Annotation of embedaddon/quagga/lib/regex.c, revision 1.1.1.1
1.1 misho 1: /* Extended regular expression matching and search library,
2: version 0.12.
3: (Implements POSIX draft P1003.2/D11.2, except for some of the
4: internationalization features.)
5: Copyright (C) 1993, 94, 95, 96, 97, 98, 99 Free Software Foundation, Inc.
6:
7: The GNU C Library is free software; you can redistribute it and/or
8: modify it under the terms of the GNU Library General Public License as
9: published by the Free Software Foundation; either version 2 of the
10: License, or (at your option) any later version.
11:
12: The GNU C Library is distributed in the hope that it will be useful,
13: but WITHOUT ANY WARRANTY; without even the implied warranty of
14: MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15: Library General Public License for more details.
16:
17: You should have received a copy of the GNU Library General Public
18: License along with the GNU C Library; see the file COPYING.LIB. If not,
19: write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20: Boston, MA 02111-1307, USA. */
21:
22: /* AIX requires this to be the first thing in the file. */
23: #if defined _AIX && !defined REGEX_MALLOC
24: #pragma alloca
25: #endif
26:
27: #undef _GNU_SOURCE
28: #define _GNU_SOURCE
29:
30: #ifdef HAVE_CONFIG_H
31: # include <config.h>
32: #endif
33: #ifdef _WIN32
34: /* Windows does not provide unistd.h, which is required for abort() */
35: #include <process.h>
36: #endif /* _WIN32 */
37:
38: #ifndef PARAMS
39: # if defined __GNUC__ || (defined __STDC__ && __STDC__)
40: # define PARAMS(args) args
41: # else
42: # define PARAMS(args) ()
43: # endif /* GCC. */
44: #endif /* Not PARAMS. */
45:
46: #if defined STDC_HEADERS && !defined emacs
47: # include <stddef.h>
48: #else
49: /* We need this for `regex.h', and perhaps for the Emacs include files. */
50: # include <sys/types.h>
51: #endif
52:
53: #define WIDE_CHAR_SUPPORT (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC)
54:
55: /* For platform which support the ISO C amendement 1 functionality we
56: support user defined character classes. */
57: #if defined _LIBC || WIDE_CHAR_SUPPORT
58: /* Solaris 2.5 has a bug: <wchar.h> must be included before <wctype.h>. */
59: # include <wchar.h>
60: # include <wctype.h>
61: #endif
62:
63: #ifdef _LIBC
64: /* We have to keep the namespace clean. */
65: # define regfree(preg) __regfree (preg)
66: # define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef)
67: # define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags)
68: # define regerror(errcode, preg, errbuf, errbuf_size) \
69: __regerror(errcode, preg, errbuf, errbuf_size)
70: # define re_set_registers(bu, re, nu, st, en) \
71: __re_set_registers (bu, re, nu, st, en)
72: # define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \
73: __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
74: # define re_match(bufp, string, size, pos, regs) \
75: __re_match (bufp, string, size, pos, regs)
76: # define re_search(bufp, string, size, startpos, range, regs) \
77: __re_search (bufp, string, size, startpos, range, regs)
78: # define re_compile_pattern(pattern, length, bufp) \
79: __re_compile_pattern (pattern, length, bufp)
80: # define re_set_syntax(syntax) __re_set_syntax (syntax)
81: # define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \
82: __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop)
83: # define re_compile_fastmap(bufp) __re_compile_fastmap (bufp)
84:
85: #define btowc __btowc
86: #endif
87:
88: /* This is for other GNU distributions with internationalized messages. */
89: #if HAVE_LIBINTL_H || defined _LIBC
90: # include <libintl.h>
91: #else
92: # define gettext(msgid) (msgid)
93: #endif
94:
95: #ifndef gettext_noop
96: /* This define is so xgettext can find the internationalizable
97: strings. */
98: # define gettext_noop(String) String
99: #endif
100:
101: /* The `emacs' switch turns on certain matching commands
102: that make sense only in Emacs. */
103: #ifdef emacs
104:
105: # include "lisp.h"
106: # include "buffer.h"
107: # include "syntax.h"
108:
109: #else /* not emacs */
110:
111: /* If we are not linking with Emacs proper,
112: we can't use the relocating allocator
113: even if config.h says that we can. */
114: # undef REL_ALLOC
115:
116: # if defined STDC_HEADERS || defined _LIBC
117: # include <stdlib.h>
118: # else
119: char *malloc ();
120: char *realloc ();
121: # endif
122:
123: /* When used in Emacs's lib-src, we need to get bzero and bcopy somehow.
124: If nothing else has been done, use the method below. */
125: # ifdef INHIBIT_STRING_HEADER
126: # if !(defined HAVE_BZERO && defined HAVE_BCOPY)
127: # if !defined bzero && !defined bcopy
128: # undef INHIBIT_STRING_HEADER
129: # endif
130: # endif
131: # endif
132:
133: /* This is the normal way of making sure we have a bcopy and a bzero.
134: This is used in most programs--a few other programs avoid this
135: by defining INHIBIT_STRING_HEADER. */
136: # ifndef INHIBIT_STRING_HEADER
137: # if defined HAVE_STRING_H || defined STDC_HEADERS || defined _LIBC
138: # include <string.h>
139: # ifndef bzero
140: # ifndef _LIBC
141: # define bzero(s, n) (memset (s, '\0', n), (s))
142: # else
143: # define bzero(s, n) __bzero (s, n)
144: # endif
145: # endif
146: # else
147: # include <strings.h>
148: # ifndef memcmp
149: # define memcmp(s1, s2, n) bcmp (s1, s2, n)
150: # endif
151: # ifndef memcpy
152: # define memcpy(d, s, n) (bcopy (s, d, n), (d))
153: # endif
154: # endif
155: # endif
156:
157: /* Define the syntax stuff for \<, \>, etc. */
158:
159: /* This must be nonzero for the wordchar and notwordchar pattern
160: commands in re_match_2. */
161: # ifndef Sword
162: # define Sword 1
163: # endif
164:
165: # ifdef SWITCH_ENUM_BUG
166: # define SWITCH_ENUM_CAST(x) ((int)(x))
167: # else
168: # define SWITCH_ENUM_CAST(x) (x)
169: # endif
170:
171: /* How many characters in the character set. */
172: # define CHAR_SET_SIZE 256
173:
174: # ifdef SYNTAX_TABLE
175:
176: extern char *re_syntax_table;
177:
178: # else /* not SYNTAX_TABLE */
179:
180: static char re_syntax_table[CHAR_SET_SIZE];
181:
182: static void
183: init_syntax_once ()
184: {
185: register int c;
186: static int done;
187:
188: if (done)
189: return;
190:
191: memset (re_syntax_table, 0, sizeof re_syntax_table);
192:
193: for (c = 'a'; c <= 'z'; c++)
194: re_syntax_table[c] = Sword;
195:
196: for (c = 'A'; c <= 'Z'; c++)
197: re_syntax_table[c] = Sword;
198:
199: for (c = '0'; c <= '9'; c++)
200: re_syntax_table[c] = Sword;
201:
202: re_syntax_table['_'] = Sword;
203:
204: done = 1;
205: }
206:
207: # endif /* not SYNTAX_TABLE */
208:
209: # define SYNTAX(c) re_syntax_table[c]
210:
211: #endif /* not emacs */
212:
213: /* Get the interface, including the syntax bits. */
214: #include <regex-gnu.h>
215:
216: /* isalpha etc. are used for the character classes. */
217: #include <ctype.h>
218:
219: /* Jim Meyering writes:
220:
221: "... Some ctype macros are valid only for character codes that
222: isascii says are ASCII (SGI's IRIX-4.0.5 is one such system --when
223: using /bin/cc or gcc but without giving an ansi option). So, all
224: ctype uses should be through macros like ISPRINT... If
225: STDC_HEADERS is defined, then autoconf has verified that the ctype
226: macros don't need to be guarded with references to isascii. ...
227: Defining isascii to 1 should let any compiler worth its salt
228: eliminate the && through constant folding."
229: Solaris defines some of these symbols so we must undefine them first. */
230:
231: #undef ISASCII
232: #if defined STDC_HEADERS || (!defined isascii && !defined HAVE_ISASCII)
233: # define ISASCII(c) 1
234: #else
235: # define ISASCII(c) isascii(c)
236: #endif
237:
238: #ifdef isblank
239: # define ISBLANK(c) (ISASCII (c) && isblank (c))
240: #else
241: # define ISBLANK(c) ((c) == ' ' || (c) == '\t')
242: #endif
243: #ifdef isgraph
244: # define ISGRAPH(c) (ISASCII (c) && isgraph (c))
245: #else
246: # define ISGRAPH(c) (ISASCII (c) && isprint (c) && !isspace (c))
247: #endif
248:
249: #undef ISPRINT
250: #define ISPRINT(c) (ISASCII (c) && isprint (c))
251: #define ISDIGIT(c) (ISASCII (c) && isdigit (c))
252: #define ISALNUM(c) (ISASCII (c) && isalnum (c))
253: #define ISALPHA(c) (ISASCII (c) && isalpha (c))
254: #define ISCNTRL(c) (ISASCII (c) && iscntrl (c))
255: #define ISLOWER(c) (ISASCII (c) && islower (c))
256: #define ISPUNCT(c) (ISASCII (c) && ispunct (c))
257: #define ISSPACE(c) (ISASCII (c) && isspace (c))
258: #define ISUPPER(c) (ISASCII (c) && isupper (c))
259: #define ISXDIGIT(c) (ISASCII (c) && isxdigit (c))
260:
261: #ifdef _tolower
262: # define TOLOWER(c) _tolower(c)
263: #else
264: # define TOLOWER(c) tolower(c)
265: #endif
266:
267: #ifndef NULL
268: # define NULL (void *)0
269: #endif
270:
271: /* We remove any previous definition of `SIGN_EXTEND_CHAR',
272: since ours (we hope) works properly with all combinations of
273: machines, compilers, `char' and `unsigned char' argument types.
274: (Per Bothner suggested the basic approach.) */
275: #undef SIGN_EXTEND_CHAR
276: #if __STDC__
277: # define SIGN_EXTEND_CHAR(c) ((signed char) (c))
278: #else /* not __STDC__ */
279: /* As in Harbison and Steele. */
280: # define SIGN_EXTEND_CHAR(c) ((((unsigned char) (c)) ^ 128) - 128)
281: #endif
282:
283: /* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we
284: use `alloca' instead of `malloc'. This is because using malloc in
285: re_search* or re_match* could cause memory leaks when C-g is used in
286: Emacs; also, malloc is slower and causes storage fragmentation. On
287: the other hand, malloc is more portable, and easier to debug.
288:
289: Because we sometimes use alloca, some routines have to be macros,
290: not functions -- `alloca'-allocated space disappears at the end of the
291: function it is called in. */
292:
293: #ifdef REGEX_MALLOC
294:
295: # define REGEX_ALLOCATE malloc
296: # define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize)
297: # define REGEX_FREE free
298:
299: #else /* not REGEX_MALLOC */
300:
301: /* Emacs already defines alloca, sometimes. */
302: # ifndef alloca
303:
304: /* Make alloca work the best possible way. */
305: # ifdef __GNUC__
306: # define alloca __builtin_alloca
307: # else /* not __GNUC__ */
308: # if HAVE_ALLOCA_H
309: # include <alloca.h>
310: # endif /* HAVE_ALLOCA_H */
311: # endif /* not __GNUC__ */
312:
313: # endif /* not alloca */
314:
315: # define REGEX_ALLOCATE alloca
316:
317: /* Assumes a `char *destination' variable. */
318: # define REGEX_REALLOCATE(source, osize, nsize) \
319: (destination = (char *) alloca (nsize), \
320: memcpy (destination, source, osize))
321:
322: /* No need to do anything to free, after alloca. */
323: # define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */
324:
325: #endif /* not REGEX_MALLOC */
326:
327: /* Define how to allocate the failure stack. */
328:
329: #if defined REL_ALLOC && defined REGEX_MALLOC
330:
331: # define REGEX_ALLOCATE_STACK(size) \
332: r_alloc (&failure_stack_ptr, (size))
333: # define REGEX_REALLOCATE_STACK(source, osize, nsize) \
334: r_re_alloc (&failure_stack_ptr, (nsize))
335: # define REGEX_FREE_STACK(ptr) \
336: r_alloc_free (&failure_stack_ptr)
337:
338: #else /* not using relocating allocator */
339:
340: # ifdef REGEX_MALLOC
341:
342: # define REGEX_ALLOCATE_STACK malloc
343: # define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize)
344: # define REGEX_FREE_STACK free
345:
346: # else /* not REGEX_MALLOC */
347:
348: # define REGEX_ALLOCATE_STACK alloca
349:
350: # define REGEX_REALLOCATE_STACK(source, osize, nsize) \
351: REGEX_REALLOCATE (source, osize, nsize)
352: /* No need to explicitly free anything. */
353: # define REGEX_FREE_STACK(arg)
354:
355: # endif /* not REGEX_MALLOC */
356: #endif /* not using relocating allocator */
357:
358:
359: /* True if `size1' is non-NULL and PTR is pointing anywhere inside
360: `string1' or just past its end. This works if PTR is NULL, which is
361: a good thing. */
362: #define FIRST_STRING_P(ptr) \
363: (size1 && string1 <= (ptr) && (ptr) <= string1 + size1)
364:
365: /* (Re)Allocate N items of type T using malloc, or fail. */
366: #define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t)))
367: #define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t)))
368: #define RETALLOC_IF(addr, n, t) \
369: if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t)
370: #define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t)))
371:
372: #define BYTEWIDTH 8 /* In bits. */
373:
374: #define STREQ(s1, s2) ((strcmp (s1, s2) == 0))
375:
376: #undef MAX
377: #undef MIN
378: #define MAX(a, b) ((a) > (b) ? (a) : (b))
379: #define MIN(a, b) ((a) < (b) ? (a) : (b))
380:
381: typedef char boolean;
382: #define false 0
383: #define true 1
384:
385: static int re_match_2_internal PARAMS ((struct re_pattern_buffer *bufp,
386: const char *string1, int size1,
387: const char *string2, int size2,
388: int pos,
389: struct re_registers *regs,
390: int stop));
391:
392: /* These are the command codes that appear in compiled regular
393: expressions. Some opcodes are followed by argument bytes. A
394: command code can specify any interpretation whatsoever for its
395: arguments. Zero bytes may appear in the compiled regular expression. */
396:
397: typedef enum
398: {
399: no_op = 0,
400:
401: /* Succeed right away--no more backtracking. */
402: succeed,
403:
404: /* Followed by one byte giving n, then by n literal bytes. */
405: exactn,
406:
407: /* Matches any (more or less) character. */
408: anychar,
409:
410: /* Matches any one char belonging to specified set. First
411: following byte is number of bitmap bytes. Then come bytes
412: for a bitmap saying which chars are in. Bits in each byte
413: are ordered low-bit-first. A character is in the set if its
414: bit is 1. A character too large to have a bit in the map is
415: automatically not in the set. */
416: charset,
417:
418: /* Same parameters as charset, but match any character that is
419: not one of those specified. */
420: charset_not,
421:
422: /* Start remembering the text that is matched, for storing in a
423: register. Followed by one byte with the register number, in
424: the range 0 to one less than the pattern buffer's re_nsub
425: field. Then followed by one byte with the number of groups
426: inner to this one. (This last has to be part of the
427: start_memory only because we need it in the on_failure_jump
428: of re_match_2.) */
429: start_memory,
430:
431: /* Stop remembering the text that is matched and store it in a
432: memory register. Followed by one byte with the register
433: number, in the range 0 to one less than `re_nsub' in the
434: pattern buffer, and one byte with the number of inner groups,
435: just like `start_memory'. (We need the number of inner
436: groups here because we don't have any easy way of finding the
437: corresponding start_memory when we're at a stop_memory.) */
438: stop_memory,
439:
440: /* Match a duplicate of something remembered. Followed by one
441: byte containing the register number. */
442: duplicate,
443:
444: /* Fail unless at beginning of line. */
445: begline,
446:
447: /* Fail unless at end of line. */
448: endline,
449:
450: /* Succeeds if at beginning of buffer (if emacs) or at beginning
451: of string to be matched (if not). */
452: begbuf,
453:
454: /* Analogously, for end of buffer/string. */
455: endbuf,
456:
457: /* Followed by two byte relative address to which to jump. */
458: jump,
459:
460: /* Same as jump, but marks the end of an alternative. */
461: jump_past_alt,
462:
463: /* Followed by two-byte relative address of place to resume at
464: in case of failure. */
465: on_failure_jump,
466:
467: /* Like on_failure_jump, but pushes a placeholder instead of the
468: current string position when executed. */
469: on_failure_keep_string_jump,
470:
471: /* Throw away latest failure point and then jump to following
472: two-byte relative address. */
473: pop_failure_jump,
474:
475: /* Change to pop_failure_jump if know won't have to backtrack to
476: match; otherwise change to jump. This is used to jump
477: back to the beginning of a repeat. If what follows this jump
478: clearly won't match what the repeat does, such that we can be
479: sure that there is no use backtracking out of repetitions
480: already matched, then we change it to a pop_failure_jump.
481: Followed by two-byte address. */
482: maybe_pop_jump,
483:
484: /* Jump to following two-byte address, and push a dummy failure
485: point. This failure point will be thrown away if an attempt
486: is made to use it for a failure. A `+' construct makes this
487: before the first repeat. Also used as an intermediary kind
488: of jump when compiling an alternative. */
489: dummy_failure_jump,
490:
491: /* Push a dummy failure point and continue. Used at the end of
492: alternatives. */
493: push_dummy_failure,
494:
495: /* Followed by two-byte relative address and two-byte number n.
496: After matching N times, jump to the address upon failure. */
497: succeed_n,
498:
499: /* Followed by two-byte relative address, and two-byte number n.
500: Jump to the address N times, then fail. */
501: jump_n,
502:
503: /* Set the following two-byte relative address to the
504: subsequent two-byte number. The address *includes* the two
505: bytes of number. */
506: set_number_at,
507:
508: wordchar, /* Matches any word-constituent character. */
509: notwordchar, /* Matches any char that is not a word-constituent. */
510:
511: wordbeg, /* Succeeds if at word beginning. */
512: wordend, /* Succeeds if at word end. */
513:
514: wordbound, /* Succeeds if at a word boundary. */
515: notwordbound /* Succeeds if not at a word boundary. */
516:
517: #ifdef emacs
518: ,before_dot, /* Succeeds if before point. */
519: at_dot, /* Succeeds if at point. */
520: after_dot, /* Succeeds if after point. */
521:
522: /* Matches any character whose syntax is specified. Followed by
523: a byte which contains a syntax code, e.g., Sword. */
524: syntaxspec,
525:
526: /* Matches any character whose syntax is not that specified. */
527: notsyntaxspec
528: #endif /* emacs */
529: } re_opcode_t;
530:
531: /* Common operations on the compiled pattern. */
532:
533: /* Store NUMBER in two contiguous bytes starting at DESTINATION. */
534:
535: #define STORE_NUMBER(destination, number) \
536: do { \
537: (destination)[0] = (number) & 0377; \
538: (destination)[1] = (number) >> 8; \
539: } while (0)
540:
541: /* Same as STORE_NUMBER, except increment DESTINATION to
542: the byte after where the number is stored. Therefore, DESTINATION
543: must be an lvalue. */
544:
545: #define STORE_NUMBER_AND_INCR(destination, number) \
546: do { \
547: STORE_NUMBER (destination, number); \
548: (destination) += 2; \
549: } while (0)
550:
551: /* Put into DESTINATION a number stored in two contiguous bytes starting
552: at SOURCE. */
553:
554: #define EXTRACT_NUMBER(destination, source) \
555: do { \
556: (destination) = *(source) & 0377; \
557: (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \
558: } while (0)
559:
560: #ifdef DEBUG
561: static void extract_number _RE_ARGS ((int *dest, unsigned char *source));
562: static void
563: extract_number (dest, source)
564: int *dest;
565: unsigned char *source;
566: {
567: int temp = SIGN_EXTEND_CHAR (*(source + 1));
568: *dest = *source & 0377;
569: *dest += temp << 8;
570: }
571:
572: # ifndef EXTRACT_MACROS /* To debug the macros. */
573: # undef EXTRACT_NUMBER
574: # define EXTRACT_NUMBER(dest, src) extract_number (&dest, src)
575: # endif /* not EXTRACT_MACROS */
576:
577: #endif /* DEBUG */
578:
579: /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number.
580: SOURCE must be an lvalue. */
581:
582: #define EXTRACT_NUMBER_AND_INCR(destination, source) \
583: do { \
584: EXTRACT_NUMBER (destination, source); \
585: (source) += 2; \
586: } while (0)
587:
588: #ifdef DEBUG
589: static void extract_number_and_incr _RE_ARGS ((int *destination,
590: unsigned char **source));
591: static void
592: extract_number_and_incr (destination, source)
593: int *destination;
594: unsigned char **source;
595: {
596: extract_number (destination, *source);
597: *source += 2;
598: }
599:
600: # ifndef EXTRACT_MACROS
601: # undef EXTRACT_NUMBER_AND_INCR
602: # define EXTRACT_NUMBER_AND_INCR(dest, src) \
603: extract_number_and_incr (&dest, &src)
604: # endif /* not EXTRACT_MACROS */
605:
606: #endif /* DEBUG */
607:
608: /* If DEBUG is defined, Regex prints many voluminous messages about what
609: it is doing (if the variable `debug' is nonzero). If linked with the
610: main program in `iregex.c', you can enter patterns and strings
611: interactively. And if linked with the main program in `main.c' and
612: the other test files, you can run the already-written tests. */
613:
614: #ifdef DEBUG
615:
616: /* We use standard I/O for debugging. */
617: # include <stdio.h>
618:
619: /* It is useful to test things that ``must'' be true when debugging. */
620: # include "zassert.h"
621:
622: static int debug;
623:
624: # define DEBUG_STATEMENT(e) e
625: # define DEBUG_PRINT1(x) if (debug) printf (x)
626: # define DEBUG_PRINT2(x1, x2) if (debug) printf (x1, x2)
627: # define DEBUG_PRINT3(x1, x2, x3) if (debug) printf (x1, x2, x3)
628: # define DEBUG_PRINT4(x1, x2, x3, x4) if (debug) printf (x1, x2, x3, x4)
629: # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \
630: if (debug) print_partial_compiled_pattern (s, e)
631: # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \
632: if (debug) print_double_string (w, s1, sz1, s2, sz2)
633:
634:
635: /* Print the fastmap in human-readable form. */
636:
637: void
638: print_fastmap (fastmap)
639: char *fastmap;
640: {
641: unsigned was_a_range = 0;
642: unsigned i = 0;
643:
644: while (i < (1 << BYTEWIDTH))
645: {
646: if (fastmap[i++])
647: {
648: was_a_range = 0;
649: putchar (i - 1);
650: while (i < (1 << BYTEWIDTH) && fastmap[i])
651: {
652: was_a_range = 1;
653: i++;
654: }
655: if (was_a_range)
656: {
657: printf ("-");
658: putchar (i - 1);
659: }
660: }
661: }
662: putchar ('\n');
663: }
664:
665:
666: /* Print a compiled pattern string in human-readable form, starting at
667: the START pointer into it and ending just before the pointer END. */
668:
669: void
670: print_partial_compiled_pattern (start, end)
671: unsigned char *start;
672: unsigned char *end;
673: {
674: int mcnt, mcnt2;
675: unsigned char *p1;
676: unsigned char *p = start;
677: unsigned char *pend = end;
678:
679: if (start == NULL)
680: {
681: printf ("(null)\n");
682: return;
683: }
684:
685: /* Loop over pattern commands. */
686: while (p < pend)
687: {
688: printf ("%d:\t", p - start);
689:
690: switch ((re_opcode_t) *p++)
691: {
692: case no_op:
693: printf ("/no_op");
694: break;
695:
696: case exactn:
697: mcnt = *p++;
698: printf ("/exactn/%d", mcnt);
699: do
700: {
701: putchar ('/');
702: putchar (*p++);
703: }
704: while (--mcnt);
705: break;
706:
707: case start_memory:
708: mcnt = *p++;
709: printf ("/start_memory/%d/%d", mcnt, *p++);
710: break;
711:
712: case stop_memory:
713: mcnt = *p++;
714: printf ("/stop_memory/%d/%d", mcnt, *p++);
715: break;
716:
717: case duplicate:
718: printf ("/duplicate/%d", *p++);
719: break;
720:
721: case anychar:
722: printf ("/anychar");
723: break;
724:
725: case charset:
726: case charset_not:
727: {
728: register int c, last = -100;
729: register int in_range = 0;
730:
731: printf ("/charset [%s",
732: (re_opcode_t) *(p - 1) == charset_not ? "^" : "");
733:
734: assert (p + *p < pend);
735:
736: for (c = 0; c < 256; c++)
737: if (c / 8 < *p
738: && (p[1 + (c/8)] & (1 << (c % 8))))
739: {
740: /* Are we starting a range? */
741: if (last + 1 == c && ! in_range)
742: {
743: putchar ('-');
744: in_range = 1;
745: }
746: /* Have we broken a range? */
747: else if (last + 1 != c && in_range)
748: {
749: putchar (last);
750: in_range = 0;
751: }
752:
753: if (! in_range)
754: putchar (c);
755:
756: last = c;
757: }
758:
759: if (in_range)
760: putchar (last);
761:
762: putchar (']');
763:
764: p += 1 + *p;
765: }
766: break;
767:
768: case begline:
769: printf ("/begline");
770: break;
771:
772: case endline:
773: printf ("/endline");
774: break;
775:
776: case on_failure_jump:
777: extract_number_and_incr (&mcnt, &p);
778: printf ("/on_failure_jump to %d", p + mcnt - start);
779: break;
780:
781: case on_failure_keep_string_jump:
782: extract_number_and_incr (&mcnt, &p);
783: printf ("/on_failure_keep_string_jump to %d", p + mcnt - start);
784: break;
785:
786: case dummy_failure_jump:
787: extract_number_and_incr (&mcnt, &p);
788: printf ("/dummy_failure_jump to %d", p + mcnt - start);
789: break;
790:
791: case push_dummy_failure:
792: printf ("/push_dummy_failure");
793: break;
794:
795: case maybe_pop_jump:
796: extract_number_and_incr (&mcnt, &p);
797: printf ("/maybe_pop_jump to %d", p + mcnt - start);
798: break;
799:
800: case pop_failure_jump:
801: extract_number_and_incr (&mcnt, &p);
802: printf ("/pop_failure_jump to %d", p + mcnt - start);
803: break;
804:
805: case jump_past_alt:
806: extract_number_and_incr (&mcnt, &p);
807: printf ("/jump_past_alt to %d", p + mcnt - start);
808: break;
809:
810: case jump:
811: extract_number_and_incr (&mcnt, &p);
812: printf ("/jump to %d", p + mcnt - start);
813: break;
814:
815: case succeed_n:
816: extract_number_and_incr (&mcnt, &p);
817: p1 = p + mcnt;
818: extract_number_and_incr (&mcnt2, &p);
819: printf ("/succeed_n to %d, %d times", p1 - start, mcnt2);
820: break;
821:
822: case jump_n:
823: extract_number_and_incr (&mcnt, &p);
824: p1 = p + mcnt;
825: extract_number_and_incr (&mcnt2, &p);
826: printf ("/jump_n to %d, %d times", p1 - start, mcnt2);
827: break;
828:
829: case set_number_at:
830: extract_number_and_incr (&mcnt, &p);
831: p1 = p + mcnt;
832: extract_number_and_incr (&mcnt2, &p);
833: printf ("/set_number_at location %d to %d", p1 - start, mcnt2);
834: break;
835:
836: case wordbound:
837: printf ("/wordbound");
838: break;
839:
840: case notwordbound:
841: printf ("/notwordbound");
842: break;
843:
844: case wordbeg:
845: printf ("/wordbeg");
846: break;
847:
848: case wordend:
849: printf ("/wordend");
850:
851: # ifdef emacs
852: case before_dot:
853: printf ("/before_dot");
854: break;
855:
856: case at_dot:
857: printf ("/at_dot");
858: break;
859:
860: case after_dot:
861: printf ("/after_dot");
862: break;
863:
864: case syntaxspec:
865: printf ("/syntaxspec");
866: mcnt = *p++;
867: printf ("/%d", mcnt);
868: break;
869:
870: case notsyntaxspec:
871: printf ("/notsyntaxspec");
872: mcnt = *p++;
873: printf ("/%d", mcnt);
874: break;
875: # endif /* emacs */
876:
877: case wordchar:
878: printf ("/wordchar");
879: break;
880:
881: case notwordchar:
882: printf ("/notwordchar");
883: break;
884:
885: case begbuf:
886: printf ("/begbuf");
887: break;
888:
889: case endbuf:
890: printf ("/endbuf");
891: break;
892:
893: default:
894: printf ("?%d", *(p-1));
895: }
896:
897: putchar ('\n');
898: }
899:
900: printf ("%d:\tend of pattern.\n", p - start);
901: }
902:
903:
904: void
905: print_compiled_pattern (bufp)
906: struct re_pattern_buffer *bufp;
907: {
908: unsigned char *buffer = bufp->buffer;
909:
910: print_partial_compiled_pattern (buffer, buffer + bufp->used);
911: printf ("%ld bytes used/%ld bytes allocated.\n",
912: bufp->used, bufp->allocated);
913:
914: if (bufp->fastmap_accurate && bufp->fastmap)
915: {
916: printf ("fastmap: ");
917: print_fastmap (bufp->fastmap);
918: }
919:
920: printf ("re_nsub: %d\t", bufp->re_nsub);
921: printf ("regs_alloc: %d\t", bufp->regs_allocated);
922: printf ("can_be_null: %d\t", bufp->can_be_null);
923: printf ("newline_anchor: %d\n", bufp->newline_anchor);
924: printf ("no_sub: %d\t", bufp->no_sub);
925: printf ("not_bol: %d\t", bufp->not_bol);
926: printf ("not_eol: %d\t", bufp->not_eol);
927: printf ("syntax: %lx\n", bufp->syntax);
928: /* Perhaps we should print the translate table? */
929: }
930:
931:
932: void
933: print_double_string (where, string1, size1, string2, size2)
934: const char *where;
935: const char *string1;
936: const char *string2;
937: int size1;
938: int size2;
939: {
940: int this_char;
941:
942: if (where == NULL)
943: printf ("(null)");
944: else
945: {
946: if (FIRST_STRING_P (where))
947: {
948: for (this_char = where - string1; this_char < size1; this_char++)
949: putchar (string1[this_char]);
950:
951: where = string2;
952: }
953:
954: for (this_char = where - string2; this_char < size2; this_char++)
955: putchar (string2[this_char]);
956: }
957: }
958:
959: void
960: printchar (c)
961: int c;
962: {
963: putc (c, stderr);
964: }
965:
966: #else /* not DEBUG */
967:
968: # undef assert
969: # define assert(e)
970:
971: # define DEBUG_STATEMENT(e)
972: # define DEBUG_PRINT1(x)
973: # define DEBUG_PRINT2(x1, x2)
974: # define DEBUG_PRINT3(x1, x2, x3)
975: # define DEBUG_PRINT4(x1, x2, x3, x4)
976: # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e)
977: # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2)
978:
979: #endif /* not DEBUG */
980:
981: /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can
982: also be assigned to arbitrarily: each pattern buffer stores its own
983: syntax, so it can be changed between regex compilations. */
984: /* This has no initializer because initialized variables in Emacs
985: become read-only after dumping. */
986: reg_syntax_t re_syntax_options;
987:
988:
989: /* Specify the precise syntax of regexps for compilation. This provides
990: for compatibility for various utilities which historically have
991: different, incompatible syntaxes.
