Require cl-lib at runtime in vc-hg
[emacs.git] / src / lread.c
blob6467043b1daa23833d6c4acbb9ca0df948a5ed14
1 /* Lisp parsing and input streams.
3 Copyright (C) 1985-1989, 1993-1995, 1997-2017 Free Software Foundation,
4 Inc.
6 This file is part of GNU Emacs.
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 (at
11 your option) any later version.
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.
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/>. */
21 /* Tell globals.h to define tables needed by init_obarray. */
22 #define DEFINE_SYMBOLS
24 #include <config.h>
25 #include "sysstdio.h"
26 #include <stdlib.h>
27 #include <sys/types.h>
28 #include <sys/stat.h>
29 #include <sys/file.h>
30 #include <errno.h>
31 #include <math.h>
32 #include <stat-time.h>
33 #include "lisp.h"
34 #include "dispextern.h"
35 #include "intervals.h"
36 #include "character.h"
37 #include "buffer.h"
38 #include "charset.h"
39 #include <epaths.h>
40 #include "commands.h"
41 #include "keyboard.h"
42 #include "systime.h"
43 #include "termhooks.h"
44 #include "blockinput.h"
45 #include <c-ctype.h>
47 #ifdef MSDOS
48 #include "msdos.h"
49 #if __DJGPP__ == 2 && __DJGPP_MINOR__ < 5
50 # define INFINITY __builtin_inf()
51 # define NAN __builtin_nan("")
52 #endif
53 #endif
55 #ifdef HAVE_NS
56 #include "nsterm.h"
57 #endif
59 #include <unistd.h>
61 #ifdef HAVE_SETLOCALE
62 #include <locale.h>
63 #endif /* HAVE_SETLOCALE */
65 #include <fcntl.h>
67 #ifdef HAVE_FSEEKO
68 #define file_offset off_t
69 #define file_tell ftello
70 #else
71 #define file_offset long
72 #define file_tell ftell
73 #endif
75 /* The association list of objects read with the #n=object form.
76 Each member of the list has the form (n . object), and is used to
77 look up the object for the corresponding #n# construct.
78 It must be set to nil before all top-level calls to read0. */
79 static Lisp_Object read_objects;
81 /* File for get_file_char to read from. Use by load. */
82 static FILE *instream;
84 /* For use within read-from-string (this reader is non-reentrant!!) */
85 static ptrdiff_t read_from_string_index;
86 static ptrdiff_t read_from_string_index_byte;
87 static ptrdiff_t read_from_string_limit;
89 /* Number of characters read in the current call to Fread or
90 Fread_from_string. */
91 static EMACS_INT readchar_count;
93 /* This contains the last string skipped with #@. */
94 static char *saved_doc_string;
95 /* Length of buffer allocated in saved_doc_string. */
96 static ptrdiff_t saved_doc_string_size;
97 /* Length of actual data in saved_doc_string. */
98 static ptrdiff_t saved_doc_string_length;
99 /* This is the file position that string came from. */
100 static file_offset saved_doc_string_position;
102 /* This contains the previous string skipped with #@.
103 We copy it from saved_doc_string when a new string
104 is put in saved_doc_string. */
105 static char *prev_saved_doc_string;
106 /* Length of buffer allocated in prev_saved_doc_string. */
107 static ptrdiff_t prev_saved_doc_string_size;
108 /* Length of actual data in prev_saved_doc_string. */
109 static ptrdiff_t prev_saved_doc_string_length;
110 /* This is the file position that string came from. */
111 static file_offset prev_saved_doc_string_position;
113 /* True means inside a new-style backquote
114 with no surrounding parentheses.
115 Fread initializes this to false, so we need not specbind it
116 or worry about what happens to it when there is an error. */
117 static bool new_backquote_flag;
119 /* A list of file names for files being loaded in Fload. Used to
120 check for recursive loads. */
122 static Lisp_Object Vloads_in_progress;
124 static int read_emacs_mule_char (int, int (*) (int, Lisp_Object),
125 Lisp_Object);
127 static void readevalloop (Lisp_Object, FILE *, Lisp_Object, bool,
128 Lisp_Object, Lisp_Object,
129 Lisp_Object, Lisp_Object);
131 /* Functions that read one byte from the current source READCHARFUN
132 or unreads one byte. If the integer argument C is -1, it returns
133 one read byte, or -1 when there's no more byte in the source. If C
134 is 0 or positive, it unreads C, and the return value is not
135 interesting. */
137 static int readbyte_for_lambda (int, Lisp_Object);
138 static int readbyte_from_file (int, Lisp_Object);
139 static int readbyte_from_string (int, Lisp_Object);
141 /* Handle unreading and rereading of characters.
142 Write READCHAR to read a character,
143 UNREAD(c) to unread c to be read again.
145 These macros correctly read/unread multibyte characters. */
147 #define READCHAR readchar (readcharfun, NULL)
148 #define UNREAD(c) unreadchar (readcharfun, c)
150 /* Same as READCHAR but set *MULTIBYTE to the multibyteness of the source. */
151 #define READCHAR_REPORT_MULTIBYTE(multibyte) readchar (readcharfun, multibyte)
153 /* When READCHARFUN is Qget_file_char, Qget_emacs_mule_file_char,
154 Qlambda, or a cons, we use this to keep an unread character because
155 a file stream can't handle multibyte-char unreading. The value -1
156 means that there's no unread character. */
157 static int unread_char;
159 static int
160 readchar (Lisp_Object readcharfun, bool *multibyte)
162 Lisp_Object tem;
163 register int c;
164 int (*readbyte) (int, Lisp_Object);
165 unsigned char buf[MAX_MULTIBYTE_LENGTH];
166 int i, len;
167 bool emacs_mule_encoding = 0;
169 if (multibyte)
170 *multibyte = 0;
172 readchar_count++;
174 if (BUFFERP (readcharfun))
176 register struct buffer *inbuffer = XBUFFER (readcharfun);
178 ptrdiff_t pt_byte = BUF_PT_BYTE (inbuffer);
180 if (! BUFFER_LIVE_P (inbuffer))
181 return -1;
183 if (pt_byte >= BUF_ZV_BYTE (inbuffer))
184 return -1;
186 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
188 /* Fetch the character code from the buffer. */
189 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, pt_byte);
190 BUF_INC_POS (inbuffer, pt_byte);
191 c = STRING_CHAR (p);
192 if (multibyte)
193 *multibyte = 1;
195 else
197 c = BUF_FETCH_BYTE (inbuffer, pt_byte);
198 if (! ASCII_CHAR_P (c))
199 c = BYTE8_TO_CHAR (c);
200 pt_byte++;
202 SET_BUF_PT_BOTH (inbuffer, BUF_PT (inbuffer) + 1, pt_byte);
204 return c;
206 if (MARKERP (readcharfun))
208 register struct buffer *inbuffer = XMARKER (readcharfun)->buffer;
210 ptrdiff_t bytepos = marker_byte_position (readcharfun);
212 if (bytepos >= BUF_ZV_BYTE (inbuffer))
213 return -1;
215 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
217 /* Fetch the character code from the buffer. */
218 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, bytepos);
219 BUF_INC_POS (inbuffer, bytepos);
220 c = STRING_CHAR (p);
221 if (multibyte)
222 *multibyte = 1;
224 else
226 c = BUF_FETCH_BYTE (inbuffer, bytepos);
227 if (! ASCII_CHAR_P (c))
228 c = BYTE8_TO_CHAR (c);
229 bytepos++;
232 XMARKER (readcharfun)->bytepos = bytepos;
233 XMARKER (readcharfun)->charpos++;
235 return c;
238 if (EQ (readcharfun, Qlambda))
240 readbyte = readbyte_for_lambda;
241 goto read_multibyte;
244 if (EQ (readcharfun, Qget_file_char))
246 readbyte = readbyte_from_file;
247 goto read_multibyte;
250 if (STRINGP (readcharfun))
252 if (read_from_string_index >= read_from_string_limit)
253 c = -1;
254 else if (STRING_MULTIBYTE (readcharfun))
256 if (multibyte)
257 *multibyte = 1;
258 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c, readcharfun,
259 read_from_string_index,
260 read_from_string_index_byte);
262 else
264 c = SREF (readcharfun, read_from_string_index_byte);
265 read_from_string_index++;
266 read_from_string_index_byte++;
268 return c;
271 if (CONSP (readcharfun) && STRINGP (XCAR (readcharfun)))
273 /* This is the case that read_vector is reading from a unibyte
274 string that contains a byte sequence previously skipped
275 because of #@NUMBER. The car part of readcharfun is that
276 string, and the cdr part is a value of readcharfun given to
277 read_vector. */
278 readbyte = readbyte_from_string;
279 if (EQ (XCDR (readcharfun), Qget_emacs_mule_file_char))
280 emacs_mule_encoding = 1;
281 goto read_multibyte;
284 if (EQ (readcharfun, Qget_emacs_mule_file_char))
286 readbyte = readbyte_from_file;
287 emacs_mule_encoding = 1;
288 goto read_multibyte;
291 tem = call0 (readcharfun);
293 if (NILP (tem))
294 return -1;
295 return XINT (tem);
297 read_multibyte:
298 if (unread_char >= 0)
300 c = unread_char;
301 unread_char = -1;
302 return c;
304 c = (*readbyte) (-1, readcharfun);
305 if (c < 0)
306 return c;
307 if (multibyte)
308 *multibyte = 1;
309 if (ASCII_CHAR_P (c))
310 return c;
311 if (emacs_mule_encoding)
312 return read_emacs_mule_char (c, readbyte, readcharfun);
313 i = 0;
314 buf[i++] = c;
315 len = BYTES_BY_CHAR_HEAD (c);
316 while (i < len)
318 c = (*readbyte) (-1, readcharfun);
319 if (c < 0 || ! TRAILING_CODE_P (c))
321 while (--i > 1)
322 (*readbyte) (buf[i], readcharfun);
323 return BYTE8_TO_CHAR (buf[0]);
325 buf[i++] = c;
327 return STRING_CHAR (buf);
330 #define FROM_FILE_P(readcharfun) \
331 (EQ (readcharfun, Qget_file_char) \
332 || EQ (readcharfun, Qget_emacs_mule_file_char))
334 static void
335 skip_dyn_bytes (Lisp_Object readcharfun, ptrdiff_t n)
337 if (FROM_FILE_P (readcharfun))
339 block_input (); /* FIXME: Not sure if it's needed. */
340 fseek (instream, n, SEEK_CUR);
341 unblock_input ();
343 else
344 { /* We're not reading directly from a file. In that case, it's difficult
345 to reliably count bytes, since these are usually meant for the file's
346 encoding, whereas we're now typically in the internal encoding.
347 But luckily, skip_dyn_bytes is used to skip over a single
348 dynamic-docstring (or dynamic byte-code) which is always quoted such
349 that \037 is the final char. */
350 int c;
351 do {
352 c = READCHAR;
353 } while (c >= 0 && c != '\037');
357 static void
358 skip_dyn_eof (Lisp_Object readcharfun)
360 if (FROM_FILE_P (readcharfun))
362 block_input (); /* FIXME: Not sure if it's needed. */
363 fseek (instream, 0, SEEK_END);
364 unblock_input ();
366 else
367 while (READCHAR >= 0);
370 /* Unread the character C in the way appropriate for the stream READCHARFUN.
371 If the stream is a user function, call it with the char as argument. */
373 static void
374 unreadchar (Lisp_Object readcharfun, int c)
376 readchar_count--;
377 if (c == -1)
378 /* Don't back up the pointer if we're unreading the end-of-input mark,
379 since readchar didn't advance it when we read it. */
381 else if (BUFFERP (readcharfun))
383 struct buffer *b = XBUFFER (readcharfun);
384 ptrdiff_t charpos = BUF_PT (b);
385 ptrdiff_t bytepos = BUF_PT_BYTE (b);
387 if (! NILP (BVAR (b, enable_multibyte_characters)))
388 BUF_DEC_POS (b, bytepos);
389 else
390 bytepos--;
392 SET_BUF_PT_BOTH (b, charpos - 1, bytepos);
394 else if (MARKERP (readcharfun))
396 struct buffer *b = XMARKER (readcharfun)->buffer;
397 ptrdiff_t bytepos = XMARKER (readcharfun)->bytepos;
399 XMARKER (readcharfun)->charpos--;
400 if (! NILP (BVAR (b, enable_multibyte_characters)))
401 BUF_DEC_POS (b, bytepos);
402 else
403 bytepos--;
405 XMARKER (readcharfun)->bytepos = bytepos;
407 else if (STRINGP (readcharfun))
409 read_from_string_index--;
410 read_from_string_index_byte
411 = string_char_to_byte (readcharfun, read_from_string_index);
413 else if (CONSP (readcharfun) && STRINGP (XCAR (readcharfun)))
415 unread_char = c;
417 else if (EQ (readcharfun, Qlambda))
419 unread_char = c;
421 else if (FROM_FILE_P (readcharfun))
423 unread_char = c;
425 else
426 call1 (readcharfun, make_number (c));
429 static int
430 readbyte_for_lambda (int c, Lisp_Object readcharfun)
432 return read_bytecode_char (c >= 0);
436 static int
437 readbyte_from_file (int c, Lisp_Object readcharfun)
439 if (c >= 0)
441 block_input ();
442 ungetc (c, instream);
443 unblock_input ();
444 return 0;
447 block_input ();
448 c = getc (instream);
450 /* Interrupted reads have been observed while reading over the network. */
451 while (c == EOF && ferror (instream) && errno == EINTR)
453 unblock_input ();
454 maybe_quit ();
455 block_input ();
456 clearerr (instream);
457 c = getc (instream);
460 unblock_input ();
462 return (c == EOF ? -1 : c);
465 static int
466 readbyte_from_string (int c, Lisp_Object readcharfun)
468 Lisp_Object string = XCAR (readcharfun);
470 if (c >= 0)
472 read_from_string_index--;
473 read_from_string_index_byte
474 = string_char_to_byte (string, read_from_string_index);
477 if (read_from_string_index >= read_from_string_limit)
478 c = -1;
479 else
480 FETCH_STRING_CHAR_ADVANCE (c, string,
481 read_from_string_index,
482 read_from_string_index_byte);
483 return c;
487 /* Read one non-ASCII character from INSTREAM. The character is
488 encoded in `emacs-mule' and the first byte is already read in
489 C. */
491 static int
492 read_emacs_mule_char (int c, int (*readbyte) (int, Lisp_Object), Lisp_Object readcharfun)
494 /* Emacs-mule coding uses at most 4-byte for one character. */
495 unsigned char buf[4];
496 int len = emacs_mule_bytes[c];
497 struct charset *charset;
498 int i;
499 unsigned code;
501 if (len == 1)
502 /* C is not a valid leading-code of `emacs-mule'. */
503 return BYTE8_TO_CHAR (c);
505 i = 0;
506 buf[i++] = c;
507 while (i < len)
509 c = (*readbyte) (-1, readcharfun);
510 if (c < 0xA0)
512 while (--i > 1)
513 (*readbyte) (buf[i], readcharfun);
514 return BYTE8_TO_CHAR (buf[0]);
516 buf[i++] = c;
519 if (len == 2)
521 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
522 code = buf[1] & 0x7F;
524 else if (len == 3)
526 if (buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_11
527 || buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_12)
529 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
530 code = buf[2] & 0x7F;
532 else
534 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
535 code = ((buf[1] << 8) | buf[2]) & 0x7F7F;
538 else
540 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
541 code = ((buf[2] << 8) | buf[3]) & 0x7F7F;
543 c = DECODE_CHAR (charset, code);
544 if (c < 0)
545 Fsignal (Qinvalid_read_syntax,
546 list1 (build_string ("invalid multibyte form")));
547 return c;
551 static Lisp_Object read_internal_start (Lisp_Object, Lisp_Object,
552 Lisp_Object);
553 static Lisp_Object read0 (Lisp_Object);
554 static Lisp_Object read1 (Lisp_Object, int *, bool);
556 static Lisp_Object read_list (bool, Lisp_Object);
557 static Lisp_Object read_vector (Lisp_Object, bool);
559 static Lisp_Object substitute_object_recurse (Lisp_Object, Lisp_Object,
560 Lisp_Object);
561 static void substitute_in_interval (INTERVAL, Lisp_Object);
564 /* Get a character from the tty. */
566 /* Read input events until we get one that's acceptable for our purposes.
568 If NO_SWITCH_FRAME, switch-frame events are stashed
569 until we get a character we like, and then stuffed into
570 unread_switch_frame.
572 If ASCII_REQUIRED, check function key events to see
573 if the unmodified version of the symbol has a Qascii_character
574 property, and use that character, if present.
576 If ERROR_NONASCII, signal an error if the input we
577 get isn't an ASCII character with modifiers. If it's false but
578 ASCII_REQUIRED is true, just re-read until we get an ASCII
579 character.
581 If INPUT_METHOD, invoke the current input method
582 if the character warrants that.
584 If SECONDS is a number, wait that many seconds for input, and
585 return Qnil if no input arrives within that time. */
587 static Lisp_Object
588 read_filtered_event (bool no_switch_frame, bool ascii_required,
589 bool error_nonascii, bool input_method, Lisp_Object seconds)
591 Lisp_Object val, delayed_switch_frame;
592 struct timespec end_time;
594 #ifdef HAVE_WINDOW_SYSTEM
595 if (display_hourglass_p)
596 cancel_hourglass ();
597 #endif
599 delayed_switch_frame = Qnil;
601 /* Compute timeout. */
602 if (NUMBERP (seconds))
604 double duration = XFLOATINT (seconds);
605 struct timespec wait_time = dtotimespec (duration);
606 end_time = timespec_add (current_timespec (), wait_time);
609 /* Read until we get an acceptable event. */
610 retry:
612 val = read_char (0, Qnil, (input_method ? Qnil : Qt), 0,
613 NUMBERP (seconds) ? &end_time : NULL);
614 while (INTEGERP (val) && XINT (val) == -2); /* wrong_kboard_jmpbuf */
616 if (BUFFERP (val))
617 goto retry;
619 /* `switch-frame' events are put off until after the next ASCII
620 character. This is better than signaling an error just because
621 the last characters were typed to a separate minibuffer frame,
622 for example. Eventually, some code which can deal with
623 switch-frame events will read it and process it. */
624 if (no_switch_frame
625 && EVENT_HAS_PARAMETERS (val)
626 && EQ (EVENT_HEAD_KIND (EVENT_HEAD (val)), Qswitch_frame))
628 delayed_switch_frame = val;
629 goto retry;
632 if (ascii_required && !(NUMBERP (seconds) && NILP (val)))
634 /* Convert certain symbols to their ASCII equivalents. */
635 if (SYMBOLP (val))
637 Lisp_Object tem, tem1;
638 tem = Fget (val, Qevent_symbol_element_mask);
639 if (!NILP (tem))
641 tem1 = Fget (Fcar (tem), Qascii_character);
642 /* Merge this symbol's modifier bits
643 with the ASCII equivalent of its basic code. */
644 if (!NILP (tem1))
645 XSETFASTINT (val, XINT (tem1) | XINT (Fcar (Fcdr (tem))));
649 /* If we don't have a character now, deal with it appropriately. */
650 if (!INTEGERP (val))
652 if (error_nonascii)
654 Vunread_command_events = list1 (val);
655 error ("Non-character input-event");
657 else
658 goto retry;
662 if (! NILP (delayed_switch_frame))
663 unread_switch_frame = delayed_switch_frame;
665 #if 0
667 #ifdef HAVE_WINDOW_SYSTEM
668 if (display_hourglass_p)
669 start_hourglass ();
670 #endif
672 #endif
674 return val;
677 DEFUN ("read-char", Fread_char, Sread_char, 0, 3, 0,
678 doc: /* Read a character from the command input (keyboard or macro).
679 It is returned as a number.
680 If the character has modifiers, they are resolved and reflected to the
681 character code if possible (e.g. C-SPC -> 0).
683 If the user generates an event which is not a character (i.e. a mouse
684 click or function key event), `read-char' signals an error. As an
685 exception, switch-frame events are put off until non-character events
686 can be read.
687 If you want to read non-character events, or ignore them, call
688 `read-event' or `read-char-exclusive' instead.
690 If the optional argument PROMPT is non-nil, display that as a prompt.
691 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
692 input method is turned on in the current buffer, that input method
693 is used for reading a character.
694 If the optional argument SECONDS is non-nil, it should be a number
695 specifying the maximum number of seconds to wait for input. If no
696 input arrives in that time, return nil. SECONDS may be a
697 floating-point value. */)
698 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
700 Lisp_Object val;
702 if (! NILP (prompt))
703 message_with_string ("%s", prompt, 0);
704 val = read_filtered_event (1, 1, 1, ! NILP (inherit_input_method), seconds);
706 return (NILP (val) ? Qnil
707 : make_number (char_resolve_modifier_mask (XINT (val))));
710 DEFUN ("read-event", Fread_event, Sread_event, 0, 3, 0,
711 doc: /* Read an event object from the input stream.
