]> code.delx.au - gnu-emacs/blob - src/lread.c
74a5fdfe67bcdaed30996823c1a8ab956f0f723e
[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 "dispextern.h"
35 #include "intervals.h"
36 #include "character.h"
37 #include "buffer.h"
38 #include "charset.h"
39 #include "coding.h"
40 #include <epaths.h>
41 #include "commands.h"
42 #include "keyboard.h"
43 #include "systime.h"
44 #include "termhooks.h"
45 #include "blockinput.h"
46
47 #ifdef MSDOS
48 #include "msdos.h"
49 #endif
50
51 #ifdef HAVE_NS
52 #include "nsterm.h"
53 #endif
54
55 #include <unistd.h>
56
57 #ifdef HAVE_SETLOCALE
58 #include <locale.h>
59 #endif /* HAVE_SETLOCALE */
60
61 #include <fcntl.h>
62
63 #ifdef HAVE_FSEEKO
64 #define file_offset off_t
65 #define file_tell ftello
66 #else
67 #define file_offset long
68 #define file_tell ftell
69 #endif
70
71 /* The association list of objects read with the #n=object form.
72 Each member of the list has the form (n . object), and is used to
73 look up the object for the corresponding #n# construct.
74 It must be set to nil before all top-level calls to read0. */
75 static Lisp_Object read_objects;
76
77 /* File for get_file_char to read from. Use by load. */
78 static FILE *instream;
79
80 /* For use within read-from-string (this reader is non-reentrant!!) */
81 static ptrdiff_t read_from_string_index;
82 static ptrdiff_t read_from_string_index_byte;
83 static ptrdiff_t read_from_string_limit;
84
85 /* Number of characters read in the current call to Fread or
86 Fread_from_string. */
87 static EMACS_INT readchar_count;
88
89 /* This contains the last string skipped with #@. */
90 static char *saved_doc_string;
91 /* Length of buffer allocated in saved_doc_string. */
92 static ptrdiff_t saved_doc_string_size;
93 /* Length of actual data in saved_doc_string. */
94 static ptrdiff_t saved_doc_string_length;
95 /* This is the file position that string came from. */
96 static file_offset saved_doc_string_position;
97
98 /* This contains the previous string skipped with #@.
99 We copy it from saved_doc_string when a new string
100 is put in saved_doc_string. */
101 static char *prev_saved_doc_string;
102 /* Length of buffer allocated in prev_saved_doc_string. */
103 static ptrdiff_t prev_saved_doc_string_size;
104 /* Length of actual data in prev_saved_doc_string. */
105 static ptrdiff_t prev_saved_doc_string_length;
106 /* This is the file position that string came from. */
107 static file_offset prev_saved_doc_string_position;
108
109 /* True means inside a new-style backquote
110 with no surrounding parentheses.
111 Fread initializes this to false, so we need not specbind it
112 or worry about what happens to it when there is an error. */
113 static bool new_backquote_flag;
114
115 /* A list of file names for files being loaded in Fload. Used to
116 check for recursive loads. */
117
118 static Lisp_Object Vloads_in_progress;
119
120 static int read_emacs_mule_char (int, int (*) (int, Lisp_Object),
121 Lisp_Object);
122
123 static void readevalloop (Lisp_Object, FILE *, Lisp_Object, bool,
124 Lisp_Object, Lisp_Object,
125 Lisp_Object, Lisp_Object);
126 \f
127 /* Functions that read one byte from the current source READCHARFUN
128 or unreads one byte. If the integer argument C is -1, it returns
129 one read byte, or -1 when there's no more byte in the source. If C
130 is 0 or positive, it unreads C, and the return value is not
131 interesting. */
132
133 static int readbyte_for_lambda (int, Lisp_Object);
134 static int readbyte_from_file (int, Lisp_Object);
135 static int readbyte_from_string (int, Lisp_Object);
136
137 /* Handle unreading and rereading of characters.
138 Write READCHAR to read a character,
139 UNREAD(c) to unread c to be read again.
140
141 These macros correctly read/unread multibyte characters. */
142
143 #define READCHAR readchar (readcharfun, NULL)
144 #define UNREAD(c) unreadchar (readcharfun, c)
145
146 /* Same as READCHAR but set *MULTIBYTE to the multibyteness of the source. */
147 #define READCHAR_REPORT_MULTIBYTE(multibyte) readchar (readcharfun, multibyte)
148
149 /* When READCHARFUN is Qget_file_char, Qget_emacs_mule_file_char,
150 Qlambda, or a cons, we use this to keep an unread character because
151 a file stream can't handle multibyte-char unreading. The value -1
152 means that there's no unread character. */
153 static int unread_char;
154
155 static int
156 readchar (Lisp_Object readcharfun, bool *multibyte)
157 {
158 Lisp_Object tem;
159 register int c;
160 int (*readbyte) (int, Lisp_Object);
161 unsigned char buf[MAX_MULTIBYTE_LENGTH];
162 int i, len;
163 bool emacs_mule_encoding = 0;
164
165 if (multibyte)
166 *multibyte = 0;
167
168 readchar_count++;
169
170 if (BUFFERP (readcharfun))
171 {
172 register struct buffer *inbuffer = XBUFFER (readcharfun);
173
174 ptrdiff_t pt_byte = BUF_PT_BYTE (inbuffer);
175
176 if (! BUFFER_LIVE_P (inbuffer))
177 return -1;
178
179 if (pt_byte >= BUF_ZV_BYTE (inbuffer))
180 return -1;
181
182 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
183 {
184 /* Fetch the character code from the buffer. */
185 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, pt_byte);
186 BUF_INC_POS (inbuffer, pt_byte);
187 c = STRING_CHAR (p);
188 if (multibyte)
189 *multibyte = 1;
190 }
191 else
192 {
193 c = BUF_FETCH_BYTE (inbuffer, pt_byte);
194 if (! ASCII_CHAR_P (c))
195 c = BYTE8_TO_CHAR (c);
196 pt_byte++;
197 }
198 SET_BUF_PT_BOTH (inbuffer, BUF_PT (inbuffer) + 1, pt_byte);
199
200 return c;
201 }
202 if (MARKERP (readcharfun))
203 {
204 register struct buffer *inbuffer = XMARKER (readcharfun)->buffer;
205
206 ptrdiff_t bytepos = marker_byte_position (readcharfun);
207
208 if (bytepos >= BUF_ZV_BYTE (inbuffer))
209 return -1;
210
211 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
212 {
213 /* Fetch the character code from the buffer. */
214 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, bytepos);
215 BUF_INC_POS (inbuffer, bytepos);
216 c = STRING_CHAR (p);
217 if (multibyte)
218 *multibyte = 1;
219 }
220 else
221 {
222 c = BUF_FETCH_BYTE (inbuffer, bytepos);
223 if (! ASCII_CHAR_P (c))
224 c = BYTE8_TO_CHAR (c);
225 bytepos++;
226 }
227
228 XMARKER (readcharfun)->bytepos = bytepos;
229 XMARKER (readcharfun)->charpos++;
230
231 return c;
232 }
233
234 if (EQ (readcharfun, Qlambda))
235 {
236 readbyte = readbyte_for_lambda;
237 goto read_multibyte;
238 }
239
240 if (EQ (readcharfun, Qget_file_char))
241 {
242 readbyte = readbyte_from_file;
243 goto read_multibyte;
244 }
245
246 if (STRINGP (readcharfun))
247 {
248 if (read_from_string_index >= read_from_string_limit)
249 c = -1;
250 else if (STRING_MULTIBYTE (readcharfun))
251 {
252 if (multibyte)
253 *multibyte = 1;
254 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c, readcharfun,
255 read_from_string_index,
256 read_from_string_index_byte);
257 }
258 else
259 {
260 c = SREF (readcharfun, read_from_string_index_byte);
261 read_from_string_index++;
262 read_from_string_index_byte++;
263 }
264 return c;
265 }
266
267 if (CONSP (readcharfun))
268 {
269 /* This is the case that read_vector is reading from a unibyte
270 string that contains a byte sequence previously skipped
271 because of #@NUMBER. The car part of readcharfun is that
272 string, and the cdr part is a value of readcharfun given to
273 read_vector. */
274 readbyte = readbyte_from_string;
275 if (EQ (XCDR (readcharfun), Qget_emacs_mule_file_char))
276 emacs_mule_encoding = 1;
277 goto read_multibyte;
278 }
279
280 if (EQ (readcharfun, Qget_emacs_mule_file_char))
281 {
282 readbyte = readbyte_from_file;
283 emacs_mule_encoding = 1;
284 goto read_multibyte;
285 }
286
287 tem = call0 (readcharfun);
288
289 if (NILP (tem))
290 return -1;
291 return XINT (tem);
292
293 read_multibyte:
294 if (unread_char >= 0)
295 {
296 c = unread_char;
297 unread_char = -1;
298 return c;
299 }
300 c = (*readbyte) (-1, readcharfun);
301 if (c < 0)
302 return c;
303 if (multibyte)
304 *multibyte = 1;
305 if (ASCII_CHAR_P (c))
306 return c;
307 if (emacs_mule_encoding)
308 return read_emacs_mule_char (c, readbyte, readcharfun);
309 i = 0;
310 buf[i++] = c;
311 len = BYTES_BY_CHAR_HEAD (c);
312 while (i < len)
313 {
314 c = (*readbyte) (-1, readcharfun);
315 if (c < 0 || ! TRAILING_CODE_P (c))
316 {
317 while (--i > 1)
318 (*readbyte) (buf[i], readcharfun);
319 return BYTE8_TO_CHAR (buf[0]);
320 }
321 buf[i++] = c;
322 }
323 return STRING_CHAR (buf);
324 }
325
326 #define FROM_FILE_P(readcharfun) \
327 (EQ (readcharfun, Qget_file_char) \
328 || EQ (readcharfun, Qget_emacs_mule_file_char))
329
330 static void
331 skip_dyn_bytes (Lisp_Object readcharfun, ptrdiff_t n)
332 {
333 if (FROM_FILE_P (readcharfun))
334 {
335 block_input (); /* FIXME: Not sure if it's needed. */
336 fseek (instream, n, SEEK_CUR);
337 unblock_input ();
338 }
339 else
340 { /* We're not reading directly from a file. In that case, it's difficult
341 to reliably count bytes, since these are usually meant for the file's
342 encoding, whereas we're now typically in the internal encoding.
343 But luckily, skip_dyn_bytes is used to skip over a single
344 dynamic-docstring (or dynamic byte-code) which is always quoted such
345 that \037 is the final char. */
346 int c;
347 do {
348 c = READCHAR;
349 } while (c >= 0 && c != '\037');
350 }
351 }
352
353 static void
354 skip_dyn_eof (Lisp_Object readcharfun)
355 {
356 if (FROM_FILE_P (readcharfun))
357 {
358 block_input (); /* FIXME: Not sure if it's needed. */
359 fseek (instream, 0, SEEK_END);
360 unblock_input ();
361 }
362 else
363 while (READCHAR >= 0);
364 }
365
366 /* Unread the character C in the way appropriate for the stream READCHARFUN.
367 If the stream is a user function, call it with the char as argument. */
368
369 static void
370 unreadchar (Lisp_Object readcharfun, int c)
371 {
372 readchar_count--;
373 if (c == -1)
374 /* Don't back up the pointer if we're unreading the end-of-input mark,
375 since readchar didn't advance it when we read it. */
376 ;
377 else if (BUFFERP (readcharfun))
378 {
379 struct buffer *b = XBUFFER (readcharfun);
380 ptrdiff_t charpos = BUF_PT (b);
381 ptrdiff_t bytepos = BUF_PT_BYTE (b);
382
383 if (! NILP (BVAR (b, enable_multibyte_characters)))
384 BUF_DEC_POS (b, bytepos);
385 else
386 bytepos--;
387
388 SET_BUF_PT_BOTH (b, charpos - 1, bytepos);
389 }
390 else if (MARKERP (readcharfun))
391 {
392 struct buffer *b = XMARKER (readcharfun)->buffer;
393 ptrdiff_t bytepos = XMARKER (readcharfun)->bytepos;
394
395 XMARKER (readcharfun)->charpos--;
396 if (! NILP (BVAR (b, enable_multibyte_characters)))
397 BUF_DEC_POS (b, bytepos);
398 else
399 bytepos--;
400
401 XMARKER (readcharfun)->bytepos = bytepos;
402 }
403 else if (STRINGP (readcharfun))
404 {
405 read_from_string_index--;
406 read_from_string_index_byte
407 = string_char_to_byte (readcharfun, read_from_string_index);
408 }
409 else if (CONSP (readcharfun))
410 {
411 unread_char = c;
412 }
413 else if (EQ (readcharfun, Qlambda))
414 {
415 unread_char = c;
416 }
417 else if (FROM_FILE_P (readcharfun))
418 {
419 unread_char = c;
420 }
421 else
422 call1 (readcharfun, make_number (c));
423 }
424
425 static int
426 readbyte_for_lambda (int c, Lisp_Object readcharfun)
427 {
428 return read_bytecode_char (c >= 0);
429 }
430
431
432 static int
433 readbyte_from_file (int c, Lisp_Object readcharfun)
434 {
435 if (c >= 0)
436 {
437 block_input ();
438 ungetc (c, instream);
439 unblock_input ();
440 return 0;
441 }
442
443 block_input ();
444 c = getc (instream);
445
446 /* Interrupted reads have been observed while reading over the network. */
447 while (c == EOF && ferror (instream) && errno == EINTR)
448 {
449 unblock_input ();
450 QUIT;
451 block_input ();
452 clearerr (instream);
453 c = getc (instream);
454 }
455
456 unblock_input ();
457
458 return (c == EOF ? -1 : c);
459 }
460
461 static int
462 readbyte_from_string (int c, Lisp_Object readcharfun)
463 {
464 Lisp_Object string = XCAR (readcharfun);
465
466 if (c >= 0)
467 {
468 read_from_string_index--;
469 read_from_string_index_byte
470 = string_char_to_byte (string, read_from_string_index);
471 }
472
473 if (read_from_string_index >= read_from_string_limit)
474 c = -1;
475 else
476 FETCH_STRING_CHAR_ADVANCE (c, string,
477 read_from_string_index,
478 read_from_string_index_byte);
479 return c;
480 }
481
482
483 /* Read one non-ASCII character from INSTREAM. The character is
484 encoded in `emacs-mule' and the first byte is already read in
485 C. */
486
487 static int
488 read_emacs_mule_char (int c, int (*readbyte) (int, Lisp_Object), Lisp_Object readcharfun)
489 {
490 /* Emacs-mule coding uses at most 4-byte for one character. */
491 unsigned char buf[4];
492 int len = emacs_mule_bytes[c];
493 struct charset *charset;
494 int i;
495 unsigned code;
496
497 if (len == 1)
498 /* C is not a valid leading-code of `emacs-mule'. */
499 return BYTE8_TO_CHAR (c);
500
501 i = 0;
502 buf[i++] = c;
503 while (i < len)
504 {
505 c = (*readbyte) (-1, readcharfun);
506 if (c < 0xA0)
507 {
508 while (--i > 1)
509 (*readbyte) (buf[i], readcharfun);
510 return BYTE8_TO_CHAR (buf[0]);
511 }
512 buf[i++] = c;
513 }
514
515 if (len == 2)
516 {
517 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
518 code = buf[1] & 0x7F;
519 }
520 else if (len == 3)
521 {
522 if (buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_11
523 || buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_12)
524 {
525 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
526 code = buf[2] & 0x7F;
527 }
528 else
529 {
530 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
531 code = ((buf[1] << 8) | buf[2]) & 0x7F7F;
532 }
533 }
534 else
535 {
536 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
537 code = ((buf[2] << 8) | buf[3]) & 0x7F7F;
538 }
539 c = DECODE_CHAR (charset, code);
540 if (c < 0)
541 Fsignal (Qinvalid_read_syntax,
542 list1 (build_string ("invalid multibyte form")));
543 return c;
544 }
545
546
547 static Lisp_Object read_internal_start (Lisp_Object, Lisp_Object,
548 Lisp_Object);
549 static Lisp_Object read0 (Lisp_Object);
550 static Lisp_Object read1 (Lisp_Object, int *, bool);
551
552 static Lisp_Object read_list (bool, Lisp_Object);
553 static Lisp_Object read_vector (Lisp_Object, bool);
554
555 static Lisp_Object substitute_object_recurse (Lisp_Object, Lisp_Object,
556 Lisp_Object);
557 static void substitute_object_in_subtree (Lisp_Object,
558 Lisp_Object);
559 static void substitute_in_interval (INTERVAL, Lisp_Object);
560
561 \f
562 /* Get a character from the tty. */
563
564 /* Read input events until we get one that's acceptable for our purposes.
565
566 If NO_SWITCH_FRAME, switch-frame events are stashed
567 until we get a character we like, and then stuffed into
568 unread_switch_frame.
569
570 If ASCII_REQUIRED, check function key events to see
571 if the unmodified version of the symbol has a Qascii_character
572 property, and use that character, if present.
573
574 If ERROR_NONASCII, signal an error if the input we
575 get isn't an ASCII character with modifiers. If it's false but
576 ASCII_REQUIRED is true, just re-read until we get an ASCII
577 character.
578
579 If INPUT_METHOD, invoke the current input method
580 if the character warrants that.
581
582 If SECONDS is a number, wait that many seconds for input, and
583 return Qnil if no input arrives within that time. */
584
585 static Lisp_Object
586 read_filtered_event (bool no_switch_frame, bool ascii_required,
587 bool error_nonascii, bool input_method, Lisp_Object seconds)
588 {
589 Lisp_Object val, delayed_switch_frame;
590 struct timespec end_time;
591
592 #ifdef HAVE_WINDOW_SYSTEM
593 if (display_hourglass_p)
594 cancel_hourglass ();
595 #endif
596
597 delayed_switch_frame = Qnil;
598
599 /* Compute timeout. */
600 if (NUMBERP (seconds))
601 {
602 double duration = extract_float (seconds);
603 struct timespec wait_time = dtotimespec (duration);
604 end_time = timespec_add (current_timespec (), wait_time);
605 }
606
607 /* Read until we get an acceptable event. */
608 retry:
609 do
610 val = read_char (0, Qnil, (input_method ? Qnil : Qt), 0,
611 NUMBERP (seconds) ? &end_time : NULL);
612 while (INTEGERP (val) && XINT (val) == -2); /* wrong_kboard_jmpbuf */
613
614 if (BUFFERP (val))
615 goto retry;
616
617 /* `switch-frame' events are put off until after the next ASCII
618 character. This is better than signaling an error just because
619 the last characters were typed to a separate minibuffer frame,
620 for example. Eventually, some code which can deal with
621 switch-frame events will read it and process it. */
622 if (no_switch_frame
623 && EVENT_HAS_PARAMETERS (val)
624 && EQ (EVENT_HEAD_KIND (EVENT_HEAD (val)), Qswitch_frame))
625 {
626 delayed_switch_frame = val;
627 goto retry;
628 }
629
630 if (ascii_required && !(NUMBERP (seconds) && NILP (val)))
631 {
632 /* Convert certain symbols to their ASCII equivalents. */
633 if (SYMBOLP (val))
634 {
635 Lisp_Object tem, tem1;
636 tem = Fget (val, Qevent_symbol_element_mask);
637 if (!NILP (tem))
638 {
639 tem1 = Fget (Fcar (tem), Qascii_character);
640 /* Merge this symbol's modifier bits
641 with the ASCII equivalent of its basic code. */
642 if (!NILP (tem1))
643 XSETFASTINT (val, XINT (tem1) | XINT (Fcar (Fcdr (tem))));
644 }
645 }
646
647 /* If we don't have a character now, deal with it appropriately. */
648 if (!INTEGERP (val))
649 {
650 if (error_nonascii)
651 {
652 Vunread_command_events = list1 (val);
653 error ("Non-character input-event");
654 }
655 else
656 goto retry;
657 }
658 }
659
660 if (! NILP (delayed_switch_frame))
661 unread_switch_frame = delayed_switch_frame;
662
663 #if 0
664
665 #ifdef HAVE_WINDOW_SYSTEM
666 if (display_hourglass_p)
667 start_hourglass ();
668 #endif
669
670 #endif
671
672 return val;
673 }
674
675 DEFUN ("read-char", Fread_char, Sread_char, 0, 3, 0,
676 doc: /* Read a character from the command input (keyboard or macro).
677 It is returned as a number.
678 If the character has modifiers, they are resolved and reflected to the
679 character code if possible (e.g. C-SPC -> 0).
680
681 If the user generates an event which is not a character (i.e. a mouse
682 click or function key event), `read-char' signals an error. As an
683 exception, switch-frame events are put off until non-character events
684 can be read.
685 If you want to read non-character events, or ignore them, call
686 `read-event' or `read-char-exclusive' instead.
687
688 If the optional argument PROMPT is non-nil, display that as a prompt.
689 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
690 input method is turned on in the current buffer, that input method
691 is used for reading a character.
