Merge from emacs-24; up to 2013-01-03T01:56:56Z!rgm@gnu.org
[bpt/emacs.git] / src / lread.c
1 /* Lisp parsing and input streams.
2
3 Copyright (C) 1985-1989, 1993-1995, 1997-2013 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
22 #include <config.h>
23 #include "sysstdio.h"
24 #include <sys/types.h>
25 #include <sys/stat.h>
26 #include <sys/file.h>
27 #include <errno.h>
28 #include <limits.h> /* For CHAR_BIT. */
29 #include <stat-time.h>
30 #include "lisp.h"
31 #include "intervals.h"
32 #include "character.h"
33 #include "buffer.h"
34 #include "charset.h"
35 #include "coding.h"
36 #include <epaths.h>
37 #include "commands.h"
38 #include "keyboard.h"
39 #include "frame.h"
40 #include "termhooks.h"
41 #include "blockinput.h"
42
43 #ifdef MSDOS
44 #include "msdos.h"
45 #endif
46
47 #ifdef HAVE_NS
48 #include "nsterm.h"
49 #endif
50
51 #include <unistd.h>
52
53 #ifdef HAVE_SETLOCALE
54 #include <locale.h>
55 #endif /* HAVE_SETLOCALE */
56
57 #include <fcntl.h>
58
59 #ifdef HAVE_FSEEKO
60 #define file_offset off_t
61 #define file_tell ftello
62 #else
63 #define file_offset long
64 #define file_tell ftell
65 #endif
66
67 /* Hash table read constants. */
68 static Lisp_Object Qhash_table, Qdata;
69 static Lisp_Object Qtest, Qsize;
70 static Lisp_Object Qweakness;
71 static Lisp_Object Qrehash_size;
72 static Lisp_Object Qrehash_threshold;
73
74 static Lisp_Object Qread_char, Qget_file_char, Qcurrent_load_list;
75 Lisp_Object Qstandard_input;
76 Lisp_Object Qvariable_documentation;
77 static Lisp_Object Qascii_character, Qload, Qload_file_name;
78 Lisp_Object Qbackquote, Qcomma, Qcomma_at, Qcomma_dot, Qfunction;
79 static Lisp_Object Qinhibit_file_name_operation;
80 static Lisp_Object Qeval_buffer_list;
81 Lisp_Object Qlexical_binding;
82 static Lisp_Object Qfile_truename, Qdo_after_load_evaluation; /* ACM 2006/5/16 */
83
84 /* Used instead of Qget_file_char while loading *.elc files compiled
85 by Emacs 21 or older. */
86 static Lisp_Object Qget_emacs_mule_file_char;
87
88 static Lisp_Object Qload_force_doc_strings;
89
90 static Lisp_Object Qload_in_progress;
91
92 /* The association list of objects read with the #n=object form.
93 Each member of the list has the form (n . object), and is used to
94 look up the object for the corresponding #n# construct.
95 It must be set to nil before all top-level calls to read0. */
96 static Lisp_Object read_objects;
97
98 /* File for get_file_char to read from. Use by load. */
99 static FILE *instream;
100
101 /* For use within read-from-string (this reader is non-reentrant!!) */
102 static ptrdiff_t read_from_string_index;
103 static ptrdiff_t read_from_string_index_byte;
104 static ptrdiff_t read_from_string_limit;
105
106 /* Number of characters read in the current call to Fread or
107 Fread_from_string. */
108 static EMACS_INT readchar_count;
109
110 /* This contains the last string skipped with #@. */
111 static char *saved_doc_string;
112 /* Length of buffer allocated in saved_doc_string. */
113 static ptrdiff_t saved_doc_string_size;
114 /* Length of actual data in saved_doc_string. */
115 static ptrdiff_t saved_doc_string_length;
116 /* This is the file position that string came from. */
117 static file_offset saved_doc_string_position;
118
119 /* This contains the previous string skipped with #@.
120 We copy it from saved_doc_string when a new string
121 is put in saved_doc_string. */
122 static char *prev_saved_doc_string;
123 /* Length of buffer allocated in prev_saved_doc_string. */
124 static ptrdiff_t prev_saved_doc_string_size;
125 /* Length of actual data in prev_saved_doc_string. */
126 static ptrdiff_t prev_saved_doc_string_length;
127 /* This is the file position that string came from. */
128 static file_offset prev_saved_doc_string_position;
129
130 /* True means inside a new-style backquote
131 with no surrounding parentheses.
132 Fread initializes this to false, so we need not specbind it
133 or worry about what happens to it when there is an error. */
134 static bool new_backquote_flag;
135 static Lisp_Object Qold_style_backquotes;
136
137 /* A list of file names for files being loaded in Fload. Used to
138 check for recursive loads. */
139
140 static Lisp_Object Vloads_in_progress;
141
142 static int read_emacs_mule_char (int, int (*) (int, Lisp_Object),
143 Lisp_Object);
144
145 static void readevalloop (Lisp_Object, FILE *, Lisp_Object, bool,
146 Lisp_Object, Lisp_Object,
147 Lisp_Object, Lisp_Object);
148 \f
149 /* Functions that read one byte from the current source READCHARFUN
150 or unreads one byte. If the integer argument C is -1, it returns
151 one read byte, or -1 when there's no more byte in the source. If C
152 is 0 or positive, it unreads C, and the return value is not
153 interesting. */
154
155 static int readbyte_for_lambda (int, Lisp_Object);
156 static int readbyte_from_file (int, Lisp_Object);
157 static int readbyte_from_string (int, Lisp_Object);
158
159 /* Handle unreading and rereading of characters.
160 Write READCHAR to read a character,
161 UNREAD(c) to unread c to be read again.
162
163 These macros correctly read/unread multibyte characters. */
164
165 #define READCHAR readchar (readcharfun, NULL)
166 #define UNREAD(c) unreadchar (readcharfun, c)
167
168 /* Same as READCHAR but set *MULTIBYTE to the multibyteness of the source. */
169 #define READCHAR_REPORT_MULTIBYTE(multibyte) readchar (readcharfun, multibyte)
170
171 /* When READCHARFUN is Qget_file_char, Qget_emacs_mule_file_char,
172 Qlambda, or a cons, we use this to keep an unread character because
173 a file stream can't handle multibyte-char unreading. The value -1
174 means that there's no unread character. */
175 static int unread_char;
176
177 static int
178 readchar (Lisp_Object readcharfun, bool *multibyte)
179 {
180 Lisp_Object tem;
181 register int c;
182 int (*readbyte) (int, Lisp_Object);
183 unsigned char buf[MAX_MULTIBYTE_LENGTH];
184 int i, len;
185 bool emacs_mule_encoding = 0;
186
187 if (multibyte)
188 *multibyte = 0;
189
190 readchar_count++;
191
192 if (BUFFERP (readcharfun))
193 {
194 register struct buffer *inbuffer = XBUFFER (readcharfun);
195
196 ptrdiff_t pt_byte = BUF_PT_BYTE (inbuffer);
197
198 if (! BUFFER_LIVE_P (inbuffer))
199 return -1;
200
201 if (pt_byte >= BUF_ZV_BYTE (inbuffer))
202 return -1;
203
204 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
205 {
206 /* Fetch the character code from the buffer. */
207 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, pt_byte);
208 BUF_INC_POS (inbuffer, pt_byte);
209 c = STRING_CHAR (p);
210 if (multibyte)
211 *multibyte = 1;
212 }
213 else
214 {
215 c = BUF_FETCH_BYTE (inbuffer, pt_byte);
216 if (! ASCII_BYTE_P (c))
217 c = BYTE8_TO_CHAR (c);
218 pt_byte++;
219 }
220 SET_BUF_PT_BOTH (inbuffer, BUF_PT (inbuffer) + 1, pt_byte);
221
222 return c;
223 }
224 if (MARKERP (readcharfun))
225 {
226 register struct buffer *inbuffer = XMARKER (readcharfun)->buffer;
227
228 ptrdiff_t bytepos = marker_byte_position (readcharfun);
229
230 if (bytepos >= BUF_ZV_BYTE (inbuffer))
231 return -1;
232
233 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
234 {
235 /* Fetch the character code from the buffer. */
236 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, bytepos);
237 BUF_INC_POS (inbuffer, bytepos);
238 c = STRING_CHAR (p);
239 if (multibyte)
240 *multibyte = 1;
241 }
242 else
243 {
244 c = BUF_FETCH_BYTE (inbuffer, bytepos);
245 if (! ASCII_BYTE_P (c))
246 c = BYTE8_TO_CHAR (c);
247 bytepos++;
248 }
249
250 XMARKER (readcharfun)->bytepos = bytepos;
251 XMARKER (readcharfun)->charpos++;
252
253 return c;
254 }
255
256 if (EQ (readcharfun, Qlambda))
257 {
258 readbyte = readbyte_for_lambda;
259 goto read_multibyte;
260 }
261
262 if (EQ (readcharfun, Qget_file_char))
263 {
264 readbyte = readbyte_from_file;
265 goto read_multibyte;
266 }
267
268 if (STRINGP (readcharfun))
269 {
270 if (read_from_string_index >= read_from_string_limit)
271 c = -1;
272 else if (STRING_MULTIBYTE (readcharfun))
273 {
274 if (multibyte)
275 *multibyte = 1;
276 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c, readcharfun,
277 read_from_string_index,
278 read_from_string_index_byte);
279 }
280 else
281 {
282 c = SREF (readcharfun, read_from_string_index_byte);
283 read_from_string_index++;
284 read_from_string_index_byte++;
285 }
286 return c;
287 }
288
289 if (CONSP (readcharfun))
290 {
291 /* This is the case that read_vector is reading from a unibyte
292 string that contains a byte sequence previously skipped
293 because of #@NUMBER. The car part of readcharfun is that
294 string, and the cdr part is a value of readcharfun given to
295 read_vector. */
296 readbyte = readbyte_from_string;
297 if (EQ (XCDR (readcharfun), Qget_emacs_mule_file_char))
298 emacs_mule_encoding = 1;
299 goto read_multibyte;
300 }
301
302 if (EQ (readcharfun, Qget_emacs_mule_file_char))
303 {
304 readbyte = readbyte_from_file;
305 emacs_mule_encoding = 1;
306 goto read_multibyte;
307 }
308
309 tem = call0 (readcharfun);
310
311 if (NILP (tem))
312 return -1;
313 return XINT (tem);
314
315 read_multibyte:
316 if (unread_char >= 0)
317 {
318 c = unread_char;
319 unread_char = -1;
320 return c;
321 }
322 c = (*readbyte) (-1, readcharfun);
323 if (c < 0)
324 return c;
325 if (multibyte)
326 *multibyte = 1;
327 if (ASCII_BYTE_P (c))
328 return c;
329 if (emacs_mule_encoding)
330 return read_emacs_mule_char (c, readbyte, readcharfun);
331 i = 0;
332 buf[i++] = c;
333 len = BYTES_BY_CHAR_HEAD (c);
334 while (i < len)
335 {
336 c = (*readbyte) (-1, readcharfun);
337 if (c < 0 || ! TRAILING_CODE_P (c))
338 {
339 while (--i > 1)
340 (*readbyte) (buf[i], readcharfun);
341 return BYTE8_TO_CHAR (buf[0]);
342 }
343 buf[i++] = c;
344 }
345 return STRING_CHAR (buf);
346 }
347
348 #define FROM_FILE_P(readcharfun) \
349 (EQ (readcharfun, Qget_file_char) \
350 || EQ (readcharfun, Qget_emacs_mule_file_char))
351
352 static void
353 skip_dyn_bytes (Lisp_Object readcharfun, ptrdiff_t n)
354 {
355 if (FROM_FILE_P (readcharfun))
356 {
357 block_input (); /* FIXME: Not sure if it's needed. */
358 fseek (instream, n, SEEK_CUR);
359 unblock_input ();
360 }
361 else
362 { /* We're not reading directly from a file. In that case, it's difficult
363 to reliably count bytes, since these are usually meant for the file's
364 encoding, whereas we're now typically in the internal encoding.
365 But luckily, skip_dyn_bytes is used to skip over a single
366 dynamic-docstring (or dynamic byte-code) which is always quoted such
367 that \037 is the final char. */
368 int c;
369 do {
370 c = READCHAR;
371 } while (c >= 0 && c != '\037');
372 }
373 }
374
375 static void
376 skip_dyn_eof (Lisp_Object readcharfun)
377 {
378 if (FROM_FILE_P (readcharfun))
379 {
380 block_input (); /* FIXME: Not sure if it's needed. */
381 fseek (instream, 0, SEEK_END);
382 unblock_input ();
383 }
384 else
385 while (READCHAR >= 0);
386 }
387
388 /* Unread the character C in the way appropriate for the stream READCHARFUN.
389 If the stream is a user function, call it with the char as argument. */
390
391 static void
392 unreadchar (Lisp_Object readcharfun, int c)
393 {
394 readchar_count--;
395 if (c == -1)
396 /* Don't back up the pointer if we're unreading the end-of-input mark,
397 since readchar didn't advance it when we read it. */
398 ;
399 else if (BUFFERP (readcharfun))
400 {
401 struct buffer *b = XBUFFER (readcharfun);
402 ptrdiff_t charpos = BUF_PT (b);
403 ptrdiff_t bytepos = BUF_PT_BYTE (b);
404
405 if (! NILP (BVAR (b, enable_multibyte_characters)))
406 BUF_DEC_POS (b, bytepos);
407 else
408 bytepos--;
409
410 SET_BUF_PT_BOTH (b, charpos - 1, bytepos);
411 }
412 else if (MARKERP (readcharfun))
413 {
414 struct buffer *b = XMARKER (readcharfun)->buffer;
415 ptrdiff_t bytepos = XMARKER (readcharfun)->bytepos;
416
417 XMARKER (readcharfun)->charpos--;
418 if (! NILP (BVAR (b, enable_multibyte_characters)))
419 BUF_DEC_POS (b, bytepos);
420 else
421 bytepos--;
422
423 XMARKER (readcharfun)->bytepos = bytepos;
424 }
425 else if (STRINGP (readcharfun))
426 {
427 read_from_string_index--;
428 read_from_string_index_byte
429 = string_char_to_byte (readcharfun, read_from_string_index);
430 }
431 else if (CONSP (readcharfun))
432 {
433 unread_char = c;
434 }
435 else if (EQ (readcharfun, Qlambda))
436 {
437 unread_char = c;
438 }
439 else if (FROM_FILE_P (readcharfun))
440 {
441 unread_char = c;
442 }
443 else
444 call1 (readcharfun, make_number (c));
445 }
446
447 static int
448 readbyte_for_lambda (int c, Lisp_Object readcharfun)
449 {
450 return read_bytecode_char (c >= 0);
451 }
452
453
454 static int
455 readbyte_from_file (int c, Lisp_Object readcharfun)
456 {
457 if (c >= 0)
458 {
459 block_input ();
460 ungetc (c, instream);
461 unblock_input ();
462 return 0;
463 }
464
465 block_input ();
466 c = getc (instream);
467
468 /* Interrupted reads have been observed while reading over the network. */
469 while (c == EOF && ferror (instream) && errno == EINTR)
470 {
471 unblock_input ();
472 QUIT;
473 block_input ();
474 clearerr (instream);
475 c = getc (instream);
476 }
477
478 unblock_input ();
479
480 return (c == EOF ? -1 : c);
481 }
482
483 static int
484 readbyte_from_string (int c, Lisp_Object readcharfun)
485 {
486 Lisp_Object string = XCAR (readcharfun);
487
488 if (c >= 0)
489 {
490 read_from_string_index--;
491 read_from_string_index_byte
492 = string_char_to_byte (string, read_from_string_index);
493 }
494
495 if (read_from_string_index >= read_from_string_limit)
496 c = -1;
497 else
498 FETCH_STRING_CHAR_ADVANCE (c, string,
499 read_from_string_index,
500 read_from_string_index_byte);
501 return c;
502 }
503
504
505 /* Read one non-ASCII character from INSTREAM. The character is
506 encoded in `emacs-mule' and the first byte is already read in
507 C. */
508
509 static int
510 read_emacs_mule_char (int c, int (*readbyte) (int, Lisp_Object), Lisp_Object readcharfun)
511 {
512 /* Emacs-mule coding uses at most 4-byte for one character. */
513 unsigned char buf[4];
514 int len = emacs_mule_bytes[c];
515 struct charset *charset;
516 int i;
517 unsigned code;
518
519 if (len == 1)
520 /* C is not a valid leading-code of `emacs-mule'. */
521 return BYTE8_TO_CHAR (c);
522
523 i = 0;
524 buf[i++] = c;
525 while (i < len)
526 {
527 c = (*readbyte) (-1, readcharfun);
528 if (c < 0xA0)
529 {
530 while (--i > 1)
531 (*readbyte) (buf[i], readcharfun);
532 return BYTE8_TO_CHAR (buf[0]);
533 }
534 buf[i++] = c;
535 }
536
537 if (len == 2)
538 {
539 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
540 code = buf[1] & 0x7F;
541 }
542 else if (len == 3)
543 {
544 if (buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_11
545 || buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_12)
546 {
547 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
548 code = buf[2] & 0x7F;
549 }
550 else
551 {
552 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
553 code = ((buf[1] << 8) | buf[2]) & 0x7F7F;
554 }
555 }
556 else
557 {
558 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
559 code = ((buf[2] << 8) | buf[3]) & 0x7F7F;
560 }
561 c = DECODE_CHAR (charset, code);
562 if (c < 0)
563 Fsignal (Qinvalid_read_syntax,
564 list1 (build_string ("invalid multibyte form")));
565 return c;
566 }
567
568
569 static Lisp_Object read_internal_start (Lisp_Object, Lisp_Object,
570 Lisp_Object);
571 static Lisp_Object read0 (Lisp_Object);
572 static Lisp_Object read1 (Lisp_Object, int *, bool);
573
574 static Lisp_Object read_list (bool, Lisp_Object);
575 static Lisp_Object read_vector (Lisp_Object, bool);
576
577 static Lisp_Object substitute_object_recurse (Lisp_Object, Lisp_Object,
578 Lisp_Object);
579 static void substitute_object_in_subtree (Lisp_Object,
580 Lisp_Object);
581 static void substitute_in_interval (INTERVAL, Lisp_Object);
582
583 \f
584 /* Get a character from the tty. */
585
586 /* Read input events until we get one that's acceptable for our purposes.
587
588 If NO_SWITCH_FRAME, switch-frame events are stashed
589 until we get a character we like, and then stuffed into
590 unread_switch_frame.
591
592 If ASCII_REQUIRED, check function key events to see
593 if the unmodified version of the symbol has a Qascii_character
594 property, and use that character, if present.
595
596 If ERROR_NONASCII, signal an error if the input we
597 get isn't an ASCII character with modifiers. If it's false but
598 ASCII_REQUIRED is true, just re-read until we get an ASCII
599 character.
600
601 If INPUT_METHOD, invoke the current input method
602 if the character warrants that.
603
604 If SECONDS is a number, wait that many seconds for input, and
605 return Qnil if no input arrives within that time. */
606
607 static Lisp_Object
608 read_filtered_event (bool no_switch_frame, bool ascii_required,
609 bool error_nonascii, bool input_method, Lisp_Object seconds)
610 {
611 Lisp_Object val, delayed_switch_frame;
612 EMACS_TIME end_time;
613
614 #ifdef HAVE_WINDOW_SYSTEM
615 if (display_hourglass_p)
616 cancel_hourglass ();
617 #endif
618
619 delayed_switch_frame = Qnil;
620
621 /* Compute timeout. */
622 if (NUMBERP (seconds))
623 {
624 double duration = extract_float (seconds);
625 EMACS_TIME wait_time = EMACS_TIME_FROM_DOUBLE (duration);
626 end_time = add_emacs_time (current_emacs_time (), wait_time);
627 }
628
629 /* Read until we get an acceptable event. */
630 retry:
631 do
632 val = read_char (0, Qnil, (input_method ? Qnil : Qt), 0,
633 NUMBERP (seconds) ? &end_time : NULL);
634 while (INTEGERP (val) && XINT (val) == -2); /* wrong_kboard_jmpbuf */
635
636 if (BUFFERP (val))
637 goto retry;
638
639 /* `switch-frame' events are put off until after the next ASCII
640 character. This is better than signaling an error just because
641 the last characters were typed to a separate minibuffer frame,
642 for example. Eventually, some code which can deal with
643 switch-frame events will read it and process it. */
644 if (no_switch_frame
645 && EVENT_HAS_PARAMETERS (val)
646 && EQ (EVENT_HEAD_KIND (EVENT_HEAD (val)), Qswitch_frame))
647 {
648 delayed_switch_frame = val;
649 goto retry;
650 }
651
652 if (ascii_required && !(NUMBERP (seconds) && NILP (val)))
653 {
654 /* Convert certain symbols to their ASCII equivalents. */
655 if (SYMBOLP (val))
656 {
657 Lisp_Object tem, tem1;
658 tem = Fget (val, Qevent_symbol_element_mask);
659 if (!NILP (tem))
660 {
661 tem1 = Fget (Fcar (tem), Qascii_character);
662 /* Merge this symbol's modifier bits
663 with the ASCII equivalent of its basic code. */
664 if (!NILP (tem1))
665 XSETFASTINT (val, XINT (tem1) | XINT (Fcar (Fcdr (tem))));
666 }
667 }
668
669 /* If we don't have a character now, deal with it appropriately. */
670 if (!INTEGERP (val))
671 {
672 if (error_nonascii)
673 {
674 Vunread_command_events = list1 (val);
675 error ("Non-character input-event");
676 }
677 else
678 goto retry;
679 }
680 }
681
682 if (! NILP (delayed_switch_frame))
683 unread_switch_frame = delayed_switch_frame;
684
685 #if 0
686
687 #ifdef HAVE_WINDOW_SYSTEM
688 if (display_hourglass_p)
689 start_hourglass ();
690 #endif
691
692 #endif
693
694 return val;
695 }
696
697 DEFUN ("read-char", Fread_char, Sread_char, 0, 3, 0,
698 doc: /* Read a character from the command input (keyboard or macro).
