]> code.delx.au - gnu-emacs/blob - src/lread.c
Merge from origin/emacs-24
[gnu-emacs] / src / lread.c
1 /* Lisp parsing and input streams.
2
3 Copyright (C) 1985-1989, 1993-1995, 1997-2015 Free Software Foundation,
4 Inc.
5
6 This file is part of GNU Emacs.
7
8 GNU Emacs is free software: you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation, either version 3 of the License, or
11 (at your option) any later version.
12
13 GNU Emacs is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
17
18 You should have received a copy of the GNU General Public License
19 along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */
20
21 /* Tell globals.h to define tables needed by init_obarray. */
22 #define DEFINE_SYMBOLS
23
24 #include <config.h>
25 #include "sysstdio.h"
26 #include <sys/types.h>
27 #include <sys/stat.h>
28 #include <sys/file.h>
29 #include <errno.h>
30 #include <limits.h> /* For CHAR_BIT. */
31 #include <math.h>
32 #include <stat-time.h>
33 #include "lisp.h"
34 #include "intervals.h"
35 #include "character.h"
36 #include "buffer.h"
37 #include "charset.h"
38 #include "coding.h"
39 #include <epaths.h>
40 #include "commands.h"
41 #include "keyboard.h"
42 #include "frame.h"
43 #include "termhooks.h"
44 #include "blockinput.h"
45
46 #ifdef MSDOS
47 #include "msdos.h"
48 #endif
49
50 #ifdef HAVE_NS
51 #include "nsterm.h"
52 #endif
53
54 #include <unistd.h>
55
56 #ifdef HAVE_SETLOCALE
57 #include <locale.h>
58 #endif /* HAVE_SETLOCALE */
59
60 #include <fcntl.h>
61
62 #ifdef HAVE_FSEEKO
63 #define file_offset off_t
64 #define file_tell ftello
65 #else
66 #define file_offset long
67 #define file_tell ftell
68 #endif
69
70 /* The association list of objects read with the #n=object form.
71 Each member of the list has the form (n . object), and is used to
72 look up the object for the corresponding #n# construct.
73 It must be set to nil before all top-level calls to read0. */
74 static Lisp_Object read_objects;
75
76 /* File for get_file_char to read from. Use by load. */
77 static FILE *instream;
78
79 /* For use within read-from-string (this reader is non-reentrant!!) */
80 static ptrdiff_t read_from_string_index;
81 static ptrdiff_t read_from_string_index_byte;
82 static ptrdiff_t read_from_string_limit;
83
84 /* Number of characters read in the current call to Fread or
85 Fread_from_string. */
86 static EMACS_INT readchar_count;
87
88 /* This contains the last string skipped with #@. */
89 static char *saved_doc_string;
90 /* Length of buffer allocated in saved_doc_string. */
91 static ptrdiff_t saved_doc_string_size;
92 /* Length of actual data in saved_doc_string. */
93 static ptrdiff_t saved_doc_string_length;
94 /* This is the file position that string came from. */
95 static file_offset saved_doc_string_position;
96
97 /* This contains the previous string skipped with #@.
98 We copy it from saved_doc_string when a new string
99 is put in saved_doc_string. */
100 static char *prev_saved_doc_string;
101 /* Length of buffer allocated in prev_saved_doc_string. */
102 static ptrdiff_t prev_saved_doc_string_size;
103 /* Length of actual data in prev_saved_doc_string. */
104 static ptrdiff_t prev_saved_doc_string_length;
105 /* This is the file position that string came from. */
106 static file_offset prev_saved_doc_string_position;
107
108 /* True means inside a new-style backquote
109 with no surrounding parentheses.
110 Fread initializes this to false, so we need not specbind it
111 or worry about what happens to it when there is an error. */
112 static bool new_backquote_flag;
113
114 /* A list of file names for files being loaded in Fload. Used to
115 check for recursive loads. */
116
117 static Lisp_Object Vloads_in_progress;
118
119 static int read_emacs_mule_char (int, int (*) (int, Lisp_Object),
120 Lisp_Object);
121
122 static void readevalloop (Lisp_Object, FILE *, Lisp_Object, bool,
123 Lisp_Object, Lisp_Object,
124 Lisp_Object, Lisp_Object);
125 \f
126 /* Functions that read one byte from the current source READCHARFUN
127 or unreads one byte. If the integer argument C is -1, it returns
128 one read byte, or -1 when there's no more byte in the source. If C
129 is 0 or positive, it unreads C, and the return value is not
130 interesting. */
131
132 static int readbyte_for_lambda (int, Lisp_Object);
133 static int readbyte_from_file (int, Lisp_Object);
134 static int readbyte_from_string (int, Lisp_Object);
135
136 /* Handle unreading and rereading of characters.
137 Write READCHAR to read a character,
138 UNREAD(c) to unread c to be read again.
139
140 These macros correctly read/unread multibyte characters. */
141
142 #define READCHAR readchar (readcharfun, NULL)
143 #define UNREAD(c) unreadchar (readcharfun, c)
144
145 /* Same as READCHAR but set *MULTIBYTE to the multibyteness of the source. */
146 #define READCHAR_REPORT_MULTIBYTE(multibyte) readchar (readcharfun, multibyte)
147
148 /* When READCHARFUN is Qget_file_char, Qget_emacs_mule_file_char,
149 Qlambda, or a cons, we use this to keep an unread character because
150 a file stream can't handle multibyte-char unreading. The value -1
151 means that there's no unread character. */
152 static int unread_char;
153
154 static int
155 readchar (Lisp_Object readcharfun, bool *multibyte)
156 {
157 Lisp_Object tem;
158 register int c;
159 int (*readbyte) (int, Lisp_Object);
160 unsigned char buf[MAX_MULTIBYTE_LENGTH];
161 int i, len;
162 bool emacs_mule_encoding = 0;
163
164 if (multibyte)
165 *multibyte = 0;
166
167 readchar_count++;
168
169 if (BUFFERP (readcharfun))
170 {
171 register struct buffer *inbuffer = XBUFFER (readcharfun);
172
173 ptrdiff_t pt_byte = BUF_PT_BYTE (inbuffer);
174
175 if (! BUFFER_LIVE_P (inbuffer))
176 return -1;
177
178 if (pt_byte >= BUF_ZV_BYTE (inbuffer))
179 return -1;
180
181 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
182 {
183 /* Fetch the character code from the buffer. */
184 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, pt_byte);
185 BUF_INC_POS (inbuffer, pt_byte);
186 c = STRING_CHAR (p);
187 if (multibyte)
188 *multibyte = 1;
189 }
190 else
191 {
192 c = BUF_FETCH_BYTE (inbuffer, pt_byte);
193 if (! ASCII_CHAR_P (c))
194 c = BYTE8_TO_CHAR (c);
195 pt_byte++;
196 }
197 SET_BUF_PT_BOTH (inbuffer, BUF_PT (inbuffer) + 1, pt_byte);
198
199 return c;
200 }
201 if (MARKERP (readcharfun))
202 {
203 register struct buffer *inbuffer = XMARKER (readcharfun)->buffer;
204
205 ptrdiff_t bytepos = marker_byte_position (readcharfun);
206
207 if (bytepos >= BUF_ZV_BYTE (inbuffer))
208 return -1;
209
210 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
211 {
212 /* Fetch the character code from the buffer. */
213 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, bytepos);
214 BUF_INC_POS (inbuffer, bytepos);
215 c = STRING_CHAR (p);
216 if (multibyte)
217 *multibyte = 1;
218 }
219 else
220 {
221 c = BUF_FETCH_BYTE (inbuffer, bytepos);
222 if (! ASCII_CHAR_P (c))
223 c = BYTE8_TO_CHAR (c);
224 bytepos++;
225 }
226
227 XMARKER (readcharfun)->bytepos = bytepos;
228 XMARKER (readcharfun)->charpos++;
229
230 return c;
231 }
232
233 if (EQ (readcharfun, Qlambda))
234 {
235 readbyte = readbyte_for_lambda;
236 goto read_multibyte;
237 }
238
239 if (EQ (readcharfun, Qget_file_char))
240 {
241 readbyte = readbyte_from_file;
242 goto read_multibyte;
243 }
244
245 if (STRINGP (readcharfun))
246 {
247 if (read_from_string_index >= read_from_string_limit)
248 c = -1;
249 else if (STRING_MULTIBYTE (readcharfun))
250 {
251 if (multibyte)
252 *multibyte = 1;
253 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c, readcharfun,
254 read_from_string_index,
255 read_from_string_index_byte);
256 }
257 else
258 {
259 c = SREF (readcharfun, read_from_string_index_byte);
260 read_from_string_index++;
261 read_from_string_index_byte++;
262 }
263 return c;
264 }
265
266 if (CONSP (readcharfun))
267 {
268 /* This is the case that read_vector is reading from a unibyte
269 string that contains a byte sequence previously skipped
270 because of #@NUMBER. The car part of readcharfun is that
271 string, and the cdr part is a value of readcharfun given to
272 read_vector. */
273 readbyte = readbyte_from_string;
274 if (EQ (XCDR (readcharfun), Qget_emacs_mule_file_char))
275 emacs_mule_encoding = 1;
276 goto read_multibyte;
277 }
278
279 if (EQ (readcharfun, Qget_emacs_mule_file_char))
280 {
281 readbyte = readbyte_from_file;
282 emacs_mule_encoding = 1;
283 goto read_multibyte;
284 }
285
286 tem = call0 (readcharfun);
287
288 if (NILP (tem))
289 return -1;
290 return XINT (tem);
291
292 read_multibyte:
293 if (unread_char >= 0)
294 {
295 c = unread_char;
296 unread_char = -1;
297 return c;
298 }
299 c = (*readbyte) (-1, readcharfun);
300 if (c < 0)
301 return c;
302 if (multibyte)
303 *multibyte = 1;
304 if (ASCII_CHAR_P (c))
305 return c;
306 if (emacs_mule_encoding)
307 return read_emacs_mule_char (c, readbyte, readcharfun);
308 i = 0;
309 buf[i++] = c;
310 len = BYTES_BY_CHAR_HEAD (c);
311 while (i < len)
312 {
313 c = (*readbyte) (-1, readcharfun);
314 if (c < 0 || ! TRAILING_CODE_P (c))
315 {
316 while (--i > 1)
317 (*readbyte) (buf[i], readcharfun);
318 return BYTE8_TO_CHAR (buf[0]);
319 }
320 buf[i++] = c;
321 }
322 return STRING_CHAR (buf);
323 }
324
325 #define FROM_FILE_P(readcharfun) \
326 (EQ (readcharfun, Qget_file_char) \
327 || EQ (readcharfun, Qget_emacs_mule_file_char))
328
329 static void
330 skip_dyn_bytes (Lisp_Object readcharfun, ptrdiff_t n)
331 {
332 if (FROM_FILE_P (readcharfun))
333 {
334 block_input (); /* FIXME: Not sure if it's needed. */
335 fseek (instream, n, SEEK_CUR);
336 unblock_input ();
337 }
338 else
339 { /* We're not reading directly from a file. In that case, it's difficult
340 to reliably count bytes, since these are usually meant for the file's
341 encoding, whereas we're now typically in the internal encoding.
342 But luckily, skip_dyn_bytes is used to skip over a single
343 dynamic-docstring (or dynamic byte-code) which is always quoted such
344 that \037 is the final char. */
345 int c;
346 do {
347 c = READCHAR;
348 } while (c >= 0 && c != '\037');
349 }
350 }
351
352 static void
353 skip_dyn_eof (Lisp_Object readcharfun)
354 {
355 if (FROM_FILE_P (readcharfun))
356 {
357 block_input (); /* FIXME: Not sure if it's needed. */
358 fseek (instream, 0, SEEK_END);
359 unblock_input ();
360 }
361 else
362 while (READCHAR >= 0);
363 }
364
365 /* Unread the character C in the way appropriate for the stream READCHARFUN.
366 If the stream is a user function, call it with the char as argument. */
367
368 static void
369 unreadchar (Lisp_Object readcharfun, int c)
370 {
371 readchar_count--;
372 if (c == -1)
373 /* Don't back up the pointer if we're unreading the end-of-input mark,
374 since readchar didn't advance it when we read it. */
375 ;
376 else if (BUFFERP (readcharfun))
377 {
378 struct buffer *b = XBUFFER (readcharfun);
379 ptrdiff_t charpos = BUF_PT (b);
380 ptrdiff_t bytepos = BUF_PT_BYTE (b);
381
382 if (! NILP (BVAR (b, enable_multibyte_characters)))
383 BUF_DEC_POS (b, bytepos);
384 else
385 bytepos--;
386
387 SET_BUF_PT_BOTH (b, charpos - 1, bytepos);
388 }
389 else if (MARKERP (readcharfun))
390 {
391 struct buffer *b = XMARKER (readcharfun)->buffer;
392 ptrdiff_t bytepos = XMARKER (readcharfun)->bytepos;
393
394 XMARKER (readcharfun)->charpos--;
395 if (! NILP (BVAR (b, enable_multibyte_characters)))
396 BUF_DEC_POS (b, bytepos);
397 else
398 bytepos--;
399
400 XMARKER (readcharfun)->bytepos = bytepos;
401 }
402 else if (STRINGP (readcharfun))
403 {
404 read_from_string_index--;
405 read_from_string_index_byte
406 = string_char_to_byte (readcharfun, read_from_string_index);
407 }
408 else if (CONSP (readcharfun))
409 {
410 unread_char = c;
411 }
412 else if (EQ (readcharfun, Qlambda))
413 {
414 unread_char = c;
415 }
416 else if (FROM_FILE_P (readcharfun))
417 {
418 unread_char = c;
419 }
420 else
421 call1 (readcharfun, make_number (c));
422 }
423
424 static int
425 readbyte_for_lambda (int c, Lisp_Object readcharfun)
426 {
427 return read_bytecode_char (c >= 0);
428 }
429
430
431 static int
432 readbyte_from_file (int c, Lisp_Object readcharfun)
433 {
434 if (c >= 0)
435 {
436 block_input ();
437 ungetc (c, instream);
438 unblock_input ();
439 return 0;
440 }
441
442 block_input ();
443 c = getc (instream);
444
445 /* Interrupted reads have been observed while reading over the network. */
446 while (c == EOF && ferror (instream) && errno == EINTR)
447 {
448 unblock_input ();
449 QUIT;
450 block_input ();
451 clearerr (instream);
452 c = getc (instream);
453 }
454
455 unblock_input ();
456
457 return (c == EOF ? -1 : c);
458 }
459
460 static int
461 readbyte_from_string (int c, Lisp_Object readcharfun)
462 {
463 Lisp_Object string = XCAR (readcharfun);
464
465 if (c >= 0)
466 {
467 read_from_string_index--;
468 read_from_string_index_byte
469 = string_char_to_byte (string, read_from_string_index);
470 }
471
472 if (read_from_string_index >= read_from_string_limit)
473 c = -1;
474 else
475 FETCH_STRING_CHAR_ADVANCE (c, string,
476 read_from_string_index,
477 read_from_string_index_byte);
478 return c;
479 }
480
481
482 /* Read one non-ASCII character from INSTREAM. The character is
483 encoded in `emacs-mule' and the first byte is already read in
484 C. */
485
486 static int
487 read_emacs_mule_char (int c, int (*readbyte) (int, Lisp_Object), Lisp_Object readcharfun)
488 {
489 /* Emacs-mule coding uses at most 4-byte for one character. */
490 unsigned char buf[4];
491 int len = emacs_mule_bytes[c];
492 struct charset *charset;
493 int i;
494 unsigned code;
495
496 if (len == 1)
497 /* C is not a valid leading-code of `emacs-mule'. */
498 return BYTE8_TO_CHAR (c);
499
500 i = 0;
501 buf[i++] = c;
502 while (i < len)
503 {
504 c = (*readbyte) (-1, readcharfun);
505 if (c < 0xA0)
506 {
507 while (--i > 1)
508 (*readbyte) (buf[i], readcharfun);
509 return BYTE8_TO_CHAR (buf[0]);
510 }
511 buf[i++] = c;
512 }
513
514 if (len == 2)
515 {
516 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
517 code = buf[1] & 0x7F;
518 }
519 else if (len == 3)
520 {
521 if (buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_11
522 || buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_12)
523 {
524 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
525 code = buf[2] & 0x7F;
526 }
527 else
528 {
529 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
530 code = ((buf[1] << 8) | buf[2]) & 0x7F7F;
531 }
532 }
533 else
534 {
535 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
536 code = ((buf[2] << 8) | buf[3]) & 0x7F7F;
537 }
538 c = DECODE_CHAR (charset, code);
539 if (c < 0)
540 Fsignal (Qinvalid_read_syntax,
541 list1 (build_string ("invalid multibyte form")));
542 return c;
543 }
544
545
546 static Lisp_Object read_internal_start (Lisp_Object, Lisp_Object,
547 Lisp_Object);
548 static Lisp_Object read0 (Lisp_Object);
549 static Lisp_Object read1 (Lisp_Object, int *, bool);
550
551 static Lisp_Object read_list (bool, Lisp_Object);
552 static Lisp_Object read_vector (Lisp_Object, bool);
553
554 static Lisp_Object substitute_object_recurse (Lisp_Object, Lisp_Object,
555 Lisp_Object);
556 static void substitute_object_in_subtree (Lisp_Object,
557 Lisp_Object);
558 static void substitute_in_interval (INTERVAL, Lisp_Object);
559
560 \f
561 /* Get a character from the tty. */
562
563 /* Read input events until we get one that's acceptable for our purposes.
564
565 If NO_SWITCH_FRAME, switch-frame events are stashed
566 until we get a character we like, and then stuffed into
567 unread_switch_frame.
568
569 If ASCII_REQUIRED, check function key events to see
570 if the unmodified version of the symbol has a Qascii_character
571 property, and use that character, if present.
572
573 If ERROR_NONASCII, signal an error if the input we
574 get isn't an ASCII character with modifiers. If it's false but
575 ASCII_REQUIRED is true, just re-read until we get an ASCII
576 character.
577
578 If INPUT_METHOD, invoke the current input method
579 if the character warrants that.
580
581 If SECONDS is a number, wait that many seconds for input, and
582 return Qnil if no input arrives within that time. */
583
584 static Lisp_Object
585 read_filtered_event (bool no_switch_frame, bool ascii_required,
586 bool error_nonascii, bool input_method, Lisp_Object seconds)
587 {
588 Lisp_Object val, delayed_switch_frame;
589 struct timespec end_time;
590
591 #ifdef HAVE_WINDOW_SYSTEM
592 if (display_hourglass_p)
593 cancel_hourglass ();
594 #endif
595
596 delayed_switch_frame = Qnil;
597
598 /* Compute timeout. */
599 if (NUMBERP (seconds))
600 {
601 double duration = extract_float (seconds);
602 struct timespec wait_time = dtotimespec (duration);
603 end_time = timespec_add (current_timespec (), wait_time);
604 }
605
606 /* Read until we get an acceptable event. */
607 retry:
608 do
609 val = read_char (0, Qnil, (input_method ? Qnil : Qt), 0,
610 NUMBERP (seconds) ? &end_time : NULL);
611 while (INTEGERP (val) && XINT (val) == -2); /* wrong_kboard_jmpbuf */
612
613 if (BUFFERP (val))
614 goto retry;
615
616 /* `switch-frame' events are put off until after the next ASCII
617 character. This is better than signaling an error just because
618 the last characters were typed to a separate minibuffer frame,
619 for example. Eventually, some code which can deal with
620 switch-frame events will read it and process it. */
621 if (no_switch_frame
622 && EVENT_HAS_PARAMETERS (val)
623 && EQ (EVENT_HEAD_KIND (EVENT_HEAD (val)), Qswitch_frame))
624 {
625 delayed_switch_frame = val;
626 goto retry;
627 }
628
629 if (ascii_required && !(NUMBERP (seconds) && NILP (val)))
630 {
631 /* Convert certain symbols to their ASCII equivalents. */
632 if (SYMBOLP (val))
633 {
634 Lisp_Object tem, tem1;
635 tem = Fget (val, Qevent_symbol_element_mask);
636 if (!NILP (tem))
637 {
638 tem1 = Fget (Fcar (tem), Qascii_character);
639 /* Merge this symbol's modifier bits
640 with the ASCII equivalent of its basic code. */
641 if (!NILP (tem1))
642 XSETFASTINT (val, XINT (tem1) | XINT (Fcar (Fcdr (tem))));
643 }
644 }
645
646 /* If we don't have a character now, deal with it appropriately. */
647 if (!INTEGERP (val))
648 {
649 if (error_nonascii)
650 {
651 Vunread_command_events = list1 (val);
652 error ("Non-character input-event");
653 }
654 else
655 goto retry;
656 }
657 }
658
659 if (! NILP (delayed_switch_frame))
660 unread_switch_frame = delayed_switch_frame;
661
662 #if 0
663
664 #ifdef HAVE_WINDOW_SYSTEM
665 if (display_hourglass_p)
666 start_hourglass ();
667 #endif
668
669 #endif
670
671 return val;
672 }
673
674 DEFUN ("read-char", Fread_char, Sread_char, 0, 3, 0,
675 doc: /* Read a character from the command input (keyboard or macro).
