Fixes: debbugs:17865
[bpt/emacs.git] / src / lread.c
1 /* Lisp parsing and input streams.
2
3 Copyright (C) 1985-1989, 1993-1995, 1997-2014 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 struct timespec 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 struct timespec wait_time = dtotimespec (duration);
626 end_time = timespec_add (current_timespec (), 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 When searching suffixes, this function normally stops at the first
1034 one that exists. If the option `load-prefer-newer' is non-nil,
1035 however, it tries all suffixes, and uses whichever file is the newest.
1036
1037 Loading a file records its definitions, and its `provide' and
1038 `require' calls, in an element of `load-history' whose
1039 car is the file name loaded. See `load-history'.
1040
1041 While the file is in the process of being loaded, the variable
1042 `load-in-progress' is non-nil and the variable `load-file-name'
1043 is bound to the file's name.
1044
1045 Return t if the file exists and loads successfully. */)
1046 (Lisp_Object file, Lisp_Object noerror, Lisp_Object nomessage,
1047 Lisp_Object nosuffix, Lisp_Object must_suffix)
1048 {
1049 FILE *stream;
1050 int fd;
1051 int fd_index;
1052 ptrdiff_t count = SPECPDL_INDEX ();
1053 struct gcpro gcpro1, gcpro2, gcpro3;
1054 Lisp_Object found, efound, hist_file_name;
1055 /* True means we printed the ".el is newer" message. */
1056 bool newer = 0;
1057 /* True means we are loading a compiled file. */
1058 bool compiled = 0;
1059 Lisp_Object handler;
1060 bool safe_p = 1;
1061 const char *fmode = "r";
1062 int version;
1063
1064 #ifdef DOS_NT
1065 fmode = "rt";
1066 #endif /* DOS_NT */
1067
1068 CHECK_STRING (file);
1069
1070 /* If file name is magic, call the handler. */
1071 /* This shouldn't be necessary any more now that `openp' handles it right.
1072 handler = Ffind_file_name_handler (file, Qload);
1073 if (!NILP (handler))
1074 return call5 (handler, Qload, file, noerror, nomessage, nosuffix); */
1075
1076 /* Do this after the handler to avoid
1077 the need to gcpro noerror, nomessage and nosuffix.
1078 (Below here, we care only whether they are nil or not.)
1079 The presence of this call is the result of a historical accident:
1080 it used to be in every file-operation and when it got removed
1081 everywhere, it accidentally stayed here. Since then, enough people
1082 supposedly have things like (load "$PROJECT/foo.el") in their .emacs
1083 that it seemed risky to remove. */
1084 if (! NILP (noerror))
1085 {
1086 file = internal_condition_case_1 (Fsubstitute_in_file_name, file,
1087 Qt, load_error_handler);
1088 if (NILP (file))
1089 return Qnil;
1090 }
1091 else
1092 file = Fsubstitute_in_file_name (file);
1093
1094 /* Avoid weird lossage with null string as arg,
1095 since it would try to load a directory as a Lisp file. */
1096 if (SCHARS (file) == 0)
1097 {
1098 fd = -1;
1099 errno = ENOENT;
1100 }
1101 else
1102 {
1103 Lisp_Object suffixes;
1104 found = Qnil;
1105 GCPRO2 (file, found);
1106
1107 if (! NILP (must_suffix))
1108 {
1109 /* Don't insist on adding a suffix if FILE already ends with one. */
1110 ptrdiff_t size = SBYTES (file);
1111 if (size > 3
1112 && !strcmp (SSDATA (file) + size - 3, ".el"))
1113 must_suffix = Qnil;
1114 else if (size > 4
1115 && !strcmp (SSDATA (file) + size - 4, ".elc"))
1116 must_suffix = Qnil;
1117 /* Don't insist on adding a suffix
1118 if the argument includes a directory name. */
1119 else if (! NILP (Ffile_name_directory (file)))
1120 must_suffix = Qnil;
1121 }
1122
1123 if (!NILP (nosuffix))
1124 suffixes = Qnil;
1125 else
1126 {
1127 suffixes = Fget_load_suffixes ();
1128 if (NILP (must_suffix))
1129 {
1130 Lisp_Object arg[2];
1131 arg[0] = suffixes;
1132 arg[1] = Vload_file_rep_suffixes;
1133 suffixes = Fappend (2, arg);
1134 }
1135 }
1136
1137 fd = openp (Vload_path, file, suffixes, &found, Qnil, load_prefer_newer);
1138 UNGCPRO;
1139 }
1140
1141 if (fd == -1)
1142 {
1143 if (NILP (noerror))
1144 report_file_error ("Cannot open load file", file);
1145 return Qnil;
1146 }
1147
1148 /* Tell startup.el whether or not we found the user's init file. */
1149 if (EQ (Qt, Vuser_init_file))
1150 Vuser_init_file = found;
1151
1152 /* If FD is -2, that means openp found a magic file. */
1153 if (fd == -2)
1154 {
1155 if (NILP (Fequal (found, file)))
1156 /* If FOUND is a different file name from FILE,
1157 find its handler even if we have already inhibited
1158 the `load' operation on FILE. */
1159 handler = Ffind_file_name_handler (found, Qt);
1160 else
1161 handler = Ffind_file_name_handler (found, Qload);
1162 if (! NILP (handler))
1163 return call5 (handler, Qload, found, noerror, nomessage, Qt);
1164 #ifdef DOS_NT
1165 /* Tramp has to deal with semi-broken packages that prepend
1166 drive letters to remote files. For that reason, Tramp
1167 catches file operations that test for file existence, which
1168 makes openp think X:/foo.elc files are remote. However,
1169 Tramp does not catch `load' operations for such files, so we
1170 end up with a nil as the `load' handler above. If we would
1171 continue with fd = -2, we will behave wrongly, and in
1172 particular try reading a .elc file in the "rt" mode instead
1173 of "rb". See bug #9311 for the results. To work around
1174 this, we try to open the file locally, and go with that if it
1175 succeeds. */
1176 fd = emacs_open (SSDATA (ENCODE_FILE (found)), O_RDONLY, 0);
1177 if (fd == -1)
1178 fd = -2;
1179 #endif
1180 }
1181
1182 if (fd < 0)
1183 {
1184 /* Pacify older GCC with --enable-gcc-warnings. */
1185 IF_LINT (fd_index = 0);
1186 }
1187 else
1188 {
1189 fd_index = SPECPDL_INDEX ();
1190 record_unwind_protect_int (close_file_unwind, fd);
1191 }
1192
1193 /* Check if we're stuck in a recursive load cycle.
1194
1195 2000-09-21: It's not possible to just check for the file loaded
1196 being a member of Vloads_in_progress. This fails because of the
1197 way the byte compiler currently works; `provide's are not
1198 evaluated, see font-lock.el/jit-lock.el as an example. This
1199 leads to a certain amount of ``normal'' recursion.
1200
1201 Also, just loading a file recursively is not always an error in
1202 the general case; the second load may do something different. */
1203 {
1204 int load_count = 0;
1205 Lisp_Object tem;
1206 for (tem = Vloads_in_progress; CONSP (tem); tem = XCDR (tem))
1207 if (!NILP (Fequal (found, XCAR (tem))) && (++load_count > 3))
1208 signal_error ("Recursive load", Fcons (found, Vloads_in_progress));
1209 record_unwind_protect (record_load_unwind, Vloads_in_progress);
1210 Vloads_in_progress = Fcons (found, Vloads_in_progress);
1211 }
1212
1213 /* All loads are by default dynamic, unless the file itself specifies
1214 otherwise using a file-variable in the first line. This is bound here
1215 so that it takes effect whether or not we use
1216 Vload_source_file_function. */
1217 specbind (Qlexical_binding, Qnil);
1218
1219 /* Get the name for load-history. */
1220 hist_file_name = (! NILP (Vpurify_flag)
1221 ? concat2 (Ffile_name_directory (file),
1222 Ffile_name_nondirectory (found))
1223 : found) ;
1224
1225 version = -1;
1226
1227 /* Check for the presence of old-style quotes and warn about them. */
1228 specbind (Qold_style_backquotes, Qnil);
1229 record_unwind_protect (load_warn_old_style_backquotes, file);
1230
1231 if (!memcmp (SDATA (found) + SBYTES (found) - 4, ".elc", 4)
1232 || (fd >= 0 && (version = safe_to_load_version (fd)) > 0))
1233 /* Load .elc files directly, but not when they are
1234 remote and have no handler! */
1235 {
1236 if (fd != -2)
1237 {
1238 struct stat s1, s2;
1239 int result;
1240
1241 GCPRO3 (file, found, hist_file_name);
1242
1243 if (version < 0
1244 && ! (version = safe_to_load_version (fd)))
1245 {
1246 safe_p = 0;
1247 if (!load_dangerous_libraries)
1248 error ("File `%s' was not compiled in Emacs", SDATA (found));
1249 else if (!NILP (nomessage) && !force_load_messages)
1250 message_with_string ("File `%s' not compiled in Emacs", found, 1);
1251 }
1252
1253 compiled = 1;
1254
1255 efound = ENCODE_FILE (found);
1256
1257 #ifdef DOS_NT
1258 fmode = "rb";
1259 #endif /* DOS_NT */
1260
1261 /* openp already checked for newness, no point doing it again.
1262 FIXME would be nice to get a message when openp
1263 ignores suffix order due to load_prefer_newer. */
1264 if (!load_prefer_newer)
1265 {
1266 result = stat (SSDATA (efound), &s1);
1267 if (result == 0)
1268 {
1269 SSET (efound, SBYTES (efound) - 1, 0);
1270 result = stat (SSDATA (efound), &s2);
1271 SSET (efound, SBYTES (efound) - 1, 'c');
1272 }
1273
1274 if (result == 0
1275 && timespec_cmp (get_stat_mtime (&s1), get_stat_mtime (&s2)) < 0)
1276 {
1277 /* Make the progress messages mention that source is newer. */
1278 newer = 1;
1279
1280 /* If we won't print another message, mention this anyway. */
1281 if (!NILP (nomessage) && !force_load_messages)
1282 {
1283 Lisp_Object msg_file;
1284 msg_file = Fsubstring (found, make_number (0), make_number (-1));
1285 message_with_string ("Source file `%s' newer than byte-compiled file",
1286 msg_file, 1);
1287 }
1288 }
1289 } /* !load_prefer_newer */
1290 UNGCPRO;
1291 }
1292 }
1293 else
1294 {
1295 /* We are loading a source file (*.el). */
1296 if (!NILP (Vload_source_file_function))
1297 {
1298 Lisp_Object val;
1299
1300 if (fd >= 0)
1301 {
1302 emacs_close (fd);
1303 clear_unwind_protect (fd_index);
1304 }
1305 val = call4 (Vload_source_file_function, found, hist_file_name,
1306 NILP (noerror) ? Qnil : Qt,
1307 (NILP (nomessage) || force_load_messages) ? Qnil : Qt);
1308 return unbind_to (count, val);
1309 }
1310 }
1311
1312 GCPRO3 (file, found, hist_file_name);
1313
1314 if (fd < 0)
1315 {
1316 /* We somehow got here with fd == -2, meaning the file is deemed
1317 to be remote. Don't even try to reopen the file locally;
1318 just force a failure. */
1319 stream = NULL;
1320 errno = EINVAL;
1321 }
1322 else
1323 {
1324 #ifdef WINDOWSNT
1325 emacs_close (fd);
1326 clear_unwind_protect (fd_index);
1327 efound = ENCODE_FILE (found);
1328 stream = emacs_fopen (SSDATA (efound), fmode);
1329 #else
1330 stream = fdopen (fd, fmode);
1331 #endif
1332 }
1333 if (! stream)
1334 report_file_error ("Opening stdio stream", file);
1335 set_unwind_protect_ptr (fd_index, fclose_unwind, stream);
1336
1337 if (! NILP (Vpurify_flag))
1338 Vpreloaded_file_list = Fcons (Fpurecopy (file), Vpreloaded_file_list);
1339
1340 if (NILP (nomessage) || force_load_messages)
1341 {
1342 if (!safe_p)
1343 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...",
1344 file, 1);
1345 else if (!compiled)
1346 message_with_string ("Loading %s (source)...", file, 1);
1347 else if (newer)
1348 message_with_string ("Loading %s (compiled; note, source file is newer)...",
1349 file, 1);
1350 else /* The typical case; compiled file newer than source file. */
1351 message_with_string ("Loading %s...", file, 1);
1352 }
1353
1354 specbind (Qload_file_name, found);
1355 specbind (Qinhibit_file_name_operation, Qnil);
1356 specbind (Qload_in_progress, Qt);
1357
1358 instream = stream;
1359 if (lisp_file_lexically_bound_p (Qget_file_char))
1360 Fset (Qlexical_binding, Qt);
1361
1362 if (! version || version >= 22)
1363 readevalloop (Qget_file_char, stream, hist_file_name,
1364 0, Qnil, Qnil, Qnil, Qnil);
1365 else
1366 {
1367 /* We can't handle a file which was compiled with
1368 byte-compile-dynamic by older version of Emacs. */
1369 specbind (Qload_force_doc_strings, Qt);
1370 readevalloop (Qget_emacs_mule_file_char, stream, hist_file_name,
1371 0, Qnil, Qnil, Qnil, Qnil);
1372 }
1373 unbind_to (count, Qnil);
1374
1375 /* Run any eval-after-load forms for this file. */
1376 if (!NILP (Ffboundp (Qdo_after_load_evaluation)))
1377 call1 (Qdo_after_load_evaluation, hist_file_name) ;
1378
1379 UNGCPRO;
1380
1381 xfree (saved_doc_string);
1382 saved_doc_string = 0;
1383 saved_doc_string_size = 0;
1384
1385 xfree (prev_saved_doc_string);
1386 prev_saved_doc_string = 0;
1387 prev_saved_doc_string_size = 0;
1388
1389 if (!noninteractive && (NILP (nomessage) || force_load_messages))
1390 {
1391 if (!safe_p)
1392 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...done",
1393 file, 1);
1394 else if (!compiled)
1395 message_with_string ("Loading %s (source)...done", file, 1);
1396 else if (newer)
1397 message_with_string ("Loading %s (compiled; note, source file is newer)...done",
1398 file, 1);
1399 else /* The typical case; compiled file newer than source file. */
1400 message_with_string ("Loading %s...done", file, 1);
1401 }
1402
1403 return Qt;
1404 }
1405 \f
1406 static bool
1407 complete_filename_p (Lisp_Object pathname)
1408 {
1409 const unsigned char *s = SDATA (pathname);
1410 return (IS_DIRECTORY_SEP (s[0])
1411 || (SCHARS (pathname) > 2
1412 && IS_DEVICE_SEP (s[1]) && IS_DIRECTORY_SEP (s[2])));
1413 }
1414
1415 DEFUN ("locate-file-internal", Flocate_file_internal, Slocate_file_internal, 2, 4, 0,
1416 doc: /* Search for FILENAME through PATH.
1417 Returns the file's name in absolute form, or nil if not found.
1418 If SUFFIXES is non-nil, it should be a list of suffixes to append to
1419 file name when searching.
1420 If non-nil, PREDICATE is used instead of `file-readable-p'.
1421 PREDICATE can also be an integer to pass to the faccessat(2) function,
1422 in which case file-name-handlers are ignored.
1423 This function will normally skip directories, so if you want it to find
1424 directories, make sure the PREDICATE function returns `dir-ok' for them. */)
1425 (Lisp_Object filename, Lisp_Object path, Lisp_Object suffixes, Lisp_Object predicate)
1426 {
1427 Lisp_Object file;
1428 int fd = openp (path, filename, suffixes, &file, predicate, false);
1429 if (NILP (predicate) && fd >= 0)
1430 emacs_close (fd);
1431 return file;
1432 }
1433
1434 static Lisp_Object Qdir_ok;
1435
1436 /* Search for a file whose name is STR, looking in directories
1437 in the Lisp list PATH, and trying suffixes from SUFFIX.
1438 On success, return a file descriptor (or 1 or -2 as described below).
1439 On failure, return -1 and set errno.
1440
1441 SUFFIXES is a list of strings containing possible suffixes.
1442 The empty suffix is automatically added if the list is empty.
1443
1444 PREDICATE non-nil means don't open the files,
1445 just look for one that satisfies the predicate. In this case,
1446 return 1 on success. The predicate can be a lisp function or
1447 an integer to pass to `access' (in which case file-name-handlers
1448 are ignored).
1449
1450 If STOREPTR is nonzero, it points to a slot where the name of
1451 the file actually found should be stored as a Lisp string.
1452 nil is stored there on failure.
1453
1454 If the file we find is remote, return -2
1455 but store the found remote file name in *STOREPTR.
