Mercurial > hg > xemacs-beta
annotate src/regex.c @ 5650:7fa8667cdaa7
Imitate GNU Emacs API for `batch-byte-recompile-directory'.
2012-04-23 Michael Sperber <mike@xemacs.org>
* bytecomp.el (batch-byte-recompile-directory): Accept an optional
argument that's passed on to `byte-recompile-directory' as the
prefix argument, thus imitating GNU Emacs's API.
| author | Mike Sperber <sperber@deinprogramm.de> |
|---|---|
| date | Mon, 23 Apr 2012 10:06:39 +0200 |
| parents | 3f4a234f4672 |
| children | 3df910176b6a |
| rev | line source |
|---|---|
| 428 | 1 /* Extended regular expression matching and search library, |
| 2 version 0.12, extended for XEmacs. | |
| 3 (Implements POSIX draft P10003.2/D11.2, except for | |
| 4 internationalization features.) | |
| 5 | |
| 6 Copyright (C) 1993, 1994, 1995 Free Software Foundation, Inc. | |
| 7 Copyright (C) 1995 Sun Microsystems, Inc. | |
| 5041 | 8 Copyright (C) 1995, 2001, 2002, 2003, 2010 Ben Wing. |
| 428 | 9 |
|
5402
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
10 This file is part of XEmacs. |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
11 |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
12 XEmacs is free software: you can redistribute it and/or modify it |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
13 under the terms of the GNU General Public License as published by the |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
14 Free Software Foundation, either version 3 of the License, or (at your |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
15 option) any later version. |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
16 |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
17 XEmacs is distributed in the hope that it will be useful, but WITHOUT |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
18 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
19 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
20 for more details. |
|
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
21 |
| 428 | 22 You should have received a copy of the GNU General Public License |
|
5402
308d34e9f07d
Changed bulk of GPLv2 or later files identified by script
Mats Lidell <matsl@xemacs.org>
parents:
5041
diff
changeset
|
23 along with XEmacs. If not, see <http://www.gnu.org/licenses/>. */ |
| 428 | 24 /* Synched up with: FSF 19.29. */ |
| 25 | |
| 26 #ifdef HAVE_CONFIG_H | |
| 27 #include <config.h> | |
| 28 #endif | |
| 29 | |
| 30 #ifndef _GNU_SOURCE | |
| 31 #define _GNU_SOURCE 1 | |
| 32 #endif | |
| 33 | |
| 34 /* We assume non-Mule if emacs isn't defined. */ | |
| 35 #ifndef emacs | |
| 36 #undef MULE | |
| 37 #endif | |
| 38 | |
| 771 | 39 /* XEmacs addition */ |
| 40 #ifdef REL_ALLOC | |
| 41 #define REGEX_REL_ALLOC /* may be undefined below */ | |
| 42 #endif | |
| 43 | |
| 428 | 44 /* XEmacs: define this to add in a speedup for patterns anchored at |
| 45 the beginning of a line. Keep the ifdefs so that it's easier to | |
| 46 tell where/why this code has diverged from v19. */ | |
| 47 #define REGEX_BEGLINE_CHECK | |
| 48 | |
| 49 /* XEmacs: the current mmap-based ralloc handles small blocks very | |
| 50 poorly, so we disable it here. */ | |
| 51 | |
| 771 | 52 #if defined (HAVE_MMAP) || defined (DOUG_LEA_MALLOC) |
| 53 # undef REGEX_REL_ALLOC | |
| 428 | 54 #endif |
| 55 | |
| 56 /* The `emacs' switch turns on certain matching commands | |
| 57 that make sense only in Emacs. */ | |
| 58 #ifdef emacs | |
| 59 | |
| 60 #include "lisp.h" | |
| 61 #include "buffer.h" | |
| 62 #include "syntax.h" | |
| 63 | |
| 64 #if (defined (DEBUG_XEMACS) && !defined (DEBUG)) | |
| 65 #define DEBUG | |
| 66 #endif | |
| 67 | |
| 867 | 68 #define RE_TRANSLATE_1(ch) TRT_TABLE_OF (translate, (Ichar) ch) |
| 446 | 69 #define TRANSLATE_P(tr) (!NILP (tr)) |
| 428 | 70 |
| 826 | 71 /* Converts the pointer to the char to BEG-based offset from the start. */ |
| 72 #define PTR_TO_OFFSET(d) (MATCHING_IN_FIRST_STRING \ | |
| 73 ? (d) - string1 : (d) - (string2 - size1)) | |
| 74 | |
| 428 | 75 #else /* not emacs */ |
| 76 | |
| 2367 | 77 #include <stdlib.h> |
| 78 #include <sys/types.h> | |
| 79 #include <stddef.h> /* needed for ptrdiff_t under Solaris */ | |
| 80 #include <string.h> | |
| 81 | |
| 2286 | 82 #include "compiler.h" /* Get compiler-specific definitions like UNUSED */ |
| 83 | |
| 2500 | 84 #define ABORT abort |
| 85 | |
| 428 | 86 /* If we are not linking with Emacs proper, |
| 87 we can't use the relocating allocator | |
| 88 even if config.h says that we can. */ | |
| 771 | 89 #undef REGEX_REL_ALLOC |
| 428 | 90 |
| 544 | 91 /* defined in lisp.h */ |
| 92 #ifdef REGEX_MALLOC | |
| 93 #ifndef DECLARE_NOTHING | |
| 94 #define DECLARE_NOTHING struct nosuchstruct | |
| 95 #endif | |
| 96 #endif | |
| 97 | |
| 867 | 98 #define itext_ichar(str) ((Ichar) (str)[0]) |
| 99 #define itext_ichar_fmt(str, fmt, object) ((Ichar) (str)[0]) | |
| 100 #define itext_ichar_ascii_fmt(str, fmt, object) ((Ichar) (str)[0]) | |
| 428 | 101 |
| 102 #if (LONGBITS > INTBITS) | |
| 103 # define EMACS_INT long | |
| 104 #else | |
| 105 # define EMACS_INT int | |
| 106 #endif | |
| 107 | |
| 867 | 108 typedef int Ichar; |
| 109 | |
| 110 #define INC_IBYTEPTR(p) ((p)++) | |
| 111 #define INC_IBYTEPTR_FMT(p, fmt) ((p)++) | |
| 112 #define DEC_IBYTEPTR(p) ((p)--) | |
| 113 #define DEC_IBYTEPTR_FMT(p, fmt) ((p)--) | |
|
4750
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
114 #define MAX_ICHAR_LEN 1 |
| 867 | 115 #define itext_ichar_len(ptr) 1 |
| 116 #define itext_ichar_len_fmt(ptr, fmt) 1 | |
| 428 | 117 |
| 118 /* Define the syntax stuff for \<, \>, etc. */ | |
| 119 | |
| 120 /* This must be nonzero for the wordchar and notwordchar pattern | |
| 121 commands in re_match_2. */ | |
| 122 #ifndef Sword | |
| 123 #define Sword 1 | |
| 124 #endif | |
| 125 | |
| 126 #ifdef SYNTAX_TABLE | |
| 127 | |
| 128 extern char *re_syntax_table; | |
| 129 | |
| 130 #else /* not SYNTAX_TABLE */ | |
| 131 | |
| 132 /* How many characters in the character set. */ | |
| 133 #define CHAR_SET_SIZE 256 | |
| 134 | |
| 135 static char re_syntax_table[CHAR_SET_SIZE]; | |
| 136 | |
| 137 static void | |
| 138 init_syntax_once (void) | |
| 139 { | |
| 140 static int done = 0; | |
| 141 | |
| 142 if (!done) | |
| 143 { | |
| 442 | 144 const char *word_syntax_chars = |
| 428 | 145 "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789_"; |
| 146 | |
| 147 memset (re_syntax_table, 0, sizeof (re_syntax_table)); | |
| 148 | |
| 149 while (*word_syntax_chars) | |
| 647 | 150 re_syntax_table[(unsigned int) (*word_syntax_chars++)] = Sword; |
| 428 | 151 |
| 152 done = 1; | |
| 153 } | |
| 154 } | |
| 155 | |
| 446 | 156 #endif /* SYNTAX_TABLE */ |
| 428 | 157 |
| 826 | 158 #define SYNTAX(ignored, c) re_syntax_table[c] |
| 460 | 159 #undef SYNTAX_FROM_CACHE |
| 826 | 160 #define SYNTAX_FROM_CACHE SYNTAX |
| 161 | |
| 162 #define RE_TRANSLATE_1(c) translate[(unsigned char) (c)] | |
| 446 | 163 #define TRANSLATE_P(tr) tr |
| 164 | |
| 165 #endif /* emacs */ | |
| 428 | 166 |
| 2201 | 167 /* This is for other GNU distributions with internationalized messages. */ |
| 168 #if defined (I18N3) && (defined (HAVE_LIBINTL_H) || defined (_LIBC)) | |
| 169 # include <libintl.h> | |
| 170 #else | |
| 171 # define gettext(msgid) (msgid) | |
| 172 #endif | |
| 173 | |
| 428 | 174 |
| 175 /* Get the interface, including the syntax bits. */ | |
| 176 #include "regex.h" | |
| 177 | |
| 178 /* isalpha etc. are used for the character classes. */ | |
| 179 #include <ctype.h> | |
| 180 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
181 #ifdef emacs |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
182 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
183 /* 1 if C is an ASCII character. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
184 #define ISASCII(c) ((c) < 0x80) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
185 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
186 /* 1 if C is a unibyte character. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
187 #define ISUNIBYTE(c) 0 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
188 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
189 /* The Emacs definitions should not be directly affected by locales. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
190 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
191 /* In Emacs, these are only used for single-byte characters. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
192 #define ISDIGIT(c) ((c) >= '0' && (c) <= '9') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
193 #define ISCNTRL(c) ((c) < ' ') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
194 #define ISXDIGIT(c) (ISDIGIT (c) || ((c) >= 'a' && (c) <= 'f') \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
195 || ((c) >= 'A' && (c) <= 'F')) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
196 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
197 /* This is only used for single-byte characters. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
198 #define ISBLANK(c) ((c) == ' ' || (c) == '\t') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
199 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
200 /* The rest must handle multibyte characters. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
201 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
202 #define ISGRAPH(c) ((c) > ' ' && (c) != 0x7f) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
203 #define ISPRINT(c) ((c) == ' ' || ISGRAPH (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
204 #define ISALPHA(c) (ISASCII (c) ? (((c) >= 'a' && (c) <= 'z') \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
205 || ((c) >= 'A' && (c) <= 'Z')) \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
206 : ISWORD (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
207 #define ISALNUM(c) (ISALPHA (c) || ISDIGIT (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
208 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
209 #define ISLOWER(c) LOWERCASEP (lispbuf, c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
210 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
211 #define ISPUNCT(c) (ISASCII (c) \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
212 ? ((c) > ' ' && (c) < 0x7F \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
213 && !(((c) >= 'a' && (c) <= 'z') \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
214 || ((c) >= 'A' && (c) <= 'Z') \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
215 || ((c) >= '0' && (c) <= '9'))) \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
216 : !ISWORD (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
217 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
218 #define ISSPACE(c) \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
219 (SYNTAX (BUFFER_MIRROR_SYNTAX_TABLE (lispbuf), c) == Swhitespace) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
220 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
221 #define ISUPPER(c) UPPERCASEP (lispbuf, c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
222 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
223 #define ISWORD(c) (SYNTAX (BUFFER_MIRROR_SYNTAX_TABLE (lispbuf), c) == Sword) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
224 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
225 #else /* not emacs */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
226 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
227 /* 1 if C is an ASCII character. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
228 #define ISASCII(c) ((c) < 0200) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
229 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
230 /* 1 if C is a unibyte character. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
231 #define ISUNIBYTE(c) 0 |
| 428 | 232 |
| 233 #ifdef isblank | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
234 # define ISBLANK(c) isblank (c) |
| 428 | 235 #else |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
236 # define ISBLANK(c) ((c) == ' ' || (c) == '\t') |
| 428 | 237 #endif |
| 238 #ifdef isgraph | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
239 # define ISGRAPH(c) isgraph (c) |
| 428 | 240 #else |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
241 # define ISGRAPH(c) (isprint (c) && !isspace (c)) |
| 428 | 242 #endif |
| 243 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
244 /* Solaris defines ISPRINT so we must undefine it first. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
245 #undef ISPRINT |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
246 #define ISPRINT(c) isprint (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
247 #define ISDIGIT(c) isdigit (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
248 #define ISALNUM(c) isalnum (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
249 #define ISALPHA(c) isalpha (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
250 #define ISCNTRL(c) iscntrl (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
251 #define ISLOWER(c) islower (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
252 #define ISPUNCT(c) ispunct (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
253 #define ISSPACE(c) isspace (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
254 #define ISUPPER(c) isupper (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
255 #define ISXDIGIT(c) isxdigit (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
256 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
257 #define ISWORD(c) ISALPHA (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
258 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
259 #ifdef _tolower |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
260 # define TOLOWER(c) _tolower (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
261 #else |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
262 # define TOLOWER(c) tolower (c) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
263 #endif |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
264 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
265 #endif /* emacs */ |
| 428 | 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 | |
| 1333 | 293 #ifndef emacs |
| 294 #define ALLOCA alloca | |
| 295 #define xmalloc malloc | |
| 296 #define xrealloc realloc | |
|
4976
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
297 #define xfree free |
| 1333 | 298 #endif |
| 299 | |
| 300 #ifdef emacs | |
| 301 #define ALLOCA_GARBAGE_COLLECT() \ | |
| 302 do \ | |
| 303 { \ | |
| 304 if (need_to_check_c_alloca) \ | |
| 305 xemacs_c_alloca (0); \ | |
| 306 } while (0) | |
| 307 #elif defined (C_ALLOCA) | |
| 308 #define ALLOCA_GARBAGE_COLLECT() alloca (0) | |
| 309 #else | |
| 310 #define ALLOCA_GARBAGE_COLLECT() | |
| 311 #endif | |
| 312 | |
| 313 #ifndef emacs | |
| 314 /* So we can use just it to conditionalize on */ | |
| 315 #undef ERROR_CHECK_MALLOC | |
| 316 #endif | |
| 317 | |
| 318 #ifdef ERROR_CHECK_MALLOC | |
| 319 /* When REL_ALLOC, malloc() is problematic because it could potentially | |
| 320 cause all rel-alloc()ed data -- including buffer text -- to be relocated. | |
| 321 We deal with this by checking for such relocation whenever we have | |
| 322 executed a statement that may call malloc() -- or alloca(), which may | |
| 323 end up calling malloc() in some circumstances -- and recomputing all | |
| 324 of our string pointers in re_match_2_internal() and re_search_2(). | |
| 325 However, if malloc() or alloca() happens and we don't know about it, | |
| 326 we could still be screwed. So we set up a system where we indicate all | |
| 327 places where we are prepared for malloc() or alloca(), and in any | |
| 328 other circumstances, calls to those functions (from anywhere inside of | |
| 2500 | 329 XEmacs!) will ABORT(). We do this even when REL_ALLOC is not defined |
| 1333 | 330 so that we catch these problems sooner, since many developers and beta |
| 331 testers will not be running with REL_ALLOC. */ | |
| 332 int regex_malloc_disallowed; | |
| 333 #define BEGIN_REGEX_MALLOC_OK() regex_malloc_disallowed = 0 | |
| 334 #define END_REGEX_MALLOC_OK() regex_malloc_disallowed = 1 | |
| 335 #define UNBIND_REGEX_MALLOC_CHECK() unbind_to (depth) | |
| 336 #else | |
| 337 #define BEGIN_REGEX_MALLOC_OK() | |
| 338 #define END_REGEX_MALLOC_OK() | |
| 339 #define UNBIND_REGEX_MALLOC_CHECK() | |
| 340 #endif | |
| 341 | |
| 342 | |
| 428 | 343 #ifdef REGEX_MALLOC |
| 344 | |
| 1333 | 345 #define REGEX_ALLOCATE xmalloc |
| 346 #define REGEX_REALLOCATE(source, osize, nsize) xrealloc (source, nsize) | |
| 347 #define REGEX_FREE xfree | |
| 428 | 348 |
| 349 #else /* not REGEX_MALLOC */ | |
| 350 | |
| 351 /* Emacs already defines alloca, sometimes. */ | |
| 352 #ifndef alloca | |
| 353 | |
| 354 /* Make alloca work the best possible way. */ | |
| 355 #ifdef __GNUC__ | |
| 356 #define alloca __builtin_alloca | |
| 771 | 357 #elif defined (__DECC) /* XEmacs: added next 3 lines, similar to config.h.in */ |
| 358 #include <alloca.h> | |
| 359 #pragma intrinsic(alloca) | |
| 428 | 360 #else /* not __GNUC__ */ |
| 361 #if HAVE_ALLOCA_H | |
| 362 #include <alloca.h> | |
| 363 #else /* not __GNUC__ or HAVE_ALLOCA_H */ | |
| 364 #ifndef _AIX /* Already did AIX, up at the top. */ | |
| 444 | 365 void *alloca (); |
| 428 | 366 #endif /* not _AIX */ |
| 446 | 367 #endif /* HAVE_ALLOCA_H */ |
| 368 #endif /* __GNUC__ */ | |
| 428 | 369 |
| 370 #endif /* not alloca */ | |
| 371 | |
| 1333 | 372 #define REGEX_ALLOCATE ALLOCA |
| 428 | 373 |
| 2367 | 374 /* !!#### Needs review */ |
| 428 | 375 /* Assumes a `char *destination' variable. */ |
| 376 #define REGEX_REALLOCATE(source, osize, nsize) \ | |
| 1333 | 377 (destination = (char *) ALLOCA (nsize), \ |
| 428 | 378 memmove (destination, source, osize), \ |
| 379 destination) | |
| 380 | |
| 1726 | 381 /* No need to do anything to free, after alloca. |
| 382 Do nothing! But inhibit gcc warning. */ | |
| 383 #define REGEX_FREE(arg,type) ((void)0) | |
| 428 | 384 |
| 446 | 385 #endif /* REGEX_MALLOC */ |
| 428 | 386 |
| 387 /* Define how to allocate the failure stack. */ | |
| 388 | |
| 771 | 389 #ifdef REGEX_REL_ALLOC |
| 428 | 390 #define REGEX_ALLOCATE_STACK(size) \ |
| 1346 | 391 r_alloc ((unsigned char **) &failure_stack_ptr, (size)) |
| 428 | 392 #define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
| 1346 | 393 r_re_alloc ((unsigned char **) &failure_stack_ptr, (nsize)) |
| 428 | 394 #define REGEX_FREE_STACK(ptr) \ |
| 1346 | 395 r_alloc_free ((unsigned char **) &failure_stack_ptr) |
| 428 | 396 |
| 771 | 397 #else /* not REGEX_REL_ALLOC */ |
| 428 | 398 |
| 399 #ifdef REGEX_MALLOC | |
| 400 | |
| 1333 | 401 #define REGEX_ALLOCATE_STACK xmalloc |
| 402 #define REGEX_REALLOCATE_STACK(source, osize, nsize) xrealloc (source, nsize) | |
|
4976
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
403 #define REGEX_FREE_STACK(arg) xfree (arg) |
| 428 | 404 |
| 405 #else /* not REGEX_MALLOC */ | |
| 406 | |
| 1333 | 407 #define REGEX_ALLOCATE_STACK ALLOCA |
| 428 | 408 |
| 409 #define REGEX_REALLOCATE_STACK(source, osize, nsize) \ | |
| 410 REGEX_REALLOCATE (source, osize, nsize) | |
| 411 /* No need to explicitly free anything. */ | |
| 412 #define REGEX_FREE_STACK(arg) | |
| 413 | |
| 446 | 414 #endif /* REGEX_MALLOC */ |
| 771 | 415 #endif /* REGEX_REL_ALLOC */ |
| 428 | 416 |
| 417 | |
| 418 /* True if `size1' is non-NULL and PTR is pointing anywhere inside | |
| 419 `string1' or just past its end. This works if PTR is NULL, which is | |
| 420 a good thing. */ | |
| 421 #define FIRST_STRING_P(ptr) \ | |
| 422 (size1 && string1 <= (ptr) && (ptr) <= string1 + size1) | |
| 423 | |
| 424 /* (Re)Allocate N items of type T using malloc, or fail. */ | |
| 1333 | 425 #define TALLOC(n, t) ((t *) xmalloc ((n) * sizeof (t))) |
| 426 #define RETALLOC(addr, n, t) ((addr) = (t *) xrealloc (addr, (n) * sizeof (t))) | |
| 428 | 427 #define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t))) |
| 428 | |
| 429 #define BYTEWIDTH 8 /* In bits. */ | |
| 430 | |
| 434 | 431 #define STREQ(s1, s2) (strcmp (s1, s2) == 0) |
| 428 | 432 |
| 433 #undef MAX | |
| 434 #undef MIN | |
| 435 #define MAX(a, b) ((a) > (b) ? (a) : (b)) | |
| 436 #define MIN(a, b) ((a) < (b) ? (a) : (b)) | |
| 437 | |
| 446 | 438 /* Type of source-pattern and string chars. */ |
| 439 typedef const unsigned char re_char; | |
| 440 | |
| 460 | 441 typedef char re_bool; |
| 428 | 442 #define false 0 |
| 443 #define true 1 | |
| 444 | |
| 445 | |
| 1346 | 446 #ifdef emacs |
| 447 | |
| 448 #ifdef MULE | |
| 449 | |
| 450 Lisp_Object Vthe_lisp_rangetab; | |
| 451 | |
| 452 void | |
| 453 vars_of_regex (void) | |
| 454 { | |
| 2421 | 455 Vthe_lisp_rangetab = Fmake_range_table (Qstart_closed_end_closed); |
| 1346 | 456 staticpro (&Vthe_lisp_rangetab); |
| 457 } | |
| 458 | |
| 459 #else /* not MULE */ | |
| 460 | |
| 461 void | |
| 462 vars_of_regex (void) | |
| 463 { | |
| 464 } | |
| 465 | |
| 466 #endif /* MULE */ | |
| 467 | |
| 468 /* Convert an offset from the start of the logical text string formed by | |
| 469 concatenating the two strings together into a character position in the | |
| 470 Lisp buffer or string that the text represents. Knows that | |
| 471 when handling buffer text, the "string" we're passed in is always | |
| 472 BEGV - ZV. */ | |
| 473 | |
| 474 static Charxpos | |
| 475 offset_to_charxpos (Lisp_Object lispobj, int off) | |
| 476 { | |
| 477 if (STRINGP (lispobj)) | |
| 478 return string_index_byte_to_char (lispobj, off); | |
| 479 else if (BUFFERP (lispobj)) | |
| 480 return bytebpos_to_charbpos (XBUFFER (lispobj), | |
| 481 off + BYTE_BUF_BEGV (XBUFFER (lispobj))); | |
| 482 else | |
| 483 return 0; | |
| 484 } | |
| 485 | |
| 486 #ifdef REL_ALLOC | |
| 487 | |
| 488 /* STRING1 is the value of STRING1 given to re_match_2(). LISPOBJ is | |
| 489 the Lisp object (if any) from which the string is taken. If LISPOBJ | |
| 490 is a buffer, return a relocation offset to be added to all pointers to | |
| 491 string data so that they will be accurate again, after an allocation or | |
| 492 reallocation that potentially relocated the buffer data. | |
| 493 */ | |
| 494 static Bytecount | |
| 495 offset_post_relocation (Lisp_Object lispobj, Ibyte *orig_buftext) | |
| 496 { | |
| 497 if (!BUFFERP (lispobj)) | |
| 498 return 0; | |
| 499 return (BYTE_BUF_BYTE_ADDRESS (XBUFFER (lispobj), | |
| 500 BYTE_BUF_BEGV (XBUFFER (lispobj))) - | |
| 501 orig_buftext); | |
| 502 } | |
| 503 | |
| 504 #endif /* REL_ALLOC */ | |
| 505 | |
| 506 #ifdef ERROR_CHECK_MALLOC | |
| 507 | |
| 508 /* NOTE that this can run malloc() so you need to adjust afterwards. */ | |
| 509 | |
| 510 static int | |
| 511 bind_regex_malloc_disallowed (int value) | |
| 512 { | |
| 513 /* Tricky, because the act of binding can run malloc(). */ | |
| 514 int old_regex_malloc_disallowed = regex_malloc_disallowed; | |
| 515 int depth; | |
| 516 regex_malloc_disallowed = 0; | |
| 517 depth = record_unwind_protect_restoring_int (®ex_malloc_disallowed, | |
| 518 old_regex_malloc_disallowed); | |
| 519 regex_malloc_disallowed = value; | |
| 520 return depth; | |
| 521 } | |
| 522 | |
| 523 #endif /* ERROR_CHECK_MALLOC */ | |
| 524 | |
| 525 #endif /* emacs */ | |
| 526 | |
| 527 | |
| 428 | 528 /* These are the command codes that appear in compiled regular |
| 529 expressions. Some opcodes are followed by argument bytes. A | |
| 530 command code can specify any interpretation whatsoever for its | |
| 531 arguments. Zero bytes may appear in the compiled regular expression. */ | |
| 532 | |
| 533 typedef enum | |
| 534 { | |
| 535 no_op = 0, | |
| 536 | |
| 537 /* Succeed right away--no more backtracking. */ | |
| 538 succeed, | |
| 539 | |
| 540 /* Followed by one byte giving n, then by n literal bytes. */ | |
| 541 exactn, | |
| 542 | |
| 543 /* Matches any (more or less) character. */ | |
| 544 anychar, | |
| 545 | |
| 546 /* Matches any one char belonging to specified set. First | |
| 547 following byte is number of bitmap bytes. Then come bytes | |
| 548 for a bitmap saying which chars are in. Bits in each byte | |
| 549 are ordered low-bit-first. A character is in the set if its | |
| 550 bit is 1. A character too large to have a bit in the map is | |
| 551 automatically not in the set. */ | |
| 552 charset, | |
| 553 | |
| 554 /* Same parameters as charset, but match any character that is | |
| 555 not one of those specified. */ | |
| 556 charset_not, | |
| 557 | |
| 558 /* Start remembering the text that is matched, for storing in a | |
| 559 register. Followed by one byte with the register number, in | |
| 502 | 560 the range 1 to the pattern buffer's re_ngroups |
| 428 | 561 field. Then followed by one byte with the number of groups |
| 562 inner to this one. (This last has to be part of the | |
| 563 start_memory only because we need it in the on_failure_jump | |
| 564 of re_match_2.) */ | |
| 565 start_memory, | |
| 566 | |
| 567 /* Stop remembering the text that is matched and store it in a | |
| 568 memory register. Followed by one byte with the register | |
| 502 | 569 number, in the range 1 to `re_ngroups' in the |
| 428 | 570 pattern buffer, and one byte with the number of inner groups, |
| 571 just like `start_memory'. (We need the number of inner | |
| 572 groups here because we don't have any easy way of finding the | |
| 573 corresponding start_memory when we're at a stop_memory.) */ | |
| 574 stop_memory, | |
| 575 | |
| 576 /* Match a duplicate of something remembered. Followed by one | |
| 577 byte containing the register number. */ | |
| 578 duplicate, | |
| 579 | |
| 580 /* Fail unless at beginning of line. */ | |
| 581 begline, | |
| 582 | |
| 583 /* Fail unless at end of line. */ | |
| 584 endline, | |
| 585 | |
| 586 /* Succeeds if at beginning of buffer (if emacs) or at beginning | |
| 587 of string to be matched (if not). */ | |
| 588 begbuf, | |
| 589 | |
| 590 /* Analogously, for end of buffer/string. */ | |
| 591 endbuf, | |
| 592 | |
| 593 /* Followed by two byte relative address to which to jump. */ | |
| 594 jump, | |
| 595 | |
| 596 /* Same as jump, but marks the end of an alternative. */ | |
| 597 jump_past_alt, | |
| 598 | |
| 599 /* Followed by two-byte relative address of place to resume at | |
| 600 in case of failure. */ | |
| 601 on_failure_jump, | |
| 602 | |
| 603 /* Like on_failure_jump, but pushes a placeholder instead of the | |
| 604 current string position when executed. */ | |
| 605 on_failure_keep_string_jump, | |
| 606 | |
| 607 /* Throw away latest failure point and then jump to following | |
| 608 two-byte relative address. */ | |
| 609 pop_failure_jump, | |
| 610 | |
| 611 /* Change to pop_failure_jump if know won't have to backtrack to | |
| 612 match; otherwise change to jump. This is used to jump | |
| 613 back to the beginning of a repeat. If what follows this jump | |
| 614 clearly won't match what the repeat does, such that we can be | |
| 615 sure that there is no use backtracking out of repetitions | |
| 616 already matched, then we change it to a pop_failure_jump. | |
| 617 Followed by two-byte address. */ | |
| 618 maybe_pop_jump, | |
| 619 | |
| 620 /* Jump to following two-byte address, and push a dummy failure | |
| 621 point. This failure point will be thrown away if an attempt | |
| 622 is made to use it for a failure. A `+' construct makes this | |
| 623 before the first repeat. Also used as an intermediary kind | |
| 624 of jump when compiling an alternative. */ | |
| 625 dummy_failure_jump, | |
| 626 | |
| 627 /* Push a dummy failure point and continue. Used at the end of | |
| 628 alternatives. */ | |
| 629 push_dummy_failure, | |
| 630 | |
| 631 /* Followed by two-byte relative address and two-byte number n. | |
| 632 After matching N times, jump to the address upon failure. */ | |
| 633 succeed_n, | |
| 634 | |
| 635 /* Followed by two-byte relative address, and two-byte number n. | |
| 636 Jump to the address N times, then fail. */ | |
| 637 jump_n, | |
| 638 | |
| 639 /* Set the following two-byte relative address to the | |
| 640 subsequent two-byte number. The address *includes* the two | |
| 641 bytes of number. */ | |
| 642 set_number_at, | |
| 643 | |
| 644 wordchar, /* Matches any word-constituent character. */ | |
| 645 notwordchar, /* Matches any char that is not a word-constituent. */ | |
| 646 | |
| 647 wordbeg, /* Succeeds if at word beginning. */ | |
| 648 wordend, /* Succeeds if at word end. */ | |
| 649 | |
| 650 wordbound, /* Succeeds if at a word boundary. */ | |
| 651 notwordbound /* Succeeds if not at a word boundary. */ | |
| 652 | |
| 653 #ifdef emacs | |
| 654 ,before_dot, /* Succeeds if before point. */ | |
| 655 at_dot, /* Succeeds if at point. */ | |
| 656 after_dot, /* Succeeds if after point. */ | |
| 657 | |
| 658 /* Matches any character whose syntax is specified. Followed by | |
| 659 a byte which contains a syntax code, e.g., Sword. */ | |
| 660 syntaxspec, | |
| 661 | |
| 662 /* Matches any character whose syntax is not that specified. */ | |
| 663 notsyntaxspec | |
| 664 | |
| 665 #endif /* emacs */ | |
| 666 | |
| 667 #ifdef MULE | |
| 668 /* need extra stuff to be able to properly work with XEmacs/Mule | |
| 669 characters (which may take up more than one byte) */ | |
| 670 | |
| 671 ,charset_mule, /* Matches any character belonging to specified set. | |
| 672 The set is stored in "unified range-table | |
| 673 format"; see rangetab.c. Unlike the `charset' | |
| 674 opcode, this can handle arbitrary characters. */ | |
| 675 | |
| 676 charset_mule_not /* Same parameters as charset_mule, but match any | |
| 677 character that is not one of those specified. */ | |
| 678 | |
| 679 /* 97/2/17 jhod: The following two were merged back in from the Mule | |
| 680 2.3 code to enable some language specific processing */ | |
| 681 ,categoryspec, /* Matches entries in the character category tables */ | |
| 682 notcategoryspec /* The opposite of the above */ | |
| 683 #endif /* MULE */ | |
| 684 | |
| 685 } re_opcode_t; | |
| 686 | |
| 687 /* Common operations on the compiled pattern. */ | |
| 688 | |
| 689 /* Store NUMBER in two contiguous bytes starting at DESTINATION. */ | |
| 690 | |
| 691 #define STORE_NUMBER(destination, number) \ | |
| 692 do { \ | |
| 693 (destination)[0] = (number) & 0377; \ | |
| 694 (destination)[1] = (number) >> 8; \ | |
| 695 } while (0) | |
| 696 | |
| 697 /* Same as STORE_NUMBER, except increment DESTINATION to | |
| 698 the byte after where the number is stored. Therefore, DESTINATION | |
| 699 must be an lvalue. */ | |
| 700 | |
| 701 #define STORE_NUMBER_AND_INCR(destination, number) \ | |
| 702 do { \ | |
| 703 STORE_NUMBER (destination, number); \ | |
| 704 (destination) += 2; \ | |
| 705 } while (0) | |
| 706 | |
| 707 /* Put into DESTINATION a number stored in two contiguous bytes starting | |
| 708 at SOURCE. */ | |
| 709 | |
| 710 #define EXTRACT_NUMBER(destination, source) \ | |
| 711 do { \ | |
| 712 (destination) = *(source) & 0377; \ | |
| 713 (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \ | |
| 714 } while (0) | |
| 715 | |
| 716 #ifdef DEBUG | |
| 717 static void | |
| 446 | 718 extract_number (int *dest, re_char *source) |
| 428 | 719 { |