676 It is returned as a number.
677 If the character has modifiers, they are resolved and reflected to the
678 character code if possible (e.g. C-SPC -> 0).
679
680 If the user generates an event which is not a character (i.e. a mouse
681 click or function key event), `read-char' signals an error. As an
682 exception, switch-frame events are put off until non-character events
683 can be read.
684 If you want to read non-character events, or ignore them, call
685 `read-event' or `read-char-exclusive' instead.
686
687 If the optional argument PROMPT is non-nil, display that as a prompt.
688 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
689 input method is turned on in the current buffer, that input method
690 is used for reading a character.
691 If the optional argument SECONDS is non-nil, it should be a number
692 specifying the maximum number of seconds to wait for input. If no
693 input arrives in that time, return nil. SECONDS may be a
694 floating-point value. */)
695 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
696 {
697 Lisp_Object val;
698
699 if (! NILP (prompt))
700 message_with_string ("%s", prompt, 0);
701 val = read_filtered_event (1, 1, 1, ! NILP (inherit_input_method), seconds);
702
703 return (NILP (val) ? Qnil
704 : make_number (char_resolve_modifier_mask (XINT (val))));
705 }
706
707 DEFUN ("read-event", Fread_event, Sread_event, 0, 3, 0,
708 doc: /* Read an event object from the input stream.
709 If the optional argument PROMPT is non-nil, display that as a prompt.
710 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
711 input method is turned on in the current buffer, that input method
712 is used for reading a character.
713 If the optional argument SECONDS is non-nil, it should be a number
714 specifying the maximum number of seconds to wait for input. If no
715 input arrives in that time, return nil. SECONDS may be a
716 floating-point value. */)
717 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
718 {
719 if (! NILP (prompt))
720 message_with_string ("%s", prompt, 0);
721 return read_filtered_event (0, 0, 0, ! NILP (inherit_input_method), seconds);
722 }
723
724 DEFUN ("read-char-exclusive", Fread_char_exclusive, Sread_char_exclusive, 0, 3, 0,
725 doc: /* Read a character from the command input (keyboard or macro).
726 It is returned as a number. Non-character events are ignored.
727 If the character has modifiers, they are resolved and reflected to the
728 character code if possible (e.g. C-SPC -> 0).
729
730 If the optional argument PROMPT is non-nil, display that as a prompt.
731 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
732 input method is turned on in the current buffer, that input method
733 is used for reading a character.
734 If the optional argument SECONDS is non-nil, it should be a number
735 specifying the maximum number of seconds to wait for input. If no
736 input arrives in that time, return nil. SECONDS may be a
737 floating-point value. */)
738 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
739 {
740 Lisp_Object val;
741
742 if (! NILP (prompt))
743 message_with_string ("%s", prompt, 0);
744
745 val = read_filtered_event (1, 1, 0, ! NILP (inherit_input_method), seconds);
746
747 return (NILP (val) ? Qnil
748 : make_number (char_resolve_modifier_mask (XINT (val))));
749 }
750
751 DEFUN ("get-file-char", Fget_file_char, Sget_file_char, 0, 0, 0,
752 doc: /* Don't use this yourself. */)
753 (void)
754 {
755 register Lisp_Object val;
756 block_input ();
757 XSETINT (val, getc (instream));
758 unblock_input ();
759 return val;
760 }
761
762
763 \f
764
765 /* Return true if the lisp code read using READCHARFUN defines a non-nil
766 `lexical-binding' file variable. After returning, the stream is
767 positioned following the first line, if it is a comment or #! line,
768 otherwise nothing is read. */
769
770 static bool
771 lisp_file_lexically_bound_p (Lisp_Object readcharfun)
772 {
773 int ch = READCHAR;
774
775 if (ch == '#')
776 {
777 ch = READCHAR;
778 if (ch != '!')
779 {
780 UNREAD (ch);
781 UNREAD ('#');
782 return 0;
783 }
784 while (ch != '\n' && ch != EOF)
785 ch = READCHAR;
786 if (ch == '\n') ch = READCHAR;
787 /* It is OK to leave the position after a #! line, since
788 that is what read1 does. */
789 }
790
791 if (ch != ';')
792 /* The first line isn't a comment, just give up. */
793 {
794 UNREAD (ch);
795 return 0;
796 }
797 else
798 /* Look for an appropriate file-variable in the first line. */
799 {
800 bool rv = 0;
801 enum {
802 NOMINAL, AFTER_FIRST_DASH, AFTER_ASTERIX
803 } beg_end_state = NOMINAL;
804 bool in_file_vars = 0;
805
806 #define UPDATE_BEG_END_STATE(ch) \
807 if (beg_end_state == NOMINAL) \
808 beg_end_state = (ch == '-' ? AFTER_FIRST_DASH : NOMINAL); \
809 else if (beg_end_state == AFTER_FIRST_DASH) \
810 beg_end_state = (ch == '*' ? AFTER_ASTERIX : NOMINAL); \
811 else if (beg_end_state == AFTER_ASTERIX) \
812 { \
813 if (ch == '-') \
814 in_file_vars = !in_file_vars; \
815 beg_end_state = NOMINAL; \
816 }
817
818 /* Skip until we get to the file vars, if any. */
819 do
820 {
821 ch = READCHAR;
822 UPDATE_BEG_END_STATE (ch);
823 }
824 while (!in_file_vars && ch != '\n' && ch != EOF);
825
826 while (in_file_vars)
827 {
828 char var[100], val[100];
829 unsigned i;
830
831 ch = READCHAR;
832
833 /* Read a variable name. */
834 while (ch == ' ' || ch == '\t')
835 ch = READCHAR;
836
837 i = 0;
838 while (ch != ':' && ch != '\n' && ch != EOF && in_file_vars)
839 {
840 if (i < sizeof var - 1)
841 var[i++] = ch;
842 UPDATE_BEG_END_STATE (ch);
843 ch = READCHAR;
844 }
845
846 /* Stop scanning if no colon was found before end marker. */
847 if (!in_file_vars || ch == '\n' || ch == EOF)
848 break;
849
850 while (i > 0 && (var[i - 1] == ' ' || var[i - 1] == '\t'))
851 i--;
852 var[i] = '\0';
853
854 if (ch == ':')
855 {
856 /* Read a variable value. */
857 ch = READCHAR;
858
859 while (ch == ' ' || ch == '\t')
860 ch = READCHAR;
861
862 i = 0;
863 while (ch != ';' && ch != '\n' && ch != EOF && in_file_vars)
864 {
865 if (i < sizeof val - 1)
866 val[i++] = ch;
867 UPDATE_BEG_END_STATE (ch);
868 ch = READCHAR;
869 }
870 if (! in_file_vars)
871 /* The value was terminated by an end-marker, which remove. */
872 i -= 3;
873 while (i > 0 && (val[i - 1] == ' ' || val[i - 1] == '\t'))
874 i--;
875 val[i] = '\0';
876
877 if (strcmp (var, "lexical-binding") == 0)
878 /* This is it... */
879 {
880 rv = (strcmp (val, "nil") != 0);
881 break;
882 }
883 }
884 }
885
886 while (ch != '\n' && ch != EOF)
887 ch = READCHAR;
888
889 return rv;
890 }
891 }
892 \f
893 /* Value is a version number of byte compiled code if the file
894 associated with file descriptor FD is a compiled Lisp file that's
895 safe to load. Only files compiled with Emacs are safe to load.
896 Files compiled with XEmacs can lead to a crash in Fbyte_code
897 because of an incompatible change in the byte compiler. */
898
899 static int
900 safe_to_load_version (int fd)
901 {
902 char buf[512];
903 int nbytes, i;
904 int version = 1;
905
906 /* Read the first few bytes from the file, and look for a line
907 specifying the byte compiler version used. */
908 nbytes = emacs_read (fd, buf, sizeof buf);
909 if (nbytes > 0)
910 {
911 /* Skip to the next newline, skipping over the initial `ELC'
912 with NUL bytes following it, but note the version. */
913 for (i = 0; i < nbytes && buf[i] != '\n'; ++i)
914 if (i == 4)
915 version = buf[i];
916
917 if (i >= nbytes
918 || fast_c_string_match_ignore_case (Vbytecomp_version_regexp,
919 buf + i, nbytes - i) < 0)
920 version = 0;
921 }
922
923 lseek (fd, 0, SEEK_SET);
924 return version;
925 }
926
927
928 /* Callback for record_unwind_protect. Restore the old load list OLD,
929 after loading a file successfully. */
930
931 static void
932 record_load_unwind (Lisp_Object old)
933 {
934 Vloads_in_progress = old;
935 }
936
937 /* This handler function is used via internal_condition_case_1. */
938
939 static Lisp_Object
940 load_error_handler (Lisp_Object data)
941 {
942 return Qnil;
943 }
944
945 static void
946 load_warn_old_style_backquotes (Lisp_Object file)
947 {
948 if (!NILP (Vold_style_backquotes))
949 {
950 AUTO_STRING (format, "Loading `%s': old-style backquotes detected!");
951 CALLN (Fmessage, format, file);
952 }
953 }
954
955 DEFUN ("get-load-suffixes", Fget_load_suffixes, Sget_load_suffixes, 0, 0, 0,
956 doc: /* Return the suffixes that `load' should try if a suffix is \
957 required.
958 This uses the variables `load-suffixes' and `load-file-rep-suffixes'. */)
959 (void)
960 {
961 Lisp_Object lst = Qnil, suffixes = Vload_suffixes, suffix, ext;
962 while (CONSP (suffixes))
963 {
964 Lisp_Object exts = Vload_file_rep_suffixes;
965 suffix = XCAR (suffixes);
966 suffixes = XCDR (suffixes);
967 while (CONSP (exts))
968 {
969 ext = XCAR (exts);
970 exts = XCDR (exts);
971 lst = Fcons (concat2 (suffix, ext), lst);
972 }
973 }
974 return Fnreverse (lst);
975 }
976
977 DEFUN ("load", Fload, Sload, 1, 5, 0,
978 doc: /* Execute a file of Lisp code named FILE.
979 First try FILE with `.elc' appended, then try with `.el',
980 then try FILE unmodified (the exact suffixes in the exact order are
981 determined by `load-suffixes'). Environment variable references in
982 FILE are replaced with their values by calling `substitute-in-file-name'.
983 This function searches the directories in `load-path'.
984
985 If optional second arg NOERROR is non-nil,
986 report no error if FILE doesn't exist.
987 Print messages at start and end of loading unless
988 optional third arg NOMESSAGE is non-nil (but `force-load-messages'
989 overrides that).
990 If optional fourth arg NOSUFFIX is non-nil, don't try adding
991 suffixes `.elc' or `.el' to the specified name FILE.
992 If optional fifth arg MUST-SUFFIX is non-nil, insist on
993 the suffix `.elc' or `.el'; don't accept just FILE unless
994 it ends in one of those suffixes or includes a directory name.
995
996 If NOSUFFIX is nil, then if a file could not be found, try looking for
997 a different representation of the file by adding non-empty suffixes to
998 its name, before trying another file. Emacs uses this feature to find
999 compressed versions of files when Auto Compression mode is enabled.
1000 If NOSUFFIX is non-nil, disable this feature.
1001
1002 The suffixes that this function tries out, when NOSUFFIX is nil, are
1003 given by the return value of `get-load-suffixes' and the values listed
1004 in `load-file-rep-suffixes'. If MUST-SUFFIX is non-nil, only the
1005 return value of `get-load-suffixes' is used, i.e. the file name is
1006 required to have a non-empty suffix.
1007
1008 When searching suffixes, this function normally stops at the first
1009 one that exists. If the option `load-prefer-newer' is non-nil,
1010 however, it tries all suffixes, and uses whichever file is the newest.
1011
1012 Loading a file records its definitions, and its `provide' and
1013 `require' calls, in an element of `load-history' whose
1014 car is the file name loaded. See `load-history'.
1015
1016 While the file is in the process of being loaded, the variable
1017 `load-in-progress' is non-nil and the variable `load-file-name'
1018 is bound to the file's name.
1019
1020 Return t if the file exists and loads successfully. */)
1021 (Lisp_Object file, Lisp_Object noerror, Lisp_Object nomessage,
1022 Lisp_Object nosuffix, Lisp_Object must_suffix)
1023 {
1024 FILE *stream;
1025 int fd;
1026 int fd_index;
1027 ptrdiff_t count = SPECPDL_INDEX ();
1028 struct gcpro gcpro1, gcpro2, gcpro3;
1029 Lisp_Object found, efound, hist_file_name;
1030 /* True means we printed the ".el is newer" message. */
1031 bool newer = 0;
1032 /* True means we are loading a compiled file. */
1033 bool compiled = 0;
1034 Lisp_Object handler;
1035 bool safe_p = 1;
1036 const char *fmode = "r" FOPEN_TEXT;
1037 int version;
1038
1039 CHECK_STRING (file);
1040
1041 /* If file name is magic, call the handler. */
1042 /* This shouldn't be necessary any more now that `openp' handles it right.
1043 handler = Ffind_file_name_handler (file, Qload);
1044 if (!NILP (handler))
1045 return call5 (handler, Qload, file, noerror, nomessage, nosuffix); */
1046
1047 /* Do this after the handler to avoid
1048 the need to gcpro noerror, nomessage and nosuffix.
1049 (Below here, we care only whether they are nil or not.)
1050 The presence of this call is the result of a historical accident:
1051 it used to be in every file-operation and when it got removed
1052 everywhere, it accidentally stayed here. Since then, enough people
1053 supposedly have things like (load "$PROJECT/foo.el") in their .emacs
1054 that it seemed risky to remove. */
1055 if (! NILP (noerror))
1056 {
1057 file = internal_condition_case_1 (Fsubstitute_in_file_name, file,
1058 Qt, load_error_handler);
1059 if (NILP (file))
1060 return Qnil;
1061 }
1062 else
1063 file = Fsubstitute_in_file_name (file);
1064
1065 /* Avoid weird lossage with null string as arg,
1066 since it would try to load a directory as a Lisp file. */
1067 if (SCHARS (file) == 0)
1068 {
1069 fd = -1;
1070 errno = ENOENT;
1071 }
1072 else
1073 {
1074 Lisp_Object suffixes;
1075 found = Qnil;
1076 GCPRO2 (file, found);
1077
1078 if (! NILP (must_suffix))
1079 {
1080 /* Don't insist on adding a suffix if FILE already ends with one. */
1081 ptrdiff_t size = SBYTES (file);
1082 if (size > 3
1083 && !strcmp (SSDATA (file) + size - 3, ".el"))
1084 must_suffix = Qnil;
1085 else if (size > 4
1086 && !strcmp (SSDATA (file) + size - 4, ".elc"))
1087 must_suffix = Qnil;
1088 /* Don't insist on adding a suffix
1089 if the argument includes a directory name. */
1090 else if (! NILP (Ffile_name_directory (file)))
1091 must_suffix = Qnil;
1092 }
1093
1094 if (!NILP (nosuffix))
1095 suffixes = Qnil;
1096 else
1097 {
1098 suffixes = Fget_load_suffixes ();
1099 if (NILP (must_suffix))
1100 suffixes = CALLN (Fappend, suffixes, Vload_file_rep_suffixes);
1101 }
1102
1103 fd = openp (Vload_path, file, suffixes, &found, Qnil, load_prefer_newer);
1104 UNGCPRO;
1105 }
1106
1107 if (fd == -1)
1108 {
1109 if (NILP (noerror))
1110 report_file_error ("Cannot open load file", file);
1111 return Qnil;
1112 }
1113
1114 /* Tell startup.el whether or not we found the user's init file. */
1115 if (EQ (Qt, Vuser_init_file))
1116 Vuser_init_file = found;
1117
1118 /* If FD is -2, that means openp found a magic file. */
1119 if (fd == -2)
1120 {
1121 if (NILP (Fequal (found, file)))
1122 /* If FOUND is a different file name from FILE,
1123 find its handler even if we have already inhibited
1124 the `load' operation on FILE. */
1125 handler = Ffind_file_name_handler (found, Qt);
1126 else
1127 handler = Ffind_file_name_handler (found, Qload);
1128 if (! NILP (handler))
1129 return call5 (handler, Qload, found, noerror, nomessage, Qt);
1130 #ifdef DOS_NT
1131 /* Tramp has to deal with semi-broken packages that prepend
1132 drive letters to remote files. For that reason, Tramp
1133 catches file operations that test for file existence, which
1134 makes openp think X:/foo.elc files are remote. However,
1135 Tramp does not catch `load' operations for such files, so we
1136 end up with a nil as the `load' handler above. If we would
1137 continue with fd = -2, we will behave wrongly, and in
1138 particular try reading a .elc file in the "rt" mode instead
1139 of "rb". See bug #9311 for the results. To work around
1140 this, we try to open the file locally, and go with that if it
1141 succeeds. */
1142 fd = emacs_open (SSDATA (ENCODE_FILE (found)), O_RDONLY, 0);
1143 if (fd == -1)
1144 fd = -2;
1145 #endif
1146 }
1147
1148 if (fd < 0)
1149 {
1150 /* Pacify older GCC with --enable-gcc-warnings. */
1151 IF_LINT (fd_index = 0);
1152 }
1153 else
1154 {
1155 fd_index = SPECPDL_INDEX ();
1156 record_unwind_protect_int (close_file_unwind, fd);
1157 }
1158
1159 /* Check if we're stuck in a recursive load cycle.
1160
1161 2000-09-21: It's not possible to just check for the file loaded
1162 being a member of Vloads_in_progress. This fails because of the
1163 way the byte compiler currently works; `provide's are not
1164 evaluated, see font-lock.el/jit-lock.el as an example. This
1165 leads to a certain amount of ``normal'' recursion.
1166
1167 Also, just loading a file recursively is not always an error in
1168 the general case; the second load may do something different. */
1169 {
1170 int load_count = 0;
1171 Lisp_Object tem;
1172 for (tem = Vloads_in_progress; CONSP (tem); tem = XCDR (tem))
1173 if (!NILP (Fequal (found, XCAR (tem))) && (++load_count > 3))
1174 signal_error ("Recursive load", Fcons (found, Vloads_in_progress));
1175 record_unwind_protect (record_load_unwind, Vloads_in_progress);
1176 Vloads_in_progress = Fcons (found, Vloads_in_progress);
1177 }
1178
1179 /* All loads are by default dynamic, unless the file itself specifies
1180 otherwise using a file-variable in the first line. This is bound here
1181 so that it takes effect whether or not we use
1182 Vload_source_file_function. */
1183 specbind (Qlexical_binding, Qnil);
1184
1185 /* Get the name for load-history. */
1186 hist_file_name = (! NILP (Vpurify_flag)
1187 ? concat2 (Ffile_name_directory (file),
1188 Ffile_name_nondirectory (found))
1189 : found) ;
1190
1191 version = -1;
1192
1193 /* Check for the presence of old-style quotes and warn about them. */
1194 specbind (Qold_style_backquotes, Qnil);
1195 record_unwind_protect (load_warn_old_style_backquotes, file);
1196
1197 if (!memcmp (SDATA (found) + SBYTES (found) - 4, ".elc", 4)
1198 || (fd >= 0 && (version = safe_to_load_version (fd)) > 0))
1199 /* Load .elc files directly, but not when they are
1200 remote and have no handler! */
1201 {
1202 if (fd != -2)
1203 {
1204 struct stat s1, s2;
1205 int result;
1206
1207 GCPRO3 (file, found, hist_file_name);
1208
1209 if (version < 0
1210 && ! (version = safe_to_load_version (fd)))
1211 {
1212 safe_p = 0;
1213 if (!load_dangerous_libraries)
1214 error ("File `%s' was not compiled in Emacs", SDATA (found));
1215 else if (!NILP (nomessage) && !force_load_messages)
1216 message_with_string ("File `%s' not compiled in Emacs", found, 1);
1217 }
1218
1219 compiled = 1;
1220
1221 efound = ENCODE_FILE (found);
1222 fmode = "r" FOPEN_BINARY;
1223
1224 /* openp already checked for newness, no point doing it again.