699 It is returned as a number.
700 If the character has modifiers, they are resolved and reflected to the
701 character code if possible (e.g. C-SPC -> 0).
702
703 If the user generates an event which is not a character (i.e. a mouse
704 click or function key event), `read-char' signals an error. As an
705 exception, switch-frame events are put off until non-character events
706 can be read.
707 If you want to read non-character events, or ignore them, call
708 `read-event' or `read-char-exclusive' instead.
709
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 Lisp_Object val;
721
722 if (! NILP (prompt))
723 message_with_string ("%s", prompt, 0);
724 val = read_filtered_event (1, 1, 1, ! NILP (inherit_input_method), seconds);
725
726 return (NILP (val) ? Qnil
727 : make_number (char_resolve_modifier_mask (XINT (val))));
728 }
729
730 DEFUN ("read-event", Fread_event, Sread_event, 0, 3, 0,
731 doc: /* Read an event object from the input stream.
732 If the optional argument PROMPT is non-nil, display that as a prompt.
733 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
734 input method is turned on in the current buffer, that input method
735 is used for reading a character.
736 If the optional argument SECONDS is non-nil, it should be a number
737 specifying the maximum number of seconds to wait for input. If no
738 input arrives in that time, return nil. SECONDS may be a
739 floating-point value. */)
740 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
741 {
742 if (! NILP (prompt))
743 message_with_string ("%s", prompt, 0);
744 return read_filtered_event (0, 0, 0, ! NILP (inherit_input_method), seconds);
745 }
746
747 DEFUN ("read-char-exclusive", Fread_char_exclusive, Sread_char_exclusive, 0, 3, 0,
748 doc: /* Read a character from the command input (keyboard or macro).
749 It is returned as a number. Non-character events are ignored.
750 If the character has modifiers, they are resolved and reflected to the
751 character code if possible (e.g. C-SPC -> 0).
752
753 If the optional argument PROMPT is non-nil, display that as a prompt.
754 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
755 input method is turned on in the current buffer, that input method
756 is used for reading a character.
757 If the optional argument SECONDS is non-nil, it should be a number
758 specifying the maximum number of seconds to wait for input. If no
759 input arrives in that time, return nil. SECONDS may be a
760 floating-point value. */)
761 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
762 {
763 Lisp_Object val;
764
765 if (! NILP (prompt))
766 message_with_string ("%s", prompt, 0);
767
768 val = read_filtered_event (1, 1, 0, ! NILP (inherit_input_method), seconds);
769
770 return (NILP (val) ? Qnil
771 : make_number (char_resolve_modifier_mask (XINT (val))));
772 }
773
774 DEFUN ("get-file-char", Fget_file_char, Sget_file_char, 0, 0, 0,
775 doc: /* Don't use this yourself. */)
776 (void)
777 {
778 register Lisp_Object val;
779 block_input ();
780 XSETINT (val, getc (instream));
781 unblock_input ();
782 return val;
783 }
784
785
786 \f
787
788 /* Return true if the lisp code read using READCHARFUN defines a non-nil
789 `lexical-binding' file variable. After returning, the stream is
790 positioned following the first line, if it is a comment or #! line,
791 otherwise nothing is read. */
792
793 static bool
794 lisp_file_lexically_bound_p (Lisp_Object readcharfun)
795 {
796 int ch = READCHAR;
797
798 if (ch == '#')
799 {
800 ch = READCHAR;
801 if (ch != '!')
802 {
803 UNREAD (ch);
804 UNREAD ('#');
805 return 0;
806 }
807 while (ch != '\n' && ch != EOF)
808 ch = READCHAR;
809 if (ch == '\n') ch = READCHAR;
810 /* It is OK to leave the position after a #! line, since
811 that is what read1 does. */
812 }
813
814 if (ch != ';')
815 /* The first line isn't a comment, just give up. */
816 {
817 UNREAD (ch);
818 return 0;
819 }
820 else
821 /* Look for an appropriate file-variable in the first line. */
822 {
823 bool rv = 0;
824 enum {
825 NOMINAL, AFTER_FIRST_DASH, AFTER_ASTERIX
826 } beg_end_state = NOMINAL;
827 bool in_file_vars = 0;
828
829 #define UPDATE_BEG_END_STATE(ch) \
830 if (beg_end_state == NOMINAL) \
831 beg_end_state = (ch == '-' ? AFTER_FIRST_DASH : NOMINAL); \
832 else if (beg_end_state == AFTER_FIRST_DASH) \
833 beg_end_state = (ch == '*' ? AFTER_ASTERIX : NOMINAL); \
834 else if (beg_end_state == AFTER_ASTERIX) \
835 { \
836 if (ch == '-') \
837 in_file_vars = !in_file_vars; \
838 beg_end_state = NOMINAL; \
839 }
840
841 /* Skip until we get to the file vars, if any. */
842 do
843 {
844 ch = READCHAR;
845 UPDATE_BEG_END_STATE (ch);
846 }
847 while (!in_file_vars && ch != '\n' && ch != EOF);
848
849 while (in_file_vars)
850 {
851 char var[100], val[100];
852 unsigned i;
853
854 ch = READCHAR;
855
856 /* Read a variable name. */
857 while (ch == ' ' || ch == '\t')
858 ch = READCHAR;
859
860 i = 0;
861 while (ch != ':' && ch != '\n' && ch != EOF && in_file_vars)
862 {
863 if (i < sizeof var - 1)
864 var[i++] = ch;
865 UPDATE_BEG_END_STATE (ch);
866 ch = READCHAR;
867 }
868
869 /* Stop scanning if no colon was found before end marker. */
870 if (!in_file_vars || ch == '\n' || ch == EOF)
871 break;
872
873 while (i > 0 && (var[i - 1] == ' ' || var[i - 1] == '\t'))
874 i--;
875 var[i] = '\0';
876
877 if (ch == ':')
878 {
879 /* Read a variable value. */
880 ch = READCHAR;
881
882 while (ch == ' ' || ch == '\t')
883 ch = READCHAR;
884
885 i = 0;
886 while (ch != ';' && ch != '\n' && ch != EOF && in_file_vars)
887 {
888 if (i < sizeof val - 1)
889 val[i++] = ch;
890 UPDATE_BEG_END_STATE (ch);
891 ch = READCHAR;
892 }
893 if (! in_file_vars)
894 /* The value was terminated by an end-marker, which remove. */
895 i -= 3;
896 while (i > 0 && (val[i - 1] == ' ' || val[i - 1] == '\t'))
897 i--;
898 val[i] = '\0';
899
900 if (strcmp (var, "lexical-binding") == 0)
901 /* This is it... */
902 {
903 rv = (strcmp (val, "nil") != 0);
904 break;
905 }
906 }
907 }
908
909 while (ch != '\n' && ch != EOF)
910 ch = READCHAR;
911
912 return rv;
913 }
914 }
915 \f
916 /* Value is a version number of byte compiled code if the file
917 associated with file descriptor FD is a compiled Lisp file that's
918 safe to load. Only files compiled with Emacs are safe to load.
919 Files compiled with XEmacs can lead to a crash in Fbyte_code
920 because of an incompatible change in the byte compiler. */
921
922 static int
923 safe_to_load_version (int fd)
924 {
925 char buf[512];
926 int nbytes, i;
927 int version = 1;
928
929 /* Read the first few bytes from the file, and look for a line
930 specifying the byte compiler version used. */
931 nbytes = emacs_read (fd, buf, sizeof buf);
932 if (nbytes > 0)
933 {
934 /* Skip to the next newline, skipping over the initial `ELC'
935 with NUL bytes following it, but note the version. */
936 for (i = 0; i < nbytes && buf[i] != '\n'; ++i)
937 if (i == 4)
938 version = buf[i];
939
940 if (i >= nbytes
941 || fast_c_string_match_ignore_case (Vbytecomp_version_regexp,
942 buf + i, nbytes - i) < 0)
943 version = 0;
944 }
945
946 lseek (fd, 0, SEEK_SET);
947 return version;
948 }
949
950
951 /* Callback for record_unwind_protect. Restore the old load list OLD,
952 after loading a file successfully. */
953
954 static void
955 record_load_unwind (Lisp_Object old)
956 {
957 Vloads_in_progress = old;
958 }
959
960 /* This handler function is used via internal_condition_case_1. */
961
962 static Lisp_Object
963 load_error_handler (Lisp_Object data)
964 {
965 return Qnil;
966 }
967
968 static void
969 load_warn_old_style_backquotes (Lisp_Object file)
970 {
971 if (!NILP (Vold_style_backquotes))
972 {
973 Lisp_Object args[2];
974 args[0] = build_string ("Loading `%s': old-style backquotes detected!");
975 args[1] = file;
976 Fmessage (2, args);
977 }
978 }
979
980 DEFUN ("get-load-suffixes", Fget_load_suffixes, Sget_load_suffixes, 0, 0, 0,
981 doc: /* Return the suffixes that `load' should try if a suffix is \
982 required.
983 This uses the variables `load-suffixes' and `load-file-rep-suffixes'. */)
984 (void)
985 {
986 Lisp_Object lst = Qnil, suffixes = Vload_suffixes, suffix, ext;
987 while (CONSP (suffixes))
988 {
989 Lisp_Object exts = Vload_file_rep_suffixes;
990 suffix = XCAR (suffixes);
991 suffixes = XCDR (suffixes);
992 while (CONSP (exts))
993 {
994 ext = XCAR (exts);
995 exts = XCDR (exts);
996 lst = Fcons (concat2 (suffix, ext), lst);
997 }
998 }
999 return Fnreverse (lst);
1000 }
1001
1002 DEFUN ("load", Fload, Sload, 1, 5, 0,
1003 doc: /* Execute a file of Lisp code named FILE.
1004 First try FILE with `.elc' appended, then try with `.el',
1005 then try FILE unmodified (the exact suffixes in the exact order are
1006 determined by `load-suffixes'). Environment variable references in
1007 FILE are replaced with their values by calling `substitute-in-file-name'.
1008 This function searches the directories in `load-path'.
1009
1010 If optional second arg NOERROR is non-nil,
1011 report no error if FILE doesn't exist.
1012 Print messages at start and end of loading unless
1013 optional third arg NOMESSAGE is non-nil (but `force-load-messages'
1014 overrides that).
1015 If optional fourth arg NOSUFFIX is non-nil, don't try adding
1016 suffixes `.elc' or `.el' to the specified name FILE.
1017 If optional fifth arg MUST-SUFFIX is non-nil, insist on
1018 the suffix `.elc' or `.el'; don't accept just FILE unless
1019 it ends in one of those suffixes or includes a directory name.
1020
1021 If NOSUFFIX is nil, then if a file could not be found, try looking for
1022 a different representation of the file by adding non-empty suffixes to
1023 its name, before trying another file. Emacs uses this feature to find
1024 compressed versions of files when Auto Compression mode is enabled.
1025 If NOSUFFIX is non-nil, disable this feature.
1026
1027 The suffixes that this function tries out, when NOSUFFIX is nil, are
1028 given by the return value of `get-load-suffixes' and the values listed
1029 in `load-file-rep-suffixes'. If MUST-SUFFIX is non-nil, only the
1030 return value of `get-load-suffixes' is used, i.e. the file name is
1031 required to have a non-empty suffix.
1032
1033 Loading a file records its definitions, and its `provide' and
1034 `require' calls, in an element of `load-history' whose
1035 car is the file name loaded. See `load-history'.
1036
1037 While the file is in the process of being loaded, the variable
1038 `load-in-progress' is non-nil and the variable `load-file-name'
1039 is bound to the file's name.
1040
1041 Return t if the file exists and loads successfully. */)
1042 (Lisp_Object file, Lisp_Object noerror, Lisp_Object nomessage,
1043 Lisp_Object nosuffix, Lisp_Object must_suffix)
1044 {
1045 FILE *stream;
1046 int fd;
1047 int fd_index;
1048 ptrdiff_t count = SPECPDL_INDEX ();
1049 struct gcpro gcpro1, gcpro2, gcpro3;
1050 Lisp_Object found, efound, hist_file_name;
1051 /* True means we printed the ".el is newer" message. */
1052 bool newer = 0;
1053 /* True means we are loading a compiled file. */
1054 bool compiled = 0;
1055 Lisp_Object handler;
1056 bool safe_p = 1;
1057 const char *fmode = "r";
1058 int version;
1059
1060 #ifdef DOS_NT
1061 fmode = "rt";
1062 #endif /* DOS_NT */
1063
1064 CHECK_STRING (file);
1065
1066 /* If file name is magic, call the handler. */
1067 /* This shouldn't be necessary any more now that `openp' handles it right.
1068 handler = Ffind_file_name_handler (file, Qload);
1069 if (!NILP (handler))
1070 return call5 (handler, Qload, file, noerror, nomessage, nosuffix); */
1071
1072 /* Do this after the handler to avoid
1073 the need to gcpro noerror, nomessage and nosuffix.
1074 (Below here, we care only whether they are nil or not.)
1075 The presence of this call is the result of a historical accident:
1076 it used to be in every file-operation and when it got removed
1077 everywhere, it accidentally stayed here. Since then, enough people
1078 supposedly have things like (load "$PROJECT/foo.el") in their .emacs
1079 that it seemed risky to remove. */
1080 if (! NILP (noerror))
1081 {
1082 file = internal_condition_case_1 (Fsubstitute_in_file_name, file,
1083 Qt, load_error_handler);
1084 if (NILP (file))
1085 return Qnil;
1086 }
1087 else
1088 file = Fsubstitute_in_file_name (file);
1089
1090 /* Avoid weird lossage with null string as arg,
1091 since it would try to load a directory as a Lisp file. */
1092 if (SCHARS (file) == 0)
1093 {
1094 fd = -1;
1095 errno = ENOENT;
1096 }
1097 else
1098 {
1099 Lisp_Object suffixes;
1100 found = Qnil;
1101 GCPRO2 (file, found);
1102
1103 if (! NILP (must_suffix))
1104 {
1105 /* Don't insist on adding a suffix if FILE already ends with one. */
1106 ptrdiff_t size = SBYTES (file);
1107 if (size > 3
1108 && !strcmp (SSDATA (file) + size - 3, ".el"))
1109 must_suffix = Qnil;
1110 else if (size > 4
1111 && !strcmp (SSDATA (file) + size - 4, ".elc"))
1112 must_suffix = Qnil;
1113 /* Don't insist on adding a suffix
1114 if the argument includes a directory name. */
1115 else if (! NILP (Ffile_name_directory (file)))
1116 must_suffix = Qnil;
1117 }
1118
1119 if (!NILP (nosuffix))
1120 suffixes = Qnil;
1121 else
1122 {
1123 suffixes = Fget_load_suffixes ();
1124 if (NILP (must_suffix))
1125 {
1126 Lisp_Object arg[2];
1127 arg[0] = suffixes;
1128 arg[1] = Vload_file_rep_suffixes;
1129 suffixes = Fappend (2, arg);
1130 }
1131 }
1132
1133 fd = openp (Vload_path, file, suffixes, &found, Qnil);
1134 UNGCPRO;
1135 }
1136
1137 if (fd == -1)
1138 {
1139 if (NILP (noerror))
1140 report_file_error ("Cannot open load file", file);
1141 return Qnil;
1142 }
1143
1144 /* Tell startup.el whether or not we found the user's init file. */
1145 if (EQ (Qt, Vuser_init_file))
1146 Vuser_init_file = found;
1147
1148 /* If FD is -2, that means openp found a magic file. */
1149 if (fd == -2)
1150 {
1151 if (NILP (Fequal (found, file)))
1152 /* If FOUND is a different file name from FILE,
1153 find its handler even if we have already inhibited
1154 the `load' operation on FILE. */
1155 handler = Ffind_file_name_handler (found, Qt);
1156 else
1157 handler = Ffind_file_name_handler (found, Qload);
1158 if (! NILP (handler))
1159 return call5 (handler, Qload, found, noerror, nomessage, Qt);
1160 #ifdef DOS_NT
1161 /* Tramp has to deal with semi-broken packages that prepend
1162 drive letters to remote files. For that reason, Tramp
1163 catches file operations that test for file existence, which
1164 makes openp think X:/foo.elc files are remote. However,
1165 Tramp does not catch `load' operations for such files, so we
1166 end up with a nil as the `load' handler above. If we would
1167 continue with fd = -2, we will behave wrongly, and in
1168 particular try reading a .elc file in the "rt" mode instead
1169 of "rb". See bug #9311 for the results. To work around
1170 this, we try to open the file locally, and go with that if it
1171 succeeds. */
1172 fd = emacs_open (SSDATA (ENCODE_FILE (found)), O_RDONLY, 0);
1173 if (fd == -1)
1174 fd = -2;
1175 #endif
1176 }
1177
1178 if (fd < 0)
1179 {
1180 /* Pacify older GCC with --enable-gcc-warnings. */
1181 IF_LINT (fd_index = 0);
1182 }
1183 else
1184 {
1185 fd_index = SPECPDL_INDEX ();
1186 record_unwind_protect_int (close_file_unwind, fd);
1187 }
1188
1189 /* Check if we're stuck in a recursive load cycle.
1190
1191 2000-09-21: It's not possible to just check for the file loaded
1192 being a member of Vloads_in_progress. This fails because of the
1193 way the byte compiler currently works; `provide's are not
1194 evaluated, see font-lock.el/jit-lock.el as an example. This
1195 leads to a certain amount of ``normal'' recursion.
