Subject: Restore correct Gnus newsgroup name after sending message
[emacs.git] / src / lread.c
blobea2a1d1d858199e994efc3ff8804500f2d2337e1
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_object_in_subtree (Lisp_Object,
562 Lisp_Object);
563 static void substitute_in_interval (INTERVAL, Lisp_Object);
566 /* Get a character from the tty. */
568 /* Read input events until we get one that's acceptable for our purposes.
570 If NO_SWITCH_FRAME, switch-frame events are stashed
571 until we get a character we like, and then stuffed into
572 unread_switch_frame.
574 If ASCII_REQUIRED, check function key events to see
575 if the unmodified version of the symbol has a Qascii_character
576 property, and use that character, if present.
578 If ERROR_NONASCII, signal an error if the input we
579 get isn't an ASCII character with modifiers. If it's false but
580 ASCII_REQUIRED is true, just re-read until we get an ASCII
581 character.
583 If INPUT_METHOD, invoke the current input method
584 if the character warrants that.
586 If SECONDS is a number, wait that many seconds for input, and
587 return Qnil if no input arrives within that time. */
589 static Lisp_Object
590 read_filtered_event (bool no_switch_frame, bool ascii_required,
591 bool error_nonascii, bool input_method, Lisp_Object seconds)
593 Lisp_Object val, delayed_switch_frame;
594 struct timespec end_time;
596 #ifdef HAVE_WINDOW_SYSTEM
597 if (display_hourglass_p)
598 cancel_hourglass ();
599 #endif
601 delayed_switch_frame = Qnil;
603 /* Compute timeout. */
604 if (NUMBERP (seconds))
606 double duration = extract_float (seconds);
607 struct timespec wait_time = dtotimespec (duration);
608 end_time = timespec_add (current_timespec (), wait_time);
611 /* Read until we get an acceptable event. */
612 retry:
614 val = read_char (0, Qnil, (input_method ? Qnil : Qt), 0,
615 NUMBERP (seconds) ? &end_time : NULL);
616 while (INTEGERP (val) && XINT (val) == -2); /* wrong_kboard_jmpbuf */
618 if (BUFFERP (val))
619 goto retry;
621 /* `switch-frame' events are put off until after the next ASCII
622 character. This is better than signaling an error just because
623 the last characters were typed to a separate minibuffer frame,
624 for example. Eventually, some code which can deal with
625 switch-frame events will read it and process it. */
626 if (no_switch_frame
627 && EVENT_HAS_PARAMETERS (val)
628 && EQ (EVENT_HEAD_KIND (EVENT_HEAD (val)), Qswitch_frame))
630 delayed_switch_frame = val;
631 goto retry;
634 if (ascii_required && !(NUMBERP (seconds) && NILP (val)))
636 /* Convert certain symbols to their ASCII equivalents. */
637 if (SYMBOLP (val))
639 Lisp_Object tem, tem1;
640 tem = Fget (val, Qevent_symbol_element_mask);
641 if (!NILP (tem))
643 tem1 = Fget (Fcar (tem), Qascii_character);
644 /* Merge this symbol's modifier bits
645 with the ASCII equivalent of its basic code. */
646 if (!NILP (tem1))
647 XSETFASTINT (val, XINT (tem1) | XINT (Fcar (Fcdr (tem))));
651 /* If we don't have a character now, deal with it appropriately. */
652 if (!INTEGERP (val))
654 if (error_nonascii)
656 Vunread_command_events = list1 (val);
657 error ("Non-character input-event");
659 else
660 goto retry;
664 if (! NILP (delayed_switch_frame))
665 unread_switch_frame = delayed_switch_frame;
667 #if 0
669 #ifdef HAVE_WINDOW_SYSTEM
670 if (display_hourglass_p)
671 start_hourglass ();
672 #endif
674 #endif
676 return val;
679 DEFUN ("read-char", Fread_char, Sread_char, 0, 3, 0,
680 doc: /* Read a character from the command input (keyboard or macro).
681 It is returned as a number.
682 If the character has modifiers, they are resolved and reflected to the
683 character code if possible (e.g. C-SPC -> 0).
685 If the user generates an event which is not a character (i.e. a mouse
686 click or function key event), `read-char' signals an error. As an
687 exception, switch-frame events are put off until non-character events
688 can be read.
689 If you want to read non-character events, or ignore them, call
690 `read-event' or `read-char-exclusive' instead.
692 If the optional argument PROMPT is non-nil, display that as a prompt.
693 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
694 input method is turned on in the current buffer, that input method
695 is used for reading a character.
696 If the optional argument SECONDS is non-nil, it should be a number
697 specifying the maximum number of seconds to wait for input. If no
698 input arrives in that time, return nil. SECONDS may be a
699 floating-point value. */)
700 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
702 Lisp_Object val;
704 if (! NILP (prompt))
705 message_with_string ("%s", prompt, 0);
706 val = read_filtered_event (1, 1, 1, ! NILP (inherit_input_method), seconds);
708 return (NILP (val) ? Qnil
709 : make_number (char_resolve_modifier_mask (XINT (val))));
712 DEFUN ("read-event", Fread_event, Sread_event, 0, 3, 0,
713 doc: /* Read an event object from the input stream.
714 If the optional argument PROMPT is non-nil, display that as a prompt.
715 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
716 input method is turned on in the current buffer, that input method
717 is used for reading a character.
718 If the optional argument SECONDS is non-nil, it should be a number
719 specifying the maximum number of seconds to wait for input. If no
720 input arrives in that time, return nil. SECONDS may be a
721 floating-point value. */)
722 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
724 if (! NILP (prompt))
725 message_with_string ("%s", prompt, 0);
726 return read_filtered_event (0, 0, 0, ! NILP (inherit_input_method), seconds);
729 DEFUN ("read-char-exclusive", Fread_char_exclusive, Sread_char_exclusive, 0, 3, 0,
730 doc: /* Read a character from the command input (keyboard or macro).
731 It is returned as a number. Non-character events are ignored.
732 If the character has modifiers, they are resolved and reflected to the
733 character code if possible (e.g. C-SPC -> 0).
735 If the optional argument PROMPT is non-nil, display that as a prompt.
736 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
737 input method is turned on in the current buffer, that input method
738 is used for reading a character.
739 If the optional argument SECONDS is non-nil, it should be a number
740 specifying the maximum number of seconds to wait for input. If no
741 input arrives in that time, return nil. SECONDS may be a
742 floating-point value. */)
743 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
745 Lisp_Object val;
747 if (! NILP (prompt))
748 message_with_string ("%s", prompt, 0);
750 val = read_filtered_event (1, 1, 0, ! NILP (inherit_input_method), seconds);
752 return (NILP (val) ? Qnil
753 : make_number (char_resolve_modifier_mask (XINT (val))));
756 DEFUN ("get-file-char", Fget_file_char, Sget_file_char, 0, 0, 0,
757 doc: /* Don't use this yourself. */)
758 (void)
760 register Lisp_Object val;
761 block_input ();
762 XSETINT (val, getc (instream));
763 unblock_input ();
764 return val;
770 /* Return true if the lisp code read using READCHARFUN defines a non-nil
771 `lexical-binding' file variable. After returning, the stream is
772 positioned following the first line, if it is a comment or #! line,
773 otherwise nothing is read. */
775 static bool
776 lisp_file_lexically_bound_p (Lisp_Object readcharfun)
778 int ch = READCHAR;
780 if (ch == '#')
782 ch = READCHAR;
783 if (ch != '!')
785 UNREAD (ch);
786 UNREAD ('#');
787 return 0;
789 while (ch != '\n' && ch != EOF)
790 ch = READCHAR;
791 if (ch == '\n') ch = READCHAR;
792 /* It is OK to leave the position after a #! line, since
793 that is what read1 does. */
796 if (ch != ';')
797 /* The first line isn't a comment, just give up. */
799 UNREAD (ch);
800 return 0;
802 else
803 /* Look for an appropriate file-variable in the first line. */
805 bool rv = 0;
806 enum {
807 NOMINAL, AFTER_FIRST_DASH, AFTER_ASTERIX
808 } beg_end_state = NOMINAL;
809 bool in_file_vars = 0;
811 #define UPDATE_BEG_END_STATE(ch) \
812 if (beg_end_state == NOMINAL) \
813 beg_end_state = (ch == '-' ? AFTER_FIRST_DASH : NOMINAL); \
814 else if (beg_end_state == AFTER_FIRST_DASH) \
815 beg_end_state = (ch == '*' ? AFTER_ASTERIX : NOMINAL); \
816 else if (beg_end_state == AFTER_ASTERIX) \
818 if (ch == '-') \
819 in_file_vars = !in_file_vars; \
820 beg_end_state = NOMINAL; \
823 /* Skip until we get to the file vars, if any. */
826 ch = READCHAR;
827 UPDATE_BEG_END_STATE (ch);
829 while (!in_file_vars && ch != '\n' && ch != EOF);
831 while (in_file_vars)
833 char var[100], val[100];
834 unsigned i;
836 ch = READCHAR;
838 /* Read a variable name. */
839 while (ch == ' ' || ch == '\t')
840 ch = READCHAR;
842 i = 0;
843 while (ch != ':' && ch != '\n' && ch != EOF && in_file_vars)
845 if (i < sizeof var - 1)
846 var[i++] = ch;
847 UPDATE_BEG_END_STATE (ch);
848 ch = READCHAR;
851 /* Stop scanning if no colon was found before end marker. */
852 if (!in_file_vars || ch == '\n' || ch == EOF)
853 break;
855 while (i > 0 && (var[i - 1] == ' ' || var[i - 1] == '\t'))
856 i--;
857 var[i] = '\0';
859 if (ch == ':')
861 /* Read a variable value. */
862 ch = READCHAR;
864 while (ch == ' ' || ch == '\t')
865 ch = READCHAR;
867 i = 0;
868 while (ch != ';' && ch != '\n' && ch != EOF && in_file_vars)
870 if (i < sizeof val - 1)
871 val[i++] = ch;
872 UPDATE_BEG_END_STATE (ch);
873 ch = READCHAR;
875 if (! in_file_vars)
876 /* The value was terminated by an end-marker, which remove. */
877 i -= 3;
878 while (i > 0 && (val[i - 1] == ' ' || val[i - 1] == '\t'))
879 i--;
880 val[i] = '\0';
882 if (strcmp (var, "lexical-binding") == 0)
883 /* This is it... */
885 rv = (strcmp (val, "nil") != 0);
886 break;
891 while (ch != '\n' && ch != EOF)
892 ch = READCHAR;
894 return rv;
898 /* Value is a version number of byte compiled code if the file
899 associated with file descriptor FD is a compiled Lisp file that's
900 safe to load. Only files compiled with Emacs are safe to load.
901 Files compiled with XEmacs can lead to a crash in Fbyte_code
902 because of an incompatible change in the byte compiler. */
904 static int
905 safe_to_load_version (int fd)
907 char buf[512];
908 int nbytes, i;
909 int version = 1;
911 /* Read the first few bytes from the file, and look for a line
912 specifying the byte compiler version used. */
913 nbytes = emacs_read (fd, buf, sizeof buf);
914 if (nbytes > 0)
916 /* Skip to the next newline, skipping over the initial `ELC'
917 with NUL bytes following it, but note the version. */
918 for (i = 0; i < nbytes && buf[i] != '\n'; ++i)
919 if (i == 4)
920 version = buf[i];
922 if (i >= nbytes
923 || fast_c_string_match_ignore_case (Vbytecomp_version_regexp,
924 buf + i, nbytes - i) < 0)
925 version = 0;
928 lseek (fd, 0, SEEK_SET);
929 return version;
933 /* Callback for record_unwind_protect. Restore the old load list OLD,
934 after loading a file successfully. */
936 static void
937 record_load_unwind (Lisp_Object old)
939 Vloads_in_progress = old;
942 /* This handler function is used via internal_condition_case_1. */
944 static Lisp_Object
945 load_error_handler (Lisp_Object data)
947 return Qnil;
950 static void
951 load_warn_old_style_backquotes (Lisp_Object file)
953 if (!NILP (Vold_style_backquotes))
955 AUTO_STRING (format, "Loading `%s': old-style backquotes detected!");
956 CALLN (Fmessage, format, file);
960 DEFUN ("get-load-suffixes", Fget_load_suffixes, Sget_load_suffixes, 0, 0, 0,
961 doc: /* Return the suffixes that `load' should try if a suffix is \
962 required.
963 This uses the variables `load-suffixes' and `load-file-rep-suffixes'. */)
964 (void)
966 Lisp_Object lst = Qnil, suffixes = Vload_suffixes, suffix, ext;
967 while (CONSP (suffixes))
969 Lisp_Object exts = Vload_file_rep_suffixes;
970 suffix = XCAR (suffixes);
971 suffixes = XCDR (suffixes);
972 while (CONSP (exts))
974 ext = XCAR (exts);
975 exts = XCDR (exts);
976 lst = Fcons (concat2 (suffix, ext), lst);
979 return Fnreverse (lst);
982 /* Returns true if STRING ends with SUFFIX */
983 static bool
984 suffix_p (Lisp_Object string, const char *suffix)
986 ptrdiff_t suffix_len = strlen (suffix);
987 ptrdiff_t string_len = SBYTES (string);
989 return string_len >= suffix_len && !strcmp (SSDATA (string) + string_len - suffix_len, suffix);
992 DEFUN ("load", Fload, Sload, 1, 5, 0,
993 doc: /* Execute a file of Lisp code named FILE.
994 First try FILE with `.elc' appended, then try with `.el', then try
995 with a system-dependent suffix of dynamic modules (see `load-suffixes'),
996 then try FILE unmodified (the exact suffixes in the exact order are
997 determined by `load-suffixes'). Environment variable references in
998 FILE are replaced with their values by calling `substitute-in-file-name'.
999 This function searches the directories in `load-path'.
1001 If optional second arg NOERROR is non-nil,
1002 report no error if FILE doesn't exist.
1003 Print messages at start and end of loading unless
1004 optional third arg NOMESSAGE is non-nil (but `force-load-messages'
1005 overrides that).
1006 If optional fourth arg NOSUFFIX is non-nil, don't try adding
1007 suffixes to the specified name FILE.
1008 If optional fifth arg MUST-SUFFIX is non-nil, insist on
1009 the suffix `.elc' or `.el' or the module suffix; don't accept just
1010 FILE unless it ends in one of those suffixes or includes a directory name.
1012 If NOSUFFIX is nil, then if a file could not be found, try looking for
1013 a different representation of the file by adding non-empty suffixes to
1014 its name, before trying another file. Emacs uses this feature to find
1015 compressed versions of files when Auto Compression mode is enabled.
1016 If NOSUFFIX is non-nil, disable this feature.
1018 The suffixes that this function tries out, when NOSUFFIX is nil, are
1019 given by the return value of `get-load-suffixes' and the values listed
1020 in `load-file-rep-suffixes'. If MUST-SUFFIX is non-nil, only the
1021 return value of `get-load-suffixes' is used, i.e. the file name is
1022 required to have a non-empty suffix.
1024 When searching suffixes, this function normally stops at the first
1025 one that exists. If the option `load-prefer-newer' is non-nil,
1026 however, it tries all suffixes, and uses whichever file is the newest.
1028 Loading a file records its definitions, and its `provide' and
1029 `require' calls, in an element of `load-history' whose
1030 car is the file name loaded. See `load-history'.
1032 While the file is in the process of being loaded, the variable
1033 `load-in-progress' is non-nil and the variable `load-file-name'
1034 is bound to the file's name.
1036 Return t if the file exists and loads successfully. */)
1037 (Lisp_Object file, Lisp_Object noerror, Lisp_Object nomessage,
1038 Lisp_Object nosuffix, Lisp_Object must_suffix)
1040 FILE *stream;
1041 int fd;
1042 int fd_index UNINIT;
1043 ptrdiff_t count = SPECPDL_INDEX ();
1044 Lisp_Object found, efound, hist_file_name;
1045 /* True means we printed the ".el is newer" message. */
1046 bool newer = 0;
1047 /* True means we are loading a compiled file. */
1048 bool compiled = 0;
1049 Lisp_Object handler;
1050 bool safe_p = 1;
1051 const char *fmode = "r" FOPEN_TEXT;
1052 int version;
1054 CHECK_STRING (file);
1056 /* If file name is magic, call the handler. */
1057 /* This shouldn't be necessary any more now that `openp' handles it right.
1058 handler = Ffind_file_name_handler (file, Qload);
1059 if (!NILP (handler))
1060 return call5 (handler, Qload, file, noerror, nomessage, nosuffix); */
1062 /* The presence of this call is the result of a historical accident:
1063 it used to be in every file-operation and when it got removed
1064 everywhere, it accidentally stayed here. Since then, enough people
1065 supposedly have things like (load "$PROJECT/foo.el") in their .emacs
1066 that it seemed risky to remove. */
1067 if (! NILP (noerror))
1069 file = internal_condition_case_1 (Fsubstitute_in_file_name, file,
1070 Qt, load_error_handler);
1071 if (NILP (file))
1072 return Qnil;
1074 else
1075 file = Fsubstitute_in_file_name (file);
1077 /* Avoid weird lossage with null string as arg,
1078 since it would try to load a directory as a Lisp file. */
1079 if (SCHARS (file) == 0)
1081 fd = -1;
1082 errno = ENOENT;
1084 else
1086 Lisp_Object suffixes;
1087 found = Qnil;
1089 if (! NILP (must_suffix))
1091 /* Don't insist on adding a suffix if FILE already ends with one. */
1092 if (suffix_p (file, ".el")
1093 || suffix_p (file, ".elc")
1094 #ifdef HAVE_MODULES
1095 || suffix_p (file, MODULES_SUFFIX)
1096 #endif
1098 must_suffix = Qnil;
1099 /* Don't insist on adding a suffix
1100 if the argument includes a directory name. */
1101 else if (! NILP (Ffile_name_directory (file)))
1102 must_suffix = Qnil;
1105 if (!NILP (nosuffix))
1106 suffixes = Qnil;
1107 else
1109 suffixes = Fget_load_suffixes ();
1110 if (NILP (must_suffix))
1111 suffixes = CALLN (Fappend, suffixes, Vload_file_rep_suffixes);
1114 fd = openp (Vload_path, file, suffixes, &found, Qnil, load_prefer_newer);
1117 if (fd == -1)
1119 if (NILP (noerror))
1120 report_file_error ("Cannot open load file", file);
1121 return Qnil;
1124 /* Tell startup.el whether or not we found the user's init file. */
1125 if (EQ (Qt, Vuser_init_file))
1126 Vuser_init_file = found;
1128 /* If FD is -2, that means openp found a magic file. */
1129 if (fd == -2)
1131 if (NILP (Fequal (found, file)))
1132 /* If FOUND is a different file name from FILE,
1133 find its handler even if we have already inhibited
1134 the `load' operation on FILE. */
1135 handler = Ffind_file_name_handler (found, Qt);
1136 else
1137 handler = Ffind_file_name_handler (found, Qload);
1138 if (! NILP (handler))
1139 return call5 (handler, Qload, found, noerror, nomessage, Qt);
1140 #ifdef DOS_NT
1141 /* Tramp has to deal with semi-broken packages that prepend
1142 drive letters to remote files. For that reason, Tramp
1143 catches file operations that test for file existence, which
1144 makes openp think X:/foo.elc files are remote. However,
1145 Tramp does not catch `load' operations for such files, so we
1146 end up with a nil as the `load' handler above. If we would
1147 continue with fd = -2, we will behave wrongly, and in
1148 particular try reading a .elc file in the "rt" mode instead
1149 of "rb". See bug #9311 for the results. To work around
1150 this, we try to open the file locally, and go with that if it
1151 succeeds. */
1152 fd = emacs_open (SSDATA (ENCODE_FILE (found)), O_RDONLY, 0);
1153 if (fd == -1)
1154 fd = -2;
1155 #endif
1158 if (0 <= fd)
1160 fd_index = SPECPDL_INDEX ();
1161 record_unwind_protect_int (close_file_unwind, fd);
1164 #ifdef HAVE_MODULES
1165 if (suffix_p (found, MODULES_SUFFIX))
1166 return unbind_to (count, Fmodule_load (found));
1167 #endif
1169 /* Check if we're stuck in a recursive load cycle.
