3686 }
3687 }
3688 \f
3689 static Lisp_Object initial_obarray;
3690
3691 /* `oblookup' stores the bucket number here, for the sake of Funintern. */
3692
3693 static size_t oblookup_last_bucket_number;
3694
3695 /* Get an error if OBARRAY is not an obarray.
3696 If it is one, return it. */
3697
3698 Lisp_Object
3699 check_obarray (Lisp_Object obarray)
3700 {
3701 if (!VECTORP (obarray) || ASIZE (obarray) == 0)
3702 {
3703 /* If Vobarray is now invalid, force it to be valid. */
3704 if (EQ (Vobarray, obarray)) Vobarray = initial_obarray;
3705 wrong_type_argument (Qvectorp, obarray);
3706 }
3707 return obarray;
3708 }
3709
3710 /* Intern symbol SYM in OBARRAY using bucket INDEX. */
3711
3712 static Lisp_Object
3713 intern_sym (Lisp_Object sym, Lisp_Object obarray, Lisp_Object index)
3714 {
3715 Lisp_Object *ptr;
3716
3717 XSYMBOL (sym)->interned = (EQ (obarray, initial_obarray)
3718 ? SYMBOL_INTERNED_IN_INITIAL_OBARRAY
3719 : SYMBOL_INTERNED);
3720
3721 if (SREF (SYMBOL_NAME (sym), 0) == ':' && EQ (obarray, initial_obarray))
3722 {
3723 XSYMBOL (sym)->constant = 1;
3724 XSYMBOL (sym)->redirect = SYMBOL_PLAINVAL;
3725 SET_SYMBOL_VAL (XSYMBOL (sym), sym);
3726 }
3727
3728 ptr = aref_addr (obarray, XINT (index));
3729 set_symbol_next (sym, SYMBOLP (*ptr) ? XSYMBOL (*ptr) : NULL);
3730 *ptr = sym;
3731 return sym;
3732 }
3733
3734 /* Intern a symbol with name STRING in OBARRAY using bucket INDEX. */
3735
3736 Lisp_Object
3737 intern_driver (Lisp_Object string, Lisp_Object obarray, Lisp_Object index)
3738 {
3739 return intern_sym (Fmake_symbol (string), obarray, index);
3740 }
3741
3742 /* Intern the C string STR: return a symbol with that name,
3743 interned in the current obarray. */
3744
3745 Lisp_Object
3746 intern_1 (const char *str, ptrdiff_t len)
3747 {
3748 Lisp_Object obarray = check_obarray (Vobarray);
3749 Lisp_Object tem = oblookup (obarray, str, len, len);
3750
3751 return (SYMBOLP (tem) ? tem
3752 /* The above `oblookup' was done on the basis of nchars==nbytes, so
3753 the string has to be unibyte. */
3754 : intern_driver (make_unibyte_string (str, len),
3755 obarray, tem));
3756 }
3757
3758 Lisp_Object
3759 intern_c_string_1 (const char *str, ptrdiff_t len)
3760 {
3761 Lisp_Object obarray = check_obarray (Vobarray);
3762 Lisp_Object tem = oblookup (obarray, str, len, len);
3763
3764 if (!SYMBOLP (tem))
3765 {
3766 /* Creating a non-pure string from a string literal not implemented yet.
3767 We could just use make_string here and live with the extra copy. */
3768 eassert (!NILP (Vpurify_flag));
3769 tem = intern_driver (make_pure_c_string (str, len), obarray, tem);
3770 }
3771 return tem;
3772 }
3773
3774 static void
3775 define_symbol (Lisp_Object sym, char const *str)
3776 {
3777 ptrdiff_t len = strlen (str);
3778 Lisp_Object string = make_pure_c_string (str, len);
3779 init_symbol (sym, string);
3780
3781 /* Qunbound is uninterned, so that it's not confused with any symbol
3782 'unbound' created by a Lisp program. */
3783 if (! EQ (sym, Qunbound))
3784 {
3785 Lisp_Object bucket = oblookup (initial_obarray, str, len, len);
3786 eassert (INTEGERP (bucket));
3787 intern_sym (sym, initial_obarray, bucket);
3788 }
3789 }
3790 \f
3791 DEFUN ("intern", Fintern, Sintern, 1, 2, 0,
3792 doc: /* Return the canonical symbol whose name is STRING.
3793 If there is none, one is created by this function and returned.
3794 A second optional argument specifies the obarray to use;
3795 it defaults to the value of `obarray'. */)
3796 (Lisp_Object string, Lisp_Object obarray)
3797 {
3798 Lisp_Object tem;
3799
3800 obarray = check_obarray (NILP (obarray) ? Vobarray : obarray);
3801 CHECK_STRING (string);
3802
3803 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3804 if (!SYMBOLP (tem))
3805 tem = intern_driver (NILP (Vpurify_flag) ? string : Fpurecopy (string),
3806 obarray, tem);
3807 return tem;
3808 }
3809
3810 DEFUN ("intern-soft", Fintern_soft, Sintern_soft, 1, 2, 0,
3811 doc: /* Return the canonical symbol named NAME, or nil if none exists.
3812 NAME may be a string or a symbol. If it is a symbol, that exact
3813 symbol is searched for.