1196
1197 Also, just loading a file recursively is not always an error in
1198 the general case; the second load may do something different. */
1199 {
1200 int load_count = 0;
1201 Lisp_Object tem;
1202 for (tem = Vloads_in_progress; CONSP (tem); tem = XCDR (tem))
1203 if (!NILP (Fequal (found, XCAR (tem))) && (++load_count > 3))
1204 signal_error ("Recursive load", Fcons (found, Vloads_in_progress));
1205 record_unwind_protect (record_load_unwind, Vloads_in_progress);
1206 Vloads_in_progress = Fcons (found, Vloads_in_progress);
1207 }
1208
1209 /* All loads are by default dynamic, unless the file itself specifies
1210 otherwise using a file-variable in the first line. This is bound here
1211 so that it takes effect whether or not we use
1212 Vload_source_file_function. */
1213 specbind (Qlexical_binding, Qnil);
1214
1215 /* Get the name for load-history. */
1216 hist_file_name = (! NILP (Vpurify_flag)
1217 ? concat2 (Ffile_name_directory (file),
1218 Ffile_name_nondirectory (found))
1219 : found) ;
1220
1221 version = -1;
1222
1223 /* Check for the presence of old-style quotes and warn about them. */
1224 specbind (Qold_style_backquotes, Qnil);
1225 record_unwind_protect (load_warn_old_style_backquotes, file);
1226
1227 if (!memcmp (SDATA (found) + SBYTES (found) - 4, ".elc", 4)
1228 || (fd >= 0 && (version = safe_to_load_version (fd)) > 0))
1229 /* Load .elc files directly, but not when they are
1230 remote and have no handler! */
1231 {
1232 if (fd != -2)
1233 {
1234 struct stat s1, s2;
1235 int result;
1236
1237 GCPRO3 (file, found, hist_file_name);
1238
1239 if (version < 0
1240 && ! (version = safe_to_load_version (fd)))
1241 {
1242 safe_p = 0;
1243 if (!load_dangerous_libraries)
1244 error ("File `%s' was not compiled in Emacs", SDATA (found));
1245 else if (!NILP (nomessage) && !force_load_messages)
1246 message_with_string ("File `%s' not compiled in Emacs", found, 1);
1247 }
1248
1249 compiled = 1;
1250
1251 efound = ENCODE_FILE (found);
1252
1253 #ifdef DOS_NT
1254 fmode = "rb";
1255 #endif /* DOS_NT */
1256 result = stat (SSDATA (efound), &s1);
1257 if (result == 0)
1258 {
1259 SSET (efound, SBYTES (efound) - 1, 0);
1260 result = stat (SSDATA (efound), &s2);
1261 SSET (efound, SBYTES (efound) - 1, 'c');
1262 }
1263
1264 if (result == 0
1265 && EMACS_TIME_LT (get_stat_mtime (&s1), get_stat_mtime (&s2)))
1266 {
1267 /* Make the progress messages mention that source is newer. */
1268 newer = 1;
1269
1270 /* If we won't print another message, mention this anyway. */
1271 if (!NILP (nomessage) && !force_load_messages)
1272 {
1273 Lisp_Object msg_file;
1274 msg_file = Fsubstring (found, make_number (0), make_number (-1));
1275 message_with_string ("Source file `%s' newer than byte-compiled file",
1276 msg_file, 1);
1277 }
1278 }
1279 UNGCPRO;
1280 }
1281 }
1282 else
1283 {
1284 /* We are loading a source file (*.el). */
1285 if (!NILP (Vload_source_file_function))
1286 {
1287 Lisp_Object val;
1288
1289 if (fd >= 0)
1290 {
1291 emacs_close (fd);
1292 clear_unwind_protect (fd_index);
1293 }
1294 val = call4 (Vload_source_file_function, found, hist_file_name,
1295 NILP (noerror) ? Qnil : Qt,
1296 (NILP (nomessage) || force_load_messages) ? Qnil : Qt);
1297 return unbind_to (count, val);
1298 }
1299 }
1300
1301 GCPRO3 (file, found, hist_file_name);
1302
1303 if (fd < 0)
1304 {
1305 /* We somehow got here with fd == -2, meaning the file is deemed
1306 to be remote. Don't even try to reopen the file locally;
1307 just force a failure. */
1308 stream = NULL;
1309 errno = EINVAL;
1310 }
1311 else
1312 {
1313 #ifdef WINDOWSNT
1314 emacs_close (fd);
1315 clear_unwind_protect (fd_index);
1316 efound = ENCODE_FILE (found);
1317 stream = emacs_fopen (SSDATA (efound), fmode);
1318 #else
1319 stream = fdopen (fd, fmode);
1320 #endif
1321 }
1322 if (! stream)
1323 report_file_error ("Opening stdio stream", file);
1324 set_unwind_protect_ptr (fd_index, fclose_unwind, stream);
1325
1326 if (! NILP (Vpurify_flag))
1327 Vpreloaded_file_list = Fcons (Fpurecopy (file), Vpreloaded_file_list);
1328
1329 if (NILP (nomessage) || force_load_messages)
1330 {
1331 if (!safe_p)
1332 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...",
1333 file, 1);
1334 else if (!compiled)
1335 message_with_string ("Loading %s (source)...", file, 1);
1336 else if (newer)
1337 message_with_string ("Loading %s (compiled; note, source file is newer)...",
1338 file, 1);
1339 else /* The typical case; compiled file newer than source file. */
1340 message_with_string ("Loading %s...", file, 1);
1341 }
1342
1343 specbind (Qload_file_name, found);
1344 specbind (Qinhibit_file_name_operation, Qnil);
1345 specbind (Qload_in_progress, Qt);
1346
1347 instream = stream;
1348 if (lisp_file_lexically_bound_p (Qget_file_char))
1349 Fset (Qlexical_binding, Qt);
1350
1351 if (! version || version >= 22)
1352 readevalloop (Qget_file_char, stream, hist_file_name,
1353 0, Qnil, Qnil, Qnil, Qnil);
1354 else
1355 {
1356 /* We can't handle a file which was compiled with
1357 byte-compile-dynamic by older version of Emacs. */
1358 specbind (Qload_force_doc_strings, Qt);
1359 readevalloop (Qget_emacs_mule_file_char, stream, hist_file_name,
1360 0, Qnil, Qnil, Qnil, Qnil);
1361 }
1362 unbind_to (count, Qnil);
1363
1364 /* Run any eval-after-load forms for this file. */
1365 if (!NILP (Ffboundp (Qdo_after_load_evaluation)))
1366 call1 (Qdo_after_load_evaluation, hist_file_name) ;
1367
1368 UNGCPRO;
1369
1370 xfree (saved_doc_string);
1371 saved_doc_string = 0;
1372 saved_doc_string_size = 0;
1373
1374 xfree (prev_saved_doc_string);
1375 prev_saved_doc_string = 0;
1376 prev_saved_doc_string_size = 0;
1377
1378 if (!noninteractive && (NILP (nomessage) || force_load_messages))
1379 {
1380 if (!safe_p)
1381 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...done",
1382 file, 1);
1383 else if (!compiled)
1384 message_with_string ("Loading %s (source)...done", file, 1);
1385 else if (newer)
1386 message_with_string ("Loading %s (compiled; note, source file is newer)...done",
1387 file, 1);
1388 else /* The typical case; compiled file newer than source file. */
1389 message_with_string ("Loading %s...done", file, 1);
1390 }
1391
1392 return Qt;
1393 }
1394 \f
1395 static bool
1396 complete_filename_p (Lisp_Object pathname)
1397 {
1398 const unsigned char *s = SDATA (pathname);
1399 return (IS_DIRECTORY_SEP (s[0])
1400 || (SCHARS (pathname) > 2
1401 && IS_DEVICE_SEP (s[1]) && IS_DIRECTORY_SEP (s[2])));
1402 }
1403
1404 DEFUN ("locate-file-internal", Flocate_file_internal, Slocate_file_internal, 2, 4, 0,
1405 doc: /* Search for FILENAME through PATH.
1406 Returns the file's name in absolute form, or nil if not found.
1407 If SUFFIXES is non-nil, it should be a list of suffixes to append to
1408 file name when searching.
1409 If non-nil, PREDICATE is used instead of `file-readable-p'.
1410 PREDICATE can also be an integer to pass to the faccessat(2) function,
1411 in which case file-name-handlers are ignored.
1412 This function will normally skip directories, so if you want it to find
1413 directories, make sure the PREDICATE function returns `dir-ok' for them. */)
1414 (Lisp_Object filename, Lisp_Object path, Lisp_Object suffixes, Lisp_Object predicate)
1415 {
1416 Lisp_Object file;
1417 int fd = openp (path, filename, suffixes, &file, predicate);
1418 if (NILP (predicate) && fd >= 0)
1419 emacs_close (fd);
1420 return file;
1421 }
1422
1423 static Lisp_Object Qdir_ok;
1424
1425 /* Search for a file whose name is STR, looking in directories
1426 in the Lisp list PATH, and trying suffixes from SUFFIX.
1427 On success, return a file descriptor (or 1 or -2 as described below).
1428 On failure, return -1 and set errno.
1429
1430 SUFFIXES is a list of strings containing possible suffixes.
1431 The empty suffix is automatically added if the list is empty.
1432
1433 PREDICATE non-nil means don't open the files,
1434 just look for one that satisfies the predicate. In this case,
1435 return 1 on success. The predicate can be a lisp function or
1436 an integer to pass to `access' (in which case file-name-handlers
1437 are ignored).
1438
1439 If STOREPTR is nonzero, it points to a slot where the name of
1440 the file actually found should be stored as a Lisp string.
1441 nil is stored there on failure.
1442
1443 If the file we find is remote, return -2
1444 but store the found remote file name in *STOREPTR. */
1445
1446 int
1447 openp (Lisp_Object path, Lisp_Object str, Lisp_Object suffixes,
1448 Lisp_Object *storeptr, Lisp_Object predicate)
1449 {
1450 ptrdiff_t fn_size = 100;
1451 char buf[100];
1452 char *fn = buf;
1453 bool absolute = 0;
1454 ptrdiff_t want_length;
1455 Lisp_Object filename;
1456 struct gcpro gcpro1, gcpro2, gcpro3, gcpro4, gcpro5, gcpro6;
1457 Lisp_Object string, tail, encoded_fn;
1458 ptrdiff_t max_suffix_len = 0;
1459 int last_errno = ENOENT;
1460
1461 CHECK_STRING (str);
1462
1463 for (tail = suffixes; CONSP (tail); tail = XCDR (tail))
1464 {
1465 CHECK_STRING_CAR (tail);
1466 max_suffix_len = max (max_suffix_len,
1467 SBYTES (XCAR (tail)));
1468 }
1469
1470 string = filename = encoded_fn = Qnil;
1471 GCPRO6 (str, string, filename, path, suffixes, encoded_fn);
1472
1473 if (storeptr)
1474 *storeptr = Qnil;
1475
1476 if (complete_filename_p (str))
1477 absolute = 1;
1478
1479 for (; CONSP (path); path = XCDR (path))
1480 {
1481 filename = Fexpand_file_name (str, XCAR (path));
1482 if (!complete_filename_p (filename))
1483 /* If there are non-absolute elts in PATH (eg "."). */
1484 /* Of course, this could conceivably lose if luser sets
1485 default-directory to be something non-absolute... */
1486 {
1487 filename = Fexpand_file_name (filename, BVAR (current_buffer, directory));
1488 if (!complete_filename_p (filename))
1489 /* Give up on this path element! */
1490 continue;
1491 }
1492
1493 /* Calculate maximum length of any filename made from
1494 this path element/specified file name and any possible suffix. */
1495 want_length = max_suffix_len + SBYTES (filename);
1496 if (fn_size <= want_length)
1497 fn = alloca (fn_size = 100 + want_length);
1498
1499 /* Loop over suffixes. */
1500 for (tail = NILP (suffixes) ? list1 (empty_unibyte_string) : suffixes;
1501 CONSP (tail); tail = XCDR (tail))
1502 {
1503 ptrdiff_t fnlen, lsuffix = SBYTES (XCAR (tail));
1504 Lisp_Object handler;
1505
1506 /* Concatenate path element/specified name with the suffix.
1507 If the directory starts with /:, remove that. */
1508 int prefixlen = ((SCHARS (filename) > 2
1509 && SREF (filename, 0) == '/'
1510 && SREF (filename, 1) == ':')
1511 ? 2 : 0);
1512 fnlen = SBYTES (filename) - prefixlen;
1513 memcpy (fn, SDATA (filename) + prefixlen, fnlen);
1514 memcpy (fn + fnlen, SDATA (XCAR (tail)), lsuffix + 1);
1515 fnlen += lsuffix;
1516 /* Check that the file exists and is not a directory. */
1517 /* We used to only check for handlers on non-absolute file names:
1518 if (absolute)
1519 handler = Qnil;
1520 else
1521 handler = Ffind_file_name_handler (filename, Qfile_exists_p);
1522 It's not clear why that was the case and it breaks things like
1523 (load "/bar.el") where the file is actually "/bar.el.gz". */
1524 string = make_string (fn, fnlen);
1525 handler = Ffind_file_name_handler (string, Qfile_exists_p);
1526 if ((!NILP (handler) || !NILP (predicate)) && !NATNUMP (predicate))
1527 {
1528 bool exists;
1529 if (NILP (predicate))
1530 exists = !NILP (Ffile_readable_p (string));
1531 else
1532 {
1533 Lisp_Object tmp = call1 (predicate, string);
1534 if (NILP (tmp))
1535 exists = 0;
1536 else if (EQ (tmp, Qdir_ok)
1537 || NILP (Ffile_directory_p (string)))
1538 exists = 1;
1539 else
1540 {
1541 exists = 0;
1542 last_errno = EISDIR;
1543 }
1544 }
1545
1546 if (exists)
1547 {
1548 /* We succeeded; return this descriptor and filename. */
1549 if (storeptr)
1550 *storeptr = string;
1551 UNGCPRO;
1552 return -2;
1553 }
1554 }
1555 else
1556 {
1557 int fd;
1558 const char *pfn;
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 fd = emacs_open (pfn, O_RDONLY, 0);
1582 if (fd < 0)
1583 {
1584 if (errno != ENOENT)
1585 last_errno = errno;
1586 }
1587 else
1588 {
1589 struct stat st;
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 /* We succeeded; return this descriptor and filename. */
1604 if (storeptr)
1605 *storeptr = string;
1606 UNGCPRO;
1607 return fd;
1608 }
1609 }
1610 }
1611 if (absolute)
1612 break;
1613 }
1614
1615 UNGCPRO;
1616 errno = last_errno;
1617 return -1;
1618 }
1619
1620 \f
1621 /* Merge the list we've accumulated of globals from the current input source
1622 into the load_history variable. The details depend on whether
1623 the source has an associated file name or not.
1624
1625 FILENAME is the file name that we are loading from.
1626
1627 ENTIRE is true if loading that entire file, false if evaluating
1628 part of it. */
1629
1630 static void
1631 build_load_history (Lisp_Object filename, bool entire)
1632 {
1633 Lisp_Object tail, prev, newelt;
1634 Lisp_Object tem, tem2;
1635 bool foundit = 0;
1636
1637 tail = Vload_history;
1638 prev = Qnil;
1639
1640 while (CONSP (tail))
1641 {
1642 tem = XCAR (tail);
1643
1644 /* Find the feature's previous assoc list... */
1645 if (!NILP (Fequal (filename, Fcar (tem))))
1646 {
1647 foundit = 1;
1648
1649 /* If we're loading the entire file, remove old data. */
1650 if (entire)
1651 {
1652 if (NILP (prev))
1653 Vload_history = XCDR (tail);
1654 else
1655 Fsetcdr (prev, XCDR (tail));
1656 }
1657
1658 /* Otherwise, cons on new symbols that are not already members. */
1659 else
1660 {
1661 tem2 = Vcurrent_load_list;
1662
1663 while (CONSP (tem2))
1664 {
1665 newelt = XCAR (tem2);
1666
1667 if (NILP (Fmember (newelt, tem)))
1668 Fsetcar (tail, Fcons (XCAR (tem),
1669 Fcons (newelt, XCDR (tem))));
1670
1671 tem2 = XCDR (tem2);
1672 QUIT;
1673 }
1674 }
1675 }
1676 else
1677 prev = tail;
1678 tail = XCDR (tail);
1679 QUIT;
1680 }
1681
1682 /* If we're loading an entire file, cons the new assoc onto the
1683 front of load-history, the most-recently-loaded position. Also
1684 do this if we didn't find an existing member for the file. */
1685 if (entire || !foundit)
1686 Vload_history = Fcons (Fnreverse (Vcurrent_load_list),
1687 Vload_history);
1688 }
1689
1690 static void
1691 readevalloop_1 (int old)
1692 {
1693 load_convert_to_unibyte = old;
1694 }
1695
1696 /* Signal an `end-of-file' error, if possible with file name
1697 information. */
1698
1699 static _Noreturn void
1700 end_of_file_error (void)
1701 {
1702 if (STRINGP (Vload_file_name))
1703 xsignal1 (Qend_of_file, Vload_file_name);
1704
1705 xsignal0 (Qend_of_file);
1706 }
1707
1708 /* UNIBYTE specifies how to set load_convert_to_unibyte
1709 for this invocation.
1710 READFUN, if non-nil, is used instead of `read'.
1711
1712 START, END specify region to read in current buffer (from eval-region).
1713 If the input is not from a buffer, they must be nil. */
1714
1715 static void
1716 readevalloop (Lisp_Object readcharfun,
1717 FILE *stream,
1718 Lisp_Object sourcename,
1719 bool printflag,
1720 Lisp_Object unibyte, Lisp_Object readfun,
1721 Lisp_Object start, Lisp_Object end)
1722 {
1723 register int c;
1724 register Lisp_Object val;
1725 ptrdiff_t count = SPECPDL_INDEX ();
1726 struct gcpro gcpro1, gcpro2, gcpro3, gcpro4;
1727 struct buffer *b = 0;
1728 bool continue_reading_p;
1729 Lisp_Object lex_bound;
1730 /* True if reading an entire buffer. */
1731 bool whole_buffer = 0;
1732 /* True on the first time around. */
1733 bool first_sexp = 1;
1734 Lisp_Object macroexpand = intern ("internal-macroexpand-for-load");
1735
1736 if (NILP (Ffboundp (macroexpand))
1737 /* Don't macroexpand in .elc files, since it should have been done
1738 already. We actually don't know whether we're in a .elc file or not,
1739 so we use circumstantial evidence: .el files normally go through
1740 Vload_source_file_function -> load-with-code-conversion
1741 -> eval-buffer. */
1742 || EQ (readcharfun, Qget_file_char)
1743 || EQ (readcharfun, Qget_emacs_mule_file_char))
1744 macroexpand = Qnil;
1745
1746 if (MARKERP (readcharfun))
1747 {
1748 if (NILP (start))
1749 start = readcharfun;
1750 }
1751
1752 if (BUFFERP (readcharfun))
1753 b = XBUFFER (readcharfun);
1754 else if (MARKERP (readcharfun))
1755 b = XMARKER (readcharfun)->buffer;
1756
1757 /* We assume START is nil when input is not from a buffer. */
1758 if (! NILP (start) && !b)
1759 emacs_abort ();
1760
1761 specbind (Qstandard_input, readcharfun); /* GCPROs readcharfun. */
1762 specbind (Qcurrent_load_list, Qnil);
1763 record_unwind_protect_int (readevalloop_1, load_convert_to_unibyte);
1764 load_convert_to_unibyte = !NILP (unibyte);
1765
1766 /* If lexical binding is active (either because it was specified in
1767 the file's header, or via a buffer-local variable), create an empty
1768 lexical environment, otherwise, turn off lexical binding. */
1769 lex_bound = find_symbol_value (Qlexical_binding);
1770 specbind (Qinternal_interpreter_environment,
1771 (NILP (lex_bound) || EQ (lex_bound, Qunbound)
1772 ? Qnil : list1 (Qt)));
1773
1774 GCPRO4 (sourcename, readfun, start, end);
1775
1776 /* Try to ensure sourcename is a truename, except whilst preloading. */
1777 if (NILP (Vpurify_flag)
1778 && !NILP (sourcename) && !NILP (Ffile_name_absolute_p (sourcename))
1779 && !NILP (Ffboundp (Qfile_truename)))
1780 sourcename = call1 (Qfile_truename, sourcename) ;
1781
1782 LOADHIST_ATTACH (sourcename);
1783
1784 continue_reading_p = 1;
1785 while (continue_reading_p)
1786 {
1787 ptrdiff_t count1 = SPECPDL_INDEX ();
1788
1789 if (b != 0 && !BUFFER_LIVE_P (b))
1790 error ("Reading from killed buffer");
1791
1792 if (!NILP (start))
1793 {
1794 /* Switch to the buffer we are reading from. */
1795 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1796 set_buffer_internal (b);
1797
1798 /* Save point in it. */
1799 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1800 /* Save ZV in it. */
1801 record_unwind_protect (save_restriction_restore, save_restriction_save ());
1802 /* Those get unbound after we read one expression. */
1803
1804 /* Set point and ZV around stuff to be read. */
1805 Fgoto_char (start);
1806 if (!NILP (end))
1807 Fnarrow_to_region (make_number (BEGV), end);
1808
1809 /* Just for cleanliness, convert END to a marker
1810 if it is an integer. */
1811 if (INTEGERP (end))
1812 end = Fpoint_max_marker ();
1813 }
1814
1815 /* On the first cycle, we can easily test here
1816 whether we are reading the whole buffer. */
1817 if (b && first_sexp)
1818 whole_buffer = (PT == BEG && ZV == Z);
1819
1820 instream = stream;
1821 read_next:
1822 c = READCHAR;
1823 if (c == ';')
1824 {
1825 while ((c = READCHAR) != '\n' && c != -1);
1826 goto read_next;
1827 }
1828 if (c < 0)
1829 {
1830 unbind_to (count1, Qnil);
1831 break;
1832 }
1833
1834 /* Ignore whitespace here, so we can detect eof. */
1835 if (c == ' ' || c == '\t' || c == '\n' || c == '\f' || c == '\r'
1836 || c == 0xa0) /* NBSP */
1837 goto read_next;
1838
1839 if (!NILP (Vpurify_flag) && c == '(')
1840 {
1841 val = read_list (0, readcharfun);
1842 }
1843 else
1844 {
1845 UNREAD (c);
1846 read_objects = Qnil;
1847 if (!NILP (readfun))
1848 {
1849 val = call1 (readfun, readcharfun);
1850
1851 /* If READCHARFUN has set point to ZV, we should
1852 stop reading, even if the form read sets point
1853 to a different value when evaluated. */
1854 if (BUFFERP (readcharfun))
1855 {
1856 struct buffer *buf = XBUFFER (readcharfun);
1857 if (BUF_PT (buf) == BUF_ZV (buf))
1858 continue_reading_p = 0;
1859 }
1860 }
1861 else if (! NILP (Vload_read_function))
1862 val = call1 (Vload_read_function, readcharfun);
1863 else
1864 val = read_internal_start (readcharfun, Qnil, Qnil);
1865 }
1866
1867 if (!NILP (start) && continue_reading_p)
1868 start = Fpoint_marker ();
1869
1870 /* Restore saved point and BEGV. */
1871 unbind_to (count1, Qnil);
1872
1873 /* Now eval what we just read. */
1874 if (!NILP (macroexpand))
1875 val = call1 (macroexpand, val);
1876 val = eval_sub (val);
1877
1878 if (printflag)
1879 {
1880 Vvalues = Fcons (val, Vvalues);
1881 if (EQ (Vstandard_output, Qt))
1882 Fprin1 (val, Qnil);
1883 else
1884 Fprint (val, Qnil);
1885 }
1886
1887 first_sexp = 0;
1888 }
1889
1890 build_load_history (sourcename,
1891 stream || whole_buffer);
1892
1893 UNGCPRO;
1894
1895 unbind_to (count, Qnil);
1896 }
1897
1898 DEFUN ("eval-buffer", Feval_buffer, Seval_buffer, 0, 5, "",
1899 doc: /* Execute the current buffer as Lisp code.