1225 FIXME would be nice to get a message when openp
1226 ignores suffix order due to load_prefer_newer. */
1227 if (!load_prefer_newer)
1228 {
1229 result = stat (SSDATA (efound), &s1);
1230 if (result == 0)
1231 {
1232 SSET (efound, SBYTES (efound) - 1, 0);
1233 result = stat (SSDATA (efound), &s2);
1234 SSET (efound, SBYTES (efound) - 1, 'c');
1235 }
1236
1237 if (result == 0
1238 && timespec_cmp (get_stat_mtime (&s1), get_stat_mtime (&s2)) < 0)
1239 {
1240 /* Make the progress messages mention that source is newer. */
1241 newer = 1;
1242
1243 /* If we won't print another message, mention this anyway. */
1244 if (!NILP (nomessage) && !force_load_messages)
1245 {
1246 Lisp_Object msg_file;
1247 msg_file = Fsubstring (found, make_number (0), make_number (-1));
1248 message_with_string ("Source file `%s' newer than byte-compiled file",
1249 msg_file, 1);
1250 }
1251 }
1252 } /* !load_prefer_newer */
1253 UNGCPRO;
1254 }
1255 }
1256 else
1257 {
1258 /* We are loading a source file (*.el). */
1259 if (!NILP (Vload_source_file_function))
1260 {
1261 Lisp_Object val;
1262
1263 if (fd >= 0)
1264 {
1265 emacs_close (fd);
1266 clear_unwind_protect (fd_index);
1267 }
1268 val = call4 (Vload_source_file_function, found, hist_file_name,
1269 NILP (noerror) ? Qnil : Qt,
1270 (NILP (nomessage) || force_load_messages) ? Qnil : Qt);
1271 return unbind_to (count, val);
1272 }
1273 }
1274
1275 GCPRO3 (file, found, hist_file_name);
1276
1277 if (fd < 0)
1278 {
1279 /* We somehow got here with fd == -2, meaning the file is deemed
1280 to be remote. Don't even try to reopen the file locally;
1281 just force a failure. */
1282 stream = NULL;
1283 errno = EINVAL;
1284 }
1285 else
1286 {
1287 #ifdef WINDOWSNT
1288 emacs_close (fd);
1289 clear_unwind_protect (fd_index);
1290 efound = ENCODE_FILE (found);
1291 stream = emacs_fopen (SSDATA (efound), fmode);
1292 #else
1293 stream = fdopen (fd, fmode);
1294 #endif
1295 }
1296 if (! stream)
1297 report_file_error ("Opening stdio stream", file);
1298 set_unwind_protect_ptr (fd_index, fclose_unwind, stream);
1299
1300 if (! NILP (Vpurify_flag))
1301 Vpreloaded_file_list = Fcons (Fpurecopy (file), Vpreloaded_file_list);
1302
1303 if (NILP (nomessage) || force_load_messages)
1304 {
1305 if (!safe_p)
1306 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...",
1307 file, 1);
1308 else if (!compiled)
1309 message_with_string ("Loading %s (source)...", file, 1);
1310 else if (newer)
1311 message_with_string ("Loading %s (compiled; note, source file is newer)...",
1312 file, 1);
1313 else /* The typical case; compiled file newer than source file. */
1314 message_with_string ("Loading %s...", file, 1);
1315 }
1316
1317 specbind (Qload_file_name, found);
1318 specbind (Qinhibit_file_name_operation, Qnil);
1319 specbind (Qload_in_progress, Qt);
1320
1321 instream = stream;
1322 if (lisp_file_lexically_bound_p (Qget_file_char))
1323 Fset (Qlexical_binding, Qt);
1324
1325 if (! version || version >= 22)
1326 readevalloop (Qget_file_char, stream, hist_file_name,
1327 0, Qnil, Qnil, Qnil, Qnil);
1328 else
1329 {
1330 /* We can't handle a file which was compiled with
1331 byte-compile-dynamic by older version of Emacs. */
1332 specbind (Qload_force_doc_strings, Qt);
1333 readevalloop (Qget_emacs_mule_file_char, stream, hist_file_name,
1334 0, Qnil, Qnil, Qnil, Qnil);
1335 }
1336 unbind_to (count, Qnil);
1337
1338 /* Run any eval-after-load forms for this file. */
1339 if (!NILP (Ffboundp (Qdo_after_load_evaluation)))
1340 call1 (Qdo_after_load_evaluation, hist_file_name) ;
1341
1342 UNGCPRO;
1343
1344 xfree (saved_doc_string);
1345 saved_doc_string = 0;
1346 saved_doc_string_size = 0;
1347
1348 xfree (prev_saved_doc_string);
1349 prev_saved_doc_string = 0;
1350 prev_saved_doc_string_size = 0;
1351
1352 if (!noninteractive && (NILP (nomessage) || force_load_messages))
1353 {
1354 if (!safe_p)
1355 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...done",
1356 file, 1);
1357 else if (!compiled)
1358 message_with_string ("Loading %s (source)...done", file, 1);
1359 else if (newer)
1360 message_with_string ("Loading %s (compiled; note, source file is newer)...done",
1361 file, 1);
1362 else /* The typical case; compiled file newer than source file. */
1363 message_with_string ("Loading %s...done", file, 1);
1364 }
1365
1366 return Qt;
1367 }
1368 \f
1369 static bool
1370 complete_filename_p (Lisp_Object pathname)
1371 {
1372 const unsigned char *s = SDATA (pathname);
1373 return (IS_DIRECTORY_SEP (s[0])
1374 || (SCHARS (pathname) > 2
1375 && IS_DEVICE_SEP (s[1]) && IS_DIRECTORY_SEP (s[2])));
1376 }
1377
1378 DEFUN ("locate-file-internal", Flocate_file_internal, Slocate_file_internal, 2, 4, 0,
1379 doc: /* Search for FILENAME through PATH.
1380 Returns the file's name in absolute form, or nil if not found.
1381 If SUFFIXES is non-nil, it should be a list of suffixes to append to
1382 file name when searching.
1383 If non-nil, PREDICATE is used instead of `file-readable-p'.
1384 PREDICATE can also be an integer to pass to the faccessat(2) function,
1385 in which case file-name-handlers are ignored.
1386 This function will normally skip directories, so if you want it to find
1387 directories, make sure the PREDICATE function returns `dir-ok' for them. */)
1388 (Lisp_Object filename, Lisp_Object path, Lisp_Object suffixes, Lisp_Object predicate)
1389 {
1390 Lisp_Object file;
1391 int fd = openp (path, filename, suffixes, &file, predicate, false);
1392 if (NILP (predicate) && fd >= 0)
1393 emacs_close (fd);
1394 return file;
1395 }
1396
1397 /* Search for a file whose name is STR, looking in directories
1398 in the Lisp list PATH, and trying suffixes from SUFFIX.
1399 On success, return a file descriptor (or 1 or -2 as described below).
1400 On failure, return -1 and set errno.
1401
1402 SUFFIXES is a list of strings containing possible suffixes.
1403 The empty suffix is automatically added if the list is empty.
1404
1405 PREDICATE non-nil means don't open the files,
1406 just look for one that satisfies the predicate. In this case,
1407 return 1 on success. The predicate can be a lisp function or
1408 an integer to pass to `access' (in which case file-name-handlers
1409 are ignored).
1410
1411 If STOREPTR is nonzero, it points to a slot where the name of
1412 the file actually found should be stored as a Lisp string.
1413 nil is stored there on failure.
1414
1415 If the file we find is remote, return -2
1416 but store the found remote file name in *STOREPTR.
1417
1418 If NEWER is true, try all SUFFIXes and return the result for the
1419 newest file that exists. Does not apply to remote files,
1420 or if PREDICATE is specified. */
1421
1422 int
1423 openp (Lisp_Object path, Lisp_Object str, Lisp_Object suffixes,
1424 Lisp_Object *storeptr, Lisp_Object predicate, bool newer)
1425 {
1426 ptrdiff_t fn_size = 100;
1427 char buf[100];
1428 char *fn = buf;
1429 bool absolute;
1430 ptrdiff_t want_length;
1431 Lisp_Object filename;
1432 struct gcpro gcpro1, gcpro2, gcpro3, gcpro4, gcpro5, gcpro6, gcpro7;
1433 Lisp_Object string, tail, encoded_fn, save_string;
1434 ptrdiff_t max_suffix_len = 0;
1435 int last_errno = ENOENT;
1436 int save_fd = -1;
1437 USE_SAFE_ALLOCA;
1438
1439 /* The last-modified time of the newest matching file found.
1440 Initialize it to something less than all valid timestamps. */
1441 struct timespec save_mtime = make_timespec (TYPE_MINIMUM (time_t), -1);
1442
1443 CHECK_STRING (str);
1444
1445 for (tail = suffixes; CONSP (tail); tail = XCDR (tail))
1446 {
1447 CHECK_STRING_CAR (tail);
1448 max_suffix_len = max (max_suffix_len,
1449 SBYTES (XCAR (tail)));
1450 }
1451
1452 string = filename = encoded_fn = save_string = Qnil;
1453 GCPRO7 (str, string, save_string, filename, path, suffixes, encoded_fn);
1454
1455 if (storeptr)
1456 *storeptr = Qnil;
1457
1458 absolute = complete_filename_p (str);
1459
1460 for (; CONSP (path); path = XCDR (path))
1461 {
1462 filename = Fexpand_file_name (str, XCAR (path));
1463 if (!complete_filename_p (filename))
1464 /* If there are non-absolute elts in PATH (eg "."). */
1465 /* Of course, this could conceivably lose if luser sets
1466 default-directory to be something non-absolute... */
1467 {
1468 filename = Fexpand_file_name (filename, BVAR (current_buffer, directory));
1469 if (!complete_filename_p (filename))
1470 /* Give up on this path element! */
1471 continue;
1472 }
1473
1474 /* Calculate maximum length of any filename made from
1475 this path element/specified file name and any possible suffix. */
1476 want_length = max_suffix_len + SBYTES (filename);
1477 if (fn_size <= want_length)
1478 {
1479 fn_size = 100 + want_length;
1480 fn = SAFE_ALLOCA (fn_size);
1481 }
1482
1483 /* Loop over suffixes. */
1484 for (tail = NILP (suffixes) ? list1 (empty_unibyte_string) : suffixes;
1485 CONSP (tail); tail = XCDR (tail))
1486 {
1487 Lisp_Object suffix = XCAR (tail);
1488 ptrdiff_t fnlen, lsuffix = SBYTES (suffix);
1489 Lisp_Object handler;
1490
1491 /* Concatenate path element/specified name with the suffix.
1492 If the directory starts with /:, remove that. */
1493 int prefixlen = ((SCHARS (filename) > 2
1494 && SREF (filename, 0) == '/'
1495 && SREF (filename, 1) == ':')
1496 ? 2 : 0);
1497 fnlen = SBYTES (filename) - prefixlen;
1498 memcpy (fn, SDATA (filename) + prefixlen, fnlen);
1499 memcpy (fn + fnlen, SDATA (suffix), lsuffix + 1);
1500 fnlen += lsuffix;
1501 /* Check that the file exists and is not a directory. */
1502 /* We used to only check for handlers on non-absolute file names:
1503 if (absolute)
1504 handler = Qnil;
1505 else
1506 handler = Ffind_file_name_handler (filename, Qfile_exists_p);
1507 It's not clear why that was the case and it breaks things like
1508 (load "/bar.el") where the file is actually "/bar.el.gz". */
1509 /* make_string has its own ideas on when to return a unibyte
1510 string and when a multibyte string, but we know better.
1511 We must have a unibyte string when dumping, since
1512 file-name encoding is shaky at best at that time, and in
1513 particular default-file-name-coding-system is reset
1514 several times during loadup. We therefore don't want to
1515 encode the file before passing it to file I/O library
1516 functions. */
1517 if (!STRING_MULTIBYTE (filename) && !STRING_MULTIBYTE (suffix))
1518 string = make_unibyte_string (fn, fnlen);
1519 else
1520 string = make_string (fn, fnlen);
1521 handler = Ffind_file_name_handler (string, Qfile_exists_p);
1522 if ((!NILP (handler) || !NILP (predicate)) && !NATNUMP (predicate))
1523 {
1524 bool exists;
1525 if (NILP (predicate))
1526 exists = !NILP (Ffile_readable_p (string));
1527 else
1528 {
1529 Lisp_Object tmp = call1 (predicate, string);
1530 if (NILP (tmp))
1531 exists = false;
1532 else if (EQ (tmp, Qdir_ok)
1533 || NILP (Ffile_directory_p (string)))
1534 exists = true;
1535 else
1536 {
1537 exists = false;
1538 last_errno = EISDIR;
1539 }
1540 }
1541
1542 if (exists)
1543 {
1544 /* We succeeded; return this descriptor and filename. */
1545 if (storeptr)
1546 *storeptr = string;
1547 SAFE_FREE ();
1548 UNGCPRO;
1549 return -2;
1550 }
1551 }
1552 else
1553 {
1554 int fd;
1555 const char *pfn;
1556 struct stat st;
1557
1558 encoded_fn = ENCODE_FILE (string);
1559 pfn = SSDATA (encoded_fn);
1560
1561 /* Check that we can access or open it. */
1562 if (NATNUMP (predicate))
1563 {
1564 fd = -1;
1565 if (INT_MAX < XFASTINT (predicate))
1566 last_errno = EINVAL;
1567 else if (faccessat (AT_FDCWD, pfn, XFASTINT (predicate),
1568 AT_EACCESS)
1569 == 0)
1570 {
1571 if (file_directory_p (pfn))
1572 last_errno = EISDIR;
1573 else
1574 fd = 1;
1575 }
1576 }
1577 else
1578 {
1579 fd = emacs_open (pfn, O_RDONLY, 0);
1580 if (fd < 0)
1581 {
1582 if (errno != ENOENT)
1583 last_errno = errno;
1584 }
1585 else
1586 {
1587 int err = (fstat (fd, &st) != 0 ? errno
1588 : S_ISDIR (st.st_mode) ? EISDIR : 0);
1589 if (err)
1590 {
1591 last_errno = err;
1592 emacs_close (fd);
1593 fd = -1;
1594 }
1595 }
1596 }
1597
1598 if (fd >= 0)
1599 {
1600 if (newer && !NATNUMP (predicate))
1601 {
1602 struct timespec mtime = get_stat_mtime (&st);
1603
1604 if (timespec_cmp (mtime, save_mtime) <= 0)
1605 emacs_close (fd);
1606 else
1607 {
1608 if (0 <= save_fd)
1609 emacs_close (save_fd);
1610 save_fd = fd;
1611 save_mtime = mtime;
1612 save_string = string;
1613 }
1614 }
1615 else
1616 {
1617 /* We succeeded; return this descriptor and filename. */
1618 if (storeptr)
1619 *storeptr = string;
1620 SAFE_FREE ();
1621 UNGCPRO;
1622 return fd;
1623 }
1624 }
1625
1626 /* No more suffixes. Return the newest. */
1627 if (0 <= save_fd && ! CONSP (XCDR (tail)))
1628 {
1629 if (storeptr)
1630 *storeptr = save_string;
1631 SAFE_FREE ();
1632 UNGCPRO;
1633 return save_fd;
1634 }
1635 }
1636 }
1637 if (absolute)
1638 break;
1639 }
1640
1641 SAFE_FREE ();
1642 UNGCPRO;
1643 errno = last_errno;
1644 return -1;
1645 }
1646
1647 \f
1648 /* Merge the list we've accumulated of globals from the current input source
1649 into the load_history variable. The details depend on whether
1650 the source has an associated file name or not.
1651
1652 FILENAME is the file name that we are loading from.
1653
1654 ENTIRE is true if loading that entire file, false if evaluating
1655 part of it. */
1656
1657 static void
1658 build_load_history (Lisp_Object filename, bool entire)
1659 {
1660 Lisp_Object tail, prev, newelt;
1661 Lisp_Object tem, tem2;
1662 bool foundit = 0;
1663
1664 tail = Vload_history;
1665 prev = Qnil;
1666
1667 while (CONSP (tail))
1668 {
1669 tem = XCAR (tail);
1670
1671 /* Find the feature's previous assoc list... */
1672 if (!NILP (Fequal (filename, Fcar (tem))))
1673 {
1674 foundit = 1;
1675
1676 /* If we're loading the entire file, remove old data. */
1677 if (entire)
1678 {
1679 if (NILP (prev))
1680 Vload_history = XCDR (tail);
1681 else
1682 Fsetcdr (prev, XCDR (tail));
1683 }
1684
1685 /* Otherwise, cons on new symbols that are not already members. */
1686 else
1687 {
1688 tem2 = Vcurrent_load_list;
1689
1690 while (CONSP (tem2))
1691 {
1692 newelt = XCAR (tem2);
1693
1694 if (NILP (Fmember (newelt, tem)))
1695 Fsetcar (tail, Fcons (XCAR (tem),
1696 Fcons (newelt, XCDR (tem))));
1697
1698 tem2 = XCDR (tem2);
1699 QUIT;
1700 }
1701 }
1702 }
1703 else
1704 prev = tail;
1705 tail = XCDR (tail);
1706 QUIT;
1707 }
1708
1709 /* If we're loading an entire file, cons the new assoc onto the
1710 front of load-history, the most-recently-loaded position. Also
1711 do this if we didn't find an existing member for the file. */
1712 if (entire || !foundit)
1713 Vload_history = Fcons (Fnreverse (Vcurrent_load_list),
1714 Vload_history);
1715 }
1716
1717 static void
1718 readevalloop_1 (int old)
1719 {
1720 load_convert_to_unibyte = old;
1721 }
1722
1723 /* Signal an `end-of-file' error, if possible with file name
1724 information. */
1725
1726 static _Noreturn void
1727 end_of_file_error (void)
1728 {
1729 if (STRINGP (Vload_file_name))
1730 xsignal1 (Qend_of_file, Vload_file_name);
1731
1732 xsignal0 (Qend_of_file);
1733 }
1734
1735 static Lisp_Object
1736 readevalloop_eager_expand_eval (Lisp_Object val, Lisp_Object macroexpand)
1737 {
1738 /* If we macroexpand the toplevel form non-recursively and it ends
1739 up being a `progn' (or if it was a progn to start), treat each
1740 form in the progn as a top-level form. This way, if one form in
1741 the progn defines a macro, that macro is in effect when we expand
1742 the remaining forms. See similar code in bytecomp.el. */
1743 val = call2 (macroexpand, val, Qnil);
1744 if (EQ (CAR_SAFE (val), Qprogn))
1745 {
1746 struct gcpro gcpro1;
1747 Lisp_Object subforms = XCDR (val);
1748
1749 GCPRO1 (subforms);
1750 for (val = Qnil; CONSP (subforms); subforms = XCDR (subforms))
1751 val = readevalloop_eager_expand_eval (XCAR (subforms),
1752 macroexpand);
1753 UNGCPRO;
1754 }
1755 else
1756 val = eval_sub (call2 (macroexpand, val, Qt));
1757 return val;
1758 }
1759
1760 /* UNIBYTE specifies how to set load_convert_to_unibyte
1761 for this invocation.