692 If the optional argument SECONDS is non-nil, it should be a number
693 specifying the maximum number of seconds to wait for input. If no
694 input arrives in that time, return nil. SECONDS may be a
695 floating-point value. */)
696 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
697 {
698 Lisp_Object val;
699
700 if (! NILP (prompt))
701 message_with_string ("%s", prompt, 0);
702 val = read_filtered_event (1, 1, 1, ! NILP (inherit_input_method), seconds);
703
704 return (NILP (val) ? Qnil
705 : make_number (char_resolve_modifier_mask (XINT (val))));
706 }
707
708 DEFUN ("read-event", Fread_event, Sread_event, 0, 3, 0,
709 doc: /* Read an event object from the input stream.
710 If the optional argument PROMPT is non-nil, display that as a prompt.
711 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
712 input method is turned on in the current buffer, that input method
713 is used for reading a character.
714 If the optional argument SECONDS is non-nil, it should be a number
715 specifying the maximum number of seconds to wait for input. If no
716 input arrives in that time, return nil. SECONDS may be a
717 floating-point value. */)
718 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
719 {
720 if (! NILP (prompt))
721 message_with_string ("%s", prompt, 0);
722 return read_filtered_event (0, 0, 0, ! NILP (inherit_input_method), seconds);
723 }
724
725 DEFUN ("read-char-exclusive", Fread_char_exclusive, Sread_char_exclusive, 0, 3, 0,
726 doc: /* Read a character from the command input (keyboard or macro).
727 It is returned as a number. Non-character events are ignored.
728 If the character has modifiers, they are resolved and reflected to the
729 character code if possible (e.g. C-SPC -> 0).
730
731 If the optional argument PROMPT is non-nil, display that as a prompt.
732 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
733 input method is turned on in the current buffer, that input method
734 is used for reading a character.
735 If the optional argument SECONDS is non-nil, it should be a number
736 specifying the maximum number of seconds to wait for input. If no
737 input arrives in that time, return nil. SECONDS may be a
738 floating-point value. */)
739 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
740 {
741 Lisp_Object val;
742
743 if (! NILP (prompt))
744 message_with_string ("%s", prompt, 0);
745
746 val = read_filtered_event (1, 1, 0, ! NILP (inherit_input_method), seconds);
747
748 return (NILP (val) ? Qnil
749 : make_number (char_resolve_modifier_mask (XINT (val))));
750 }
751
752 DEFUN ("get-file-char", Fget_file_char, Sget_file_char, 0, 0, 0,
753 doc: /* Don't use this yourself. */)
754 (void)
755 {
756 register Lisp_Object val;
757 block_input ();
758 XSETINT (val, getc (instream));
759 unblock_input ();
760 return val;
761 }
762
763
764 \f
765
766 /* Return true if the lisp code read using READCHARFUN defines a non-nil
767 `lexical-binding' file variable. After returning, the stream is
768 positioned following the first line, if it is a comment or #! line,
769 otherwise nothing is read. */
770
771 static bool
772 lisp_file_lexically_bound_p (Lisp_Object readcharfun)
773 {
774 int ch = READCHAR;
775
776 if (ch == '#')
777 {
778 ch = READCHAR;
779 if (ch != '!')
780 {
781 UNREAD (ch);
782 UNREAD ('#');
783 return 0;
784 }
785 while (ch != '\n' && ch != EOF)
786 ch = READCHAR;
787 if (ch == '\n') ch = READCHAR;
788 /* It is OK to leave the position after a #! line, since
789 that is what read1 does. */
790 }
791
792 if (ch != ';')
793 /* The first line isn't a comment, just give up. */
794 {
795 UNREAD (ch);
796 return 0;
797 }
798 else
799 /* Look for an appropriate file-variable in the first line. */
800 {
801 bool rv = 0;
802 enum {
803 NOMINAL, AFTER_FIRST_DASH, AFTER_ASTERIX
804 } beg_end_state = NOMINAL;
805 bool in_file_vars = 0;
806
807 #define UPDATE_BEG_END_STATE(ch) \
808 if (beg_end_state == NOMINAL) \
809 beg_end_state = (ch == '-' ? AFTER_FIRST_DASH : NOMINAL); \
810 else if (beg_end_state == AFTER_FIRST_DASH) \
811 beg_end_state = (ch == '*' ? AFTER_ASTERIX : NOMINAL); \
812 else if (beg_end_state == AFTER_ASTERIX) \
813 { \
814 if (ch == '-') \
815 in_file_vars = !in_file_vars; \
816 beg_end_state = NOMINAL; \
817 }
818
819 /* Skip until we get to the file vars, if any. */
820 do
821 {
822 ch = READCHAR;
823 UPDATE_BEG_END_STATE (ch);
824 }
825 while (!in_file_vars && ch != '\n' && ch != EOF);
826
827 while (in_file_vars)
828 {
829 char var[100], val[100];
830 unsigned i;
831
832 ch = READCHAR;
833
834 /* Read a variable name. */
835 while (ch == ' ' || ch == '\t')
836 ch = READCHAR;
837
838 i = 0;
839 while (ch != ':' && ch != '\n' && ch != EOF && in_file_vars)
840 {
841 if (i < sizeof var - 1)
842 var[i++] = ch;
843 UPDATE_BEG_END_STATE (ch);
844 ch = READCHAR;
845 }
846
847 /* Stop scanning if no colon was found before end marker. */
848 if (!in_file_vars || ch == '\n' || ch == EOF)
849 break;
850
851 while (i > 0 && (var[i - 1] == ' ' || var[i - 1] == '\t'))
852 i--;
853 var[i] = '\0';
854
855 if (ch == ':')
856 {
857 /* Read a variable value. */
858 ch = READCHAR;
859
860 while (ch == ' ' || ch == '\t')
861 ch = READCHAR;
862
863 i = 0;
864 while (ch != ';' && ch != '\n' && ch != EOF && in_file_vars)
865 {
866 if (i < sizeof val - 1)
867 val[i++] = ch;
868 UPDATE_BEG_END_STATE (ch);
869 ch = READCHAR;
870 }
871 if (! in_file_vars)
872 /* The value was terminated by an end-marker, which remove. */
873 i -= 3;
874 while (i > 0 && (val[i - 1] == ' ' || val[i - 1] == '\t'))
875 i--;
876 val[i] = '\0';
877
878 if (strcmp (var, "lexical-binding") == 0)
879 /* This is it... */
880 {
881 rv = (strcmp (val, "nil") != 0);
882 break;
883 }
884 }
885 }
886
887 while (ch != '\n' && ch != EOF)
888 ch = READCHAR;
889
890 return rv;
891 }
892 }
893 \f
894 /* Value is a version number of byte compiled code if the file
895 associated with file descriptor FD is a compiled Lisp file that's
896 safe to load. Only files compiled with Emacs are safe to load.
897 Files compiled with XEmacs can lead to a crash in Fbyte_code
898 because of an incompatible change in the byte compiler. */
899
900 static int
901 safe_to_load_version (int fd)
902 {
903 char buf[512];
904 int nbytes, i;
905 int version = 1;
906
907 /* Read the first few bytes from the file, and look for a line
908 specifying the byte compiler version used. */
909 nbytes = emacs_read (fd, buf, sizeof buf);
910 if (nbytes > 0)
911 {
912 /* Skip to the next newline, skipping over the initial `ELC'
913 with NUL bytes following it, but note the version. */
914 for (i = 0; i < nbytes && buf[i] != '\n'; ++i)
915 if (i == 4)
916 version = buf[i];
917
918 if (i >= nbytes
919 || fast_c_string_match_ignore_case (Vbytecomp_version_regexp,
920 buf + i, nbytes - i) < 0)
921 version = 0;
922 }
923
924 lseek (fd, 0, SEEK_SET);
925 return version;
926 }
927
928
929 /* Callback for record_unwind_protect. Restore the old load list OLD,
930 after loading a file successfully. */
931
932 static void
933 record_load_unwind (Lisp_Object old)
934 {
935 Vloads_in_progress = old;
936 }
937
938 /* This handler function is used via internal_condition_case_1. */
939
940 static Lisp_Object
941 load_error_handler (Lisp_Object data)
942 {
943 return Qnil;
944 }
945
946 static void
947 load_warn_old_style_backquotes (Lisp_Object file)
948 {
949 if (!NILP (Vold_style_backquotes))
950 {
951 AUTO_STRING (format, "Loading `%s': old-style backquotes detected!");
952 CALLN (Fmessage, format, file);
953 }
954 }
955
956 DEFUN ("get-load-suffixes", Fget_load_suffixes, Sget_load_suffixes, 0, 0, 0,
957 doc: /* Return the suffixes that `load' should try if a suffix is \
958 required.
959 This uses the variables `load-suffixes' and `load-file-rep-suffixes'. */)
960 (void)
961 {
962 Lisp_Object lst = Qnil, suffixes = Vload_suffixes, suffix, ext;
963 while (CONSP (suffixes))
964 {
965 Lisp_Object exts = Vload_file_rep_suffixes;
966 suffix = XCAR (suffixes);
967 suffixes = XCDR (suffixes);
968 while (CONSP (exts))
969 {
970 ext = XCAR (exts);
971 exts = XCDR (exts);
972 lst = Fcons (concat2 (suffix, ext), lst);
973 }
974 }
975 return Fnreverse (lst);
976 }
977
978 /* Returns true if STRING ends with SUFFIX */
979 static bool
980 suffix_p (Lisp_Object string, const char *suffix)
981 {
982 ptrdiff_t suffix_len = strlen (suffix);
983 ptrdiff_t string_len = SBYTES (string);
984
985 return string_len >= suffix_len && !strcmp (SSDATA (string) + string_len - suffix_len, suffix);
986 }
987
988 DEFUN ("load", Fload, Sload, 1, 5, 0,
989 doc: /* Execute a file of Lisp code named FILE.
990 First try FILE with `.elc' appended, then try with `.el', then try
991 with a system-dependent suffix of dynamic modules (see `load-suffixes'),
992 then try FILE unmodified (the exact suffixes in the exact order are
993 determined by `load-suffixes'). Environment variable references in
994 FILE are replaced with their values by calling `substitute-in-file-name'.
995 This function searches the directories in `load-path'.
996
997 If optional second arg NOERROR is non-nil,
998 report no error if FILE doesn't exist.
999 Print messages at start and end of loading unless
1000 optional third arg NOMESSAGE is non-nil (but `force-load-messages'
1001 overrides that).
1002 If optional fourth arg NOSUFFIX is non-nil, don't try adding
1003 suffixes to the specified name FILE.
1004 If optional fifth arg MUST-SUFFIX is non-nil, insist on
1005 the suffix `.elc' or `.el' or the module suffix; don't accept just
1006 FILE unless it ends in one of those suffixes or includes a directory name.
1007
1008 If NOSUFFIX is nil, then if a file could not be found, try looking for
1009 a different representation of the file by adding non-empty suffixes to
1010 its name, before trying another file. Emacs uses this feature to find
1011 compressed versions of files when Auto Compression mode is enabled.
1012 If NOSUFFIX is non-nil, disable this feature.
1013
1014 The suffixes that this function tries out, when NOSUFFIX is nil, are
1015 given by the return value of `get-load-suffixes' and the values listed
1016 in `load-file-rep-suffixes'. If MUST-SUFFIX is non-nil, only the
1017 return value of `get-load-suffixes' is used, i.e. the file name is
1018 required to have a non-empty suffix.
1019
1020 When searching suffixes, this function normally stops at the first
1021 one that exists. If the option `load-prefer-newer' is non-nil,
1022 however, it tries all suffixes, and uses whichever file is the newest.
1023
1024 Loading a file records its definitions, and its `provide' and
1025 `require' calls, in an element of `load-history' whose
1026 car is the file name loaded. See `load-history'.
1027
1028 While the file is in the process of being loaded, the variable
1029 `load-in-progress' is non-nil and the variable `load-file-name'
1030 is bound to the file's name.
1031
1032 Return t if the file exists and loads successfully. */)
1033 (Lisp_Object file, Lisp_Object noerror, Lisp_Object nomessage,
1034 Lisp_Object nosuffix, Lisp_Object must_suffix)
1035 {
1036 FILE *stream;
1037 int fd;
1038 int fd_index;
1039 ptrdiff_t count = SPECPDL_INDEX ();
1040 Lisp_Object found, efound, hist_file_name;
1041 /* True means we printed the ".el is newer" message. */
1042 bool newer = 0;
1043 /* True means we are loading a compiled file. */
1044 bool compiled = 0;
1045 Lisp_Object handler;
1046 bool safe_p = 1;
1047 const char *fmode = "r" FOPEN_TEXT;
1048 int version;
1049
1050 CHECK_STRING (file);
1051
1052 /* If file name is magic, call the handler. */
1053 /* This shouldn't be necessary any more now that `openp' handles it right.
1054 handler = Ffind_file_name_handler (file, Qload);
1055 if (!NILP (handler))
1056 return call5 (handler, Qload, file, noerror, nomessage, nosuffix); */
1057
1058 /* The presence of this call is the result of a historical accident:
1059 it used to be in every file-operation and when it got removed
1060 everywhere, it accidentally stayed here. Since then, enough people
1061 supposedly have things like (load "$PROJECT/foo.el") in their .emacs
1062 that it seemed risky to remove. */
1063 if (! NILP (noerror))
1064 {
1065 file = internal_condition_case_1 (Fsubstitute_in_file_name, file,
1066 Qt, load_error_handler);
1067 if (NILP (file))
1068 return Qnil;
1069 }
1070 else
1071 file = Fsubstitute_in_file_name (file);
1072
1073 /* Avoid weird lossage with null string as arg,
1074 since it would try to load a directory as a Lisp file. */
1075 if (SCHARS (file) == 0)
1076 {
1077 fd = -1;
1078 errno = ENOENT;
1079 }
1080 else
1081 {
1082 Lisp_Object suffixes;
1083 found = Qnil;
1084
1085 if (! NILP (must_suffix))
1086 {
1087 /* Don't insist on adding a suffix if FILE already ends with one. */
1088 if (suffix_p (file, ".el")
1089 || suffix_p (file, ".elc")
1090 #ifdef HAVE_MODULES
1091 || suffix_p (file, MODULES_SUFFIX)
1092 #endif
1093 )
1094 must_suffix = Qnil;
1095 /* Don't insist on adding a suffix
1096 if the argument includes a directory name. */
1097 else if (! NILP (Ffile_name_directory (file)))
1098 must_suffix = Qnil;
1099 }
1100
1101 if (!NILP (nosuffix))
1102 suffixes = Qnil;
1103 else
1104 {
1105 suffixes = Fget_load_suffixes ();
1106 if (NILP (must_suffix))
1107 suffixes = CALLN (Fappend, suffixes, Vload_file_rep_suffixes);
1108 }
1109
1110 fd = openp (Vload_path, file, suffixes, &found, Qnil, load_prefer_newer);
1111 }
1112
1113 if (fd == -1)
1114 {
1115 if (NILP (noerror))
1116 report_file_error ("Cannot open load file", file);
1117 return Qnil;
1118 }
1119
1120 /* Tell startup.el whether or not we found the user's init file. */
1121 if (EQ (Qt, Vuser_init_file))
1122 Vuser_init_file = found;
1123
1124 /* If FD is -2, that means openp found a magic file. */
1125 if (fd == -2)
1126 {
1127 if (NILP (Fequal (found, file)))
1128 /* If FOUND is a different file name from FILE,
1129 find its handler even if we have already inhibited
1130 the `load' operation on FILE. */
1131 handler = Ffind_file_name_handler (found, Qt);
1132 else
1133 handler = Ffind_file_name_handler (found, Qload);
1134 if (! NILP (handler))
1135 return call5 (handler, Qload, found, noerror, nomessage, Qt);
1136 #ifdef DOS_NT
1137 /* Tramp has to deal with semi-broken packages that prepend
1138 drive letters to remote files. For that reason, Tramp
1139 catches file operations that test for file existence, which
1140 makes openp think X:/foo.elc files are remote. However,
1141 Tramp does not catch `load' operations for such files, so we
1142 end up with a nil as the `load' handler above. If we would
1143 continue with fd = -2, we will behave wrongly, and in
1144 particular try reading a .elc file in the "rt" mode instead
1145 of "rb". See bug #9311 for the results. To work around
1146 this, we try to open the file locally, and go with that if it
1147 succeeds. */
1148 fd = emacs_open (SSDATA (ENCODE_FILE (found)), O_RDONLY, 0);
1149 if (fd == -1)
1150 fd = -2;
1151 #endif
1152 }
1153
1154 if (fd < 0)
1155 {
1156 /* Pacify older GCC with --enable-gcc-warnings. */
1157 IF_LINT (fd_index = 0);
1158 }
1159 else
1160 {
1161 fd_index = SPECPDL_INDEX ();
1162 record_unwind_protect_int (close_file_unwind, fd);
1163 }
1164
1165 #ifdef HAVE_MODULES
1166 if (suffix_p (found, MODULES_SUFFIX))
1167 return unbind_to (count, Fmodule_load (found));
1168 #endif
1169
1170 /* Check if we're stuck in a recursive load cycle.
1171
1172 2000-09-21: It's not possible to just check for the file loaded
1173 being a member of Vloads_in_progress. This fails because of the
1174 way the byte compiler currently works; `provide's are not
1175 evaluated, see font-lock.el/jit-lock.el as an example. This
1176 leads to a certain amount of ``normal'' recursion.
1177
1178 Also, just loading a file recursively is not always an error in
1179 the general case; the second load may do something different. */
1180 {
1181 int load_count = 0;
1182 Lisp_Object tem;
1183 for (tem = Vloads_in_progress; CONSP (tem); tem = XCDR (tem))
1184 if (!NILP (Fequal (found, XCAR (tem))) && (++load_count > 3))
1185 signal_error ("Recursive load", Fcons (found, Vloads_in_progress));
1186 record_unwind_protect (record_load_unwind, Vloads_in_progress);
1187 Vloads_in_progress = Fcons (found, Vloads_in_progress);
1188 }
1189
1190 /* All loads are by default dynamic, unless the file itself specifies
1191 otherwise using a file-variable in the first line. This is bound here
1192 so that it takes effect whether or not we use
1193 Vload_source_file_function. */
1194 specbind (Qlexical_binding, Qnil);
1195
1196 /* Get the name for load-history. */
1197 hist_file_name = (! NILP (Vpurify_flag)
1198 ? concat2 (Ffile_name_directory (file),
1199 Ffile_name_nondirectory (found))
1200 : found) ;
1201
1202 version = -1;
1203
1204 /* Check for the presence of old-style quotes and warn about them. */
1205 specbind (Qold_style_backquotes, Qnil);
1206 record_unwind_protect (load_warn_old_style_backquotes, file);
1207
1208 if (suffix_p (found, ".elc") || (fd >= 0 && (version = safe_to_load_version (fd)) > 0))
1209 /* Load .elc files directly, but not when they are
1210 remote and have no handler! */
1211 {
1212 if (fd != -2)
1213 {
1214 struct stat s1, s2;
1215 int result;
1216
1217 if (version < 0
1218 && ! (version = safe_to_load_version (fd)))
1219 {
1220 safe_p = 0;
1221 if (!load_dangerous_libraries)
1222 error ("File `%s' was not compiled in Emacs", SDATA (found));
1223 else if (!NILP (nomessage) && !force_load_messages)
1224 message_with_string ("File `%s' not compiled in Emacs", found, 1);
1225 }
1226
1227 compiled = 1;
1228
1229 efound = ENCODE_FILE (found);
1230 fmode = "r" FOPEN_BINARY;
1231
1232 /* openp already checked for newness, no point doing it again.