1900 When called from a Lisp program (i.e., not interactively), this
1901 function accepts up to five optional arguments:
1902 BUFFER is the buffer to evaluate (nil means use current buffer).
1903 PRINTFLAG controls printing of output:
1904 A value of nil means discard it; anything else is stream for print.
1905 FILENAME specifies the file name to use for `load-history'.
1906 UNIBYTE, if non-nil, specifies `load-convert-to-unibyte' for this
1907 invocation.
1908 DO-ALLOW-PRINT, if non-nil, specifies that `print' and related
1909 functions should work normally even if PRINTFLAG is nil.
1910
1911 This function preserves the position of point. */)
1912 (Lisp_Object buffer, Lisp_Object printflag, Lisp_Object filename, Lisp_Object unibyte, Lisp_Object do_allow_print)
1913 {
1914 ptrdiff_t count = SPECPDL_INDEX ();
1915 Lisp_Object tem, buf;
1916
1917 if (NILP (buffer))
1918 buf = Fcurrent_buffer ();
1919 else
1920 buf = Fget_buffer (buffer);
1921 if (NILP (buf))
1922 error ("No such buffer");
1923
1924 if (NILP (printflag) && NILP (do_allow_print))
1925 tem = Qsymbolp;
1926 else
1927 tem = printflag;
1928
1929 if (NILP (filename))
1930 filename = BVAR (XBUFFER (buf), filename);
1931
1932 specbind (Qeval_buffer_list, Fcons (buf, Veval_buffer_list));
1933 specbind (Qstandard_output, tem);
1934 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1935 BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
1936 specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
1937 readevalloop (buf, 0, filename,
1938 !NILP (printflag), unibyte, Qnil, Qnil, Qnil);
1939 unbind_to (count, Qnil);
1940
1941 return Qnil;
1942 }
1943
1944 DEFUN ("eval-region", Feval_region, Seval_region, 2, 4, "r",
1945 doc: /* Execute the region as Lisp code.
1946 When called from programs, expects two arguments,
1947 giving starting and ending indices in the current buffer
1948 of the text to be executed.
1949 Programs can pass third argument PRINTFLAG which controls output:
1950 A value of nil means discard it; anything else is stream for printing it.
1951 Also the fourth argument READ-FUNCTION, if non-nil, is used
1952 instead of `read' to read each expression. It gets one argument
1953 which is the input stream for reading characters.
1954
1955 This function does not move point. */)
1956 (Lisp_Object start, Lisp_Object end, Lisp_Object printflag, Lisp_Object read_function)
1957 {
1958 /* FIXME: Do the eval-sexp-add-defvars dance! */
1959 ptrdiff_t count = SPECPDL_INDEX ();
1960 Lisp_Object tem, cbuf;
1961
1962 cbuf = Fcurrent_buffer ();
1963
1964 if (NILP (printflag))
1965 tem = Qsymbolp;
1966 else
1967 tem = printflag;
1968 specbind (Qstandard_output, tem);
1969 specbind (Qeval_buffer_list, Fcons (cbuf, Veval_buffer_list));
1970
1971 /* `readevalloop' calls functions which check the type of start and end. */
1972 readevalloop (cbuf, 0, BVAR (XBUFFER (cbuf), filename),
1973 !NILP (printflag), Qnil, read_function,
1974 start, end);
1975
1976 return unbind_to (count, Qnil);
1977 }
1978
1979 \f
1980 DEFUN ("read", Fread, Sread, 0, 1, 0,
1981 doc: /* Read one Lisp expression as text from STREAM, return as Lisp object.
1982 If STREAM is nil, use the value of `standard-input' (which see).
1983 STREAM or the value of `standard-input' may be:
1984 a buffer (read from point and advance it)
1985 a marker (read from where it points and advance it)
1986 a function (call it with no arguments for each character,
1987 call it with a char as argument to push a char back)
1988 a string (takes text from string, starting at the beginning)
1989 t (read text line using minibuffer and use it, or read from
1990 standard input in batch mode). */)
1991 (Lisp_Object stream)
1992 {
1993 if (NILP (stream))
1994 stream = Vstandard_input;
1995 if (EQ (stream, Qt))
1996 stream = Qread_char;
1997 if (EQ (stream, Qread_char))
1998 /* FIXME: ¿¡ When is this used !? */
1999 return call1 (intern ("read-minibuffer"),
2000 build_string ("Lisp expression: "));
2001
2002 return read_internal_start (stream, Qnil, Qnil);
2003 }
2004
2005 DEFUN ("read-from-string", Fread_from_string, Sread_from_string, 1, 3, 0,
2006 doc: /* Read one Lisp expression which is represented as text by STRING.
2007 Returns a cons: (OBJECT-READ . FINAL-STRING-INDEX).
2008 FINAL-STRING-INDEX is an integer giving the position of the next
2009 remaining character in STRING.
2010 START and END optionally delimit a substring of STRING from which to read;
2011 they default to 0 and (length STRING) respectively. */)
2012 (Lisp_Object string, Lisp_Object start, Lisp_Object end)
2013 {
2014 Lisp_Object ret;
2015 CHECK_STRING (string);
2016 /* `read_internal_start' sets `read_from_string_index'. */
2017 ret = read_internal_start (string, start, end);
2018 return Fcons (ret, make_number (read_from_string_index));
2019 }
2020
2021 /* Function to set up the global context we need in toplevel read
2022 calls. */
2023 static Lisp_Object
2024 read_internal_start (Lisp_Object stream, Lisp_Object start, Lisp_Object end)
2025 /* `start', `end' only used when stream is a string. */
2026 {
2027 Lisp_Object retval;
2028
2029 readchar_count = 0;
2030 new_backquote_flag = 0;
2031 read_objects = Qnil;
2032 if (EQ (Vread_with_symbol_positions, Qt)
2033 || EQ (Vread_with_symbol_positions, stream))
2034 Vread_symbol_positions_list = Qnil;
2035
2036 if (STRINGP (stream)
2037 || ((CONSP (stream) && STRINGP (XCAR (stream)))))
2038 {
2039 ptrdiff_t startval, endval;
2040 Lisp_Object string;
2041
2042 if (STRINGP (stream))
2043 string = stream;
2044 else
2045 string = XCAR (stream);
2046
2047 if (NILP (end))
2048 endval = SCHARS (string);
2049 else
2050 {
2051 CHECK_NUMBER (end);
2052 if (! (0 <= XINT (end) && XINT (end) <= SCHARS (string)))
2053 args_out_of_range (string, end);
2054 endval = XINT (end);
2055 }
2056
2057 if (NILP (start))
2058 startval = 0;
2059 else
2060 {
2061 CHECK_NUMBER (start);
2062 if (! (0 <= XINT (start) && XINT (start) <= endval))
2063 args_out_of_range (string, start);
2064 startval = XINT (start);
2065 }
2066 read_from_string_index = startval;
2067 read_from_string_index_byte = string_char_to_byte (string, startval);
2068 read_from_string_limit = endval;
2069 }
2070
2071 retval = read0 (stream);
2072 if (EQ (Vread_with_symbol_positions, Qt)
2073 || EQ (Vread_with_symbol_positions, stream))
2074 Vread_symbol_positions_list = Fnreverse (Vread_symbol_positions_list);
2075 return retval;
2076 }
2077 \f
2078
2079 /* Signal Qinvalid_read_syntax error.
2080 S is error string of length N (if > 0) */
2081
2082 static _Noreturn void
2083 invalid_syntax (const char *s)
2084 {
2085 xsignal1 (Qinvalid_read_syntax, build_string (s));
2086 }
2087
2088
2089 /* Use this for recursive reads, in contexts where internal tokens
2090 are not allowed. */
2091
2092 static Lisp_Object
2093 read0 (Lisp_Object readcharfun)
2094 {
2095 register Lisp_Object val;
2096 int c;
2097
2098 val = read1 (readcharfun, &c, 0);
2099 if (!c)
2100 return val;
2101
2102 xsignal1 (Qinvalid_read_syntax,
2103 Fmake_string (make_number (1), make_number (c)));
2104 }
2105 \f
2106 static ptrdiff_t read_buffer_size;
2107 static char *read_buffer;
2108
2109 /* Read a \-escape sequence, assuming we already read the `\'.
2110 If the escape sequence forces unibyte, return eight-bit char. */
2111
2112 static int
2113 read_escape (Lisp_Object readcharfun, bool stringp)
2114 {
2115 int c = READCHAR;
2116 /* \u allows up to four hex digits, \U up to eight. Default to the
2117 behavior for \u, and change this value in the case that \U is seen. */
2118 int unicode_hex_count = 4;
2119
2120 switch (c)
2121 {
2122 case -1:
2123 end_of_file_error ();
2124
2125 case 'a':
2126 return '\007';
2127 case 'b':
2128 return '\b';
2129 case 'd':
2130 return 0177;
2131 case 'e':
2132 return 033;
2133 case 'f':
2134 return '\f';
2135 case 'n':
2136 return '\n';
2137 case 'r':
2138 return '\r';
2139 case 't':
2140 return '\t';
2141 case 'v':
2142 return '\v';
2143 case '\n':
2144 return -1;
2145 case ' ':
2146 if (stringp)
2147 return -1;
2148 return ' ';
2149
2150 case 'M':
2151 c = READCHAR;
2152 if (c != '-')
2153 error ("Invalid escape character syntax");
2154 c = READCHAR;
2155 if (c == '\\')
2156 c = read_escape (readcharfun, 0);
2157 return c | meta_modifier;
2158
2159 case 'S':
2160 c = READCHAR;
2161 if (c != '-')
2162 error ("Invalid escape character syntax");
2163 c = READCHAR;
2164 if (c == '\\')
2165 c = read_escape (readcharfun, 0);
2166 return c | shift_modifier;
2167
2168 case 'H':
2169 c = READCHAR;
2170 if (c != '-')
2171 error ("Invalid escape character syntax");
2172 c = READCHAR;
2173 if (c == '\\')
2174 c = read_escape (readcharfun, 0);
2175 return c | hyper_modifier;
2176
2177 case 'A':
2178 c = READCHAR;
2179 if (c != '-')
2180 error ("Invalid escape character syntax");
2181 c = READCHAR;
2182 if (c == '\\')
2183 c = read_escape (readcharfun, 0);
2184 return c | alt_modifier;
2185
2186 case 's':
2187 c = READCHAR;
2188 if (stringp || c != '-')
2189 {
2190 UNREAD (c);
2191 return ' ';
2192 }
2193 c = READCHAR;
2194 if (c == '\\')
2195 c = read_escape (readcharfun, 0);
2196 return c | super_modifier;
2197
2198 case 'C':
2199 c = READCHAR;
2200 if (c != '-')
2201 error ("Invalid escape character syntax");
2202 case '^':
2203 c = READCHAR;
2204 if (c == '\\')
2205 c = read_escape (readcharfun, 0);
2206 if ((c & ~CHAR_MODIFIER_MASK) == '?')
2207 return 0177 | (c & CHAR_MODIFIER_MASK);
2208 else if (! SINGLE_BYTE_CHAR_P ((c & ~CHAR_MODIFIER_MASK)))
2209 return c | ctrl_modifier;
2210 /* ASCII control chars are made from letters (both cases),
2211 as well as the non-letters within 0100...0137. */
2212 else if ((c & 0137) >= 0101 && (c & 0137) <= 0132)
2213 return (c & (037 | ~0177));
2214 else if ((c & 0177) >= 0100 && (c & 0177) <= 0137)
2215 return (c & (037 | ~0177));
2216 else
2217 return c | ctrl_modifier;
2218
2219 case '0':
2220 case '1':
2221 case '2':
2222 case '3':
2223 case '4':
2224 case '5':
2225 case '6':
2226 case '7':
2227 /* An octal escape, as in ANSI C. */
2228 {
2229 register int i = c - '0';
2230 register int count = 0;
2231 while (++count < 3)
2232 {
2233 if ((c = READCHAR) >= '0' && c <= '7')
2234 {
2235 i *= 8;
2236 i += c - '0';
2237 }
2238 else
2239 {
2240 UNREAD (c);
2241 break;
2242 }
2243 }
2244
2245 if (i >= 0x80 && i < 0x100)
2246 i = BYTE8_TO_CHAR (i);
2247 return i;
2248 }
2249
2250 case 'x':
2251 /* A hex escape, as in ANSI C. */
2252 {
2253 unsigned int i = 0;
2254 int count = 0;
2255 while (1)
2256 {
2257 c = READCHAR;
2258 if (c >= '0' && c <= '9')
2259 {
2260 i *= 16;
2261 i += c - '0';
2262 }
2263 else if ((c >= 'a' && c <= 'f')
2264 || (c >= 'A' && c <= 'F'))
2265 {
2266 i *= 16;
2267 if (c >= 'a' && c <= 'f')
2268 i += c - 'a' + 10;
2269 else
2270 i += c - 'A' + 10;
2271 }
2272 else
2273 {
2274 UNREAD (c);
2275 break;
2276 }
2277 /* Allow hex escapes as large as ?\xfffffff, because some
2278 packages use them to denote characters with modifiers. */
2279 if ((CHAR_META | (CHAR_META - 1)) < i)
2280 error ("Hex character out of range: \\x%x...", i);
2281 count += count < 3;
2282 }
2283
2284 if (count < 3 && i >= 0x80)
2285 return BYTE8_TO_CHAR (i);
2286 return i;
2287 }
2288
2289 case 'U':
2290 /* Post-Unicode-2.0: Up to eight hex chars. */
2291 unicode_hex_count = 8;
2292 case 'u':
2293
2294 /* A Unicode escape. We only permit them in strings and characters,
2295 not arbitrarily in the source code, as in some other languages. */
2296 {
2297 unsigned int i = 0;
2298 int count = 0;
2299
2300 while (++count <= unicode_hex_count)
2301 {
2302 c = READCHAR;
2303 /* `isdigit' and `isalpha' may be locale-specific, which we don't
2304 want. */
2305 if (c >= '0' && c <= '9') i = (i << 4) + (c - '0');
2306 else if (c >= 'a' && c <= 'f') i = (i << 4) + (c - 'a') + 10;
2307 else if (c >= 'A' && c <= 'F') i = (i << 4) + (c - 'A') + 10;
2308 else
2309 error ("Non-hex digit used for Unicode escape");
2310 }
2311 if (i > 0x10FFFF)
2312 error ("Non-Unicode character: 0x%x", i);
2313 return i;
2314 }
2315
2316 default:
2317 return c;
2318 }
2319 }
2320
2321 /* Return the digit that CHARACTER stands for in the given BASE.
2322 Return -1 if CHARACTER is out of range for BASE,
2323 and -2 if CHARACTER is not valid for any supported BASE. */
2324 static int
2325 digit_to_number (int character, int base)
2326 {
2327 int digit;
2328
2329 if ('0' <= character && character <= '9')
2330 digit = character - '0';
2331 else if ('a' <= character && character <= 'z')
2332 digit = character - 'a' + 10;
2333 else if ('A' <= character && character <= 'Z')
2334 digit = character - 'A' + 10;
2335 else
2336 return -2;
2337
2338 return digit < base ? digit : -1;
2339 }
2340
2341 /* Read an integer in radix RADIX using READCHARFUN to read
2342 characters. RADIX must be in the interval [2..36]; if it isn't, a
2343 read error is signaled . Value is the integer read. Signals an
2344 error if encountering invalid read syntax or if RADIX is out of
2345 range. */
2346
2347 static Lisp_Object
2348 read_integer (Lisp_Object readcharfun, EMACS_INT radix)
2349 {
2350 /* Room for sign, leading 0, other digits, trailing null byte.
2351 Also, room for invalid syntax diagnostic. */
2352 char buf[max (1 + 1 + sizeof (uintmax_t) * CHAR_BIT + 1,
2353 sizeof "integer, radix " + INT_STRLEN_BOUND (EMACS_INT))];
2354
2355 int valid = -1; /* 1 if valid, 0 if not, -1 if incomplete. */
2356
2357 if (radix < 2 || radix > 36)
2358 valid = 0;
2359 else
2360 {
2361 char *p = buf;
2362 int c, digit;
2363
2364 c = READCHAR;
2365 if (c == '-' || c == '+')
2366 {
2367 *p++ = c;
2368 c = READCHAR;
2369 }
2370
2371 if (c == '0')
2372 {
2373 *p++ = c;
2374 valid = 1;
2375
2376 /* Ignore redundant leading zeros, so the buffer doesn't
2377 fill up with them. */
2378 do
2379 c = READCHAR;
2380 while (c == '0');
2381 }
2382
2383 while ((digit = digit_to_number (c, radix)) >= -1)
2384 {
2385 if (digit == -1)
2386 valid = 0;
2387 if (valid < 0)
2388 valid = 1;
2389
2390 if (p < buf + sizeof buf - 1)
2391 *p++ = c;
2392 else
2393 valid = 0;
2394
2395 c = READCHAR;
2396 }
2397
2398 UNREAD (c);
2399 *p = '\0';
2400 }
2401
2402 if (! valid)
2403 {
2404 sprintf (buf, "integer, radix %"pI"d", radix);
2405 invalid_syntax (buf);
2406 }
2407
2408 return string_to_number (buf, radix, 0);
2409 }
2410
2411
2412 /* If the next token is ')' or ']' or '.', we store that character
2413 in *PCH and the return value is not interesting. Else, we store
2414 zero in *PCH and we read and return one lisp object.
2415
2416 FIRST_IN_LIST is true if this is the first element of a list. */
2417
2418 static Lisp_Object
2419 read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
2420 {
2421 int c;
2422 bool uninterned_symbol = 0;
2423 bool multibyte;
2424
2425 *pch = 0;
2426
2427 retry:
2428
2429 c = READCHAR_REPORT_MULTIBYTE (&multibyte);
2430 if (c < 0)
2431 end_of_file_error ();
2432
2433 switch (c)
2434 {
2435 case '(':
2436 return read_list (0, readcharfun);
2437
2438 case '[':
2439 return read_vector (readcharfun, 0);
2440
2441 case ')':
2442 case ']':
2443 {
2444 *pch = c;
2445 return Qnil;
2446 }
2447
2448 case '#':
2449 c = READCHAR;
2450 if (c == 's')
2451 {
2452 c = READCHAR;
2453 if (c == '(')
2454 {
2455 /* Accept extended format for hashtables (extensible to
2456 other types), e.g.