1171 2000-09-21: It's not possible to just check for the file loaded
1172 being a member of Vloads_in_progress. This fails because of the
1173 way the byte compiler currently works; `provide's are not
1174 evaluated, see font-lock.el/jit-lock.el as an example. This
1175 leads to a certain amount of ``normal'' recursion.
1177 Also, just loading a file recursively is not always an error in
1178 the general case; the second load may do something different. */
1180 int load_count = 0;
1181 Lisp_Object tem;
1182 for (tem = Vloads_in_progress; CONSP (tem); tem = XCDR (tem))
1183 if (!NILP (Fequal (found, XCAR (tem))) && (++load_count > 3))
1184 signal_error ("Recursive load", Fcons (found, Vloads_in_progress));
1185 record_unwind_protect (record_load_unwind, Vloads_in_progress);
1186 Vloads_in_progress = Fcons (found, Vloads_in_progress);
1189 /* All loads are by default dynamic, unless the file itself specifies
1190 otherwise using a file-variable in the first line. This is bound here
1191 so that it takes effect whether or not we use
1192 Vload_source_file_function. */
1193 specbind (Qlexical_binding, Qnil);
1195 /* Get the name for load-history. */
1196 hist_file_name = (! NILP (Vpurify_flag)
1197 ? concat2 (Ffile_name_directory (file),
1198 Ffile_name_nondirectory (found))
1199 : found) ;
1201 version = -1;
1203 /* Check for the presence of old-style quotes and warn about them. */
1204 specbind (Qold_style_backquotes, Qnil);
1205 record_unwind_protect (load_warn_old_style_backquotes, file);
1207 int is_elc;
1208 if ((is_elc = suffix_p (found, ".elc")) != 0
1209 /* version = 1 means the file is empty, in which case we can
1210 treat it as not byte-compiled. */
1211 || (fd >= 0 && (version = safe_to_load_version (fd)) > 1))
1212 /* Load .elc files directly, but not when they are
1213 remote and have no handler! */
1215 if (fd != -2)
1217 struct stat s1, s2;
1218 int result;
1220 if (version < 0
1221 && ! (version = safe_to_load_version (fd)))
1223 safe_p = 0;
1224 if (!load_dangerous_libraries)
1225 error ("File `%s' was not compiled in Emacs", SDATA (found));
1226 else if (!NILP (nomessage) && !force_load_messages)
1227 message_with_string ("File `%s' not compiled in Emacs", found, 1);
1230 compiled = 1;
1232 efound = ENCODE_FILE (found);
1233 fmode = "r" FOPEN_BINARY;
1235 /* openp already checked for newness, no point doing it again.
1236 FIXME would be nice to get a message when openp
1237 ignores suffix order due to load_prefer_newer. */
1238 if (!load_prefer_newer && is_elc)
1240 result = stat (SSDATA (efound), &s1);
1241 if (result == 0)
1243 SSET (efound, SBYTES (efound) - 1, 0);
1244 result = stat (SSDATA (efound), &s2);
1245 SSET (efound, SBYTES (efound) - 1, 'c');
1248 if (result == 0
1249 && timespec_cmp (get_stat_mtime (&s1), get_stat_mtime (&s2)) < 0)
1251 /* Make the progress messages mention that source is newer. */
1252 newer = 1;
1254 /* If we won't print another message, mention this anyway. */
1255 if (!NILP (nomessage) && !force_load_messages)
1257 Lisp_Object msg_file;
1258 msg_file = Fsubstring (found, make_number (0), make_number (-1));
1259 message_with_string ("Source file `%s' newer than byte-compiled file",
1260 msg_file, 1);
1263 } /* !load_prefer_newer */
1266 else
1268 /* We are loading a source file (*.el). */
1269 if (!NILP (Vload_source_file_function))
1271 Lisp_Object val;
1273 if (fd >= 0)
1275 emacs_close (fd);
1276 clear_unwind_protect (fd_index);
1278 val = call4 (Vload_source_file_function, found, hist_file_name,
1279 NILP (noerror) ? Qnil : Qt,
1280 (NILP (nomessage) || force_load_messages) ? Qnil : Qt);
1281 return unbind_to (count, val);
1285 if (fd < 0)
1287 /* We somehow got here with fd == -2, meaning the file is deemed
1288 to be remote. Don't even try to reopen the file locally;
1289 just force a failure. */
1290 stream = NULL;
1291 errno = EINVAL;
1293 else
1295 #ifdef WINDOWSNT
1296 emacs_close (fd);
1297 clear_unwind_protect (fd_index);
1298 efound = ENCODE_FILE (found);
1299 stream = emacs_fopen (SSDATA (efound), fmode);
1300 #else
1301 stream = fdopen (fd, fmode);
1302 #endif
1304 if (! stream)
1305 report_file_error ("Opening stdio stream", file);
1306 set_unwind_protect_ptr (fd_index, fclose_unwind, stream);
1308 if (! NILP (Vpurify_flag))
1309 Vpreloaded_file_list = Fcons (Fpurecopy (file), Vpreloaded_file_list);
1311 if (NILP (nomessage) || force_load_messages)
1313 if (!safe_p)
1314 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...",
1315 file, 1);
1316 else if (!compiled)
1317 message_with_string ("Loading %s (source)...", file, 1);
1318 else if (newer)
1319 message_with_string ("Loading %s (compiled; note, source file is newer)...",
1320 file, 1);
1321 else /* The typical case; compiled file newer than source file. */
1322 message_with_string ("Loading %s...", file, 1);
1325 specbind (Qload_file_name, found);
1326 specbind (Qinhibit_file_name_operation, Qnil);
1327 specbind (Qload_in_progress, Qt);
1329 instream = stream;
1330 if (lisp_file_lexically_bound_p (Qget_file_char))
1331 Fset (Qlexical_binding, Qt);
1333 if (! version || version >= 22)
1334 readevalloop (Qget_file_char, stream, hist_file_name,
1335 0, Qnil, Qnil, Qnil, Qnil);
1336 else
1338 /* We can't handle a file which was compiled with
1339 byte-compile-dynamic by older version of Emacs. */
1340 specbind (Qload_force_doc_strings, Qt);
1341 readevalloop (Qget_emacs_mule_file_char, stream, hist_file_name,
1342 0, Qnil, Qnil, Qnil, Qnil);
1344 unbind_to (count, Qnil);
1346 /* Run any eval-after-load forms for this file. */
1347 if (!NILP (Ffboundp (Qdo_after_load_evaluation)))
1348 call1 (Qdo_after_load_evaluation, hist_file_name) ;
1350 xfree (saved_doc_string);
1351 saved_doc_string = 0;
1352 saved_doc_string_size = 0;
1354 xfree (prev_saved_doc_string);
1355 prev_saved_doc_string = 0;
1356 prev_saved_doc_string_size = 0;
1358 if (!noninteractive && (NILP (nomessage) || force_load_messages))
1360 if (!safe_p)
1361 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...done",
1362 file, 1);
1363 else if (!compiled)
1364 message_with_string ("Loading %s (source)...done", file, 1);
1365 else if (newer)
1366 message_with_string ("Loading %s (compiled; note, source file is newer)...done",
1367 file, 1);
1368 else /* The typical case; compiled file newer than source file. */
1369 message_with_string ("Loading %s...done", file, 1);
1372 return Qt;
1375 static bool
1376 complete_filename_p (Lisp_Object pathname)
1378 const unsigned char *s = SDATA (pathname);
1379 return (IS_DIRECTORY_SEP (s[0])
1380 || (SCHARS (pathname) > 2
1381 && IS_DEVICE_SEP (s[1]) && IS_DIRECTORY_SEP (s[2])));
1384 DEFUN ("locate-file-internal", Flocate_file_internal, Slocate_file_internal, 2, 4, 0,
1385 doc: /* Search for FILENAME through PATH.
1386 Returns the file's name in absolute form, or nil if not found.
1387 If SUFFIXES is non-nil, it should be a list of suffixes to append to
1388 file name when searching.
1389 If non-nil, PREDICATE is used instead of `file-readable-p'.
1390 PREDICATE can also be an integer to pass to the faccessat(2) function,
1391 in which case file-name-handlers are ignored.
1392 This function will normally skip directories, so if you want it to find
1393 directories, make sure the PREDICATE function returns `dir-ok' for them. */)
1394 (Lisp_Object filename, Lisp_Object path, Lisp_Object suffixes, Lisp_Object predicate)
1396 Lisp_Object file;
1397 int fd = openp (path, filename, suffixes, &file, predicate, false);
1398 if (NILP (predicate) && fd >= 0)
1399 emacs_close (fd);
1400 return file;
1403 /* Search for a file whose name is STR, looking in directories
1404 in the Lisp list PATH, and trying suffixes from SUFFIX.
1405 On success, return a file descriptor (or 1 or -2 as described below).
1406 On failure, return -1 and set errno.
1408 SUFFIXES is a list of strings containing possible suffixes.
1409 The empty suffix is automatically added if the list is empty.
1411 PREDICATE t means the files are binary.
1412 PREDICATE non-nil and non-t means don't open the files,
1413 just look for one that satisfies the predicate. In this case,
1414 return -2 on success. The predicate can be a lisp function or
1415 an integer to pass to `access' (in which case file-name-handlers
1416 are ignored).
1418 If STOREPTR is nonzero, it points to a slot where the name of
1419 the file actually found should be stored as a Lisp string.
1420 nil is stored there on failure.
1422 If the file we find is remote, return -2
1423 but store the found remote file name in *STOREPTR.
1425 If NEWER is true, try all SUFFIXes and return the result for the
1426 newest file that exists. Does not apply to remote files,
1427 or if a non-nil and non-t PREDICATE is specified. */
1430 openp (Lisp_Object path, Lisp_Object str, Lisp_Object suffixes,
1431 Lisp_Object *storeptr, Lisp_Object predicate, bool newer)
1433 ptrdiff_t fn_size = 100;
1434 char buf[100];
1435 char *fn = buf;
1436 bool absolute;
1437 ptrdiff_t want_length;
1438 Lisp_Object filename;
1439 Lisp_Object string, tail, encoded_fn, save_string;
1440 ptrdiff_t max_suffix_len = 0;
1441 int last_errno = ENOENT;
1442 int save_fd = -1;
1443 USE_SAFE_ALLOCA;
1445 /* The last-modified time of the newest matching file found.
1446 Initialize it to something less than all valid timestamps. */
1447 struct timespec save_mtime = make_timespec (TYPE_MINIMUM (time_t), -1);
1449 CHECK_STRING (str);
1451 for (tail = suffixes; CONSP (tail); tail = XCDR (tail))
1453 CHECK_STRING_CAR (tail);
1454 max_suffix_len = max (max_suffix_len,
1455 SBYTES (XCAR (tail)));
1458 string = filename = encoded_fn = save_string = Qnil;
1460 if (storeptr)
1461 *storeptr = Qnil;
1463 absolute = complete_filename_p (str);
1465 for (; CONSP (path); path = XCDR (path))
1467 ptrdiff_t baselen, prefixlen;
1469 filename = Fexpand_file_name (str, XCAR (path));
1470 if (!complete_filename_p (filename))
1471 /* If there are non-absolute elts in PATH (eg "."). */
1472 /* Of course, this could conceivably lose if luser sets
1473 default-directory to be something non-absolute... */
1475 filename = Fexpand_file_name (filename, BVAR (current_buffer, directory));
1476 if (!complete_filename_p (filename))
1477 /* Give up on this path element! */
1478 continue;
1481 /* Calculate maximum length of any filename made from
1482 this path element/specified file name and any possible suffix. */
1483 want_length = max_suffix_len + SBYTES (filename);
1484 if (fn_size <= want_length)
1486 fn_size = 100 + want_length;
1487 fn = SAFE_ALLOCA (fn_size);
1490 /* Copy FILENAME's data to FN but remove starting /: if any. */
1491 prefixlen = ((SCHARS (filename) > 2
1492 && SREF (filename, 0) == '/'
1493 && SREF (filename, 1) == ':')
1494 ? 2 : 0);
1495 baselen = SBYTES (filename) - prefixlen;
1496 memcpy (fn, SDATA (filename) + prefixlen, baselen);
1498 /* Loop over suffixes. */
1499 for (tail = NILP (suffixes) ? list1 (empty_unibyte_string) : suffixes;
1500 CONSP (tail); tail = XCDR (tail))
1502 Lisp_Object suffix = XCAR (tail);
1503 ptrdiff_t fnlen, lsuffix = SBYTES (suffix);
1504 Lisp_Object handler;
1506 /* Make complete filename by appending SUFFIX. */
1507 memcpy (fn + baselen, SDATA (suffix), lsuffix + 1);
1508 fnlen = baselen + lsuffix;
1510 /* Check that the file exists and is not a directory. */
1511 /* We used to only check for handlers on non-absolute file names:
1512 if (absolute)
1513 handler = Qnil;
1514 else
1515 handler = Ffind_file_name_handler (filename, Qfile_exists_p);
1516 It's not clear why that was the case and it breaks things like
1517 (load "/bar.el") where the file is actually "/bar.el.gz". */
1518 /* make_string has its own ideas on when to return a unibyte
1519 string and when a multibyte string, but we know better.
1520 We must have a unibyte string when dumping, since
1521 file-name encoding is shaky at best at that time, and in
1522 particular default-file-name-coding-system is reset
1523 several times during loadup. We therefore don't want to
1524 encode the file before passing it to file I/O library
1525 functions. */
1526 if (!STRING_MULTIBYTE (filename) && !STRING_MULTIBYTE (suffix))
1527 string = make_unibyte_string (fn, fnlen);
1528 else
1529 string = make_string (fn, fnlen);
1530 handler = Ffind_file_name_handler (string, Qfile_exists_p);
1531 if ((!NILP (handler) || (!NILP (predicate) && !EQ (predicate, Qt)))
1532 && !NATNUMP (predicate))
1534 bool exists;
1535 if (NILP (predicate) || EQ (predicate, Qt))
1536 exists = !NILP (Ffile_readable_p (string));
1537 else
1539 Lisp_Object tmp = call1 (predicate, string);
1540 if (NILP (tmp))
1541 exists = false;
1542 else if (EQ (tmp, Qdir_ok)
1543 || NILP (Ffile_directory_p (string)))
1544 exists = true;
1545 else
1547 exists = false;
1548 last_errno = EISDIR;
1552 if (exists)
1554 /* We succeeded; return this descriptor and filename. */
1555 if (storeptr)
1556 *storeptr = string;
1557 SAFE_FREE ();
1558 return -2;
1561 else
1563 int fd;
1564 const char *pfn;
1565 struct stat st;
1567 encoded_fn = ENCODE_FILE (string);
1568 pfn = SSDATA (encoded_fn);
1570 /* Check that we can access or open it. */
1571 if (NATNUMP (predicate))
1573 fd = -1;
1574 if (INT_MAX < XFASTINT (predicate))
1575 last_errno = EINVAL;
1576 else if (faccessat (AT_FDCWD, pfn, XFASTINT (predicate),
1577 AT_EACCESS)
1578 == 0)
1580 if (file_directory_p (pfn))
1581 last_errno = EISDIR;
1582 else
1583 fd = 1;
1586 else
1588 fd = emacs_open (pfn, O_RDONLY, 0);
1589 if (fd < 0)
1591 if (errno != ENOENT)
1592 last_errno = errno;
1594 else
1596 int err = (fstat (fd, &st) != 0 ? errno
1597 : S_ISDIR (st.st_mode) ? EISDIR : 0);
1598 if (err)
1600 last_errno = err;
1601 emacs_close (fd);
1602 fd = -1;
1607 if (fd >= 0)
1609 if (newer && !NATNUMP (predicate))
1611 struct timespec mtime = get_stat_mtime (&st);
1613 if (timespec_cmp (mtime, save_mtime) <= 0)
1614 emacs_close (fd);
1615 else
1617 if (0 <= save_fd)
1618 emacs_close (save_fd);
1619 save_fd = fd;
1620 save_mtime = mtime;
1621 save_string = string;
1624 else
1626 /* We succeeded; return this descriptor and filename. */
1627 if (storeptr)
1628 *storeptr = string;
1629 SAFE_FREE ();
1630 return fd;
1634 /* No more suffixes. Return the newest. */
1635 if (0 <= save_fd && ! CONSP (XCDR (tail)))
1637 if (storeptr)
1638 *storeptr = save_string;
1639 SAFE_FREE ();
1640 return save_fd;
1644 if (absolute)
1645 break;
1648 SAFE_FREE ();
1649 errno = last_errno;
1650 return -1;
1654 /* Merge the list we've accumulated of globals from the current input source
1655 into the load_history variable. The details depend on whether
1656 the source has an associated file name or not.