1456
1457 If NEWER is true, try all SUFFIXes and return the result for the
1458 newest file that exists. Does not apply to remote files,
1459 or if PREDICATE is specified. */
1460
1461 int
1462 openp (Lisp_Object path, Lisp_Object str, Lisp_Object suffixes,
1463 Lisp_Object *storeptr, Lisp_Object predicate, bool newer)
1464 {
1465 ptrdiff_t fn_size = 100;
1466 char buf[100];
1467 char *fn = buf;
1468 bool absolute;
1469 ptrdiff_t want_length;
1470 Lisp_Object filename;
1471 struct gcpro gcpro1, gcpro2, gcpro3, gcpro4, gcpro5, gcpro6, gcpro7;
1472 Lisp_Object string, tail, encoded_fn, save_string;
1473 ptrdiff_t max_suffix_len = 0;
1474 int last_errno = ENOENT;
1475 int save_fd = -1;
1476
1477 /* The last-modified time of the newest matching file found.
1478 Initialize it to something less than all valid timestamps. */
1479 struct timespec save_mtime = make_timespec (TYPE_MINIMUM (time_t), -1);
1480
1481 CHECK_STRING (str);
1482
1483 for (tail = suffixes; CONSP (tail); tail = XCDR (tail))
1484 {
1485 CHECK_STRING_CAR (tail);
1486 max_suffix_len = max (max_suffix_len,
1487 SBYTES (XCAR (tail)));
1488 }
1489
1490 string = filename = encoded_fn = save_string = Qnil;
1491 GCPRO7 (str, string, save_string, filename, path, suffixes, encoded_fn);
1492
1493 if (storeptr)
1494 *storeptr = Qnil;
1495
1496 absolute = complete_filename_p (str);
1497
1498 for (; CONSP (path); path = XCDR (path))
1499 {
1500 filename = Fexpand_file_name (str, XCAR (path));
1501 if (!complete_filename_p (filename))
1502 /* If there are non-absolute elts in PATH (eg "."). */
1503 /* Of course, this could conceivably lose if luser sets
1504 default-directory to be something non-absolute... */
1505 {
1506 filename = Fexpand_file_name (filename, BVAR (current_buffer, directory));
1507 if (!complete_filename_p (filename))
1508 /* Give up on this path element! */
1509 continue;
1510 }
1511
1512 /* Calculate maximum length of any filename made from
1513 this path element/specified file name and any possible suffix. */
1514 want_length = max_suffix_len + SBYTES (filename);
1515 if (fn_size <= want_length)
1516 fn = alloca (fn_size = 100 + want_length);
1517
1518 /* Loop over suffixes. */
1519 for (tail = NILP (suffixes) ? list1 (empty_unibyte_string) : suffixes;
1520 CONSP (tail); tail = XCDR (tail))
1521 {
1522 Lisp_Object suffix = XCAR (tail);
1523 ptrdiff_t fnlen, lsuffix = SBYTES (suffix);
1524 Lisp_Object handler;
1525
1526 /* Concatenate path element/specified name with the suffix.
1527 If the directory starts with /:, remove that. */
1528 int prefixlen = ((SCHARS (filename) > 2
1529 && SREF (filename, 0) == '/'
1530 && SREF (filename, 1) == ':')
1531 ? 2 : 0);
1532 fnlen = SBYTES (filename) - prefixlen;
1533 memcpy (fn, SDATA (filename) + prefixlen, fnlen);
1534 memcpy (fn + fnlen, SDATA (suffix), lsuffix + 1);
1535 fnlen += lsuffix;
1536 /* Check that the file exists and is not a directory. */
1537 /* We used to only check for handlers on non-absolute file names:
1538 if (absolute)
1539 handler = Qnil;
1540 else
1541 handler = Ffind_file_name_handler (filename, Qfile_exists_p);
1542 It's not clear why that was the case and it breaks things like
1543 (load "/bar.el") where the file is actually "/bar.el.gz". */
1544 /* make_string has its own ideas on when to return a unibyte
1545 string and when a multibyte string, but we know better.
1546 We must have a unibyte string when dumping, since
1547 file-name encoding is shaky at best at that time, and in
1548 particular default-file-name-coding-system is reset
1549 several times during loadup. We therefore don't want to
1550 encode the file before passing it to file I/O library
1551 functions. */
1552 if (!STRING_MULTIBYTE (filename) && !STRING_MULTIBYTE (suffix))
1553 string = make_unibyte_string (fn, fnlen);
1554 else
1555 string = make_string (fn, fnlen);
1556 handler = Ffind_file_name_handler (string, Qfile_exists_p);
1557 if ((!NILP (handler) || !NILP (predicate)) && !NATNUMP (predicate))
1558 {
1559 bool exists;
1560 if (NILP (predicate))
1561 exists = !NILP (Ffile_readable_p (string));
1562 else
1563 {
1564 Lisp_Object tmp = call1 (predicate, string);
1565 if (NILP (tmp))
1566 exists = false;
1567 else if (EQ (tmp, Qdir_ok)
1568 || NILP (Ffile_directory_p (string)))
1569 exists = true;
1570 else
1571 {
1572 exists = false;
1573 last_errno = EISDIR;
1574 }
1575 }
1576
1577 if (exists)
1578 {
1579 /* We succeeded; return this descriptor and filename. */
1580 if (storeptr)
1581 *storeptr = string;
1582 UNGCPRO;
1583 return -2;
1584 }
1585 }
1586 else
1587 {
1588 int fd;
1589 const char *pfn;
1590 struct stat st;
1591
1592 encoded_fn = ENCODE_FILE (string);
1593 pfn = SSDATA (encoded_fn);
1594
1595 /* Check that we can access or open it. */
1596 if (NATNUMP (predicate))
1597 {
1598 fd = -1;
1599 if (INT_MAX < XFASTINT (predicate))
1600 last_errno = EINVAL;
1601 else if (faccessat (AT_FDCWD, pfn, XFASTINT (predicate),
1602 AT_EACCESS)
1603 == 0)
1604 {
1605 if (file_directory_p (pfn))
1606 last_errno = EISDIR;
1607 else
1608 fd = 1;
1609 }
1610 }
1611 else
1612 {
1613 fd = emacs_open (pfn, O_RDONLY, 0);
1614 if (fd < 0)
1615 {
1616 if (errno != ENOENT)
1617 last_errno = errno;
1618 }
1619 else
1620 {
1621 int err = (fstat (fd, &st) != 0 ? errno
1622 : S_ISDIR (st.st_mode) ? EISDIR : 0);
1623 if (err)
1624 {
1625 last_errno = err;
1626 emacs_close (fd);
1627 fd = -1;
1628 }
1629 }
1630 }
1631
1632 if (fd >= 0)
1633 {
1634 if (newer && !NATNUMP (predicate))
1635 {
1636 struct timespec mtime = get_stat_mtime (&st);
1637
1638 if (timespec_cmp (mtime, save_mtime) <= 0)
1639 emacs_close (fd);
1640 else
1641 {
1642 if (0 <= save_fd)
1643 emacs_close (save_fd);
1644 save_fd = fd;
1645 save_mtime = mtime;
1646 save_string = string;
1647 }
1648 }
1649 else
1650 {
1651 /* We succeeded; return this descriptor and filename. */
1652 if (storeptr)
1653 *storeptr = string;
1654 UNGCPRO;
1655 return fd;
1656 }
1657 }
1658
1659 /* No more suffixes. Return the newest. */
1660 if (0 <= save_fd && ! CONSP (XCDR (tail)))
1661 {
1662 if (storeptr)
1663 *storeptr = save_string;
1664 UNGCPRO;
1665 return save_fd;
1666 }
1667 }
1668 }
1669 if (absolute)
1670 break;
1671 }
1672
1673 UNGCPRO;
1674 errno = last_errno;
1675 return -1;
1676 }
1677
1678 \f
1679 /* Merge the list we've accumulated of globals from the current input source
1680 into the load_history variable. The details depend on whether
1681 the source has an associated file name or not.
1682
1683 FILENAME is the file name that we are loading from.
1684
1685 ENTIRE is true if loading that entire file, false if evaluating
1686 part of it. */
1687
1688 static void
1689 build_load_history (Lisp_Object filename, bool entire)
1690 {
1691 Lisp_Object tail, prev, newelt;
1692 Lisp_Object tem, tem2;
1693 bool foundit = 0;
1694
1695 tail = Vload_history;
1696 prev = Qnil;
1697
1698 while (CONSP (tail))
1699 {
1700 tem = XCAR (tail);
1701
1702 /* Find the feature's previous assoc list... */
1703 if (!NILP (Fequal (filename, Fcar (tem))))
1704 {
1705 foundit = 1;
1706
1707 /* If we're loading the entire file, remove old data. */
1708 if (entire)
1709 {
1710 if (NILP (prev))
1711 Vload_history = XCDR (tail);
1712 else
1713 Fsetcdr (prev, XCDR (tail));
1714 }
1715
1716 /* Otherwise, cons on new symbols that are not already members. */
1717 else
1718 {
1719 tem2 = Vcurrent_load_list;
1720
1721 while (CONSP (tem2))
1722 {
1723 newelt = XCAR (tem2);
1724
1725 if (NILP (Fmember (newelt, tem)))
1726 Fsetcar (tail, Fcons (XCAR (tem),
1727 Fcons (newelt, XCDR (tem))));
1728
1729 tem2 = XCDR (tem2);
1730 QUIT;
1731 }
1732 }
1733 }
1734 else
1735 prev = tail;
1736 tail = XCDR (tail);
1737 QUIT;
1738 }
1739
1740 /* If we're loading an entire file, cons the new assoc onto the
1741 front of load-history, the most-recently-loaded position. Also
1742 do this if we didn't find an existing member for the file. */
1743 if (entire || !foundit)
1744 Vload_history = Fcons (Fnreverse (Vcurrent_load_list),
1745 Vload_history);
1746 }
1747
1748 static void
1749 readevalloop_1 (int old)
1750 {
1751 load_convert_to_unibyte = old;
1752 }
1753
1754 /* Signal an `end-of-file' error, if possible with file name
1755 information. */
1756
1757 static _Noreturn void
1758 end_of_file_error (void)
1759 {
1760 if (STRINGP (Vload_file_name))
1761 xsignal1 (Qend_of_file, Vload_file_name);
1762
1763 xsignal0 (Qend_of_file);
1764 }
1765
1766 /* UNIBYTE specifies how to set load_convert_to_unibyte
1767 for this invocation.
1768 READFUN, if non-nil, is used instead of `read'.
1769
1770 START, END specify region to read in current buffer (from eval-region).
1771 If the input is not from a buffer, they must be nil. */
1772
1773 static void
1774 readevalloop (Lisp_Object readcharfun,
1775 FILE *stream,
1776 Lisp_Object sourcename,
1777 bool printflag,
1778 Lisp_Object unibyte, Lisp_Object readfun,
1779 Lisp_Object start, Lisp_Object end)
1780 {
1781 register int c;
1782 register Lisp_Object val;
1783 ptrdiff_t count = SPECPDL_INDEX ();
1784 struct gcpro gcpro1, gcpro2, gcpro3, gcpro4;
1785 struct buffer *b = 0;
1786 bool continue_reading_p;
1787 Lisp_Object lex_bound;
1788 /* True if reading an entire buffer. */
1789 bool whole_buffer = 0;
1790 /* True on the first time around. */
1791 bool first_sexp = 1;
1792 Lisp_Object macroexpand = intern ("internal-macroexpand-for-load");
1793
1794 if (NILP (Ffboundp (macroexpand))
1795 /* Don't macroexpand in .elc files, since it should have been done
1796 already. We actually don't know whether we're in a .elc file or not,
1797 so we use circumstantial evidence: .el files normally go through
1798 Vload_source_file_function -> load-with-code-conversion
1799 -> eval-buffer. */
1800 || EQ (readcharfun, Qget_file_char)
1801 || EQ (readcharfun, Qget_emacs_mule_file_char))
1802 macroexpand = Qnil;
1803
1804 if (MARKERP (readcharfun))
1805 {
1806 if (NILP (start))
1807 start = readcharfun;
1808 }
1809
1810 if (BUFFERP (readcharfun))
1811 b = XBUFFER (readcharfun);
1812 else if (MARKERP (readcharfun))
1813 b = XMARKER (readcharfun)->buffer;
1814
1815 /* We assume START is nil when input is not from a buffer. */
1816 if (! NILP (start) && !b)
1817 emacs_abort ();
1818
1819 specbind (Qstandard_input, readcharfun); /* GCPROs readcharfun. */
1820 specbind (Qcurrent_load_list, Qnil);
1821 record_unwind_protect_int (readevalloop_1, load_convert_to_unibyte);
1822 load_convert_to_unibyte = !NILP (unibyte);
1823
1824 /* If lexical binding is active (either because it was specified in
1825 the file's header, or via a buffer-local variable), create an empty
1826 lexical environment, otherwise, turn off lexical binding. */
1827 lex_bound = find_symbol_value (Qlexical_binding);
1828 specbind (Qinternal_interpreter_environment,
1829 (NILP (lex_bound) || EQ (lex_bound, Qunbound)
1830 ? Qnil : list1 (Qt)));
1831
1832 GCPRO4 (sourcename, readfun, start, end);
1833
1834 /* Try to ensure sourcename is a truename, except whilst preloading. */
1835 if (NILP (Vpurify_flag)
1836 && !NILP (sourcename) && !NILP (Ffile_name_absolute_p (sourcename))
1837 && !NILP (Ffboundp (Qfile_truename)))
1838 sourcename = call1 (Qfile_truename, sourcename) ;
1839
1840 LOADHIST_ATTACH (sourcename);
1841
1842 continue_reading_p = 1;
1843 while (continue_reading_p)
1844 {
1845 ptrdiff_t count1 = SPECPDL_INDEX ();
1846
1847 if (b != 0 && !BUFFER_LIVE_P (b))
1848 error ("Reading from killed buffer");
1849
1850 if (!NILP (start))
1851 {
1852 /* Switch to the buffer we are reading from. */
1853 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1854 set_buffer_internal (b);
1855
1856 /* Save point in it. */
1857 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1858 /* Save ZV in it. */
1859 record_unwind_protect (save_restriction_restore, save_restriction_save ());
1860 /* Those get unbound after we read one expression. */
1861
1862 /* Set point and ZV around stuff to be read. */
1863 Fgoto_char (start);
1864 if (!NILP (end))
1865 Fnarrow_to_region (make_number (BEGV), end);
1866
1867 /* Just for cleanliness, convert END to a marker
1868 if it is an integer. */
1869 if (INTEGERP (end))
1870 end = Fpoint_max_marker ();
1871 }
1872
1873 /* On the first cycle, we can easily test here
1874 whether we are reading the whole buffer. */
1875 if (b && first_sexp)
1876 whole_buffer = (PT == BEG && ZV == Z);
1877
1878 instream = stream;
1879 read_next:
1880 c = READCHAR;
1881 if (c == ';')
1882 {
1883 while ((c = READCHAR) != '\n' && c != -1);
1884 goto read_next;
1885 }
1886 if (c < 0)
1887 {
1888 unbind_to (count1, Qnil);
1889 break;
1890 }
1891
1892 /* Ignore whitespace here, so we can detect eof. */
1893 if (c == ' ' || c == '\t' || c == '\n' || c == '\f' || c == '\r'
1894 || c == 0xa0) /* NBSP */
1895 goto read_next;
1896
1897 if (!NILP (Vpurify_flag) && c == '(')
1898 {
1899 val = read_list (0, readcharfun);
1900 }
1901 else
1902 {
1903 UNREAD (c);
1904 read_objects = Qnil;
1905 if (!NILP (readfun))
1906 {
1907 val = call1 (readfun, readcharfun);
1908
1909 /* If READCHARFUN has set point to ZV, we should
1910 stop reading, even if the form read sets point
1911 to a different value when evaluated. */
1912 if (BUFFERP (readcharfun))
1913 {
1914 struct buffer *buf = XBUFFER (readcharfun);
1915 if (BUF_PT (buf) == BUF_ZV (buf))
1916 continue_reading_p = 0;
1917 }
1918 }
1919 else if (! NILP (Vload_read_function))
1920 val = call1 (Vload_read_function, readcharfun);
1921 else
1922 val = read_internal_start (readcharfun, Qnil, Qnil);
1923 }
1924
1925 if (!NILP (start) && continue_reading_p)
1926 start = Fpoint_marker ();
1927
1928 /* Restore saved point and BEGV. */
1929 unbind_to (count1, Qnil);
1930
1931 /* Now eval what we just read. */
1932 if (!NILP (macroexpand))
1933 val = call1 (macroexpand, val);
1934 val = eval_sub (val);
1935
1936 if (printflag)
1937 {
1938 Vvalues = Fcons (val, Vvalues);
1939 if (EQ (Vstandard_output, Qt))
1940 Fprin1 (val, Qnil);
1941 else
1942 Fprint (val, Qnil);
1943 }
1944
1945 first_sexp = 0;
1946 }
1947
1948 build_load_history (sourcename,
1949 stream || whole_buffer);
1950
1951 UNGCPRO;
1952
1953 unbind_to (count, Qnil);
1954 }
1955
1956 DEFUN ("eval-buffer", Feval_buffer, Seval_buffer, 0, 5, "",
1957 doc: /* Execute the current buffer as Lisp code.