992:
993: The argument SYNTAX is a bit mask comprised of the various bits
994: defined in regex.h. We return the old syntax. */
995:
996: reg_syntax_t
997: re_set_syntax (syntax)
998: reg_syntax_t syntax;
999: {
1000: reg_syntax_t ret = re_syntax_options;
1001:
1002: re_syntax_options = syntax;
1003: #ifdef DEBUG
1004: if (syntax & RE_DEBUG)
1005: debug = 1;
1006: else if (debug) /* was on but now is not */
1007: debug = 0;
1008: #endif /* DEBUG */
1009: return ret;
1010: }
1011: #ifdef _LIBC
1012: weak_alias (__re_set_syntax, re_set_syntax)
1013: #endif
1014:
1015: /* This table gives an error message for each of the error codes listed
1016: in regex.h. Obviously the order here has to be same as there.
1017: POSIX doesn't require that we do anything for REG_NOERROR,
1018: but why not be nice? */
1019:
1020: static const char re_error_msgid[] =
1021: {
1022: #define REG_NOERROR_IDX 0
1023: gettext_noop ("Success") /* REG_NOERROR */
1024: "\0"
1025: #define REG_NOMATCH_IDX (REG_NOERROR_IDX + sizeof "Success")
1026: gettext_noop ("No match") /* REG_NOMATCH */
1027: "\0"
1028: #define REG_BADPAT_IDX (REG_NOMATCH_IDX + sizeof "No match")
1029: gettext_noop ("Invalid regular expression") /* REG_BADPAT */
1030: "\0"
1031: #define REG_ECOLLATE_IDX (REG_BADPAT_IDX + sizeof "Invalid regular expression")
1032: gettext_noop ("Invalid collation character") /* REG_ECOLLATE */
1033: "\0"
1034: #define REG_ECTYPE_IDX (REG_ECOLLATE_IDX + sizeof "Invalid collation character")
1035: gettext_noop ("Invalid character class name") /* REG_ECTYPE */
1036: "\0"
1037: #define REG_EESCAPE_IDX (REG_ECTYPE_IDX + sizeof "Invalid character class name")
1038: gettext_noop ("Trailing backslash") /* REG_EESCAPE */
1039: "\0"
1040: #define REG_ESUBREG_IDX (REG_EESCAPE_IDX + sizeof "Trailing backslash")
1041: gettext_noop ("Invalid back reference") /* REG_ESUBREG */
1042: "\0"
1043: #define REG_EBRACK_IDX (REG_ESUBREG_IDX + sizeof "Invalid back reference")
1044: gettext_noop ("Unmatched [ or [^") /* REG_EBRACK */
1045: "\0"
1046: #define REG_EPAREN_IDX (REG_EBRACK_IDX + sizeof "Unmatched [ or [^")
1047: gettext_noop ("Unmatched ( or \\(") /* REG_EPAREN */
1048: "\0"
1049: #define REG_EBRACE_IDX (REG_EPAREN_IDX + sizeof "Unmatched ( or \\(")
1050: gettext_noop ("Unmatched \\{") /* REG_EBRACE */
1051: "\0"
1052: #define REG_BADBR_IDX (REG_EBRACE_IDX + sizeof "Unmatched \\{")
1053: gettext_noop ("Invalid content of \\{\\}") /* REG_BADBR */
1054: "\0"
1055: #define REG_ERANGE_IDX (REG_BADBR_IDX + sizeof "Invalid content of \\{\\}")
1056: gettext_noop ("Invalid range end") /* REG_ERANGE */
1057: "\0"
1058: #define REG_ESPACE_IDX (REG_ERANGE_IDX + sizeof "Invalid range end")
1059: gettext_noop ("Memory exhausted") /* REG_ESPACE */
1060: "\0"
1061: #define REG_BADRPT_IDX (REG_ESPACE_IDX + sizeof "Memory exhausted")
1062: gettext_noop ("Invalid preceding regular expression") /* REG_BADRPT */
1063: "\0"
1064: #define REG_EEND_IDX (REG_BADRPT_IDX + sizeof "Invalid preceding regular expression")
1065: gettext_noop ("Premature end of regular expression") /* REG_EEND */
1066: "\0"
1067: #define REG_ESIZE_IDX (REG_EEND_IDX + sizeof "Premature end of regular expression")
1068: gettext_noop ("Regular expression too big") /* REG_ESIZE */
1069: "\0"
1070: #define REG_ERPAREN_IDX (REG_ESIZE_IDX + sizeof "Regular expression too big")
1071: gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */
1072: };
1073:
1074: static const size_t re_error_msgid_idx[] =
1075: {
1076: REG_NOERROR_IDX,
1077: REG_NOMATCH_IDX,
1078: REG_BADPAT_IDX,
1079: REG_ECOLLATE_IDX,
1080: REG_ECTYPE_IDX,
1081: REG_EESCAPE_IDX,
1082: REG_ESUBREG_IDX,
1083: REG_EBRACK_IDX,
1084: REG_EPAREN_IDX,
1085: REG_EBRACE_IDX,
1086: REG_BADBR_IDX,
1087: REG_ERANGE_IDX,
1088: REG_ESPACE_IDX,
1089: REG_BADRPT_IDX,
1090: REG_EEND_IDX,
1091: REG_ESIZE_IDX,
1092: REG_ERPAREN_IDX
1093: };
1094:
1095: /* Avoiding alloca during matching, to placate r_alloc. */
1096:
1097: /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the
1098: searching and matching functions should not call alloca. On some
1099: systems, alloca is implemented in terms of malloc, and if we're
1100: using the relocating allocator routines, then malloc could cause a
1101: relocation, which might (if the strings being searched are in the
1102: ralloc heap) shift the data out from underneath the regexp
1103: routines.
1104:
1105: Here's another reason to avoid allocation: Emacs
1106: processes input from X in a signal handler; processing X input may
1107: call malloc; if input arrives while a matching routine is calling
1108: malloc, then we're scrod. But Emacs can't just block input while
1109: calling matching routines; then we don't notice interrupts when
1110: they come in. So, Emacs blocks input around all regexp calls
1111: except the matching calls, which it leaves unprotected, in the
1112: faith that they will not malloc. */
1113:
1114: /* Normally, this is fine. */
1115: #define MATCH_MAY_ALLOCATE
1116:
1117: /* When using GNU C, we are not REALLY using the C alloca, no matter
1118: what config.h may say. So don't take precautions for it. */
1119: #ifdef __GNUC__
1120: # undef C_ALLOCA
1121: #endif
1122:
1123: /* The match routines may not allocate if (1) they would do it with malloc
1124: and (2) it's not safe for them to use malloc.
1125: Note that if REL_ALLOC is defined, matching would not use malloc for the
1126: failure stack, but we would still use it for the register vectors;
1127: so REL_ALLOC should not affect this. */
1128: #if (defined C_ALLOCA || defined REGEX_MALLOC) && defined emacs
1129: # undef MATCH_MAY_ALLOCATE
1130: #endif
1131:
1132:
1133: /* Failure stack declarations and macros; both re_compile_fastmap and
1134: re_match_2 use a failure stack. These have to be macros because of
1135: REGEX_ALLOCATE_STACK. */
1136:
1137:
1138: /* Number of failure points for which to initially allocate space
1139: when matching. If this number is exceeded, we allocate more
1140: space, so it is not a hard limit. */
1141: #ifndef INIT_FAILURE_ALLOC
1142: # define INIT_FAILURE_ALLOC 5
1143: #endif
1144:
1145: /* Roughly the maximum number of failure points on the stack. Would be
1146: exactly that if always used MAX_FAILURE_ITEMS items each time we failed.
1147: This is a variable only so users of regex can assign to it; we never
1148: change it ourselves. */
1149:
1150: #ifdef INT_IS_16BIT
1151:
1152: # if defined MATCH_MAY_ALLOCATE
1153: /* 4400 was enough to cause a crash on Alpha OSF/1,
1154: whose default stack limit is 2mb. */
1155: long int re_max_failures = 4000;
1156: # else
1157: long int re_max_failures = 2000;
1158: # endif
1159:
1160: union fail_stack_elt
1161: {
1162: unsigned char *pointer;
1163: long int integer;
1164: };
1165:
1166: typedef union fail_stack_elt fail_stack_elt_t;
1167:
1168: typedef struct
1169: {
1170: fail_stack_elt_t *stack;
1171: unsigned long int size;
1172: unsigned long int avail; /* Offset of next open position. */
1173: } fail_stack_type;
1174:
1175: #else /* not INT_IS_16BIT */
1176:
1177: # if defined MATCH_MAY_ALLOCATE
1178: /* 4400 was enough to cause a crash on Alpha OSF/1,
1179: whose default stack limit is 2mb. */
1180: int re_max_failures = 20000;
1181: # else
1182: int re_max_failures = 2000;
1183: # endif
1184:
1185: union fail_stack_elt
1186: {
1187: unsigned char *pointer;
1188: int integer;
1189: };
1190:
1191: typedef union fail_stack_elt fail_stack_elt_t;
1192:
1193: typedef struct
1194: {
1195: fail_stack_elt_t *stack;
1196: unsigned size;
1197: unsigned avail; /* Offset of next open position. */
1198: } fail_stack_type;
1199:
1200: #endif /* INT_IS_16BIT */
1201:
1202: #define FAIL_STACK_EMPTY() (fail_stack.avail == 0)
1203: #define FAIL_STACK_PTR_EMPTY() (fail_stack_ptr->avail == 0)
1204: #define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size)
1205:
1206:
1207: /* Define macros to initialize and free the failure stack.
1208: Do `return -2' if the alloc fails. */
1209:
1210: #ifdef MATCH_MAY_ALLOCATE
1211: # define INIT_FAIL_STACK() \
1212: do { \
1213: fail_stack.stack = (fail_stack_elt_t *) \
1214: REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * sizeof (fail_stack_elt_t)); \
1215: \
1216: if (fail_stack.stack == NULL) \
1217: return -2; \
1218: \
1219: fail_stack.size = INIT_FAILURE_ALLOC; \
1220: fail_stack.avail = 0; \
1221: } while (0)
1222:
1223: # define RESET_FAIL_STACK() REGEX_FREE_STACK (fail_stack.stack)
1224: #else
1225: # define INIT_FAIL_STACK() \
1226: do { \
1227: fail_stack.avail = 0; \
1228: } while (0)
1229:
1230: # define RESET_FAIL_STACK()
1231: #endif
1232:
1233:
1234: /* Double the size of FAIL_STACK, up to approximately `re_max_failures' items.
1235:
1236: Return 1 if succeeds, and 0 if either ran out of memory
1237: allocating space for it or it was already too large.
1238:
1239: REGEX_REALLOCATE_STACK requires `destination' be declared. */
1240:
1241: #define DOUBLE_FAIL_STACK(fail_stack) \
1242: ((fail_stack).size > (unsigned) (re_max_failures * MAX_FAILURE_ITEMS) \
1243: ? 0 \
1244: : ((fail_stack).stack = (fail_stack_elt_t *) \
1245: REGEX_REALLOCATE_STACK ((fail_stack).stack, \
1246: (fail_stack).size * sizeof (fail_stack_elt_t), \
1247: ((fail_stack).size << 1) * sizeof (fail_stack_elt_t)), \
1248: \
1249: (fail_stack).stack == NULL \
1250: ? 0 \
1251: : ((fail_stack).size <<= 1, \
1252: 1)))
1253:
1254:
1255: /* Push pointer POINTER on FAIL_STACK.
1256: Return 1 if was able to do so and 0 if ran out of memory allocating
1257: space to do so. */
1258: #define PUSH_PATTERN_OP(POINTER, FAIL_STACK) \
1259: ((FAIL_STACK_FULL () \
1260: && !DOUBLE_FAIL_STACK (FAIL_STACK)) \
1261: ? 0 \
1262: : ((FAIL_STACK).stack[(FAIL_STACK).avail++].pointer = POINTER, \
1263: 1))
1264:
1265: /* Push a pointer value onto the failure stack.
1266: Assumes the variable `fail_stack'. Probably should only
1267: be called from within `PUSH_FAILURE_POINT'. */
1268: #define PUSH_FAILURE_POINTER(item) \
1269: fail_stack.stack[fail_stack.avail++].pointer = (unsigned char *) (item)
1270:
1271: /* This pushes an integer-valued item onto the failure stack.
1272: Assumes the variable `fail_stack'. Probably should only
1273: be called from within `PUSH_FAILURE_POINT'. */
1274: #define PUSH_FAILURE_INT(item) \
1275: fail_stack.stack[fail_stack.avail++].integer = (item)
1276:
1277: /* Push a fail_stack_elt_t value onto the failure stack.
1278: Assumes the variable `fail_stack'. Probably should only
1279: be called from within `PUSH_FAILURE_POINT'. */
1280: #define PUSH_FAILURE_ELT(item) \
1281: fail_stack.stack[fail_stack.avail++] = (item)
1282:
1283: /* These three POP... operations complement the three PUSH... operations.
1284: All assume that `fail_stack' is nonempty. */
1285: #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer
1286: #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer
1287: #define POP_FAILURE_ELT() fail_stack.stack[--fail_stack.avail]
1288:
1289: /* Used to omit pushing failure point id's when we're not debugging. */
1290: #ifdef DEBUG
1291: # define DEBUG_PUSH PUSH_FAILURE_INT
1292: # define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_INT ()
1293: #else
1294: # define DEBUG_PUSH(item)
1295: # define DEBUG_POP(item_addr)
1296: #endif
1297:
1298:
1299: /* Push the information about the state we will need
1300: if we ever fail back to it.
1301:
1302: Requires variables fail_stack, regstart, regend, reg_info, and
1303: num_regs_pushed be declared. DOUBLE_FAIL_STACK requires `destination'
1304: be declared.
1305:
1306: Does `return FAILURE_CODE' if runs out of memory. */
1307:
1308: #define PUSH_FAILURE_POINT(pattern_place, string_place, failure_code) \
1309: do { \
1310: char *destination; \
1311: /* Must be int, so when we don't save any registers, the arithmetic \
1312: of 0 + -1 isn't done as unsigned. */ \
1313: /* Can't be int, since there is not a shred of a guarantee that int \
1314: is wide enough to hold a value of something to which pointer can \
1315: be assigned */ \
1316: active_reg_t this_reg; \
1317: \
1318: DEBUG_STATEMENT (failure_id++); \
1319: DEBUG_STATEMENT (nfailure_points_pushed++); \
1320: DEBUG_PRINT2 ("\nPUSH_FAILURE_POINT #%u:\n", failure_id); \
1321: DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail);\
1322: DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\
1323: \
1324: DEBUG_PRINT2 (" slots needed: %ld\n", NUM_FAILURE_ITEMS); \
1325: DEBUG_PRINT2 (" available: %d\n", REMAINING_AVAIL_SLOTS); \
1326: \
1327: /* Ensure we have enough space allocated for what we will push. */ \
1328: while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \
1329: { \
1330: if (!DOUBLE_FAIL_STACK (fail_stack)) \
1331: return failure_code; \
1332: \
1333: DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", \
1334: (fail_stack).size); \
1335: DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\
1336: } \
1337: \
1338: /* Push the info, starting with the registers. */ \
1339: DEBUG_PRINT1 ("\n"); \
1340: \
1341: if (1) \
1342: for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \
1343: this_reg++) \
1344: { \
1345: DEBUG_PRINT2 (" Pushing reg: %lu\n", this_reg); \
1346: DEBUG_STATEMENT (num_regs_pushed++); \
1347: \
1348: DEBUG_PRINT2 (" start: %p\n", regstart[this_reg]); \
1349: PUSH_FAILURE_POINTER (regstart[this_reg]); \
1350: \
1351: DEBUG_PRINT2 (" end: %p\n", regend[this_reg]); \
1352: PUSH_FAILURE_POINTER (regend[this_reg]); \
1353: \
1354: DEBUG_PRINT2 (" info: %p\n ", \
1355: reg_info[this_reg].word.pointer); \
1356: DEBUG_PRINT2 (" match_null=%d", \
1357: REG_MATCH_NULL_STRING_P (reg_info[this_reg])); \
1358: DEBUG_PRINT2 (" active=%d", IS_ACTIVE (reg_info[this_reg])); \
1359: DEBUG_PRINT2 (" matched_something=%d", \
1360: MATCHED_SOMETHING (reg_info[this_reg])); \
1361: DEBUG_PRINT2 (" ever_matched=%d", \
1362: EVER_MATCHED_SOMETHING (reg_info[this_reg])); \
1363: DEBUG_PRINT1 ("\n"); \
1364: PUSH_FAILURE_ELT (reg_info[this_reg].word); \
1365: } \
1366: \
1367: DEBUG_PRINT2 (" Pushing low active reg: %ld\n", lowest_active_reg);\
1368: PUSH_FAILURE_INT (lowest_active_reg); \
1369: \
1370: DEBUG_PRINT2 (" Pushing high active reg: %ld\n", highest_active_reg);\
1371: PUSH_FAILURE_INT (highest_active_reg); \
1372: \
1373: DEBUG_PRINT2 (" Pushing pattern %p:\n", pattern_place); \
1374: DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \
1375: PUSH_FAILURE_POINTER (pattern_place); \
1376: \
1377: DEBUG_PRINT2 (" Pushing string %p: `", string_place); \
1378: DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, \
1379: size2); \
1380: DEBUG_PRINT1 ("'\n"); \
1381: PUSH_FAILURE_POINTER (string_place); \
1382: \
1383: DEBUG_PRINT2 (" Pushing failure id: %u\n", failure_id); \
1384: DEBUG_PUSH (failure_id); \
1385: } while (0)
1386:
1387: /* This is the number of items that are pushed and popped on the stack
1388: for each register. */
1389: #define NUM_REG_ITEMS 3
1390:
1391: /* Individual items aside from the registers. */
1392: #ifdef DEBUG
1393: # define NUM_NONREG_ITEMS 5 /* Includes failure point id. */
1394: #else
1395: # define NUM_NONREG_ITEMS 4
1396: #endif
1397:
1398: /* We push at most this many items on the stack. */
1399: /* We used to use (num_regs - 1), which is the number of registers
1400: this regexp will save; but that was changed to 5
1401: to avoid stack overflow for a regexp with lots of parens. */
1402: #define MAX_FAILURE_ITEMS (5 * NUM_REG_ITEMS + NUM_NONREG_ITEMS)
1403:
1404: /* We actually push this many items. */
1405: #define NUM_FAILURE_ITEMS \
1406: (((0 \
1407: ? 0 : highest_active_reg - lowest_active_reg + 1) \
1408: * NUM_REG_ITEMS) \
1409: + NUM_NONREG_ITEMS)
1410:
1411: /* How many items can still be added to the stack without overflowing it. */
1412: #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail)
1413:
1414:
1415: /* Pops what PUSH_FAIL_STACK pushes.
1416:
1417: We restore into the parameters, all of which should be lvalues:
1418: STR -- the saved data position.
1419: PAT -- the saved pattern position.
1420: LOW_REG, HIGH_REG -- the highest and lowest active registers.
1421: REGSTART, REGEND -- arrays of string positions.
1422: REG_INFO -- array of information about each subexpression.