1658 FILENAME is the file name that we are loading from.
1660 ENTIRE is true if loading that entire file, false if evaluating
1661 part of it. */
1663 static void
1664 build_load_history (Lisp_Object filename, bool entire)
1666 Lisp_Object tail, prev, newelt;
1667 Lisp_Object tem, tem2;
1668 bool foundit = 0;
1670 tail = Vload_history;
1671 prev = Qnil;
1673 while (CONSP (tail))
1675 tem = XCAR (tail);
1677 /* Find the feature's previous assoc list... */
1678 if (!NILP (Fequal (filename, Fcar (tem))))
1680 foundit = 1;
1682 /* If we're loading the entire file, remove old data. */
1683 if (entire)
1685 if (NILP (prev))
1686 Vload_history = XCDR (tail);
1687 else
1688 Fsetcdr (prev, XCDR (tail));
1691 /* Otherwise, cons on new symbols that are not already members. */
1692 else
1694 tem2 = Vcurrent_load_list;
1696 while (CONSP (tem2))
1698 newelt = XCAR (tem2);
1700 if (NILP (Fmember (newelt, tem)))
1701 Fsetcar (tail, Fcons (XCAR (tem),
1702 Fcons (newelt, XCDR (tem))));
1704 tem2 = XCDR (tem2);
1705 maybe_quit ();
1709 else
1710 prev = tail;
1711 tail = XCDR (tail);
1712 maybe_quit ();
1715 /* If we're loading an entire file, cons the new assoc onto the
1716 front of load-history, the most-recently-loaded position. Also
1717 do this if we didn't find an existing member for the file. */
1718 if (entire || !foundit)
1719 Vload_history = Fcons (Fnreverse (Vcurrent_load_list),
1720 Vload_history);
1723 static void
1724 readevalloop_1 (int old)
1726 load_convert_to_unibyte = old;
1729 /* Signal an `end-of-file' error, if possible with file name
1730 information. */
1732 static _Noreturn void
1733 end_of_file_error (void)
1735 if (STRINGP (Vload_file_name))
1736 xsignal1 (Qend_of_file, Vload_file_name);
1738 xsignal0 (Qend_of_file);
1741 static Lisp_Object
1742 readevalloop_eager_expand_eval (Lisp_Object val, Lisp_Object macroexpand)
1744 /* If we macroexpand the toplevel form non-recursively and it ends
1745 up being a `progn' (or if it was a progn to start), treat each
1746 form in the progn as a top-level form. This way, if one form in
1747 the progn defines a macro, that macro is in effect when we expand
1748 the remaining forms. See similar code in bytecomp.el. */
1749 val = call2 (macroexpand, val, Qnil);
1750 if (EQ (CAR_SAFE (val), Qprogn))
1752 Lisp_Object subforms = XCDR (val);
1754 for (val = Qnil; CONSP (subforms); subforms = XCDR (subforms))
1755 val = readevalloop_eager_expand_eval (XCAR (subforms),
1756 macroexpand);
1758 else
1759 val = eval_sub (call2 (macroexpand, val, Qt));
1760 return val;
1763 /* UNIBYTE specifies how to set load_convert_to_unibyte
1764 for this invocation.
1765 READFUN, if non-nil, is used instead of `read'.
1767 START, END specify region to read in current buffer (from eval-region).
1768 If the input is not from a buffer, they must be nil. */
1770 static void
1771 readevalloop (Lisp_Object readcharfun,
1772 FILE *stream,
1773 Lisp_Object sourcename,
1774 bool printflag,
1775 Lisp_Object unibyte, Lisp_Object readfun,
1776 Lisp_Object start, Lisp_Object end)
1778 int c;
1779 Lisp_Object val;
1780 ptrdiff_t count = SPECPDL_INDEX ();
1781 struct buffer *b = 0;
1782 bool continue_reading_p;
1783 Lisp_Object lex_bound;
1784 /* True if reading an entire buffer. */
1785 bool whole_buffer = 0;
1786 /* True on the first time around. */
1787 bool first_sexp = 1;
1788 Lisp_Object macroexpand = intern ("internal-macroexpand-for-load");
1790 if (NILP (Ffboundp (macroexpand))
1791 /* Don't macroexpand in .elc files, since it should have been done
1792 already. We actually don't know whether we're in a .elc file or not,
1793 so we use circumstantial evidence: .el files normally go through
1794 Vload_source_file_function -> load-with-code-conversion
1795 -> eval-buffer. */
1796 || EQ (readcharfun, Qget_file_char)
1797 || EQ (readcharfun, Qget_emacs_mule_file_char))
1798 macroexpand = Qnil;
1800 if (MARKERP (readcharfun))
1802 if (NILP (start))
1803 start = readcharfun;
1806 if (BUFFERP (readcharfun))
1807 b = XBUFFER (readcharfun);
1808 else if (MARKERP (readcharfun))
1809 b = XMARKER (readcharfun)->buffer;
1811 /* We assume START is nil when input is not from a buffer. */
1812 if (! NILP (start) && !b)
1813 emacs_abort ();
1815 specbind (Qstandard_input, readcharfun);
1816 specbind (Qcurrent_load_list, Qnil);
1817 record_unwind_protect_int (readevalloop_1, load_convert_to_unibyte);
1818 load_convert_to_unibyte = !NILP (unibyte);
1820 /* If lexical binding is active (either because it was specified in
1821 the file's header, or via a buffer-local variable), create an empty
1822 lexical environment, otherwise, turn off lexical binding. */
1823 lex_bound = find_symbol_value (Qlexical_binding);
1824 specbind (Qinternal_interpreter_environment,
1825 (NILP (lex_bound) || EQ (lex_bound, Qunbound)
1826 ? Qnil : list1 (Qt)));
1828 /* Try to ensure sourcename is a truename, except whilst preloading. */
1829 if (NILP (Vpurify_flag)
1830 && !NILP (sourcename) && !NILP (Ffile_name_absolute_p (sourcename))
1831 && !NILP (Ffboundp (Qfile_truename)))
1832 sourcename = call1 (Qfile_truename, sourcename) ;
1834 LOADHIST_ATTACH (sourcename);
1836 continue_reading_p = 1;
1837 while (continue_reading_p)
1839 ptrdiff_t count1 = SPECPDL_INDEX ();
1841 if (b != 0 && !BUFFER_LIVE_P (b))
1842 error ("Reading from killed buffer");
1844 if (!NILP (start))
1846 /* Switch to the buffer we are reading from. */
1847 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1848 set_buffer_internal (b);
1850 /* Save point in it. */
1851 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1852 /* Save ZV in it. */
1853 record_unwind_protect (save_restriction_restore, save_restriction_save ());
1854 /* Those get unbound after we read one expression. */
1856 /* Set point and ZV around stuff to be read. */
1857 Fgoto_char (start);
1858 if (!NILP (end))
1859 Fnarrow_to_region (make_number (BEGV), end);
1861 /* Just for cleanliness, convert END to a marker
1862 if it is an integer. */
1863 if (INTEGERP (end))
1864 end = Fpoint_max_marker ();
1867 /* On the first cycle, we can easily test here
1868 whether we are reading the whole buffer. */
1869 if (b && first_sexp)
1870 whole_buffer = (PT == BEG && ZV == Z);
1872 instream = stream;
1873 read_next:
1874 c = READCHAR;
1875 if (c == ';')
1877 while ((c = READCHAR) != '\n' && c != -1);
1878 goto read_next;
1880 if (c < 0)
1882 unbind_to (count1, Qnil);
1883 break;
1886 /* Ignore whitespace here, so we can detect eof. */
1887 if (c == ' ' || c == '\t' || c == '\n' || c == '\f' || c == '\r'
1888 || c == NO_BREAK_SPACE)
1889 goto read_next;
1891 if (!NILP (Vpurify_flag) && c == '(')
1893 val = read_list (0, readcharfun);
1895 else
1897 UNREAD (c);
1898 read_objects = Qnil;
1899 if (!NILP (readfun))
1901 val = call1 (readfun, readcharfun);
1903 /* If READCHARFUN has set point to ZV, we should
1904 stop reading, even if the form read sets point
1905 to a different value when evaluated. */
1906 if (BUFFERP (readcharfun))
1908 struct buffer *buf = XBUFFER (readcharfun);
1909 if (BUF_PT (buf) == BUF_ZV (buf))
1910 continue_reading_p = 0;
1913 else if (! NILP (Vload_read_function))
1914 val = call1 (Vload_read_function, readcharfun);
1915 else
1916 val = read_internal_start (readcharfun, Qnil, Qnil);
1919 if (!NILP (start) && continue_reading_p)
1920 start = Fpoint_marker ();
1922 /* Restore saved point and BEGV. */
1923 unbind_to (count1, Qnil);
1925 /* Now eval what we just read. */
1926 if (!NILP (macroexpand))
1927 val = readevalloop_eager_expand_eval (val, macroexpand);
1928 else
1929 val = eval_sub (val);
1931 if (printflag)
1933 Vvalues = Fcons (val, Vvalues);
1934 if (EQ (Vstandard_output, Qt))
1935 Fprin1 (val, Qnil);
1936 else
1937 Fprint (val, Qnil);
1940 first_sexp = 0;
1943 build_load_history (sourcename,
1944 stream || whole_buffer);
1946 unbind_to (count, Qnil);
1949 DEFUN ("eval-buffer", Feval_buffer, Seval_buffer, 0, 5, "",
1950 doc: /* Execute the accessible portion of current buffer as Lisp code.
1951 You can use \\[narrow-to-region] to limit the part of buffer to be evaluated.
1952 When called from a Lisp program (i.e., not interactively), this
1953 function accepts up to five optional arguments:
1954 BUFFER is the buffer to evaluate (nil means use current buffer),
1955 or a name of a buffer (a string).
1956 PRINTFLAG controls printing of output by any output functions in the
1957 evaluated code, such as `print', `princ', and `prin1':
1958 a value of nil means discard it; anything else is the stream to print to.
1959 See Info node `(elisp)Output Streams' for details on streams.
1960 FILENAME specifies the file name to use for `load-history'.
1961 UNIBYTE, if non-nil, specifies `load-convert-to-unibyte' for this
1962 invocation.
1963 DO-ALLOW-PRINT, if non-nil, specifies that output functions in the
1964 evaluated code should work normally even if PRINTFLAG is nil, in
1965 which case the output is displayed in the echo area.
1967 This function preserves the position of point. */)
1968 (Lisp_Object buffer, Lisp_Object printflag, Lisp_Object filename, Lisp_Object unibyte, Lisp_Object do_allow_print)
1970 ptrdiff_t count = SPECPDL_INDEX ();
1971 Lisp_Object tem, buf;
1973 if (NILP (buffer))
1974 buf = Fcurrent_buffer ();
1975 else
1976 buf = Fget_buffer (buffer);
1977 if (NILP (buf))
1978 error ("No such buffer");
1980 if (NILP (printflag) && NILP (do_allow_print))
1981 tem = Qsymbolp;
1982 else
1983 tem = printflag;
1985 if (NILP (filename))
1986 filename = BVAR (XBUFFER (buf), filename);
1988 specbind (Qeval_buffer_list, Fcons (buf, Veval_buffer_list));
1989 specbind (Qstandard_output, tem);
1990 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1991 BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
1992 specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
1993 readevalloop (buf, 0, filename,
1994 !NILP (printflag), unibyte, Qnil, Qnil, Qnil);
1995 unbind_to (count, Qnil);
1997 return Qnil;
2000 DEFUN ("eval-region", Feval_region, Seval_region, 2, 4, "r",
2001 doc: /* Execute the region as Lisp code.
2002 When called from programs, expects two arguments,
2003 giving starting and ending indices in the current buffer
2004 of the text to be executed.
2005 Programs can pass third argument PRINTFLAG which controls output:
2006 a value of nil means discard it; anything else is stream for printing it.
2007 See Info node `(elisp)Output Streams' for details on streams.
2008 Also the fourth argument READ-FUNCTION, if non-nil, is used
2009 instead of `read' to read each expression. It gets one argument
2010 which is the input stream for reading characters.
2012 This function does not move point. */)
2013 (Lisp_Object start, Lisp_Object end, Lisp_Object printflag, Lisp_Object read_function)
2015 /* FIXME: Do the eval-sexp-add-defvars dance! */
2016 ptrdiff_t count = SPECPDL_INDEX ();
2017 Lisp_Object tem, cbuf;
2019 cbuf = Fcurrent_buffer ();
2021 if (NILP (printflag))
2022 tem = Qsymbolp;
2023 else
2024 tem = printflag;
2025 specbind (Qstandard_output, tem);
2026 specbind (Qeval_buffer_list, Fcons (cbuf, Veval_buffer_list));
2028 /* `readevalloop' calls functions which check the type of start and end. */
2029 readevalloop (cbuf, 0, BVAR (XBUFFER (cbuf), filename),
2030 !NILP (printflag), Qnil, read_function,
2031 start, end);
2033 return unbind_to (count, Qnil);
2037 DEFUN ("read", Fread, Sread, 0, 1, 0,
2038 doc: /* Read one Lisp expression as text from STREAM, return as Lisp object.
2039 If STREAM is nil, use the value of `standard-input' (which see).
2040 STREAM or the value of `standard-input' may be:
2041 a buffer (read from point and advance it)
2042 a marker (read from where it points and advance it)
2043 a function (call it with no arguments for each character,
2044 call it with a char as argument to push a char back)
2045 a string (takes text from string, starting at the beginning)
2046 t (read text line using minibuffer and use it, or read from
2047 standard input in batch mode). */)
2048 (Lisp_Object stream)
2050 if (NILP (stream))
2051 stream = Vstandard_input;
2052 if (EQ (stream, Qt))
2053 stream = Qread_char;
2054 if (EQ (stream, Qread_char))
2055 /* FIXME: ?! When is this used !? */
2056 return call1 (intern ("read-minibuffer"),
2057 build_string ("Lisp expression: "));
2059 return read_internal_start (stream, Qnil, Qnil);
2062 DEFUN ("read-from-string", Fread_from_string, Sread_from_string, 1, 3, 0,
2063 doc: /* Read one Lisp expression which is represented as text by STRING.
2064 Returns a cons: (OBJECT-READ . FINAL-STRING-INDEX).
2065 FINAL-STRING-INDEX is an integer giving the position of the next
2066 remaining character in STRING. START and END optionally delimit
2067 a substring of STRING from which to read; they default to 0 and
2068 \(length STRING) respectively. Negative values are counted from
2069 the end of STRING. */)
2070 (Lisp_Object string, Lisp_Object start, Lisp_Object end)
2072 Lisp_Object ret;
2073 CHECK_STRING (string);
2074 /* `read_internal_start' sets `read_from_string_index'. */
2075 ret = read_internal_start (string, start, end);
2076 return Fcons (ret, make_number (read_from_string_index));
2079 /* Function to set up the global context we need in toplevel read
2080 calls. START and END only used when STREAM is a string. */
2081 static Lisp_Object
2082 read_internal_start (Lisp_Object stream, Lisp_Object start, Lisp_Object end)
2084 Lisp_Object retval;
2086 readchar_count = 0;
2087 new_backquote_flag = 0;
2088 read_objects = Qnil;
2089 if (EQ (Vread_with_symbol_positions, Qt)
2090 || EQ (Vread_with_symbol_positions, stream))
2091 Vread_symbol_positions_list = Qnil;
2093 if (STRINGP (stream)
2094 || ((CONSP (stream) && STRINGP (XCAR (stream)))))
2096 ptrdiff_t startval, endval;
2097 Lisp_Object string;
2099 if (STRINGP (stream))
2100 string = stream;
2101 else
2102 string = XCAR (stream);
2104 validate_subarray (string, start, end, SCHARS (string),
2105 &startval, &endval);
2107 read_from_string_index = startval;
2108 read_from_string_index_byte = string_char_to_byte (string, startval);
2109 read_from_string_limit = endval;
2112 retval = read0 (stream);
2113 if (EQ (Vread_with_symbol_positions, Qt)
2114 || EQ (Vread_with_symbol_positions, stream))
2115 Vread_symbol_positions_list = Fnreverse (Vread_symbol_positions_list);
2116 return retval;
2120 /* Signal Qinvalid_read_syntax error.
2121 S is error string of length N (if > 0) */
2123 static _Noreturn void
2124 invalid_syntax (const char *s)
2126 xsignal1 (Qinvalid_read_syntax, build_string (s));
2130 /* Use this for recursive reads, in contexts where internal tokens
2131 are not allowed. */
2133 static Lisp_Object
2134 read0 (Lisp_Object readcharfun)
2136 register Lisp_Object val;
2137 int c;
2139 val = read1 (readcharfun, &c, 0);
2140 if (!c)
2141 return val;
2143 xsignal1 (Qinvalid_read_syntax,
2144 Fmake_string (make_number (1), make_number (c)));
2147 /* Grow a read buffer BUF that contains OFFSET useful bytes of data,
2148 by at least MAX_MULTIBYTE_LENGTH bytes. Update *BUF_ADDR and
2149 *BUF_SIZE accordingly; 0 <= OFFSET <= *BUF_SIZE. If *BUF_ADDR is
2150 initially null, BUF is on the stack: copy its data to the new heap
2151 buffer. Otherwise, BUF must equal *BUF_ADDR and can simply be
2152 reallocated. Either way, remember the heap allocation (which is at
2153 pdl slot COUNT) so that it can be freed when unwinding the stack.*/
2155 static char *
2156 grow_read_buffer (char *buf, ptrdiff_t offset,
2157 char **buf_addr, ptrdiff_t *buf_size, ptrdiff_t count)
2159 char *p = xpalloc (*buf_addr, buf_size, MAX_MULTIBYTE_LENGTH, -1, 1);
2160 if (!*buf_addr)
2162 memcpy (p, buf, offset);
2163 record_unwind_protect_ptr (xfree, p);
2165 else
2166 set_unwind_protect_ptr (count, xfree, p);
2167 *buf_addr = p;
2168 return p;
2171 /* Return the scalar value that has the Unicode character name NAME.