1958 When called from a Lisp program (i.e., not interactively), this
1959 function accepts up to five optional arguments:
1960 BUFFER is the buffer to evaluate (nil means use current buffer).
1961 PRINTFLAG controls printing of output:
1962 A value of nil means discard it; anything else is stream for print.
1963 FILENAME specifies the file name to use for `load-history'.
1964 UNIBYTE, if non-nil, specifies `load-convert-to-unibyte' for this
1965 invocation.
1966 DO-ALLOW-PRINT, if non-nil, specifies that `print' and related
1967 functions should work normally even if PRINTFLAG is nil.
1968
1969 This function preserves the position of point. */)
1970 (Lisp_Object buffer, Lisp_Object printflag, Lisp_Object filename, Lisp_Object unibyte, Lisp_Object do_allow_print)
1971 {
1972 ptrdiff_t count = SPECPDL_INDEX ();
1973 Lisp_Object tem, buf;
1974
1975 if (NILP (buffer))
1976 buf = Fcurrent_buffer ();
1977 else
1978 buf = Fget_buffer (buffer);
1979 if (NILP (buf))
1980 error ("No such buffer");
1981
1982 if (NILP (printflag) && NILP (do_allow_print))
1983 tem = Qsymbolp;
1984 else
1985 tem = printflag;
1986
1987 if (NILP (filename))
1988 filename = BVAR (XBUFFER (buf), filename);
1989
1990 specbind (Qeval_buffer_list, Fcons (buf, Veval_buffer_list));
1991 specbind (Qstandard_output, tem);
1992 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1993 BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
1994 specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
1995 readevalloop (buf, 0, filename,
1996 !NILP (printflag), unibyte, Qnil, Qnil, Qnil);
1997 unbind_to (count, Qnil);
1998
1999 return Qnil;
2000 }
2001
2002 DEFUN ("eval-region", Feval_region, Seval_region, 2, 4, "r",
2003 doc: /* Execute the region as Lisp code.
2004 When called from programs, expects two arguments,
2005 giving starting and ending indices in the current buffer
2006 of the text to be executed.
2007 Programs can pass third argument PRINTFLAG which controls output:
2008 A value of nil means discard it; anything else is stream for printing it.
2009 Also the fourth argument READ-FUNCTION, if non-nil, is used
2010 instead of `read' to read each expression. It gets one argument
2011 which is the input stream for reading characters.
2012
2013 This function does not move point. */)
2014 (Lisp_Object start, Lisp_Object end, Lisp_Object printflag, Lisp_Object read_function)
2015 {
2016 /* FIXME: Do the eval-sexp-add-defvars dance! */
2017 ptrdiff_t count = SPECPDL_INDEX ();
2018 Lisp_Object tem, cbuf;
2019
2020 cbuf = Fcurrent_buffer ();
2021
2022 if (NILP (printflag))
2023 tem = Qsymbolp;
2024 else
2025 tem = printflag;
2026 specbind (Qstandard_output, tem);
2027 specbind (Qeval_buffer_list, Fcons (cbuf, Veval_buffer_list));
2028
2029 /* `readevalloop' calls functions which check the type of start and end. */
2030 readevalloop (cbuf, 0, BVAR (XBUFFER (cbuf), filename),
2031 !NILP (printflag), Qnil, read_function,
2032 start, end);
2033
2034 return unbind_to (count, Qnil);
2035 }
2036
2037 \f
2038 DEFUN ("read", Fread, Sread, 0, 1, 0,
2039 doc: /* Read one Lisp expression as text from STREAM, return as Lisp object.
2040 If STREAM is nil, use the value of `standard-input' (which see).
2041 STREAM or the value of `standard-input' may be:
2042 a buffer (read from point and advance it)
2043 a marker (read from where it points and advance it)
2044 a function (call it with no arguments for each character,
2045 call it with a char as argument to push a char back)
2046 a string (takes text from string, starting at the beginning)
2047 t (read text line using minibuffer and use it, or read from
2048 standard input in batch mode). */)
2049 (Lisp_Object stream)
2050 {
2051 if (NILP (stream))
2052 stream = Vstandard_input;
2053 if (EQ (stream, Qt))
2054 stream = Qread_char;
2055 if (EQ (stream, Qread_char))
2056 /* FIXME: ?! When is this used !? */
2057 return call1 (intern ("read-minibuffer"),
2058 build_string ("Lisp expression: "));
2059
2060 return read_internal_start (stream, Qnil, Qnil);
2061 }
2062
2063 DEFUN ("read-from-string", Fread_from_string, Sread_from_string, 1, 3, 0,
2064 doc: /* Read one Lisp expression which is represented as text by STRING.
2065 Returns a cons: (OBJECT-READ . FINAL-STRING-INDEX).
2066 FINAL-STRING-INDEX is an integer giving the position of the next
2067 remaining character in STRING.
2068 START and END optionally delimit a substring of STRING from which to read;
2069 they default to 0 and (length STRING) respectively. */)
2070 (Lisp_Object string, Lisp_Object start, Lisp_Object end)
2071 {
2072 Lisp_Object ret;
2073 CHECK_STRING (string);
2074 /* `read_internal_start' sets `read_from_string_index'. */
2075 ret = read_internal_start (string, start, end);
2076 return Fcons (ret, make_number (read_from_string_index));
2077 }
2078
2079 /* Function to set up the global context we need in toplevel read
2080 calls. */
2081 static Lisp_Object
2082 read_internal_start (Lisp_Object stream, Lisp_Object start, Lisp_Object end)
2083 /* `start', `end' only used when stream is a string. */
2084 {
2085 Lisp_Object retval;
2086
2087 readchar_count = 0;
2088 new_backquote_flag = 0;
2089 read_objects = Qnil;
2090 if (EQ (Vread_with_symbol_positions, Qt)
2091 || EQ (Vread_with_symbol_positions, stream))
2092 Vread_symbol_positions_list = Qnil;
2093
2094 if (STRINGP (stream)
2095 || ((CONSP (stream) && STRINGP (XCAR (stream)))))
2096 {
2097 ptrdiff_t startval, endval;
2098 Lisp_Object string;
2099
2100 if (STRINGP (stream))
2101 string = stream;
2102 else
2103 string = XCAR (stream);
2104
2105 if (NILP (end))
2106 endval = SCHARS (string);
2107 else
2108 {
2109 CHECK_NUMBER (end);
2110 if (! (0 <= XINT (end) && XINT (end) <= SCHARS (string)))
2111 args_out_of_range (string, end);
2112 endval = XINT (end);
2113 }
2114
2115 if (NILP (start))
2116 startval = 0;
2117 else
2118 {
2119 CHECK_NUMBER (start);
2120 if (! (0 <= XINT (start) && XINT (start) <= endval))
2121 args_out_of_range (string, start);
2122 startval = XINT (start);
2123 }
2124 read_from_string_index = startval;
2125 read_from_string_index_byte = string_char_to_byte (string, startval);
2126 read_from_string_limit = endval;
2127 }
2128
2129 retval = read0 (stream);
2130 if (EQ (Vread_with_symbol_positions, Qt)
2131 || EQ (Vread_with_symbol_positions, stream))
2132 Vread_symbol_positions_list = Fnreverse (Vread_symbol_positions_list);
2133 return retval;
2134 }
2135 \f
2136
2137 /* Signal Qinvalid_read_syntax error.
2138 S is error string of length N (if > 0) */
2139
2140 static _Noreturn void
2141 invalid_syntax (const char *s)
2142 {
2143 xsignal1 (Qinvalid_read_syntax, build_string (s));
2144 }
2145
2146
2147 /* Use this for recursive reads, in contexts where internal tokens
2148 are not allowed. */
2149
2150 static Lisp_Object
2151 read0 (Lisp_Object readcharfun)
2152 {
2153 register Lisp_Object val;
2154 int c;
2155
2156 val = read1 (readcharfun, &c, 0);
2157 if (!c)
2158 return val;
2159
2160 xsignal1 (Qinvalid_read_syntax,
2161 Fmake_string (make_number (1), make_number (c)));
2162 }
2163 \f
2164 static ptrdiff_t read_buffer_size;
2165 static char *read_buffer;
2166
2167 /* Read a \-escape sequence, assuming we already read the `\'.
2168 If the escape sequence forces unibyte, return eight-bit char. */
2169
2170 static int
2171 read_escape (Lisp_Object readcharfun, bool stringp)
2172 {
2173 int c = READCHAR;
2174 /* \u allows up to four hex digits, \U up to eight. Default to the
2175 behavior for \u, and change this value in the case that \U is seen. */
2176 int unicode_hex_count = 4;
2177
2178 switch (c)
2179 {
2180 case -1:
2181 end_of_file_error ();
2182
2183 case 'a':
2184 return '\007';
2185 case 'b':
2186 return '\b';
2187 case 'd':
2188 return 0177;
2189 case 'e':
2190 return 033;
2191 case 'f':
2192 return '\f';
2193 case 'n':
2194 return '\n';
2195 case 'r':
2196 return '\r';
2197 case 't':
2198 return '\t';
2199 case 'v':
2200 return '\v';
2201 case '\n':
2202 return -1;
2203 case ' ':
2204 if (stringp)
2205 return -1;
2206 return ' ';
2207
2208 case 'M':
2209 c = READCHAR;
2210 if (c != '-')
2211 error ("Invalid escape character syntax");
2212 c = READCHAR;
2213 if (c == '\\')
2214 c = read_escape (readcharfun, 0);
2215 return c | meta_modifier;
2216
2217 case 'S':
2218 c = READCHAR;
2219 if (c != '-')
2220 error ("Invalid escape character syntax");
2221 c = READCHAR;
2222 if (c == '\\')
2223 c = read_escape (readcharfun, 0);
2224 return c | shift_modifier;
2225
2226 case 'H':
2227 c = READCHAR;
2228 if (c != '-')
2229 error ("Invalid escape character syntax");
2230 c = READCHAR;
2231 if (c == '\\')
2232 c = read_escape (readcharfun, 0);
2233 return c | hyper_modifier;
2234
2235 case 'A':
2236 c = READCHAR;
2237 if (c != '-')
2238 error ("Invalid escape character syntax");
2239 c = READCHAR;
2240 if (c == '\\')
2241 c = read_escape (readcharfun, 0);
2242 return c | alt_modifier;
2243
2244 case 's':
2245 c = READCHAR;
2246 if (stringp || c != '-')
2247 {
2248 UNREAD (c);
2249 return ' ';
2250 }
2251 c = READCHAR;
2252 if (c == '\\')
2253 c = read_escape (readcharfun, 0);
2254 return c | super_modifier;
2255
2256 case 'C':
2257 c = READCHAR;
2258 if (c != '-')
2259 error ("Invalid escape character syntax");
2260 case '^':
2261 c = READCHAR;
2262 if (c == '\\')
2263 c = read_escape (readcharfun, 0);
2264 if ((c & ~CHAR_MODIFIER_MASK) == '?')
2265 return 0177 | (c & CHAR_MODIFIER_MASK);
2266 else if (! SINGLE_BYTE_CHAR_P ((c & ~CHAR_MODIFIER_MASK)))
2267 return c | ctrl_modifier;
2268 /* ASCII control chars are made from letters (both cases),
2269 as well as the non-letters within 0100...0137. */
2270 else if ((c & 0137) >= 0101 && (c & 0137) <= 0132)
2271 return (c & (037 | ~0177));
2272 else if ((c & 0177) >= 0100 && (c & 0177) <= 0137)
2273 return (c & (037 | ~0177));
2274 else
2275 return c | ctrl_modifier;
2276
2277 case '0':
2278 case '1':
2279 case '2':
2280 case '3':
2281 case '4':
2282 case '5':
2283 case '6':
2284 case '7':
2285 /* An octal escape, as in ANSI C. */
2286 {
2287 register int i = c - '0';
2288 register int count = 0;
2289 while (++count < 3)
2290 {
2291 if ((c = READCHAR) >= '0' && c <= '7')
2292 {
2293 i *= 8;
2294 i += c - '0';
2295 }
2296 else
2297 {
2298 UNREAD (c);
2299 break;
2300 }
2301 }
2302
2303 if (i >= 0x80 && i < 0x100)
2304 i = BYTE8_TO_CHAR (i);
2305 return i;
2306 }
2307
2308 case 'x':
2309 /* A hex escape, as in ANSI C. */
2310 {
2311 unsigned int i = 0;
2312 int count = 0;
2313 while (1)
2314 {
2315 c = READCHAR;
2316 if (c >= '0' && c <= '9')
2317 {
2318 i *= 16;
2319 i += c - '0';
2320 }
2321 else if ((c >= 'a' && c <= 'f')
2322 || (c >= 'A' && c <= 'F'))
2323 {
2324 i *= 16;
2325 if (c >= 'a' && c <= 'f')
2326 i += c - 'a' + 10;
2327 else
2328 i += c - 'A' + 10;
2329 }
2330 else
2331 {
2332 UNREAD (c);
2333 break;
2334 }
2335 /* Allow hex escapes as large as ?\xfffffff, because some
2336 packages use them to denote characters with modifiers. */
2337 if ((CHAR_META | (CHAR_META - 1)) < i)
2338 error ("Hex character out of range: \\x%x...", i);
2339 count += count < 3;
2340 }
2341
2342 if (count < 3 && i >= 0x80)
2343 return BYTE8_TO_CHAR (i);
2344 return i;
2345 }
2346
2347 case 'U':
2348 /* Post-Unicode-2.0: Up to eight hex chars. */
2349 unicode_hex_count = 8;
2350 case 'u':
2351
2352 /* A Unicode escape. We only permit them in strings and characters,
2353 not arbitrarily in the source code, as in some other languages. */
2354 {
2355 unsigned int i = 0;
2356 int count = 0;
2357
2358 while (++count <= unicode_hex_count)
2359 {
2360 c = READCHAR;
2361 /* `isdigit' and `isalpha' may be locale-specific, which we don't
2362 want. */
2363 if (c >= '0' && c <= '9') i = (i << 4) + (c - '0');
2364 else if (c >= 'a' && c <= 'f') i = (i << 4) + (c - 'a') + 10;
2365 else if (c >= 'A' && c <= 'F') i = (i << 4) + (c - 'A') + 10;
2366 else
2367 error ("Non-hex digit used for Unicode escape");
2368 }
2369 if (i > 0x10FFFF)
2370 error ("Non-Unicode character: 0x%x", i);
2371 return i;
2372 }
2373
2374 default:
2375 return c;
2376 }
2377 }
2378
2379 /* Return the digit that CHARACTER stands for in the given BASE.
2380 Return -1 if CHARACTER is out of range for BASE,
2381 and -2 if CHARACTER is not valid for any supported BASE. */
2382 static int
2383 digit_to_number (int character, int base)
2384 {
2385 int digit;
2386
2387 if ('0' <= character && character <= '9')
2388 digit = character - '0';
2389 else if ('a' <= character && character <= 'z')
2390 digit = character - 'a' + 10;
2391 else if ('A' <= character && character <= 'Z')
2392 digit = character - 'A' + 10;
2393 else
2394 return -2;
2395
2396 return digit < base ? digit : -1;
2397 }
2398
2399 /* Read an integer in radix RADIX using READCHARFUN to read
2400 characters. RADIX must be in the interval [2..36]; if it isn't, a
2401 read error is signaled . Value is the integer read. Signals an
2402 error if encountering invalid read syntax or if RADIX is out of
2403 range. */
2404
2405 static Lisp_Object
2406 read_integer (Lisp_Object readcharfun, EMACS_INT radix)
2407 {
2408 /* Room for sign, leading 0, other digits, trailing null byte.
2409 Also, room for invalid syntax diagnostic. */
2410 char buf[max (1 + 1 + sizeof (uintmax_t) * CHAR_BIT + 1,
2411 sizeof "integer, radix " + INT_STRLEN_BOUND (EMACS_INT))];
2412
2413 int valid = -1; /* 1 if valid, 0 if not, -1 if incomplete. */
2414
2415 if (radix < 2 || radix > 36)
2416 valid = 0;
2417 else
2418 {
2419 char *p = buf;
2420 int c, digit;
2421
2422 c = READCHAR;
2423 if (c == '-' || c == '+')
2424 {
2425 *p++ = c;
2426 c = READCHAR;
2427 }
2428
2429 if (c == '0')
2430 {
2431 *p++ = c;
2432 valid = 1;
2433
2434 /* Ignore redundant leading zeros, so the buffer doesn't
2435 fill up with them. */
2436 do
2437 c = READCHAR;
2438 while (c == '0');
2439 }
2440
2441 while ((digit = digit_to_number (c, radix)) >= -1)
2442 {
2443 if (digit == -1)
2444 valid = 0;
2445 if (valid < 0)
2446 valid = 1;
2447
2448 if (p < buf + sizeof buf - 1)
2449 *p++ = c;
2450 else
2451 valid = 0;
2452
2453 c = READCHAR;
2454 }
2455
2456 UNREAD (c);
2457 *p = '\0';
2458 }
2459
2460 if (! valid)
2461 {
2462 sprintf (buf, "integer, radix %"pI"d", radix);
2463 invalid_syntax (buf);
2464 }
2465
2466 return string_to_number (buf, radix, 0);
2467 }
2468
2469
2470 /* If the next token is ')' or ']' or '.', we store that character
2471 in *PCH and the return value is not interesting. Else, we store
2472 zero in *PCH and we read and return one lisp object.