3814 A second optional argument specifies the obarray to use;
3815 it defaults to the value of `obarray'. */)
3816 (Lisp_Object name, Lisp_Object obarray)
3817 {
3818 register Lisp_Object tem, string;
3819
3820 if (NILP (obarray)) obarray = Vobarray;
3821 obarray = check_obarray (obarray);
3822
3823 if (!SYMBOLP (name))
3824 {
3825 CHECK_STRING (name);
3826 string = name;
3827 }
3828 else
3829 string = SYMBOL_NAME (name);
3830
3831 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3832 if (INTEGERP (tem) || (SYMBOLP (name) && !EQ (name, tem)))
3833 return Qnil;
3834 else
3835 return tem;
3836 }
3837 \f
3838 DEFUN ("unintern", Funintern, Sunintern, 1, 2, 0,
3839 doc: /* Delete the symbol named NAME, if any, from OBARRAY.
3840 The value is t if a symbol was found and deleted, nil otherwise.
3841 NAME may be a string or a symbol. If it is a symbol, that symbol
3842 is deleted, if it belongs to OBARRAY--no other symbol is deleted.
3843 OBARRAY, if nil, defaults to the value of the variable `obarray'.
3844 usage: (unintern NAME OBARRAY) */)
3845 (Lisp_Object name, Lisp_Object obarray)
3846 {
3847 register Lisp_Object string, tem;
3848 size_t hash;
3849
3850 if (NILP (obarray)) obarray = Vobarray;
3851 obarray = check_obarray (obarray);
3852
3853 if (SYMBOLP (name))
3854 string = SYMBOL_NAME (name);
3855 else
3856 {
3857 CHECK_STRING (name);
3858 string = name;
3859 }
3860
3861 tem = oblookup (obarray, SSDATA (string),
3862 SCHARS (string),
3863 SBYTES (string));
3864 if (INTEGERP (tem))
3865 return Qnil;
3866 /* If arg was a symbol, don't delete anything but that symbol itself. */
3867 if (SYMBOLP (name) && !EQ (name, tem))
3868 return Qnil;
3869
3870 /* There are plenty of other symbols which will screw up the Emacs
3871 session if we unintern them, as well as even more ways to use
3872 `setq' or `fset' or whatnot to make the Emacs session
3873 unusable. Let's not go down this silly road. --Stef */
3874 /* if (EQ (tem, Qnil) || EQ (tem, Qt))
3875 error ("Attempt to unintern t or nil"); */
3876
3877 XSYMBOL (tem)->interned = SYMBOL_UNINTERNED;
3878
3879 hash = oblookup_last_bucket_number;
3880
3881 if (EQ (AREF (obarray, hash), tem))
3882 {
3883 if (XSYMBOL (tem)->next)
3884 {
3885 Lisp_Object sym;
3886 XSETSYMBOL (sym, XSYMBOL (tem)->next);
3887 ASET (obarray, hash, sym);
3888 }
3889 else
3890 ASET (obarray, hash, make_number (0));
3891 }
3892 else
3893 {
3894 Lisp_Object tail, following;
3895
3896 for (tail = AREF (obarray, hash);
3897 XSYMBOL (tail)->next;
3898 tail = following)
3899 {
3900 XSETSYMBOL (following, XSYMBOL (tail)->next);
3901 if (EQ (following, tem))
3902 {
3903 set_symbol_next (tail, XSYMBOL (following)->next);
3904 break;
3905 }
3906 }
3907 }
3908
3909 return Qt;
3910 }
3911 \f
3912 /* Return the symbol in OBARRAY whose names matches the string
3913 of SIZE characters (SIZE_BYTE bytes) at PTR.
3914 If there is no such symbol, return the integer bucket number of
3915 where the symbol would be if it were present.
3916
3917 Also store the bucket number in oblookup_last_bucket_number. */
3918
3919 Lisp_Object
3920 oblookup (Lisp_Object obarray, register const char *ptr, ptrdiff_t size, ptrdiff_t size_byte)
3921 {
3922 size_t hash;
3923 size_t obsize;
3924 register Lisp_Object tail;
3925 Lisp_Object bucket, tem;
3926
3927 obarray = check_obarray (obarray);
3928 obsize = ASIZE (obarray);
3929
3930 /* This is sometimes needed in the middle of GC. */
3931 obsize &= ~ARRAY_MARK_FLAG;
3932 hash = hash_string (ptr, size_byte) % obsize;
3933 bucket = AREF (obarray, hash);
3934 oblookup_last_bucket_number = hash;
3935 if (EQ (bucket, make_number (0)))
3936 ;
3937 else if (!SYMBOLP (bucket))
3938 error ("Bad data in guts of obarray"); /* Like CADR error message. */
3939 else
3940 for (tail = bucket; ; XSETSYMBOL (tail, XSYMBOL (tail)->next))
3941 {
3942 if (SBYTES (SYMBOL_NAME (tail)) == size_byte
3943 && SCHARS (SYMBOL_NAME (tail)) == size
3944 && !memcmp (SDATA (SYMBOL_NAME (tail)), ptr, size_byte))
3945 return tail;
3946 else if (XSYMBOL (tail)->next == 0)
3947 break;
3948 }
3949 XSETINT (tem, hash);
3950 return tem;
3951 }
3952 \f
3953 void
3954 map_obarray (Lisp_Object obarray, void (*fn) (Lisp_Object, Lisp_Object), Lisp_Object arg)
3955 {
3956 ptrdiff_t i;
3957 register Lisp_Object tail;
3958 CHECK_VECTOR (obarray);
3959 for (i = ASIZE (obarray) - 1; i >= 0; i--)
3960 {
3961 tail = AREF (obarray, i);
3962 if (SYMBOLP (tail))
3963 while (1)
3964 {
3965 (*fn) (tail, arg);
3966 if (XSYMBOL (tail)->next == 0)
3967 break;
3968 XSETSYMBOL (tail, XSYMBOL (tail)->next);
3969 }
3970 }
3971 }
3972
3973 static void
3974 mapatoms_1 (Lisp_Object sym, Lisp_Object function)
3975 {
3976 call1 (function, sym);
3977 }
3978
3979 DEFUN ("mapatoms", Fmapatoms, Smapatoms, 1, 2, 0,
3980 doc: /* Call FUNCTION on every symbol in OBARRAY.