712 If the optional argument PROMPT is non-nil, display that as a prompt.
713 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
714 input method is turned on in the current buffer, that input method
715 is used for reading a character.
716 If the optional argument SECONDS is non-nil, it should be a number
717 specifying the maximum number of seconds to wait for input. If no
718 input arrives in that time, return nil. SECONDS may be a
719 floating-point value. */)
720 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
722 if (! NILP (prompt))
723 message_with_string ("%s", prompt, 0);
724 return read_filtered_event (0, 0, 0, ! NILP (inherit_input_method), seconds);
727 DEFUN ("read-char-exclusive", Fread_char_exclusive, Sread_char_exclusive, 0, 3, 0,
728 doc: /* Read a character from the command input (keyboard or macro).
729 It is returned as a number. Non-character events are ignored.
730 If the character has modifiers, they are resolved and reflected to the
731 character code if possible (e.g. C-SPC -> 0).
733 If the optional argument PROMPT is non-nil, display that as a prompt.
734 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
735 input method is turned on in the current buffer, that input method
736 is used for reading a character.
737 If the optional argument SECONDS is non-nil, it should be a number
738 specifying the maximum number of seconds to wait for input. If no
739 input arrives in that time, return nil. SECONDS may be a
740 floating-point value. */)
741 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
743 Lisp_Object val;
745 if (! NILP (prompt))
746 message_with_string ("%s", prompt, 0);
748 val = read_filtered_event (1, 1, 0, ! NILP (inherit_input_method), seconds);
750 return (NILP (val) ? Qnil
751 : make_number (char_resolve_modifier_mask (XINT (val))));
754 DEFUN ("get-file-char", Fget_file_char, Sget_file_char, 0, 0, 0,
755 doc: /* Don't use this yourself. */)
756 (void)
758 register Lisp_Object val;
759 block_input ();
760 XSETINT (val, getc (instream));
761 unblock_input ();
762 return val;
768 /* Return true if the lisp code read using READCHARFUN defines a non-nil
769 `lexical-binding' file variable. After returning, the stream is
770 positioned following the first line, if it is a comment or #! line,
771 otherwise nothing is read. */
773 static bool
774 lisp_file_lexically_bound_p (Lisp_Object readcharfun)
776 int ch = READCHAR;
778 if (ch == '#')
780 ch = READCHAR;
781 if (ch != '!')
783 UNREAD (ch);
784 UNREAD ('#');
785 return 0;
787 while (ch != '\n' && ch != EOF)
788 ch = READCHAR;
789 if (ch == '\n') ch = READCHAR;
790 /* It is OK to leave the position after a #! line, since
791 that is what read1 does. */
794 if (ch != ';')
795 /* The first line isn't a comment, just give up. */
797 UNREAD (ch);
798 return 0;
800 else
801 /* Look for an appropriate file-variable in the first line. */
803 bool rv = 0;
804 enum {
805 NOMINAL, AFTER_FIRST_DASH, AFTER_ASTERIX
806 } beg_end_state = NOMINAL;
807 bool in_file_vars = 0;
809 #define UPDATE_BEG_END_STATE(ch) \
810 if (beg_end_state == NOMINAL) \
811 beg_end_state = (ch == '-' ? AFTER_FIRST_DASH : NOMINAL); \
812 else if (beg_end_state == AFTER_FIRST_DASH) \
813 beg_end_state = (ch == '*' ? AFTER_ASTERIX : NOMINAL); \
814 else if (beg_end_state == AFTER_ASTERIX) \
816 if (ch == '-') \
817 in_file_vars = !in_file_vars; \
818 beg_end_state = NOMINAL; \
821 /* Skip until we get to the file vars, if any. */
824 ch = READCHAR;
825 UPDATE_BEG_END_STATE (ch);
827 while (!in_file_vars && ch != '\n' && ch != EOF);
829 while (in_file_vars)
831 char var[100], val[100];
832 unsigned i;
834 ch = READCHAR;
836 /* Read a variable name. */
837 while (ch == ' ' || ch == '\t')
838 ch = READCHAR;
840 i = 0;
841 while (ch != ':' && ch != '\n' && ch != EOF && in_file_vars)
843 if (i < sizeof var - 1)
844 var[i++] = ch;
845 UPDATE_BEG_END_STATE (ch);
846 ch = READCHAR;
849 /* Stop scanning if no colon was found before end marker. */
850 if (!in_file_vars || ch == '\n' || ch == EOF)
851 break;
853 while (i > 0 && (var[i - 1] == ' ' || var[i - 1] == '\t'))
854 i--;
855 var[i] = '\0';
857 if (ch == ':')
859 /* Read a variable value. */
860 ch = READCHAR;
862 while (ch == ' ' || ch == '\t')
863 ch = READCHAR;
865 i = 0;
866 while (ch != ';' && ch != '\n' && ch != EOF && in_file_vars)
868 if (i < sizeof val - 1)
869 val[i++] = ch;
870 UPDATE_BEG_END_STATE (ch);
871 ch = READCHAR;
873 if (! in_file_vars)
874 /* The value was terminated by an end-marker, which remove. */
875 i -= 3;
876 while (i > 0 && (val[i - 1] == ' ' || val[i - 1] == '\t'))
877 i--;
878 val[i] = '\0';
880 if (strcmp (var, "lexical-binding") == 0)
881 /* This is it... */
883 rv = (strcmp (val, "nil") != 0);
884 break;
889 while (ch != '\n' && ch != EOF)
890 ch = READCHAR;
892 return rv;
896 /* Value is a version number of byte compiled code if the file
897 associated with file descriptor FD is a compiled Lisp file that's
898 safe to load. Only files compiled with Emacs are safe to load.
899 Files compiled with XEmacs can lead to a crash in Fbyte_code
900 because of an incompatible change in the byte compiler. */
902 static int
903 safe_to_load_version (int fd)
905 char buf[512];
906 int nbytes, i;
907 int version = 1;
909 /* Read the first few bytes from the file, and look for a line
910 specifying the byte compiler version used. */
911 nbytes = emacs_read_quit (fd, buf, sizeof buf);
912 if (nbytes > 0)
914 /* Skip to the next newline, skipping over the initial `ELC'
915 with NUL bytes following it, but note the version. */
916 for (i = 0; i < nbytes && buf[i] != '\n'; ++i)
917 if (i == 4)
918 version = buf[i];
920 if (i >= nbytes
921 || fast_c_string_match_ignore_case (Vbytecomp_version_regexp,
922 buf + i, nbytes - i) < 0)
923 version = 0;
926 lseek (fd, 0, SEEK_SET);
927 return version;
931 /* Callback for record_unwind_protect. Restore the old load list OLD,
932 after loading a file successfully. */
934 static void
935 record_load_unwind (Lisp_Object old)
937 Vloads_in_progress = old;
940 /* This handler function is used via internal_condition_case_1. */
942 static Lisp_Object
943 load_error_handler (Lisp_Object data)
945 return Qnil;
948 static void
949 load_warn_old_style_backquotes (Lisp_Object file)
951 if (!NILP (Vold_style_backquotes))
953 AUTO_STRING (format, "Loading `%s': old-style backquotes detected!");
954 CALLN (Fmessage, format, file);
958 static void
959 load_warn_unescaped_character_literals (Lisp_Object file)
961 if (NILP (Vlread_unescaped_character_literals)) return;
962 CHECK_CONS (Vlread_unescaped_character_literals);
963 AUTO_STRING (format,
964 "Loading `%s': unescaped character literals %s detected!");
965 AUTO_STRING (separator, ", ");
966 CALLN (Fmessage,
967 format, file,
968 Fmapconcat (Qstring,
969 Fsort (Vlread_unescaped_character_literals, Qlss),
970 separator));
973 DEFUN ("get-load-suffixes", Fget_load_suffixes, Sget_load_suffixes, 0, 0, 0,
974 doc: /* Return the suffixes that `load' should try if a suffix is \
975 required.
976 This uses the variables `load-suffixes' and `load-file-rep-suffixes'. */)
977 (void)
979 Lisp_Object lst = Qnil, suffixes = Vload_suffixes, suffix, ext;
980 while (CONSP (suffixes))
982 Lisp_Object exts = Vload_file_rep_suffixes;
983 suffix = XCAR (suffixes);
984 suffixes = XCDR (suffixes);
985 while (CONSP (exts))
987 ext = XCAR (exts);
988 exts = XCDR (exts);
989 lst = Fcons (concat2 (suffix, ext), lst);
992 return Fnreverse (lst);
995 /* Returns true if STRING ends with SUFFIX */
996 static bool
997 suffix_p (Lisp_Object string, const char *suffix)
999 ptrdiff_t suffix_len = strlen (suffix);
1000 ptrdiff_t string_len = SBYTES (string);
1002 return string_len >= suffix_len && !strcmp (SSDATA (string) + string_len - suffix_len, suffix);
1005 DEFUN ("load", Fload, Sload, 1, 5, 0,
1006 doc: /* Execute a file of Lisp code named FILE.
1007 First try FILE with `.elc' appended, then try with `.el', then try
1008 with a system-dependent suffix of dynamic modules (see `load-suffixes'),
1009 then try FILE unmodified (the exact suffixes in the exact order are
1010 determined by `load-suffixes'). Environment variable references in
1011 FILE are replaced with their values by calling `substitute-in-file-name'.
1012 This function searches the directories in `load-path'.
1014 If optional second arg NOERROR is non-nil,
1015 report no error if FILE doesn't exist.
1016 Print messages at start and end of loading unless
1017 optional third arg NOMESSAGE is non-nil (but `force-load-messages'
1018 overrides that).
1019 If optional fourth arg NOSUFFIX is non-nil, don't try adding
1020 suffixes to the specified name FILE.
1021 If optional fifth arg MUST-SUFFIX is non-nil, insist on
1022 the suffix `.elc' or `.el' or the module suffix; don't accept just
1023 FILE unless it ends in one of those suffixes or includes a directory name.
1025 If NOSUFFIX is nil, then if a file could not be found, try looking for
1026 a different representation of the file by adding non-empty suffixes to
1027 its name, before trying another file. Emacs uses this feature to find
1028 compressed versions of files when Auto Compression mode is enabled.
1029 If NOSUFFIX is non-nil, disable this feature.
1031 The suffixes that this function tries out, when NOSUFFIX is nil, are
1032 given by the return value of `get-load-suffixes' and the values listed
1033 in `load-file-rep-suffixes'. If MUST-SUFFIX is non-nil, only the
1034 return value of `get-load-suffixes' is used, i.e. the file name is
1035 required to have a non-empty suffix.
1037 When searching suffixes, this function normally stops at the first
1038 one that exists. If the option `load-prefer-newer' is non-nil,
1039 however, it tries all suffixes, and uses whichever file is the newest.
1041 Loading a file records its definitions, and its `provide' and
1042 `require' calls, in an element of `load-history' whose
1043 car is the file name loaded. See `load-history'.
1045 While the file is in the process of being loaded, the variable
1046 `load-in-progress' is non-nil and the variable `load-file-name'
1047 is bound to the file's name.
1049 Return t if the file exists and loads successfully. */)
1050 (Lisp_Object file, Lisp_Object noerror, Lisp_Object nomessage,
1051 Lisp_Object nosuffix, Lisp_Object must_suffix)
1053 FILE *stream;
1054 int fd;
1055 int fd_index UNINIT;
1056 ptrdiff_t count = SPECPDL_INDEX ();
1057 Lisp_Object found, efound, hist_file_name;
1058 /* True means we printed the ".el is newer" message. */
1059 bool newer = 0;
1060 /* True means we are loading a compiled file. */
1061 bool compiled = 0;
1062 Lisp_Object handler;
1063 bool safe_p = 1;
1064 const char *fmode = "r" FOPEN_TEXT;
1065 int version;
1067 CHECK_STRING (file);
1069 /* If file name is magic, call the handler. */
1070 /* This shouldn't be necessary any more now that `openp' handles it right.
1071 handler = Ffind_file_name_handler (file, Qload);
1072 if (!NILP (handler))
1073 return call5 (handler, Qload, file, noerror, nomessage, nosuffix); */
1075 /* The presence of this call is the result of a historical accident:
1076 it used to be in every file-operation and when it got removed
1077 everywhere, it accidentally stayed here. Since then, enough people
1078 supposedly have things like (load "$PROJECT/foo.el") in their .emacs
1079 that it seemed risky to remove. */
1080 if (! NILP (noerror))
1082 file = internal_condition_case_1 (Fsubstitute_in_file_name, file,
1083 Qt, load_error_handler);
1084 if (NILP (file))
1085 return Qnil;
1087 else
1088 file = Fsubstitute_in_file_name (file);
1090 /* Avoid weird lossage with null string as arg,
1091 since it would try to load a directory as a Lisp file. */
1092 if (SCHARS (file) == 0)
1094 fd = -1;
1095 errno = ENOENT;
1097 else
1099 Lisp_Object suffixes;
1100 found = Qnil;
1102 if (! NILP (must_suffix))
1104 /* Don't insist on adding a suffix if FILE already ends with one. */
1105 if (suffix_p (file, ".el")
1106 || suffix_p (file, ".elc")
1107 #ifdef HAVE_MODULES
1108 || suffix_p (file, MODULES_SUFFIX)
1109 #endif
1111 must_suffix = Qnil;
1112 /* Don't insist on adding a suffix
1113 if the argument includes a directory name. */
1114 else if (! NILP (Ffile_name_directory (file)))
1115 must_suffix = Qnil;
1118 if (!NILP (nosuffix))
1119 suffixes = Qnil;
1120 else
1122 suffixes = Fget_load_suffixes ();
1123 if (NILP (must_suffix))
1124 suffixes = CALLN (Fappend, suffixes, Vload_file_rep_suffixes);
1127 fd = openp (Vload_path, file, suffixes, &found, Qnil, load_prefer_newer);
1130 if (fd == -1)
1132 if (NILP (noerror))
1133 report_file_error ("Cannot open load file", file);
1134 return Qnil;
1137 /* Tell startup.el whether or not we found the user's init file. */
1138 if (EQ (Qt, Vuser_init_file))
1139 Vuser_init_file = found;
1141 /* If FD is -2, that means openp found a magic file. */
1142 if (fd == -2)
1144 if (NILP (Fequal (found, file)))
1145 /* If FOUND is a different file name from FILE,
1146 find its handler even if we have already inhibited
1147 the `load' operation on FILE. */
1148 handler = Ffind_file_name_handler (found, Qt);
1149 else
1150 handler = Ffind_file_name_handler (found, Qload);
1151 if (! NILP (handler))
1152 return call5 (handler, Qload, found, noerror, nomessage, Qt);
1153 #ifdef DOS_NT
1154 /* Tramp has to deal with semi-broken packages that prepend
1155 drive letters to remote files. For that reason, Tramp
1156 catches file operations that test for file existence, which
1157 makes openp think X:/foo.elc files are remote. However,
1158 Tramp does not catch `load' operations for such files, so we
1159 end up with a nil as the `load' handler above. If we would
1160 continue with fd = -2, we will behave wrongly, and in
1161 particular try reading a .elc file in the "rt" mode instead
1162 of "rb". See bug #9311 for the results. To work around
1163 this, we try to open the file locally, and go with that if it
1164 succeeds. */
1165 fd = emacs_open (SSDATA (ENCODE_FILE (found)), O_RDONLY, 0);
1166 if (fd == -1)
1167 fd = -2;
1168 #endif
1171 if (0 <= fd)
1173 fd_index = SPECPDL_INDEX ();
1174 record_unwind_protect_int (close_file_unwind, fd);
1177 #ifdef HAVE_MODULES
1178 if (suffix_p (found, MODULES_SUFFIX))
1179 return unbind_to (count, Fmodule_load (found));
1180 #endif
1182 /* Check if we're stuck in a recursive load cycle.
1184 2000-09-21: It's not possible to just check for the file loaded
1185 being a member of Vloads_in_progress. This fails because of the
1186 way the byte compiler currently works; `provide's are not
1187 evaluated, see font-lock.el/jit-lock.el as an example. This
1188 leads to a certain amount of ``normal'' recursion.
1190 Also, just loading a file recursively is not always an error in
1191 the general case; the second load may do something different. */
1193 int load_count = 0;
1194 Lisp_Object tem;
1195 for (tem = Vloads_in_progress; CONSP (tem); tem = XCDR (tem))
1196 if (!NILP (Fequal (found, XCAR (tem))) && (++load_count > 3))
1197 signal_error ("Recursive load", Fcons (found, Vloads_in_progress));
1198 record_unwind_protect (record_load_unwind, Vloads_in_progress);
1199 Vloads_in_progress = Fcons (found, Vloads_in_progress);
1202 /* All loads are by default dynamic, unless the file itself specifies
1203 otherwise using a file-variable in the first line. This is bound here
1204 so that it takes effect whether or not we use
1205 Vload_source_file_function. */
1206 specbind (Qlexical_binding, Qnil);
1208 /* Get the name for load-history. */
1209 hist_file_name = (! NILP (Vpurify_flag)
1210 ? concat2 (Ffile_name_directory (file),
1211 Ffile_name_nondirectory (found))
1212 : found) ;
1214 version = -1;
1216 /* Check for the presence of old-style quotes and warn about them. */
1217 specbind (Qold_style_backquotes, Qnil);
1218 record_unwind_protect (load_warn_old_style_backquotes, file);
1220 /* Check for the presence of unescaped character literals and warn
1221 about them. */
1222 specbind (Qlread_unescaped_character_literals, Qnil);
1223 record_unwind_protect (load_warn_unescaped_character_literals, file);
1225 int is_elc;
1226 if ((is_elc = suffix_p (found, ".elc")) != 0
1227 /* version = 1 means the file is empty, in which case we can
1228 treat it as not byte-compiled. */
1229 || (fd >= 0 && (version = safe_to_load_version (fd)) > 1))
1230 /* Load .elc files directly, but not when they are
1231 remote and have no handler! */
1233 if (fd != -2)
1235 struct stat s1, s2;
1236 int result;
1238 if (version < 0
1239 && ! (version = safe_to_load_version (fd)))
1241 safe_p = 0;
1242 if (!load_dangerous_libraries)
1243 error ("File `%s' was not compiled in Emacs", SDATA (found));
1244 else if (!NILP (nomessage) && !force_load_messages)
1245 message_with_string ("File `%s' not compiled in Emacs", found, 1);
1248 compiled = 1;
1250 efound = ENCODE_FILE (found);
1251 fmode = "r" FOPEN_BINARY;
1253 /* openp already checked for newness, no point doing it again.