| 720 int temp = SIGN_EXTEND_CHAR (*(source + 1)); | |
| 721 *dest = *source & 0377; | |
| 722 *dest += temp << 8; | |
| 723 } | |
| 724 | |
| 725 #ifndef EXTRACT_MACROS /* To debug the macros. */ | |
| 726 #undef EXTRACT_NUMBER | |
| 727 #define EXTRACT_NUMBER(dest, src) extract_number (&dest, src) | |
| 728 #endif /* not EXTRACT_MACROS */ | |
| 729 | |
| 730 #endif /* DEBUG */ | |
| 731 | |
| 732 /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. | |
| 733 SOURCE must be an lvalue. */ | |
| 734 | |
| 735 #define EXTRACT_NUMBER_AND_INCR(destination, source) \ | |
| 736 do { \ | |
| 737 EXTRACT_NUMBER (destination, source); \ | |
| 738 (source) += 2; \ | |
| 739 } while (0) | |
| 740 | |
| 741 #ifdef DEBUG | |
| 742 static void | |
| 743 extract_number_and_incr (int *destination, unsigned char **source) | |
| 744 { | |
| 745 extract_number (destination, *source); | |
| 746 *source += 2; | |
| 747 } | |
| 748 | |
| 749 #ifndef EXTRACT_MACROS | |
| 750 #undef EXTRACT_NUMBER_AND_INCR | |
| 751 #define EXTRACT_NUMBER_AND_INCR(dest, src) \ | |
| 752 extract_number_and_incr (&dest, &src) | |
| 753 #endif /* not EXTRACT_MACROS */ | |
| 754 | |
| 755 #endif /* DEBUG */ | |
| 756 | |
| 757 /* If DEBUG is defined, Regex prints many voluminous messages about what | |
| 758 it is doing (if the variable `debug' is nonzero). If linked with the | |
| 759 main program in `iregex.c', you can enter patterns and strings | |
| 760 interactively. And if linked with the main program in `main.c' and | |
| 761 the other test files, you can run the already-written tests. */ | |
| 762 | |
| 763 #if defined (DEBUG) | |
| 764 | |
| 765 /* We use standard I/O for debugging. */ | |
| 766 #include <stdio.h> | |
| 767 | |
| 768 #ifndef emacs | |
| 769 /* XEmacs provides its own version of assert() */ | |
| 770 /* It is useful to test things that ``must'' be true when debugging. */ | |
| 771 #include <assert.h> | |
| 772 #endif | |
| 773 | |
| 5041 | 774 extern int debug_regexps; |
| 428 | 775 |
| 776 #define DEBUG_STATEMENT(e) e | |
| 5041 | 777 |
| 778 #define DEBUG_PRINT1(x) if (debug_regexps) printf (x) | |
| 779 #define DEBUG_PRINT2(x1, x2) if (debug_regexps) printf (x1, x2) | |
| 780 #define DEBUG_PRINT3(x1, x2, x3) if (debug_regexps) printf (x1, x2, x3) | |
| 781 #define DEBUG_PRINT4(x1, x2, x3, x4) if (debug_regexps) printf (x1, x2, x3, x4) | |
| 428 | 782 #define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ |
| 5041 | 783 if (debug_regexps) print_partial_compiled_pattern (s, e) |
| 428 | 784 #define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ |
| 5041 | 785 if (debug_regexps) print_double_string (w, s1, sz1, s2, sz2) |
| 786 | |
| 787 #define DEBUG_FAIL_PRINT1(x) \ | |
| 788 if (debug_regexps & RE_DEBUG_FAILURE_POINT) printf (x) | |
| 789 #define DEBUG_FAIL_PRINT2(x1, x2) \ | |
| 790 if (debug_regexps & RE_DEBUG_FAILURE_POINT) printf (x1, x2) | |
| 791 #define DEBUG_FAIL_PRINT3(x1, x2, x3) \ | |
| 792 if (debug_regexps & RE_DEBUG_FAILURE_POINT) printf (x1, x2, x3) | |
| 793 #define DEBUG_FAIL_PRINT4(x1, x2, x3, x4) \ | |
| 794 if (debug_regexps & RE_DEBUG_FAILURE_POINT) printf (x1, x2, x3, x4) | |
| 795 #define DEBUG_FAIL_PRINT_COMPILED_PATTERN(p, s, e) \ | |
| 796 if (debug_regexps & RE_DEBUG_FAILURE_POINT) \ | |
| 797 print_partial_compiled_pattern (s, e) | |
| 798 #define DEBUG_FAIL_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ | |
| 799 if (debug_regexps & RE_DEBUG_FAILURE_POINT) \ | |
| 800 print_double_string (w, s1, sz1, s2, sz2) | |
| 801 | |
| 802 #define DEBUG_MATCH_PRINT1(x) \ | |
| 803 if (debug_regexps & RE_DEBUG_MATCHING) printf (x) | |
| 804 #define DEBUG_MATCH_PRINT2(x1, x2) \ | |
| 805 if (debug_regexps & RE_DEBUG_MATCHING) printf (x1, x2) | |
| 806 #define DEBUG_MATCH_PRINT3(x1, x2, x3) \ | |
| 807 if (debug_regexps & RE_DEBUG_MATCHING) printf (x1, x2, x3) | |
| 808 #define DEBUG_MATCH_PRINT4(x1, x2, x3, x4) \ | |
| 809 if (debug_regexps & RE_DEBUG_MATCHING) printf (x1, x2, x3, x4) | |
| 810 #define DEBUG_MATCH_PRINT_COMPILED_PATTERN(p, s, e) \ | |
| 811 if (debug_regexps & RE_DEBUG_MATCHING) \ | |
| 812 print_partial_compiled_pattern (s, e) | |
| 813 #define DEBUG_MATCH_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ | |
| 814 if (debug_regexps & RE_DEBUG_MATCHING) \ | |
| 815 print_double_string (w, s1, sz1, s2, sz2) | |
| 428 | 816 |
| 817 | |
| 818 /* Print the fastmap in human-readable form. */ | |
| 819 | |
| 820 static void | |
| 821 print_fastmap (char *fastmap) | |
| 822 { | |
| 647 | 823 int was_a_range = 0; |
| 824 int i = 0; | |
| 428 | 825 |
| 826 while (i < (1 << BYTEWIDTH)) | |
| 827 { | |
| 828 if (fastmap[i++]) | |
| 829 { | |
| 830 was_a_range = 0; | |
| 831 putchar (i - 1); | |
| 832 while (i < (1 << BYTEWIDTH) && fastmap[i]) | |
| 833 { | |
| 834 was_a_range = 1; | |
| 835 i++; | |
| 836 } | |
| 837 if (was_a_range) | |
| 838 { | |
| 839 putchar ('-'); | |
| 840 putchar (i - 1); | |
| 841 } | |
| 842 } | |
| 843 } | |
| 844 putchar ('\n'); | |
| 845 } | |
| 846 | |
| 847 | |
| 848 /* Print a compiled pattern string in human-readable form, starting at | |
| 849 the START pointer into it and ending just before the pointer END. */ | |
| 850 | |
| 851 static void | |
| 446 | 852 print_partial_compiled_pattern (re_char *start, re_char *end) |
| 428 | 853 { |
| 854 int mcnt, mcnt2; | |
| 446 | 855 unsigned char *p = (unsigned char *) start; |
| 856 re_char *pend = end; | |
| 428 | 857 |
| 858 if (start == NULL) | |
| 859 { | |
| 860 puts ("(null)"); | |
| 861 return; | |
| 862 } | |
| 863 | |
| 864 /* Loop over pattern commands. */ | |
| 865 while (p < pend) | |
| 866 { | |
| 867 printf ("%ld:\t", (long)(p - start)); | |
| 868 | |
| 869 switch ((re_opcode_t) *p++) | |
| 870 { | |
| 871 case no_op: | |
| 872 printf ("/no_op"); | |
| 873 break; | |
| 874 | |
| 875 case exactn: | |
| 876 mcnt = *p++; | |
| 877 printf ("/exactn/%d", mcnt); | |
|
4750
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
878 while (mcnt--) |
| 428 | 879 { |
|
4750
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
880 putchar ('/'); |
| 428 | 881 putchar (*p++); |
| 882 } | |
| 883 break; | |
| 884 | |
| 885 case start_memory: | |
| 886 mcnt = *p++; | |
| 887 printf ("/start_memory/%d/%d", mcnt, *p++); | |
| 888 break; | |
| 889 | |
| 890 case stop_memory: | |
| 891 mcnt = *p++; | |
| 892 printf ("/stop_memory/%d/%d", mcnt, *p++); | |
| 893 break; | |
| 894 | |
| 895 case duplicate: | |
| 896 printf ("/duplicate/%d", *p++); | |
| 897 break; | |
| 898 | |
| 899 case anychar: | |
| 900 printf ("/anychar"); | |
| 901 break; | |
| 902 | |
| 903 case charset: | |
| 904 case charset_not: | |
| 905 { | |
| 906 REGISTER int c, last = -100; | |
| 907 REGISTER int in_range = 0; | |
| 908 | |
| 909 printf ("/charset [%s", | |
| 910 (re_opcode_t) *(p - 1) == charset_not ? "^" : ""); | |
| 911 | |
| 912 assert (p + *p < pend); | |
| 913 | |
| 914 for (c = 0; c < 256; c++) | |
| 915 if (((unsigned char) (c / 8) < *p) | |
| 916 && (p[1 + (c/8)] & (1 << (c % 8)))) | |
| 917 { | |
| 918 /* Are we starting a range? */ | |
| 919 if (last + 1 == c && ! in_range) | |
| 920 { | |
| 921 putchar ('-'); | |
| 922 in_range = 1; | |
| 923 } | |
| 924 /* Have we broken a range? */ | |
| 925 else if (last + 1 != c && in_range) | |
| 926 { | |
| 927 putchar (last); | |
| 928 in_range = 0; | |
| 929 } | |
| 930 | |
| 931 if (! in_range) | |
| 932 putchar (c); | |
| 933 | |
| 934 last = c; | |
| 935 } | |
| 936 | |
| 937 if (in_range) | |
| 938 putchar (last); | |
| 939 | |
| 940 putchar (']'); | |
| 941 | |
| 942 p += 1 + *p; | |
| 943 } | |
| 944 break; | |
| 945 | |
| 946 #ifdef MULE | |
| 947 case charset_mule: | |
| 948 case charset_mule_not: | |
| 949 { | |
| 950 int nentries, i; | |
| 951 | |
| 952 printf ("/charset_mule [%s", | |
| 953 (re_opcode_t) *(p - 1) == charset_mule_not ? "^" : ""); | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
954 printf (" flags: 0x%02x ", *p++); |
| 428 | 955 nentries = unified_range_table_nentries (p); |
| 956 for (i = 0; i < nentries; i++) | |
| 957 { | |
| 958 EMACS_INT first, last; | |
| 959 Lisp_Object dummy_val; | |
| 960 | |
| 961 unified_range_table_get_range (p, i, &first, &last, | |
| 962 &dummy_val); | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
963 if (first < 0x80) |
| 428 | 964 putchar (first); |
| 965 else | |
| 966 printf ("(0x%lx)", (long)first); | |
| 967 if (first != last) | |
| 968 { | |
| 969 putchar ('-'); | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
970 if (last < 0x80) |
| 428 | 971 putchar (last); |
| 972 else | |
| 973 printf ("(0x%lx)", (long)last); | |
| 974 } | |
| 975 } | |
| 976 putchar (']'); | |
| 977 p += unified_range_table_bytes_used (p); | |
| 978 } | |
| 979 break; | |
| 980 #endif | |
| 981 | |
| 982 case begline: | |
| 983 printf ("/begline"); | |
| 984 break; | |
| 985 | |
| 986 case endline: | |
| 987 printf ("/endline"); | |
| 988 break; | |
| 989 | |
| 990 case on_failure_jump: | |
| 991 extract_number_and_incr (&mcnt, &p); | |
| 992 printf ("/on_failure_jump to %ld", (long)(p + mcnt - start)); | |
| 993 break; | |
| 994 | |
| 995 case on_failure_keep_string_jump: | |
| 996 extract_number_and_incr (&mcnt, &p); | |
| 997 printf ("/on_failure_keep_string_jump to %ld", (long)(p + mcnt - start)); | |
| 998 break; | |
| 999 | |
| 1000 case dummy_failure_jump: | |
| 1001 extract_number_and_incr (&mcnt, &p); | |
| 1002 printf ("/dummy_failure_jump to %ld", (long)(p + mcnt - start)); | |
| 1003 break; | |
| 1004 | |
| 1005 case push_dummy_failure: | |
| 1006 printf ("/push_dummy_failure"); | |
| 1007 break; | |
| 1008 | |
| 1009 case maybe_pop_jump: | |
| 1010 extract_number_and_incr (&mcnt, &p); | |
| 1011 printf ("/maybe_pop_jump to %ld", (long)(p + mcnt - start)); | |
| 1012 break; | |
| 1013 | |
| 1014 case pop_failure_jump: | |
| 1015 extract_number_and_incr (&mcnt, &p); | |
| 1016 printf ("/pop_failure_jump to %ld", (long)(p + mcnt - start)); | |
| 1017 break; | |
| 1018 | |
| 1019 case jump_past_alt: | |
| 1020 extract_number_and_incr (&mcnt, &p); | |
| 1021 printf ("/jump_past_alt to %ld", (long)(p + mcnt - start)); | |
| 1022 break; | |
| 1023 | |
| 1024 case jump: | |
| 1025 extract_number_and_incr (&mcnt, &p); | |
| 1026 printf ("/jump to %ld", (long)(p + mcnt - start)); | |
| 1027 break; | |
| 1028 | |
| 1029 case succeed_n: | |
| 1030 extract_number_and_incr (&mcnt, &p); | |
| 1031 extract_number_and_incr (&mcnt2, &p); | |
| 1032 printf ("/succeed_n to %ld, %d times", (long)(p + mcnt - start), mcnt2); | |
| 1033 break; | |
| 1034 | |
| 1035 case jump_n: | |
| 1036 extract_number_and_incr (&mcnt, &p); | |
| 1037 extract_number_and_incr (&mcnt2, &p); | |
| 1038 printf ("/jump_n to %ld, %d times", (long)(p + mcnt - start), mcnt2); | |
| 1039 break; | |
| 1040 | |
| 1041 case set_number_at: | |
| 1042 extract_number_and_incr (&mcnt, &p); | |
| 1043 extract_number_and_incr (&mcnt2, &p); | |
| 1044 printf ("/set_number_at location %ld to %d", (long)(p + mcnt - start), mcnt2); | |
| 1045 break; | |
| 1046 | |
| 1047 case wordbound: | |
| 1048 printf ("/wordbound"); | |
| 1049 break; | |
| 1050 | |
| 1051 case notwordbound: | |
| 1052 printf ("/notwordbound"); | |
| 1053 break; | |
| 1054 | |
| 1055 case wordbeg: | |
| 1056 printf ("/wordbeg"); | |
| 1057 break; | |
| 1058 | |
| 1059 case wordend: | |
| 1060 printf ("/wordend"); | |
| 1061 | |
| 1062 #ifdef emacs | |
| 1063 case before_dot: | |
| 1064 printf ("/before_dot"); | |
| 1065 break; | |
| 1066 | |
| 1067 case at_dot: | |
| 1068 printf ("/at_dot"); | |
| 1069 break; | |
| 1070 | |
| 1071 case after_dot: | |
| 1072 printf ("/after_dot"); | |
| 1073 break; | |
| 1074 | |
| 1075 case syntaxspec: | |
| 1076 printf ("/syntaxspec"); | |
| 1077 mcnt = *p++; | |
| 1078 printf ("/%d", mcnt); | |
| 1079 break; | |
| 1080 | |
| 1081 case notsyntaxspec: | |
| 1082 printf ("/notsyntaxspec"); | |
| 1083 mcnt = *p++; | |
| 1084 printf ("/%d", mcnt); | |
| 1085 break; | |
| 1086 | |
| 1087 #ifdef MULE | |
| 1088 /* 97/2/17 jhod Mule category patch */ | |
| 1089 case categoryspec: | |
| 1090 printf ("/categoryspec"); | |
| 1091 mcnt = *p++; | |
| 1092 printf ("/%d", mcnt); | |
| 1093 break; | |
| 1094 | |
| 1095 case notcategoryspec: | |
| 1096 printf ("/notcategoryspec"); | |
| 1097 mcnt = *p++; | |
| 1098 printf ("/%d", mcnt); | |
| 1099 break; | |
| 1100 /* end of category patch */ | |
| 1101 #endif /* MULE */ | |
| 1102 #endif /* emacs */ | |
| 1103 | |
| 1104 case wordchar: | |
| 1105 printf ("/wordchar"); | |
| 1106 break; | |
| 1107 | |
| 1108 case notwordchar: | |
| 1109 printf ("/notwordchar"); | |
| 1110 break; | |
| 1111 | |
| 1112 case begbuf: | |
| 1113 printf ("/begbuf"); | |
| 1114 break; | |
| 1115 | |
| 1116 case endbuf: | |
| 1117 printf ("/endbuf"); | |
| 1118 break; | |
| 1119 | |
| 1120 default: | |
| 1121 printf ("?%d", *(p-1)); | |
| 1122 } | |
| 1123 | |
| 1124 putchar ('\n'); | |
| 1125 } | |
| 1126 | |
| 1127 printf ("%ld:\tend of pattern.\n", (long)(p - start)); | |
| 1128 } | |
| 1129 | |
| 1130 | |
| 1131 static void | |
| 1132 print_compiled_pattern (struct re_pattern_buffer *bufp) | |
| 1133 { | |
| 446 | 1134 re_char *buffer = bufp->buffer; |
| 428 | 1135 |
| 1136 print_partial_compiled_pattern (buffer, buffer + bufp->used); | |
| 1137 printf ("%ld bytes used/%ld bytes allocated.\n", bufp->used, | |
| 1138 bufp->allocated); | |
| 1139 | |
| 1140 if (bufp->fastmap_accurate && bufp->fastmap) | |
| 1141 { | |
| 1142 printf ("fastmap: "); | |
| 1143 print_fastmap (bufp->fastmap); | |
| 1144 } | |
| 1145 | |
| 1146 printf ("re_nsub: %ld\t", (long)bufp->re_nsub); | |
| 502 | 1147 printf ("re_ngroups: %ld\t", (long)bufp->re_ngroups); |
| 428 | 1148 printf ("regs_alloc: %d\t", bufp->regs_allocated); |
| 1149 printf ("can_be_null: %d\t", bufp->can_be_null); | |
| 1150 printf ("newline_anchor: %d\n", bufp->newline_anchor); | |
| 1151 printf ("no_sub: %d\t", bufp->no_sub); | |
| 1152 printf ("not_bol: %d\t", bufp->not_bol); | |
| 1153 printf ("not_eol: %d\t", bufp->not_eol); | |
| 1154 printf ("syntax: %d\n", bufp->syntax); | |
| 1155 /* Perhaps we should print the translate table? */ | |
| 1156 /* and maybe the category table? */ | |
| 502 | 1157 |
| 1158 if (bufp->external_to_internal_register) | |
| 1159 { | |
| 1160 int i; | |
| 1161 | |
| 1162 printf ("external_to_internal_register:\n"); | |
| 1163 for (i = 0; i <= bufp->re_nsub; i++) | |
| 1164 { | |
| 1165 if (i > 0) | |
| 1166 printf (", "); | |
| 1167 printf ("%d -> %d", i, bufp->external_to_internal_register[i]); | |
| 1168 } | |
| 1169 printf ("\n"); | |
| 1170 } | |
| 428 | 1171 } |
| 1172 | |
| 1173 | |
| 1174 static void | |
| 446 | 1175 print_double_string (re_char *where, re_char *string1, int size1, |
| 1176 re_char *string2, int size2) | |
| 428 | 1177 { |
| 1178 if (where == NULL) | |
| 1179 printf ("(null)"); | |
| 1180 else | |
| 1181 { | |
| 647 | 1182 int this_char; |
| 428 | 1183 |
| 1184 if (FIRST_STRING_P (where)) | |
| 1185 { | |
| 1186 for (this_char = where - string1; this_char < size1; this_char++) | |
| 1187 putchar (string1[this_char]); | |
| 1188 | |
| 1189 where = string2; | |
| 1190 } | |
| 1191 | |
| 1192 for (this_char = where - string2; this_char < size2; this_char++) | |
| 1193 putchar (string2[this_char]); | |
| 1194 } | |
| 1195 } | |
| 1196 | |
| 1197 #else /* not DEBUG */ | |
| 1198 | |
| 771 | 1199 #ifndef emacs |
| 428 | 1200 #undef assert |
| 771 | 1201 #define assert(e) ((void) (1)) |
| 1202 #endif | |
| 428 | 1203 |
| 1204 #define DEBUG_STATEMENT(e) | |
| 5041 | 1205 |
| 428 | 1206 #define DEBUG_PRINT1(x) |
| 1207 #define DEBUG_PRINT2(x1, x2) | |
| 1208 #define DEBUG_PRINT3(x1, x2, x3) | |
| 1209 #define DEBUG_PRINT4(x1, x2, x3, x4) | |
| 1210 #define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) | |
| 1211 #define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | |
| 1212 | |
| 5041 | 1213 #define DEBUG_FAIL_PRINT1(x) |
| 1214 #define DEBUG_FAIL_PRINT2(x1, x2) | |
| 1215 #define DEBUG_FAIL_PRINT3(x1, x2, x3) | |
| 1216 #define DEBUG_FAIL_PRINT4(x1, x2, x3, x4) | |
| 1217 #define DEBUG_FAIL_PRINT_COMPILED_PATTERN(p, s, e) | |
| 1218 #define DEBUG_FAIL_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | |
| 1219 | |
| 1220 #define DEBUG_MATCH_PRINT1(x) | |
| 1221 #define DEBUG_MATCH_PRINT2(x1, x2) | |
| 1222 #define DEBUG_MATCH_PRINT3(x1, x2, x3) | |
| 1223 #define DEBUG_MATCH_PRINT4(x1, x2, x3, x4) | |
| 1224 #define DEBUG_MATCH_PRINT_COMPILED_PATTERN(p, s, e) | |
| 1225 #define DEBUG_MATCH_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | |
| 1226 | |
| 446 | 1227 #endif /* DEBUG */ |
| 428 | 1228 |
| 1229 /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can | |
| 1230 also be assigned to arbitrarily: each pattern buffer stores its own | |
| 1231 syntax, so it can be changed between regex compilations. */ | |
| 1232 /* This has no initializer because initialized variables in Emacs | |
| 1233 become read-only after dumping. */ | |
| 1234 reg_syntax_t re_syntax_options; | |
| 1235 | |
| 1236 | |
| 1237 /* Specify the precise syntax of regexps for compilation. This provides | |
| 1238 for compatibility for various utilities which historically have | |
| 1239 different, incompatible syntaxes. | |
| 1240 | |
| 1241 The argument SYNTAX is a bit mask comprised of the various bits | |
| 1242 defined in regex.h. We return the old syntax. */ | |
| 1243 | |
| 1244 reg_syntax_t | |
| 1245 re_set_syntax (reg_syntax_t syntax) | |
| 1246 { | |
| 1247 reg_syntax_t ret = re_syntax_options; | |
| 1248 | |
| 1249 re_syntax_options = syntax; | |
| 1250 return ret; | |
| 1251 } | |
| 1252 | |
| 1253 /* This table gives an error message for each of the error codes listed | |
| 1254 in regex.h. Obviously the order here has to be same as there. | |
| 1255 POSIX doesn't require that we do anything for REG_NOERROR, | |
| 1256 but why not be nice? */ | |
| 1257 | |
| 442 | 1258 static const char *re_error_msgid[] = |
| 428 | 1259 { |
| 1260 "Success", /* REG_NOERROR */ | |
| 1261 "No match", /* REG_NOMATCH */ | |
| 1262 "Invalid regular expression", /* REG_BADPAT */ | |
| 1263 "Invalid collation character", /* REG_ECOLLATE */ | |
| 1264 "Invalid character class name", /* REG_ECTYPE */ | |
| 1265 "Trailing backslash", /* REG_EESCAPE */ | |
| 1266 "Invalid back reference", /* REG_ESUBREG */ | |
| 1267 "Unmatched [ or [^", /* REG_EBRACK */ | |
| 1268 "Unmatched ( or \\(", /* REG_EPAREN */ | |
| 1269 "Unmatched \\{", /* REG_EBRACE */ | |
| 1270 "Invalid content of \\{\\}", /* REG_BADBR */ | |
| 1271 "Invalid range end", /* REG_ERANGE */ | |
| 1272 "Memory exhausted", /* REG_ESPACE */ | |
| 1273 "Invalid preceding regular expression", /* REG_BADRPT */ | |
| 1274 "Premature end of regular expression", /* REG_EEND */ | |
| 1275 "Regular expression too big", /* REG_ESIZE */ | |
| 1276 "Unmatched ) or \\)", /* REG_ERPAREN */ | |
| 1277 #ifdef emacs | |
| 1278 "Invalid syntax designator", /* REG_ESYNTAX */ | |
| 1279 #endif | |
| 1280 #ifdef MULE | |
| 1281 "Ranges may not span charsets", /* REG_ERANGESPAN */ | |
| 1282 "Invalid category designator", /* REG_ECATEGORY */ | |
| 1283 #endif | |
| 1284 }; | |
| 1285 | |
| 1286 /* Avoiding alloca during matching, to placate r_alloc. */ | |
| 1287 | |
| 1333 | 1288 /* About these various flags: |
| 1289 | |
| 1290 MATCH_MAY_ALLOCATE indicates that it's OK to do allocation in the | |
| 1291 searching and matching functions. In this case, we use local variables | |
| 1292 to hold the values allocated. If not, we use *global* variables, which | |
| 1293 are pre-allocated. NOTE: XEmacs ***MUST*** run with MATCH_MAY_ALLOCATE, | |
| 1294 because the regexp routines may get called reentrantly as a result of | |
| 1295 QUIT processing (e.g. under Windows: re_match -> QUIT -> quit_p -> drain | |
| 1296 events -> process WM_INITMENU -> call filter -> re_match; see stack | |
| 1297 trace in signal.c), so we cannot have any global variables (unless we do | |
| 1298 lots of trickiness including some unwind-protects, which isn't worth it | |
| 1299 at this point). | |
| 1300 | |
| 1301 REL_ALLOC means that the relocating allocator is in use, for buffers | |
| 1302 and such. REGEX_REL_ALLOC means that we use rel-alloc to manage the | |
| 1303 fail stack, which may grow quite large. REGEX_MALLOC means we use | |
| 1304 malloc() in place of alloca() to allocate the fail stack -- only | |
| 1305 applicable if REGEX_REL_ALLOC is not defined. | |
| 1306 */ | |
| 1307 | |
| 428 | 1308 /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the |
| 1309 searching and matching functions should not call alloca. On some | |
| 1310 systems, alloca is implemented in terms of malloc, and if we're | |
| 1311 using the relocating allocator routines, then malloc could cause a | |
| 1312 relocation, which might (if the strings being searched are in the | |
| 1313 ralloc heap) shift the data out from underneath the regexp | |
| 771 | 1314 routines. [To clarify: The purpose of rel-alloc is to allow data to |
| 1315 be moved in memory from one place to another so that all data | |
| 1316 blocks can be consolidated together and excess memory released back | |
| 1317 to the operating system. This requires that all the blocks that | |
| 1318 are managed by rel-alloc go at the very end of the program's heap, | |
| 1319 after all regularly malloc()ed data. malloc(), however, is used to | |
| 1320 owning the end of the heap, so that when more memory is needed, it | |
| 1321 just expands the heap using sbrk(). This is reconciled by using a | |
| 1322 malloc() (such as malloc.c, gmalloc.c, or recent versions of | |
| 1323 malloc() in libc) where the sbrk() call can be replaced with a | |
| 1324 user-specified call -- in this case, to rel-alloc's r_alloc_sbrk() | |
| 1325 routine. This routine calls the real sbrk(), but then shifts all | |
| 1326 the rel-alloc-managed blocks forward to the end of the heap again, | |
| 1327 so that malloc() gets the memory it needs in the location it needs | |
| 1328 it at. The regex routines may well have pointers to buffer data as | |
| 1329 their arguments, and buffers are managed by rel-alloc if rel-alloc | |
| 1330 has been enabled, so calling malloc() may potentially screw things | |
| 1331 up badly if it runs out of space and asks for more from the OS.] | |
| 1332 | |
| 1333 [[Here's another reason to avoid allocation: Emacs processes input | |
| 1334 from X in a signal handler; processing X input may call malloc; if | |
| 1335 input arrives while a matching routine is calling malloc, then | |
| 1336 we're scrod. But Emacs can't just block input while calling | |
| 1337 matching routines; then we don't notice interrupts when they come | |
| 1338 in. So, Emacs blocks input around all regexp calls except the | |
| 1339 matching calls, which it leaves unprotected, in the faith that they | |
| 1333 | 1340 will not malloc.]] This previous paragraph is irrelevant under XEmacs, |
| 1341 as we *do not* do anything so stupid as process input from within a | |
| 1342 signal handler. | |
| 1343 | |
| 1344 However, the regexp routines may get called reentrantly as a result of | |
| 1345 QUIT processing (e.g. under Windows: re_match -> QUIT -> quit_p -> drain | |
| 1346 events -> process WM_INITMENU -> call filter -> re_match; see stack | |
| 1347 trace in signal.c), so we cannot have any global variables (unless we do | |
| 1348 lots of trickiness including some unwind-protects, which isn't worth it | |
| 1349 at this point). Hence we MUST have MATCH_MAY_ALLOCATE defined. | |
| 1350 | |
| 1351 Also, the first paragraph does not make complete sense to me -- what | |
| 1352 about the use of rel-alloc to handle the fail stacks? Shouldn't these | |
| 1353 reallocations potentially cause buffer data to be relocated as well? I | |
| 826 | 1354 must be missing something, though -- perhaps the writer above is |
| 1355 assuming that the failure stack(s) will always be allocated after the | |
| 1356 buffer data, and thus reallocating them with rel-alloc won't move buffer | |
| 1333 | 1357 data. (In fact, a cursory glance at the code in ralloc.c seems to |
| 1358 confirm this.) --ben */ | |
| 428 | 1359 |
| 1360 /* Normally, this is fine. */ | |
| 1361 #define MATCH_MAY_ALLOCATE | |
| 1362 | |
| 1363 /* When using GNU C, we are not REALLY using the C alloca, no matter | |
| 1364 what config.h may say. So don't take precautions for it. */ | |
| 1365 #ifdef __GNUC__ | |
| 1366 #undef C_ALLOCA | |
| 1367 #endif | |
| 1368 | |
| 1369 /* The match routines may not allocate if (1) they would do it with malloc | |
| 1370 and (2) it's not safe for them to use malloc. | |
| 1371 Note that if REL_ALLOC is defined, matching would not use malloc for the | |
| 1372 failure stack, but we would still use it for the register vectors; | |
| 1373 so REL_ALLOC should not affect this. */ | |
| 771 | 1374 |
| 1333 | 1375 /* XEmacs can handle REL_ALLOC and malloc() OK */ |
| 1376 #if !defined (emacs) && (defined (C_ALLOCA) || defined (REGEX_MALLOC)) && defined (REL_ALLOC) | |
| 428 | 1377 #undef MATCH_MAY_ALLOCATE |
| 1378 #endif | |
| 1379 | |
| 1333 | 1380 #if !defined (MATCH_MAY_ALLOCATE) && defined (emacs) |
| 771 | 1381 #error regex must be handle reentrancy; MATCH_MAY_ALLOCATE must be defined |
| 1382 #endif | |
| 1383 | |
| 428 | 1384 |
| 1385 /* Failure stack declarations and macros; both re_compile_fastmap and | |
| 1386 re_match_2 use a failure stack. These have to be macros because of | |
| 1387 REGEX_ALLOCATE_STACK. */ | |
| 1388 | |
| 1389 | |
| 1390 /* Number of failure points for which to initially allocate space | |
| 1391 when matching. If this number is exceeded, we allocate more | |
| 1392 space, so it is not a hard limit. */ | |
| 1393 #ifndef INIT_FAILURE_ALLOC | |
| 3300 | 1394 #define INIT_FAILURE_ALLOC 20 |
| 428 | 1395 #endif |
| 1396 | |
| 1397 /* Roughly the maximum number of failure points on the stack. Would be | |
| 1398 exactly that if always used MAX_FAILURE_SPACE each time we failed. | |
| 1399 This is a variable only so users of regex can assign to it; we never | |
| 1400 change it ourselves. */ | |
| 1401 #if defined (MATCH_MAY_ALLOCATE) | |
| 1402 /* 4400 was enough to cause a crash on Alpha OSF/1, | |
| 1403 whose default stack limit is 2mb. */ | |
| 3300 | 1404 int re_max_failures = 40000; |
| 428 | 1405 #else |
| 3300 | 1406 int re_max_failures = 4000; |
| 428 | 1407 #endif |
| 1408 | |
| 1409 union fail_stack_elt | |
| 1410 { | |
| 446 | 1411 re_char *pointer; |
| 428 | 1412 int integer; |
| 1413 }; | |
| 1414 | |
| 1415 typedef union fail_stack_elt fail_stack_elt_t; | |
| 1416 | |
| 1417 typedef struct | |
| 1418 { | |
| 1419 fail_stack_elt_t *stack; | |
| 665 | 1420 Elemcount size; |
| 1421 Elemcount avail; /* Offset of next open position. */ | |
| 428 | 1422 } fail_stack_type; |
| 1423 | |
| 1424 #define FAIL_STACK_EMPTY() (fail_stack.avail == 0) | |
| 1425 #define FAIL_STACK_PTR_EMPTY() (fail_stack_ptr->avail == 0) | |
| 1426 #define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size) | |
| 1427 | |
| 1428 | |
| 1429 /* Define macros to initialize and free the failure stack. | |
| 1430 Do `return -2' if the alloc fails. */ | |
| 1431 | |
| 1432 #ifdef MATCH_MAY_ALLOCATE | |
| 1333 | 1433 #define INIT_FAIL_STACK() \ |
| 1434 do { \ | |
| 1435 fail_stack.stack = (fail_stack_elt_t *) \ | |
| 1436 REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * \ | |
| 1437 sizeof (fail_stack_elt_t)); \ | |
| 1438 \ | |
| 1439 if (fail_stack.stack == NULL) \ | |
| 1440 { \ | |
| 1441 UNBIND_REGEX_MALLOC_CHECK (); \ | |
| 1442 return -2; \ | |
| 1443 } \ | |
| 1444 \ | |
| 1445 fail_stack.size = INIT_FAILURE_ALLOC; \ | |
| 1446 fail_stack.avail = 0; \ | |
| 428 | 1447 } while (0) |
| 1448 | |
| 1449 #define RESET_FAIL_STACK() REGEX_FREE_STACK (fail_stack.stack) | |
| 1450 #else | |
| 1451 #define INIT_FAIL_STACK() \ | |
| 1452 do { \ | |
| 1453 fail_stack.avail = 0; \ | |
| 1454 } while (0) | |
| 1455 | |
| 1456 #define RESET_FAIL_STACK() | |
| 1457 #endif | |
| 1458 | |
| 1459 | |
| 1460 /* Double the size of FAIL_STACK, up to approximately `re_max_failures' items. | |
| 1461 | |
| 1462 Return 1 if succeeds, and 0 if either ran out of memory | |
| 1463 allocating space for it or it was already too large. | |
| 1464 | |
| 1465 REGEX_REALLOCATE_STACK requires `destination' be declared. */ | |
| 1466 | |
| 1467 #define DOUBLE_FAIL_STACK(fail_stack) \ | |
| 1468 ((fail_stack).size > re_max_failures * MAX_FAILURE_ITEMS \ | |
| 1469 ? 0 \ | |
| 1470 : ((fail_stack).stack = (fail_stack_elt_t *) \ | |
| 1471 REGEX_REALLOCATE_STACK ((fail_stack).stack, \ | |
| 1472 (fail_stack).size * sizeof (fail_stack_elt_t), \ | |
| 1473 ((fail_stack).size << 1) * sizeof (fail_stack_elt_t)), \ | |
| 1474 \ | |
| 1475 (fail_stack).stack == NULL \ | |
| 1476 ? 0 \ | |
| 1477 : ((fail_stack).size <<= 1, \ | |
| 1478 1))) | |
| 1479 | |
| 1333 | 1480 #if !defined (emacs) || !defined (REL_ALLOC) |
| 1481 #define RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS() | |
| 1482 #else | |
| 1483 /* Don't change NULL pointers */ | |
| 1484 #define ADD_IF_NZ(val) if (val) val += rmdp_offset | |
| 1346 | 1485 #define RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS() \ |
| 1486 do \ | |
| 1487 { \ | |
| 1488 Bytecount rmdp_offset = offset_post_relocation (lispobj, orig_buftext); \ | |
| 1489 \ | |
| 1490 if (rmdp_offset) \ | |
| 1491 { \ | |
| 1492 int i; \ | |
| 1493 \ | |
| 1494 ADD_IF_NZ (string1); \ | |
| 1495 ADD_IF_NZ (string2); \ | |
| 1496 ADD_IF_NZ (d); \ | |
| 1497 ADD_IF_NZ (dend); \ | |
| 1498 ADD_IF_NZ (end1); \ | |
| 1499 ADD_IF_NZ (end2); \ | |
| 1500 ADD_IF_NZ (end_match_1); \ | |
| 1501 ADD_IF_NZ (end_match_2); \ | |
| 1502 \ | |
| 1503 if (bufp->re_ngroups) \ | |
| 1504 { \ | |
| 1505 for (i = 0; i < num_regs; i++) \ | |
| 1506 { \ | |
| 1507 ADD_IF_NZ (regstart[i]); \ | |
| 1508 ADD_IF_NZ (regend[i]); \ | |
| 1509 ADD_IF_NZ (old_regstart[i]); \ | |
| 1510 ADD_IF_NZ (old_regend[i]); \ | |
| 1511 ADD_IF_NZ (best_regstart[i]); \ | |
| 1512 ADD_IF_NZ (best_regend[i]); \ | |
| 1513 ADD_IF_NZ (reg_dummy[i]); \ | |
| 1514 } \ | |
| 1515 } \ | |
| 1516 \ | |
| 1517 ADD_IF_NZ (match_end); \ | |
| 1518 } \ | |
| 1333 | 1519 } while (0) |
| 1520 #endif /* !defined (emacs) || !defined (REL_ALLOC) */ | |
| 1521 | |
| 1522 #if !defined (emacs) || !defined (REL_ALLOC) | |
| 1523 #define RE_SEARCH_RELOCATE_MOVEABLE_DATA_POINTERS() | |
| 1524 #else | |
| 1346 | 1525 #define RE_SEARCH_RELOCATE_MOVEABLE_DATA_POINTERS() \ |
| 1526 do \ | |
| 1527 { \ | |
| 1528 Bytecount rmdp_offset = offset_post_relocation (lispobj, orig_buftext); \ | |
| 1529 \ | |
| 1530 if (rmdp_offset) \ | |
| 1531 { \ | |
| 1532 ADD_IF_NZ (str1); \ | |
| 1533 ADD_IF_NZ (str2); \ | |
| 1534 ADD_IF_NZ (string1); \ | |
| 1535 ADD_IF_NZ (string2); \ | |
| 1536 ADD_IF_NZ (d); \ | |
| 1537 } \ | |
| 1333 | 1538 } while (0) |
| 1539 | |
| 1540 #endif /* emacs */ | |
| 428 | 1541 |
| 1542 /* Push pointer POINTER on FAIL_STACK. | |
| 1543 Return 1 if was able to do so and 0 if ran out of memory allocating | |
| 1544 space to do so. */ | |
| 1545 #define PUSH_PATTERN_OP(POINTER, FAIL_STACK) \ | |
| 1546 ((FAIL_STACK_FULL () \ | |
| 1547 && !DOUBLE_FAIL_STACK (FAIL_STACK)) \ | |
| 1548 ? 0 \ | |
| 1549 : ((FAIL_STACK).stack[(FAIL_STACK).avail++].pointer = POINTER, \ | |