2172 Raise 'invalid-read-syntax' if there is no such character. */
2173 static int
2174 character_name_to_code (char const *name, ptrdiff_t name_len)
2176 /* For "U+XXXX", pass the leading '+' to string_to_number to reject
2177 monstrosities like "U+-0000". */
2178 Lisp_Object code
2179 = (name[0] == 'U' && name[1] == '+'
2180 ? string_to_number (name + 1, 16, false)
2181 : call2 (Qchar_from_name, make_unibyte_string (name, name_len), Qt));
2183 if (! RANGED_INTEGERP (0, code, MAX_UNICODE_CHAR)
2184 || char_surrogate_p (XINT (code)))
2186 AUTO_STRING (format, "\\N{%s}");
2187 AUTO_STRING_WITH_LEN (namestr, name, name_len);
2188 xsignal1 (Qinvalid_read_syntax, CALLN (Fformat, format, namestr));
2191 return XINT (code);
2194 /* Bound on the length of a Unicode character name. As of
2195 Unicode 9.0.0 the maximum is 83, so this should be safe. */
2196 enum { UNICODE_CHARACTER_NAME_LENGTH_BOUND = 200 };
2198 /* Read a \-escape sequence, assuming we already read the `\'.
2199 If the escape sequence forces unibyte, return eight-bit char. */
2201 static int
2202 read_escape (Lisp_Object readcharfun, bool stringp)
2204 int c = READCHAR;
2205 /* \u allows up to four hex digits, \U up to eight. Default to the
2206 behavior for \u, and change this value in the case that \U is seen. */
2207 int unicode_hex_count = 4;
2209 switch (c)
2211 case -1:
2212 end_of_file_error ();
2214 case 'a':
2215 return '\007';
2216 case 'b':
2217 return '\b';
2218 case 'd':
2219 return 0177;
2220 case 'e':
2221 return 033;
2222 case 'f':
2223 return '\f';
2224 case 'n':
2225 return '\n';
2226 case 'r':
2227 return '\r';
2228 case 't':
2229 return '\t';
2230 case 'v':
2231 return '\v';
2232 case '\n':
2233 return -1;
2234 case ' ':
2235 if (stringp)
2236 return -1;
2237 return ' ';
2239 case 'M':
2240 c = READCHAR;
2241 if (c != '-')
2242 error ("Invalid escape character syntax");
2243 c = READCHAR;
2244 if (c == '\\')
2245 c = read_escape (readcharfun, 0);
2246 return c | meta_modifier;
2248 case 'S':
2249 c = READCHAR;
2250 if (c != '-')
2251 error ("Invalid escape character syntax");
2252 c = READCHAR;
2253 if (c == '\\')
2254 c = read_escape (readcharfun, 0);
2255 return c | shift_modifier;
2257 case 'H':
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 | hyper_modifier;
2266 case 'A':
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 | alt_modifier;
2275 case 's':
2276 c = READCHAR;
2277 if (stringp || c != '-')
2279 UNREAD (c);
2280 return ' ';
2282 c = READCHAR;
2283 if (c == '\\')
2284 c = read_escape (readcharfun, 0);
2285 return c | super_modifier;
2287 case 'C':
2288 c = READCHAR;
2289 if (c != '-')
2290 error ("Invalid escape character syntax");
2291 case '^':
2292 c = READCHAR;
2293 if (c == '\\')
2294 c = read_escape (readcharfun, 0);
2295 if ((c & ~CHAR_MODIFIER_MASK) == '?')
2296 return 0177 | (c & CHAR_MODIFIER_MASK);
2297 else if (! SINGLE_BYTE_CHAR_P ((c & ~CHAR_MODIFIER_MASK)))
2298 return c | ctrl_modifier;
2299 /* ASCII control chars are made from letters (both cases),
2300 as well as the non-letters within 0100...0137. */
2301 else if ((c & 0137) >= 0101 && (c & 0137) <= 0132)
2302 return (c & (037 | ~0177));
2303 else if ((c & 0177) >= 0100 && (c & 0177) <= 0137)
2304 return (c & (037 | ~0177));
2305 else
2306 return c | ctrl_modifier;
2308 case '0':
2309 case '1':
2310 case '2':
2311 case '3':
2312 case '4':
2313 case '5':
2314 case '6':
2315 case '7':
2316 /* An octal escape, as in ANSI C. */
2318 register int i = c - '0';
2319 register int count = 0;
2320 while (++count < 3)
2322 if ((c = READCHAR) >= '0' && c <= '7')
2324 i *= 8;
2325 i += c - '0';
2327 else
2329 UNREAD (c);
2330 break;
2334 if (i >= 0x80 && i < 0x100)
2335 i = BYTE8_TO_CHAR (i);
2336 return i;
2339 case 'x':
2340 /* A hex escape, as in ANSI C. */
2342 unsigned int i = 0;
2343 int count = 0;
2344 while (1)
2346 c = READCHAR;
2347 if (c >= '0' && c <= '9')
2349 i *= 16;
2350 i += c - '0';
2352 else if ((c >= 'a' && c <= 'f')
2353 || (c >= 'A' && c <= 'F'))
2355 i *= 16;
2356 if (c >= 'a' && c <= 'f')
2357 i += c - 'a' + 10;
2358 else
2359 i += c - 'A' + 10;
2361 else
2363 UNREAD (c);
2364 break;
2366 /* Allow hex escapes as large as ?\xfffffff, because some
2367 packages use them to denote characters with modifiers. */
2368 if ((CHAR_META | (CHAR_META - 1)) < i)
2369 error ("Hex character out of range: \\x%x...", i);
2370 count += count < 3;
2373 if (count < 3 && i >= 0x80)
2374 return BYTE8_TO_CHAR (i);
2375 return i;
2378 case 'U':
2379 /* Post-Unicode-2.0: Up to eight hex chars. */
2380 unicode_hex_count = 8;
2381 case 'u':
2383 /* A Unicode escape. We only permit them in strings and characters,
2384 not arbitrarily in the source code, as in some other languages. */
2386 unsigned int i = 0;
2387 int count = 0;
2389 while (++count <= unicode_hex_count)
2391 c = READCHAR;
2392 /* `isdigit' and `isalpha' may be locale-specific, which we don't
2393 want. */
2394 if (c >= '0' && c <= '9') i = (i << 4) + (c - '0');
2395 else if (c >= 'a' && c <= 'f') i = (i << 4) + (c - 'a') + 10;
2396 else if (c >= 'A' && c <= 'F') i = (i << 4) + (c - 'A') + 10;
2397 else
2398 error ("Non-hex digit used for Unicode escape");
2400 if (i > 0x10FFFF)
2401 error ("Non-Unicode character: 0x%x", i);
2402 return i;
2405 case 'N':
2406 /* Named character. */
2408 c = READCHAR;
2409 if (c != '{')
2410 invalid_syntax ("Expected opening brace after \\N");
2411 char name[UNICODE_CHARACTER_NAME_LENGTH_BOUND + 1];
2412 bool whitespace = false;
2413 ptrdiff_t length = 0;
2414 while (true)
2416 c = READCHAR;
2417 if (c < 0)
2418 end_of_file_error ();
2419 if (c == '}')
2420 break;
2421 if (! (0 < c && c < 0x80))
2423 AUTO_STRING (format,
2424 "Invalid character U+%04X in character name");
2425 xsignal1 (Qinvalid_read_syntax,
2426 CALLN (Fformat, format, make_natnum (c)));
2428 /* Treat multiple adjacent whitespace characters as a
2429 single space character. This makes it easier to use
2430 character names in e.g. multi-line strings. */
2431 if (c_isspace (c))
2433 if (whitespace)
2434 continue;
2435 c = ' ';
2436 whitespace = true;
2438 else
2439 whitespace = false;
2440 name[length++] = c;
2441 if (length >= sizeof name)
2442 invalid_syntax ("Character name too long");
2444 if (length == 0)
2445 invalid_syntax ("Empty character name");
2446 name[length] = '\0';
2448 /* character_name_to_code can invoke read1, recursively.
2449 This is why read1's buffer is not static. */
2450 return character_name_to_code (name, length);
2453 default:
2454 return c;
2458 /* Return the digit that CHARACTER stands for in the given BASE.
2459 Return -1 if CHARACTER is out of range for BASE,
2460 and -2 if CHARACTER is not valid for any supported BASE. */
2461 static int
2462 digit_to_number (int character, int base)
2464 int digit;
2466 if ('0' <= character && character <= '9')
2467 digit = character - '0';
2468 else if ('a' <= character && character <= 'z')
2469 digit = character - 'a' + 10;
2470 else if ('A' <= character && character <= 'Z')
2471 digit = character - 'A' + 10;
2472 else
2473 return -2;
2475 return digit < base ? digit : -1;
2478 /* Read an integer in radix RADIX using READCHARFUN to read
2479 characters. RADIX must be in the interval [2..36]; if it isn't, a
2480 read error is signaled . Value is the integer read. Signals an
2481 error if encountering invalid read syntax or if RADIX is out of
2482 range. */
2484 static Lisp_Object
2485 read_integer (Lisp_Object readcharfun, EMACS_INT radix)
2487 /* Room for sign, leading 0, other digits, trailing null byte.
2488 Also, room for invalid syntax diagnostic. */
2489 char buf[max (1 + 1 + UINTMAX_WIDTH + 1,
2490 sizeof "integer, radix " + INT_STRLEN_BOUND (EMACS_INT))];
2492 int valid = -1; /* 1 if valid, 0 if not, -1 if incomplete. */
2494 if (radix < 2 || radix > 36)
2495 valid = 0;
2496 else
2498 char *p = buf;
2499 int c, digit;
2501 c = READCHAR;
2502 if (c == '-' || c == '+')
2504 *p++ = c;
2505 c = READCHAR;
2508 if (c == '0')
2510 *p++ = c;
2511 valid = 1;
2513 /* Ignore redundant leading zeros, so the buffer doesn't
2514 fill up with them. */
2516 c = READCHAR;
2517 while (c == '0');
2520 while ((digit = digit_to_number (c, radix)) >= -1)
2522 if (digit == -1)
2523 valid = 0;
2524 if (valid < 0)
2525 valid = 1;
2527 if (p < buf + sizeof buf - 1)
2528 *p++ = c;
2529 else
2530 valid = 0;
2532 c = READCHAR;
2535 UNREAD (c);
2536 *p = '\0';
2539 if (valid != 1)
2541 sprintf (buf, "integer, radix %"pI"d", radix);
2542 invalid_syntax (buf);
2545 return string_to_number (buf, radix, 0);
2549 /* If the next token is ')' or ']' or '.', we store that character
2550 in *PCH and the return value is not interesting. Else, we store
2551 zero in *PCH and we read and return one lisp object.
2553 FIRST_IN_LIST is true if this is the first element of a list. */
2555 static Lisp_Object
2556 read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
2558 int c;
2559 bool uninterned_symbol = false;
2560 bool multibyte;
2561 char stackbuf[MAX_ALLOCA];
2563 *pch = 0;
2565 retry:
2567 c = READCHAR_REPORT_MULTIBYTE (&multibyte);
2568 if (c < 0)
2569 end_of_file_error ();
2571 switch (c)
2573 case '(':
2574 return read_list (0, readcharfun);
2576 case '[':
2577 return read_vector (readcharfun, 0);
2579 case ')':
2580 case ']':
2582 *pch = c;
2583 return Qnil;
2586 case '#':
2587 c = READCHAR;
2588 if (c == 's')
2590 c = READCHAR;
2591 if (c == '(')
2593 /* Accept extended format for hash tables (extensible to
2594 other types), e.g.
2595 #s(hash-table size 2 test equal data (k1 v1 k2 v2)) */
2596 Lisp_Object tmp = read_list (0, readcharfun);
2597 Lisp_Object head = CAR_SAFE (tmp);
2598 Lisp_Object data = Qnil;
2599 Lisp_Object val = Qnil;
2600 /* The size is 2 * number of allowed keywords to
2601 make-hash-table. */
2602 Lisp_Object params[10];
2603 Lisp_Object ht;
2604 Lisp_Object key = Qnil;
2605 int param_count = 0;
2607 if (!EQ (head, Qhash_table))
2608 error ("Invalid extended read marker at head of #s list "
2609 "(only hash-table allowed)");
2611 tmp = CDR_SAFE (tmp);
2613 /* This is repetitive but fast and simple. */
2614 params[param_count] = QCsize;
2615 params[param_count + 1] = Fplist_get (tmp, Qsize);
2616 if (!NILP (params[param_count + 1]))
2617 param_count += 2;
2619 params[param_count] = QCtest;
2620 params[param_count + 1] = Fplist_get (tmp, Qtest);
2621 if (!NILP (params[param_count + 1]))
2622 param_count += 2;
2624 params[param_count] = QCweakness;
2625 params[param_count + 1] = Fplist_get (tmp, Qweakness);
2626 if (!NILP (params[param_count + 1]))
2627 param_count += 2;
2629 params[param_count] = QCrehash_size;
2630 params[param_count + 1] = Fplist_get (tmp, Qrehash_size);
2631 if (!NILP (params[param_count + 1]))
2632 param_count += 2;
2634 params[param_count] = QCrehash_threshold;
2635 params[param_count + 1] = Fplist_get (tmp, Qrehash_threshold);
2636 if (!NILP (params[param_count + 1]))
2637 param_count += 2;
2639 /* This is the hash table data. */
2640 data = Fplist_get (tmp, Qdata);
2642 /* Now use params to make a new hash table and fill it. */
2643 ht = Fmake_hash_table (param_count, params);
2645 while (CONSP (data))
2647 key = XCAR (data);
2648 data = XCDR (data);
2649 if (!CONSP (data))
2650 error ("Odd number of elements in hash table data");
2651 val = XCAR (data);
2652 data = XCDR (data);
2653 Fputhash (key, val, ht);
2656 return ht;
2658 UNREAD (c);
2659 invalid_syntax ("#");
2661 if (c == '^')
2663 c = READCHAR;
2664 if (c == '[')
2666 Lisp_Object tmp;
2667 tmp = read_vector (readcharfun, 0);
2668 if (ASIZE (tmp) < CHAR_TABLE_STANDARD_SLOTS)
2669 error ("Invalid size char-table");
2670 XSETPVECTYPE (XVECTOR (tmp), PVEC_CHAR_TABLE);
2671 return tmp;
2673 else if (c == '^')
2675 c = READCHAR;
2676 if (c == '[')
2678 /* Sub char-table can't be read as a regular
2679 vector because of a two C integer fields. */
2680 Lisp_Object tbl, tmp = read_list (1, readcharfun);
2681 ptrdiff_t size = XINT (Flength (tmp));
2682 int i, depth, min_char;
2683 struct Lisp_Cons *cell;
2685 if (size == 0)
2686 error ("Zero-sized sub char-table");
2688 if (! RANGED_INTEGERP (1, XCAR (tmp), 3))
2689 error ("Invalid depth in sub char-table");
2690 depth = XINT (XCAR (tmp));
2691 if (chartab_size[depth] != size - 2)
2692 error ("Invalid size in sub char-table");
2693 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2694 free_cons (cell);
2696 if (! RANGED_INTEGERP (0, XCAR (tmp), MAX_CHAR))
2697 error ("Invalid minimum character in sub-char-table");
2698 min_char = XINT (XCAR (tmp));
2699 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2700 free_cons (cell);
2702 tbl = make_uninit_sub_char_table (depth, min_char);
2703 for (i = 0; i < size; i++)
2705 XSUB_CHAR_TABLE (tbl)->contents[i] = XCAR (tmp);
2706 cell = XCONS (tmp), tmp = XCDR (tmp);
2707 free_cons (cell);
2709 return tbl;
2711 invalid_syntax ("#^^");
2713 invalid_syntax ("#^");
2715 if (c == '&')
2717 Lisp_Object length;
2718 length = read1 (readcharfun, pch, first_in_list);
2719 c = READCHAR;
2720 if (c == '"')
2722 Lisp_Object tmp, val;
2723 EMACS_INT size_in_chars = bool_vector_bytes (XFASTINT (length));
2724 unsigned char *data;
2726 UNREAD (c);
2727 tmp = read1 (readcharfun, pch, first_in_list);
2728 if (STRING_MULTIBYTE (tmp)
2729 || (size_in_chars != SCHARS (tmp)
2730 /* We used to print 1 char too many
2731 when the number of bits was a multiple of 8.
2732 Accept such input in case it came from an old
2733 version. */
2734 && ! (XFASTINT (length)
2735 == (SCHARS (tmp) - 1) * BOOL_VECTOR_BITS_PER_CHAR)))
2736 invalid_syntax ("#&...");
2738 val = make_uninit_bool_vector (XFASTINT (length));
2739 data = bool_vector_uchar_data (val);
2740 memcpy (data, SDATA (tmp), size_in_chars);
2741 /* Clear the extraneous bits in the last byte. */
2742 if (XINT (length) != size_in_chars * BOOL_VECTOR_BITS_PER_CHAR)
2743 data[size_in_chars - 1]
2744 &= (1 << (XINT (length) % BOOL_VECTOR_BITS_PER_CHAR)) - 1;
2745 return val;
2747 invalid_syntax ("#&...");
2749 if (c == '[')
2751 /* Accept compiled functions at read-time so that we don't have to
2752 build them using function calls. */
2753 Lisp_Object tmp;
2754 struct Lisp_Vector *vec;
2755 tmp = read_vector (readcharfun, 1);
2756 vec = XVECTOR (tmp);
2757 if (vec->header.size == 0)
2758 invalid_syntax ("Empty byte-code object");
2759 make_byte_code (vec);
2760 return tmp;
2762 if (c == '(')
2764 Lisp_Object tmp;
2765 int ch;
2767 /* Read the string itself. */
2768 tmp = read1 (readcharfun, &ch, 0);
2769 if (ch != 0 || !STRINGP (tmp))
2770 invalid_syntax ("#");
2771 /* Read the intervals and their properties. */
2772 while (1)
2774 Lisp_Object beg, end, plist;
2776 beg = read1 (readcharfun, &ch, 0);
2777 end = plist = Qnil;
2778 if (ch == ')')
2779 break;
2780 if (ch == 0)
2781 end = read1 (readcharfun, &ch, 0);
2782 if (ch == 0)
2783 plist = read1 (readcharfun, &ch, 0);
2784 if (ch)
2785 invalid_syntax ("Invalid string property list");
2786 Fset_text_properties (beg, end, plist, tmp);
2789 return tmp;
2792 /* #@NUMBER is used to skip NUMBER following bytes.