1233 FIXME would be nice to get a message when openp
1234 ignores suffix order due to load_prefer_newer. */
1235 if (!load_prefer_newer)
1236 {
1237 result = stat (SSDATA (efound), &s1);
1238 if (result == 0)
1239 {
1240 SSET (efound, SBYTES (efound) - 1, 0);
1241 result = stat (SSDATA (efound), &s2);
1242 SSET (efound, SBYTES (efound) - 1, 'c');
1243 }
1244
1245 if (result == 0
1246 && timespec_cmp (get_stat_mtime (&s1), get_stat_mtime (&s2)) < 0)
1247 {
1248 /* Make the progress messages mention that source is newer. */
1249 newer = 1;
1250
1251 /* If we won't print another message, mention this anyway. */
1252 if (!NILP (nomessage) && !force_load_messages)
1253 {
1254 Lisp_Object msg_file;
1255 msg_file = Fsubstring (found, make_number (0), make_number (-1));
1256 message_with_string ("Source file `%s' newer than byte-compiled file",
1257 msg_file, 1);
1258 }
1259 }
1260 } /* !load_prefer_newer */
1261 }
1262 }
1263 else
1264 {
1265 /* We are loading a source file (*.el). */
1266 if (!NILP (Vload_source_file_function))
1267 {
1268 Lisp_Object val;
1269
1270 if (fd >= 0)
1271 {
1272 emacs_close (fd);
1273 clear_unwind_protect (fd_index);
1274 }
1275 val = call4 (Vload_source_file_function, found, hist_file_name,
1276 NILP (noerror) ? Qnil : Qt,
1277 (NILP (nomessage) || force_load_messages) ? Qnil : Qt);
1278 return unbind_to (count, val);
1279 }
1280 }
1281
1282 if (fd < 0)
1283 {
1284 /* We somehow got here with fd == -2, meaning the file is deemed
1285 to be remote. Don't even try to reopen the file locally;
1286 just force a failure. */
1287 stream = NULL;
1288 errno = EINVAL;
1289 }
1290 else
1291 {
1292 #ifdef WINDOWSNT
1293 emacs_close (fd);
1294 clear_unwind_protect (fd_index);
1295 efound = ENCODE_FILE (found);
1296 stream = emacs_fopen (SSDATA (efound), fmode);
1297 #else
1298 stream = fdopen (fd, fmode);
1299 #endif
1300 }
1301 if (! stream)
1302 report_file_error ("Opening stdio stream", file);
1303 set_unwind_protect_ptr (fd_index, fclose_unwind, stream);
1304
1305 if (! NILP (Vpurify_flag))
1306 Vpreloaded_file_list = Fcons (Fpurecopy (file), Vpreloaded_file_list);
1307
1308 if (NILP (nomessage) || force_load_messages)
1309 {
1310 if (!safe_p)
1311 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...",
1312 file, 1);
1313 else if (!compiled)
1314 message_with_string ("Loading %s (source)...", file, 1);
1315 else if (newer)
1316 message_with_string ("Loading %s (compiled; note, source file is newer)...",
1317 file, 1);
1318 else /* The typical case; compiled file newer than source file. */
1319 message_with_string ("Loading %s...", file, 1);
1320 }
1321
1322 specbind (Qload_file_name, found);
1323 specbind (Qinhibit_file_name_operation, Qnil);
1324 specbind (Qload_in_progress, Qt);
1325
1326 instream = stream;
1327 if (lisp_file_lexically_bound_p (Qget_file_char))
1328 Fset (Qlexical_binding, Qt);
1329
1330 if (! version || version >= 22)
1331 readevalloop (Qget_file_char, stream, hist_file_name,
1332 0, Qnil, Qnil, Qnil, Qnil);
1333 else
1334 {
1335 /* We can't handle a file which was compiled with
1336 byte-compile-dynamic by older version of Emacs. */
1337 specbind (Qload_force_doc_strings, Qt);
1338 readevalloop (Qget_emacs_mule_file_char, stream, hist_file_name,
1339 0, Qnil, Qnil, Qnil, Qnil);
1340 }
1341 unbind_to (count, Qnil);
1342
1343 /* Run any eval-after-load forms for this file. */
1344 if (!NILP (Ffboundp (Qdo_after_load_evaluation)))
1345 call1 (Qdo_after_load_evaluation, hist_file_name) ;
1346
1347 xfree (saved_doc_string);
1348 saved_doc_string = 0;
1349 saved_doc_string_size = 0;
1350
1351 xfree (prev_saved_doc_string);
1352 prev_saved_doc_string = 0;
1353 prev_saved_doc_string_size = 0;
1354
1355 if (!noninteractive && (NILP (nomessage) || force_load_messages))
1356 {
1357 if (!safe_p)
1358 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...done",
1359 file, 1);
1360 else if (!compiled)
1361 message_with_string ("Loading %s (source)...done", file, 1);
1362 else if (newer)
1363 message_with_string ("Loading %s (compiled; note, source file is newer)...done",
1364 file, 1);
1365 else /* The typical case; compiled file newer than source file. */
1366 message_with_string ("Loading %s...done", file, 1);
1367 }
1368
1369 return Qt;
1370 }
1371 \f
1372 static bool
1373 complete_filename_p (Lisp_Object pathname)
1374 {
1375 const unsigned char *s = SDATA (pathname);
1376 return (IS_DIRECTORY_SEP (s[0])
1377 || (SCHARS (pathname) > 2
1378 && IS_DEVICE_SEP (s[1]) && IS_DIRECTORY_SEP (s[2])));
1379 }
1380
1381 DEFUN ("locate-file-internal", Flocate_file_internal, Slocate_file_internal, 2, 4, 0,
1382 doc: /* Search for FILENAME through PATH.
1383 Returns the file's name in absolute form, or nil if not found.
1384 If SUFFIXES is non-nil, it should be a list of suffixes to append to
1385 file name when searching.
1386 If non-nil, PREDICATE is used instead of `file-readable-p'.
1387 PREDICATE can also be an integer to pass to the faccessat(2) function,
1388 in which case file-name-handlers are ignored.
1389 This function will normally skip directories, so if you want it to find
1390 directories, make sure the PREDICATE function returns `dir-ok' for them. */)
1391 (Lisp_Object filename, Lisp_Object path, Lisp_Object suffixes, Lisp_Object predicate)
1392 {
1393 Lisp_Object file;
1394 int fd = openp (path, filename, suffixes, &file, predicate, false);
1395 if (NILP (predicate) && fd >= 0)
1396 emacs_close (fd);
1397 return file;
1398 }
1399
1400 /* Search for a file whose name is STR, looking in directories
1401 in the Lisp list PATH, and trying suffixes from SUFFIX.
1402 On success, return a file descriptor (or 1 or -2 as described below).
1403 On failure, return -1 and set errno.
1404
1405 SUFFIXES is a list of strings containing possible suffixes.
1406 The empty suffix is automatically added if the list is empty.
1407
1408 PREDICATE t means the files are binary.
1409 PREDICATE non-nil and non-t means don't open the files,
1410 just look for one that satisfies the predicate. In this case,
1411 return 1 on success. The predicate can be a lisp function or
1412 an integer to pass to `access' (in which case file-name-handlers
1413 are ignored).
1414
1415 If STOREPTR is nonzero, it points to a slot where the name of
1416 the file actually found should be stored as a Lisp string.
1417 nil is stored there on failure.
1418
1419 If the file we find is remote, return -2
1420 but store the found remote file name in *STOREPTR.
1421
1422 If NEWER is true, try all SUFFIXes and return the result for the
1423 newest file that exists. Does not apply to remote files,
1424 or if a non-nil and non-t PREDICATE is specified. */
1425
1426 int
1427 openp (Lisp_Object path, Lisp_Object str, Lisp_Object suffixes,
1428 Lisp_Object *storeptr, Lisp_Object predicate, bool newer)
1429 {
1430 ptrdiff_t fn_size = 100;
1431 char buf[100];
1432 char *fn = buf;
1433 bool absolute;
1434 ptrdiff_t want_length;
1435 Lisp_Object filename;
1436 Lisp_Object string, tail, encoded_fn, save_string;
1437 ptrdiff_t max_suffix_len = 0;
1438 int last_errno = ENOENT;
1439 int save_fd = -1;
1440 USE_SAFE_ALLOCA;
1441
1442 /* The last-modified time of the newest matching file found.
1443 Initialize it to something less than all valid timestamps. */
1444 struct timespec save_mtime = make_timespec (TYPE_MINIMUM (time_t), -1);
1445
1446 CHECK_STRING (str);
1447
1448 for (tail = suffixes; CONSP (tail); tail = XCDR (tail))
1449 {
1450 CHECK_STRING_CAR (tail);
1451 max_suffix_len = max (max_suffix_len,
1452 SBYTES (XCAR (tail)));
1453 }
1454
1455 string = filename = encoded_fn = save_string = Qnil;
1456
1457 if (storeptr)
1458 *storeptr = Qnil;
1459
1460 absolute = complete_filename_p (str);
1461
1462 for (; CONSP (path); path = XCDR (path))
1463 {
1464 filename = Fexpand_file_name (str, XCAR (path));
1465 if (!complete_filename_p (filename))
1466 /* If there are non-absolute elts in PATH (eg "."). */
1467 /* Of course, this could conceivably lose if luser sets
1468 default-directory to be something non-absolute... */
1469 {
1470 filename = Fexpand_file_name (filename, BVAR (current_buffer, directory));
1471 if (!complete_filename_p (filename))
1472 /* Give up on this path element! */
1473 continue;
1474 }
1475
1476 /* Calculate maximum length of any filename made from
1477 this path element/specified file name and any possible suffix. */
1478 want_length = max_suffix_len + SBYTES (filename);
1479 if (fn_size <= want_length)
1480 {
1481 fn_size = 100 + want_length;
1482 fn = SAFE_ALLOCA (fn_size);
1483 }
1484
1485 /* Loop over suffixes. */
1486 for (tail = NILP (suffixes) ? list1 (empty_unibyte_string) : suffixes;
1487 CONSP (tail); tail = XCDR (tail))
1488 {
1489 Lisp_Object suffix = XCAR (tail);
1490 ptrdiff_t fnlen, lsuffix = SBYTES (suffix);
1491 Lisp_Object handler;
1492
1493 /* Concatenate path element/specified name with the suffix.
1494 If the directory starts with /:, remove that. */
1495 int prefixlen = ((SCHARS (filename) > 2
1496 && SREF (filename, 0) == '/'
1497 && SREF (filename, 1) == ':')
1498 ? 2 : 0);
1499 fnlen = SBYTES (filename) - prefixlen;
1500 memcpy (fn, SDATA (filename) + prefixlen, fnlen);
1501 memcpy (fn + fnlen, SDATA (suffix), lsuffix + 1);
1502 fnlen += lsuffix;
1503 /* Check that the file exists and is not a directory. */
1504 /* We used to only check for handlers on non-absolute file names:
1505 if (absolute)
1506 handler = Qnil;
1507 else
1508 handler = Ffind_file_name_handler (filename, Qfile_exists_p);
1509 It's not clear why that was the case and it breaks things like
1510 (load "/bar.el") where the file is actually "/bar.el.gz". */
1511 /* make_string has its own ideas on when to return a unibyte
1512 string and when a multibyte string, but we know better.
1513 We must have a unibyte string when dumping, since
1514 file-name encoding is shaky at best at that time, and in
1515 particular default-file-name-coding-system is reset
1516 several times during loadup. We therefore don't want to
1517 encode the file before passing it to file I/O library
1518 functions. */
1519 if (!STRING_MULTIBYTE (filename) && !STRING_MULTIBYTE (suffix))
1520 string = make_unibyte_string (fn, fnlen);
1521 else
1522 string = make_string (fn, fnlen);
1523 handler = Ffind_file_name_handler (string, Qfile_exists_p);
1524 if ((!NILP (handler) || (!NILP (predicate) && !EQ (predicate, Qt)))
1525 && !NATNUMP (predicate))
1526 {
1527 bool exists;
1528 if (NILP (predicate) || EQ (predicate, Qt))
1529 exists = !NILP (Ffile_readable_p (string));
1530 else
1531 {
1532 Lisp_Object tmp = call1 (predicate, string);
1533 if (NILP (tmp))
1534 exists = false;
1535 else if (EQ (tmp, Qdir_ok)
1536 || NILP (Ffile_directory_p (string)))
1537 exists = true;
1538 else
1539 {
1540 exists = false;
1541 last_errno = EISDIR;
1542 }
1543 }
1544
1545 if (exists)
1546 {
1547 /* We succeeded; return this descriptor and filename. */
1548 if (storeptr)
1549 *storeptr = string;
1550 SAFE_FREE ();
1551 return -2;
1552 }
1553 }
1554 else
1555 {
1556 int fd;
1557 const char *pfn;
1558 struct stat st;
1559
1560 encoded_fn = ENCODE_FILE (string);
1561 pfn = SSDATA (encoded_fn);
1562
1563 /* Check that we can access or open it. */
1564 if (NATNUMP (predicate))
1565 {
1566 fd = -1;
1567 if (INT_MAX < XFASTINT (predicate))
1568 last_errno = EINVAL;
1569 else if (faccessat (AT_FDCWD, pfn, XFASTINT (predicate),
1570 AT_EACCESS)
1571 == 0)
1572 {
1573 if (file_directory_p (pfn))
1574 last_errno = EISDIR;
1575 else
1576 fd = 1;
1577 }
1578 }
1579 else
1580 {
1581 int oflags = O_RDONLY + (NILP (predicate) ? 0 : O_BINARY);
1582 fd = emacs_open (pfn, oflags, 0);
1583 if (fd < 0)
1584 {
1585 if (errno != ENOENT)
1586 last_errno = errno;
1587 }
1588 else
1589 {
1590 int err = (fstat (fd, &st) != 0 ? errno
1591 : S_ISDIR (st.st_mode) ? EISDIR : 0);
1592 if (err)
1593 {
1594 last_errno = err;
1595 emacs_close (fd);
1596 fd = -1;
1597 }
1598 }
1599 }
1600
1601 if (fd >= 0)
1602 {
1603 if (newer && !NATNUMP (predicate))
1604 {
1605 struct timespec mtime = get_stat_mtime (&st);
1606
1607 if (timespec_cmp (mtime, save_mtime) <= 0)
1608 emacs_close (fd);
1609 else
1610 {
1611 if (0 <= save_fd)
1612 emacs_close (save_fd);
1613 save_fd = fd;
1614 save_mtime = mtime;
1615 save_string = string;
1616 }
1617 }
1618 else
1619 {
1620 /* We succeeded; return this descriptor and filename. */
1621 if (storeptr)
1622 *storeptr = string;
1623 SAFE_FREE ();
1624 return fd;
1625 }
1626 }
1627
1628 /* No more suffixes. Return the newest. */
1629 if (0 <= save_fd && ! CONSP (XCDR (tail)))
1630 {
1631 if (storeptr)
1632 *storeptr = save_string;
1633 SAFE_FREE ();
1634 return save_fd;
1635 }
1636 }
1637 }
1638 if (absolute)
1639 break;
1640 }
1641
1642 SAFE_FREE ();
1643 errno = last_errno;
1644 return -1;
1645 }
1646
1647 \f
1648 /* Merge the list we've accumulated of globals from the current input source
1649 into the load_history variable. The details depend on whether
1650 the source has an associated file name or not.
1651
1652 FILENAME is the file name that we are loading from.
1653
1654 ENTIRE is true if loading that entire file, false if evaluating
1655 part of it. */
1656
1657 static void
1658 build_load_history (Lisp_Object filename, bool entire)
1659 {
1660 Lisp_Object tail, prev, newelt;
1661 Lisp_Object tem, tem2;
1662 bool foundit = 0;
1663
1664 tail = Vload_history;
1665 prev = Qnil;
1666
1667 while (CONSP (tail))
1668 {
1669 tem = XCAR (tail);
1670
1671 /* Find the feature's previous assoc list... */
1672 if (!NILP (Fequal (filename, Fcar (tem))))
1673 {
1674 foundit = 1;
1675
1676 /* If we're loading the entire file, remove old data. */
1677 if (entire)
1678 {
1679 if (NILP (prev))
1680 Vload_history = XCDR (tail);
1681 else
1682 Fsetcdr (prev, XCDR (tail));
1683 }
1684
1685 /* Otherwise, cons on new symbols that are not already members. */
1686 else
1687 {
1688 tem2 = Vcurrent_load_list;
1689
1690 while (CONSP (tem2))
1691 {
1692 newelt = XCAR (tem2);
1693
1694 if (NILP (Fmember (newelt, tem)))
1695 Fsetcar (tail, Fcons (XCAR (tem),
1696 Fcons (newelt, XCDR (tem))));
1697
1698 tem2 = XCDR (tem2);
1699 QUIT;
1700 }
1701 }
1702 }
1703 else
1704 prev = tail;
1705 tail = XCDR (tail);
1706 QUIT;
1707 }
1708
1709 /* If we're loading an entire file, cons the new assoc onto the
1710 front of load-history, the most-recently-loaded position. Also
1711 do this if we didn't find an existing member for the file. */
1712 if (entire || !foundit)
1713 Vload_history = Fcons (Fnreverse (Vcurrent_load_list),
1714 Vload_history);
1715 }
1716
1717 static void
1718 readevalloop_1 (int old)
1719 {
1720 load_convert_to_unibyte = old;
1721 }
1722
1723 /* Signal an `end-of-file' error, if possible with file name
1724 information. */
1725
1726 static _Noreturn void
1727 end_of_file_error (void)
1728 {
1729 if (STRINGP (Vload_file_name))
1730 xsignal1 (Qend_of_file, Vload_file_name);
1731
1732 xsignal0 (Qend_of_file);
1733 }
1734
1735 static Lisp_Object
1736 readevalloop_eager_expand_eval (Lisp_Object val, Lisp_Object macroexpand)
1737 {
1738 /* If we macroexpand the toplevel form non-recursively and it ends
1739 up being a `progn' (or if it was a progn to start), treat each
1740 form in the progn as a top-level form. This way, if one form in
1741 the progn defines a macro, that macro is in effect when we expand
1742 the remaining forms. See similar code in bytecomp.el. */
1743 val = call2 (macroexpand, val, Qnil);
1744 if (EQ (CAR_SAFE (val), Qprogn))
1745 {
1746 Lisp_Object subforms = XCDR (val);
1747
1748 for (val = Qnil; CONSP (subforms); subforms = XCDR (subforms))
1749 val = readevalloop_eager_expand_eval (XCAR (subforms),
1750 macroexpand);
1751 }
1752 else
1753 val = eval_sub (call2 (macroexpand, val, Qt));
1754 return val;
1755 }
1756
1757 /* UNIBYTE specifies how to set load_convert_to_unibyte
1758 for this invocation.
1759 READFUN, if non-nil, is used instead of `read'.
1760
1761 START, END specify region to read in current buffer (from eval-region).
1762 If the input is not from a buffer, they must be nil. */
1763
1764 static void
1765 readevalloop (Lisp_Object readcharfun,
1766 FILE *stream,
1767 Lisp_Object sourcename,
1768 bool printflag,
1769 Lisp_Object unibyte, Lisp_Object readfun,
1770 Lisp_Object start, Lisp_Object end)
1771 {
1772 int c;
1773 Lisp_Object val;
1774 ptrdiff_t count = SPECPDL_INDEX ();
1775 struct buffer *b = 0;
1776 bool continue_reading_p;
1777 Lisp_Object lex_bound;
1778 /* True if reading an entire buffer. */
1779 bool whole_buffer = 0;
1780 /* True on the first time around. */
1781 bool first_sexp = 1;
1782 Lisp_Object macroexpand = intern ("internal-macroexpand-for-load");
1783
1784 if (NILP (Ffboundp (macroexpand))
1785 /* Don't macroexpand in .elc files, since it should have been done
1786 already. We actually don't know whether we're in a .elc file or not,
1787 so we use circumstantial evidence: .el files normally go through
1788 Vload_source_file_function -> load-with-code-conversion
1789 -> eval-buffer. */
1790 || EQ (readcharfun, Qget_file_char)
1791 || EQ (readcharfun, Qget_emacs_mule_file_char))
1792 macroexpand = Qnil;
1793
1794 if (MARKERP (readcharfun))
1795 {
1796 if (NILP (start))
1797 start = readcharfun;
1798 }
1799
1800 if (BUFFERP (readcharfun))
1801 b = XBUFFER (readcharfun);
1802 else if (MARKERP (readcharfun))
1803 b = XMARKER (readcharfun)->buffer;
1804
1805 /* We assume START is nil when input is not from a buffer. */
1806 if (! NILP (start) && !b)
1807 emacs_abort ();
1808
1809 specbind (Qstandard_input, readcharfun);
1810 specbind (Qcurrent_load_list, Qnil);
1811 record_unwind_protect_int (readevalloop_1, load_convert_to_unibyte);
1812 load_convert_to_unibyte = !NILP (unibyte);
1813
1814 /* If lexical binding is active (either because it was specified in
1815 the file's header, or via a buffer-local variable), create an empty
1816 lexical environment, otherwise, turn off lexical binding. */
1817 lex_bound = find_symbol_value (Qlexical_binding);
1818 specbind (Qinternal_interpreter_environment,
1819 (NILP (lex_bound) || EQ (lex_bound, Qunbound)
1820 ? Qnil : list1 (Qt)));
1821
1822 /* Try to ensure sourcename is a truename, except whilst preloading. */
1823 if (NILP (Vpurify_flag)
1824 && !NILP (sourcename) && !NILP (Ffile_name_absolute_p (sourcename))
1825 && !NILP (Ffboundp (Qfile_truename)))
1826 sourcename = call1 (Qfile_truename, sourcename) ;
1827
1828 LOADHIST_ATTACH (sourcename);
1829
1830 continue_reading_p = 1;
1831 while (continue_reading_p)
1832 {
1833 ptrdiff_t count1 = SPECPDL_INDEX ();
1834
1835 if (b != 0 && !BUFFER_LIVE_P (b))
1836 error ("Reading from killed buffer");
1837
1838 if (!NILP (start))
1839 {
1840 /* Switch to the buffer we are reading from. */
1841 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1842 set_buffer_internal (b);
1843
1844 /* Save point in it. */
1845 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1846 /* Save ZV in it. */
1847 record_unwind_protect (save_restriction_restore, save_restriction_save ());
1848 /* Those get unbound after we read one expression. */
1849
1850 /* Set point and ZV around stuff to be read. */
1851 Fgoto_char (start);
1852 if (!NILP (end))
1853 Fnarrow_to_region (make_number (BEGV), end);
1854
1855 /* Just for cleanliness, convert END to a marker
1856 if it is an integer. */
1857 if (INTEGERP (end))
1858 end = Fpoint_max_marker ();
1859 }
1860
1861 /* On the first cycle, we can easily test here
1862 whether we are reading the whole buffer. */
1863 if (b && first_sexp)
1864 whole_buffer = (PT == BEG && ZV == Z);
1865
1866 instream = stream;
1867 read_next:
1868 c = READCHAR;
1869 if (c == ';')
1870 {
1871 while ((c = READCHAR) != '\n' && c != -1);
1872 goto read_next;
1873 }
1874 if (c < 0)
1875 {
1876 unbind_to (count1, Qnil);
1877 break;
1878 }
1879
1880 /* Ignore whitespace here, so we can detect eof. */
1881 if (c == ' ' || c == '\t' || c == '\n' || c == '\f' || c == '\r'
1882 || c == NO_BREAK_SPACE)
1883 goto read_next;
1884
1885 if (!NILP (Vpurify_flag) && c == '(')
1886 {
1887 val = read_list (0, readcharfun);
1888 }
1889 else
1890 {
1891 UNREAD (c);
1892 read_objects = Qnil;
1893 if (!NILP (readfun))
1894 {
1895 val = call1 (readfun, readcharfun);
1896
1897 /* If READCHARFUN has set point to ZV, we should
1898 stop reading, even if the form read sets point
1899 to a different value when evaluated. */
1900 if (BUFFERP (readcharfun))
1901 {
1902 struct buffer *buf = XBUFFER (readcharfun);
1903 if (BUF_PT (buf) == BUF_ZV (buf))
1904 continue_reading_p = 0;
1905 }
1906 }
1907 else if (! NILP (Vload_read_function))
1908 val = call1 (Vload_read_function, readcharfun);
1909 else
1910 val = read_internal_start (readcharfun, Qnil, Qnil);
1911 }
1912
1913 if (!NILP (start) && continue_reading_p)
1914 start = Fpoint_marker ();
1915
1916 /* Restore saved point and BEGV. */
1917 unbind_to (count1, Qnil);
1918
1919 /* Now eval what we just read. */
1920 if (!NILP (macroexpand))
1921 val = readevalloop_eager_expand_eval (val, macroexpand);
1922 else
1923 val = eval_sub (val);
1924
1925 if (printflag)
1926 {
1927 Vvalues = Fcons (val, Vvalues);
1928 if (EQ (Vstandard_output, Qt))
1929 Fprin1 (val, Qnil);
1930 else
1931 Fprint (val, Qnil);
1932 }
1933
1934 first_sexp = 0;
1935 }
1936
1937 build_load_history (sourcename,
1938 stream || whole_buffer);
1939
1940 unbind_to (count, Qnil);
1941 }
1942
1943 DEFUN ("eval-buffer", Feval_buffer, Seval_buffer, 0, 5, "",
1944 doc: /* Execute the accessible portion of current buffer as Lisp code.