3981 OBARRAY defaults to the value of `obarray'. */)
3982 (Lisp_Object function, Lisp_Object obarray)
3983 {
3984 if (NILP (obarray)) obarray = Vobarray;
3985 obarray = check_obarray (obarray);
3986
3987 map_obarray (obarray, mapatoms_1, function);
3988 return Qnil;
3989 }
3990
3991 #define OBARRAY_SIZE 1511
3992
3993 void
3994 init_obarray (void)
3995 {
3996 Lisp_Object oblength;
3997 ptrdiff_t size = 100 + MAX_MULTIBYTE_LENGTH;
3998
3999 XSETFASTINT (oblength, OBARRAY_SIZE);
4000
4001 Vobarray = Fmake_vector (oblength, make_number (0));
4002 initial_obarray = Vobarray;
4003 staticpro (&initial_obarray);
4004
4005 for (int i = 0; i < ARRAYELTS (lispsym); i++)
4006 define_symbol (builtin_lisp_symbol (i), defsym_name[i]);
4007
4008 DEFSYM (Qunbound, "unbound");
4009
4010 DEFSYM (Qnil, "nil");
4011 SET_SYMBOL_VAL (XSYMBOL (Qnil), Qnil);
4012 XSYMBOL (Qnil)->constant = 1;
4013 XSYMBOL (Qnil)->declared_special = true;
4014
4015 DEFSYM (Qt, "t");
4016 SET_SYMBOL_VAL (XSYMBOL (Qt), Qt);
4017 XSYMBOL (Qt)->constant = 1;
4018 XSYMBOL (Qt)->declared_special = true;
4019
4020 /* Qt is correct even if CANNOT_DUMP. loadup.el will set to nil at end. */
4021 Vpurify_flag = Qt;
4022
4023 DEFSYM (Qvariable_documentation, "variable-documentation");
4024
4025 read_buffer = xmalloc (size);
4026 read_buffer_size = size;
4027 }
4028 \f
4029 void
4030 defsubr (struct Lisp_Subr *sname)
4031 {
4032 Lisp_Object sym, tem;
4033 sym = intern_c_string (sname->symbol_name);
4034 XSETPVECTYPE (sname, PVEC_SUBR);
4035 XSETSUBR (tem, sname);
4036 set_symbol_function (sym, tem);
4037 }
4038
4039 #ifdef NOTDEF /* Use fset in subr.el now! */
4040 void
4041 defalias (struct Lisp_Subr *sname, char *string)
4042 {
4043 Lisp_Object sym;
4044 sym = intern (string);
4045 XSETSUBR (XSYMBOL (sym)->function, sname);
4046 }
4047 #endif /* NOTDEF */
4048
4049 /* Define an "integer variable"; a symbol whose value is forwarded to a
4050 C variable of type EMACS_INT. Sample call (with "xx" to fool make-docfile):
4051 DEFxxVAR_INT ("emacs-priority", &emacs_priority, "Documentation"); */
4052 void
4053 defvar_int (struct Lisp_Intfwd *i_fwd,
4054 const char *namestring, EMACS_INT *address)
4055 {
4056 Lisp_Object sym;
4057 sym = intern_c_string (namestring);
4058 i_fwd->type = Lisp_Fwd_Int;
4059 i_fwd->intvar = address;
4060 XSYMBOL (sym)->declared_special = 1;
4061 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4062 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)i_fwd);
4063 }
4064
4065 /* Similar but define a variable whose value is t if address contains 1,
4066 nil if address contains 0. */
4067 void
4068 defvar_bool (struct Lisp_Boolfwd *b_fwd,
4069 const char *namestring, bool *address)
4070 {
4071 Lisp_Object sym;
4072 sym = intern_c_string (namestring);
4073 b_fwd->type = Lisp_Fwd_Bool;
4074 b_fwd->boolvar = address;
4075 XSYMBOL (sym)->declared_special = 1;
4076 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4077 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)b_fwd);
4078 Vbyte_boolean_vars = Fcons (sym, Vbyte_boolean_vars);
4079 }
4080
4081 /* Similar but define a variable whose value is the Lisp Object stored
4082 at address. Two versions: with and without gc-marking of the C
4083 variable. The nopro version is used when that variable will be
4084 gc-marked for some other reason, since marking the same slot twice
4085 can cause trouble with strings. */
4086 void
4087 defvar_lisp_nopro (struct Lisp_Objfwd *o_fwd,
4088 const char *namestring, Lisp_Object *address)
4089 {
4090 Lisp_Object sym;
4091 sym = intern_c_string (namestring);
4092 o_fwd->type = Lisp_Fwd_Obj;
4093 o_fwd->objvar = address;
4094 XSYMBOL (sym)->declared_special = 1;
4095 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4096 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)o_fwd);
4097 }
4098
4099 void
4100 defvar_lisp (struct Lisp_Objfwd *o_fwd,
4101 const char *namestring, Lisp_Object *address)
4102 {
4103 defvar_lisp_nopro (o_fwd, namestring, address);
4104 staticpro (address);
4105 }
4106
4107 /* Similar but define a variable whose value is the Lisp Object stored