1423:
1424: Also assumes the variables `fail_stack' and (if debugging), `bufp',
1425: `pend', `string1', `size1', `string2', and `size2'. */
1426:
1427: #define POP_FAILURE_POINT(str, pat, low_reg, high_reg, regstart, regend, reg_info)\
1428: { \
1429: DEBUG_STATEMENT (unsigned failure_id;) \
1430: active_reg_t this_reg; \
1431: const unsigned char *string_temp; \
1432: \
1433: assert (!FAIL_STACK_EMPTY ()); \
1434: \
1435: /* Remove failure points and point to how many regs pushed. */ \
1436: DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \
1437: DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \
1438: DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \
1439: \
1440: assert (fail_stack.avail >= NUM_NONREG_ITEMS); \
1441: \
1442: DEBUG_POP (&failure_id); \
1443: DEBUG_PRINT2 (" Popping failure id: %u\n", failure_id); \
1444: \
1445: /* If the saved string location is NULL, it came from an \
1446: on_failure_keep_string_jump opcode, and we want to throw away the \
1447: saved NULL, thus retaining our current position in the string. */ \
1448: string_temp = POP_FAILURE_POINTER (); \
1449: if (string_temp != NULL) \
1450: str = (const char *) string_temp; \
1451: \
1452: DEBUG_PRINT2 (" Popping string %p: `", str); \
1453: DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \
1454: DEBUG_PRINT1 ("'\n"); \
1455: \
1456: pat = (unsigned char *) POP_FAILURE_POINTER (); \
1457: DEBUG_PRINT2 (" Popping pattern %p:\n", pat); \
1458: DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \
1459: \
1460: /* Restore register info. */ \
1461: high_reg = (active_reg_t) POP_FAILURE_INT (); \
1462: DEBUG_PRINT2 (" Popping high active reg: %ld\n", high_reg); \
1463: \
1464: low_reg = (active_reg_t) POP_FAILURE_INT (); \
1465: DEBUG_PRINT2 (" Popping low active reg: %ld\n", low_reg); \
1466: \
1467: if (1) \
1468: for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \
1469: { \
1470: DEBUG_PRINT2 (" Popping reg: %ld\n", this_reg); \
1471: \
1472: reg_info[this_reg].word = POP_FAILURE_ELT (); \
1473: DEBUG_PRINT2 (" info: %p\n", \
1474: reg_info[this_reg].word.pointer); \
1475: \
1476: regend[this_reg] = (const char *) POP_FAILURE_POINTER (); \
1477: DEBUG_PRINT2 (" end: %p\n", regend[this_reg]); \
1478: \
1479: regstart[this_reg] = (const char *) POP_FAILURE_POINTER (); \
1480: DEBUG_PRINT2 (" start: %p\n", regstart[this_reg]); \
1481: } \
1482: else \
1483: { \
1484: for (this_reg = highest_active_reg; this_reg > high_reg; this_reg--) \
1485: { \
1486: reg_info[this_reg].word.integer = 0; \
1487: regend[this_reg] = 0; \
1488: regstart[this_reg] = 0; \
1489: } \
1490: highest_active_reg = high_reg; \
1491: } \
1492: \
1493: set_regs_matched_done = 0; \
1494: DEBUG_STATEMENT (nfailure_points_popped++); \
1495: } /* POP_FAILURE_POINT */
1496:
1497:
1498:
1499: /* Structure for per-register (a.k.a. per-group) information.
1500: Other register information, such as the
1501: starting and ending positions (which are addresses), and the list of
1502: inner groups (which is a bits list) are maintained in separate
1503: variables.
1504:
1505: We are making a (strictly speaking) nonportable assumption here: that
1506: the compiler will pack our bit fields into something that fits into
1507: the type of `word', i.e., is something that fits into one item on the
1508: failure stack. */
1509:
1510:
1511: /* Declarations and macros for re_match_2. */
1512:
1513: typedef union
1514: {
1515: fail_stack_elt_t word;
1516: struct
1517: {
1518: /* This field is one if this group can match the empty string,
1519: zero if not. If not yet determined, `MATCH_NULL_UNSET_VALUE'. */
1520: #define MATCH_NULL_UNSET_VALUE 3
1521: unsigned match_null_string_p : 2;
1522: unsigned is_active : 1;
1523: unsigned matched_something : 1;
1524: unsigned ever_matched_something : 1;
1525: } bits;
1526: } register_info_type;
1527:
1528: #define REG_MATCH_NULL_STRING_P(R) ((R).bits.match_null_string_p)
1529: #define IS_ACTIVE(R) ((R).bits.is_active)
1530: #define MATCHED_SOMETHING(R) ((R).bits.matched_something)
1531: #define EVER_MATCHED_SOMETHING(R) ((R).bits.ever_matched_something)
1532:
1533:
1534: /* Call this when have matched a real character; it sets `matched' flags
1535: for the subexpressions which we are currently inside. Also records
1536: that those subexprs have matched. */
1537: #define SET_REGS_MATCHED() \
1538: do \
1539: { \
1540: if (!set_regs_matched_done) \
1541: { \
1542: active_reg_t r; \
1543: set_regs_matched_done = 1; \
1544: for (r = lowest_active_reg; r <= highest_active_reg; r++) \
1545: { \
1546: MATCHED_SOMETHING (reg_info[r]) \
1547: = EVER_MATCHED_SOMETHING (reg_info[r]) \
1548: = 1; \
1549: } \
1550: } \
1551: } \
1552: while (0)
1553:
1554: /* Registers are set to a sentinel when they haven't yet matched. */
1555: static char reg_unset_dummy;
1556: #define REG_UNSET_VALUE (®_unset_dummy)
1557: #define REG_UNSET(e) ((e) == REG_UNSET_VALUE)
1558:
1559: /* Subroutine declarations and macros for regex_compile. */
1560:
1561: static reg_errcode_t regex_compile _RE_ARGS ((const char *pattern, size_t size,
1562: reg_syntax_t syntax,
1563: struct re_pattern_buffer *bufp));
1564: static void store_op1 _RE_ARGS ((re_opcode_t op, unsigned char *loc, int arg));
1565: static void store_op2 _RE_ARGS ((re_opcode_t op, unsigned char *loc,
1566: int arg1, int arg2));
1567: static void insert_op1 _RE_ARGS ((re_opcode_t op, unsigned char *loc,
1568: int arg, unsigned char *end));
1569: static void insert_op2 _RE_ARGS ((re_opcode_t op, unsigned char *loc,
1570: int arg1, int arg2, unsigned char *end));
1571: static boolean at_begline_loc_p _RE_ARGS ((const char *pattern, const char *p,
1572: reg_syntax_t syntax));
1573: static boolean at_endline_loc_p _RE_ARGS ((const char *p, const char *pend,
1574: reg_syntax_t syntax));
1575: static reg_errcode_t compile_range _RE_ARGS ((const char **p_ptr,
1576: const char *pend,
1577: char *translate,
1578: reg_syntax_t syntax,
1579: unsigned char *b));
1580:
1581: /* Fetch the next character in the uncompiled pattern---translating it
1582: if necessary. Also cast from a signed character in the constant
1583: string passed to us by the user to an unsigned char that we can use
1584: as an array index (in, e.g., `translate'). */
1585: #ifndef PATFETCH
1586: # define PATFETCH(c) \
1587: do {if (p == pend) return REG_EEND; \
1588: c = (unsigned char) *p++; \
1589: if (translate) c = (unsigned char) translate[c]; \
1590: } while (0)
1591: #endif
1592:
1593: /* Fetch the next character in the uncompiled pattern, with no
1594: translation. */
1595: #define PATFETCH_RAW(c) \
1596: do {if (p == pend) return REG_EEND; \
1597: c = (unsigned char) *p++; \
1598: } while (0)
1599:
1600: /* Go backwards one character in the pattern. */
1601: #define PATUNFETCH p--
1602:
1603:
1604: /* If `translate' is non-null, return translate[D], else just D. We
1605: cast the subscript to translate because some data is declared as
1606: `char *', to avoid warnings when a string constant is passed. But
1607: when we use a character as a subscript we must make it unsigned. */
1608: #ifndef TRANSLATE
1609: # define TRANSLATE(d) \
1610: (translate ? (char) translate[(unsigned char) (d)] : (d))
1611: #endif
1612:
1613:
1614: /* Macros for outputting the compiled pattern into `buffer'. */
1615:
1616: /* If the buffer isn't allocated when it comes in, use this. */
1617: #define INIT_BUF_SIZE 32
1618:
1619: /* Make sure we have at least N more bytes of space in buffer. */
1620: #define GET_BUFFER_SPACE(n) \
1621: while ((unsigned long) (b - bufp->buffer + (n)) > bufp->allocated) \
1622: EXTEND_BUFFER ()
1623:
1624: /* Make sure we have one more byte of buffer space and then add C to it. */
1625: #define BUF_PUSH(c) \
1626: do { \
1627: GET_BUFFER_SPACE (1); \
1628: *b++ = (unsigned char) (c); \
1629: } while (0)
1630:
1631:
1632: /* Ensure we have two more bytes of buffer space and then append C1 and C2. */
1633: #define BUF_PUSH_2(c1, c2) \
1634: do { \
1635: GET_BUFFER_SPACE (2); \
1636: *b++ = (unsigned char) (c1); \
1637: *b++ = (unsigned char) (c2); \
1638: } while (0)
1639:
1640:
1641: /* As with BUF_PUSH_2, except for three bytes. */
1642: #define BUF_PUSH_3(c1, c2, c3) \
1643: do { \
1644: GET_BUFFER_SPACE (3); \
1645: *b++ = (unsigned char) (c1); \
1646: *b++ = (unsigned char) (c2); \
1647: *b++ = (unsigned char) (c3); \
1648: } while (0)
1649:
1650:
1651: /* Store a jump with opcode OP at LOC to location TO. We store a
1652: relative address offset by the three bytes the jump itself occupies. */
1653: #define STORE_JUMP(op, loc, to) \
1654: store_op1 (op, loc, (int) ((to) - (loc) - 3))
1655:
1656: /* Likewise, for a two-argument jump. */
1657: #define STORE_JUMP2(op, loc, to, arg) \
1658: store_op2 (op, loc, (int) ((to) - (loc) - 3), arg)
1659:
1660: /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */
1661: #define INSERT_JUMP(op, loc, to) \
1662: insert_op1 (op, loc, (int) ((to) - (loc) - 3), b)
1663:
1664: /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */
1665: #define INSERT_JUMP2(op, loc, to, arg) \
1666: insert_op2 (op, loc, (int) ((to) - (loc) - 3), arg, b)
1667:
1668:
1669: /* This is not an arbitrary limit: the arguments which represent offsets
1670: into the pattern are two bytes long. So if 2^16 bytes turns out to
1671: be too small, many things would have to change. */
1672: /* Any other compiler which, like MSC, has allocation limit below 2^16
1673: bytes will have to use approach similar to what was done below for
1674: MSC and drop MAX_BUF_SIZE a bit. Otherwise you may end up
1675: reallocating to 0 bytes. Such thing is not going to work too well.
1676: You have been warned!! */
1677: #if defined _MSC_VER && !defined _WIN32
1678: /* Microsoft C 16-bit versions limit malloc to approx 65512 bytes.
1679: The REALLOC define eliminates a flurry of conversion warnings,
1680: but is not required. */
1681: # define MAX_BUF_SIZE 65500L
1682: # define REALLOC(p,s) realloc ((p), (size_t) (s))
1683: #else
1684: # define MAX_BUF_SIZE (1L << 16)
1685: # define REALLOC(p,s) realloc ((p), (s))
1686: #endif
1687:
1688: /* Extend the buffer by twice its current size via realloc and
1689: reset the pointers that pointed into the old block to point to the
1690: correct places in the new one. If extending the buffer results in it
1691: being larger than MAX_BUF_SIZE, then flag memory exhausted. */
1692: #define EXTEND_BUFFER() \
1693: do { \
1694: unsigned char *old_buffer = bufp->buffer; \
1695: if (bufp->allocated == MAX_BUF_SIZE) \
1696: return REG_ESIZE; \
1697: bufp->allocated <<= 1; \
1698: if (bufp->allocated > MAX_BUF_SIZE) \
1699: bufp->allocated = MAX_BUF_SIZE; \
1700: bufp->buffer = (unsigned char *) REALLOC (bufp->buffer, bufp->allocated);\
1701: if (bufp->buffer == NULL) \
1702: return REG_ESPACE; \
1703: /* If the buffer moved, move all the pointers into it. */ \
1704: if (old_buffer != bufp->buffer) \
1705: { \
1706: b = (b - old_buffer) + bufp->buffer; \
1707: begalt = (begalt - old_buffer) + bufp->buffer; \
1708: if (fixup_alt_jump) \
1709: fixup_alt_jump = (fixup_alt_jump - old_buffer) + bufp->buffer;\
1710: if (laststart) \
1711: laststart = (laststart - old_buffer) + bufp->buffer; \
1712: if (pending_exact) \
1713: pending_exact = (pending_exact - old_buffer) + bufp->buffer; \
1714: } \
1715: } while (0)
1716:
1717:
1718: /* Since we have one byte reserved for the register number argument to
1719: {start,stop}_memory, the maximum number of groups we can report
1720: things about is what fits in that byte. */
1721: #define MAX_REGNUM 255
1722:
1723: /* But patterns can have more than `MAX_REGNUM' registers. We just
1724: ignore the excess. */
1725: typedef unsigned regnum_t;
1726:
1727:
1728: /* Macros for the compile stack. */
1729:
1730: /* Since offsets can go either forwards or backwards, this type needs to
1731: be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */
1732: /* int may be not enough when sizeof(int) == 2. */
1733: typedef long pattern_offset_t;
1734:
1735: typedef struct
1736: {
1737: pattern_offset_t begalt_offset;
1738: pattern_offset_t fixup_alt_jump;
1739: pattern_offset_t inner_group_offset;
1740: pattern_offset_t laststart_offset;
1741: regnum_t regnum;
1742: } compile_stack_elt_t;
1743:
1744:
1745: typedef struct
1746: {
1747: compile_stack_elt_t *stack;
1748: unsigned size;
1749: unsigned avail; /* Offset of next open position. */
1750: } compile_stack_type;
1751:
1752:
1753: #define INIT_COMPILE_STACK_SIZE 32
1754:
1755: #define COMPILE_STACK_EMPTY (compile_stack.avail == 0)
1756: #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size)
1757:
1758: /* The next available element. */
1759: #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail])
1760:
1761:
1762: /* Set the bit for character C in a list. */
1763: #define SET_LIST_BIT(c) \
1764: (b[((unsigned char) (c)) / BYTEWIDTH] \
1765: |= 1 << (((unsigned char) c) % BYTEWIDTH))
1766:
1767:
1768: /* Get the next unsigned number in the uncompiled pattern. */
1769: #define GET_UNSIGNED_NUMBER(num) \
1770: { if (p != pend) \
1771: { \
1772: PATFETCH (c); \
1773: while (ISDIGIT (c)) \
1774: { \
1775: if (num < 0) \
1776: num = 0; \
1777: num = num * 10 + c - '0'; \
1778: if (p == pend) \
1779: break; \
1780: PATFETCH (c); \
1781: } \
1782: } \
1783: }
1784:
1785: #if defined _LIBC || WIDE_CHAR_SUPPORT
1786: /* The GNU C library provides support for user-defined character classes
1787: and the functions from ISO C amendement 1. */
1788: # ifdef CHARCLASS_NAME_MAX
1789: # define CHAR_CLASS_MAX_LENGTH CHARCLASS_NAME_MAX
1790: # else
1791: /* This shouldn't happen but some implementation might still have this
1792: problem. Use a reasonable default value. */
1793: # define CHAR_CLASS_MAX_LENGTH 256
1794: # endif
1795:
1796: # ifdef _LIBC
1797: # define IS_CHAR_CLASS(string) __wctype (string)
1798: # else
1799: # define IS_CHAR_CLASS(string) wctype (string)
1800: # endif
1801: #else
1802: # define CHAR_CLASS_MAX_LENGTH 6 /* Namely, `xdigit'. */
1803:
1804: # define IS_CHAR_CLASS(string) \
1805: (STREQ (string, "alpha") || STREQ (string, "upper") \
1806: || STREQ (string, "lower") || STREQ (string, "digit") \
1807: || STREQ (string, "alnum") || STREQ (string, "xdigit") \
1808: || STREQ (string, "space") || STREQ (string, "print") \
1809: || STREQ (string, "punct") || STREQ (string, "graph") \
1810: || STREQ (string, "cntrl") || STREQ (string, "blank"))
1811: #endif
1812:
1813: #ifndef MATCH_MAY_ALLOCATE
1814:
1815: /* If we cannot allocate large objects within re_match_2_internal,
1816: we make the fail stack and register vectors global.
1817: The fail stack, we grow to the maximum size when a regexp
1818: is compiled.
1819: The register vectors, we adjust in size each time we
1820: compile a regexp, according to the number of registers it needs. */
1821:
1822: static fail_stack_type fail_stack;
1823:
1824: /* Size with which the following vectors are currently allocated.
1825: That is so we can make them bigger as needed,
1826: but never make them smaller. */
1827: static int regs_allocated_size;
1828:
1829: static const char ** regstart, ** regend;
1830: static const char ** old_regstart, ** old_regend;
1831: static const char **best_regstart, **best_regend;
1832: static register_info_type *reg_info;
1833: static const char **reg_dummy;
1834: static register_info_type *reg_info_dummy;
1835:
1836: /* Make the register vectors big enough for NUM_REGS registers,
1837: but don't make them smaller. */
1838:
1839: static
1840: regex_grow_registers (num_regs)
1841: int num_regs;
1842: {
1843: if (num_regs > regs_allocated_size)
1844: {
1845: RETALLOC_IF (regstart, num_regs, const char *);
1846: RETALLOC_IF (regend, num_regs, const char *);
1847: RETALLOC_IF (old_regstart, num_regs, const char *);
1848: RETALLOC_IF (old_regend, num_regs, const char *);
1849: RETALLOC_IF (best_regstart, num_regs, const char *);
1850: RETALLOC_IF (best_regend, num_regs, const char *);
1851: RETALLOC_IF (reg_info, num_regs, register_info_type);
1852: RETALLOC_IF (reg_dummy, num_regs, const char *);
1853: RETALLOC_IF (reg_info_dummy, num_regs, register_info_type);
1854:
1855: regs_allocated_size = num_regs;
1856: }
1857: }
1858:
1859: #endif /* not MATCH_MAY_ALLOCATE */
1860:
1861: static boolean group_in_compile_stack _RE_ARGS ((compile_stack_type
1862: compile_stack,
1863: regnum_t regnum));
1864:
1865: /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX.
1866: Returns one of error codes defined in `regex.h', or zero for success.
1867:
1868: Assumes the `allocated' (and perhaps `buffer') and `translate'
1869: fields are set in BUFP on entry.
1870:
1871: If it succeeds, results are put in BUFP (if it returns an error, the
1872: contents of BUFP are undefined):
1873: `buffer' is the compiled pattern;
1874: `syntax' is set to SYNTAX;
1875: `used' is set to the length of the compiled pattern;
1876: `fastmap_accurate' is zero;
1877: `re_nsub' is the number of subexpressions in PATTERN;
1878: `not_bol' and `not_eol' are zero;
1879:
1880: The `fastmap' and `newline_anchor' fields are neither
1881: examined nor set. */
1882:
1883: /* Return, freeing storage we allocated. */
1884: #define FREE_STACK_RETURN(value) \
1885: return (free (compile_stack.stack), value)
1886:
1887: static reg_errcode_t
1888: regex_compile (pattern, size, syntax, bufp)
1889: const char *pattern;
1890: size_t size;
1891: reg_syntax_t syntax;
1892: struct re_pattern_buffer *bufp;
1893: {
1894: /* We fetch characters from PATTERN here. Even though PATTERN is
1895: `char *' (i.e., signed), we declare these variables as unsigned, so
1896: they can be reliably used as array indices. */
1897: register unsigned char c, c1;
1898:
1899: /* A random temporary spot in PATTERN. */
1900: const char *p1;
1901:
1902: /* Points to the end of the buffer, where we should append. */
1903: register unsigned char *b;
1904:
1905: /* Keeps track of unclosed groups. */
1906: compile_stack_type compile_stack;
1907:
1908: /* Points to the current (ending) position in the pattern. */
1909: const char *p = pattern;
1910: const char *pend = pattern + size;
1911:
1912: /* How to translate the characters in the pattern. */
1913: RE_TRANSLATE_TYPE translate = bufp->translate;
1914:
1915: /* Address of the count-byte of the most recently inserted `exactn'
1916: command. This makes it possible to tell if a new exact-match
1917: character can be added to that command or if the character requires
1918: a new `exactn' command. */
1919: unsigned char *pending_exact = 0;
1920:
1921: /* Address of start of the most recently finished expression.
1922: This tells, e.g., postfix * where to find the start of its
1923: operand. Reset at the beginning of groups and alternatives. */
1924: unsigned char *laststart = 0;
1925:
1926: /* Address of beginning of regexp, or inside of last group. */
1927: unsigned char *begalt;
1928:
1929: /* Place in the uncompiled pattern (i.e., the {) to
1930: which to go back if the interval is invalid. */
1931: const char *beg_interval;
1932:
1933: /* Address of the place where a forward jump should go to the end of
1934: the containing expression. Each alternative of an `or' -- except the
1935: last -- ends with a forward jump of this sort. */
1936: unsigned char *fixup_alt_jump = 0;
1937:
1938: /* Counts open-groups as they are encountered. Remembered for the
1939: matching close-group on the compile stack, so the same register
1940: number is put in the stop_memory as the start_memory. */
1941: regnum_t regnum = 0;
1942:
1943: #ifdef DEBUG
1944: DEBUG_PRINT1 ("\nCompiling pattern: ");
1945: if (debug)
1946: {
1947: unsigned debug_count;
1948:
1949: for (debug_count = 0; debug_count < size; debug_count++)
1950: putchar (pattern[debug_count]);
1951: putchar ('\n');
1952: }
1953: #endif /* DEBUG */
1954:
1955: /* Initialize the compile stack. */
1956: compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t);
1957: if (compile_stack.stack == NULL)
1958: return REG_ESPACE;
1959:
1960: compile_stack.size = INIT_COMPILE_STACK_SIZE;
1961: compile_stack.avail = 0;
1962:
1963: /* Initialize the pattern buffer. */
1964: bufp->syntax = syntax;
1965: bufp->fastmap_accurate = 0;
1966: bufp->not_bol = bufp->not_eol = 0;
1967:
1968: /* Set `used' to zero, so that if we return an error, the pattern
1969: printer (for debugging) will think there's no pattern. We reset it
1970: at the end. */
1971: bufp->used = 0;
1972:
1973: /* Always count groups, whether or not bufp->no_sub is set. */
1974: bufp->re_nsub = 0;
1975:
1976: #if !defined emacs && !defined SYNTAX_TABLE
1977: /* Initialize the syntax table. */
1978: init_syntax_once ();
1979: #endif
1980:
1981: if (bufp->allocated == 0)
1982: {
1983: if (bufp->buffer)
1984: { /* If zero allocated, but buffer is non-null, try to realloc
1985: enough space. This loses if buffer's address is bogus, but
1986: that is the user's responsibility. */
1987: RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char);
1988: }
1989: else
1990: { /* Caller did not allocate a buffer. Do it for them. */
1991: bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char);
1992: }
1993: if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE);
1994:
1995: bufp->allocated = INIT_BUF_SIZE;
1996: }
1997:
1998: begalt = b = bufp->buffer;
1999:
2000: /* Loop through the uncompiled pattern until we're at the end. */
2001: while (p != pend)
2002: {
2003: PATFETCH (c);
2004:
2005: switch (c)
2006: {
2007: case '^':
2008: {
2009: if ( /* If at start of pattern, it's an operator. */
2010: p == pattern + 1
2011: /* If context independent, it's an operator. */
2012: || syntax & RE_CONTEXT_INDEP_ANCHORS
2013: /* Otherwise, depends on what's come before. */
2014: || at_begline_loc_p (pattern, p, syntax))
2015: BUF_PUSH (begline);
2016: else
2017: goto normal_char;
2018: }
2019: break;
2020:
2021:
2022: case '$':
2023: {
2024: if ( /* If at end of pattern, it's an operator. */
2025: p == pend
2026: /* If context independent, it's an operator. */
2027: || syntax & RE_CONTEXT_INDEP_ANCHORS
2028: /* Otherwise, depends on what's next. */
2029: || at_endline_loc_p (p, pend, syntax))
2030: BUF_PUSH (endline);
2031: else
2032: goto normal_char;
2033: }
2034: break;
2035:
2036:
2037: case '+':
2038: case '?':
2039: if ((syntax & RE_BK_PLUS_QM)
2040: || (syntax & RE_LIMITED_OPS))
2041: goto normal_char;
2042: handle_plus:
2043: case '*':
2044: /* If there is no previous pattern... */
2045: if (!laststart)
2046: {
2047: if (syntax & RE_CONTEXT_INVALID_OPS)
2048: FREE_STACK_RETURN (REG_BADRPT);
2049: else if (!(syntax & RE_CONTEXT_INDEP_OPS))
2050: goto normal_char;
2051: }
2052:
2053: {
2054: /* Are we optimizing this jump? */
2055: boolean keep_string_p = false;
2056:
2057: /* 1 means zero (many) matches is allowed. */
2058: char zero_times_ok = 0, many_times_ok = 0;
2059:
2060: /* If there is a sequence of repetition chars, collapse it
2061: down to just one (the right one). We can't combine
2062: interval operators with these because of, e.g., `a{2}*',
2063: which should only match an even number of `a's. */
2064:
2065: for (;;)
2066: {
2067: zero_times_ok |= c != '+';
2068: many_times_ok |= c != '?';
2069:
2070: if (p == pend)
2071: break;
2072:
2073: PATFETCH (c);
2074:
2075: if (c == '*'
2076: || (!(syntax & RE_BK_PLUS_QM) && (c == '+' || c == '?')))
2077: ;
2078:
2079: else if (syntax & RE_BK_PLUS_QM && c == '\\')
2080: {
2081: if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
2082:
2083: PATFETCH (c1);
2084: if (!(c1 == '+' || c1 == '?'))
2085: {
2086: PATUNFETCH;
2087: PATUNFETCH;
2088: break;
2089: }
2090:
2091: c = c1;
2092: }
2093: else
2094: {
2095: PATUNFETCH;
2096: break;
2097: }
2098:
2099: /* If we get here, we found another repeat character. */
2100: }
2101:
2102: /* Star, etc. applied to an empty pattern is equivalent
2103: to an empty pattern. */
2104: if (!laststart)
2105: break;
2106:
2107: /* Now we know whether or not zero matches is allowed
2108: and also whether or not two or more matches is allowed. */
2109: if (many_times_ok)
2110: { /* More than one repetition is allowed, so put in at the
2111: end a backward relative jump from `b' to before the next
2112: jump we're going to put in below (which jumps from
2113: laststart to after this jump).
2114:
2115: But if we are at the `*' in the exact sequence `.*\n',
2116: insert an unconditional jump backwards to the .,
2117: instead of the beginning of the loop. This way we only
2118: push a failure point once, instead of every time
2119: through the loop. */
2120: assert (p - 1 > pattern);
2121:
2122: /* Allocate the space for the jump. */
2123: GET_BUFFER_SPACE (3);
2124:
2125: /* We know we are not at the first character of the pattern,
2126: because laststart was nonzero. And we've already
2127: incremented `p', by the way, to be the character after
2128: the `*'. Do we have to do something analogous here
2129: for null bytes, because of RE_DOT_NOT_NULL? */
2130: if (TRANSLATE (*(p - 2)) == TRANSLATE ('.')