1945 You can use \\[narrow-to-region] to limit the part of buffer to be evaluated.
1946 When called from a Lisp program (i.e., not interactively), this
1947 function accepts up to five optional arguments:
1948 BUFFER is the buffer to evaluate (nil means use current buffer),
1949 or a name of a buffer (a string).
1950 PRINTFLAG controls printing of output by any output functions in the
1951 evaluated code, such as `print', `princ', and `prin1':
1952 a value of nil means discard it; anything else is the stream to print to.
1953 See Info node `(elisp)Output Streams' for details on streams.
1954 FILENAME specifies the file name to use for `load-history'.
1955 UNIBYTE, if non-nil, specifies `load-convert-to-unibyte' for this
1956 invocation.
1957 DO-ALLOW-PRINT, if non-nil, specifies that output functions in the
1958 evaluated code should work normally even if PRINTFLAG is nil, in
1959 which case the output is displayed in the echo area.
1960
1961 This function preserves the position of point. */)
1962 (Lisp_Object buffer, Lisp_Object printflag, Lisp_Object filename, Lisp_Object unibyte, Lisp_Object do_allow_print)
1963 {
1964 ptrdiff_t count = SPECPDL_INDEX ();
1965 Lisp_Object tem, buf;
1966
1967 if (NILP (buffer))
1968 buf = Fcurrent_buffer ();
1969 else
1970 buf = Fget_buffer (buffer);
1971 if (NILP (buf))
1972 error ("No such buffer");
1973
1974 if (NILP (printflag) && NILP (do_allow_print))
1975 tem = Qsymbolp;
1976 else
1977 tem = printflag;
1978
1979 if (NILP (filename))
1980 filename = BVAR (XBUFFER (buf), filename);
1981
1982 specbind (Qeval_buffer_list, Fcons (buf, Veval_buffer_list));
1983 specbind (Qstandard_output, tem);
1984 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1985 BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
1986 specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
1987 readevalloop (buf, 0, filename,
1988 !NILP (printflag), unibyte, Qnil, Qnil, Qnil);
1989 unbind_to (count, Qnil);
1990
1991 return Qnil;
1992 }
1993
1994 DEFUN ("eval-region", Feval_region, Seval_region, 2, 4, "r",
1995 doc: /* Execute the region as Lisp code.
1996 When called from programs, expects two arguments,
1997 giving starting and ending indices in the current buffer
1998 of the text to be executed.
1999 Programs can pass third argument PRINTFLAG which controls output:
2000 a value of nil means discard it; anything else is stream for printing it.
2001 See Info node `(elisp)Output Streams' for details on streams.
2002 Also the fourth argument READ-FUNCTION, if non-nil, is used
2003 instead of `read' to read each expression. It gets one argument
2004 which is the input stream for reading characters.
2005
2006 This function does not move point. */)
2007 (Lisp_Object start, Lisp_Object end, Lisp_Object printflag, Lisp_Object read_function)
2008 {
2009 /* FIXME: Do the eval-sexp-add-defvars dance! */
2010 ptrdiff_t count = SPECPDL_INDEX ();
2011 Lisp_Object tem, cbuf;
2012
2013 cbuf = Fcurrent_buffer ();
2014
2015 if (NILP (printflag))
2016 tem = Qsymbolp;
2017 else
2018 tem = printflag;
2019 specbind (Qstandard_output, tem);
2020 specbind (Qeval_buffer_list, Fcons (cbuf, Veval_buffer_list));
2021
2022 /* `readevalloop' calls functions which check the type of start and end. */
2023 readevalloop (cbuf, 0, BVAR (XBUFFER (cbuf), filename),
2024 !NILP (printflag), Qnil, read_function,
2025 start, end);
2026
2027 return unbind_to (count, Qnil);
2028 }
2029
2030 \f
2031 DEFUN ("read", Fread, Sread, 0, 1, 0,
2032 doc: /* Read one Lisp expression as text from STREAM, return as Lisp object.
2033 If STREAM is nil, use the value of `standard-input' (which see).
2034 STREAM or the value of `standard-input' may be:
2035 a buffer (read from point and advance it)
2036 a marker (read from where it points and advance it)
2037 a function (call it with no arguments for each character,
2038 call it with a char as argument to push a char back)
2039 a string (takes text from string, starting at the beginning)
2040 t (read text line using minibuffer and use it, or read from
2041 standard input in batch mode). */)
2042 (Lisp_Object stream)
2043 {
2044 if (NILP (stream))
2045 stream = Vstandard_input;
2046 if (EQ (stream, Qt))
2047 stream = Qread_char;
2048 if (EQ (stream, Qread_char))
2049 /* FIXME: ?! When is this used !? */
2050 return call1 (intern ("read-minibuffer"),
2051 build_string ("Lisp expression: "));
2052
2053 return read_internal_start (stream, Qnil, Qnil);
2054 }
2055
2056 DEFUN ("read-from-string", Fread_from_string, Sread_from_string, 1, 3, 0,
2057 doc: /* Read one Lisp expression which is represented as text by STRING.
2058 Returns a cons: (OBJECT-READ . FINAL-STRING-INDEX).
2059 FINAL-STRING-INDEX is an integer giving the position of the next
2060 remaining character in STRING. START and END optionally delimit
2061 a substring of STRING from which to read; they default to 0 and
2062 (length STRING) respectively. Negative values are counted from
2063 the end of STRING. */)
2064 (Lisp_Object string, Lisp_Object start, Lisp_Object end)
2065 {
2066 Lisp_Object ret;
2067 CHECK_STRING (string);
2068 /* `read_internal_start' sets `read_from_string_index'. */
2069 ret = read_internal_start (string, start, end);
2070 return Fcons (ret, make_number (read_from_string_index));
2071 }
2072
2073 /* Function to set up the global context we need in toplevel read
2074 calls. START and END only used when STREAM is a string. */
2075 static Lisp_Object
2076 read_internal_start (Lisp_Object stream, Lisp_Object start, Lisp_Object end)
2077 {
2078 Lisp_Object retval;
2079
2080 readchar_count = 0;
2081 new_backquote_flag = 0;
2082 read_objects = Qnil;
2083 if (EQ (Vread_with_symbol_positions, Qt)
2084 || EQ (Vread_with_symbol_positions, stream))
2085 Vread_symbol_positions_list = Qnil;
2086
2087 if (STRINGP (stream)
2088 || ((CONSP (stream) && STRINGP (XCAR (stream)))))
2089 {
2090 ptrdiff_t startval, endval;
2091 Lisp_Object string;
2092
2093 if (STRINGP (stream))
2094 string = stream;
2095 else
2096 string = XCAR (stream);
2097
2098 validate_subarray (string, start, end, SCHARS (string),
2099 &startval, &endval);
2100
2101 read_from_string_index = startval;
2102 read_from_string_index_byte = string_char_to_byte (string, startval);
2103 read_from_string_limit = endval;
2104 }
2105
2106 retval = read0 (stream);
2107 if (EQ (Vread_with_symbol_positions, Qt)
2108 || EQ (Vread_with_symbol_positions, stream))
2109 Vread_symbol_positions_list = Fnreverse (Vread_symbol_positions_list);
2110 return retval;
2111 }
2112 \f
2113
2114 /* Signal Qinvalid_read_syntax error.
2115 S is error string of length N (if > 0) */
2116
2117 static _Noreturn void
2118 invalid_syntax (const char *s)
2119 {
2120 xsignal1 (Qinvalid_read_syntax, build_string (s));
2121 }
2122
2123
2124 /* Use this for recursive reads, in contexts where internal tokens
2125 are not allowed. */
2126
2127 static Lisp_Object
2128 read0 (Lisp_Object readcharfun)
2129 {
2130 register Lisp_Object val;
2131 int c;
2132
2133 val = read1 (readcharfun, &c, 0);
2134 if (!c)
2135 return val;
2136
2137 xsignal1 (Qinvalid_read_syntax,
2138 Fmake_string (make_number (1), make_number (c)));
2139 }
2140 \f
2141 static ptrdiff_t read_buffer_size;
2142 static char *read_buffer;
2143
2144 /* Grow the read buffer by at least MAX_MULTIBYTE_LENGTH bytes. */
2145
2146 static void
2147 grow_read_buffer (void)
2148 {
2149 read_buffer = xpalloc (read_buffer, &read_buffer_size,
2150 MAX_MULTIBYTE_LENGTH, -1, 1);
2151 }
2152
2153 /* Read a \-escape sequence, assuming we already read the `\'.
2154 If the escape sequence forces unibyte, return eight-bit char. */
2155
2156 static int
2157 read_escape (Lisp_Object readcharfun, bool stringp)
2158 {
2159 int c = READCHAR;
2160 /* \u allows up to four hex digits, \U up to eight. Default to the
2161 behavior for \u, and change this value in the case that \U is seen. */
2162 int unicode_hex_count = 4;
2163
2164 switch (c)
2165 {
2166 case -1:
2167 end_of_file_error ();
2168
2169 case 'a':
2170 return '\007';
2171 case 'b':
2172 return '\b';
2173 case 'd':
2174 return 0177;
2175 case 'e':
2176 return 033;
2177 case 'f':
2178 return '\f';
2179 case 'n':
2180 return '\n';
2181 case 'r':
2182 return '\r';
2183 case 't':
2184 return '\t';
2185 case 'v':
2186 return '\v';
2187 case '\n':
2188 return -1;
2189 case ' ':
2190 if (stringp)
2191 return -1;
2192 return ' ';
2193
2194 case 'M':
2195 c = READCHAR;
2196 if (c != '-')
2197 error ("Invalid escape character syntax");
2198 c = READCHAR;
2199 if (c == '\\')
2200 c = read_escape (readcharfun, 0);
2201 return c | meta_modifier;
2202
2203 case 'S':
2204 c = READCHAR;
2205 if (c != '-')
2206 error ("Invalid escape character syntax");
2207 c = READCHAR;
2208 if (c == '\\')
2209 c = read_escape (readcharfun, 0);
2210 return c | shift_modifier;
2211
2212 case 'H':
2213 c = READCHAR;
2214 if (c != '-')
2215 error ("Invalid escape character syntax");
2216 c = READCHAR;
2217 if (c == '\\')
2218 c = read_escape (readcharfun, 0);
2219 return c | hyper_modifier;
2220
2221 case 'A':
2222 c = READCHAR;
2223 if (c != '-')
2224 error ("Invalid escape character syntax");
2225 c = READCHAR;
2226 if (c == '\\')
2227 c = read_escape (readcharfun, 0);
2228 return c | alt_modifier;
2229
2230 case 's':
2231 c = READCHAR;
2232 if (stringp || c != '-')
2233 {
2234 UNREAD (c);
2235 return ' ';
2236 }
2237 c = READCHAR;
2238 if (c == '\\')
2239 c = read_escape (readcharfun, 0);
2240 return c | super_modifier;
2241
2242 case 'C':
2243 c = READCHAR;
2244 if (c != '-')
2245 error ("Invalid escape character syntax");
2246 case '^':
2247 c = READCHAR;
2248 if (c == '\\')
2249 c = read_escape (readcharfun, 0);
2250 if ((c & ~CHAR_MODIFIER_MASK) == '?')
2251 return 0177 | (c & CHAR_MODIFIER_MASK);
2252 else if (! SINGLE_BYTE_CHAR_P ((c & ~CHAR_MODIFIER_MASK)))
2253 return c | ctrl_modifier;
2254 /* ASCII control chars are made from letters (both cases),
2255 as well as the non-letters within 0100...0137. */
2256 else if ((c & 0137) >= 0101 && (c & 0137) <= 0132)
2257 return (c & (037 | ~0177));
2258 else if ((c & 0177) >= 0100 && (c & 0177) <= 0137)
2259 return (c & (037 | ~0177));
2260 else
2261 return c | ctrl_modifier;
2262
2263 case '0':
2264 case '1':
2265 case '2':
2266 case '3':
2267 case '4':
2268 case '5':
2269 case '6':
2270 case '7':
2271 /* An octal escape, as in ANSI C. */
2272 {
2273 register int i = c - '0';
2274 register int count = 0;
2275 while (++count < 3)
2276 {
2277 if ((c = READCHAR) >= '0' && c <= '7')
2278 {
2279 i *= 8;
2280 i += c - '0';
2281 }
2282 else
2283 {
2284 UNREAD (c);
2285 break;
2286 }
2287 }
2288
2289 if (i >= 0x80 && i < 0x100)
2290 i = BYTE8_TO_CHAR (i);
2291 return i;
2292 }
2293
2294 case 'x':
2295 /* A hex escape, as in ANSI C. */
2296 {
2297 unsigned int i = 0;
2298 int count = 0;
2299 while (1)
2300 {
2301 c = READCHAR;
2302 if (c >= '0' && c <= '9')
2303 {
2304 i *= 16;
2305 i += c - '0';
2306 }
2307 else if ((c >= 'a' && c <= 'f')
2308 || (c >= 'A' && c <= 'F'))
2309 {
2310 i *= 16;
2311 if (c >= 'a' && c <= 'f')
2312 i += c - 'a' + 10;
2313 else
2314 i += c - 'A' + 10;
2315 }
2316 else
2317 {
2318 UNREAD (c);
2319 break;
2320 }
2321 /* Allow hex escapes as large as ?\xfffffff, because some
2322 packages use them to denote characters with modifiers. */
2323 if ((CHAR_META | (CHAR_META - 1)) < i)
2324 error ("Hex character out of range: \\x%x...", i);
2325 count += count < 3;
2326 }
2327
2328 if (count < 3 && i >= 0x80)
2329 return BYTE8_TO_CHAR (i);
2330 return i;
2331 }
2332
2333 case 'U':
2334 /* Post-Unicode-2.0: Up to eight hex chars. */
2335 unicode_hex_count = 8;
2336 case 'u':
2337
2338 /* A Unicode escape. We only permit them in strings and characters,
2339 not arbitrarily in the source code, as in some other languages. */
2340 {
2341 unsigned int i = 0;
2342 int count = 0;
2343
2344 while (++count <= unicode_hex_count)
2345 {
2346 c = READCHAR;
2347 /* `isdigit' and `isalpha' may be locale-specific, which we don't
2348 want. */
2349 if (c >= '0' && c <= '9') i = (i << 4) + (c - '0');
2350 else if (c >= 'a' && c <= 'f') i = (i << 4) + (c - 'a') + 10;
2351 else if (c >= 'A' && c <= 'F') i = (i << 4) + (c - 'A') + 10;
2352 else
2353 error ("Non-hex digit used for Unicode escape");
2354 }
2355 if (i > 0x10FFFF)
2356 error ("Non-Unicode character: 0x%x", i);
2357 return i;
2358 }
2359
2360 default:
2361 return c;
2362 }
2363 }
2364
2365 /* Return the digit that CHARACTER stands for in the given BASE.
2366 Return -1 if CHARACTER is out of range for BASE,
2367 and -2 if CHARACTER is not valid for any supported BASE. */
2368 static int
2369 digit_to_number (int character, int base)
2370 {
2371 int digit;
2372
2373 if ('0' <= character && character <= '9')
2374 digit = character - '0';
2375 else if ('a' <= character && character <= 'z')
2376 digit = character - 'a' + 10;
2377 else if ('A' <= character && character <= 'Z')
2378 digit = character - 'A' + 10;
2379 else
2380 return -2;
2381
2382 return digit < base ? digit : -1;
2383 }
2384
2385 /* Read an integer in radix RADIX using READCHARFUN to read
2386 characters. RADIX must be in the interval [2..36]; if it isn't, a
2387 read error is signaled . Value is the integer read. Signals an
2388 error if encountering invalid read syntax or if RADIX is out of
2389 range. */
2390
2391 static Lisp_Object
2392 read_integer (Lisp_Object readcharfun, EMACS_INT radix)
2393 {
2394 /* Room for sign, leading 0, other digits, trailing null byte.
2395 Also, room for invalid syntax diagnostic. */
2396 char buf[max (1 + 1 + sizeof (uintmax_t) * CHAR_BIT + 1,
2397 sizeof "integer, radix " + INT_STRLEN_BOUND (EMACS_INT))];
2398
2399 int valid = -1; /* 1 if valid, 0 if not, -1 if incomplete. */
2400
2401 if (radix < 2 || radix > 36)
2402 valid = 0;
2403 else
2404 {
2405 char *p = buf;
2406 int c, digit;
2407
2408 c = READCHAR;
2409 if (c == '-' || c == '+')
2410 {
2411 *p++ = c;
2412 c = READCHAR;
2413 }
2414
2415 if (c == '0')
2416 {
2417 *p++ = c;
2418 valid = 1;
2419
2420 /* Ignore redundant leading zeros, so the buffer doesn't
2421 fill up with them. */
2422 do
2423 c = READCHAR;
2424 while (c == '0');
2425 }
2426
2427 while ((digit = digit_to_number (c, radix)) >= -1)
2428 {
2429 if (digit == -1)
2430 valid = 0;
2431 if (valid < 0)
2432 valid = 1;
2433
2434 if (p < buf + sizeof buf - 1)
2435 *p++ = c;
2436 else
2437 valid = 0;
2438
2439 c = READCHAR;
2440 }
2441
2442 UNREAD (c);
2443 *p = '\0';
2444 }
2445
2446 if (! valid)
2447 {
2448 sprintf (buf, "integer, radix %"pI"d", radix);
2449 invalid_syntax (buf);
2450 }
2451
2452 return string_to_number (buf, radix, 0);
2453 }
2454
2455
2456 /* If the next token is ')' or ']' or '.', we store that character
2457 in *PCH and the return value is not interesting. Else, we store
2458 zero in *PCH and we read and return one lisp object.
2459
2460 FIRST_IN_LIST is true if this is the first element of a list. */
2461
2462 static Lisp_Object
2463 read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
2464 {
2465 int c;
2466 bool uninterned_symbol = 0;
2467 bool multibyte;
2468
2469 *pch = 0;
2470
2471 retry:
2472
2473 c = READCHAR_REPORT_MULTIBYTE (&multibyte);
2474 if (c < 0)
2475 end_of_file_error ();
2476
2477 switch (c)
2478 {
2479 case '(':
2480 return read_list (0, readcharfun);
2481
2482 case '[':
2483 return read_vector (readcharfun, 0);
2484
2485 case ')':
2486 case ']':
2487 {
2488 *pch = c;
2489 return Qnil;
2490 }
2491
2492 case '#':
2493 c = READCHAR;
2494 if (c == 's')
2495 {
2496 c = READCHAR;
2497 if (c == '(')
2498 {
2499 /* Accept extended format for hashtables (extensible to
2500 other types), e.g.