2457 #s(hash-table size 2 test equal data (k1 v1 k2 v2)) */
2458 Lisp_Object tmp = read_list (0, readcharfun);
2459 Lisp_Object head = CAR_SAFE (tmp);
2460 Lisp_Object data = Qnil;
2461 Lisp_Object val = Qnil;
2462 /* The size is 2 * number of allowed keywords to
2463 make-hash-table. */
2464 Lisp_Object params[10];
2465 Lisp_Object ht;
2466 Lisp_Object key = Qnil;
2467 int param_count = 0;
2468
2469 if (!EQ (head, Qhash_table))
2470 error ("Invalid extended read marker at head of #s list "
2471 "(only hash-table allowed)");
2472
2473 tmp = CDR_SAFE (tmp);
2474
2475 /* This is repetitive but fast and simple. */
2476 params[param_count] = QCsize;
2477 params[param_count + 1] = Fplist_get (tmp, Qsize);
2478 if (!NILP (params[param_count + 1]))
2479 param_count += 2;
2480
2481 params[param_count] = QCtest;
2482 params[param_count + 1] = Fplist_get (tmp, Qtest);
2483 if (!NILP (params[param_count + 1]))
2484 param_count += 2;
2485
2486 params[param_count] = QCweakness;
2487 params[param_count + 1] = Fplist_get (tmp, Qweakness);
2488 if (!NILP (params[param_count + 1]))
2489 param_count += 2;
2490
2491 params[param_count] = QCrehash_size;
2492 params[param_count + 1] = Fplist_get (tmp, Qrehash_size);
2493 if (!NILP (params[param_count + 1]))
2494 param_count += 2;
2495
2496 params[param_count] = QCrehash_threshold;
2497 params[param_count + 1] = Fplist_get (tmp, Qrehash_threshold);
2498 if (!NILP (params[param_count + 1]))
2499 param_count += 2;
2500
2501 /* This is the hashtable data. */
2502 data = Fplist_get (tmp, Qdata);
2503
2504 /* Now use params to make a new hashtable and fill it. */
2505 ht = Fmake_hash_table (param_count, params);
2506
2507 while (CONSP (data))
2508 {
2509 key = XCAR (data);
2510 data = XCDR (data);
2511 if (!CONSP (data))
2512 error ("Odd number of elements in hashtable data");
2513 val = XCAR (data);
2514 data = XCDR (data);
2515 Fputhash (key, val, ht);
2516 }
2517
2518 return ht;
2519 }
2520 UNREAD (c);
2521 invalid_syntax ("#");
2522 }
2523 if (c == '^')
2524 {
2525 c = READCHAR;
2526 if (c == '[')
2527 {
2528 Lisp_Object tmp;
2529 tmp = read_vector (readcharfun, 0);
2530 if (ASIZE (tmp) < CHAR_TABLE_STANDARD_SLOTS)
2531 error ("Invalid size char-table");
2532 XSETPVECTYPE (XVECTOR (tmp), PVEC_CHAR_TABLE);
2533 return tmp;
2534 }
2535 else if (c == '^')
2536 {
2537 c = READCHAR;
2538 if (c == '[')
2539 {
2540 Lisp_Object tmp;
2541 int depth;
2542 ptrdiff_t size;
2543
2544 tmp = read_vector (readcharfun, 0);
2545 size = ASIZE (tmp);
2546 if (size == 0)
2547 error ("Invalid size char-table");
2548 if (! RANGED_INTEGERP (1, AREF (tmp, 0), 3))
2549 error ("Invalid depth in char-table");
2550 depth = XINT (AREF (tmp, 0));
2551 if (chartab_size[depth] != size - 2)
2552 error ("Invalid size char-table");
2553 XSETPVECTYPE (XVECTOR (tmp), PVEC_SUB_CHAR_TABLE);
2554 return tmp;
2555 }
2556 invalid_syntax ("#^^");
2557 }
2558 invalid_syntax ("#^");
2559 }
2560 if (c == '&')
2561 {
2562 Lisp_Object length;
2563 length = read1 (readcharfun, pch, first_in_list);
2564 c = READCHAR;
2565 if (c == '"')
2566 {
2567 Lisp_Object tmp, val;
2568 EMACS_INT size_in_chars
2569 = ((XFASTINT (length) + BOOL_VECTOR_BITS_PER_CHAR - 1)
2570 / BOOL_VECTOR_BITS_PER_CHAR);
2571
2572 UNREAD (c);
2573 tmp = read1 (readcharfun, pch, first_in_list);
2574 if (STRING_MULTIBYTE (tmp)
2575 || (size_in_chars != SCHARS (tmp)
2576 /* We used to print 1 char too many
2577 when the number of bits was a multiple of 8.
2578 Accept such input in case it came from an old
2579 version. */
2580 && ! (XFASTINT (length)
2581 == (SCHARS (tmp) - 1) * BOOL_VECTOR_BITS_PER_CHAR)))
2582 invalid_syntax ("#&...");
2583
2584 val = Fmake_bool_vector (length, Qnil);
2585 memcpy (XBOOL_VECTOR (val)->data, SDATA (tmp), size_in_chars);
2586 /* Clear the extraneous bits in the last byte. */
2587 if (XINT (length) != size_in_chars * BOOL_VECTOR_BITS_PER_CHAR)
2588 XBOOL_VECTOR (val)->data[size_in_chars - 1]
2589 &= (1 << (XINT (length) % BOOL_VECTOR_BITS_PER_CHAR)) - 1;
2590 return val;
2591 }
2592 invalid_syntax ("#&...");
2593 }
2594 if (c == '[')
2595 {
2596 /* Accept compiled functions at read-time so that we don't have to
2597 build them using function calls. */
2598 Lisp_Object tmp;
2599 tmp = read_vector (readcharfun, 1);
2600 make_byte_code (XVECTOR (tmp));
2601 return tmp;
2602 }
2603 if (c == '(')
2604 {
2605 Lisp_Object tmp;
2606 struct gcpro gcpro1;
2607 int ch;
2608
2609 /* Read the string itself. */
2610 tmp = read1 (readcharfun, &ch, 0);
2611 if (ch != 0 || !STRINGP (tmp))
2612 invalid_syntax ("#");
2613 GCPRO1 (tmp);
2614 /* Read the intervals and their properties. */
2615 while (1)
2616 {
2617 Lisp_Object beg, end, plist;
2618
2619 beg = read1 (readcharfun, &ch, 0);
2620 end = plist = Qnil;
2621 if (ch == ')')
2622 break;
2623 if (ch == 0)
2624 end = read1 (readcharfun, &ch, 0);
2625 if (ch == 0)
2626 plist = read1 (readcharfun, &ch, 0);
2627 if (ch)
2628 invalid_syntax ("Invalid string property list");
2629 Fset_text_properties (beg, end, plist, tmp);
2630 }
2631 UNGCPRO;
2632 return tmp;
2633 }
2634
2635 /* #@NUMBER is used to skip NUMBER following bytes.
2636 That's used in .elc files to skip over doc strings
2637 and function definitions. */
2638 if (c == '@')
2639 {
2640 enum { extra = 100 };
2641 ptrdiff_t i, nskip = 0, digits = 0;
2642
2643 /* Read a decimal integer. */
2644 while ((c = READCHAR) >= 0
2645 && c >= '0' && c <= '9')
2646 {
2647 if ((STRING_BYTES_BOUND - extra) / 10 <= nskip)
2648 string_overflow ();
2649 digits++;
2650 nskip *= 10;
2651 nskip += c - '0';
2652 if (digits == 2 && nskip == 0)
2653 { /* We've just seen #@00, which means "skip to end". */
2654 skip_dyn_eof (readcharfun);
2655 return Qnil;
2656 }
2657 }
2658 if (nskip > 0)
2659 /* We can't use UNREAD here, because in the code below we side-step
2660 READCHAR. Instead, assume the first char after #@NNN occupies
2661 a single byte, which is the case normally since it's just
2662 a space. */
2663 nskip--;
2664 else
2665 UNREAD (c);
2666
2667 if (load_force_doc_strings
2668 && (FROM_FILE_P (readcharfun)))
2669 {
2670 /* If we are supposed to force doc strings into core right now,
2671 record the last string that we skipped,
2672 and record where in the file it comes from. */
2673
2674 /* But first exchange saved_doc_string
2675 with prev_saved_doc_string, so we save two strings. */
2676 {
2677 char *temp = saved_doc_string;
2678 ptrdiff_t temp_size = saved_doc_string_size;
2679 file_offset temp_pos = saved_doc_string_position;
2680 ptrdiff_t temp_len = saved_doc_string_length;
2681
2682 saved_doc_string = prev_saved_doc_string;
2683 saved_doc_string_size = prev_saved_doc_string_size;
2684 saved_doc_string_position = prev_saved_doc_string_position;
2685 saved_doc_string_length = prev_saved_doc_string_length;
2686
2687 prev_saved_doc_string = temp;
2688 prev_saved_doc_string_size = temp_size;
2689 prev_saved_doc_string_position = temp_pos;
2690 prev_saved_doc_string_length = temp_len;
2691 }
2692
2693 if (saved_doc_string_size == 0)
2694 {
2695 saved_doc_string = xmalloc (nskip + extra);
2696 saved_doc_string_size = nskip + extra;
2697 }
2698 if (nskip > saved_doc_string_size)
2699 {
2700 saved_doc_string = xrealloc (saved_doc_string, nskip + extra);
2701 saved_doc_string_size = nskip + extra;
2702 }
2703
2704 saved_doc_string_position = file_tell (instream);
2705
2706 /* Copy that many characters into saved_doc_string. */
2707 block_input ();
2708 for (i = 0; i < nskip && c >= 0; i++)
2709 saved_doc_string[i] = c = getc (instream);
2710 unblock_input ();
2711
2712 saved_doc_string_length = i;
2713 }
2714 else
2715 /* Skip that many bytes. */
2716 skip_dyn_bytes (readcharfun, nskip);
2717
2718 goto retry;
2719 }
2720 if (c == '!')
2721 {
2722 /* #! appears at the beginning of an executable file.
2723 Skip the first line. */
2724 while (c != '\n' && c >= 0)
2725 c = READCHAR;
2726 goto retry;
2727 }
2728 if (c == '$')
2729 return Vload_file_name;
2730 if (c == '\'')
2731 return list2 (Qfunction, read0 (readcharfun));
2732 /* #:foo is the uninterned symbol named foo. */
2733 if (c == ':')
2734 {
2735 uninterned_symbol = 1;
2736 c = READCHAR;
2737 if (!(c > 040
2738 && c != 0xa0 /* NBSP */
2739 && (c >= 0200
2740 || strchr ("\"';()[]#`,", c) == NULL)))
2741 {
2742 /* No symbol character follows, this is the empty
2743 symbol. */
2744 UNREAD (c);
2745 return Fmake_symbol (empty_unibyte_string);
2746 }
2747 goto read_symbol;
2748 }
2749 /* ## is the empty symbol. */
2750 if (c == '#')
2751 return Fintern (empty_unibyte_string, Qnil);
2752 /* Reader forms that can reuse previously read objects. */
2753 if (c >= '0' && c <= '9')
2754 {
2755 EMACS_INT n = 0;
2756 Lisp_Object tem;
2757
2758 /* Read a non-negative integer. */
2759 while (c >= '0' && c <= '9')
2760 {
2761 if (MOST_POSITIVE_FIXNUM / 10 < n
2762 || MOST_POSITIVE_FIXNUM < n * 10 + c - '0')
2763 n = MOST_POSITIVE_FIXNUM + 1;
2764 else
2765 n = n * 10 + c - '0';
2766 c = READCHAR;
2767 }
2768
2769 if (n <= MOST_POSITIVE_FIXNUM)
2770 {
2771 if (c == 'r' || c == 'R')
2772 return read_integer (readcharfun, n);
2773
2774 if (! NILP (Vread_circle))
2775 {
2776 /* #n=object returns object, but associates it with
2777 n for #n#. */
2778 if (c == '=')
2779 {
2780 /* Make a placeholder for #n# to use temporarily. */
2781 Lisp_Object placeholder;
2782 Lisp_Object cell;
2783
2784 placeholder = Fcons (Qnil, Qnil);
2785 cell = Fcons (make_number (n), placeholder);
2786 read_objects = Fcons (cell, read_objects);
2787
2788 /* Read the object itself. */
2789 tem = read0 (readcharfun);
2790
2791 /* Now put it everywhere the placeholder was... */
2792 substitute_object_in_subtree (tem, placeholder);
2793
2794 /* ...and #n# will use the real value from now on. */
2795 Fsetcdr (cell, tem);
2796
2797 return tem;
2798 }
2799
2800 /* #n# returns a previously read object. */
2801 if (c == '#')
2802 {
2803 tem = Fassq (make_number (n), read_objects);
2804 if (CONSP (tem))
2805 return XCDR (tem);
2806 }
2807 }
2808 }
2809 /* Fall through to error message. */
2810 }
2811 else if (c == 'x' || c == 'X')
2812 return read_integer (readcharfun, 16);
2813 else if (c == 'o' || c == 'O')
2814 return read_integer (readcharfun, 8);
2815 else if (c == 'b' || c == 'B')
2816 return read_integer (readcharfun, 2);
2817
2818 UNREAD (c);
2819 invalid_syntax ("#");
2820
2821 case ';':
2822 while ((c = READCHAR) >= 0 && c != '\n');
2823 goto retry;
2824
2825 case '\'':
2826 return list2 (Qquote, read0 (readcharfun));
2827
2828 case '`':
2829 {
2830 int next_char = READCHAR;
2831 UNREAD (next_char);
2832 /* Transition from old-style to new-style:
2833 If we see "(`" it used to mean old-style, which usually works
2834 fine because ` should almost never appear in such a position
2835 for new-style. But occasionally we need "(`" to mean new
2836 style, so we try to distinguish the two by the fact that we
2837 can either write "( `foo" or "(` foo", where the first
2838 intends to use new-style whereas the second intends to use
2839 old-style. For Emacs-25, we should completely remove this
2840 first_in_list exception (old-style can still be obtained via
2841 "(\`" anyway). */
2842 if (!new_backquote_flag && first_in_list && next_char == ' ')
2843 {
2844 Vold_style_backquotes = Qt;
2845 goto default_label;
2846 }
2847 else
2848 {
2849 Lisp_Object value;
2850 bool saved_new_backquote_flag = new_backquote_flag;
2851
2852 new_backquote_flag = 1;
2853 value = read0 (readcharfun);
2854 new_backquote_flag = saved_new_backquote_flag;
2855
2856 return list2 (Qbackquote, value);
2857 }
2858 }
2859 case ',':
2860 {
2861 int next_char = READCHAR;
2862 UNREAD (next_char);
2863 /* Transition from old-style to new-style:
2864 It used to be impossible to have a new-style , other than within
2865 a new-style `. This is sufficient when ` and , are used in the
2866 normal way, but ` and , can also appear in args to macros that
2867 will not interpret them in the usual way, in which case , may be
2868 used without any ` anywhere near.
2869 So we now use the same heuristic as for backquote: old-style
2870 unquotes are only recognized when first on a list, and when
2871 followed by a space.
2872 Because it's more difficult to peek 2 chars ahead, a new-style
2873 ,@ can still not be used outside of a `, unless it's in the middle
2874 of a list. */
2875 if (new_backquote_flag
2876 || !first_in_list
2877 || (next_char != ' ' && next_char != '@'))
2878 {
2879 Lisp_Object comma_type = Qnil;
2880 Lisp_Object value;
2881 int ch = READCHAR;
2882
2883 if (ch == '@')
2884 comma_type = Qcomma_at;
2885 else if (ch == '.')
2886 comma_type = Qcomma_dot;
2887 else
2888 {
2889 if (ch >= 0) UNREAD (ch);
2890 comma_type = Qcomma;
2891 }
2892
2893 value = read0 (readcharfun);
2894 return list2 (comma_type, value);
2895 }
2896 else
2897 {
2898 Vold_style_backquotes = Qt;
2899 goto default_label;
2900 }
2901 }
2902 case '?':
2903 {
2904 int modifiers;
2905 int next_char;
2906 bool ok;
2907
2908 c = READCHAR;
2909 if (c < 0)
2910 end_of_file_error ();
2911
2912 /* Accept `single space' syntax like (list ? x) where the
2913 whitespace character is SPC or TAB.
2914 Other literal whitespace like NL, CR, and FF are not accepted,
2915 as there are well-established escape sequences for these. */
2916 if (c == ' ' || c == '\t')
2917 return make_number (c);
2918
2919 if (c == '\\')
2920 c = read_escape (readcharfun, 0);
2921 modifiers = c & CHAR_MODIFIER_MASK;
2922 c &= ~CHAR_MODIFIER_MASK;
2923 if (CHAR_BYTE8_P (c))
2924 c = CHAR_TO_BYTE8 (c);
2925 c |= modifiers;
2926
2927 next_char = READCHAR;
2928 ok = (next_char <= 040
2929 || (next_char < 0200
2930 && strchr ("\"';()[]#?`,.", next_char) != NULL));
2931 UNREAD (next_char);
2932 if (ok)
2933 return make_number (c);
2934
2935 invalid_syntax ("?");
2936 }
2937
2938 case '"':
2939 {
2940 char *p = read_buffer;
2941 char *end = read_buffer + read_buffer_size;
2942 int ch;
2943 /* True if we saw an escape sequence specifying
2944 a multibyte character. */
2945 bool force_multibyte = 0;
2946 /* True if we saw an escape sequence specifying
2947 a single-byte character. */
2948 bool force_singlebyte = 0;
2949 bool cancel = 0;
2950 ptrdiff_t nchars = 0;
2951
2952 while ((ch = READCHAR) >= 0
2953 && ch != '\"')
2954 {
2955 if (end - p < MAX_MULTIBYTE_LENGTH)
2956 {
2957 ptrdiff_t offset = p - read_buffer;
2958 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
2959 memory_full (SIZE_MAX);
2960 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
2961 read_buffer_size *= 2;
2962 p = read_buffer + offset;
2963 end = read_buffer + read_buffer_size;
2964 }
2965
2966 if (ch == '\\')
2967 {
2968 int modifiers;
2969
2970 ch = read_escape (readcharfun, 1);
2971
2972 /* CH is -1 if \ newline has just been seen. */
2973 if (ch == -1)
2974 {
2975 if (p == read_buffer)
2976 cancel = 1;
2977 continue;
2978 }
2979
2980 modifiers = ch & CHAR_MODIFIER_MASK;
2981 ch = ch & ~CHAR_MODIFIER_MASK;
2982
2983 if (CHAR_BYTE8_P (ch))
2984 force_singlebyte = 1;
2985 else if (! ASCII_CHAR_P (ch))
2986 force_multibyte = 1;
2987 else /* I.e. ASCII_CHAR_P (ch). */
2988 {
2989 /* Allow `\C- ' and `\C-?'. */
2990 if (modifiers == CHAR_CTL)
2991 {
2992 if (ch == ' ')
2993 ch = 0, modifiers = 0;
2994 else if (ch == '?')