2793 That's used in .elc files to skip over doc strings
2794 and function definitions. */
2795 if (c == '@')
2797 enum { extra = 100 };
2798 ptrdiff_t i, nskip = 0, digits = 0;
2800 /* Read a decimal integer. */
2801 while ((c = READCHAR) >= 0
2802 && c >= '0' && c <= '9')
2804 if ((STRING_BYTES_BOUND - extra) / 10 <= nskip)
2805 string_overflow ();
2806 digits++;
2807 nskip *= 10;
2808 nskip += c - '0';
2809 if (digits == 2 && nskip == 0)
2810 { /* We've just seen #@00, which means "skip to end". */
2811 skip_dyn_eof (readcharfun);
2812 return Qnil;
2815 if (nskip > 0)
2816 /* We can't use UNREAD here, because in the code below we side-step
2817 READCHAR. Instead, assume the first char after #@NNN occupies
2818 a single byte, which is the case normally since it's just
2819 a space. */
2820 nskip--;
2821 else
2822 UNREAD (c);
2824 if (load_force_doc_strings
2825 && (FROM_FILE_P (readcharfun)))
2827 /* If we are supposed to force doc strings into core right now,
2828 record the last string that we skipped,
2829 and record where in the file it comes from. */
2831 /* But first exchange saved_doc_string
2832 with prev_saved_doc_string, so we save two strings. */
2834 char *temp = saved_doc_string;
2835 ptrdiff_t temp_size = saved_doc_string_size;
2836 file_offset temp_pos = saved_doc_string_position;
2837 ptrdiff_t temp_len = saved_doc_string_length;
2839 saved_doc_string = prev_saved_doc_string;
2840 saved_doc_string_size = prev_saved_doc_string_size;
2841 saved_doc_string_position = prev_saved_doc_string_position;
2842 saved_doc_string_length = prev_saved_doc_string_length;
2844 prev_saved_doc_string = temp;
2845 prev_saved_doc_string_size = temp_size;
2846 prev_saved_doc_string_position = temp_pos;
2847 prev_saved_doc_string_length = temp_len;
2850 if (saved_doc_string_size == 0)
2852 saved_doc_string = xmalloc (nskip + extra);
2853 saved_doc_string_size = nskip + extra;
2855 if (nskip > saved_doc_string_size)
2857 saved_doc_string = xrealloc (saved_doc_string, nskip + extra);
2858 saved_doc_string_size = nskip + extra;
2861 saved_doc_string_position = file_tell (instream);
2863 /* Copy that many characters into saved_doc_string. */
2864 block_input ();
2865 for (i = 0; i < nskip && c >= 0; i++)
2866 saved_doc_string[i] = c = getc (instream);
2867 unblock_input ();
2869 saved_doc_string_length = i;
2871 else
2872 /* Skip that many bytes. */
2873 skip_dyn_bytes (readcharfun, nskip);
2875 goto retry;
2877 if (c == '!')
2879 /* #! appears at the beginning of an executable file.
2880 Skip the first line. */
2881 while (c != '\n' && c >= 0)
2882 c = READCHAR;
2883 goto retry;
2885 if (c == '$')
2886 return Vload_file_name;
2887 if (c == '\'')
2888 return list2 (Qfunction, read0 (readcharfun));
2889 /* #:foo is the uninterned symbol named foo. */
2890 if (c == ':')
2892 uninterned_symbol = true;
2893 c = READCHAR;
2894 if (!(c > 040
2895 && c != NO_BREAK_SPACE
2896 && (c >= 0200
2897 || strchr ("\"';()[]#`,", c) == NULL)))
2899 /* No symbol character follows, this is the empty
2900 symbol. */
2901 UNREAD (c);
2902 return Fmake_symbol (empty_unibyte_string);
2904 goto read_symbol;
2906 /* ## is the empty symbol. */
2907 if (c == '#')
2908 return Fintern (empty_unibyte_string, Qnil);
2909 /* Reader forms that can reuse previously read objects. */
2910 if (c >= '0' && c <= '9')
2912 EMACS_INT n = 0;
2913 Lisp_Object tem;
2914 bool overflow = false;
2916 /* Read a non-negative integer. */
2917 while (c >= '0' && c <= '9')
2919 overflow |= INT_MULTIPLY_WRAPV (n, 10, &n);
2920 overflow |= INT_ADD_WRAPV (n, c - '0', &n);
2921 c = READCHAR;
2924 if (!overflow && n <= MOST_POSITIVE_FIXNUM)
2926 if (c == 'r' || c == 'R')
2927 return read_integer (readcharfun, n);
2929 if (! NILP (Vread_circle))
2931 /* #n=object returns object, but associates it with
2932 n for #n#. */
2933 if (c == '=')
2935 /* Make a placeholder for #n# to use temporarily. */
2936 /* Note: We used to use AUTO_CONS to allocate
2937 placeholder, but that is a bad idea, since it
2938 will place a stack-allocated cons cell into
2939 the list in read_objects, which is a
2940 staticpro'd global variable, and thus each of
2941 its elements is marked during each GC. A
2942 stack-allocated object will become garbled
2943 when its stack slot goes out of scope, and
2944 some other function reuses it for entirely
2945 different purposes, which will cause crashes
2946 in GC. */
2947 Lisp_Object placeholder = Fcons (Qnil, Qnil);
2948 Lisp_Object cell = Fcons (make_number (n), placeholder);
2949 read_objects = Fcons (cell, read_objects);
2951 /* Read the object itself. */
2952 tem = read0 (readcharfun);
2954 /* Now put it everywhere the placeholder was... */
2955 substitute_object_in_subtree (tem, placeholder);
2957 /* ...and #n# will use the real value from now on. */
2958 Fsetcdr (cell, tem);
2960 return tem;
2963 /* #n# returns a previously read object. */
2964 if (c == '#')
2966 tem = Fassq (make_number (n), read_objects);
2967 if (CONSP (tem))
2968 return XCDR (tem);
2972 /* Fall through to error message. */
2974 else if (c == 'x' || c == 'X')
2975 return read_integer (readcharfun, 16);
2976 else if (c == 'o' || c == 'O')
2977 return read_integer (readcharfun, 8);
2978 else if (c == 'b' || c == 'B')
2979 return read_integer (readcharfun, 2);
2981 UNREAD (c);
2982 invalid_syntax ("#");
2984 case ';':
2985 while ((c = READCHAR) >= 0 && c != '\n');
2986 goto retry;
2988 case '\'':
2989 return list2 (Qquote, read0 (readcharfun));
2991 case '`':
2993 int next_char = READCHAR;
2994 UNREAD (next_char);
2995 /* Transition from old-style to new-style:
2996 If we see "(`" it used to mean old-style, which usually works
2997 fine because ` should almost never appear in such a position
2998 for new-style. But occasionally we need "(`" to mean new
2999 style, so we try to distinguish the two by the fact that we
3000 can either write "( `foo" or "(` foo", where the first
3001 intends to use new-style whereas the second intends to use
3002 old-style. For Emacs-25, we should completely remove this
3003 first_in_list exception (old-style can still be obtained via
3004 "(\`" anyway). */
3005 if (!new_backquote_flag && first_in_list && next_char == ' ')
3007 Vold_style_backquotes = Qt;
3008 goto default_label;
3010 else
3012 Lisp_Object value;
3013 bool saved_new_backquote_flag = new_backquote_flag;
3015 new_backquote_flag = 1;
3016 value = read0 (readcharfun);
3017 new_backquote_flag = saved_new_backquote_flag;
3019 return list2 (Qbackquote, value);
3022 case ',':
3024 int next_char = READCHAR;
3025 UNREAD (next_char);
3026 /* Transition from old-style to new-style:
3027 It used to be impossible to have a new-style , other than within
3028 a new-style `. This is sufficient when ` and , are used in the
3029 normal way, but ` and , can also appear in args to macros that
3030 will not interpret them in the usual way, in which case , may be
3031 used without any ` anywhere near.
3032 So we now use the same heuristic as for backquote: old-style
3033 unquotes are only recognized when first on a list, and when
3034 followed by a space.
3035 Because it's more difficult to peek 2 chars ahead, a new-style
3036 ,@ can still not be used outside of a `, unless it's in the middle
3037 of a list. */
3038 if (new_backquote_flag
3039 || !first_in_list
3040 || (next_char != ' ' && next_char != '@'))
3042 Lisp_Object comma_type = Qnil;
3043 Lisp_Object value;
3044 int ch = READCHAR;
3046 if (ch == '@')
3047 comma_type = Qcomma_at;
3048 else if (ch == '.')
3049 comma_type = Qcomma_dot;
3050 else
3052 if (ch >= 0) UNREAD (ch);
3053 comma_type = Qcomma;
3056 value = read0 (readcharfun);
3057 return list2 (comma_type, value);
3059 else
3061 Vold_style_backquotes = Qt;
3062 goto default_label;
3065 case '?':
3067 int modifiers;
3068 int next_char;
3069 bool ok;
3071 c = READCHAR;
3072 if (c < 0)
3073 end_of_file_error ();
3075 /* Accept `single space' syntax like (list ? x) where the
3076 whitespace character is SPC or TAB.
3077 Other literal whitespace like NL, CR, and FF are not accepted,
3078 as there are well-established escape sequences for these. */
3079 if (c == ' ' || c == '\t')
3080 return make_number (c);
3082 if (c == '\\')
3083 c = read_escape (readcharfun, 0);
3084 modifiers = c & CHAR_MODIFIER_MASK;
3085 c &= ~CHAR_MODIFIER_MASK;
3086 if (CHAR_BYTE8_P (c))
3087 c = CHAR_TO_BYTE8 (c);
3088 c |= modifiers;
3090 next_char = READCHAR;
3091 ok = (next_char <= 040
3092 || (next_char < 0200
3093 && strchr ("\"';()[]#?`,.", next_char) != NULL));
3094 UNREAD (next_char);
3095 if (ok)
3096 return make_number (c);
3098 invalid_syntax ("?");
3101 case '"':
3103 ptrdiff_t count = SPECPDL_INDEX ();
3104 char *read_buffer = stackbuf;
3105 ptrdiff_t read_buffer_size = sizeof stackbuf;
3106 char *heapbuf = NULL;
3107 char *p = read_buffer;
3108 char *end = read_buffer + read_buffer_size;
3109 int ch;
3110 /* True if we saw an escape sequence specifying
3111 a multibyte character. */
3112 bool force_multibyte = false;
3113 /* True if we saw an escape sequence specifying
3114 a single-byte character. */
3115 bool force_singlebyte = false;
3116 bool cancel = false;
3117 ptrdiff_t nchars = 0;
3119 while ((ch = READCHAR) >= 0
3120 && ch != '\"')
3122 if (end - p < MAX_MULTIBYTE_LENGTH)
3124 ptrdiff_t offset = p - read_buffer;
3125 read_buffer = grow_read_buffer (read_buffer, offset,
3126 &heapbuf, &read_buffer_size,
3127 count);
3128 p = read_buffer + offset;
3129 end = read_buffer + read_buffer_size;
3132 if (ch == '\\')
3134 int modifiers;
3136 ch = read_escape (readcharfun, 1);
3138 /* CH is -1 if \ newline or \ space has just been seen. */
3139 if (ch == -1)
3141 if (p == read_buffer)
3142 cancel = true;
3143 continue;
3146 modifiers = ch & CHAR_MODIFIER_MASK;
3147 ch = ch & ~CHAR_MODIFIER_MASK;
3149 if (CHAR_BYTE8_P (ch))
3150 force_singlebyte = true;
3151 else if (! ASCII_CHAR_P (ch))
3152 force_multibyte = true;
3153 else /* I.e. ASCII_CHAR_P (ch). */
3155 /* Allow `\C- ' and `\C-?'. */
3156 if (modifiers == CHAR_CTL)
3158 if (ch == ' ')
3159 ch = 0, modifiers = 0;
3160 else if (ch == '?')
3161 ch = 127, modifiers = 0;
3163 if (modifiers & CHAR_SHIFT)
3165 /* Shift modifier is valid only with [A-Za-z]. */
3166 if (ch >= 'A' && ch <= 'Z')
3167 modifiers &= ~CHAR_SHIFT;
3168 else if (ch >= 'a' && ch <= 'z')
3169 ch -= ('a' - 'A'), modifiers &= ~CHAR_SHIFT;
3172 if (modifiers & CHAR_META)
3174 /* Move the meta bit to the right place for a
3175 string. */
3176 modifiers &= ~CHAR_META;
3177 ch = BYTE8_TO_CHAR (ch | 0x80);
3178 force_singlebyte = true;
3182 /* Any modifiers remaining are invalid. */
3183 if (modifiers)
3184 error ("Invalid modifier in string");
3185 p += CHAR_STRING (ch, (unsigned char *) p);
3187 else
3189 p += CHAR_STRING (ch, (unsigned char *) p);
3190 if (CHAR_BYTE8_P (ch))
3191 force_singlebyte = true;
3192 else if (! ASCII_CHAR_P (ch))
3193 force_multibyte = true;
3195 nchars++;
3198 if (ch < 0)
3199 end_of_file_error ();
3201 /* If purifying, and string starts with \ newline,
3202 return zero instead. This is for doc strings
3203 that we are really going to find in etc/DOC.nn.nn. */
3204 if (!NILP (Vpurify_flag) && NILP (Vdoc_file_name) && cancel)
3205 return unbind_to (count, make_number (0));
3207 if (! force_multibyte && force_singlebyte)
3209 /* READ_BUFFER contains raw 8-bit bytes and no multibyte
3210 forms. Convert it to unibyte. */
3211 nchars = str_as_unibyte ((unsigned char *) read_buffer,
3212 p - read_buffer);
3213 p = read_buffer + nchars;
3216 Lisp_Object result
3217 = make_specified_string (read_buffer, nchars, p - read_buffer,
3218 (force_multibyte
3219 || (p - read_buffer != nchars)));
3220 return unbind_to (count, result);
3223 case '.':
3225 int next_char = READCHAR;
3226 UNREAD (next_char);
3228 if (next_char <= 040
3229 || (next_char < 0200
3230 && strchr ("\"';([#?`,", next_char) != NULL))
3232 *pch = c;
3233 return Qnil;
3236 /* Otherwise, we fall through! Note that the atom-reading loop
3237 below will now loop at least once, assuring that we will not
3238 try to UNREAD two characters in a row. */
3240 default:
3241 default_label:
3242 if (c <= 040) goto retry;
3243 if (c == NO_BREAK_SPACE)
3244 goto retry;
3246 read_symbol:
3248 ptrdiff_t count = SPECPDL_INDEX ();
3249 char *read_buffer = stackbuf;
3250 ptrdiff_t read_buffer_size = sizeof stackbuf;
3251 char *heapbuf = NULL;
3252 char *p = read_buffer;
3253 char *end = read_buffer + read_buffer_size;
3254 bool quoted = false;
3255 EMACS_INT start_position = readchar_count - 1;
3259 if (end - p < MAX_MULTIBYTE_LENGTH + 1)
3261 ptrdiff_t offset = p - read_buffer;
3262 read_buffer = grow_read_buffer (read_buffer, offset,
3263 &heapbuf, &read_buffer_size,
3264 count);
3265 p = read_buffer + offset;
3266 end = read_buffer + read_buffer_size;
3269 if (c == '\\')
3271 c = READCHAR;
3272 if (c == -1)
3273 end_of_file_error ();
3274 quoted = true;
3277 if (multibyte)
3278 p += CHAR_STRING (c, (unsigned char *) p);
3279 else
3280 *p++ = c;
3281 c = READCHAR;
3283 while (c > 040
3284 && c != NO_BREAK_SPACE
3285 && (c >= 0200
3286 || strchr ("\"';()[]#`,", c) == NULL));
3288 *p = 0;
3289 UNREAD (c);
3291 if (!quoted && !uninterned_symbol)
3293 Lisp_Object result = string_to_number (read_buffer, 10, 0);
3294 if (! NILP (result))
3295 return unbind_to (count, result);
3298 ptrdiff_t nbytes = p - read_buffer;
3299 ptrdiff_t nchars
3300 = (multibyte
3301 ? multibyte_chars_in_text ((unsigned char *) read_buffer,
3302 nbytes)
3303 : nbytes);
3304 Lisp_Object name = ((uninterned_symbol && ! NILP (Vpurify_flag)
3305 ? make_pure_string : make_specified_string)
3306 (read_buffer, nchars, nbytes, multibyte));
3307 Lisp_Object result = (uninterned_symbol ? Fmake_symbol (name)
3308 : Fintern (name, Qnil));
3310 if (EQ (Vread_with_symbol_positions, Qt)
3311 || EQ (Vread_with_symbol_positions, readcharfun))
3312 Vread_symbol_positions_list
3313 = Fcons (Fcons (result, make_number (start_position)),
3314 Vread_symbol_positions_list);
3315 return unbind_to (count, result);
3321 /* List of nodes we've seen during substitute_object_in_subtree. */
3322 static Lisp_Object seen_list;
3324 static void
3325 substitute_object_in_subtree (Lisp_Object object, Lisp_Object placeholder)
3327 Lisp_Object check_object;
3329 /* We haven't seen any objects when we start. */
3330 seen_list = Qnil;
3332 /* Make all the substitutions. */
3333 check_object
3334 = substitute_object_recurse (object, placeholder, object);
3336 /* Clear seen_list because we're done with it. */
3337 seen_list = Qnil;
3339 /* The returned object here is expected to always eq the
3340 original. */
3341 if (!EQ (check_object, object))
3342 error ("Unexpected mutation error in reader");
3345 /* Feval doesn't get called from here, so no gc protection is needed. */
3346 #define SUBSTITUTE(get_val, set_val) \
3347 do { \
3348 Lisp_Object old_value = get_val; \
3349 Lisp_Object true_value \
3350 = substitute_object_recurse (object, placeholder, \
3351 old_value); \
3353 if (!EQ (old_value, true_value)) \
3355 set_val; \
3357 } while (0)
3359 static Lisp_Object
3360 substitute_object_recurse (Lisp_Object object, Lisp_Object placeholder, Lisp_Object subtree)
3362 /* If we find the placeholder, return the target object. */
3363 if (EQ (placeholder, subtree))
3364 return object;
3366 /* If we've been to this node before, don't explore it again. */
3367 if (!EQ (Qnil, Fmemq (subtree, seen_list)))
3368 return subtree;
3370 /* If this node can be the entry point to a cycle, remember that
3371 we've seen it. It can only be such an entry point if it was made
3372 by #n=, which means that we can find it as a value in
3373 read_objects. */
3374 if (!EQ (Qnil, Frassq (subtree, read_objects)))
3375 seen_list = Fcons (subtree, seen_list);
3377 /* Recurse according to subtree's type.