1762 READFUN, if non-nil, is used instead of `read'.
1763
1764 START, END specify region to read in current buffer (from eval-region).
1765 If the input is not from a buffer, they must be nil. */
1766
1767 static void
1768 readevalloop (Lisp_Object readcharfun,
1769 FILE *stream,
1770 Lisp_Object sourcename,
1771 bool printflag,
1772 Lisp_Object unibyte, Lisp_Object readfun,
1773 Lisp_Object start, Lisp_Object end)
1774 {
1775 register int c;
1776 register Lisp_Object val;
1777 ptrdiff_t count = SPECPDL_INDEX ();
1778 struct gcpro gcpro1, gcpro2, gcpro3, gcpro4;
1779 struct buffer *b = 0;
1780 bool continue_reading_p;
1781 Lisp_Object lex_bound;
1782 /* True if reading an entire buffer. */
1783 bool whole_buffer = 0;
1784 /* True on the first time around. */
1785 bool first_sexp = 1;
1786 Lisp_Object macroexpand = intern ("internal-macroexpand-for-load");
1787
1788 if (NILP (Ffboundp (macroexpand))
1789 /* Don't macroexpand in .elc files, since it should have been done
1790 already. We actually don't know whether we're in a .elc file or not,
1791 so we use circumstantial evidence: .el files normally go through
1792 Vload_source_file_function -> load-with-code-conversion
1793 -> eval-buffer. */
1794 || EQ (readcharfun, Qget_file_char)
1795 || EQ (readcharfun, Qget_emacs_mule_file_char))
1796 macroexpand = Qnil;
1797
1798 if (MARKERP (readcharfun))
1799 {
1800 if (NILP (start))
1801 start = readcharfun;
1802 }
1803
1804 if (BUFFERP (readcharfun))
1805 b = XBUFFER (readcharfun);
1806 else if (MARKERP (readcharfun))
1807 b = XMARKER (readcharfun)->buffer;
1808
1809 /* We assume START is nil when input is not from a buffer. */
1810 if (! NILP (start) && !b)
1811 emacs_abort ();
1812
1813 specbind (Qstandard_input, readcharfun); /* GCPROs readcharfun. */
1814 specbind (Qcurrent_load_list, Qnil);
1815 record_unwind_protect_int (readevalloop_1, load_convert_to_unibyte);
1816 load_convert_to_unibyte = !NILP (unibyte);
1817
1818 /* If lexical binding is active (either because it was specified in
1819 the file's header, or via a buffer-local variable), create an empty
1820 lexical environment, otherwise, turn off lexical binding. */
1821 lex_bound = find_symbol_value (Qlexical_binding);
1822 specbind (Qinternal_interpreter_environment,
1823 (NILP (lex_bound) || EQ (lex_bound, Qunbound)
1824 ? Qnil : list1 (Qt)));
1825
1826 GCPRO4 (sourcename, readfun, start, end);
1827
1828 /* Try to ensure sourcename is a truename, except whilst preloading. */
1829 if (NILP (Vpurify_flag)
1830 && !NILP (sourcename) && !NILP (Ffile_name_absolute_p (sourcename))
1831 && !NILP (Ffboundp (Qfile_truename)))
1832 sourcename = call1 (Qfile_truename, sourcename) ;
1833
1834 LOADHIST_ATTACH (sourcename);
1835
1836 continue_reading_p = 1;
1837 while (continue_reading_p)
1838 {
1839 ptrdiff_t count1 = SPECPDL_INDEX ();
1840
1841 if (b != 0 && !BUFFER_LIVE_P (b))
1842 error ("Reading from killed buffer");
1843
1844 if (!NILP (start))
1845 {
1846 /* Switch to the buffer we are reading from. */
1847 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1848 set_buffer_internal (b);
1849
1850 /* Save point in it. */
1851 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1852 /* Save ZV in it. */
1853 record_unwind_protect (save_restriction_restore, save_restriction_save ());
1854 /* Those get unbound after we read one expression. */
1855
1856 /* Set point and ZV around stuff to be read. */
1857 Fgoto_char (start);
1858 if (!NILP (end))
1859 Fnarrow_to_region (make_number (BEGV), end);
1860
1861 /* Just for cleanliness, convert END to a marker
1862 if it is an integer. */
1863 if (INTEGERP (end))
1864 end = Fpoint_max_marker ();
1865 }
1866
1867 /* On the first cycle, we can easily test here
1868 whether we are reading the whole buffer. */
1869 if (b && first_sexp)
1870 whole_buffer = (PT == BEG && ZV == Z);
1871
1872 instream = stream;
1873 read_next:
1874 c = READCHAR;
1875 if (c == ';')
1876 {
1877 while ((c = READCHAR) != '\n' && c != -1);
1878 goto read_next;
1879 }
1880 if (c < 0)
1881 {
1882 unbind_to (count1, Qnil);
1883 break;
1884 }
1885
1886 /* Ignore whitespace here, so we can detect eof. */
1887 if (c == ' ' || c == '\t' || c == '\n' || c == '\f' || c == '\r'
1888 || c == 0xa0) /* NBSP */
1889 goto read_next;
1890
1891 if (!NILP (Vpurify_flag) && c == '(')
1892 {
1893 val = read_list (0, readcharfun);
1894 }
1895 else
1896 {
1897 UNREAD (c);
1898 read_objects = Qnil;
1899 if (!NILP (readfun))
1900 {
1901 val = call1 (readfun, readcharfun);
1902
1903 /* If READCHARFUN has set point to ZV, we should
1904 stop reading, even if the form read sets point
1905 to a different value when evaluated. */
1906 if (BUFFERP (readcharfun))
1907 {
1908 struct buffer *buf = XBUFFER (readcharfun);
1909 if (BUF_PT (buf) == BUF_ZV (buf))
1910 continue_reading_p = 0;
1911 }
1912 }
1913 else if (! NILP (Vload_read_function))
1914 val = call1 (Vload_read_function, readcharfun);
1915 else
1916 val = read_internal_start (readcharfun, Qnil, Qnil);
1917 }
1918
1919 if (!NILP (start) && continue_reading_p)
1920 start = Fpoint_marker ();
1921
1922 /* Restore saved point and BEGV. */
1923 unbind_to (count1, Qnil);
1924
1925 /* Now eval what we just read. */
1926 if (!NILP (macroexpand))
1927 val = readevalloop_eager_expand_eval (val, macroexpand);
1928 else
1929 val = eval_sub (val);
1930
1931 if (printflag)
1932 {
1933 Vvalues = Fcons (val, Vvalues);
1934 if (EQ (Vstandard_output, Qt))
1935 Fprin1 (val, Qnil);
1936 else
1937 Fprint (val, Qnil);
1938 }
1939
1940 first_sexp = 0;
1941 }
1942
1943 build_load_history (sourcename,
1944 stream || whole_buffer);
1945
1946 UNGCPRO;
1947
1948 unbind_to (count, Qnil);
1949 }
1950
1951 DEFUN ("eval-buffer", Feval_buffer, Seval_buffer, 0, 5, "",
1952 doc: /* Execute the current buffer as Lisp code.
1953 When called from a Lisp program (i.e., not interactively), this
1954 function accepts up to five optional arguments:
1955 BUFFER is the buffer to evaluate (nil means use current buffer).
1956 PRINTFLAG controls printing of output:
1957 A value of nil means discard it; anything else is stream for print.
1958 FILENAME specifies the file name to use for `load-history'.
1959 UNIBYTE, if non-nil, specifies `load-convert-to-unibyte' for this
1960 invocation.
1961 DO-ALLOW-PRINT, if non-nil, specifies that `print' and related
1962 functions should work normally even if PRINTFLAG is nil.
1963
1964 This function preserves the position of point. */)
1965 (Lisp_Object buffer, Lisp_Object printflag, Lisp_Object filename, Lisp_Object unibyte, Lisp_Object do_allow_print)
1966 {
1967 ptrdiff_t count = SPECPDL_INDEX ();
1968 Lisp_Object tem, buf;
1969
1970 if (NILP (buffer))
1971 buf = Fcurrent_buffer ();
1972 else
1973 buf = Fget_buffer (buffer);
1974 if (NILP (buf))
1975 error ("No such buffer");
1976
1977 if (NILP (printflag) && NILP (do_allow_print))
1978 tem = Qsymbolp;
1979 else
1980 tem = printflag;
1981
1982 if (NILP (filename))
1983 filename = BVAR (XBUFFER (buf), filename);
1984
1985 specbind (Qeval_buffer_list, Fcons (buf, Veval_buffer_list));
1986 specbind (Qstandard_output, tem);
1987 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1988 BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
1989 specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
1990 readevalloop (buf, 0, filename,
1991 !NILP (printflag), unibyte, Qnil, Qnil, Qnil);
1992 unbind_to (count, Qnil);
1993
1994 return Qnil;
1995 }
1996
1997 DEFUN ("eval-region", Feval_region, Seval_region, 2, 4, "r",
1998 doc: /* Execute the region as Lisp code.
1999 When called from programs, expects two arguments,
2000 giving starting and ending indices in the current buffer
2001 of the text to be executed.
2002 Programs can pass third argument PRINTFLAG which controls output:
2003 A value of nil means discard it; anything else is stream for printing it.
2004 Also the fourth argument READ-FUNCTION, if non-nil, is used
2005 instead of `read' to read each expression. It gets one argument
2006 which is the input stream for reading characters.
2007
2008 This function does not move point. */)
2009 (Lisp_Object start, Lisp_Object end, Lisp_Object printflag, Lisp_Object read_function)
2010 {
2011 /* FIXME: Do the eval-sexp-add-defvars dance! */
2012 ptrdiff_t count = SPECPDL_INDEX ();
2013 Lisp_Object tem, cbuf;
2014
2015 cbuf = Fcurrent_buffer ();
2016
2017 if (NILP (printflag))
2018 tem = Qsymbolp;
2019 else
2020 tem = printflag;
2021 specbind (Qstandard_output, tem);
2022 specbind (Qeval_buffer_list, Fcons (cbuf, Veval_buffer_list));
2023
2024 /* `readevalloop' calls functions which check the type of start and end. */
2025 readevalloop (cbuf, 0, BVAR (XBUFFER (cbuf), filename),
2026 !NILP (printflag), Qnil, read_function,
2027 start, end);
2028
2029 return unbind_to (count, Qnil);
2030 }
2031
2032 \f
2033 DEFUN ("read", Fread, Sread, 0, 1, 0,
2034 doc: /* Read one Lisp expression as text from STREAM, return as Lisp object.
2035 If STREAM is nil, use the value of `standard-input' (which see).
2036 STREAM or the value of `standard-input' may be:
2037 a buffer (read from point and advance it)
2038 a marker (read from where it points and advance it)
2039 a function (call it with no arguments for each character,
2040 call it with a char as argument to push a char back)
2041 a string (takes text from string, starting at the beginning)
2042 t (read text line using minibuffer and use it, or read from
2043 standard input in batch mode). */)
2044 (Lisp_Object stream)
2045 {
2046 if (NILP (stream))
2047 stream = Vstandard_input;
2048 if (EQ (stream, Qt))
2049 stream = Qread_char;
2050 if (EQ (stream, Qread_char))
2051 /* FIXME: ?! When is this used !? */
2052 return call1 (intern ("read-minibuffer"),
2053 build_string ("Lisp expression: "));
2054
2055 return read_internal_start (stream, Qnil, Qnil);
2056 }
2057
2058 DEFUN ("read-from-string", Fread_from_string, Sread_from_string, 1, 3, 0,
2059 doc: /* Read one Lisp expression which is represented as text by STRING.
2060 Returns a cons: (OBJECT-READ . FINAL-STRING-INDEX).
2061 FINAL-STRING-INDEX is an integer giving the position of the next
2062 remaining character in STRING. START and END optionally delimit
2063 a substring of STRING from which to read; they default to 0 and
2064 (length STRING) respectively. Negative values are counted from
2065 the end of STRING. */)
2066 (Lisp_Object string, Lisp_Object start, Lisp_Object end)
2067 {
2068 Lisp_Object ret;
2069 CHECK_STRING (string);
2070 /* `read_internal_start' sets `read_from_string_index'. */
2071 ret = read_internal_start (string, start, end);
2072 return Fcons (ret, make_number (read_from_string_index));
2073 }
2074
2075 /* Function to set up the global context we need in toplevel read
2076 calls. START and END only used when STREAM is a string. */
2077 static Lisp_Object
2078 read_internal_start (Lisp_Object stream, Lisp_Object start, Lisp_Object end)
2079 {
2080 Lisp_Object retval;
2081
2082 readchar_count = 0;
2083 new_backquote_flag = 0;
2084 read_objects = Qnil;
2085 if (EQ (Vread_with_symbol_positions, Qt)
2086 || EQ (Vread_with_symbol_positions, stream))
2087 Vread_symbol_positions_list = Qnil;
2088
2089 if (STRINGP (stream)
2090 || ((CONSP (stream) && STRINGP (XCAR (stream)))))
2091 {
2092 ptrdiff_t startval, endval;
2093 Lisp_Object string;
2094
2095 if (STRINGP (stream))
2096 string = stream;
2097 else
2098 string = XCAR (stream);
2099
2100 validate_subarray (string, start, end, SCHARS (string),
2101 &startval, &endval);
2102
2103 read_from_string_index = startval;
2104 read_from_string_index_byte = string_char_to_byte (string, startval);
2105 read_from_string_limit = endval;
2106 }
2107
2108 retval = read0 (stream);
2109 if (EQ (Vread_with_symbol_positions, Qt)
2110 || EQ (Vread_with_symbol_positions, stream))
2111 Vread_symbol_positions_list = Fnreverse (Vread_symbol_positions_list);
2112 return retval;
2113 }
2114 \f
2115
2116 /* Signal Qinvalid_read_syntax error.
2117 S is error string of length N (if > 0) */
2118
2119 static _Noreturn void
2120 invalid_syntax (const char *s)
2121 {
2122 xsignal1 (Qinvalid_read_syntax, build_string (s));
2123 }
2124
2125
2126 /* Use this for recursive reads, in contexts where internal tokens
2127 are not allowed. */
2128
2129 static Lisp_Object
2130 read0 (Lisp_Object readcharfun)
2131 {
2132 register Lisp_Object val;
2133 int c;
2134
2135 val = read1 (readcharfun, &c, 0);
2136 if (!c)
2137 return val;
2138
2139 xsignal1 (Qinvalid_read_syntax,
2140 Fmake_string (make_number (1), make_number (c)));
2141 }
2142 \f
2143 static ptrdiff_t read_buffer_size;
2144 static char *read_buffer;
2145
2146 /* Read a \-escape sequence, assuming we already read the `\'.
2147 If the escape sequence forces unibyte, return eight-bit char. */
2148
2149 static int
2150 read_escape (Lisp_Object readcharfun, bool stringp)
2151 {
2152 int c = READCHAR;
2153 /* \u allows up to four hex digits, \U up to eight. Default to the
2154 behavior for \u, and change this value in the case that \U is seen. */
2155 int unicode_hex_count = 4;
2156
2157 switch (c)
2158 {
2159 case -1:
2160 end_of_file_error ();
2161
2162 case 'a':
2163 return '\007';
2164 case 'b':
2165 return '\b';
2166 case 'd':
2167 return 0177;
2168 case 'e':
2169 return 033;
2170 case 'f':
2171 return '\f';
2172 case 'n':
2173 return '\n';
2174 case 'r':
2175 return '\r';
2176 case 't':
2177 return '\t';
2178 case 'v':
2179 return '\v';
2180 case '\n':
2181 return -1;
2182 case ' ':
2183 if (stringp)
2184 return -1;
2185 return ' ';
2186
2187 case 'M':
2188 c = READCHAR;
2189 if (c != '-')
2190 error ("Invalid escape character syntax");
2191 c = READCHAR;
2192 if (c == '\\')
2193 c = read_escape (readcharfun, 0);
2194 return c | meta_modifier;
2195
2196 case 'S':
2197 c = READCHAR;
2198 if (c != '-')
2199 error ("Invalid escape character syntax");
2200 c = READCHAR;
2201 if (c == '\\')
2202 c = read_escape (readcharfun, 0);
2203 return c | shift_modifier;
2204
2205 case 'H':
2206 c = READCHAR;
2207 if (c != '-')
2208 error ("Invalid escape character syntax");
2209 c = READCHAR;
2210 if (c == '\\')
2211 c = read_escape (readcharfun, 0);
2212 return c | hyper_modifier;
2213
2214 case 'A':
2215 c = READCHAR;
2216 if (c != '-')
2217 error ("Invalid escape character syntax");
2218 c = READCHAR;
2219 if (c == '\\')
2220 c = read_escape (readcharfun, 0);
2221 return c | alt_modifier;
2222
2223 case 's':
2224 c = READCHAR;
2225 if (stringp || c != '-')
2226 {
2227 UNREAD (c);
2228 return ' ';
2229 }
2230 c = READCHAR;
2231 if (c == '\\')
2232 c = read_escape (readcharfun, 0);
2233 return c | super_modifier;
2234
2235 case 'C':
2236 c = READCHAR;
2237 if (c != '-')
2238 error ("Invalid escape character syntax");
2239 case '^':
2240 c = READCHAR;
2241 if (c == '\\')
2242 c = read_escape (readcharfun, 0);
2243 if ((c & ~CHAR_MODIFIER_MASK) == '?')
2244 return 0177 | (c & CHAR_MODIFIER_MASK);
2245 else if (! SINGLE_BYTE_CHAR_P ((c & ~CHAR_MODIFIER_MASK)))
2246 return c | ctrl_modifier;
2247 /* ASCII control chars are made from letters (both cases),
2248 as well as the non-letters within 0100...0137. */
2249 else if ((c & 0137) >= 0101 && (c & 0137) <= 0132)
2250 return (c & (037 | ~0177));
2251 else if ((c & 0177) >= 0100 && (c & 0177) <= 0137)
2252 return (c & (037 | ~0177));
2253 else
2254 return c | ctrl_modifier;
2255
2256 case '0':
2257 case '1':
2258 case '2':
2259 case '3':
2260 case '4':
2261 case '5':
2262 case '6':
2263 case '7':
2264 /* An octal escape, as in ANSI C. */
2265 {
2266 register int i = c - '0';
2267 register int count = 0;
2268 while (++count < 3)
2269 {
2270 if ((c = READCHAR) >= '0' && c <= '7')
2271 {
2272 i *= 8;
2273 i += c - '0';
2274 }
2275 else
2276 {
2277 UNREAD (c);
2278 break;
2279 }
2280 }
2281
2282 if (i >= 0x80 && i < 0x100)
2283 i = BYTE8_TO_CHAR (i);
2284 return i;
2285 }
2286
2287 case 'x':
2288 /* A hex escape, as in ANSI C. */
2289 {
2290 unsigned int i = 0;
2291 int count = 0;
2292 while (1)
2293 {
2294 c = READCHAR;
2295 if (c >= '0' && c <= '9')
2296 {
2297 i *= 16;
2298 i += c - '0';
2299 }
2300 else if ((c >= 'a' && c <= 'f')
2301 || (c >= 'A' && c <= 'F'))
2302 {
2303 i *= 16;
2304 if (c >= 'a' && c <= 'f')
2305 i += c - 'a' + 10;
2306 else
2307 i += c - 'A' + 10;
2308 }
2309 else
2310 {
2311 UNREAD (c);
2312 break;
2313 }
2314 /* Allow hex escapes as large as ?\xfffffff, because some
2315 packages use them to denote characters with modifiers. */
2316 if ((CHAR_META | (CHAR_META - 1)) < i)
2317 error ("Hex character out of range: \\x%x...", i);
2318 count += count < 3;
2319 }
2320
2321 if (count < 3 && i >= 0x80)
2322 return BYTE8_TO_CHAR (i);
2323 return i;
2324 }
2325
2326 case 'U':
2327 /* Post-Unicode-2.0: Up to eight hex chars. */
2328 unicode_hex_count = 8;
2329 case 'u':
2330
2331 /* A Unicode escape. We only permit them in strings and characters,
2332 not arbitrarily in the source code, as in some other languages. */
2333 {
2334 unsigned int i = 0;
2335 int count = 0;
2336
2337 while (++count <= unicode_hex_count)
2338 {
2339 c = READCHAR;
2340 /* `isdigit' and `isalpha' may be locale-specific, which we don't
2341 want. */
2342 if (c >= '0' && c <= '9') i = (i << 4) + (c - '0');
2343 else if (c >= 'a' && c <= 'f') i = (i << 4) + (c - 'a') + 10;
2344 else if (c >= 'A' && c <= 'F') i = (i << 4) + (c - 'A') + 10;
2345 else
2346 error ("Non-hex digit used for Unicode escape");
2347 }
2348 if (i > 0x10FFFF)
2349 error ("Non-Unicode character: 0x%x", i);
2350 return i;
2351 }
2352
2353 default:
2354 return c;
2355 }
2356 }
2357
2358 /* Return the digit that CHARACTER stands for in the given BASE.
2359 Return -1 if CHARACTER is out of range for BASE,
2360 and -2 if CHARACTER is not valid for any supported BASE. */
2361 static int
2362 digit_to_number (int character, int base)
2363 {
2364 int digit;
2365
2366 if ('0' <= character && character <= '9')
2367 digit = character - '0';
2368 else if ('a' <= character && character <= 'z')
2369 digit = character - 'a' + 10;
2370 else if ('A' <= character && character <= 'Z')
2371 digit = character - 'A' + 10;
2372 else
2373 return -2;
2374
2375 return digit < base ? digit : -1;
2376 }
2377
2378 /* Read an integer in radix RADIX using READCHARFUN to read
2379 characters. RADIX must be in the interval [2..36]; if it isn't, a
2380 read error is signaled . Value is the integer read. Signals an
2381 error if encountering invalid read syntax or if RADIX is out of
2382 range. */
2383
2384 static Lisp_Object
2385 read_integer (Lisp_Object readcharfun, EMACS_INT radix)
2386 {
2387 /* Room for sign, leading 0, other digits, trailing null byte.
2388 Also, room for invalid syntax diagnostic. */
2389 char buf[max (1 + 1 + sizeof (uintmax_t) * CHAR_BIT + 1,
2390 sizeof "integer, radix " + INT_STRLEN_BOUND (EMACS_INT))];
2391
2392 int valid = -1; /* 1 if valid, 0 if not, -1 if incomplete. */
2393
2394 if (radix < 2 || radix > 36)
2395 valid = 0;
2396 else
2397 {
2398 char *p = buf;
2399 int c, digit;
2400
2401 c = READCHAR;
2402 if (c == '-' || c == '+')
2403 {
2404 *p++ = c;
2405 c = READCHAR;
2406 }
2407
2408 if (c == '0')
2409 {
2410 *p++ = c;
2411 valid = 1;
2412
2413 /* Ignore redundant leading zeros, so the buffer doesn't
2414 fill up with them. */
2415 do
2416 c = READCHAR;
2417 while (c == '0');
2418 }
2419
2420 while ((digit = digit_to_number (c, radix)) >= -1)
2421 {
2422 if (digit == -1)
2423 valid = 0;
2424 if (valid < 0)
2425 valid = 1;
2426
2427 if (p < buf + sizeof buf - 1)
2428 *p++ = c;
2429 else
2430 valid = 0;
2431
2432 c = READCHAR;
2433 }
2434
2435 UNREAD (c);
2436 *p = '\0';
2437 }
2438
2439 if (! valid)
2440 {
2441 sprintf (buf, "integer, radix %"pI"d", radix);
2442 invalid_syntax (buf);
2443 }
2444
2445 return string_to_number (buf, radix, 0);
2446 }
2447
2448
2449 /* If the next token is ')' or ']' or '.', we store that character
2450 in *PCH and the return value is not interesting. Else, we store
2451 zero in *PCH and we read and return one lisp object.
2452
2453 FIRST_IN_LIST is true if this is the first element of a list. */
2454
2455 static Lisp_Object
2456 read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
2457 {
2458 int c;
2459 bool uninterned_symbol = 0;
2460 bool multibyte;
2461
2462 *pch = 0;
2463
2464 retry:
2465
2466 c = READCHAR_REPORT_MULTIBYTE (&multibyte);
2467 if (c < 0)
2468 end_of_file_error ();
2469
2470 switch (c)
2471 {
2472 case '(':
2473 return read_list (0, readcharfun);
2474
2475 case '[':
2476 return read_vector (readcharfun, 0);
2477
2478 case ')':
2479 case ']':
2480 {
2481 *pch = c;
2482 return Qnil;
2483 }
2484
2485 case '#':
2486 c = READCHAR;
2487 if (c == 's')
2488 {
2489 c = READCHAR;
2490 if (c == '(')
2491 {
2492 /* Accept extended format for hashtables (extensible to
2493 other types), e.g.