2131: && zero_times_ok
2132: && p < pend && TRANSLATE (*p) == TRANSLATE ('\n')
2133: && !(syntax & RE_DOT_NEWLINE))
2134: { /* We have .*\n. */
2135: STORE_JUMP (jump, b, laststart);
2136: keep_string_p = true;
2137: }
2138: else
2139: /* Anything else. */
2140: STORE_JUMP (maybe_pop_jump, b, laststart - 3);
2141:
2142: /* We've added more stuff to the buffer. */
2143: b += 3;
2144: }
2145:
2146: /* On failure, jump from laststart to b + 3, which will be the
2147: end of the buffer after this jump is inserted. */
2148: GET_BUFFER_SPACE (3);
2149: INSERT_JUMP (keep_string_p ? on_failure_keep_string_jump
2150: : on_failure_jump,
2151: laststart, b + 3);
2152: pending_exact = 0;
2153: b += 3;
2154:
2155: if (!zero_times_ok)
2156: {
2157: /* At least one repetition is required, so insert a
2158: `dummy_failure_jump' before the initial
2159: `on_failure_jump' instruction of the loop. This
2160: effects a skip over that instruction the first time
2161: we hit that loop. */
2162: GET_BUFFER_SPACE (3);
2163: INSERT_JUMP (dummy_failure_jump, laststart, laststart + 6);
2164: b += 3;
2165: }
2166: }
2167: break;
2168:
2169:
2170: case '.':
2171: laststart = b;
2172: BUF_PUSH (anychar);
2173: break;
2174:
2175:
2176: case '[':
2177: {
2178: boolean had_char_class = false;
2179:
2180: if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
2181:
2182: /* Ensure that we have enough space to push a charset: the
2183: opcode, the length count, and the bitset; 34 bytes in all. */
2184: GET_BUFFER_SPACE (34);
2185:
2186: laststart = b;
2187:
2188: /* We test `*p == '^' twice, instead of using an if
2189: statement, so we only need one BUF_PUSH. */
2190: BUF_PUSH (*p == '^' ? charset_not : charset);
2191: if (*p == '^')
2192: p++;
2193:
2194: /* Remember the first position in the bracket expression. */
2195: p1 = p;
2196:
2197: /* Push the number of bytes in the bitmap. */
2198: BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH);
2199:
2200: /* Clear the whole map. */
2201: memset (b, 0, (1 << BYTEWIDTH) / BYTEWIDTH);
2202:
2203: /* charset_not matches newline according to a syntax bit. */
2204: if ((re_opcode_t) b[-2] == charset_not
2205: && (syntax & RE_HAT_LISTS_NOT_NEWLINE))
2206: SET_LIST_BIT ('\n');
2207:
2208: /* Read in characters and ranges, setting map bits. */
2209: for (;;)
2210: {
2211: if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
2212:
2213: PATFETCH (c);
2214:
2215: /* \ might escape characters inside [...] and [^...]. */
2216: if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\')
2217: {
2218: if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
2219:
2220: PATFETCH (c1);
2221: SET_LIST_BIT (c1);
2222: continue;
2223: }
2224:
2225: /* Could be the end of the bracket expression. If it's
2226: not (i.e., when the bracket expression is `[]' so
2227: far), the ']' character bit gets set way below. */
2228: if (c == ']' && p != p1 + 1)
2229: break;
2230:
2231: /* Look ahead to see if it's a range when the last thing
2232: was a character class. */
2233: if (had_char_class && c == '-' && *p != ']')
2234: FREE_STACK_RETURN (REG_ERANGE);
2235:
2236: /* Look ahead to see if it's a range when the last thing
2237: was a character: if this is a hyphen not at the
2238: beginning or the end of a list, then it's the range
2239: operator. */
2240: if (c == '-'
2241: && !(p - 2 >= pattern && p[-2] == '[')
2242: && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^')
2243: && *p != ']')
2244: {
2245: reg_errcode_t ret
2246: = compile_range (&p, pend, translate, syntax, b);
2247: if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
2248: }
2249:
2250: else if (p[0] == '-' && p[1] != ']')
2251: { /* This handles ranges made up of characters only. */
2252: reg_errcode_t ret;
2253:
2254: /* Move past the `-'. */
2255: PATFETCH (c1);
2256:
2257: ret = compile_range (&p, pend, translate, syntax, b);
2258: if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
2259: }
2260:
2261: /* See if we're at the beginning of a possible character
2262: class. */
2263:
2264: else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':')
2265: { /* Leave room for the null. */
2266: char str[CHAR_CLASS_MAX_LENGTH + 1];
2267:
2268: PATFETCH (c);
2269: c1 = 0;
2270:
2271: /* If pattern is `[[:'. */
2272: if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
2273:
2274: for (;;)
2275: {
2276: PATFETCH (c);
2277: if ((c == ':' && *p == ']') || p == pend)
2278: break;
2279: if (c1 < CHAR_CLASS_MAX_LENGTH)
2280: str[c1++] = c;
2281: else
2282: /* This is in any case an invalid class name. */
2283: str[0] = '\0';
2284: }
2285: str[c1] = '\0';
2286:
2287: /* If isn't a word bracketed by `[:' and `:]':
2288: undo the ending character, the letters, and leave
2289: the leading `:' and `[' (but set bits for them). */
2290: if (c == ':' && *p == ']')
2291: {
2292: #if defined _LIBC || WIDE_CHAR_SUPPORT
2293: boolean is_lower = STREQ (str, "lower");
2294: boolean is_upper = STREQ (str, "upper");
2295: wctype_t wt;
2296: int ch;
2297:
2298: wt = IS_CHAR_CLASS (str);
2299: if (wt == 0)
2300: FREE_STACK_RETURN (REG_ECTYPE);
2301:
2302: /* Throw away the ] at the end of the character
2303: class. */
2304: PATFETCH (c);
2305:
2306: if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
2307:
2308: for (ch = 0; ch < 1 << BYTEWIDTH; ++ch)
2309: {
2310: # ifdef _LIBC
2311: if (__iswctype (__btowc (ch), wt))
2312: SET_LIST_BIT (ch);
2313: # else
2314: if (iswctype (btowc (ch), wt))
2315: SET_LIST_BIT (ch);
2316: # endif
2317:
2318: if (translate && (is_upper || is_lower)
2319: && (ISUPPER (ch) || ISLOWER (ch)))
2320: SET_LIST_BIT (ch);
2321: }
2322:
2323: had_char_class = true;
2324: #else
2325: int ch;
2326: boolean is_alnum = STREQ (str, "alnum");
2327: boolean is_alpha = STREQ (str, "alpha");
2328: boolean is_blank = STREQ (str, "blank");
2329: boolean is_cntrl = STREQ (str, "cntrl");
2330: boolean is_digit = STREQ (str, "digit");
2331: boolean is_graph = STREQ (str, "graph");
2332: boolean is_lower = STREQ (str, "lower");
2333: boolean is_print = STREQ (str, "print");
2334: boolean is_punct = STREQ (str, "punct");
2335: boolean is_space = STREQ (str, "space");
2336: boolean is_upper = STREQ (str, "upper");
2337: boolean is_xdigit = STREQ (str, "xdigit");
2338:
2339: if (!IS_CHAR_CLASS (str))
2340: FREE_STACK_RETURN (REG_ECTYPE);
2341:
2342: /* Throw away the ] at the end of the character
2343: class. */
2344: PATFETCH (c);
2345:
2346: if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
2347:
2348: for (ch = 0; ch < 1 << BYTEWIDTH; ch++)
2349: {
2350: /* This was split into 3 if's to
2351: avoid an arbitrary limit in some compiler. */
2352: if ( (is_alnum && ISALNUM (ch))
2353: || (is_alpha && ISALPHA (ch))
2354: || (is_blank && ISBLANK (ch))
2355: || (is_cntrl && ISCNTRL (ch)))
2356: SET_LIST_BIT (ch);
2357: if ( (is_digit && ISDIGIT (ch))
2358: || (is_graph && ISGRAPH (ch))
2359: || (is_lower && ISLOWER (ch))
2360: || (is_print && ISPRINT (ch)))
2361: SET_LIST_BIT (ch);
2362: if ( (is_punct && ISPUNCT (ch))
2363: || (is_space && ISSPACE (ch))
2364: || (is_upper && ISUPPER (ch))
2365: || (is_xdigit && ISXDIGIT (ch)))
2366: SET_LIST_BIT (ch);
2367: if ( translate && (is_upper || is_lower)
2368: && (ISUPPER (ch) || ISLOWER (ch)))
2369: SET_LIST_BIT (ch);
2370: }
2371: had_char_class = true;
2372: #endif /* libc || wctype.h */
2373: }
2374: else
2375: {
2376: c1++;
2377: while (c1--)
2378: PATUNFETCH;
2379: SET_LIST_BIT ('[');
2380: SET_LIST_BIT (':');
2381: had_char_class = false;
2382: }
2383: }
2384: else
2385: {
2386: had_char_class = false;
2387: SET_LIST_BIT (c);
2388: }
2389: }
2390:
2391: /* Discard any (non)matching list bytes that are all 0 at the
2392: end of the map. Decrease the map-length byte too. */
2393: while ((int) b[-1] > 0 && b[b[-1] - 1] == 0)
2394: b[-1]--;
2395: b += b[-1];
2396: }
2397: break;
2398:
2399:
2400: case '(':
2401: if (syntax & RE_NO_BK_PARENS)
2402: goto handle_open;
2403: else
2404: goto normal_char;
2405:
2406:
2407: case ')':
2408: if (syntax & RE_NO_BK_PARENS)
2409: goto handle_close;
2410: else
2411: goto normal_char;
2412:
2413:
2414: case '\n':
2415: if (syntax & RE_NEWLINE_ALT)
2416: goto handle_alt;
2417: else
2418: goto normal_char;
2419:
2420:
2421: case '|':
2422: if (syntax & RE_NO_BK_VBAR)
2423: goto handle_alt;
2424: else
2425: goto normal_char;
2426:
2427:
2428: case '{':
2429: if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES)
2430: goto handle_interval;
2431: else
2432: goto normal_char;
2433:
2434:
2435: case '\\':
2436: if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
2437:
2438: /* Do not translate the character after the \, so that we can
2439: distinguish, e.g., \B from \b, even if we normally would
2440: translate, e.g., B to b. */
2441: PATFETCH_RAW (c);
2442:
2443: switch (c)
2444: {
2445: case '(':
2446: if (syntax & RE_NO_BK_PARENS)
2447: goto normal_backslash;
2448:
2449: handle_open:
2450: bufp->re_nsub++;
2451: regnum++;
2452:
2453: if (COMPILE_STACK_FULL)
2454: {
2455: RETALLOC (compile_stack.stack, compile_stack.size << 1,
2456: compile_stack_elt_t);
2457: if (compile_stack.stack == NULL) return REG_ESPACE;
2458:
2459: compile_stack.size <<= 1;
2460: }
2461:
2462: /* These are the values to restore when we hit end of this
2463: group. They are all relative offsets, so that if the
2464: whole pattern moves because of realloc, they will still
2465: be valid. */
2466: COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer;
2467: COMPILE_STACK_TOP.fixup_alt_jump
2468: = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0;
2469: COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer;
2470: COMPILE_STACK_TOP.regnum = regnum;
2471:
2472: /* We will eventually replace the 0 with the number of
2473: groups inner to this one. But do not push a
2474: start_memory for groups beyond the last one we can
2475: represent in the compiled pattern. */
2476: if (regnum <= MAX_REGNUM)
2477: {
2478: COMPILE_STACK_TOP.inner_group_offset = b - bufp->buffer + 2;
2479: BUF_PUSH_3 (start_memory, regnum, 0);
2480: }
2481:
2482: compile_stack.avail++;
2483:
2484: fixup_alt_jump = 0;
2485: laststart = 0;
2486: begalt = b;
2487: /* If we've reached MAX_REGNUM groups, then this open
2488: won't actually generate any code, so we'll have to
2489: clear pending_exact explicitly. */
2490: pending_exact = 0;
2491: break;
2492:
2493:
2494: case ')':
2495: if (syntax & RE_NO_BK_PARENS) goto normal_backslash;
2496:
2497: if (COMPILE_STACK_EMPTY)
2498: {
2499: if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
2500: goto normal_backslash;
2501: else
2502: FREE_STACK_RETURN (REG_ERPAREN);
2503: }
2504:
2505: handle_close:
2506: if (fixup_alt_jump)
2507: { /* Push a dummy failure point at the end of the
2508: alternative for a possible future
2509: `pop_failure_jump' to pop. See comments at
2510: `push_dummy_failure' in `re_match_2'. */
2511: BUF_PUSH (push_dummy_failure);
2512:
2513: /* We allocated space for this jump when we assigned
2514: to `fixup_alt_jump', in the `handle_alt' case below. */
2515: STORE_JUMP (jump_past_alt, fixup_alt_jump, b - 1);
2516: }
2517:
2518: /* See similar code for backslashed left paren above. */
2519: if (COMPILE_STACK_EMPTY)
2520: {
2521: if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
2522: goto normal_char;
2523: else
2524: FREE_STACK_RETURN (REG_ERPAREN);
2525: }
2526:
2527: /* Since we just checked for an empty stack above, this
2528: ``can't happen''. */
2529: assert (compile_stack.avail != 0);
2530: {
2531: /* We don't just want to restore into `regnum', because
2532: later groups should continue to be numbered higher,
2533: as in `(ab)c(de)' -- the second group is #2. */
2534: regnum_t this_group_regnum;
2535:
2536: compile_stack.avail--;
2537: begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset;
2538: fixup_alt_jump
2539: = COMPILE_STACK_TOP.fixup_alt_jump
2540: ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1
2541: : 0;
2542: laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset;
2543: this_group_regnum = COMPILE_STACK_TOP.regnum;
2544: /* If we've reached MAX_REGNUM groups, then this open
2545: won't actually generate any code, so we'll have to
2546: clear pending_exact explicitly. */
2547: pending_exact = 0;
2548:
2549: /* We're at the end of the group, so now we know how many
2550: groups were inside this one. */
2551: if (this_group_regnum <= MAX_REGNUM)
2552: {
2553: unsigned char *inner_group_loc
2554: = bufp->buffer + COMPILE_STACK_TOP.inner_group_offset;
2555:
2556: *inner_group_loc = regnum - this_group_regnum;
2557: BUF_PUSH_3 (stop_memory, this_group_regnum,
2558: regnum - this_group_regnum);
2559: }
2560: }
2561: break;
2562:
2563:
2564: case '|': /* `\|'. */
2565: if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR)
2566: goto normal_backslash;
2567: handle_alt:
2568: if (syntax & RE_LIMITED_OPS)
2569: goto normal_char;
2570:
2571: /* Insert before the previous alternative a jump which
2572: jumps to this alternative if the former fails. */
2573: GET_BUFFER_SPACE (3);
2574: INSERT_JUMP (on_failure_jump, begalt, b + 6);
2575: pending_exact = 0;
2576: b += 3;
2577:
2578: /* The alternative before this one has a jump after it
2579: which gets executed if it gets matched. Adjust that
2580: jump so it will jump to this alternative's analogous
2581: jump (put in below, which in turn will jump to the next
2582: (if any) alternative's such jump, etc.). The last such
2583: jump jumps to the correct final destination. A picture:
2584: _____ _____
2585: | | | |
2586: | v | v
2587: a | b | c
2588:
2589: If we are at `b', then fixup_alt_jump right now points to a
2590: three-byte space after `a'. We'll put in the jump, set
2591: fixup_alt_jump to right after `b', and leave behind three
2592: bytes which we'll fill in when we get to after `c'. */
2593:
2594: if (fixup_alt_jump)
2595: STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
2596:
2597: /* Mark and leave space for a jump after this alternative,
2598: to be filled in later either by next alternative or
2599: when know we're at the end of a series of alternatives. */
2600: fixup_alt_jump = b;
2601: GET_BUFFER_SPACE (3);
2602: b += 3;
2603:
2604: laststart = 0;
2605: begalt = b;
2606: break;
2607:
2608:
2609: case '{':
2610: /* If \{ is a literal. */
2611: if (!(syntax & RE_INTERVALS)
2612: /* If we're at `\{' and it's not the open-interval
2613: operator. */
2614: || ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES))
2615: || (p - 2 == pattern && p == pend))
2616: goto normal_backslash;
2617:
2618: handle_interval:
2619: {
2620: /* If got here, then the syntax allows intervals. */
2621:
2622: /* At least (most) this many matches must be made. */
2623: int lower_bound = -1, upper_bound = -1;
2624:
2625: beg_interval = p - 1;
2626:
2627: if (p == pend)
2628: {
2629: if (syntax & RE_NO_BK_BRACES)
2630: goto unfetch_interval;
2631: else
2632: FREE_STACK_RETURN (REG_EBRACE);
2633: }
2634:
2635: GET_UNSIGNED_NUMBER (lower_bound);
2636:
2637: if (c == ',')
2638: {
2639: GET_UNSIGNED_NUMBER (upper_bound);
2640: if (upper_bound < 0) upper_bound = RE_DUP_MAX;
2641: }
2642: else
2643: /* Interval such as `{1}' => match exactly once. */
2644: upper_bound = lower_bound;
2645:
2646: if (lower_bound < 0 || upper_bound > RE_DUP_MAX
2647: || lower_bound > upper_bound)
2648: {
2649: if (syntax & RE_NO_BK_BRACES)
2650: goto unfetch_interval;
2651: else
2652: FREE_STACK_RETURN (REG_BADBR);
2653: }
2654:
2655: if (!(syntax & RE_NO_BK_BRACES))
2656: {
2657: if (c != '\\') FREE_STACK_RETURN (REG_EBRACE);
2658:
2659: PATFETCH (c);
2660: }
2661:
2662: if (c != '}')
2663: {
2664: if (syntax & RE_NO_BK_BRACES)
2665: goto unfetch_interval;
2666: else
2667: FREE_STACK_RETURN (REG_BADBR);
2668: }
2669:
2670: /* We just parsed a valid interval. */
2671:
2672: /* If it's invalid to have no preceding re. */
2673: if (!laststart)
2674: {
2675: if (syntax & RE_CONTEXT_INVALID_OPS)
2676: FREE_STACK_RETURN (REG_BADRPT);
2677: else if (syntax & RE_CONTEXT_INDEP_OPS)
2678: laststart = b;
2679: else
2680: goto unfetch_interval;
2681: }
2682:
2683: /* If the upper bound is zero, don't want to succeed at
2684: all; jump from `laststart' to `b + 3', which will be
2685: the end of the buffer after we insert the jump. */
2686: if (upper_bound == 0)
2687: {
2688: GET_BUFFER_SPACE (3);
2689: INSERT_JUMP (jump, laststart, b + 3);
2690: b += 3;
2691: }
2692:
2693: /* Otherwise, we have a nontrivial interval. When
2694: we're all done, the pattern will look like:
2695: set_number_at <jump count> <upper bound>
2696: set_number_at <succeed_n count> <lower bound>
2697: succeed_n <after jump addr> <succeed_n count>
2698: <body of loop>
2699: jump_n <succeed_n addr> <jump count>
2700: (The upper bound and `jump_n' are omitted if
2701: `upper_bound' is 1, though.) */
2702: else
2703: { /* If the upper bound is > 1, we need to insert
2704: more at the end of the loop. */
2705: unsigned nbytes = 10 + (upper_bound > 1) * 10;
2706:
2707: GET_BUFFER_SPACE (nbytes);
2708:
2709: /* Initialize lower bound of the `succeed_n', even
2710: though it will be set during matching by its
2711: attendant `set_number_at' (inserted next),
2712: because `re_compile_fastmap' needs to know.
2713: Jump to the `jump_n' we might insert below. */
2714: INSERT_JUMP2 (succeed_n, laststart,
2715: b + 5 + (upper_bound > 1) * 5,
2716: lower_bound);
2717: b += 5;
2718:
2719: /* Code to initialize the lower bound. Insert
2720: before the `succeed_n'. The `5' is the last two
2721: bytes of this `set_number_at', plus 3 bytes of
2722: the following `succeed_n'. */
2723: insert_op2 (set_number_at, laststart, 5, lower_bound, b);
2724: b += 5;
2725:
2726: if (upper_bound > 1)
2727: { /* More than one repetition is allowed, so
2728: append a backward jump to the `succeed_n'
2729: that starts this interval.
2730:
2731: When we've reached this during matching,
2732: we'll have matched the interval once, so
2733: jump back only `upper_bound - 1' times. */
2734: STORE_JUMP2 (jump_n, b, laststart + 5,
2735: upper_bound - 1);
2736: b += 5;
2737:
2738: /* The location we want to set is the second
2739: parameter of the `jump_n'; that is `b-2' as
2740: an absolute address. `laststart' will be
2741: the `set_number_at' we're about to insert;
2742: `laststart+3' the number to set, the source
2743: for the relative address. But we are
2744: inserting into the middle of the pattern --
2745: so everything is getting moved up by 5.
2746: Conclusion: (b - 2) - (laststart + 3) + 5,
2747: i.e., b - laststart.
2748:
2749: We insert this at the beginning of the loop
2750: so that if we fail during matching, we'll
2751: reinitialize the bounds. */
2752: insert_op2 (set_number_at, laststart, b - laststart,
2753: upper_bound - 1, b);
2754: b += 5;
2755: }
2756: }
2757: pending_exact = 0;
2758: beg_interval = NULL;
2759: }
2760: break;
2761:
2762: unfetch_interval:
2763: /* If an invalid interval, match the characters as literals. */
2764: assert (beg_interval);
2765: p = beg_interval;
2766: beg_interval = NULL;
2767:
2768: /* normal_char and normal_backslash need `c'. */
2769: PATFETCH (c);
2770:
2771: if (!(syntax & RE_NO_BK_BRACES))
2772: {
2773: if (p > pattern && p[-1] == '\\')
2774: goto normal_backslash;
2775: }
2776: goto normal_char;
2777:
2778: #ifdef emacs
2779: /* There is no way to specify the before_dot and after_dot
2780: operators. rms says this is ok. --karl */
2781: case '=':
2782: BUF_PUSH (at_dot);
2783: break;
2784:
2785: case 's':
2786: laststart = b;
2787: PATFETCH (c);
2788: BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]);
2789: break;
2790:
2791: case 'S':
2792: laststart = b;
2793: PATFETCH (c);
2794: BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]);
2795: break;
2796: #endif /* emacs */
2797:
2798:
2799: case 'w':
2800: if (syntax & RE_NO_GNU_OPS)
2801: goto normal_char;
2802: laststart = b;
2803: BUF_PUSH (wordchar);
2804: break;
2805:
2806:
2807: case 'W':
2808: if (syntax & RE_NO_GNU_OPS)
2809: goto normal_char;
2810: laststart = b;
2811: BUF_PUSH (notwordchar);
2812: break;
2813:
2814:
2815: case '<':
2816: if (syntax & RE_NO_GNU_OPS)
2817: goto normal_char;
2818: BUF_PUSH (wordbeg);
2819: break;
2820:
2821: case '>':
2822: if (syntax & RE_NO_GNU_OPS)
2823: goto normal_char;
2824: BUF_PUSH (wordend);
2825: break;
2826:
2827: case 'b':
2828: if (syntax & RE_NO_GNU_OPS)
2829: goto normal_char;
2830: BUF_PUSH (wordbound);
2831: break;
2832:
2833: case 'B':
2834: if (syntax & RE_NO_GNU_OPS)
2835: goto normal_char;
2836: BUF_PUSH (notwordbound);
2837: break;
2838:
2839: case '`':
2840: if (syntax & RE_NO_GNU_OPS)
2841: goto normal_char;
2842: BUF_PUSH (begbuf);
2843: break;
2844:
2845: case '\'':
2846: if (syntax & RE_NO_GNU_OPS)
2847: goto normal_char;
2848: BUF_PUSH (endbuf);
2849: break;
2850:
2851: case '1': case '2': case '3': case '4': case '5':
2852: case '6': case '7': case '8': case '9':
2853: if (syntax & RE_NO_BK_REFS)
2854: goto normal_char;
2855:
2856: c1 = c - '0';
2857:
2858: if (c1 > regnum)
2859: FREE_STACK_RETURN (REG_ESUBREG);
2860:
2861: /* Can't back reference to a subexpression if inside of it. */
2862: if (group_in_compile_stack (compile_stack, (regnum_t) c1))
2863: goto normal_char;
2864:
2865: laststart = b;
2866: BUF_PUSH_2 (duplicate, c1);
2867: break;
2868:
2869:
2870: case '+':
2871: case '?':
2872: if (syntax & RE_BK_PLUS_QM)
2873: goto handle_plus;
2874: else
2875: goto normal_backslash;
2876:
2877: default:
2878: normal_backslash:
2879: /* You might think it would be useful for \ to mean
2880: not to translate; but if we don't translate it
2881: it will never match anything. */
2882: c = TRANSLATE (c);
2883: goto normal_char;
2884: }
2885: break;
2886:
2887:
2888: default:
2889: /* Expects the character in `c'. */
2890: normal_char:
2891: /* If no exactn currently being built. */
2892: if (!pending_exact
2893:
2894: /* If last exactn not at current position. */
2895: || pending_exact + *pending_exact + 1 != b
2896:
2897: /* We have only one byte following the exactn for the count. */
2898: || *pending_exact == (1 << BYTEWIDTH) - 1
2899:
2900: /* If followed by a repetition operator. */
2901: || *p == '*' || *p == '^'
2902: || ((syntax & RE_BK_PLUS_QM)
2903: ? *p == '\\' && (p[1] == '+' || p[1] == '?')
2904: : (*p == '+' || *p == '?'))