1254 FIXME would be nice to get a message when openp
1255 ignores suffix order due to load_prefer_newer. */
1256 if (!load_prefer_newer && is_elc)
1258 result = stat (SSDATA (efound), &s1);
1259 if (result == 0)
1261 SSET (efound, SBYTES (efound) - 1, 0);
1262 result = stat (SSDATA (efound), &s2);
1263 SSET (efound, SBYTES (efound) - 1, 'c');
1266 if (result == 0
1267 && timespec_cmp (get_stat_mtime (&s1), get_stat_mtime (&s2)) < 0)
1269 /* Make the progress messages mention that source is newer. */
1270 newer = 1;
1272 /* If we won't print another message, mention this anyway. */
1273 if (!NILP (nomessage) && !force_load_messages)
1275 Lisp_Object msg_file;
1276 msg_file = Fsubstring (found, make_number (0), make_number (-1));
1277 message_with_string ("Source file `%s' newer than byte-compiled file",
1278 msg_file, 1);
1281 } /* !load_prefer_newer */
1284 else
1286 /* We are loading a source file (*.el). */
1287 if (!NILP (Vload_source_file_function))
1289 Lisp_Object val;
1291 if (fd >= 0)
1293 emacs_close (fd);
1294 clear_unwind_protect (fd_index);
1296 val = call4 (Vload_source_file_function, found, hist_file_name,
1297 NILP (noerror) ? Qnil : Qt,
1298 (NILP (nomessage) || force_load_messages) ? Qnil : Qt);
1299 return unbind_to (count, val);
1303 if (fd < 0)
1305 /* We somehow got here with fd == -2, meaning the file is deemed
1306 to be remote. Don't even try to reopen the file locally;
1307 just force a failure. */
1308 stream = NULL;
1309 errno = EINVAL;
1311 else
1313 #ifdef WINDOWSNT
1314 emacs_close (fd);
1315 clear_unwind_protect (fd_index);
1316 efound = ENCODE_FILE (found);
1317 stream = emacs_fopen (SSDATA (efound), fmode);
1318 #else
1319 stream = fdopen (fd, fmode);
1320 #endif
1322 if (! stream)
1323 report_file_error ("Opening stdio stream", file);
1324 set_unwind_protect_ptr (fd_index, fclose_unwind, stream);
1326 if (! NILP (Vpurify_flag))
1327 Vpreloaded_file_list = Fcons (Fpurecopy (file), Vpreloaded_file_list);
1329 if (NILP (nomessage) || force_load_messages)
1331 if (!safe_p)
1332 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...",
1333 file, 1);
1334 else if (!compiled)
1335 message_with_string ("Loading %s (source)...", file, 1);
1336 else if (newer)
1337 message_with_string ("Loading %s (compiled; note, source file is newer)...",
1338 file, 1);
1339 else /* The typical case; compiled file newer than source file. */
1340 message_with_string ("Loading %s...", file, 1);
1343 specbind (Qload_file_name, found);
1344 specbind (Qinhibit_file_name_operation, Qnil);
1345 specbind (Qload_in_progress, Qt);
1347 instream = stream;
1348 if (lisp_file_lexically_bound_p (Qget_file_char))
1349 Fset (Qlexical_binding, Qt);
1351 if (! version || version >= 22)
1352 readevalloop (Qget_file_char, stream, hist_file_name,
1353 0, Qnil, Qnil, Qnil, Qnil);
1354 else
1356 /* We can't handle a file which was compiled with
1357 byte-compile-dynamic by older version of Emacs. */
1358 specbind (Qload_force_doc_strings, Qt);
1359 readevalloop (Qget_emacs_mule_file_char, stream, hist_file_name,
1360 0, Qnil, Qnil, Qnil, Qnil);
1362 unbind_to (count, Qnil);
1364 /* Run any eval-after-load forms for this file. */
1365 if (!NILP (Ffboundp (Qdo_after_load_evaluation)))
1366 call1 (Qdo_after_load_evaluation, hist_file_name) ;
1368 xfree (saved_doc_string);
1369 saved_doc_string = 0;
1370 saved_doc_string_size = 0;
1372 xfree (prev_saved_doc_string);
1373 prev_saved_doc_string = 0;
1374 prev_saved_doc_string_size = 0;
1376 if (!noninteractive && (NILP (nomessage) || force_load_messages))
1378 if (!safe_p)
1379 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...done",
1380 file, 1);
1381 else if (!compiled)
1382 message_with_string ("Loading %s (source)...done", file, 1);
1383 else if (newer)
1384 message_with_string ("Loading %s (compiled; note, source file is newer)...done",
1385 file, 1);
1386 else /* The typical case; compiled file newer than source file. */
1387 message_with_string ("Loading %s...done", file, 1);
1390 return Qt;
1393 static bool
1394 complete_filename_p (Lisp_Object pathname)
1396 const unsigned char *s = SDATA (pathname);
1397 return (IS_DIRECTORY_SEP (s[0])
1398 || (SCHARS (pathname) > 2
1399 && IS_DEVICE_SEP (s[1]) && IS_DIRECTORY_SEP (s[2])));
1402 DEFUN ("locate-file-internal", Flocate_file_internal, Slocate_file_internal, 2, 4, 0,
1403 doc: /* Search for FILENAME through PATH.
1404 Returns the file's name in absolute form, or nil if not found.
1405 If SUFFIXES is non-nil, it should be a list of suffixes to append to
1406 file name when searching.
1407 If non-nil, PREDICATE is used instead of `file-readable-p'.
1408 PREDICATE can also be an integer to pass to the faccessat(2) function,
1409 in which case file-name-handlers are ignored.
1410 This function will normally skip directories, so if you want it to find
1411 directories, make sure the PREDICATE function returns `dir-ok' for them. */)
1412 (Lisp_Object filename, Lisp_Object path, Lisp_Object suffixes, Lisp_Object predicate)
1414 Lisp_Object file;
1415 int fd = openp (path, filename, suffixes, &file, predicate, false);
1416 if (NILP (predicate) && fd >= 0)
1417 emacs_close (fd);
1418 return file;
1421 /* Search for a file whose name is STR, looking in directories
1422 in the Lisp list PATH, and trying suffixes from SUFFIX.
1423 On success, return a file descriptor (or 1 or -2 as described below).
1424 On failure, return -1 and set errno.
1426 SUFFIXES is a list of strings containing possible suffixes.
1427 The empty suffix is automatically added if the list is empty.
1429 PREDICATE t means the files are binary.
1430 PREDICATE non-nil and non-t means don't open the files,
1431 just look for one that satisfies the predicate. In this case,
1432 return -2 on success. The predicate can be a lisp function or
1433 an integer to pass to `access' (in which case file-name-handlers
1434 are ignored).
1436 If STOREPTR is nonzero, it points to a slot where the name of
1437 the file actually found should be stored as a Lisp string.
1438 nil is stored there on failure.
1440 If the file we find is remote, return -2
1441 but store the found remote file name in *STOREPTR.
1443 If NEWER is true, try all SUFFIXes and return the result for the
1444 newest file that exists. Does not apply to remote files,
1445 or if a non-nil and non-t PREDICATE is specified. */
1448 openp (Lisp_Object path, Lisp_Object str, Lisp_Object suffixes,
1449 Lisp_Object *storeptr, Lisp_Object predicate, bool newer)
1451 ptrdiff_t fn_size = 100;
1452 char buf[100];
1453 char *fn = buf;
1454 bool absolute;
1455 ptrdiff_t want_length;
1456 Lisp_Object filename;
1457 Lisp_Object string, tail, encoded_fn, save_string;
1458 ptrdiff_t max_suffix_len = 0;
1459 int last_errno = ENOENT;
1460 int save_fd = -1;
1461 USE_SAFE_ALLOCA;
1463 /* The last-modified time of the newest matching file found.
1464 Initialize it to something less than all valid timestamps. */
1465 struct timespec save_mtime = make_timespec (TYPE_MINIMUM (time_t), -1);
1467 CHECK_STRING (str);
1469 for (tail = suffixes; CONSP (tail); tail = XCDR (tail))
1471 CHECK_STRING_CAR (tail);
1472 max_suffix_len = max (max_suffix_len,
1473 SBYTES (XCAR (tail)));
1476 string = filename = encoded_fn = save_string = Qnil;
1478 if (storeptr)
1479 *storeptr = Qnil;
1481 absolute = complete_filename_p (str);
1483 for (; CONSP (path); path = XCDR (path))
1485 ptrdiff_t baselen, prefixlen;
1487 filename = Fexpand_file_name (str, XCAR (path));
1488 if (!complete_filename_p (filename))
1489 /* If there are non-absolute elts in PATH (eg "."). */
1490 /* Of course, this could conceivably lose if luser sets
1491 default-directory to be something non-absolute... */
1493 filename = Fexpand_file_name (filename, BVAR (current_buffer, directory));
1494 if (!complete_filename_p (filename))
1495 /* Give up on this path element! */
1496 continue;
1499 /* Calculate maximum length of any filename made from
1500 this path element/specified file name and any possible suffix. */
1501 want_length = max_suffix_len + SBYTES (filename);
1502 if (fn_size <= want_length)
1504 fn_size = 100 + want_length;
1505 fn = SAFE_ALLOCA (fn_size);
1508 /* Copy FILENAME's data to FN but remove starting /: if any. */
1509 prefixlen = ((SCHARS (filename) > 2
1510 && SREF (filename, 0) == '/'
1511 && SREF (filename, 1) == ':')
1512 ? 2 : 0);
1513 baselen = SBYTES (filename) - prefixlen;
1514 memcpy (fn, SDATA (filename) + prefixlen, baselen);
1516 /* Loop over suffixes. */
1517 for (tail = NILP (suffixes) ? list1 (empty_unibyte_string) : suffixes;
1518 CONSP (tail); tail = XCDR (tail))
1520 Lisp_Object suffix = XCAR (tail);
1521 ptrdiff_t fnlen, lsuffix = SBYTES (suffix);
1522 Lisp_Object handler;
1524 /* Make complete filename by appending SUFFIX. */
1525 memcpy (fn + baselen, SDATA (suffix), lsuffix + 1);
1526 fnlen = baselen + lsuffix;
1528 /* Check that the file exists and is not a directory. */
1529 /* We used to only check for handlers on non-absolute file names:
1530 if (absolute)
1531 handler = Qnil;
1532 else
1533 handler = Ffind_file_name_handler (filename, Qfile_exists_p);
1534 It's not clear why that was the case and it breaks things like
1535 (load "/bar.el") where the file is actually "/bar.el.gz". */
1536 /* make_string has its own ideas on when to return a unibyte
1537 string and when a multibyte string, but we know better.
1538 We must have a unibyte string when dumping, since
1539 file-name encoding is shaky at best at that time, and in
1540 particular default-file-name-coding-system is reset
1541 several times during loadup. We therefore don't want to
1542 encode the file before passing it to file I/O library
1543 functions. */
1544 if (!STRING_MULTIBYTE (filename) && !STRING_MULTIBYTE (suffix))
1545 string = make_unibyte_string (fn, fnlen);
1546 else
1547 string = make_string (fn, fnlen);
1548 handler = Ffind_file_name_handler (string, Qfile_exists_p);
1549 if ((!NILP (handler) || (!NILP (predicate) && !EQ (predicate, Qt)))
1550 && !NATNUMP (predicate))
1552 bool exists;
1553 if (NILP (predicate) || EQ (predicate, Qt))
1554 exists = !NILP (Ffile_readable_p (string));
1555 else
1557 Lisp_Object tmp = call1 (predicate, string);
1558 if (NILP (tmp))
1559 exists = false;
1560 else if (EQ (tmp, Qdir_ok)
1561 || NILP (Ffile_directory_p (string)))
1562 exists = true;
1563 else
1565 exists = false;
1566 last_errno = EISDIR;
1570 if (exists)
1572 /* We succeeded; return this descriptor and filename. */
1573 if (storeptr)
1574 *storeptr = string;
1575 SAFE_FREE ();
1576 return -2;
1579 else
1581 int fd;
1582 const char *pfn;
1583 struct stat st;
1585 encoded_fn = ENCODE_FILE (string);
1586 pfn = SSDATA (encoded_fn);
1588 /* Check that we can access or open it. */
1589 if (NATNUMP (predicate))
1591 fd = -1;
1592 if (INT_MAX < XFASTINT (predicate))
1593 last_errno = EINVAL;
1594 else if (faccessat (AT_FDCWD, pfn, XFASTINT (predicate),
1595 AT_EACCESS)
1596 == 0)
1598 if (file_directory_p (pfn))
1599 last_errno = EISDIR;
1600 else
1601 fd = 1;
1604 else
1606 fd = emacs_open (pfn, O_RDONLY, 0);
1607 if (fd < 0)
1609 if (errno != ENOENT)
1610 last_errno = errno;
1612 else
1614 int err = (fstat (fd, &st) != 0 ? errno
1615 : S_ISDIR (st.st_mode) ? EISDIR : 0);
1616 if (err)
1618 last_errno = err;
1619 emacs_close (fd);
1620 fd = -1;
1625 if (fd >= 0)
1627 if (newer && !NATNUMP (predicate))
1629 struct timespec mtime = get_stat_mtime (&st);
1631 if (timespec_cmp (mtime, save_mtime) <= 0)
1632 emacs_close (fd);
1633 else
1635 if (0 <= save_fd)
1636 emacs_close (save_fd);
1637 save_fd = fd;
1638 save_mtime = mtime;
1639 save_string = string;
1642 else
1644 /* We succeeded; return this descriptor and filename. */
1645 if (storeptr)
1646 *storeptr = string;
1647 SAFE_FREE ();
1648 return fd;
1652 /* No more suffixes. Return the newest. */
1653 if (0 <= save_fd && ! CONSP (XCDR (tail)))
1655 if (storeptr)
1656 *storeptr = save_string;
1657 SAFE_FREE ();
1658 return save_fd;
1662 if (absolute)
1663 break;
1666 SAFE_FREE ();
1667 errno = last_errno;
1668 return -1;
1672 /* Merge the list we've accumulated of globals from the current input source
1673 into the load_history variable. The details depend on whether
1674 the source has an associated file name or not.
1676 FILENAME is the file name that we are loading from.
1678 ENTIRE is true if loading that entire file, false if evaluating
1679 part of it. */
1681 static void
1682 build_load_history (Lisp_Object filename, bool entire)
1684 Lisp_Object tail, prev, newelt;
1685 Lisp_Object tem, tem2;
1686 bool foundit = 0;
1688 tail = Vload_history;
1689 prev = Qnil;
1691 while (CONSP (tail))
1693 tem = XCAR (tail);
1695 /* Find the feature's previous assoc list... */
1696 if (!NILP (Fequal (filename, Fcar (tem))))
1698 foundit = 1;
1700 /* If we're loading the entire file, remove old data. */
1701 if (entire)
1703 if (NILP (prev))
1704 Vload_history = XCDR (tail);
1705 else
1706 Fsetcdr (prev, XCDR (tail));
1709 /* Otherwise, cons on new symbols that are not already members. */
1710 else
1712 tem2 = Vcurrent_load_list;
1714 while (CONSP (tem2))
1716 newelt = XCAR (tem2);
1718 if (NILP (Fmember (newelt, tem)))
1719 Fsetcar (tail, Fcons (XCAR (tem),
1720 Fcons (newelt, XCDR (tem))));
1722 tem2 = XCDR (tem2);
1723 maybe_quit ();
1727 else
1728 prev = tail;
1729 tail = XCDR (tail);
1730 maybe_quit ();
1733 /* If we're loading an entire file, cons the new assoc onto the
1734 front of load-history, the most-recently-loaded position. Also
1735 do this if we didn't find an existing member for the file. */
1736 if (entire || !foundit)
1737 Vload_history = Fcons (Fnreverse (Vcurrent_load_list),
1738 Vload_history);
1741 static void
1742 readevalloop_1 (int old)
1744 load_convert_to_unibyte = old;
1747 /* Signal an `end-of-file' error, if possible with file name
1748 information. */
1750 static _Noreturn void
1751 end_of_file_error (void)
1753 if (STRINGP (Vload_file_name))
1754 xsignal1 (Qend_of_file, Vload_file_name);
1756 xsignal0 (Qend_of_file);
1759 static Lisp_Object
1760 readevalloop_eager_expand_eval (Lisp_Object val, Lisp_Object macroexpand)
1762 /* If we macroexpand the toplevel form non-recursively and it ends
1763 up being a `progn' (or if it was a progn to start), treat each
1764 form in the progn as a top-level form. This way, if one form in
1765 the progn defines a macro, that macro is in effect when we expand
1766 the remaining forms. See similar code in bytecomp.el. */
1767 val = call2 (macroexpand, val, Qnil);
1768 if (EQ (CAR_SAFE (val), Qprogn))
1770 Lisp_Object subforms = XCDR (val);
1772 for (val = Qnil; CONSP (subforms); subforms = XCDR (subforms))
1773 val = readevalloop_eager_expand_eval (XCAR (subforms),
1774 macroexpand);
1776 else
1777 val = eval_sub (call2 (macroexpand, val, Qt));
1778 return val;
1781 /* UNIBYTE specifies how to set load_convert_to_unibyte
1782 for this invocation.
1783 READFUN, if non-nil, is used instead of `read'.
1785 START, END specify region to read in current buffer (from eval-region).
1786 If the input is not from a buffer, they must be nil. */
1788 static void
1789 readevalloop (Lisp_Object readcharfun,
1790 FILE *stream,
1791 Lisp_Object sourcename,
1792 bool printflag,
1793 Lisp_Object unibyte, Lisp_Object readfun,
1794 Lisp_Object start, Lisp_Object end)
1796 int c;
1797 Lisp_Object val;
1798 ptrdiff_t count = SPECPDL_INDEX ();
1799 struct buffer *b = 0;
1800 bool continue_reading_p;
1801 Lisp_Object lex_bound;
1802 /* True if reading an entire buffer. */
1803 bool whole_buffer = 0;
1804 /* True on the first time around. */
1805 bool first_sexp = 1;
1806 Lisp_Object macroexpand = intern ("internal-macroexpand-for-load");
1808 if (NILP (Ffboundp (macroexpand))
1809 /* Don't macroexpand in .elc files, since it should have been done
1810 already. We actually don't know whether we're in a .elc file or not,
1811 so we use circumstantial evidence: .el files normally go through
1812 Vload_source_file_function -> load-with-code-conversion
1813 -> eval-buffer. */
1814 || EQ (readcharfun, Qget_file_char)
1815 || EQ (readcharfun, Qget_emacs_mule_file_char))
1816 macroexpand = Qnil;
1818 if (MARKERP (readcharfun))
1820 if (NILP (start))
1821 start = readcharfun;
1824 if (BUFFERP (readcharfun))
1825 b = XBUFFER (readcharfun);
1826 else if (MARKERP (readcharfun))
1827 b = XMARKER (readcharfun)->buffer;
1829 /* We assume START is nil when input is not from a buffer. */
1830 if (! NILP (start) && !b)
1831 emacs_abort ();
1833 specbind (Qstandard_input, readcharfun);
1834 specbind (Qcurrent_load_list, Qnil);
1835 record_unwind_protect_int (readevalloop_1, load_convert_to_unibyte);
1836 load_convert_to_unibyte = !NILP (unibyte);
1838 /* If lexical binding is active (either because it was specified in
1839 the file's header, or via a buffer-local variable), create an empty
1840 lexical environment, otherwise, turn off lexical binding. */
1841 lex_bound = find_symbol_value (Qlexical_binding);
1842 specbind (Qinternal_interpreter_environment,
1843 (NILP (lex_bound) || EQ (lex_bound, Qunbound)
1844 ? Qnil : list1 (Qt)));
1846 /* Try to ensure sourcename is a truename, except whilst preloading. */
1847 if (NILP (Vpurify_flag)
1848 && !NILP (sourcename) && !NILP (Ffile_name_absolute_p (sourcename))
1849 && !NILP (Ffboundp (Qfile_truename)))
1850 sourcename = call1 (Qfile_truename, sourcename) ;
1852 LOADHIST_ATTACH (sourcename);
1854 continue_reading_p = 1;
1855 while (continue_reading_p)
1857 ptrdiff_t count1 = SPECPDL_INDEX ();
1859 if (b != 0 && !BUFFER_LIVE_P (b))
1860 error ("Reading from killed buffer");
1862 if (!NILP (start))
1864 /* Switch to the buffer we are reading from. */
1865 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1866 set_buffer_internal (b);
1868 /* Save point in it. */
1869 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1870 /* Save ZV in it. */
1871 record_unwind_protect (save_restriction_restore, save_restriction_save ());
1872 /* Those get unbound after we read one expression. */
1874 /* Set point and ZV around stuff to be read. */
1875 Fgoto_char (start);
1876 if (!NILP (end))
1877 Fnarrow_to_region (make_number (BEGV), end);
1879 /* Just for cleanliness, convert END to a marker
1880 if it is an integer. */
1881 if (INTEGERP (end))
1882 end = Fpoint_max_marker ();
1885 /* On the first cycle, we can easily test here
1886 whether we are reading the whole buffer. */
1887 if (b && first_sexp)
1888 whole_buffer = (PT == BEG && ZV == Z);
1890 instream = stream;
1891 read_next:
1892 c = READCHAR;
1893 if (c == ';')
1895 while ((c = READCHAR) != '\n' && c != -1);
1896 goto read_next;
1898 if (c < 0)
1900 unbind_to (count1, Qnil);
1901 break;
1904 /* Ignore whitespace here, so we can detect eof. */
1905 if (c == ' ' || c == '\t' || c == '\n' || c == '\f' || c == '\r'
1906 || c == NO_BREAK_SPACE)
1907 goto read_next;
1909 if (!NILP (Vpurify_flag) && c == '(')
1911 val = read_list (0, readcharfun);
1913 else
1915 UNREAD (c);
1916 read_objects = Qnil;
1917 if (!NILP (readfun))
1919 val = call1 (readfun, readcharfun);
1921 /* If READCHARFUN has set point to ZV, we should
1922 stop reading, even if the form read sets point
1923 to a different value when evaluated. */
1924 if (BUFFERP (readcharfun))
1926 struct buffer *buf = XBUFFER (readcharfun);
1927 if (BUF_PT (buf) == BUF_ZV (buf))
1928 continue_reading_p = 0;
1931 else if (! NILP (Vload_read_function))
1932 val = call1 (Vload_read_function, readcharfun);
1933 else
1934 val = read_internal_start (readcharfun, Qnil, Qnil);
1937 if (!NILP (start) && continue_reading_p)
1938 start = Fpoint_marker ();
1940 /* Restore saved point and BEGV. */
1941 unbind_to (count1, Qnil);
1943 /* Now eval what we just read. */
1944 if (!NILP (macroexpand))
1945 val = readevalloop_eager_expand_eval (val, macroexpand);
1946 else
1947 val = eval_sub (val);
1949 if (printflag)
1951 Vvalues = Fcons (val, Vvalues);
1952 if (EQ (Vstandard_output, Qt))
1953 Fprin1 (val, Qnil);
1954 else
1955 Fprint (val, Qnil);
1958 first_sexp = 0;
1961 build_load_history (sourcename,
1962 stream || whole_buffer);
1964 unbind_to (count, Qnil);
1967 DEFUN ("eval-buffer", Feval_buffer, Seval_buffer, 0, 5, "",
1968 doc: /* Execute the accessible portion of current buffer as Lisp code.