2494 #s(hash-table size 2 test equal data (k1 v1 k2 v2)) */
2495 Lisp_Object tmp = read_list (0, readcharfun);
2496 Lisp_Object head = CAR_SAFE (tmp);
2497 Lisp_Object data = Qnil;
2498 Lisp_Object val = Qnil;
2499 /* The size is 2 * number of allowed keywords to
2500 make-hash-table. */
2501 Lisp_Object params[10];
2502 Lisp_Object ht;
2503 Lisp_Object key = Qnil;
2504 int param_count = 0;
2505
2506 if (!EQ (head, Qhash_table))
2507 error ("Invalid extended read marker at head of #s list "
2508 "(only hash-table allowed)");
2509
2510 tmp = CDR_SAFE (tmp);
2511
2512 /* This is repetitive but fast and simple. */
2513 params[param_count] = QCsize;
2514 params[param_count + 1] = Fplist_get (tmp, Qsize);
2515 if (!NILP (params[param_count + 1]))
2516 param_count += 2;
2517
2518 params[param_count] = QCtest;
2519 params[param_count + 1] = Fplist_get (tmp, Qtest);
2520 if (!NILP (params[param_count + 1]))
2521 param_count += 2;
2522
2523 params[param_count] = QCweakness;
2524 params[param_count + 1] = Fplist_get (tmp, Qweakness);
2525 if (!NILP (params[param_count + 1]))
2526 param_count += 2;
2527
2528 params[param_count] = QCrehash_size;
2529 params[param_count + 1] = Fplist_get (tmp, Qrehash_size);
2530 if (!NILP (params[param_count + 1]))
2531 param_count += 2;
2532
2533 params[param_count] = QCrehash_threshold;
2534 params[param_count + 1] = Fplist_get (tmp, Qrehash_threshold);
2535 if (!NILP (params[param_count + 1]))
2536 param_count += 2;
2537
2538 /* This is the hashtable data. */
2539 data = Fplist_get (tmp, Qdata);
2540
2541 /* Now use params to make a new hashtable and fill it. */
2542 ht = Fmake_hash_table (param_count, params);
2543
2544 while (CONSP (data))
2545 {
2546 key = XCAR (data);
2547 data = XCDR (data);
2548 if (!CONSP (data))
2549 error ("Odd number of elements in hashtable data");
2550 val = XCAR (data);
2551 data = XCDR (data);
2552 Fputhash (key, val, ht);
2553 }
2554
2555 return ht;
2556 }
2557 UNREAD (c);
2558 invalid_syntax ("#");
2559 }
2560 if (c == '^')
2561 {
2562 c = READCHAR;
2563 if (c == '[')
2564 {
2565 Lisp_Object tmp;
2566 tmp = read_vector (readcharfun, 0);
2567 if (ASIZE (tmp) < CHAR_TABLE_STANDARD_SLOTS)
2568 error ("Invalid size char-table");
2569 XSETPVECTYPE (XVECTOR (tmp), PVEC_CHAR_TABLE);
2570 return tmp;
2571 }
2572 else if (c == '^')
2573 {
2574 c = READCHAR;
2575 if (c == '[')
2576 {
2577 /* Sub char-table can't be read as a regular
2578 vector because of a two C integer fields. */
2579 Lisp_Object tbl, tmp = read_list (1, readcharfun);
2580 ptrdiff_t size = XINT (Flength (tmp));
2581 int i, depth, min_char;
2582 struct Lisp_Cons *cell;
2583
2584 if (size == 0)
2585 error ("Zero-sized sub char-table");
2586
2587 if (! RANGED_INTEGERP (1, XCAR (tmp), 3))
2588 error ("Invalid depth in sub char-table");
2589 depth = XINT (XCAR (tmp));
2590 if (chartab_size[depth] != size - 2)
2591 error ("Invalid size in sub char-table");
2592 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2593 free_cons (cell);
2594
2595 if (! RANGED_INTEGERP (0, XCAR (tmp), MAX_CHAR))
2596 error ("Invalid minimum character in sub-char-table");
2597 min_char = XINT (XCAR (tmp));
2598 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2599 free_cons (cell);
2600
2601 tbl = make_uninit_sub_char_table (depth, min_char);
2602 for (i = 0; i < size; i++)
2603 {
2604 XSUB_CHAR_TABLE (tbl)->contents[i] = XCAR (tmp);
2605 cell = XCONS (tmp), tmp = XCDR (tmp);
2606 free_cons (cell);
2607 }
2608 return tbl;
2609 }
2610 invalid_syntax ("#^^");
2611 }
2612 invalid_syntax ("#^");
2613 }
2614 if (c == '&')
2615 {
2616 Lisp_Object length;
2617 length = read1 (readcharfun, pch, first_in_list);
2618 c = READCHAR;
2619 if (c == '"')
2620 {
2621 Lisp_Object tmp, val;
2622 EMACS_INT size_in_chars = bool_vector_bytes (XFASTINT (length));
2623 unsigned char *data;
2624
2625 UNREAD (c);
2626 tmp = read1 (readcharfun, pch, first_in_list);
2627 if (STRING_MULTIBYTE (tmp)
2628 || (size_in_chars != SCHARS (tmp)
2629 /* We used to print 1 char too many
2630 when the number of bits was a multiple of 8.
2631 Accept such input in case it came from an old
2632 version. */
2633 && ! (XFASTINT (length)
2634 == (SCHARS (tmp) - 1) * BOOL_VECTOR_BITS_PER_CHAR)))
2635 invalid_syntax ("#&...");
2636
2637 val = make_uninit_bool_vector (XFASTINT (length));
2638 data = bool_vector_uchar_data (val);
2639 memcpy (data, SDATA (tmp), size_in_chars);
2640 /* Clear the extraneous bits in the last byte. */
2641 if (XINT (length) != size_in_chars * BOOL_VECTOR_BITS_PER_CHAR)
2642 data[size_in_chars - 1]
2643 &= (1 << (XINT (length) % BOOL_VECTOR_BITS_PER_CHAR)) - 1;
2644 return val;
2645 }
2646 invalid_syntax ("#&...");
2647 }
2648 if (c == '[')
2649 {
2650 /* Accept compiled functions at read-time so that we don't have to
2651 build them using function calls. */
2652 Lisp_Object tmp;
2653 struct Lisp_Vector *vec;
2654 tmp = read_vector (readcharfun, 1);
2655 vec = XVECTOR (tmp);
2656 if (vec->header.size == 0)
2657 invalid_syntax ("Empty byte-code object");
2658 make_byte_code (vec);
2659 return tmp;
2660 }
2661 if (c == '(')
2662 {
2663 Lisp_Object tmp;
2664 struct gcpro gcpro1;
2665 int ch;
2666
2667 /* Read the string itself. */
2668 tmp = read1 (readcharfun, &ch, 0);
2669 if (ch != 0 || !STRINGP (tmp))
2670 invalid_syntax ("#");
2671 GCPRO1 (tmp);
2672 /* Read the intervals and their properties. */
2673 while (1)
2674 {
2675 Lisp_Object beg, end, plist;
2676
2677 beg = read1 (readcharfun, &ch, 0);
2678 end = plist = Qnil;
2679 if (ch == ')')
2680 break;
2681 if (ch == 0)
2682 end = read1 (readcharfun, &ch, 0);
2683 if (ch == 0)
2684 plist = read1 (readcharfun, &ch, 0);
2685 if (ch)
2686 invalid_syntax ("Invalid string property list");
2687 Fset_text_properties (beg, end, plist, tmp);
2688 }
2689 UNGCPRO;
2690 return tmp;
2691 }
2692
2693 /* #@NUMBER is used to skip NUMBER following bytes.
2694 That's used in .elc files to skip over doc strings
2695 and function definitions. */
2696 if (c == '@')
2697 {
2698 enum { extra = 100 };
2699 ptrdiff_t i, nskip = 0, digits = 0;
2700
2701 /* Read a decimal integer. */
2702 while ((c = READCHAR) >= 0
2703 && c >= '0' && c <= '9')
2704 {
2705 if ((STRING_BYTES_BOUND - extra) / 10 <= nskip)
2706 string_overflow ();
2707 digits++;
2708 nskip *= 10;
2709 nskip += c - '0';
2710 if (digits == 2 && nskip == 0)
2711 { /* We've just seen #@00, which means "skip to end". */
2712 skip_dyn_eof (readcharfun);
2713 return Qnil;
2714 }
2715 }
2716 if (nskip > 0)
2717 /* We can't use UNREAD here, because in the code below we side-step
2718 READCHAR. Instead, assume the first char after #@NNN occupies
2719 a single byte, which is the case normally since it's just
2720 a space. */
2721 nskip--;
2722 else
2723 UNREAD (c);
2724
2725 if (load_force_doc_strings
2726 && (FROM_FILE_P (readcharfun)))
2727 {
2728 /* If we are supposed to force doc strings into core right now,
2729 record the last string that we skipped,
2730 and record where in the file it comes from. */
2731
2732 /* But first exchange saved_doc_string
2733 with prev_saved_doc_string, so we save two strings. */
2734 {
2735 char *temp = saved_doc_string;
2736 ptrdiff_t temp_size = saved_doc_string_size;
2737 file_offset temp_pos = saved_doc_string_position;
2738 ptrdiff_t temp_len = saved_doc_string_length;
2739
2740 saved_doc_string = prev_saved_doc_string;
2741 saved_doc_string_size = prev_saved_doc_string_size;
2742 saved_doc_string_position = prev_saved_doc_string_position;
2743 saved_doc_string_length = prev_saved_doc_string_length;
2744
2745 prev_saved_doc_string = temp;
2746 prev_saved_doc_string_size = temp_size;
2747 prev_saved_doc_string_position = temp_pos;
2748 prev_saved_doc_string_length = temp_len;
2749 }
2750
2751 if (saved_doc_string_size == 0)
2752 {
2753 saved_doc_string = xmalloc (nskip + extra);
2754 saved_doc_string_size = nskip + extra;
2755 }
2756 if (nskip > saved_doc_string_size)
2757 {
2758 saved_doc_string = xrealloc (saved_doc_string, nskip + extra);
2759 saved_doc_string_size = nskip + extra;
2760 }
2761
2762 saved_doc_string_position = file_tell (instream);
2763
2764 /* Copy that many characters into saved_doc_string. */
2765 block_input ();
2766 for (i = 0; i < nskip && c >= 0; i++)
2767 saved_doc_string[i] = c = getc (instream);
2768 unblock_input ();
2769
2770 saved_doc_string_length = i;
2771 }
2772 else
2773 /* Skip that many bytes. */
2774 skip_dyn_bytes (readcharfun, nskip);
2775
2776 goto retry;
2777 }
2778 if (c == '!')
2779 {
2780 /* #! appears at the beginning of an executable file.
2781 Skip the first line. */
2782 while (c != '\n' && c >= 0)
2783 c = READCHAR;
2784 goto retry;
2785 }
2786 if (c == '$')
2787 return Vload_file_name;
2788 if (c == '\'')
2789 return list2 (Qfunction, read0 (readcharfun));
2790 /* #:foo is the uninterned symbol named foo. */
2791 if (c == ':')
2792 {
2793 uninterned_symbol = 1;
2794 c = READCHAR;
2795 if (!(c > 040
2796 && c != 0xa0 /* NBSP */
2797 && (c >= 0200
2798 || strchr ("\"';()[]#`,", c) == NULL)))
2799 {
2800 /* No symbol character follows, this is the empty
2801 symbol. */
2802 UNREAD (c);
2803 return Fmake_symbol (empty_unibyte_string);
2804 }
2805 goto read_symbol;
2806 }
2807 /* ## is the empty symbol. */
2808 if (c == '#')
2809 return Fintern (empty_unibyte_string, Qnil);
2810 /* Reader forms that can reuse previously read objects. */
2811 if (c >= '0' && c <= '9')
2812 {
2813 EMACS_INT n = 0;
2814 Lisp_Object tem;
2815
2816 /* Read a non-negative integer. */
2817 while (c >= '0' && c <= '9')
2818 {
2819 if (MOST_POSITIVE_FIXNUM / 10 < n
2820 || MOST_POSITIVE_FIXNUM < n * 10 + c - '0')
2821 n = MOST_POSITIVE_FIXNUM + 1;
2822 else
2823 n = n * 10 + c - '0';
2824 c = READCHAR;
2825 }
2826
2827 if (n <= MOST_POSITIVE_FIXNUM)
2828 {
2829 if (c == 'r' || c == 'R')
2830 return read_integer (readcharfun, n);
2831
2832 if (! NILP (Vread_circle))
2833 {
2834 /* #n=object returns object, but associates it with
2835 n for #n#. */
2836 if (c == '=')
2837 {
2838 /* Make a placeholder for #n# to use temporarily. */
2839 AUTO_CONS (placeholder, Qnil, Qnil);
2840 Lisp_Object cell = Fcons (make_number (n), placeholder);
2841 read_objects = Fcons (cell, read_objects);
2842
2843 /* Read the object itself. */
2844 tem = read0 (readcharfun);
2845
2846 /* Now put it everywhere the placeholder was... */
2847 substitute_object_in_subtree (tem, placeholder);
2848
2849 /* ...and #n# will use the real value from now on. */
2850 Fsetcdr (cell, tem);
2851
2852 return tem;
2853 }
2854
2855 /* #n# returns a previously read object. */
2856 if (c == '#')
2857 {
2858 tem = Fassq (make_number (n), read_objects);
2859 if (CONSP (tem))
2860 return XCDR (tem);
2861 }
2862 }
2863 }
2864 /* Fall through to error message. */
2865 }
2866 else if (c == 'x' || c == 'X')
2867 return read_integer (readcharfun, 16);
2868 else if (c == 'o' || c == 'O')
2869 return read_integer (readcharfun, 8);
2870 else if (c == 'b' || c == 'B')
2871 return read_integer (readcharfun, 2);
2872
2873 UNREAD (c);
2874 invalid_syntax ("#");
2875
2876 case ';':
2877 while ((c = READCHAR) >= 0 && c != '\n');
2878 goto retry;
2879
2880 case '\'':
2881 return list2 (Qquote, read0 (readcharfun));
2882
2883 case '`':
2884 {
2885 int next_char = READCHAR;
2886 UNREAD (next_char);
2887 /* Transition from old-style to new-style:
2888 If we see "(`" it used to mean old-style, which usually works
2889 fine because ` should almost never appear in such a position
2890 for new-style. But occasionally we need "(`" to mean new
2891 style, so we try to distinguish the two by the fact that we
2892 can either write "( `foo" or "(` foo", where the first
2893 intends to use new-style whereas the second intends to use
2894 old-style. For Emacs-25, we should completely remove this
2895 first_in_list exception (old-style can still be obtained via
2896 "(\`" anyway). */
2897 if (!new_backquote_flag && first_in_list && next_char == ' ')
2898 {
2899 Vold_style_backquotes = Qt;
2900 goto default_label;
2901 }
2902 else
2903 {
2904 Lisp_Object value;
2905 bool saved_new_backquote_flag = new_backquote_flag;
2906
2907 new_backquote_flag = 1;
2908 value = read0 (readcharfun);
2909 new_backquote_flag = saved_new_backquote_flag;
2910
2911 return list2 (Qbackquote, value);
2912 }
2913 }
2914 case ',':
2915 {
2916 int next_char = READCHAR;
2917 UNREAD (next_char);
2918 /* Transition from old-style to new-style:
2919 It used to be impossible to have a new-style , other than within
2920 a new-style `. This is sufficient when ` and , are used in the
2921 normal way, but ` and , can also appear in args to macros that
2922 will not interpret them in the usual way, in which case , may be
2923 used without any ` anywhere near.
2924 So we now use the same heuristic as for backquote: old-style
2925 unquotes are only recognized when first on a list, and when
2926 followed by a space.
2927 Because it's more difficult to peek 2 chars ahead, a new-style
2928 ,@ can still not be used outside of a `, unless it's in the middle
2929 of a list. */
2930 if (new_backquote_flag
2931 || !first_in_list
2932 || (next_char != ' ' && next_char != '@'))
2933 {
2934 Lisp_Object comma_type = Qnil;
2935 Lisp_Object value;
2936 int ch = READCHAR;
2937
2938 if (ch == '@')
2939 comma_type = Qcomma_at;
2940 else if (ch == '.')
2941 comma_type = Qcomma_dot;
2942 else
2943 {
2944 if (ch >= 0) UNREAD (ch);
2945 comma_type = Qcomma;
2946 }
2947
2948 value = read0 (readcharfun);
2949 return list2 (comma_type, value);
2950 }
2951 else
2952 {
2953 Vold_style_backquotes = Qt;
2954 goto default_label;
2955 }
2956 }
2957 case '?':
2958 {
2959 int modifiers;
2960 int next_char;
2961 bool ok;
2962
2963 c = READCHAR;
2964 if (c < 0)
2965 end_of_file_error ();
2966
2967 /* Accept `single space' syntax like (list ? x) where the
2968 whitespace character is SPC or TAB.
2969 Other literal whitespace like NL, CR, and FF are not accepted,
2970 as there are well-established escape sequences for these. */
2971 if (c == ' ' || c == '\t')
2972 return make_number (c);
2973
2974 if (c == '\\')
2975 c = read_escape (readcharfun, 0);
2976 modifiers = c & CHAR_MODIFIER_MASK;
2977 c &= ~CHAR_MODIFIER_MASK;
2978 if (CHAR_BYTE8_P (c))
2979 c = CHAR_TO_BYTE8 (c);
2980 c |= modifiers;
2981
2982 next_char = READCHAR;
2983 ok = (next_char <= 040
2984 || (next_char < 0200
2985 && strchr ("\"';()[]#?`,.", next_char) != NULL));
2986 UNREAD (next_char);
2987 if (ok)
2988 return make_number (c);
2989
2990 invalid_syntax ("?");
2991 }
2992
2993 case '"':
2994 {
2995 char *p = read_buffer;
2996 char *end = read_buffer + read_buffer_size;
2997 int ch;
2998 /* True if we saw an escape sequence specifying
2999 a multibyte character. */
3000 bool force_multibyte = 0;
3001 /* True if we saw an escape sequence specifying
3002 a single-byte character. */
3003 bool force_singlebyte = 0;
3004 bool cancel = 0;
3005 ptrdiff_t nchars = 0;
3006
3007 while ((ch = READCHAR) >= 0
3008 && ch != '\"')
3009 {
3010 if (end - p < MAX_MULTIBYTE_LENGTH)
3011 {
3012 ptrdiff_t offset = p - read_buffer;
3013 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3014 memory_full (SIZE_MAX);
3015 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3016 read_buffer_size *= 2;
3017 p = read_buffer + offset;
3018 end = read_buffer + read_buffer_size;
3019 }
3020
3021 if (ch == '\\')
3022 {
3023 int modifiers;
3024
3025 ch = read_escape (readcharfun, 1);
3026
3027 /* CH is -1 if \ newline has just been seen. */
3028 if (ch == -1)
3029 {
3030 if (p == read_buffer)
3031 cancel = 1;
3032 continue;
3033 }
3034
3035 modifiers = ch & CHAR_MODIFIER_MASK;
3036 ch = ch & ~CHAR_MODIFIER_MASK;
3037
3038 if (CHAR_BYTE8_P (ch))
3039 force_singlebyte = 1;
3040 else if (! ASCII_CHAR_P (ch))
3041 force_multibyte = 1;
3042 else /* I.e. ASCII_CHAR_P (ch). */
3043 {
3044 /* Allow `\C- ' and `\C-?'. */
3045 if (modifiers == CHAR_CTL)
3046 {
3047 if (ch == ' ')
3048 ch = 0, modifiers = 0;
3049 else if (ch == '?')