3378 Every branch must return a Lisp_Object. */
3379 switch (XTYPE (subtree))
3381 case Lisp_Vectorlike:
3383 ptrdiff_t i = 0, length = 0;
3384 if (BOOL_VECTOR_P (subtree))
3385 return subtree; /* No sub-objects anyway. */
3386 else if (CHAR_TABLE_P (subtree) || SUB_CHAR_TABLE_P (subtree)
3387 || COMPILEDP (subtree) || HASH_TABLE_P (subtree))
3388 length = ASIZE (subtree) & PSEUDOVECTOR_SIZE_MASK;
3389 else if (VECTORP (subtree))
3390 length = ASIZE (subtree);
3391 else
3392 /* An unknown pseudovector may contain non-Lisp fields, so we
3393 can't just blindly traverse all its fields. We used to call
3394 `Flength' which signaled `sequencep', so I just preserved this
3395 behavior. */
3396 wrong_type_argument (Qsequencep, subtree);
3398 if (SUB_CHAR_TABLE_P (subtree))
3399 i = 2;
3400 for ( ; i < length; i++)
3401 SUBSTITUTE (AREF (subtree, i),
3402 ASET (subtree, i, true_value));
3403 return subtree;
3406 case Lisp_Cons:
3408 SUBSTITUTE (XCAR (subtree),
3409 XSETCAR (subtree, true_value));
3410 SUBSTITUTE (XCDR (subtree),
3411 XSETCDR (subtree, true_value));
3412 return subtree;
3415 case Lisp_String:
3417 /* Check for text properties in each interval.
3418 substitute_in_interval contains part of the logic. */
3420 INTERVAL root_interval = string_intervals (subtree);
3421 AUTO_CONS (arg, object, placeholder);
3423 traverse_intervals_noorder (root_interval,
3424 &substitute_in_interval, arg);
3426 return subtree;
3429 /* Other types don't recurse any further. */
3430 default:
3431 return subtree;
3435 /* Helper function for substitute_object_recurse. */
3436 static void
3437 substitute_in_interval (INTERVAL interval, Lisp_Object arg)
3439 Lisp_Object object = Fcar (arg);
3440 Lisp_Object placeholder = Fcdr (arg);
3442 SUBSTITUTE (interval->plist, set_interval_plist (interval, true_value));
3446 #define LEAD_INT 1
3447 #define DOT_CHAR 2
3448 #define TRAIL_INT 4
3449 #define E_EXP 16
3452 /* Convert STRING to a number, assuming base BASE. Return a fixnum if CP has
3453 integer syntax and fits in a fixnum, else return the nearest float if CP has
3454 either floating point or integer syntax and BASE is 10, else return nil. If
3455 IGNORE_TRAILING, consider just the longest prefix of CP that has
3456 valid floating point syntax. Signal an overflow if BASE is not 10 and the
3457 number has integer syntax but does not fit. */
3459 Lisp_Object
3460 string_to_number (char const *string, int base, bool ignore_trailing)
3462 int state;
3463 char const *cp = string;
3464 int leading_digit;
3465 bool float_syntax = 0;
3466 double value = 0;
3468 /* Negate the value ourselves. This treats 0, NaNs, and infinity properly on
3469 IEEE floating point hosts, and works around a formerly-common bug where
3470 atof ("-0.0") drops the sign. */
3471 bool negative = *cp == '-';
3473 bool signedp = negative || *cp == '+';
3474 cp += signedp;
3476 state = 0;
3478 leading_digit = digit_to_number (*cp, base);
3479 if (leading_digit >= 0)
3481 state |= LEAD_INT;
3483 ++cp;
3484 while (digit_to_number (*cp, base) >= 0);
3486 if (*cp == '.')
3488 state |= DOT_CHAR;
3489 cp++;
3492 if (base == 10)
3494 if ('0' <= *cp && *cp <= '9')
3496 state |= TRAIL_INT;
3498 cp++;
3499 while ('0' <= *cp && *cp <= '9');
3501 if (*cp == 'e' || *cp == 'E')
3503 char const *ecp = cp;
3504 cp++;
3505 if (*cp == '+' || *cp == '-')
3506 cp++;
3507 if ('0' <= *cp && *cp <= '9')
3509 state |= E_EXP;
3511 cp++;
3512 while ('0' <= *cp && *cp <= '9');
3514 else if (cp[-1] == '+'
3515 && cp[0] == 'I' && cp[1] == 'N' && cp[2] == 'F')
3517 state |= E_EXP;
3518 cp += 3;
3519 value = INFINITY;
3521 else if (cp[-1] == '+'
3522 && cp[0] == 'N' && cp[1] == 'a' && cp[2] == 'N')
3524 state |= E_EXP;
3525 cp += 3;
3526 /* NAN is a "positive" NaN on all known Emacs hosts. */
3527 value = NAN;
3529 else
3530 cp = ecp;
3533 float_syntax = ((state & (DOT_CHAR|TRAIL_INT)) == (DOT_CHAR|TRAIL_INT)
3534 || state == (LEAD_INT|E_EXP));
3537 /* Return nil if the number uses invalid syntax. If IGNORE_TRAILING, accept
3538 any prefix that matches. Otherwise, the entire string must match. */
3539 if (! (ignore_trailing
3540 ? ((state & LEAD_INT) != 0 || float_syntax)
3541 : (!*cp && ((state & ~DOT_CHAR) == LEAD_INT || float_syntax))))
3542 return Qnil;
3544 /* If the number uses integer and not float syntax, and is in C-language
3545 range, use its value, preferably as a fixnum. */
3546 if (leading_digit >= 0 && ! float_syntax)
3548 uintmax_t n;
3550 /* Fast special case for single-digit integers. This also avoids a
3551 glitch when BASE is 16 and IGNORE_TRAILING, because in that
3552 case some versions of strtoumax accept numbers like "0x1" that Emacs
3553 does not allow. */
3554 if (digit_to_number (string[signedp + 1], base) < 0)
3555 return make_number (negative ? -leading_digit : leading_digit);
3557 errno = 0;
3558 n = strtoumax (string + signedp, NULL, base);
3559 if (errno == ERANGE)
3561 /* Unfortunately there's no simple and accurate way to convert
3562 non-base-10 numbers that are out of C-language range. */
3563 if (base != 10)
3564 xsignal1 (Qoverflow_error, build_string (string));
3566 else if (n <= (negative ? -MOST_NEGATIVE_FIXNUM : MOST_POSITIVE_FIXNUM))
3568 EMACS_INT signed_n = n;
3569 return make_number (negative ? -signed_n : signed_n);
3571 else
3572 value = n;
3575 /* Either the number uses float syntax, or it does not fit into a fixnum.
3576 Convert it from string to floating point, unless the value is already
3577 known because it is an infinity, a NAN, or its absolute value fits in
3578 uintmax_t. */
3579 if (! value)
3580 value = atof (string + signedp);
3582 return make_float (negative ? -value : value);
3586 static Lisp_Object
3587 read_vector (Lisp_Object readcharfun, bool bytecodeflag)
3589 ptrdiff_t i, size;
3590 Lisp_Object *ptr;
3591 Lisp_Object tem, item, vector;
3592 struct Lisp_Cons *otem;
3593 Lisp_Object len;
3595 tem = read_list (1, readcharfun);
3596 len = Flength (tem);
3597 vector = Fmake_vector (len, Qnil);
3599 size = ASIZE (vector);
3600 ptr = XVECTOR (vector)->contents;
3601 for (i = 0; i < size; i++)
3603 item = Fcar (tem);
3604 /* If `load-force-doc-strings' is t when reading a lazily-loaded
3605 bytecode object, the docstring containing the bytecode and
3606 constants values must be treated as unibyte and passed to
3607 Fread, to get the actual bytecode string and constants vector. */
3608 if (bytecodeflag && load_force_doc_strings)
3610 if (i == COMPILED_BYTECODE)
3612 if (!STRINGP (item))
3613 error ("Invalid byte code");
3615 /* Delay handling the bytecode slot until we know whether
3616 it is lazily-loaded (we can tell by whether the
3617 constants slot is nil). */
3618 ASET (vector, COMPILED_CONSTANTS, item);
3619 item = Qnil;
3621 else if (i == COMPILED_CONSTANTS)
3623 Lisp_Object bytestr = ptr[COMPILED_CONSTANTS];
3625 if (NILP (item))
3627 /* Coerce string to unibyte (like string-as-unibyte,
3628 but without generating extra garbage and
3629 guaranteeing no change in the contents). */
3630 STRING_SET_CHARS (bytestr, SBYTES (bytestr));
3631 STRING_SET_UNIBYTE (bytestr);
3633 item = Fread (Fcons (bytestr, readcharfun));
3634 if (!CONSP (item))
3635 error ("Invalid byte code");
3637 otem = XCONS (item);
3638 bytestr = XCAR (item);
3639 item = XCDR (item);
3640 free_cons (otem);
3643 /* Now handle the bytecode slot. */
3644 ASET (vector, COMPILED_BYTECODE, bytestr);
3646 else if (i == COMPILED_DOC_STRING
3647 && STRINGP (item)
3648 && ! STRING_MULTIBYTE (item))
3650 if (EQ (readcharfun, Qget_emacs_mule_file_char))
3651 item = Fdecode_coding_string (item, Qemacs_mule, Qnil, Qnil);
3652 else
3653 item = Fstring_as_multibyte (item);
3656 ASET (vector, i, item);
3657 otem = XCONS (tem);
3658 tem = Fcdr (tem);
3659 free_cons (otem);
3661 return vector;
3664 /* FLAG means check for ']' to terminate rather than ')' and '.'. */
3666 static Lisp_Object
3667 read_list (bool flag, Lisp_Object readcharfun)
3669 Lisp_Object val, tail;
3670 Lisp_Object elt, tem;
3671 /* 0 is the normal case.
3672 1 means this list is a doc reference; replace it with the number 0.
3673 2 means this list is a doc reference; replace it with the doc string. */
3674 int doc_reference = 0;
3676 /* Initialize this to 1 if we are reading a list. */
3677 bool first_in_list = flag <= 0;
3679 val = Qnil;
3680 tail = Qnil;
3682 while (1)
3684 int ch;
3685 elt = read1 (readcharfun, &ch, first_in_list);
3687 first_in_list = 0;
3689 /* While building, if the list starts with #$, treat it specially. */
3690 if (EQ (elt, Vload_file_name)
3691 && ! NILP (elt)
3692 && !NILP (Vpurify_flag))
3694 if (NILP (Vdoc_file_name))
3695 /* We have not yet called Snarf-documentation, so assume
3696 this file is described in the DOC file
3697 and Snarf-documentation will fill in the right value later.
3698 For now, replace the whole list with 0. */
3699 doc_reference = 1;
3700 else
3701 /* We have already called Snarf-documentation, so make a relative
3702 file name for this file, so it can be found properly
3703 in the installed Lisp directory.
3704 We don't use Fexpand_file_name because that would make
3705 the directory absolute now. */
3707 AUTO_STRING (dot_dot_lisp, "../lisp/");
3708 elt = concat2 (dot_dot_lisp, Ffile_name_nondirectory (elt));
3711 else if (EQ (elt, Vload_file_name)
3712 && ! NILP (elt)
3713 && load_force_doc_strings)
3714 doc_reference = 2;
3716 if (ch)
3718 if (flag > 0)
3720 if (ch == ']')
3721 return val;
3722 invalid_syntax (") or . in a vector");
3724 if (ch == ')')
3725 return val;
3726 if (ch == '.')
3728 if (!NILP (tail))
3729 XSETCDR (tail, read0 (readcharfun));
3730 else
3731 val = read0 (readcharfun);
3732 read1 (readcharfun, &ch, 0);
3734 if (ch == ')')
3736 if (doc_reference == 1)
3737 return make_number (0);
3738 if (doc_reference == 2 && INTEGERP (XCDR (val)))
3740 char *saved = NULL;
3741 file_offset saved_position;
3742 /* Get a doc string from the file we are loading.
3743 If it's in saved_doc_string, get it from there.
3745 Here, we don't know if the string is a
3746 bytecode string or a doc string. As a
3747 bytecode string must be unibyte, we always
3748 return a unibyte string. If it is actually a
3749 doc string, caller must make it
3750 multibyte. */
3752 /* Position is negative for user variables. */
3753 EMACS_INT pos = eabs (XINT (XCDR (val)));
3754 if (pos >= saved_doc_string_position
3755 && pos < (saved_doc_string_position
3756 + saved_doc_string_length))
3758 saved = saved_doc_string;
3759 saved_position = saved_doc_string_position;
3761 /* Look in prev_saved_doc_string the same way. */
3762 else if (pos >= prev_saved_doc_string_position
3763 && pos < (prev_saved_doc_string_position
3764 + prev_saved_doc_string_length))
3766 saved = prev_saved_doc_string;
3767 saved_position = prev_saved_doc_string_position;
3769 if (saved)
3771 ptrdiff_t start = pos - saved_position;
3772 ptrdiff_t from, to;
3774 /* Process quoting with ^A,
3775 and find the end of the string,
3776 which is marked with ^_ (037). */
3777 for (from = start, to = start;
3778 saved[from] != 037;)
3780 int c = saved[from++];
3781 if (c == 1)
3783 c = saved[from++];
3784 saved[to++] = (c == 1 ? c
3785 : c == '0' ? 0
3786 : c == '_' ? 037
3787 : c);
3789 else
3790 saved[to++] = c;
3793 return make_unibyte_string (saved + start,
3794 to - start);
3796 else
3797 return get_doc_string (val, 1, 0);
3800 return val;
3802 invalid_syntax (". in wrong context");
3804 invalid_syntax ("] in a list");
3806 tem = list1 (elt);
3807 if (!NILP (tail))
3808 XSETCDR (tail, tem);
3809 else
3810 val = tem;
3811 tail = tem;
3815 static Lisp_Object initial_obarray;
3817 /* `oblookup' stores the bucket number here, for the sake of Funintern. */
3819 static size_t oblookup_last_bucket_number;
3821 /* Get an error if OBARRAY is not an obarray.
3822 If it is one, return it. */
3824 Lisp_Object
3825 check_obarray (Lisp_Object obarray)
3827 /* We don't want to signal a wrong-type-argument error when we are
3828 shutting down due to a fatal error, and we don't want to hit
3829 assertions in VECTORP and ASIZE if the fatal error was during GC. */
3830 if (!fatal_error_in_progress
3831 && (!VECTORP (obarray) || ASIZE (obarray) == 0))
3833 /* If Vobarray is now invalid, force it to be valid. */
3834 if (EQ (Vobarray, obarray)) Vobarray = initial_obarray;
3835 wrong_type_argument (Qvectorp, obarray);
3837 return obarray;
3840 /* Intern symbol SYM in OBARRAY using bucket INDEX. */
3842 static Lisp_Object
3843 intern_sym (Lisp_Object sym, Lisp_Object obarray, Lisp_Object index)
3845 Lisp_Object *ptr;
3847 XSYMBOL (sym)->interned = (EQ (obarray, initial_obarray)
3848 ? SYMBOL_INTERNED_IN_INITIAL_OBARRAY
3849 : SYMBOL_INTERNED);
3851 if (SREF (SYMBOL_NAME (sym), 0) == ':' && EQ (obarray, initial_obarray))
3853 make_symbol_constant (sym);
3854 XSYMBOL (sym)->redirect = SYMBOL_PLAINVAL;
3855 SET_SYMBOL_VAL (XSYMBOL (sym), sym);
3858 ptr = aref_addr (obarray, XINT (index));
3859 set_symbol_next (sym, SYMBOLP (*ptr) ? XSYMBOL (*ptr) : NULL);
3860 *ptr = sym;
3861 return sym;
3864 /* Intern a symbol with name STRING in OBARRAY using bucket INDEX. */
3866 Lisp_Object
3867 intern_driver (Lisp_Object string, Lisp_Object obarray, Lisp_Object index)
3869 return intern_sym (Fmake_symbol (string), obarray, index);
3872 /* Intern the C string STR: return a symbol with that name,
3873 interned in the current obarray. */
3875 Lisp_Object
3876 intern_1 (const char *str, ptrdiff_t len)
3878 Lisp_Object obarray = check_obarray (Vobarray);
3879 Lisp_Object tem = oblookup (obarray, str, len, len);
3881 return (SYMBOLP (tem) ? tem
3882 /* The above `oblookup' was done on the basis of nchars==nbytes, so
3883 the string has to be unibyte. */
3884 : intern_driver (make_unibyte_string (str, len),
3885 obarray, tem));
3888 Lisp_Object
3889 intern_c_string_1 (const char *str, ptrdiff_t len)
3891 Lisp_Object obarray = check_obarray (Vobarray);
3892 Lisp_Object tem = oblookup (obarray, str, len, len);
3894 if (!SYMBOLP (tem))
3896 /* Creating a non-pure string from a string literal not implemented yet.
3897 We could just use make_string here and live with the extra copy. */
3898 eassert (!NILP (Vpurify_flag));
3899 tem = intern_driver (make_pure_c_string (str, len), obarray, tem);
3901 return tem;
3904 static void
3905 define_symbol (Lisp_Object sym, char const *str)
3907 ptrdiff_t len = strlen (str);
3908 Lisp_Object string = make_pure_c_string (str, len);
3909 init_symbol (sym, string);
3911 /* Qunbound is uninterned, so that it's not confused with any symbol
3912 'unbound' created by a Lisp program. */
3913 if (! EQ (sym, Qunbound))
3915 Lisp_Object bucket = oblookup (initial_obarray, str, len, len);
3916 eassert (INTEGERP (bucket));
3917 intern_sym (sym, initial_obarray, bucket);
3921 DEFUN ("intern", Fintern, Sintern, 1, 2, 0,
3922 doc: /* Return the canonical symbol whose name is STRING.
3923 If there is none, one is created by this function and returned.
3924 A second optional argument specifies the obarray to use;
3925 it defaults to the value of `obarray'. */)
3926 (Lisp_Object string, Lisp_Object obarray)
3928 Lisp_Object tem;
3930 obarray = check_obarray (NILP (obarray) ? Vobarray : obarray);
3931 CHECK_STRING (string);
3933 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3934 if (!SYMBOLP (tem))
3935 tem = intern_driver (NILP (Vpurify_flag) ? string : Fpurecopy (string),
3936 obarray, tem);
3937 return tem;
3940 DEFUN ("intern-soft", Fintern_soft, Sintern_soft, 1, 2, 0,
3941 doc: /* Return the canonical symbol named NAME, or nil if none exists.