2473
2474 FIRST_IN_LIST is true if this is the first element of a list. */
2475
2476 static Lisp_Object
2477 read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
2478 {
2479 int c;
2480 bool uninterned_symbol = 0;
2481 bool multibyte;
2482
2483 *pch = 0;
2484
2485 retry:
2486
2487 c = READCHAR_REPORT_MULTIBYTE (&multibyte);
2488 if (c < 0)
2489 end_of_file_error ();
2490
2491 switch (c)
2492 {
2493 case '(':
2494 return read_list (0, readcharfun);
2495
2496 case '[':
2497 return read_vector (readcharfun, 0);
2498
2499 case ')':
2500 case ']':
2501 {
2502 *pch = c;
2503 return Qnil;
2504 }
2505
2506 case '#':
2507 c = READCHAR;
2508 if (c == 's')
2509 {
2510 c = READCHAR;
2511 if (c == '(')
2512 {
2513 /* Accept extended format for hashtables (extensible to
2514 other types), e.g.
2515 #s(hash-table size 2 test equal data (k1 v1 k2 v2)) */
2516 Lisp_Object tmp = read_list (0, readcharfun);
2517 Lisp_Object head = CAR_SAFE (tmp);
2518 Lisp_Object data = Qnil;
2519 Lisp_Object val = Qnil;
2520 /* The size is 2 * number of allowed keywords to
2521 make-hash-table. */
2522 Lisp_Object params[10];
2523 Lisp_Object ht;
2524 Lisp_Object key = Qnil;
2525 int param_count = 0;
2526
2527 if (!EQ (head, Qhash_table))
2528 error ("Invalid extended read marker at head of #s list "
2529 "(only hash-table allowed)");
2530
2531 tmp = CDR_SAFE (tmp);
2532
2533 /* This is repetitive but fast and simple. */
2534 params[param_count] = QCsize;
2535 params[param_count + 1] = Fplist_get (tmp, Qsize);
2536 if (!NILP (params[param_count + 1]))
2537 param_count += 2;
2538
2539 params[param_count] = QCtest;
2540 params[param_count + 1] = Fplist_get (tmp, Qtest);
2541 if (!NILP (params[param_count + 1]))
2542 param_count += 2;
2543
2544 params[param_count] = QCweakness;
2545 params[param_count + 1] = Fplist_get (tmp, Qweakness);
2546 if (!NILP (params[param_count + 1]))
2547 param_count += 2;
2548
2549 params[param_count] = QCrehash_size;
2550 params[param_count + 1] = Fplist_get (tmp, Qrehash_size);
2551 if (!NILP (params[param_count + 1]))
2552 param_count += 2;
2553
2554 params[param_count] = QCrehash_threshold;
2555 params[param_count + 1] = Fplist_get (tmp, Qrehash_threshold);
2556 if (!NILP (params[param_count + 1]))
2557 param_count += 2;
2558
2559 /* This is the hashtable data. */
2560 data = Fplist_get (tmp, Qdata);
2561
2562 /* Now use params to make a new hashtable and fill it. */
2563 ht = Fmake_hash_table (param_count, params);
2564
2565 while (CONSP (data))
2566 {
2567 key = XCAR (data);
2568 data = XCDR (data);
2569 if (!CONSP (data))
2570 error ("Odd number of elements in hashtable data");
2571 val = XCAR (data);
2572 data = XCDR (data);
2573 Fputhash (key, val, ht);
2574 }
2575
2576 return ht;
2577 }
2578 UNREAD (c);
2579 invalid_syntax ("#");
2580 }
2581 if (c == '^')
2582 {
2583 c = READCHAR;
2584 if (c == '[')
2585 {
2586 Lisp_Object tmp;
2587 tmp = read_vector (readcharfun, 0);
2588 if (ASIZE (tmp) < CHAR_TABLE_STANDARD_SLOTS)
2589 error ("Invalid size char-table");
2590 XSETPVECTYPE (XVECTOR (tmp), PVEC_CHAR_TABLE);
2591 return tmp;
2592 }
2593 else if (c == '^')
2594 {
2595 c = READCHAR;
2596 if (c == '[')
2597 {
2598 Lisp_Object tmp;
2599 int depth;
2600 ptrdiff_t size;
2601
2602 tmp = read_vector (readcharfun, 0);
2603 size = ASIZE (tmp);
2604 if (size == 0)
2605 error ("Invalid size char-table");
2606 if (! RANGED_INTEGERP (1, AREF (tmp, 0), 3))
2607 error ("Invalid depth in char-table");
2608 depth = XINT (AREF (tmp, 0));
2609 if (chartab_size[depth] != size - 2)
2610 error ("Invalid size char-table");
2611 XSETPVECTYPE (XVECTOR (tmp), PVEC_SUB_CHAR_TABLE);
2612 return tmp;
2613 }
2614 invalid_syntax ("#^^");
2615 }
2616 invalid_syntax ("#^");
2617 }
2618 if (c == '&')
2619 {
2620 Lisp_Object length;
2621 length = read1 (readcharfun, pch, first_in_list);
2622 c = READCHAR;
2623 if (c == '"')
2624 {
2625 Lisp_Object tmp, val;
2626 EMACS_INT size_in_chars = bool_vector_bytes (XFASTINT (length));
2627 unsigned char *data;
2628
2629 UNREAD (c);
2630 tmp = read1 (readcharfun, pch, first_in_list);
2631 if (STRING_MULTIBYTE (tmp)
2632 || (size_in_chars != SCHARS (tmp)
2633 /* We used to print 1 char too many
2634 when the number of bits was a multiple of 8.
2635 Accept such input in case it came from an old
2636 version. */
2637 && ! (XFASTINT (length)
2638 == (SCHARS (tmp) - 1) * BOOL_VECTOR_BITS_PER_CHAR)))
2639 invalid_syntax ("#&...");
2640
2641 val = make_uninit_bool_vector (XFASTINT (length));
2642 data = bool_vector_uchar_data (val);
2643 memcpy (data, SDATA (tmp), size_in_chars);
2644 /* Clear the extraneous bits in the last byte. */
2645 if (XINT (length) != size_in_chars * BOOL_VECTOR_BITS_PER_CHAR)
2646 data[size_in_chars - 1]
2647 &= (1 << (XINT (length) % BOOL_VECTOR_BITS_PER_CHAR)) - 1;
2648 return val;
2649 }
2650 invalid_syntax ("#&...");
2651 }
2652 if (c == '[')
2653 {
2654 /* Accept compiled functions at read-time so that we don't have to
2655 build them using function calls. */
2656 Lisp_Object tmp;
2657 struct Lisp_Vector *vec;
2658 tmp = read_vector (readcharfun, 1);
2659 vec = XVECTOR (tmp);
2660 if (vec->header.size == 0)
2661 invalid_syntax ("Empty byte-code object");
2662 make_byte_code (vec);
2663 return tmp;
2664 }
2665 if (c == '(')
2666 {
2667 Lisp_Object tmp;
2668 struct gcpro gcpro1;
2669 int ch;
2670
2671 /* Read the string itself. */
2672 tmp = read1 (readcharfun, &ch, 0);
2673 if (ch != 0 || !STRINGP (tmp))
2674 invalid_syntax ("#");
2675 GCPRO1 (tmp);
2676 /* Read the intervals and their properties. */
2677 while (1)
2678 {
2679 Lisp_Object beg, end, plist;
2680
2681 beg = read1 (readcharfun, &ch, 0);
2682 end = plist = Qnil;
2683 if (ch == ')')
2684 break;
2685 if (ch == 0)
2686 end = read1 (readcharfun, &ch, 0);
2687 if (ch == 0)
2688 plist = read1 (readcharfun, &ch, 0);
2689 if (ch)
2690 invalid_syntax ("Invalid string property list");
2691 Fset_text_properties (beg, end, plist, tmp);
2692 }
2693 UNGCPRO;
2694 return tmp;
2695 }
2696
2697 /* #@NUMBER is used to skip NUMBER following bytes.
2698 That's used in .elc files to skip over doc strings
2699 and function definitions. */
2700 if (c == '@')
2701 {
2702 enum { extra = 100 };
2703 ptrdiff_t i, nskip = 0, digits = 0;
2704
2705 /* Read a decimal integer. */
2706 while ((c = READCHAR) >= 0
2707 && c >= '0' && c <= '9')
2708 {
2709 if ((STRING_BYTES_BOUND - extra) / 10 <= nskip)
2710 string_overflow ();
2711 digits++;
2712 nskip *= 10;
2713 nskip += c - '0';
2714 if (digits == 2 && nskip == 0)
2715 { /* We've just seen #@00, which means "skip to end". */
2716 skip_dyn_eof (readcharfun);
2717 return Qnil;
2718 }
2719 }
2720 if (nskip > 0)
2721 /* We can't use UNREAD here, because in the code below we side-step
2722 READCHAR. Instead, assume the first char after #@NNN occupies
2723 a single byte, which is the case normally since it's just
2724 a space. */
2725 nskip--;
2726 else
2727 UNREAD (c);
2728
2729 if (load_force_doc_strings
2730 && (FROM_FILE_P (readcharfun)))
2731 {
2732 /* If we are supposed to force doc strings into core right now,
2733 record the last string that we skipped,
2734 and record where in the file it comes from. */
2735
2736 /* But first exchange saved_doc_string
2737 with prev_saved_doc_string, so we save two strings. */
2738 {
2739 char *temp = saved_doc_string;
2740 ptrdiff_t temp_size = saved_doc_string_size;
2741 file_offset temp_pos = saved_doc_string_position;
2742 ptrdiff_t temp_len = saved_doc_string_length;
2743
2744 saved_doc_string = prev_saved_doc_string;
2745 saved_doc_string_size = prev_saved_doc_string_size;
2746 saved_doc_string_position = prev_saved_doc_string_position;
2747 saved_doc_string_length = prev_saved_doc_string_length;
2748
2749 prev_saved_doc_string = temp;
2750 prev_saved_doc_string_size = temp_size;
2751 prev_saved_doc_string_position = temp_pos;
2752 prev_saved_doc_string_length = temp_len;
2753 }
2754
2755 if (saved_doc_string_size == 0)
2756 {
2757 saved_doc_string = xmalloc (nskip + extra);
2758 saved_doc_string_size = nskip + extra;
2759 }
2760 if (nskip > saved_doc_string_size)
2761 {
2762 saved_doc_string = xrealloc (saved_doc_string, nskip + extra);
2763 saved_doc_string_size = nskip + extra;
2764 }
2765
2766 saved_doc_string_position = file_tell (instream);
2767
2768 /* Copy that many characters into saved_doc_string. */
2769 block_input ();
2770 for (i = 0; i < nskip && c >= 0; i++)
2771 saved_doc_string[i] = c = getc (instream);
2772 unblock_input ();
2773
2774 saved_doc_string_length = i;
2775 }
2776 else
2777 /* Skip that many bytes. */
2778 skip_dyn_bytes (readcharfun, nskip);
2779
2780 goto retry;
2781 }
2782 if (c == '!')
2783 {
2784 /* #! appears at the beginning of an executable file.
2785 Skip the first line. */
2786 while (c != '\n' && c >= 0)
2787 c = READCHAR;
2788 goto retry;
2789 }
2790 if (c == '$')
2791 return Vload_file_name;
2792 if (c == '\'')
2793 return list2 (Qfunction, read0 (readcharfun));
2794 /* #:foo is the uninterned symbol named foo. */
2795 if (c == ':')
2796 {
2797 uninterned_symbol = 1;
2798 c = READCHAR;
2799 if (!(c > 040
2800 && c != 0xa0 /* NBSP */
2801 && (c >= 0200
2802 || strchr ("\"';()[]#`,", c) == NULL)))
2803 {
2804 /* No symbol character follows, this is the empty
2805 symbol. */
2806 UNREAD (c);
2807 return Fmake_symbol (empty_unibyte_string);
2808 }
2809 goto read_symbol;
2810 }
2811 /* ## is the empty symbol. */
2812 if (c == '#')
2813 return Fintern (empty_unibyte_string, Qnil);
2814 /* Reader forms that can reuse previously read objects. */
2815 if (c >= '0' && c <= '9')
2816 {
2817 EMACS_INT n = 0;
2818 Lisp_Object tem;
2819
2820 /* Read a non-negative integer. */
2821 while (c >= '0' && c <= '9')
2822 {
2823 if (MOST_POSITIVE_FIXNUM / 10 < n
2824 || MOST_POSITIVE_FIXNUM < n * 10 + c - '0')
2825 n = MOST_POSITIVE_FIXNUM + 1;
2826 else
2827 n = n * 10 + c - '0';
2828 c = READCHAR;
2829 }
2830
2831 if (n <= MOST_POSITIVE_FIXNUM)
2832 {
2833 if (c == 'r' || c == 'R')
2834 return read_integer (readcharfun, n);
2835
2836 if (! NILP (Vread_circle))
2837 {
2838 /* #n=object returns object, but associates it with
2839 n for #n#. */
2840 if (c == '=')
2841 {
2842 /* Make a placeholder for #n# to use temporarily. */
2843 Lisp_Object placeholder;
2844 Lisp_Object cell;
2845
2846 placeholder = Fcons (Qnil, Qnil);
2847 cell = Fcons (make_number (n), placeholder);
2848 read_objects = Fcons (cell, read_objects);
2849
2850 /* Read the object itself. */
2851 tem = read0 (readcharfun);
2852
2853 /* Now put it everywhere the placeholder was... */
2854 substitute_object_in_subtree (tem, placeholder);
2855
2856 /* ...and #n# will use the real value from now on. */
2857 Fsetcdr (cell, tem);
2858
2859 return tem;
2860 }
2861
2862 /* #n# returns a previously read object. */
2863 if (c == '#')
2864 {
2865 tem = Fassq (make_number (n), read_objects);
2866 if (CONSP (tem))
2867 return XCDR (tem);
2868 }
2869 }
2870 }
2871 /* Fall through to error message. */
2872 }
2873 else if (c == 'x' || c == 'X')
2874 return read_integer (readcharfun, 16);
2875 else if (c == 'o' || c == 'O')
2876 return read_integer (readcharfun, 8);
2877 else if (c == 'b' || c == 'B')
2878 return read_integer (readcharfun, 2);
2879
2880 UNREAD (c);
2881 invalid_syntax ("#");
2882
2883 case ';':
2884 while ((c = READCHAR) >= 0 && c != '\n');
2885 goto retry;
2886
2887 case '\'':
2888 return list2 (Qquote, read0 (readcharfun));
2889
2890 case '`':
2891 {
2892 int next_char = READCHAR;
2893 UNREAD (next_char);
2894 /* Transition from old-style to new-style:
2895 If we see "(`" it used to mean old-style, which usually works
2896 fine because ` should almost never appear in such a position
2897 for new-style. But occasionally we need "(`" to mean new
2898 style, so we try to distinguish the two by the fact that we
2899 can either write "( `foo" or "(` foo", where the first
2900 intends to use new-style whereas the second intends to use
2901 old-style. For Emacs-25, we should completely remove this
2902 first_in_list exception (old-style can still be obtained via
2903 "(\`" anyway). */
2904 if (!new_backquote_flag && first_in_list && next_char == ' ')
2905 {
2906 Vold_style_backquotes = Qt;
2907 goto default_label;
2908 }
2909 else
2910 {
2911 Lisp_Object value;
2912 bool saved_new_backquote_flag = new_backquote_flag;
2913
2914 new_backquote_flag = 1;
2915 value = read0 (readcharfun);
2916 new_backquote_flag = saved_new_backquote_flag;
2917
2918 return list2 (Qbackquote, value);
2919 }
2920 }
2921 case ',':
2922 {
2923 int next_char = READCHAR;
2924 UNREAD (next_char);
2925 /* Transition from old-style to new-style:
2926 It used to be impossible to have a new-style , other than within
2927 a new-style `. This is sufficient when ` and , are used in the
2928 normal way, but ` and , can also appear in args to macros that
2929 will not interpret them in the usual way, in which case , may be
2930 used without any ` anywhere near.
2931 So we now use the same heuristic as for backquote: old-style
2932 unquotes are only recognized when first on a list, and when
2933 followed by a space.
2934 Because it's more difficult to peek 2 chars ahead, a new-style
2935 ,@ can still not be used outside of a `, unless it's in the middle
2936 of a list. */
2937 if (new_backquote_flag
2938 || !first_in_list
2939 || (next_char != ' ' && next_char != '@'))
2940 {
2941 Lisp_Object comma_type = Qnil;
2942 Lisp_Object value;
2943 int ch = READCHAR;
2944
2945 if (ch == '@')
2946 comma_type = Qcomma_at;
2947 else if (ch == '.')