3050 ch = 127, modifiers = 0;
3051 }
3052 if (modifiers & CHAR_SHIFT)
3053 {
3054 /* Shift modifier is valid only with [A-Za-z]. */
3055 if (ch >= 'A' && ch <= 'Z')
3056 modifiers &= ~CHAR_SHIFT;
3057 else if (ch >= 'a' && ch <= 'z')
3058 ch -= ('a' - 'A'), modifiers &= ~CHAR_SHIFT;
3059 }
3060
3061 if (modifiers & CHAR_META)
3062 {
3063 /* Move the meta bit to the right place for a
3064 string. */
3065 modifiers &= ~CHAR_META;
3066 ch = BYTE8_TO_CHAR (ch | 0x80);
3067 force_singlebyte = 1;
3068 }
3069 }
3070
3071 /* Any modifiers remaining are invalid. */
3072 if (modifiers)
3073 error ("Invalid modifier in string");
3074 p += CHAR_STRING (ch, (unsigned char *) p);
3075 }
3076 else
3077 {
3078 p += CHAR_STRING (ch, (unsigned char *) p);
3079 if (CHAR_BYTE8_P (ch))
3080 force_singlebyte = 1;
3081 else if (! ASCII_CHAR_P (ch))
3082 force_multibyte = 1;
3083 }
3084 nchars++;
3085 }
3086
3087 if (ch < 0)
3088 end_of_file_error ();
3089
3090 /* If purifying, and string starts with \ newline,
3091 return zero instead. This is for doc strings
3092 that we are really going to find in etc/DOC.nn.nn. */
3093 if (!NILP (Vpurify_flag) && NILP (Vdoc_file_name) && cancel)
3094 return make_number (0);
3095
3096 if (! force_multibyte && force_singlebyte)
3097 {
3098 /* READ_BUFFER contains raw 8-bit bytes and no multibyte
3099 forms. Convert it to unibyte. */
3100 nchars = str_as_unibyte ((unsigned char *) read_buffer,
3101 p - read_buffer);
3102 p = read_buffer + nchars;
3103 }
3104
3105 return make_specified_string (read_buffer, nchars, p - read_buffer,
3106 (force_multibyte
3107 || (p - read_buffer != nchars)));
3108 }
3109
3110 case '.':
3111 {
3112 int next_char = READCHAR;
3113 UNREAD (next_char);
3114
3115 if (next_char <= 040
3116 || (next_char < 0200
3117 && strchr ("\"';([#?`,", next_char) != NULL))
3118 {
3119 *pch = c;
3120 return Qnil;
3121 }
3122
3123 /* Otherwise, we fall through! Note that the atom-reading loop
3124 below will now loop at least once, assuring that we will not
3125 try to UNREAD two characters in a row. */
3126 }
3127 default:
3128 default_label:
3129 if (c <= 040) goto retry;
3130 if (c == 0xa0) /* NBSP */
3131 goto retry;
3132
3133 read_symbol:
3134 {
3135 char *p = read_buffer;
3136 bool quoted = 0;
3137 EMACS_INT start_position = readchar_count - 1;
3138
3139 {
3140 char *end = read_buffer + read_buffer_size;
3141
3142 do
3143 {
3144 if (end - p < MAX_MULTIBYTE_LENGTH)
3145 {
3146 ptrdiff_t offset = p - read_buffer;
3147 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3148 memory_full (SIZE_MAX);
3149 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3150 read_buffer_size *= 2;
3151 p = read_buffer + offset;
3152 end = read_buffer + read_buffer_size;
3153 }
3154
3155 if (c == '\\')
3156 {
3157 c = READCHAR;
3158 if (c == -1)
3159 end_of_file_error ();
3160 quoted = 1;
3161 }
3162
3163 if (multibyte)
3164 p += CHAR_STRING (c, (unsigned char *) p);
3165 else
3166 *p++ = c;
3167 c = READCHAR;
3168 }
3169 while (c > 040
3170 && c != 0xa0 /* NBSP */
3171 && (c >= 0200
3172 || strchr ("\"';()[]#`,", c) == NULL));
3173
3174 if (p == end)
3175 {
3176 ptrdiff_t offset = p - read_buffer;
3177 if (min (PTRDIFF_MAX, SIZE_MAX) / 2 < read_buffer_size)
3178 memory_full (SIZE_MAX);
3179 read_buffer = xrealloc (read_buffer, read_buffer_size * 2);
3180 read_buffer_size *= 2;
3181 p = read_buffer + offset;
3182 end = read_buffer + read_buffer_size;
3183 }
3184 *p = 0;
3185 UNREAD (c);
3186 }
3187
3188 if (!quoted && !uninterned_symbol)
3189 {
3190 Lisp_Object result = string_to_number (read_buffer, 10, 0);
3191 if (! NILP (result))
3192 return result;
3193 }
3194 {
3195 Lisp_Object name, result;
3196 ptrdiff_t nbytes = p - read_buffer;
3197 ptrdiff_t nchars
3198 = (multibyte
3199 ? multibyte_chars_in_text ((unsigned char *) read_buffer,
3200 nbytes)
3201 : nbytes);
3202
3203 name = ((uninterned_symbol && ! NILP (Vpurify_flag)
3204 ? make_pure_string : make_specified_string)
3205 (read_buffer, nchars, nbytes, multibyte));
3206 result = (uninterned_symbol ? Fmake_symbol (name)
3207 : Fintern (name, Qnil));
3208
3209 if (EQ (Vread_with_symbol_positions, Qt)
3210 || EQ (Vread_with_symbol_positions, readcharfun))
3211 Vread_symbol_positions_list
3212 = Fcons (Fcons (result, make_number (start_position)),
3213 Vread_symbol_positions_list);
3214 return result;
3215 }
3216 }
3217 }
3218 }
3219 \f
3220
3221 /* List of nodes we've seen during substitute_object_in_subtree. */
3222 static Lisp_Object seen_list;
3223
3224 static void
3225 substitute_object_in_subtree (Lisp_Object object, Lisp_Object placeholder)
3226 {
3227 Lisp_Object check_object;
3228
3229 /* We haven't seen any objects when we start. */
3230 seen_list = Qnil;
3231
3232 /* Make all the substitutions. */
3233 check_object
3234 = substitute_object_recurse (object, placeholder, object);
3235
3236 /* Clear seen_list because we're done with it. */
3237 seen_list = Qnil;
3238
3239 /* The returned object here is expected to always eq the
3240 original. */
3241 if (!EQ (check_object, object))
3242 error ("Unexpected mutation error in reader");
3243 }
3244
3245 /* Feval doesn't get called from here, so no gc protection is needed. */
3246 #define SUBSTITUTE(get_val, set_val) \
3247 do { \
3248 Lisp_Object old_value = get_val; \
3249 Lisp_Object true_value \
3250 = substitute_object_recurse (object, placeholder, \
3251 old_value); \
3252 \
3253 if (!EQ (old_value, true_value)) \
3254 { \
3255 set_val; \
3256 } \
3257 } while (0)
3258
3259 static Lisp_Object
3260 substitute_object_recurse (Lisp_Object object, Lisp_Object placeholder, Lisp_Object subtree)
3261 {
3262 /* If we find the placeholder, return the target object. */
3263 if (EQ (placeholder, subtree))
3264 return object;
3265
3266 /* If we've been to this node before, don't explore it again. */
3267 if (!EQ (Qnil, Fmemq (subtree, seen_list)))
3268 return subtree;
3269
3270 /* If this node can be the entry point to a cycle, remember that
3271 we've seen it. It can only be such an entry point if it was made
3272 by #n=, which means that we can find it as a value in
3273 read_objects. */
3274 if (!EQ (Qnil, Frassq (subtree, read_objects)))
3275 seen_list = Fcons (subtree, seen_list);
3276
3277 /* Recurse according to subtree's type.
3278 Every branch must return a Lisp_Object. */
3279 switch (XTYPE (subtree))
3280 {
3281 case Lisp_Vectorlike:
3282 {
3283 ptrdiff_t i, length = 0;
3284 if (BOOL_VECTOR_P (subtree))
3285 return subtree; /* No sub-objects anyway. */
3286 else if (CHAR_TABLE_P (subtree) || SUB_CHAR_TABLE_P (subtree)
3287 || COMPILEDP (subtree) || HASH_TABLE_P (subtree))
3288 length = ASIZE (subtree) & PSEUDOVECTOR_SIZE_MASK;
3289 else if (VECTORP (subtree))
3290 length = ASIZE (subtree);
3291 else
3292 /* An unknown pseudovector may contain non-Lisp fields, so we
3293 can't just blindly traverse all its fields. We used to call
3294 `Flength' which signaled `sequencep', so I just preserved this
3295 behavior. */
3296 wrong_type_argument (Qsequencep, subtree);
3297
3298 for (i = 0; i < length; i++)
3299 SUBSTITUTE (AREF (subtree, i),
3300 ASET (subtree, i, true_value));
3301 return subtree;
3302 }
3303
3304 case Lisp_Cons:
3305 {
3306 SUBSTITUTE (XCAR (subtree),
3307 XSETCAR (subtree, true_value));
3308 SUBSTITUTE (XCDR (subtree),
3309 XSETCDR (subtree, true_value));
3310 return subtree;
3311 }
3312
3313 case Lisp_String:
3314 {
3315 /* Check for text properties in each interval.
3316 substitute_in_interval contains part of the logic. */
3317
3318 INTERVAL root_interval = string_intervals (subtree);
3319 AUTO_CONS (arg, object, placeholder);
3320
3321 traverse_intervals_noorder (root_interval,
3322 &substitute_in_interval, arg);
3323
3324 return subtree;
3325 }
3326
3327 /* Other types don't recurse any further. */
3328 default:
3329 return subtree;
3330 }
3331 }
3332
3333 /* Helper function for substitute_object_recurse. */
3334 static void
3335 substitute_in_interval (INTERVAL interval, Lisp_Object arg)
3336 {
3337 Lisp_Object object = Fcar (arg);
3338 Lisp_Object placeholder = Fcdr (arg);
3339
3340 SUBSTITUTE (interval->plist, set_interval_plist (interval, true_value));
3341 }
3342
3343 \f
3344 #define LEAD_INT 1
3345 #define DOT_CHAR 2
3346 #define TRAIL_INT 4
3347 #define E_EXP 16
3348
3349
3350 /* Convert STRING to a number, assuming base BASE. Return a fixnum if CP has
3351 integer syntax and fits in a fixnum, else return the nearest float if CP has
3352 either floating point or integer syntax and BASE is 10, else return nil. If
3353 IGNORE_TRAILING, consider just the longest prefix of CP that has
3354 valid floating point syntax. Signal an overflow if BASE is not 10 and the
3355 number has integer syntax but does not fit. */
3356
3357 Lisp_Object
3358 string_to_number (char const *string, int base, bool ignore_trailing)
3359 {
3360 int state;
3361 char const *cp = string;
3362 int leading_digit;
3363 bool float_syntax = 0;
3364 double value = 0;
3365
3366 /* Negate the value ourselves. This treats 0, NaNs, and infinity properly on
3367 IEEE floating point hosts, and works around a formerly-common bug where
3368 atof ("-0.0") drops the sign. */
3369 bool negative = *cp == '-';
3370
3371 bool signedp = negative || *cp == '+';
3372 cp += signedp;
3373
3374 state = 0;
3375
3376 leading_digit = digit_to_number (*cp, base);
3377 if (leading_digit >= 0)
3378 {
3379 state |= LEAD_INT;
3380 do
3381 ++cp;
3382 while (digit_to_number (*cp, base) >= 0);
3383 }
3384 if (*cp == '.')
3385 {
3386 state |= DOT_CHAR;
3387 cp++;
3388 }
3389
3390 if (base == 10)
3391 {
3392 if ('0' <= *cp && *cp <= '9')
3393 {
3394 state |= TRAIL_INT;
3395 do
3396 cp++;
3397 while ('0' <= *cp && *cp <= '9');
3398 }
3399 if (*cp == 'e' || *cp == 'E')
3400 {
3401 char const *ecp = cp;
3402 cp++;
3403 if (*cp == '+' || *cp == '-')
3404 cp++;
3405 if ('0' <= *cp && *cp <= '9')
3406 {
3407 state |= E_EXP;
3408 do
3409 cp++;
3410 while ('0' <= *cp && *cp <= '9');
3411 }
3412 else if (cp[-1] == '+'
3413 && cp[0] == 'I' && cp[1] == 'N' && cp[2] == 'F')
3414 {
3415 state |= E_EXP;
3416 cp += 3;
3417 value = INFINITY;
3418 }
3419 else if (cp[-1] == '+'
3420 && cp[0] == 'N' && cp[1] == 'a' && cp[2] == 'N')
3421 {
3422 state |= E_EXP;
3423 cp += 3;
3424 /* NAN is a "positive" NaN on all known Emacs hosts. */
3425 value = NAN;
3426 }
3427 else
3428 cp = ecp;
3429 }
3430
3431 float_syntax = ((state & (DOT_CHAR|TRAIL_INT)) == (DOT_CHAR|TRAIL_INT)
3432 || state == (LEAD_INT|E_EXP));
3433 }
3434
3435 /* Return nil if the number uses invalid syntax. If IGNORE_TRAILING, accept
3436 any prefix that matches. Otherwise, the entire string must match. */
3437 if (! (ignore_trailing
3438 ? ((state & LEAD_INT) != 0 || float_syntax)
3439 : (!*cp && ((state & ~DOT_CHAR) == LEAD_INT || float_syntax))))
3440 return Qnil;
3441
3442 /* If the number uses integer and not float syntax, and is in C-language
3443 range, use its value, preferably as a fixnum. */
3444 if (leading_digit >= 0 && ! float_syntax)
3445 {
3446 uintmax_t n;
3447
3448 /* Fast special case for single-digit integers. This also avoids a
3449 glitch when BASE is 16 and IGNORE_TRAILING, because in that
3450 case some versions of strtoumax accept numbers like "0x1" that Emacs
3451 does not allow. */
3452 if (digit_to_number (string[signedp + 1], base) < 0)
3453 return make_number (negative ? -leading_digit : leading_digit);
3454
3455 errno = 0;
3456 n = strtoumax (string + signedp, NULL, base);
3457 if (errno == ERANGE)
3458 {
3459 /* Unfortunately there's no simple and accurate way to convert
3460 non-base-10 numbers that are out of C-language range. */
3461 if (base != 10)
3462 xsignal1 (Qoverflow_error, build_string (string));
3463 }
3464 else if (n <= (negative ? -MOST_NEGATIVE_FIXNUM : MOST_POSITIVE_FIXNUM))
3465 {
3466 EMACS_INT signed_n = n;
3467 return make_number (negative ? -signed_n : signed_n);
3468 }
3469 else
3470 value = n;
3471 }
3472
3473 /* Either the number uses float syntax, or it does not fit into a fixnum.
3474 Convert it from string to floating point, unless the value is already
3475 known because it is an infinity, a NAN, or its absolute value fits in
3476 uintmax_t. */
3477 if (! value)
3478 value = atof (string + signedp);
3479
3480 return make_float (negative ? -value : value);
3481 }
3482
3483 \f
3484 static Lisp_Object
3485 read_vector (Lisp_Object readcharfun, bool bytecodeflag)
3486 {
3487 ptrdiff_t i, size;
3488 Lisp_Object *ptr;
3489 Lisp_Object tem, item, vector;
3490 struct Lisp_Cons *otem;
3491 Lisp_Object len;
3492
3493 tem = read_list (1, readcharfun);
3494 len = Flength (tem);
3495 vector = Fmake_vector (len, Qnil);
3496
3497 size = ASIZE (vector);
3498 ptr = XVECTOR (vector)->contents;
3499 for (i = 0; i < size; i++)
3500 {
3501 item = Fcar (tem);
3502 /* If `load-force-doc-strings' is t when reading a lazily-loaded
3503 bytecode object, the docstring containing the bytecode and
3504 constants values must be treated as unibyte and passed to
3505 Fread, to get the actual bytecode string and constants vector. */
3506 if (bytecodeflag && load_force_doc_strings)
3507 {
3508 if (i == COMPILED_BYTECODE)
3509 {
3510 if (!STRINGP (item))
3511 error ("Invalid byte code");
3512
3513 /* Delay handling the bytecode slot until we know whether
3514 it is lazily-loaded (we can tell by whether the
3515 constants slot is nil). */
3516 ASET (vector, COMPILED_CONSTANTS, item);
3517 item = Qnil;
3518 }
3519 else if (i == COMPILED_CONSTANTS)
3520 {
3521 Lisp_Object bytestr = ptr[COMPILED_CONSTANTS];
3522
3523 if (NILP (item))
3524 {
3525 /* Coerce string to unibyte (like string-as-unibyte,
3526 but without generating extra garbage and
3527 guaranteeing no change in the contents). */
3528 STRING_SET_CHARS (bytestr, SBYTES (bytestr));
3529 STRING_SET_UNIBYTE (bytestr);
3530
3531 item = Fread (Fcons (bytestr, readcharfun));
3532 if (!CONSP (item))
3533 error ("Invalid byte code");
3534
3535 otem = XCONS (item);
3536 bytestr = XCAR (item);
3537 item = XCDR (item);
3538 free_cons (otem);
3539 }
3540
3541 /* Now handle the bytecode slot. */
3542 ASET (vector, COMPILED_BYTECODE, bytestr);
3543 }
3544 else if (i == COMPILED_DOC_STRING
3545 && STRINGP (item)
3546 && ! STRING_MULTIBYTE (item))
3547 {
3548 if (EQ (readcharfun, Qget_emacs_mule_file_char))
3549 item = Fdecode_coding_string (item, Qemacs_mule, Qnil, Qnil);
3550 else
3551 item = Fstring_as_multibyte (item);
3552 }
3553 }
3554 ASET (vector, i, item);
3555 otem = XCONS (tem);
3556 tem = Fcdr (tem);
3557 free_cons (otem);
3558 }
3559 return vector;
3560 }
3561
3562 /* FLAG means check for ']' to terminate rather than ')' and '.'. */
3563
3564 static Lisp_Object
3565 read_list (bool flag, Lisp_Object readcharfun)
3566 {
3567 Lisp_Object val, tail;
3568 Lisp_Object elt, tem;
3569 struct gcpro gcpro1, gcpro2;
3570 /* 0 is the normal case.
3571 1 means this list is a doc reference; replace it with the number 0.
3572 2 means this list is a doc reference; replace it with the doc string. */
3573 int doc_reference = 0;
3574
3575 /* Initialize this to 1 if we are reading a list. */
3576 bool first_in_list = flag <= 0;
3577
3578 val = Qnil;
3579 tail = Qnil;
3580
3581 while (1)
3582 {
3583 int ch;
3584 GCPRO2 (val, tail);
3585 elt = read1 (readcharfun, &ch, first_in_list);
3586 UNGCPRO;
3587
3588 first_in_list = 0;
3589
3590 /* While building, if the list starts with #$, treat it specially. */
3591 if (EQ (elt, Vload_file_name)
3592 && ! NILP (elt)
3593 && !NILP (Vpurify_flag))
3594 {
3595 if (NILP (Vdoc_file_name))
3596 /* We have not yet called Snarf-documentation, so assume
3597 this file is described in the DOC file
3598 and Snarf-documentation will fill in the right value later.
3599 For now, replace the whole list with 0. */
3600 doc_reference = 1;
3601 else
3602 /* We have already called Snarf-documentation, so make a relative
3603 file name for this file, so it can be found properly
3604 in the installed Lisp directory.
3605 We don't use Fexpand_file_name because that would make
3606 the directory absolute now. */
3607 {
3608 AUTO_STRING (dot_dot_lisp, "../lisp/");
3609 elt = concat2 (dot_dot_lisp, Ffile_name_nondirectory (elt));
3610 }
3611 }
3612 else if (EQ (elt, Vload_file_name)
3613 && ! NILP (elt)
3614 && load_force_doc_strings)
3615 doc_reference = 2;
3616
3617 if (ch)
3618 {
3619 if (flag > 0)
3620 {
3621 if (ch == ']')
3622 return val;
3623 invalid_syntax (") or . in a vector");
3624 }
3625 if (ch == ')')
3626 return val;
3627 if (ch == '.')
3628 {
3629 GCPRO2 (val, tail);
3630 if (!NILP (tail))
3631 XSETCDR (tail, read0 (readcharfun));
3632 else
3633 val = read0 (readcharfun);
3634 read1 (readcharfun, &ch, 0);
3635 UNGCPRO;
3636 if (ch == ')')
3637 {
3638 if (doc_reference == 1)
3639 return make_number (0);
3640 if (doc_reference == 2 && INTEGERP (XCDR (val)))
3641 {
3642 char *saved = NULL;
3643 file_offset saved_position;
3644 /* Get a doc string from the file we are loading.
3645 If it's in saved_doc_string, get it from there.
3646
3647 Here, we don't know if the string is a
3648 bytecode string or a doc string. As a
3649 bytecode string must be unibyte, we always
3650 return a unibyte string. If it is actually a
3651 doc string, caller must make it
3652 multibyte. */
3653
3654 /* Position is negative for user variables. */
3655 EMACS_INT pos = eabs (XINT (XCDR (val)));
3656 if (pos >= saved_doc_string_position
3657 && pos < (saved_doc_string_position
3658 + saved_doc_string_length))
3659 {
3660 saved = saved_doc_string;
3661 saved_position = saved_doc_string_position;
3662 }
3663 /* Look in prev_saved_doc_string the same way. */
3664 else if (pos >= prev_saved_doc_string_position
3665 && pos < (prev_saved_doc_string_position
3666 + prev_saved_doc_string_length))
3667 {
3668 saved = prev_saved_doc_string;
3669 saved_position = prev_saved_doc_string_position;
3670 }
3671 if (saved)
3672 {
3673 ptrdiff_t start = pos - saved_position;
3674 ptrdiff_t from, to;
3675
3676 /* Process quoting with ^A,
3677 and find the end of the string,
3678 which is marked with ^_ (037). */
3679 for (from = start, to = start;
3680 saved[from] != 037;)
3681 {
3682 int c = saved[from++];
3683 if (c == 1)
3684 {
3685 c = saved[from++];
3686 saved[to++] = (c == 1 ? c
3687 : c == '0' ? 0
3688 : c == '_' ? 037
3689 : c);
3690 }
3691 else
3692 saved[to++] = c;
3693 }
3694
3695 return make_unibyte_string (saved + start,
3696 to - start);
3697 }
3698 else
3699 return get_doc_string (val, 1, 0);
3700 }
3701
3702 return val;
3703 }
3704 invalid_syntax (". in wrong context");
3705 }
3706 invalid_syntax ("] in a list");
3707 }
3708 tem = list1 (elt);
3709 if (!NILP (tail))
3710 XSETCDR (tail, tem);
3711 else
3712 val = tem;
3713 tail = tem;
3714 }
3715 }
3716 \f
3717 static Lisp_Object initial_obarray;
3718
3719 /* `oblookup' stores the bucket number here, for the sake of Funintern. */
3720
3721 static size_t oblookup_last_bucket_number;
3722
3723 /* Get an error if OBARRAY is not an obarray.