1969 You can use \\[narrow-to-region] to limit the part of buffer to be evaluated.
1970 When called from a Lisp program (i.e., not interactively), this
1971 function accepts up to five optional arguments:
1972 BUFFER is the buffer to evaluate (nil means use current buffer),
1973 or a name of a buffer (a string).
1974 PRINTFLAG controls printing of output by any output functions in the
1975 evaluated code, such as `print', `princ', and `prin1':
1976 a value of nil means discard it; anything else is the stream to print to.
1977 See Info node `(elisp)Output Streams' for details on streams.
1978 FILENAME specifies the file name to use for `load-history'.
1979 UNIBYTE, if non-nil, specifies `load-convert-to-unibyte' for this
1980 invocation.
1981 DO-ALLOW-PRINT, if non-nil, specifies that output functions in the
1982 evaluated code should work normally even if PRINTFLAG is nil, in
1983 which case the output is displayed in the echo area.
1985 This function preserves the position of point. */)
1986 (Lisp_Object buffer, Lisp_Object printflag, Lisp_Object filename, Lisp_Object unibyte, Lisp_Object do_allow_print)
1988 ptrdiff_t count = SPECPDL_INDEX ();
1989 Lisp_Object tem, buf;
1991 if (NILP (buffer))
1992 buf = Fcurrent_buffer ();
1993 else
1994 buf = Fget_buffer (buffer);
1995 if (NILP (buf))
1996 error ("No such buffer");
1998 if (NILP (printflag) && NILP (do_allow_print))
1999 tem = Qsymbolp;
2000 else
2001 tem = printflag;
2003 if (NILP (filename))
2004 filename = BVAR (XBUFFER (buf), filename);
2006 specbind (Qeval_buffer_list, Fcons (buf, Veval_buffer_list));
2007 specbind (Qstandard_output, tem);
2008 record_unwind_protect (save_excursion_restore, save_excursion_save ());
2009 BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
2010 specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
2011 readevalloop (buf, 0, filename,
2012 !NILP (printflag), unibyte, Qnil, Qnil, Qnil);
2013 unbind_to (count, Qnil);
2015 return Qnil;
2018 DEFUN ("eval-region", Feval_region, Seval_region, 2, 4, "r",
2019 doc: /* Execute the region as Lisp code.
2020 When called from programs, expects two arguments,
2021 giving starting and ending indices in the current buffer
2022 of the text to be executed.
2023 Programs can pass third argument PRINTFLAG which controls output:
2024 a value of nil means discard it; anything else is stream for printing it.
2025 See Info node `(elisp)Output Streams' for details on streams.
2026 Also the fourth argument READ-FUNCTION, if non-nil, is used
2027 instead of `read' to read each expression. It gets one argument
2028 which is the input stream for reading characters.
2030 This function does not move point. */)
2031 (Lisp_Object start, Lisp_Object end, Lisp_Object printflag, Lisp_Object read_function)
2033 /* FIXME: Do the eval-sexp-add-defvars dance! */
2034 ptrdiff_t count = SPECPDL_INDEX ();
2035 Lisp_Object tem, cbuf;
2037 cbuf = Fcurrent_buffer ();
2039 if (NILP (printflag))
2040 tem = Qsymbolp;
2041 else
2042 tem = printflag;
2043 specbind (Qstandard_output, tem);
2044 specbind (Qeval_buffer_list, Fcons (cbuf, Veval_buffer_list));
2046 /* `readevalloop' calls functions which check the type of start and end. */
2047 readevalloop (cbuf, 0, BVAR (XBUFFER (cbuf), filename),
2048 !NILP (printflag), Qnil, read_function,
2049 start, end);
2051 return unbind_to (count, Qnil);
2055 DEFUN ("read", Fread, Sread, 0, 1, 0,
2056 doc: /* Read one Lisp expression as text from STREAM, return as Lisp object.
2057 If STREAM is nil, use the value of `standard-input' (which see).
2058 STREAM or the value of `standard-input' may be:
2059 a buffer (read from point and advance it)
2060 a marker (read from where it points and advance it)
2061 a function (call it with no arguments for each character,
2062 call it with a char as argument to push a char back)
2063 a string (takes text from string, starting at the beginning)
2064 t (read text line using minibuffer and use it, or read from
2065 standard input in batch mode). */)
2066 (Lisp_Object stream)
2068 if (NILP (stream))
2069 stream = Vstandard_input;
2070 if (EQ (stream, Qt))
2071 stream = Qread_char;
2072 if (EQ (stream, Qread_char))
2073 /* FIXME: ?! When is this used !? */
2074 return call1 (intern ("read-minibuffer"),
2075 build_string ("Lisp expression: "));
2077 return read_internal_start (stream, Qnil, Qnil);
2080 DEFUN ("read-from-string", Fread_from_string, Sread_from_string, 1, 3, 0,
2081 doc: /* Read one Lisp expression which is represented as text by STRING.
2082 Returns a cons: (OBJECT-READ . FINAL-STRING-INDEX).
2083 FINAL-STRING-INDEX is an integer giving the position of the next
2084 remaining character in STRING. START and END optionally delimit
2085 a substring of STRING from which to read; they default to 0 and
2086 \(length STRING) respectively. Negative values are counted from
2087 the end of STRING. */)
2088 (Lisp_Object string, Lisp_Object start, Lisp_Object end)
2090 Lisp_Object ret;
2091 CHECK_STRING (string);
2092 /* `read_internal_start' sets `read_from_string_index'. */
2093 ret = read_internal_start (string, start, end);
2094 return Fcons (ret, make_number (read_from_string_index));
2097 /* Function to set up the global context we need in toplevel read
2098 calls. START and END only used when STREAM is a string. */
2099 static Lisp_Object
2100 read_internal_start (Lisp_Object stream, Lisp_Object start, Lisp_Object end)
2102 Lisp_Object retval;
2104 readchar_count = 0;
2105 new_backquote_flag = 0;
2106 read_objects = Qnil;
2107 if (EQ (Vread_with_symbol_positions, Qt)
2108 || EQ (Vread_with_symbol_positions, stream))
2109 Vread_symbol_positions_list = Qnil;
2111 if (STRINGP (stream)
2112 || ((CONSP (stream) && STRINGP (XCAR (stream)))))
2114 ptrdiff_t startval, endval;
2115 Lisp_Object string;
2117 if (STRINGP (stream))
2118 string = stream;
2119 else
2120 string = XCAR (stream);
2122 validate_subarray (string, start, end, SCHARS (string),
2123 &startval, &endval);
2125 read_from_string_index = startval;
2126 read_from_string_index_byte = string_char_to_byte (string, startval);
2127 read_from_string_limit = endval;
2130 retval = read0 (stream);
2131 if (EQ (Vread_with_symbol_positions, Qt)
2132 || EQ (Vread_with_symbol_positions, stream))
2133 Vread_symbol_positions_list = Fnreverse (Vread_symbol_positions_list);
2134 return retval;
2138 /* Signal Qinvalid_read_syntax error.
2139 S is error string of length N (if > 0) */
2141 static _Noreturn void
2142 invalid_syntax (const char *s)
2144 xsignal1 (Qinvalid_read_syntax, build_string (s));
2148 /* Use this for recursive reads, in contexts where internal tokens
2149 are not allowed. */
2151 static Lisp_Object
2152 read0 (Lisp_Object readcharfun)
2154 register Lisp_Object val;
2155 int c;
2157 val = read1 (readcharfun, &c, 0);
2158 if (!c)
2159 return val;
2161 xsignal1 (Qinvalid_read_syntax,
2162 Fmake_string (make_number (1), make_number (c)));
2165 /* Grow a read buffer BUF that contains OFFSET useful bytes of data,
2166 by at least MAX_MULTIBYTE_LENGTH bytes. Update *BUF_ADDR and
2167 *BUF_SIZE accordingly; 0 <= OFFSET <= *BUF_SIZE. If *BUF_ADDR is
2168 initially null, BUF is on the stack: copy its data to the new heap
2169 buffer. Otherwise, BUF must equal *BUF_ADDR and can simply be
2170 reallocated. Either way, remember the heap allocation (which is at
2171 pdl slot COUNT) so that it can be freed when unwinding the stack.*/
2173 static char *
2174 grow_read_buffer (char *buf, ptrdiff_t offset,
2175 char **buf_addr, ptrdiff_t *buf_size, ptrdiff_t count)
2177 char *p = xpalloc (*buf_addr, buf_size, MAX_MULTIBYTE_LENGTH, -1, 1);
2178 if (!*buf_addr)
2180 memcpy (p, buf, offset);
2181 record_unwind_protect_ptr (xfree, p);
2183 else
2184 set_unwind_protect_ptr (count, xfree, p);
2185 *buf_addr = p;
2186 return p;
2189 /* Return the scalar value that has the Unicode character name NAME.
2190 Raise 'invalid-read-syntax' if there is no such character. */
2191 static int
2192 character_name_to_code (char const *name, ptrdiff_t name_len)
2194 /* For "U+XXXX", pass the leading '+' to string_to_number to reject
2195 monstrosities like "U+-0000". */
2196 Lisp_Object code
2197 = (name[0] == 'U' && name[1] == '+'
2198 ? string_to_number (name + 1, 16, false)
2199 : call2 (Qchar_from_name, make_unibyte_string (name, name_len), Qt));
2201 if (! RANGED_INTEGERP (0, code, MAX_UNICODE_CHAR)
2202 || char_surrogate_p (XINT (code)))
2204 AUTO_STRING (format, "\\N{%s}");
2205 AUTO_STRING_WITH_LEN (namestr, name, name_len);
2206 xsignal1 (Qinvalid_read_syntax, CALLN (Fformat, format, namestr));
2209 return XINT (code);
2212 /* Bound on the length of a Unicode character name. As of
2213 Unicode 9.0.0 the maximum is 83, so this should be safe. */
2214 enum { UNICODE_CHARACTER_NAME_LENGTH_BOUND = 200 };
2216 /* Read a \-escape sequence, assuming we already read the `\'.
2217 If the escape sequence forces unibyte, return eight-bit char. */
2219 static int
2220 read_escape (Lisp_Object readcharfun, bool stringp)
2222 int c = READCHAR;
2223 /* \u allows up to four hex digits, \U up to eight. Default to the
2224 behavior for \u, and change this value in the case that \U is seen. */
2225 int unicode_hex_count = 4;
2227 switch (c)
2229 case -1:
2230 end_of_file_error ();
2232 case 'a':
2233 return '\007';
2234 case 'b':
2235 return '\b';
2236 case 'd':
2237 return 0177;
2238 case 'e':
2239 return 033;
2240 case 'f':
2241 return '\f';
2242 case 'n':
2243 return '\n';
2244 case 'r':
2245 return '\r';
2246 case 't':
2247 return '\t';
2248 case 'v':
2249 return '\v';
2250 case '\n':
2251 return -1;
2252 case ' ':
2253 if (stringp)
2254 return -1;
2255 return ' ';
2257 case 'M':
2258 c = READCHAR;
2259 if (c != '-')
2260 error ("Invalid escape character syntax");
2261 c = READCHAR;
2262 if (c == '\\')
2263 c = read_escape (readcharfun, 0);
2264 return c | meta_modifier;
2266 case 'S':
2267 c = READCHAR;
2268 if (c != '-')
2269 error ("Invalid escape character syntax");
2270 c = READCHAR;
2271 if (c == '\\')
2272 c = read_escape (readcharfun, 0);
2273 return c | shift_modifier;
2275 case 'H':
2276 c = READCHAR;
2277 if (c != '-')
2278 error ("Invalid escape character syntax");
2279 c = READCHAR;
2280 if (c == '\\')
2281 c = read_escape (readcharfun, 0);
2282 return c | hyper_modifier;
2284 case 'A':
2285 c = READCHAR;
2286 if (c != '-')
2287 error ("Invalid escape character syntax");
2288 c = READCHAR;
2289 if (c == '\\')
2290 c = read_escape (readcharfun, 0);
2291 return c | alt_modifier;
2293 case 's':
2294 c = READCHAR;
2295 if (stringp || c != '-')
2297 UNREAD (c);
2298 return ' ';
2300 c = READCHAR;
2301 if (c == '\\')
2302 c = read_escape (readcharfun, 0);
2303 return c | super_modifier;
2305 case 'C':
2306 c = READCHAR;
2307 if (c != '-')
2308 error ("Invalid escape character syntax");
2309 case '^':
2310 c = READCHAR;
2311 if (c == '\\')
2312 c = read_escape (readcharfun, 0);
2313 if ((c & ~CHAR_MODIFIER_MASK) == '?')
2314 return 0177 | (c & CHAR_MODIFIER_MASK);
2315 else if (! SINGLE_BYTE_CHAR_P ((c & ~CHAR_MODIFIER_MASK)))
2316 return c | ctrl_modifier;
2317 /* ASCII control chars are made from letters (both cases),
2318 as well as the non-letters within 0100...0137. */
2319 else if ((c & 0137) >= 0101 && (c & 0137) <= 0132)
2320 return (c & (037 | ~0177));
2321 else if ((c & 0177) >= 0100 && (c & 0177) <= 0137)
2322 return (c & (037 | ~0177));
2323 else
2324 return c | ctrl_modifier;
2326 case '0':
2327 case '1':
2328 case '2':
2329 case '3':
2330 case '4':
2331 case '5':
2332 case '6':
2333 case '7':
2334 /* An octal escape, as in ANSI C. */
2336 register int i = c - '0';
2337 register int count = 0;
2338 while (++count < 3)
2340 if ((c = READCHAR) >= '0' && c <= '7')
2342 i *= 8;
2343 i += c - '0';
2345 else
2347 UNREAD (c);
2348 break;
2352 if (i >= 0x80 && i < 0x100)
2353 i = BYTE8_TO_CHAR (i);
2354 return i;
2357 case 'x':
2358 /* A hex escape, as in ANSI C. */
2360 unsigned int i = 0;
2361 int count = 0;
2362 while (1)
2364 c = READCHAR;
2365 if (c >= '0' && c <= '9')
2367 i *= 16;
2368 i += c - '0';
2370 else if ((c >= 'a' && c <= 'f')
2371 || (c >= 'A' && c <= 'F'))
2373 i *= 16;
2374 if (c >= 'a' && c <= 'f')
2375 i += c - 'a' + 10;
2376 else
2377 i += c - 'A' + 10;
2379 else
2381 UNREAD (c);
2382 break;
2384 /* Allow hex escapes as large as ?\xfffffff, because some
2385 packages use them to denote characters with modifiers. */
2386 if ((CHAR_META | (CHAR_META - 1)) < i)
2387 error ("Hex character out of range: \\x%x...", i);
2388 count += count < 3;
2391 if (count < 3 && i >= 0x80)
2392 return BYTE8_TO_CHAR (i);
2393 return i;
2396 case 'U':
2397 /* Post-Unicode-2.0: Up to eight hex chars. */
2398 unicode_hex_count = 8;
2399 case 'u':
2401 /* A Unicode escape. We only permit them in strings and characters,
2402 not arbitrarily in the source code, as in some other languages. */
2404 unsigned int i = 0;
2405 int count = 0;
2407 while (++count <= unicode_hex_count)
2409 c = READCHAR;
2410 /* `isdigit' and `isalpha' may be locale-specific, which we don't
2411 want. */
2412 if (c >= '0' && c <= '9') i = (i << 4) + (c - '0');
2413 else if (c >= 'a' && c <= 'f') i = (i << 4) + (c - 'a') + 10;
2414 else if (c >= 'A' && c <= 'F') i = (i << 4) + (c - 'A') + 10;
2415 else
2416 error ("Non-hex digit used for Unicode escape");
2418 if (i > 0x10FFFF)
2419 error ("Non-Unicode character: 0x%x", i);
2420 return i;
2423 case 'N':
2424 /* Named character. */
2426 c = READCHAR;
2427 if (c != '{')
2428 invalid_syntax ("Expected opening brace after \\N");
2429 char name[UNICODE_CHARACTER_NAME_LENGTH_BOUND + 1];
2430 bool whitespace = false;
2431 ptrdiff_t length = 0;
2432 while (true)
2434 c = READCHAR;
2435 if (c < 0)
2436 end_of_file_error ();
2437 if (c == '}')
2438 break;
2439 if (! (0 < c && c < 0x80))
2441 AUTO_STRING (format,
2442 "Invalid character U+%04X in character name");
2443 xsignal1 (Qinvalid_read_syntax,
2444 CALLN (Fformat, format, make_natnum (c)));
2446 /* Treat multiple adjacent whitespace characters as a
2447 single space character. This makes it easier to use
2448 character names in e.g. multi-line strings. */
2449 if (c_isspace (c))
2451 if (whitespace)
2452 continue;
2453 c = ' ';
2454 whitespace = true;
2456 else
2457 whitespace = false;
2458 name[length++] = c;
2459 if (length >= sizeof name)
2460 invalid_syntax ("Character name too long");
2462 if (length == 0)
2463 invalid_syntax ("Empty character name");
2464 name[length] = '\0';
2466 /* character_name_to_code can invoke read1, recursively.
2467 This is why read1's buffer is not static. */
2468 return character_name_to_code (name, length);
2471 default:
2472 return c;
2476 /* Return the digit that CHARACTER stands for in the given BASE.
2477 Return -1 if CHARACTER is out of range for BASE,
2478 and -2 if CHARACTER is not valid for any supported BASE. */
2479 static int
2480 digit_to_number (int character, int base)
2482 int digit;
2484 if ('0' <= character && character <= '9')
2485 digit = character - '0';
2486 else if ('a' <= character && character <= 'z')
2487 digit = character - 'a' + 10;
2488 else if ('A' <= character && character <= 'Z')
2489 digit = character - 'A' + 10;
2490 else
2491 return -2;
2493 return digit < base ? digit : -1;
2496 /* Read an integer in radix RADIX using READCHARFUN to read
2497 characters. RADIX must be in the interval [2..36]; if it isn't, a
2498 read error is signaled . Value is the integer read. Signals an
2499 error if encountering invalid read syntax or if RADIX is out of
2500 range. */
2502 static Lisp_Object
2503 read_integer (Lisp_Object readcharfun, EMACS_INT radix)
2505 /* Room for sign, leading 0, other digits, trailing null byte.
2506 Also, room for invalid syntax diagnostic. */
2507 char buf[max (1 + 1 + UINTMAX_WIDTH + 1,
2508 sizeof "integer, radix " + INT_STRLEN_BOUND (EMACS_INT))];
2510 int valid = -1; /* 1 if valid, 0 if not, -1 if incomplete. */
2512 if (radix < 2 || radix > 36)
2513 valid = 0;
2514 else
2516 char *p = buf;
2517 int c, digit;
2519 c = READCHAR;
2520 if (c == '-' || c == '+')
2522 *p++ = c;
2523 c = READCHAR;
2526 if (c == '0')
2528 *p++ = c;
2529 valid = 1;
2531 /* Ignore redundant leading zeros, so the buffer doesn't
2532 fill up with them. */
2534 c = READCHAR;
2535 while (c == '0');
2538 while ((digit = digit_to_number (c, radix)) >= -1)
2540 if (digit == -1)
2541 valid = 0;
2542 if (valid < 0)
2543 valid = 1;
2545 if (p < buf + sizeof buf - 1)
2546 *p++ = c;
2547 else
2548 valid = 0;
2550 c = READCHAR;
2553 UNREAD (c);
2554 *p = '\0';
2557 if (valid != 1)
2559 sprintf (buf, "integer, radix %"pI"d", radix);
2560 invalid_syntax (buf);
2563 return string_to_number (buf, radix, 0);
2567 /* If the next token is ')' or ']' or '.', we store that character
2568 in *PCH and the return value is not interesting. Else, we store
2569 zero in *PCH and we read and return one lisp object.