2948 comma_type = Qcomma_dot;
2949 else
2950 {
2951 if (ch >= 0) UNREAD (ch);
2952 comma_type = Qcomma;
2953 }
2954
2955 value = read0 (readcharfun);
2956 return list2 (comma_type, value);
2957 }
2958 else
2959 {
2960 Vold_style_backquotes = Qt;
2961 goto default_label;
2962 }
2963 }
2964 case '?':
2965 {
2966 int modifiers;
2967 int next_char;
2968 bool ok;
2969
2970 c = READCHAR;
2971 if (c < 0)
2972 end_of_file_error ();
2973
2974 /* Accept `single space' syntax like (list ? x) where the
2975 whitespace character is SPC or TAB.
2976 Other literal whitespace like NL, CR, and FF are not accepted,
2977 as there are well-established escape sequences for these. */
2978 if (c == ' ' || c == '\t')
2979 return make_number (c);
2980
2981 if (c == '\\')
2982 c = read_escape (readcharfun, 0);
2983 modifiers = c & CHAR_MODIFIER_MASK;
2984 c &= ~CHAR_MODIFIER_MASK;
2985 if (CHAR_BYTE8_P (c))
2986 c = CHAR_TO_BYTE8 (c);
2987 c |= modifiers;
2988
2989 next_char = READCHAR;
2990 ok = (next_char <= 040
2991 || (next_char < 0200
2992 && strchr ("\"';()[]#?`,.", next_char) != NULL));
2993 UNREAD (next_char);
2994 if (ok)
2995 return make_number (c);
2996
2997 invalid_syntax ("?");
2998 }
2999
3000 case '"':
3001 {
3002 char *p = read_buffer;
3003 char *end = read_buffer + read_buffer_size;
3004 int ch;
3005 /* True if we saw an escape sequence specifying
3006 a multibyte character. */
3007 bool force_multibyte = 0;
3008 /* True if we saw an escape sequence specifying
3009 a single-byte character. */
3010 bool force_singlebyte = 0;
3011 bool cancel = 0;
3012 ptrdiff_t nchars = 0;
3013
3014 while ((ch = READCHAR) >= 0
3015 && ch != '\"')
3016 {
3017 if (end - p < MAX_MULTIBYTE_LENGTH)
3018 {
3019 ptrdiff_t offset = p - read_buffer;
3020 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3021 memory_full (SIZE_MAX);
3022 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3023 read_buffer_size *= 2;
3024 p = read_buffer + offset;
3025 end = read_buffer + read_buffer_size;
3026 }
3027
3028 if (ch == '\\')
3029 {
3030 int modifiers;
3031
3032 ch = read_escape (readcharfun, 1);
3033
3034 /* CH is -1 if \ newline has just been seen. */
3035 if (ch == -1)
3036 {
3037 if (p == read_buffer)
3038 cancel = 1;
3039 continue;
3040 }
3041
3042 modifiers = ch & CHAR_MODIFIER_MASK;
3043 ch = ch & ~CHAR_MODIFIER_MASK;
3044
3045 if (CHAR_BYTE8_P (ch))
3046 force_singlebyte = 1;
3047 else if (! ASCII_CHAR_P (ch))
3048 force_multibyte = 1;
3049 else /* I.e. ASCII_CHAR_P (ch). */
3050 {
3051 /* Allow `\C- ' and `\C-?'. */
3052 if (modifiers == CHAR_CTL)
3053 {
3054 if (ch == ' ')
3055 ch = 0, modifiers = 0;
3056 else if (ch == '?')
3057 ch = 127, modifiers = 0;
3058 }
3059 if (modifiers & CHAR_SHIFT)
3060 {
3061 /* Shift modifier is valid only with [A-Za-z]. */
3062 if (ch >= 'A' && ch <= 'Z')
3063 modifiers &= ~CHAR_SHIFT;
3064 else if (ch >= 'a' && ch <= 'z')
3065 ch -= ('a' - 'A'), modifiers &= ~CHAR_SHIFT;
3066 }
3067
3068 if (modifiers & CHAR_META)
3069 {
3070 /* Move the meta bit to the right place for a
3071 string. */
3072 modifiers &= ~CHAR_META;
3073 ch = BYTE8_TO_CHAR (ch | 0x80);
3074 force_singlebyte = 1;
3075 }
3076 }
3077
3078 /* Any modifiers remaining are invalid. */
3079 if (modifiers)
3080 error ("Invalid modifier in string");
3081 p += CHAR_STRING (ch, (unsigned char *) p);
3082 }
3083 else
3084 {
3085 p += CHAR_STRING (ch, (unsigned char *) p);
3086 if (CHAR_BYTE8_P (ch))
3087 force_singlebyte = 1;
3088 else if (! ASCII_CHAR_P (ch))
3089 force_multibyte = 1;
3090 }
3091 nchars++;
3092 }
3093
3094 if (ch < 0)
3095 end_of_file_error ();
3096
3097 /* If purifying, and string starts with \ newline,
3098 return zero instead. This is for doc strings
3099 that we are really going to find in etc/DOC.nn.nn. */
3100 if (!NILP (Vpurify_flag) && NILP (Vdoc_file_name) && cancel)
3101 return make_number (0);
3102
3103 if (! force_multibyte && force_singlebyte)
3104 {
3105 /* READ_BUFFER contains raw 8-bit bytes and no multibyte
3106 forms. Convert it to unibyte. */
3107 nchars = str_as_unibyte ((unsigned char *) read_buffer,
3108 p - read_buffer);
3109 p = read_buffer + nchars;
3110 }
3111
3112 return make_specified_string (read_buffer, nchars, p - read_buffer,
3113 (force_multibyte
3114 || (p - read_buffer != nchars)));
3115 }
3116
3117 case '.':
3118 {
3119 int next_char = READCHAR;
3120 UNREAD (next_char);
3121
3122 if (next_char <= 040
3123 || (next_char < 0200
3124 && strchr ("\"';([#?`,", next_char) != NULL))
3125 {
3126 *pch = c;
3127 return Qnil;
3128 }
3129
3130 /* Otherwise, we fall through! Note that the atom-reading loop
3131 below will now loop at least once, assuring that we will not
3132 try to UNREAD two characters in a row. */
3133 }
3134 default:
3135 default_label:
3136 if (c <= 040) goto retry;
3137 if (c == 0xa0) /* NBSP */
3138 goto retry;
3139
3140 read_symbol:
3141 {
3142 char *p = read_buffer;
3143 bool quoted = 0;
3144 EMACS_INT start_position = readchar_count - 1;
3145
3146 {
3147 char *end = read_buffer + read_buffer_size;
3148
3149 do
3150 {
3151 if (end - p < MAX_MULTIBYTE_LENGTH)
3152 {
3153 ptrdiff_t offset = p - read_buffer;
3154 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3155 memory_full (SIZE_MAX);
3156 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3157 read_buffer_size *= 2;
3158 p = read_buffer + offset;
3159 end = read_buffer + read_buffer_size;
3160 }
3161
3162 if (c == '\\')
3163 {
3164 c = READCHAR;
3165 if (c == -1)
3166 end_of_file_error ();
3167 quoted = 1;
3168 }
3169
3170 if (multibyte)
3171 p += CHAR_STRING (c, (unsigned char *) p);
3172 else
3173 *p++ = c;
3174 c = READCHAR;
3175 }
3176 while (c > 040
3177 && c != 0xa0 /* NBSP */
3178 && (c >= 0200
3179 || strchr ("\"';()[]#`,", c) == NULL));
3180
3181 if (p == end)
3182 {
3183 ptrdiff_t offset = p - read_buffer;
3184 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3185 memory_full (SIZE_MAX);
3186 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3187 read_buffer_size *= 2;
3188 p = read_buffer + offset;
3189 end = read_buffer + read_buffer_size;
3190 }
3191 *p = 0;
3192 UNREAD (c);
3193 }
3194
3195 if (!quoted && !uninterned_symbol)
3196 {
3197 Lisp_Object result = string_to_number (read_buffer, 10, 0);
3198 if (! NILP (result))
3199 return result;
3200 }
3201 {
3202 Lisp_Object name, result;
3203 ptrdiff_t nbytes = p - read_buffer;
3204 ptrdiff_t nchars
3205 = (multibyte
3206 ? multibyte_chars_in_text ((unsigned char *) read_buffer,
3207 nbytes)
3208 : nbytes);
3209
3210 name = ((uninterned_symbol && ! NILP (Vpurify_flag)
3211 ? make_pure_string : make_specified_string)
3212 (read_buffer, nchars, nbytes, multibyte));
3213 result = (uninterned_symbol ? Fmake_symbol (name)
3214 : Fintern (name, Qnil));
3215
3216 if (EQ (Vread_with_symbol_positions, Qt)
3217 || EQ (Vread_with_symbol_positions, readcharfun))
3218 Vread_symbol_positions_list
3219 = Fcons (Fcons (result, make_number (start_position)),
3220 Vread_symbol_positions_list);
3221 return result;
3222 }
3223 }
3224 }
3225 }
3226 \f
3227
3228 /* List of nodes we've seen during substitute_object_in_subtree. */
3229 static Lisp_Object seen_list;
3230
3231 static void
3232 substitute_object_in_subtree (Lisp_Object object, Lisp_Object placeholder)
3233 {
3234 Lisp_Object check_object;
3235
3236 /* We haven't seen any objects when we start. */
3237 seen_list = Qnil;
3238
3239 /* Make all the substitutions. */
3240 check_object
3241 = substitute_object_recurse (object, placeholder, object);
3242
3243 /* Clear seen_list because we're done with it. */
3244 seen_list = Qnil;
3245
3246 /* The returned object here is expected to always eq the
3247 original. */
3248 if (!EQ (check_object, object))
3249 error ("Unexpected mutation error in reader");
3250 }
3251
3252 /* Feval doesn't get called from here, so no gc protection is needed. */
3253 #define SUBSTITUTE(get_val, set_val) \
3254 do { \
3255 Lisp_Object old_value = get_val; \
3256 Lisp_Object true_value \
3257 = substitute_object_recurse (object, placeholder, \
3258 old_value); \
3259 \
3260 if (!EQ (old_value, true_value)) \
3261 { \
3262 set_val; \
3263 } \
3264 } while (0)
3265
3266 static Lisp_Object
3267 substitute_object_recurse (Lisp_Object object, Lisp_Object placeholder, Lisp_Object subtree)
3268 {
3269 /* If we find the placeholder, return the target object. */
3270 if (EQ (placeholder, subtree))
3271 return object;
3272
3273 /* If we've been to this node before, don't explore it again. */
3274 if (!EQ (Qnil, Fmemq (subtree, seen_list)))
3275 return subtree;
3276
3277 /* If this node can be the entry point to a cycle, remember that
3278 we've seen it. It can only be such an entry point if it was made
3279 by #n=, which means that we can find it as a value in
3280 read_objects. */
3281 if (!EQ (Qnil, Frassq (subtree, read_objects)))
3282 seen_list = Fcons (subtree, seen_list);
3283
3284 /* Recurse according to subtree's type.
3285 Every branch must return a Lisp_Object. */
3286 switch (XTYPE (subtree))
3287 {
3288 case Lisp_Vectorlike:
3289 {
3290 ptrdiff_t i, length = 0;
3291 if (BOOL_VECTOR_P (subtree))
3292 return subtree; /* No sub-objects anyway. */
3293 else if (CHAR_TABLE_P (subtree) || SUB_CHAR_TABLE_P (subtree)
3294 || COMPILEDP (subtree) || HASH_TABLE_P (subtree))
3295 length = ASIZE (subtree) & PSEUDOVECTOR_SIZE_MASK;
3296 else if (VECTORP (subtree))
3297 length = ASIZE (subtree);
3298 else
3299 /* An unknown pseudovector may contain non-Lisp fields, so we
3300 can't just blindly traverse all its fields. We used to call
3301 `Flength' which signaled `sequencep', so I just preserved this
3302 behavior. */
3303 wrong_type_argument (Qsequencep, subtree);
3304
3305 for (i = 0; i < length; i++)
3306 SUBSTITUTE (AREF (subtree, i),
3307 ASET (subtree, i, true_value));
3308 return subtree;
3309 }
3310
3311 case Lisp_Cons:
3312 {
3313 SUBSTITUTE (XCAR (subtree),
3314 XSETCAR (subtree, true_value));
3315 SUBSTITUTE (XCDR (subtree),
3316 XSETCDR (subtree, true_value));
3317 return subtree;
3318 }
3319
3320 case Lisp_String:
3321 {
3322 /* Check for text properties in each interval.
3323 substitute_in_interval contains part of the logic. */
3324
3325 INTERVAL root_interval = string_intervals (subtree);
3326 Lisp_Object arg = Fcons (object, placeholder);
3327
3328 traverse_intervals_noorder (root_interval,
3329 &substitute_in_interval, arg);
3330
3331 return subtree;
3332 }
3333
3334 /* Other types don't recurse any further. */
3335 default:
3336 return subtree;
3337 }
3338 }
3339
3340 /* Helper function for substitute_object_recurse. */
3341 static void
3342 substitute_in_interval (INTERVAL interval, Lisp_Object arg)
3343 {
3344 Lisp_Object object = Fcar (arg);
3345 Lisp_Object placeholder = Fcdr (arg);
3346
3347 SUBSTITUTE (interval->plist, set_interval_plist (interval, true_value));
3348 }
3349
3350 \f
3351 #define LEAD_INT 1
3352 #define DOT_CHAR 2
3353 #define TRAIL_INT 4
3354 #define E_EXP 16
3355
3356
3357 /* Convert STRING to a number, assuming base BASE. Return a fixnum if CP has
3358 integer syntax and fits in a fixnum, else return the nearest float if CP has
3359 either floating point or integer syntax and BASE is 10, else return nil. If
3360 IGNORE_TRAILING, consider just the longest prefix of CP that has
3361 valid floating point syntax. Signal an overflow if BASE is not 10 and the
3362 number has integer syntax but does not fit. */
3363
3364 Lisp_Object
3365 string_to_number (char const *string, int base, bool ignore_trailing)
3366 {
3367 int state;
3368 char const *cp = string;
3369 int leading_digit;
3370 bool float_syntax = 0;
3371 double value = 0;
3372
3373 /* Compute NaN and infinities using a variable, to cope with compilers that
3374 think they are smarter than we are. */
3375 double zero = 0;
3376
3377 /* Negate the value ourselves. This treats 0, NaNs, and infinity properly on
3378 IEEE floating point hosts, and works around a formerly-common bug where
3379 atof ("-0.0") drops the sign. */
3380 bool negative = *cp == '-';
3381
3382 bool signedp = negative || *cp == '+';
3383 cp += signedp;
3384
3385 state = 0;
3386
3387 leading_digit = digit_to_number (*cp, base);
3388 if (leading_digit >= 0)
3389 {
3390 state |= LEAD_INT;
3391 do
3392 ++cp;
3393 while (digit_to_number (*cp, base) >= 0);
3394 }
3395 if (*cp == '.')
3396 {
3397 state |= DOT_CHAR;
3398 cp++;
3399 }
3400
3401 if (base == 10)
3402 {
3403 if ('0' <= *cp && *cp <= '9')
3404 {
3405 state |= TRAIL_INT;
3406 do
3407 cp++;
3408 while ('0' <= *cp && *cp <= '9');
3409 }
3410 if (*cp == 'e' || *cp == 'E')
3411 {
3412 char const *ecp = cp;
3413 cp++;
3414 if (*cp == '+' || *cp == '-')
3415 cp++;
3416 if ('0' <= *cp && *cp <= '9')
3417 {
3418 state |= E_EXP;
3419 do
3420 cp++;
3421 while ('0' <= *cp && *cp <= '9');
3422 }
3423 else if (cp[-1] == '+'
3424 && cp[0] == 'I' && cp[1] == 'N' && cp[2] == 'F')
3425 {
3426 state |= E_EXP;
3427 cp += 3;
3428 value = 1.0 / zero;
3429 }
3430 else if (cp[-1] == '+'
3431 && cp[0] == 'N' && cp[1] == 'a' && cp[2] == 'N')
3432 {
3433 state |= E_EXP;
3434 cp += 3;
3435 value = zero / zero;
3436
3437 /* If that made a "negative" NaN, negate it. */
3438 {
3439 int i;
3440 union { double d; char c[sizeof (double)]; }
3441 u_data, u_minus_zero;
3442 u_data.d = value;
3443 u_minus_zero.d = -0.0;
3444 for (i = 0; i < sizeof (double); i++)
3445 if (u_data.c[i] & u_minus_zero.c[i])
3446 {
3447 value = -value;
3448 break;
3449 }
3450 }
3451 /* Now VALUE is a positive NaN. */
3452 }
3453 else
3454 cp = ecp;
3455 }
3456
3457 float_syntax = ((state & (DOT_CHAR|TRAIL_INT)) == (DOT_CHAR|TRAIL_INT)
3458 || state == (LEAD_INT|E_EXP));
3459 }
3460
3461 /* Return nil if the number uses invalid syntax. If IGNORE_TRAILING, accept
3462 any prefix that matches. Otherwise, the entire string must match. */
3463 if (! (ignore_trailing
3464 ? ((state & LEAD_INT) != 0 || float_syntax)
3465 : (!*cp && ((state & ~DOT_CHAR) == LEAD_INT || float_syntax))))
3466 return Qnil;
3467
3468 /* If the number uses integer and not float syntax, and is in C-language
3469 range, use its value, preferably as a fixnum. */
3470 if (leading_digit >= 0 && ! float_syntax)
3471 {
3472 uintmax_t n;
3473
3474 /* Fast special case for single-digit integers. This also avoids a
3475 glitch when BASE is 16 and IGNORE_TRAILING, because in that
3476 case some versions of strtoumax accept numbers like "0x1" that Emacs
3477 does not allow. */
3478 if (digit_to_number (string[signedp + 1], base) < 0)
3479 return make_number (negative ? -leading_digit : leading_digit);
3480
3481 errno = 0;
3482 n = strtoumax (string + signedp, NULL, base);
3483 if (errno == ERANGE)
3484 {
3485 /* Unfortunately there's no simple and accurate way to convert
3486 non-base-10 numbers that are out of C-language range. */
3487 if (base != 10)
3488 xsignal1 (Qoverflow_error, build_string (string));
3489 }
3490 else if (n <= (negative ? -MOST_NEGATIVE_FIXNUM : MOST_POSITIVE_FIXNUM))
3491 {
3492 EMACS_INT signed_n = n;
3493 return make_number (negative ? -signed_n : signed_n);
3494 }
3495 else
3496 value = n;
3497 }
3498
3499 /* Either the number uses float syntax, or it does not fit into a fixnum.