3724 If it is one, return it. */
3725
3726 Lisp_Object
3727 check_obarray (Lisp_Object obarray)
3728 {
3729 if (!VECTORP (obarray) || ASIZE (obarray) == 0)
3730 {
3731 /* If Vobarray is now invalid, force it to be valid. */
3732 if (EQ (Vobarray, obarray)) Vobarray = initial_obarray;
3733 wrong_type_argument (Qvectorp, obarray);
3734 }
3735 return obarray;
3736 }
3737
3738 /* Intern symbol SYM in OBARRAY using bucket INDEX. */
3739
3740 static Lisp_Object
3741 intern_sym (Lisp_Object sym, Lisp_Object obarray, Lisp_Object index)
3742 {
3743 Lisp_Object *ptr;
3744
3745 XSYMBOL (sym)->interned = (EQ (obarray, initial_obarray)
3746 ? SYMBOL_INTERNED_IN_INITIAL_OBARRAY
3747 : SYMBOL_INTERNED);
3748
3749 if (SREF (SYMBOL_NAME (sym), 0) == ':' && EQ (obarray, initial_obarray))
3750 {
3751 XSYMBOL (sym)->constant = 1;
3752 XSYMBOL (sym)->redirect = SYMBOL_PLAINVAL;
3753 SET_SYMBOL_VAL (XSYMBOL (sym), sym);
3754 }
3755
3756 ptr = aref_addr (obarray, XINT (index));
3757 set_symbol_next (sym, SYMBOLP (*ptr) ? XSYMBOL (*ptr) : NULL);
3758 *ptr = sym;
3759 return sym;
3760 }
3761
3762 /* Intern a symbol with name STRING in OBARRAY using bucket INDEX. */
3763
3764 Lisp_Object
3765 intern_driver (Lisp_Object string, Lisp_Object obarray, Lisp_Object index)
3766 {
3767 return intern_sym (Fmake_symbol (string), obarray, index);
3768 }
3769
3770 /* Intern the C string STR: return a symbol with that name,
3771 interned in the current obarray. */
3772
3773 Lisp_Object
3774 intern_1 (const char *str, ptrdiff_t len)
3775 {
3776 Lisp_Object obarray = check_obarray (Vobarray);
3777 Lisp_Object tem = oblookup (obarray, str, len, len);
3778
3779 return SYMBOLP (tem) ? tem : intern_driver (make_string (str, len),
3780 obarray, tem);
3781 }
3782
3783 Lisp_Object
3784 intern_c_string_1 (const char *str, ptrdiff_t len)
3785 {
3786 Lisp_Object obarray = check_obarray (Vobarray);
3787 Lisp_Object tem = oblookup (obarray, str, len, len);
3788
3789 if (!SYMBOLP (tem))
3790 {
3791 /* Creating a non-pure string from a string literal not implemented yet.
3792 We could just use make_string here and live with the extra copy. */
3793 eassert (!NILP (Vpurify_flag));
3794 tem = intern_driver (make_pure_c_string (str, len), obarray, tem);
3795 }
3796 return tem;
3797 }
3798
3799 static void
3800 define_symbol (Lisp_Object sym, char const *str)
3801 {
3802 ptrdiff_t len = strlen (str);
3803 Lisp_Object string = make_pure_c_string (str, len);
3804 init_symbol (sym, string);
3805
3806 /* Qunbound is uninterned, so that it's not confused with any symbol
3807 'unbound' created by a Lisp program. */
3808 if (! EQ (sym, Qunbound))
3809 {
3810 Lisp_Object bucket = oblookup (initial_obarray, str, len, len);
3811 eassert (INTEGERP (bucket));
3812 intern_sym (sym, initial_obarray, bucket);
3813 }
3814 }
3815 \f
3816 DEFUN ("intern", Fintern, Sintern, 1, 2, 0,
3817 doc: /* Return the canonical symbol whose name is STRING.
3818 If there is none, one is created by this function and returned.
3819 A second optional argument specifies the obarray to use;
3820 it defaults to the value of `obarray'. */)
3821 (Lisp_Object string, Lisp_Object obarray)
3822 {
3823 Lisp_Object tem;
3824
3825 obarray = check_obarray (NILP (obarray) ? Vobarray : obarray);
3826 CHECK_STRING (string);
3827
3828 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3829 if (!SYMBOLP (tem))
3830 tem = intern_driver (NILP (Vpurify_flag) ? string : Fpurecopy (string),
3831 obarray, tem);
3832 return tem;
3833 }
3834
3835 DEFUN ("intern-soft", Fintern_soft, Sintern_soft, 1, 2, 0,
3836 doc: /* Return the canonical symbol named NAME, or nil if none exists.
3837 NAME may be a string or a symbol. If it is a symbol, that exact
3838 symbol is searched for.
3839 A second optional argument specifies the obarray to use;
3840 it defaults to the value of `obarray'. */)
3841 (Lisp_Object name, Lisp_Object obarray)
3842 {
3843 register Lisp_Object tem, string;
3844
3845 if (NILP (obarray)) obarray = Vobarray;
3846 obarray = check_obarray (obarray);
3847
3848 if (!SYMBOLP (name))
3849 {
3850 CHECK_STRING (name);
3851 string = name;
3852 }
3853 else
3854 string = SYMBOL_NAME (name);
3855
3856 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3857 if (INTEGERP (tem) || (SYMBOLP (name) && !EQ (name, tem)))
3858 return Qnil;
3859 else
3860 return tem;
3861 }
3862 \f
3863 DEFUN ("unintern", Funintern, Sunintern, 1, 2, 0,
3864 doc: /* Delete the symbol named NAME, if any, from OBARRAY.
3865 The value is t if a symbol was found and deleted, nil otherwise.
3866 NAME may be a string or a symbol. If it is a symbol, that symbol
3867 is deleted, if it belongs to OBARRAY--no other symbol is deleted.
3868 OBARRAY, if nil, defaults to the value of the variable `obarray'.
3869 usage: (unintern NAME OBARRAY) */)
3870 (Lisp_Object name, Lisp_Object obarray)
3871 {
3872 register Lisp_Object string, tem;
3873 size_t hash;
3874
3875 if (NILP (obarray)) obarray = Vobarray;
3876 obarray = check_obarray (obarray);
3877
3878 if (SYMBOLP (name))
3879 string = SYMBOL_NAME (name);
3880 else
3881 {
3882 CHECK_STRING (name);
3883 string = name;
3884 }
3885
3886 tem = oblookup (obarray, SSDATA (string),
3887 SCHARS (string),
3888 SBYTES (string));
3889 if (INTEGERP (tem))
3890 return Qnil;
3891 /* If arg was a symbol, don't delete anything but that symbol itself. */
3892 if (SYMBOLP (name) && !EQ (name, tem))
3893 return Qnil;
3894
3895 /* There are plenty of other symbols which will screw up the Emacs
3896 session if we unintern them, as well as even more ways to use
3897 `setq' or `fset' or whatnot to make the Emacs session
3898 unusable. Let's not go down this silly road. --Stef */
3899 /* if (EQ (tem, Qnil) || EQ (tem, Qt))
3900 error ("Attempt to unintern t or nil"); */
3901
3902 XSYMBOL (tem)->interned = SYMBOL_UNINTERNED;
3903
3904 hash = oblookup_last_bucket_number;
3905
3906 if (EQ (AREF (obarray, hash), tem))
3907 {
3908 if (XSYMBOL (tem)->next)
3909 {
3910 Lisp_Object sym;
3911 XSETSYMBOL (sym, XSYMBOL (tem)->next);
3912 ASET (obarray, hash, sym);
3913 }
3914 else
3915 ASET (obarray, hash, make_number (0));
3916 }
3917 else
3918 {
3919 Lisp_Object tail, following;
3920
3921 for (tail = AREF (obarray, hash);
3922 XSYMBOL (tail)->next;
3923 tail = following)
3924 {
3925 XSETSYMBOL (following, XSYMBOL (tail)->next);
3926 if (EQ (following, tem))
3927 {
3928 set_symbol_next (tail, XSYMBOL (following)->next);
3929 break;
3930 }
3931 }
3932 }
3933
3934 return Qt;
3935 }
3936 \f
3937 /* Return the symbol in OBARRAY whose names matches the string
3938 of SIZE characters (SIZE_BYTE bytes) at PTR.
3939 If there is no such symbol, return the integer bucket number of
3940 where the symbol would be if it were present.
3941
3942 Also store the bucket number in oblookup_last_bucket_number. */
3943
3944 Lisp_Object
3945 oblookup (Lisp_Object obarray, register const char *ptr, ptrdiff_t size, ptrdiff_t size_byte)
3946 {
3947 size_t hash;
3948 size_t obsize;
3949 register Lisp_Object tail;
3950 Lisp_Object bucket, tem;
3951
3952 obarray = check_obarray (obarray);
3953 obsize = ASIZE (obarray);
3954
3955 /* This is sometimes needed in the middle of GC. */
3956 obsize &= ~ARRAY_MARK_FLAG;
3957 hash = hash_string (ptr, size_byte) % obsize;
3958 bucket = AREF (obarray, hash);
3959 oblookup_last_bucket_number = hash;
3960 if (EQ (bucket, make_number (0)))
3961 ;
3962 else if (!SYMBOLP (bucket))
3963 error ("Bad data in guts of obarray"); /* Like CADR error message. */
3964 else
3965 for (tail = bucket; ; XSETSYMBOL (tail, XSYMBOL (tail)->next))
3966 {
3967 if (SBYTES (SYMBOL_NAME (tail)) == size_byte
3968 && SCHARS (SYMBOL_NAME (tail)) == size
3969 && !memcmp (SDATA (SYMBOL_NAME (tail)), ptr, size_byte))
3970 return tail;
3971 else if (XSYMBOL (tail)->next == 0)
3972 break;
3973 }
3974 XSETINT (tem, hash);
3975 return tem;
3976 }
3977 \f
3978 void
3979 map_obarray (Lisp_Object obarray, void (*fn) (Lisp_Object, Lisp_Object), Lisp_Object arg)
3980 {
3981 ptrdiff_t i;
3982 register Lisp_Object tail;
3983 CHECK_VECTOR (obarray);
3984 for (i = ASIZE (obarray) - 1; i >= 0; i--)
3985 {
3986 tail = AREF (obarray, i);
3987 if (SYMBOLP (tail))
3988 while (1)
3989 {
3990 (*fn) (tail, arg);
3991 if (XSYMBOL (tail)->next == 0)
3992 break;
3993 XSETSYMBOL (tail, XSYMBOL (tail)->next);
3994 }
3995 }
3996 }
3997
3998 static void
3999 mapatoms_1 (Lisp_Object sym, Lisp_Object function)
4000 {
4001 call1 (function, sym);
4002 }
4003
4004 DEFUN ("mapatoms", Fmapatoms, Smapatoms, 1, 2, 0,
4005 doc: /* Call FUNCTION on every symbol in OBARRAY.
4006 OBARRAY defaults to the value of `obarray'. */)
4007 (Lisp_Object function, Lisp_Object obarray)
4008 {
4009 if (NILP (obarray)) obarray = Vobarray;
4010 obarray = check_obarray (obarray);
4011
4012 map_obarray (obarray, mapatoms_1, function);
4013 return Qnil;
4014 }
4015
4016 #define OBARRAY_SIZE 1511
4017
4018 void
4019 init_obarray (void)
4020 {
4021 Lisp_Object oblength;
4022 ptrdiff_t size = 100 + MAX_MULTIBYTE_LENGTH;
4023
4024 XSETFASTINT (oblength, OBARRAY_SIZE);
4025
4026 Vobarray = Fmake_vector (oblength, make_number (0));
4027 initial_obarray = Vobarray;
4028 staticpro (&initial_obarray);
4029
4030 for (int i = 0; i < ARRAYELTS (lispsym); i++)
4031 define_symbol (builtin_lisp_symbol (i), defsym_name[i]);
4032
4033 DEFSYM (Qunbound, "unbound");
4034
4035 DEFSYM (Qnil, "nil");
4036 SET_SYMBOL_VAL (XSYMBOL (Qnil), Qnil);
4037 XSYMBOL (Qnil)->constant = 1;
4038 XSYMBOL (Qnil)->declared_special = true;
4039
4040 DEFSYM (Qt, "t");
4041 SET_SYMBOL_VAL (XSYMBOL (Qt), Qt);
4042 XSYMBOL (Qt)->constant = 1;
4043 XSYMBOL (Qt)->declared_special = true;
4044
4045 /* Qt is correct even if CANNOT_DUMP. loadup.el will set to nil at end. */
4046 Vpurify_flag = Qt;
4047
4048 DEFSYM (Qvariable_documentation, "variable-documentation");
4049
4050 read_buffer = xmalloc (size);
4051 read_buffer_size = size;
4052 }
4053 \f
4054 void
4055 defsubr (struct Lisp_Subr *sname)
4056 {
4057 Lisp_Object sym, tem;
4058 sym = intern_c_string (sname->symbol_name);
4059 XSETPVECTYPE (sname, PVEC_SUBR);
4060 XSETSUBR (tem, sname);
4061 set_symbol_function (sym, tem);
4062 }
4063
4064 #ifdef NOTDEF /* Use fset in subr.el now! */
4065 void
4066 defalias (struct Lisp_Subr *sname, char *string)
4067 {
4068 Lisp_Object sym;
4069 sym = intern (string);
4070 XSETSUBR (XSYMBOL (sym)->function, sname);
4071 }
4072 #endif /* NOTDEF */
4073
4074 /* Define an "integer variable"; a symbol whose value is forwarded to a
4075 C variable of type EMACS_INT. Sample call (with "xx" to fool make-docfile):
4076 DEFxxVAR_INT ("emacs-priority", &emacs_priority, "Documentation"); */
4077 void
4078 defvar_int (struct Lisp_Intfwd *i_fwd,
4079 const char *namestring, EMACS_INT *address)
4080 {
4081 Lisp_Object sym;
4082 sym = intern_c_string (namestring);
4083 i_fwd->type = Lisp_Fwd_Int;
4084 i_fwd->intvar = address;
4085 XSYMBOL (sym)->declared_special = 1;
4086 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4087 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)i_fwd);
4088 }
4089
4090 /* Similar but define a variable whose value is t if address contains 1,
4091 nil if address contains 0. */
4092 void
4093 defvar_bool (struct Lisp_Boolfwd *b_fwd,
4094 const char *namestring, bool *address)
4095 {
4096 Lisp_Object sym;
4097 sym = intern_c_string (namestring);
4098 b_fwd->type = Lisp_Fwd_Bool;
4099 b_fwd->boolvar = address;
4100 XSYMBOL (sym)->declared_special = 1;
4101 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4102 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)b_fwd);
4103 Vbyte_boolean_vars = Fcons (sym, Vbyte_boolean_vars);
4104 }
4105
4106 /* Similar but define a variable whose value is the Lisp Object stored
4107 at address. Two versions: with and without gc-marking of the C
4108 variable. The nopro version is used when that variable will be
4109 gc-marked for some other reason, since marking the same slot twice
4110 can cause trouble with strings. */
4111 void
4112 defvar_lisp_nopro (struct Lisp_Objfwd *o_fwd,
4113 const char *namestring, Lisp_Object *address)
4114 {
4115 Lisp_Object sym;
4116 sym = intern_c_string (namestring);
4117 o_fwd->type = Lisp_Fwd_Obj;
4118 o_fwd->objvar = address;
4119 XSYMBOL (sym)->declared_special = 1;
4120 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4121 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)o_fwd);
4122 }
4123
4124 void
4125 defvar_lisp (struct Lisp_Objfwd *o_fwd,
4126 const char *namestring, Lisp_Object *address)
4127 {
4128 defvar_lisp_nopro (o_fwd, namestring, address);
4129 staticpro (address);
4130 }
4131
4132 /* Similar but define a variable whose value is the Lisp Object stored
4133 at a particular offset in the current kboard object. */
4134
4135 void
4136 defvar_kboard (struct Lisp_Kboard_Objfwd *ko_fwd,
4137 const char *namestring, int offset)
4138 {
4139 Lisp_Object sym;
4140 sym = intern_c_string (namestring);
4141 ko_fwd->type = Lisp_Fwd_Kboard_Obj;
4142 ko_fwd->offset = offset;
4143 XSYMBOL (sym)->declared_special = 1;
4144 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4145 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)ko_fwd);
4146 }
4147 \f
4148 /* Check that the elements of lpath exist. */
4149
4150 static void
4151 load_path_check (Lisp_Object lpath)
4152 {
4153 Lisp_Object path_tail;
4154
4155 /* The only elements that might not exist are those from
4156 PATH_LOADSEARCH, EMACSLOADPATH. Anything else is only added if
4157 it exists. */
4158 for (path_tail = lpath; !NILP (path_tail); path_tail = XCDR (path_tail))
4159 {
4160 Lisp_Object dirfile;
4161 dirfile = Fcar (path_tail);
4162 if (STRINGP (dirfile))
4163 {
4164 dirfile = Fdirectory_file_name (dirfile);
4165 if (! file_accessible_directory_p (dirfile))
4166 dir_warning ("Lisp directory", XCAR (path_tail));
4167 }
4168 }
4169 }
4170
4171 /* Return the default load-path, to be used if EMACSLOADPATH is unset.
4172 This does not include the standard site-lisp directories
4173 under the installation prefix (i.e., PATH_SITELOADSEARCH),
4174 but it does (unless no_site_lisp is set) include site-lisp
4175 directories in the source/build directories if those exist and we
4176 are running uninstalled.
4177
4178 Uses the following logic:
4179 If CANNOT_DUMP: Use PATH_LOADSEARCH.
4180 The remainder is what happens when dumping works:
4181 If purify-flag (ie dumping) just use PATH_DUMPLOADSEARCH.
4182 Otherwise use PATH_LOADSEARCH.
4183
4184 If !initialized, then just return PATH_DUMPLOADSEARCH.
4185 If initialized:
4186 If Vinstallation_directory is not nil (ie, running uninstalled):
4187 If installation-dir/lisp exists and not already a member,
4188 we must be running uninstalled. Reset the load-path
4189 to just installation-dir/lisp. (The default PATH_LOADSEARCH
4190 refers to the eventual installation directories. Since we
4191 are not yet installed, we should not use them, even if they exist.)
4192 If installation-dir/lisp does not exist, just add
4193 PATH_DUMPLOADSEARCH at the end instead.
4194 Add installation-dir/site-lisp (if !no_site_lisp, and exists
4195 and not already a member) at the front.
4196 If installation-dir != source-dir (ie running an uninstalled,
4197 out-of-tree build) AND install-dir/src/Makefile exists BUT
4198 install-dir/src/Makefile.in does NOT exist (this is a sanity
4199 check), then repeat the above steps for source-dir/lisp, site-lisp. */
4200
4201 static Lisp_Object
4202 load_path_default (void)
4203 {
4204 Lisp_Object lpath = Qnil;
4205 const char *normal;
4206
4207 #ifdef CANNOT_DUMP
4208 #ifdef HAVE_NS
4209 const char *loadpath = ns_load_path ();
4210 #endif
4211
4212 normal = PATH_LOADSEARCH;
4213 #ifdef HAVE_NS
4214 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4215 #else
4216 lpath = decode_env_path (0, normal, 0);
4217 #endif
4218
4219 #else /* !CANNOT_DUMP */
4220
4221 normal = NILP (Vpurify_flag) ? PATH_LOADSEARCH : PATH_DUMPLOADSEARCH;
4222
4223 if (initialized)
4224 {
4225 #ifdef HAVE_NS
4226 const char *loadpath = ns_load_path ();
4227 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4228 #else
4229 lpath = decode_env_path (0, normal, 0);
4230 #endif
4231 if (!NILP (Vinstallation_directory))
4232 {
4233 Lisp_Object tem, tem1;
4234
4235 /* Add to the path the lisp subdir of the installation
4236 dir, if it is accessible. Note: in out-of-tree builds,
4237 this directory is empty save for Makefile. */
4238 tem = Fexpand_file_name (build_string ("lisp"),
4239 Vinstallation_directory);
4240 tem1 = Ffile_accessible_directory_p (tem);
4241 if (!NILP (tem1))
4242 {
4243 if (NILP (Fmember (tem, lpath)))
4244 {
4245 /* We are running uninstalled. The default load-path
4246 points to the eventual installed lisp directories.