2905: || ((syntax & RE_INTERVALS)
2906: && ((syntax & RE_NO_BK_BRACES)
2907: ? *p == '{'
2908: : (p[0] == '\\' && p[1] == '{'))))
2909: {
2910: /* Start building a new exactn. */
2911:
2912: laststart = b;
2913:
2914: BUF_PUSH_2 (exactn, 0);
2915: pending_exact = b - 1;
2916: }
2917:
2918: BUF_PUSH (c);
2919: (*pending_exact)++;
2920: break;
2921: } /* switch (c) */
2922: } /* while p != pend */
2923:
2924:
2925: /* Through the pattern now. */
2926:
2927: if (fixup_alt_jump)
2928: STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
2929:
2930: if (!COMPILE_STACK_EMPTY)
2931: FREE_STACK_RETURN (REG_EPAREN);
2932:
2933: /* If we don't want backtracking, force success
2934: the first time we reach the end of the compiled pattern. */
2935: if (syntax & RE_NO_POSIX_BACKTRACKING)
2936: BUF_PUSH (succeed);
2937:
2938: free (compile_stack.stack);
2939:
2940: /* We have succeeded; set the length of the buffer. */
2941: bufp->used = b - bufp->buffer;
2942:
2943: #ifdef DEBUG
2944: if (debug)
2945: {
2946: DEBUG_PRINT1 ("\nCompiled pattern: \n");
2947: print_compiled_pattern (bufp);
2948: }
2949: #endif /* DEBUG */
2950:
2951: #ifndef MATCH_MAY_ALLOCATE
2952: /* Initialize the failure stack to the largest possible stack. This
2953: isn't necessary unless we're trying to avoid calling alloca in
2954: the search and match routines. */
2955: {
2956: int num_regs = bufp->re_nsub + 1;
2957:
2958: /* Since DOUBLE_FAIL_STACK refuses to double only if the current size
2959: is strictly greater than re_max_failures, the largest possible stack
2960: is 2 * re_max_failures failure points. */
2961: if (fail_stack.size < (2 * re_max_failures * MAX_FAILURE_ITEMS))
2962: {
2963: fail_stack.size = (2 * re_max_failures * MAX_FAILURE_ITEMS);
2964:
2965: # ifdef emacs
2966: if (! fail_stack.stack)
2967: fail_stack.stack
2968: = (fail_stack_elt_t *) xmalloc (fail_stack.size
2969: * sizeof (fail_stack_elt_t));
2970: else
2971: fail_stack.stack
2972: = (fail_stack_elt_t *) xrealloc (fail_stack.stack,
2973: (fail_stack.size
2974: * sizeof (fail_stack_elt_t)));
2975: # else /* not emacs */
2976: if (! fail_stack.stack)
2977: fail_stack.stack
2978: = (fail_stack_elt_t *) malloc (fail_stack.size
2979: * sizeof (fail_stack_elt_t));
2980: else
2981: fail_stack.stack
2982: = (fail_stack_elt_t *) realloc (fail_stack.stack,
2983: (fail_stack.size
2984: * sizeof (fail_stack_elt_t)));
2985: # endif /* not emacs */
2986: }
2987:
2988: regex_grow_registers (num_regs);
2989: }
2990: #endif /* not MATCH_MAY_ALLOCATE */
2991:
2992: return REG_NOERROR;
2993: } /* regex_compile */
2994:
2995: /* Subroutines for `regex_compile'. */
2996:
2997: /* Store OP at LOC followed by two-byte integer parameter ARG. */
2998:
2999: static void
3000: store_op1 (op, loc, arg)
3001: re_opcode_t op;
3002: unsigned char *loc;
3003: int arg;
3004: {
3005: *loc = (unsigned char) op;
3006: STORE_NUMBER (loc + 1, arg);
3007: }
3008:
3009:
3010: /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */
3011:
3012: static void
3013: store_op2 (op, loc, arg1, arg2)
3014: re_opcode_t op;
3015: unsigned char *loc;
3016: int arg1, arg2;
3017: {
3018: *loc = (unsigned char) op;
3019: STORE_NUMBER (loc + 1, arg1);
3020: STORE_NUMBER (loc + 3, arg2);
3021: }
3022:
3023:
3024: /* Copy the bytes from LOC to END to open up three bytes of space at LOC
3025: for OP followed by two-byte integer parameter ARG. */
3026:
3027: static void
3028: insert_op1 (op, loc, arg, end)
3029: re_opcode_t op;
3030: unsigned char *loc;
3031: int arg;
3032: unsigned char *end;
3033: {
3034: register unsigned char *pfrom = end;
3035: register unsigned char *pto = end + 3;
3036:
3037: while (pfrom != loc)
3038: *--pto = *--pfrom;
3039:
3040: store_op1 (op, loc, arg);
3041: }
3042:
3043:
3044: /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */
3045:
3046: static void
3047: insert_op2 (op, loc, arg1, arg2, end)
3048: re_opcode_t op;
3049: unsigned char *loc;
3050: int arg1, arg2;
3051: unsigned char *end;
3052: {
3053: register unsigned char *pfrom = end;
3054: register unsigned char *pto = end + 5;
3055:
3056: while (pfrom != loc)
3057: *--pto = *--pfrom;
3058:
3059: store_op2 (op, loc, arg1, arg2);
3060: }
3061:
3062:
3063: /* P points to just after a ^ in PATTERN. Return true if that ^ comes
3064: after an alternative or a begin-subexpression. We assume there is at
3065: least one character before the ^. */
3066:
3067: static boolean
3068: at_begline_loc_p (pattern, p, syntax)
3069: const char *pattern, *p;
3070: reg_syntax_t syntax;
3071: {
3072: const char *prev = p - 2;
3073: boolean prev_prev_backslash = prev > pattern && prev[-1] == '\\';
3074:
3075: return
3076: /* After a subexpression? */
3077: (*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash))
3078: /* After an alternative? */
3079: || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash));
3080: }
3081:
3082:
3083: /* The dual of at_begline_loc_p. This one is for $. We assume there is
3084: at least one character after the $, i.e., `P < PEND'. */
3085:
3086: static boolean
3087: at_endline_loc_p (p, pend, syntax)
3088: const char *p, *pend;
3089: reg_syntax_t syntax;
3090: {
3091: const char *next = p;
3092: boolean next_backslash = *next == '\\';
3093: const char *next_next = p + 1 < pend ? p + 1 : 0;
3094:
3095: return
3096: /* Before a subexpression? */
3097: (syntax & RE_NO_BK_PARENS ? *next == ')'
3098: : next_backslash && next_next && *next_next == ')')
3099: /* Before an alternative? */
3100: || (syntax & RE_NO_BK_VBAR ? *next == '|'
3101: : next_backslash && next_next && *next_next == '|');
3102: }
3103:
3104:
3105: /* Returns true if REGNUM is in one of COMPILE_STACK's elements and
3106: false if it's not. */
3107:
3108: static boolean
3109: group_in_compile_stack (compile_stack, regnum)
3110: compile_stack_type compile_stack;
3111: regnum_t regnum;
3112: {
3113: int this_element;
3114:
3115: for (this_element = compile_stack.avail - 1;
3116: this_element >= 0;
3117: this_element--)
3118: if (compile_stack.stack[this_element].regnum == regnum)
3119: return true;
3120:
3121: return false;
3122: }
3123:
3124:
3125: /* Read the ending character of a range (in a bracket expression) from the
3126: uncompiled pattern *P_PTR (which ends at PEND). We assume the
3127: starting character is in `P[-2]'. (`P[-1]' is the character `-'.)
3128: Then we set the translation of all bits between the starting and
3129: ending characters (inclusive) in the compiled pattern B.
3130:
3131: Return an error code.
3132:
3133: We use these short variable names so we can use the same macros as
3134: `regex_compile' itself. */
3135:
3136: static reg_errcode_t
3137: compile_range (p_ptr, pend, translate, syntax, b)
3138: const char **p_ptr, *pend;
3139: RE_TRANSLATE_TYPE translate;
3140: reg_syntax_t syntax;
3141: unsigned char *b;
3142: {
3143: unsigned this_char;
3144:
3145: const char *p = *p_ptr;
3146: unsigned int range_start, range_end;
3147:
3148: if (p == pend)
3149: return REG_ERANGE;
3150:
3151: /* Even though the pattern is a signed `char *', we need to fetch
3152: with unsigned char *'s; if the high bit of the pattern character
3153: is set, the range endpoints will be negative if we fetch using a
3154: signed char *.
3155:
3156: We also want to fetch the endpoints without translating them; the
3157: appropriate translation is done in the bit-setting loop below. */
3158: /* The SVR4 compiler on the 3B2 had trouble with unsigned const char *. */
3159: range_start = ((const unsigned char *) p)[-2];
3160: range_end = ((const unsigned char *) p)[0];
3161:
3162: /* Have to increment the pointer into the pattern string, so the
3163: caller isn't still at the ending character. */
3164: (*p_ptr)++;
3165:
3166: /* If the start is after the end, the range is empty. */
3167: if (range_start > range_end)
3168: return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR;
3169:
3170: /* Here we see why `this_char' has to be larger than an `unsigned
3171: char' -- the range is inclusive, so if `range_end' == 0xff
3172: (assuming 8-bit characters), we would otherwise go into an infinite
3173: loop, since all characters <= 0xff. */
3174: for (this_char = range_start; this_char <= range_end; this_char++)
3175: {
3176: SET_LIST_BIT (TRANSLATE (this_char));
3177: }
3178:
3179: return REG_NOERROR;
3180: }
3181:
3182: /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in
3183: BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible
3184: characters can start a string that matches the pattern. This fastmap
3185: is used by re_search to skip quickly over impossible starting points.
3186:
3187: The caller must supply the address of a (1 << BYTEWIDTH)-byte data
3188: area as BUFP->fastmap.
3189:
3190: We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in
3191: the pattern buffer.
3192:
3193: Returns 0 if we succeed, -2 if an internal error. */
3194:
3195: int
3196: re_compile_fastmap (bufp)
3197: struct re_pattern_buffer *bufp;
3198: {
3199: int j, k;
3200: #ifdef MATCH_MAY_ALLOCATE
3201: fail_stack_type fail_stack;
3202: #endif
3203: #ifndef REGEX_MALLOC
3204: char *destination;
3205: #endif
3206:
3207: register char *fastmap = bufp->fastmap;
3208: unsigned char *pattern = bufp->buffer;
3209: unsigned char *p = pattern;
3210: register unsigned char *pend = pattern + bufp->used;
3211:
3212: #ifdef REL_ALLOC
3213: /* This holds the pointer to the failure stack, when
3214: it is allocated relocatably. */
3215: fail_stack_elt_t *failure_stack_ptr;
3216: #endif
3217:
3218: /* Assume that each path through the pattern can be null until
3219: proven otherwise. We set this false at the bottom of switch
3220: statement, to which we get only if a particular path doesn't
3221: match the empty string. */
3222: boolean path_can_be_null = true;
3223:
3224: /* We aren't doing a `succeed_n' to begin with. */
3225: boolean succeed_n_p = false;
3226:
3227: assert (fastmap != NULL && p != NULL);
3228:
3229: INIT_FAIL_STACK ();
3230: memset (fastmap, 0, 1 << BYTEWIDTH); /* Assume nothing's valid. */
3231: bufp->fastmap_accurate = 1; /* It will be when we're done. */
3232: bufp->can_be_null = 0;
3233:
3234: while (1)
3235: {
3236: if (p == pend || *p == succeed)
3237: {
3238: /* We have reached the (effective) end of pattern. */
3239: if (!FAIL_STACK_EMPTY ())
3240: {
3241: bufp->can_be_null |= path_can_be_null;
3242:
3243: /* Reset for next path. */
3244: path_can_be_null = true;
3245:
3246: p = fail_stack.stack[--fail_stack.avail].pointer;
3247:
3248: continue;
3249: }
3250: else
3251: break;
3252: }
3253:
3254: /* We should never be about to go beyond the end of the pattern. */
3255: assert (p < pend);
3256:
3257: switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++))
3258: {
3259:
3260: /* I guess the idea here is to simply not bother with a fastmap
3261: if a backreference is used, since it's too hard to figure out
3262: the fastmap for the corresponding group. Setting
3263: `can_be_null' stops `re_search_2' from using the fastmap, so
3264: that is all we do. */
3265: case duplicate:
3266: bufp->can_be_null = 1;
3267: goto done;
3268:
3269:
3270: /* Following are the cases which match a character. These end
3271: with `break'. */
3272:
3273: case exactn:
3274: fastmap[p[1]] = 1;
3275: break;
3276:
3277:
3278: case charset:
3279: for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--)
3280: if (p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH)))
3281: fastmap[j] = 1;
3282: break;
3283:
3284:
3285: case charset_not:
3286: /* Chars beyond end of map must be allowed. */
3287: for (j = *p * BYTEWIDTH; j < (1 << BYTEWIDTH); j++)
3288: fastmap[j] = 1;
3289:
3290: for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--)
3291: if (!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))))
3292: fastmap[j] = 1;
3293: break;
3294:
3295:
3296: case wordchar:
3297: for (j = 0; j < (1 << BYTEWIDTH); j++)
3298: if (SYNTAX (j) == Sword)
3299: fastmap[j] = 1;
3300: break;
3301:
3302:
3303: case notwordchar:
3304: for (j = 0; j < (1 << BYTEWIDTH); j++)
3305: if (SYNTAX (j) != Sword)
3306: fastmap[j] = 1;
3307: break;
3308:
3309:
3310: case anychar:
3311: {
3312: int fastmap_newline = fastmap['\n'];
3313:
3314: /* `.' matches anything ... */
3315: for (j = 0; j < (1 << BYTEWIDTH); j++)
3316: fastmap[j] = 1;
3317:
3318: /* ... except perhaps newline. */
3319: if (!(bufp->syntax & RE_DOT_NEWLINE))
3320: fastmap['\n'] = fastmap_newline;
3321:
3322: /* Return if we have already set `can_be_null'; if we have,
3323: then the fastmap is irrelevant. Something's wrong here. */
3324: else if (bufp->can_be_null)
3325: goto done;
3326:
3327: /* Otherwise, have to check alternative paths. */
3328: break;
3329: }
3330:
3331: #ifdef emacs
3332: case syntaxspec:
3333: k = *p++;
3334: for (j = 0; j < (1 << BYTEWIDTH); j++)
3335: if (SYNTAX (j) == (enum syntaxcode) k)
3336: fastmap[j] = 1;
3337: break;
3338:
3339:
3340: case notsyntaxspec:
3341: k = *p++;
3342: for (j = 0; j < (1 << BYTEWIDTH); j++)
3343: if (SYNTAX (j) != (enum syntaxcode) k)
3344: fastmap[j] = 1;
3345: break;
3346:
3347:
3348: /* All cases after this match the empty string. These end with
3349: `continue'. */
3350:
3351:
3352: case before_dot:
3353: case at_dot:
3354: case after_dot:
3355: continue;
3356: #endif /* emacs */
3357:
3358:
3359: case no_op:
3360: case begline:
3361: case endline:
3362: case begbuf:
3363: case endbuf:
3364: case wordbound:
3365: case notwordbound:
3366: case wordbeg:
3367: case wordend:
3368: case push_dummy_failure:
3369: continue;
3370:
3371:
3372: case jump_n:
3373: case pop_failure_jump:
3374: case maybe_pop_jump:
3375: case jump:
3376: case jump_past_alt:
3377: case dummy_failure_jump:
3378: EXTRACT_NUMBER_AND_INCR (j, p);
3379: p += j;
3380: if (j > 0)
3381: continue;
3382:
3383: /* Jump backward implies we just went through the body of a
3384: loop and matched nothing. Opcode jumped to should be
3385: `on_failure_jump' or `succeed_n'. Just treat it like an
3386: ordinary jump. For a * loop, it has pushed its failure
3387: point already; if so, discard that as redundant. */
3388: if ((re_opcode_t) *p != on_failure_jump
3389: && (re_opcode_t) *p != succeed_n)
3390: continue;
3391:
3392: p++;
3393: EXTRACT_NUMBER_AND_INCR (j, p);
3394: p += j;
3395:
3396: /* If what's on the stack is where we are now, pop it. */
3397: if (!FAIL_STACK_EMPTY ()
3398: && fail_stack.stack[fail_stack.avail - 1].pointer == p)
3399: fail_stack.avail--;
3400:
3401: continue;
3402:
3403:
3404: case on_failure_jump:
3405: case on_failure_keep_string_jump:
3406: handle_on_failure_jump:
3407: EXTRACT_NUMBER_AND_INCR (j, p);
3408:
3409: /* For some patterns, e.g., `(a?)?', `p+j' here points to the
3410: end of the pattern. We don't want to push such a point,
3411: since when we restore it above, entering the switch will
3412: increment `p' past the end of the pattern. We don't need
3413: to push such a point since we obviously won't find any more
3414: fastmap entries beyond `pend'. Such a pattern can match
3415: the null string, though. */
3416: if (p + j < pend)
3417: {
3418: if (!PUSH_PATTERN_OP (p + j, fail_stack))
3419: {
3420: RESET_FAIL_STACK ();
3421: return -2;
3422: }
3423: }
3424: else
3425: bufp->can_be_null = 1;
3426:
3427: if (succeed_n_p)
3428: {
3429: EXTRACT_NUMBER_AND_INCR (k, p); /* Skip the n. */
3430: succeed_n_p = false;
3431: }
3432:
3433: continue;
3434:
3435:
3436: case succeed_n:
3437: /* Get to the number of times to succeed. */
3438: p += 2;
3439:
3440: /* Increment p past the n for when k != 0. */
3441: EXTRACT_NUMBER_AND_INCR (k, p);
3442: if (k == 0)
3443: {
3444: p -= 4;
3445: succeed_n_p = true; /* Spaghetti code alert. */
3446: goto handle_on_failure_jump;
3447: }
3448: continue;
3449:
3450:
3451: case set_number_at:
3452: p += 4;
3453: continue;
3454:
3455:
3456: case start_memory:
3457: case stop_memory:
3458: p += 2;
3459: continue;
3460:
3461:
3462: default:
3463: abort (); /* We have listed all the cases. */
3464: } /* switch *p++ */
3465:
3466: /* Getting here means we have found the possible starting
3467: characters for one path of the pattern -- and that the empty
3468: string does not match. We need not follow this path further.
3469: Instead, look at the next alternative (remembered on the
3470: stack), or quit if no more. The test at the top of the loop
3471: does these things. */
3472: path_can_be_null = false;
3473: p = pend;
3474: } /* while p */
3475:
3476: /* Set `can_be_null' for the last path (also the first path, if the
3477: pattern is empty). */
3478: bufp->can_be_null |= path_can_be_null;
3479:
3480: done:
3481: RESET_FAIL_STACK ();
3482: return 0;
3483: } /* re_compile_fastmap */
3484: #ifdef _LIBC
3485: weak_alias (__re_compile_fastmap, re_compile_fastmap)
3486: #endif
3487:
3488: /* Set REGS to hold NUM_REGS registers, storing them in STARTS and
3489: ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use
3490: this memory for recording register information. STARTS and ENDS
3491: must be allocated using the malloc library routine, and must each
3492: be at least NUM_REGS * sizeof (regoff_t) bytes long.
3493:
3494: If NUM_REGS == 0, then subsequent matches should allocate their own
3495: register data.
3496:
3497: Unless this function is called, the first search or match using
3498: PATTERN_BUFFER will allocate its own register data, without
3499: freeing the old data. */
3500:
3501: void
3502: re_set_registers (bufp, regs, num_regs, starts, ends)
3503: struct re_pattern_buffer *bufp;
3504: struct re_registers *regs;
3505: unsigned num_regs;
3506: regoff_t *starts, *ends;
3507: {
3508: if (num_regs)
3509: {
3510: bufp->regs_allocated = REGS_REALLOCATE;
3511: regs->num_regs = num_regs;
3512: regs->start = starts;
3513: regs->end = ends;
3514: }
3515: else
3516: {
3517: bufp->regs_allocated = REGS_UNALLOCATED;
3518: regs->num_regs = 0;
3519: regs->start = regs->end = (regoff_t *) 0;
3520: }
3521: }
3522: #ifdef _LIBC
3523: weak_alias (__re_set_registers, re_set_registers)
3524: #endif
3525:
3526: /* Searching routines. */
3527:
3528: /* Like re_search_2, below, but only one string is specified, and
3529: doesn't let you say where to stop matching. */
3530:
3531: int
3532: re_search (bufp, string, size, startpos, range, regs)
3533: struct re_pattern_buffer *bufp;
3534: const char *string;
3535: int size, startpos, range;
3536: struct re_registers *regs;
3537: {
3538: return re_search_2 (bufp, NULL, 0, string, size, startpos, range,
3539: regs, size);
3540: }
3541: #ifdef _LIBC
3542: weak_alias (__re_search, re_search)
3543: #endif
3544:
3545:
3546: /* Using the compiled pattern in BUFP->buffer, first tries to match the
3547: virtual concatenation of STRING1 and STRING2, starting first at index
3548: STARTPOS, then at STARTPOS + 1, and so on.
3549:
3550: STRING1 and STRING2 have length SIZE1 and SIZE2, respectively.
3551:
3552: RANGE is how far to scan while trying to match. RANGE = 0 means try
3553: only at STARTPOS; in general, the last start tried is STARTPOS +
3554: RANGE.
3555:
3556: In REGS, return the indices of the virtual concatenation of STRING1
3557: and STRING2 that matched the entire BUFP->buffer and its contained
3558: subexpressions.
3559:
3560: Do not consider matching one past the index STOP in the virtual
3561: concatenation of STRING1 and STRING2.
3562:
3563: We return either the position in the strings at which the match was
3564: found, -1 if no match, or -2 if error (such as failure
3565: stack overflow). */
3566:
3567: int
3568: re_search_2 (bufp, string1, size1, string2, size2, startpos, range, regs, stop)
3569: struct re_pattern_buffer *bufp;
3570: const char *string1, *string2;
3571: int size1, size2;
3572: int startpos;
3573: int range;
3574: struct re_registers *regs;
3575: int stop;
3576: {
3577: int val;
3578: register char *fastmap = bufp->fastmap;
3579: register RE_TRANSLATE_TYPE translate = bufp->translate;
3580: int total_size = size1 + size2;
3581: int endpos = startpos + range;
3582:
3583: /* Check for out-of-range STARTPOS. */
3584: if (startpos < 0 || startpos > total_size)
3585: return -1;
3586:
3587: /* Fix up RANGE if it might eventually take us outside
3588: the virtual concatenation of STRING1 and STRING2.
3589: Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */
3590: if (endpos < 0)
3591: range = 0 - startpos;
3592: else if (endpos > total_size)
3593: range = total_size - startpos;
3594:
3595: /* If the search isn't to be a backwards one, don't waste time in a
3596: search for a pattern that must be anchored. */
3597: if (bufp->used > 0 && range > 0
3598: && ((re_opcode_t) bufp->buffer[0] == begbuf
3599: /* `begline' is like `begbuf' if it cannot match at newlines. */
3600: || ((re_opcode_t) bufp->buffer[0] == begline
3601: && !bufp->newline_anchor)))
3602: {
3603: if (startpos > 0)
3604: return -1;
3605: else
3606: range = 1;
3607: }
3608:
3609: #ifdef emacs
3610: /* In a forward search for something that starts with \=.