2501 #s(hash-table size 2 test equal data (k1 v1 k2 v2)) */
2502 Lisp_Object tmp = read_list (0, readcharfun);
2503 Lisp_Object head = CAR_SAFE (tmp);
2504 Lisp_Object data = Qnil;
2505 Lisp_Object val = Qnil;
2506 /* The size is 2 * number of allowed keywords to
2507 make-hash-table. */
2508 Lisp_Object params[10];
2509 Lisp_Object ht;
2510 Lisp_Object key = Qnil;
2511 int param_count = 0;
2512
2513 if (!EQ (head, Qhash_table))
2514 error ("Invalid extended read marker at head of #s list "
2515 "(only hash-table allowed)");
2516
2517 tmp = CDR_SAFE (tmp);
2518
2519 /* This is repetitive but fast and simple. */
2520 params[param_count] = QCsize;
2521 params[param_count + 1] = Fplist_get (tmp, Qsize);
2522 if (!NILP (params[param_count + 1]))
2523 param_count += 2;
2524
2525 params[param_count] = QCtest;
2526 params[param_count + 1] = Fplist_get (tmp, Qtest);
2527 if (!NILP (params[param_count + 1]))
2528 param_count += 2;
2529
2530 params[param_count] = QCweakness;
2531 params[param_count + 1] = Fplist_get (tmp, Qweakness);
2532 if (!NILP (params[param_count + 1]))
2533 param_count += 2;
2534
2535 params[param_count] = QCrehash_size;
2536 params[param_count + 1] = Fplist_get (tmp, Qrehash_size);
2537 if (!NILP (params[param_count + 1]))
2538 param_count += 2;
2539
2540 params[param_count] = QCrehash_threshold;
2541 params[param_count + 1] = Fplist_get (tmp, Qrehash_threshold);
2542 if (!NILP (params[param_count + 1]))
2543 param_count += 2;
2544
2545 /* This is the hashtable data. */
2546 data = Fplist_get (tmp, Qdata);
2547
2548 /* Now use params to make a new hashtable and fill it. */
2549 ht = Fmake_hash_table (param_count, params);
2550
2551 while (CONSP (data))
2552 {
2553 key = XCAR (data);
2554 data = XCDR (data);
2555 if (!CONSP (data))
2556 error ("Odd number of elements in hashtable data");
2557 val = XCAR (data);
2558 data = XCDR (data);
2559 Fputhash (key, val, ht);
2560 }
2561
2562 return ht;
2563 }
2564 UNREAD (c);
2565 invalid_syntax ("#");
2566 }
2567 if (c == '^')
2568 {
2569 c = READCHAR;
2570 if (c == '[')
2571 {
2572 Lisp_Object tmp;
2573 tmp = read_vector (readcharfun, 0);
2574 if (ASIZE (tmp) < CHAR_TABLE_STANDARD_SLOTS)
2575 error ("Invalid size char-table");
2576 XSETPVECTYPE (XVECTOR (tmp), PVEC_CHAR_TABLE);
2577 return tmp;
2578 }
2579 else if (c == '^')
2580 {
2581 c = READCHAR;
2582 if (c == '[')
2583 {
2584 /* Sub char-table can't be read as a regular
2585 vector because of a two C integer fields. */
2586 Lisp_Object tbl, tmp = read_list (1, readcharfun);
2587 ptrdiff_t size = XINT (Flength (tmp));
2588 int i, depth, min_char;
2589 struct Lisp_Cons *cell;
2590
2591 if (size == 0)
2592 error ("Zero-sized sub char-table");
2593
2594 if (! RANGED_INTEGERP (1, XCAR (tmp), 3))
2595 error ("Invalid depth in sub char-table");
2596 depth = XINT (XCAR (tmp));
2597 if (chartab_size[depth] != size - 2)
2598 error ("Invalid size in sub char-table");
2599 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2600 free_cons (cell);
2601
2602 if (! RANGED_INTEGERP (0, XCAR (tmp), MAX_CHAR))
2603 error ("Invalid minimum character in sub-char-table");
2604 min_char = XINT (XCAR (tmp));
2605 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2606 free_cons (cell);
2607
2608 tbl = make_uninit_sub_char_table (depth, min_char);
2609 for (i = 0; i < size; i++)
2610 {
2611 XSUB_CHAR_TABLE (tbl)->contents[i] = XCAR (tmp);
2612 cell = XCONS (tmp), tmp = XCDR (tmp);
2613 free_cons (cell);
2614 }
2615 return tbl;
2616 }
2617 invalid_syntax ("#^^");
2618 }
2619 invalid_syntax ("#^");
2620 }
2621 if (c == '&')
2622 {
2623 Lisp_Object length;
2624 length = read1 (readcharfun, pch, first_in_list);
2625 c = READCHAR;
2626 if (c == '"')
2627 {
2628 Lisp_Object tmp, val;
2629 EMACS_INT size_in_chars = bool_vector_bytes (XFASTINT (length));
2630 unsigned char *data;
2631
2632 UNREAD (c);
2633 tmp = read1 (readcharfun, pch, first_in_list);
2634 if (STRING_MULTIBYTE (tmp)
2635 || (size_in_chars != SCHARS (tmp)
2636 /* We used to print 1 char too many
2637 when the number of bits was a multiple of 8.
2638 Accept such input in case it came from an old
2639 version. */
2640 && ! (XFASTINT (length)
2641 == (SCHARS (tmp) - 1) * BOOL_VECTOR_BITS_PER_CHAR)))
2642 invalid_syntax ("#&...");
2643
2644 val = make_uninit_bool_vector (XFASTINT (length));
2645 data = bool_vector_uchar_data (val);
2646 memcpy (data, SDATA (tmp), size_in_chars);
2647 /* Clear the extraneous bits in the last byte. */
2648 if (XINT (length) != size_in_chars * BOOL_VECTOR_BITS_PER_CHAR)
2649 data[size_in_chars - 1]
2650 &= (1 << (XINT (length) % BOOL_VECTOR_BITS_PER_CHAR)) - 1;
2651 return val;
2652 }
2653 invalid_syntax ("#&...");
2654 }
2655 if (c == '[')
2656 {
2657 /* Accept compiled functions at read-time so that we don't have to
2658 build them using function calls. */
2659 Lisp_Object tmp;
2660 struct Lisp_Vector *vec;
2661 tmp = read_vector (readcharfun, 1);
2662 vec = XVECTOR (tmp);
2663 if (vec->header.size == 0)
2664 invalid_syntax ("Empty byte-code object");
2665 make_byte_code (vec);
2666 return tmp;
2667 }
2668 if (c == '(')
2669 {
2670 Lisp_Object tmp;
2671 int ch;
2672
2673 /* Read the string itself. */
2674 tmp = read1 (readcharfun, &ch, 0);
2675 if (ch != 0 || !STRINGP (tmp))
2676 invalid_syntax ("#");
2677 /* Read the intervals and their properties. */
2678 while (1)
2679 {
2680 Lisp_Object beg, end, plist;
2681
2682 beg = read1 (readcharfun, &ch, 0);
2683 end = plist = Qnil;
2684 if (ch == ')')
2685 break;
2686 if (ch == 0)
2687 end = read1 (readcharfun, &ch, 0);
2688 if (ch == 0)
2689 plist = read1 (readcharfun, &ch, 0);
2690 if (ch)
2691 invalid_syntax ("Invalid string property list");
2692 Fset_text_properties (beg, end, plist, tmp);
2693 }
2694
2695 return tmp;
2696 }
2697
2698 /* #@NUMBER is used to skip NUMBER following bytes.
2699 That's used in .elc files to skip over doc strings
2700 and function definitions. */
2701 if (c == '@')
2702 {
2703 enum { extra = 100 };
2704 ptrdiff_t i, nskip = 0, digits = 0;
2705
2706 /* Read a decimal integer. */
2707 while ((c = READCHAR) >= 0
2708 && c >= '0' && c <= '9')
2709 {
2710 if ((STRING_BYTES_BOUND - extra) / 10 <= nskip)
2711 string_overflow ();
2712 digits++;
2713 nskip *= 10;
2714 nskip += c - '0';
2715 if (digits == 2 && nskip == 0)
2716 { /* We've just seen #@00, which means "skip to end". */
2717 skip_dyn_eof (readcharfun);
2718 return Qnil;
2719 }
2720 }
2721 if (nskip > 0)
2722 /* We can't use UNREAD here, because in the code below we side-step
2723 READCHAR. Instead, assume the first char after #@NNN occupies
2724 a single byte, which is the case normally since it's just
2725 a space. */
2726 nskip--;
2727 else
2728 UNREAD (c);
2729
2730 if (load_force_doc_strings
2731 && (FROM_FILE_P (readcharfun)))
2732 {
2733 /* If we are supposed to force doc strings into core right now,
2734 record the last string that we skipped,
2735 and record where in the file it comes from. */
2736
2737 /* But first exchange saved_doc_string
2738 with prev_saved_doc_string, so we save two strings. */
2739 {
2740 char *temp = saved_doc_string;
2741 ptrdiff_t temp_size = saved_doc_string_size;
2742 file_offset temp_pos = saved_doc_string_position;
2743 ptrdiff_t temp_len = saved_doc_string_length;
2744
2745 saved_doc_string = prev_saved_doc_string;
2746 saved_doc_string_size = prev_saved_doc_string_size;
2747 saved_doc_string_position = prev_saved_doc_string_position;
2748 saved_doc_string_length = prev_saved_doc_string_length;
2749
2750 prev_saved_doc_string = temp;
2751 prev_saved_doc_string_size = temp_size;
2752 prev_saved_doc_string_position = temp_pos;
2753 prev_saved_doc_string_length = temp_len;
2754 }
2755
2756 if (saved_doc_string_size == 0)
2757 {
2758 saved_doc_string = xmalloc (nskip + extra);
2759 saved_doc_string_size = nskip + extra;
2760 }
2761 if (nskip > saved_doc_string_size)
2762 {
2763 saved_doc_string = xrealloc (saved_doc_string, nskip + extra);
2764 saved_doc_string_size = nskip + extra;
2765 }
2766
2767 saved_doc_string_position = file_tell (instream);
2768
2769 /* Copy that many characters into saved_doc_string. */
2770 block_input ();
2771 for (i = 0; i < nskip && c >= 0; i++)
2772 saved_doc_string[i] = c = getc (instream);
2773 unblock_input ();
2774
2775 saved_doc_string_length = i;
2776 }
2777 else
2778 /* Skip that many bytes. */
2779 skip_dyn_bytes (readcharfun, nskip);
2780
2781 goto retry;
2782 }
2783 if (c == '!')
2784 {
2785 /* #! appears at the beginning of an executable file.
2786 Skip the first line. */
2787 while (c != '\n' && c >= 0)
2788 c = READCHAR;
2789 goto retry;
2790 }
2791 if (c == '$')
2792 return Vload_file_name;
2793 if (c == '\'')
2794 return list2 (Qfunction, read0 (readcharfun));
2795 /* #:foo is the uninterned symbol named foo. */
2796 if (c == ':')
2797 {
2798 uninterned_symbol = 1;
2799 c = READCHAR;
2800 if (!(c > 040
2801 && c != NO_BREAK_SPACE
2802 && (c >= 0200
2803 || strchr ("\"';()[]#`,", c) == NULL)))
2804 {
2805 /* No symbol character follows, this is the empty
2806 symbol. */
2807 UNREAD (c);
2808 return Fmake_symbol (empty_unibyte_string);
2809 }
2810 goto read_symbol;
2811 }
2812 /* ## is the empty symbol. */
2813 if (c == '#')
2814 return Fintern (empty_unibyte_string, Qnil);
2815 /* Reader forms that can reuse previously read objects. */
2816 if (c >= '0' && c <= '9')
2817 {
2818 EMACS_INT n = 0;
2819 Lisp_Object tem;
2820
2821 /* Read a non-negative integer. */
2822 while (c >= '0' && c <= '9')
2823 {
2824 if (MOST_POSITIVE_FIXNUM / 10 < n
2825 || MOST_POSITIVE_FIXNUM < n * 10 + c - '0')
2826 n = MOST_POSITIVE_FIXNUM + 1;
2827 else
2828 n = n * 10 + c - '0';
2829 c = READCHAR;
2830 }
2831
2832 if (n <= MOST_POSITIVE_FIXNUM)
2833 {
2834 if (c == 'r' || c == 'R')
2835 return read_integer (readcharfun, n);
2836
2837 if (! NILP (Vread_circle))
2838 {
2839 /* #n=object returns object, but associates it with
2840 n for #n#. */
2841 if (c == '=')
2842 {
2843 /* Make a placeholder for #n# to use temporarily. */
2844 AUTO_CONS (placeholder, Qnil, Qnil);
2845 Lisp_Object cell = Fcons (make_number (n), placeholder);
2846 read_objects = Fcons (cell, read_objects);
2847
2848 /* Read the object itself. */
2849 tem = read0 (readcharfun);
2850
2851 /* Now put it everywhere the placeholder was... */
2852 substitute_object_in_subtree (tem, placeholder);
2853
2854 /* ...and #n# will use the real value from now on. */
2855 Fsetcdr (cell, tem);
2856
2857 return tem;
2858 }
2859
2860 /* #n# returns a previously read object. */
2861 if (c == '#')
2862 {
2863 tem = Fassq (make_number (n), read_objects);
2864 if (CONSP (tem))
2865 return XCDR (tem);
2866 }
2867 }
2868 }
2869 /* Fall through to error message. */
2870 }
2871 else if (c == 'x' || c == 'X')
2872 return read_integer (readcharfun, 16);
2873 else if (c == 'o' || c == 'O')
2874 return read_integer (readcharfun, 8);
2875 else if (c == 'b' || c == 'B')
2876 return read_integer (readcharfun, 2);
2877
2878 UNREAD (c);
2879 invalid_syntax ("#");
2880
2881 case ';':
2882 while ((c = READCHAR) >= 0 && c != '\n');
2883 goto retry;
2884
2885 case '\'':
2886 return list2 (Qquote, read0 (readcharfun));
2887
2888 case '`':
2889 {
2890 int next_char = READCHAR;
2891 UNREAD (next_char);
2892 /* Transition from old-style to new-style:
2893 If we see "(`" it used to mean old-style, which usually works
2894 fine because ` should almost never appear in such a position
2895 for new-style. But occasionally we need "(`" to mean new
2896 style, so we try to distinguish the two by the fact that we
2897 can either write "( `foo" or "(` foo", where the first
2898 intends to use new-style whereas the second intends to use
2899 old-style. For Emacs-25, we should completely remove this
2900 first_in_list exception (old-style can still be obtained via
2901 "(\`" anyway). */
2902 if (!new_backquote_flag && first_in_list && next_char == ' ')
2903 {
2904 Vold_style_backquotes = Qt;
2905 goto default_label;
2906 }
2907 else
2908 {
2909 Lisp_Object value;
2910 bool saved_new_backquote_flag = new_backquote_flag;
2911
2912 new_backquote_flag = 1;
2913 value = read0 (readcharfun);
2914 new_backquote_flag = saved_new_backquote_flag;
2915
2916 return list2 (Qbackquote, value);
2917 }
2918 }
2919 case ',':
2920 {
2921 int next_char = READCHAR;
2922 UNREAD (next_char);
2923 /* Transition from old-style to new-style:
2924 It used to be impossible to have a new-style , other than within
2925 a new-style `. This is sufficient when ` and , are used in the
2926 normal way, but ` and , can also appear in args to macros that
2927 will not interpret them in the usual way, in which case , may be
2928 used without any ` anywhere near.
2929 So we now use the same heuristic as for backquote: old-style
2930 unquotes are only recognized when first on a list, and when
2931 followed by a space.
2932 Because it's more difficult to peek 2 chars ahead, a new-style
2933 ,@ can still not be used outside of a `, unless it's in the middle
2934 of a list. */
2935 if (new_backquote_flag
2936 || !first_in_list
2937 || (next_char != ' ' && next_char != '@'))
2938 {
2939 Lisp_Object comma_type = Qnil;
2940 Lisp_Object value;
2941 int ch = READCHAR;
2942
2943 if (ch == '@')
2944 comma_type = Qcomma_at;
2945 else if (ch == '.')
2946 comma_type = Qcomma_dot;
2947 else
2948 {
2949 if (ch >= 0) UNREAD (ch);
2950 comma_type = Qcomma;
2951 }
2952
2953 value = read0 (readcharfun);
2954 return list2 (comma_type, value);
2955 }
2956 else
2957 {
2958 Vold_style_backquotes = Qt;
2959 goto default_label;
2960 }
2961 }
2962 case '?':
2963 {
2964 int modifiers;
2965 int next_char;
2966 bool ok;
2967
2968 c = READCHAR;
2969 if (c < 0)
2970 end_of_file_error ();
2971
2972 /* Accept `single space' syntax like (list ? x) where the
2973 whitespace character is SPC or TAB.
2974 Other literal whitespace like NL, CR, and FF are not accepted,
2975 as there are well-established escape sequences for these. */
2976 if (c == ' ' || c == '\t')
2977 return make_number (c);
2978
2979 if (c == '\\')
2980 c = read_escape (readcharfun, 0);
2981 modifiers = c & CHAR_MODIFIER_MASK;
2982 c &= ~CHAR_MODIFIER_MASK;
2983 if (CHAR_BYTE8_P (c))
2984 c = CHAR_TO_BYTE8 (c);
2985 c |= modifiers;
2986
2987 next_char = READCHAR;
2988 ok = (next_char <= 040
2989 || (next_char < 0200
2990 && strchr ("\"';()[]#?`,.", next_char) != NULL));
2991 UNREAD (next_char);
2992 if (ok)
2993 return make_number (c);
2994
2995 invalid_syntax ("?");
2996 }
2997
2998 case '"':
2999 {
3000 char *p = read_buffer;
3001 char *end = read_buffer + read_buffer_size;
3002 int ch;
3003 /* True if we saw an escape sequence specifying
3004 a multibyte character. */
3005 bool force_multibyte = 0;
3006 /* True if we saw an escape sequence specifying
3007 a single-byte character. */
3008 bool force_singlebyte = 0;
3009 bool cancel = 0;
3010 ptrdiff_t nchars = 0;
3011
3012 while ((ch = READCHAR) >= 0
3013 && ch != '\"')
3014 {
3015 if (end - p < MAX_MULTIBYTE_LENGTH)
3016 {
3017 ptrdiff_t offset = p - read_buffer;
3018 grow_read_buffer ();
3019 p = read_buffer + offset;
3020 end = read_buffer + read_buffer_size;
3021 }
3022
3023 if (ch == '\\')
3024 {
3025 int modifiers;
3026
3027 ch = read_escape (readcharfun, 1);
3028
3029 /* CH is -1 if \ newline or \ space has just been seen. */
3030 if (ch == -1)
3031 {
3032 if (p == read_buffer)
3033 cancel = 1;
3034 continue;
3035 }
3036
3037 modifiers = ch & CHAR_MODIFIER_MASK;
3038 ch = ch & ~CHAR_MODIFIER_MASK;
3039
3040 if (CHAR_BYTE8_P (ch))
3041 force_singlebyte = 1;
3042 else if (! ASCII_CHAR_P (ch))
3043 force_multibyte = 1;
3044 else /* I.e. ASCII_CHAR_P (ch). */
3045 {
3046 /* Allow `\C- ' and `\C-?'. */
3047 if (modifiers == CHAR_CTL)
3048 {
3049 if (ch == ' ')
3050 ch = 0, modifiers = 0;
3051 else if (ch == '?')