3942 NAME may be a string or a symbol. If it is a symbol, that exact
3943 symbol is searched for.
3944 A second optional argument specifies the obarray to use;
3945 it defaults to the value of `obarray'. */)
3946 (Lisp_Object name, Lisp_Object obarray)
3948 register Lisp_Object tem, string;
3950 if (NILP (obarray)) obarray = Vobarray;
3951 obarray = check_obarray (obarray);
3953 if (!SYMBOLP (name))
3955 CHECK_STRING (name);
3956 string = name;
3958 else
3959 string = SYMBOL_NAME (name);
3961 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3962 if (INTEGERP (tem) || (SYMBOLP (name) && !EQ (name, tem)))
3963 return Qnil;
3964 else
3965 return tem;
3968 DEFUN ("unintern", Funintern, Sunintern, 1, 2, 0,
3969 doc: /* Delete the symbol named NAME, if any, from OBARRAY.
3970 The value is t if a symbol was found and deleted, nil otherwise.
3971 NAME may be a string or a symbol. If it is a symbol, that symbol
3972 is deleted, if it belongs to OBARRAY--no other symbol is deleted.
3973 OBARRAY, if nil, defaults to the value of the variable `obarray'.
3974 usage: (unintern NAME OBARRAY) */)
3975 (Lisp_Object name, Lisp_Object obarray)
3977 register Lisp_Object string, tem;
3978 size_t hash;
3980 if (NILP (obarray)) obarray = Vobarray;
3981 obarray = check_obarray (obarray);
3983 if (SYMBOLP (name))
3984 string = SYMBOL_NAME (name);
3985 else
3987 CHECK_STRING (name);
3988 string = name;
3991 tem = oblookup (obarray, SSDATA (string),
3992 SCHARS (string),
3993 SBYTES (string));
3994 if (INTEGERP (tem))
3995 return Qnil;
3996 /* If arg was a symbol, don't delete anything but that symbol itself. */
3997 if (SYMBOLP (name) && !EQ (name, tem))
3998 return Qnil;
4000 /* There are plenty of other symbols which will screw up the Emacs
4001 session if we unintern them, as well as even more ways to use
4002 `setq' or `fset' or whatnot to make the Emacs session
4003 unusable. Let's not go down this silly road. --Stef */
4004 /* if (EQ (tem, Qnil) || EQ (tem, Qt))
4005 error ("Attempt to unintern t or nil"); */
4007 XSYMBOL (tem)->interned = SYMBOL_UNINTERNED;
4009 hash = oblookup_last_bucket_number;
4011 if (EQ (AREF (obarray, hash), tem))
4013 if (XSYMBOL (tem)->next)
4015 Lisp_Object sym;
4016 XSETSYMBOL (sym, XSYMBOL (tem)->next);
4017 ASET (obarray, hash, sym);
4019 else
4020 ASET (obarray, hash, make_number (0));
4022 else
4024 Lisp_Object tail, following;
4026 for (tail = AREF (obarray, hash);
4027 XSYMBOL (tail)->next;
4028 tail = following)
4030 XSETSYMBOL (following, XSYMBOL (tail)->next);
4031 if (EQ (following, tem))
4033 set_symbol_next (tail, XSYMBOL (following)->next);
4034 break;
4039 return Qt;
4042 /* Return the symbol in OBARRAY whose names matches the string
4043 of SIZE characters (SIZE_BYTE bytes) at PTR.
4044 If there is no such symbol, return the integer bucket number of
4045 where the symbol would be if it were present.
4047 Also store the bucket number in oblookup_last_bucket_number. */
4049 Lisp_Object
4050 oblookup (Lisp_Object obarray, register const char *ptr, ptrdiff_t size, ptrdiff_t size_byte)
4052 size_t hash;
4053 size_t obsize;
4054 register Lisp_Object tail;
4055 Lisp_Object bucket, tem;
4057 obarray = check_obarray (obarray);
4058 /* This is sometimes needed in the middle of GC. */
4059 obsize = gc_asize (obarray);
4060 hash = hash_string (ptr, size_byte) % obsize;
4061 bucket = AREF (obarray, hash);
4062 oblookup_last_bucket_number = hash;
4063 if (EQ (bucket, make_number (0)))
4065 else if (!SYMBOLP (bucket))
4066 error ("Bad data in guts of obarray"); /* Like CADR error message. */
4067 else
4068 for (tail = bucket; ; XSETSYMBOL (tail, XSYMBOL (tail)->next))
4070 if (SBYTES (SYMBOL_NAME (tail)) == size_byte
4071 && SCHARS (SYMBOL_NAME (tail)) == size
4072 && !memcmp (SDATA (SYMBOL_NAME (tail)), ptr, size_byte))
4073 return tail;
4074 else if (XSYMBOL (tail)->next == 0)
4075 break;
4077 XSETINT (tem, hash);
4078 return tem;
4081 void
4082 map_obarray (Lisp_Object obarray, void (*fn) (Lisp_Object, Lisp_Object), Lisp_Object arg)
4084 ptrdiff_t i;
4085 register Lisp_Object tail;
4086 CHECK_VECTOR (obarray);
4087 for (i = ASIZE (obarray) - 1; i >= 0; i--)
4089 tail = AREF (obarray, i);
4090 if (SYMBOLP (tail))
4091 while (1)
4093 (*fn) (tail, arg);
4094 if (XSYMBOL (tail)->next == 0)
4095 break;
4096 XSETSYMBOL (tail, XSYMBOL (tail)->next);
4101 static void
4102 mapatoms_1 (Lisp_Object sym, Lisp_Object function)
4104 call1 (function, sym);
4107 DEFUN ("mapatoms", Fmapatoms, Smapatoms, 1, 2, 0,
4108 doc: /* Call FUNCTION on every symbol in OBARRAY.
4109 OBARRAY defaults to the value of `obarray'. */)
4110 (Lisp_Object function, Lisp_Object obarray)
4112 if (NILP (obarray)) obarray = Vobarray;
4113 obarray = check_obarray (obarray);
4115 map_obarray (obarray, mapatoms_1, function);
4116 return Qnil;
4119 #define OBARRAY_SIZE 15121
4121 void
4122 init_obarray (void)
4124 Vobarray = Fmake_vector (make_number (OBARRAY_SIZE), make_number (0));
4125 initial_obarray = Vobarray;
4126 staticpro (&initial_obarray);
4128 for (int i = 0; i < ARRAYELTS (lispsym); i++)
4129 define_symbol (builtin_lisp_symbol (i), defsym_name[i]);
4131 DEFSYM (Qunbound, "unbound");
4133 DEFSYM (Qnil, "nil");
4134 SET_SYMBOL_VAL (XSYMBOL (Qnil), Qnil);
4135 make_symbol_constant (Qnil);
4136 XSYMBOL (Qnil)->declared_special = true;
4138 DEFSYM (Qt, "t");
4139 SET_SYMBOL_VAL (XSYMBOL (Qt), Qt);
4140 make_symbol_constant (Qt);
4141 XSYMBOL (Qt)->declared_special = true;
4143 /* Qt is correct even if CANNOT_DUMP. loadup.el will set to nil at end. */
4144 Vpurify_flag = Qt;
4146 DEFSYM (Qvariable_documentation, "variable-documentation");
4149 void
4150 defsubr (struct Lisp_Subr *sname)
4152 Lisp_Object sym, tem;
4153 sym = intern_c_string (sname->symbol_name);
4154 XSETPVECTYPE (sname, PVEC_SUBR);
4155 XSETSUBR (tem, sname);
4156 set_symbol_function (sym, tem);
4159 #ifdef NOTDEF /* Use fset in subr.el now! */
4160 void
4161 defalias (struct Lisp_Subr *sname, char *string)
4163 Lisp_Object sym;
4164 sym = intern (string);
4165 XSETSUBR (XSYMBOL (sym)->function, sname);
4167 #endif /* NOTDEF */
4169 /* Define an "integer variable"; a symbol whose value is forwarded to a
4170 C variable of type EMACS_INT. Sample call (with "xx" to fool make-docfile):
4171 DEFxxVAR_INT ("emacs-priority", &emacs_priority, "Documentation"); */
4172 void
4173 defvar_int (struct Lisp_Intfwd *i_fwd,
4174 const char *namestring, EMACS_INT *address)
4176 Lisp_Object sym;
4177 sym = intern_c_string (namestring);
4178 i_fwd->type = Lisp_Fwd_Int;
4179 i_fwd->intvar = address;
4180 XSYMBOL (sym)->declared_special = 1;
4181 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4182 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)i_fwd);
4185 /* Similar but define a variable whose value is t if address contains 1,
4186 nil if address contains 0. */
4187 void
4188 defvar_bool (struct Lisp_Boolfwd *b_fwd,
4189 const char *namestring, bool *address)
4191 Lisp_Object sym;
4192 sym = intern_c_string (namestring);
4193 b_fwd->type = Lisp_Fwd_Bool;
4194 b_fwd->boolvar = address;
4195 XSYMBOL (sym)->declared_special = 1;
4196 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4197 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)b_fwd);
4198 Vbyte_boolean_vars = Fcons (sym, Vbyte_boolean_vars);
4201 /* Similar but define a variable whose value is the Lisp Object stored
4202 at address. Two versions: with and without gc-marking of the C
4203 variable. The nopro version is used when that variable will be
4204 gc-marked for some other reason, since marking the same slot twice
4205 can cause trouble with strings. */
4206 void
4207 defvar_lisp_nopro (struct Lisp_Objfwd *o_fwd,
4208 const char *namestring, Lisp_Object *address)
4210 Lisp_Object sym;
4211 sym = intern_c_string (namestring);
4212 o_fwd->type = Lisp_Fwd_Obj;
4213 o_fwd->objvar = address;
4214 XSYMBOL (sym)->declared_special = 1;
4215 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4216 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)o_fwd);
4219 void
4220 defvar_lisp (struct Lisp_Objfwd *o_fwd,
4221 const char *namestring, Lisp_Object *address)
4223 defvar_lisp_nopro (o_fwd, namestring, address);
4224 staticpro (address);
4227 /* Similar but define a variable whose value is the Lisp Object stored
4228 at a particular offset in the current kboard object. */
4230 void
4231 defvar_kboard (struct Lisp_Kboard_Objfwd *ko_fwd,
4232 const char *namestring, int offset)
4234 Lisp_Object sym;
4235 sym = intern_c_string (namestring);
4236 ko_fwd->type = Lisp_Fwd_Kboard_Obj;
4237 ko_fwd->offset = offset;
4238 XSYMBOL (sym)->declared_special = 1;
4239 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4240 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)ko_fwd);
4243 /* Check that the elements of lpath exist. */
4245 static void
4246 load_path_check (Lisp_Object lpath)
4248 Lisp_Object path_tail;
4250 /* The only elements that might not exist are those from
4251 PATH_LOADSEARCH, EMACSLOADPATH. Anything else is only added if
4252 it exists. */
4253 for (path_tail = lpath; !NILP (path_tail); path_tail = XCDR (path_tail))
4255 Lisp_Object dirfile;
4256 dirfile = Fcar (path_tail);
4257 if (STRINGP (dirfile))
4259 dirfile = Fdirectory_file_name (dirfile);
4260 if (! file_accessible_directory_p (dirfile))
4261 dir_warning ("Lisp directory", XCAR (path_tail));
4266 /* Return the default load-path, to be used if EMACSLOADPATH is unset.
4267 This does not include the standard site-lisp directories
4268 under the installation prefix (i.e., PATH_SITELOADSEARCH),
4269 but it does (unless no_site_lisp is set) include site-lisp
4270 directories in the source/build directories if those exist and we
4271 are running uninstalled.
4273 Uses the following logic:
4274 If CANNOT_DUMP:
4275 If Vinstallation_directory is not nil (ie, running uninstalled),
4276 use PATH_DUMPLOADSEARCH (ie, build path). Else use PATH_LOADSEARCH.
4277 The remainder is what happens when dumping works:
4278 If purify-flag (ie dumping) just use PATH_DUMPLOADSEARCH.
4279 Otherwise use PATH_LOADSEARCH.
4281 If !initialized, then just return PATH_DUMPLOADSEARCH.
4282 If initialized:
4283 If Vinstallation_directory is not nil (ie, running uninstalled):
4284 If installation-dir/lisp exists and not already a member,
4285 we must be running uninstalled. Reset the load-path
4286 to just installation-dir/lisp. (The default PATH_LOADSEARCH
4287 refers to the eventual installation directories. Since we
4288 are not yet installed, we should not use them, even if they exist.)
4289 If installation-dir/lisp does not exist, just add
4290 PATH_DUMPLOADSEARCH at the end instead.
4291 Add installation-dir/site-lisp (if !no_site_lisp, and exists
4292 and not already a member) at the front.
4293 If installation-dir != source-dir (ie running an uninstalled,
4294 out-of-tree build) AND install-dir/src/Makefile exists BUT
4295 install-dir/src/Makefile.in does NOT exist (this is a sanity
4296 check), then repeat the above steps for source-dir/lisp, site-lisp. */
4298 static Lisp_Object
4299 load_path_default (void)
4301 Lisp_Object lpath = Qnil;
4302 const char *normal;
4304 #ifdef CANNOT_DUMP
4305 #ifdef HAVE_NS
4306 const char *loadpath = ns_load_path ();
4307 #endif
4309 normal = PATH_LOADSEARCH;
4310 if (!NILP (Vinstallation_directory)) normal = PATH_DUMPLOADSEARCH;
4312 #ifdef HAVE_NS
4313 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4314 #else
4315 lpath = decode_env_path (0, normal, 0);
4316 #endif
4318 #else /* !CANNOT_DUMP */
4320 normal = NILP (Vpurify_flag) ? PATH_LOADSEARCH : PATH_DUMPLOADSEARCH;
4322 if (initialized)
4324 #ifdef HAVE_NS
4325 const char *loadpath = ns_load_path ();
4326 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4327 #else
4328 lpath = decode_env_path (0, normal, 0);
4329 #endif
4330 if (!NILP (Vinstallation_directory))
4332 Lisp_Object tem, tem1;
4334 /* Add to the path the lisp subdir of the installation
4335 dir, if it is accessible. Note: in out-of-tree builds,
4336 this directory is empty save for Makefile. */
4337 tem = Fexpand_file_name (build_string ("lisp"),
4338 Vinstallation_directory);
4339 tem1 = Ffile_accessible_directory_p (tem);
4340 if (!NILP (tem1))
4342 if (NILP (Fmember (tem, lpath)))
4344 /* We are running uninstalled. The default load-path
4345 points to the eventual installed lisp directories.
4346 We should not use those now, even if they exist,
4347 so start over from a clean slate. */
4348 lpath = list1 (tem);
4351 else
4352 /* That dir doesn't exist, so add the build-time
4353 Lisp dirs instead. */
4355 Lisp_Object dump_path =
4356 decode_env_path (0, PATH_DUMPLOADSEARCH, 0);
4357 lpath = nconc2 (lpath, dump_path);
4360 /* Add site-lisp under the installation dir, if it exists. */
4361 if (!no_site_lisp)
4363 tem = Fexpand_file_name (build_string ("site-lisp"),
4364 Vinstallation_directory);
4365 tem1 = Ffile_accessible_directory_p (tem);
4366 if (!NILP (tem1))
4368 if (NILP (Fmember (tem, lpath)))
4369 lpath = Fcons (tem, lpath);
4373 /* If Emacs was not built in the source directory,
4374 and it is run from where it was built, add to load-path
4375 the lisp and site-lisp dirs under that directory. */
4377 if (NILP (Fequal (Vinstallation_directory, Vsource_directory)))
4379 Lisp_Object tem2;
4381 tem = Fexpand_file_name (build_string ("src/Makefile"),
4382 Vinstallation_directory);
4383 tem1 = Ffile_exists_p (tem);
4385 /* Don't be fooled if they moved the entire source tree
4386 AFTER dumping Emacs. If the build directory is indeed
4387 different from the source dir, src/Makefile.in and
4388 src/Makefile will not be found together. */
4389 tem = Fexpand_file_name (build_string ("src/Makefile.in"),
4390 Vinstallation_directory);
4391 tem2 = Ffile_exists_p (tem);
4392 if (!NILP (tem1) && NILP (tem2))
4394 tem = Fexpand_file_name (build_string ("lisp"),
4395 Vsource_directory);
4397 if (NILP (Fmember (tem, lpath)))
4398 lpath = Fcons (tem, lpath);
4400 if (!no_site_lisp)
4402 tem = Fexpand_file_name (build_string ("site-lisp"),
4403 Vsource_directory);
4404 tem1 = Ffile_accessible_directory_p (tem);
4405 if (!NILP (tem1))
4407 if (NILP (Fmember (tem, lpath)))
4408 lpath = Fcons (tem, lpath);
4412 } /* Vinstallation_directory != Vsource_directory */
4414 } /* if Vinstallation_directory */
4416 else /* !initialized */
4418 /* NORMAL refers to PATH_DUMPLOADSEARCH, ie the lisp dir in the
4419 source directory. We used to add ../lisp (ie the lisp dir in
4420 the build directory) at the front here, but that should not
4421 be necessary, since in out of tree builds lisp/ is empty, save
4422 for Makefile. */
4423 lpath = decode_env_path (0, normal, 0);
4425 #endif /* !CANNOT_DUMP */
4427 return lpath;
4430 void
4431 init_lread (void)
4433 /* First, set Vload_path. */
4435 /* Ignore EMACSLOADPATH when dumping. */
4436 #ifdef CANNOT_DUMP
4437 bool use_loadpath = true;
4438 #else
4439 bool use_loadpath = NILP (Vpurify_flag);
4440 #endif
4442 if (use_loadpath && egetenv ("EMACSLOADPATH"))
4444 Vload_path = decode_env_path ("EMACSLOADPATH", 0, 1);
4446 /* Check (non-nil) user-supplied elements. */
4447 load_path_check (Vload_path);
4449 /* If no nils in the environment variable, use as-is.