3611: don't keep searching past point. */
3612: if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0)
3613: {
3614: range = PT - startpos;
3615: if (range <= 0)
3616: return -1;
3617: }
3618: #endif /* emacs */
3619:
3620: /* Update the fastmap now if not correct already. */
3621: if (fastmap && !bufp->fastmap_accurate)
3622: if (re_compile_fastmap (bufp) == -2)
3623: return -2;
3624:
3625: /* Loop through the string, looking for a place to start matching. */
3626: for (;;)
3627: {
3628: /* If a fastmap is supplied, skip quickly over characters that
3629: cannot be the start of a match. If the pattern can match the
3630: null string, however, we don't need to skip characters; we want
3631: the first null string. */
3632: if (fastmap && startpos < total_size && !bufp->can_be_null)
3633: {
3634: if (range > 0) /* Searching forwards. */
3635: {
3636: register const char *d;
3637: register int lim = 0;
3638: int irange = range;
3639:
3640: if (startpos < size1 && startpos + range >= size1)
3641: lim = range - (size1 - startpos);
3642:
3643: d = (startpos >= size1 ? string2 - size1 : string1) + startpos;
3644:
3645: /* Written out as an if-else to avoid testing `translate'
3646: inside the loop. */
3647: if (translate)
3648: while (range > lim
3649: && !fastmap[(unsigned char)
3650: translate[(unsigned char) *d++]])
3651: range--;
3652: else
3653: while (range > lim && !fastmap[(unsigned char) *d++])
3654: range--;
3655:
3656: startpos += irange - range;
3657: }
3658: else /* Searching backwards. */
3659: {
3660: register char c = (size1 == 0 || startpos >= size1
3661: ? string2[startpos - size1]
3662: : string1[startpos]);
3663:
3664: if (!fastmap[(unsigned char) TRANSLATE (c)])
3665: goto advance;
3666: }
3667: }
3668:
3669: /* If can't match the null string, and that's all we have left, fail. */
3670: if (range >= 0 && startpos == total_size && fastmap
3671: && !bufp->can_be_null)
3672: return -1;
3673:
3674: val = re_match_2_internal (bufp, string1, size1, string2, size2,
3675: startpos, regs, stop);
3676: #ifndef REGEX_MALLOC
3677: # ifdef C_ALLOCA
3678: alloca (0);
3679: # endif
3680: #endif
3681:
3682: if (val >= 0)
3683: return startpos;
3684:
3685: if (val == -2)
3686: return -2;
3687:
3688: advance:
3689: if (!range)
3690: break;
3691: else if (range > 0)
3692: {
3693: range--;
3694: startpos++;
3695: }
3696: else
3697: {
3698: range++;
3699: startpos--;
3700: }
3701: }
3702: return -1;
3703: } /* re_search_2 */
3704: #ifdef _LIBC
3705: weak_alias (__re_search_2, re_search_2)
3706: #endif
3707:
3708: /* This converts PTR, a pointer into one of the search strings `string1'
3709: and `string2' into an offset from the beginning of that string. */
3710: #define POINTER_TO_OFFSET(ptr) \
3711: (FIRST_STRING_P (ptr) \
3712: ? ((regoff_t) ((ptr) - string1)) \
3713: : ((regoff_t) ((ptr) - string2 + size1)))
3714:
3715: /* Macros for dealing with the split strings in re_match_2. */
3716:
3717: #define MATCHING_IN_FIRST_STRING (dend == end_match_1)
3718:
3719: /* Call before fetching a character with *d. This switches over to
3720: string2 if necessary. */
3721: #define PREFETCH() \
3722: while (d == dend) \
3723: { \
3724: /* End of string2 => fail. */ \
3725: if (dend == end_match_2) \
3726: goto fail; \
3727: /* End of string1 => advance to string2. */ \
3728: d = string2; \
3729: dend = end_match_2; \
3730: }
3731:
3732:
3733: /* Test if at very beginning or at very end of the virtual concatenation
3734: of `string1' and `string2'. If only one string, it's `string2'. */
3735: #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2)
3736: #define AT_STRINGS_END(d) ((d) == end2)
3737:
3738:
3739: /* Test if D points to a character which is word-constituent. We have
3740: two special cases to check for: if past the end of string1, look at
3741: the first character in string2; and if before the beginning of
3742: string2, look at the last character in string1. */
3743: #define WORDCHAR_P(d) \
3744: (SYNTAX ((d) == end1 ? *string2 \
3745: : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \
3746: == Sword)
3747:
3748: /* Disabled due to a compiler bug -- see comment at case wordbound */
3749: #if 0
3750: /* Test if the character before D and the one at D differ with respect
3751: to being word-constituent. */
3752: #define AT_WORD_BOUNDARY(d) \
3753: (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \
3754: || WORDCHAR_P (d - 1) != WORDCHAR_P (d))
3755: #endif
3756:
3757: /* Free everything we malloc. */
3758: #ifdef MATCH_MAY_ALLOCATE
3759: # define FREE_VAR(var) if (var) REGEX_FREE (var); var = NULL
3760: # define FREE_VARIABLES() \
3761: do { \
3762: REGEX_FREE_STACK (fail_stack.stack); \
3763: FREE_VAR (regstart); \
3764: FREE_VAR (regend); \
3765: FREE_VAR (old_regstart); \
3766: FREE_VAR (old_regend); \
3767: FREE_VAR (best_regstart); \
3768: FREE_VAR (best_regend); \
3769: FREE_VAR (reg_info); \
3770: FREE_VAR (reg_dummy); \
3771: FREE_VAR (reg_info_dummy); \
3772: } while (0)
3773: #else
3774: # define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */
3775: #endif /* not MATCH_MAY_ALLOCATE */
3776:
3777: /* These values must meet several constraints. They must not be valid
3778: register values; since we have a limit of 255 registers (because
3779: we use only one byte in the pattern for the register number), we can
3780: use numbers larger than 255. They must differ by 1, because of
3781: NUM_FAILURE_ITEMS above. And the value for the lowest register must
3782: be larger than the value for the highest register, so we do not try
3783: to actually save any registers when none are active. */
3784: #define NO_HIGHEST_ACTIVE_REG (1 << BYTEWIDTH)
3785: #define NO_LOWEST_ACTIVE_REG (NO_HIGHEST_ACTIVE_REG + 1)
3786:
3787: /* Matching routines. */
3788:
3789: #ifndef emacs /* Emacs never uses this. */
3790: /* re_match is like re_match_2 except it takes only a single string. */
3791:
3792: int
3793: re_match (bufp, string, size, pos, regs)
3794: struct re_pattern_buffer *bufp;
3795: const char *string;
3796: int size, pos;
3797: struct re_registers *regs;
3798: {
3799: int result = re_match_2_internal (bufp, NULL, 0, string, size,
3800: pos, regs, size);
3801: # ifndef REGEX_MALLOC
3802: # ifdef C_ALLOCA
3803: alloca (0);
3804: # endif
3805: # endif
3806: return result;
3807: }
3808: # ifdef _LIBC
3809: weak_alias (__re_match, re_match)
3810: # endif
3811: #endif /* not emacs */
3812:
3813: static boolean group_match_null_string_p _RE_ARGS ((unsigned char **p,
3814: unsigned char *end,
3815: register_info_type *reg_info));
3816: static boolean alt_match_null_string_p _RE_ARGS ((unsigned char *p,
3817: unsigned char *end,
3818: register_info_type *reg_info));
3819: static boolean common_op_match_null_string_p _RE_ARGS ((unsigned char **p,
3820: unsigned char *end,
3821: register_info_type *reg_info));
3822: static int bcmp_translate _RE_ARGS ((const char *s1, const char *s2,
3823: int len, char *translate));
3824:
3825: /* re_match_2 matches the compiled pattern in BUFP against the
3826: the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1
3827: and SIZE2, respectively). We start matching at POS, and stop
3828: matching at STOP.
3829:
3830: If REGS is non-null and the `no_sub' field of BUFP is nonzero, we
3831: store offsets for the substring each group matched in REGS. See the
3832: documentation for exactly how many groups we fill.
3833:
3834: We return -1 if no match, -2 if an internal error (such as the
3835: failure stack overflowing). Otherwise, we return the length of the
3836: matched substring. */
3837:
3838: int
3839: re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
3840: struct re_pattern_buffer *bufp;
3841: const char *string1, *string2;
3842: int size1, size2;
3843: int pos;
3844: struct re_registers *regs;
3845: int stop;
3846: {
3847: int result = re_match_2_internal (bufp, string1, size1, string2, size2,
3848: pos, regs, stop);
3849: #ifndef REGEX_MALLOC
3850: # ifdef C_ALLOCA
3851: alloca (0);
3852: # endif
3853: #endif
3854: return result;
3855: }
3856: #ifdef _LIBC
3857: weak_alias (__re_match_2, re_match_2)
3858: #endif
3859:
3860: /* This is a separate function so that we can force an alloca cleanup
3861: afterwards. */
3862: static int
3863: re_match_2_internal (bufp, string1, size1, string2, size2, pos, regs, stop)
3864: struct re_pattern_buffer *bufp;
3865: const char *string1, *string2;
3866: int size1, size2;
3867: int pos;
3868: struct re_registers *regs;
3869: int stop;
3870: {
3871: /* General temporaries. */
3872: int mcnt;
3873: unsigned char *p1;
3874:
3875: /* Just past the end of the corresponding string. */
3876: const char *end1, *end2;
3877:
3878: /* Pointers into string1 and string2, just past the last characters in
3879: each to consider matching. */
3880: const char *end_match_1, *end_match_2;
3881:
3882: /* Where we are in the data, and the end of the current string. */
3883: const char *d, *dend;
3884:
3885: /* Where we are in the pattern, and the end of the pattern. */
3886: unsigned char *p = bufp->buffer;
3887: register unsigned char *pend = p + bufp->used;
3888:
3889: /* Mark the opcode just after a start_memory, so we can test for an
3890: empty subpattern when we get to the stop_memory. */
3891: unsigned char *just_past_start_mem = 0;
3892:
3893: /* We use this to map every character in the string. */
3894: RE_TRANSLATE_TYPE translate = bufp->translate;
3895:
3896: /* Failure point stack. Each place that can handle a failure further
3897: down the line pushes a failure point on this stack. It consists of
3898: restart, regend, and reg_info for all registers corresponding to
3899: the subexpressions we're currently inside, plus the number of such
3900: registers, and, finally, two char *'s. The first char * is where
3901: to resume scanning the pattern; the second one is where to resume
3902: scanning the strings. If the latter is zero, the failure point is
3903: a ``dummy''; if a failure happens and the failure point is a dummy,
3904: it gets discarded and the next next one is tried. */
3905: #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */
3906: fail_stack_type fail_stack;
3907: #endif
3908: #ifdef DEBUG
3909: static unsigned failure_id;
3910: unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0;
3911: #endif
3912:
3913: #ifdef REL_ALLOC
3914: /* This holds the pointer to the failure stack, when
3915: it is allocated relocatably. */
3916: fail_stack_elt_t *failure_stack_ptr;
3917: #endif
3918:
3919: /* We fill all the registers internally, independent of what we
3920: return, for use in backreferences. The number here includes
3921: an element for register zero. */
3922: size_t num_regs = bufp->re_nsub + 1;
3923:
3924: /* The currently active registers. */
3925: active_reg_t lowest_active_reg = NO_LOWEST_ACTIVE_REG;
3926: active_reg_t highest_active_reg = NO_HIGHEST_ACTIVE_REG;
3927:
3928: /* Information on the contents of registers. These are pointers into
3929: the input strings; they record just what was matched (on this
3930: attempt) by a subexpression part of the pattern, that is, the
3931: regnum-th regstart pointer points to where in the pattern we began
3932: matching and the regnum-th regend points to right after where we
3933: stopped matching the regnum-th subexpression. (The zeroth register
3934: keeps track of what the whole pattern matches.) */
3935: #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
3936: const char **regstart, **regend;
3937: #endif
3938:
3939: /* If a group that's operated upon by a repetition operator fails to
3940: match anything, then the register for its start will need to be
3941: restored because it will have been set to wherever in the string we
3942: are when we last see its open-group operator. Similarly for a
3943: register's end. */
3944: #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
3945: const char **old_regstart, **old_regend;
3946: #endif
3947:
3948: /* The is_active field of reg_info helps us keep track of which (possibly
3949: nested) subexpressions we are currently in. The matched_something
3950: field of reg_info[reg_num] helps us tell whether or not we have
3951: matched any of the pattern so far this time through the reg_num-th
3952: subexpression. These two fields get reset each time through any
3953: loop their register is in. */
3954: #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */
3955: register_info_type *reg_info;
3956: #endif
3957:
3958: /* The following record the register info as found in the above
3959: variables when we find a match better than any we've seen before.
3960: This happens as we backtrack through the failure points, which in
3961: turn happens only if we have not yet matched the entire string. */
3962: unsigned best_regs_set = false;
3963: #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
3964: const char **best_regstart, **best_regend;
3965: #endif
3966:
3967: /* Logically, this is `best_regend[0]'. But we don't want to have to
3968: allocate space for that if we're not allocating space for anything
3969: else (see below). Also, we never need info about register 0 for
3970: any of the other register vectors, and it seems rather a kludge to
3971: treat `best_regend' differently than the rest. So we keep track of
3972: the end of the best match so far in a separate variable. We
3973: initialize this to NULL so that when we backtrack the first time
3974: and need to test it, it's not garbage. */
3975: const char *match_end = NULL;
3976:
3977: /* This helps SET_REGS_MATCHED avoid doing redundant work. */
3978: int set_regs_matched_done = 0;
3979:
3980: /* Used when we pop values we don't care about. */
3981: #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
3982: const char **reg_dummy;
3983: register_info_type *reg_info_dummy;
3984: #endif
3985:
3986: #ifdef DEBUG
3987: /* Counts the total number of registers pushed. */
3988: unsigned num_regs_pushed = 0;
3989: #endif
3990:
3991: DEBUG_PRINT1 ("\n\nEntering re_match_2.\n");
3992:
3993: INIT_FAIL_STACK ();
3994:
3995: #ifdef MATCH_MAY_ALLOCATE
3996: /* Do not bother to initialize all the register variables if there are
3997: no groups in the pattern, as it takes a fair amount of time. If
3998: there are groups, we include space for register 0 (the whole
3999: pattern), even though we never use it, since it simplifies the
4000: array indexing. We should fix this. */
4001: if (bufp->re_nsub)
4002: {
4003: regstart = REGEX_TALLOC (num_regs, const char *);
4004: regend = REGEX_TALLOC (num_regs, const char *);
4005: old_regstart = REGEX_TALLOC (num_regs, const char *);
4006: old_regend = REGEX_TALLOC (num_regs, const char *);
4007: best_regstart = REGEX_TALLOC (num_regs, const char *);
4008: best_regend = REGEX_TALLOC (num_regs, const char *);
4009: reg_info = REGEX_TALLOC (num_regs, register_info_type);
4010: reg_dummy = REGEX_TALLOC (num_regs, const char *);
4011: reg_info_dummy = REGEX_TALLOC (num_regs, register_info_type);
4012:
4013: if (!(regstart && regend && old_regstart && old_regend && reg_info
4014: && best_regstart && best_regend && reg_dummy && reg_info_dummy))
4015: {
4016: FREE_VARIABLES ();
4017: return -2;
4018: }
4019: }
4020: else
4021: {
4022: /* We must initialize all our variables to NULL, so that
4023: `FREE_VARIABLES' doesn't try to free them. */
4024: regstart = regend = old_regstart = old_regend = best_regstart
4025: = best_regend = reg_dummy = NULL;
4026: reg_info = reg_info_dummy = (register_info_type *) NULL;
4027: }
4028: #endif /* MATCH_MAY_ALLOCATE */
4029:
4030: /* The starting position is bogus. */
4031: if (pos < 0 || pos > size1 + size2)
4032: {
4033: FREE_VARIABLES ();
4034: return -1;
4035: }
4036:
4037: /* Initialize subexpression text positions to -1 to mark ones that no
4038: start_memory/stop_memory has been seen for. Also initialize the
4039: register information struct. */
4040: for (mcnt = 1; (unsigned) mcnt < num_regs; mcnt++)
4041: {
4042: regstart[mcnt] = regend[mcnt]
4043: = old_regstart[mcnt] = old_regend[mcnt] = REG_UNSET_VALUE;
4044:
4045: REG_MATCH_NULL_STRING_P (reg_info[mcnt]) = MATCH_NULL_UNSET_VALUE;
4046: IS_ACTIVE (reg_info[mcnt]) = 0;
4047: MATCHED_SOMETHING (reg_info[mcnt]) = 0;
4048: EVER_MATCHED_SOMETHING (reg_info[mcnt]) = 0;
4049: }
4050:
4051: /* We move `string1' into `string2' if the latter's empty -- but not if
4052: `string1' is null. */
4053: if (size2 == 0 && string1 != NULL)
4054: {
4055: string2 = string1;
4056: size2 = size1;
4057: string1 = 0;
4058: size1 = 0;
4059: }
4060: end1 = string1 + size1;
4061: end2 = string2 + size2;
4062:
4063: /* Compute where to stop matching, within the two strings. */
4064: if (stop <= size1)
4065: {
4066: end_match_1 = string1 + stop;
4067: end_match_2 = string2;
4068: }
4069: else
4070: {
4071: end_match_1 = end1;
4072: end_match_2 = string2 + stop - size1;
4073: }
4074:
4075: /* `p' scans through the pattern as `d' scans through the data.
4076: `dend' is the end of the input string that `d' points within. `d'
4077: is advanced into the following input string whenever necessary, but
4078: this happens before fetching; therefore, at the beginning of the
4079: loop, `d' can be pointing at the end of a string, but it cannot
4080: equal `string2'. */
4081: if (size1 > 0 && pos <= size1)
4082: {
4083: d = string1 + pos;
4084: dend = end_match_1;
4085: }
4086: else
4087: {
4088: d = string2 + pos - size1;
4089: dend = end_match_2;
4090: }
4091:
4092: DEBUG_PRINT1 ("The compiled pattern is:\n");
4093: DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend);
4094: DEBUG_PRINT1 ("The string to match is: `");
4095: DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2);
4096: DEBUG_PRINT1 ("'\n");
4097:
4098: /* This loops over pattern commands. It exits by returning from the
4099: function if the match is complete, or it drops through if the match
4100: fails at this starting point in the input data. */
4101: for (;;)
4102: {
4103: #ifdef _LIBC
4104: DEBUG_PRINT2 ("\n%p: ", p);
4105: #else
4106: DEBUG_PRINT2 ("\n0x%x: ", p);
4107: #endif
4108:
4109: if (p == pend)
4110: { /* End of pattern means we might have succeeded. */
4111: DEBUG_PRINT1 ("end of pattern ... ");
4112:
4113: /* If we haven't matched the entire string, and we want the
4114: longest match, try backtracking. */
4115: if (d != end_match_2)
4116: {
4117: /* 1 if this match ends in the same string (string1 or string2)
4118: as the best previous match. */
4119: boolean same_str_p = (FIRST_STRING_P (match_end)
4120: == MATCHING_IN_FIRST_STRING);
4121: /* 1 if this match is the best seen so far. */
4122: boolean best_match_p;
4123:
4124: /* AIX compiler got confused when this was combined
4125: with the previous declaration. */
4126: if (same_str_p)
4127: best_match_p = d > match_end;
4128: else
4129: best_match_p = !MATCHING_IN_FIRST_STRING;
4130:
4131: DEBUG_PRINT1 ("backtracking.\n");
4132:
4133: if (!FAIL_STACK_EMPTY ())
4134: { /* More failure points to try. */
4135:
4136: /* If exceeds best match so far, save it. */
4137: if (!best_regs_set || best_match_p)
4138: {
4139: best_regs_set = true;
4140: match_end = d;
4141:
4142: DEBUG_PRINT1 ("\nSAVING match as best so far.\n");
4143:
4144: for (mcnt = 1; (unsigned) mcnt < num_regs; mcnt++)
4145: {
4146: best_regstart[mcnt] = regstart[mcnt];
4147: best_regend[mcnt] = regend[mcnt];
4148: }
4149: }
4150: goto fail;
4151: }
4152:
4153: /* If no failure points, don't restore garbage. And if
4154: last match is real best match, don't restore second
4155: best one. */
4156: else if (best_regs_set && !best_match_p)
4157: {
4158: restore_best_regs:
4159: /* Restore best match. It may happen that `dend ==
4160: end_match_1' while the restored d is in string2.
4161: For example, the pattern `x.*y.*z' against the
4162: strings `x-' and `y-z-', if the two strings are
4163: not consecutive in memory. */
4164: DEBUG_PRINT1 ("Restoring best registers.\n");
4165:
4166: d = match_end;
4167: dend = ((d >= string1 && d <= end1)
4168: ? end_match_1 : end_match_2);
4169:
4170: for (mcnt = 1; (unsigned) mcnt < num_regs; mcnt++)
4171: {
4172: regstart[mcnt] = best_regstart[mcnt];
4173: regend[mcnt] = best_regend[mcnt];
4174: }
4175: }
4176: } /* d != end_match_2 */
4177:
4178: succeed_label:
4179: DEBUG_PRINT1 ("Accepting match.\n");
4180:
4181: /* If caller wants register contents data back, do it. */
4182: if (regs && !bufp->no_sub)
4183: {
4184: /* Have the register data arrays been allocated? */
4185: if (bufp->regs_allocated == REGS_UNALLOCATED)
4186: { /* No. So allocate them with malloc. We need one
4187: extra element beyond `num_regs' for the `-1' marker
4188: GNU code uses. */
4189: regs->num_regs = MAX (RE_NREGS, num_regs + 1);
4190: regs->start = TALLOC (regs->num_regs, regoff_t);
4191: regs->end = TALLOC (regs->num_regs, regoff_t);
4192: if (regs->start == NULL || regs->end == NULL)
4193: {
4194: FREE_VARIABLES ();
4195: return -2;
4196: }
4197: bufp->regs_allocated = REGS_REALLOCATE;
4198: }
4199: else if (bufp->regs_allocated == REGS_REALLOCATE)
4200: { /* Yes. If we need more elements than were already
4201: allocated, reallocate them. If we need fewer, just
4202: leave it alone. */
4203: if (regs->num_regs < num_regs + 1)
4204: {
4205: regs->num_regs = num_regs + 1;
4206: RETALLOC (regs->start, regs->num_regs, regoff_t);
4207: RETALLOC (regs->end, regs->num_regs, regoff_t);
4208: if (regs->start == NULL || regs->end == NULL)
4209: {
4210: FREE_VARIABLES ();
4211: return -2;
4212: }
4213: }
4214: }
4215: else
4216: {
4217: /* These braces fend off a "empty body in an else-statement"
4218: warning under GCC when assert expands to nothing. */
4219: assert (bufp->regs_allocated == REGS_FIXED);
4220: }
4221:
4222: /* Convert the pointer data in `regstart' and `regend' to
4223: indices. Register zero has to be set differently,
4224: since we haven't kept track of any info for it. */
4225: if (regs->num_regs > 0)
4226: {
4227: regs->start[0] = pos;
4228: regs->end[0] = (MATCHING_IN_FIRST_STRING
4229: ? ((regoff_t) (d - string1))
4230: : ((regoff_t) (d - string2 + size1)));
4231: }
4232:
4233: /* Go through the first `min (num_regs, regs->num_regs)'
4234: registers, since that is all we initialized. */
4235: for (mcnt = 1; (unsigned) mcnt < MIN (num_regs, regs->num_regs);
4236: mcnt++)
4237: {
4238: if (REG_UNSET (regstart[mcnt]) || REG_UNSET (regend[mcnt]))
4239: regs->start[mcnt] = regs->end[mcnt] = -1;
4240: else
4241: {
4242: regs->start[mcnt]
4243: = (regoff_t) POINTER_TO_OFFSET (regstart[mcnt]);
4244: regs->end[mcnt]
4245: = (regoff_t) POINTER_TO_OFFSET (regend[mcnt]);
4246: }
4247: }
4248:
4249: /* If the regs structure we return has more elements than
4250: were in the pattern, set the extra elements to -1. If
4251: we (re)allocated the registers, this is the case,
4252: because we always allocate enough to have at least one
4253: -1 at the end. */
4254: for (mcnt = num_regs; (unsigned) mcnt < regs->num_regs; mcnt++)
4255: regs->start[mcnt] = regs->end[mcnt] = -1;
4256: } /* regs && !bufp->no_sub */
4257:
4258: DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n",
4259: nfailure_points_pushed, nfailure_points_popped,
4260: nfailure_points_pushed - nfailure_points_popped);
4261: DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed);
4262:
4263: mcnt = d - pos - (MATCHING_IN_FIRST_STRING
4264: ? string1
4265: : string2 - size1);
4266:
4267: DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt);
4268:
4269: FREE_VARIABLES ();
4270: return mcnt;
4271: }
4272:
4273: /* Otherwise match next pattern command. */
4274: switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++))
4275: {
4276: /* Ignore these. Used to ignore the n of succeed_n's which
4277: currently have n == 0. */
4278: case no_op:
4279: DEBUG_PRINT1 ("EXECUTING no_op.\n");
4280: break;
4281:
4282: case succeed:
4283: DEBUG_PRINT1 ("EXECUTING succeed.\n");
4284: goto succeed_label;
4285:
4286: /* Match the next n pattern characters exactly. The following
4287: byte in the pattern defines n, and the n bytes after that
4288: are the characters to match. */
4289: case exactn:
4290: mcnt = *p++;
4291: DEBUG_PRINT2 ("EXECUTING exactn %d.\n", mcnt);
4292:
4293: /* This is written out as an if-else so we don't waste time
4294: testing `translate' inside the loop. */
4295: if (translate)
4296: {
4297: do
4298: {
4299: PREFETCH ();
4300: if ((unsigned char) translate[(unsigned char) *d++]
4301: != (unsigned char) *p++)
4302: goto fail;
4303: }
4304: while (--mcnt);
4305: }
4306: else
4307: {
4308: do
4309: {
4310: PREFETCH ();
4311: if (*d++ != (char) *p++) goto fail;
4312: }
4313: while (--mcnt);
4314: }
4315: SET_REGS_MATCHED ();
4316: break;
4317:
4318:
4319: /* Match any character except possibly a newline or a null. */
4320: case anychar:
4321: DEBUG_PRINT1 ("EXECUTING anychar.\n");
4322:
4323: PREFETCH ();
4324:
4325: if ((!(bufp->syntax & RE_DOT_NEWLINE) && TRANSLATE (*d) == '\n')
4326: || (bufp->syntax & RE_DOT_NOT_NULL && TRANSLATE (*d) == '\000'))
4327: goto fail;
4328:
4329: SET_REGS_MATCHED ();
4330: DEBUG_PRINT2 (" Matched `%d'.\n", *d);
4331: d++;
4332: break;
4333:
4334:
4335: case charset:
4336: case charset_not:
4337: {
4338: register unsigned char c;
4339: boolean not = (re_opcode_t) *(p - 1) == charset_not;
4340:
4341: DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : "");
4342:
4343: PREFETCH ();
4344: c = TRANSLATE (*d); /* The character to match. */
4345:
4346: /* Cast to `unsigned' instead of `unsigned char' in case the
4347: bit list is a full 32 bytes long. */
4348: if (c < (unsigned) (*p * BYTEWIDTH)
4349: && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
4350: not = !not;
4351:
4352: p += 1 + *p;
4353:
4354: if (!not) goto fail;
4355:
4356: SET_REGS_MATCHED ();
4357: d++;
4358: break;
4359: }
4360:
4361:
4362: /* The beginning of a group is represented by start_memory.