4108 at a particular offset in the current kboard object. */
4109
4110 void
4111 defvar_kboard (struct Lisp_Kboard_Objfwd *ko_fwd,
4112 const char *namestring, int offset)
4113 {
4114 Lisp_Object sym;
4115 sym = intern_c_string (namestring);
4116 ko_fwd->type = Lisp_Fwd_Kboard_Obj;
4117 ko_fwd->offset = offset;
4118 XSYMBOL (sym)->declared_special = 1;
4119 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4120 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)ko_fwd);
4121 }
4122 \f
4123 /* Check that the elements of lpath exist. */
4124
4125 static void
4126 load_path_check (Lisp_Object lpath)
4127 {
4128 Lisp_Object path_tail;
4129
4130 /* The only elements that might not exist are those from
4131 PATH_LOADSEARCH, EMACSLOADPATH. Anything else is only added if
4132 it exists. */
4133 for (path_tail = lpath; !NILP (path_tail); path_tail = XCDR (path_tail))
4134 {
4135 Lisp_Object dirfile;
4136 dirfile = Fcar (path_tail);
4137 if (STRINGP (dirfile))
4138 {
4139 dirfile = Fdirectory_file_name (dirfile);
4140 if (! file_accessible_directory_p (dirfile))
4141 dir_warning ("Lisp directory", XCAR (path_tail));
4142 }
4143 }
4144 }
4145
4146 /* Return the default load-path, to be used if EMACSLOADPATH is unset.
4147 This does not include the standard site-lisp directories
4148 under the installation prefix (i.e., PATH_SITELOADSEARCH),
4149 but it does (unless no_site_lisp is set) include site-lisp
4150 directories in the source/build directories if those exist and we
4151 are running uninstalled.
4152
4153 Uses the following logic:
4154 If CANNOT_DUMP: Use PATH_LOADSEARCH.
4155 The remainder is what happens when dumping works:
4156 If purify-flag (ie dumping) just use PATH_DUMPLOADSEARCH.
4157 Otherwise use PATH_LOADSEARCH.
4158
4159 If !initialized, then just return PATH_DUMPLOADSEARCH.
4160 If initialized:
4161 If Vinstallation_directory is not nil (ie, running uninstalled):
4162 If installation-dir/lisp exists and not already a member,
4163 we must be running uninstalled. Reset the load-path
4164 to just installation-dir/lisp. (The default PATH_LOADSEARCH
4165 refers to the eventual installation directories. Since we
4166 are not yet installed, we should not use them, even if they exist.)
4167 If installation-dir/lisp does not exist, just add
4168 PATH_DUMPLOADSEARCH at the end instead.
4169 Add installation-dir/site-lisp (if !no_site_lisp, and exists
4170 and not already a member) at the front.
4171 If installation-dir != source-dir (ie running an uninstalled,
4172 out-of-tree build) AND install-dir/src/Makefile exists BUT
4173 install-dir/src/Makefile.in does NOT exist (this is a sanity
4174 check), then repeat the above steps for source-dir/lisp, site-lisp. */
4175
4176 static Lisp_Object
4177 load_path_default (void)
4178 {
4179 Lisp_Object lpath = Qnil;
4180 const char *normal;
4181
4182 #ifdef CANNOT_DUMP
4183 #ifdef HAVE_NS
4184 const char *loadpath = ns_load_path ();
4185 #endif
4186
4187 normal = PATH_LOADSEARCH;
4188 #ifdef HAVE_NS
4189 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4190 #else
4191 lpath = decode_env_path (0, normal, 0);
4192 #endif
4193
4194 #else /* !CANNOT_DUMP */
4195
4196 normal = NILP (Vpurify_flag) ? PATH_LOADSEARCH : PATH_DUMPLOADSEARCH;
4197
4198 if (initialized)
4199 {
4200 #ifdef HAVE_NS
4201 const char *loadpath = ns_load_path ();
4202 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4203 #else
4204 lpath = decode_env_path (0, normal, 0);
4205 #endif
4206 if (!NILP (Vinstallation_directory))
4207 {
4208 Lisp_Object tem, tem1;
4209
4210 /* Add to the path the lisp subdir of the installation
4211 dir, if it is accessible. Note: in out-of-tree builds,
4212 this directory is empty save for Makefile. */
4213 tem = Fexpand_file_name (build_string ("lisp"),
4214 Vinstallation_directory);
4215 tem1 = Ffile_accessible_directory_p (tem);
4216 if (!NILP (tem1))
4217 {
4218 if (NILP (Fmember (tem, lpath)))
4219 {
4220 /* We are running uninstalled. The default load-path
4221 points to the eventual installed lisp directories.