| 1550 1)) | |
| 1551 | |
| 1552 /* Push a pointer value onto the failure stack. | |
| 1553 Assumes the variable `fail_stack'. Probably should only | |
| 1554 be called from within `PUSH_FAILURE_POINT'. */ | |
| 1555 #define PUSH_FAILURE_POINTER(item) \ | |
| 1556 fail_stack.stack[fail_stack.avail++].pointer = (unsigned char *) (item) | |
| 1557 | |
| 1558 /* This pushes an integer-valued item onto the failure stack. | |
| 1559 Assumes the variable `fail_stack'. Probably should only | |
| 1560 be called from within `PUSH_FAILURE_POINT'. */ | |
| 1561 #define PUSH_FAILURE_INT(item) \ | |
| 1562 fail_stack.stack[fail_stack.avail++].integer = (item) | |
| 1563 | |
| 1564 /* Push a fail_stack_elt_t value onto the failure stack. | |
| 1565 Assumes the variable `fail_stack'. Probably should only | |
| 1566 be called from within `PUSH_FAILURE_POINT'. */ | |
| 1567 #define PUSH_FAILURE_ELT(item) \ | |
| 1568 fail_stack.stack[fail_stack.avail++] = (item) | |
| 1569 | |
| 1570 /* These three POP... operations complement the three PUSH... operations. | |
| 1571 All assume that `fail_stack' is nonempty. */ | |
| 1572 #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer | |
| 1573 #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer | |
| 1574 #define POP_FAILURE_ELT() fail_stack.stack[--fail_stack.avail] | |
| 1575 | |
| 1576 /* Used to omit pushing failure point id's when we're not debugging. */ | |
| 1577 #ifdef DEBUG | |
| 1578 #define DEBUG_PUSH PUSH_FAILURE_INT | |
| 1579 #define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_INT () | |
| 1580 #else | |
| 1581 #define DEBUG_PUSH(item) | |
| 1582 #define DEBUG_POP(item_addr) | |
| 1583 #endif | |
| 1584 | |
| 1585 | |
| 1586 /* Push the information about the state we will need | |
| 1587 if we ever fail back to it. | |
| 1588 | |
| 1589 Requires variables fail_stack, regstart, regend, reg_info, and | |
| 1590 num_regs be declared. DOUBLE_FAIL_STACK requires `destination' be | |
| 1591 declared. | |
| 1592 | |
| 1593 Does `return FAILURE_CODE' if runs out of memory. */ | |
| 1594 | |
| 771 | 1595 #if !defined (REGEX_MALLOC) && !defined (REGEX_REL_ALLOC) |
| 456 | 1596 #define DECLARE_DESTINATION char *destination |
| 428 | 1597 #else |
| 456 | 1598 #define DECLARE_DESTINATION DECLARE_NOTHING |
| 428 | 1599 #endif |
| 1600 | |
| 1601 #define PUSH_FAILURE_POINT(pattern_place, string_place, failure_code) \ | |
| 456 | 1602 do { \ |
| 1603 DECLARE_DESTINATION; \ | |
| 1604 /* Must be int, so when we don't save any registers, the arithmetic \ | |
| 1605 of 0 + -1 isn't done as unsigned. */ \ | |
| 1606 int this_reg; \ | |
| 428 | 1607 \ |
| 456 | 1608 DEBUG_STATEMENT (failure_id++); \ |
| 1609 DEBUG_STATEMENT (nfailure_points_pushed++); \ | |
| 5041 | 1610 DEBUG_FAIL_PRINT2 ("\nPUSH_FAILURE_POINT #%d:\n", failure_id); \ |
| 1611 DEBUG_FAIL_PRINT2 (" Before push, next avail: %ld\n", \ | |
| 647 | 1612 (long) (fail_stack).avail); \ |
| 5041 | 1613 DEBUG_FAIL_PRINT2 (" size: %ld\n", \ |
| 647 | 1614 (long) (fail_stack).size); \ |
| 456 | 1615 \ |
| 5041 | 1616 DEBUG_FAIL_PRINT2 (" slots needed: %d\n", NUM_FAILURE_ITEMS); \ |
| 1617 DEBUG_FAIL_PRINT2 (" available: %ld\n", \ | |
| 456 | 1618 (long) REMAINING_AVAIL_SLOTS); \ |
| 428 | 1619 \ |
| 456 | 1620 /* Ensure we have enough space allocated for what we will push. */ \ |
| 1621 while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \ | |
| 1622 { \ | |
| 1333 | 1623 BEGIN_REGEX_MALLOC_OK (); \ |
| 456 | 1624 if (!DOUBLE_FAIL_STACK (fail_stack)) \ |
| 1333 | 1625 { \ |
| 1626 END_REGEX_MALLOC_OK (); \ | |
| 1627 UNBIND_REGEX_MALLOC_CHECK (); \ | |
| 1628 return failure_code; \ | |
| 1629 } \ | |
| 1630 END_REGEX_MALLOC_OK (); \ | |
| 5041 | 1631 DEBUG_FAIL_PRINT2 ("\n Doubled stack; size now: %ld\n", \ |
| 647 | 1632 (long) (fail_stack).size); \ |
| 5041 | 1633 DEBUG_FAIL_PRINT2 (" slots available: %ld\n", \ |
| 456 | 1634 (long) REMAINING_AVAIL_SLOTS); \ |
| 1333 | 1635 \ |
| 1636 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); \ | |
| 456 | 1637 } \ |
| 428 | 1638 \ |
| 456 | 1639 /* Push the info, starting with the registers. */ \ |
| 5041 | 1640 DEBUG_FAIL_PRINT1 ("\n"); \ |
| 428 | 1641 \ |
| 456 | 1642 for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \ |
| 1643 this_reg++) \ | |
| 1644 { \ | |
| 5041 | 1645 DEBUG_FAIL_PRINT2 (" Pushing reg: %d\n", this_reg); \ |
| 456 | 1646 DEBUG_STATEMENT (num_regs_pushed++); \ |
| 428 | 1647 \ |
| 5041 | 1648 DEBUG_FAIL_PRINT2 (" start: 0x%lx\n", (long) regstart[this_reg]); \ |
| 456 | 1649 PUSH_FAILURE_POINTER (regstart[this_reg]); \ |
| 1650 \ | |
| 5041 | 1651 DEBUG_FAIL_PRINT2 (" end: 0x%lx\n", (long) regend[this_reg]); \ |
| 456 | 1652 PUSH_FAILURE_POINTER (regend[this_reg]); \ |
| 428 | 1653 \ |
| 5041 | 1654 DEBUG_FAIL_PRINT2 (" info: 0x%lx\n ", \ |
| 456 | 1655 * (long *) (®_info[this_reg])); \ |
| 5041 | 1656 DEBUG_FAIL_PRINT2 (" match_null=%d", \ |
| 456 | 1657 REG_MATCH_NULL_STRING_P (reg_info[this_reg])); \ |
| 5041 | 1658 DEBUG_FAIL_PRINT2 (" active=%d", IS_ACTIVE (reg_info[this_reg])); \ |
| 1659 DEBUG_FAIL_PRINT2 (" matched_something=%d", \ | |
| 456 | 1660 MATCHED_SOMETHING (reg_info[this_reg])); \ |
| 5041 | 1661 DEBUG_FAIL_PRINT2 (" ever_matched_something=%d", \ |
| 456 | 1662 EVER_MATCHED_SOMETHING (reg_info[this_reg])); \ |
| 5041 | 1663 DEBUG_FAIL_PRINT1 ("\n"); \ |
| 456 | 1664 PUSH_FAILURE_ELT (reg_info[this_reg].word); \ |
| 1665 } \ | |
| 428 | 1666 \ |
| 5041 | 1667 DEBUG_FAIL_PRINT2 (" Pushing low active reg: %d\n", lowest_active_reg); \ |
| 456 | 1668 PUSH_FAILURE_INT (lowest_active_reg); \ |
| 428 | 1669 \ |
| 5041 | 1670 DEBUG_FAIL_PRINT2 (" Pushing high active reg: %d\n", highest_active_reg); \ |
| 456 | 1671 PUSH_FAILURE_INT (highest_active_reg); \ |
| 428 | 1672 \ |
| 5041 | 1673 DEBUG_FAIL_PRINT2 (" Pushing pattern 0x%lx: \n", (long) pattern_place); \ |
| 1674 DEBUG_FAIL_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \ | |
| 456 | 1675 PUSH_FAILURE_POINTER (pattern_place); \ |
| 428 | 1676 \ |
| 5041 | 1677 DEBUG_FAIL_PRINT2 (" Pushing string 0x%lx: `", (long) string_place); \ |
| 1678 DEBUG_FAIL_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, \ | |
| 456 | 1679 size2); \ |
| 5041 | 1680 DEBUG_FAIL_PRINT1 ("'\n"); \ |
| 456 | 1681 PUSH_FAILURE_POINTER (string_place); \ |
| 428 | 1682 \ |
| 5041 | 1683 DEBUG_FAIL_PRINT2 (" Pushing failure id: %u\n", failure_id); \ |
| 456 | 1684 DEBUG_PUSH (failure_id); \ |
| 1685 } while (0) | |
| 428 | 1686 |
| 1687 /* This is the number of items that are pushed and popped on the stack | |
| 1688 for each register. */ | |
| 1689 #define NUM_REG_ITEMS 3 | |
| 1690 | |
| 1691 /* Individual items aside from the registers. */ | |
| 1692 #ifdef DEBUG | |
| 1693 #define NUM_NONREG_ITEMS 5 /* Includes failure point id. */ | |
| 1694 #else | |
| 1695 #define NUM_NONREG_ITEMS 4 | |
| 1696 #endif | |
| 1697 | |
| 1698 /* We push at most this many items on the stack. */ | |
| 1699 /* We used to use (num_regs - 1), which is the number of registers | |
| 1700 this regexp will save; but that was changed to 5 | |
| 1701 to avoid stack overflow for a regexp with lots of parens. */ | |
| 1702 #define MAX_FAILURE_ITEMS (5 * NUM_REG_ITEMS + NUM_NONREG_ITEMS) | |
| 1703 | |
| 1704 /* We actually push this many items. */ | |
| 1705 #define NUM_FAILURE_ITEMS \ | |
| 1706 ((highest_active_reg - lowest_active_reg + 1) * NUM_REG_ITEMS \ | |
| 1707 + NUM_NONREG_ITEMS) | |
| 1708 | |
| 1709 /* How many items can still be added to the stack without overflowing it. */ | |
| 1710 #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) | |
| 1711 | |
| 1712 | |
| 1713 /* Pops what PUSH_FAIL_STACK pushes. | |
| 1714 | |
| 1715 We restore into the parameters, all of which should be lvalues: | |
| 1716 STR -- the saved data position. | |
| 1717 PAT -- the saved pattern position. | |
| 1718 LOW_REG, HIGH_REG -- the highest and lowest active registers. | |
| 1719 REGSTART, REGEND -- arrays of string positions. | |
| 1720 REG_INFO -- array of information about each subexpression. | |
| 1721 | |
| 1722 Also assumes the variables `fail_stack' and (if debugging), `bufp', | |
| 1723 `pend', `string1', `size1', `string2', and `size2'. */ | |
| 1724 | |
| 456 | 1725 #define POP_FAILURE_POINT(str, pat, low_reg, high_reg, \ |
| 1726 regstart, regend, reg_info) \ | |
| 1727 do { \ | |
| 428 | 1728 DEBUG_STATEMENT (fail_stack_elt_t ffailure_id;) \ |
| 1729 int this_reg; \ | |
| 442 | 1730 const unsigned char *string_temp; \ |
| 428 | 1731 \ |
| 1732 assert (!FAIL_STACK_EMPTY ()); \ | |
| 1733 \ | |
| 1734 /* Remove failure points and point to how many regs pushed. */ \ | |
| 5041 | 1735 DEBUG_FAIL_PRINT1 ("POP_FAILURE_POINT:\n"); \ |
| 1736 DEBUG_FAIL_PRINT2 (" Before pop, next avail: %ld\n", \ | |
| 647 | 1737 (long) fail_stack.avail); \ |
| 5041 | 1738 DEBUG_FAIL_PRINT2 (" size: %ld\n", \ |
| 647 | 1739 (long) fail_stack.size); \ |
| 428 | 1740 \ |
| 1741 assert (fail_stack.avail >= NUM_NONREG_ITEMS); \ | |
| 1742 \ | |
| 1743 DEBUG_POP (&ffailure_id.integer); \ | |
| 5041 | 1744 DEBUG_FAIL_PRINT2 (" Popping failure id: %d\n", \ |
| 647 | 1745 * (int *) &ffailure_id); \ |
| 428 | 1746 \ |
| 1747 /* If the saved string location is NULL, it came from an \ | |
| 1748 on_failure_keep_string_jump opcode, and we want to throw away the \ | |
| 1749 saved NULL, thus retaining our current position in the string. */ \ | |
| 1750 string_temp = POP_FAILURE_POINTER (); \ | |
| 1751 if (string_temp != NULL) \ | |
| 446 | 1752 str = string_temp; \ |
| 428 | 1753 \ |
| 5041 | 1754 DEBUG_FAIL_PRINT2 (" Popping string 0x%lx: `", (long) str); \ |
| 1755 DEBUG_FAIL_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ | |
| 1756 DEBUG_FAIL_PRINT1 ("'\n"); \ | |
| 428 | 1757 \ |
| 1758 pat = (unsigned char *) POP_FAILURE_POINTER (); \ | |
| 5041 | 1759 DEBUG_FAIL_PRINT2 (" Popping pattern 0x%lx: ", (long) pat); \ |
| 1760 DEBUG_FAIL_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ | |
| 428 | 1761 \ |
| 1762 /* Restore register info. */ \ | |
| 647 | 1763 high_reg = POP_FAILURE_INT (); \ |
| 5041 | 1764 DEBUG_FAIL_PRINT2 (" Popping high active reg: %d\n", high_reg); \ |
| 428 | 1765 \ |
| 647 | 1766 low_reg = POP_FAILURE_INT (); \ |
| 5041 | 1767 DEBUG_FAIL_PRINT2 (" Popping low active reg: %d\n", low_reg); \ |
| 428 | 1768 \ |
| 1769 for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \ | |
| 1770 { \ | |
| 5041 | 1771 DEBUG_FAIL_PRINT2 (" Popping reg: %d\n", this_reg); \ |
| 428 | 1772 \ |
| 1773 reg_info[this_reg].word = POP_FAILURE_ELT (); \ | |
| 5041 | 1774 DEBUG_FAIL_PRINT2 (" info: 0x%lx\n", \ |
| 428 | 1775 * (long *) ®_info[this_reg]); \ |
| 1776 \ | |
| 446 | 1777 regend[this_reg] = POP_FAILURE_POINTER (); \ |
| 5041 | 1778 DEBUG_FAIL_PRINT2 (" end: 0x%lx\n", (long) regend[this_reg]); \ |
| 428 | 1779 \ |
| 446 | 1780 regstart[this_reg] = POP_FAILURE_POINTER (); \ |
| 5041 | 1781 DEBUG_FAIL_PRINT2 (" start: 0x%lx\n", (long) regstart[this_reg]); \ |
| 428 | 1782 } \ |
| 1783 \ | |
| 1784 set_regs_matched_done = 0; \ | |
| 1785 DEBUG_STATEMENT (nfailure_points_popped++); \ | |
| 456 | 1786 } while (0) /* POP_FAILURE_POINT */ |
| 428 | 1787 |
| 1788 | |
| 1789 | |
| 1790 /* Structure for per-register (a.k.a. per-group) information. | |
| 1791 Other register information, such as the | |
| 1792 starting and ending positions (which are addresses), and the list of | |
| 1793 inner groups (which is a bits list) are maintained in separate | |
| 1794 variables. | |
| 1795 | |
| 1796 We are making a (strictly speaking) nonportable assumption here: that | |
| 1797 the compiler will pack our bit fields into something that fits into | |
| 1798 the type of `word', i.e., is something that fits into one item on the | |
| 1799 failure stack. */ | |
| 1800 | |
| 1801 typedef union | |
| 1802 { | |
| 1803 fail_stack_elt_t word; | |
| 1804 struct | |
| 1805 { | |
| 1806 /* This field is one if this group can match the empty string, | |
| 1807 zero if not. If not yet determined, `MATCH_NULL_UNSET_VALUE'. */ | |
| 1808 #define MATCH_NULL_UNSET_VALUE 3 | |
| 647 | 1809 unsigned int match_null_string_p : 2; |
| 1810 unsigned int is_active : 1; | |
| 1811 unsigned int matched_something : 1; | |
| 1812 unsigned int ever_matched_something : 1; | |
| 428 | 1813 } bits; |
| 1814 } register_info_type; | |
| 1815 | |
| 1816 #define REG_MATCH_NULL_STRING_P(R) ((R).bits.match_null_string_p) | |
| 1817 #define IS_ACTIVE(R) ((R).bits.is_active) | |
| 1818 #define MATCHED_SOMETHING(R) ((R).bits.matched_something) | |
| 1819 #define EVER_MATCHED_SOMETHING(R) ((R).bits.ever_matched_something) | |
| 1820 | |
| 1821 | |
| 1822 /* Call this when have matched a real character; it sets `matched' flags | |
| 1823 for the subexpressions which we are currently inside. Also records | |
| 1824 that those subexprs have matched. */ | |
| 1825 #define SET_REGS_MATCHED() \ | |
| 1826 do \ | |
| 1827 { \ | |
| 1828 if (!set_regs_matched_done) \ | |
| 1829 { \ | |
| 647 | 1830 int r; \ |
| 428 | 1831 set_regs_matched_done = 1; \ |
| 1832 for (r = lowest_active_reg; r <= highest_active_reg; r++) \ | |
| 1833 { \ | |
| 1834 MATCHED_SOMETHING (reg_info[r]) \ | |
| 1835 = EVER_MATCHED_SOMETHING (reg_info[r]) \ | |
| 1836 = 1; \ | |
| 1837 } \ | |
| 1838 } \ | |
| 1839 } \ | |
| 1840 while (0) | |
| 1841 | |
| 1842 /* Registers are set to a sentinel when they haven't yet matched. */ | |
| 446 | 1843 static unsigned char reg_unset_dummy; |
| 428 | 1844 #define REG_UNSET_VALUE (®_unset_dummy) |
| 1845 #define REG_UNSET(e) ((e) == REG_UNSET_VALUE) | |
| 1846 | |
| 1847 /* Subroutine declarations and macros for regex_compile. */ | |
| 1848 | |
| 1849 /* Fetch the next character in the uncompiled pattern---translating it | |
| 826 | 1850 if necessary. */ |
| 428 | 1851 #define PATFETCH(c) \ |
| 446 | 1852 do { \ |
| 1853 PATFETCH_RAW (c); \ | |
| 826 | 1854 c = RE_TRANSLATE (c); \ |
| 428 | 1855 } while (0) |
| 1856 | |
| 1857 /* Fetch the next character in the uncompiled pattern, with no | |
| 1858 translation. */ | |
| 1859 #define PATFETCH_RAW(c) \ | |
| 1860 do {if (p == pend) return REG_EEND; \ | |
| 1861 assert (p < pend); \ | |
| 867 | 1862 c = itext_ichar (p); \ |
| 1863 INC_IBYTEPTR (p); \ | |
| 428 | 1864 } while (0) |
| 1865 | |
| 1866 /* Go backwards one character in the pattern. */ | |
| 867 | 1867 #define PATUNFETCH DEC_IBYTEPTR (p) |
| 428 | 1868 |
| 1869 /* If `translate' is non-null, return translate[D], else just D. We | |
| 1870 cast the subscript to translate because some data is declared as | |
| 1871 `char *', to avoid warnings when a string constant is passed. But | |
| 1872 when we use a character as a subscript we must make it unsigned. */ | |
| 826 | 1873 #define RE_TRANSLATE(d) \ |
| 1874 (TRANSLATE_P (translate) ? RE_TRANSLATE_1 (d) : (d)) | |
| 428 | 1875 |
| 1876 /* Macros for outputting the compiled pattern into `buffer'. */ | |
| 1877 | |
| 1878 /* If the buffer isn't allocated when it comes in, use this. */ | |
| 1879 #define INIT_BUF_SIZE 32 | |
| 1880 | |
| 1881 /* Make sure we have at least N more bytes of space in buffer. */ | |
| 1882 #define GET_BUFFER_SPACE(n) \ | |
| 647 | 1883 while (buf_end - bufp->buffer + (n) > (ptrdiff_t) bufp->allocated) \ |
| 428 | 1884 EXTEND_BUFFER () |
| 1885 | |
| 1886 /* Make sure we have one more byte of buffer space and then add C to it. */ | |
| 1887 #define BUF_PUSH(c) \ | |
| 1888 do { \ | |
| 1889 GET_BUFFER_SPACE (1); \ | |
| 446 | 1890 *buf_end++ = (unsigned char) (c); \ |
| 428 | 1891 } while (0) |
| 1892 | |
| 1893 | |
| 1894 /* Ensure we have two more bytes of buffer space and then append C1 and C2. */ | |
| 1895 #define BUF_PUSH_2(c1, c2) \ | |
| 1896 do { \ | |
| 1897 GET_BUFFER_SPACE (2); \ | |
| 446 | 1898 *buf_end++ = (unsigned char) (c1); \ |
| 1899 *buf_end++ = (unsigned char) (c2); \ | |
| 428 | 1900 } while (0) |
| 1901 | |
| 1902 | |
| 1903 /* As with BUF_PUSH_2, except for three bytes. */ | |
| 1904 #define BUF_PUSH_3(c1, c2, c3) \ | |
| 1905 do { \ | |
| 1906 GET_BUFFER_SPACE (3); \ | |
| 446 | 1907 *buf_end++ = (unsigned char) (c1); \ |
| 1908 *buf_end++ = (unsigned char) (c2); \ | |
| 1909 *buf_end++ = (unsigned char) (c3); \ | |
| 428 | 1910 } while (0) |
| 1911 | |
| 1912 | |
| 1913 /* Store a jump with opcode OP at LOC to location TO. We store a | |
| 1914 relative address offset by the three bytes the jump itself occupies. */ | |
| 1915 #define STORE_JUMP(op, loc, to) \ | |
| 1916 store_op1 (op, loc, (to) - (loc) - 3) | |
| 1917 | |
| 1918 /* Likewise, for a two-argument jump. */ | |
| 1919 #define STORE_JUMP2(op, loc, to, arg) \ | |
| 1920 store_op2 (op, loc, (to) - (loc) - 3, arg) | |
| 1921 | |
| 446 | 1922 /* Like `STORE_JUMP', but for inserting. Assume `buf_end' is the |
| 1923 buffer end. */ | |
| 428 | 1924 #define INSERT_JUMP(op, loc, to) \ |
| 446 | 1925 insert_op1 (op, loc, (to) - (loc) - 3, buf_end) |
| 1926 | |
| 1927 /* Like `STORE_JUMP2', but for inserting. Assume `buf_end' is the | |
| 1928 buffer end. */ | |
| 428 | 1929 #define INSERT_JUMP2(op, loc, to, arg) \ |
| 446 | 1930 insert_op2 (op, loc, (to) - (loc) - 3, arg, buf_end) |
| 428 | 1931 |
| 1932 | |
| 1933 /* This is not an arbitrary limit: the arguments which represent offsets | |
| 1934 into the pattern are two bytes long. So if 2^16 bytes turns out to | |
| 1935 be too small, many things would have to change. */ | |
| 1936 #define MAX_BUF_SIZE (1L << 16) | |
| 1937 | |
| 1938 | |
| 1939 /* Extend the buffer by twice its current size via realloc and | |
| 1940 reset the pointers that pointed into the old block to point to the | |
| 1941 correct places in the new one. If extending the buffer results in it | |
| 1942 being larger than MAX_BUF_SIZE, then flag memory exhausted. */ | |
| 1333 | 1943 #define EXTEND_BUFFER() \ |
| 1944 do { \ | |
| 1945 re_char *old_buffer = bufp->buffer; \ | |
| 1946 if (bufp->allocated == MAX_BUF_SIZE) \ | |
| 1947 return REG_ESIZE; \ | |
| 1948 bufp->allocated <<= 1; \ | |
| 1949 if (bufp->allocated > MAX_BUF_SIZE) \ | |
| 1950 bufp->allocated = MAX_BUF_SIZE; \ | |
| 1951 bufp->buffer = \ | |
| 1952 (unsigned char *) xrealloc (bufp->buffer, bufp->allocated); \ | |
| 1953 if (bufp->buffer == NULL) \ | |
| 1954 return REG_ESPACE; \ | |
| 1955 /* If the buffer moved, move all the pointers into it. */ \ | |
| 1956 if (old_buffer != bufp->buffer) \ | |
| 1957 { \ | |
| 1958 buf_end = (buf_end - old_buffer) + bufp->buffer; \ | |
| 1959 begalt = (begalt - old_buffer) + bufp->buffer; \ | |
| 1960 if (fixup_alt_jump) \ | |
| 1961 fixup_alt_jump = (fixup_alt_jump - old_buffer) + bufp->buffer; \ | |
| 1962 if (laststart) \ | |
| 1963 laststart = (laststart - old_buffer) + bufp->buffer; \ | |
| 1964 if (pending_exact) \ | |
| 1965 pending_exact = (pending_exact - old_buffer) + bufp->buffer; \ | |
| 1966 } \ | |
| 428 | 1967 } while (0) |
| 1968 | |
| 1969 | |
| 1970 /* Since we have one byte reserved for the register number argument to | |
| 1971 {start,stop}_memory, the maximum number of groups we can report | |
| 1972 things about is what fits in that byte. */ | |
| 1973 #define MAX_REGNUM 255 | |
| 1974 | |
| 1975 /* But patterns can have more than `MAX_REGNUM' registers. We just | |
| 502 | 1976 ignore the excess. |
| 1977 #### not true! groups past this will fail in lots of ways, if we | |
| 1978 ever have to backtrack. | |
| 1979 */ | |
| 647 | 1980 typedef int regnum_t; |
| 428 | 1981 |
| 502 | 1982 #define INIT_REG_TRANSLATE_SIZE 5 |
| 428 | 1983 |
| 1984 /* Macros for the compile stack. */ | |
| 1985 | |
| 1986 /* Since offsets can go either forwards or backwards, this type needs to | |
| 1987 be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ | |
| 1988 typedef int pattern_offset_t; | |
| 1989 | |
| 1990 typedef struct | |
| 1991 { | |
| 1992 pattern_offset_t begalt_offset; | |
| 1993 pattern_offset_t fixup_alt_jump; | |
| 1994 pattern_offset_t inner_group_offset; | |
| 1995 pattern_offset_t laststart_offset; | |
| 1996 regnum_t regnum; | |
| 1997 } compile_stack_elt_t; | |
| 1998 | |
| 1999 | |
| 2000 typedef struct | |
| 2001 { | |
| 2002 compile_stack_elt_t *stack; | |
| 647 | 2003 int size; |
| 2004 int avail; /* Offset of next open position. */ | |
| 428 | 2005 } compile_stack_type; |
| 2006 | |
| 2007 | |
| 2008 #define INIT_COMPILE_STACK_SIZE 32 | |
| 2009 | |
| 2010 #define COMPILE_STACK_EMPTY (compile_stack.avail == 0) | |
| 2011 #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) | |
| 2012 | |
| 2013 /* The next available element. */ | |
| 2014 #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) | |
| 2015 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2016 /* Bits used to implement the multibyte-part of the various character |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2017 classes such as [:alnum:] in a charset's range table. XEmacs; use an |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2018 enum, so they're visible in the debugger. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2019 enum |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2020 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2021 BIT_WORD = (1 << 0), |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2022 BIT_LOWER = (1 << 1), |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2023 BIT_PUNCT = (1 << 2), |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2024 BIT_SPACE = (1 << 3), |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2025 BIT_UPPER = (1 << 4), |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2026 /* XEmacs; we need this, because we unify treatment of ASCII and non-ASCII |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2027 (possible matches) in charset_mule. [:alpha:] matches all characters |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2028 with word syntax, with the exception of [0-9]. We don't need |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2029 BIT_MULTIBYTE. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2030 BIT_ALPHA = (1 << 5) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2031 }; |
| 428 | 2032 |
| 2033 /* Set the bit for character C in a bit vector. */ | |
| 2034 #define SET_LIST_BIT(c) \ | |
| 446 | 2035 (buf_end[((unsigned char) (c)) / BYTEWIDTH] \ |
| 428 | 2036 |= 1 << (((unsigned char) c) % BYTEWIDTH)) |
| 2037 | |
| 2038 #ifdef MULE | |
| 2039 | |
| 2040 /* Set the "bit" for character C in a range table. */ | |
| 2041 #define SET_RANGETAB_BIT(c) put_range_table (rtab, c, c, Qt) | |
| 2042 | |
| 2043 #endif | |
| 2044 | |
| 2045 /* Get the next unsigned number in the uncompiled pattern. */ | |
| 2046 #define GET_UNSIGNED_NUMBER(num) \ | |
| 2047 { if (p != pend) \ | |
| 2048 { \ | |
| 2049 PATFETCH (c); \ | |
| 2050 while (ISDIGIT (c)) \ | |
| 2051 { \ | |
| 2052 if (num < 0) \ | |
| 2053 num = 0; \ | |
| 2054 num = num * 10 + c - '0'; \ | |
| 2055 if (p == pend) \ | |
| 2056 break; \ | |
| 2057 PATFETCH (c); \ | |
| 2058 } \ | |
| 2059 } \ | |
| 2060 } | |
| 2061 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2062 #define CHAR_CLASS_MAX_LENGTH 9 /* Namely, `multibyte'. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2063 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2064 /* Map a string to the char class it names (if any). */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2065 static re_wctype_t |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2066 re_wctype (const char *string) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2067 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2068 if (STREQ (string, "alnum")) return RECC_ALNUM; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2069 else if (STREQ (string, "alpha")) return RECC_ALPHA; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2070 else if (STREQ (string, "word")) return RECC_WORD; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2071 else if (STREQ (string, "ascii")) return RECC_ASCII; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2072 else if (STREQ (string, "nonascii")) return RECC_NONASCII; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2073 else if (STREQ (string, "graph")) return RECC_GRAPH; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2074 else if (STREQ (string, "lower")) return RECC_LOWER; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2075 else if (STREQ (string, "print")) return RECC_PRINT; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2076 else if (STREQ (string, "punct")) return RECC_PUNCT; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2077 else if (STREQ (string, "space")) return RECC_SPACE; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2078 else if (STREQ (string, "upper")) return RECC_UPPER; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2079 else if (STREQ (string, "unibyte")) return RECC_UNIBYTE; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2080 else if (STREQ (string, "multibyte")) return RECC_MULTIBYTE; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2081 else if (STREQ (string, "digit")) return RECC_DIGIT; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2082 else if (STREQ (string, "xdigit")) return RECC_XDIGIT; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2083 else if (STREQ (string, "cntrl")) return RECC_CNTRL; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2084 else if (STREQ (string, "blank")) return RECC_BLANK; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2085 else return RECC_ERROR; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2086 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2087 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2088 /* True if CH is in the char class CC. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2089 static re_bool |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2090 re_iswctype (int ch, re_wctype_t cc) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2091 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2092 #ifdef emacs |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2093 /* This is cheesy, lispbuf isn't available to us when compiling the |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2094 pattern. It's effectively only called (on Mule builds) when the current |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2095 buffer doesn't matter (e.g. for RECC_ASCII, RECC_CNTRL), so it's not a |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2096 big deal. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2097 struct buffer *lispbuf = current_buffer; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2098 #endif |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2099 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2100 switch (cc) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2101 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2102 case RECC_ALNUM: return ISALNUM (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2103 case RECC_ALPHA: return ISALPHA (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2104 case RECC_BLANK: return ISBLANK (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2105 case RECC_CNTRL: return ISCNTRL (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2106 case RECC_DIGIT: return ISDIGIT (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2107 case RECC_GRAPH: return ISGRAPH (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2108 case RECC_LOWER: return ISLOWER (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2109 case RECC_PRINT: return ISPRINT (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2110 case RECC_PUNCT: return ISPUNCT (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2111 case RECC_SPACE: return ISSPACE (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2112 case RECC_UPPER: return ISUPPER (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2113 case RECC_XDIGIT: return ISXDIGIT (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2114 case RECC_ASCII: return ISASCII (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2115 case RECC_NONASCII: case RECC_MULTIBYTE: return !ISASCII (ch); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2116 case RECC_UNIBYTE: return ISUNIBYTE (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2117 case RECC_WORD: return ISWORD (ch) != 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2118 case RECC_ERROR: return false; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2119 default: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2120 abort (); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2121 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2122 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2123 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2124 #ifdef MULE |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2125 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2126 static re_bool |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2127 re_wctype_can_match_non_ascii (re_wctype_t cc) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2128 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2129 switch (cc) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2130 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2131 case RECC_ASCII: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2132 case RECC_UNIBYTE: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2133 case RECC_CNTRL: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2134 case RECC_DIGIT: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2135 case RECC_XDIGIT: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2136 case RECC_BLANK: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2137 return false; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2138 default: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2139 return true; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2140 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2141 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2142 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2143 /* Return a bit-pattern to use in the range-table bits to match multibyte |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2144 chars of class CC. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2145 static unsigned char |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2146 re_wctype_to_bit (re_wctype_t cc) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2147 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2148 switch (cc) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2149 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2150 case RECC_PRINT: case RECC_GRAPH: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2151 case RECC_ALPHA: return BIT_ALPHA; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2152 case RECC_ALNUM: case RECC_WORD: return BIT_WORD; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2153 case RECC_LOWER: return BIT_LOWER; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2154 case RECC_UPPER: return BIT_UPPER; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2155 case RECC_PUNCT: return BIT_PUNCT; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2156 case RECC_SPACE: return BIT_SPACE; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2157 case RECC_MULTIBYTE: case RECC_NONASCII: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2158 case RECC_ASCII: case RECC_DIGIT: case RECC_XDIGIT: case RECC_CNTRL: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2159 case RECC_BLANK: case RECC_UNIBYTE: case RECC_ERROR: return 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2160 default: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2161 abort (); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2162 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2163 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2164 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2165 #endif /* emacs */ |