3052 ch = 127, modifiers = 0;
3053 }
3054 if (modifiers & CHAR_SHIFT)
3055 {
3056 /* Shift modifier is valid only with [A-Za-z]. */
3057 if (ch >= 'A' && ch <= 'Z')
3058 modifiers &= ~CHAR_SHIFT;
3059 else if (ch >= 'a' && ch <= 'z')
3060 ch -= ('a' - 'A'), modifiers &= ~CHAR_SHIFT;
3061 }
3062
3063 if (modifiers & CHAR_META)
3064 {
3065 /* Move the meta bit to the right place for a
3066 string. */
3067 modifiers &= ~CHAR_META;
3068 ch = BYTE8_TO_CHAR (ch | 0x80);
3069 force_singlebyte = 1;
3070 }
3071 }
3072
3073 /* Any modifiers remaining are invalid. */
3074 if (modifiers)
3075 error ("Invalid modifier in string");
3076 p += CHAR_STRING (ch, (unsigned char *) p);
3077 }
3078 else
3079 {
3080 p += CHAR_STRING (ch, (unsigned char *) p);
3081 if (CHAR_BYTE8_P (ch))
3082 force_singlebyte = 1;
3083 else if (! ASCII_CHAR_P (ch))
3084 force_multibyte = 1;
3085 }
3086 nchars++;
3087 }
3088
3089 if (ch < 0)
3090 end_of_file_error ();
3091
3092 /* If purifying, and string starts with \ newline,
3093 return zero instead. This is for doc strings
3094 that we are really going to find in etc/DOC.nn.nn. */
3095 if (!NILP (Vpurify_flag) && NILP (Vdoc_file_name) && cancel)
3096 return make_number (0);
3097
3098 if (! force_multibyte && force_singlebyte)
3099 {
3100 /* READ_BUFFER contains raw 8-bit bytes and no multibyte
3101 forms. Convert it to unibyte. */
3102 nchars = str_as_unibyte ((unsigned char *) read_buffer,
3103 p - read_buffer);
3104 p = read_buffer + nchars;
3105 }
3106
3107 return make_specified_string (read_buffer, nchars, p - read_buffer,
3108 (force_multibyte
3109 || (p - read_buffer != nchars)));
3110 }
3111
3112 case '.':
3113 {
3114 int next_char = READCHAR;
3115 UNREAD (next_char);
3116
3117 if (next_char <= 040
3118 || (next_char < 0200
3119 && strchr ("\"';([#?`,", next_char) != NULL))
3120 {
3121 *pch = c;
3122 return Qnil;
3123 }
3124
3125 /* Otherwise, we fall through! Note that the atom-reading loop
3126 below will now loop at least once, assuring that we will not
3127 try to UNREAD two characters in a row. */
3128 }
3129 default:
3130 default_label:
3131 if (c <= 040) goto retry;
3132 if (c == NO_BREAK_SPACE)
3133 goto retry;
3134
3135 read_symbol:
3136 {
3137 char *p = read_buffer;
3138 bool quoted = 0;
3139 EMACS_INT start_position = readchar_count - 1;
3140
3141 {
3142 char *end = read_buffer + read_buffer_size;
3143
3144 do
3145 {
3146 if (end - p < MAX_MULTIBYTE_LENGTH)
3147 {
3148 ptrdiff_t offset = p - read_buffer;
3149 grow_read_buffer ();
3150 p = read_buffer + offset;
3151 end = read_buffer + read_buffer_size;
3152 }
3153
3154 if (c == '\\')
3155 {
3156 c = READCHAR;
3157 if (c == -1)
3158 end_of_file_error ();
3159 quoted = 1;
3160 }
3161
3162 if (multibyte)
3163 p += CHAR_STRING (c, (unsigned char *) p);
3164 else
3165 *p++ = c;
3166 c = READCHAR;
3167 }
3168 while (c > 040
3169 && c != NO_BREAK_SPACE
3170 && (c >= 0200
3171 || strchr ("\"';()[]#`,", c) == NULL));
3172
3173 if (p == end)
3174 {
3175 ptrdiff_t offset = p - read_buffer;
3176 grow_read_buffer ();
3177 p = read_buffer + offset;
3178 end = read_buffer + read_buffer_size;
3179 }
3180 *p = 0;
3181 UNREAD (c);
3182 }
3183
3184 if (!quoted && !uninterned_symbol)
3185 {
3186 Lisp_Object result = string_to_number (read_buffer, 10, 0);
3187 if (! NILP (result))
3188 return result;
3189 }
3190 {
3191 Lisp_Object name, result;
3192 ptrdiff_t nbytes = p - read_buffer;
3193 ptrdiff_t nchars
3194 = (multibyte
3195 ? multibyte_chars_in_text ((unsigned char *) read_buffer,
3196 nbytes)
3197 : nbytes);
3198
3199 name = ((uninterned_symbol && ! NILP (Vpurify_flag)
3200 ? make_pure_string : make_specified_string)
3201 (read_buffer, nchars, nbytes, multibyte));
3202 result = (uninterned_symbol ? Fmake_symbol (name)
3203 : Fintern (name, Qnil));
3204
3205 if (EQ (Vread_with_symbol_positions, Qt)
3206 || EQ (Vread_with_symbol_positions, readcharfun))
3207 Vread_symbol_positions_list
3208 = Fcons (Fcons (result, make_number (start_position)),
3209 Vread_symbol_positions_list);
3210 return result;
3211 }
3212 }
3213 }
3214 }
3215 \f
3216
3217 /* List of nodes we've seen during substitute_object_in_subtree. */
3218 static Lisp_Object seen_list;
3219
3220 static void
3221 substitute_object_in_subtree (Lisp_Object object, Lisp_Object placeholder)
3222 {
3223 Lisp_Object check_object;
3224
3225 /* We haven't seen any objects when we start. */
3226 seen_list = Qnil;
3227
3228 /* Make all the substitutions. */
3229 check_object
3230 = substitute_object_recurse (object, placeholder, object);
3231
3232 /* Clear seen_list because we're done with it. */
3233 seen_list = Qnil;
3234
3235 /* The returned object here is expected to always eq the
3236 original. */
3237 if (!EQ (check_object, object))
3238 error ("Unexpected mutation error in reader");
3239 }
3240
3241 /* Feval doesn't get called from here, so no gc protection is needed. */
3242 #define SUBSTITUTE(get_val, set_val) \
3243 do { \
3244 Lisp_Object old_value = get_val; \
3245 Lisp_Object true_value \
3246 = substitute_object_recurse (object, placeholder, \
3247 old_value); \
3248 \
3249 if (!EQ (old_value, true_value)) \
3250 { \
3251 set_val; \
3252 } \
3253 } while (0)
3254
3255 static Lisp_Object
3256 substitute_object_recurse (Lisp_Object object, Lisp_Object placeholder, Lisp_Object subtree)
3257 {
3258 /* If we find the placeholder, return the target object. */
3259 if (EQ (placeholder, subtree))
3260 return object;
3261
3262 /* If we've been to this node before, don't explore it again. */
3263 if (!EQ (Qnil, Fmemq (subtree, seen_list)))
3264 return subtree;
3265
3266 /* If this node can be the entry point to a cycle, remember that
3267 we've seen it. It can only be such an entry point if it was made
3268 by #n=, which means that we can find it as a value in
3269 read_objects. */
3270 if (!EQ (Qnil, Frassq (subtree, read_objects)))
3271 seen_list = Fcons (subtree, seen_list);
3272
3273 /* Recurse according to subtree's type.
3274 Every branch must return a Lisp_Object. */
3275 switch (XTYPE (subtree))
3276 {
3277 case Lisp_Vectorlike:
3278 {
3279 ptrdiff_t i = 0, length = 0;
3280 if (BOOL_VECTOR_P (subtree))
3281 return subtree; /* No sub-objects anyway. */
3282 else if (CHAR_TABLE_P (subtree) || SUB_CHAR_TABLE_P (subtree)
3283 || COMPILEDP (subtree) || HASH_TABLE_P (subtree))
3284 length = ASIZE (subtree) & PSEUDOVECTOR_SIZE_MASK;
3285 else if (VECTORP (subtree))
3286 length = ASIZE (subtree);
3287 else
3288 /* An unknown pseudovector may contain non-Lisp fields, so we
3289 can't just blindly traverse all its fields. We used to call
3290 `Flength' which signaled `sequencep', so I just preserved this
3291 behavior. */
3292 wrong_type_argument (Qsequencep, subtree);
3293
3294 if (SUB_CHAR_TABLE_P (subtree))
3295 i = 2;
3296 for ( ; i < length; i++)
3297 SUBSTITUTE (AREF (subtree, i),
3298 ASET (subtree, i, true_value));
3299 return subtree;
3300 }
3301
3302 case Lisp_Cons:
3303 {
3304 SUBSTITUTE (XCAR (subtree),
3305 XSETCAR (subtree, true_value));
3306 SUBSTITUTE (XCDR (subtree),
3307 XSETCDR (subtree, true_value));
3308 return subtree;
3309 }
3310
3311 case Lisp_String:
3312 {
3313 /* Check for text properties in each interval.
3314 substitute_in_interval contains part of the logic. */
3315
3316 INTERVAL root_interval = string_intervals (subtree);
3317 AUTO_CONS (arg, object, placeholder);
3318
3319 traverse_intervals_noorder (root_interval,
3320 &substitute_in_interval, arg);
3321
3322 return subtree;
3323 }
3324
3325 /* Other types don't recurse any further. */
3326 default:
3327 return subtree;
3328 }
3329 }
3330
3331 /* Helper function for substitute_object_recurse. */
3332 static void
3333 substitute_in_interval (INTERVAL interval, Lisp_Object arg)
3334 {
3335 Lisp_Object object = Fcar (arg);
3336 Lisp_Object placeholder = Fcdr (arg);
3337
3338 SUBSTITUTE (interval->plist, set_interval_plist (interval, true_value));
3339 }
3340
3341 \f
3342 #define LEAD_INT 1
3343 #define DOT_CHAR 2
3344 #define TRAIL_INT 4
3345 #define E_EXP 16
3346
3347
3348 /* Convert STRING to a number, assuming base BASE. Return a fixnum if CP has
3349 integer syntax and fits in a fixnum, else return the nearest float if CP has
3350 either floating point or integer syntax and BASE is 10, else return nil. If
3351 IGNORE_TRAILING, consider just the longest prefix of CP that has
3352 valid floating point syntax. Signal an overflow if BASE is not 10 and the
3353 number has integer syntax but does not fit. */
3354
3355 Lisp_Object
3356 string_to_number (char const *string, int base, bool ignore_trailing)
3357 {
3358 int state;
3359 char const *cp = string;
3360 int leading_digit;
3361 bool float_syntax = 0;
3362 double value = 0;
3363
3364 /* Negate the value ourselves. This treats 0, NaNs, and infinity properly on
3365 IEEE floating point hosts, and works around a formerly-common bug where
3366 atof ("-0.0") drops the sign. */
3367 bool negative = *cp == '-';
3368
3369 bool signedp = negative || *cp == '+';
3370 cp += signedp;
3371
3372 state = 0;
3373
3374 leading_digit = digit_to_number (*cp, base);
3375 if (leading_digit >= 0)
3376 {
3377 state |= LEAD_INT;
3378 do
3379 ++cp;
3380 while (digit_to_number (*cp, base) >= 0);
3381 }
3382 if (*cp == '.')
3383 {
3384 state |= DOT_CHAR;
3385 cp++;
3386 }
3387
3388 if (base == 10)
3389 {
3390 if ('0' <= *cp && *cp <= '9')
3391 {
3392 state |= TRAIL_INT;
3393 do
3394 cp++;
3395 while ('0' <= *cp && *cp <= '9');
3396 }
3397 if (*cp == 'e' || *cp == 'E')
3398 {
3399 char const *ecp = cp;
3400 cp++;
3401 if (*cp == '+' || *cp == '-')
3402 cp++;
3403 if ('0' <= *cp && *cp <= '9')
3404 {
3405 state |= E_EXP;
3406 do
3407 cp++;
3408 while ('0' <= *cp && *cp <= '9');
3409 }
3410 else if (cp[-1] == '+'
3411 && cp[0] == 'I' && cp[1] == 'N' && cp[2] == 'F')
3412 {
3413 state |= E_EXP;
3414 cp += 3;
3415 value = INFINITY;
3416 }
3417 else if (cp[-1] == '+'
3418 && cp[0] == 'N' && cp[1] == 'a' && cp[2] == 'N')
3419 {
3420 state |= E_EXP;
3421 cp += 3;
3422 /* NAN is a "positive" NaN on all known Emacs hosts. */
3423 value = NAN;
3424 }
3425 else
3426 cp = ecp;
3427 }
3428
3429 float_syntax = ((state & (DOT_CHAR|TRAIL_INT)) == (DOT_CHAR|TRAIL_INT)
3430 || state == (LEAD_INT|E_EXP));
3431 }
3432
3433 /* Return nil if the number uses invalid syntax. If IGNORE_TRAILING, accept
3434 any prefix that matches. Otherwise, the entire string must match. */
3435 if (! (ignore_trailing
3436 ? ((state & LEAD_INT) != 0 || float_syntax)
3437 : (!*cp && ((state & ~DOT_CHAR) == LEAD_INT || float_syntax))))
3438 return Qnil;
3439
3440 /* If the number uses integer and not float syntax, and is in C-language
3441 range, use its value, preferably as a fixnum. */
3442 if (leading_digit >= 0 && ! float_syntax)
3443 {
3444 uintmax_t n;
3445
3446 /* Fast special case for single-digit integers. This also avoids a
3447 glitch when BASE is 16 and IGNORE_TRAILING, because in that
3448 case some versions of strtoumax accept numbers like "0x1" that Emacs
3449 does not allow. */
3450 if (digit_to_number (string[signedp + 1], base) < 0)
3451 return make_number (negative ? -leading_digit : leading_digit);
3452
3453 errno = 0;
3454 n = strtoumax (string + signedp, NULL, base);
3455 if (errno == ERANGE)
3456 {
3457 /* Unfortunately there's no simple and accurate way to convert
3458 non-base-10 numbers that are out of C-language range. */
3459 if (base != 10)
3460 xsignal1 (Qoverflow_error, build_string (string));
3461 }
3462 else if (n <= (negative ? -MOST_NEGATIVE_FIXNUM : MOST_POSITIVE_FIXNUM))
3463 {
3464 EMACS_INT signed_n = n;
3465 return make_number (negative ? -signed_n : signed_n);
3466 }
3467 else
3468 value = n;
3469 }
3470
3471 /* Either the number uses float syntax, or it does not fit into a fixnum.
3472 Convert it from string to floating point, unless the value is already
3473 known because it is an infinity, a NAN, or its absolute value fits in
3474 uintmax_t. */
3475 if (! value)
3476 value = atof (string + signedp);
3477
3478 return make_float (negative ? -value : value);
3479 }
3480
3481 \f
3482 static Lisp_Object
3483 read_vector (Lisp_Object readcharfun, bool bytecodeflag)
3484 {
3485 ptrdiff_t i, size;
3486 Lisp_Object *ptr;
3487 Lisp_Object tem, item, vector;
3488 struct Lisp_Cons *otem;
3489 Lisp_Object len;
3490
3491 tem = read_list (1, readcharfun);
3492 len = Flength (tem);
3493 vector = Fmake_vector (len, Qnil);
3494
3495 size = ASIZE (vector);
3496 ptr = XVECTOR (vector)->contents;
3497 for (i = 0; i < size; i++)
3498 {
3499 item = Fcar (tem);
3500 /* If `load-force-doc-strings' is t when reading a lazily-loaded
3501 bytecode object, the docstring containing the bytecode and
3502 constants values must be treated as unibyte and passed to
3503 Fread, to get the actual bytecode string and constants vector. */
3504 if (bytecodeflag && load_force_doc_strings)
3505 {
3506 if (i == COMPILED_BYTECODE)
3507 {
3508 if (!STRINGP (item))
3509 error ("Invalid byte code");
3510
3511 /* Delay handling the bytecode slot until we know whether
3512 it is lazily-loaded (we can tell by whether the
3513 constants slot is nil). */
3514 ASET (vector, COMPILED_CONSTANTS, item);
3515 item = Qnil;
3516 }
3517 else if (i == COMPILED_CONSTANTS)
3518 {
3519 Lisp_Object bytestr = ptr[COMPILED_CONSTANTS];
3520
3521 if (NILP (item))
3522 {
3523 /* Coerce string to unibyte (like string-as-unibyte,
3524 but without generating extra garbage and
3525 guaranteeing no change in the contents). */
3526 STRING_SET_CHARS (bytestr, SBYTES (bytestr));
3527 STRING_SET_UNIBYTE (bytestr);
3528
3529 item = Fread (Fcons (bytestr, readcharfun));
3530 if (!CONSP (item))
3531 error ("Invalid byte code");
3532
3533 otem = XCONS (item);
3534 bytestr = XCAR (item);
3535 item = XCDR (item);
3536 free_cons (otem);
3537 }
3538
3539 /* Now handle the bytecode slot. */
3540 ASET (vector, COMPILED_BYTECODE, bytestr);
3541 }
3542 else if (i == COMPILED_DOC_STRING
3543 && STRINGP (item)
3544 && ! STRING_MULTIBYTE (item))
3545 {
3546 if (EQ (readcharfun, Qget_emacs_mule_file_char))
3547 item = Fdecode_coding_string (item, Qemacs_mule, Qnil, Qnil);
3548 else
3549 item = Fstring_as_multibyte (item);
3550 }
3551 }
3552 ASET (vector, i, item);
3553 otem = XCONS (tem);
3554 tem = Fcdr (tem);
3555 free_cons (otem);
3556 }
3557 return vector;
3558 }
3559
3560 /* FLAG means check for ']' to terminate rather than ')' and '.'. */
3561
3562 static Lisp_Object
3563 read_list (bool flag, Lisp_Object readcharfun)
3564 {
3565 Lisp_Object val, tail;
3566 Lisp_Object elt, tem;
3567 /* 0 is the normal case.
3568 1 means this list is a doc reference; replace it with the number 0.
3569 2 means this list is a doc reference; replace it with the doc string. */
3570 int doc_reference = 0;
3571
3572 /* Initialize this to 1 if we are reading a list. */
3573 bool first_in_list = flag <= 0;
3574
3575 val = Qnil;
3576 tail = Qnil;
3577
3578 while (1)
3579 {
3580 int ch;
3581 elt = read1 (readcharfun, &ch, first_in_list);
3582
3583 first_in_list = 0;
3584
3585 /* While building, if the list starts with #$, treat it specially. */
3586 if (EQ (elt, Vload_file_name)
3587 && ! NILP (elt)
3588 && !NILP (Vpurify_flag))
3589 {
3590 if (NILP (Vdoc_file_name))
3591 /* We have not yet called Snarf-documentation, so assume
3592 this file is described in the DOC file
3593 and Snarf-documentation will fill in the right value later.
3594 For now, replace the whole list with 0. */
3595 doc_reference = 1;
3596 else
3597 /* We have already called Snarf-documentation, so make a relative
3598 file name for this file, so it can be found properly
3599 in the installed Lisp directory.
3600 We don't use Fexpand_file_name because that would make
3601 the directory absolute now. */
3602 {
3603 AUTO_STRING (dot_dot_lisp, "../lisp/");
3604 elt = concat2 (dot_dot_lisp, Ffile_name_nondirectory (elt));
3605 }
3606 }
3607 else if (EQ (elt, Vload_file_name)
3608 && ! NILP (elt)
3609 && load_force_doc_strings)
3610 doc_reference = 2;
3611
3612 if (ch)
3613 {
3614 if (flag > 0)
3615 {
3616 if (ch == ']')
3617 return val;
3618 invalid_syntax (") or . in a vector");
3619 }
3620 if (ch == ')')
3621 return val;
3622 if (ch == '.')
3623 {
3624 if (!NILP (tail))
3625 XSETCDR (tail, read0 (readcharfun));
3626 else
3627 val = read0 (readcharfun);
3628 read1 (readcharfun, &ch, 0);
3629
3630 if (ch == ')')
3631 {
3632 if (doc_reference == 1)
3633 return make_number (0);
3634 if (doc_reference == 2 && INTEGERP (XCDR (val)))
3635 {
3636 char *saved = NULL;
3637 file_offset saved_position;
3638 /* Get a doc string from the file we are loading.
3639 If it's in saved_doc_string, get it from there.
3640
3641 Here, we don't know if the string is a
3642 bytecode string or a doc string. As a
3643 bytecode string must be unibyte, we always
3644 return a unibyte string. If it is actually a
3645 doc string, caller must make it
3646 multibyte. */
3647
3648 /* Position is negative for user variables. */
3649 EMACS_INT pos = eabs (XINT (XCDR (val)));
3650 if (pos >= saved_doc_string_position
3651 && pos < (saved_doc_string_position
3652 + saved_doc_string_length))
3653 {
3654 saved = saved_doc_string;
3655 saved_position = saved_doc_string_position;
3656 }
3657 /* Look in prev_saved_doc_string the same way. */
3658 else if (pos >= prev_saved_doc_string_position
3659 && pos < (prev_saved_doc_string_position
3660 + prev_saved_doc_string_length))
3661 {
3662 saved = prev_saved_doc_string;
3663 saved_position = prev_saved_doc_string_position;
3664 }
3665 if (saved)
3666 {
3667 ptrdiff_t start = pos - saved_position;
3668 ptrdiff_t from, to;
3669
3670 /* Process quoting with ^A,
3671 and find the end of the string,
3672 which is marked with ^_ (037). */
3673 for (from = start, to = start;
3674 saved[from] != 037;)
3675 {
3676 int c = saved[from++];
3677 if (c == 1)
3678 {
3679 c = saved[from++];
3680 saved[to++] = (c == 1 ? c
3681 : c == '0' ? 0
3682 : c == '_' ? 037
3683 : c);
3684 }
3685 else
3686 saved[to++] = c;
3687 }
3688
3689 return make_unibyte_string (saved + start,
3690 to - start);
3691 }
3692 else
3693 return get_doc_string (val, 1, 0);
3694 }
3695
3696 return val;
3697 }
3698 invalid_syntax (". in wrong context");
3699 }
3700 invalid_syntax ("] in a list");
3701 }
3702 tem = list1 (elt);
3703 if (!NILP (tail))
3704 XSETCDR (tail, tem);
3705 else
3706 val = tem;
3707 tail = tem;
3708 }
3709 }
3710 \f
3711 static Lisp_Object initial_obarray;
3712
3713 /* `oblookup' stores the bucket number here, for the sake of Funintern. */
3714
3715 static size_t oblookup_last_bucket_number;
3716
3717 /* Get an error if OBARRAY is not an obarray.