4222 We should not use those now, even if they exist,
4223 so start over from a clean slate. */
4224 lpath = list1 (tem);
4225 }
4226 }
4227 else
4228 /* That dir doesn't exist, so add the build-time
4229 Lisp dirs instead. */
4230 {
4231 Lisp_Object dump_path =
4232 decode_env_path (0, PATH_DUMPLOADSEARCH, 0);
4233 lpath = nconc2 (lpath, dump_path);
4234 }
4235
4236 /* Add site-lisp under the installation dir, if it exists. */
4237 if (!no_site_lisp)
4238 {
4239 tem = Fexpand_file_name (build_string ("site-lisp"),
4240 Vinstallation_directory);
4241 tem1 = Ffile_accessible_directory_p (tem);
4242 if (!NILP (tem1))
4243 {
4244 if (NILP (Fmember (tem, lpath)))
4245 lpath = Fcons (tem, lpath);
4246 }
4247 }
4248
4249 /* If Emacs was not built in the source directory,
4250 and it is run from where it was built, add to load-path
4251 the lisp and site-lisp dirs under that directory. */
4252
4253 if (NILP (Fequal (Vinstallation_directory, Vsource_directory)))
4254 {
4255 Lisp_Object tem2;
4256
4257 tem = Fexpand_file_name (build_string ("src/Makefile"),
4258 Vinstallation_directory);
4259 tem1 = Ffile_exists_p (tem);
4260
4261 /* Don't be fooled if they moved the entire source tree
4262 AFTER dumping Emacs. If the build directory is indeed
4263 different from the source dir, src/Makefile.in and
4264 src/Makefile will not be found together. */
4265 tem = Fexpand_file_name (build_string ("src/Makefile.in"),
4266 Vinstallation_directory);
4267 tem2 = Ffile_exists_p (tem);
4268 if (!NILP (tem1) && NILP (tem2))
4269 {
4270 tem = Fexpand_file_name (build_string ("lisp"),
4271 Vsource_directory);
4272
4273 if (NILP (Fmember (tem, lpath)))
4274 lpath = Fcons (tem, lpath);
4275
4276 if (!no_site_lisp)
4277 {
4278 tem = Fexpand_file_name (build_string ("site-lisp"),
4279 Vsource_directory);
4280 tem1 = Ffile_accessible_directory_p (tem);
4281 if (!NILP (tem1))
4282 {
4283 if (NILP (Fmember (tem, lpath)))
4284 lpath = Fcons (tem, lpath);
4285 }
4286 }
4287 }
4288 } /* Vinstallation_directory != Vsource_directory */
4289
4290 } /* if Vinstallation_directory */
4291 }
4292 else /* !initialized */
4293 {
4294 /* NORMAL refers to PATH_DUMPLOADSEARCH, ie the lisp dir in the
4295 source directory. We used to add ../lisp (ie the lisp dir in
4296 the build directory) at the front here, but that should not
4297 be necessary, since in out of tree builds lisp/ is empty, save
4298 for Makefile. */
4299 lpath = decode_env_path (0, normal, 0);
4300 }
4301 #endif /* !CANNOT_DUMP */
4302
4303 return lpath;
4304 }
4305
4306 void
4307 init_lread (void)
4308 {
4309 /* First, set Vload_path. */
4310
4311 /* Ignore EMACSLOADPATH when dumping. */
4312 #ifdef CANNOT_DUMP
4313 bool use_loadpath = true;
4314 #else
4315 bool use_loadpath = NILP (Vpurify_flag);
4316 #endif
4317
4318 if (use_loadpath && egetenv ("EMACSLOADPATH"))
4319 {
4320 Vload_path = decode_env_path ("EMACSLOADPATH", 0, 1);
4321
4322 /* Check (non-nil) user-supplied elements. */
4323 load_path_check (Vload_path);
4324
4325 /* If no nils in the environment variable, use as-is.
4326 Otherwise, replace any nils with the default. */
4327 if (! NILP (Fmemq (Qnil, Vload_path)))
4328 {
4329 Lisp_Object elem, elpath = Vload_path;
4330 Lisp_Object default_lpath = load_path_default ();
4331
4332 /* Check defaults, before adding site-lisp. */
4333 load_path_check (default_lpath);
4334
4335 /* Add the site-lisp directories to the front of the default. */
4336 if (!no_site_lisp)
4337 {
4338 Lisp_Object sitelisp;
4339 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4340 if (! NILP (sitelisp))
4341 default_lpath = nconc2 (sitelisp, default_lpath);
4342 }
4343
4344 Vload_path = Qnil;
4345
4346 /* Replace nils from EMACSLOADPATH by default. */
4347 while (CONSP (elpath))
4348 {
4349 elem = XCAR (elpath);
4350 elpath = XCDR (elpath);
4351 Vload_path = CALLN (Fappend, Vload_path,
4352 NILP (elem) ? default_lpath : list1 (elem));
4353 }
4354 } /* Fmemq (Qnil, Vload_path) */
4355 }
4356 else
4357 {
4358 Vload_path = load_path_default ();
4359
4360 /* Check before adding site-lisp directories.
4361 The install should have created them, but they are not
4362 required, so no need to warn if they are absent.