2995 ch = 127, modifiers = 0;
2996 }
2997 if (modifiers & CHAR_SHIFT)
2998 {
2999 /* Shift modifier is valid only with [A-Za-z]. */
3000 if (ch >= 'A' && ch <= 'Z')
3001 modifiers &= ~CHAR_SHIFT;
3002 else if (ch >= 'a' && ch <= 'z')
3003 ch -= ('a' - 'A'), modifiers &= ~CHAR_SHIFT;
3004 }
3005
3006 if (modifiers & CHAR_META)
3007 {
3008 /* Move the meta bit to the right place for a
3009 string. */
3010 modifiers &= ~CHAR_META;
3011 ch = BYTE8_TO_CHAR (ch | 0x80);
3012 force_singlebyte = 1;
3013 }
3014 }
3015
3016 /* Any modifiers remaining are invalid. */
3017 if (modifiers)
3018 error ("Invalid modifier in string");
3019 p += CHAR_STRING (ch, (unsigned char *) p);
3020 }
3021 else
3022 {
3023 p += CHAR_STRING (ch, (unsigned char *) p);
3024 if (CHAR_BYTE8_P (ch))
3025 force_singlebyte = 1;
3026 else if (! ASCII_CHAR_P (ch))
3027 force_multibyte = 1;
3028 }
3029 nchars++;
3030 }
3031
3032 if (ch < 0)
3033 end_of_file_error ();
3034
3035 /* If purifying, and string starts with \ newline,
3036 return zero instead. This is for doc strings
3037 that we are really going to find in etc/DOC.nn.nn. */
3038 if (!NILP (Vpurify_flag) && NILP (Vdoc_file_name) && cancel)
3039 return make_number (0);
3040
3041 if (! force_multibyte && force_singlebyte)
3042 {
3043 /* READ_BUFFER contains raw 8-bit bytes and no multibyte
3044 forms. Convert it to unibyte. */
3045 nchars = str_as_unibyte ((unsigned char *) read_buffer,
3046 p - read_buffer);
3047 p = read_buffer + nchars;
3048 }
3049
3050 return make_specified_string (read_buffer, nchars, p - read_buffer,
3051 (force_multibyte
3052 || (p - read_buffer != nchars)));
3053 }
3054
3055 case '.':
3056 {
3057 int next_char = READCHAR;
3058 UNREAD (next_char);
3059
3060 if (next_char <= 040
3061 || (next_char < 0200
3062 && strchr ("\"';([#?`,", next_char) != NULL))
3063 {
3064 *pch = c;
3065 return Qnil;
3066 }
3067
3068 /* Otherwise, we fall through! Note that the atom-reading loop
3069 below will now loop at least once, assuring that we will not
3070 try to UNREAD two characters in a row. */
3071 }
3072 default:
3073 default_label:
3074 if (c <= 040) goto retry;
3075 if (c == 0xa0) /* NBSP */
3076 goto retry;
3077
3078 read_symbol:
3079 {
3080 char *p = read_buffer;
3081 bool quoted = 0;
3082 EMACS_INT start_position = readchar_count - 1;
3083
3084 {
3085 char *end = read_buffer + read_buffer_size;
3086
3087 do
3088 {
3089 if (end - p < MAX_MULTIBYTE_LENGTH)
3090 {
3091 ptrdiff_t offset = p - read_buffer;
3092 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3093 memory_full (SIZE_MAX);
3094 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3095 read_buffer_size *= 2;
3096 p = read_buffer + offset;
3097 end = read_buffer + read_buffer_size;
3098 }
3099
3100 if (c == '\\')
3101 {
3102 c = READCHAR;
3103 if (c == -1)
3104 end_of_file_error ();
3105 quoted = 1;
3106 }
3107
3108 if (multibyte)
3109 p += CHAR_STRING (c, (unsigned char *) p);
3110 else
3111 *p++ = c;
3112 c = READCHAR;
3113 }
3114 while (c > 040
3115 && c != 0xa0 /* NBSP */
3116 && (c >= 0200
3117 || strchr ("\"';()[]#`,", c) == NULL));
3118
3119 if (p == end)
3120 {
3121 ptrdiff_t offset = p - read_buffer;
3122 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3123 memory_full (SIZE_MAX);
3124 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3125 read_buffer_size *= 2;
3126 p = read_buffer + offset;
3127 end = read_buffer + read_buffer_size;
3128 }
3129 *p = 0;
3130 UNREAD (c);
3131 }
3132
3133 if (!quoted && !uninterned_symbol)
3134 {
3135 Lisp_Object result = string_to_number (read_buffer, 10, 0);
3136 if (! NILP (result))
3137 return result;
3138 }
3139 {
3140 Lisp_Object name, result;
3141 ptrdiff_t nbytes = p - read_buffer;
3142 ptrdiff_t nchars
3143 = (multibyte
3144 ? multibyte_chars_in_text ((unsigned char *) read_buffer,
3145 nbytes)
3146 : nbytes);
3147
3148 name = ((uninterned_symbol && ! NILP (Vpurify_flag)
3149 ? make_pure_string : make_specified_string)
3150 (read_buffer, nchars, nbytes, multibyte));
3151 result = (uninterned_symbol ? Fmake_symbol (name)
3152 : Fintern (name, Qnil));
3153
3154 if (EQ (Vread_with_symbol_positions, Qt)
3155 || EQ (Vread_with_symbol_positions, readcharfun))
3156 Vread_symbol_positions_list
3157 = Fcons (Fcons (result, make_number (start_position)),
3158 Vread_symbol_positions_list);
3159 return result;
3160 }
3161 }
3162 }
3163 }
3164 \f
3165
3166 /* List of nodes we've seen during substitute_object_in_subtree. */
3167 static Lisp_Object seen_list;
3168
3169 static void
3170 substitute_object_in_subtree (Lisp_Object object, Lisp_Object placeholder)
3171 {
3172 Lisp_Object check_object;
3173
3174 /* We haven't seen any objects when we start. */
3175 seen_list = Qnil;
3176
3177 /* Make all the substitutions. */
3178 check_object
3179 = substitute_object_recurse (object, placeholder, object);
3180
3181 /* Clear seen_list because we're done with it. */
3182 seen_list = Qnil;
3183
3184 /* The returned object here is expected to always eq the
3185 original. */
3186 if (!EQ (check_object, object))
3187 error ("Unexpected mutation error in reader");
3188 }
3189
3190 /* Feval doesn't get called from here, so no gc protection is needed. */
3191 #define SUBSTITUTE(get_val, set_val) \
3192 do { \
3193 Lisp_Object old_value = get_val; \
3194 Lisp_Object true_value \
3195 = substitute_object_recurse (object, placeholder, \
3196 old_value); \
3197 \
3198 if (!EQ (old_value, true_value)) \
3199 { \
3200 set_val; \
3201 } \
3202 } while (0)
3203
3204 static Lisp_Object
3205 substitute_object_recurse (Lisp_Object object, Lisp_Object placeholder, Lisp_Object subtree)
3206 {
3207 /* If we find the placeholder, return the target object. */
3208 if (EQ (placeholder, subtree))
3209 return object;
3210
3211 /* If we've been to this node before, don't explore it again. */
3212 if (!EQ (Qnil, Fmemq (subtree, seen_list)))
3213 return subtree;
3214
3215 /* If this node can be the entry point to a cycle, remember that
3216 we've seen it. It can only be such an entry point if it was made
3217 by #n=, which means that we can find it as a value in
3218 read_objects. */
3219 if (!EQ (Qnil, Frassq (subtree, read_objects)))
3220 seen_list = Fcons (subtree, seen_list);
3221
3222 /* Recurse according to subtree's type.
3223 Every branch must return a Lisp_Object. */
3224 switch (XTYPE (subtree))
3225 {
3226 case Lisp_Vectorlike:
3227 {
3228 ptrdiff_t i, length = 0;
3229 if (BOOL_VECTOR_P (subtree))
3230 return subtree; /* No sub-objects anyway. */
3231 else if (CHAR_TABLE_P (subtree) || SUB_CHAR_TABLE_P (subtree)
3232 || COMPILEDP (subtree) || HASH_TABLE_P (subtree))
3233 length = ASIZE (subtree) & PSEUDOVECTOR_SIZE_MASK;
3234 else if (VECTORP (subtree))
3235 length = ASIZE (subtree);
3236 else
3237 /* An unknown pseudovector may contain non-Lisp fields, so we
3238 can't just blindly traverse all its fields. We used to call
3239 `Flength' which signaled `sequencep', so I just preserved this
3240 behavior. */
3241 wrong_type_argument (Qsequencep, subtree);
3242
3243 for (i = 0; i < length; i++)
3244 SUBSTITUTE (AREF (subtree, i),
3245 ASET (subtree, i, true_value));
3246 return subtree;
3247 }
3248
3249 case Lisp_Cons:
3250 {
3251 SUBSTITUTE (XCAR (subtree),
3252 XSETCAR (subtree, true_value));
3253 SUBSTITUTE (XCDR (subtree),
3254 XSETCDR (subtree, true_value));
3255 return subtree;
3256 }
3257
3258 case Lisp_String:
3259 {
3260 /* Check for text properties in each interval.
3261 substitute_in_interval contains part of the logic. */
3262
3263 INTERVAL root_interval = string_intervals (subtree);
3264 Lisp_Object arg = Fcons (object, placeholder);
3265
3266 traverse_intervals_noorder (root_interval,
3267 &substitute_in_interval, arg);
3268
3269 return subtree;
3270 }
3271
3272 /* Other types don't recurse any further. */
3273 default:
3274 return subtree;
3275 }
3276 }
3277
3278 /* Helper function for substitute_object_recurse. */
3279 static void
3280 substitute_in_interval (INTERVAL interval, Lisp_Object arg)
3281 {
3282 Lisp_Object object = Fcar (arg);
3283 Lisp_Object placeholder = Fcdr (arg);
3284
3285 SUBSTITUTE (interval->plist, set_interval_plist (interval, true_value));
3286 }
3287
3288 \f
3289 #define LEAD_INT 1
3290 #define DOT_CHAR 2
3291 #define TRAIL_INT 4
3292 #define E_EXP 16
3293
3294
3295 /* Convert STRING to a number, assuming base BASE. Return a fixnum if CP has
3296 integer syntax and fits in a fixnum, else return the nearest float if CP has
3297 either floating point or integer syntax and BASE is 10, else return nil. If
3298 IGNORE_TRAILING, consider just the longest prefix of CP that has
3299 valid floating point syntax. Signal an overflow if BASE is not 10 and the
3300 number has integer syntax but does not fit. */
3301
3302 Lisp_Object
3303 string_to_number (char const *string, int base, bool ignore_trailing)
3304 {
3305 int state;
3306 char const *cp = string;
3307 int leading_digit;
3308 bool float_syntax = 0;
3309 double value = 0;
3310
3311 /* Compute NaN and infinities using a variable, to cope with compilers that
3312 think they are smarter than we are. */
3313 double zero = 0;
3314
3315 /* Negate the value ourselves. This treats 0, NaNs, and infinity properly on
3316 IEEE floating point hosts, and works around a formerly-common bug where
3317 atof ("-0.0") drops the sign. */
3318 bool negative = *cp == '-';
3319
3320 bool signedp = negative || *cp == '+';
3321 cp += signedp;
3322
3323 state = 0;
3324
3325 leading_digit = digit_to_number (*cp, base);
3326 if (leading_digit >= 0)
3327 {
3328 state |= LEAD_INT;
3329 do
3330 ++cp;
3331 while (digit_to_number (*cp, base) >= 0);
3332 }
3333 if (*cp == '.')
3334 {
3335 state |= DOT_CHAR;
3336 cp++;
3337 }
3338
3339 if (base == 10)
3340 {
3341 if ('0' <= *cp && *cp <= '9')
3342 {
3343 state |= TRAIL_INT;
3344 do
3345 cp++;
3346 while ('0' <= *cp && *cp <= '9');
3347 }
3348 if (*cp == 'e' || *cp == 'E')
3349 {
3350 char const *ecp = cp;
3351 cp++;
3352 if (*cp == '+' || *cp == '-')
3353 cp++;
3354 if ('0' <= *cp && *cp <= '9')
3355 {
3356 state |= E_EXP;
3357 do
3358 cp++;
3359 while ('0' <= *cp && *cp <= '9');
3360 }
3361 else if (cp[-1] == '+'
3362 && cp[0] == 'I' && cp[1] == 'N' && cp[2] == 'F')
3363 {
3364 state |= E_EXP;
3365 cp += 3;
3366 value = 1.0 / zero;
3367 }
3368 else if (cp[-1] == '+'
3369 && cp[0] == 'N' && cp[1] == 'a' && cp[2] == 'N')
3370 {
3371 state |= E_EXP;
3372 cp += 3;
3373 value = zero / zero;
3374
3375 /* If that made a "negative" NaN, negate it. */
3376 {
3377 int i;
3378 union { double d; char c[sizeof (double)]; }
3379 u_data, u_minus_zero;
3380 u_data.d = value;
3381 u_minus_zero.d = -0.0;
3382 for (i = 0; i < sizeof (double); i++)
3383 if (u_data.c[i] & u_minus_zero.c[i])
3384 {
3385 value = -value;
3386 break;
3387 }
3388 }
3389 /* Now VALUE is a positive NaN. */
3390 }
3391 else
3392 cp = ecp;
3393 }
3394
3395 float_syntax = ((state & (DOT_CHAR|TRAIL_INT)) == (DOT_CHAR|TRAIL_INT)
3396 || state == (LEAD_INT|E_EXP));
3397 }
3398
3399 /* Return nil if the number uses invalid syntax. If IGNORE_TRAILING, accept
3400 any prefix that matches. Otherwise, the entire string must match. */
3401 if (! (ignore_trailing
3402 ? ((state & LEAD_INT) != 0 || float_syntax)
3403 : (!*cp && ((state & ~DOT_CHAR) == LEAD_INT || float_syntax))))
3404 return Qnil;
3405
3406 /* If the number uses integer and not float syntax, and is in C-language
3407 range, use its value, preferably as a fixnum. */
3408 if (leading_digit >= 0 && ! float_syntax)
3409 {
3410 uintmax_t n;
3411
3412 /* Fast special case for single-digit integers. This also avoids a
3413 glitch when BASE is 16 and IGNORE_TRAILING, because in that
3414 case some versions of strtoumax accept numbers like "0x1" that Emacs
3415 does not allow. */
3416 if (digit_to_number (string[signedp + 1], base) < 0)
3417 return make_number (negative ? -leading_digit : leading_digit);
3418
3419 errno = 0;
3420 n = strtoumax (string + signedp, NULL, base);
3421 if (errno == ERANGE)
3422 {
3423 /* Unfortunately there's no simple and accurate way to convert
3424 non-base-10 numbers that are out of C-language range. */
3425 if (base != 10)
3426 xsignal1 (Qoverflow_error, build_string (string));
3427 }
3428 else if (n <= (negative ? -MOST_NEGATIVE_FIXNUM : MOST_POSITIVE_FIXNUM))
3429 {
3430 EMACS_INT signed_n = n;
3431 return make_number (negative ? -signed_n : signed_n);
3432 }
3433 else
3434 value = n;
3435 }
3436
3437 /* Either the number uses float syntax, or it does not fit into a fixnum.
3438 Convert it from string to floating point, unless the value is already
3439 known because it is an infinity, a NAN, or its absolute value fits in
3440 uintmax_t. */
3441 if (! value)
3442 value = atof (string + signedp);
3443
3444 return make_float (negative ? -value : value);
3445 }
3446
3447 \f
3448 static Lisp_Object
3449 read_vector (Lisp_Object readcharfun, bool bytecodeflag)
3450 {
3451 ptrdiff_t i, size;
3452 Lisp_Object *ptr;
3453 Lisp_Object tem, item, vector;
3454 struct Lisp_Cons *otem;
3455 Lisp_Object len;
3456
3457 tem = read_list (1, readcharfun);
3458 len = Flength (tem);
3459 vector = Fmake_vector (len, Qnil);
3460
3461 size = ASIZE (vector);
3462 ptr = XVECTOR (vector)->contents;
3463 for (i = 0; i < size; i++)
3464 {
3465 item = Fcar (tem);
3466 /* If `load-force-doc-strings' is t when reading a lazily-loaded
3467 bytecode object, the docstring containing the bytecode and
3468 constants values must be treated as unibyte and passed to
3469 Fread, to get the actual bytecode string and constants vector. */
3470 if (bytecodeflag && load_force_doc_strings)
3471 {
3472 if (i == COMPILED_BYTECODE)
3473 {
3474 if (!STRINGP (item))
3475 error ("Invalid byte code");
3476
3477 /* Delay handling the bytecode slot until we know whether
3478 it is lazily-loaded (we can tell by whether the
3479 constants slot is nil). */
3480 ASET (vector, COMPILED_CONSTANTS, item);
3481 item = Qnil;
3482 }
3483 else if (i == COMPILED_CONSTANTS)
3484 {
3485 Lisp_Object bytestr = ptr[COMPILED_CONSTANTS];
3486
3487 if (NILP (item))
3488 {
3489 /* Coerce string to unibyte (like string-as-unibyte,
3490 but without generating extra garbage and
3491 guaranteeing no change in the contents). */
3492 STRING_SET_CHARS (bytestr, SBYTES (bytestr));
3493 STRING_SET_UNIBYTE (bytestr);
3494
3495 item = Fread (Fcons (bytestr, readcharfun));
3496 if (!CONSP (item))
3497 error ("Invalid byte code");
3498
3499 otem = XCONS (item);
3500 bytestr = XCAR (item);
3501 item = XCDR (item);
3502 free_cons (otem);
3503 }
3504
3505 /* Now handle the bytecode slot. */
3506 ASET (vector, COMPILED_BYTECODE, bytestr);
3507 }
3508 else if (i == COMPILED_DOC_STRING
3509 && STRINGP (item)
3510 && ! STRING_MULTIBYTE (item))
3511 {
3512 if (EQ (readcharfun, Qget_emacs_mule_file_char))
3513 item = Fdecode_coding_string (item, Qemacs_mule, Qnil, Qnil);
3514 else
3515 item = Fstring_as_multibyte (item);
3516 }
3517 }
3518 ASET (vector, i, item);
3519 otem = XCONS (tem);
3520 tem = Fcdr (tem);
3521 free_cons (otem);
3522 }
3523 return vector;
3524 }
3525
3526 /* FLAG means check for ] to terminate rather than ) and . */
3527
3528 static Lisp_Object
3529 read_list (bool flag, Lisp_Object readcharfun)
3530 {
3531 Lisp_Object val, tail;
3532 Lisp_Object elt, tem;
3533 struct gcpro gcpro1, gcpro2;
3534 /* 0 is the normal case.
3535 1 means this list is a doc reference; replace it with the number 0.
3536 2 means this list is a doc reference; replace it with the doc string. */
3537 int doc_reference = 0;
3538
3539 /* Initialize this to 1 if we are reading a list. */
3540 bool first_in_list = flag <= 0;
3541
3542 val = Qnil;
3543 tail = Qnil;
3544
3545 while (1)
3546 {
3547 int ch;
3548 GCPRO2 (val, tail);
3549 elt = read1 (readcharfun, &ch, first_in_list);
3550 UNGCPRO;
3551
3552 first_in_list = 0;
3553
3554 /* While building, if the list starts with #$, treat it specially. */
3555 if (EQ (elt, Vload_file_name)
3556 && ! NILP (elt)
3557 && !NILP (Vpurify_flag))
3558 {
3559 if (NILP (Vdoc_file_name))
3560 /* We have not yet called Snarf-documentation, so assume
3561 this file is described in the DOC file
3562 and Snarf-documentation will fill in the right value later.
3563 For now, replace the whole list with 0. */
3564 doc_reference = 1;
3565 else
3566 /* We have already called Snarf-documentation, so make a relative
3567 file name for this file, so it can be found properly
3568 in the installed Lisp directory.
3569 We don't use Fexpand_file_name because that would make
3570 the directory absolute now. */
3571 elt = concat2 (build_string ("../lisp/"),
3572 Ffile_name_nondirectory (elt));
3573 }
3574 else if (EQ (elt, Vload_file_name)
3575 && ! NILP (elt)
3576 && load_force_doc_strings)
3577 doc_reference = 2;
3578
3579 if (ch)
3580 {
3581 if (flag > 0)
3582 {
3583 if (ch == ']')
3584 return val;
3585 invalid_syntax (") or . in a vector");
3586 }
3587 if (ch == ')')
3588 return val;
3589 if (ch == '.')
3590 {
3591 GCPRO2 (val, tail);
3592 if (!NILP (tail))
3593 XSETCDR (tail, read0 (readcharfun));
3594 else
3595 val = read0 (readcharfun);
3596 read1 (readcharfun, &ch, 0);
3597 UNGCPRO;
3598 if (ch == ')')
3599 {
3600 if (doc_reference == 1)
3601 return make_number (0);
3602 if (doc_reference == 2 && INTEGERP (XCDR (val)))
3603 {
3604 char *saved = NULL;
3605 file_offset saved_position;
3606 /* Get a doc string from the file we are loading.
3607 If it's in saved_doc_string, get it from there.
3608
3609 Here, we don't know if the string is a
3610 bytecode string or a doc string. As a
3611 bytecode string must be unibyte, we always
3612 return a unibyte string. If it is actually a
3613 doc string, caller must make it
3614 multibyte. */
3615
3616 /* Position is negative for user variables. */
3617 EMACS_INT pos = eabs (XINT (XCDR (val)));
3618 if (pos >= saved_doc_string_position
3619 && pos < (saved_doc_string_position
3620 + saved_doc_string_length))
3621 {
3622 saved = saved_doc_string;
3623 saved_position = saved_doc_string_position;
3624 }
3625 /* Look in prev_saved_doc_string the same way. */
3626 else if (pos >= prev_saved_doc_string_position
3627 && pos < (prev_saved_doc_string_position
3628 + prev_saved_doc_string_length))
3629 {
3630 saved = prev_saved_doc_string;
3631 saved_position = prev_saved_doc_string_position;
3632 }
3633 if (saved)
3634 {
3635 ptrdiff_t start = pos - saved_position;
3636 ptrdiff_t from, to;
3637
3638 /* Process quoting with ^A,
3639 and find the end of the string,
3640 which is marked with ^_ (037). */
3641 for (from = start, to = start;
3642 saved[from] != 037;)
3643 {
3644 int c = saved[from++];
3645 if (c == 1)
3646 {
3647 c = saved[from++];
3648 saved[to++] = (c == 1 ? c
3649 : c == '0' ? 0
3650 : c == '_' ? 037
3651 : c);
3652 }
3653 else
3654 saved[to++] = c;
3655 }
3656
3657 return make_unibyte_string (saved + start,
3658 to - start);
3659 }
3660 else
3661 return get_doc_string (val, 1, 0);
3662 }
3663
3664 return val;
3665 }
3666 invalid_syntax (". in wrong context");
3667 }
3668 invalid_syntax ("] in a list");
3669 }
3670 tem = list1 (elt);
3671 if (!NILP (tail))
3672 XSETCDR (tail, tem);
3673 else
3674 val = tem;
3675 tail = tem;
3676 }
3677 }
3678 \f
3679 static Lisp_Object initial_obarray;
3680
3681 /* `oblookup' stores the bucket number here, for the sake of Funintern. */
3682
3683 static size_t oblookup_last_bucket_number;
3684
3685 /* Get an error if OBARRAY is not an obarray.