4363: The arguments are the register number in the next byte, and the
4364: number of groups inner to this one in the next. The text
4365: matched within the group is recorded (in the internal
4366: registers data structure) under the register number. */
4367: case start_memory:
4368: DEBUG_PRINT3 ("EXECUTING start_memory %d (%d):\n", *p, p[1]);
4369:
4370: /* Find out if this group can match the empty string. */
4371: p1 = p; /* To send to group_match_null_string_p. */
4372:
4373: if (REG_MATCH_NULL_STRING_P (reg_info[*p]) == MATCH_NULL_UNSET_VALUE)
4374: REG_MATCH_NULL_STRING_P (reg_info[*p])
4375: = group_match_null_string_p (&p1, pend, reg_info);
4376:
4377: /* Save the position in the string where we were the last time
4378: we were at this open-group operator in case the group is
4379: operated upon by a repetition operator, e.g., with `(a*)*b'
4380: against `ab'; then we want to ignore where we are now in
4381: the string in case this attempt to match fails. */
4382: old_regstart[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p])
4383: ? REG_UNSET (regstart[*p]) ? d : regstart[*p]
4384: : regstart[*p];
4385: DEBUG_PRINT2 (" old_regstart: %d\n",
4386: POINTER_TO_OFFSET (old_regstart[*p]));
4387:
4388: regstart[*p] = d;
4389: DEBUG_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p]));
4390:
4391: IS_ACTIVE (reg_info[*p]) = 1;
4392: MATCHED_SOMETHING (reg_info[*p]) = 0;
4393:
4394: /* Clear this whenever we change the register activity status. */
4395: set_regs_matched_done = 0;
4396:
4397: /* This is the new highest active register. */
4398: highest_active_reg = *p;
4399:
4400: /* If nothing was active before, this is the new lowest active
4401: register. */
4402: if (lowest_active_reg == NO_LOWEST_ACTIVE_REG)
4403: lowest_active_reg = *p;
4404:
4405: /* Move past the register number and inner group count. */
4406: p += 2;
4407: just_past_start_mem = p;
4408:
4409: break;
4410:
4411:
4412: /* The stop_memory opcode represents the end of a group. Its
4413: arguments are the same as start_memory's: the register
4414: number, and the number of inner groups. */
4415: case stop_memory:
4416: DEBUG_PRINT3 ("EXECUTING stop_memory %d (%d):\n", *p, p[1]);
4417:
4418: /* We need to save the string position the last time we were at
4419: this close-group operator in case the group is operated
4420: upon by a repetition operator, e.g., with `((a*)*(b*)*)*'
4421: against `aba'; then we want to ignore where we are now in
4422: the string in case this attempt to match fails. */
4423: old_regend[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p])
4424: ? REG_UNSET (regend[*p]) ? d : regend[*p]
4425: : regend[*p];
4426: DEBUG_PRINT2 (" old_regend: %d\n",
4427: POINTER_TO_OFFSET (old_regend[*p]));
4428:
4429: regend[*p] = d;
4430: DEBUG_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p]));
4431:
4432: /* This register isn't active anymore. */
4433: IS_ACTIVE (reg_info[*p]) = 0;
4434:
4435: /* Clear this whenever we change the register activity status. */
4436: set_regs_matched_done = 0;
4437:
4438: /* If this was the only register active, nothing is active
4439: anymore. */
4440: if (lowest_active_reg == highest_active_reg)
4441: {
4442: lowest_active_reg = NO_LOWEST_ACTIVE_REG;
4443: highest_active_reg = NO_HIGHEST_ACTIVE_REG;
4444: }
4445: else
4446: { /* We must scan for the new highest active register, since
4447: it isn't necessarily one less than now: consider
4448: (a(b)c(d(e)f)g). When group 3 ends, after the f), the
4449: new highest active register is 1. */
4450: unsigned char r = *p - 1;
4451: while (r > 0 && !IS_ACTIVE (reg_info[r]))
4452: r--;
4453:
4454: /* If we end up at register zero, that means that we saved
4455: the registers as the result of an `on_failure_jump', not
4456: a `start_memory', and we jumped to past the innermost
4457: `stop_memory'. For example, in ((.)*) we save
4458: registers 1 and 2 as a result of the *, but when we pop
4459: back to the second ), we are at the stop_memory 1.
4460: Thus, nothing is active. */
4461: if (r == 0)
4462: {
4463: lowest_active_reg = NO_LOWEST_ACTIVE_REG;
4464: highest_active_reg = NO_HIGHEST_ACTIVE_REG;
4465: }
4466: else
4467: highest_active_reg = r;
4468: }
4469:
4470: /* If just failed to match something this time around with a
4471: group that's operated on by a repetition operator, try to
4472: force exit from the ``loop'', and restore the register
4473: information for this group that we had before trying this
4474: last match. */
4475: if ((!MATCHED_SOMETHING (reg_info[*p])
4476: || just_past_start_mem == p - 1)
4477: && (p + 2) < pend)
4478: {
4479: boolean is_a_jump_n = false;
4480:
4481: p1 = p + 2;
4482: mcnt = 0;
4483: switch ((re_opcode_t) *p1++)
4484: {
4485: case jump_n:
4486: is_a_jump_n = true;
4487: case pop_failure_jump:
4488: case maybe_pop_jump:
4489: case jump:
4490: case dummy_failure_jump:
4491: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
4492: if (is_a_jump_n)
4493: p1 += 2;
4494: break;
4495:
4496: default:
4497: /* do nothing */ ;
4498: }
4499: p1 += mcnt;
4500:
4501: /* If the next operation is a jump backwards in the pattern
4502: to an on_failure_jump right before the start_memory
4503: corresponding to this stop_memory, exit from the loop
4504: by forcing a failure after pushing on the stack the
4505: on_failure_jump's jump in the pattern, and d. */
4506: if (mcnt < 0 && (re_opcode_t) *p1 == on_failure_jump
4507: && (re_opcode_t) p1[3] == start_memory && p1[4] == *p)
4508: {
4509: /* If this group ever matched anything, then restore
4510: what its registers were before trying this last
4511: failed match, e.g., with `(a*)*b' against `ab' for
4512: regstart[1], and, e.g., with `((a*)*(b*)*)*'
4513: against `aba' for regend[3].
4514:
4515: Also restore the registers for inner groups for,
4516: e.g., `((a*)(b*))*' against `aba' (register 3 would
4517: otherwise get trashed). */
4518:
4519: if (EVER_MATCHED_SOMETHING (reg_info[*p]))
4520: {
4521: unsigned r;
4522:
4523: EVER_MATCHED_SOMETHING (reg_info[*p]) = 0;
4524:
4525: /* Restore this and inner groups' (if any) registers. */
4526: for (r = *p; r < (unsigned) *p + (unsigned) *(p + 1);
4527: r++)
4528: {
4529: regstart[r] = old_regstart[r];
4530:
4531: /* xx why this test? */
4532: if (old_regend[r] >= regstart[r])
4533: regend[r] = old_regend[r];
4534: }
4535: }
4536: p1++;
4537: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
4538: PUSH_FAILURE_POINT (p1 + mcnt, d, -2);
4539:
4540: goto fail;
4541: }
4542: }
4543:
4544: /* Move past the register number and the inner group count. */
4545: p += 2;
4546: break;
4547:
4548:
4549: /* \<digit> has been turned into a `duplicate' command which is
4550: followed by the numeric value of <digit> as the register number. */
4551: case duplicate:
4552: {
4553: register const char *d2, *dend2;
4554: int regno = *p++; /* Get which register to match against. */
4555: DEBUG_PRINT2 ("EXECUTING duplicate %d.\n", regno);
4556:
4557: /* Can't back reference a group which we've never matched. */
4558: if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno]))
4559: goto fail;
4560:
4561: /* Where in input to try to start matching. */
4562: d2 = regstart[regno];
4563:
4564: /* Where to stop matching; if both the place to start and
4565: the place to stop matching are in the same string, then
4566: set to the place to stop, otherwise, for now have to use
4567: the end of the first string. */
4568:
4569: dend2 = ((FIRST_STRING_P (regstart[regno])
4570: == FIRST_STRING_P (regend[regno]))
4571: ? regend[regno] : end_match_1);
4572: for (;;)
4573: {
4574: /* If necessary, advance to next segment in register
4575: contents. */
4576: while (d2 == dend2)
4577: {
4578: if (dend2 == end_match_2) break;
4579: if (dend2 == regend[regno]) break;
4580:
4581: /* End of string1 => advance to string2. */
4582: d2 = string2;
4583: dend2 = regend[regno];
4584: }
4585: /* At end of register contents => success */
4586: if (d2 == dend2) break;
4587:
4588: /* If necessary, advance to next segment in data. */
4589: PREFETCH ();
4590:
4591: /* How many characters left in this segment to match. */
4592: mcnt = dend - d;
4593:
4594: /* Want how many consecutive characters we can match in
4595: one shot, so, if necessary, adjust the count. */
4596: if (mcnt > dend2 - d2)
4597: mcnt = dend2 - d2;
4598:
4599: /* Compare that many; failure if mismatch, else move
4600: past them. */
4601: if (translate
4602: ? bcmp_translate (d, d2, mcnt, translate)
4603: : memcmp (d, d2, mcnt))
4604: goto fail;
4605: d += mcnt, d2 += mcnt;
4606:
4607: /* Do this because we've match some characters. */
4608: SET_REGS_MATCHED ();
4609: }
4610: }
4611: break;
4612:
4613:
4614: /* begline matches the empty string at the beginning of the string
4615: (unless `not_bol' is set in `bufp'), and, if
4616: `newline_anchor' is set, after newlines. */
4617: case begline:
4618: DEBUG_PRINT1 ("EXECUTING begline.\n");
4619:
4620: if (AT_STRINGS_BEG (d))
4621: {
4622: if (!bufp->not_bol) break;
4623: }
4624: else if (d[-1] == '\n' && bufp->newline_anchor)
4625: {
4626: break;
4627: }
4628: /* In all other cases, we fail. */
4629: goto fail;
4630:
4631:
4632: /* endline is the dual of begline. */
4633: case endline:
4634: DEBUG_PRINT1 ("EXECUTING endline.\n");
4635:
4636: if (AT_STRINGS_END (d))
4637: {
4638: if (!bufp->not_eol) break;
4639: }
4640:
4641: /* We have to ``prefetch'' the next character. */
4642: else if ((d == end1 ? *string2 : *d) == '\n'
4643: && bufp->newline_anchor)
4644: {
4645: break;
4646: }
4647: goto fail;
4648:
4649:
4650: /* Match at the very beginning of the data. */
4651: case begbuf:
4652: DEBUG_PRINT1 ("EXECUTING begbuf.\n");
4653: if (AT_STRINGS_BEG (d))
4654: break;
4655: goto fail;
4656:
4657:
4658: /* Match at the very end of the data. */
4659: case endbuf:
4660: DEBUG_PRINT1 ("EXECUTING endbuf.\n");
4661: if (AT_STRINGS_END (d))
4662: break;
4663: goto fail;
4664:
4665:
4666: /* on_failure_keep_string_jump is used to optimize `.*\n'. It
4667: pushes NULL as the value for the string on the stack. Then
4668: `pop_failure_point' will keep the current value for the
4669: string, instead of restoring it. To see why, consider
4670: matching `foo\nbar' against `.*\n'. The .* matches the foo;
4671: then the . fails against the \n. But the next thing we want
4672: to do is match the \n against the \n; if we restored the
4673: string value, we would be back at the foo.
4674:
4675: Because this is used only in specific cases, we don't need to
4676: check all the things that `on_failure_jump' does, to make
4677: sure the right things get saved on the stack. Hence we don't
4678: share its code. The only reason to push anything on the
4679: stack at all is that otherwise we would have to change
4680: `anychar's code to do something besides goto fail in this
4681: case; that seems worse than this. */
4682: case on_failure_keep_string_jump:
4683: DEBUG_PRINT1 ("EXECUTING on_failure_keep_string_jump");
4684:
4685: EXTRACT_NUMBER_AND_INCR (mcnt, p);
4686: #ifdef _LIBC
4687: DEBUG_PRINT3 (" %d (to %p):\n", mcnt, p + mcnt);
4688: #else
4689: DEBUG_PRINT3 (" %d (to 0x%x):\n", mcnt, p + mcnt);
4690: #endif
4691:
4692: PUSH_FAILURE_POINT (p + mcnt, NULL, -2);
4693: break;
4694:
4695:
4696: /* Uses of on_failure_jump:
4697:
4698: Each alternative starts with an on_failure_jump that points
4699: to the beginning of the next alternative. Each alternative
4700: except the last ends with a jump that in effect jumps past
4701: the rest of the alternatives. (They really jump to the
4702: ending jump of the following alternative, because tensioning
4703: these jumps is a hassle.)
4704:
4705: Repeats start with an on_failure_jump that points past both
4706: the repetition text and either the following jump or
4707: pop_failure_jump back to this on_failure_jump. */
4708: case on_failure_jump:
4709: on_failure:
4710: DEBUG_PRINT1 ("EXECUTING on_failure_jump");
4711:
4712: EXTRACT_NUMBER_AND_INCR (mcnt, p);
4713: #ifdef _LIBC
4714: DEBUG_PRINT3 (" %d (to %p)", mcnt, p + mcnt);
4715: #else
4716: DEBUG_PRINT3 (" %d (to 0x%x)", mcnt, p + mcnt);
4717: #endif
4718:
4719: /* If this on_failure_jump comes right before a group (i.e.,
4720: the original * applied to a group), save the information
4721: for that group and all inner ones, so that if we fail back
4722: to this point, the group's information will be correct.
4723: For example, in \(a*\)*\1, we need the preceding group,
4724: and in \(zz\(a*\)b*\)\2, we need the inner group. */
4725:
4726: /* We can't use `p' to check ahead because we push
4727: a failure point to `p + mcnt' after we do this. */
4728: p1 = p;
4729:
4730: /* We need to skip no_op's before we look for the
4731: start_memory in case this on_failure_jump is happening as
4732: the result of a completed succeed_n, as in \(a\)\{1,3\}b\1
4733: against aba. */
4734: while (p1 < pend && (re_opcode_t) *p1 == no_op)
4735: p1++;
4736:
4737: if (p1 < pend && (re_opcode_t) *p1 == start_memory)
4738: {
4739: /* We have a new highest active register now. This will
4740: get reset at the start_memory we are about to get to,
4741: but we will have saved all the registers relevant to
4742: this repetition op, as described above. */
4743: highest_active_reg = *(p1 + 1) + *(p1 + 2);
4744: if (lowest_active_reg == NO_LOWEST_ACTIVE_REG)
4745: lowest_active_reg = *(p1 + 1);
4746: }
4747:
4748: DEBUG_PRINT1 (":\n");
4749: PUSH_FAILURE_POINT (p + mcnt, d, -2);
4750: break;
4751:
4752:
4753: /* A smart repeat ends with `maybe_pop_jump'.
4754: We change it to either `pop_failure_jump' or `jump'. */
4755: case maybe_pop_jump:
4756: EXTRACT_NUMBER_AND_INCR (mcnt, p);
4757: DEBUG_PRINT2 ("EXECUTING maybe_pop_jump %d.\n", mcnt);
4758: {
4759: register unsigned char *p2 = p;
4760:
4761: /* Compare the beginning of the repeat with what in the
4762: pattern follows its end. If we can establish that there
4763: is nothing that they would both match, i.e., that we
4764: would have to backtrack because of (as in, e.g., `a*a')
4765: then we can change to pop_failure_jump, because we'll
4766: never have to backtrack.
4767:
4768: This is not true in the case of alternatives: in
4769: `(a|ab)*' we do need to backtrack to the `ab' alternative
4770: (e.g., if the string was `ab'). But instead of trying to
4771: detect that here, the alternative has put on a dummy
4772: failure point which is what we will end up popping. */
4773:
4774: /* Skip over open/close-group commands.
4775: If what follows this loop is a ...+ construct,
4776: look at what begins its body, since we will have to
4777: match at least one of that. */
4778: while (1)
4779: {
4780: if (p2 + 2 < pend
4781: && ((re_opcode_t) *p2 == stop_memory
4782: || (re_opcode_t) *p2 == start_memory))
4783: p2 += 3;
4784: else if (p2 + 6 < pend
4785: && (re_opcode_t) *p2 == dummy_failure_jump)
4786: p2 += 6;
4787: else
4788: break;
4789: }
4790:
4791: p1 = p + mcnt;
4792: /* p1[0] ... p1[2] are the `on_failure_jump' corresponding
4793: to the `maybe_finalize_jump' of this case. Examine what
4794: follows. */
4795:
4796: /* If we're at the end of the pattern, we can change. */
4797: if (p2 == pend)
4798: {
4799: /* Consider what happens when matching ":\(.*\)"
4800: against ":/". I don't really understand this code
4801: yet. */
4802: p[-3] = (unsigned char) pop_failure_jump;
4803: DEBUG_PRINT1
4804: (" End of pattern: change to `pop_failure_jump'.\n");
4805: }
4806:
4807: else if ((re_opcode_t) *p2 == exactn
4808: || (bufp->newline_anchor && (re_opcode_t) *p2 == endline))
4809: {
4810: register unsigned char c
4811: = *p2 == (unsigned char) endline ? '\n' : p2[2];
4812:
4813: if ((re_opcode_t) p1[3] == exactn && p1[5] != c)
4814: {
4815: p[-3] = (unsigned char) pop_failure_jump;
4816: DEBUG_PRINT3 (" %c != %c => pop_failure_jump.\n",
4817: c, p1[5]);
4818: }
4819:
4820: else if ((re_opcode_t) p1[3] == charset
4821: || (re_opcode_t) p1[3] == charset_not)
4822: {
4823: int not = (re_opcode_t) p1[3] == charset_not;
4824:
4825: if (c < (unsigned char) (p1[4] * BYTEWIDTH)
4826: && p1[5 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
4827: not = !not;
4828:
4829: /* `not' is equal to 1 if c would match, which means
4830: that we can't change to pop_failure_jump. */
4831: if (!not)
4832: {
4833: p[-3] = (unsigned char) pop_failure_jump;
4834: DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
4835: }
4836: }
4837: }
4838: else if ((re_opcode_t) *p2 == charset)
4839: {
4840: #ifdef DEBUG
4841: register unsigned char c
4842: = *p2 == (unsigned char) endline ? '\n' : p2[2];
4843: #endif
4844:
4845: #if 0
4846: if ((re_opcode_t) p1[3] == exactn
4847: && ! ((int) p2[1] * BYTEWIDTH > (int) p1[5]
4848: && (p2[2 + p1[5] / BYTEWIDTH]
4849: & (1 << (p1[5] % BYTEWIDTH)))))
4850: #else
4851: if ((re_opcode_t) p1[3] == exactn
4852: && ! ((int) p2[1] * BYTEWIDTH > (int) p1[4]
4853: && (p2[2 + p1[4] / BYTEWIDTH]
4854: & (1 << (p1[4] % BYTEWIDTH)))))
4855: #endif
4856: {
4857: p[-3] = (unsigned char) pop_failure_jump;
4858: DEBUG_PRINT3 (" %c != %c => pop_failure_jump.\n",
4859: c, p1[5]);
4860: }
4861:
4862: else if ((re_opcode_t) p1[3] == charset_not)
4863: {
4864: int idx;
4865: /* We win if the charset_not inside the loop
4866: lists every character listed in the charset after. */
4867: for (idx = 0; idx < (int) p2[1]; idx++)
4868: if (! (p2[2 + idx] == 0
4869: || (idx < (int) p1[4]
4870: && ((p2[2 + idx] & ~ p1[5 + idx]) == 0))))
4871: break;
4872:
4873: if (idx == p2[1])
4874: {
4875: p[-3] = (unsigned char) pop_failure_jump;
4876: DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
4877: }
4878: }
4879: else if ((re_opcode_t) p1[3] == charset)
4880: {
4881: int idx;
4882: /* We win if the charset inside the loop
4883: has no overlap with the one after the loop. */
4884: for (idx = 0;
4885: idx < (int) p2[1] && idx < (int) p1[4];
4886: idx++)
4887: if ((p2[2 + idx] & p1[5 + idx]) != 0)
4888: break;
4889:
4890: if (idx == p2[1] || idx == p1[4])
4891: {
4892: p[-3] = (unsigned char) pop_failure_jump;
4893: DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
4894: }
4895: }
4896: }
4897: }
4898: p -= 2; /* Point at relative address again. */
4899: if ((re_opcode_t) p[-1] != pop_failure_jump)
4900: {
4901: p[-1] = (unsigned char) jump;
4902: DEBUG_PRINT1 (" Match => jump.\n");
4903: goto unconditional_jump;
4904: }
4905: /* Note fall through. */
4906:
4907:
4908: /* The end of a simple repeat has a pop_failure_jump back to
4909: its matching on_failure_jump, where the latter will push a
4910: failure point. The pop_failure_jump takes off failure
4911: points put on by this pop_failure_jump's matching
4912: on_failure_jump; we got through the pattern to here from the
4913: matching on_failure_jump, so didn't fail. */
4914: case pop_failure_jump:
4915: {
4916: /* We need to pass separate storage for the lowest and
4917: highest registers, even though we don't care about the
4918: actual values. Otherwise, we will restore only one
4919: register from the stack, since lowest will == highest in
4920: `pop_failure_point'. */
4921: active_reg_t dummy_low_reg, dummy_high_reg;
4922: unsigned char *pdummy;
4923: const char *sdummy;
4924:
4925: DEBUG_PRINT1 ("EXECUTING pop_failure_jump.\n");
4926: POP_FAILURE_POINT (sdummy, pdummy,
4927: dummy_low_reg, dummy_high_reg,
4928: reg_dummy, reg_dummy, reg_info_dummy);
4929: }
4930: /* Note fall through. */
4931:
4932: unconditional_jump:
4933: #ifdef _LIBC
4934: DEBUG_PRINT2 ("\n%p: ", p);
4935: #else
4936: DEBUG_PRINT2 ("\n0x%x: ", p);
4937: #endif
4938: /* Note fall through. */
4939:
4940: /* Unconditionally jump (without popping any failure points). */
4941: case jump:
4942: EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */
4943: DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt);
4944: p += mcnt; /* Do the jump. */
4945: #ifdef _LIBC
4946: DEBUG_PRINT2 ("(to %p).\n", p);
4947: #else
4948: DEBUG_PRINT2 ("(to 0x%x).\n", p);
4949: #endif
4950: break;
4951:
4952:
4953: /* We need this opcode so we can detect where alternatives end
4954: in `group_match_null_string_p' et al. */
4955: case jump_past_alt:
4956: DEBUG_PRINT1 ("EXECUTING jump_past_alt.\n");
4957: goto unconditional_jump;
4958:
4959:
4960: /* Normally, the on_failure_jump pushes a failure point, which
4961: then gets popped at pop_failure_jump. We will end up at
4962: pop_failure_jump, also, and with a pattern of, say, `a+', we
4963: are skipping over the on_failure_jump, so we have to push
4964: something meaningless for pop_failure_jump to pop. */
4965: case dummy_failure_jump:
4966: DEBUG_PRINT1 ("EXECUTING dummy_failure_jump.\n");
4967: /* It doesn't matter what we push for the string here. What
4968: the code at `fail' tests is the value for the pattern. */
4969: PUSH_FAILURE_POINT (NULL, NULL, -2);
4970: goto unconditional_jump;
4971:
4972:
4973: /* At the end of an alternative, we need to push a dummy failure
4974: point in case we are followed by a `pop_failure_jump', because
4975: we don't want the failure point for the alternative to be
4976: popped. For example, matching `(a|ab)*' against `aab'
4977: requires that we match the `ab' alternative. */
4978: case push_dummy_failure:
4979: DEBUG_PRINT1 ("EXECUTING push_dummy_failure.\n");
4980: /* See comments just above at `dummy_failure_jump' about the
4981: two zeroes. */
4982: PUSH_FAILURE_POINT (NULL, NULL, -2);
4983: break;
4984:
4985: /* Have to succeed matching what follows at least n times.