4363 Or we might be running before installation. */
4364 load_path_check (Vload_path);
4365
4366 /* Add the site-lisp directories at the front. */
4367 if (initialized && !no_site_lisp)
4368 {
4369 Lisp_Object sitelisp;
4370 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4371 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4372 }
4373 }
4374
4375 Vvalues = Qnil;
4376
4377 load_in_progress = 0;
4378 Vload_file_name = Qnil;
4379 Vstandard_input = Qt;
4380 Vloads_in_progress = Qnil;
4381 }
4382
4383 /* Print a warning that directory intended for use USE and with name
4384 DIRNAME cannot be accessed. On entry, errno should correspond to
4385 the access failure. Print the warning on stderr and put it in
4386 *Messages*. */
4387
4388 void
4389 dir_warning (char const *use, Lisp_Object dirname)
4390 {
4391 static char const format[] = "Warning: %s '%s': %s\n";
4392 int access_errno = errno;
4393 fprintf (stderr, format, use, SSDATA (ENCODE_SYSTEM (dirname)),
4394 strerror (access_errno));
4395
4396 /* Don't log the warning before we've initialized!! */
4397 if (initialized)
4398 {
4399 char const *diagnostic = emacs_strerror (access_errno);
4400 USE_SAFE_ALLOCA;
4401 char *buffer = SAFE_ALLOCA (sizeof format - 3 * (sizeof "%s" - 1)
4402 + strlen (use) + SBYTES (dirname)
4403 + strlen (diagnostic));
4404 ptrdiff_t message_len = esprintf (buffer, format, use, SSDATA (dirname),
4405 diagnostic);
4406 message_dolog (buffer, message_len, 0, STRING_MULTIBYTE (dirname));
4407 SAFE_FREE ();
4408 }
4409 }
4410
4411 void
4412 syms_of_lread (void)
4413 {
4414 defsubr (&Sread);
4415 defsubr (&Sread_from_string);
4416 defsubr (&Sintern);
4417 defsubr (&Sintern_soft);
4418 defsubr (&Sunintern);
4419 defsubr (&Sget_load_suffixes);
4420 defsubr (&Sload);
4421 defsubr (&Seval_buffer);
4422 defsubr (&Seval_region);
4423 defsubr (&Sread_char);
4424 defsubr (&Sread_char_exclusive);
4425 defsubr (&Sread_event);
4426 defsubr (&Sget_file_char);
4427 defsubr (&Smapatoms);
4428 defsubr (&Slocate_file_internal);
4429
4430 DEFVAR_LISP ("obarray", Vobarray,
4431 doc: /* Symbol table for use by `intern' and `read'.
4432 It is a vector whose length ought to be prime for best results.
4433 The vector's contents don't make sense if examined from Lisp programs;
4434 to find all the symbols in an obarray, use `mapatoms'. */);
4435
4436 DEFVAR_LISP ("values", Vvalues,
4437 doc: /* List of values of all expressions which were read, evaluated and printed.
4438 Order is reverse chronological. */);
4439 XSYMBOL (intern ("values"))->declared_special = 0;
4440
4441 DEFVAR_LISP ("standard-input", Vstandard_input,
4442 doc: /* Stream for read to get input from.
4443 See documentation of `read' for possible values. */);
4444 Vstandard_input = Qt;
4445
4446 DEFVAR_LISP ("read-with-symbol-positions", Vread_with_symbol_positions,
4447 doc: /* If non-nil, add position of read symbols to `read-symbol-positions-list'.
4448
4449 If this variable is a buffer, then only forms read from that buffer
4450 will be added to `read-symbol-positions-list'.
4451 If this variable is t, then all read forms will be added.
4452 The effect of all other values other than nil are not currently
4453 defined, although they may be in the future.
4454
4455 The positions are relative to the last call to `read' or
4456 `read-from-string'. It is probably a bad idea to set this variable at
4457 the toplevel; bind it instead. */);
4458 Vread_with_symbol_positions = Qnil;
4459
4460 DEFVAR_LISP ("read-symbol-positions-list", Vread_symbol_positions_list,
4461 doc: /* A list mapping read symbols to their positions.
4462 This variable is modified during calls to `read' or
4463 `read-from-string', but only when `read-with-symbol-positions' is
4464 non-nil.
4465
4466 Each element of the list looks like (SYMBOL . CHAR-POSITION), where
4467 CHAR-POSITION is an integer giving the offset of that occurrence of the
4468 symbol from the position where `read' or `read-from-string' started.
4469
4470 Note that a symbol will appear multiple times in this list, if it was
4471 read multiple times. The list is in the same order as the symbols
4472 were read in. */);
4473 Vread_symbol_positions_list = Qnil;
4474
4475 DEFVAR_LISP ("read-circle", Vread_circle,
4476 doc: /* Non-nil means read recursive structures using #N= and #N# syntax. */);
4477 Vread_circle = Qt;
4478
4479 DEFVAR_LISP ("load-path", Vload_path,
4480 doc: /* List of directories to search for files to load.
4481 Each element is a string (directory name) or nil (meaning `default-directory').
4482 Initialized during startup as described in Info node `(elisp)Library Search'. */);
4483
4484 DEFVAR_LISP ("load-suffixes", Vload_suffixes,
4485 doc: /* List of suffixes for (compiled or source) Emacs Lisp files.
4486 This list should not include the empty string.
4487 `load' and related functions try to append these suffixes, in order,
4488 to the specified file name if a Lisp suffix is allowed or required. */);
4489 Vload_suffixes = list2 (build_pure_c_string (".elc"),
4490 build_pure_c_string (".el"));
4491 DEFVAR_LISP ("load-file-rep-suffixes", Vload_file_rep_suffixes,
4492 doc: /* List of suffixes that indicate representations of \
4493 the same file.
4494 This list should normally start with the empty string.
4495
4496 Enabling Auto Compression mode appends the suffixes in
4497 `jka-compr-load-suffixes' to this list and disabling Auto Compression
4498 mode removes them again. `load' and related functions use this list to
4499 determine whether they should look for compressed versions of a file
4500 and, if so, which suffixes they should try to append to the file name
4501 in order to do so. However, if you want to customize which suffixes
4502 the loading functions recognize as compression suffixes, you should
4503 customize `jka-compr-load-suffixes' rather than the present variable. */);
4504 Vload_file_rep_suffixes = list1 (empty_unibyte_string);
4505
4506 DEFVAR_BOOL ("load-in-progress", load_in_progress,
4507 doc: /* Non-nil if inside of `load'. */);
4508 DEFSYM (Qload_in_progress, "load-in-progress");
4509
4510 DEFVAR_LISP ("after-load-alist", Vafter_load_alist,
4511 doc: /* An alist of functions to be evalled when particular files are loaded.