| 428 | 2166 |
| 2167 static void store_op1 (re_opcode_t op, unsigned char *loc, int arg); | |
| 2168 static void store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2); | |
| 2169 static void insert_op1 (re_opcode_t op, unsigned char *loc, int arg, | |
| 2170 unsigned char *end); | |
| 2171 static void insert_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2, | |
| 2172 unsigned char *end); | |
| 460 | 2173 static re_bool at_begline_loc_p (re_char *pattern, re_char *p, |
| 428 | 2174 reg_syntax_t syntax); |
| 460 | 2175 static re_bool at_endline_loc_p (re_char *p, re_char *pend, int syntax); |
| 2176 static re_bool group_in_compile_stack (compile_stack_type compile_stack, | |
| 428 | 2177 regnum_t regnum); |
| 446 | 2178 static reg_errcode_t compile_range (re_char **p_ptr, re_char *pend, |
| 2179 RE_TRANSLATE_TYPE translate, | |
| 2180 reg_syntax_t syntax, | |
| 428 | 2181 unsigned char *b); |
| 2182 #ifdef MULE | |
| 446 | 2183 static reg_errcode_t compile_extended_range (re_char **p_ptr, |
| 2184 re_char *pend, | |
| 2185 RE_TRANSLATE_TYPE translate, | |
| 428 | 2186 reg_syntax_t syntax, |
| 2187 Lisp_Object rtab); | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2188 static reg_errcode_t compile_char_class (re_wctype_t cc, Lisp_Object rtab, |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2189 Bitbyte *flags_out); |
| 428 | 2190 #endif /* MULE */ |
| 460 | 2191 static re_bool group_match_null_string_p (unsigned char **p, |
| 428 | 2192 unsigned char *end, |
| 2193 register_info_type *reg_info); | |
| 460 | 2194 static re_bool alt_match_null_string_p (unsigned char *p, unsigned char *end, |
| 428 | 2195 register_info_type *reg_info); |
| 460 | 2196 static re_bool common_op_match_null_string_p (unsigned char **p, |
| 428 | 2197 unsigned char *end, |
| 2198 register_info_type *reg_info); | |
| 826 | 2199 static int bcmp_translate (re_char *s1, re_char *s2, |
| 2200 REGISTER int len, RE_TRANSLATE_TYPE translate | |
| 2201 #ifdef emacs | |
| 2202 , Internal_Format fmt, Lisp_Object lispobj | |
| 2203 #endif | |
| 2204 ); | |
| 428 | 2205 static int re_match_2_internal (struct re_pattern_buffer *bufp, |
| 446 | 2206 re_char *string1, int size1, |
| 2207 re_char *string2, int size2, int pos, | |
| 826 | 2208 struct re_registers *regs, int stop |
| 2209 RE_LISP_CONTEXT_ARGS_DECL); | |
| 428 | 2210 |
| 2211 #ifndef MATCH_MAY_ALLOCATE | |
| 2212 | |
| 2213 /* If we cannot allocate large objects within re_match_2_internal, | |
| 2214 we make the fail stack and register vectors global. | |
| 2215 The fail stack, we grow to the maximum size when a regexp | |
| 2216 is compiled. | |
| 2217 The register vectors, we adjust in size each time we | |
| 2218 compile a regexp, according to the number of registers it needs. */ | |
| 2219 | |
| 2220 static fail_stack_type fail_stack; | |
| 2221 | |
| 2222 /* Size with which the following vectors are currently allocated. | |
| 2223 That is so we can make them bigger as needed, | |
| 2224 but never make them smaller. */ | |
| 2225 static int regs_allocated_size; | |
| 2226 | |
| 446 | 2227 static re_char ** regstart, ** regend; |
| 2228 static re_char ** old_regstart, ** old_regend; | |
| 2229 static re_char **best_regstart, **best_regend; | |
| 428 | 2230 static register_info_type *reg_info; |
| 446 | 2231 static re_char **reg_dummy; |
| 428 | 2232 static register_info_type *reg_info_dummy; |
| 2233 | |
| 2234 /* Make the register vectors big enough for NUM_REGS registers, | |
| 2235 but don't make them smaller. */ | |
| 2236 | |
| 2237 static | |
| 2238 regex_grow_registers (int num_regs) | |
| 2239 { | |
| 2240 if (num_regs > regs_allocated_size) | |
| 2241 { | |
| 551 | 2242 RETALLOC (regstart, num_regs, re_char *); |
| 2243 RETALLOC (regend, num_regs, re_char *); | |
| 2244 RETALLOC (old_regstart, num_regs, re_char *); | |
| 2245 RETALLOC (old_regend, num_regs, re_char *); | |
| 2246 RETALLOC (best_regstart, num_regs, re_char *); | |
| 2247 RETALLOC (best_regend, num_regs, re_char *); | |
| 2248 RETALLOC (reg_info, num_regs, register_info_type); | |
| 2249 RETALLOC (reg_dummy, num_regs, re_char *); | |
| 2250 RETALLOC (reg_info_dummy, num_regs, register_info_type); | |
| 428 | 2251 |
| 2252 regs_allocated_size = num_regs; | |
| 2253 } | |
| 2254 } | |
| 2255 | |
| 2256 #endif /* not MATCH_MAY_ALLOCATE */ | |
| 2257 | |
| 2258 /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. | |
| 2259 Returns one of error codes defined in `regex.h', or zero for success. | |
| 2260 | |
| 2261 Assumes the `allocated' (and perhaps `buffer') and `translate' | |
| 2262 fields are set in BUFP on entry. | |
| 2263 | |
| 2264 If it succeeds, results are put in BUFP (if it returns an error, the | |
| 2265 contents of BUFP are undefined): | |
| 2266 `buffer' is the compiled pattern; | |
| 2267 `syntax' is set to SYNTAX; | |
| 2268 `used' is set to the length of the compiled pattern; | |
| 2269 `fastmap_accurate' is zero; | |
| 502 | 2270 `re_ngroups' is the number of groups/subexpressions (including shy |
| 2271 groups) in PATTERN; | |
| 2272 `re_nsub' is the number of non-shy groups in PATTERN; | |
| 428 | 2273 `not_bol' and `not_eol' are zero; |
| 2274 | |
| 2275 The `fastmap' and `newline_anchor' fields are neither | |
| 2276 examined nor set. */ | |
| 2277 | |
| 2278 /* Return, freeing storage we allocated. */ | |
| 1726 | 2279 #define FREE_STACK_RETURN(value) \ |
| 2280 do \ | |
| 2281 { \ | |
|
4976
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
2282 xfree (compile_stack.stack); \ |
| 1726 | 2283 return value; \ |
| 1333 | 2284 } while (0) |
| 428 | 2285 |
| 2286 static reg_errcode_t | |
| 446 | 2287 regex_compile (re_char *pattern, int size, reg_syntax_t syntax, |
| 428 | 2288 struct re_pattern_buffer *bufp) |
| 2289 { | |
| 2290 /* We fetch characters from PATTERN here. We declare these as int | |
| 2291 (or possibly long) so that chars above 127 can be used as | |
| 2292 array indices. The macros that fetch a character from the pattern | |
| 2293 make sure to coerce to unsigned char before assigning, so we won't | |
| 2294 get bitten by negative numbers here. */ | |
| 2295 /* XEmacs change: used to be unsigned char. */ | |
| 2296 REGISTER EMACS_INT c, c1; | |
| 2297 | |
| 2298 /* A random temporary spot in PATTERN. */ | |
| 446 | 2299 re_char *p1; |
| 428 | 2300 |
| 2301 /* Points to the end of the buffer, where we should append. */ | |
| 446 | 2302 REGISTER unsigned char *buf_end; |
| 428 | 2303 |
| 2304 /* Keeps track of unclosed groups. */ | |
| 2305 compile_stack_type compile_stack; | |
| 2306 | |
| 2307 /* Points to the current (ending) position in the pattern. */ | |
| 446 | 2308 re_char *p = pattern; |
| 2309 re_char *pend = pattern + size; | |
| 428 | 2310 |
| 2311 /* How to translate the characters in the pattern. */ | |
| 446 | 2312 RE_TRANSLATE_TYPE translate = bufp->translate; |
| 428 | 2313 |
| 2314 /* Address of the count-byte of the most recently inserted `exactn' | |
| 2315 command. This makes it possible to tell if a new exact-match | |
| 2316 character can be added to that command or if the character requires | |
| 2317 a new `exactn' command. */ | |
| 2318 unsigned char *pending_exact = 0; | |
| 2319 | |
| 2320 /* Address of start of the most recently finished expression. | |
| 2321 This tells, e.g., postfix * where to find the start of its | |
| 2322 operand. Reset at the beginning of groups and alternatives. */ | |
| 2323 unsigned char *laststart = 0; | |
| 2324 | |
| 2325 /* Address of beginning of regexp, or inside of last group. */ | |
| 2326 unsigned char *begalt; | |
| 2327 | |
| 2328 /* Place in the uncompiled pattern (i.e., the {) to | |
| 2329 which to go back if the interval is invalid. */ | |
| 446 | 2330 re_char *beg_interval; |
| 428 | 2331 |
| 2332 /* Address of the place where a forward jump should go to the end of | |
| 2333 the containing expression. Each alternative of an `or' -- except the | |
| 2334 last -- ends with a forward jump of this sort. */ | |
| 2335 unsigned char *fixup_alt_jump = 0; | |
| 2336 | |
| 2337 /* Counts open-groups as they are encountered. Remembered for the | |
| 2338 matching close-group on the compile stack, so the same register | |
| 2339 number is put in the stop_memory as the start_memory. */ | |
| 2340 regnum_t regnum = 0; | |
| 2341 | |
| 2342 #ifdef DEBUG | |
| 5041 | 2343 if (debug_regexps & RE_DEBUG_COMPILATION) |
| 428 | 2344 { |
| 647 | 2345 int debug_count; |
| 428 | 2346 |
| 5041 | 2347 DEBUG_PRINT1 ("\nCompiling pattern: "); |
| 428 | 2348 for (debug_count = 0; debug_count < size; debug_count++) |
| 2349 putchar (pattern[debug_count]); | |
| 2350 putchar ('\n'); | |
| 2351 } | |
| 2352 #endif /* DEBUG */ | |
| 2353 | |
| 2354 /* Initialize the compile stack. */ | |
| 2355 compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); | |
| 2356 if (compile_stack.stack == NULL) | |
| 2357 return REG_ESPACE; | |
| 2358 | |
| 2359 compile_stack.size = INIT_COMPILE_STACK_SIZE; | |
| 2360 compile_stack.avail = 0; | |
| 2361 | |
| 2362 /* Initialize the pattern buffer. */ | |
| 2363 bufp->syntax = syntax; | |
| 2364 bufp->fastmap_accurate = 0; | |
| 2365 bufp->not_bol = bufp->not_eol = 0; | |
| 2366 | |
| 2367 /* Set `used' to zero, so that if we return an error, the pattern | |
| 2368 printer (for debugging) will think there's no pattern. We reset it | |
| 2369 at the end. */ | |
| 2370 bufp->used = 0; | |
| 2371 | |
| 2372 /* Always count groups, whether or not bufp->no_sub is set. */ | |
| 2373 bufp->re_nsub = 0; | |
| 502 | 2374 bufp->re_ngroups = 0; |
| 2375 | |
| 2376 bufp->warned_about_incompatible_back_references = 0; | |
| 2377 | |
| 2378 if (bufp->external_to_internal_register == 0) | |
| 2379 { | |
| 2380 bufp->external_to_internal_register_size = INIT_REG_TRANSLATE_SIZE; | |
| 2381 RETALLOC (bufp->external_to_internal_register, | |
| 2382 bufp->external_to_internal_register_size, | |
| 2383 int); | |
| 2384 } | |
| 2385 | |
| 2386 { | |
| 2387 int i; | |
| 2388 | |
| 2389 bufp->external_to_internal_register[0] = 0; | |
| 2390 for (i = 1; i < bufp->external_to_internal_register_size; i++) | |
| 2391 bufp->external_to_internal_register[i] = (int) 0xDEADBEEF; | |
| 2392 } | |
| 428 | 2393 |
| 2394 #if !defined (emacs) && !defined (SYNTAX_TABLE) | |
| 2395 /* Initialize the syntax table. */ | |
| 2396 init_syntax_once (); | |
| 2397 #endif | |
| 2398 | |
| 2399 if (bufp->allocated == 0) | |
| 2400 { | |
| 2401 if (bufp->buffer) | |
| 2402 { /* If zero allocated, but buffer is non-null, try to realloc | |
| 2403 enough space. This loses if buffer's address is bogus, but | |
| 2404 that is the user's responsibility. */ | |
| 2405 RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char); | |
| 2406 } | |
| 2407 else | |
| 2408 { /* Caller did not allocate a buffer. Do it for them. */ | |
| 2409 bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char); | |
| 2410 } | |
| 2411 if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE); | |
| 2412 | |
| 2413 bufp->allocated = INIT_BUF_SIZE; | |
| 2414 } | |
| 2415 | |
| 446 | 2416 begalt = buf_end = bufp->buffer; |
| 428 | 2417 |
| 2418 /* Loop through the uncompiled pattern until we're at the end. */ | |
| 2419 while (p != pend) | |
| 2420 { | |
| 2421 PATFETCH (c); | |
| 2422 | |
| 2423 switch (c) | |
| 2424 { | |
| 2425 case '^': | |
| 2426 { | |
| 2427 if ( /* If at start of pattern, it's an operator. */ | |
| 2428 p == pattern + 1 | |
| 2429 /* If context independent, it's an operator. */ | |
| 2430 || syntax & RE_CONTEXT_INDEP_ANCHORS | |
| 2431 /* Otherwise, depends on what's come before. */ | |
| 2432 || at_begline_loc_p (pattern, p, syntax)) | |
| 2433 BUF_PUSH (begline); | |
| 2434 else | |
| 2435 goto normal_char; | |
| 2436 } | |
| 2437 break; | |
| 2438 | |
| 2439 | |
| 2440 case '$': | |
| 2441 { | |
| 2442 if ( /* If at end of pattern, it's an operator. */ | |
| 2443 p == pend | |
| 2444 /* If context independent, it's an operator. */ | |
| 2445 || syntax & RE_CONTEXT_INDEP_ANCHORS | |
| 2446 /* Otherwise, depends on what's next. */ | |
| 2447 || at_endline_loc_p (p, pend, syntax)) | |
| 2448 BUF_PUSH (endline); | |
| 2449 else | |
| 2450 goto normal_char; | |
| 2451 } | |
| 2452 break; | |
| 2453 | |
| 2454 | |
| 2455 case '+': | |
| 2456 case '?': | |
| 2457 if ((syntax & RE_BK_PLUS_QM) | |
| 2458 || (syntax & RE_LIMITED_OPS)) | |
| 2459 goto normal_char; | |
| 2460 handle_plus: | |
| 2461 case '*': | |
| 2462 /* If there is no previous pattern... */ | |
| 2463 if (!laststart) | |
| 2464 { | |
| 2465 if (syntax & RE_CONTEXT_INVALID_OPS) | |
| 2466 FREE_STACK_RETURN (REG_BADRPT); | |
| 2467 else if (!(syntax & RE_CONTEXT_INDEP_OPS)) | |
| 2468 goto normal_char; | |
| 2469 } | |
| 2470 | |
| 2471 { | |
| 2472 /* true means zero/many matches are allowed. */ | |
| 460 | 2473 re_bool zero_times_ok = c != '+'; |
| 2474 re_bool many_times_ok = c != '?'; | |
| 428 | 2475 |
| 2476 /* true means match shortest string possible. */ | |
| 460 | 2477 re_bool minimal = false; |
| 428 | 2478 |
| 2479 /* If there is a sequence of repetition chars, collapse it | |
| 2480 down to just one (the right one). We can't combine | |
| 2481 interval operators with these because of, e.g., `a{2}*', | |
| 2482 which should only match an even number of `a's. */ | |
| 2483 while (p != pend) | |
| 2484 { | |
| 2485 PATFETCH (c); | |
| 2486 | |
| 2487 if (c == '*' || (!(syntax & RE_BK_PLUS_QM) | |
| 2488 && (c == '+' || c == '?'))) | |
| 2489 ; | |
| 2490 | |
| 2491 else if (syntax & RE_BK_PLUS_QM && c == '\\') | |
| 2492 { | |
| 2493 if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
| 2494 | |
| 2495 PATFETCH (c1); | |
| 2496 if (!(c1 == '+' || c1 == '?')) | |
| 2497 { | |
| 2498 PATUNFETCH; | |
| 2499 PATUNFETCH; | |
| 2500 break; | |
| 2501 } | |
| 2502 | |
| 2503 c = c1; | |
| 2504 } | |
| 2505 else | |
| 2506 { | |
| 2507 PATUNFETCH; | |
| 2508 break; | |
| 2509 } | |
| 2510 | |
| 2511 /* If we get here, we found another repeat character. */ | |
| 2512 if (!(syntax & RE_NO_MINIMAL_MATCHING)) | |
| 2513 { | |
| 440 | 2514 /* "*?" and "+?" and "??" are okay (and mean match |
| 2515 minimally), but other sequences (such as "*??" and | |
| 2516 "+++") are rejected (reserved for future use). */ | |
| 428 | 2517 if (minimal || c != '?') |
| 2518 FREE_STACK_RETURN (REG_BADRPT); | |
| 2519 minimal = true; | |
| 2520 } | |
| 2521 else | |
| 2522 { | |
| 2523 zero_times_ok |= c != '+'; | |
| 2524 many_times_ok |= c != '?'; | |
| 2525 } | |
| 2526 } | |
| 2527 | |
| 2528 /* Star, etc. applied to an empty pattern is equivalent | |
| 2529 to an empty pattern. */ | |
| 2530 if (!laststart) | |
| 2531 break; | |
| 2532 | |
| 2533 /* Now we know whether zero matches is allowed | |
| 2534 and whether two or more matches is allowed | |
| 2535 and whether we want minimal or maximal matching. */ | |
| 2536 if (minimal) | |
| 2537 { | |
| 2538 if (!many_times_ok) | |
| 2539 { | |
| 2540 /* "a??" becomes: | |
| 2541 0: /on_failure_jump to 6 | |
| 2542 3: /jump to 9 | |
| 2543 6: /exactn/1/A | |
| 2544 9: end of pattern. | |
| 2545 */ | |
| 2546 GET_BUFFER_SPACE (6); | |
| 446 | 2547 INSERT_JUMP (jump, laststart, buf_end + 3); |
| 2548 buf_end += 3; | |
| 428 | 2549 INSERT_JUMP (on_failure_jump, laststart, laststart + 6); |
| 446 | 2550 buf_end += 3; |
| 428 | 2551 } |
| 2552 else if (zero_times_ok) | |
| 2553 { | |
| 2554 /* "a*?" becomes: | |
| 2555 0: /jump to 6 | |
| 2556 3: /exactn/1/A | |
| 2557 6: /on_failure_jump to 3 | |
| 2558 9: end of pattern. | |
| 2559 */ | |
| 2560 GET_BUFFER_SPACE (6); | |
| 446 | 2561 INSERT_JUMP (jump, laststart, buf_end + 3); |
| 2562 buf_end += 3; | |
| 2563 STORE_JUMP (on_failure_jump, buf_end, laststart + 3); | |
| 2564 buf_end += 3; | |
| 428 | 2565 } |
| 2566 else | |
| 2567 { | |
| 2568 /* "a+?" becomes: | |
| 2569 0: /exactn/1/A | |
| 2570 3: /on_failure_jump to 0 | |
| 2571 6: end of pattern. | |
| 2572 */ | |
| 2573 GET_BUFFER_SPACE (3); | |
| 446 | 2574 STORE_JUMP (on_failure_jump, buf_end, laststart); |
| 2575 buf_end += 3; | |
| 428 | 2576 } |
| 2577 } | |
| 2578 else | |
| 2579 { | |
| 2580 /* Are we optimizing this jump? */ | |
| 460 | 2581 re_bool keep_string_p = false; |
| 428 | 2582 |
| 2583 if (many_times_ok) | |
| 446 | 2584 { /* More than one repetition is allowed, so put in |
| 2585 at the end a backward relative jump from | |
| 2586 `buf_end' to before the next jump we're going | |
| 2587 to put in below (which jumps from laststart to | |
| 2588 after this jump). | |
| 428 | 2589 |
| 2590 But if we are at the `*' in the exact sequence `.*\n', | |
| 2591 insert an unconditional jump backwards to the ., | |
| 2592 instead of the beginning of the loop. This way we only | |
| 2593 push a failure point once, instead of every time | |
| 2594 through the loop. */ | |
| 2595 assert (p - 1 > pattern); | |
| 2596 | |
| 2597 /* Allocate the space for the jump. */ | |
| 2598 GET_BUFFER_SPACE (3); | |
| 2599 | |
| 2600 /* We know we are not at the first character of the | |
| 2601 pattern, because laststart was nonzero. And we've | |
| 2602 already incremented `p', by the way, to be the | |
| 2603 character after the `*'. Do we have to do something | |
| 2604 analogous here for null bytes, because of | |
| 2605 RE_DOT_NOT_NULL? */ | |
| 446 | 2606 if (*(p - 2) == '.' |
| 428 | 2607 && zero_times_ok |
| 446 | 2608 && p < pend && *p == '\n' |
| 428 | 2609 && !(syntax & RE_DOT_NEWLINE)) |
| 2610 { /* We have .*\n. */ | |
| 446 | 2611 STORE_JUMP (jump, buf_end, laststart); |
| 428 | 2612 keep_string_p = true; |
| 2613 } | |
| 2614 else | |
| 2615 /* Anything else. */ | |
| 446 | 2616 STORE_JUMP (maybe_pop_jump, buf_end, laststart - 3); |
| 428 | 2617 |
| 2618 /* We've added more stuff to the buffer. */ | |
| 446 | 2619 buf_end += 3; |
| 428 | 2620 } |
| 2621 | |
| 446 | 2622 /* On failure, jump from laststart to buf_end + 3, |
| 2623 which will be the end of the buffer after this jump | |
| 2624 is inserted. */ | |
| 428 | 2625 GET_BUFFER_SPACE (3); |
| 2626 INSERT_JUMP (keep_string_p ? on_failure_keep_string_jump | |
| 2627 : on_failure_jump, | |
| 446 | 2628 laststart, buf_end + 3); |
| 2629 buf_end += 3; | |
| 428 | 2630 |
| 2631 if (!zero_times_ok) | |
| 2632 { | |
| 2633 /* At least one repetition is required, so insert a | |
| 2634 `dummy_failure_jump' before the initial | |
| 2635 `on_failure_jump' instruction of the loop. This | |
| 2636 effects a skip over that instruction the first time | |
| 2637 we hit that loop. */ | |
| 2638 GET_BUFFER_SPACE (3); | |
| 2639 INSERT_JUMP (dummy_failure_jump, laststart, laststart + 6); | |
| 446 | 2640 buf_end += 3; |
| 428 | 2641 } |
| 2642 } | |
| 2643 pending_exact = 0; | |
| 2644 } | |
| 2645 break; | |
| 2646 | |
| 2647 | |
| 2648 case '.': | |
| 446 | 2649 laststart = buf_end; |
| 428 | 2650 BUF_PUSH (anychar); |
| 2651 break; | |
| 2652 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2653 #ifdef MULE |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2654 #define MAYBE_START_OVER_WITH_EXTENDED(ch) \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2655 if (ch >= 0x80) \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2656 { \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2657 goto start_over_with_extended; \ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2658 } while (0) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2659 #else |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2660 #define MAYBE_START_OVER_WITH_EXTENDED(ch) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2661 #endif |
| 428 | 2662 |
| 2663 case '[': | |
| 2664 { | |
| 2665 /* XEmacs change: this whole section */ | |
| 460 | 2666 re_bool had_char_class = false; |
| 428 | 2667 |
| 2668 if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
| 2669 | |
| 2670 /* Ensure that we have enough space to push a charset: the | |
| 2671 opcode, the length count, and the bitset; 34 bytes in all. */ | |
| 2672 GET_BUFFER_SPACE (34); | |
| 2673 | |
| 446 | 2674 laststart = buf_end; |
| 428 | 2675 |
| 2676 /* We test `*p == '^' twice, instead of using an if | |
| 2677 statement, so we only need one BUF_PUSH. */ | |
| 2678 BUF_PUSH (*p == '^' ? charset_not : charset); | |
| 2679 if (*p == '^') | |
| 2680 p++; | |
| 2681 | |
| 2682 /* Remember the first position in the bracket expression. */ | |
| 2683 p1 = p; | |
| 2684 | |
| 2685 /* Push the number of bytes in the bitmap. */ | |
| 2686 BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); | |
| 2687 | |
| 2688 /* Clear the whole map. */ | |
| 446 | 2689 memset (buf_end, 0, (1 << BYTEWIDTH) / BYTEWIDTH); |
| 428 | 2690 |
| 2691 /* charset_not matches newline according to a syntax bit. */ | |
| 446 | 2692 if ((re_opcode_t) buf_end[-2] == charset_not |
| 428 | 2693 && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) |
| 2694 SET_LIST_BIT ('\n'); | |
| 2695 | |
| 2696 /* Read in characters and ranges, setting map bits. */ | |
| 2697 for (;;) | |
| 2698 { | |
| 2699 if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
| 2700 | |
| 446 | 2701 PATFETCH (c); |
| 428 | 2702 |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2703 /* Frumble-bumble, we may have found some extended chars. |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2704 Need to start over, process everything using the general |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2705 extended-char mechanism, and need to use charset_mule and |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2706 charset_mule_not instead of charset and charset_not. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2707 MAYBE_START_OVER_WITH_EXTENDED (c); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2708 |
| 428 | 2709 /* \ might escape characters inside [...] and [^...]. */ |
| 2710 if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') | |
| 2711 { | |
| 2712 if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
| 2713 | |
| 446 | 2714 PATFETCH (c1); |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2715 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2716 MAYBE_START_OVER_WITH_EXTENDED (c1); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2717 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2718 SET_LIST_BIT (c1); |
| 428 | 2719 continue; |
| 2720 } | |
| 2721 | |
| 2722 /* Could be the end of the bracket expression. If it's | |
| 2723 not (i.e., when the bracket expression is `[]' so | |
| 2724 far), the ']' character bit gets set way below. */ | |
| 2725 if (c == ']' && p != p1 + 1) | |
| 2726 break; | |
| 2727 | |
| 2728 /* Look ahead to see if it's a range when the last thing | |
| 2729 was a character class. */ | |
| 2730 if (had_char_class && c == '-' && *p != ']') | |
| 2731 FREE_STACK_RETURN (REG_ERANGE); | |
| 2732 | |
| 2733 /* Look ahead to see if it's a range when the last thing | |
| 2734 was a character: if this is a hyphen not at the | |
| 2735 beginning or the end of a list, then it's the range | |
| 2736 operator. */ | |
| 2737 if (c == '-' | |
| 2738 && !(p - 2 >= pattern && p[-2] == '[') | |
| 446 | 2739 && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^') |
| 428 | 2740 && *p != ']') |
| 2741 { | |
| 2742 reg_errcode_t ret; | |
| 2743 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2744 MAYBE_START_OVER_WITH_EXTENDED (*(unsigned char *)p); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2745 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2746 ret = compile_range (&p, pend, translate, syntax, |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2747 buf_end); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2748 |
| 428 | 2749 if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); |
| 2750 } | |
| 2751 | |
| 2752 else if (p[0] == '-' && p[1] != ']') | |
| 2753 { /* This handles ranges made up of characters only. */ | |
| 2754 reg_errcode_t ret; | |
| 2755 | |
| 2756 /* Move past the `-'. */ | |
| 2757 PATFETCH (c1); | |
| 2758 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2759 MAYBE_START_OVER_WITH_EXTENDED (*(unsigned char *)p); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2760 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2761 ret = compile_range (&p, pend, translate, syntax, buf_end); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2762 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2763 if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2764 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2765 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2766 /* See if we're at the beginning of a possible character |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2767 class. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2768 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2769 else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2770 { /* Leave room for the null. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2771 char str[CHAR_CLASS_MAX_LENGTH + 1]; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2772 int ch = 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2773 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2774 PATFETCH (c); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2775 c1 = 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2776 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2777 /* If pattern is `[[:'. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2778 if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2779 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2780 for (;;) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2781 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2782 PATFETCH (c); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2783 if ((c == ':' && *p == ']') || p == pend) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2784 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2785 if (c1 < CHAR_CLASS_MAX_LENGTH) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2786 str[c1++] = c; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2787 else |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2788 /* This is in any case an invalid class name. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2789 str[0] = '\0'; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2790 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2791 str[c1] = '\0'; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2792 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2793 /* If isn't a word bracketed by `[:' and `:]': |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2794 undo the ending character, the letters, and leave |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2795 the leading `:' and `[' (but set bits for them). */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2796 if (c == ':' && *p == ']') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2797 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2798 re_wctype_t cc = re_wctype (str); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2799 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2800 if (cc == RECC_ERROR) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2801 FREE_STACK_RETURN (REG_ECTYPE); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2802 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2803 /* Throw away the ] at the end of the character |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2804 class. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2805 PATFETCH (c); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2806 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2807 if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2808 |
| 428 | 2809 #ifdef MULE |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2810 if (re_wctype_can_match_non_ascii (cc)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2811 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2812 goto start_over_with_extended; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2813 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2814 #endif /* MULE */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2815 for (ch = 0; ch < (1 << BYTEWIDTH); ++ch) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2816 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2817 if (re_iswctype (ch, cc)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2818 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2819 SET_LIST_BIT (ch); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2820 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2821 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2822 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2823 had_char_class = true; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2824 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2825 else |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2826 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2827 c1++; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2828 while (c1--) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2829 PATUNFETCH; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2830 SET_LIST_BIT ('['); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2831 SET_LIST_BIT (':'); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2832 had_char_class = false; |