3686 If it is one, return it. */
3687
3688 Lisp_Object
3689 check_obarray (Lisp_Object obarray)
3690 {
3691 if (!VECTORP (obarray) || ASIZE (obarray) == 0)
3692 {
3693 /* If Vobarray is now invalid, force it to be valid. */
3694 if (EQ (Vobarray, obarray)) Vobarray = initial_obarray;
3695 wrong_type_argument (Qvectorp, obarray);
3696 }
3697 return obarray;
3698 }
3699
3700 /* Intern the C string STR: return a symbol with that name,
3701 interned in the current obarray. */
3702
3703 Lisp_Object
3704 intern_1 (const char *str, ptrdiff_t len)
3705 {
3706 Lisp_Object obarray = check_obarray (Vobarray);
3707 Lisp_Object tem = oblookup (obarray, str, len, len);
3708
3709 return SYMBOLP (tem) ? tem : Fintern (make_string (str, len), obarray);
3710 }
3711
3712 Lisp_Object
3713 intern_c_string_1 (const char *str, ptrdiff_t len)
3714 {
3715 Lisp_Object obarray = check_obarray (Vobarray);
3716 Lisp_Object tem = oblookup (obarray, str, len, len);
3717
3718 if (SYMBOLP (tem))
3719 return tem;
3720
3721 if (NILP (Vpurify_flag))
3722 /* Creating a non-pure string from a string literal not
3723 implemented yet. We could just use make_string here and live
3724 with the extra copy. */
3725 emacs_abort ();
3726
3727 return Fintern (make_pure_c_string (str, len), obarray);
3728 }
3729 \f
3730 DEFUN ("intern", Fintern, Sintern, 1, 2, 0,
3731 doc: /* Return the canonical symbol whose name is STRING.
3732 If there is none, one is created by this function and returned.
3733 A second optional argument specifies the obarray to use;
3734 it defaults to the value of `obarray'. */)
3735 (Lisp_Object string, Lisp_Object obarray)
3736 {
3737 register Lisp_Object tem, sym, *ptr;
3738
3739 if (NILP (obarray)) obarray = Vobarray;
3740 obarray = check_obarray (obarray);
3741
3742 CHECK_STRING (string);
3743
3744 tem = oblookup (obarray, SSDATA (string),
3745 SCHARS (string),
3746 SBYTES (string));
3747 if (!INTEGERP (tem))
3748 return tem;
3749
3750 if (!NILP (Vpurify_flag))
3751 string = Fpurecopy (string);
3752 sym = Fmake_symbol (string);
3753
3754 if (EQ (obarray, initial_obarray))
3755 XSYMBOL (sym)->interned = SYMBOL_INTERNED_IN_INITIAL_OBARRAY;
3756 else
3757 XSYMBOL (sym)->interned = SYMBOL_INTERNED;
3758
3759 if ((SREF (string, 0) == ':')
3760 && EQ (obarray, initial_obarray))
3761 {
3762 XSYMBOL (sym)->constant = 1;
3763 XSYMBOL (sym)->redirect = SYMBOL_PLAINVAL;
3764 SET_SYMBOL_VAL (XSYMBOL (sym), sym);
3765 }
3766
3767 ptr = aref_addr (obarray, XINT(tem));
3768 if (SYMBOLP (*ptr))
3769 set_symbol_next (sym, XSYMBOL (*ptr));
3770 else
3771 set_symbol_next (sym, NULL);
3772 *ptr = sym;
3773 return sym;
3774 }
3775
3776 DEFUN ("intern-soft", Fintern_soft, Sintern_soft, 1, 2, 0,
3777 doc: /* Return the canonical symbol named NAME, or nil if none exists.
3778 NAME may be a string or a symbol. If it is a symbol, that exact
3779 symbol is searched for.
3780 A second optional argument specifies the obarray to use;
3781 it defaults to the value of `obarray'. */)
3782 (Lisp_Object name, Lisp_Object obarray)
3783 {
3784 register Lisp_Object tem, string;
3785
3786 if (NILP (obarray)) obarray = Vobarray;
3787 obarray = check_obarray (obarray);
3788
3789 if (!SYMBOLP (name))
3790 {
3791 CHECK_STRING (name);
3792 string = name;
3793 }
3794 else
3795 string = SYMBOL_NAME (name);
3796
3797 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3798 if (INTEGERP (tem) || (SYMBOLP (name) && !EQ (name, tem)))
3799 return Qnil;
3800 else
3801 return tem;
3802 }
3803 \f
3804 DEFUN ("unintern", Funintern, Sunintern, 1, 2, 0,
3805 doc: /* Delete the symbol named NAME, if any, from OBARRAY.
3806 The value is t if a symbol was found and deleted, nil otherwise.
3807 NAME may be a string or a symbol. If it is a symbol, that symbol
3808 is deleted, if it belongs to OBARRAY--no other symbol is deleted.
3809 OBARRAY defaults to the value of the variable `obarray'. */)
3810 (Lisp_Object name, Lisp_Object obarray)
3811 {
3812 register Lisp_Object string, tem;
3813 size_t hash;
3814
3815 if (NILP (obarray)) obarray = Vobarray;
3816 obarray = check_obarray (obarray);
3817
3818 if (SYMBOLP (name))
3819 string = SYMBOL_NAME (name);
3820 else
3821 {
3822 CHECK_STRING (name);
3823 string = name;
3824 }
3825
3826 tem = oblookup (obarray, SSDATA (string),
3827 SCHARS (string),
3828 SBYTES (string));
3829 if (INTEGERP (tem))
3830 return Qnil;
3831 /* If arg was a symbol, don't delete anything but that symbol itself. */
3832 if (SYMBOLP (name) && !EQ (name, tem))
3833 return Qnil;
3834
3835 /* There are plenty of other symbols which will screw up the Emacs
3836 session if we unintern them, as well as even more ways to use
3837 `setq' or `fset' or whatnot to make the Emacs session
3838 unusable. Let's not go down this silly road. --Stef */
3839 /* if (EQ (tem, Qnil) || EQ (tem, Qt))
3840 error ("Attempt to unintern t or nil"); */
3841
3842 XSYMBOL (tem)->interned = SYMBOL_UNINTERNED;
3843
3844 hash = oblookup_last_bucket_number;
3845
3846 if (EQ (AREF (obarray, hash), tem))
3847 {
3848 if (XSYMBOL (tem)->next)
3849 {
3850 Lisp_Object sym;
3851 XSETSYMBOL (sym, XSYMBOL (tem)->next);
3852 ASET (obarray, hash, sym);
3853 }
3854 else
3855 ASET (obarray, hash, make_number (0));
3856 }
3857 else
3858 {
3859 Lisp_Object tail, following;
3860
3861 for (tail = AREF (obarray, hash);
3862 XSYMBOL (tail)->next;
3863 tail = following)
3864 {
3865 XSETSYMBOL (following, XSYMBOL (tail)->next);
3866 if (EQ (following, tem))
3867 {
3868 set_symbol_next (tail, XSYMBOL (following)->next);
3869 break;
3870 }
3871 }
3872 }
3873
3874 return Qt;
3875 }
3876 \f
3877 /* Return the symbol in OBARRAY whose names matches the string
3878 of SIZE characters (SIZE_BYTE bytes) at PTR.
3879 If there is no such symbol in OBARRAY, return nil.
3880
3881 Also store the bucket number in oblookup_last_bucket_number. */
3882
3883 Lisp_Object
3884 oblookup (Lisp_Object obarray, register const char *ptr, ptrdiff_t size, ptrdiff_t size_byte)
3885 {
3886 size_t hash;
3887 size_t obsize;
3888 register Lisp_Object tail;
3889 Lisp_Object bucket, tem;
3890
3891 obarray = check_obarray (obarray);
3892 obsize = ASIZE (obarray);
3893
3894 /* This is sometimes needed in the middle of GC. */
3895 obsize &= ~ARRAY_MARK_FLAG;
3896 hash = hash_string (ptr, size_byte) % obsize;
3897 bucket = AREF (obarray, hash);
3898 oblookup_last_bucket_number = hash;
3899 if (EQ (bucket, make_number (0)))
3900 ;
3901 else if (!SYMBOLP (bucket))
3902 error ("Bad data in guts of obarray"); /* Like CADR error message. */
3903 else
3904 for (tail = bucket; ; XSETSYMBOL (tail, XSYMBOL (tail)->next))
3905 {
3906 if (SBYTES (SYMBOL_NAME (tail)) == size_byte
3907 && SCHARS (SYMBOL_NAME (tail)) == size
3908 && !memcmp (SDATA (SYMBOL_NAME (tail)), ptr, size_byte))
3909 return tail;
3910 else if (XSYMBOL (tail)->next == 0)
3911 break;
3912 }
3913 XSETINT (tem, hash);
3914 return tem;
3915 }
3916 \f
3917 void
3918 map_obarray (Lisp_Object obarray, void (*fn) (Lisp_Object, Lisp_Object), Lisp_Object arg)
3919 {
3920 ptrdiff_t i;
3921 register Lisp_Object tail;
3922 CHECK_VECTOR (obarray);
3923 for (i = ASIZE (obarray) - 1; i >= 0; i--)
3924 {
3925 tail = AREF (obarray, i);
3926 if (SYMBOLP (tail))
3927 while (1)
3928 {
3929 (*fn) (tail, arg);
3930 if (XSYMBOL (tail)->next == 0)
3931 break;
3932 XSETSYMBOL (tail, XSYMBOL (tail)->next);
3933 }
3934 }
3935 }
3936
3937 static void
3938 mapatoms_1 (Lisp_Object sym, Lisp_Object function)
3939 {
3940 call1 (function, sym);
3941 }
3942
3943 DEFUN ("mapatoms", Fmapatoms, Smapatoms, 1, 2, 0,
3944 doc: /* Call FUNCTION on every symbol in OBARRAY.
3945 OBARRAY defaults to the value of `obarray'. */)
3946 (Lisp_Object function, Lisp_Object obarray)
3947 {
3948 if (NILP (obarray)) obarray = Vobarray;
3949 obarray = check_obarray (obarray);
3950
3951 map_obarray (obarray, mapatoms_1, function);
3952 return Qnil;
3953 }
3954
3955 #define OBARRAY_SIZE 1511
3956
3957 void
3958 init_obarray (void)
3959 {
3960 Lisp_Object oblength;
3961 ptrdiff_t size = 100 + MAX_MULTIBYTE_LENGTH;
3962
3963 XSETFASTINT (oblength, OBARRAY_SIZE);
3964
3965 Vobarray = Fmake_vector (oblength, make_number (0));
3966 initial_obarray = Vobarray;
3967 staticpro (&initial_obarray);
3968
3969 Qunbound = Fmake_symbol (build_pure_c_string ("unbound"));
3970 /* Set temporary dummy values to Qnil and Vpurify_flag to satisfy the
3971 NILP (Vpurify_flag) check in intern_c_string. */
3972 Qnil = make_number (-1); Vpurify_flag = make_number (1);
3973 Qnil = intern_c_string ("nil");
3974
3975 /* Fmake_symbol inits fields of new symbols with Qunbound and Qnil,
3976 so those two need to be fixed manually. */
3977 SET_SYMBOL_VAL (XSYMBOL (Qunbound), Qunbound);
3978 set_symbol_function (Qunbound, Qnil);
3979 set_symbol_plist (Qunbound, Qnil);
3980 SET_SYMBOL_VAL (XSYMBOL (Qnil), Qnil);
3981 XSYMBOL (Qnil)->constant = 1;
3982 XSYMBOL (Qnil)->declared_special = 1;
3983 set_symbol_plist (Qnil, Qnil);
3984 set_symbol_function (Qnil, Qnil);
3985
3986 Qt = intern_c_string ("t");
3987 SET_SYMBOL_VAL (XSYMBOL (Qt), Qt);
3988 XSYMBOL (Qnil)->declared_special = 1;
3989 XSYMBOL (Qt)->constant = 1;
3990
3991 /* Qt is correct even if CANNOT_DUMP. loadup.el will set to nil at end. */
3992 Vpurify_flag = Qt;
3993
3994 DEFSYM (Qvariable_documentation, "variable-documentation");
3995
3996 read_buffer = xmalloc (size);
3997 read_buffer_size = size;
3998 }
3999 \f
4000 void
4001 defsubr (struct Lisp_Subr *sname)
4002 {
4003 Lisp_Object sym, tem;
4004 sym = intern_c_string (sname->symbol_name);
4005 XSETPVECTYPE (sname, PVEC_SUBR);
4006 XSETSUBR (tem, sname);
4007 set_symbol_function (sym, tem);
4008 }
4009
4010 #ifdef NOTDEF /* Use fset in subr.el now! */
4011 void
4012 defalias (struct Lisp_Subr *sname, char *string)
4013 {
4014 Lisp_Object sym;
4015 sym = intern (string);
4016 XSETSUBR (XSYMBOL (sym)->function, sname);
4017 }
4018 #endif /* NOTDEF */
4019
4020 /* Define an "integer variable"; a symbol whose value is forwarded to a
4021 C variable of type EMACS_INT. Sample call (with "xx" to fool make-docfile):
4022 DEFxxVAR_INT ("emacs-priority", &emacs_priority, "Documentation"); */
4023 void
4024 defvar_int (struct Lisp_Intfwd *i_fwd,
4025 const char *namestring, EMACS_INT *address)
4026 {
4027 Lisp_Object sym;
4028 sym = intern_c_string (namestring);
4029 i_fwd->type = Lisp_Fwd_Int;
4030 i_fwd->intvar = address;
4031 XSYMBOL (sym)->declared_special = 1;
4032 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4033 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)i_fwd);
4034 }
4035
4036 /* Similar but define a variable whose value is t if address contains 1,
4037 nil if address contains 0. */
4038 void
4039 defvar_bool (struct Lisp_Boolfwd *b_fwd,
4040 const char *namestring, bool *address)
4041 {
4042 Lisp_Object sym;
4043 sym = intern_c_string (namestring);
4044 b_fwd->type = Lisp_Fwd_Bool;
4045 b_fwd->boolvar = address;
4046 XSYMBOL (sym)->declared_special = 1;
4047 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4048 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)b_fwd);
4049 Vbyte_boolean_vars = Fcons (sym, Vbyte_boolean_vars);
4050 }
4051
4052 /* Similar but define a variable whose value is the Lisp Object stored
4053 at address. Two versions: with and without gc-marking of the C
4054 variable. The nopro version is used when that variable will be
4055 gc-marked for some other reason, since marking the same slot twice
4056 can cause trouble with strings. */
4057 void
4058 defvar_lisp_nopro (struct Lisp_Objfwd *o_fwd,
4059 const char *namestring, Lisp_Object *address)
4060 {
4061 Lisp_Object sym;
4062 sym = intern_c_string (namestring);
4063 o_fwd->type = Lisp_Fwd_Obj;
4064 o_fwd->objvar = address;
4065 XSYMBOL (sym)->declared_special = 1;
4066 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4067 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)o_fwd);
4068 }
4069
4070 void
4071 defvar_lisp (struct Lisp_Objfwd *o_fwd,
4072 const char *namestring, Lisp_Object *address)
4073 {
4074 defvar_lisp_nopro (o_fwd, namestring, address);
4075 staticpro (address);
4076 }
4077
4078 /* Similar but define a variable whose value is the Lisp Object stored
4079 at a particular offset in the current kboard object. */
4080
4081 void
4082 defvar_kboard (struct Lisp_Kboard_Objfwd *ko_fwd,
4083 const char *namestring, int offset)
4084 {
4085 Lisp_Object sym;
4086 sym = intern_c_string (namestring);
4087 ko_fwd->type = Lisp_Fwd_Kboard_Obj;
4088 ko_fwd->offset = offset;
4089 XSYMBOL (sym)->declared_special = 1;
4090 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4091 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)ko_fwd);
4092 }
4093 \f
4094 /* Check that the elements of Vload_path exist. */
4095
4096 static void
4097 load_path_check (void)
4098 {
4099 Lisp_Object path_tail;
4100
4101 /* The only elements that might not exist are those from
4102 PATH_LOADSEARCH, EMACSLOADPATH. Anything else is only added if
4103 it exists. */
4104 for (path_tail = Vload_path; !NILP (path_tail); path_tail = XCDR (path_tail))
4105 {
4106 Lisp_Object dirfile;
4107 dirfile = Fcar (path_tail);
4108 if (STRINGP (dirfile))
4109 {
4110 dirfile = Fdirectory_file_name (dirfile);
4111 if (! file_accessible_directory_p (SSDATA (dirfile)))
4112 dir_warning ("Lisp directory", XCAR (path_tail));
4113 }
4114 }
4115 }
4116
4117 /* Record the value of load-path used at the start of dumping
4118 so we can see if the site changed it later during dumping. */
4119 static Lisp_Object dump_path;
4120
4121 /* Compute the default Vload_path, with the following logic:
4122 If CANNOT_DUMP:
4123 use EMACSLOADPATH env-var if set; otherwise use PATH_LOADSEARCH,
4124 prepending PATH_SITELOADSEARCH unless --no-site-lisp.
4125 The remainder is what happens when dumping works:
4126 If purify-flag (ie dumping) just use PATH_DUMPLOADSEARCH.
4127 Otherwise use EMACSLOADPATH if set, else PATH_LOADSEARCH.
4128
4129 If !initialized, then just set both Vload_path and dump_path.
4130 If initialized, then if Vload_path != dump_path, do nothing.
4131 (Presumably the load-path has already been changed by something.
4132 This can only be from a site-load file during dumping,
4133 or because EMACSLOADPATH is set.)
4134 If Vinstallation_directory is not nil (ie, running uninstalled):
4135 If installation-dir/lisp exists and not already a member,
4136 we must be running uninstalled. Reset the load-path
4137 to just installation-dir/lisp. (The default PATH_LOADSEARCH
4138 refers to the eventual installation directories. Since we
4139 are not yet installed, we should not use them, even if they exist.)
4140 If installation-dir/lisp does not exist, just add dump_path at the
4141 end instead.
4142 Add installation-dir/leim (if exists and not already a member) at the front.
4143 Add installation-dir/site-lisp (if !no_site_lisp, and exists
4144 and not already a member) at the front.
4145 If installation-dir != source-dir (ie running an uninstalled,
4146 out-of-tree build) AND install-dir/src/Makefile exists BUT
4147 install-dir/src/Makefile.in does NOT exist (this is a sanity
4148 check), then repeat the above steps for source-dir/lisp,
4149 leim and site-lisp.
4150 Finally, add the site-lisp directories at the front (if !no_site_lisp).
4151 */
4152
4153 void
4154 init_lread (void)
4155 {
4156 const char *normal;
4157
4158 #ifdef CANNOT_DUMP
4159 #ifdef HAVE_NS
4160 const char *loadpath = ns_load_path ();
4161 #endif
4162
4163 normal = PATH_LOADSEARCH;
4164 #ifdef HAVE_NS
4165 Vload_path = decode_env_path ("EMACSLOADPATH", loadpath ? loadpath : normal);
4166 #else
4167 Vload_path = decode_env_path ("EMACSLOADPATH", normal);
4168 #endif
4169
4170 load_path_check ();
4171
4172 /* FIXME CANNOT_DUMP platforms should get source-dir/lisp etc added
4173 to their load-path too, AFAICS. I don't think we can tell the
4174 difference between initialized and !initialized in this case,
4175 so we'll have to do it unconditionally when Vinstallation_directory
4176 is non-nil. */
4177 if (!no_site_lisp && !egetenv ("EMACSLOADPATH"))
4178 {
4179 Lisp_Object sitelisp;
4180 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH);
4181 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4182 }
4183 #else /* !CANNOT_DUMP */
4184 if (NILP (Vpurify_flag))
4185 {
4186 normal = PATH_LOADSEARCH;
4187 /* If the EMACSLOADPATH environment variable is set, use its value.
4188 This doesn't apply if we're dumping. */
4189 if (egetenv ("EMACSLOADPATH"))
4190 Vload_path = decode_env_path ("EMACSLOADPATH", normal);
4191 }
4192 else
4193 normal = PATH_DUMPLOADSEARCH;
4194
4195 /* In a dumped Emacs, we normally reset the value of Vload_path using
4196 PATH_LOADSEARCH, since the value that was dumped uses lisp/ in
4197 the source directory, instead of the path of the installed elisp
4198 libraries. However, if it appears that Vload_path has already been
4199 changed from the default that was saved before dumping, don't
4200 change it further. Changes can only be due to EMACSLOADPATH, or
4201 site-lisp files that were processed during dumping. */
4202 if (initialized)
4203 {
4204 if (NILP (Fequal (dump_path, Vload_path)))
4205 {
4206 /* Do not make any changes, just check the elements exist. */
4207 /* Note: --no-site-lisp is ignored.