3500 Convert it from string to floating point, unless the value is already
3501 known because it is an infinity, a NAN, or its absolute value fits in
3502 uintmax_t. */
3503 if (! value)
3504 value = atof (string + signedp);
3505
3506 return make_float (negative ? -value : value);
3507 }
3508
3509 \f
3510 static Lisp_Object
3511 read_vector (Lisp_Object readcharfun, bool bytecodeflag)
3512 {
3513 ptrdiff_t i, size;
3514 Lisp_Object *ptr;
3515 Lisp_Object tem, item, vector;
3516 struct Lisp_Cons *otem;
3517 Lisp_Object len;
3518
3519 tem = read_list (1, readcharfun);
3520 len = Flength (tem);
3521 vector = Fmake_vector (len, Qnil);
3522
3523 size = ASIZE (vector);
3524 ptr = XVECTOR (vector)->contents;
3525 for (i = 0; i < size; i++)
3526 {
3527 item = Fcar (tem);
3528 /* If `load-force-doc-strings' is t when reading a lazily-loaded
3529 bytecode object, the docstring containing the bytecode and
3530 constants values must be treated as unibyte and passed to
3531 Fread, to get the actual bytecode string and constants vector. */
3532 if (bytecodeflag && load_force_doc_strings)
3533 {
3534 if (i == COMPILED_BYTECODE)
3535 {
3536 if (!STRINGP (item))
3537 error ("Invalid byte code");
3538
3539 /* Delay handling the bytecode slot until we know whether
3540 it is lazily-loaded (we can tell by whether the
3541 constants slot is nil). */
3542 ASET (vector, COMPILED_CONSTANTS, item);
3543 item = Qnil;
3544 }
3545 else if (i == COMPILED_CONSTANTS)
3546 {
3547 Lisp_Object bytestr = ptr[COMPILED_CONSTANTS];
3548
3549 if (NILP (item))
3550 {
3551 /* Coerce string to unibyte (like string-as-unibyte,
3552 but without generating extra garbage and
3553 guaranteeing no change in the contents). */
3554 STRING_SET_CHARS (bytestr, SBYTES (bytestr));
3555 STRING_SET_UNIBYTE (bytestr);
3556
3557 item = Fread (Fcons (bytestr, readcharfun));
3558 if (!CONSP (item))
3559 error ("Invalid byte code");
3560
3561 otem = XCONS (item);
3562 bytestr = XCAR (item);
3563 item = XCDR (item);
3564 free_cons (otem);
3565 }
3566
3567 /* Now handle the bytecode slot. */
3568 ASET (vector, COMPILED_BYTECODE, bytestr);
3569 }
3570 else if (i == COMPILED_DOC_STRING
3571 && STRINGP (item)
3572 && ! STRING_MULTIBYTE (item))
3573 {
3574 if (EQ (readcharfun, Qget_emacs_mule_file_char))
3575 item = Fdecode_coding_string (item, Qemacs_mule, Qnil, Qnil);
3576 else
3577 item = Fstring_as_multibyte (item);
3578 }
3579 }
3580 ASET (vector, i, item);
3581 otem = XCONS (tem);
3582 tem = Fcdr (tem);
3583 free_cons (otem);
3584 }
3585 return vector;
3586 }
3587
3588 /* FLAG means check for ']' to terminate rather than ')' and '.'. */
3589
3590 static Lisp_Object
3591 read_list (bool flag, Lisp_Object readcharfun)
3592 {
3593 Lisp_Object val, tail;
3594 Lisp_Object elt, tem;
3595 struct gcpro gcpro1, gcpro2;
3596 /* 0 is the normal case.
3597 1 means this list is a doc reference; replace it with the number 0.
3598 2 means this list is a doc reference; replace it with the doc string. */
3599 int doc_reference = 0;
3600
3601 /* Initialize this to 1 if we are reading a list. */
3602 bool first_in_list = flag <= 0;
3603
3604 val = Qnil;
3605 tail = Qnil;
3606
3607 while (1)
3608 {
3609 int ch;
3610 GCPRO2 (val, tail);
3611 elt = read1 (readcharfun, &ch, first_in_list);
3612 UNGCPRO;
3613
3614 first_in_list = 0;
3615
3616 /* While building, if the list starts with #$, treat it specially. */
3617 if (EQ (elt, Vload_file_name)
3618 && ! NILP (elt)
3619 && !NILP (Vpurify_flag))
3620 {
3621 if (NILP (Vdoc_file_name))
3622 /* We have not yet called Snarf-documentation, so assume
3623 this file is described in the DOC file
3624 and Snarf-documentation will fill in the right value later.
3625 For now, replace the whole list with 0. */
3626 doc_reference = 1;
3627 else
3628 /* We have already called Snarf-documentation, so make a relative
3629 file name for this file, so it can be found properly
3630 in the installed Lisp directory.
3631 We don't use Fexpand_file_name because that would make
3632 the directory absolute now. */
3633 elt = concat2 (build_string ("../lisp/"),
3634 Ffile_name_nondirectory (elt));
3635 }
3636 else if (EQ (elt, Vload_file_name)
3637 && ! NILP (elt)
3638 && load_force_doc_strings)
3639 doc_reference = 2;
3640
3641 if (ch)
3642 {
3643 if (flag > 0)
3644 {
3645 if (ch == ']')
3646 return val;
3647 invalid_syntax (") or . in a vector");
3648 }
3649 if (ch == ')')
3650 return val;
3651 if (ch == '.')
3652 {
3653 GCPRO2 (val, tail);
3654 if (!NILP (tail))
3655 XSETCDR (tail, read0 (readcharfun));
3656 else
3657 val = read0 (readcharfun);
3658 read1 (readcharfun, &ch, 0);
3659 UNGCPRO;
3660 if (ch == ')')
3661 {
3662 if (doc_reference == 1)
3663 return make_number (0);
3664 if (doc_reference == 2 && INTEGERP (XCDR (val)))
3665 {
3666 char *saved = NULL;
3667 file_offset saved_position;
3668 /* Get a doc string from the file we are loading.
3669 If it's in saved_doc_string, get it from there.
3670
3671 Here, we don't know if the string is a
3672 bytecode string or a doc string. As a
3673 bytecode string must be unibyte, we always
3674 return a unibyte string. If it is actually a
3675 doc string, caller must make it
3676 multibyte. */
3677
3678 /* Position is negative for user variables. */
3679 EMACS_INT pos = eabs (XINT (XCDR (val)));
3680 if (pos >= saved_doc_string_position
3681 && pos < (saved_doc_string_position
3682 + saved_doc_string_length))
3683 {
3684 saved = saved_doc_string;
3685 saved_position = saved_doc_string_position;
3686 }
3687 /* Look in prev_saved_doc_string the same way. */
3688 else if (pos >= prev_saved_doc_string_position
3689 && pos < (prev_saved_doc_string_position
3690 + prev_saved_doc_string_length))
3691 {
3692 saved = prev_saved_doc_string;
3693 saved_position = prev_saved_doc_string_position;
3694 }
3695 if (saved)
3696 {
3697 ptrdiff_t start = pos - saved_position;
3698 ptrdiff_t from, to;
3699
3700 /* Process quoting with ^A,
3701 and find the end of the string,
3702 which is marked with ^_ (037). */
3703 for (from = start, to = start;
3704 saved[from] != 037;)
3705 {
3706 int c = saved[from++];
3707 if (c == 1)
3708 {
3709 c = saved[from++];
3710 saved[to++] = (c == 1 ? c
3711 : c == '0' ? 0
3712 : c == '_' ? 037
3713 : c);
3714 }
3715 else
3716 saved[to++] = c;
3717 }
3718
3719 return make_unibyte_string (saved + start,
3720 to - start);
3721 }
3722 else
3723 return get_doc_string (val, 1, 0);
3724 }
3725
3726 return val;
3727 }
3728 invalid_syntax (". in wrong context");
3729 }
3730 invalid_syntax ("] in a list");
3731 }
3732 tem = list1 (elt);
3733 if (!NILP (tail))
3734 XSETCDR (tail, tem);
3735 else
3736 val = tem;
3737 tail = tem;
3738 }
3739 }
3740 \f
3741 static Lisp_Object initial_obarray;
3742
3743 /* `oblookup' stores the bucket number here, for the sake of Funintern. */
3744
3745 static size_t oblookup_last_bucket_number;
3746
3747 /* Get an error if OBARRAY is not an obarray.
3748 If it is one, return it. */
3749
3750 Lisp_Object
3751 check_obarray (Lisp_Object obarray)
3752 {
3753 if (!VECTORP (obarray) || ASIZE (obarray) == 0)
3754 {
3755 /* If Vobarray is now invalid, force it to be valid. */
3756 if (EQ (Vobarray, obarray)) Vobarray = initial_obarray;
3757 wrong_type_argument (Qvectorp, obarray);
3758 }
3759 return obarray;
3760 }
3761
3762 /* Intern the C string STR: return a symbol with that name,
3763 interned in the current obarray. */
3764
3765 Lisp_Object
3766 intern_1 (const char *str, ptrdiff_t len)
3767 {
3768 Lisp_Object obarray = check_obarray (Vobarray);
3769 Lisp_Object tem = oblookup (obarray, str, len, len);
3770
3771 return SYMBOLP (tem) ? tem : Fintern (make_string (str, len), obarray);
3772 }
3773
3774 Lisp_Object
3775 intern_c_string_1 (const char *str, ptrdiff_t len)
3776 {
3777 Lisp_Object obarray = check_obarray (Vobarray);
3778 Lisp_Object tem = oblookup (obarray, str, len, len);
3779
3780 if (SYMBOLP (tem))
3781 return tem;
3782
3783 if (NILP (Vpurify_flag))
3784 /* Creating a non-pure string from a string literal not
3785 implemented yet. We could just use make_string here and live
3786 with the extra copy. */
3787 emacs_abort ();
3788
3789 return Fintern (make_pure_c_string (str, len), obarray);
3790 }
3791 \f
3792 DEFUN ("intern", Fintern, Sintern, 1, 2, 0,
3793 doc: /* Return the canonical symbol whose name is STRING.
3794 If there is none, one is created by this function and returned.
3795 A second optional argument specifies the obarray to use;
3796 it defaults to the value of `obarray'. */)
3797 (Lisp_Object string, Lisp_Object obarray)
3798 {
3799 register Lisp_Object tem, sym, *ptr;
3800
3801 if (NILP (obarray)) obarray = Vobarray;
3802 obarray = check_obarray (obarray);
3803
3804 CHECK_STRING (string);
3805
3806 tem = oblookup (obarray, SSDATA (string),
3807 SCHARS (string),
3808 SBYTES (string));
3809 if (!INTEGERP (tem))
3810 return tem;
3811
3812 if (!NILP (Vpurify_flag))
3813 string = Fpurecopy (string);
3814 sym = Fmake_symbol (string);
3815
3816 if (EQ (obarray, initial_obarray))
3817 XSYMBOL (sym)->interned = SYMBOL_INTERNED_IN_INITIAL_OBARRAY;
3818 else
3819 XSYMBOL (sym)->interned = SYMBOL_INTERNED;
3820
3821 if ((SREF (string, 0) == ':')
3822 && EQ (obarray, initial_obarray))
3823 {
3824 XSYMBOL (sym)->constant = 1;
3825 XSYMBOL (sym)->redirect = SYMBOL_PLAINVAL;
3826 SET_SYMBOL_VAL (XSYMBOL (sym), sym);
3827 }
3828
3829 ptr = aref_addr (obarray, XINT(tem));
3830 if (SYMBOLP (*ptr))
3831 set_symbol_next (sym, XSYMBOL (*ptr));
3832 else
3833 set_symbol_next (sym, NULL);
3834 *ptr = sym;
3835 return sym;
3836 }
3837
3838 DEFUN ("intern-soft", Fintern_soft, Sintern_soft, 1, 2, 0,
3839 doc: /* Return the canonical symbol named NAME, or nil if none exists.
3840 NAME may be a string or a symbol. If it is a symbol, that exact
3841 symbol is searched for.
3842 A second optional argument specifies the obarray to use;
3843 it defaults to the value of `obarray'. */)
3844 (Lisp_Object name, Lisp_Object obarray)
3845 {
3846 register Lisp_Object tem, string;
3847
3848 if (NILP (obarray)) obarray = Vobarray;
3849 obarray = check_obarray (obarray);
3850
3851 if (!SYMBOLP (name))
3852 {
3853 CHECK_STRING (name);
3854 string = name;
3855 }
3856 else
3857 string = SYMBOL_NAME (name);
3858
3859 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3860 if (INTEGERP (tem) || (SYMBOLP (name) && !EQ (name, tem)))
3861 return Qnil;
3862 else
3863 return tem;
3864 }
3865 \f
3866 DEFUN ("unintern", Funintern, Sunintern, 1, 2, 0,
3867 doc: /* Delete the symbol named NAME, if any, from OBARRAY.
3868 The value is t if a symbol was found and deleted, nil otherwise.
3869 NAME may be a string or a symbol. If it is a symbol, that symbol
3870 is deleted, if it belongs to OBARRAY--no other symbol is deleted.
3871 OBARRAY, if nil, defaults to the value of the variable `obarray'.
3872 usage: (unintern NAME OBARRAY) */)
3873 (Lisp_Object name, Lisp_Object obarray)
3874 {
3875 register Lisp_Object string, tem;
3876 size_t hash;
3877
3878 if (NILP (obarray)) obarray = Vobarray;
3879 obarray = check_obarray (obarray);
3880
3881 if (SYMBOLP (name))
3882 string = SYMBOL_NAME (name);
3883 else
3884 {
3885 CHECK_STRING (name);
3886 string = name;
3887 }
3888
3889 tem = oblookup (obarray, SSDATA (string),
3890 SCHARS (string),
3891 SBYTES (string));
3892 if (INTEGERP (tem))
3893 return Qnil;
3894 /* If arg was a symbol, don't delete anything but that symbol itself. */
3895 if (SYMBOLP (name) && !EQ (name, tem))
3896 return Qnil;
3897
3898 /* There are plenty of other symbols which will screw up the Emacs
3899 session if we unintern them, as well as even more ways to use
3900 `setq' or `fset' or whatnot to make the Emacs session
3901 unusable. Let's not go down this silly road. --Stef */
3902 /* if (EQ (tem, Qnil) || EQ (tem, Qt))
3903 error ("Attempt to unintern t or nil"); */
3904
3905 XSYMBOL (tem)->interned = SYMBOL_UNINTERNED;
3906
3907 hash = oblookup_last_bucket_number;
3908
3909 if (EQ (AREF (obarray, hash), tem))
3910 {
3911 if (XSYMBOL (tem)->next)
3912 {
3913 Lisp_Object sym;
3914 XSETSYMBOL (sym, XSYMBOL (tem)->next);
3915 ASET (obarray, hash, sym);
3916 }
3917 else
3918 ASET (obarray, hash, make_number (0));
3919 }
3920 else
3921 {
3922 Lisp_Object tail, following;
3923
3924 for (tail = AREF (obarray, hash);
3925 XSYMBOL (tail)->next;
3926 tail = following)
3927 {
3928 XSETSYMBOL (following, XSYMBOL (tail)->next);
3929 if (EQ (following, tem))
3930 {
3931 set_symbol_next (tail, XSYMBOL (following)->next);
3932 break;
3933 }
3934 }
3935 }
3936
3937 return Qt;
3938 }
3939 \f
3940 /* Return the symbol in OBARRAY whose names matches the string
3941 of SIZE characters (SIZE_BYTE bytes) at PTR.
3942 If there is no such symbol, return the integer bucket number of
3943 where the symbol would be if it were present.