| 428 | 2833 } |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2834 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2835 else |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2836 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2837 had_char_class = false; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2838 SET_LIST_BIT (c); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2839 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2840 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2841 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2842 /* Discard any (non)matching list bytes that are all 0 at the |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2843 end of the map. Decrease the map-length byte too. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2844 while ((int) buf_end[-1] > 0 && buf_end[buf_end[-1] - 1] == 0) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2845 buf_end[-1]--; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2846 buf_end += buf_end[-1]; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2847 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2848 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2849 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2850 #ifdef MULE |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2851 start_over_with_extended: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2852 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2853 REGISTER Lisp_Object rtab = Qnil; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2854 Bitbyte flags = 0; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2855 int bytes_needed = sizeof (flags); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2856 re_bool had_char_class = false; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2857 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2858 /* There are extended chars here, which means we need to use the |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2859 unified range-table format. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2860 if (buf_end[-2] == charset) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2861 buf_end[-2] = charset_mule; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2862 else |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2863 buf_end[-2] = charset_mule_not; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2864 buf_end--; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2865 p = p1; /* go back to the beginning of the charset, after |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2866 a possible ^. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2867 rtab = Vthe_lisp_rangetab; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2868 Fclear_range_table (rtab); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2869 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2870 /* charset_not matches newline according to a syntax bit. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2871 if ((re_opcode_t) buf_end[-1] == charset_mule_not |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2872 && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2873 SET_RANGETAB_BIT ('\n'); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2874 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2875 /* Read in characters and ranges, setting map bits. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2876 for (;;) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2877 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2878 if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2879 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2880 PATFETCH (c); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2881 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2882 /* \ might escape characters inside [...] and [^...]. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2883 if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2884 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2885 if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2886 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2887 PATFETCH (c1); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2888 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2889 SET_RANGETAB_BIT (c1); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2890 continue; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2891 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2892 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2893 /* Could be the end of the bracket expression. If it's |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2894 not (i.e., when the bracket expression is `[]' so |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2895 far), the ']' character bit gets set way below. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2896 if (c == ']' && p != p1 + 1) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2897 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2898 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2899 /* Look ahead to see if it's a range when the last thing |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2900 was a character class. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2901 if (had_char_class && c == '-' && *p != ']') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2902 FREE_STACK_RETURN (REG_ERANGE); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2903 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2904 /* Look ahead to see if it's a range when the last thing |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2905 was a character: if this is a hyphen not at the |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2906 beginning or the end of a list, then it's the range |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2907 operator. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2908 if (c == '-' |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2909 && !(p - 2 >= pattern && p[-2] == '[') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2910 && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2911 && *p != ']') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2912 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2913 reg_errcode_t ret; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2914 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2915 ret = compile_extended_range (&p, pend, translate, syntax, |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2916 rtab); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2917 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2918 if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2919 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2920 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2921 else if (p[0] == '-' && p[1] != ']') |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2922 { /* This handles ranges made up of characters only. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2923 reg_errcode_t ret; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2924 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2925 /* Move past the `-'. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2926 PATFETCH (c1); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2927 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2928 ret = compile_extended_range (&p, pend, translate, |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2929 syntax, rtab); |
| 428 | 2930 if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); |
| 2931 } | |
| 2932 | |
| 2933 /* See if we're at the beginning of a possible character | |
| 2934 class. */ | |
| 2935 | |
| 2936 else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') | |
| 2937 { /* Leave room for the null. */ | |
| 2938 char str[CHAR_CLASS_MAX_LENGTH + 1]; | |
| 2939 | |
| 2940 PATFETCH (c); | |
| 2941 c1 = 0; | |
| 2942 | |
| 2943 /* If pattern is `[[:'. */ | |
| 2944 if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
| 2945 | |
| 2946 for (;;) | |
| 2947 { | |
| 2948 PATFETCH (c); | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2949 if ((c == ':' && *p == ']') || p == pend) |
| 428 | 2950 break; |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2951 if (c1 < CHAR_CLASS_MAX_LENGTH) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2952 str[c1++] = c; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2953 else |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2954 /* This is in any case an invalid class name. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2955 str[0] = '\0'; |
| 428 | 2956 } |
| 2957 str[c1] = '\0'; | |
| 2958 | |
| 446 | 2959 /* If isn't a word bracketed by `[:' and `:]': |
| 428 | 2960 undo the ending character, the letters, and leave |
| 2961 the leading `:' and `[' (but set bits for them). */ | |
| 2962 if (c == ':' && *p == ']') | |
| 2963 { | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2964 re_wctype_t cc = re_wctype (str); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2965 reg_errcode_t ret = REG_NOERROR; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2966 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2967 if (cc == RECC_ERROR) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2968 FREE_STACK_RETURN (REG_ECTYPE); |
| 428 | 2969 |
| 2970 /* Throw away the ] at the end of the character | |
| 2971 class. */ | |
| 2972 PATFETCH (c); | |
| 2973 | |
| 2974 if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
| 2975 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2976 ret = compile_char_class (cc, rtab, &flags); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2977 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2978 if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2979 |
| 428 | 2980 had_char_class = true; |
| 2981 } | |
| 2982 else | |
| 2983 { | |
| 2984 c1++; | |
| 2985 while (c1--) | |
| 2986 PATUNFETCH; | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2987 SET_RANGETAB_BIT ('['); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2988 SET_RANGETAB_BIT (':'); |
| 428 | 2989 had_char_class = false; |
| 2990 } | |
| 2991 } | |
| 2992 else | |
| 2993 { | |
| 2994 had_char_class = false; | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2995 SET_RANGETAB_BIT (c); |
| 428 | 2996 } |
| 2997 } | |
| 2998 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
2999 bytes_needed += unified_range_table_bytes_needed (rtab); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3000 GET_BUFFER_SPACE (bytes_needed); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3001 *buf_end++ = flags; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3002 unified_range_table_copy_data (rtab, buf_end); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3003 buf_end += unified_range_table_bytes_used (buf_end); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3004 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3005 } |
| 428 | 3006 #endif /* MULE */ |
| 3007 | |
| 3008 case '(': | |
| 3009 if (syntax & RE_NO_BK_PARENS) | |
| 3010 goto handle_open; | |
| 3011 else | |
| 3012 goto normal_char; | |
| 3013 | |
| 3014 | |
| 3015 case ')': | |
| 3016 if (syntax & RE_NO_BK_PARENS) | |
| 3017 goto handle_close; | |
| 3018 else | |
| 3019 goto normal_char; | |
| 3020 | |
| 3021 | |
| 3022 case '\n': | |
| 3023 if (syntax & RE_NEWLINE_ALT) | |
| 3024 goto handle_alt; | |
| 3025 else | |
| 3026 goto normal_char; | |
| 3027 | |
| 3028 | |
| 3029 case '|': | |
| 3030 if (syntax & RE_NO_BK_VBAR) | |
| 3031 goto handle_alt; | |
| 3032 else | |
| 3033 goto normal_char; | |
| 3034 | |
| 3035 | |
| 3036 case '{': | |
| 3037 if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) | |
| 3038 goto handle_interval; | |
| 3039 else | |
| 3040 goto normal_char; | |
| 3041 | |
| 3042 | |
| 3043 case '\\': | |
| 3044 if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
| 3045 | |
| 3046 /* Do not translate the character after the \, so that we can | |
| 3047 distinguish, e.g., \B from \b, even if we normally would | |
| 3048 translate, e.g., B to b. */ | |
| 3049 PATFETCH_RAW (c); | |
| 3050 | |
| 3051 switch (c) | |
| 3052 { | |
| 3053 case '(': | |
| 3054 if (syntax & RE_NO_BK_PARENS) | |
| 3055 goto normal_backslash; | |
| 3056 | |
| 3057 handle_open: | |
| 3058 { | |
| 3059 regnum_t r; | |
| 502 | 3060 int shy = 0; |
| 428 | 3061 |
| 3062 if (!(syntax & RE_NO_SHY_GROUPS) | |
| 3063 && p != pend | |
| 446 | 3064 && *p == '?') |
| 428 | 3065 { |
| 3066 p++; | |
| 446 | 3067 PATFETCH (c); |
| 428 | 3068 switch (c) |
| 3069 { | |
| 3070 case ':': /* shy groups */ | |
| 502 | 3071 shy = 1; |
| 428 | 3072 break; |
| 3073 | |
| 3074 /* All others are reserved for future constructs. */ | |
| 3075 default: | |
| 3076 FREE_STACK_RETURN (REG_BADPAT); | |
| 3077 } | |
| 3078 } | |
| 502 | 3079 |
| 3080 r = ++regnum; | |
| 3081 bufp->re_ngroups++; | |
| 3082 if (!shy) | |
| 3083 { | |
| 3084 bufp->re_nsub++; | |
| 3085 while (bufp->external_to_internal_register_size <= | |
| 3086 bufp->re_nsub) | |
| 3087 { | |
| 3088 int i; | |
| 3089 int old_size = | |
| 3090 bufp->external_to_internal_register_size; | |
| 3091 bufp->external_to_internal_register_size += 5; | |
| 3092 RETALLOC (bufp->external_to_internal_register, | |
| 3093 bufp->external_to_internal_register_size, | |
| 3094 int); | |
| 3095 /* debugging */ | |
| 3096 for (i = old_size; | |
| 3097 i < bufp->external_to_internal_register_size; i++) | |
| 3098 bufp->external_to_internal_register[i] = | |
| 3099 (int) 0xDEADBEEF; | |
| 3100 } | |
| 3101 | |
| 3102 bufp->external_to_internal_register[bufp->re_nsub] = | |
| 3103 bufp->re_ngroups; | |
| 3104 } | |
| 428 | 3105 |
| 3106 if (COMPILE_STACK_FULL) | |
| 3107 { | |
| 3108 RETALLOC (compile_stack.stack, compile_stack.size << 1, | |
| 3109 compile_stack_elt_t); | |
| 3110 if (compile_stack.stack == NULL) return REG_ESPACE; | |
| 3111 | |
| 3112 compile_stack.size <<= 1; | |
| 3113 } | |
| 3114 | |
| 3115 /* These are the values to restore when we hit end of this | |
| 3116 group. They are all relative offsets, so that if the | |
| 3117 whole pattern moves because of realloc, they will still | |
| 3118 be valid. */ | |
| 3119 COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer; | |
| 3120 COMPILE_STACK_TOP.fixup_alt_jump | |
| 3121 = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0; | |
| 446 | 3122 COMPILE_STACK_TOP.laststart_offset = buf_end - bufp->buffer; |
| 428 | 3123 COMPILE_STACK_TOP.regnum = r; |
| 3124 | |
| 3125 /* We will eventually replace the 0 with the number of | |
| 3126 groups inner to this one. But do not push a | |
| 3127 start_memory for groups beyond the last one we can | |
| 502 | 3128 represent in the compiled pattern. |
| 3129 #### bad bad bad. this will fail in lots of ways, if we | |
| 3130 ever have to backtrack for these groups. | |
| 3131 */ | |
| 428 | 3132 if (r <= MAX_REGNUM) |
| 3133 { | |
| 3134 COMPILE_STACK_TOP.inner_group_offset | |
| 446 | 3135 = buf_end - bufp->buffer + 2; |
| 428 | 3136 BUF_PUSH_3 (start_memory, r, 0); |
| 3137 } | |
| 3138 | |
| 3139 compile_stack.avail++; | |
| 3140 | |
| 3141 fixup_alt_jump = 0; | |
| 3142 laststart = 0; | |
| 446 | 3143 begalt = buf_end; |
| 428 | 3144 /* If we've reached MAX_REGNUM groups, then this open |
| 3145 won't actually generate any code, so we'll have to | |
| 3146 clear pending_exact explicitly. */ | |
| 3147 pending_exact = 0; | |
| 3148 } | |
| 3149 break; | |
| 3150 | |
| 3151 | |
| 3152 case ')': | |
| 3153 if (syntax & RE_NO_BK_PARENS) goto normal_backslash; | |
| 3154 | |
| 3155 if (COMPILE_STACK_EMPTY) { | |
| 3156 if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
| 3157 goto normal_backslash; | |
| 3158 else | |
| 3159 FREE_STACK_RETURN (REG_ERPAREN); | |
| 3160 } | |
| 3161 | |
| 3162 handle_close: | |
| 3163 if (fixup_alt_jump) | |
| 3164 { /* Push a dummy failure point at the end of the | |
| 3165 alternative for a possible future | |
| 3166 `pop_failure_jump' to pop. See comments at | |
| 3167 `push_dummy_failure' in `re_match_2'. */ | |
| 3168 BUF_PUSH (push_dummy_failure); | |
| 3169 | |
| 3170 /* We allocated space for this jump when we assigned | |
| 3171 to `fixup_alt_jump', in the `handle_alt' case below. */ | |
| 446 | 3172 STORE_JUMP (jump_past_alt, fixup_alt_jump, buf_end - 1); |
| 428 | 3173 } |
| 3174 | |
| 3175 /* See similar code for backslashed left paren above. */ | |
| 3176 if (COMPILE_STACK_EMPTY) { | |
| 3177 if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
| 3178 goto normal_char; | |
| 3179 else | |
| 3180 FREE_STACK_RETURN (REG_ERPAREN); | |
| 3181 } | |
| 3182 | |
| 3183 /* Since we just checked for an empty stack above, this | |
| 3184 ``can't happen''. */ | |
| 3185 assert (compile_stack.avail != 0); | |
| 3186 { | |
| 3187 /* We don't just want to restore into `regnum', because | |
| 3188 later groups should continue to be numbered higher, | |
| 3189 as in `(ab)c(de)' -- the second group is #2. */ | |
| 3190 regnum_t this_group_regnum; | |
| 3191 | |
| 3192 compile_stack.avail--; | |
| 3193 begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset; | |
| 3194 fixup_alt_jump | |
| 3195 = COMPILE_STACK_TOP.fixup_alt_jump | |
| 3196 ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1 | |
| 3197 : 0; | |
| 3198 laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset; | |
| 3199 this_group_regnum = COMPILE_STACK_TOP.regnum; | |
| 3200 /* If we've reached MAX_REGNUM groups, then this open | |
| 3201 won't actually generate any code, so we'll have to | |
| 3202 clear pending_exact explicitly. */ | |
| 3203 pending_exact = 0; | |
| 3204 | |
| 3205 /* We're at the end of the group, so now we know how many | |
| 3206 groups were inside this one. */ | |
| 3207 if (this_group_regnum <= MAX_REGNUM) | |
| 3208 { | |
| 3209 unsigned char *inner_group_loc | |
| 3210 = bufp->buffer + COMPILE_STACK_TOP.inner_group_offset; | |
| 3211 | |
| 3212 *inner_group_loc = regnum - this_group_regnum; | |
| 3213 BUF_PUSH_3 (stop_memory, this_group_regnum, | |
| 3214 regnum - this_group_regnum); | |
| 3215 } | |
| 3216 } | |
| 3217 break; | |
| 3218 | |
| 3219 | |
| 3220 case '|': /* `\|'. */ | |
| 3221 if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) | |
| 3222 goto normal_backslash; | |
| 3223 handle_alt: | |
| 3224 if (syntax & RE_LIMITED_OPS) | |
| 3225 goto normal_char; | |
| 3226 | |
| 3227 /* Insert before the previous alternative a jump which | |
| 3228 jumps to this alternative if the former fails. */ | |
| 3229 GET_BUFFER_SPACE (3); | |
| 446 | 3230 INSERT_JUMP (on_failure_jump, begalt, buf_end + 6); |
| 428 | 3231 pending_exact = 0; |
| 446 | 3232 buf_end += 3; |
| 428 | 3233 |
| 3234 /* The alternative before this one has a jump after it | |
| 3235 which gets executed if it gets matched. Adjust that | |
| 3236 jump so it will jump to this alternative's analogous | |
| 3237 jump (put in below, which in turn will jump to the next | |
| 3238 (if any) alternative's such jump, etc.). The last such | |
| 3239 jump jumps to the correct final destination. A picture: | |
| 3240 _____ _____ | |
| 3241 | | | | | |
| 3242 | v | v | |
| 3243 a | b | c | |
| 3244 | |
| 3245 If we are at `b', then fixup_alt_jump right now points to a | |
| 3246 three-byte space after `a'. We'll put in the jump, set | |
| 3247 fixup_alt_jump to right after `b', and leave behind three | |
| 3248 bytes which we'll fill in when we get to after `c'. */ | |
| 3249 | |
| 3250 if (fixup_alt_jump) | |
| 446 | 3251 STORE_JUMP (jump_past_alt, fixup_alt_jump, buf_end); |
| 428 | 3252 |
| 3253 /* Mark and leave space for a jump after this alternative, | |
| 3254 to be filled in later either by next alternative or | |
| 3255 when know we're at the end of a series of alternatives. */ | |
| 446 | 3256 fixup_alt_jump = buf_end; |
| 428 | 3257 GET_BUFFER_SPACE (3); |
| 446 | 3258 buf_end += 3; |
| 428 | 3259 |
| 3260 laststart = 0; | |
| 446 | 3261 begalt = buf_end; |
| 428 | 3262 break; |
| 3263 | |
| 3264 | |
| 3265 case '{': | |
| 3266 /* If \{ is a literal. */ | |
| 3267 if (!(syntax & RE_INTERVALS) | |
| 3268 /* If we're at `\{' and it's not the open-interval | |
| 3269 operator. */ | |
| 3270 || ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES)) | |
| 3271 || (p - 2 == pattern && p == pend)) | |
| 3272 goto normal_backslash; | |
| 3273 | |
| 3274 handle_interval: | |
| 3275 { | |
| 3276 /* If got here, then the syntax allows intervals. */ | |
| 3277 | |
| 3278 /* At least (most) this many matches must be made. */ | |
| 3279 int lower_bound = -1, upper_bound = -1; | |
| 3280 | |
| 3281 beg_interval = p - 1; | |
| 3282 | |
| 3283 if (p == pend) | |
| 3284 { | |
| 3285 if (syntax & RE_NO_BK_BRACES) | |
| 3286 goto unfetch_interval; | |
| 3287 else | |
| 3288 FREE_STACK_RETURN (REG_EBRACE); | |
| 3289 } | |
| 3290 | |
| 3291 GET_UNSIGNED_NUMBER (lower_bound); | |
| 3292 | |
| 3293 if (c == ',') | |
| 3294 { | |
| 3295 GET_UNSIGNED_NUMBER (upper_bound); | |
| 3296 if (upper_bound < 0) upper_bound = RE_DUP_MAX; | |
| 3297 } | |
| 3298 else | |
| 3299 /* Interval such as `{1}' => match exactly once. */ | |
| 3300 upper_bound = lower_bound; | |
| 3301 | |
| 3302 if (lower_bound < 0 || upper_bound > RE_DUP_MAX | |
| 3303 || lower_bound > upper_bound) | |
| 3304 { | |
| 3305 if (syntax & RE_NO_BK_BRACES) | |
| 3306 goto unfetch_interval; | |
| 3307 else | |
| 3308 FREE_STACK_RETURN (REG_BADBR); | |
| 3309 } | |
| 3310 | |
| 3311 if (!(syntax & RE_NO_BK_BRACES)) | |
| 3312 { | |
| 3313 if (c != '\\') FREE_STACK_RETURN (REG_EBRACE); | |
| 3314 | |
| 3315 PATFETCH (c); | |
| 3316 } | |
| 3317 | |
| 3318 if (c != '}') | |
| 3319 { | |
| 3320 if (syntax & RE_NO_BK_BRACES) | |
| 3321 goto unfetch_interval; | |
| 3322 else | |
| 3323 FREE_STACK_RETURN (REG_BADBR); | |
| 3324 } | |
| 3325 | |
| 3326 /* We just parsed a valid interval. */ | |
| 3327 | |
| 3328 /* If it's invalid to have no preceding re. */ | |
| 3329 if (!laststart) | |
| 3330 { | |
| 3331 if (syntax & RE_CONTEXT_INVALID_OPS) | |
| 3332 FREE_STACK_RETURN (REG_BADRPT); | |
| 3333 else if (syntax & RE_CONTEXT_INDEP_OPS) | |
| 446 | 3334 laststart = buf_end; |
| 428 | 3335 else |
| 3336 goto unfetch_interval; | |
| 3337 } | |
| 3338 | |
| 3339 /* If the upper bound is zero, don't want to succeed at | |
| 3340 all; jump from `laststart' to `b + 3', which will be | |
| 3341 the end of the buffer after we insert the jump. */ | |
| 3342 if (upper_bound == 0) | |
| 3343 { | |
| 3344 GET_BUFFER_SPACE (3); | |
| 446 | 3345 INSERT_JUMP (jump, laststart, buf_end + 3); |
| 3346 buf_end += 3; | |
| 428 | 3347 } |
| 3348 | |
| 3349 /* Otherwise, we have a nontrivial interval. When | |
| 3350 we're all done, the pattern will look like: | |
| 3351 set_number_at <jump count> <upper bound> | |
| 3352 set_number_at <succeed_n count> <lower bound> | |
| 3353 succeed_n <after jump addr> <succeed_n count> | |
| 3354 <body of loop> | |
| 3355 jump_n <succeed_n addr> <jump count> | |
| 3356 (The upper bound and `jump_n' are omitted if | |
| 3357 `upper_bound' is 1, though.) */ | |
| 3358 else | |
| 3359 { /* If the upper bound is > 1, we need to insert | |
| 3360 more at the end of the loop. */ | |
| 647 | 3361 int nbytes = 10 + (upper_bound > 1) * 10; |
| 428 | 3362 |
| 3363 GET_BUFFER_SPACE (nbytes); | |
| 3364 | |
| 3365 /* Initialize lower bound of the `succeed_n', even | |
| 3366 though it will be set during matching by its | |
| 3367 attendant `set_number_at' (inserted next), | |
| 3368 because `re_compile_fastmap' needs to know. | |
| 3369 Jump to the `jump_n' we might insert below. */ | |
| 3370 INSERT_JUMP2 (succeed_n, laststart, | |
| 446 | 3371 buf_end + 5 + (upper_bound > 1) * 5, |
| 428 | 3372 lower_bound); |
| 446 | 3373 buf_end += 5; |
| 428 | 3374 |
| 3375 /* Code to initialize the lower bound. Insert | |
| 3376 before the `succeed_n'. The `5' is the last two | |
| 3377 bytes of this `set_number_at', plus 3 bytes of | |
| 3378 the following `succeed_n'. */ | |
| 446 | 3379 insert_op2 (set_number_at, laststart, 5, lower_bound, buf_end); |
| 3380 buf_end += 5; | |
| 428 | 3381 |
| 3382 if (upper_bound > 1) | |
| 3383 { /* More than one repetition is allowed, so | |
| 3384 append a backward jump to the `succeed_n' | |
| 3385 that starts this interval. | |
| 3386 | |
| 3387 When we've reached this during matching, | |
| 3388 we'll have matched the interval once, so | |
| 3389 jump back only `upper_bound - 1' times. */ | |
| 446 | 3390 STORE_JUMP2 (jump_n, buf_end, laststart + 5, |
| 428 | 3391 upper_bound - 1); |
| 446 | 3392 buf_end += 5; |
| 428 | 3393 |
| 3394 /* The location we want to set is the second | |
| 3395 parameter of the `jump_n'; that is `b-2' as | |
| 3396 an absolute address. `laststart' will be | |
| 3397 the `set_number_at' we're about to insert; | |
| 3398 `laststart+3' the number to set, the source | |
| 3399 for the relative address. But we are | |
| 3400 inserting into the middle of the pattern -- | |
| 3401 so everything is getting moved up by 5. | |
| 3402 Conclusion: (b - 2) - (laststart + 3) + 5, | |
| 3403 i.e., b - laststart. | |
| 3404 | |
| 3405 We insert this at the beginning of the loop | |
| 3406 so that if we fail during matching, we'll | |
| 3407 reinitialize the bounds. */ | |
| 446 | 3408 insert_op2 (set_number_at, laststart, |
| 3409 buf_end - laststart, | |
| 3410 upper_bound - 1, buf_end); | |
| 3411 buf_end += 5; | |
| 428 | 3412 } |
| 3413 } | |
| 3414 pending_exact = 0; | |
| 3415 beg_interval = NULL; | |
| 3416 } | |
| 3417 break; | |
| 3418 | |
| 3419 unfetch_interval: | |
| 3420 /* If an invalid interval, match the characters as literals. */ | |
| 3421 assert (beg_interval); | |
| 3422 p = beg_interval; | |
| 3423 beg_interval = NULL; | |
| 3424 | |
| 3425 /* normal_char and normal_backslash need `c'. */ | |
| 3426 PATFETCH (c); | |
| 3427 | |
| 3428 if (!(syntax & RE_NO_BK_BRACES)) | |
| 3429 { | |
| 3430 if (p > pattern && p[-1] == '\\') | |
| 3431 goto normal_backslash; | |
| 3432 } | |
| 3433 goto normal_char; | |
| 3434 | |
| 3435 #ifdef emacs | |
| 3436 /* There is no way to specify the before_dot and after_dot | |
| 3437 operators. rms says this is ok. --karl */ | |
| 3438 case '=': | |
| 3439 BUF_PUSH (at_dot); | |
| 3440 break; | |
| 3441 | |
| 3442 case 's': | |
| 446 | 3443 laststart = buf_end; |
| 428 | 3444 PATFETCH (c); |
| 3445 /* XEmacs addition */ | |
| 3446 if (c >= 0x80 || syntax_spec_code[c] == 0377) | |
| 3447 FREE_STACK_RETURN (REG_ESYNTAX); | |
| 3448 BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); | |
| 3449 break; | |
| 3450 | |
| 3451 case 'S': | |
| 446 | 3452 laststart = buf_end; |
| 428 | 3453 PATFETCH (c); |
| 3454 /* XEmacs addition */ | |
| 3455 if (c >= 0x80 || syntax_spec_code[c] == 0377) | |
| 3456 FREE_STACK_RETURN (REG_ESYNTAX); | |
| 3457 BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); | |
| 3458 break; | |
| 3459 | |
| 3460 #ifdef MULE | |
| 3461 /* 97.2.17 jhod merged in to XEmacs from mule-2.3 */ | |
| 3462 case 'c': | |
| 446 | 3463 laststart = buf_end; |
| 428 | 3464 PATFETCH_RAW (c); |
| 3465 if (c < 32 || c > 127) | |
| 3466 FREE_STACK_RETURN (REG_ECATEGORY); | |
| 3467 BUF_PUSH_2 (categoryspec, c); | |
| 3468 break; | |
| 3469 | |
| 3470 case 'C': | |
| 446 | 3471 laststart = buf_end; |
| 428 | 3472 PATFETCH_RAW (c); |
| 3473 if (c < 32 || c > 127) | |
| 3474 FREE_STACK_RETURN (REG_ECATEGORY); | |
| 3475 BUF_PUSH_2 (notcategoryspec, c); | |
| 3476 break; | |
| 3477 /* end of category patch */ | |
| 3478 #endif /* MULE */ | |
| 3479 #endif /* emacs */ | |
| 3480 | |
| 3481 | |
| 3482 case 'w': | |
| 446 | 3483 laststart = buf_end; |
| 428 | 3484 BUF_PUSH (wordchar); |
| 3485 break; | |
| 3486 | |
| 3487 | |
| 3488 case 'W': | |
| 446 | 3489 laststart = buf_end; |
| 428 | 3490 BUF_PUSH (notwordchar); |
| 3491 break; | |
| 3492 | |
| 3493 | |
| 3494 case '<': | |
| 3495 BUF_PUSH (wordbeg); | |
| 3496 break; | |
| 3497 | |
| 3498 case '>': | |
| 3499 BUF_PUSH (wordend); | |
| 3500 break; | |
| 3501 | |
| 3502 case 'b': | |
| 3503 BUF_PUSH (wordbound); | |
| 3504 break; | |
| 3505 | |
| 3506 case 'B': | |
| 3507 BUF_PUSH (notwordbound); | |
| 3508 break; | |
| 3509 | |
| 3510 case '`': | |
| 3511 BUF_PUSH (begbuf); | |
| 3512 break; | |
| 3513 | |
| 3514 case '\'': | |
| 3515 BUF_PUSH (endbuf); | |
| 3516 break; | |
| 3517 | |
| 3518 case '1': case '2': case '3': case '4': case '5': | |
| 3519 case '6': case '7': case '8': case '9': | |
| 446 | 3520 { |
| 502 | 3521 regnum_t reg, regint; |
| 3522 int may_need_to_unfetch = 0; | |
| 446 | 3523 if (syntax & RE_NO_BK_REFS) |
| 3524 goto normal_char; | |
| 3525 | |
| 502 | 3526 /* This only goes up to 99. It could be extended to work |
| 3527 up to 255 (the maximum number of registers that can be | |
| 3528 handled by the current regexp engine, because it stores | |
| 3529 its register numbers in the compiled pattern as one byte, | |
| 3530 ugh). Doing that's a bit trickier, because you might | |
| 3531 have the case where \25 a back-ref but \255 is not, ... */ | |
| 446 | 3532 reg = c - '0'; |
| 502 | 3533 if (p < pend) |
| 3534 { | |
| 3535 PATFETCH (c); | |
| 3536 if (c >= '0' && c <= '9') | |
| 3537 { | |
| 3538 regnum_t new_reg = reg * 10 + c - '0'; | |
| 3539 if (new_reg <= bufp->re_nsub) | |
| 3540 { | |
| 3541 reg = new_reg; | |
| 3542 may_need_to_unfetch = 1; | |
| 3543 } | |
| 3544 else | |
| 3545 PATUNFETCH; | |
| 3546 } | |
| 523 | 3547 else |
| 3548 PATUNFETCH; | |
| 502 | 3549 } |
| 3550 | |
| 3551 if (reg > bufp->re_nsub) | |
| 446 | 3552 FREE_STACK_RETURN (REG_ESUBREG); |
| 3553 | |
| 502 | 3554 regint = bufp->external_to_internal_register[reg]; |
| 446 | 3555 /* Can't back reference to a subexpression if inside of it. */ |
| 502 | 3556 if (group_in_compile_stack (compile_stack, regint)) |
| 3557 { | |
| 3558 if (may_need_to_unfetch) | |
| 3559 PATUNFETCH; | |
| 3560 goto normal_char; | |
| 3561 } | |
| 3562 | |
| 3563 #ifdef emacs | |
| 3564 if (reg > 9 && | |
| 3565 bufp->warned_about_incompatible_back_references == 0) | |
| 3566 { | |
| 3567 bufp->warned_about_incompatible_back_references = 1; | |
| 3568 warn_when_safe (intern ("regex"), Qinfo, | |
| 3569 "Back reference \\%d now has new " | |
| 3570 "semantics in %s", reg, pattern); | |
| 3571 } | |
| 3572 #endif | |
| 446 | 3573 |
| 3574 laststart = buf_end; | |
| 502 | 3575 BUF_PUSH_2 (duplicate, regint); |
| 446 | 3576 } |
| 428 | 3577 break; |
| 3578 | |
| 3579 | |
| 3580 case '+': | |
| 3581 case '?': | |
| 3582 if (syntax & RE_BK_PLUS_QM) | |
| 3583 goto handle_plus; | |
| 3584 else | |
| 3585 goto normal_backslash; | |
| 3586 | |
| 3587 default: | |
| 3588 normal_backslash: | |
| 3589 /* You might think it would be useful for \ to mean | |
| 3590 not to translate; but if we don't translate it, | |
| 3591 it will never match anything. */ | |
| 826 | 3592 c = RE_TRANSLATE (c); |
| 428 | 3593 goto normal_char; |
| 3594 } | |
| 3595 break; | |
| 3596 | |
| 3597 | |
| 3598 default: | |
| 3599 /* Expects the character in `c'. */ | |
| 3600 /* `p' points to the location after where `c' came from. */ | |
| 3601 normal_char: | |
| 3602 { | |
|
4750
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3603 /* The following conditional synced to GNU Emacs 22.1. */ |
| 428 | 3604 /* If no exactn currently being built. */ |
| 3605 if (!pending_exact | |
| 3606 | |
| 3607 /* If last exactn not at current position. */ | |
| 446 | 3608 || pending_exact + *pending_exact + 1 != buf_end |
| 428 | 3609 |
| 3610 /* We have only one byte following the exactn for the count. */ | |
|
4750
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3611 || *pending_exact >= (1 << BYTEWIDTH) - MAX_ICHAR_LEN |
|
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3612 |
|
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3613 /* If followed by a repetition operator. |
|
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3614 If the lookahead fails because of end of pattern, any |
|
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3615 trailing backslash will get caught later. */ |
|
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3616 || (p != pend && (*p == '*' || *p == '^')) |
| 428 | 3617 || ((syntax & RE_BK_PLUS_QM) |
|
4750
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3618 ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?') |
|
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3619 : p != pend && (*p == '+' || *p == '?')) |
| 428 | 3620 || ((syntax & RE_INTERVALS) |
| 3621 && ((syntax & RE_NO_BK_BRACES) | |
|
4750
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3622 ? p != pend && *p == '{' |
|
b5f21bb36684
Fix crash in regex.c (closes issue630).