2571 FIRST_IN_LIST is true if this is the first element of a list. */
2573 static Lisp_Object
2574 read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
2576 int c;
2577 bool uninterned_symbol = false;
2578 bool multibyte;
2579 char stackbuf[MAX_ALLOCA];
2581 *pch = 0;
2583 retry:
2585 c = READCHAR_REPORT_MULTIBYTE (&multibyte);
2586 if (c < 0)
2587 end_of_file_error ();
2589 switch (c)
2591 case '(':
2592 return read_list (0, readcharfun);
2594 case '[':
2595 return read_vector (readcharfun, 0);
2597 case ')':
2598 case ']':
2600 *pch = c;
2601 return Qnil;
2604 case '#':
2605 c = READCHAR;
2606 if (c == 's')
2608 c = READCHAR;
2609 if (c == '(')
2611 /* Accept extended format for hash tables (extensible to
2612 other types), e.g.
2613 #s(hash-table size 2 test equal data (k1 v1 k2 v2)) */
2614 Lisp_Object tmp = read_list (0, readcharfun);
2615 Lisp_Object head = CAR_SAFE (tmp);
2616 Lisp_Object data = Qnil;
2617 Lisp_Object val = Qnil;
2618 /* The size is 2 * number of allowed keywords to
2619 make-hash-table. */
2620 Lisp_Object params[12];
2621 Lisp_Object ht;
2622 Lisp_Object key = Qnil;
2623 int param_count = 0;
2625 if (!EQ (head, Qhash_table))
2627 ptrdiff_t size = XINT (Flength (tmp));
2628 Lisp_Object record = Fmake_record (CAR_SAFE (tmp),
2629 make_number (size - 1),
2630 Qnil);
2631 for (int i = 1; i < size; i++)
2633 tmp = Fcdr (tmp);
2634 ASET (record, i, Fcar (tmp));
2636 return record;
2639 tmp = CDR_SAFE (tmp);
2641 /* This is repetitive but fast and simple. */
2642 params[param_count] = QCsize;
2643 params[param_count + 1] = Fplist_get (tmp, Qsize);
2644 if (!NILP (params[param_count + 1]))
2645 param_count += 2;
2647 params[param_count] = QCtest;
2648 params[param_count + 1] = Fplist_get (tmp, Qtest);
2649 if (!NILP (params[param_count + 1]))
2650 param_count += 2;
2652 params[param_count] = QCweakness;
2653 params[param_count + 1] = Fplist_get (tmp, Qweakness);
2654 if (!NILP (params[param_count + 1]))
2655 param_count += 2;
2657 params[param_count] = QCrehash_size;
2658 params[param_count + 1] = Fplist_get (tmp, Qrehash_size);
2659 if (!NILP (params[param_count + 1]))
2660 param_count += 2;
2662 params[param_count] = QCrehash_threshold;
2663 params[param_count + 1] = Fplist_get (tmp, Qrehash_threshold);
2664 if (!NILP (params[param_count + 1]))
2665 param_count += 2;
2667 params[param_count] = QCpurecopy;
2668 params[param_count + 1] = Fplist_get (tmp, Qpurecopy);
2669 if (!NILP (params[param_count + 1]))
2670 param_count += 2;
2672 /* This is the hash table data. */
2673 data = Fplist_get (tmp, Qdata);
2675 /* Now use params to make a new hash table and fill it. */
2676 ht = Fmake_hash_table (param_count, params);
2678 while (CONSP (data))
2680 key = XCAR (data);
2681 data = XCDR (data);
2682 if (!CONSP (data))
2683 error ("Odd number of elements in hash table data");
2684 val = XCAR (data);
2685 data = XCDR (data);
2686 Fputhash (key, val, ht);
2689 return ht;
2691 UNREAD (c);
2692 invalid_syntax ("#");
2694 if (c == '^')
2696 c = READCHAR;
2697 if (c == '[')
2699 Lisp_Object tmp;
2700 tmp = read_vector (readcharfun, 0);
2701 if (ASIZE (tmp) < CHAR_TABLE_STANDARD_SLOTS)
2702 error ("Invalid size char-table");
2703 XSETPVECTYPE (XVECTOR (tmp), PVEC_CHAR_TABLE);
2704 return tmp;
2706 else if (c == '^')
2708 c = READCHAR;
2709 if (c == '[')
2711 /* Sub char-table can't be read as a regular
2712 vector because of a two C integer fields. */
2713 Lisp_Object tbl, tmp = read_list (1, readcharfun);
2714 ptrdiff_t size = XINT (Flength (tmp));
2715 int i, depth, min_char;
2716 struct Lisp_Cons *cell;
2718 if (size == 0)
2719 error ("Zero-sized sub char-table");
2721 if (! RANGED_INTEGERP (1, XCAR (tmp), 3))
2722 error ("Invalid depth in sub char-table");
2723 depth = XINT (XCAR (tmp));
2724 if (chartab_size[depth] != size - 2)
2725 error ("Invalid size in sub char-table");
2726 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2727 free_cons (cell);
2729 if (! RANGED_INTEGERP (0, XCAR (tmp), MAX_CHAR))
2730 error ("Invalid minimum character in sub-char-table");
2731 min_char = XINT (XCAR (tmp));
2732 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2733 free_cons (cell);
2735 tbl = make_uninit_sub_char_table (depth, min_char);
2736 for (i = 0; i < size; i++)
2738 XSUB_CHAR_TABLE (tbl)->contents[i] = XCAR (tmp);
2739 cell = XCONS (tmp), tmp = XCDR (tmp);
2740 free_cons (cell);
2742 return tbl;
2744 invalid_syntax ("#^^");
2746 invalid_syntax ("#^");
2748 if (c == '&')
2750 Lisp_Object length;
2751 length = read1 (readcharfun, pch, first_in_list);
2752 c = READCHAR;
2753 if (c == '"')
2755 Lisp_Object tmp, val;
2756 EMACS_INT size_in_chars = bool_vector_bytes (XFASTINT (length));
2757 unsigned char *data;
2759 UNREAD (c);
2760 tmp = read1 (readcharfun, pch, first_in_list);
2761 if (STRING_MULTIBYTE (tmp)
2762 || (size_in_chars != SCHARS (tmp)
2763 /* We used to print 1 char too many
2764 when the number of bits was a multiple of 8.
2765 Accept such input in case it came from an old
2766 version. */
2767 && ! (XFASTINT (length)
2768 == (SCHARS (tmp) - 1) * BOOL_VECTOR_BITS_PER_CHAR)))
2769 invalid_syntax ("#&...");
2771 val = make_uninit_bool_vector (XFASTINT (length));
2772 data = bool_vector_uchar_data (val);
2773 memcpy (data, SDATA (tmp), size_in_chars);
2774 /* Clear the extraneous bits in the last byte. */
2775 if (XINT (length) != size_in_chars * BOOL_VECTOR_BITS_PER_CHAR)
2776 data[size_in_chars - 1]
2777 &= (1 << (XINT (length) % BOOL_VECTOR_BITS_PER_CHAR)) - 1;
2778 return val;
2780 invalid_syntax ("#&...");
2782 if (c == '[')
2784 /* Accept compiled functions at read-time so that we don't have to
2785 build them using function calls. */
2786 Lisp_Object tmp;
2787 struct Lisp_Vector *vec;
2788 tmp = read_vector (readcharfun, 1);
2789 vec = XVECTOR (tmp);
2790 if (vec->header.size == 0)
2791 invalid_syntax ("Empty byte-code object");
2792 make_byte_code (vec);
2793 return tmp;
2795 if (c == '(')
2797 Lisp_Object tmp;
2798 int ch;
2800 /* Read the string itself. */
2801 tmp = read1 (readcharfun, &ch, 0);
2802 if (ch != 0 || !STRINGP (tmp))
2803 invalid_syntax ("#");
2804 /* Read the intervals and their properties. */
2805 while (1)
2807 Lisp_Object beg, end, plist;
2809 beg = read1 (readcharfun, &ch, 0);
2810 end = plist = Qnil;
2811 if (ch == ')')
2812 break;
2813 if (ch == 0)
2814 end = read1 (readcharfun, &ch, 0);
2815 if (ch == 0)
2816 plist = read1 (readcharfun, &ch, 0);
2817 if (ch)
2818 invalid_syntax ("Invalid string property list");
2819 Fset_text_properties (beg, end, plist, tmp);
2822 return tmp;
2825 /* #@NUMBER is used to skip NUMBER following bytes.
2826 That's used in .elc files to skip over doc strings
2827 and function definitions. */
2828 if (c == '@')
2830 enum { extra = 100 };
2831 ptrdiff_t i, nskip = 0, digits = 0;
2833 /* Read a decimal integer. */
2834 while ((c = READCHAR) >= 0
2835 && c >= '0' && c <= '9')
2837 if ((STRING_BYTES_BOUND - extra) / 10 <= nskip)
2838 string_overflow ();
2839 digits++;
2840 nskip *= 10;
2841 nskip += c - '0';
2842 if (digits == 2 && nskip == 0)
2843 { /* We've just seen #@00, which means "skip to end". */
2844 skip_dyn_eof (readcharfun);
2845 return Qnil;
2848 if (nskip > 0)
2849 /* We can't use UNREAD here, because in the code below we side-step
2850 READCHAR. Instead, assume the first char after #@NNN occupies
2851 a single byte, which is the case normally since it's just
2852 a space. */
2853 nskip--;
2854 else
2855 UNREAD (c);
2857 if (load_force_doc_strings
2858 && (FROM_FILE_P (readcharfun)))
2860 /* If we are supposed to force doc strings into core right now,
2861 record the last string that we skipped,
2862 and record where in the file it comes from. */
2864 /* But first exchange saved_doc_string
2865 with prev_saved_doc_string, so we save two strings. */
2867 char *temp = saved_doc_string;
2868 ptrdiff_t temp_size = saved_doc_string_size;
2869 file_offset temp_pos = saved_doc_string_position;
2870 ptrdiff_t temp_len = saved_doc_string_length;
2872 saved_doc_string = prev_saved_doc_string;
2873 saved_doc_string_size = prev_saved_doc_string_size;
2874 saved_doc_string_position = prev_saved_doc_string_position;
2875 saved_doc_string_length = prev_saved_doc_string_length;
2877 prev_saved_doc_string = temp;
2878 prev_saved_doc_string_size = temp_size;
2879 prev_saved_doc_string_position = temp_pos;
2880 prev_saved_doc_string_length = temp_len;
2883 if (saved_doc_string_size == 0)
2885 saved_doc_string = xmalloc (nskip + extra);
2886 saved_doc_string_size = nskip + extra;
2888 if (nskip > saved_doc_string_size)
2890 saved_doc_string = xrealloc (saved_doc_string, nskip + extra);
2891 saved_doc_string_size = nskip + extra;
2894 saved_doc_string_position = file_tell (instream);
2896 /* Copy that many characters into saved_doc_string. */
2897 block_input ();
2898 for (i = 0; i < nskip && c >= 0; i++)
2899 saved_doc_string[i] = c = getc (instream);
2900 unblock_input ();
2902 saved_doc_string_length = i;
2904 else
2905 /* Skip that many bytes. */
2906 skip_dyn_bytes (readcharfun, nskip);
2908 goto retry;
2910 if (c == '!')
2912 /* #! appears at the beginning of an executable file.
2913 Skip the first line. */
2914 while (c != '\n' && c >= 0)
2915 c = READCHAR;
2916 goto retry;
2918 if (c == '$')
2919 return Vload_file_name;
2920 if (c == '\'')
2921 return list2 (Qfunction, read0 (readcharfun));
2922 /* #:foo is the uninterned symbol named foo. */
2923 if (c == ':')
2925 uninterned_symbol = true;
2926 c = READCHAR;
2927 if (!(c > 040
2928 && c != NO_BREAK_SPACE
2929 && (c >= 0200
2930 || strchr ("\"';()[]#`,", c) == NULL)))
2932 /* No symbol character follows, this is the empty
2933 symbol. */
2934 UNREAD (c);
2935 return Fmake_symbol (empty_unibyte_string);
2937 goto read_symbol;
2939 /* ## is the empty symbol. */
2940 if (c == '#')
2941 return Fintern (empty_unibyte_string, Qnil);
2942 /* Reader forms that can reuse previously read objects. */
2943 if (c >= '0' && c <= '9')
2945 EMACS_INT n = 0;
2946 Lisp_Object tem;
2947 bool overflow = false;
2949 /* Read a non-negative integer. */
2950 while (c >= '0' && c <= '9')
2952 overflow |= INT_MULTIPLY_WRAPV (n, 10, &n);
2953 overflow |= INT_ADD_WRAPV (n, c - '0', &n);
2954 c = READCHAR;
2957 if (!overflow && n <= MOST_POSITIVE_FIXNUM)
2959 if (c == 'r' || c == 'R')
2960 return read_integer (readcharfun, n);
2962 if (! NILP (Vread_circle))
2964 /* #n=object returns object, but associates it with
2965 n for #n#. */
2966 if (c == '=')
2968 /* Make a placeholder for #n# to use temporarily. */
2969 /* Note: We used to use AUTO_CONS to allocate
2970 placeholder, but that is a bad idea, since it
2971 will place a stack-allocated cons cell into
2972 the list in read_objects, which is a
2973 staticpro'd global variable, and thus each of
2974 its elements is marked during each GC. A
2975 stack-allocated object will become garbled
2976 when its stack slot goes out of scope, and
2977 some other function reuses it for entirely
2978 different purposes, which will cause crashes
2979 in GC. */
2980 Lisp_Object placeholder = Fcons (Qnil, Qnil);
2981 Lisp_Object cell = Fcons (make_number (n), placeholder);
2982 read_objects = Fcons (cell, read_objects);
2984 /* Read the object itself. */
2985 tem = read0 (readcharfun);
2987 /* Now put it everywhere the placeholder was... */
2988 Fsubstitute_object_in_subtree (tem, placeholder);
2990 /* ...and #n# will use the real value from now on. */
2991 Fsetcdr (cell, tem);
2993 return tem;
2996 /* #n# returns a previously read object. */
2997 if (c == '#')
2999 tem = Fassq (make_number (n), read_objects);
3000 if (CONSP (tem))
3001 return XCDR (tem);
3005 /* Fall through to error message. */
3007 else if (c == 'x' || c == 'X')
3008 return read_integer (readcharfun, 16);
3009 else if (c == 'o' || c == 'O')
3010 return read_integer (readcharfun, 8);
3011 else if (c == 'b' || c == 'B')
3012 return read_integer (readcharfun, 2);
3014 UNREAD (c);
3015 invalid_syntax ("#");
3017 case ';':
3018 while ((c = READCHAR) >= 0 && c != '\n');
3019 goto retry;
3021 case '\'':
3022 return list2 (Qquote, read0 (readcharfun));
3024 case '`':
3026 int next_char = READCHAR;
3027 UNREAD (next_char);
3028 /* Transition from old-style to new-style:
3029 If we see "(`" it used to mean old-style, which usually works
3030 fine because ` should almost never appear in such a position
3031 for new-style. But occasionally we need "(`" to mean new
3032 style, so we try to distinguish the two by the fact that we
3033 can either write "( `foo" or "(` foo", where the first
3034 intends to use new-style whereas the second intends to use
3035 old-style. For Emacs-25, we should completely remove this
3036 first_in_list exception (old-style can still be obtained via
3037 "(\`" anyway). */
3038 if (!new_backquote_flag && first_in_list && next_char == ' ')
3040 Vold_style_backquotes = Qt;
3041 goto default_label;
3043 else
3045 Lisp_Object value;
3046 bool saved_new_backquote_flag = new_backquote_flag;
3048 new_backquote_flag = 1;
3049 value = read0 (readcharfun);
3050 new_backquote_flag = saved_new_backquote_flag;
3052 return list2 (Qbackquote, value);
3055 case ',':
3057 int next_char = READCHAR;
3058 UNREAD (next_char);
3059 /* Transition from old-style to new-style:
3060 It used to be impossible to have a new-style , other than within
3061 a new-style `. This is sufficient when ` and , are used in the
3062 normal way, but ` and , can also appear in args to macros that
3063 will not interpret them in the usual way, in which case , may be
3064 used without any ` anywhere near.
3065 So we now use the same heuristic as for backquote: old-style
3066 unquotes are only recognized when first on a list, and when
3067 followed by a space.
3068 Because it's more difficult to peek 2 chars ahead, a new-style
3069 ,@ can still not be used outside of a `, unless it's in the middle
3070 of a list. */
3071 if (new_backquote_flag
3072 || !first_in_list
3073 || (next_char != ' ' && next_char != '@'))
3075 Lisp_Object comma_type = Qnil;
3076 Lisp_Object value;
3077 int ch = READCHAR;
3079 if (ch == '@')
3080 comma_type = Qcomma_at;
3081 else if (ch == '.')
3082 comma_type = Qcomma_dot;
3083 else
3085 if (ch >= 0) UNREAD (ch);
3086 comma_type = Qcomma;
3089 value = read0 (readcharfun);
3090 return list2 (comma_type, value);
3092 else
3094 Vold_style_backquotes = Qt;
3095 goto default_label;
3098 case '?':
3100 int modifiers;
3101 int next_char;
3102 bool ok;
3104 c = READCHAR;
3105 if (c < 0)
3106 end_of_file_error ();
3108 /* Accept `single space' syntax like (list ? x) where the
3109 whitespace character is SPC or TAB.
3110 Other literal whitespace like NL, CR, and FF are not accepted,
3111 as there are well-established escape sequences for these. */
3112 if (c == ' ' || c == '\t')
3113 return make_number (c);
3115 if (c == '(' || c == ')' || c == '[' || c == ']'
3116 || c == '"' || c == ';')
3118 CHECK_LIST (Vlread_unescaped_character_literals);
3119 Lisp_Object char_obj = make_natnum (c);
3120 if (NILP (Fmemq (char_obj, Vlread_unescaped_character_literals)))
3121 Vlread_unescaped_character_literals =
3122 Fcons (char_obj, Vlread_unescaped_character_literals);
3125 if (c == '\\')
3126 c = read_escape (readcharfun, 0);
3127 modifiers = c & CHAR_MODIFIER_MASK;
3128 c &= ~CHAR_MODIFIER_MASK;
3129 if (CHAR_BYTE8_P (c))
3130 c = CHAR_TO_BYTE8 (c);
3131 c |= modifiers;
3133 next_char = READCHAR;
3134 ok = (next_char <= 040
3135 || (next_char < 0200
3136 && strchr ("\"';()[]#?`,.", next_char) != NULL));
3137 UNREAD (next_char);
3138 if (ok)
3139 return make_number (c);
3141 invalid_syntax ("?");
3144 case '"':
3146 ptrdiff_t count = SPECPDL_INDEX ();
3147 char *read_buffer = stackbuf;
3148 ptrdiff_t read_buffer_size = sizeof stackbuf;
3149 char *heapbuf = NULL;
3150 char *p = read_buffer;
3151 char *end = read_buffer + read_buffer_size;
3152 int ch;
3153 /* True if we saw an escape sequence specifying
3154 a multibyte character. */
3155 bool force_multibyte = false;
3156 /* True if we saw an escape sequence specifying
3157 a single-byte character. */
3158 bool force_singlebyte = false;
3159 bool cancel = false;
3160 ptrdiff_t nchars = 0;
3162 while ((ch = READCHAR) >= 0
3163 && ch != '\"')
3165 if (end - p < MAX_MULTIBYTE_LENGTH)
3167 ptrdiff_t offset = p - read_buffer;
3168 read_buffer = grow_read_buffer (read_buffer, offset,
3169 &heapbuf, &read_buffer_size,
3170 count);
3171 p = read_buffer + offset;
3172 end = read_buffer + read_buffer_size;
3175 if (ch == '\\')
3177 int modifiers;
3179 ch = read_escape (readcharfun, 1);
3181 /* CH is -1 if \ newline or \ space has just been seen. */
3182 if (ch == -1)
3184 if (p == read_buffer)
3185 cancel = true;
3186 continue;
3189 modifiers = ch & CHAR_MODIFIER_MASK;
3190 ch = ch & ~CHAR_MODIFIER_MASK;
3192 if (CHAR_BYTE8_P (ch))
3193 force_singlebyte = true;
3194 else if (! ASCII_CHAR_P (ch))
3195 force_multibyte = true;
3196 else /* I.e. ASCII_CHAR_P (ch). */
3198 /* Allow `\C- ' and `\C-?'. */
3199 if (modifiers == CHAR_CTL)
3201 if (ch == ' ')
3202 ch = 0, modifiers = 0;
3203 else if (ch == '?')