4512 Each element looks like (REGEXP-OR-FEATURE FUNCS...).
4513
4514 REGEXP-OR-FEATURE is either a regular expression to match file names, or
4515 a symbol (a feature name).
4516
4517 When `load' is run and the file-name argument matches an element's
4518 REGEXP-OR-FEATURE, or when `provide' is run and provides the symbol
4519 REGEXP-OR-FEATURE, the FUNCS in the element are called.
4520
4521 An error in FORMS does not undo the load, but does prevent execution of
4522 the rest of the FORMS. */);
4523 Vafter_load_alist = Qnil;
4524
4525 DEFVAR_LISP ("load-history", Vload_history,
4526 doc: /* Alist mapping loaded file names to symbols and features.
4527 Each alist element should be a list (FILE-NAME ENTRIES...), where
4528 FILE-NAME is the name of a file that has been loaded into Emacs.
4529 The file name is absolute and true (i.e. it doesn't contain symlinks).
4530 As an exception, one of the alist elements may have FILE-NAME nil,
4531 for symbols and features not associated with any file.
4532
4533 The remaining ENTRIES in the alist element describe the functions and
4534 variables defined in that file, the features provided, and the
4535 features required. Each entry has the form `(provide . FEATURE)',
4536 `(require . FEATURE)', `(defun . FUNCTION)', `(autoload . SYMBOL)',
4537 `(defface . SYMBOL)', or `(t . SYMBOL)'. Entries like `(t . SYMBOL)'
4538 may precede a `(defun . FUNCTION)' entry, and means that SYMBOL was an
4539 autoload before this file redefined it as a function. In addition,
4540 entries may also be single symbols, which means that SYMBOL was
4541 defined by `defvar' or `defconst'.
4542
4543 During preloading, the file name recorded is relative to the main Lisp
4544 directory. These file names are converted to absolute at startup. */);
4545 Vload_history = Qnil;
4546
4547 DEFVAR_LISP ("load-file-name", Vload_file_name,
4548 doc: /* Full name of file being loaded by `load'. */);
4549 Vload_file_name = Qnil;
4550
4551 DEFVAR_LISP ("user-init-file", Vuser_init_file,
4552 doc: /* File name, including directory, of user's initialization file.
4553 If the file loaded had extension `.elc', and the corresponding source file
4554 exists, this variable contains the name of source file, suitable for use
4555 by functions like `custom-save-all' which edit the init file.
4556 While Emacs loads and evaluates the init file, value is the real name
4557 of the file, regardless of whether or not it has the `.elc' extension. */);
4558 Vuser_init_file = Qnil;
4559
4560 DEFVAR_LISP ("current-load-list", Vcurrent_load_list,
4561 doc: /* Used for internal purposes by `load'. */);
4562 Vcurrent_load_list = Qnil;
4563
4564 DEFVAR_LISP ("load-read-function", Vload_read_function,
4565 doc: /* Function used by `load' and `eval-region' for reading expressions.
4566 Called with a single argument (the stream from which to read).
4567 The default is to use the function `read'. */);
4568 DEFSYM (Qread, "read");
4569 Vload_read_function = Qread;
4570
4571 DEFVAR_LISP ("load-source-file-function", Vload_source_file_function,
4572 doc: /* Function called in `load' to load an Emacs Lisp source file.
4573 The value should be a function for doing code conversion before
4574 reading a source file. It can also be nil, in which case loading is
4575 done without any code conversion.
4576
4577 If the value is a function, it is called with four arguments,
4578 FULLNAME, FILE, NOERROR, NOMESSAGE. FULLNAME is the absolute name of
4579 the file to load, FILE is the non-absolute name (for messages etc.),
4580 and NOERROR and NOMESSAGE are the corresponding arguments passed to
4581 `load'. The function should return t if the file was loaded. */);
4582 Vload_source_file_function = Qnil;
4583
4584 DEFVAR_BOOL ("load-force-doc-strings", load_force_doc_strings,
4585 doc: /* Non-nil means `load' should force-load all dynamic doc strings.
4586 This is useful when the file being loaded is a temporary copy. */);
4587 load_force_doc_strings = 0;
4588
4589 DEFVAR_BOOL ("load-convert-to-unibyte", load_convert_to_unibyte,
4590 doc: /* Non-nil means `read' converts strings to unibyte whenever possible.
4591 This is normally bound by `load' and `eval-buffer' to control `read',
4592 and is not meant for users to change. */);
4593 load_convert_to_unibyte = 0;
4594
4595 DEFVAR_LISP ("source-directory", Vsource_directory,
4596 doc: /* Directory in which Emacs sources were found when Emacs was built.