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4527
diff
changeset
|
3623 : p + 1 < pend && (p[0] == '\\' && p[1] == '{')))) |
| 428 | 3624 { |
| 3625 /* Start building a new exactn. */ | |
| 3626 | |
| 446 | 3627 laststart = buf_end; |
| 428 | 3628 |
| 3629 BUF_PUSH_2 (exactn, 0); | |
| 446 | 3630 pending_exact = buf_end - 1; |
| 428 | 3631 } |
| 3632 | |
| 446 | 3633 #ifndef MULE |
| 428 | 3634 BUF_PUSH (c); |
| 3635 (*pending_exact)++; | |
| 446 | 3636 #else |
| 3637 { | |
| 3638 Bytecount bt_count; | |
| 867 | 3639 Ibyte tmp_buf[MAX_ICHAR_LEN]; |
| 446 | 3640 int i; |
| 3641 | |
| 867 | 3642 bt_count = set_itext_ichar (tmp_buf, c); |
| 446 | 3643 |
| 3644 for (i = 0; i < bt_count; i++) | |
| 3645 { | |
| 3646 BUF_PUSH (tmp_buf[i]); | |
| 3647 (*pending_exact)++; | |
| 3648 } | |
| 3649 } | |
| 3650 #endif | |
| 428 | 3651 break; |
| 3652 } | |
| 3653 } /* switch (c) */ | |
| 3654 } /* while p != pend */ | |
| 3655 | |
| 3656 | |
| 3657 /* Through the pattern now. */ | |
| 3658 | |
| 3659 if (fixup_alt_jump) | |
| 446 | 3660 STORE_JUMP (jump_past_alt, fixup_alt_jump, buf_end); |
| 428 | 3661 |
| 3662 if (!COMPILE_STACK_EMPTY) | |
| 3663 FREE_STACK_RETURN (REG_EPAREN); | |
| 3664 | |
| 3665 /* If we don't want backtracking, force success | |
| 3666 the first time we reach the end of the compiled pattern. */ | |
| 3667 if (syntax & RE_NO_POSIX_BACKTRACKING) | |
| 3668 BUF_PUSH (succeed); | |
| 3669 | |
|
4976
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
3670 xfree (compile_stack.stack); |
| 428 | 3671 |
| 3672 /* We have succeeded; set the length of the buffer. */ | |
| 446 | 3673 bufp->used = buf_end - bufp->buffer; |
| 428 | 3674 |
| 3675 #ifdef DEBUG | |
| 5041 | 3676 if (debug_regexps & RE_DEBUG_COMPILATION) |
| 428 | 3677 { |
| 3678 DEBUG_PRINT1 ("\nCompiled pattern: \n"); | |
| 3679 print_compiled_pattern (bufp); | |
| 3680 } | |
| 3681 #endif /* DEBUG */ | |
| 3682 | |
| 3683 #ifndef MATCH_MAY_ALLOCATE | |
| 3684 /* Initialize the failure stack to the largest possible stack. This | |
| 3685 isn't necessary unless we're trying to avoid calling alloca in | |
| 3686 the search and match routines. */ | |
| 3687 { | |
| 502 | 3688 int num_regs = bufp->re_ngroups + 1; |
| 428 | 3689 |
| 3690 /* Since DOUBLE_FAIL_STACK refuses to double only if the current size | |
| 3691 is strictly greater than re_max_failures, the largest possible stack | |
| 3692 is 2 * re_max_failures failure points. */ | |
| 3693 if (fail_stack.size < (2 * re_max_failures * MAX_FAILURE_ITEMS)) | |
| 3694 { | |
| 3695 fail_stack.size = (2 * re_max_failures * MAX_FAILURE_ITEMS); | |
| 3696 | |
| 3697 if (! fail_stack.stack) | |
| 3698 fail_stack.stack | |
| 3699 = (fail_stack_elt_t *) xmalloc (fail_stack.size | |
| 3700 * sizeof (fail_stack_elt_t)); | |
| 3701 else | |
| 3702 fail_stack.stack | |
| 3703 = (fail_stack_elt_t *) xrealloc (fail_stack.stack, | |
| 3704 (fail_stack.size | |
| 3705 * sizeof (fail_stack_elt_t))); | |
| 3706 } | |
| 3707 | |
| 3708 regex_grow_registers (num_regs); | |
| 3709 } | |
| 3710 #endif /* not MATCH_MAY_ALLOCATE */ | |
| 3711 | |
| 3712 return REG_NOERROR; | |
| 3713 } /* regex_compile */ | |
| 3714 | |
| 3715 /* Subroutines for `regex_compile'. */ | |
| 3716 | |
| 3717 /* Store OP at LOC followed by two-byte integer parameter ARG. */ | |
| 3718 | |
| 3719 static void | |
| 3720 store_op1 (re_opcode_t op, unsigned char *loc, int arg) | |
| 3721 { | |
| 3722 *loc = (unsigned char) op; | |
| 3723 STORE_NUMBER (loc + 1, arg); | |
| 3724 } | |
| 3725 | |
| 3726 | |
| 3727 /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
| 3728 | |
| 3729 static void | |
| 3730 store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2) | |
| 3731 { | |
| 3732 *loc = (unsigned char) op; | |
| 3733 STORE_NUMBER (loc + 1, arg1); | |
| 3734 STORE_NUMBER (loc + 3, arg2); | |
| 3735 } | |
| 3736 | |
| 3737 | |
| 3738 /* Copy the bytes from LOC to END to open up three bytes of space at LOC | |
| 3739 for OP followed by two-byte integer parameter ARG. */ | |
| 3740 | |
| 3741 static void | |
| 3742 insert_op1 (re_opcode_t op, unsigned char *loc, int arg, unsigned char *end) | |
| 3743 { | |
| 3744 REGISTER unsigned char *pfrom = end; | |
| 3745 REGISTER unsigned char *pto = end + 3; | |
| 3746 | |
| 3747 while (pfrom != loc) | |
| 3748 *--pto = *--pfrom; | |
| 3749 | |
| 3750 store_op1 (op, loc, arg); | |
| 3751 } | |
| 3752 | |
| 3753 | |
| 3754 /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
| 3755 | |
| 3756 static void | |
| 3757 insert_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2, | |
| 3758 unsigned char *end) | |
| 3759 { | |
| 3760 REGISTER unsigned char *pfrom = end; | |
| 3761 REGISTER unsigned char *pto = end + 5; | |
| 3762 | |
| 3763 while (pfrom != loc) | |
| 3764 *--pto = *--pfrom; | |
| 3765 | |
| 3766 store_op2 (op, loc, arg1, arg2); | |
| 3767 } | |
| 3768 | |
| 3769 | |
| 3770 /* P points to just after a ^ in PATTERN. Return true if that ^ comes | |
| 3771 after an alternative or a begin-subexpression. We assume there is at | |
| 3772 least one character before the ^. */ | |
| 3773 | |
| 460 | 3774 static re_bool |
| 446 | 3775 at_begline_loc_p (re_char *pattern, re_char *p, reg_syntax_t syntax) |
| 428 | 3776 { |
| 446 | 3777 re_char *prev = p - 2; |
| 460 | 3778 re_bool prev_prev_backslash = prev > pattern && prev[-1] == '\\'; |
| 428 | 3779 |
| 3780 return | |
| 3781 /* After a subexpression? */ | |
| 3782 (*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash)) | |
| 3783 /* After an alternative? */ | |
| 3784 || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash)); | |
| 3785 } | |
| 3786 | |
| 3787 | |
| 3788 /* The dual of at_begline_loc_p. This one is for $. We assume there is | |
| 3789 at least one character after the $, i.e., `P < PEND'. */ | |
| 3790 | |
| 460 | 3791 static re_bool |
| 446 | 3792 at_endline_loc_p (re_char *p, re_char *pend, int syntax) |
| 428 | 3793 { |
| 446 | 3794 re_char *next = p; |
| 460 | 3795 re_bool next_backslash = *next == '\\'; |
| 446 | 3796 re_char *next_next = p + 1 < pend ? p + 1 : 0; |
| 428 | 3797 |
| 3798 return | |
| 3799 /* Before a subexpression? */ | |
| 3800 (syntax & RE_NO_BK_PARENS ? *next == ')' | |
| 3801 : next_backslash && next_next && *next_next == ')') | |
| 3802 /* Before an alternative? */ | |
| 3803 || (syntax & RE_NO_BK_VBAR ? *next == '|' | |
| 3804 : next_backslash && next_next && *next_next == '|'); | |
| 3805 } | |
| 3806 | |
| 3807 | |
| 3808 /* Returns true if REGNUM is in one of COMPILE_STACK's elements and | |
| 3809 false if it's not. */ | |
| 3810 | |
| 460 | 3811 static re_bool |
| 428 | 3812 group_in_compile_stack (compile_stack_type compile_stack, regnum_t regnum) |
| 3813 { | |
| 3814 int this_element; | |
| 3815 | |
| 3816 for (this_element = compile_stack.avail - 1; | |
| 3817 this_element >= 0; | |
| 3818 this_element--) | |
| 3819 if (compile_stack.stack[this_element].regnum == regnum) | |
| 3820 return true; | |
| 3821 | |
| 3822 return false; | |
| 3823 } | |
| 3824 | |
| 3825 | |
| 3826 /* Read the ending character of a range (in a bracket expression) from the | |
| 3827 uncompiled pattern *P_PTR (which ends at PEND). We assume the | |
| 3828 starting character is in `P[-2]'. (`P[-1]' is the character `-'.) | |
| 3829 Then we set the translation of all bits between the starting and | |
| 3830 ending characters (inclusive) in the compiled pattern B. | |
| 3831 | |
| 3832 Return an error code. | |
| 3833 | |
| 3834 We use these short variable names so we can use the same macros as | |
| 826 | 3835 `regex_compile' itself. |
| 3836 | |
| 3837 Under Mule, this is only called when both chars of the range are | |
| 3838 ASCII. */ | |
| 428 | 3839 |
| 3840 static reg_errcode_t | |
| 446 | 3841 compile_range (re_char **p_ptr, re_char *pend, RE_TRANSLATE_TYPE translate, |
| 3842 reg_syntax_t syntax, unsigned char *buf_end) | |
| 428 | 3843 { |
| 867 | 3844 Ichar this_char; |
| 428 | 3845 |
| 446 | 3846 re_char *p = *p_ptr; |
| 428 | 3847 int range_start, range_end; |
| 3848 | |
| 3849 if (p == pend) | |
| 3850 return REG_ERANGE; | |
| 3851 | |
| 3852 /* Even though the pattern is a signed `char *', we need to fetch | |
| 3853 with unsigned char *'s; if the high bit of the pattern character | |
| 3854 is set, the range endpoints will be negative if we fetch using a | |
| 3855 signed char *. | |
| 3856 | |
| 3857 We also want to fetch the endpoints without translating them; the | |
| 3858 appropriate translation is done in the bit-setting loop below. */ | |
| 442 | 3859 /* The SVR4 compiler on the 3B2 had trouble with unsigned const char *. */ |
| 3860 range_start = ((const unsigned char *) p)[-2]; | |
| 3861 range_end = ((const unsigned char *) p)[0]; | |
| 428 | 3862 |
| 3863 /* Have to increment the pointer into the pattern string, so the | |
| 3864 caller isn't still at the ending character. */ | |
| 3865 (*p_ptr)++; | |
| 3866 | |
| 3867 /* If the start is after the end, the range is empty. */ | |
| 3868 if (range_start > range_end) | |
| 3869 return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR; | |
| 3870 | |
| 3871 /* Here we see why `this_char' has to be larger than an `unsigned | |
| 3872 char' -- the range is inclusive, so if `range_end' == 0xff | |
| 3873 (assuming 8-bit characters), we would otherwise go into an infinite | |
| 3874 loop, since all characters <= 0xff. */ | |
| 3875 for (this_char = range_start; this_char <= range_end; this_char++) | |
| 3876 { | |
| 826 | 3877 SET_LIST_BIT (RE_TRANSLATE (this_char)); |
| 428 | 3878 } |
| 3879 | |
| 3880 return REG_NOERROR; | |
| 3881 } | |
| 3882 | |
| 3883 #ifdef MULE | |
| 3884 | |
| 3885 static reg_errcode_t | |
| 446 | 3886 compile_extended_range (re_char **p_ptr, re_char *pend, |
| 3887 RE_TRANSLATE_TYPE translate, | |
| 428 | 3888 reg_syntax_t syntax, Lisp_Object rtab) |
| 3889 { | |
| 867 | 3890 Ichar this_char, range_start, range_end; |
| 3891 const Ibyte *p; | |
| 428 | 3892 |
| 3893 if (*p_ptr == pend) | |
| 3894 return REG_ERANGE; | |
| 3895 | |
| 867 | 3896 p = (const Ibyte *) *p_ptr; |
| 3897 range_end = itext_ichar (p); | |
| 428 | 3898 p--; /* back to '-' */ |
| 867 | 3899 DEC_IBYTEPTR (p); /* back to start of range */ |
| 428 | 3900 /* We also want to fetch the endpoints without translating them; the |
| 3901 appropriate translation is done in the bit-setting loop below. */ | |
| 867 | 3902 range_start = itext_ichar (p); |
| 3903 INC_IBYTEPTR (*p_ptr); | |
| 428 | 3904 |
| 3905 /* If the start is after the end, the range is empty. */ | |
| 3906 if (range_start > range_end) | |
| 3907 return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR; | |
| 3908 | |
| 3909 /* Can't have ranges spanning different charsets, except maybe for | |
| 3910 ranges entirely within the first 256 chars. */ | |
| 3911 | |
| 3912 if ((range_start >= 0x100 || range_end >= 0x100) | |
| 867 | 3913 && ichar_leading_byte (range_start) != |
| 3914 ichar_leading_byte (range_end)) | |
| 428 | 3915 return REG_ERANGESPAN; |
| 3916 | |
| 826 | 3917 /* #### This might be way inefficient if the range encompasses 10,000 |
| 3918 chars or something. To be efficient, you'd have to do something like | |
| 3919 this: | |
| 428 | 3920 |
| 3921 range_table a; | |
| 3922 range_table b; | |
| 3923 map over translation table in [range_start, range_end] of | |
| 3924 (put the mapped range in a; | |
| 3925 put the translation in b) | |
| 3926 invert the range in a and truncate to [range_start, range_end] | |
| 3927 compute the union of a, b | |
| 3928 union the result into rtab | |
| 3929 */ | |
| 826 | 3930 for (this_char = range_start; this_char <= range_end; this_char++) |
| 428 | 3931 { |
| 826 | 3932 SET_RANGETAB_BIT (RE_TRANSLATE (this_char)); |
| 428 | 3933 } |
| 3934 | |
| 3935 if (this_char <= range_end) | |
| 3936 put_range_table (rtab, this_char, range_end, Qt); | |
| 3937 | |
| 3938 return REG_NOERROR; | |
| 3939 } | |
| 3940 | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3941 static reg_errcode_t |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3942 compile_char_class (re_wctype_t cc, Lisp_Object rtab, Bitbyte *flags_out) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3943 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3944 *flags_out |= re_wctype_to_bit (cc); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3945 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3946 switch (cc) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3947 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3948 case RECC_ASCII: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3949 put_range_table (rtab, 0, 0x7f, Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3950 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3951 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3952 case RECC_XDIGIT: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3953 put_range_table (rtab, 'a', 'f', Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3954 put_range_table (rtab, 'A', 'f', Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3955 /* fallthrough */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3956 case RECC_DIGIT: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3957 put_range_table (rtab, '0', '9', Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3958 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3959 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3960 case RECC_BLANK: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3961 put_range_table (rtab, ' ', ' ', Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3962 put_range_table (rtab, '\t', '\t', Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3963 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3964 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3965 case RECC_PRINT: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3966 put_range_table (rtab, ' ', 0x7e, Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3967 put_range_table (rtab, 0x80, MOST_POSITIVE_FIXNUM, Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3968 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3969 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3970 case RECC_GRAPH: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3971 put_range_table (rtab, '!', 0x7e, Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3972 put_range_table (rtab, 0x80, MOST_POSITIVE_FIXNUM, Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3973 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3974 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3975 case RECC_NONASCII: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3976 case RECC_MULTIBYTE: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3977 put_range_table (rtab, 0x80, MOST_POSITIVE_FIXNUM, Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3978 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3979 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3980 case RECC_CNTRL: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3981 put_range_table (rtab, 0x00, 0x1f, Qt); |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3982 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3983 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3984 case RECC_UNIBYTE: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3985 /* Never true in XEmacs. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3986 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3987 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3988 /* The following all have their own bits in the class_bits argument to |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3989 charset_mule and charset_mule_not, they don't use the range table |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3990 information. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3991 case RECC_ALPHA: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3992 case RECC_WORD: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3993 case RECC_ALNUM: /* Equivalent to RECC_WORD */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3994 case RECC_LOWER: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3995 case RECC_PUNCT: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3996 case RECC_SPACE: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3997 case RECC_UPPER: |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3998 break; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
3999 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4000 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4001 return REG_NOERROR; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4002 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4003 |
| 428 | 4004 #endif /* MULE */ |
| 4005 | |
| 4006 /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in | |
| 4007 BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible | |
| 4008 characters can start a string that matches the pattern. This fastmap | |
| 4009 is used by re_search to skip quickly over impossible starting points. | |
| 4010 | |
| 4011 The caller must supply the address of a (1 << BYTEWIDTH)-byte data | |
| 4012 area as BUFP->fastmap. | |
| 4013 | |
| 4014 We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in | |
| 4015 the pattern buffer. | |
| 4016 | |
| 4017 Returns 0 if we succeed, -2 if an internal error. */ | |
| 4018 | |
| 4019 int | |
| 826 | 4020 re_compile_fastmap (struct re_pattern_buffer *bufp |
| 4021 RE_LISP_SHORT_CONTEXT_ARGS_DECL) | |
| 428 | 4022 { |
| 4023 int j, k; | |
| 4024 #ifdef MATCH_MAY_ALLOCATE | |
| 4025 fail_stack_type fail_stack; | |
| 4026 #endif | |
| 456 | 4027 DECLARE_DESTINATION; |
| 428 | 4028 /* We don't push any register information onto the failure stack. */ |
| 4029 | |
| 826 | 4030 /* &&#### this should be changed for 8-bit-fixed, for efficiency. see |
| 4031 comment marked with &&#### in re_search_2. */ | |
| 4032 | |
| 428 | 4033 REGISTER char *fastmap = bufp->fastmap; |
| 4034 unsigned char *pattern = bufp->buffer; | |
| 647 | 4035 long size = bufp->used; |
| 428 | 4036 unsigned char *p = pattern; |
| 4037 REGISTER unsigned char *pend = pattern + size; | |
| 4038 | |
| 771 | 4039 #ifdef REGEX_REL_ALLOC |
| 428 | 4040 /* This holds the pointer to the failure stack, when |
| 4041 it is allocated relocatably. */ | |
| 4042 fail_stack_elt_t *failure_stack_ptr; | |
| 4043 #endif | |
| 4044 | |
| 4045 /* Assume that each path through the pattern can be null until | |
| 4046 proven otherwise. We set this false at the bottom of switch | |
| 4047 statement, to which we get only if a particular path doesn't | |
| 4048 match the empty string. */ | |
| 460 | 4049 re_bool path_can_be_null = true; |
| 428 | 4050 |
| 4051 /* We aren't doing a `succeed_n' to begin with. */ | |
| 460 | 4052 re_bool succeed_n_p = false; |
| 428 | 4053 |
| 1333 | 4054 #ifdef ERROR_CHECK_MALLOC |
| 4055 /* The pattern comes from string data, not buffer data. We don't access | |
| 4056 any buffer data, so we don't have to worry about malloc() (but the | |
| 4057 disallowed flag may have been set by a caller). */ | |
| 4058 int depth = bind_regex_malloc_disallowed (0); | |
| 4059 #endif | |
| 4060 | |
| 428 | 4061 assert (fastmap != NULL && p != NULL); |
| 4062 | |
| 4063 INIT_FAIL_STACK (); | |
| 4064 memset (fastmap, 0, 1 << BYTEWIDTH); /* Assume nothing's valid. */ | |
| 4065 bufp->fastmap_accurate = 1; /* It will be when we're done. */ | |
| 4066 bufp->can_be_null = 0; | |
| 4067 | |
| 4068 while (1) | |
| 4069 { | |
| 4070 if (p == pend || *p == succeed) | |
| 4071 { | |
| 4072 /* We have reached the (effective) end of pattern. */ | |
| 4073 if (!FAIL_STACK_EMPTY ()) | |
| 4074 { | |
| 4075 bufp->can_be_null |= path_can_be_null; | |
| 4076 | |
| 4077 /* Reset for next path. */ | |
| 4078 path_can_be_null = true; | |
| 4079 | |
| 446 | 4080 p = (unsigned char *) fail_stack.stack[--fail_stack.avail].pointer; |
| 428 | 4081 |
| 4082 continue; | |
| 4083 } | |
| 4084 else | |
| 4085 break; | |
| 4086 } | |
| 4087 | |
| 4088 /* We should never be about to go beyond the end of the pattern. */ | |
| 4089 assert (p < pend); | |
| 4090 | |
|
4759
aa5ed11f473b
Remove support for obsolete systems. See xemacs-patches message with ID
Jerry James <james@xemacs.org>
parents:
4750
diff
changeset
|
4091 switch ((re_opcode_t) *p++) |
| 428 | 4092 { |
| 4093 | |
| 4094 /* I guess the idea here is to simply not bother with a fastmap | |
| 4095 if a backreference is used, since it's too hard to figure out | |
| 4096 the fastmap for the corresponding group. Setting | |
| 4097 `can_be_null' stops `re_search_2' from using the fastmap, so | |
| 4098 that is all we do. */ | |
| 4099 case duplicate: | |
| 4100 bufp->can_be_null = 1; | |
| 4101 goto done; | |
| 4102 | |
| 4103 | |
| 4104 /* Following are the cases which match a character. These end | |
| 4105 with `break'. */ | |
| 4106 | |
| 4107 case exactn: | |
| 4108 fastmap[p[1]] = 1; | |
| 4109 break; | |
| 4110 | |
| 4111 | |
| 4112 case charset: | |
| 4113 /* XEmacs: Under Mule, these bit vectors will | |
| 4114 only contain values for characters below 0x80. */ | |
| 4115 for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--) | |
| 4116 if (p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) | |
| 4117 fastmap[j] = 1; | |
| 4118 break; | |
| 4119 | |
| 4120 | |
| 4121 case charset_not: | |
| 4122 /* Chars beyond end of map must be allowed. */ | |
| 4123 #ifdef MULE | |
| 4124 for (j = *p * BYTEWIDTH; j < 0x80; j++) | |
| 4125 fastmap[j] = 1; | |
| 4126 /* And all extended characters must be allowed, too. */ | |
| 4127 for (j = 0x80; j < 0xA0; j++) | |
| 4128 fastmap[j] = 1; | |
| 446 | 4129 #else /* not MULE */ |
| 428 | 4130 for (j = *p * BYTEWIDTH; j < (1 << BYTEWIDTH); j++) |
| 4131 fastmap[j] = 1; | |
| 446 | 4132 #endif /* MULE */ |
| 428 | 4133 |
| 4134 for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--) | |
| 4135 if (!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH)))) | |
| 4136 fastmap[j] = 1; | |
| 4137 break; | |
| 4138 | |
| 4139 #ifdef MULE | |
| 4140 case charset_mule: | |
| 4141 { | |
| 4142 int nentries; | |
| 4143 int i; | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4144 Bitbyte flags = *p++; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4145 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4146 if (flags) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4147 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4148 /* We need to consult the syntax table, fastmap won't |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4149 work. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4150 bufp->can_be_null = 1; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4151 goto done; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4152 } |
| 428 | 4153 |
| 4154 nentries = unified_range_table_nentries (p); | |
| 4155 for (i = 0; i < nentries; i++) | |
| 4156 { | |
| 4157 EMACS_INT first, last; | |
| 4158 Lisp_Object dummy_val; | |
| 4159 int jj; | |
| 867 | 4160 Ibyte strr[MAX_ICHAR_LEN]; |
| 428 | 4161 |
| 4162 unified_range_table_get_range (p, i, &first, &last, | |
| 4163 &dummy_val); | |
| 4164 for (jj = first; jj <= last && jj < 0x80; jj++) | |
| 4165 fastmap[jj] = 1; | |
| 4166 /* Ranges below 0x100 can span charsets, but there | |
| 4167 are only two (Control-1 and Latin-1), and | |
| 4168 either first or last has to be in them. */ | |
| 867 | 4169 set_itext_ichar (strr, first); |
| 428 | 4170 fastmap[*strr] = 1; |
| 4171 if (last < 0x100) | |
| 4172 { | |
| 867 | 4173 set_itext_ichar (strr, last); |
| 428 | 4174 fastmap[*strr] = 1; |
| 4175 } | |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4176 else if (MOST_POSITIVE_FIXNUM == last) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4177 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4178 /* This is RECC_MULTIBYTE or RECC_NONASCII; true for all |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4179 non-ASCII characters. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4180 jj = 0x80; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4181 while (jj < 0xA0) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4182 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4183 fastmap[jj++] = 1; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4184 } |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4185 } |
| 428 | 4186 } |
| 4187 } | |
| 4188 break; | |
| 4189 | |
| 4190 case charset_mule_not: | |
| 4191 { | |
| 4192 int nentries; | |
| 4193 int i; | |
|
4832
07fa38c30fdf
fix messed-up fastmap calculation in charset_mule_not
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
4194 int smallest_prev = 0; |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4195 Bitbyte flags = *p++; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4196 |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4197 if (flags) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4198 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4199 /* We need to consult the syntax table, fastmap won't |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4200 work. */ |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4201 bufp->can_be_null = 1; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4202 goto done; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
4203 } |
| 428 | 4204 |
| 4205 nentries = unified_range_table_nentries (p); | |
| 4206 for (i = 0; i < nentries; i++) | |
| 4207 { | |
| 4208 EMACS_INT first, last; | |
| 4209 Lisp_Object dummy_val; | |
| 4210 int jj; | |
| 4211 | |
| 4212 unified_range_table_get_range (p, i, &first, &last, | |
| 4213 &dummy_val); | |
| 4214 for (jj = smallest_prev; jj < first && jj < 0x80; jj++) | |
| 4215 fastmap[jj] = 1; | |
| 4216 smallest_prev = last + 1; | |
| 4217 if (smallest_prev >= 0x80) | |
| 4218 break; | |
| 4219 } | |
|
4832
07fa38c30fdf
fix messed-up fastmap calculation in charset_mule_not
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
4220 |
|
07fa38c30fdf
fix messed-up fastmap calculation in charset_mule_not
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
4221 /* Also set lead bytes after the end */ |
|
07fa38c30fdf
fix messed-up fastmap calculation in charset_mule_not
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
4222 for (i = smallest_prev; i < 0x80; i++) |
|
07fa38c30fdf
fix messed-up fastmap calculation in charset_mule_not
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
4223 fastmap[i] = 1; |
|
07fa38c30fdf
fix messed-up fastmap calculation in charset_mule_not
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
4224 |
| 428 | 4225 /* Calculating which leading bytes are actually allowed |
| 4226 here is rather difficult, so we just punt and allow | |
| 4227 all of them. */ | |
| 4228 for (i = 0x80; i < 0xA0; i++) | |
| 4229 fastmap[i] = 1; | |
| 4230 } | |
| 4231 break; | |
| 4232 #endif /* MULE */ | |
| 4233 | |
| 4234 | |
| 4235 case anychar: | |
| 4236 { | |
| 4237 int fastmap_newline = fastmap['\n']; | |
| 4238 | |
| 4239 /* `.' matches anything ... */ | |
| 4240 #ifdef MULE | |
| 4241 /* "anything" only includes bytes that can be the | |
| 4242 first byte of a character. */ | |
| 4243 for (j = 0; j < 0xA0; j++) | |
| 4244 fastmap[j] = 1; | |
| 4245 #else | |
| 4246 for (j = 0; j < (1 << BYTEWIDTH); j++) | |
| 4247 fastmap[j] = 1; | |
| 4248 #endif | |
| 4249 | |
| 4250 /* ... except perhaps newline. */ | |
| 4251 if (!(bufp->syntax & RE_DOT_NEWLINE)) | |
| 4252 fastmap['\n'] = fastmap_newline; | |
| 4253 | |
| 4254 /* Return if we have already set `can_be_null'; if we have, | |
| 4255 then the fastmap is irrelevant. Something's wrong here. */ | |
| 4256 else if (bufp->can_be_null) | |
| 4257 goto done; | |
| 4258 | |
| 4259 /* Otherwise, have to check alternative paths. */ | |
| 4260 break; | |
| 4261 } | |
| 4262 | |
| 826 | 4263 #ifndef emacs |
| 4264 case wordchar: | |
| 4265 for (j = 0; j < (1 << BYTEWIDTH); j++) | |
| 4266 if (SYNTAX (ignored, j) == Sword) | |
| 4267 fastmap[j] = 1; | |
| 4268 break; | |
| 4269 | |
| 4270 case notwordchar: | |
| 4271 for (j = 0; j < (1 << BYTEWIDTH); j++) | |
| 4272 if (SYNTAX (ignored, j) != Sword) | |
| 4273 fastmap[j] = 1; | |
| 4274 break; | |
| 4275 #else /* emacs */ | |
| 4276 case wordchar: | |
| 4277 case notwordchar: | |
| 460 | 4278 case wordbound: |
| 4279 case notwordbound: | |
| 4280 case wordbeg: | |
| 4281 case wordend: | |
| 4282 case notsyntaxspec: | |
| 4283 case syntaxspec: | |
| 4284 /* This match depends on text properties. These end with | |
| 4285 aborting optimizations. */ | |
| 4286 bufp->can_be_null = 1; | |
| 4287 goto done; | |
| 826 | 4288 #if 0 /* all of the following code is unused now that the `syntax-table' |