3204 ch = 127, modifiers = 0;
3206 if (modifiers & CHAR_SHIFT)
3208 /* Shift modifier is valid only with [A-Za-z]. */
3209 if (ch >= 'A' && ch <= 'Z')
3210 modifiers &= ~CHAR_SHIFT;
3211 else if (ch >= 'a' && ch <= 'z')
3212 ch -= ('a' - 'A'), modifiers &= ~CHAR_SHIFT;
3215 if (modifiers & CHAR_META)
3217 /* Move the meta bit to the right place for a
3218 string. */
3219 modifiers &= ~CHAR_META;
3220 ch = BYTE8_TO_CHAR (ch | 0x80);
3221 force_singlebyte = true;
3225 /* Any modifiers remaining are invalid. */
3226 if (modifiers)
3227 error ("Invalid modifier in string");
3228 p += CHAR_STRING (ch, (unsigned char *) p);
3230 else
3232 p += CHAR_STRING (ch, (unsigned char *) p);
3233 if (CHAR_BYTE8_P (ch))
3234 force_singlebyte = true;
3235 else if (! ASCII_CHAR_P (ch))
3236 force_multibyte = true;
3238 nchars++;
3241 if (ch < 0)
3242 end_of_file_error ();
3244 /* If purifying, and string starts with \ newline,
3245 return zero instead. This is for doc strings
3246 that we are really going to find in etc/DOC.nn.nn. */
3247 if (!NILP (Vpurify_flag) && NILP (Vdoc_file_name) && cancel)
3248 return unbind_to (count, make_number (0));
3250 if (! force_multibyte && force_singlebyte)
3252 /* READ_BUFFER contains raw 8-bit bytes and no multibyte
3253 forms. Convert it to unibyte. */
3254 nchars = str_as_unibyte ((unsigned char *) read_buffer,
3255 p - read_buffer);
3256 p = read_buffer + nchars;
3259 Lisp_Object result
3260 = make_specified_string (read_buffer, nchars, p - read_buffer,
3261 (force_multibyte
3262 || (p - read_buffer != nchars)));
3263 return unbind_to (count, result);
3266 case '.':
3268 int next_char = READCHAR;
3269 UNREAD (next_char);
3271 if (next_char <= 040
3272 || (next_char < 0200
3273 && strchr ("\"';([#?`,", next_char) != NULL))
3275 *pch = c;
3276 return Qnil;
3279 /* Otherwise, we fall through! Note that the atom-reading loop
3280 below will now loop at least once, assuring that we will not
3281 try to UNREAD two characters in a row. */
3283 default:
3284 default_label:
3285 if (c <= 040) goto retry;
3286 if (c == NO_BREAK_SPACE)
3287 goto retry;
3289 read_symbol:
3291 ptrdiff_t count = SPECPDL_INDEX ();
3292 char *read_buffer = stackbuf;
3293 ptrdiff_t read_buffer_size = sizeof stackbuf;
3294 char *heapbuf = NULL;
3295 char *p = read_buffer;
3296 char *end = read_buffer + read_buffer_size;
3297 bool quoted = false;
3298 EMACS_INT start_position = readchar_count - 1;
3302 if (end - p < MAX_MULTIBYTE_LENGTH + 1)
3304 ptrdiff_t offset = p - read_buffer;
3305 read_buffer = grow_read_buffer (read_buffer, offset,
3306 &heapbuf, &read_buffer_size,
3307 count);
3308 p = read_buffer + offset;
3309 end = read_buffer + read_buffer_size;
3312 if (c == '\\')
3314 c = READCHAR;
3315 if (c == -1)
3316 end_of_file_error ();
3317 quoted = true;
3320 if (multibyte)
3321 p += CHAR_STRING (c, (unsigned char *) p);
3322 else
3323 *p++ = c;
3324 c = READCHAR;
3326 while (c > 040
3327 && c != NO_BREAK_SPACE
3328 && (c >= 0200
3329 || strchr ("\"';()[]#`,", c) == NULL));
3331 *p = 0;
3332 UNREAD (c);
3334 if (!quoted && !uninterned_symbol)
3336 Lisp_Object result = string_to_number (read_buffer, 10, 0);
3337 if (! NILP (result))
3338 return unbind_to (count, result);
3341 ptrdiff_t nbytes = p - read_buffer;
3342 ptrdiff_t nchars
3343 = (multibyte
3344 ? multibyte_chars_in_text ((unsigned char *) read_buffer,
3345 nbytes)
3346 : nbytes);
3347 Lisp_Object name = ((uninterned_symbol && ! NILP (Vpurify_flag)
3348 ? make_pure_string : make_specified_string)
3349 (read_buffer, nchars, nbytes, multibyte));
3350 Lisp_Object result = (uninterned_symbol ? Fmake_symbol (name)
3351 : Fintern (name, Qnil));
3353 if (EQ (Vread_with_symbol_positions, Qt)
3354 || EQ (Vread_with_symbol_positions, readcharfun))
3355 Vread_symbol_positions_list
3356 = Fcons (Fcons (result, make_number (start_position)),
3357 Vread_symbol_positions_list);
3358 return unbind_to (count, result);
3364 /* List of nodes we've seen during substitute_object_in_subtree. */
3365 static Lisp_Object seen_list;
3367 DEFUN ("substitute-object-in-subtree", Fsubstitute_object_in_subtree,
3368 Ssubstitute_object_in_subtree, 2, 2, 0,
3369 doc: /* Replace every reference to PLACEHOLDER in OBJECT with OBJECT. */)
3370 (Lisp_Object object, Lisp_Object placeholder)
3372 Lisp_Object check_object;
3374 /* We haven't seen any objects when we start. */
3375 seen_list = Qnil;
3377 /* Make all the substitutions. */
3378 check_object
3379 = substitute_object_recurse (object, placeholder, object);
3381 /* Clear seen_list because we're done with it. */
3382 seen_list = Qnil;
3384 /* The returned object here is expected to always eq the
3385 original. */
3386 if (!EQ (check_object, object))
3387 error ("Unexpected mutation error in reader");
3388 return Qnil;
3391 /* Feval doesn't get called from here, so no gc protection is needed. */
3392 #define SUBSTITUTE(get_val, set_val) \
3393 do { \
3394 Lisp_Object old_value = get_val; \
3395 Lisp_Object true_value \
3396 = substitute_object_recurse (object, placeholder, \
3397 old_value); \
3399 if (!EQ (old_value, true_value)) \
3401 set_val; \
3403 } while (0)
3405 static Lisp_Object
3406 substitute_object_recurse (Lisp_Object object, Lisp_Object placeholder, Lisp_Object subtree)
3408 /* If we find the placeholder, return the target object. */
3409 if (EQ (placeholder, subtree))
3410 return object;
3412 /* If we've been to this node before, don't explore it again. */
3413 if (!EQ (Qnil, Fmemq (subtree, seen_list)))
3414 return subtree;
3416 /* If this node can be the entry point to a cycle, remember that
3417 we've seen it. It can only be such an entry point if it was made
3418 by #n=, which means that we can find it as a value in
3419 read_objects. */
3420 if (!EQ (Qnil, Frassq (subtree, read_objects)))
3421 seen_list = Fcons (subtree, seen_list);
3423 /* Recurse according to subtree's type.
3424 Every branch must return a Lisp_Object. */
3425 switch (XTYPE (subtree))
3427 case Lisp_Vectorlike:
3429 ptrdiff_t i = 0, length = 0;
3430 if (BOOL_VECTOR_P (subtree))
3431 return subtree; /* No sub-objects anyway. */
3432 else if (CHAR_TABLE_P (subtree) || SUB_CHAR_TABLE_P (subtree)
3433 || COMPILEDP (subtree) || HASH_TABLE_P (subtree)
3434 || RECORDP (subtree))
3435 length = PVSIZE (subtree);
3436 else if (VECTORP (subtree))
3437 length = ASIZE (subtree);
3438 else
3439 /* An unknown pseudovector may contain non-Lisp fields, so we
3440 can't just blindly traverse all its fields. We used to call
3441 `Flength' which signaled `sequencep', so I just preserved this
3442 behavior. */
3443 wrong_type_argument (Qsequencep, subtree);
3445 if (SUB_CHAR_TABLE_P (subtree))
3446 i = 2;
3447 for ( ; i < length; i++)
3448 SUBSTITUTE (AREF (subtree, i),
3449 ASET (subtree, i, true_value));
3450 return subtree;
3453 case Lisp_Cons:
3455 SUBSTITUTE (XCAR (subtree),
3456 XSETCAR (subtree, true_value));
3457 SUBSTITUTE (XCDR (subtree),
3458 XSETCDR (subtree, true_value));
3459 return subtree;
3462 case Lisp_String:
3464 /* Check for text properties in each interval.
3465 substitute_in_interval contains part of the logic. */
3467 INTERVAL root_interval = string_intervals (subtree);
3468 AUTO_CONS (arg, object, placeholder);
3470 traverse_intervals_noorder (root_interval,
3471 &substitute_in_interval, arg);
3473 return subtree;
3476 /* Other types don't recurse any further. */
3477 default:
3478 return subtree;
3482 /* Helper function for substitute_object_recurse. */
3483 static void
3484 substitute_in_interval (INTERVAL interval, Lisp_Object arg)
3486 Lisp_Object object = Fcar (arg);
3487 Lisp_Object placeholder = Fcdr (arg);
3489 SUBSTITUTE (interval->plist, set_interval_plist (interval, true_value));
3493 #define LEAD_INT 1
3494 #define DOT_CHAR 2
3495 #define TRAIL_INT 4
3496 #define E_EXP 16
3499 /* Convert STRING to a number, assuming base BASE. Return a fixnum if CP has
3500 integer syntax and fits in a fixnum, else return the nearest float if CP has
3501 either floating point or integer syntax and BASE is 10, else return nil. If
3502 IGNORE_TRAILING, consider just the longest prefix of CP that has
3503 valid floating point syntax. Signal an overflow if BASE is not 10 and the
3504 number has integer syntax but does not fit. */
3506 Lisp_Object
3507 string_to_number (char const *string, int base, bool ignore_trailing)
3509 int state;
3510 char const *cp = string;
3511 int leading_digit;
3512 bool float_syntax = 0;
3513 double value = 0;
3515 /* Negate the value ourselves. This treats 0, NaNs, and infinity properly on
3516 IEEE floating point hosts, and works around a formerly-common bug where
3517 atof ("-0.0") drops the sign. */
3518 bool negative = *cp == '-';
3520 bool signedp = negative || *cp == '+';
3521 cp += signedp;
3523 state = 0;
3525 leading_digit = digit_to_number (*cp, base);
3526 if (leading_digit >= 0)
3528 state |= LEAD_INT;
3530 ++cp;
3531 while (digit_to_number (*cp, base) >= 0);
3533 if (*cp == '.')
3535 state |= DOT_CHAR;
3536 cp++;
3539 if (base == 10)
3541 if ('0' <= *cp && *cp <= '9')
3543 state |= TRAIL_INT;
3545 cp++;
3546 while ('0' <= *cp && *cp <= '9');
3548 if (*cp == 'e' || *cp == 'E')
3550 char const *ecp = cp;
3551 cp++;
3552 if (*cp == '+' || *cp == '-')
3553 cp++;
3554 if ('0' <= *cp && *cp <= '9')
3556 state |= E_EXP;
3558 cp++;
3559 while ('0' <= *cp && *cp <= '9');
3561 else if (cp[-1] == '+'
3562 && cp[0] == 'I' && cp[1] == 'N' && cp[2] == 'F')
3564 state |= E_EXP;
3565 cp += 3;
3566 value = INFINITY;
3568 else if (cp[-1] == '+'
3569 && cp[0] == 'N' && cp[1] == 'a' && cp[2] == 'N')
3571 state |= E_EXP;
3572 cp += 3;
3573 /* NAN is a "positive" NaN on all known Emacs hosts. */
3574 value = NAN;
3576 else
3577 cp = ecp;
3580 float_syntax = ((state & (DOT_CHAR|TRAIL_INT)) == (DOT_CHAR|TRAIL_INT)
3581 || state == (LEAD_INT|E_EXP));
3584 /* Return nil if the number uses invalid syntax. If IGNORE_TRAILING, accept
3585 any prefix that matches. Otherwise, the entire string must match. */
3586 if (! (ignore_trailing
3587 ? ((state & LEAD_INT) != 0 || float_syntax)
3588 : (!*cp && ((state & ~DOT_CHAR) == LEAD_INT || float_syntax))))
3589 return Qnil;
3591 /* If the number uses integer and not float syntax, and is in C-language
3592 range, use its value, preferably as a fixnum. */
3593 if (leading_digit >= 0 && ! float_syntax)
3595 uintmax_t n;
3597 /* Fast special case for single-digit integers. This also avoids a
3598 glitch when BASE is 16 and IGNORE_TRAILING, because in that
3599 case some versions of strtoumax accept numbers like "0x1" that Emacs
3600 does not allow. */
3601 if (digit_to_number (string[signedp + 1], base) < 0)
3602 return make_number (negative ? -leading_digit : leading_digit);
3604 errno = 0;
3605 n = strtoumax (string + signedp, NULL, base);
3606 if (errno == ERANGE)
3608 /* Unfortunately there's no simple and accurate way to convert
3609 non-base-10 numbers that are out of C-language range. */
3610 if (base != 10)
3611 xsignal1 (Qoverflow_error, build_string (string));
3613 else if (n <= (negative ? -MOST_NEGATIVE_FIXNUM : MOST_POSITIVE_FIXNUM))
3615 EMACS_INT signed_n = n;
3616 return make_number (negative ? -signed_n : signed_n);
3618 else
3619 value = n;
3622 /* Either the number uses float syntax, or it does not fit into a fixnum.
3623 Convert it from string to floating point, unless the value is already
3624 known because it is an infinity, a NAN, or its absolute value fits in
3625 uintmax_t. */
3626 if (! value)
3627 value = atof (string + signedp);
3629 return make_float (negative ? -value : value);
3633 static Lisp_Object
3634 read_vector (Lisp_Object readcharfun, bool bytecodeflag)
3636 ptrdiff_t i, size;
3637 Lisp_Object *ptr;
3638 Lisp_Object tem, item, vector;
3639 struct Lisp_Cons *otem;
3640 Lisp_Object len;
3642 tem = read_list (1, readcharfun);
3643 len = Flength (tem);
3644 vector = Fmake_vector (len, Qnil);
3646 size = ASIZE (vector);
3647 ptr = XVECTOR (vector)->contents;
3648 for (i = 0; i < size; i++)
3650 item = Fcar (tem);
3651 /* If `load-force-doc-strings' is t when reading a lazily-loaded
3652 bytecode object, the docstring containing the bytecode and
3653 constants values must be treated as unibyte and passed to
3654 Fread, to get the actual bytecode string and constants vector. */
3655 if (bytecodeflag && load_force_doc_strings)
3657 if (i == COMPILED_BYTECODE)
3659 if (!STRINGP (item))
3660 error ("Invalid byte code");
3662 /* Delay handling the bytecode slot until we know whether
3663 it is lazily-loaded (we can tell by whether the
3664 constants slot is nil). */
3665 ASET (vector, COMPILED_CONSTANTS, item);
3666 item = Qnil;
3668 else if (i == COMPILED_CONSTANTS)
3670 Lisp_Object bytestr = ptr[COMPILED_CONSTANTS];
3672 if (NILP (item))
3674 /* Coerce string to unibyte (like string-as-unibyte,
3675 but without generating extra garbage and
3676 guaranteeing no change in the contents). */
3677 STRING_SET_CHARS (bytestr, SBYTES (bytestr));
3678 STRING_SET_UNIBYTE (bytestr);
3680 item = Fread (Fcons (bytestr, readcharfun));
3681 if (!CONSP (item))
3682 error ("Invalid byte code");
3684 otem = XCONS (item);
3685 bytestr = XCAR (item);
3686 item = XCDR (item);
3687 free_cons (otem);
3690 /* Now handle the bytecode slot. */
3691 ASET (vector, COMPILED_BYTECODE, bytestr);
3693 else if (i == COMPILED_DOC_STRING
3694 && STRINGP (item)
3695 && ! STRING_MULTIBYTE (item))
3697 if (EQ (readcharfun, Qget_emacs_mule_file_char))
3698 item = Fdecode_coding_string (item, Qemacs_mule, Qnil, Qnil);
3699 else
3700 item = Fstring_as_multibyte (item);
3703 ASET (vector, i, item);
3704 otem = XCONS (tem);
3705 tem = Fcdr (tem);
3706 free_cons (otem);
3708 return vector;
3711 /* FLAG means check for ']' to terminate rather than ')' and '.'. */
3713 static Lisp_Object
3714 read_list (bool flag, Lisp_Object readcharfun)
3716 Lisp_Object val, tail;
3717 Lisp_Object elt, tem;
3718 /* 0 is the normal case.
3719 1 means this list is a doc reference; replace it with the number 0.
3720 2 means this list is a doc reference; replace it with the doc string. */
3721 int doc_reference = 0;
3723 /* Initialize this to 1 if we are reading a list. */
3724 bool first_in_list = flag <= 0;
3726 val = Qnil;
3727 tail = Qnil;
3729 while (1)
3731 int ch;
3732 elt = read1 (readcharfun, &ch, first_in_list);
3734 first_in_list = 0;
3736 /* While building, if the list starts with #$, treat it specially. */
3737 if (EQ (elt, Vload_file_name)
3738 && ! NILP (elt)
3739 && !NILP (Vpurify_flag))
3741 if (NILP (Vdoc_file_name))
3742 /* We have not yet called Snarf-documentation, so assume
3743 this file is described in the DOC file
3744 and Snarf-documentation will fill in the right value later.
3745 For now, replace the whole list with 0. */
3746 doc_reference = 1;
3747 else
3748 /* We have already called Snarf-documentation, so make a relative
3749 file name for this file, so it can be found properly
3750 in the installed Lisp directory.
3751 We don't use Fexpand_file_name because that would make
3752 the directory absolute now. */
3754 AUTO_STRING (dot_dot_lisp, "../lisp/");
3755 elt = concat2 (dot_dot_lisp, Ffile_name_nondirectory (elt));
3758 else if (EQ (elt, Vload_file_name)
3759 && ! NILP (elt)
3760 && load_force_doc_strings)
3761 doc_reference = 2;
3763 if (ch)
3765 if (flag > 0)
3767 if (ch == ']')
3768 return val;
3769 invalid_syntax (") or . in a vector");
3771 if (ch == ')')
3772 return val;
3773 if (ch == '.')
3775 if (!NILP (tail))
3776 XSETCDR (tail, read0 (readcharfun));
3777 else
3778 val = read0 (readcharfun);
3779 read1 (readcharfun, &ch, 0);
3781 if (ch == ')')
3783 if (doc_reference == 1)
3784 return make_number (0);
3785 if (doc_reference == 2 && INTEGERP (XCDR (val)))
3787 char *saved = NULL;
3788 file_offset saved_position;
3789 /* Get a doc string from the file we are loading.
3790 If it's in saved_doc_string, get it from there.
3792 Here, we don't know if the string is a
3793 bytecode string or a doc string. As a
3794 bytecode string must be unibyte, we always
3795 return a unibyte string. If it is actually a
3796 doc string, caller must make it
3797 multibyte. */
3799 /* Position is negative for user variables. */
3800 EMACS_INT pos = eabs (XINT (XCDR (val)));
3801 if (pos >= saved_doc_string_position
3802 && pos < (saved_doc_string_position
3803 + saved_doc_string_length))
3805 saved = saved_doc_string;
3806 saved_position = saved_doc_string_position;
3808 /* Look in prev_saved_doc_string the same way. */
3809 else if (pos >= prev_saved_doc_string_position
3810 && pos < (prev_saved_doc_string_position
3811 + prev_saved_doc_string_length))
3813 saved = prev_saved_doc_string;
3814 saved_position = prev_saved_doc_string_position;
3816 if (saved)
3818 ptrdiff_t start = pos - saved_position;
3819 ptrdiff_t from, to;
3821 /* Process quoting with ^A,
3822 and find the end of the string,
3823 which is marked with ^_ (037). */
3824 for (from = start, to = start;
3825 saved[from] != 037;)
3827 int c = saved[from++];
3828 if (c == 1)
3830 c = saved[from++];
3831 saved[to++] = (c == 1 ? c
3832 : c == '0' ? 0
3833 : c == '_' ? 037
3834 : c);
3836 else
3837 saved[to++] = c;
3840 return make_unibyte_string (saved + start,
3841 to - start);
3843 else
3844 return get_doc_string (val, 1, 0);
3847 return val;
3849 invalid_syntax (". in wrong context");
3851 invalid_syntax ("] in a list");
3853 tem = list1 (elt);
3854 if (!NILP (tail))
3855 XSETCDR (tail, tem);
3856 else
3857 val = tem;
3858 tail = tem;
3862 static Lisp_Object initial_obarray;
3864 /* `oblookup' stores the bucket number here, for the sake of Funintern. */
3866 static size_t oblookup_last_bucket_number;
3868 /* Get an error if OBARRAY is not an obarray.