4450 Otherwise, replace any nils with the default. */
4451 if (! NILP (Fmemq (Qnil, Vload_path)))
4453 Lisp_Object elem, elpath = Vload_path;
4454 Lisp_Object default_lpath = load_path_default ();
4456 /* Check defaults, before adding site-lisp. */
4457 load_path_check (default_lpath);
4459 /* Add the site-lisp directories to the front of the default. */
4460 if (!no_site_lisp && PATH_SITELOADSEARCH[0] != '\0')
4462 Lisp_Object sitelisp;
4463 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4464 if (! NILP (sitelisp))
4465 default_lpath = nconc2 (sitelisp, default_lpath);
4468 Vload_path = Qnil;
4470 /* Replace nils from EMACSLOADPATH by default. */
4471 while (CONSP (elpath))
4473 elem = XCAR (elpath);
4474 elpath = XCDR (elpath);
4475 Vload_path = CALLN (Fappend, Vload_path,
4476 NILP (elem) ? default_lpath : list1 (elem));
4478 } /* Fmemq (Qnil, Vload_path) */
4480 else
4482 Vload_path = load_path_default ();
4484 /* Check before adding site-lisp directories.
4485 The install should have created them, but they are not
4486 required, so no need to warn if they are absent.
4487 Or we might be running before installation. */
4488 load_path_check (Vload_path);
4490 /* Add the site-lisp directories at the front. */
4491 if (initialized && !no_site_lisp && PATH_SITELOADSEARCH[0] != '\0')
4493 Lisp_Object sitelisp;
4494 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4495 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4499 Vvalues = Qnil;
4501 load_in_progress = 0;
4502 Vload_file_name = Qnil;
4503 Vstandard_input = Qt;
4504 Vloads_in_progress = Qnil;
4507 /* Print a warning that directory intended for use USE and with name
4508 DIRNAME cannot be accessed. On entry, errno should correspond to
4509 the access failure. Print the warning on stderr and put it in
4510 *Messages*. */
4512 void
4513 dir_warning (char const *use, Lisp_Object dirname)
4515 static char const format[] = "Warning: %s '%s': %s\n";
4516 char *diagnostic = emacs_strerror (errno);
4517 fprintf (stderr, format, use, SSDATA (ENCODE_SYSTEM (dirname)), diagnostic);
4519 /* Don't log the warning before we've initialized!! */
4520 if (initialized)
4522 ptrdiff_t diaglen = strlen (diagnostic);
4523 AUTO_STRING_WITH_LEN (diag, diagnostic, diaglen);
4524 if (! NILP (Vlocale_coding_system))
4526 Lisp_Object s
4527 = code_convert_string_norecord (diag, Vlocale_coding_system, false);
4528 diagnostic = SSDATA (s);
4529 diaglen = SBYTES (s);
4531 USE_SAFE_ALLOCA;
4532 char *buffer = SAFE_ALLOCA (sizeof format - 3 * (sizeof "%s" - 1)
4533 + strlen (use) + SBYTES (dirname) + diaglen);
4534 ptrdiff_t message_len = esprintf (buffer, format, use, SSDATA (dirname),
4535 diagnostic);
4536 message_dolog (buffer, message_len, 0, STRING_MULTIBYTE (dirname));
4537 SAFE_FREE ();
4541 void
4542 syms_of_lread (void)
4544 defsubr (&Sread);
4545 defsubr (&Sread_from_string);
4546 defsubr (&Sintern);
4547 defsubr (&Sintern_soft);
4548 defsubr (&Sunintern);
4549 defsubr (&Sget_load_suffixes);
4550 defsubr (&Sload);
4551 defsubr (&Seval_buffer);
4552 defsubr (&Seval_region);
4553 defsubr (&Sread_char);
4554 defsubr (&Sread_char_exclusive);
4555 defsubr (&Sread_event);
4556 defsubr (&Sget_file_char);
4557 defsubr (&Smapatoms);
4558 defsubr (&Slocate_file_internal);
4560 DEFVAR_LISP ("obarray", Vobarray,
4561 doc: /* Symbol table for use by `intern' and `read'.
4562 It is a vector whose length ought to be prime for best results.
4563 The vector's contents don't make sense if examined from Lisp programs;
4564 to find all the symbols in an obarray, use `mapatoms'. */);
4566 DEFVAR_LISP ("values", Vvalues,
4567 doc: /* List of values of all expressions which were read, evaluated and printed.
4568 Order is reverse chronological. */);
4569 XSYMBOL (intern ("values"))->declared_special = 0;
4571 DEFVAR_LISP ("standard-input", Vstandard_input,
4572 doc: /* Stream for read to get input from.
4573 See documentation of `read' for possible values. */);
4574 Vstandard_input = Qt;
4576 DEFVAR_LISP ("read-with-symbol-positions", Vread_with_symbol_positions,
4577 doc: /* If non-nil, add position of read symbols to `read-symbol-positions-list'.
4579 If this variable is a buffer, then only forms read from that buffer
4580 will be added to `read-symbol-positions-list'.
4581 If this variable is t, then all read forms will be added.
4582 The effect of all other values other than nil are not currently
4583 defined, although they may be in the future.
4585 The positions are relative to the last call to `read' or
4586 `read-from-string'. It is probably a bad idea to set this variable at
4587 the toplevel; bind it instead. */);
4588 Vread_with_symbol_positions = Qnil;
4590 DEFVAR_LISP ("read-symbol-positions-list", Vread_symbol_positions_list,
4591 doc: /* A list mapping read symbols to their positions.
4592 This variable is modified during calls to `read' or
4593 `read-from-string', but only when `read-with-symbol-positions' is
4594 non-nil.
4596 Each element of the list looks like (SYMBOL . CHAR-POSITION), where
4597 CHAR-POSITION is an integer giving the offset of that occurrence of the
4598 symbol from the position where `read' or `read-from-string' started.
4600 Note that a symbol will appear multiple times in this list, if it was
4601 read multiple times. The list is in the same order as the symbols
4602 were read in. */);
4603 Vread_symbol_positions_list = Qnil;
4605 DEFVAR_LISP ("read-circle", Vread_circle,
4606 doc: /* Non-nil means read recursive structures using #N= and #N# syntax. */);
4607 Vread_circle = Qt;
4609 DEFVAR_LISP ("load-path", Vload_path,
4610 doc: /* List of directories to search for files to load.
4611 Each element is a string (directory file name) or nil (meaning
4612 `default-directory').
4613 This list is consulted by the `require' function.
4614 Initialized during startup as described in Info node `(elisp)Library Search'.
4615 Use `directory-file-name' when adding items to this path. However, Lisp
4616 programs that process this list should tolerate directories both with
4617 and without trailing slashes. */);
4619 DEFVAR_LISP ("load-suffixes", Vload_suffixes,
4620 doc: /* List of suffixes for Emacs Lisp files and dynamic modules.
4621 This list includes suffixes for both compiled and source Emacs Lisp files.
4622 This list should not include the empty string.
4623 `load' and related functions try to append these suffixes, in order,
4624 to the specified file name if a suffix is allowed or required. */);
4625 #ifdef HAVE_MODULES
4626 Vload_suffixes = list3 (build_pure_c_string (".elc"),
4627 build_pure_c_string (".el"),
4628 build_pure_c_string (MODULES_SUFFIX));
4629 #else
4630 Vload_suffixes = list2 (build_pure_c_string (".elc"),
4631 build_pure_c_string (".el"));
4632 #endif
4633 DEFVAR_LISP ("module-file-suffix", Vmodule_file_suffix,
4634 doc: /* Suffix of loadable module file, or nil of modules are not supported. */);
4635 #ifdef HAVE_MODULES
4636 Vmodule_file_suffix = build_pure_c_string (MODULES_SUFFIX);
4637 #else
4638 Vmodule_file_suffix = Qnil;
4639 #endif
4640 DEFVAR_LISP ("load-file-rep-suffixes", Vload_file_rep_suffixes,
4641 doc: /* List of suffixes that indicate representations of \
4642 the same file.
4643 This list should normally start with the empty string.
4645 Enabling Auto Compression mode appends the suffixes in
4646 `jka-compr-load-suffixes' to this list and disabling Auto Compression
4647 mode removes them again. `load' and related functions use this list to
4648 determine whether they should look for compressed versions of a file
4649 and, if so, which suffixes they should try to append to the file name
4650 in order to do so. However, if you want to customize which suffixes
4651 the loading functions recognize as compression suffixes, you should
4652 customize `jka-compr-load-suffixes' rather than the present variable. */);
4653 Vload_file_rep_suffixes = list1 (empty_unibyte_string);
4655 DEFVAR_BOOL ("load-in-progress", load_in_progress,
4656 doc: /* Non-nil if inside of `load'. */);
4657 DEFSYM (Qload_in_progress, "load-in-progress");
4659 DEFVAR_LISP ("after-load-alist", Vafter_load_alist,
4660 doc: /* An alist of functions to be evalled when particular files are loaded.
4661 Each element looks like (REGEXP-OR-FEATURE FUNCS...).
4663 REGEXP-OR-FEATURE is either a regular expression to match file names, or
4664 a symbol (a feature name).
4666 When `load' is run and the file-name argument matches an element's
4667 REGEXP-OR-FEATURE, or when `provide' is run and provides the symbol
4668 REGEXP-OR-FEATURE, the FUNCS in the element are called.
4670 An error in FORMS does not undo the load, but does prevent execution of
4671 the rest of the FORMS. */);
4672 Vafter_load_alist = Qnil;
4674 DEFVAR_LISP ("load-history", Vload_history,
4675 doc: /* Alist mapping loaded file names to symbols and features.
4676 Each alist element should be a list (FILE-NAME ENTRIES...), where
4677 FILE-NAME is the name of a file that has been loaded into Emacs.
4678 The file name is absolute and true (i.e. it doesn't contain symlinks).
4679 As an exception, one of the alist elements may have FILE-NAME nil,
4680 for symbols and features not associated with any file.
4682 The remaining ENTRIES in the alist element describe the functions and
4683 variables defined in that file, the features provided, and the
4684 features required. Each entry has the form `(provide . FEATURE)',
4685 `(require . FEATURE)', `(defun . FUNCTION)', `(autoload . SYMBOL)',
4686 `(defface . SYMBOL)', or `(t . SYMBOL)'. Entries like `(t . SYMBOL)'
4687 may precede a `(defun . FUNCTION)' entry, and means that SYMBOL was an
4688 autoload before this file redefined it as a function. In addition,
4689 entries may also be single symbols, which means that SYMBOL was
4690 defined by `defvar' or `defconst'.
4692 During preloading, the file name recorded is relative to the main Lisp
4693 directory. These file names are converted to absolute at startup. */);
4694 Vload_history = Qnil;
4696 DEFVAR_LISP ("load-file-name", Vload_file_name,
4697 doc: /* Full name of file being loaded by `load'. */);
4698 Vload_file_name = Qnil;
4700 DEFVAR_LISP ("user-init-file", Vuser_init_file,
4701 doc: /* File name, including directory, of user's initialization file.
4702 If the file loaded had extension `.elc', and the corresponding source file
4703 exists, this variable contains the name of source file, suitable for use
4704 by functions like `custom-save-all' which edit the init file.
4705 While Emacs loads and evaluates the init file, value is the real name
4706 of the file, regardless of whether or not it has the `.elc' extension. */);
4707 Vuser_init_file = Qnil;
4709 DEFVAR_LISP ("current-load-list", Vcurrent_load_list,
4710 doc: /* Used for internal purposes by `load'. */);
4711 Vcurrent_load_list = Qnil;
4713 DEFVAR_LISP ("load-read-function", Vload_read_function,
4714 doc: /* Function used by `load' and `eval-region' for reading expressions.
4715 Called with a single argument (the stream from which to read).
4716 The default is to use the function `read'. */);
4717 DEFSYM (Qread, "read");
4718 Vload_read_function = Qread;
4720 DEFVAR_LISP ("load-source-file-function", Vload_source_file_function,
4721 doc: /* Function called in `load' to load an Emacs Lisp source file.
4722 The value should be a function for doing code conversion before
4723 reading a source file. It can also be nil, in which case loading is
4724 done without any code conversion.
4726 If the value is a function, it is called with four arguments,
4727 FULLNAME, FILE, NOERROR, NOMESSAGE. FULLNAME is the absolute name of
4728 the file to load, FILE is the non-absolute name (for messages etc.),
4729 and NOERROR and NOMESSAGE are the corresponding arguments passed to
4730 `load'. The function should return t if the file was loaded. */);
4731 Vload_source_file_function = Qnil;
4733 DEFVAR_BOOL ("load-force-doc-strings", load_force_doc_strings,
4734 doc: /* Non-nil means `load' should force-load all dynamic doc strings.
4735 This is useful when the file being loaded is a temporary copy. */);
4736 load_force_doc_strings = 0;
4738 DEFVAR_BOOL ("load-convert-to-unibyte", load_convert_to_unibyte,
4739 doc: /* Non-nil means `read' converts strings to unibyte whenever possible.
4740 This is normally bound by `load' and `eval-buffer' to control `read',
4741 and is not meant for users to change. */);
4742 load_convert_to_unibyte = 0;
4744 DEFVAR_LISP ("source-directory", Vsource_directory,
4745 doc: /* Directory in which Emacs sources were found when Emacs was built.
4746 You cannot count on them to still be there! */);
4747 Vsource_directory
4748 = Fexpand_file_name (build_string ("../"),
4749 Fcar (decode_env_path (0, PATH_DUMPLOADSEARCH, 0)));
4751 DEFVAR_LISP ("preloaded-file-list", Vpreloaded_file_list,
4752 doc: /* List of files that were preloaded (when dumping Emacs). */);
4753 Vpreloaded_file_list = Qnil;
4755 DEFVAR_LISP ("byte-boolean-vars", Vbyte_boolean_vars,
4756 doc: /* List of all DEFVAR_BOOL variables, used by the byte code optimizer. */);
4757 Vbyte_boolean_vars = Qnil;
4759 DEFVAR_BOOL ("load-dangerous-libraries", load_dangerous_libraries,
4760 doc: /* Non-nil means load dangerous compiled Lisp files.
4761 Some versions of XEmacs use different byte codes than Emacs. These
4762 incompatible byte codes can make Emacs crash when it tries to execute
4763 them. */);
4764 load_dangerous_libraries = 0;
4766 DEFVAR_BOOL ("force-load-messages", force_load_messages,
4767 doc: /* Non-nil means force printing messages when loading Lisp files.
4768 This overrides the value of the NOMESSAGE argument to `load'. */);
4769 force_load_messages = 0;
4771 DEFVAR_LISP ("bytecomp-version-regexp", Vbytecomp_version_regexp,
4772 doc: /* Regular expression matching safe to load compiled Lisp files.
4773 When Emacs loads a compiled Lisp file, it reads the first 512 bytes
4774 from the file, and matches them against this regular expression.
4775 When the regular expression matches, the file is considered to be safe
4776 to load. See also `load-dangerous-libraries'. */);
4777 Vbytecomp_version_regexp
4778 = build_pure_c_string ("^;;;.\\(in Emacs version\\|bytecomp version FSF\\)");
4780 DEFSYM (Qlexical_binding, "lexical-binding");
4781 DEFVAR_LISP ("lexical-binding", Vlexical_binding,
4782 doc: /* Whether to use lexical binding when evaluating code.
4783 Non-nil means that the code in the current buffer should be evaluated
4784 with lexical binding.
4785 This variable is automatically set from the file variables of an
4786 interpreted Lisp file read using `load'. Unlike other file local
4787 variables, this must be set in the first line of a file. */);
4788 Vlexical_binding = Qnil;
4789 Fmake_variable_buffer_local (Qlexical_binding);
4791 DEFVAR_LISP ("eval-buffer-list", Veval_buffer_list,
4792 doc: /* List of buffers being read from by calls to `eval-buffer' and `eval-region'. */);
4793 Veval_buffer_list = Qnil;
4795 DEFVAR_LISP ("old-style-backquotes", Vold_style_backquotes,
4796 doc: /* Set to non-nil when `read' encounters an old-style backquote. */);
4797 Vold_style_backquotes = Qnil;
4798 DEFSYM (Qold_style_backquotes, "old-style-backquotes");
4800 DEFVAR_BOOL ("load-prefer-newer", load_prefer_newer,
4801 doc: /* Non-nil means `load' prefers the newest version of a file.
4802 This applies when a filename suffix is not explicitly specified and
4803 `load' is trying various possible suffixes (see `load-suffixes' and
4804 `load-file-rep-suffixes'). Normally, it stops at the first file
4805 that exists unless you explicitly specify one or the other. If this
4806 option is non-nil, it checks all suffixes and uses whichever file is
4807 newest.
4808 Note that if you customize this, obviously it will not affect files
4809 that are loaded before your customizations are read! */);
4810 load_prefer_newer = 0;
4812 /* Vsource_directory was initialized in init_lread. */
4814 DEFSYM (Qcurrent_load_list, "current-load-list");
4815 DEFSYM (Qstandard_input, "standard-input");
4816 DEFSYM (Qread_char, "read-char");
4817 DEFSYM (Qget_file_char, "get-file-char");
4819 /* Used instead of Qget_file_char while loading *.elc files compiled
4820 by Emacs 21 or older. */
4821 DEFSYM (Qget_emacs_mule_file_char, "get-emacs-mule-file-char");
4823 DEFSYM (Qload_force_doc_strings, "load-force-doc-strings");
4825 DEFSYM (Qbackquote, "`");
4826 DEFSYM (Qcomma, ",");
4827 DEFSYM (Qcomma_at, ",@");
4828 DEFSYM (Qcomma_dot, ",.");
4830 DEFSYM (Qinhibit_file_name_operation, "inhibit-file-name-operation");
4831 DEFSYM (Qascii_character, "ascii-character");
4832 DEFSYM (Qfunction, "function");
4833 DEFSYM (Qload, "load");
4834 DEFSYM (Qload_file_name, "load-file-name");
4835 DEFSYM (Qeval_buffer_list, "eval-buffer-list");
4836 DEFSYM (Qfile_truename, "file-truename");
4837 DEFSYM (Qdir_ok, "dir-ok");
4838 DEFSYM (Qdo_after_load_evaluation, "do-after-load-evaluation");
4840 staticpro (&read_objects);
4841 read_objects = Qnil;
4842 staticpro (&seen_list);
4843 seen_list = Qnil;
4845 Vloads_in_progress = Qnil;
4846 staticpro (&Vloads_in_progress);
4848 DEFSYM (Qhash_table, "hash-table");
4849 DEFSYM (Qdata, "data");
4850 DEFSYM (Qtest, "test");
4851 DEFSYM (Qsize, "size");
4852 DEFSYM (Qweakness, "weakness");
4853 DEFSYM (Qrehash_size, "rehash-size");
4854 DEFSYM (Qrehash_threshold, "rehash-threshold");
4856 DEFSYM (Qchar_from_name, "char-from-name");