| 4289 property exists -- it's trickier to do this than just look in | |
| 4290 the buffer. &&#### but we could just use the syntax-cache stuff | |
| 4291 instead; why don't we? --ben */ | |
| 4292 case wordchar: | |
| 4293 k = (int) Sword; | |
| 4294 goto matchsyntax; | |
| 4295 | |
| 4296 case notwordchar: | |
| 4297 k = (int) Sword; | |
| 4298 goto matchnotsyntax; | |
| 4299 | |
| 428 | 4300 case syntaxspec: |
| 4301 k = *p++; | |
| 826 | 4302 matchsyntax: |
| 428 | 4303 #ifdef MULE |
| 4304 for (j = 0; j < 0x80; j++) | |
| 826 | 4305 if (SYNTAX |
| 4306 (XCHAR_TABLE (BUFFER_MIRROR_SYNTAX_TABLE (lispbuf)), j) == | |
| 428 | 4307 (enum syntaxcode) k) |
| 4308 fastmap[j] = 1; | |
| 4309 for (j = 0x80; j < 0xA0; j++) | |
| 4310 { | |
| 826 | 4311 if (leading_byte_prefix_p ((unsigned char) j)) |
| 428 | 4312 /* too complicated to calculate this right */ |
| 4313 fastmap[j] = 1; | |
| 4314 else | |
| 4315 { | |
| 4316 int multi_p; | |
| 4317 Lisp_Object cset; | |
| 4318 | |
| 826 | 4319 cset = charset_by_leading_byte (j); |
| 428 | 4320 if (CHARSETP (cset)) |
| 4321 { | |
| 826 | 4322 if (charset_syntax (lispbuf, cset, &multi_p) |
| 428 | 4323 == Sword || multi_p) |
| 4324 fastmap[j] = 1; | |
| 4325 } | |
| 4326 } | |
| 4327 } | |
| 446 | 4328 #else /* not MULE */ |
| 428 | 4329 for (j = 0; j < (1 << BYTEWIDTH); j++) |
| 826 | 4330 if (SYNTAX |
| 4331 (XCHAR_TABLE (BUFFER_MIRROR_SYNTAX_TABLE (lispbuf)), j) == | |
| 428 | 4332 (enum syntaxcode) k) |
| 4333 fastmap[j] = 1; | |
| 446 | 4334 #endif /* MULE */ |
| 428 | 4335 break; |
| 4336 | |
| 4337 | |
| 4338 case notsyntaxspec: | |
| 4339 k = *p++; | |
| 826 | 4340 matchnotsyntax: |
| 428 | 4341 #ifdef MULE |
| 4342 for (j = 0; j < 0x80; j++) | |
| 826 | 4343 if (SYNTAX |
| 428 | 4344 (XCHAR_TABLE |
| 826 | 4345 (BUFFER_MIRROR_SYNTAX_TABLE (lispbuf)), j) != |
| 428 | 4346 (enum syntaxcode) k) |
| 4347 fastmap[j] = 1; | |
| 4348 for (j = 0x80; j < 0xA0; j++) | |
| 4349 { | |
| 826 | 4350 if (leading_byte_prefix_p ((unsigned char) j)) |
| 428 | 4351 /* too complicated to calculate this right */ |
| 4352 fastmap[j] = 1; | |
| 4353 else | |
| 4354 { | |
| 4355 int multi_p; | |
| 4356 Lisp_Object cset; | |
| 4357 | |
| 826 | 4358 cset = charset_by_leading_byte (j); |
| 428 | 4359 if (CHARSETP (cset)) |
| 4360 { | |
| 826 | 4361 if (charset_syntax (lispbuf, cset, &multi_p) |
| 428 | 4362 != Sword || multi_p) |
| 4363 fastmap[j] = 1; | |
| 4364 } | |
| 4365 } | |
| 4366 } | |
| 446 | 4367 #else /* not MULE */ |
| 428 | 4368 for (j = 0; j < (1 << BYTEWIDTH); j++) |
| 826 | 4369 if (SYNTAX |
| 428 | 4370 (XCHAR_TABLE |
| 826 | 4371 (BUFFER_MIRROR_SYNTAX_TABLE (lispbuf)), j) != |
| 428 | 4372 (enum syntaxcode) k) |
| 4373 fastmap[j] = 1; | |
| 446 | 4374 #endif /* MULE */ |
| 428 | 4375 break; |
| 826 | 4376 #endif /* 0 */ |
| 428 | 4377 |
| 4378 #ifdef MULE | |
| 4379 /* 97/2/17 jhod category patch */ | |
| 4380 case categoryspec: | |
| 4381 case notcategoryspec: | |
| 4382 bufp->can_be_null = 1; | |
| 1333 | 4383 UNBIND_REGEX_MALLOC_CHECK (); |
| 428 | 4384 return 0; |
| 4385 /* end if category patch */ | |
| 4386 #endif /* MULE */ | |
| 4387 | |
| 4388 /* All cases after this match the empty string. These end with | |
| 4389 `continue'. */ | |
| 4390 case before_dot: | |
| 4391 case at_dot: | |
| 4392 case after_dot: | |
| 4393 continue; | |
| 826 | 4394 #endif /* emacs */ |
| 428 | 4395 |
| 4396 | |
| 4397 case no_op: | |
| 4398 case begline: | |
| 4399 case endline: | |
| 4400 case begbuf: | |
| 4401 case endbuf: | |
| 460 | 4402 #ifndef emacs |
| 428 | 4403 case wordbound: |
| 4404 case notwordbound: | |
| 4405 case wordbeg: | |
| 4406 case wordend: | |
| 460 | 4407 #endif |
| 428 | 4408 case push_dummy_failure: |
| 4409 continue; | |
| 4410 | |
| 4411 | |
| 4412 case jump_n: | |
| 4413 case pop_failure_jump: | |
| 4414 case maybe_pop_jump: | |
| 4415 case jump: | |
| 4416 case jump_past_alt: | |
| 4417 case dummy_failure_jump: | |
| 4418 EXTRACT_NUMBER_AND_INCR (j, p); | |
| 4419 p += j; | |
| 4420 if (j > 0) | |
| 4421 continue; | |
| 4422 | |
| 4423 /* Jump backward implies we just went through the body of a | |
| 4424 loop and matched nothing. Opcode jumped to should be | |
| 4425 `on_failure_jump' or `succeed_n'. Just treat it like an | |
| 4426 ordinary jump. For a * loop, it has pushed its failure | |
| 4427 point already; if so, discard that as redundant. */ | |
| 4428 if ((re_opcode_t) *p != on_failure_jump | |
| 4429 && (re_opcode_t) *p != succeed_n) | |
| 4430 continue; | |
| 4431 | |
| 4432 p++; | |
| 4433 EXTRACT_NUMBER_AND_INCR (j, p); | |
| 4434 p += j; | |
| 4435 | |
| 4436 /* If what's on the stack is where we are now, pop it. */ | |
| 4437 if (!FAIL_STACK_EMPTY () | |
| 4438 && fail_stack.stack[fail_stack.avail - 1].pointer == p) | |
| 4439 fail_stack.avail--; | |
| 4440 | |
| 4441 continue; | |
| 4442 | |
| 4443 | |
| 4444 case on_failure_jump: | |
| 4445 case on_failure_keep_string_jump: | |
| 4446 handle_on_failure_jump: | |
| 4447 EXTRACT_NUMBER_AND_INCR (j, p); | |
| 4448 | |
| 4449 /* For some patterns, e.g., `(a?)?', `p+j' here points to the | |
| 4450 end of the pattern. We don't want to push such a point, | |
| 4451 since when we restore it above, entering the switch will | |
| 4452 increment `p' past the end of the pattern. We don't need | |
| 4453 to push such a point since we obviously won't find any more | |
| 4454 fastmap entries beyond `pend'. Such a pattern can match | |
| 4455 the null string, though. */ | |
| 4456 if (p + j < pend) | |
| 4457 { | |
| 4458 if (!PUSH_PATTERN_OP (p + j, fail_stack)) | |
| 4459 { | |
| 4460 RESET_FAIL_STACK (); | |
| 1333 | 4461 UNBIND_REGEX_MALLOC_CHECK (); |
| 428 | 4462 return -2; |
| 4463 } | |
| 4464 } | |
| 4465 else | |
| 4466 bufp->can_be_null = 1; | |
| 4467 | |
| 4468 if (succeed_n_p) | |
| 4469 { | |
| 4470 EXTRACT_NUMBER_AND_INCR (k, p); /* Skip the n. */ | |
| 4471 succeed_n_p = false; | |
| 4472 } | |
| 4473 | |
| 4474 continue; | |
| 4475 | |
| 4476 | |
| 4477 case succeed_n: | |
| 4478 /* Get to the number of times to succeed. */ | |
| 4479 p += 2; | |
| 4480 | |
| 4481 /* Increment p past the n for when k != 0. */ | |
| 4482 EXTRACT_NUMBER_AND_INCR (k, p); | |
| 4483 if (k == 0) | |
| 4484 { | |
| 4485 p -= 4; | |
| 4486 succeed_n_p = true; /* Spaghetti code alert. */ | |
| 4487 goto handle_on_failure_jump; | |
| 4488 } | |
| 4489 continue; | |
| 4490 | |
| 4491 | |
| 4492 case set_number_at: | |
| 4493 p += 4; | |
| 4494 continue; | |
| 4495 | |
| 4496 | |
| 4497 case start_memory: | |
| 4498 case stop_memory: | |
| 4499 p += 2; | |
| 4500 continue; | |
| 4501 | |
| 4502 | |
| 4503 default: | |
| 2500 | 4504 ABORT (); /* We have listed all the cases. */ |
| 428 | 4505 } /* switch *p++ */ |
| 4506 | |
| 4507 /* Getting here means we have found the possible starting | |
| 4508 characters for one path of the pattern -- and that the empty | |
| 4509 string does not match. We need not follow this path further. | |
| 4510 Instead, look at the next alternative (remembered on the | |
| 4511 stack), or quit if no more. The test at the top of the loop | |
| 4512 does these things. */ | |
| 4513 path_can_be_null = false; | |
| 4514 p = pend; | |
| 4515 } /* while p */ | |
| 4516 | |
| 4517 /* Set `can_be_null' for the last path (also the first path, if the | |
| 4518 pattern is empty). */ | |
| 4519 bufp->can_be_null |= path_can_be_null; | |
| 4520 | |
| 4521 done: | |
| 4522 RESET_FAIL_STACK (); | |
| 1333 | 4523 UNBIND_REGEX_MALLOC_CHECK (); |
| 428 | 4524 return 0; |
| 4525 } /* re_compile_fastmap */ | |
| 4526 | |
| 4527 /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | |
| 4528 ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use | |
| 4529 this memory for recording register information. STARTS and ENDS | |
| 4530 must be allocated using the malloc library routine, and must each | |
| 4531 be at least NUM_REGS * sizeof (regoff_t) bytes long. | |
| 4532 | |
| 4533 If NUM_REGS == 0, then subsequent matches should allocate their own | |
| 4534 register data. | |
| 4535 | |
| 4536 Unless this function is called, the first search or match using | |
| 4537 PATTERN_BUFFER will allocate its own register data, without | |
| 4538 freeing the old data. */ | |
| 4539 | |
| 4540 void | |
| 4541 re_set_registers (struct re_pattern_buffer *bufp, struct re_registers *regs, | |
| 647 | 4542 int num_regs, regoff_t *starts, regoff_t *ends) |
| 428 | 4543 { |
| 4544 if (num_regs) | |
| 4545 { | |
| 4546 bufp->regs_allocated = REGS_REALLOCATE; | |
| 4547 regs->num_regs = num_regs; | |
| 4548 regs->start = starts; | |
| 4549 regs->end = ends; | |
| 4550 } | |
| 4551 else | |
| 4552 { | |
| 4553 bufp->regs_allocated = REGS_UNALLOCATED; | |
| 4554 regs->num_regs = 0; | |
| 4555 regs->start = regs->end = (regoff_t *) 0; | |
| 4556 } | |
| 4557 } | |
| 4558 | |
| 4559 /* Searching routines. */ | |
| 4560 | |
| 4561 /* Like re_search_2, below, but only one string is specified, and | |
| 4562 doesn't let you say where to stop matching. */ | |
| 4563 | |
| 4564 int | |
| 442 | 4565 re_search (struct re_pattern_buffer *bufp, const char *string, int size, |
| 826 | 4566 int startpos, int range, struct re_registers *regs |
| 4567 RE_LISP_CONTEXT_ARGS_DECL) | |
| 428 | 4568 { |
| 4569 return re_search_2 (bufp, NULL, 0, string, size, startpos, range, | |
| 826 | 4570 regs, size RE_LISP_CONTEXT_ARGS); |
| 428 | 4571 } |
| 4572 | |
| 4573 /* Using the compiled pattern in BUFP->buffer, first tries to match the | |
| 4574 virtual concatenation of STRING1 and STRING2, starting first at index | |
| 4575 STARTPOS, then at STARTPOS + 1, and so on. | |
| 4576 | |
| 4577 STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. | |
| 4578 | |
| 4579 RANGE is how far to scan while trying to match. RANGE = 0 means try | |
| 4580 only at STARTPOS; in general, the last start tried is STARTPOS + | |
| 4581 RANGE. | |
| 4582 | |
| 826 | 4583 All sizes and positions refer to bytes (not chars); under Mule, the code |
| 4584 knows about the format of the text and will only check at positions | |
| 4585 where a character starts. | |
| 4586 | |
| 428 | 4587 With MULE, RANGE is a byte position, not a char position. The last |
| 4588 start tried is the character starting <= STARTPOS + RANGE. | |
| 4589 | |
| 4590 In REGS, return the indices of the virtual concatenation of STRING1 | |
| 4591 and STRING2 that matched the entire BUFP->buffer and its contained | |
| 4592 subexpressions. | |
| 4593 | |
| 4594 Do not consider matching one past the index STOP in the virtual | |
| 4595 concatenation of STRING1 and STRING2. | |
| 4596 | |
| 4597 We return either the position in the strings at which the match was | |
| 4598 found, -1 if no match, or -2 if error (such as failure | |
| 4599 stack overflow). */ | |
| 4600 | |
| 4601 int | |
| 446 | 4602 re_search_2 (struct re_pattern_buffer *bufp, const char *str1, |
| 4603 int size1, const char *str2, int size2, int startpos, | |
| 826 | 4604 int range, struct re_registers *regs, int stop |
| 4605 RE_LISP_CONTEXT_ARGS_DECL) | |
| 428 | 4606 { |
| 4607 int val; | |
| 446 | 4608 re_char *string1 = (re_char *) str1; |
| 4609 re_char *string2 = (re_char *) str2; | |
| 428 | 4610 REGISTER char *fastmap = bufp->fastmap; |
| 446 | 4611 REGISTER RE_TRANSLATE_TYPE translate = bufp->translate; |
| 428 | 4612 int total_size = size1 + size2; |
| 4613 int endpos = startpos + range; | |
| 4614 #ifdef REGEX_BEGLINE_CHECK | |
| 4615 int anchored_at_begline = 0; | |
| 4616 #endif | |
| 446 | 4617 re_char *d; |
| 826 | 4618 #ifdef emacs |
| 4619 Internal_Format fmt = buffer_or_other_internal_format (lispobj); | |
| 1346 | 4620 #ifdef REL_ALLOC |
| 4621 Ibyte *orig_buftext = | |
| 4622 BUFFERP (lispobj) ? | |
| 4623 BYTE_BUF_BYTE_ADDRESS (XBUFFER (lispobj), | |
| 4624 BYTE_BUF_BEGV (XBUFFER (lispobj))) : | |
| 4625 0; | |
| 4626 #endif | |
| 1333 | 4627 #ifdef ERROR_CHECK_MALLOC |
| 4628 int depth; | |
| 4629 #endif | |
| 826 | 4630 #endif /* emacs */ |
| 4631 #if 1 | |
| 4632 int forward_search_p; | |
| 4633 #endif | |
| 428 | 4634 |
| 4635 /* Check for out-of-range STARTPOS. */ | |
| 4636 if (startpos < 0 || startpos > total_size) | |
| 4637 return -1; | |
| 4638 | |
| 4639 /* Fix up RANGE if it might eventually take us outside | |
| 4640 the virtual concatenation of STRING1 and STRING2. */ | |
| 4641 if (endpos < 0) | |
| 4642 range = 0 - startpos; | |
| 4643 else if (endpos > total_size) | |
| 4644 range = total_size - startpos; | |
| 4645 | |
| 826 | 4646 #if 1 |
| 4647 forward_search_p = range > 0; | |
| 4648 #endif | |
| 4649 | |
| 428 | 4650 /* If the search isn't to be a backwards one, don't waste time in a |
| 4651 search for a pattern that must be anchored. */ | |
| 4652 if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0) | |
| 4653 { | |
| 4654 if (startpos > 0) | |
| 4655 return -1; | |
| 4656 else | |
| 4657 { | |
| 442 | 4658 d = ((const unsigned char *) |
| 428 | 4659 (startpos >= size1 ? string2 - size1 : string1) + startpos); |
| 867 | 4660 range = itext_ichar_len_fmt (d, fmt); |
| 428 | 4661 } |
| 4662 } | |
| 4663 | |
| 460 | 4664 #ifdef emacs |
| 4665 /* In a forward search for something that starts with \=. | |
| 4666 don't keep searching past point. */ | |
| 4667 if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0) | |
| 4668 { | |
| 826 | 4669 if (!BUFFERP (lispobj)) |
| 4670 return -1; | |
|
4527
8418d1ad4944
Fix at_dot regex under Mule. <87hc6rv53v.fsf@uwakimon.sk.tsukuba.ac.jp>
Stephen J. Turnbull <stephen@xemacs.org>
parents:
3300
diff
changeset
|
4671 range = (BYTE_BUF_PT (XBUFFER (lispobj)) |
|
8418d1ad4944
Fix at_dot regex under Mule. <87hc6rv53v.fsf@uwakimon.sk.tsukuba.ac.jp>
Stephen J. Turnbull <stephen@xemacs.org>
parents:
3300
diff
changeset
|
4672 - BYTE_BUF_BEGV (XBUFFER (lispobj)) - startpos); |
| 460 | 4673 if (range < 0) |
| 4674 return -1; | |
| 4675 } | |
| 4676 #endif /* emacs */ | |
| 4677 | |
| 1333 | 4678 #ifdef ERROR_CHECK_MALLOC |
| 4679 /* Do this after the above return()s. */ | |
| 4680 depth = bind_regex_malloc_disallowed (1); | |
| 4681 #endif | |
| 4682 | |
| 428 | 4683 /* Update the fastmap now if not correct already. */ |
| 1333 | 4684 BEGIN_REGEX_MALLOC_OK (); |
| 428 | 4685 if (fastmap && !bufp->fastmap_accurate) |
| 826 | 4686 if (re_compile_fastmap (bufp RE_LISP_SHORT_CONTEXT_ARGS) == -2) |
| 1333 | 4687 { |
| 4688 END_REGEX_MALLOC_OK (); | |
| 4689 UNBIND_REGEX_MALLOC_CHECK (); | |
| 4690 return -2; | |
| 4691 } | |
| 4692 | |
| 4693 END_REGEX_MALLOC_OK (); | |
| 4694 RE_SEARCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 428 | 4695 |
| 4696 #ifdef REGEX_BEGLINE_CHECK | |
| 4697 { | |
| 647 | 4698 long i = 0; |
| 428 | 4699 |
| 4700 while (i < bufp->used) | |
| 4701 { | |
| 4702 if (bufp->buffer[i] == start_memory || | |
| 4703 bufp->buffer[i] == stop_memory) | |
| 4704 i += 2; | |
| 4705 else | |
| 4706 break; | |
| 4707 } | |
| 4708 anchored_at_begline = i < bufp->used && bufp->buffer[i] == begline; | |
| 4709 } | |
| 4710 #endif | |
| 4711 | |
| 460 | 4712 #ifdef emacs |
| 1333 | 4713 BEGIN_REGEX_MALLOC_OK (); |
| 826 | 4714 scache = setup_syntax_cache (scache, lispobj, lispbuf, |
| 4715 offset_to_charxpos (lispobj, startpos), | |
| 4716 1); | |
| 1333 | 4717 END_REGEX_MALLOC_OK (); |
| 4718 RE_SEARCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 460 | 4719 #endif |
| 4720 | |
| 428 | 4721 /* Loop through the string, looking for a place to start matching. */ |
| 4722 for (;;) | |
| 4723 { | |
| 4724 #ifdef REGEX_BEGLINE_CHECK | |
| 826 | 4725 /* If the regex is anchored at the beginning of a line (i.e. with a |
| 4726 ^), then we can speed things up by skipping to the next | |
| 4727 beginning-of-line. However, to determine "beginning of line" we | |
| 4728 need to look at the previous char, so can't do this check if at | |
| 4729 beginning of either string. (Well, we could if at the beginning of | |
| 4730 the second string, but it would require additional code, and this | |
| 4731 is just an optimization.) */ | |
| 4732 if (anchored_at_begline && startpos > 0 && startpos != size1) | |
| 428 | 4733 { |
| 826 | 4734 if (range > 0) |
| 4735 { | |
| 4736 /* whose stupid idea was it anyway to make this | |
| 4737 function take two strings to match?? */ | |
| 4738 int lim = 0; | |
| 4739 re_char *orig_d; | |
| 4740 re_char *stop_d; | |
| 4741 | |
| 4742 /* Compute limit as below in fastmap code, so we are guaranteed | |
| 4743 to remain within a single string. */ | |
| 4744 if (startpos < size1 && startpos + range >= size1) | |
| 4745 lim = range - (size1 - startpos); | |
| 4746 | |
| 4747 d = ((const unsigned char *) | |
| 4748 (startpos >= size1 ? string2 - size1 : string1) + startpos); | |
| 4749 orig_d = d; | |
| 4750 stop_d = d + range - lim; | |
| 4751 | |
| 4752 /* We want to find the next location (including the current | |
| 4753 one) where the previous char is a newline, so back up one | |
| 4754 and search forward for a newline. */ | |
| 867 | 4755 DEC_IBYTEPTR_FMT (d, fmt); /* Ok, since startpos != size1. */ |
| 826 | 4756 |
| 4757 /* Written out as an if-else to avoid testing `translate' | |
| 4758 inside the loop. */ | |
| 4759 if (TRANSLATE_P (translate)) | |
| 4760 while (d < stop_d && | |
| 867 | 4761 RE_TRANSLATE_1 (itext_ichar_fmt (d, fmt, lispobj)) |
| 826 | 4762 != '\n') |
| 867 | 4763 INC_IBYTEPTR_FMT (d, fmt); |
| 826 | 4764 else |
| 4765 while (d < stop_d && | |
| 867 | 4766 itext_ichar_ascii_fmt (d, fmt, lispobj) != '\n') |
| 4767 INC_IBYTEPTR_FMT (d, fmt); | |
| 826 | 4768 |
| 4769 /* If we were stopped by a newline, skip forward over it. | |
| 4770 Otherwise we will get in an infloop when our start position | |
| 4771 was at begline. */ | |
| 4772 if (d < stop_d) | |
| 867 | 4773 INC_IBYTEPTR_FMT (d, fmt); |
| 826 | 4774 range -= d - orig_d; |
| 4775 startpos += d - orig_d; | |
| 4776 #if 1 | |
| 4777 assert (!forward_search_p || range >= 0); | |
| 4778 #endif | |
| 4779 } | |
| 4780 else if (range < 0) | |
| 4781 { | |
| 4782 /* We're lazy, like in the fastmap code below */ | |
| 867 | 4783 Ichar c; |
| 826 | 4784 |
| 4785 d = ((const unsigned char *) | |
| 4786 (startpos >= size1 ? string2 - size1 : string1) + startpos); | |
| 867 | 4787 DEC_IBYTEPTR_FMT (d, fmt); |
| 4788 c = itext_ichar_fmt (d, fmt, lispobj); | |
| 826 | 4789 c = RE_TRANSLATE (c); |
| 4790 if (c != '\n') | |
| 4791 goto advance; | |
| 4792 } | |
| 428 | 4793 } |
| 4794 #endif /* REGEX_BEGLINE_CHECK */ | |
| 4795 | |
| 4796 /* If a fastmap is supplied, skip quickly over characters that | |
| 4797 cannot be the start of a match. If the pattern can match the | |
| 4798 null string, however, we don't need to skip characters; we want | |
| 4799 the first null string. */ | |
| 4800 if (fastmap && startpos < total_size && !bufp->can_be_null) | |
| 4801 { | |
| 826 | 4802 /* For the moment, fastmap always works as if buffer |
| 4803 is in default format, so convert chars in the search strings | |
| 4804 into default format as we go along, if necessary. | |
| 4805 | |
| 4806 &&#### fastmap needs rethinking for 8-bit-fixed so | |
| 4807 it's faster. We need it to reflect the raw | |
| 4808 8-bit-fixed values. That isn't so hard if we assume | |
| 4809 that the top 96 bytes represent a single 1-byte | |
| 4810 charset. For 16-bit/32-bit stuff it's probably not | |
| 4811 worth it to make the fastmap represent the raw, due to | |
| 4812 its nature -- we'd have to use the LSB for the | |
| 4813 fastmap, and that causes lots of problems with Mule | |
| 4814 chars, where it essentially wipes out the usefulness | |
| 4815 of the fastmap entirely. */ | |
| 428 | 4816 if (range > 0) /* Searching forwards. */ |
| 4817 { | |
| 4818 int lim = 0; | |
| 4819 int irange = range; | |
| 4820 | |
| 4821 if (startpos < size1 && startpos + range >= size1) | |
| 4822 lim = range - (size1 - startpos); | |
| 4823 | |
| 442 | 4824 d = ((const unsigned char *) |
| 428 | 4825 (startpos >= size1 ? string2 - size1 : string1) + startpos); |
| 4826 | |
| 4827 /* Written out as an if-else to avoid testing `translate' | |
| 4828 inside the loop. */ | |
| 446 | 4829 if (TRANSLATE_P (translate)) |
| 826 | 4830 { |
| 4831 while (range > lim) | |
| 4832 { | |
| 4833 re_char *old_d = d; | |
| 428 | 4834 #ifdef MULE |
| 867 | 4835 Ibyte tempch[MAX_ICHAR_LEN]; |
| 4836 Ichar buf_ch = | |
| 4837 RE_TRANSLATE_1 (itext_ichar_fmt (d, fmt, lispobj)); | |
| 4838 set_itext_ichar (tempch, buf_ch); | |
| 826 | 4839 if (fastmap[*tempch]) |
| 4840 break; | |
| 446 | 4841 #else |
| 826 | 4842 if (fastmap[(unsigned char) RE_TRANSLATE_1 (*d)]) |
| 4843 break; | |
| 446 | 4844 #endif /* MULE */ |
| 867 | 4845 INC_IBYTEPTR_FMT (d, fmt); |
| 826 | 4846 range -= (d - old_d); |
| 4847 #if 1 | |
| 1333 | 4848 assert (!forward_search_p || range >= 0); |
| 826 | 4849 #endif |
| 4850 } | |
| 4851 } | |
| 4852 #ifdef MULE | |
| 4853 else if (fmt != FORMAT_DEFAULT) | |
| 4854 { | |
| 4855 while (range > lim) | |
| 4856 { | |
| 4857 re_char *old_d = d; | |
| 867 | 4858 Ibyte tempch[MAX_ICHAR_LEN]; |
| 4859 Ichar buf_ch = itext_ichar_fmt (d, fmt, lispobj); | |
| 4860 set_itext_ichar (tempch, buf_ch); | |
| 826 | 4861 if (fastmap[*tempch]) |
| 4862 break; | |
| 867 | 4863 INC_IBYTEPTR_FMT (d, fmt); |
| 826 | 4864 range -= (d - old_d); |
| 4865 #if 1 | |
| 1333 | 4866 assert (!forward_search_p || range >= 0); |
| 826 | 4867 #endif |
| 4868 } | |
| 4869 } | |
| 4870 #endif /* MULE */ | |
| 428 | 4871 else |
| 826 | 4872 { |
| 4873 while (range > lim && !fastmap[*d]) | |
| 4874 { | |
| 4875 re_char *old_d = d; | |
| 867 | 4876 INC_IBYTEPTR (d); |
| 826 | 4877 range -= (d - old_d); |
| 4878 #if 1 | |
| 4879 assert (!forward_search_p || range >= 0); | |
| 4880 #endif | |
| 4881 } | |
| 4882 } | |
| 428 | 4883 |
| 4884 startpos += irange - range; | |
| 4885 } | |
| 4886 else /* Searching backwards. */ | |
| 4887 { | |
| 826 | 4888 /* #### It's not clear why we don't just write a loop, like |
| 4889 for the moving-forward case. Perhaps the writer got lazy, | |
| 4890 since backward searches aren't so common. */ | |
| 4891 d = ((const unsigned char *) | |
| 4892 (startpos >= size1 ? string2 - size1 : string1) + startpos); | |
| 428 | 4893 #ifdef MULE |
| 826 | 4894 { |
| 867 | 4895 Ibyte tempch[MAX_ICHAR_LEN]; |
| 4896 Ichar buf_ch = | |
| 4897 RE_TRANSLATE (itext_ichar_fmt (d, fmt, lispobj)); | |
| 4898 set_itext_ichar (tempch, buf_ch); | |
| 826 | 4899 if (!fastmap[*tempch]) |
| 4900 goto advance; | |
| 4901 } | |
| 428 | 4902 #else |
| 826 | 4903 if (!fastmap[(unsigned char) RE_TRANSLATE (*d)]) |
| 446 | 4904 goto advance; |
| 826 | 4905 #endif /* MULE */ |
| 428 | 4906 } |
| 4907 } | |
| 4908 | |
| 4909 /* If can't match the null string, and that's all we have left, fail. */ | |
| 4910 if (range >= 0 && startpos == total_size && fastmap | |
| 4911 && !bufp->can_be_null) | |
| 1333 | 4912 { |
| 4913 UNBIND_REGEX_MALLOC_CHECK (); | |
| 4914 return -1; | |
| 4915 } | |
| 428 | 4916 |
| 4917 #ifdef emacs /* XEmacs added, w/removal of immediate_quit */ | |
| 4918 if (!no_quit_in_re_search) | |
| 1333 | 4919 { |
| 4920 BEGIN_REGEX_MALLOC_OK (); | |
| 4921 QUIT; | |
| 4922 END_REGEX_MALLOC_OK (); | |
| 4923 RE_SEARCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 4924 } | |
| 4925 | |
| 428 | 4926 #endif |
| 1333 | 4927 BEGIN_REGEX_MALLOC_OK (); |
| 428 | 4928 val = re_match_2_internal (bufp, string1, size1, string2, size2, |
| 826 | 4929 startpos, regs, stop |
| 4930 RE_LISP_CONTEXT_ARGS); | |
| 428 | 4931 #ifndef REGEX_MALLOC |
| 1333 | 4932 ALLOCA_GARBAGE_COLLECT (); |
| 428 | 4933 #endif |
| 1333 | 4934 END_REGEX_MALLOC_OK (); |
| 4935 RE_SEARCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 428 | 4936 |
| 4937 if (val >= 0) | |
| 1333 | 4938 { |
| 4939 UNBIND_REGEX_MALLOC_CHECK (); | |
| 4940 return startpos; | |
| 4941 } | |
| 428 | 4942 |
| 4943 if (val == -2) | |
| 1333 | 4944 { |
| 4945 UNBIND_REGEX_MALLOC_CHECK (); | |
| 4946 return -2; | |
| 4947 } | |
| 4948 | |
| 4949 RE_SEARCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 428 | 4950 advance: |
| 4951 if (!range) | |
| 4952 break; | |
| 4953 else if (range > 0) | |
| 4954 { | |
| 826 | 4955 Bytecount d_size; |
| 442 | 4956 d = ((const unsigned char *) |
| 428 | 4957 (startpos >= size1 ? string2 - size1 : string1) + startpos); |
| 867 | 4958 d_size = itext_ichar_len_fmt (d, fmt); |
| 428 | 4959 range -= d_size; |
| 826 | 4960 #if 1 |
| 4961 assert (!forward_search_p || range >= 0); | |
| 4962 #endif | |
| 428 | 4963 startpos += d_size; |
| 4964 } | |
| 4965 else | |
| 4966 { | |
| 826 | 4967 Bytecount d_size; |
| 428 | 4968 /* Note startpos > size1 not >=. If we are on the |
| 4969 string1/string2 boundary, we want to backup into string1. */ | |
| 442 | 4970 d = ((const unsigned char *) |
| 428 | 4971 (startpos > size1 ? string2 - size1 : string1) + startpos); |
| 867 | 4972 DEC_IBYTEPTR_FMT (d, fmt); |
| 4973 d_size = itext_ichar_len_fmt (d, fmt); | |
| 428 | 4974 range += d_size; |
| 826 | 4975 #if 1 |
| 4976 assert (!forward_search_p || range >= 0); | |
| 4977 #endif | |
| 428 | 4978 startpos -= d_size; |
| 4979 } | |
| 4980 } | |
| 1333 | 4981 UNBIND_REGEX_MALLOC_CHECK (); |
| 428 | 4982 return -1; |
| 4983 } /* re_search_2 */ | |
| 826 | 4984 |
| 428 | 4985 |
| 4986 /* Declarations and macros for re_match_2. */ | |
| 4987 | |
| 4988 /* This converts PTR, a pointer into one of the search strings `string1' | |
| 4989 and `string2' into an offset from the beginning of that string. */ | |
| 4990 #define POINTER_TO_OFFSET(ptr) \ | |
| 4991 (FIRST_STRING_P (ptr) \ | |
| 4992 ? ((regoff_t) ((ptr) - string1)) \ | |
| 4993 : ((regoff_t) ((ptr) - string2 + size1))) | |
| 4994 | |
| 4995 /* Macros for dealing with the split strings in re_match_2. */ | |
| 4996 | |
| 4997 #define MATCHING_IN_FIRST_STRING (dend == end_match_1) | |
| 4998 | |
| 4999 /* Call before fetching a character with *d. This switches over to | |
| 5000 string2 if necessary. */ | |
| 826 | 5001 #define REGEX_PREFETCH() \ |
| 428 | 5002 while (d == dend) \ |
| 5003 { \ | |
| 5004 /* End of string2 => fail. */ \ | |
| 5005 if (dend == end_match_2) \ | |
| 5006 goto fail; \ | |
| 5007 /* End of string1 => advance to string2. */ \ | |
| 5008 d = string2; \ | |
| 5009 dend = end_match_2; \ | |
| 5010 } | |
| 5011 | |
| 5012 | |
| 5013 /* Test if at very beginning or at very end of the virtual concatenation | |
| 5014 of `string1' and `string2'. If only one string, it's `string2'. */ | |
| 5015 #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) | |
| 5016 #define AT_STRINGS_END(d) ((d) == end2) | |
| 5017 | |
| 5018 /* XEmacs change: | |
| 5019 If the given position straddles the string gap, return the equivalent | |
| 5020 position that is before or after the gap, respectively; otherwise, | |
| 5021 return the same position. */ | |
| 5022 #define POS_BEFORE_GAP_UNSAFE(d) ((d) == string2 ? end1 : (d)) | |
| 5023 #define POS_AFTER_GAP_UNSAFE(d) ((d) == end1 ? string2 : (d)) | |
| 5024 | |
| 5025 /* Test if CH is a word-constituent character. (XEmacs change) */ | |
| 826 | 5026 #define WORDCHAR_P(ch) \ |
| 5027 (SYNTAX (BUFFER_MIRROR_SYNTAX_TABLE (lispbuf), ch) == Sword) | |
| 428 | 5028 |
| 5029 /* Free everything we malloc. */ | |
| 5030 #ifdef MATCH_MAY_ALLOCATE | |
| 1726 | 5031 #define FREE_VAR(var,type) if (var) REGEX_FREE (var, type); var = NULL |
| 428 | 5032 #define FREE_VARIABLES() \ |
| 5033 do { \ | |
| 1333 | 5034 UNBIND_REGEX_MALLOC_CHECK (); \ |
| 428 | 5035 REGEX_FREE_STACK (fail_stack.stack); \ |
| 1726 | 5036 FREE_VAR (regstart, re_char **); \ |
| 5037 FREE_VAR (regend, re_char **); \ | |
| 5038 FREE_VAR (old_regstart, re_char **); \ | |
| 5039 FREE_VAR (old_regend, re_char **); \ | |
| 5040 FREE_VAR (best_regstart, re_char **); \ | |
| 5041 FREE_VAR (best_regend, re_char **); \ | |
| 5042 FREE_VAR (reg_info, register_info_type *); \ | |
| 5043 FREE_VAR (reg_dummy, re_char **); \ | |
| 5044 FREE_VAR (reg_info_dummy, register_info_type *); \ | |
| 428 | 5045 } while (0) |
| 446 | 5046 #else /* not MATCH_MAY_ALLOCATE */ |
| 1333 | 5047 #define FREE_VARIABLES() \ |
| 5048 do { \ | |
| 5049 UNBIND_REGEX_MALLOC_CHECK (); \ | |
| 5050 } while (0) | |
| 446 | 5051 #endif /* MATCH_MAY_ALLOCATE */ |
| 428 | 5052 |
| 5053 /* These values must meet several constraints. They must not be valid | |
| 5054 register values; since we have a limit of 255 registers (because | |
| 5055 we use only one byte in the pattern for the register number), we can | |
| 5056 use numbers larger than 255. They must differ by 1, because of | |
| 5057 NUM_FAILURE_ITEMS above. And the value for the lowest register must | |
| 5058 be larger than the value for the highest register, so we do not try | |
| 5059 to actually save any registers when none are active. */ | |