3869 If it is one, return it. */
3871 Lisp_Object
3872 check_obarray (Lisp_Object obarray)
3874 /* We don't want to signal a wrong-type-argument error when we are
3875 shutting down due to a fatal error, and we don't want to hit
3876 assertions in VECTORP and ASIZE if the fatal error was during GC. */
3877 if (!fatal_error_in_progress
3878 && (!VECTORP (obarray) || ASIZE (obarray) == 0))
3880 /* If Vobarray is now invalid, force it to be valid. */
3881 if (EQ (Vobarray, obarray)) Vobarray = initial_obarray;
3882 wrong_type_argument (Qvectorp, obarray);
3884 return obarray;
3887 /* Intern symbol SYM in OBARRAY using bucket INDEX. */
3889 static Lisp_Object
3890 intern_sym (Lisp_Object sym, Lisp_Object obarray, Lisp_Object index)
3892 Lisp_Object *ptr;
3894 XSYMBOL (sym)->interned = (EQ (obarray, initial_obarray)
3895 ? SYMBOL_INTERNED_IN_INITIAL_OBARRAY
3896 : SYMBOL_INTERNED);
3898 if (SREF (SYMBOL_NAME (sym), 0) == ':' && EQ (obarray, initial_obarray))
3900 make_symbol_constant (sym);
3901 XSYMBOL (sym)->redirect = SYMBOL_PLAINVAL;
3902 SET_SYMBOL_VAL (XSYMBOL (sym), sym);
3905 ptr = aref_addr (obarray, XINT (index));
3906 set_symbol_next (sym, SYMBOLP (*ptr) ? XSYMBOL (*ptr) : NULL);
3907 *ptr = sym;
3908 return sym;
3911 /* Intern a symbol with name STRING in OBARRAY using bucket INDEX. */
3913 Lisp_Object
3914 intern_driver (Lisp_Object string, Lisp_Object obarray, Lisp_Object index)
3916 return intern_sym (Fmake_symbol (string), obarray, index);
3919 /* Intern the C string STR: return a symbol with that name,
3920 interned in the current obarray. */
3922 Lisp_Object
3923 intern_1 (const char *str, ptrdiff_t len)
3925 Lisp_Object obarray = check_obarray (Vobarray);
3926 Lisp_Object tem = oblookup (obarray, str, len, len);
3928 return (SYMBOLP (tem) ? tem
3929 /* The above `oblookup' was done on the basis of nchars==nbytes, so
3930 the string has to be unibyte. */
3931 : intern_driver (make_unibyte_string (str, len),
3932 obarray, tem));
3935 Lisp_Object
3936 intern_c_string_1 (const char *str, ptrdiff_t len)
3938 Lisp_Object obarray = check_obarray (Vobarray);
3939 Lisp_Object tem = oblookup (obarray, str, len, len);
3941 if (!SYMBOLP (tem))
3943 /* Creating a non-pure string from a string literal not implemented yet.
3944 We could just use make_string here and live with the extra copy. */
3945 eassert (!NILP (Vpurify_flag));
3946 tem = intern_driver (make_pure_c_string (str, len), obarray, tem);
3948 return tem;
3951 static void
3952 define_symbol (Lisp_Object sym, char const *str)
3954 ptrdiff_t len = strlen (str);
3955 Lisp_Object string = make_pure_c_string (str, len);
3956 init_symbol (sym, string);
3958 /* Qunbound is uninterned, so that it's not confused with any symbol
3959 'unbound' created by a Lisp program. */
3960 if (! EQ (sym, Qunbound))
3962 Lisp_Object bucket = oblookup (initial_obarray, str, len, len);
3963 eassert (INTEGERP (bucket));
3964 intern_sym (sym, initial_obarray, bucket);
3968 DEFUN ("intern", Fintern, Sintern, 1, 2, 0,
3969 doc: /* Return the canonical symbol whose name is STRING.
3970 If there is none, one is created by this function and returned.
3971 A second optional argument specifies the obarray to use;
3972 it defaults to the value of `obarray'. */)
3973 (Lisp_Object string, Lisp_Object obarray)
3975 Lisp_Object tem;
3977 obarray = check_obarray (NILP (obarray) ? Vobarray : obarray);
3978 CHECK_STRING (string);
3980 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3981 if (!SYMBOLP (tem))
3982 tem = intern_driver (NILP (Vpurify_flag) ? string : Fpurecopy (string),
3983 obarray, tem);
3984 return tem;
3987 DEFUN ("intern-soft", Fintern_soft, Sintern_soft, 1, 2, 0,
3988 doc: /* Return the canonical symbol named NAME, or nil if none exists.
3989 NAME may be a string or a symbol. If it is a symbol, that exact
3990 symbol is searched for.
3991 A second optional argument specifies the obarray to use;
3992 it defaults to the value of `obarray'. */)
3993 (Lisp_Object name, Lisp_Object obarray)
3995 register Lisp_Object tem, string;
3997 if (NILP (obarray)) obarray = Vobarray;
3998 obarray = check_obarray (obarray);
4000 if (!SYMBOLP (name))
4002 CHECK_STRING (name);
4003 string = name;
4005 else
4006 string = SYMBOL_NAME (name);
4008 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
4009 if (INTEGERP (tem) || (SYMBOLP (name) && !EQ (name, tem)))
4010 return Qnil;
4011 else
4012 return tem;
4015 DEFUN ("unintern", Funintern, Sunintern, 1, 2, 0,
4016 doc: /* Delete the symbol named NAME, if any, from OBARRAY.
4017 The value is t if a symbol was found and deleted, nil otherwise.
4018 NAME may be a string or a symbol. If it is a symbol, that symbol
4019 is deleted, if it belongs to OBARRAY--no other symbol is deleted.
4020 OBARRAY, if nil, defaults to the value of the variable `obarray'.
4021 usage: (unintern NAME OBARRAY) */)
4022 (Lisp_Object name, Lisp_Object obarray)
4024 register Lisp_Object string, tem;
4025 size_t hash;
4027 if (NILP (obarray)) obarray = Vobarray;
4028 obarray = check_obarray (obarray);
4030 if (SYMBOLP (name))
4031 string = SYMBOL_NAME (name);
4032 else
4034 CHECK_STRING (name);
4035 string = name;
4038 tem = oblookup (obarray, SSDATA (string),
4039 SCHARS (string),
4040 SBYTES (string));
4041 if (INTEGERP (tem))
4042 return Qnil;
4043 /* If arg was a symbol, don't delete anything but that symbol itself. */
4044 if (SYMBOLP (name) && !EQ (name, tem))
4045 return Qnil;
4047 /* There are plenty of other symbols which will screw up the Emacs
4048 session if we unintern them, as well as even more ways to use
4049 `setq' or `fset' or whatnot to make the Emacs session
4050 unusable. Let's not go down this silly road. --Stef */
4051 /* if (EQ (tem, Qnil) || EQ (tem, Qt))
4052 error ("Attempt to unintern t or nil"); */
4054 XSYMBOL (tem)->interned = SYMBOL_UNINTERNED;
4056 hash = oblookup_last_bucket_number;
4058 if (EQ (AREF (obarray, hash), tem))
4060 if (XSYMBOL (tem)->next)
4062 Lisp_Object sym;
4063 XSETSYMBOL (sym, XSYMBOL (tem)->next);
4064 ASET (obarray, hash, sym);
4066 else
4067 ASET (obarray, hash, make_number (0));
4069 else
4071 Lisp_Object tail, following;
4073 for (tail = AREF (obarray, hash);
4074 XSYMBOL (tail)->next;
4075 tail = following)
4077 XSETSYMBOL (following, XSYMBOL (tail)->next);
4078 if (EQ (following, tem))
4080 set_symbol_next (tail, XSYMBOL (following)->next);
4081 break;
4086 return Qt;
4089 /* Return the symbol in OBARRAY whose names matches the string
4090 of SIZE characters (SIZE_BYTE bytes) at PTR.
4091 If there is no such symbol, return the integer bucket number of
4092 where the symbol would be if it were present.
4094 Also store the bucket number in oblookup_last_bucket_number. */
4096 Lisp_Object
4097 oblookup (Lisp_Object obarray, register const char *ptr, ptrdiff_t size, ptrdiff_t size_byte)
4099 size_t hash;
4100 size_t obsize;
4101 register Lisp_Object tail;
4102 Lisp_Object bucket, tem;
4104 obarray = check_obarray (obarray);
4105 /* This is sometimes needed in the middle of GC. */
4106 obsize = gc_asize (obarray);
4107 hash = hash_string (ptr, size_byte) % obsize;
4108 bucket = AREF (obarray, hash);
4109 oblookup_last_bucket_number = hash;
4110 if (EQ (bucket, make_number (0)))
4112 else if (!SYMBOLP (bucket))
4113 error ("Bad data in guts of obarray"); /* Like CADR error message. */
4114 else
4115 for (tail = bucket; ; XSETSYMBOL (tail, XSYMBOL (tail)->next))
4117 if (SBYTES (SYMBOL_NAME (tail)) == size_byte
4118 && SCHARS (SYMBOL_NAME (tail)) == size
4119 && !memcmp (SDATA (SYMBOL_NAME (tail)), ptr, size_byte))
4120 return tail;
4121 else if (XSYMBOL (tail)->next == 0)
4122 break;
4124 XSETINT (tem, hash);
4125 return tem;
4128 void
4129 map_obarray (Lisp_Object obarray, void (*fn) (Lisp_Object, Lisp_Object), Lisp_Object arg)
4131 ptrdiff_t i;
4132 register Lisp_Object tail;
4133 CHECK_VECTOR (obarray);
4134 for (i = ASIZE (obarray) - 1; i >= 0; i--)
4136 tail = AREF (obarray, i);
4137 if (SYMBOLP (tail))
4138 while (1)
4140 (*fn) (tail, arg);
4141 if (XSYMBOL (tail)->next == 0)
4142 break;
4143 XSETSYMBOL (tail, XSYMBOL (tail)->next);
4148 static void
4149 mapatoms_1 (Lisp_Object sym, Lisp_Object function)
4151 call1 (function, sym);
4154 DEFUN ("mapatoms", Fmapatoms, Smapatoms, 1, 2, 0,
4155 doc: /* Call FUNCTION on every symbol in OBARRAY.
4156 OBARRAY defaults to the value of `obarray'. */)
4157 (Lisp_Object function, Lisp_Object obarray)
4159 if (NILP (obarray)) obarray = Vobarray;
4160 obarray = check_obarray (obarray);
4162 map_obarray (obarray, mapatoms_1, function);
4163 return Qnil;
4166 #define OBARRAY_SIZE 15121
4168 void
4169 init_obarray (void)
4171 Vobarray = Fmake_vector (make_number (OBARRAY_SIZE), make_number (0));
4172 initial_obarray = Vobarray;
4173 staticpro (&initial_obarray);
4175 for (int i = 0; i < ARRAYELTS (lispsym); i++)
4176 define_symbol (builtin_lisp_symbol (i), defsym_name[i]);
4178 DEFSYM (Qunbound, "unbound");
4180 DEFSYM (Qnil, "nil");
4181 SET_SYMBOL_VAL (XSYMBOL (Qnil), Qnil);
4182 make_symbol_constant (Qnil);
4183 XSYMBOL (Qnil)->declared_special = true;
4185 DEFSYM (Qt, "t");
4186 SET_SYMBOL_VAL (XSYMBOL (Qt), Qt);
4187 make_symbol_constant (Qt);
4188 XSYMBOL (Qt)->declared_special = true;
4190 /* Qt is correct even if CANNOT_DUMP. loadup.el will set to nil at end. */
4191 Vpurify_flag = Qt;
4193 DEFSYM (Qvariable_documentation, "variable-documentation");
4196 void
4197 defsubr (struct Lisp_Subr *sname)
4199 Lisp_Object sym, tem;
4200 sym = intern_c_string (sname->symbol_name);
4201 XSETPVECTYPE (sname, PVEC_SUBR);
4202 XSETSUBR (tem, sname);
4203 set_symbol_function (sym, tem);
4206 #ifdef NOTDEF /* Use fset in subr.el now! */
4207 void
4208 defalias (struct Lisp_Subr *sname, char *string)
4210 Lisp_Object sym;
4211 sym = intern (string);
4212 XSETSUBR (XSYMBOL (sym)->function, sname);
4214 #endif /* NOTDEF */
4216 /* Define an "integer variable"; a symbol whose value is forwarded to a
4217 C variable of type EMACS_INT. Sample call (with "xx" to fool make-docfile):
4218 DEFxxVAR_INT ("emacs-priority", &emacs_priority, "Documentation"); */
4219 void
4220 defvar_int (struct Lisp_Intfwd *i_fwd,
4221 const char *namestring, EMACS_INT *address)
4223 Lisp_Object sym;
4224 sym = intern_c_string (namestring);
4225 i_fwd->type = Lisp_Fwd_Int;
4226 i_fwd->intvar = address;
4227 XSYMBOL (sym)->declared_special = 1;
4228 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4229 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)i_fwd);
4232 /* Similar but define a variable whose value is t if address contains 1,
4233 nil if address contains 0. */
4234 void
4235 defvar_bool (struct Lisp_Boolfwd *b_fwd,
4236 const char *namestring, bool *address)
4238 Lisp_Object sym;
4239 sym = intern_c_string (namestring);
4240 b_fwd->type = Lisp_Fwd_Bool;
4241 b_fwd->boolvar = address;
4242 XSYMBOL (sym)->declared_special = 1;
4243 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4244 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)b_fwd);
4245 Vbyte_boolean_vars = Fcons (sym, Vbyte_boolean_vars);
4248 /* Similar but define a variable whose value is the Lisp Object stored
4249 at address. Two versions: with and without gc-marking of the C
4250 variable. The nopro version is used when that variable will be
4251 gc-marked for some other reason, since marking the same slot twice
4252 can cause trouble with strings. */
4253 void
4254 defvar_lisp_nopro (struct Lisp_Objfwd *o_fwd,
4255 const char *namestring, Lisp_Object *address)
4257 Lisp_Object sym;
4258 sym = intern_c_string (namestring);
4259 o_fwd->type = Lisp_Fwd_Obj;
4260 o_fwd->objvar = address;
4261 XSYMBOL (sym)->declared_special = 1;
4262 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4263 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)o_fwd);
4266 void
4267 defvar_lisp (struct Lisp_Objfwd *o_fwd,
4268 const char *namestring, Lisp_Object *address)
4270 defvar_lisp_nopro (o_fwd, namestring, address);
4271 staticpro (address);
4274 /* Similar but define a variable whose value is the Lisp Object stored
4275 at a particular offset in the current kboard object. */
4277 void
4278 defvar_kboard (struct Lisp_Kboard_Objfwd *ko_fwd,
4279 const char *namestring, int offset)
4281 Lisp_Object sym;
4282 sym = intern_c_string (namestring);
4283 ko_fwd->type = Lisp_Fwd_Kboard_Obj;
4284 ko_fwd->offset = offset;
4285 XSYMBOL (sym)->declared_special = 1;
4286 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4287 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)ko_fwd);
4290 /* Check that the elements of lpath exist. */
4292 static void
4293 load_path_check (Lisp_Object lpath)
4295 Lisp_Object path_tail;
4297 /* The only elements that might not exist are those from
4298 PATH_LOADSEARCH, EMACSLOADPATH. Anything else is only added if
4299 it exists. */
4300 for (path_tail = lpath; !NILP (path_tail); path_tail = XCDR (path_tail))
4302 Lisp_Object dirfile;
4303 dirfile = Fcar (path_tail);
4304 if (STRINGP (dirfile))
4306 dirfile = Fdirectory_file_name (dirfile);
4307 if (! file_accessible_directory_p (dirfile))
4308 dir_warning ("Lisp directory", XCAR (path_tail));
4313 /* Return the default load-path, to be used if EMACSLOADPATH is unset.
4314 This does not include the standard site-lisp directories
4315 under the installation prefix (i.e., PATH_SITELOADSEARCH),
4316 but it does (unless no_site_lisp is set) include site-lisp
4317 directories in the source/build directories if those exist and we
4318 are running uninstalled.
4320 Uses the following logic:
4321 If CANNOT_DUMP:
4322 If Vinstallation_directory is not nil (ie, running uninstalled),
4323 use PATH_DUMPLOADSEARCH (ie, build path). Else use PATH_LOADSEARCH.
4324 The remainder is what happens when dumping works:
4325 If purify-flag (ie dumping) just use PATH_DUMPLOADSEARCH.
4326 Otherwise use PATH_LOADSEARCH.
4328 If !initialized, then just return PATH_DUMPLOADSEARCH.
4329 If initialized:
4330 If Vinstallation_directory is not nil (ie, running uninstalled):
4331 If installation-dir/lisp exists and not already a member,
4332 we must be running uninstalled. Reset the load-path
4333 to just installation-dir/lisp. (The default PATH_LOADSEARCH
4334 refers to the eventual installation directories. Since we
4335 are not yet installed, we should not use them, even if they exist.)
4336 If installation-dir/lisp does not exist, just add
4337 PATH_DUMPLOADSEARCH at the end instead.
4338 Add installation-dir/site-lisp (if !no_site_lisp, and exists
4339 and not already a member) at the front.
4340 If installation-dir != source-dir (ie running an uninstalled,
4341 out-of-tree build) AND install-dir/src/Makefile exists BUT
4342 install-dir/src/Makefile.in does NOT exist (this is a sanity
4343 check), then repeat the above steps for source-dir/lisp, site-lisp. */
4345 static Lisp_Object
4346 load_path_default (void)
4348 Lisp_Object lpath = Qnil;
4349 const char *normal;
4351 #ifdef CANNOT_DUMP
4352 #ifdef HAVE_NS
4353 const char *loadpath = ns_load_path ();
4354 #endif
4356 normal = PATH_LOADSEARCH;
4357 if (!NILP (Vinstallation_directory)) normal = PATH_DUMPLOADSEARCH;
4359 #ifdef HAVE_NS
4360 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4361 #else
4362 lpath = decode_env_path (0, normal, 0);
4363 #endif
4365 #else /* !CANNOT_DUMP */
4367 normal = NILP (Vpurify_flag) ? PATH_LOADSEARCH : PATH_DUMPLOADSEARCH;
4369 if (initialized)
4371 #ifdef HAVE_NS
4372 const char *loadpath = ns_load_path ();
4373 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4374 #else
4375 lpath = decode_env_path (0, normal, 0);
4376 #endif
4377 if (!NILP (Vinstallation_directory))
4379 Lisp_Object tem, tem1;
4381 /* Add to the path the lisp subdir of the installation
4382 dir, if it is accessible. Note: in out-of-tree builds,
4383 this directory is empty save for Makefile. */
4384 tem = Fexpand_file_name (build_string ("lisp"),
4385 Vinstallation_directory);
4386 tem1 = Ffile_accessible_directory_p (tem);
4387 if (!NILP (tem1))
4389 if (NILP (Fmember (tem, lpath)))
4391 /* We are running uninstalled. The default load-path
4392 points to the eventual installed lisp directories.
4393 We should not use those now, even if they exist,
4394 so start over from a clean slate. */
4395 lpath = list1 (tem);
4398 else
4399 /* That dir doesn't exist, so add the build-time
4400 Lisp dirs instead. */
4402 Lisp_Object dump_path =
4403 decode_env_path (0, PATH_DUMPLOADSEARCH, 0);
4404 lpath = nconc2 (lpath, dump_path);
4407 /* Add site-lisp under the installation dir, if it exists. */
4408 if (!no_site_lisp)
4410 tem = Fexpand_file_name (build_string ("site-lisp"),
4411 Vinstallation_directory);
4412 tem1 = Ffile_accessible_directory_p (tem);
4413 if (!NILP (tem1))
4415 if (NILP (Fmember (tem, lpath)))
4416 lpath = Fcons (tem, lpath);
4420 /* If Emacs was not built in the source directory,
4421 and it is run from where it was built, add to load-path
4422 the lisp and site-lisp dirs under that directory. */
4424 if (NILP (Fequal (Vinstallation_directory, Vsource_directory)))
4426 Lisp_Object tem2;
4428 tem = Fexpand_file_name (build_string ("src/Makefile"),
4429 Vinstallation_directory);
4430 tem1 = Ffile_exists_p (tem);
4432 /* Don't be fooled if they moved the entire source tree
4433 AFTER dumping Emacs. If the build directory is indeed
4434 different from the source dir, src/Makefile.in and
4435 src/Makefile will not be found together. */
4436 tem = Fexpand_file_name (build_string ("src/Makefile.in"),
4437 Vinstallation_directory);
4438 tem2 = Ffile_exists_p (tem);
4439 if (!NILP (tem1) && NILP (tem2))
4441 tem = Fexpand_file_name (build_string ("lisp"),
4442 Vsource_directory);
4444 if (NILP (Fmember (tem, lpath)))
4445 lpath = Fcons (tem, lpath);
4447 if (!no_site_lisp)
4449 tem = Fexpand_file_name (build_string ("site-lisp"),
4450 Vsource_directory);
4451 tem1 = Ffile_accessible_directory_p (tem);
4452 if (!NILP (tem1))
4454 if (NILP (Fmember (tem, lpath)))
4455 lpath = Fcons (tem, lpath);
4459 } /* Vinstallation_directory != Vsource_directory */
4461 } /* if Vinstallation_directory */
4463 else /* !initialized */
4465 /* NORMAL refers to PATH_DUMPLOADSEARCH, ie the lisp dir in the
4466 source directory. We used to add ../lisp (ie the lisp dir in
4467 the build directory) at the front here, but that should not
4468 be necessary, since in out of tree builds lisp/ is empty, save
4469 for Makefile. */
4470 lpath = decode_env_path (0, normal, 0);
4472 #endif /* !CANNOT_DUMP */
4474 return lpath;
4477 void
4478 init_lread (void)
4480 /* First, set Vload_path. */
4482 /* Ignore EMACSLOADPATH when dumping. */
4483 #ifdef CANNOT_DUMP
4484 bool use_loadpath = true;
4485 #else
4486 bool use_loadpath = NILP (Vpurify_flag);
4487 #endif
4489 if (use_loadpath && egetenv ("EMACSLOADPATH"))
4491 Vload_path = decode_env_path ("EMACSLOADPATH", 0, 1);
4493 /* Check (non-nil) user-supplied elements. */
4494 load_path_check (Vload_path);
4496 /* If no nils in the environment variable, use as-is.