4597 You cannot count on them to still be there! */);
4598 Vsource_directory
4599 = Fexpand_file_name (build_string ("../"),
4600 Fcar (decode_env_path (0, PATH_DUMPLOADSEARCH, 0)));
4601
4602 DEFVAR_LISP ("preloaded-file-list", Vpreloaded_file_list,
4603 doc: /* List of files that were preloaded (when dumping Emacs). */);
4604 Vpreloaded_file_list = Qnil;
4605
4606 DEFVAR_LISP ("byte-boolean-vars", Vbyte_boolean_vars,
4607 doc: /* List of all DEFVAR_BOOL variables, used by the byte code optimizer. */);
4608 Vbyte_boolean_vars = Qnil;
4609
4610 DEFVAR_BOOL ("load-dangerous-libraries", load_dangerous_libraries,
4611 doc: /* Non-nil means load dangerous compiled Lisp files.
4612 Some versions of XEmacs use different byte codes than Emacs. These
4613 incompatible byte codes can make Emacs crash when it tries to execute
4614 them. */);
4615 load_dangerous_libraries = 0;
4616
4617 DEFVAR_BOOL ("force-load-messages", force_load_messages,
4618 doc: /* Non-nil means force printing messages when loading Lisp files.
4619 This overrides the value of the NOMESSAGE argument to `load'. */);
4620 force_load_messages = 0;
4621
4622 DEFVAR_LISP ("bytecomp-version-regexp", Vbytecomp_version_regexp,
4623 doc: /* Regular expression matching safe to load compiled Lisp files.
4624 When Emacs loads a compiled Lisp file, it reads the first 512 bytes
4625 from the file, and matches them against this regular expression.
4626 When the regular expression matches, the file is considered to be safe
4627 to load. See also `load-dangerous-libraries'. */);
4628 Vbytecomp_version_regexp
4629 = build_pure_c_string ("^;;;.\\(in Emacs version\\|bytecomp version FSF\\)");
4630
4631 DEFSYM (Qlexical_binding, "lexical-binding");
4632 DEFVAR_LISP ("lexical-binding", Vlexical_binding,
4633 doc: /* Whether to use lexical binding when evaluating code.
4634 Non-nil means that the code in the current buffer should be evaluated
4635 with lexical binding.
4636 This variable is automatically set from the file variables of an
4637 interpreted Lisp file read using `load'. Unlike other file local
4638 variables, this must be set in the first line of a file. */);
4639 Vlexical_binding = Qnil;
4640 Fmake_variable_buffer_local (Qlexical_binding);
4641
4642 DEFVAR_LISP ("eval-buffer-list", Veval_buffer_list,
4643 doc: /* List of buffers being read from by calls to `eval-buffer' and `eval-region'. */);
4644 Veval_buffer_list = Qnil;
4645
4646 DEFVAR_LISP ("old-style-backquotes", Vold_style_backquotes,
4647 doc: /* Set to non-nil when `read' encounters an old-style backquote. */);
4648 Vold_style_backquotes = Qnil;
4649 DEFSYM (Qold_style_backquotes, "old-style-backquotes");
4650
4651 DEFVAR_BOOL ("load-prefer-newer", load_prefer_newer,
4652 doc: /* Non-nil means `load' prefers the newest version of a file.
4653 This applies when a filename suffix is not explicitly specified and
4654 `load' is trying various possible suffixes (see `load-suffixes' and
4655 `load-file-rep-suffixes'). Normally, it stops at the first file
4656 that exists unless you explicitly specify one or the other. If this
4657 option is non-nil, it checks all suffixes and uses whichever file is
4658 newest.
4659 Note that if you customize this, obviously it will not affect files
4660 that are loaded before your customizations are read! */);
4661 load_prefer_newer = 0;
4662
4663 /* Vsource_directory was initialized in init_lread. */
4664
4665 DEFSYM (Qcurrent_load_list, "current-load-list");
4666 DEFSYM (Qstandard_input, "standard-input");
4667 DEFSYM (Qread_char, "read-char");
4668 DEFSYM (Qget_file_char, "get-file-char");
4669
4670 /* Used instead of Qget_file_char while loading *.elc files compiled
4671 by Emacs 21 or older. */
4672 DEFSYM (Qget_emacs_mule_file_char, "get-emacs-mule-file-char");
4673
4674 DEFSYM (Qload_force_doc_strings, "load-force-doc-strings");
4675
4676 DEFSYM (Qbackquote, "`");
4677 DEFSYM (Qcomma, ",");
4678 DEFSYM (Qcomma_at, ",@");
4679 DEFSYM (Qcomma_dot, ",.");
4680
4681 DEFSYM (Qinhibit_file_name_operation, "inhibit-file-name-operation");
4682 DEFSYM (Qascii_character, "ascii-character");
4683 DEFSYM (Qfunction, "function");
4684 DEFSYM (Qload, "load");
4685 DEFSYM (Qload_file_name, "load-file-name");
4686 DEFSYM (Qeval_buffer_list, "eval-buffer-list");
4687 DEFSYM (Qfile_truename, "file-truename");
4688 DEFSYM (Qdir_ok, "dir-ok");
4689 DEFSYM (Qdo_after_load_evaluation, "do-after-load-evaluation");
4690
4691 staticpro (&read_objects);
4692 read_objects = Qnil;
4693 staticpro (&seen_list);
4694 seen_list = Qnil;
4695
4696 Vloads_in_progress = Qnil;
4697 staticpro (&Vloads_in_progress);
4698
4699 DEFSYM (Qhash_table, "hash-table");
4700 DEFSYM (Qdata, "data");
4701 DEFSYM (Qtest, "test");
4702 DEFSYM (Qsize, "size");
4703 DEFSYM (Qweakness, "weakness");
4704 DEFSYM (Qrehash_size, "rehash-size");
4705 DEFSYM (Qrehash_threshold, "rehash-threshold");
4706 }