3718 If it is one, return it. */
3719
3720 Lisp_Object
3721 check_obarray (Lisp_Object obarray)
3722 {
3723 if (!VECTORP (obarray) || ASIZE (obarray) == 0)
3724 {
3725 /* If Vobarray is now invalid, force it to be valid. */
3726 if (EQ (Vobarray, obarray)) Vobarray = initial_obarray;
3727 wrong_type_argument (Qvectorp, obarray);
3728 }
3729 return obarray;
3730 }
3731
3732 /* Intern symbol SYM in OBARRAY using bucket INDEX. */
3733
3734 static Lisp_Object
3735 intern_sym (Lisp_Object sym, Lisp_Object obarray, Lisp_Object index)
3736 {
3737 Lisp_Object *ptr;
3738
3739 XSYMBOL (sym)->interned = (EQ (obarray, initial_obarray)
3740 ? SYMBOL_INTERNED_IN_INITIAL_OBARRAY
3741 : SYMBOL_INTERNED);
3742
3743 if (SREF (SYMBOL_NAME (sym), 0) == ':' && EQ (obarray, initial_obarray))
3744 {
3745 XSYMBOL (sym)->constant = 1;
3746 XSYMBOL (sym)->redirect = SYMBOL_PLAINVAL;
3747 SET_SYMBOL_VAL (XSYMBOL (sym), sym);
3748 }
3749
3750 ptr = aref_addr (obarray, XINT (index));
3751 set_symbol_next (sym, SYMBOLP (*ptr) ? XSYMBOL (*ptr) : NULL);
3752 *ptr = sym;
3753 return sym;
3754 }
3755
3756 /* Intern a symbol with name STRING in OBARRAY using bucket INDEX. */
3757
3758 Lisp_Object
3759 intern_driver (Lisp_Object string, Lisp_Object obarray, Lisp_Object index)
3760 {
3761 return intern_sym (Fmake_symbol (string), obarray, index);
3762 }
3763
3764 /* Intern the C string STR: return a symbol with that name,
3765 interned in the current obarray. */
3766
3767 Lisp_Object
3768 intern_1 (const char *str, ptrdiff_t len)
3769 {
3770 Lisp_Object obarray = check_obarray (Vobarray);
3771 Lisp_Object tem = oblookup (obarray, str, len, len);
3772
3773 return (SYMBOLP (tem) ? tem
3774 /* The above `oblookup' was done on the basis of nchars==nbytes, so
3775 the string has to be unibyte. */
3776 : intern_driver (make_unibyte_string (str, len),
3777 obarray, tem));
3778 }
3779
3780 Lisp_Object
3781 intern_c_string_1 (const char *str, ptrdiff_t len)
3782 {
3783 Lisp_Object obarray = check_obarray (Vobarray);
3784 Lisp_Object tem = oblookup (obarray, str, len, len);
3785
3786 if (!SYMBOLP (tem))
3787 {
3788 /* Creating a non-pure string from a string literal not implemented yet.
3789 We could just use make_string here and live with the extra copy. */
3790 eassert (!NILP (Vpurify_flag));
3791 tem = intern_driver (make_pure_c_string (str, len), obarray, tem);
3792 }
3793 return tem;
3794 }
3795
3796 static void
3797 define_symbol (Lisp_Object sym, char const *str)
3798 {
3799 ptrdiff_t len = strlen (str);
3800 Lisp_Object string = make_pure_c_string (str, len);
3801 init_symbol (sym, string);
3802
3803 /* Qunbound is uninterned, so that it's not confused with any symbol
3804 'unbound' created by a Lisp program. */
3805 if (! EQ (sym, Qunbound))
3806 {
3807 Lisp_Object bucket = oblookup (initial_obarray, str, len, len);
3808 eassert (INTEGERP (bucket));
3809 intern_sym (sym, initial_obarray, bucket);
3810 }
3811 }
3812 \f
3813 DEFUN ("intern", Fintern, Sintern, 1, 2, 0,
3814 doc: /* Return the canonical symbol whose name is STRING.
3815 If there is none, one is created by this function and returned.
3816 A second optional argument specifies the obarray to use;
3817 it defaults to the value of `obarray'. */)
3818 (Lisp_Object string, Lisp_Object obarray)
3819 {
3820 Lisp_Object tem;
3821
3822 obarray = check_obarray (NILP (obarray) ? Vobarray : obarray);
3823 CHECK_STRING (string);
3824
3825 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3826 if (!SYMBOLP (tem))
3827 tem = intern_driver (NILP (Vpurify_flag) ? string : Fpurecopy (string),
3828 obarray, tem);
3829 return tem;
3830 }
3831
3832 DEFUN ("intern-soft", Fintern_soft, Sintern_soft, 1, 2, 0,
3833 doc: /* Return the canonical symbol named NAME, or nil if none exists.
3834 NAME may be a string or a symbol. If it is a symbol, that exact
3835 symbol is searched for.
3836 A second optional argument specifies the obarray to use;
3837 it defaults to the value of `obarray'. */)
3838 (Lisp_Object name, Lisp_Object obarray)
3839 {
3840 register Lisp_Object tem, string;
3841
3842 if (NILP (obarray)) obarray = Vobarray;
3843 obarray = check_obarray (obarray);
3844
3845 if (!SYMBOLP (name))
3846 {
3847 CHECK_STRING (name);
3848 string = name;
3849 }
3850 else
3851 string = SYMBOL_NAME (name);
3852
3853 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3854 if (INTEGERP (tem) || (SYMBOLP (name) && !EQ (name, tem)))
3855 return Qnil;
3856 else
3857 return tem;
3858 }
3859 \f
3860 DEFUN ("unintern", Funintern, Sunintern, 1, 2, 0,
3861 doc: /* Delete the symbol named NAME, if any, from OBARRAY.
3862 The value is t if a symbol was found and deleted, nil otherwise.
3863 NAME may be a string or a symbol. If it is a symbol, that symbol
3864 is deleted, if it belongs to OBARRAY--no other symbol is deleted.
3865 OBARRAY, if nil, defaults to the value of the variable `obarray'.
3866 usage: (unintern NAME OBARRAY) */)
3867 (Lisp_Object name, Lisp_Object obarray)
3868 {
3869 register Lisp_Object string, tem;
3870 size_t hash;
3871
3872 if (NILP (obarray)) obarray = Vobarray;
3873 obarray = check_obarray (obarray);
3874
3875 if (SYMBOLP (name))
3876 string = SYMBOL_NAME (name);
3877 else
3878 {
3879 CHECK_STRING (name);
3880 string = name;
3881 }
3882
3883 tem = oblookup (obarray, SSDATA (string),
3884 SCHARS (string),
3885 SBYTES (string));
3886 if (INTEGERP (tem))
3887 return Qnil;
3888 /* If arg was a symbol, don't delete anything but that symbol itself. */
3889 if (SYMBOLP (name) && !EQ (name, tem))
3890 return Qnil;
3891
3892 /* There are plenty of other symbols which will screw up the Emacs
3893 session if we unintern them, as well as even more ways to use
3894 `setq' or `fset' or whatnot to make the Emacs session
3895 unusable. Let's not go down this silly road. --Stef */
3896 /* if (EQ (tem, Qnil) || EQ (tem, Qt))
3897 error ("Attempt to unintern t or nil"); */
3898
3899 XSYMBOL (tem)->interned = SYMBOL_UNINTERNED;
3900
3901 hash = oblookup_last_bucket_number;
3902
3903 if (EQ (AREF (obarray, hash), tem))
3904 {
3905 if (XSYMBOL (tem)->next)
3906 {
3907 Lisp_Object sym;
3908 XSETSYMBOL (sym, XSYMBOL (tem)->next);
3909 ASET (obarray, hash, sym);
3910 }
3911 else
3912 ASET (obarray, hash, make_number (0));
3913 }
3914 else
3915 {
3916 Lisp_Object tail, following;
3917
3918 for (tail = AREF (obarray, hash);
3919 XSYMBOL (tail)->next;
3920 tail = following)
3921 {
3922 XSETSYMBOL (following, XSYMBOL (tail)->next);
3923 if (EQ (following, tem))
3924 {
3925 set_symbol_next (tail, XSYMBOL (following)->next);
3926 break;
3927 }
3928 }
3929 }
3930
3931 return Qt;
3932 }
3933 \f
3934 /* Return the symbol in OBARRAY whose names matches the string
3935 of SIZE characters (SIZE_BYTE bytes) at PTR.
3936 If there is no such symbol, return the integer bucket number of
3937 where the symbol would be if it were present.
3938
3939 Also store the bucket number in oblookup_last_bucket_number. */
3940
3941 Lisp_Object
3942 oblookup (Lisp_Object obarray, register const char *ptr, ptrdiff_t size, ptrdiff_t size_byte)
3943 {
3944 size_t hash;
3945 size_t obsize;
3946 register Lisp_Object tail;
3947 Lisp_Object bucket, tem;
3948
3949 obarray = check_obarray (obarray);
3950 /* This is sometimes needed in the middle of GC. */
3951 obsize = gc_asize (obarray);
3952 hash = hash_string (ptr, size_byte) % obsize;
3953 bucket = AREF (obarray, hash);
3954 oblookup_last_bucket_number = hash;
3955 if (EQ (bucket, make_number (0)))
3956 ;
3957 else if (!SYMBOLP (bucket))
3958 error ("Bad data in guts of obarray"); /* Like CADR error message. */
3959 else
3960 for (tail = bucket; ; XSETSYMBOL (tail, XSYMBOL (tail)->next))
3961 {
3962 if (SBYTES (SYMBOL_NAME (tail)) == size_byte
3963 && SCHARS (SYMBOL_NAME (tail)) == size
3964 && !memcmp (SDATA (SYMBOL_NAME (tail)), ptr, size_byte))
3965 return tail;
3966 else if (XSYMBOL (tail)->next == 0)
3967 break;
3968 }
3969 XSETINT (tem, hash);
3970 return tem;
3971 }
3972 \f
3973 void
3974 map_obarray (Lisp_Object obarray, void (*fn) (Lisp_Object, Lisp_Object), Lisp_Object arg)
3975 {
3976 ptrdiff_t i;
3977 register Lisp_Object tail;
3978 CHECK_VECTOR (obarray);
3979 for (i = ASIZE (obarray) - 1; i >= 0; i--)
3980 {
3981 tail = AREF (obarray, i);
3982 if (SYMBOLP (tail))
3983 while (1)
3984 {
3985 (*fn) (tail, arg);
3986 if (XSYMBOL (tail)->next == 0)
3987 break;
3988 XSETSYMBOL (tail, XSYMBOL (tail)->next);
3989 }
3990 }
3991 }
3992
3993 static void
3994 mapatoms_1 (Lisp_Object sym, Lisp_Object function)
3995 {
3996 call1 (function, sym);
3997 }
3998
3999 DEFUN ("mapatoms", Fmapatoms, Smapatoms, 1, 2, 0,
4000 doc: /* Call FUNCTION on every symbol in OBARRAY.
4001 OBARRAY defaults to the value of `obarray'. */)
4002 (Lisp_Object function, Lisp_Object obarray)
4003 {
4004 if (NILP (obarray)) obarray = Vobarray;
4005 obarray = check_obarray (obarray);
4006
4007 map_obarray (obarray, mapatoms_1, function);
4008 return Qnil;
4009 }
4010
4011 #define OBARRAY_SIZE 1511
4012
4013 void
4014 init_obarray (void)
4015 {
4016 Lisp_Object oblength;
4017 ptrdiff_t size = 100 + MAX_MULTIBYTE_LENGTH;
4018
4019 XSETFASTINT (oblength, OBARRAY_SIZE);
4020
4021 Vobarray = Fmake_vector (oblength, make_number (0));
4022 initial_obarray = Vobarray;
4023 staticpro (&initial_obarray);
4024
4025 for (int i = 0; i < ARRAYELTS (lispsym); i++)
4026 define_symbol (builtin_lisp_symbol (i), defsym_name[i]);
4027
4028 DEFSYM (Qunbound, "unbound");
4029
4030 DEFSYM (Qnil, "nil");
4031 SET_SYMBOL_VAL (XSYMBOL (Qnil), Qnil);
4032 XSYMBOL (Qnil)->constant = 1;
4033 XSYMBOL (Qnil)->declared_special = true;
4034
4035 DEFSYM (Qt, "t");
4036 SET_SYMBOL_VAL (XSYMBOL (Qt), Qt);
4037 XSYMBOL (Qt)->constant = 1;
4038 XSYMBOL (Qt)->declared_special = true;
4039
4040 /* Qt is correct even if CANNOT_DUMP. loadup.el will set to nil at end. */
4041 Vpurify_flag = Qt;
4042
4043 DEFSYM (Qvariable_documentation, "variable-documentation");
4044
4045 read_buffer = xmalloc (size);
4046 read_buffer_size = size;
4047 }
4048 \f
4049 void
4050 defsubr (struct Lisp_Subr *sname)
4051 {
4052 Lisp_Object sym, tem;
4053 sym = intern_c_string (sname->symbol_name);
4054 XSETPVECTYPE (sname, PVEC_SUBR);
4055 XSETSUBR (tem, sname);
4056 set_symbol_function (sym, tem);
4057 }
4058
4059 #ifdef NOTDEF /* Use fset in subr.el now! */
4060 void
4061 defalias (struct Lisp_Subr *sname, char *string)
4062 {
4063 Lisp_Object sym;
4064 sym = intern (string);
4065 XSETSUBR (XSYMBOL (sym)->function, sname);
4066 }
4067 #endif /* NOTDEF */
4068
4069 /* Define an "integer variable"; a symbol whose value is forwarded to a
4070 C variable of type EMACS_INT. Sample call (with "xx" to fool make-docfile):
4071 DEFxxVAR_INT ("emacs-priority", &emacs_priority, "Documentation"); */
4072 void
4073 defvar_int (struct Lisp_Intfwd *i_fwd,
4074 const char *namestring, EMACS_INT *address)
4075 {
4076 Lisp_Object sym;
4077 sym = intern_c_string (namestring);
4078 i_fwd->type = Lisp_Fwd_Int;
4079 i_fwd->intvar = address;
4080 XSYMBOL (sym)->declared_special = 1;
4081 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4082 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)i_fwd);
4083 }
4084
4085 /* Similar but define a variable whose value is t if address contains 1,
4086 nil if address contains 0. */
4087 void
4088 defvar_bool (struct Lisp_Boolfwd *b_fwd,
4089 const char *namestring, bool *address)
4090 {
4091 Lisp_Object sym;
4092 sym = intern_c_string (namestring);
4093 b_fwd->type = Lisp_Fwd_Bool;
4094 b_fwd->boolvar = address;
4095 XSYMBOL (sym)->declared_special = 1;
4096 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4097 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)b_fwd);
4098 Vbyte_boolean_vars = Fcons (sym, Vbyte_boolean_vars);
4099 }
4100
4101 /* Similar but define a variable whose value is the Lisp Object stored
4102 at address. Two versions: with and without gc-marking of the C
4103 variable. The nopro version is used when that variable will be
4104 gc-marked for some other reason, since marking the same slot twice
4105 can cause trouble with strings. */
4106 void
4107 defvar_lisp_nopro (struct Lisp_Objfwd *o_fwd,
4108 const char *namestring, Lisp_Object *address)
4109 {
4110 Lisp_Object sym;
4111 sym = intern_c_string (namestring);
4112 o_fwd->type = Lisp_Fwd_Obj;
4113 o_fwd->objvar = address;
4114 XSYMBOL (sym)->declared_special = 1;
4115 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4116 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)o_fwd);
4117 }
4118
4119 void
4120 defvar_lisp (struct Lisp_Objfwd *o_fwd,
4121 const char *namestring, Lisp_Object *address)
4122 {
4123 defvar_lisp_nopro (o_fwd, namestring, address);
4124 staticpro (address);
4125 }
4126
4127 /* Similar but define a variable whose value is the Lisp Object stored
4128 at a particular offset in the current kboard object. */
4129
4130 void
4131 defvar_kboard (struct Lisp_Kboard_Objfwd *ko_fwd,
4132 const char *namestring, int offset)
4133 {
4134 Lisp_Object sym;
4135 sym = intern_c_string (namestring);
4136 ko_fwd->type = Lisp_Fwd_Kboard_Obj;
4137 ko_fwd->offset = offset;
4138 XSYMBOL (sym)->declared_special = 1;
4139 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4140 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)ko_fwd);
4141 }
4142 \f
4143 /* Check that the elements of lpath exist. */
4144
4145 static void
4146 load_path_check (Lisp_Object lpath)
4147 {
4148 Lisp_Object path_tail;
4149
4150 /* The only elements that might not exist are those from
4151 PATH_LOADSEARCH, EMACSLOADPATH. Anything else is only added if
4152 it exists. */
4153 for (path_tail = lpath; !NILP (path_tail); path_tail = XCDR (path_tail))
4154 {
4155 Lisp_Object dirfile;
4156 dirfile = Fcar (path_tail);
4157 if (STRINGP (dirfile))
4158 {
4159 dirfile = Fdirectory_file_name (dirfile);
4160 if (! file_accessible_directory_p (dirfile))
4161 dir_warning ("Lisp directory", XCAR (path_tail));
4162 }
4163 }
4164 }
4165
4166 /* Return the default load-path, to be used if EMACSLOADPATH is unset.
4167 This does not include the standard site-lisp directories
4168 under the installation prefix (i.e., PATH_SITELOADSEARCH),
4169 but it does (unless no_site_lisp is set) include site-lisp
4170 directories in the source/build directories if those exist and we
4171 are running uninstalled.
4172
4173 Uses the following logic:
4174 If CANNOT_DUMP: Use PATH_LOADSEARCH.
4175 The remainder is what happens when dumping works:
4176 If purify-flag (ie dumping) just use PATH_DUMPLOADSEARCH.
4177 Otherwise use PATH_LOADSEARCH.
4178
4179 If !initialized, then just return PATH_DUMPLOADSEARCH.
4180 If initialized:
4181 If Vinstallation_directory is not nil (ie, running uninstalled):
4182 If installation-dir/lisp exists and not already a member,
4183 we must be running uninstalled. Reset the load-path
4184 to just installation-dir/lisp. (The default PATH_LOADSEARCH
4185 refers to the eventual installation directories. Since we
4186 are not yet installed, we should not use them, even if they exist.)
4187 If installation-dir/lisp does not exist, just add
4188 PATH_DUMPLOADSEARCH at the end instead.
4189 Add installation-dir/site-lisp (if !no_site_lisp, and exists
4190 and not already a member) at the front.
4191 If installation-dir != source-dir (ie running an uninstalled,
4192 out-of-tree build) AND install-dir/src/Makefile exists BUT
4193 install-dir/src/Makefile.in does NOT exist (this is a sanity
4194 check), then repeat the above steps for source-dir/lisp, site-lisp. */
4195
4196 static Lisp_Object
4197 load_path_default (void)
4198 {
4199 Lisp_Object lpath = Qnil;
4200 const char *normal;
4201
4202 #ifdef CANNOT_DUMP
4203 #ifdef HAVE_NS
4204 const char *loadpath = ns_load_path ();
4205 #endif
4206
4207 normal = PATH_LOADSEARCH;
4208 #ifdef HAVE_NS
4209 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4210 #else
4211 lpath = decode_env_path (0, normal, 0);
4212 #endif
4213
4214 #else /* !CANNOT_DUMP */
4215
4216 normal = NILP (Vpurify_flag) ? PATH_LOADSEARCH : PATH_DUMPLOADSEARCH;
4217
4218 if (initialized)
4219 {
4220 #ifdef HAVE_NS
4221 const char *loadpath = ns_load_path ();
4222 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4223 #else
4224 lpath = decode_env_path (0, normal, 0);
4225 #endif
4226 if (!NILP (Vinstallation_directory))
4227 {
4228 Lisp_Object tem, tem1;
4229
4230 /* Add to the path the lisp subdir of the installation
4231 dir, if it is accessible. Note: in out-of-tree builds,
4232 this directory is empty save for Makefile. */
4233 tem = Fexpand_file_name (build_string ("lisp"),
4234 Vinstallation_directory);
4235 tem1 = Ffile_accessible_directory_p (tem);
4236 if (!NILP (tem1))
4237 {
4238 if (NILP (Fmember (tem, lpath)))
4239 {
4240 /* We are running uninstalled. The default load-path
4241 points to the eventual installed lisp directories.