4497 Otherwise, replace any nils with the default. */
4498 if (! NILP (Fmemq (Qnil, Vload_path)))
4500 Lisp_Object elem, elpath = Vload_path;
4501 Lisp_Object default_lpath = load_path_default ();
4503 /* Check defaults, before adding site-lisp. */
4504 load_path_check (default_lpath);
4506 /* Add the site-lisp directories to the front of the default. */
4507 if (!no_site_lisp && PATH_SITELOADSEARCH[0] != '\0')
4509 Lisp_Object sitelisp;
4510 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4511 if (! NILP (sitelisp))
4512 default_lpath = nconc2 (sitelisp, default_lpath);
4515 Vload_path = Qnil;
4517 /* Replace nils from EMACSLOADPATH by default. */
4518 while (CONSP (elpath))
4520 elem = XCAR (elpath);
4521 elpath = XCDR (elpath);
4522 Vload_path = CALLN (Fappend, Vload_path,
4523 NILP (elem) ? default_lpath : list1 (elem));
4525 } /* Fmemq (Qnil, Vload_path) */
4527 else
4529 Vload_path = load_path_default ();
4531 /* Check before adding site-lisp directories.
4532 The install should have created them, but they are not
4533 required, so no need to warn if they are absent.
4534 Or we might be running before installation. */
4535 load_path_check (Vload_path);
4537 /* Add the site-lisp directories at the front. */
4538 if (initialized && !no_site_lisp && PATH_SITELOADSEARCH[0] != '\0')
4540 Lisp_Object sitelisp;
4541 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4542 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4546 Vvalues = Qnil;
4548 load_in_progress = 0;
4549 Vload_file_name = Qnil;
4550 Vstandard_input = Qt;
4551 Vloads_in_progress = Qnil;
4554 /* Print a warning that directory intended for use USE and with name
4555 DIRNAME cannot be accessed. On entry, errno should correspond to
4556 the access failure. Print the warning on stderr and put it in
4557 *Messages*. */
4559 void
4560 dir_warning (char const *use, Lisp_Object dirname)
4562 static char const format[] = "Warning: %s '%s': %s\n";
4563 char *diagnostic = emacs_strerror (errno);
4564 fprintf (stderr, format, use, SSDATA (ENCODE_SYSTEM (dirname)), diagnostic);
4566 /* Don't log the warning before we've initialized!! */
4567 if (initialized)
4569 ptrdiff_t diaglen = strlen (diagnostic);
4570 AUTO_STRING_WITH_LEN (diag, diagnostic, diaglen);
4571 if (! NILP (Vlocale_coding_system))
4573 Lisp_Object s
4574 = code_convert_string_norecord (diag, Vlocale_coding_system, false);
4575 diagnostic = SSDATA (s);
4576 diaglen = SBYTES (s);
4578 USE_SAFE_ALLOCA;
4579 char *buffer = SAFE_ALLOCA (sizeof format - 3 * (sizeof "%s" - 1)
4580 + strlen (use) + SBYTES (dirname) + diaglen);
4581 ptrdiff_t message_len = esprintf (buffer, format, use, SSDATA (dirname),
4582 diagnostic);
4583 message_dolog (buffer, message_len, 0, STRING_MULTIBYTE (dirname));
4584 SAFE_FREE ();
4588 void
4589 syms_of_lread (void)
4591 defsubr (&Sread);
4592 defsubr (&Sread_from_string);
4593 defsubr (&Ssubstitute_object_in_subtree);
4594 defsubr (&Sintern);
4595 defsubr (&Sintern_soft);
4596 defsubr (&Sunintern);
4597 defsubr (&Sget_load_suffixes);
4598 defsubr (&Sload);
4599 defsubr (&Seval_buffer);
4600 defsubr (&Seval_region);
4601 defsubr (&Sread_char);
4602 defsubr (&Sread_char_exclusive);
4603 defsubr (&Sread_event);
4604 defsubr (&Sget_file_char);
4605 defsubr (&Smapatoms);
4606 defsubr (&Slocate_file_internal);
4608 DEFVAR_LISP ("obarray", Vobarray,
4609 doc: /* Symbol table for use by `intern' and `read'.
4610 It is a vector whose length ought to be prime for best results.
4611 The vector's contents don't make sense if examined from Lisp programs;
4612 to find all the symbols in an obarray, use `mapatoms'. */);
4614 DEFVAR_LISP ("values", Vvalues,
4615 doc: /* List of values of all expressions which were read, evaluated and printed.
4616 Order is reverse chronological. */);
4617 XSYMBOL (intern ("values"))->declared_special = 0;
4619 DEFVAR_LISP ("standard-input", Vstandard_input,
4620 doc: /* Stream for read to get input from.
4621 See documentation of `read' for possible values. */);
4622 Vstandard_input = Qt;
4624 DEFVAR_LISP ("read-with-symbol-positions", Vread_with_symbol_positions,
4625 doc: /* If non-nil, add position of read symbols to `read-symbol-positions-list'.
4627 If this variable is a buffer, then only forms read from that buffer
4628 will be added to `read-symbol-positions-list'.
4629 If this variable is t, then all read forms will be added.
4630 The effect of all other values other than nil are not currently
4631 defined, although they may be in the future.
4633 The positions are relative to the last call to `read' or
4634 `read-from-string'. It is probably a bad idea to set this variable at
4635 the toplevel; bind it instead. */);
4636 Vread_with_symbol_positions = Qnil;
4638 DEFVAR_LISP ("read-symbol-positions-list", Vread_symbol_positions_list,
4639 doc: /* A list mapping read symbols to their positions.
4640 This variable is modified during calls to `read' or
4641 `read-from-string', but only when `read-with-symbol-positions' is
4642 non-nil.
4644 Each element of the list looks like (SYMBOL . CHAR-POSITION), where
4645 CHAR-POSITION is an integer giving the offset of that occurrence of the
4646 symbol from the position where `read' or `read-from-string' started.
4648 Note that a symbol will appear multiple times in this list, if it was
4649 read multiple times. The list is in the same order as the symbols
4650 were read in. */);
4651 Vread_symbol_positions_list = Qnil;
4653 DEFVAR_LISP ("read-circle", Vread_circle,
4654 doc: /* Non-nil means read recursive structures using #N= and #N# syntax. */);
4655 Vread_circle = Qt;
4657 DEFVAR_LISP ("load-path", Vload_path,
4658 doc: /* List of directories to search for files to load.
4659 Each element is a string (directory file name) or nil (meaning
4660 `default-directory').
4661 This list is consulted by the `require' function.
4662 Initialized during startup as described in Info node `(elisp)Library Search'.
4663 Use `directory-file-name' when adding items to this path. However, Lisp
4664 programs that process this list should tolerate directories both with
4665 and without trailing slashes. */);
4667 DEFVAR_LISP ("load-suffixes", Vload_suffixes,
4668 doc: /* List of suffixes for Emacs Lisp files and dynamic modules.
4669 This list includes suffixes for both compiled and source Emacs Lisp files.
4670 This list should not include the empty string.
4671 `load' and related functions try to append these suffixes, in order,
4672 to the specified file name if a suffix is allowed or required. */);
4673 #ifdef HAVE_MODULES
4674 Vload_suffixes = list3 (build_pure_c_string (".elc"),
4675 build_pure_c_string (".el"),
4676 build_pure_c_string (MODULES_SUFFIX));
4677 #else
4678 Vload_suffixes = list2 (build_pure_c_string (".elc"),
4679 build_pure_c_string (".el"));
4680 #endif
4681 DEFVAR_LISP ("module-file-suffix", Vmodule_file_suffix,
4682 doc: /* Suffix of loadable module file, or nil of modules are not supported. */);
4683 #ifdef HAVE_MODULES
4684 Vmodule_file_suffix = build_pure_c_string (MODULES_SUFFIX);
4685 #else
4686 Vmodule_file_suffix = Qnil;
4687 #endif
4688 DEFVAR_LISP ("load-file-rep-suffixes", Vload_file_rep_suffixes,
4689 doc: /* List of suffixes that indicate representations of \
4690 the same file.
4691 This list should normally start with the empty string.
4693 Enabling Auto Compression mode appends the suffixes in
4694 `jka-compr-load-suffixes' to this list and disabling Auto Compression
4695 mode removes them again. `load' and related functions use this list to
4696 determine whether they should look for compressed versions of a file
4697 and, if so, which suffixes they should try to append to the file name
4698 in order to do so. However, if you want to customize which suffixes
4699 the loading functions recognize as compression suffixes, you should
4700 customize `jka-compr-load-suffixes' rather than the present variable. */);
4701 Vload_file_rep_suffixes = list1 (empty_unibyte_string);
4703 DEFVAR_BOOL ("load-in-progress", load_in_progress,
4704 doc: /* Non-nil if inside of `load'. */);
4705 DEFSYM (Qload_in_progress, "load-in-progress");
4707 DEFVAR_LISP ("after-load-alist", Vafter_load_alist,
4708 doc: /* An alist of functions to be evalled when particular files are loaded.
4709 Each element looks like (REGEXP-OR-FEATURE FUNCS...).
4711 REGEXP-OR-FEATURE is either a regular expression to match file names, or
4712 a symbol (a feature name).
4714 When `load' is run and the file-name argument matches an element's
4715 REGEXP-OR-FEATURE, or when `provide' is run and provides the symbol
4716 REGEXP-OR-FEATURE, the FUNCS in the element are called.
4718 An error in FORMS does not undo the load, but does prevent execution of
4719 the rest of the FORMS. */);
4720 Vafter_load_alist = Qnil;
4722 DEFVAR_LISP ("load-history", Vload_history,
4723 doc: /* Alist mapping loaded file names to symbols and features.
4724 Each alist element should be a list (FILE-NAME ENTRIES...), where
4725 FILE-NAME is the name of a file that has been loaded into Emacs.
4726 The file name is absolute and true (i.e. it doesn't contain symlinks).
4727 As an exception, one of the alist elements may have FILE-NAME nil,
4728 for symbols and features not associated with any file.
4730 The remaining ENTRIES in the alist element describe the functions and
4731 variables defined in that file, the features provided, and the
4732 features required. Each entry has the form `(provide . FEATURE)',
4733 `(require . FEATURE)', `(defun . FUNCTION)', `(autoload . SYMBOL)',
4734 `(defface . SYMBOL)', or `(t . SYMBOL)'. Entries like `(t . SYMBOL)'
4735 may precede a `(defun . FUNCTION)' entry, and means that SYMBOL was an
4736 autoload before this file redefined it as a function. In addition,
4737 entries may also be single symbols, which means that SYMBOL was
4738 defined by `defvar' or `defconst'.
4740 During preloading, the file name recorded is relative to the main Lisp
4741 directory. These file names are converted to absolute at startup. */);
4742 Vload_history = Qnil;
4744 DEFVAR_LISP ("load-file-name", Vload_file_name,
4745 doc: /* Full name of file being loaded by `load'. */);
4746 Vload_file_name = Qnil;
4748 DEFVAR_LISP ("user-init-file", Vuser_init_file,
4749 doc: /* File name, including directory, of user's initialization file.
4750 If the file loaded had extension `.elc', and the corresponding source file
4751 exists, this variable contains the name of source file, suitable for use
4752 by functions like `custom-save-all' which edit the init file.
4753 While Emacs loads and evaluates the init file, value is the real name
4754 of the file, regardless of whether or not it has the `.elc' extension. */);
4755 Vuser_init_file = Qnil;
4757 DEFVAR_LISP ("current-load-list", Vcurrent_load_list,
4758 doc: /* Used for internal purposes by `load'. */);
4759 Vcurrent_load_list = Qnil;
4761 DEFVAR_LISP ("load-read-function", Vload_read_function,
4762 doc: /* Function used by `load' and `eval-region' for reading expressions.
4763 Called with a single argument (the stream from which to read).
4764 The default is to use the function `read'. */);
4765 DEFSYM (Qread, "read");
4766 Vload_read_function = Qread;
4768 DEFVAR_LISP ("load-source-file-function", Vload_source_file_function,
4769 doc: /* Function called in `load' to load an Emacs Lisp source file.
4770 The value should be a function for doing code conversion before
4771 reading a source file. It can also be nil, in which case loading is
4772 done without any code conversion.
4774 If the value is a function, it is called with four arguments,
4775 FULLNAME, FILE, NOERROR, NOMESSAGE. FULLNAME is the absolute name of
4776 the file to load, FILE is the non-absolute name (for messages etc.),
4777 and NOERROR and NOMESSAGE are the corresponding arguments passed to
4778 `load'. The function should return t if the file was loaded. */);
4779 Vload_source_file_function = Qnil;
4781 DEFVAR_BOOL ("load-force-doc-strings", load_force_doc_strings,
4782 doc: /* Non-nil means `load' should force-load all dynamic doc strings.
4783 This is useful when the file being loaded is a temporary copy. */);
4784 load_force_doc_strings = 0;
4786 DEFVAR_BOOL ("load-convert-to-unibyte", load_convert_to_unibyte,
4787 doc: /* Non-nil means `read' converts strings to unibyte whenever possible.
4788 This is normally bound by `load' and `eval-buffer' to control `read',
4789 and is not meant for users to change. */);
4790 load_convert_to_unibyte = 0;
4792 DEFVAR_LISP ("source-directory", Vsource_directory,
4793 doc: /* Directory in which Emacs sources were found when Emacs was built.
4794 You cannot count on them to still be there! */);
4795 Vsource_directory
4796 = Fexpand_file_name (build_string ("../"),
4797 Fcar (decode_env_path (0, PATH_DUMPLOADSEARCH, 0)));
4799 DEFVAR_LISP ("preloaded-file-list", Vpreloaded_file_list,
4800 doc: /* List of files that were preloaded (when dumping Emacs). */);
4801 Vpreloaded_file_list = Qnil;
4803 DEFVAR_LISP ("byte-boolean-vars", Vbyte_boolean_vars,
4804 doc: /* List of all DEFVAR_BOOL variables, used by the byte code optimizer. */);
4805 Vbyte_boolean_vars = Qnil;
4807 DEFVAR_BOOL ("load-dangerous-libraries", load_dangerous_libraries,
4808 doc: /* Non-nil means load dangerous compiled Lisp files.
4809 Some versions of XEmacs use different byte codes than Emacs. These
4810 incompatible byte codes can make Emacs crash when it tries to execute
4811 them. */);
4812 load_dangerous_libraries = 0;
4814 DEFVAR_BOOL ("force-load-messages", force_load_messages,
4815 doc: /* Non-nil means force printing messages when loading Lisp files.
4816 This overrides the value of the NOMESSAGE argument to `load'. */);
4817 force_load_messages = 0;
4819 DEFVAR_LISP ("bytecomp-version-regexp", Vbytecomp_version_regexp,
4820 doc: /* Regular expression matching safe to load compiled Lisp files.
4821 When Emacs loads a compiled Lisp file, it reads the first 512 bytes
4822 from the file, and matches them against this regular expression.
4823 When the regular expression matches, the file is considered to be safe
4824 to load. See also `load-dangerous-libraries'. */);
4825 Vbytecomp_version_regexp
4826 = build_pure_c_string ("^;;;.\\(in Emacs version\\|bytecomp version FSF\\)");
4828 DEFSYM (Qlexical_binding, "lexical-binding");
4829 DEFVAR_LISP ("lexical-binding", Vlexical_binding,
4830 doc: /* Whether to use lexical binding when evaluating code.
4831 Non-nil means that the code in the current buffer should be evaluated
4832 with lexical binding.
4833 This variable is automatically set from the file variables of an
4834 interpreted Lisp file read using `load'. Unlike other file local
4835 variables, this must be set in the first line of a file. */);
4836 Vlexical_binding = Qnil;
4837 Fmake_variable_buffer_local (Qlexical_binding);
4839 DEFVAR_LISP ("eval-buffer-list", Veval_buffer_list,
4840 doc: /* List of buffers being read from by calls to `eval-buffer' and `eval-region'. */);
4841 Veval_buffer_list = Qnil;
4843 DEFVAR_LISP ("old-style-backquotes", Vold_style_backquotes,
4844 doc: /* Set to non-nil when `read' encounters an old-style backquote. */);
4845 Vold_style_backquotes = Qnil;
4846 DEFSYM (Qold_style_backquotes, "old-style-backquotes");
4848 DEFVAR_LISP ("lread--unescaped-character-literals",
4849 Vlread_unescaped_character_literals,
4850 doc: /* List of deprecated unescaped character literals encountered by `read'.
4851 For internal use only. */);
4852 Vlread_unescaped_character_literals = Qnil;
4853 DEFSYM (Qlread_unescaped_character_literals,
4854 "lread--unescaped-character-literals");
4856 DEFSYM (Qlss, "<");
4858 DEFVAR_BOOL ("load-prefer-newer", load_prefer_newer,
4859 doc: /* Non-nil means `load' prefers the newest version of a file.
4860 This applies when a filename suffix is not explicitly specified and
4861 `load' is trying various possible suffixes (see `load-suffixes' and
4862 `load-file-rep-suffixes'). Normally, it stops at the first file
4863 that exists unless you explicitly specify one or the other. If this
4864 option is non-nil, it checks all suffixes and uses whichever file is
4865 newest.
4866 Note that if you customize this, obviously it will not affect files
4867 that are loaded before your customizations are read! */);
4868 load_prefer_newer = 0;
4870 /* Vsource_directory was initialized in init_lread. */
4872 DEFSYM (Qcurrent_load_list, "current-load-list");
4873 DEFSYM (Qstandard_input, "standard-input");
4874 DEFSYM (Qread_char, "read-char");
4875 DEFSYM (Qget_file_char, "get-file-char");
4877 /* Used instead of Qget_file_char while loading *.elc files compiled
4878 by Emacs 21 or older. */
4879 DEFSYM (Qget_emacs_mule_file_char, "get-emacs-mule-file-char");
4881 DEFSYM (Qload_force_doc_strings, "load-force-doc-strings");
4883 DEFSYM (Qbackquote, "`");
4884 DEFSYM (Qcomma, ",");
4885 DEFSYM (Qcomma_at, ",@");
4886 DEFSYM (Qcomma_dot, ",.");
4888 DEFSYM (Qinhibit_file_name_operation, "inhibit-file-name-operation");
4889 DEFSYM (Qascii_character, "ascii-character");
4890 DEFSYM (Qfunction, "function");
4891 DEFSYM (Qload, "load");
4892 DEFSYM (Qload_file_name, "load-file-name");
4893 DEFSYM (Qeval_buffer_list, "eval-buffer-list");
4894 DEFSYM (Qfile_truename, "file-truename");
4895 DEFSYM (Qdir_ok, "dir-ok");
4896 DEFSYM (Qdo_after_load_evaluation, "do-after-load-evaluation");
4898 staticpro (&read_objects);
4899 read_objects = Qnil;
4900 staticpro (&seen_list);
4901 seen_list = Qnil;
4903 Vloads_in_progress = Qnil;
4904 staticpro (&Vloads_in_progress);
4906 DEFSYM (Qhash_table, "hash-table");
4907 DEFSYM (Qdata, "data");
4908 DEFSYM (Qtest, "test");
4909 DEFSYM (Qsize, "size");
4910 DEFSYM (Qpurecopy, "purecopy");
4911 DEFSYM (Qweakness, "weakness");
4912 DEFSYM (Qrehash_size, "rehash-size");
4913 DEFSYM (Qrehash_threshold, "rehash-threshold");
4915 DEFSYM (Qchar_from_name, "char-from-name");