4208 I don't know what to do about this. */
4209 load_path_check ();
4210 }
4211 else
4212 {
4213 #ifdef HAVE_NS
4214 const char *loadpath = ns_load_path ();
4215 Vload_path = decode_env_path (0, loadpath ? loadpath : normal);
4216 #else
4217 Vload_path = decode_env_path (0, normal);
4218 #endif
4219 if (!NILP (Vinstallation_directory))
4220 {
4221 Lisp_Object tem, tem1;
4222
4223 /* Add to the path the lisp subdir of the installation
4224 dir, if it is accessible. Note: in out-of-tree builds,
4225 this directory is empty save for Makefile. */
4226 tem = Fexpand_file_name (build_string ("lisp"),
4227 Vinstallation_directory);
4228 tem1 = Ffile_accessible_directory_p (tem);
4229 if (!NILP (tem1))
4230 {
4231 if (NILP (Fmember (tem, Vload_path)))
4232 {
4233 /* We are running uninstalled. The default load-path
4234 points to the eventual installed lisp, leim
4235 directories. We should not use those now, even
4236 if they exist, so start over from a clean slate. */
4237 Vload_path = list1 (tem);
4238 }
4239 }
4240 else
4241 /* That dir doesn't exist, so add the build-time
4242 Lisp dirs instead. */
4243 Vload_path = nconc2 (Vload_path, dump_path);
4244
4245 /* Add leim under the installation dir, if it is accessible. */
4246 tem = Fexpand_file_name (build_string ("leim"),
4247 Vinstallation_directory);
4248 tem1 = Ffile_accessible_directory_p (tem);
4249 if (!NILP (tem1))
4250 {
4251 if (NILP (Fmember (tem, Vload_path)))
4252 Vload_path = Fcons (tem, Vload_path);
4253 }
4254
4255 /* Add site-lisp under the installation dir, if it exists. */
4256 if (!no_site_lisp)
4257 {
4258 tem = Fexpand_file_name (build_string ("site-lisp"),
4259 Vinstallation_directory);
4260 tem1 = Ffile_accessible_directory_p (tem);
4261 if (!NILP (tem1))
4262 {
4263 if (NILP (Fmember (tem, Vload_path)))
4264 Vload_path = Fcons (tem, Vload_path);
4265 }
4266 }
4267
4268 /* If Emacs was not built in the source directory,
4269 and it is run from where it was built, add to load-path
4270 the lisp, leim and site-lisp dirs under that directory. */
4271
4272 if (NILP (Fequal (Vinstallation_directory, Vsource_directory)))
4273 {
4274 Lisp_Object tem2;
4275
4276 tem = Fexpand_file_name (build_string ("src/Makefile"),
4277 Vinstallation_directory);
4278 tem1 = Ffile_exists_p (tem);
4279
4280 /* Don't be fooled if they moved the entire source tree
4281 AFTER dumping Emacs. If the build directory is indeed
4282 different from the source dir, src/Makefile.in and
4283 src/Makefile will not be found together. */
4284 tem = Fexpand_file_name (build_string ("src/Makefile.in"),
4285 Vinstallation_directory);
4286 tem2 = Ffile_exists_p (tem);
4287 if (!NILP (tem1) && NILP (tem2))
4288 {
4289 tem = Fexpand_file_name (build_string ("lisp"),
4290 Vsource_directory);
4291
4292 if (NILP (Fmember (tem, Vload_path)))
4293 Vload_path = Fcons (tem, Vload_path);
4294
4295 tem = Fexpand_file_name (build_string ("leim"),
4296 Vsource_directory);
4297
4298 if (NILP (Fmember (tem, Vload_path)))
4299 Vload_path = Fcons (tem, Vload_path);
4300
4301 if (!no_site_lisp)
4302 {
4303 tem = Fexpand_file_name (build_string ("site-lisp"),
4304 Vsource_directory);
4305 tem1 = Ffile_accessible_directory_p (tem);
4306 if (!NILP (tem1))
4307 {
4308 if (NILP (Fmember (tem, Vload_path)))
4309 Vload_path = Fcons (tem, Vload_path);
4310 }
4311 }
4312 }
4313 } /* Vinstallation_directory != Vsource_directory */
4314
4315 } /* if Vinstallation_directory */
4316
4317 /* Check before adding the site-lisp directories.
4318 The install should have created them, but they are not
4319 required, so no need to warn if they are absent.
4320 Or we might be running before installation. */
4321 load_path_check ();
4322
4323 /* Add the site-lisp directories at the front. */
4324 if (!no_site_lisp)
4325 {
4326 Lisp_Object sitelisp;
4327 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH);
4328 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4329 }
4330 } /* if dump_path == Vload_path */
4331 }
4332 else /* !initialized */
4333 {
4334 /* NORMAL refers to PATH_DUMPLOADSEARCH, ie the lisp dir in the
4335 source directory. We used to add ../lisp (ie the lisp dir in
4336 the build directory) at the front here, but that caused trouble
4337 because it was copied from dump_path into Vload_path, above,
4338 when Vinstallation_directory was non-nil. It should not be
4339 necessary, since in out of tree builds lisp/ is empty, save
4340 for Makefile. */
4341 Vload_path = decode_env_path (0, normal);
4342 dump_path = Vload_path;
4343 /* No point calling load_path_check; load-path only contains essential
4344 elements from the source directory at this point. They cannot
4345 be missing unless something went extremely (and improbably)
4346 wrong, in which case the build will fail in obvious ways. */
4347 }
4348 #endif /* !CANNOT_DUMP */
4349
4350 Vvalues = Qnil;
4351
4352 load_in_progress = 0;
4353 Vload_file_name = Qnil;
4354 Vstandard_input = Qt;
4355 Vloads_in_progress = Qnil;
4356 }
4357
4358 /* Print a warning that directory intended for use USE and with name
4359 DIRNAME cannot be accessed. On entry, errno should correspond to
4360 the access failure. Print the warning on stderr and put it in
4361 *Messages*. */
4362
4363 void
4364 dir_warning (char const *use, Lisp_Object dirname)
4365 {
4366 static char const format[] = "Warning: %s `%s': %s\n";
4367 int access_errno = errno;
4368 fprintf (stderr, format, use, SSDATA (dirname), strerror (access_errno));
4369
4370 /* Don't log the warning before we've initialized!! */
4371 if (initialized)
4372 {
4373 char const *diagnostic = emacs_strerror (access_errno);
4374 USE_SAFE_ALLOCA;
4375 char *buffer = SAFE_ALLOCA (sizeof format - 3 * (sizeof "%s" - 1)
4376 + strlen (use) + SBYTES (dirname)
4377 + strlen (diagnostic));
4378 ptrdiff_t message_len = esprintf (buffer, format, use, SSDATA (dirname),
4379 diagnostic);
4380 message_dolog (buffer, message_len, 0, STRING_MULTIBYTE (dirname));
4381 SAFE_FREE ();
4382 }
4383 }
4384
4385 void
4386 syms_of_lread (void)
4387 {
4388 defsubr (&Sread);
4389 defsubr (&Sread_from_string);
4390 defsubr (&Sintern);
4391 defsubr (&Sintern_soft);
4392 defsubr (&Sunintern);
4393 defsubr (&Sget_load_suffixes);
4394 defsubr (&Sload);
4395 defsubr (&Seval_buffer);
4396 defsubr (&Seval_region);
4397 defsubr (&Sread_char);
4398 defsubr (&Sread_char_exclusive);
4399 defsubr (&Sread_event);
4400 defsubr (&Sget_file_char);
4401 defsubr (&Smapatoms);
4402 defsubr (&Slocate_file_internal);
4403
4404 DEFVAR_LISP ("obarray", Vobarray,
4405 doc: /* Symbol table for use by `intern' and `read'.
4406 It is a vector whose length ought to be prime for best results.
4407 The vector's contents don't make sense if examined from Lisp programs;
4408 to find all the symbols in an obarray, use `mapatoms'. */);
4409
4410 DEFVAR_LISP ("values", Vvalues,
4411 doc: /* List of values of all expressions which were read, evaluated and printed.
4412 Order is reverse chronological. */);
4413 XSYMBOL (intern ("values"))->declared_special = 0;
4414
4415 DEFVAR_LISP ("standard-input", Vstandard_input,
4416 doc: /* Stream for read to get input from.
4417 See documentation of `read' for possible values. */);
4418 Vstandard_input = Qt;
4419
4420 DEFVAR_LISP ("read-with-symbol-positions", Vread_with_symbol_positions,
4421 doc: /* If non-nil, add position of read symbols to `read-symbol-positions-list'.
4422
4423 If this variable is a buffer, then only forms read from that buffer
4424 will be added to `read-symbol-positions-list'.
4425 If this variable is t, then all read forms will be added.
4426 The effect of all other values other than nil are not currently
4427 defined, although they may be in the future.
4428
4429 The positions are relative to the last call to `read' or
4430 `read-from-string'. It is probably a bad idea to set this variable at
4431 the toplevel; bind it instead. */);
4432 Vread_with_symbol_positions = Qnil;
4433
4434 DEFVAR_LISP ("read-symbol-positions-list", Vread_symbol_positions_list,
4435 doc: /* A list mapping read symbols to their positions.
4436 This variable is modified during calls to `read' or
4437 `read-from-string', but only when `read-with-symbol-positions' is
4438 non-nil.
4439
4440 Each element of the list looks like (SYMBOL . CHAR-POSITION), where
4441 CHAR-POSITION is an integer giving the offset of that occurrence of the
4442 symbol from the position where `read' or `read-from-string' started.
4443
4444 Note that a symbol will appear multiple times in this list, if it was
4445 read multiple times. The list is in the same order as the symbols
4446 were read in. */);
4447 Vread_symbol_positions_list = Qnil;
4448
4449 DEFVAR_LISP ("read-circle", Vread_circle,
4450 doc: /* Non-nil means read recursive structures using #N= and #N# syntax. */);
4451 Vread_circle = Qt;
4452
4453 DEFVAR_LISP ("load-path", Vload_path,
4454 doc: /* List of directories to search for files to load.
4455 Each element is a string (directory name) or nil (try default directory).
4456 Initialized based on EMACSLOADPATH environment variable, if any,
4457 otherwise to default specified by file `epaths.h' when Emacs was built. */);
4458
4459 DEFVAR_LISP ("load-suffixes", Vload_suffixes,
4460 doc: /* List of suffixes for (compiled or source) Emacs Lisp files.
4461 This list should not include the empty string.
4462 `load' and related functions try to append these suffixes, in order,
4463 to the specified file name if a Lisp suffix is allowed or required. */);
4464 Vload_suffixes = list2 (build_pure_c_string (".elc"),
4465 build_pure_c_string (".el"));
4466 DEFVAR_LISP ("load-file-rep-suffixes", Vload_file_rep_suffixes,
4467 doc: /* List of suffixes that indicate representations of \
4468 the same file.
4469 This list should normally start with the empty string.
4470
4471 Enabling Auto Compression mode appends the suffixes in
4472 `jka-compr-load-suffixes' to this list and disabling Auto Compression
4473 mode removes them again. `load' and related functions use this list to
4474 determine whether they should look for compressed versions of a file
4475 and, if so, which suffixes they should try to append to the file name
4476 in order to do so. However, if you want to customize which suffixes
4477 the loading functions recognize as compression suffixes, you should
4478 customize `jka-compr-load-suffixes' rather than the present variable. */);
4479 Vload_file_rep_suffixes = list1 (empty_unibyte_string);
4480
4481 DEFVAR_BOOL ("load-in-progress", load_in_progress,
4482 doc: /* Non-nil if inside of `load'. */);
4483 DEFSYM (Qload_in_progress, "load-in-progress");
4484
4485 DEFVAR_LISP ("after-load-alist", Vafter_load_alist,
4486 doc: /* An alist of functions to be evalled when particular files are loaded.
4487 Each element looks like (REGEXP-OR-FEATURE FUNCS...).
4488
4489 REGEXP-OR-FEATURE is either a regular expression to match file names, or
4490 a symbol \(a feature name).
4491
4492 When `load' is run and the file-name argument matches an element's
4493 REGEXP-OR-FEATURE, or when `provide' is run and provides the symbol
4494 REGEXP-OR-FEATURE, the FUNCS in the element are called.
4495
4496 An error in FORMS does not undo the load, but does prevent execution of
4497 the rest of the FORMS. */);
4498 Vafter_load_alist = Qnil;
4499
4500 DEFVAR_LISP ("load-history", Vload_history,
4501 doc: /* Alist mapping loaded file names to symbols and features.
4502 Each alist element should be a list (FILE-NAME ENTRIES...), where
4503 FILE-NAME is the name of a file that has been loaded into Emacs.
4504 The file name is absolute and true (i.e. it doesn't contain symlinks).
4505 As an exception, one of the alist elements may have FILE-NAME nil,
4506 for symbols and features not associated with any file.
4507
4508 The remaining ENTRIES in the alist element describe the functions and
4509 variables defined in that file, the features provided, and the
4510 features required. Each entry has the form `(provide . FEATURE)',
4511 `(require . FEATURE)', `(defun . FUNCTION)', `(autoload . SYMBOL)',
4512 `(defface . SYMBOL)', or `(t . SYMBOL)'. Entries like `(t . SYMBOL)'
4513 may precede a `(defun . FUNCTION)' entry, and means that SYMBOL was an
4514 autoload before this file redefined it as a function. In addition,
4515 entries may also be single symbols, which means that SYMBOL was
4516 defined by `defvar' or `defconst'.
4517
4518 During preloading, the file name recorded is relative to the main Lisp
4519 directory. These file names are converted to absolute at startup. */);
4520 Vload_history = Qnil;
4521
4522 DEFVAR_LISP ("load-file-name", Vload_file_name,
4523 doc: /* Full name of file being loaded by `load'. */);
4524 Vload_file_name = Qnil;
4525
4526 DEFVAR_LISP ("user-init-file", Vuser_init_file,
4527 doc: /* File name, including directory, of user's initialization file.
4528 If the file loaded had extension `.elc', and the corresponding source file
4529 exists, this variable contains the name of source file, suitable for use
4530 by functions like `custom-save-all' which edit the init file.
4531 While Emacs loads and evaluates the init file, value is the real name
4532 of the file, regardless of whether or not it has the `.elc' extension. */);
4533 Vuser_init_file = Qnil;
4534
4535 DEFVAR_LISP ("current-load-list", Vcurrent_load_list,
4536 doc: /* Used for internal purposes by `load'. */);
4537 Vcurrent_load_list = Qnil;
4538
4539 DEFVAR_LISP ("load-read-function", Vload_read_function,
4540 doc: /* Function used by `load' and `eval-region' for reading expressions.
4541 The default is nil, which means use the function `read'. */);
4542 Vload_read_function = Qnil;
4543
4544 DEFVAR_LISP ("load-source-file-function", Vload_source_file_function,
4545 doc: /* Function called in `load' to load an Emacs Lisp source file.
4546 The value should be a function for doing code conversion before
4547 reading a source file. It can also be nil, in which case loading is
4548 done without any code conversion.
4549
4550 If the value is a function, it is called with four arguments,
4551 FULLNAME, FILE, NOERROR, NOMESSAGE. FULLNAME is the absolute name of
4552 the file to load, FILE is the non-absolute name (for messages etc.),
4553 and NOERROR and NOMESSAGE are the corresponding arguments passed to
4554 `load'. The function should return t if the file was loaded. */);
4555 Vload_source_file_function = Qnil;
4556
4557 DEFVAR_BOOL ("load-force-doc-strings", load_force_doc_strings,
4558 doc: /* Non-nil means `load' should force-load all dynamic doc strings.
4559 This is useful when the file being loaded is a temporary copy. */);
4560 load_force_doc_strings = 0;
4561
4562 DEFVAR_BOOL ("load-convert-to-unibyte", load_convert_to_unibyte,
4563 doc: /* Non-nil means `read' converts strings to unibyte whenever possible.
4564 This is normally bound by `load' and `eval-buffer' to control `read',
4565 and is not meant for users to change. */);
4566 load_convert_to_unibyte = 0;
4567
4568 DEFVAR_LISP ("source-directory", Vsource_directory,
4569 doc: /* Directory in which Emacs sources were found when Emacs was built.
4570 You cannot count on them to still be there! */);
4571 Vsource_directory
4572 = Fexpand_file_name (build_string ("../"),
4573 Fcar (decode_env_path (0, PATH_DUMPLOADSEARCH)));
4574
4575 DEFVAR_LISP ("preloaded-file-list", Vpreloaded_file_list,
4576 doc: /* List of files that were preloaded (when dumping Emacs). */);
4577 Vpreloaded_file_list = Qnil;
4578
4579 DEFVAR_LISP ("byte-boolean-vars", Vbyte_boolean_vars,
4580 doc: /* List of all DEFVAR_BOOL variables, used by the byte code optimizer. */);
4581 Vbyte_boolean_vars = Qnil;
4582
4583 DEFVAR_BOOL ("load-dangerous-libraries", load_dangerous_libraries,
4584 doc: /* Non-nil means load dangerous compiled Lisp files.
4585 Some versions of XEmacs use different byte codes than Emacs. These
4586 incompatible byte codes can make Emacs crash when it tries to execute
4587 them. */);
4588 load_dangerous_libraries = 0;
4589
4590 DEFVAR_BOOL ("force-load-messages", force_load_messages,
4591 doc: /* Non-nil means force printing messages when loading Lisp files.
4592 This overrides the value of the NOMESSAGE argument to `load'. */);
4593 force_load_messages = 0;
4594
4595 DEFVAR_LISP ("bytecomp-version-regexp", Vbytecomp_version_regexp,
4596 doc: /* Regular expression matching safe to load compiled Lisp files.
4597 When Emacs loads a compiled Lisp file, it reads the first 512 bytes
4598 from the file, and matches them against this regular expression.
4599 When the regular expression matches, the file is considered to be safe
4600 to load. See also `load-dangerous-libraries'. */);
4601 Vbytecomp_version_regexp
4602 = build_pure_c_string ("^;;;.\\(in Emacs version\\|bytecomp version FSF\\)");
4603
4604 DEFSYM (Qlexical_binding, "lexical-binding");
4605 DEFVAR_LISP ("lexical-binding", Vlexical_binding,
4606 doc: /* Whether to use lexical binding when evaluating code.
4607 Non-nil means that the code in the current buffer should be evaluated
4608 with lexical binding.
4609 This variable is automatically set from the file variables of an
4610 interpreted Lisp file read using `load'. Unlike other file local
4611 variables, this must be set in the first line of a file. */);
4612 Vlexical_binding = Qnil;
4613 Fmake_variable_buffer_local (Qlexical_binding);
4614
4615 DEFVAR_LISP ("eval-buffer-list", Veval_buffer_list,
4616 doc: /* List of buffers being read from by calls to `eval-buffer' and `eval-region'. */);
4617 Veval_buffer_list = Qnil;
4618
4619 DEFVAR_LISP ("old-style-backquotes", Vold_style_backquotes,
4620 doc: /* Set to non-nil when `read' encounters an old-style backquote. */);
4621 Vold_style_backquotes = Qnil;
4622 DEFSYM (Qold_style_backquotes, "old-style-backquotes");
4623
4624 /* Vsource_directory was initialized in init_lread. */
4625
4626 DEFSYM (Qcurrent_load_list, "current-load-list");
4627 DEFSYM (Qstandard_input, "standard-input");
4628 DEFSYM (Qread_char, "read-char");
4629 DEFSYM (Qget_file_char, "get-file-char");
4630 DEFSYM (Qget_emacs_mule_file_char, "get-emacs-mule-file-char");
4631 DEFSYM (Qload_force_doc_strings, "load-force-doc-strings");
4632
4633 DEFSYM (Qbackquote, "`");
4634 DEFSYM (Qcomma, ",");
4635 DEFSYM (Qcomma_at, ",@");
4636 DEFSYM (Qcomma_dot, ",.");
4637
4638 DEFSYM (Qinhibit_file_name_operation, "inhibit-file-name-operation");
4639 DEFSYM (Qascii_character, "ascii-character");
4640 DEFSYM (Qfunction, "function");
4641 DEFSYM (Qload, "load");
4642 DEFSYM (Qload_file_name, "load-file-name");
4643 DEFSYM (Qeval_buffer_list, "eval-buffer-list");
4644 DEFSYM (Qfile_truename, "file-truename");
4645 DEFSYM (Qdir_ok, "dir-ok");
4646 DEFSYM (Qdo_after_load_evaluation, "do-after-load-evaluation");
4647
4648 staticpro (&dump_path);
4649
4650 staticpro (&read_objects);
4651 read_objects = Qnil;
4652 staticpro (&seen_list);
4653 seen_list = Qnil;
4654
4655 Vloads_in_progress = Qnil;
4656 staticpro (&Vloads_in_progress);
4657
4658 DEFSYM (Qhash_table, "hash-table");
4659 DEFSYM (Qdata, "data");
4660 DEFSYM (Qtest, "test");
4661 DEFSYM (Qsize, "size");
4662 DEFSYM (Qweakness, "weakness");
4663 DEFSYM (Qrehash_size, "rehash-size");
4664 DEFSYM (Qrehash_threshold, "rehash-threshold");
4665 }