4242 We should not use those now, even if they exist,
4243 so start over from a clean slate. */
4244 lpath = list1 (tem);
4245 }
4246 }
4247 else
4248 /* That dir doesn't exist, so add the build-time
4249 Lisp dirs instead. */
4250 {
4251 Lisp_Object dump_path =
4252 decode_env_path (0, PATH_DUMPLOADSEARCH, 0);
4253 lpath = nconc2 (lpath, dump_path);
4254 }
4255
4256 /* Add site-lisp under the installation dir, if it exists. */
4257 if (!no_site_lisp)
4258 {
4259 tem = Fexpand_file_name (build_string ("site-lisp"),
4260 Vinstallation_directory);
4261 tem1 = Ffile_accessible_directory_p (tem);
4262 if (!NILP (tem1))
4263 {
4264 if (NILP (Fmember (tem, lpath)))
4265 lpath = Fcons (tem, lpath);
4266 }
4267 }
4268
4269 /* If Emacs was not built in the source directory,
4270 and it is run from where it was built, add to load-path
4271 the lisp and site-lisp dirs under that directory. */
4272
4273 if (NILP (Fequal (Vinstallation_directory, Vsource_directory)))
4274 {
4275 Lisp_Object tem2;
4276
4277 tem = Fexpand_file_name (build_string ("src/Makefile"),
4278 Vinstallation_directory);
4279 tem1 = Ffile_exists_p (tem);
4280
4281 /* Don't be fooled if they moved the entire source tree
4282 AFTER dumping Emacs. If the build directory is indeed
4283 different from the source dir, src/Makefile.in and
4284 src/Makefile will not be found together. */
4285 tem = Fexpand_file_name (build_string ("src/Makefile.in"),
4286 Vinstallation_directory);
4287 tem2 = Ffile_exists_p (tem);
4288 if (!NILP (tem1) && NILP (tem2))
4289 {
4290 tem = Fexpand_file_name (build_string ("lisp"),
4291 Vsource_directory);
4292
4293 if (NILP (Fmember (tem, lpath)))
4294 lpath = Fcons (tem, lpath);
4295
4296 if (!no_site_lisp)
4297 {
4298 tem = Fexpand_file_name (build_string ("site-lisp"),
4299 Vsource_directory);
4300 tem1 = Ffile_accessible_directory_p (tem);
4301 if (!NILP (tem1))
4302 {
4303 if (NILP (Fmember (tem, lpath)))
4304 lpath = Fcons (tem, lpath);
4305 }
4306 }
4307 }
4308 } /* Vinstallation_directory != Vsource_directory */
4309
4310 } /* if Vinstallation_directory */
4311 }
4312 else /* !initialized */
4313 {
4314 /* NORMAL refers to PATH_DUMPLOADSEARCH, ie the lisp dir in the
4315 source directory. We used to add ../lisp (ie the lisp dir in
4316 the build directory) at the front here, but that should not
4317 be necessary, since in out of tree builds lisp/ is empty, save
4318 for Makefile. */
4319 lpath = decode_env_path (0, normal, 0);
4320 }
4321 #endif /* !CANNOT_DUMP */
4322
4323 return lpath;
4324 }
4325
4326 void
4327 init_lread (void)
4328 {
4329 /* First, set Vload_path. */
4330
4331 /* Ignore EMACSLOADPATH when dumping. */
4332 #ifdef CANNOT_DUMP
4333 bool use_loadpath = true;
4334 #else
4335 bool use_loadpath = NILP (Vpurify_flag);
4336 #endif
4337
4338 if (use_loadpath && egetenv ("EMACSLOADPATH"))
4339 {
4340 Vload_path = decode_env_path ("EMACSLOADPATH", 0, 1);
4341
4342 /* Check (non-nil) user-supplied elements. */
4343 load_path_check (Vload_path);
4344
4345 /* If no nils in the environment variable, use as-is.
4346 Otherwise, replace any nils with the default. */
4347 if (! NILP (Fmemq (Qnil, Vload_path)))
4348 {
4349 Lisp_Object elem, elpath = Vload_path;
4350 Lisp_Object default_lpath = load_path_default ();
4351
4352 /* Check defaults, before adding site-lisp. */
4353 load_path_check (default_lpath);
4354
4355 /* Add the site-lisp directories to the front of the default. */
4356 if (!no_site_lisp && PATH_SITELOADSEARCH[0] != '\0')
4357 {
4358 Lisp_Object sitelisp;
4359 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4360 if (! NILP (sitelisp))
4361 default_lpath = nconc2 (sitelisp, default_lpath);
4362 }
4363
4364 Vload_path = Qnil;
4365
4366 /* Replace nils from EMACSLOADPATH by default. */
4367 while (CONSP (elpath))
4368 {
4369 elem = XCAR (elpath);
4370 elpath = XCDR (elpath);
4371 Vload_path = CALLN (Fappend, Vload_path,
4372 NILP (elem) ? default_lpath : list1 (elem));
4373 }
4374 } /* Fmemq (Qnil, Vload_path) */
4375 }
4376 else
4377 {
4378 Vload_path = load_path_default ();
4379
4380 /* Check before adding site-lisp directories.
4381 The install should have created them, but they are not
4382 required, so no need to warn if they are absent.
4383 Or we might be running before installation. */
4384 load_path_check (Vload_path);
4385
4386 /* Add the site-lisp directories at the front. */
4387 if (initialized && !no_site_lisp && PATH_SITELOADSEARCH[0] != '\0')
4388 {
4389 Lisp_Object sitelisp;
4390 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4391 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4392 }
4393 }
4394
4395 Vvalues = Qnil;
4396
4397 load_in_progress = 0;
4398 Vload_file_name = Qnil;
4399 Vstandard_input = Qt;
4400 Vloads_in_progress = Qnil;
4401 }
4402
4403 /* Print a warning that directory intended for use USE and with name
4404 DIRNAME cannot be accessed. On entry, errno should correspond to
4405 the access failure. Print the warning on stderr and put it in
4406 *Messages*. */
4407
4408 void
4409 dir_warning (char const *use, Lisp_Object dirname)
4410 {
4411 static char const format[] = "Warning: %s '%s': %s\n";
4412 int access_errno = errno;
4413 fprintf (stderr, format, use, SSDATA (ENCODE_SYSTEM (dirname)),
4414 strerror (access_errno));
4415
4416 /* Don't log the warning before we've initialized!! */
4417 if (initialized)
4418 {
4419 char const *diagnostic = emacs_strerror (access_errno);
4420 USE_SAFE_ALLOCA;
4421 char *buffer = SAFE_ALLOCA (sizeof format - 3 * (sizeof "%s" - 1)
4422 + strlen (use) + SBYTES (dirname)
4423 + strlen (diagnostic));
4424 ptrdiff_t message_len = esprintf (buffer, format, use, SSDATA (dirname),
4425 diagnostic);
4426 message_dolog (buffer, message_len, 0, STRING_MULTIBYTE (dirname));
4427 SAFE_FREE ();
4428 }
4429 }
4430
4431 void
4432 syms_of_lread (void)
4433 {
4434 defsubr (&Sread);
4435 defsubr (&Sread_from_string);
4436 defsubr (&Sintern);
4437 defsubr (&Sintern_soft);
4438 defsubr (&Sunintern);
4439 defsubr (&Sget_load_suffixes);
4440 defsubr (&Sload);
4441 defsubr (&Seval_buffer);
4442 defsubr (&Seval_region);
4443 defsubr (&Sread_char);
4444 defsubr (&Sread_char_exclusive);
4445 defsubr (&Sread_event);
4446 defsubr (&Sget_file_char);
4447 defsubr (&Smapatoms);
4448 defsubr (&Slocate_file_internal);
4449
4450 DEFVAR_LISP ("obarray", Vobarray,
4451 doc: /* Symbol table for use by `intern' and `read'.
4452 It is a vector whose length ought to be prime for best results.
4453 The vector's contents don't make sense if examined from Lisp programs;
4454 to find all the symbols in an obarray, use `mapatoms'. */);
4455
4456 DEFVAR_LISP ("values", Vvalues,
4457 doc: /* List of values of all expressions which were read, evaluated and printed.
4458 Order is reverse chronological. */);
4459 XSYMBOL (intern ("values"))->declared_special = 0;
4460
4461 DEFVAR_LISP ("standard-input", Vstandard_input,
4462 doc: /* Stream for read to get input from.
4463 See documentation of `read' for possible values. */);
4464 Vstandard_input = Qt;
4465
4466 DEFVAR_LISP ("read-with-symbol-positions", Vread_with_symbol_positions,
4467 doc: /* If non-nil, add position of read symbols to `read-symbol-positions-list'.
4468
4469 If this variable is a buffer, then only forms read from that buffer
4470 will be added to `read-symbol-positions-list'.
4471 If this variable is t, then all read forms will be added.
4472 The effect of all other values other than nil are not currently
4473 defined, although they may be in the future.
4474
4475 The positions are relative to the last call to `read' or
4476 `read-from-string'. It is probably a bad idea to set this variable at
4477 the toplevel; bind it instead. */);
4478 Vread_with_symbol_positions = Qnil;
4479
4480 DEFVAR_LISP ("read-symbol-positions-list", Vread_symbol_positions_list,
4481 doc: /* A list mapping read symbols to their positions.
4482 This variable is modified during calls to `read' or
4483 `read-from-string', but only when `read-with-symbol-positions' is
4484 non-nil.
4485
4486 Each element of the list looks like (SYMBOL . CHAR-POSITION), where
4487 CHAR-POSITION is an integer giving the offset of that occurrence of the
4488 symbol from the position where `read' or `read-from-string' started.
4489
4490 Note that a symbol will appear multiple times in this list, if it was
4491 read multiple times. The list is in the same order as the symbols
4492 were read in. */);
4493 Vread_symbol_positions_list = Qnil;
4494
4495 DEFVAR_LISP ("read-circle", Vread_circle,
4496 doc: /* Non-nil means read recursive structures using #N= and #N# syntax. */);
4497 Vread_circle = Qt;
4498
4499 DEFVAR_LISP ("load-path", Vload_path,
4500 doc: /* List of directories to search for files to load.
4501 Each element is a string (directory file name) or nil (meaning
4502 `default-directory').
4503 Initialized during startup as described in Info node `(elisp)Library Search'.
4504 Use `directory-file-name' when adding items to this path. However, Lisp
4505 programs that process this list should tolerate directories both with
4506 and without trailing slashes. */);
4507
4508 DEFVAR_LISP ("load-suffixes", Vload_suffixes,
4509 doc: /* List of suffixes for Emacs Lisp files and dynamic modules.
4510 This list includes suffixes for both compiled and source Emacs Lisp files.
4511 This list should not include the empty string.
4512 `load' and related functions try to append these suffixes, in order,
4513 to the specified file name if a suffix is allowed or required. */);
4514 #ifdef HAVE_MODULES
4515 Vload_suffixes = list3 (build_pure_c_string (".elc"),
4516 build_pure_c_string (".el"),
4517 build_pure_c_string (MODULES_SUFFIX));
4518 #else
4519 Vload_suffixes = list2 (build_pure_c_string (".elc"),
4520 build_pure_c_string (".el"));
4521 #endif
4522 DEFVAR_LISP ("module-file-suffix", Vmodule_file_suffix,
4523 doc: /* Suffix of loadable module file, or nil of modules are not supported. */);
4524 #ifdef HAVE_MODULES
4525 Vmodule_file_suffix = build_pure_c_string (MODULES_SUFFIX);
4526 #else
4527 Vmodule_file_suffix = Qnil;
4528 #endif
4529 DEFVAR_LISP ("load-file-rep-suffixes", Vload_file_rep_suffixes,
4530 doc: /* List of suffixes that indicate representations of \
4531 the same file.
4532 This list should normally start with the empty string.
4533
4534 Enabling Auto Compression mode appends the suffixes in
4535 `jka-compr-load-suffixes' to this list and disabling Auto Compression
4536 mode removes them again. `load' and related functions use this list to
4537 determine whether they should look for compressed versions of a file
4538 and, if so, which suffixes they should try to append to the file name
4539 in order to do so. However, if you want to customize which suffixes
4540 the loading functions recognize as compression suffixes, you should
4541 customize `jka-compr-load-suffixes' rather than the present variable. */);
4542 Vload_file_rep_suffixes = list1 (empty_unibyte_string);
4543
4544 DEFVAR_BOOL ("load-in-progress", load_in_progress,
4545 doc: /* Non-nil if inside of `load'. */);
4546 DEFSYM (Qload_in_progress, "load-in-progress");
4547
4548 DEFVAR_LISP ("after-load-alist", Vafter_load_alist,
4549 doc: /* An alist of functions to be evalled when particular files are loaded.
4550 Each element looks like (REGEXP-OR-FEATURE FUNCS...).
4551
4552 REGEXP-OR-FEATURE is either a regular expression to match file names, or
4553 a symbol (a feature name).
4554
4555 When `load' is run and the file-name argument matches an element's
4556 REGEXP-OR-FEATURE, or when `provide' is run and provides the symbol
4557 REGEXP-OR-FEATURE, the FUNCS in the element are called.
4558
4559 An error in FORMS does not undo the load, but does prevent execution of
4560 the rest of the FORMS. */);
4561 Vafter_load_alist = Qnil;
4562
4563 DEFVAR_LISP ("load-history", Vload_history,
4564 doc: /* Alist mapping loaded file names to symbols and features.
4565 Each alist element should be a list (FILE-NAME ENTRIES...), where
4566 FILE-NAME is the name of a file that has been loaded into Emacs.
4567 The file name is absolute and true (i.e. it doesn't contain symlinks).
4568 As an exception, one of the alist elements may have FILE-NAME nil,
4569 for symbols and features not associated with any file.
4570
4571 The remaining ENTRIES in the alist element describe the functions and
4572 variables defined in that file, the features provided, and the
4573 features required. Each entry has the form `(provide . FEATURE)',
4574 `(require . FEATURE)', `(defun . FUNCTION)', `(autoload . SYMBOL)',
4575 `(defface . SYMBOL)', or `(t . SYMBOL)'. Entries like `(t . SYMBOL)'
4576 may precede a `(defun . FUNCTION)' entry, and means that SYMBOL was an
4577 autoload before this file redefined it as a function. In addition,
4578 entries may also be single symbols, which means that SYMBOL was
4579 defined by `defvar' or `defconst'.
4580
4581 During preloading, the file name recorded is relative to the main Lisp
4582 directory. These file names are converted to absolute at startup. */);
4583 Vload_history = Qnil;
4584
4585 DEFVAR_LISP ("load-file-name", Vload_file_name,
4586 doc: /* Full name of file being loaded by `load'. */);
4587 Vload_file_name = Qnil;
4588
4589 DEFVAR_LISP ("user-init-file", Vuser_init_file,
4590 doc: /* File name, including directory, of user's initialization file.
4591 If the file loaded had extension `.elc', and the corresponding source file
4592 exists, this variable contains the name of source file, suitable for use
4593 by functions like `custom-save-all' which edit the init file.
4594 While Emacs loads and evaluates the init file, value is the real name
4595 of the file, regardless of whether or not it has the `.elc' extension. */);
4596 Vuser_init_file = Qnil;
4597
4598 DEFVAR_LISP ("current-load-list", Vcurrent_load_list,
4599 doc: /* Used for internal purposes by `load'. */);
4600 Vcurrent_load_list = Qnil;
4601
4602 DEFVAR_LISP ("load-read-function", Vload_read_function,
4603 doc: /* Function used by `load' and `eval-region' for reading expressions.
4604 Called with a single argument (the stream from which to read).
4605 The default is to use the function `read'. */);
4606 DEFSYM (Qread, "read");
4607 Vload_read_function = Qread;
4608
4609 DEFVAR_LISP ("load-source-file-function", Vload_source_file_function,
4610 doc: /* Function called in `load' to load an Emacs Lisp source file.
4611 The value should be a function for doing code conversion before
4612 reading a source file. It can also be nil, in which case loading is
4613 done without any code conversion.
4614
4615 If the value is a function, it is called with four arguments,
4616 FULLNAME, FILE, NOERROR, NOMESSAGE. FULLNAME is the absolute name of
4617 the file to load, FILE is the non-absolute name (for messages etc.),
4618 and NOERROR and NOMESSAGE are the corresponding arguments passed to
4619 `load'. The function should return t if the file was loaded. */);
4620 Vload_source_file_function = Qnil;
4621
4622 DEFVAR_BOOL ("load-force-doc-strings", load_force_doc_strings,
4623 doc: /* Non-nil means `load' should force-load all dynamic doc strings.
4624 This is useful when the file being loaded is a temporary copy. */);
4625 load_force_doc_strings = 0;
4626
4627 DEFVAR_BOOL ("load-convert-to-unibyte", load_convert_to_unibyte,
4628 doc: /* Non-nil means `read' converts strings to unibyte whenever possible.
4629 This is normally bound by `load' and `eval-buffer' to control `read',
4630 and is not meant for users to change. */);
4631 load_convert_to_unibyte = 0;
4632
4633 DEFVAR_LISP ("source-directory", Vsource_directory,
4634 doc: /* Directory in which Emacs sources were found when Emacs was built.
4635 You cannot count on them to still be there! */);
4636 Vsource_directory
4637 = Fexpand_file_name (build_string ("../"),
4638 Fcar (decode_env_path (0, PATH_DUMPLOADSEARCH, 0)));
4639
4640 DEFVAR_LISP ("preloaded-file-list", Vpreloaded_file_list,
4641 doc: /* List of files that were preloaded (when dumping Emacs). */);
4642 Vpreloaded_file_list = Qnil;
4643
4644 DEFVAR_LISP ("byte-boolean-vars", Vbyte_boolean_vars,
4645 doc: /* List of all DEFVAR_BOOL variables, used by the byte code optimizer. */);
4646 Vbyte_boolean_vars = Qnil;
4647
4648 DEFVAR_BOOL ("load-dangerous-libraries", load_dangerous_libraries,
4649 doc: /* Non-nil means load dangerous compiled Lisp files.
4650 Some versions of XEmacs use different byte codes than Emacs. These
4651 incompatible byte codes can make Emacs crash when it tries to execute
4652 them. */);
4653 load_dangerous_libraries = 0;
4654
4655 DEFVAR_BOOL ("force-load-messages", force_load_messages,
4656 doc: /* Non-nil means force printing messages when loading Lisp files.
4657 This overrides the value of the NOMESSAGE argument to `load'. */);
4658 force_load_messages = 0;
4659
4660 DEFVAR_LISP ("bytecomp-version-regexp", Vbytecomp_version_regexp,
4661 doc: /* Regular expression matching safe to load compiled Lisp files.
4662 When Emacs loads a compiled Lisp file, it reads the first 512 bytes
4663 from the file, and matches them against this regular expression.
4664 When the regular expression matches, the file is considered to be safe
4665 to load. See also `load-dangerous-libraries'. */);
4666 Vbytecomp_version_regexp
4667 = build_pure_c_string ("^;;;.\\(in Emacs version\\|bytecomp version FSF\\)");
4668
4669 DEFSYM (Qlexical_binding, "lexical-binding");
4670 DEFVAR_LISP ("lexical-binding", Vlexical_binding,
4671 doc: /* Whether to use lexical binding when evaluating code.
4672 Non-nil means that the code in the current buffer should be evaluated
4673 with lexical binding.
4674 This variable is automatically set from the file variables of an
4675 interpreted Lisp file read using `load'. Unlike other file local
4676 variables, this must be set in the first line of a file. */);
4677 Vlexical_binding = Qnil;
4678 Fmake_variable_buffer_local (Qlexical_binding);
4679
4680 DEFVAR_LISP ("eval-buffer-list", Veval_buffer_list,
4681 doc: /* List of buffers being read from by calls to `eval-buffer' and `eval-region'. */);
4682 Veval_buffer_list = Qnil;
4683
4684 DEFVAR_LISP ("old-style-backquotes", Vold_style_backquotes,
4685 doc: /* Set to non-nil when `read' encounters an old-style backquote. */);
4686 Vold_style_backquotes = Qnil;
4687 DEFSYM (Qold_style_backquotes, "old-style-backquotes");
4688
4689 DEFVAR_BOOL ("load-prefer-newer", load_prefer_newer,
4690 doc: /* Non-nil means `load' prefers the newest version of a file.
4691 This applies when a filename suffix is not explicitly specified and
4692 `load' is trying various possible suffixes (see `load-suffixes' and
4693 `load-file-rep-suffixes'). Normally, it stops at the first file
4694 that exists unless you explicitly specify one or the other. If this
4695 option is non-nil, it checks all suffixes and uses whichever file is
4696 newest.
4697 Note that if you customize this, obviously it will not affect files
4698 that are loaded before your customizations are read! */);
4699 load_prefer_newer = 0;
4700
4701 /* Vsource_directory was initialized in init_lread. */
4702
4703 DEFSYM (Qcurrent_load_list, "current-load-list");
4704 DEFSYM (Qstandard_input, "standard-input");
4705 DEFSYM (Qread_char, "read-char");
4706 DEFSYM (Qget_file_char, "get-file-char");
4707
4708 /* Used instead of Qget_file_char while loading *.elc files compiled
4709 by Emacs 21 or older. */
4710 DEFSYM (Qget_emacs_mule_file_char, "get-emacs-mule-file-char");
4711
4712 DEFSYM (Qload_force_doc_strings, "load-force-doc-strings");
4713
4714 DEFSYM (Qbackquote, "`");
4715 DEFSYM (Qcomma, ",");
4716 DEFSYM (Qcomma_at, ",@");
4717 DEFSYM (Qcomma_dot, ",.");
4718
4719 DEFSYM (Qinhibit_file_name_operation, "inhibit-file-name-operation");
4720 DEFSYM (Qascii_character, "ascii-character");
4721 DEFSYM (Qfunction, "function");
4722 DEFSYM (Qload, "load");
4723 DEFSYM (Qload_file_name, "load-file-name");
4724 DEFSYM (Qeval_buffer_list, "eval-buffer-list");
4725 DEFSYM (Qfile_truename, "file-truename");
4726 DEFSYM (Qdir_ok, "dir-ok");
4727 DEFSYM (Qdo_after_load_evaluation, "do-after-load-evaluation");
4728
4729 staticpro (&read_objects);
4730 read_objects = Qnil;
4731 staticpro (&seen_list);
4732 seen_list = Qnil;
4733
4734 Vloads_in_progress = Qnil;
4735 staticpro (&Vloads_in_progress);
4736
4737 DEFSYM (Qhash_table, "hash-table");
4738 DEFSYM (Qdata, "data");
4739 DEFSYM (Qtest, "test");
4740 DEFSYM (Qsize, "size");
4741 DEFSYM (Qweakness, "weakness");
4742 DEFSYM (Qrehash_size, "rehash-size");
4743 DEFSYM (Qrehash_threshold, "rehash-threshold");
4744 }