4986: After that, handle like `on_failure_jump'. */
4987: case succeed_n:
4988: EXTRACT_NUMBER (mcnt, p + 2);
4989: DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt);
4990:
4991: assert (mcnt >= 0);
4992: /* Originally, this is how many times we HAVE to succeed. */
4993: if (mcnt > 0)
4994: {
4995: mcnt--;
4996: p += 2;
4997: STORE_NUMBER_AND_INCR (p, mcnt);
4998: #ifdef _LIBC
4999: DEBUG_PRINT3 (" Setting %p to %d.\n", p - 2, mcnt);
5000: #else
5001: DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p - 2, mcnt);
5002: #endif
5003: }
5004: else if (mcnt == 0)
5005: {
5006: #ifdef _LIBC
5007: DEBUG_PRINT2 (" Setting two bytes from %p to no_op.\n", p+2);
5008: #else
5009: DEBUG_PRINT2 (" Setting two bytes from 0x%x to no_op.\n", p+2);
5010: #endif
5011: p[2] = (unsigned char) no_op;
5012: p[3] = (unsigned char) no_op;
5013: goto on_failure;
5014: }
5015: break;
5016:
5017: case jump_n:
5018: EXTRACT_NUMBER (mcnt, p + 2);
5019: DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt);
5020:
5021: /* Originally, this is how many times we CAN jump. */
5022: if (mcnt)
5023: {
5024: mcnt--;
5025: STORE_NUMBER (p + 2, mcnt);
5026: #ifdef _LIBC
5027: DEBUG_PRINT3 (" Setting %p to %d.\n", p + 2, mcnt);
5028: #else
5029: DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p + 2, mcnt);
5030: #endif
5031: goto unconditional_jump;
5032: }
5033: /* If don't have to jump any more, skip over the rest of command. */
5034: else
5035: p += 4;
5036: break;
5037:
5038: case set_number_at:
5039: {
5040: DEBUG_PRINT1 ("EXECUTING set_number_at.\n");
5041:
5042: EXTRACT_NUMBER_AND_INCR (mcnt, p);
5043: p1 = p + mcnt;
5044: EXTRACT_NUMBER_AND_INCR (mcnt, p);
5045: #ifdef _LIBC
5046: DEBUG_PRINT3 (" Setting %p to %d.\n", p1, mcnt);
5047: #else
5048: DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p1, mcnt);
5049: #endif
5050: STORE_NUMBER (p1, mcnt);
5051: break;
5052: }
5053:
5054: #if 0
5055: /* The DEC Alpha C compiler 3.x generates incorrect code for the
5056: test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of
5057: AT_WORD_BOUNDARY, so this code is disabled. Expanding the
5058: macro and introducing temporary variables works around the bug. */
5059:
5060: case wordbound:
5061: DEBUG_PRINT1 ("EXECUTING wordbound.\n");
5062: if (AT_WORD_BOUNDARY (d))
5063: break;
5064: goto fail;
5065:
5066: case notwordbound:
5067: DEBUG_PRINT1 ("EXECUTING notwordbound.\n");
5068: if (AT_WORD_BOUNDARY (d))
5069: goto fail;
5070: break;
5071: #else
5072: case wordbound:
5073: {
5074: boolean prevchar, thischar;
5075:
5076: DEBUG_PRINT1 ("EXECUTING wordbound.\n");
5077: if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d))
5078: break;
5079:
5080: prevchar = WORDCHAR_P (d - 1);
5081: thischar = WORDCHAR_P (d);
5082: if (prevchar != thischar)
5083: break;
5084: goto fail;
5085: }
5086:
5087: case notwordbound:
5088: {
5089: boolean prevchar, thischar;
5090:
5091: DEBUG_PRINT1 ("EXECUTING notwordbound.\n");
5092: if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d))
5093: goto fail;
5094:
5095: prevchar = WORDCHAR_P (d - 1);
5096: thischar = WORDCHAR_P (d);
5097: if (prevchar != thischar)
5098: goto fail;
5099: break;
5100: }
5101: #endif
5102:
5103: case wordbeg:
5104: DEBUG_PRINT1 ("EXECUTING wordbeg.\n");
5105: if (WORDCHAR_P (d) && (AT_STRINGS_BEG (d) || !WORDCHAR_P (d - 1)))
5106: break;
5107: goto fail;
5108:
5109: case wordend:
5110: DEBUG_PRINT1 ("EXECUTING wordend.\n");
5111: if (!AT_STRINGS_BEG (d) && WORDCHAR_P (d - 1)
5112: && (!WORDCHAR_P (d) || AT_STRINGS_END (d)))
5113: break;
5114: goto fail;
5115:
5116: #ifdef emacs
5117: case before_dot:
5118: DEBUG_PRINT1 ("EXECUTING before_dot.\n");
5119: if (PTR_CHAR_POS ((unsigned char *) d) >= point)
5120: goto fail;
5121: break;
5122:
5123: case at_dot:
5124: DEBUG_PRINT1 ("EXECUTING at_dot.\n");
5125: if (PTR_CHAR_POS ((unsigned char *) d) != point)
5126: goto fail;
5127: break;
5128:
5129: case after_dot:
5130: DEBUG_PRINT1 ("EXECUTING after_dot.\n");
5131: if (PTR_CHAR_POS ((unsigned char *) d) <= point)
5132: goto fail;
5133: break;
5134:
5135: case syntaxspec:
5136: DEBUG_PRINT2 ("EXECUTING syntaxspec %d.\n", mcnt);
5137: mcnt = *p++;
5138: goto matchsyntax;
5139:
5140: case wordchar:
5141: DEBUG_PRINT1 ("EXECUTING Emacs wordchar.\n");
5142: mcnt = (int) Sword;
5143: matchsyntax:
5144: PREFETCH ();
5145: /* Can't use *d++ here; SYNTAX may be an unsafe macro. */
5146: d++;
5147: if (SYNTAX (d[-1]) != (enum syntaxcode) mcnt)
5148: goto fail;
5149: SET_REGS_MATCHED ();
5150: break;
5151:
5152: case notsyntaxspec:
5153: DEBUG_PRINT2 ("EXECUTING notsyntaxspec %d.\n", mcnt);
5154: mcnt = *p++;
5155: goto matchnotsyntax;
5156:
5157: case notwordchar:
5158: DEBUG_PRINT1 ("EXECUTING Emacs notwordchar.\n");
5159: mcnt = (int) Sword;
5160: matchnotsyntax:
5161: PREFETCH ();
5162: /* Can't use *d++ here; SYNTAX may be an unsafe macro. */
5163: d++;
5164: if (SYNTAX (d[-1]) == (enum syntaxcode) mcnt)
5165: goto fail;
5166: SET_REGS_MATCHED ();
5167: break;
5168:
5169: #else /* not emacs */
5170: case wordchar:
5171: DEBUG_PRINT1 ("EXECUTING non-Emacs wordchar.\n");
5172: PREFETCH ();
5173: if (!WORDCHAR_P (d))
5174: goto fail;
5175: SET_REGS_MATCHED ();
5176: d++;
5177: break;
5178:
5179: case notwordchar:
5180: DEBUG_PRINT1 ("EXECUTING non-Emacs notwordchar.\n");
5181: PREFETCH ();
5182: if (WORDCHAR_P (d))
5183: goto fail;
5184: SET_REGS_MATCHED ();
5185: d++;
5186: break;
5187: #endif /* not emacs */
5188:
5189: default:
5190: abort ();
5191: }
5192: continue; /* Successfully executed one pattern command; keep going. */
5193:
5194:
5195: /* We goto here if a matching operation fails. */
5196: fail:
5197: if (!FAIL_STACK_EMPTY ())
5198: { /* A restart point is known. Restore to that state. */
5199: DEBUG_PRINT1 ("\nFAIL:\n");
5200: POP_FAILURE_POINT (d, p,
5201: lowest_active_reg, highest_active_reg,
5202: regstart, regend, reg_info);
5203:
5204: /* If this failure point is a dummy, try the next one. */
5205: if (!p)
5206: goto fail;
5207:
5208: /* If we failed to the end of the pattern, don't examine *p. */
5209: assert (p <= pend);
5210: if (p < pend)
5211: {
5212: boolean is_a_jump_n = false;
5213:
5214: /* If failed to a backwards jump that's part of a repetition
5215: loop, need to pop this failure point and use the next one. */
5216: switch ((re_opcode_t) *p)
5217: {
5218: case jump_n:
5219: is_a_jump_n = true;
5220: case maybe_pop_jump:
5221: case pop_failure_jump:
5222: case jump:
5223: p1 = p + 1;
5224: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
5225: p1 += mcnt;
5226:
5227: if ((is_a_jump_n && (re_opcode_t) *p1 == succeed_n)
5228: || (!is_a_jump_n
5229: && (re_opcode_t) *p1 == on_failure_jump))
5230: goto fail;
5231: break;
5232: default:
5233: /* do nothing */ ;
5234: }
5235: }
5236:
5237: if (d >= string1 && d <= end1)
5238: dend = end_match_1;
5239: }
5240: else
5241: break; /* Matching at this starting point really fails. */
5242: } /* for (;;) */
5243:
5244: if (best_regs_set)
5245: goto restore_best_regs;
5246:
5247: FREE_VARIABLES ();
5248:
5249: return -1; /* Failure to match. */
5250: } /* re_match_2 */
5251:
5252: /* Subroutine definitions for re_match_2. */
5253:
5254:
5255: /* We are passed P pointing to a register number after a start_memory.
5256:
5257: Return true if the pattern up to the corresponding stop_memory can
5258: match the empty string, and false otherwise.
5259:
5260: If we find the matching stop_memory, sets P to point to one past its number.
5261: Otherwise, sets P to an undefined byte less than or equal to END.
5262:
5263: We don't handle duplicates properly (yet). */
5264:
5265: static boolean
5266: group_match_null_string_p (p, end, reg_info)
5267: unsigned char **p, *end;
5268: register_info_type *reg_info;
5269: {
5270: int mcnt;
5271: /* Point to after the args to the start_memory. */
5272: unsigned char *p1 = *p + 2;
5273:
5274: while (p1 < end)
5275: {
5276: /* Skip over opcodes that can match nothing, and return true or
5277: false, as appropriate, when we get to one that can't, or to the
5278: matching stop_memory. */
5279:
5280: switch ((re_opcode_t) *p1)
5281: {
5282: /* Could be either a loop or a series of alternatives. */
5283: case on_failure_jump:
5284: p1++;
5285: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
5286:
5287: /* If the next operation is not a jump backwards in the
5288: pattern. */
5289:
5290: if (mcnt >= 0)
5291: {
5292: /* Go through the on_failure_jumps of the alternatives,
5293: seeing if any of the alternatives cannot match nothing.
5294: The last alternative starts with only a jump,
5295: whereas the rest start with on_failure_jump and end
5296: with a jump, e.g., here is the pattern for `a|b|c':
5297:
5298: /on_failure_jump/0/6/exactn/1/a/jump_past_alt/0/6
5299: /on_failure_jump/0/6/exactn/1/b/jump_past_alt/0/3
5300: /exactn/1/c
5301:
5302: So, we have to first go through the first (n-1)
5303: alternatives and then deal with the last one separately. */
5304:
5305:
5306: /* Deal with the first (n-1) alternatives, which start
5307: with an on_failure_jump (see above) that jumps to right
5308: past a jump_past_alt. */
5309:
5310: while ((re_opcode_t) p1[mcnt-3] == jump_past_alt)
5311: {
5312: /* `mcnt' holds how many bytes long the alternative
5313: is, including the ending `jump_past_alt' and
5314: its number. */
5315:
5316: if (!alt_match_null_string_p (p1, p1 + mcnt - 3,
5317: reg_info))
5318: return false;
5319:
5320: /* Move to right after this alternative, including the
5321: jump_past_alt. */
5322: p1 += mcnt;
5323:
5324: /* Break if it's the beginning of an n-th alternative
5325: that doesn't begin with an on_failure_jump. */
5326: if ((re_opcode_t) *p1 != on_failure_jump)
5327: break;
5328:
5329: /* Still have to check that it's not an n-th
5330: alternative that starts with an on_failure_jump. */
5331: p1++;
5332: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
5333: if ((re_opcode_t) p1[mcnt-3] != jump_past_alt)
5334: {
5335: /* Get to the beginning of the n-th alternative. */
5336: p1 -= 3;
5337: break;
5338: }
5339: }
5340:
5341: /* Deal with the last alternative: go back and get number
5342: of the `jump_past_alt' just before it. `mcnt' contains
5343: the length of the alternative. */
5344: EXTRACT_NUMBER (mcnt, p1 - 2);
5345:
5346: if (!alt_match_null_string_p (p1, p1 + mcnt, reg_info))
5347: return false;
5348:
5349: p1 += mcnt; /* Get past the n-th alternative. */
5350: } /* if mcnt > 0 */
5351: break;
5352:
5353:
5354: case stop_memory:
5355: assert (p1[1] == **p);
5356: *p = p1 + 2;
5357: return true;
5358:
5359:
5360: default:
5361: if (!common_op_match_null_string_p (&p1, end, reg_info))
5362: return false;
5363: }
5364: } /* while p1 < end */
5365:
5366: return false;
5367: } /* group_match_null_string_p */
5368:
5369:
5370: /* Similar to group_match_null_string_p, but doesn't deal with alternatives:
5371: It expects P to be the first byte of a single alternative and END one
5372: byte past the last. The alternative can contain groups. */
5373:
5374: static boolean
5375: alt_match_null_string_p (p, end, reg_info)
5376: unsigned char *p, *end;
5377: register_info_type *reg_info;
5378: {
5379: int mcnt;
5380: unsigned char *p1 = p;
5381:
5382: while (p1 < end)
5383: {
5384: /* Skip over opcodes that can match nothing, and break when we get
5385: to one that can't. */
5386:
5387: switch ((re_opcode_t) *p1)
5388: {
5389: /* It's a loop. */
5390: case on_failure_jump:
5391: p1++;
5392: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
5393: p1 += mcnt;
5394: break;
5395:
5396: default:
5397: if (!common_op_match_null_string_p (&p1, end, reg_info))
5398: return false;
5399: }
5400: } /* while p1 < end */
5401:
5402: return true;
5403: } /* alt_match_null_string_p */
5404:
5405:
5406: /* Deals with the ops common to group_match_null_string_p and
5407: alt_match_null_string_p.
5408:
5409: Sets P to one after the op and its arguments, if any. */
5410:
5411: static boolean
5412: common_op_match_null_string_p (p, end, reg_info)
5413: unsigned char **p, *end;
5414: register_info_type *reg_info;
5415: {
5416: int mcnt;
5417: boolean ret;
5418: int reg_no;
5419: unsigned char *p1 = *p;
5420:
5421: switch ((re_opcode_t) *p1++)
5422: {
5423: case no_op:
5424: case begline:
5425: case endline:
5426: case begbuf:
5427: case endbuf:
5428: case wordbeg:
5429: case wordend:
5430: case wordbound:
5431: case notwordbound:
5432: #ifdef emacs
5433: case before_dot:
5434: case at_dot:
5435: case after_dot:
5436: #endif
5437: break;
5438:
5439: case start_memory:
5440: reg_no = *p1;
5441: assert (reg_no > 0 && reg_no <= MAX_REGNUM);
5442: ret = group_match_null_string_p (&p1, end, reg_info);
5443:
5444: /* Have to set this here in case we're checking a group which
5445: contains a group and a back reference to it. */
5446:
5447: if (REG_MATCH_NULL_STRING_P (reg_info[reg_no]) == MATCH_NULL_UNSET_VALUE)
5448: REG_MATCH_NULL_STRING_P (reg_info[reg_no]) = ret;
5449:
5450: if (!ret)
5451: return false;
5452: break;
5453:
5454: /* If this is an optimized succeed_n for zero times, make the jump. */
5455: case jump:
5456: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
5457: if (mcnt >= 0)
5458: p1 += mcnt;
5459: else
5460: return false;
5461: break;
5462:
5463: case succeed_n:
5464: /* Get to the number of times to succeed. */
5465: p1 += 2;
5466: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
5467:
5468: if (mcnt == 0)
5469: {
5470: p1 -= 4;
5471: EXTRACT_NUMBER_AND_INCR (mcnt, p1);
5472: p1 += mcnt;
5473: }
5474: else
5475: return false;
5476: break;
5477:
5478: case duplicate:
5479: if (!REG_MATCH_NULL_STRING_P (reg_info[*p1]))
5480: return false;
5481: break;
5482:
5483: case set_number_at:
5484: p1 += 4;
5485:
5486: default:
5487: /* All other opcodes mean we cannot match the empty string. */
5488: return false;
5489: }
5490:
5491: *p = p1;
5492: return true;
5493: } /* common_op_match_null_string_p */
5494:
5495:
5496: /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN
5497: bytes; nonzero otherwise. */
5498:
5499: static int
5500: bcmp_translate (s1, s2, len, translate)
5501: const char *s1, *s2;
5502: register int len;
5503: RE_TRANSLATE_TYPE translate;
5504: {
5505: register const unsigned char *p1 = (const unsigned char *) s1;
5506: register const unsigned char *p2 = (const unsigned char *) s2;
5507: while (len)
5508: {
5509: if (translate[*p1++] != translate[*p2++]) return 1;
5510: len--;
5511: }
5512: return 0;
5513: }
5514:
5515: /* Entry points for GNU code. */
5516:
5517: /* re_compile_pattern is the GNU regular expression compiler: it
5518: compiles PATTERN (of length SIZE) and puts the result in BUFP.
5519: Returns 0 if the pattern was valid, otherwise an error string.
5520:
5521: Assumes the `allocated' (and perhaps `buffer') and `translate' fields
5522: are set in BUFP on entry.
5523:
5524: We call regex_compile to do the actual compilation. */
5525:
5526: const char *
5527: re_compile_pattern (pattern, length, bufp)
5528: const char *pattern;
5529: size_t length;
5530: struct re_pattern_buffer *bufp;
5531: {
5532: reg_errcode_t ret;
5533:
5534: /* GNU code is written to assume at least RE_NREGS registers will be set
5535: (and at least one extra will be -1). */
5536: bufp->regs_allocated = REGS_UNALLOCATED;
5537:
5538: /* And GNU code determines whether or not to get register information
5539: by passing null for the REGS argument to re_match, etc., not by
5540: setting no_sub. */
5541: bufp->no_sub = 0;
5542:
5543: /* Match anchors at newline. */
5544: bufp->newline_anchor = 1;
5545:
5546: ret = regex_compile (pattern, length, re_syntax_options, bufp);
5547:
5548: if (!ret)
5549: return NULL;
5550: return gettext (re_error_msgid + re_error_msgid_idx[(int) ret]);
5551: }
5552: #ifdef _LIBC
5553: weak_alias (__re_compile_pattern, re_compile_pattern)
5554: #endif
5555:
5556: /* Entry points compatible with 4.2 BSD regex library. We don't define
5557: them unless specifically requested. */
5558:
5559: #if defined _REGEX_RE_COMP || defined _LIBC
5560:
5561: /* BSD has one and only one pattern buffer. */
5562: static struct re_pattern_buffer re_comp_buf;
5563:
5564: char *
5565: #ifdef _LIBC
5566: /* Make these definitions weak in libc, so POSIX programs can redefine
5567: these names if they don't use our functions, and still use
5568: regcomp/regexec below without link errors. */
5569: weak_function
5570: #endif
5571: re_comp (s)
5572: const char *s;
5573: {
5574: reg_errcode_t ret;
5575:
5576: if (!s)
5577: {
5578: if (!re_comp_buf.buffer)
5579: return gettext ("No previous regular expression");
5580: return 0;
5581: }
5582:
5583: if (!re_comp_buf.buffer)
5584: {
5585: re_comp_buf.buffer = (unsigned char *) malloc (200);
5586: if (re_comp_buf.buffer == NULL)
5587: return (char *) gettext (re_error_msgid
5588: + re_error_msgid_idx[(int) REG_ESPACE]);
5589: re_comp_buf.allocated = 200;
5590:
5591: re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH);
5592: if (re_comp_buf.fastmap == NULL)
5593: return (char *) gettext (re_error_msgid
5594: + re_error_msgid_idx[(int) REG_ESPACE]);
5595: }
5596:
5597: /* Since `re_exec' always passes NULL for the `regs' argument, we
5598: don't need to initialize the pattern buffer fields which affect it. */
5599:
5600: /* Match anchors at newlines. */
5601: re_comp_buf.newline_anchor = 1;
5602:
5603: ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf);
5604:
5605: if (!ret)
5606: return NULL;
5607:
5608: /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
5609: return (char *) gettext (re_error_msgid + re_error_msgid_idx[(int) ret]);
5610: }
5611:
5612:
5613: int
5614: #ifdef _LIBC
5615: weak_function
5616: #endif
5617: re_exec (s)
5618: const char *s;
5619: {
5620: const int len = strlen (s);
5621: return
5622: 0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0);
5623: }
5624:
5625: #endif /* _REGEX_RE_COMP */
5626:
5627: /* POSIX.2 functions. Don't define these for Emacs. */
5628:
5629: #ifndef emacs
5630:
5631: /* regcomp takes a regular expression as a string and compiles it.
5632:
5633: PREG is a regex_t *. We do not expect any fields to be initialized,
5634: since POSIX says we shouldn't. Thus, we set
5635:
5636: `buffer' to the compiled pattern;
5637: `used' to the length of the compiled pattern;
5638: `syntax' to RE_SYNTAX_POSIX_EXTENDED if the
5639: REG_EXTENDED bit in CFLAGS is set; otherwise, to
5640: RE_SYNTAX_POSIX_BASIC;
5641: `newline_anchor' to REG_NEWLINE being set in CFLAGS;
5642: `fastmap' to an allocated space for the fastmap;
5643: `fastmap_accurate' to zero;
5644: `re_nsub' to the number of subexpressions in PATTERN.
5645:
5646: PATTERN is the address of the pattern string.
5647:
5648: CFLAGS is a series of bits which affect compilation.
5649:
5650: If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we
5651: use POSIX basic syntax.
5652:
5653: If REG_NEWLINE is set, then . and [^...] don't match newline.
5654: Also, regexec will try a match beginning after every newline.
5655:
5656: If REG_ICASE is set, then we considers upper- and lowercase
5657: versions of letters to be equivalent when matching.
5658:
5659: If REG_NOSUB is set, then when PREG is passed to regexec, that
5660: routine will report only success or failure, and nothing about the
5661: registers.
5662:
5663: It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for
5664: the return codes and their meanings.) */
5665:
5666: int
5667: regcomp (preg, pattern, cflags)
5668: regex_t *preg;
5669: const char *pattern;
5670: int cflags;
5671: {
5672: reg_errcode_t ret;
5673: reg_syntax_t syntax
5674: = (cflags & REG_EXTENDED) ?
5675: RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC;
5676:
5677: /* regex_compile will allocate the space for the compiled pattern. */
5678: preg->buffer = 0;
5679: preg->allocated = 0;
5680: preg->used = 0;
5681:
5682: /* Try to allocate space for the fastmap. */
5683: preg->fastmap = (char *) malloc (1 << BYTEWIDTH);
5684:
5685: if (cflags & REG_ICASE)
5686: {
5687: unsigned i;
5688:
5689: preg->translate
5690: = (RE_TRANSLATE_TYPE) malloc (CHAR_SET_SIZE
5691: * sizeof (*(RE_TRANSLATE_TYPE)0));
5692: if (preg->translate == NULL)
5693: return (int) REG_ESPACE;
5694:
5695: /* Map uppercase characters to corresponding lowercase ones. */
5696: for (i = 0; i < CHAR_SET_SIZE; i++)
5697: preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i;
5698: }
5699: else
5700: preg->translate = NULL;
5701:
5702: /* If REG_NEWLINE is set, newlines are treated differently. */
5703: if (cflags & REG_NEWLINE)
5704: { /* REG_NEWLINE implies neither . nor [^...] match newline. */
5705: syntax &= ~RE_DOT_NEWLINE;
5706: syntax |= RE_HAT_LISTS_NOT_NEWLINE;
5707: /* It also changes the matching behavior. */
5708: preg->newline_anchor = 1;
5709: }
5710: else
5711: preg->newline_anchor = 0;
5712:
5713: preg->no_sub = !!(cflags & REG_NOSUB);
5714:
5715: /* POSIX says a null character in the pattern terminates it, so we
5716: can use strlen here in compiling the pattern. */
5717: ret = regex_compile (pattern, strlen (pattern), syntax, preg);
5718:
5719: /* POSIX doesn't distinguish between an unmatched open-group and an
5720: unmatched close-group: both are REG_EPAREN. */
5721: if (ret == REG_ERPAREN) ret = REG_EPAREN;
5722:
5723: if (ret == REG_NOERROR && preg->fastmap)
5724: {
5725: /* Compute the fastmap now, since regexec cannot modify the pattern
5726: buffer. */
5727: if (re_compile_fastmap (preg) == -2)
5728: {
5729: /* Some error occured while computing the fastmap, just forget
5730: about it. */
5731: free (preg->fastmap);
5732: preg->fastmap = NULL;
5733: }
5734: }
5735:
5736: return (int) ret;
5737: }
5738: #ifdef _LIBC
5739: weak_alias (__regcomp, regcomp)
5740: #endif
5741:
5742:
5743: /* regexec searches for a given pattern, specified by PREG, in the
5744: string STRING.
5745:
5746: If NMATCH is zero or REG_NOSUB was set in the cflags argument to
5747: `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at
5748: least NMATCH elements, and we set them to the offsets of the
5749: corresponding matched substrings.
5750:
5751: EFLAGS specifies `execution flags' which affect matching: if
5752: REG_NOTBOL is set, then ^ does not match at the beginning of the
5753: string; if REG_NOTEOL is set, then $ does not match at the end.
5754:
5755: We return 0 if we find a match and REG_NOMATCH if not. */
5756:
5757: int
5758: regexec (preg, string, nmatch, pmatch, eflags)
5759: const regex_t *preg;
5760: const char *string;
5761: size_t nmatch;
5762: regmatch_t pmatch[];
5763: int eflags;
5764: {
5765: int ret;
5766: struct re_registers regs;
5767: regex_t private_preg;
5768: int len = strlen (string);
5769: boolean want_reg_info = !preg->no_sub && nmatch > 0;
5770:
5771: private_preg = *preg;
5772:
5773: private_preg.not_bol = !!(eflags & REG_NOTBOL);
5774: private_preg.not_eol = !!(eflags & REG_NOTEOL);
5775:
5776: /* The user has told us exactly how many registers to return
5777: information about, via `nmatch'. We have to pass that on to the
5778: matching routines. */
5779: private_preg.regs_allocated = REGS_FIXED;
5780:
5781: if (want_reg_info)
5782: {
5783: regs.num_regs = nmatch;
5784: regs.start = TALLOC (nmatch * 2, regoff_t);
5785: if (regs.start == NULL)
5786: return (int) REG_NOMATCH;
5787: regs.end = regs.start + nmatch;
5788: }
5789:
5790: /* Perform the searching operation. */
5791: ret = re_search (&private_preg, string, len,
5792: /* start: */ 0, /* range: */ len,
5793: want_reg_info ? ®s : (struct re_registers *) 0);
5794:
5795: /* Copy the register information to the POSIX structure. */
5796: if (want_reg_info)
5797: {
5798: if (ret >= 0)
5799: {
5800: unsigned r;
5801:
5802: for (r = 0; r < nmatch; r++)
5803: {
5804: pmatch[r].rm_so = regs.start[r];
5805: pmatch[r].rm_eo = regs.end[r];
5806: }
5807: }
5808:
5809: /* If we needed the temporary register info, free the space now. */
5810: free (regs.start);
5811: }
5812:
5813: /* We want zero return to mean success, unlike `re_search'. */
5814: return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH;
5815: }
5816: #ifdef _LIBC
5817: weak_alias (__regexec, regexec)
5818: #endif
5819:
5820:
5821: /* Returns a message corresponding to an error code, ERRCODE, returned
5822: from either regcomp or regexec. We don't use PREG here. */
5823:
5824: size_t
5825: regerror (err, preg, errbuf, errbuf_size)
5826: int err;
5827: const regex_t *preg;
5828: char *errbuf;
5829: size_t errbuf_size;
5830: {
5831: const char *msg;
5832: size_t msg_size;
5833:
5834: if (err < 0
5835: || err >= (int) (sizeof (re_error_msgid_idx)
5836: / sizeof (re_error_msgid_idx[0])))
5837: /* Only error codes returned by the rest of the code should be passed
5838: to this routine. If we are given anything else, or if other regex
5839: code generates an invalid error code, then the program has a bug.
5840: Dump core so we can fix it. */
5841: abort ();
5842:
5843: msg = gettext (re_error_msgid + re_error_msgid_idx[err]);
5844:
5845: msg_size = strlen (msg) + 1; /* Includes the null. */
5846:
5847: if (errbuf_size != 0)
5848: {
5849: if (msg_size > errbuf_size)
5850: {
5851: #if defined HAVE_MEMPCPY || defined _LIBC
5852: *((char *) __mempcpy (errbuf, msg, errbuf_size - 1)) = '\0';
5853: #else
5854: memcpy (errbuf, msg, errbuf_size - 1);
5855: errbuf[errbuf_size - 1] = 0;
5856: #endif
5857: }
5858: else
5859: memcpy (errbuf, msg, msg_size);
5860: }
5861:
5862: return msg_size;
5863: }
5864: #ifdef _LIBC
5865: weak_alias (__regerror, regerror)
5866: #endif
5867:
5868:
5869: /* Free dynamically allocated space used by PREG. */
5870:
5871: void
5872: regfree (preg)
5873: regex_t *preg;
5874: {
5875: if (preg->buffer != NULL)
5876: free (preg->buffer);
5877: preg->buffer = NULL;
5878:
5879: preg->allocated = 0;
5880: preg->used = 0;
5881:
5882: if (preg->fastmap != NULL)
5883: free (preg->fastmap);
5884: preg->fastmap = NULL;
5885: preg->fastmap_accurate = 0;
5886:
5887: if (preg->translate != NULL)
5888: free (preg->translate);
5889: preg->translate = NULL;
5890: }
5891: #ifdef _LIBC
5892: weak_alias (__regfree, regfree)
5893: #endif
5894:
5895: #endif /* not emacs */
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to regex.c CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [ELWIX - Embedded LightWeight unIX -] / embedaddon / quagga / lib