4247 We should not use those now, even if they exist,
4248 so start over from a clean slate. */
4249 lpath = list1 (tem);
4250 }
4251 }
4252 else
4253 /* That dir doesn't exist, so add the build-time
4254 Lisp dirs instead. */
4255 {
4256 Lisp_Object dump_path =
4257 decode_env_path (0, PATH_DUMPLOADSEARCH, 0);
4258 lpath = nconc2 (lpath, dump_path);
4259 }
4260
4261 /* Add site-lisp under the installation dir, if it exists. */
4262 if (!no_site_lisp)
4263 {
4264 tem = Fexpand_file_name (build_string ("site-lisp"),
4265 Vinstallation_directory);
4266 tem1 = Ffile_accessible_directory_p (tem);
4267 if (!NILP (tem1))
4268 {
4269 if (NILP (Fmember (tem, lpath)))
4270 lpath = Fcons (tem, lpath);
4271 }
4272 }
4273
4274 /* If Emacs was not built in the source directory,
4275 and it is run from where it was built, add to load-path
4276 the lisp and site-lisp dirs under that directory. */
4277
4278 if (NILP (Fequal (Vinstallation_directory, Vsource_directory)))
4279 {
4280 Lisp_Object tem2;
4281
4282 tem = Fexpand_file_name (build_string ("src/Makefile"),
4283 Vinstallation_directory);
4284 tem1 = Ffile_exists_p (tem);
4285
4286 /* Don't be fooled if they moved the entire source tree
4287 AFTER dumping Emacs. If the build directory is indeed
4288 different from the source dir, src/Makefile.in and
4289 src/Makefile will not be found together. */
4290 tem = Fexpand_file_name (build_string ("src/Makefile.in"),
4291 Vinstallation_directory);
4292 tem2 = Ffile_exists_p (tem);
4293 if (!NILP (tem1) && NILP (tem2))
4294 {
4295 tem = Fexpand_file_name (build_string ("lisp"),
4296 Vsource_directory);
4297
4298 if (NILP (Fmember (tem, lpath)))
4299 lpath = Fcons (tem, lpath);
4300
4301 if (!no_site_lisp)
4302 {
4303 tem = Fexpand_file_name (build_string ("site-lisp"),
4304 Vsource_directory);
4305 tem1 = Ffile_accessible_directory_p (tem);
4306 if (!NILP (tem1))
4307 {
4308 if (NILP (Fmember (tem, lpath)))
4309 lpath = Fcons (tem, lpath);
4310 }
4311 }
4312 }
4313 } /* Vinstallation_directory != Vsource_directory */
4314
4315 } /* if Vinstallation_directory */
4316 }
4317 else /* !initialized */
4318 {
4319 /* NORMAL refers to PATH_DUMPLOADSEARCH, ie the lisp dir in the
4320 source directory. We used to add ../lisp (ie the lisp dir in
4321 the build directory) at the front here, but that should not
4322 be necessary, since in out of tree builds lisp/ is empty, save
4323 for Makefile. */
4324 lpath = decode_env_path (0, normal, 0);
4325 }
4326 #endif /* !CANNOT_DUMP */
4327
4328 return lpath;
4329 }
4330
4331 void
4332 init_lread (void)
4333 {
4334 /* First, set Vload_path. */
4335
4336 /* Ignore EMACSLOADPATH when dumping. */
4337 #ifdef CANNOT_DUMP
4338 bool use_loadpath = true;
4339 #else
4340 bool use_loadpath = NILP (Vpurify_flag);
4341 #endif
4342
4343 if (use_loadpath && egetenv ("EMACSLOADPATH"))
4344 {
4345 Vload_path = decode_env_path ("EMACSLOADPATH", 0, 1);
4346
4347 /* Check (non-nil) user-supplied elements. */
4348 load_path_check (Vload_path);
4349
4350 /* If no nils in the environment variable, use as-is.
4351 Otherwise, replace any nils with the default. */
4352 if (! NILP (Fmemq (Qnil, Vload_path)))
4353 {
4354 Lisp_Object elem, elpath = Vload_path;
4355 Lisp_Object default_lpath = load_path_default ();
4356
4357 /* Check defaults, before adding site-lisp. */
4358 load_path_check (default_lpath);
4359
4360 /* Add the site-lisp directories to the front of the default. */
4361 if (!no_site_lisp)
4362 {
4363 Lisp_Object sitelisp;
4364 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4365 if (! NILP (sitelisp))
4366 default_lpath = nconc2 (sitelisp, default_lpath);
4367 }
4368
4369 Vload_path = Qnil;
4370
4371 /* Replace nils from EMACSLOADPATH by default. */
4372 while (CONSP (elpath))
4373 {
4374 elem = XCAR (elpath);
4375 elpath = XCDR (elpath);
4376 Vload_path = CALLN (Fappend, Vload_path,
4377 NILP (elem) ? default_lpath : list1 (elem));
4378 }
4379 } /* Fmemq (Qnil, Vload_path) */
4380 }
4381 else
4382 {
4383 Vload_path = load_path_default ();
4384
4385 /* Check before adding site-lisp directories.
4386 The install should have created them, but they are not
4387 required, so no need to warn if they are absent.
4388 Or we might be running before installation. */
4389 load_path_check (Vload_path);
4390
4391 /* Add the site-lisp directories at the front. */
4392 if (initialized && !no_site_lisp)
4393 {
4394 Lisp_Object sitelisp;
4395 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4396 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4397 }
4398 }
4399
4400 Vvalues = Qnil;
4401
4402 load_in_progress = 0;
4403 Vload_file_name = Qnil;
4404 Vstandard_input = Qt;
4405 Vloads_in_progress = Qnil;
4406 }
4407
4408 /* Print a warning that directory intended for use USE and with name
4409 DIRNAME cannot be accessed. On entry, errno should correspond to
4410 the access failure. Print the warning on stderr and put it in
4411 *Messages*. */
4412
4413 void
4414 dir_warning (char const *use, Lisp_Object dirname)
4415 {
4416 static char const format[] = "Warning: %s `%s': %s\n";
4417 int access_errno = errno;
4418 fprintf (stderr, format, use, SSDATA (dirname), strerror (access_errno));
4419
4420 /* Don't log the warning before we've initialized!! */
4421 if (initialized)
4422 {
4423 char const *diagnostic = emacs_strerror (access_errno);
4424 USE_SAFE_ALLOCA;
4425 char *buffer = SAFE_ALLOCA (sizeof format - 3 * (sizeof "%s" - 1)
4426 + strlen (use) + SBYTES (dirname)
4427 + strlen (diagnostic));
4428 ptrdiff_t message_len = esprintf (buffer, format, use, SSDATA (dirname),
4429 diagnostic);
4430 message_dolog (buffer, message_len, 0, STRING_MULTIBYTE (dirname));
4431 SAFE_FREE ();
4432 }
4433 }
4434
4435 void
4436 syms_of_lread (void)
4437 {
4438 defsubr (&Sread);
4439 defsubr (&Sread_from_string);
4440 defsubr (&Sintern);
4441 defsubr (&Sintern_soft);
4442 defsubr (&Sunintern);
4443 defsubr (&Sget_load_suffixes);
4444 defsubr (&Sload);
4445 defsubr (&Seval_buffer);
4446 defsubr (&Seval_region);
4447 defsubr (&Sread_char);
4448 defsubr (&Sread_char_exclusive);
4449 defsubr (&Sread_event);
4450 defsubr (&Sget_file_char);
4451 defsubr (&Smapatoms);
4452 defsubr (&Slocate_file_internal);
4453
4454 DEFVAR_LISP ("obarray", Vobarray,
4455 doc: /* Symbol table for use by `intern' and `read'.
4456 It is a vector whose length ought to be prime for best results.
4457 The vector's contents don't make sense if examined from Lisp programs;
4458 to find all the symbols in an obarray, use `mapatoms'. */);
4459
4460 DEFVAR_LISP ("values", Vvalues,
4461 doc: /* List of values of all expressions which were read, evaluated and printed.
4462 Order is reverse chronological. */);
4463 XSYMBOL (intern ("values"))->declared_special = 0;
4464
4465 DEFVAR_LISP ("standard-input", Vstandard_input,
4466 doc: /* Stream for read to get input from.
4467 See documentation of `read' for possible values. */);
4468 Vstandard_input = Qt;
4469
4470 DEFVAR_LISP ("read-with-symbol-positions", Vread_with_symbol_positions,
4471 doc: /* If non-nil, add position of read symbols to `read-symbol-positions-list'.
4472
4473 If this variable is a buffer, then only forms read from that buffer
4474 will be added to `read-symbol-positions-list'.
4475 If this variable is t, then all read forms will be added.
4476 The effect of all other values other than nil are not currently
4477 defined, although they may be in the future.
4478
4479 The positions are relative to the last call to `read' or
4480 `read-from-string'. It is probably a bad idea to set this variable at
4481 the toplevel; bind it instead. */);
4482 Vread_with_symbol_positions = Qnil;
4483
4484 DEFVAR_LISP ("read-symbol-positions-list", Vread_symbol_positions_list,
4485 doc: /* A list mapping read symbols to their positions.
4486 This variable is modified during calls to `read' or
4487 `read-from-string', but only when `read-with-symbol-positions' is
4488 non-nil.
4489
4490 Each element of the list looks like (SYMBOL . CHAR-POSITION), where
4491 CHAR-POSITION is an integer giving the offset of that occurrence of the
4492 symbol from the position where `read' or `read-from-string' started.
4493
4494 Note that a symbol will appear multiple times in this list, if it was
4495 read multiple times. The list is in the same order as the symbols
4496 were read in. */);
4497 Vread_symbol_positions_list = Qnil;
4498
4499 DEFVAR_LISP ("read-circle", Vread_circle,
4500 doc: /* Non-nil means read recursive structures using #N= and #N# syntax. */);
4501 Vread_circle = Qt;
4502
4503 DEFVAR_LISP ("load-path", Vload_path,
4504 doc: /* List of directories to search for files to load.
4505 Each element is a string (directory name) or nil (meaning `default-directory').
4506 Initialized during startup as described in Info node `(elisp)Library Search'. */);
4507
4508 DEFVAR_LISP ("load-suffixes", Vload_suffixes,
4509 doc: /* List of suffixes for (compiled or source) Emacs Lisp files.
4510 This list should not include the empty string.
4511 `load' and related functions try to append these suffixes, in order,
4512 to the specified file name if a Lisp suffix is allowed or required. */);
4513 Vload_suffixes = list2 (build_pure_c_string (".elc"),
4514 build_pure_c_string (".el"));
4515 DEFVAR_LISP ("load-file-rep-suffixes", Vload_file_rep_suffixes,
4516 doc: /* List of suffixes that indicate representations of \
4517 the same file.
4518 This list should normally start with the empty string.
4519
4520 Enabling Auto Compression mode appends the suffixes in
4521 `jka-compr-load-suffixes' to this list and disabling Auto Compression
4522 mode removes them again. `load' and related functions use this list to
4523 determine whether they should look for compressed versions of a file
4524 and, if so, which suffixes they should try to append to the file name
4525 in order to do so. However, if you want to customize which suffixes
4526 the loading functions recognize as compression suffixes, you should
4527 customize `jka-compr-load-suffixes' rather than the present variable. */);
4528 Vload_file_rep_suffixes = list1 (empty_unibyte_string);
4529
4530 DEFVAR_BOOL ("load-in-progress", load_in_progress,
4531 doc: /* Non-nil if inside of `load'. */);
4532 DEFSYM (Qload_in_progress, "load-in-progress");
4533
4534 DEFVAR_LISP ("after-load-alist", Vafter_load_alist,
4535 doc: /* An alist of functions to be evalled when particular files are loaded.
4536 Each element looks like (REGEXP-OR-FEATURE FUNCS...).
4537
4538 REGEXP-OR-FEATURE is either a regular expression to match file names, or
4539 a symbol \(a feature name).
4540
4541 When `load' is run and the file-name argument matches an element's
4542 REGEXP-OR-FEATURE, or when `provide' is run and provides the symbol
4543 REGEXP-OR-FEATURE, the FUNCS in the element are called.
4544
4545 An error in FORMS does not undo the load, but does prevent execution of
4546 the rest of the FORMS. */);
4547 Vafter_load_alist = Qnil;
4548
4549 DEFVAR_LISP ("load-history", Vload_history,
4550 doc: /* Alist mapping loaded file names to symbols and features.
4551 Each alist element should be a list (FILE-NAME ENTRIES...), where
4552 FILE-NAME is the name of a file that has been loaded into Emacs.
4553 The file name is absolute and true (i.e. it doesn't contain symlinks).
4554 As an exception, one of the alist elements may have FILE-NAME nil,
4555 for symbols and features not associated with any file.
4556
4557 The remaining ENTRIES in the alist element describe the functions and
4558 variables defined in that file, the features provided, and the
4559 features required. Each entry has the form `(provide . FEATURE)',
4560 `(require . FEATURE)', `(defun . FUNCTION)', `(autoload . SYMBOL)',
4561 `(defface . SYMBOL)', or `(t . SYMBOL)'. Entries like `(t . SYMBOL)'
4562 may precede a `(defun . FUNCTION)' entry, and means that SYMBOL was an
4563 autoload before this file redefined it as a function. In addition,
4564 entries may also be single symbols, which means that SYMBOL was
4565 defined by `defvar' or `defconst'.
4566
4567 During preloading, the file name recorded is relative to the main Lisp
4568 directory. These file names are converted to absolute at startup. */);
4569 Vload_history = Qnil;
4570
4571 DEFVAR_LISP ("load-file-name", Vload_file_name,
4572 doc: /* Full name of file being loaded by `load'. */);
4573 Vload_file_name = Qnil;
4574
4575 DEFVAR_LISP ("user-init-file", Vuser_init_file,
4576 doc: /* File name, including directory, of user's initialization file.
4577 If the file loaded had extension `.elc', and the corresponding source file
4578 exists, this variable contains the name of source file, suitable for use
4579 by functions like `custom-save-all' which edit the init file.
4580 While Emacs loads and evaluates the init file, value is the real name
4581 of the file, regardless of whether or not it has the `.elc' extension. */);
4582 Vuser_init_file = Qnil;
4583
4584 DEFVAR_LISP ("current-load-list", Vcurrent_load_list,
4585 doc: /* Used for internal purposes by `load'. */);
4586 Vcurrent_load_list = Qnil;
4587
4588 DEFVAR_LISP ("load-read-function", Vload_read_function,
4589 doc: /* Function used by `load' and `eval-region' for reading expressions.
4590 The default is nil, which means use the function `read'. */);
4591 Vload_read_function = Qnil;
4592
4593 DEFVAR_LISP ("load-source-file-function", Vload_source_file_function,
4594 doc: /* Function called in `load' to load an Emacs Lisp source file.
4595 The value should be a function for doing code conversion before
4596 reading a source file. It can also be nil, in which case loading is
4597 done without any code conversion.
4598
4599 If the value is a function, it is called with four arguments,
4600 FULLNAME, FILE, NOERROR, NOMESSAGE. FULLNAME is the absolute name of
4601 the file to load, FILE is the non-absolute name (for messages etc.),
4602 and NOERROR and NOMESSAGE are the corresponding arguments passed to
4603 `load'. The function should return t if the file was loaded. */);
4604 Vload_source_file_function = Qnil;
4605
4606 DEFVAR_BOOL ("load-force-doc-strings", load_force_doc_strings,
4607 doc: /* Non-nil means `load' should force-load all dynamic doc strings.
4608 This is useful when the file being loaded is a temporary copy. */);
4609 load_force_doc_strings = 0;
4610
4611 DEFVAR_BOOL ("load-convert-to-unibyte", load_convert_to_unibyte,
4612 doc: /* Non-nil means `read' converts strings to unibyte whenever possible.
4613 This is normally bound by `load' and `eval-buffer' to control `read',
4614 and is not meant for users to change. */);
4615 load_convert_to_unibyte = 0;
4616
4617 DEFVAR_LISP ("source-directory", Vsource_directory,
4618 doc: /* Directory in which Emacs sources were found when Emacs was built.
4619 You cannot count on them to still be there! */);
4620 Vsource_directory
4621 = Fexpand_file_name (build_string ("../"),
4622 Fcar (decode_env_path (0, PATH_DUMPLOADSEARCH, 0)));
4623
4624 DEFVAR_LISP ("preloaded-file-list", Vpreloaded_file_list,
4625 doc: /* List of files that were preloaded (when dumping Emacs). */);
4626 Vpreloaded_file_list = Qnil;
4627
4628 DEFVAR_LISP ("byte-boolean-vars", Vbyte_boolean_vars,
4629 doc: /* List of all DEFVAR_BOOL variables, used by the byte code optimizer. */);
4630 Vbyte_boolean_vars = Qnil;
4631
4632 DEFVAR_BOOL ("load-dangerous-libraries", load_dangerous_libraries,
4633 doc: /* Non-nil means load dangerous compiled Lisp files.
4634 Some versions of XEmacs use different byte codes than Emacs. These
4635 incompatible byte codes can make Emacs crash when it tries to execute
4636 them. */);
4637 load_dangerous_libraries = 0;
4638
4639 DEFVAR_BOOL ("force-load-messages", force_load_messages,
4640 doc: /* Non-nil means force printing messages when loading Lisp files.
4641 This overrides the value of the NOMESSAGE argument to `load'. */);
4642 force_load_messages = 0;
4643
4644 DEFVAR_LISP ("bytecomp-version-regexp", Vbytecomp_version_regexp,
4645 doc: /* Regular expression matching safe to load compiled Lisp files.
4646 When Emacs loads a compiled Lisp file, it reads the first 512 bytes
4647 from the file, and matches them against this regular expression.
4648 When the regular expression matches, the file is considered to be safe
4649 to load. See also `load-dangerous-libraries'. */);
4650 Vbytecomp_version_regexp
4651 = build_pure_c_string ("^;;;.\\(in Emacs version\\|bytecomp version FSF\\)");
4652
4653 DEFSYM (Qlexical_binding, "lexical-binding");
4654 DEFVAR_LISP ("lexical-binding", Vlexical_binding,
4655 doc: /* Whether to use lexical binding when evaluating code.
4656 Non-nil means that the code in the current buffer should be evaluated
4657 with lexical binding.
4658 This variable is automatically set from the file variables of an
4659 interpreted Lisp file read using `load'. Unlike other file local
4660 variables, this must be set in the first line of a file. */);
4661 Vlexical_binding = Qnil;
4662 Fmake_variable_buffer_local (Qlexical_binding);
4663
4664 DEFVAR_LISP ("eval-buffer-list", Veval_buffer_list,
4665 doc: /* List of buffers being read from by calls to `eval-buffer' and `eval-region'. */);
4666 Veval_buffer_list = Qnil;
4667
4668 DEFVAR_LISP ("old-style-backquotes", Vold_style_backquotes,
4669 doc: /* Set to non-nil when `read' encounters an old-style backquote. */);
4670 Vold_style_backquotes = Qnil;
4671 DEFSYM (Qold_style_backquotes, "old-style-backquotes");
4672
4673 DEFVAR_BOOL ("load-prefer-newer", load_prefer_newer,
4674 doc: /* Non-nil means `load' prefers the newest version of a file.
4675 This applies when a filename suffix is not explicitly specified and
4676 `load' is trying various possible suffixes (see `load-suffixes' and
4677 `load-file-rep-suffixes'). Normally, it stops at the first file
4678 that exists unless you explicitly specify one or the other. If this
4679 option is non-nil, it checks all suffixes and uses whichever file is
4680 newest.
4681 Note that if you customize this, obviously it will not affect files
4682 that are loaded before your customizations are read! */);
4683 load_prefer_newer = 0;
4684
4685 /* Vsource_directory was initialized in init_lread. */
4686
4687 DEFSYM (Qcurrent_load_list, "current-load-list");
4688 DEFSYM (Qstandard_input, "standard-input");
4689 DEFSYM (Qread_char, "read-char");
4690 DEFSYM (Qget_file_char, "get-file-char");
4691
4692 /* Used instead of Qget_file_char while loading *.elc files compiled
4693 by Emacs 21 or older. */
4694 DEFSYM (Qget_emacs_mule_file_char, "get-emacs-mule-file-char");
4695
4696 DEFSYM (Qload_force_doc_strings, "load-force-doc-strings");
4697
4698 DEFSYM (Qbackquote, "`");
4699 DEFSYM (Qcomma, ",");
4700 DEFSYM (Qcomma_at, ",@");
4701 DEFSYM (Qcomma_dot, ",.");
4702
4703 DEFSYM (Qinhibit_file_name_operation, "inhibit-file-name-operation");
4704 DEFSYM (Qascii_character, "ascii-character");
4705 DEFSYM (Qfunction, "function");
4706 DEFSYM (Qload, "load");
4707 DEFSYM (Qload_file_name, "load-file-name");
4708 DEFSYM (Qeval_buffer_list, "eval-buffer-list");
4709 DEFSYM (Qfile_truename, "file-truename");
4710 DEFSYM (Qdir_ok, "dir-ok");
4711 DEFSYM (Qdo_after_load_evaluation, "do-after-load-evaluation");
4712
4713 staticpro (&read_objects);
4714 read_objects = Qnil;
4715 staticpro (&seen_list);
4716 seen_list = Qnil;
4717
4718 Vloads_in_progress = Qnil;
4719 staticpro (&Vloads_in_progress);
4720
4721 DEFSYM (Qhash_table, "hash-table");
4722 DEFSYM (Qdata, "data");
4723 DEFSYM (Qtest, "test");
4724 DEFSYM (Qsize, "size");
4725 DEFSYM (Qweakness, "weakness");
4726 DEFSYM (Qrehash_size, "rehash-size");
4727 DEFSYM (Qrehash_threshold, "rehash-threshold");
4728 }