3944
3945 Also store the bucket number in oblookup_last_bucket_number. */
3946
3947 Lisp_Object
3948 oblookup (Lisp_Object obarray, register const char *ptr, ptrdiff_t size, ptrdiff_t size_byte)
3949 {
3950 size_t hash;
3951 size_t obsize;
3952 register Lisp_Object tail;
3953 Lisp_Object bucket, tem;
3954
3955 obarray = check_obarray (obarray);
3956 obsize = ASIZE (obarray);
3957
3958 /* This is sometimes needed in the middle of GC. */
3959 obsize &= ~ARRAY_MARK_FLAG;
3960 hash = hash_string (ptr, size_byte) % obsize;
3961 bucket = AREF (obarray, hash);
3962 oblookup_last_bucket_number = hash;
3963 if (EQ (bucket, make_number (0)))
3964 ;
3965 else if (!SYMBOLP (bucket))
3966 error ("Bad data in guts of obarray"); /* Like CADR error message. */
3967 else
3968 for (tail = bucket; ; XSETSYMBOL (tail, XSYMBOL (tail)->next))
3969 {
3970 if (SBYTES (SYMBOL_NAME (tail)) == size_byte
3971 && SCHARS (SYMBOL_NAME (tail)) == size
3972 && !memcmp (SDATA (SYMBOL_NAME (tail)), ptr, size_byte))
3973 return tail;
3974 else if (XSYMBOL (tail)->next == 0)
3975 break;
3976 }
3977 XSETINT (tem, hash);
3978 return tem;
3979 }
3980 \f
3981 void
3982 map_obarray (Lisp_Object obarray, void (*fn) (Lisp_Object, Lisp_Object), Lisp_Object arg)
3983 {
3984 ptrdiff_t i;
3985 register Lisp_Object tail;
3986 CHECK_VECTOR (obarray);
3987 for (i = ASIZE (obarray) - 1; i >= 0; i--)
3988 {
3989 tail = AREF (obarray, i);
3990 if (SYMBOLP (tail))
3991 while (1)
3992 {
3993 (*fn) (tail, arg);
3994 if (XSYMBOL (tail)->next == 0)
3995 break;
3996 XSETSYMBOL (tail, XSYMBOL (tail)->next);
3997 }
3998 }
3999 }
4000
4001 static void
4002 mapatoms_1 (Lisp_Object sym, Lisp_Object function)
4003 {
4004 call1 (function, sym);
4005 }
4006
4007 DEFUN ("mapatoms", Fmapatoms, Smapatoms, 1, 2, 0,
4008 doc: /* Call FUNCTION on every symbol in OBARRAY.
4009 OBARRAY defaults to the value of `obarray'. */)
4010 (Lisp_Object function, Lisp_Object obarray)
4011 {
4012 if (NILP (obarray)) obarray = Vobarray;
4013 obarray = check_obarray (obarray);
4014
4015 map_obarray (obarray, mapatoms_1, function);
4016 return Qnil;
4017 }
4018
4019 #define OBARRAY_SIZE 1511
4020
4021 void
4022 init_obarray (void)
4023 {
4024 Lisp_Object oblength;
4025 ptrdiff_t size = 100 + MAX_MULTIBYTE_LENGTH;
4026
4027 XSETFASTINT (oblength, OBARRAY_SIZE);
4028
4029 Vobarray = Fmake_vector (oblength, make_number (0));
4030 initial_obarray = Vobarray;
4031 staticpro (&initial_obarray);
4032
4033 Qunbound = Fmake_symbol (build_pure_c_string ("unbound"));
4034 /* Set temporary dummy values to Qnil and Vpurify_flag to satisfy the
4035 NILP (Vpurify_flag) check in intern_c_string. */
4036 Qnil = make_number (-1); Vpurify_flag = make_number (1);
4037 Qnil = intern_c_string ("nil");
4038
4039 /* Fmake_symbol inits fields of new symbols with Qunbound and Qnil,
4040 so those two need to be fixed manually. */
4041 SET_SYMBOL_VAL (XSYMBOL (Qunbound), Qunbound);
4042 set_symbol_function (Qunbound, Qnil);
4043 set_symbol_plist (Qunbound, Qnil);
4044 SET_SYMBOL_VAL (XSYMBOL (Qnil), Qnil);
4045 XSYMBOL (Qnil)->constant = 1;
4046 XSYMBOL (Qnil)->declared_special = 1;
4047 set_symbol_plist (Qnil, Qnil);
4048 set_symbol_function (Qnil, Qnil);
4049
4050 Qt = intern_c_string ("t");
4051 SET_SYMBOL_VAL (XSYMBOL (Qt), Qt);
4052 XSYMBOL (Qnil)->declared_special = 1;
4053 XSYMBOL (Qt)->constant = 1;
4054
4055 /* Qt is correct even if CANNOT_DUMP. loadup.el will set to nil at end. */
4056 Vpurify_flag = Qt;
4057
4058 DEFSYM (Qvariable_documentation, "variable-documentation");
4059
4060 read_buffer = xmalloc (size);
4061 read_buffer_size = size;
4062 }
4063 \f
4064 void
4065 defsubr (struct Lisp_Subr *sname)
4066 {
4067 Lisp_Object sym, tem;
4068 sym = intern_c_string (sname->symbol_name);
4069 XSETPVECTYPE (sname, PVEC_SUBR);
4070 XSETSUBR (tem, sname);
4071 set_symbol_function (sym, tem);
4072 }
4073
4074 #ifdef NOTDEF /* Use fset in subr.el now! */
4075 void
4076 defalias (struct Lisp_Subr *sname, char *string)
4077 {
4078 Lisp_Object sym;
4079 sym = intern (string);
4080 XSETSUBR (XSYMBOL (sym)->function, sname);
4081 }
4082 #endif /* NOTDEF */
4083
4084 /* Define an "integer variable"; a symbol whose value is forwarded to a
4085 C variable of type EMACS_INT. Sample call (with "xx" to fool make-docfile):
4086 DEFxxVAR_INT ("emacs-priority", &emacs_priority, "Documentation"); */
4087 void
4088 defvar_int (struct Lisp_Intfwd *i_fwd,
4089 const char *namestring, EMACS_INT *address)
4090 {
4091 Lisp_Object sym;
4092 sym = intern_c_string (namestring);
4093 i_fwd->type = Lisp_Fwd_Int;
4094 i_fwd->intvar = address;
4095 XSYMBOL (sym)->declared_special = 1;
4096 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4097 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)i_fwd);
4098 }
4099
4100 /* Similar but define a variable whose value is t if address contains 1,
4101 nil if address contains 0. */
4102 void
4103 defvar_bool (struct Lisp_Boolfwd *b_fwd,
4104 const char *namestring, bool *address)
4105 {
4106 Lisp_Object sym;
4107 sym = intern_c_string (namestring);
4108 b_fwd->type = Lisp_Fwd_Bool;
4109 b_fwd->boolvar = address;
4110 XSYMBOL (sym)->declared_special = 1;
4111 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4112 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)b_fwd);
4113 Vbyte_boolean_vars = Fcons (sym, Vbyte_boolean_vars);
4114 }
4115
4116 /* Similar but define a variable whose value is the Lisp Object stored
4117 at address. Two versions: with and without gc-marking of the C
4118 variable. The nopro version is used when that variable will be
4119 gc-marked for some other reason, since marking the same slot twice
4120 can cause trouble with strings. */
4121 void
4122 defvar_lisp_nopro (struct Lisp_Objfwd *o_fwd,
4123 const char *namestring, Lisp_Object *address)
4124 {
4125 Lisp_Object sym;
4126 sym = intern_c_string (namestring);
4127 o_fwd->type = Lisp_Fwd_Obj;
4128 o_fwd->objvar = address;
4129 XSYMBOL (sym)->declared_special = 1;
4130 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4131 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)o_fwd);
4132 }
4133
4134 void
4135 defvar_lisp (struct Lisp_Objfwd *o_fwd,
4136 const char *namestring, Lisp_Object *address)
4137 {
4138 defvar_lisp_nopro (o_fwd, namestring, address);
4139 staticpro (address);
4140 }
4141
4142 /* Similar but define a variable whose value is the Lisp Object stored
4143 at a particular offset in the current kboard object. */
4144
4145 void
4146 defvar_kboard (struct Lisp_Kboard_Objfwd *ko_fwd,
4147 const char *namestring, int offset)
4148 {
4149 Lisp_Object sym;
4150 sym = intern_c_string (namestring);
4151 ko_fwd->type = Lisp_Fwd_Kboard_Obj;
4152 ko_fwd->offset = offset;
4153 XSYMBOL (sym)->declared_special = 1;
4154 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4155 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)ko_fwd);
4156 }
4157 \f
4158 /* Check that the elements of lpath exist. */
4159
4160 static void
4161 load_path_check (Lisp_Object lpath)
4162 {
4163 Lisp_Object path_tail;
4164
4165 /* The only elements that might not exist are those from
4166 PATH_LOADSEARCH, EMACSLOADPATH. Anything else is only added if
4167 it exists. */
4168 for (path_tail = lpath; !NILP (path_tail); path_tail = XCDR (path_tail))
4169 {
4170 Lisp_Object dirfile;
4171 dirfile = Fcar (path_tail);
4172 if (STRINGP (dirfile))
4173 {
4174 dirfile = Fdirectory_file_name (dirfile);
4175 if (! file_accessible_directory_p (SSDATA (dirfile)))
4176 dir_warning ("Lisp directory", XCAR (path_tail));
4177 }
4178 }
4179 }
4180
4181 /* Return the default load-path, to be used if EMACSLOADPATH is unset.
4182 This does not include the standard site-lisp directories
4183 under the installation prefix (i.e., PATH_SITELOADSEARCH),
4184 but it does (unless no_site_lisp is set) include site-lisp
4185 directories in the source/build directories if those exist and we
4186 are running uninstalled.
4187
4188 Uses the following logic:
4189 If CANNOT_DUMP: Use PATH_LOADSEARCH.
4190 The remainder is what happens when dumping works:
4191 If purify-flag (ie dumping) just use PATH_DUMPLOADSEARCH.
4192 Otherwise use PATH_LOADSEARCH.
4193
4194 If !initialized, then just return PATH_DUMPLOADSEARCH.
4195 If initialized:
4196 If Vinstallation_directory is not nil (ie, running uninstalled):
4197 If installation-dir/lisp exists and not already a member,
4198 we must be running uninstalled. Reset the load-path
4199 to just installation-dir/lisp. (The default PATH_LOADSEARCH
4200 refers to the eventual installation directories. Since we
4201 are not yet installed, we should not use them, even if they exist.)
4202 If installation-dir/lisp does not exist, just add
4203 PATH_DUMPLOADSEARCH at the end instead.
4204 Add installation-dir/site-lisp (if !no_site_lisp, and exists
4205 and not already a member) at the front.
4206 If installation-dir != source-dir (ie running an uninstalled,
4207 out-of-tree build) AND install-dir/src/Makefile exists BUT
4208 install-dir/src/Makefile.in does NOT exist (this is a sanity
4209 check), then repeat the above steps for source-dir/lisp, site-lisp. */
4210
4211 static Lisp_Object
4212 load_path_default (void)
4213 {
4214 Lisp_Object lpath = Qnil;
4215 const char *normal;
4216
4217 #ifdef CANNOT_DUMP
4218 #ifdef HAVE_NS
4219 const char *loadpath = ns_load_path ();
4220 #endif
4221
4222 normal = PATH_LOADSEARCH;
4223 #ifdef HAVE_NS
4224 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4225 #else
4226 lpath = decode_env_path (0, normal, 0);
4227 #endif
4228
4229 #else /* !CANNOT_DUMP */
4230
4231 normal = NILP (Vpurify_flag) ? PATH_LOADSEARCH : PATH_DUMPLOADSEARCH;
4232
4233 if (initialized)
4234 {
4235 #ifdef HAVE_NS
4236 const char *loadpath = ns_load_path ();
4237 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4238 #else
4239 lpath = decode_env_path (0, normal, 0);
4240 #endif
4241 if (!NILP (Vinstallation_directory))
4242 {
4243 Lisp_Object tem, tem1;
4244
4245 /* Add to the path the lisp subdir of the installation
4246 dir, if it is accessible. Note: in out-of-tree builds,
4247 this directory is empty save for Makefile. */
4248 tem = Fexpand_file_name (build_string ("lisp"),
4249 Vinstallation_directory);
4250 tem1 = Ffile_accessible_directory_p (tem);
4251 if (!NILP (tem1))
4252 {
4253 if (NILP (Fmember (tem, lpath)))
4254 {
4255 /* We are running uninstalled. The default load-path
4256 points to the eventual installed lisp directories.
4257 We should not use those now, even if they exist,
4258 so start over from a clean slate. */
4259 lpath = list1 (tem);
4260 }
4261 }
4262 else
4263 /* That dir doesn't exist, so add the build-time
4264 Lisp dirs instead. */
4265 {
4266 Lisp_Object dump_path =
4267 decode_env_path (0, PATH_DUMPLOADSEARCH, 0);
4268 lpath = nconc2 (lpath, dump_path);
4269 }
4270
4271 /* Add site-lisp under the installation dir, if it exists. */
4272 if (!no_site_lisp)
4273 {
4274 tem = Fexpand_file_name (build_string ("site-lisp"),
4275 Vinstallation_directory);
4276 tem1 = Ffile_accessible_directory_p (tem);
4277 if (!NILP (tem1))
4278 {
4279 if (NILP (Fmember (tem, lpath)))
4280 lpath = Fcons (tem, lpath);
4281 }
4282 }
4283
4284 /* If Emacs was not built in the source directory,
4285 and it is run from where it was built, add to load-path
4286 the lisp and site-lisp dirs under that directory. */
4287
4288 if (NILP (Fequal (Vinstallation_directory, Vsource_directory)))
4289 {
4290 Lisp_Object tem2;
4291
4292 tem = Fexpand_file_name (build_string ("src/Makefile"),
4293 Vinstallation_directory);
4294 tem1 = Ffile_exists_p (tem);
4295
4296 /* Don't be fooled if they moved the entire source tree
4297 AFTER dumping Emacs. If the build directory is indeed
4298 different from the source dir, src/Makefile.in and
4299 src/Makefile will not be found together. */
4300 tem = Fexpand_file_name (build_string ("src/Makefile.in"),
4301 Vinstallation_directory);
4302 tem2 = Ffile_exists_p (tem);
4303 if (!NILP (tem1) && NILP (tem2))
4304 {
4305 tem = Fexpand_file_name (build_string ("lisp"),
4306 Vsource_directory);
4307
4308 if (NILP (Fmember (tem, lpath)))
4309 lpath = Fcons (tem, lpath);
4310
4311 if (!no_site_lisp)
4312 {
4313 tem = Fexpand_file_name (build_string ("site-lisp"),
4314 Vsource_directory);
4315 tem1 = Ffile_accessible_directory_p (tem);
4316 if (!NILP (tem1))
4317 {
4318 if (NILP (Fmember (tem, lpath)))
4319 lpath = Fcons (tem, lpath);
4320 }
4321 }
4322 }
4323 } /* Vinstallation_directory != Vsource_directory */
4324
4325 } /* if Vinstallation_directory */
4326 }
4327 else /* !initialized */
4328 {
4329 /* NORMAL refers to PATH_DUMPLOADSEARCH, ie the lisp dir in the
4330 source directory. We used to add ../lisp (ie the lisp dir in
4331 the build directory) at the front here, but that should not
4332 be necessary, since in out of tree builds lisp/ is empty, save
4333 for Makefile. */
4334 lpath = decode_env_path (0, normal, 0);
4335 }
4336 #endif /* !CANNOT_DUMP */
4337
4338 return lpath;
4339 }
4340
4341 void
4342 init_lread (void)
4343 {
4344 /* First, set Vload_path. */
4345
4346 /* Ignore EMACSLOADPATH when dumping. */
4347 #ifdef CANNOT_DUMP
4348 bool use_loadpath = true;
4349 #else
4350 bool use_loadpath = NILP (Vpurify_flag);
4351 #endif
4352
4353 if (use_loadpath && egetenv ("EMACSLOADPATH"))
4354 {
4355 Vload_path = decode_env_path ("EMACSLOADPATH", 0, 1);
4356
4357 /* Check (non-nil) user-supplied elements. */
4358 load_path_check (Vload_path);
4359
4360 /* If no nils in the environment variable, use as-is.
4361 Otherwise, replace any nils with the default. */
4362 if (! NILP (Fmemq (Qnil, Vload_path)))
4363 {
4364 Lisp_Object elem, elpath = Vload_path;
4365 Lisp_Object default_lpath = load_path_default ();
4366
4367 /* Check defaults, before adding site-lisp. */
4368 load_path_check (default_lpath);
4369
4370 /* Add the site-lisp directories to the front of the default. */
4371 if (!no_site_lisp)
4372 {
4373 Lisp_Object sitelisp;
4374 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4375 if (! NILP (sitelisp))
4376 default_lpath = nconc2 (sitelisp, default_lpath);
4377 }
4378
4379 Vload_path = Qnil;
4380
4381 /* Replace nils from EMACSLOADPATH by default. */
4382 while (CONSP (elpath))
4383 {
4384 Lisp_Object arg[2];
4385 elem = XCAR (elpath);
4386 elpath = XCDR (elpath);
4387 arg[0] = Vload_path;
4388 arg[1] = NILP (elem) ? default_lpath : Fcons (elem, Qnil);
4389 Vload_path = Fappend (2, arg);
4390 }
4391 } /* Fmemq (Qnil, Vload_path) */
4392 }
4393 else
4394 {
4395 Vload_path = load_path_default ();
4396
4397 /* Check before adding site-lisp directories.
4398 The install should have created them, but they are not
4399 required, so no need to warn if they are absent.
4400 Or we might be running before installation. */
4401 load_path_check (Vload_path);
4402
4403 /* Add the site-lisp directories at the front. */
4404 if (initialized && !no_site_lisp)
4405 {
4406 Lisp_Object sitelisp;
4407 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4408 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4409 }
4410 }
4411
4412 Vvalues = Qnil;
4413
4414 load_in_progress = 0;
4415 Vload_file_name = Qnil;
4416 Vstandard_input = Qt;
4417 Vloads_in_progress = Qnil;
4418 }
4419
4420 /* Print a warning that directory intended for use USE and with name
4421 DIRNAME cannot be accessed. On entry, errno should correspond to
4422 the access failure. Print the warning on stderr and put it in
4423 *Messages*. */
4424
4425 void
4426 dir_warning (char const *use, Lisp_Object dirname)
4427 {
4428 static char const format[] = "Warning: %s `%s': %s\n";
4429 int access_errno = errno;
4430 fprintf (stderr, format, use, SSDATA (dirname), strerror (access_errno));
4431
4432 /* Don't log the warning before we've initialized!! */
4433 if (initialized)
4434 {
4435 char const *diagnostic = emacs_strerror (access_errno);
4436 USE_SAFE_ALLOCA;
4437 char *buffer = SAFE_ALLOCA (sizeof format - 3 * (sizeof "%s" - 1)
4438 + strlen (use) + SBYTES (dirname)
4439 + strlen (diagnostic));
4440 ptrdiff_t message_len = esprintf (buffer, format, use, SSDATA (dirname),
4441 diagnostic);
4442 message_dolog (buffer, message_len, 0, STRING_MULTIBYTE (dirname));
4443 SAFE_FREE ();
4444 }
4445 }
4446
4447 void
4448 syms_of_lread (void)
4449 {
4450 defsubr (&Sread);
4451 defsubr (&Sread_from_string);
4452 defsubr (&Sintern);
4453 defsubr (&Sintern_soft);
4454 defsubr (&Sunintern);
4455 defsubr (&Sget_load_suffixes);
4456 defsubr (&Sload);
4457 defsubr (&Seval_buffer);
4458 defsubr (&Seval_region);
4459 defsubr (&Sread_char);
4460 defsubr (&Sread_char_exclusive);
4461 defsubr (&Sread_event);
4462 defsubr (&Sget_file_char);
4463 defsubr (&Smapatoms);
4464 defsubr (&Slocate_file_internal);
4465
4466 DEFVAR_LISP ("obarray", Vobarray,
4467 doc: /* Symbol table for use by `intern' and `read'.
4468 It is a vector whose length ought to be prime for best results.
4469 The vector's contents don't make sense if examined from Lisp programs;
4470 to find all the symbols in an obarray, use `mapatoms'. */);
4471
4472 DEFVAR_LISP ("values", Vvalues,
4473 doc: /* List of values of all expressions which were read, evaluated and printed.
4474 Order is reverse chronological. */);
4475 XSYMBOL (intern ("values"))->declared_special = 0;
4476
4477 DEFVAR_LISP ("standard-input", Vstandard_input,
4478 doc: /* Stream for read to get input from.
4479 See documentation of `read' for possible values. */);
4480 Vstandard_input = Qt;
4481
4482 DEFVAR_LISP ("read-with-symbol-positions", Vread_with_symbol_positions,
4483 doc: /* If non-nil, add position of read symbols to `read-symbol-positions-list'.
4484
4485 If this variable is a buffer, then only forms read from that buffer
4486 will be added to `read-symbol-positions-list'.
4487 If this variable is t, then all read forms will be added.
4488 The effect of all other values other than nil are not currently
4489 defined, although they may be in the future.
4490
4491 The positions are relative to the last call to `read' or
4492 `read-from-string'. It is probably a bad idea to set this variable at
4493 the toplevel; bind it instead. */);
4494 Vread_with_symbol_positions = Qnil;
4495
4496 DEFVAR_LISP ("read-symbol-positions-list", Vread_symbol_positions_list,
4497 doc: /* A list mapping read symbols to their positions.
4498 This variable is modified during calls to `read' or
4499 `read-from-string', but only when `read-with-symbol-positions' is
4500 non-nil.
4501
4502 Each element of the list looks like (SYMBOL . CHAR-POSITION), where
4503 CHAR-POSITION is an integer giving the offset of that occurrence of the
4504 symbol from the position where `read' or `read-from-string' started.
4505
4506 Note that a symbol will appear multiple times in this list, if it was
4507 read multiple times. The list is in the same order as the symbols
4508 were read in. */);
4509 Vread_symbol_positions_list = Qnil;
4510
4511 DEFVAR_LISP ("read-circle", Vread_circle,
4512 doc: /* Non-nil means read recursive structures using #N= and #N# syntax. */);
4513 Vread_circle = Qt;
4514
4515 DEFVAR_LISP ("load-path", Vload_path,
4516 doc: /* List of directories to search for files to load.
4517 Each element is a string (directory name) or nil (meaning `default-directory').
4518 Initialized during startup as described in Info node `(elisp)Library Search'. */);
4519
4520 DEFVAR_LISP ("load-suffixes", Vload_suffixes,
4521 doc: /* List of suffixes for (compiled or source) Emacs Lisp files.
4522 This list should not include the empty string.
4523 `load' and related functions try to append these suffixes, in order,
4524 to the specified file name if a Lisp suffix is allowed or required. */);
4525 Vload_suffixes = list2 (build_pure_c_string (".elc"),
4526 build_pure_c_string (".el"));
4527 DEFVAR_LISP ("load-file-rep-suffixes", Vload_file_rep_suffixes,
4528 doc: /* List of suffixes that indicate representations of \
4529 the same file.
4530 This list should normally start with the empty string.
4531
4532 Enabling Auto Compression mode appends the suffixes in
4533 `jka-compr-load-suffixes' to this list and disabling Auto Compression
4534 mode removes them again. `load' and related functions use this list to
4535 determine whether they should look for compressed versions of a file
4536 and, if so, which suffixes they should try to append to the file name
4537 in order to do so. However, if you want to customize which suffixes
4538 the loading functions recognize as compression suffixes, you should
4539 customize `jka-compr-load-suffixes' rather than the present variable. */);
4540 Vload_file_rep_suffixes = list1 (empty_unibyte_string);
4541
4542 DEFVAR_BOOL ("load-in-progress", load_in_progress,
4543 doc: /* Non-nil if inside of `load'. */);
4544 DEFSYM (Qload_in_progress, "load-in-progress");
4545
4546 DEFVAR_LISP ("after-load-alist", Vafter_load_alist,
4547 doc: /* An alist of functions to be evalled when particular files are loaded.
4548 Each element looks like (REGEXP-OR-FEATURE FUNCS...).
4549
4550 REGEXP-OR-FEATURE is either a regular expression to match file names, or
4551 a symbol \(a feature name).
4552
4553 When `load' is run and the file-name argument matches an element's
4554 REGEXP-OR-FEATURE, or when `provide' is run and provides the symbol
4555 REGEXP-OR-FEATURE, the FUNCS in the element are called.
4556
4557 An error in FORMS does not undo the load, but does prevent execution of
4558 the rest of the FORMS. */);
4559 Vafter_load_alist = Qnil;
4560
4561 DEFVAR_LISP ("load-history", Vload_history,
4562 doc: /* Alist mapping loaded file names to symbols and features.
4563 Each alist element should be a list (FILE-NAME ENTRIES...), where
4564 FILE-NAME is the name of a file that has been loaded into Emacs.
4565 The file name is absolute and true (i.e. it doesn't contain symlinks).
4566 As an exception, one of the alist elements may have FILE-NAME nil,
4567 for symbols and features not associated with any file.
4568
4569 The remaining ENTRIES in the alist element describe the functions and
4570 variables defined in that file, the features provided, and the
4571 features required. Each entry has the form `(provide . FEATURE)',
4572 `(require . FEATURE)', `(defun . FUNCTION)', `(autoload . SYMBOL)',
4573 `(defface . SYMBOL)', or `(t . SYMBOL)'. Entries like `(t . SYMBOL)'
4574 may precede a `(defun . FUNCTION)' entry, and means that SYMBOL was an
4575 autoload before this file redefined it as a function. In addition,
4576 entries may also be single symbols, which means that SYMBOL was
4577 defined by `defvar' or `defconst'.
4578
4579 During preloading, the file name recorded is relative to the main Lisp
4580 directory. These file names are converted to absolute at startup. */);
4581 Vload_history = Qnil;
4582
4583 DEFVAR_LISP ("load-file-name", Vload_file_name,
4584 doc: /* Full name of file being loaded by `load'. */);
4585 Vload_file_name = Qnil;
4586
4587 DEFVAR_LISP ("user-init-file", Vuser_init_file,
4588 doc: /* File name, including directory, of user's initialization file.
4589 If the file loaded had extension `.elc', and the corresponding source file
4590 exists, this variable contains the name of source file, suitable for use
4591 by functions like `custom-save-all' which edit the init file.
4592 While Emacs loads and evaluates the init file, value is the real name
4593 of the file, regardless of whether or not it has the `.elc' extension. */);
4594 Vuser_init_file = Qnil;
4595
4596 DEFVAR_LISP ("current-load-list", Vcurrent_load_list,
4597 doc: /* Used for internal purposes by `load'. */);
4598 Vcurrent_load_list = Qnil;
4599
4600 DEFVAR_LISP ("load-read-function", Vload_read_function,
4601 doc: /* Function used by `load' and `eval-region' for reading expressions.
4602 The default is nil, which means use the function `read'. */);
4603 Vload_read_function = Qnil;
4604
4605 DEFVAR_LISP ("load-source-file-function", Vload_source_file_function,
4606 doc: /* Function called in `load' to load an Emacs Lisp source file.
4607 The value should be a function for doing code conversion before
4608 reading a source file. It can also be nil, in which case loading is
4609 done without any code conversion.
4610
4611 If the value is a function, it is called with four arguments,
4612 FULLNAME, FILE, NOERROR, NOMESSAGE. FULLNAME is the absolute name of
4613 the file to load, FILE is the non-absolute name (for messages etc.),
4614 and NOERROR and NOMESSAGE are the corresponding arguments passed to
4615 `load'. The function should return t if the file was loaded. */);
4616 Vload_source_file_function = Qnil;
4617
4618 DEFVAR_BOOL ("load-force-doc-strings", load_force_doc_strings,
4619 doc: /* Non-nil means `load' should force-load all dynamic doc strings.
4620 This is useful when the file being loaded is a temporary copy. */);
4621 load_force_doc_strings = 0;
4622
4623 DEFVAR_BOOL ("load-convert-to-unibyte", load_convert_to_unibyte,
4624 doc: /* Non-nil means `read' converts strings to unibyte whenever possible.
4625 This is normally bound by `load' and `eval-buffer' to control `read',
4626 and is not meant for users to change. */);
4627 load_convert_to_unibyte = 0;
4628
4629 DEFVAR_LISP ("source-directory", Vsource_directory,
4630 doc: /* Directory in which Emacs sources were found when Emacs was built.
4631 You cannot count on them to still be there! */);
4632 Vsource_directory
4633 = Fexpand_file_name (build_string ("../"),
4634 Fcar (decode_env_path (0, PATH_DUMPLOADSEARCH, 0)));
4635
4636 DEFVAR_LISP ("preloaded-file-list", Vpreloaded_file_list,
4637 doc: /* List of files that were preloaded (when dumping Emacs). */);
4638 Vpreloaded_file_list = Qnil;
4639
4640 DEFVAR_LISP ("byte-boolean-vars", Vbyte_boolean_vars,
4641 doc: /* List of all DEFVAR_BOOL variables, used by the byte code optimizer. */);
4642 Vbyte_boolean_vars = Qnil;
4643
4644 DEFVAR_BOOL ("load-dangerous-libraries", load_dangerous_libraries,
4645 doc: /* Non-nil means load dangerous compiled Lisp files.
4646 Some versions of XEmacs use different byte codes than Emacs. These
4647 incompatible byte codes can make Emacs crash when it tries to execute
4648 them. */);
4649 load_dangerous_libraries = 0;
4650
4651 DEFVAR_BOOL ("force-load-messages", force_load_messages,
4652 doc: /* Non-nil means force printing messages when loading Lisp files.
4653 This overrides the value of the NOMESSAGE argument to `load'. */);
4654 force_load_messages = 0;
4655
4656 DEFVAR_LISP ("bytecomp-version-regexp", Vbytecomp_version_regexp,
4657 doc: /* Regular expression matching safe to load compiled Lisp files.
4658 When Emacs loads a compiled Lisp file, it reads the first 512 bytes
4659 from the file, and matches them against this regular expression.
4660 When the regular expression matches, the file is considered to be safe
4661 to load. See also `load-dangerous-libraries'. */);
4662 Vbytecomp_version_regexp
4663 = build_pure_c_string ("^;;;.\\(in Emacs version\\|bytecomp version FSF\\)");
4664
4665 DEFSYM (Qlexical_binding, "lexical-binding");
4666 DEFVAR_LISP ("lexical-binding", Vlexical_binding,
4667 doc: /* Whether to use lexical binding when evaluating code.
4668 Non-nil means that the code in the current buffer should be evaluated
4669 with lexical binding.
4670 This variable is automatically set from the file variables of an
4671 interpreted Lisp file read using `load'. Unlike other file local
4672 variables, this must be set in the first line of a file. */);
4673 Vlexical_binding = Qnil;
4674 Fmake_variable_buffer_local (Qlexical_binding);
4675
4676 DEFVAR_LISP ("eval-buffer-list", Veval_buffer_list,
4677 doc: /* List of buffers being read from by calls to `eval-buffer' and `eval-region'. */);
4678 Veval_buffer_list = Qnil;
4679
4680 DEFVAR_LISP ("old-style-backquotes", Vold_style_backquotes,
4681 doc: /* Set to non-nil when `read' encounters an old-style backquote. */);
4682 Vold_style_backquotes = Qnil;
4683 DEFSYM (Qold_style_backquotes, "old-style-backquotes");
4684
4685 DEFVAR_BOOL ("load-prefer-newer", load_prefer_newer,
4686 doc: /* Non-nil means `load' prefers the newest version of a file.
4687 This applies when a filename suffix is not explicitly specified and
4688 `load' is trying various possible suffixes (see `load-suffixes' and
4689 `load-file-rep-suffixes'). Normally, it stops at the first file
4690 that exists unless you explicitly specify one or the other. If this
4691 option is non-nil, it checks all suffixes and uses whichever file is
4692 newest.
4693 Note that if you customize this, obviously it will not affect files
4694 that are loaded before your customizations are read! */);
4695 load_prefer_newer = 0;
4696
4697 /* Vsource_directory was initialized in init_lread. */
4698
4699 DEFSYM (Qcurrent_load_list, "current-load-list");
4700 DEFSYM (Qstandard_input, "standard-input");
4701 DEFSYM (Qread_char, "read-char");
4702 DEFSYM (Qget_file_char, "get-file-char");
4703 DEFSYM (Qget_emacs_mule_file_char, "get-emacs-mule-file-char");
4704 DEFSYM (Qload_force_doc_strings, "load-force-doc-strings");
4705
4706 DEFSYM (Qbackquote, "`");
4707 DEFSYM (Qcomma, ",");
4708 DEFSYM (Qcomma_at, ",@");
4709 DEFSYM (Qcomma_dot, ",.");
4710
4711 DEFSYM (Qinhibit_file_name_operation, "inhibit-file-name-operation");
4712 DEFSYM (Qascii_character, "ascii-character");
4713 DEFSYM (Qfunction, "function");
4714 DEFSYM (Qload, "load");
4715 DEFSYM (Qload_file_name, "load-file-name");
4716 DEFSYM (Qeval_buffer_list, "eval-buffer-list");
4717 DEFSYM (Qfile_truename, "file-truename");
4718 DEFSYM (Qdir_ok, "dir-ok");
4719 DEFSYM (Qdo_after_load_evaluation, "do-after-load-evaluation");
4720
4721 staticpro (&read_objects);
4722 read_objects = Qnil;
4723 staticpro (&seen_list);
4724 seen_list = Qnil;
4725
4726 Vloads_in_progress = Qnil;
4727 staticpro (&Vloads_in_progress);
4728
4729 DEFSYM (Qhash_table, "hash-table");
4730 DEFSYM (Qdata, "data");
4731 DEFSYM (Qtest, "test");
4732 DEFSYM (Qsize, "size");
4733 DEFSYM (Qweakness, "weakness");
4734 DEFSYM (Qrehash_size, "rehash-size");
4735 DEFSYM (Qrehash_threshold, "rehash-threshold");
4736 }