| 5060 #define NO_HIGHEST_ACTIVE_REG (1 << BYTEWIDTH) | |
| 5061 #define NO_LOWEST_ACTIVE_REG (NO_HIGHEST_ACTIVE_REG + 1) | |
| 5062 | |
| 5063 /* Matching routines. */ | |
| 5064 | |
| 826 | 5065 #ifndef emacs /* XEmacs never uses this. */ |
| 428 | 5066 /* re_match is like re_match_2 except it takes only a single string. */ |
| 5067 | |
| 5068 int | |
| 442 | 5069 re_match (struct re_pattern_buffer *bufp, const char *string, int size, |
| 826 | 5070 int pos, struct re_registers *regs |
| 5071 RE_LISP_CONTEXT_ARGS_DECL) | |
| 428 | 5072 { |
| 446 | 5073 int result = re_match_2_internal (bufp, NULL, 0, (re_char *) string, size, |
| 826 | 5074 pos, regs, size |
| 5075 RE_LISP_CONTEXT_ARGS); | |
| 1333 | 5076 ALLOCA_GARBAGE_COLLECT (); |
| 428 | 5077 return result; |
| 5078 } | |
| 5079 #endif /* not emacs */ | |
| 5080 | |
| 5081 /* re_match_2 matches the compiled pattern in BUFP against the | |
| 5082 (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 and | |
| 5083 SIZE2, respectively). We start matching at POS, and stop matching | |
| 5084 at STOP. | |
| 5085 | |
| 5086 If REGS is non-null and the `no_sub' field of BUFP is nonzero, we | |
| 5087 store offsets for the substring each group matched in REGS. See the | |
| 5088 documentation for exactly how many groups we fill. | |
| 5089 | |
| 5090 We return -1 if no match, -2 if an internal error (such as the | |
| 5091 failure stack overflowing). Otherwise, we return the length of the | |
| 5092 matched substring. */ | |
| 5093 | |
| 5094 int | |
| 442 | 5095 re_match_2 (struct re_pattern_buffer *bufp, const char *string1, |
| 5096 int size1, const char *string2, int size2, int pos, | |
| 826 | 5097 struct re_registers *regs, int stop |
| 5098 RE_LISP_CONTEXT_ARGS_DECL) | |
| 428 | 5099 { |
| 460 | 5100 int result; |
| 5101 | |
| 5102 #ifdef emacs | |
| 826 | 5103 scache = setup_syntax_cache (scache, lispobj, lispbuf, |
| 5104 offset_to_charxpos (lispobj, pos), | |
| 5105 1); | |
| 460 | 5106 #endif |
| 5107 | |
| 5108 result = re_match_2_internal (bufp, (re_char *) string1, size1, | |
| 5109 (re_char *) string2, size2, | |
| 826 | 5110 pos, regs, stop |
| 5111 RE_LISP_CONTEXT_ARGS); | |
| 460 | 5112 |
| 1333 | 5113 ALLOCA_GARBAGE_COLLECT (); |
| 428 | 5114 return result; |
| 5115 } | |
| 5116 | |
| 5117 /* This is a separate function so that we can force an alloca cleanup | |
| 5118 afterwards. */ | |
| 5119 static int | |
| 446 | 5120 re_match_2_internal (struct re_pattern_buffer *bufp, re_char *string1, |
| 5121 int size1, re_char *string2, int size2, int pos, | |
| 826 | 5122 struct re_registers *regs, int stop |
| 2333 | 5123 RE_LISP_CONTEXT_ARGS_MULE_DECL) |
| 428 | 5124 { |
| 5125 /* General temporaries. */ | |
| 5126 int mcnt; | |
| 5127 unsigned char *p1; | |
| 5128 int should_succeed; /* XEmacs change */ | |
| 5129 | |
| 5130 /* Just past the end of the corresponding string. */ | |
| 446 | 5131 re_char *end1, *end2; |
| 428 | 5132 |
| 5133 /* Pointers into string1 and string2, just past the last characters in | |
| 5134 each to consider matching. */ | |
| 446 | 5135 re_char *end_match_1, *end_match_2; |
| 428 | 5136 |
| 5137 /* Where we are in the data, and the end of the current string. */ | |
| 446 | 5138 re_char *d, *dend; |
| 428 | 5139 |
| 5140 /* Where we are in the pattern, and the end of the pattern. */ | |
| 5141 unsigned char *p = bufp->buffer; | |
| 5142 REGISTER unsigned char *pend = p + bufp->used; | |
| 5143 | |
| 5144 /* Mark the opcode just after a start_memory, so we can test for an | |
| 5145 empty subpattern when we get to the stop_memory. */ | |
| 446 | 5146 re_char *just_past_start_mem = 0; |
| 428 | 5147 |
| 5148 /* We use this to map every character in the string. */ | |
| 446 | 5149 RE_TRANSLATE_TYPE translate = bufp->translate; |
| 428 | 5150 |
| 5151 /* Failure point stack. Each place that can handle a failure further | |
| 5152 down the line pushes a failure point on this stack. It consists of | |
| 5153 restart, regend, and reg_info for all registers corresponding to | |
| 5154 the subexpressions we're currently inside, plus the number of such | |
| 5155 registers, and, finally, two char *'s. The first char * is where | |
| 5156 to resume scanning the pattern; the second one is where to resume | |
| 5157 scanning the strings. If the latter is zero, the failure point is | |
| 5158 a ``dummy''; if a failure happens and the failure point is a dummy, | |
| 5159 it gets discarded and the next one is tried. */ | |
| 5160 #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ | |
| 5161 fail_stack_type fail_stack; | |
| 5162 #endif | |
| 5163 #ifdef DEBUG | |
| 647 | 5164 static int failure_id; |
| 5165 int nfailure_points_pushed = 0, nfailure_points_popped = 0; | |
| 428 | 5166 #endif |
| 5167 | |
| 771 | 5168 #ifdef REGEX_REL_ALLOC |
| 428 | 5169 /* This holds the pointer to the failure stack, when |
| 5170 it is allocated relocatably. */ | |
| 5171 fail_stack_elt_t *failure_stack_ptr; | |
| 5172 #endif | |
| 5173 | |
| 5174 /* We fill all the registers internally, independent of what we | |
| 5175 return, for use in backreferences. The number here includes | |
| 5176 an element for register zero. */ | |
| 647 | 5177 int num_regs = bufp->re_ngroups + 1; |
| 428 | 5178 |
| 5179 /* The currently active registers. */ | |
| 647 | 5180 int lowest_active_reg = NO_LOWEST_ACTIVE_REG; |
| 5181 int highest_active_reg = NO_HIGHEST_ACTIVE_REG; | |
| 428 | 5182 |
| 5183 /* Information on the contents of registers. These are pointers into | |
| 5184 the input strings; they record just what was matched (on this | |
| 5185 attempt) by a subexpression part of the pattern, that is, the | |
| 5186 regnum-th regstart pointer points to where in the pattern we began | |
| 5187 matching and the regnum-th regend points to right after where we | |
| 5188 stopped matching the regnum-th subexpression. (The zeroth register | |
| 5189 keeps track of what the whole pattern matches.) */ | |
| 5190 #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
| 446 | 5191 re_char **regstart, **regend; |
| 428 | 5192 #endif |
| 5193 | |
| 5194 /* If a group that's operated upon by a repetition operator fails to | |
| 5195 match anything, then the register for its start will need to be | |
| 5196 restored because it will have been set to wherever in the string we | |
| 5197 are when we last see its open-group operator. Similarly for a | |
| 5198 register's end. */ | |
| 5199 #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
| 446 | 5200 re_char **old_regstart, **old_regend; |
| 428 | 5201 #endif |
| 5202 | |
| 5203 /* The is_active field of reg_info helps us keep track of which (possibly | |
| 5204 nested) subexpressions we are currently in. The matched_something | |
| 5205 field of reg_info[reg_num] helps us tell whether or not we have | |
| 5206 matched any of the pattern so far this time through the reg_num-th | |
| 5207 subexpression. These two fields get reset each time through any | |
| 5208 loop their register is in. */ | |
| 5209 #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ | |
| 5210 register_info_type *reg_info; | |
| 5211 #endif | |
| 5212 | |
| 5213 /* The following record the register info as found in the above | |
| 5214 variables when we find a match better than any we've seen before. | |
| 5215 This happens as we backtrack through the failure points, which in | |
| 5216 turn happens only if we have not yet matched the entire string. */ | |
| 647 | 5217 int best_regs_set = false; |
| 428 | 5218 #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ |
| 446 | 5219 re_char **best_regstart, **best_regend; |
| 428 | 5220 #endif |
| 5221 | |
| 5222 /* Logically, this is `best_regend[0]'. But we don't want to have to | |
| 5223 allocate space for that if we're not allocating space for anything | |
| 5224 else (see below). Also, we never need info about register 0 for | |
| 5225 any of the other register vectors, and it seems rather a kludge to | |
| 5226 treat `best_regend' differently than the rest. So we keep track of | |
| 5227 the end of the best match so far in a separate variable. We | |
| 5228 initialize this to NULL so that when we backtrack the first time | |
| 5229 and need to test it, it's not garbage. */ | |
| 446 | 5230 re_char *match_end = NULL; |
| 428 | 5231 |
| 5232 /* This helps SET_REGS_MATCHED avoid doing redundant work. */ | |
| 5233 int set_regs_matched_done = 0; | |
| 5234 | |
| 5235 /* Used when we pop values we don't care about. */ | |
| 5236 #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
| 446 | 5237 re_char **reg_dummy; |
| 428 | 5238 register_info_type *reg_info_dummy; |
| 5239 #endif | |
| 5240 | |
| 5241 #ifdef DEBUG | |
| 5242 /* Counts the total number of registers pushed. */ | |
| 647 | 5243 int num_regs_pushed = 0; |
| 428 | 5244 #endif |
| 5245 | |
| 5246 /* 1 if this match ends in the same string (string1 or string2) | |
| 5247 as the best previous match. */ | |
| 460 | 5248 re_bool same_str_p; |
| 428 | 5249 |
| 5250 /* 1 if this match is the best seen so far. */ | |
| 460 | 5251 re_bool best_match_p; |
| 428 | 5252 |
| 826 | 5253 #ifdef emacs |
| 5254 Internal_Format fmt = buffer_or_other_internal_format (lispobj); | |
| 1346 | 5255 #ifdef REL_ALLOC |
| 5256 Ibyte *orig_buftext = | |
| 5257 BUFFERP (lispobj) ? | |
| 5258 BYTE_BUF_BYTE_ADDRESS (XBUFFER (lispobj), | |
| 5259 BYTE_BUF_BEGV (XBUFFER (lispobj))) : | |
| 5260 0; | |
| 5261 #endif | |
| 5262 | |
| 1333 | 5263 #ifdef ERROR_CHECK_MALLOC |
| 5264 int depth = bind_regex_malloc_disallowed (1); | |
| 5265 #endif | |
| 826 | 5266 #endif /* emacs */ |
| 771 | 5267 |
| 5041 | 5268 DEBUG_MATCH_PRINT1 ("\n\nEntering re_match_2.\n"); |
| 428 | 5269 |
| 1333 | 5270 BEGIN_REGEX_MALLOC_OK (); |
| 428 | 5271 INIT_FAIL_STACK (); |
| 1333 | 5272 END_REGEX_MALLOC_OK (); |
| 428 | 5273 |
| 5274 #ifdef MATCH_MAY_ALLOCATE | |
| 5275 /* Do not bother to initialize all the register variables if there are | |
| 5276 no groups in the pattern, as it takes a fair amount of time. If | |
| 5277 there are groups, we include space for register 0 (the whole | |
| 5278 pattern), even though we never use it, since it simplifies the | |
| 5279 array indexing. We should fix this. */ | |
| 502 | 5280 if (bufp->re_ngroups) |
| 428 | 5281 { |
| 1333 | 5282 BEGIN_REGEX_MALLOC_OK (); |
| 446 | 5283 regstart = REGEX_TALLOC (num_regs, re_char *); |
| 5284 regend = REGEX_TALLOC (num_regs, re_char *); | |
| 5285 old_regstart = REGEX_TALLOC (num_regs, re_char *); | |
| 5286 old_regend = REGEX_TALLOC (num_regs, re_char *); | |
| 5287 best_regstart = REGEX_TALLOC (num_regs, re_char *); | |
| 5288 best_regend = REGEX_TALLOC (num_regs, re_char *); | |
| 428 | 5289 reg_info = REGEX_TALLOC (num_regs, register_info_type); |
| 446 | 5290 reg_dummy = REGEX_TALLOC (num_regs, re_char *); |
| 428 | 5291 reg_info_dummy = REGEX_TALLOC (num_regs, register_info_type); |
| 1333 | 5292 END_REGEX_MALLOC_OK (); |
| 428 | 5293 |
| 5294 if (!(regstart && regend && old_regstart && old_regend && reg_info | |
| 5295 && best_regstart && best_regend && reg_dummy && reg_info_dummy)) | |
| 5296 { | |
| 5297 FREE_VARIABLES (); | |
| 5298 return -2; | |
| 5299 } | |
| 5300 } | |
| 5301 else | |
| 5302 { | |
| 5303 /* We must initialize all our variables to NULL, so that | |
| 5304 `FREE_VARIABLES' doesn't try to free them. */ | |
| 5305 regstart = regend = old_regstart = old_regend = best_regstart | |
| 5306 = best_regend = reg_dummy = NULL; | |
| 5307 reg_info = reg_info_dummy = (register_info_type *) NULL; | |
| 5308 } | |
| 5309 #endif /* MATCH_MAY_ALLOCATE */ | |
| 5310 | |
| 1333 | 5311 #if defined (emacs) && defined (REL_ALLOC) |
| 5312 { | |
| 5313 /* If the allocations above (or the call to setup_syntax_cache() in | |
| 5314 re_match_2) caused a rel-alloc relocation, then fix up the data | |
| 5315 pointers */ | |
| 1346 | 5316 Bytecount offset = offset_post_relocation (lispobj, orig_buftext); |
| 1333 | 5317 if (offset) |
| 5318 { | |
| 5319 string1 += offset; | |
| 5320 string2 += offset; | |
| 5321 } | |
| 5322 } | |
| 5323 #endif /* defined (emacs) && defined (REL_ALLOC) */ | |
| 5324 | |
| 428 | 5325 /* The starting position is bogus. */ |
| 5326 if (pos < 0 || pos > size1 + size2) | |
| 5327 { | |
| 5328 FREE_VARIABLES (); | |
| 5329 return -1; | |
| 5330 } | |
| 5331 | |
| 5332 /* Initialize subexpression text positions to -1 to mark ones that no | |
| 5333 start_memory/stop_memory has been seen for. Also initialize the | |
| 5334 register information struct. */ | |
| 5335 for (mcnt = 1; mcnt < num_regs; mcnt++) | |
| 5336 { | |
| 5337 regstart[mcnt] = regend[mcnt] | |
| 5338 = old_regstart[mcnt] = old_regend[mcnt] = REG_UNSET_VALUE; | |
| 5339 | |
| 5340 REG_MATCH_NULL_STRING_P (reg_info[mcnt]) = MATCH_NULL_UNSET_VALUE; | |
| 5341 IS_ACTIVE (reg_info[mcnt]) = 0; | |
| 5342 MATCHED_SOMETHING (reg_info[mcnt]) = 0; | |
| 5343 EVER_MATCHED_SOMETHING (reg_info[mcnt]) = 0; | |
| 5344 } | |
| 5345 /* We move `string1' into `string2' if the latter's empty -- but not if | |
| 5346 `string1' is null. */ | |
| 5347 if (size2 == 0 && string1 != NULL) | |
| 5348 { | |
| 5349 string2 = string1; | |
| 5350 size2 = size1; | |
| 5351 string1 = 0; | |
| 5352 size1 = 0; | |
| 5353 } | |
| 5354 end1 = string1 + size1; | |
| 5355 end2 = string2 + size2; | |
| 5356 | |
| 5357 /* Compute where to stop matching, within the two strings. */ | |
| 5358 if (stop <= size1) | |
| 5359 { | |
| 5360 end_match_1 = string1 + stop; | |
| 5361 end_match_2 = string2; | |
| 5362 } | |
| 5363 else | |
| 5364 { | |
| 5365 end_match_1 = end1; | |
| 5366 end_match_2 = string2 + stop - size1; | |
| 5367 } | |
| 5368 | |
| 5369 /* `p' scans through the pattern as `d' scans through the data. | |
| 5370 `dend' is the end of the input string that `d' points within. `d' | |
| 5371 is advanced into the following input string whenever necessary, but | |
| 5372 this happens before fetching; therefore, at the beginning of the | |
| 5373 loop, `d' can be pointing at the end of a string, but it cannot | |
| 5374 equal `string2'. */ | |
| 5375 if (size1 > 0 && pos <= size1) | |
| 5376 { | |
| 5377 d = string1 + pos; | |
| 5378 dend = end_match_1; | |
| 5379 } | |
| 5380 else | |
| 5381 { | |
| 5382 d = string2 + pos - size1; | |
| 5383 dend = end_match_2; | |
| 5384 } | |
| 5385 | |
| 5041 | 5386 DEBUG_MATCH_PRINT1 ("The compiled pattern is: \n"); |
| 5387 DEBUG_MATCH_PRINT_COMPILED_PATTERN (bufp, p, pend); | |
| 5388 DEBUG_MATCH_PRINT1 ("The string to match is: `"); | |
| 5389 DEBUG_MATCH_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); | |
| 5390 DEBUG_MATCH_PRINT1 ("'\n"); | |
| 428 | 5391 |
| 5392 /* This loops over pattern commands. It exits by returning from the | |
| 5393 function if the match is complete, or it drops through if the match | |
| 5394 fails at this starting point in the input data. */ | |
| 5395 for (;;) | |
| 5396 { | |
| 5041 | 5397 DEBUG_MATCH_PRINT2 ("\n0x%lx: ", (long) p); |
| 428 | 5398 #ifdef emacs /* XEmacs added, w/removal of immediate_quit */ |
| 5399 if (!no_quit_in_re_search) | |
| 1333 | 5400 { |
| 5401 BEGIN_REGEX_MALLOC_OK (); | |
| 5402 QUIT; | |
| 5403 END_REGEX_MALLOC_OK (); | |
| 1346 | 5404 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); |
| 1333 | 5405 } |
| 428 | 5406 #endif |
| 5407 | |
| 5408 if (p == pend) | |
| 5409 { /* End of pattern means we might have succeeded. */ | |
| 5041 | 5410 DEBUG_MATCH_PRINT1 ("end of pattern ... "); |
| 428 | 5411 |
| 5412 /* If we haven't matched the entire string, and we want the | |
| 5413 longest match, try backtracking. */ | |
| 5414 if (d != end_match_2) | |
| 5415 { | |
| 5416 same_str_p = (FIRST_STRING_P (match_end) | |
| 5417 == MATCHING_IN_FIRST_STRING); | |
| 5418 | |
| 5419 /* AIX compiler got confused when this was combined | |
| 5420 with the previous declaration. */ | |
| 5421 if (same_str_p) | |
| 5422 best_match_p = d > match_end; | |
| 5423 else | |
| 5424 best_match_p = !MATCHING_IN_FIRST_STRING; | |
| 5425 | |
| 5041 | 5426 DEBUG_MATCH_PRINT1 ("backtracking.\n"); |
| 428 | 5427 |
| 5428 if (!FAIL_STACK_EMPTY ()) | |
| 5429 { /* More failure points to try. */ | |
| 5430 | |
| 5431 /* If exceeds best match so far, save it. */ | |
| 5432 if (!best_regs_set || best_match_p) | |
| 5433 { | |
| 5434 best_regs_set = true; | |
| 5435 match_end = d; | |
| 5436 | |
| 5041 | 5437 DEBUG_MATCH_PRINT1 ("\nSAVING match as best so far.\n"); |
| 428 | 5438 |
| 5439 for (mcnt = 1; mcnt < num_regs; mcnt++) | |
| 5440 { | |
| 5441 best_regstart[mcnt] = regstart[mcnt]; | |
| 5442 best_regend[mcnt] = regend[mcnt]; | |
| 5443 } | |
| 5444 } | |
| 5445 goto fail; | |
| 5446 } | |
| 5447 | |
| 5448 /* If no failure points, don't restore garbage. And if | |
| 5449 last match is real best match, don't restore second | |
| 5450 best one. */ | |
| 5451 else if (best_regs_set && !best_match_p) | |
| 5452 { | |
| 5453 restore_best_regs: | |
| 5454 /* Restore best match. It may happen that `dend == | |
| 5455 end_match_1' while the restored d is in string2. | |
| 5456 For example, the pattern `x.*y.*z' against the | |
| 5457 strings `x-' and `y-z-', if the two strings are | |
| 5458 not consecutive in memory. */ | |
| 5041 | 5459 DEBUG_MATCH_PRINT1 ("Restoring best registers.\n"); |
| 428 | 5460 |
| 5461 d = match_end; | |
| 5462 dend = ((d >= string1 && d <= end1) | |
| 5463 ? end_match_1 : end_match_2); | |
| 5464 | |
| 5465 for (mcnt = 1; mcnt < num_regs; mcnt++) | |
| 5466 { | |
| 5467 regstart[mcnt] = best_regstart[mcnt]; | |
| 5468 regend[mcnt] = best_regend[mcnt]; | |
| 5469 } | |
| 5470 } | |
| 5471 } /* d != end_match_2 */ | |
| 5472 | |
| 5473 succeed_label: | |
| 5041 | 5474 DEBUG_MATCH_PRINT1 ("Accepting match.\n"); |
| 428 | 5475 |
| 5476 /* If caller wants register contents data back, do it. */ | |
| 1028 | 5477 { |
| 5478 int num_nonshy_regs = bufp->re_nsub + 1; | |
| 5479 if (regs && !bufp->no_sub) | |
| 5480 { | |
| 5481 /* Have the register data arrays been allocated? */ | |
| 5482 if (bufp->regs_allocated == REGS_UNALLOCATED) | |
| 5483 { /* No. So allocate them with malloc. We need one | |
| 5484 extra element beyond `num_regs' for the `-1' marker | |
| 5485 GNU code uses. */ | |
| 5486 regs->num_regs = MAX (RE_NREGS, num_nonshy_regs + 1); | |
| 1333 | 5487 BEGIN_REGEX_MALLOC_OK (); |
| 1028 | 5488 regs->start = TALLOC (regs->num_regs, regoff_t); |
| 5489 regs->end = TALLOC (regs->num_regs, regoff_t); | |
| 1333 | 5490 END_REGEX_MALLOC_OK (); |
| 5491 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 1028 | 5492 if (regs->start == NULL || regs->end == NULL) |
| 5493 { | |
| 5494 FREE_VARIABLES (); | |
| 5495 return -2; | |
| 5496 } | |
| 5497 bufp->regs_allocated = REGS_REALLOCATE; | |
| 5498 } | |
| 5499 else if (bufp->regs_allocated == REGS_REALLOCATE) | |
| 5500 { /* Yes. If we need more elements than were already | |
| 5501 allocated, reallocate them. If we need fewer, just | |
| 5502 leave it alone. */ | |
| 5503 if (regs->num_regs < num_nonshy_regs + 1) | |
| 5504 { | |
| 5505 regs->num_regs = num_nonshy_regs + 1; | |
| 1333 | 5506 BEGIN_REGEX_MALLOC_OK (); |
| 1028 | 5507 RETALLOC (regs->start, regs->num_regs, regoff_t); |
| 5508 RETALLOC (regs->end, regs->num_regs, regoff_t); | |
| 1333 | 5509 END_REGEX_MALLOC_OK (); |
| 5510 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 1028 | 5511 if (regs->start == NULL || regs->end == NULL) |
| 5512 { | |
| 5513 FREE_VARIABLES (); | |
| 5514 return -2; | |
| 5515 } | |
| 5516 } | |
| 5517 } | |
| 5518 else | |
| 5519 { | |
| 5520 /* The braces fend off a "empty body in an else-statement" | |
| 5521 warning under GCC when assert expands to nothing. */ | |
| 5522 assert (bufp->regs_allocated == REGS_FIXED); | |
| 5523 } | |
| 5524 | |
| 5525 /* Convert the pointer data in `regstart' and `regend' to | |
| 5526 indices. Register zero has to be set differently, | |
| 5527 since we haven't kept track of any info for it. */ | |
| 5528 if (regs->num_regs > 0) | |
| 5529 { | |
| 5530 regs->start[0] = pos; | |
| 5531 regs->end[0] = (MATCHING_IN_FIRST_STRING | |
| 5532 ? ((regoff_t) (d - string1)) | |
| 5533 : ((regoff_t) (d - string2 + size1))); | |
| 5534 } | |
| 5535 | |
| 2639 | 5536 /* Map over the NUM_NONSHY_REGS non-shy internal registers. |
| 5537 Copy each into the corresponding external register. | |
| 5538 MCNT indexes external registers. */ | |
| 1028 | 5539 for (mcnt = 1; mcnt < MIN (num_nonshy_regs, regs->num_regs); |
| 5540 mcnt++) | |
| 5541 { | |
| 5542 int internal_reg = bufp->external_to_internal_register[mcnt]; | |
| 5543 if (REG_UNSET (regstart[internal_reg]) || | |
| 5544 REG_UNSET (regend[internal_reg])) | |
| 5545 regs->start[mcnt] = regs->end[mcnt] = -1; | |
| 5546 else | |
| 5547 { | |
| 5548 regs->start[mcnt] = | |
| 5549 (regoff_t) POINTER_TO_OFFSET (regstart[internal_reg]); | |
| 5550 regs->end[mcnt] = | |
| 5551 (regoff_t) POINTER_TO_OFFSET (regend[internal_reg]); | |
| 5552 } | |
| 5553 } | |
| 5554 } /* regs && !bufp->no_sub */ | |
| 5555 | |
| 5556 /* If we have regs and the regs structure has more elements than | |
| 2639 | 5557 were in the pattern, set the extra elements starting with |
| 5558 NUM_NONSHY_REGS to -1. If we (re)allocated the registers, | |
| 5559 this is the case, because we always allocate enough to have | |
| 5560 at least one -1 at the end. | |
| 1028 | 5561 |
| 5562 We do this even when no_sub is set because some applications | |
| 5563 (XEmacs) reuse register structures which may contain stale | |
| 5564 information, and permit attempts to access those registers. | |
| 5565 | |
| 5566 It would be possible to require the caller to do this, but we'd | |
| 5567 have to change the API for this function to reflect that, and | |
| 1425 | 5568 audit all callers. Note: as of 2003-04-17 callers in XEmacs |
| 5569 do clear the registers, but it's safer to leave this code in | |
| 5570 because of reallocation. | |
| 5571 */ | |
| 1028 | 5572 if (regs && regs->num_regs > 0) |
| 5573 for (mcnt = num_nonshy_regs; mcnt < regs->num_regs; mcnt++) | |
| 5574 regs->start[mcnt] = regs->end[mcnt] = -1; | |
| 5575 } | |
| 5041 | 5576 DEBUG_MATCH_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n", |
| 428 | 5577 nfailure_points_pushed, nfailure_points_popped, |
| 5578 nfailure_points_pushed - nfailure_points_popped); | |
| 5041 | 5579 DEBUG_MATCH_PRINT2 ("%u registers pushed.\n", num_regs_pushed); |
| 428 | 5580 |
| 5581 mcnt = d - pos - (MATCHING_IN_FIRST_STRING | |
| 5582 ? string1 | |
| 5583 : string2 - size1); | |
| 5584 | |
| 5041 | 5585 DEBUG_MATCH_PRINT2 ("Returning %d from re_match_2.\n", mcnt); |
| 428 | 5586 |
| 5587 FREE_VARIABLES (); | |
| 5588 return mcnt; | |
| 5589 } | |
| 5590 | |
| 5591 /* Otherwise match next pattern command. */ | |
|
4759
aa5ed11f473b
Remove support for obsolete systems. See xemacs-patches message with ID
Jerry James <james@xemacs.org>
parents:
4750
diff
changeset
|
5592 switch ((re_opcode_t) *p++) |
| 428 | 5593 { |
| 5594 /* Ignore these. Used to ignore the n of succeed_n's which | |
| 5595 currently have n == 0. */ | |
| 5596 case no_op: | |
| 5041 | 5597 DEBUG_MATCH_PRINT1 ("EXECUTING no_op.\n"); |
| 428 | 5598 break; |
| 5599 | |
| 5600 case succeed: | |
| 5041 | 5601 DEBUG_MATCH_PRINT1 ("EXECUTING succeed.\n"); |
| 428 | 5602 goto succeed_label; |
| 5603 | |
| 826 | 5604 /* Match exactly a string of length n in the pattern. The |
| 5605 following byte in the pattern defines n, and the n bytes after | |
| 5606 that make up the string to match. (Under Mule, this will be in | |
| 5607 the default internal format.) */ | |
| 428 | 5608 case exactn: |
| 5609 mcnt = *p++; | |
| 5041 | 5610 DEBUG_MATCH_PRINT2 ("EXECUTING exactn %d.\n", mcnt); |
| 428 | 5611 |
| 5612 /* This is written out as an if-else so we don't waste time | |
| 5613 testing `translate' inside the loop. */ | |
| 446 | 5614 if (TRANSLATE_P (translate)) |
| 428 | 5615 { |
| 5616 do | |
| 5617 { | |
| 446 | 5618 #ifdef MULE |
| 5619 Bytecount pat_len; | |
| 5620 | |
| 450 | 5621 REGEX_PREFETCH (); |
| 867 | 5622 if (RE_TRANSLATE_1 (itext_ichar_fmt (d, fmt, lispobj)) |
| 5623 != itext_ichar (p)) | |
| 428 | 5624 goto fail; |
| 446 | 5625 |
| 867 | 5626 pat_len = itext_ichar_len (p); |
| 446 | 5627 p += pat_len; |
| 867 | 5628 INC_IBYTEPTR_FMT (d, fmt); |
| 446 | 5629 |
| 5630 mcnt -= pat_len; | |
| 5631 #else /* not MULE */ | |
| 450 | 5632 REGEX_PREFETCH (); |
| 826 | 5633 if ((unsigned char) RE_TRANSLATE_1 (*d++) != *p++) |
| 446 | 5634 goto fail; |
| 5635 mcnt--; | |
| 5636 #endif | |
| 428 | 5637 } |
| 446 | 5638 while (mcnt > 0); |
| 428 | 5639 } |
| 5640 else | |
| 5641 { | |
| 826 | 5642 #ifdef MULE |
| 5643 /* If buffer format is default, then we can shortcut and just | |
| 5644 compare the text directly, byte by byte. Otherwise, we | |
| 5645 need to go character by character. */ | |
| 5646 if (fmt != FORMAT_DEFAULT) | |
| 428 | 5647 { |
| 826 | 5648 do |
| 5649 { | |
| 5650 Bytecount pat_len; | |
| 5651 | |
| 5652 REGEX_PREFETCH (); | |
| 867 | 5653 if (itext_ichar_fmt (d, fmt, lispobj) != |
| 5654 itext_ichar (p)) | |
| 826 | 5655 goto fail; |
| 5656 | |
| 867 | 5657 pat_len = itext_ichar_len (p); |
| 826 | 5658 p += pat_len; |
| 867 | 5659 INC_IBYTEPTR_FMT (d, fmt); |
| 826 | 5660 |
| 5661 mcnt -= pat_len; | |
| 5662 } | |
| 5663 while (mcnt > 0); | |
| 428 | 5664 } |
| 826 | 5665 else |
| 5666 #endif | |
| 5667 { | |
| 5668 do | |
| 5669 { | |
| 5670 REGEX_PREFETCH (); | |
| 5671 if (*d++ != *p++) goto fail; | |
| 5672 mcnt--; | |
| 5673 } | |
| 5674 while (mcnt > 0); | |
| 5675 } | |
| 428 | 5676 } |
| 5677 SET_REGS_MATCHED (); | |
| 5678 break; | |
| 5679 | |
| 5680 | |
| 5681 /* Match any character except possibly a newline or a null. */ | |
| 5682 case anychar: | |
| 5041 | 5683 DEBUG_MATCH_PRINT1 ("EXECUTING anychar.\n"); |
| 428 | 5684 |
| 450 | 5685 REGEX_PREFETCH (); |
| 428 | 5686 |
| 826 | 5687 if ((!(bufp->syntax & RE_DOT_NEWLINE) && |
| 867 | 5688 RE_TRANSLATE (itext_ichar_fmt (d, fmt, lispobj)) == '\n') |
| 826 | 5689 || (bufp->syntax & RE_DOT_NOT_NULL && |
| 867 | 5690 RE_TRANSLATE (itext_ichar_fmt (d, fmt, lispobj)) == |
| 826 | 5691 '\000')) |
| 428 | 5692 goto fail; |
| 5693 | |
| 5694 SET_REGS_MATCHED (); | |
| 5041 | 5695 DEBUG_MATCH_PRINT2 (" Matched `%d'.\n", *d); |
| 867 | 5696 INC_IBYTEPTR_FMT (d, fmt); /* XEmacs change */ |
| 428 | 5697 break; |
| 5698 | |
| 5699 | |
| 5700 case charset: | |
| 5701 case charset_not: | |
| 5702 { | |
| 1414 | 5703 REGISTER Ichar c; |
| 460 | 5704 re_bool not_p = (re_opcode_t) *(p - 1) == charset_not; |
| 458 | 5705 |
| 5041 | 5706 DEBUG_MATCH_PRINT2 ("EXECUTING charset%s.\n", not_p ? "_not" : ""); |
| 428 | 5707 |
| 450 | 5708 REGEX_PREFETCH (); |
| 867 | 5709 c = itext_ichar_fmt (d, fmt, lispobj); |
| 826 | 5710 c = RE_TRANSLATE (c); /* The character to match. */ |
| 428 | 5711 |
| 647 | 5712 /* Cast to `unsigned int' instead of `unsigned char' in case the |
| 428 | 5713 bit list is a full 32 bytes long. */ |
| 1414 | 5714 if ((unsigned int)c < (unsigned int) (*p * BYTEWIDTH) |
| 428 | 5715 && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) |
| 458 | 5716 not_p = !not_p; |
| 428 | 5717 |
| 5718 p += 1 + *p; | |
| 5719 | |
| 458 | 5720 if (!not_p) goto fail; |
| 428 | 5721 |
| 5722 SET_REGS_MATCHED (); | |
| 867 | 5723 INC_IBYTEPTR_FMT (d, fmt); /* XEmacs change */ |
| 428 | 5724 break; |
| 5725 } | |
| 5726 | |
| 5727 #ifdef MULE | |
| 5728 case charset_mule: | |
| 5729 case charset_mule_not: | |
| 5730 { | |
| 867 | 5731 REGISTER Ichar c; |
| 460 | 5732 re_bool not_p = (re_opcode_t) *(p - 1) == charset_mule_not; |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5733 Bitbyte class_bits = *p++; |
| 458 | 5734 |
| 5041 | 5735 DEBUG_MATCH_PRINT2 ("EXECUTING charset_mule%s.\n", not_p ? "_not" : ""); |
| 450 | 5736 REGEX_PREFETCH (); |
| 867 | 5737 c = itext_ichar_fmt (d, fmt, lispobj); |
| 826 | 5738 c = RE_TRANSLATE (c); /* The character to match. */ |
| 428 | 5739 |
|
5648
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5740 if ((class_bits && |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5741 ((class_bits & BIT_ALPHA && ISALPHA (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5742 || (class_bits & BIT_SPACE && ISSPACE (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5743 || (class_bits & BIT_PUNCT && ISPUNCT (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5744 || (class_bits & BIT_WORD && ISWORD (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5745 || (TRANSLATE_P (translate) ? |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5746 (class_bits & (BIT_UPPER | BIT_LOWER) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5747 && !NOCASEP (lispbuf, c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5748 : ((class_bits & BIT_UPPER && ISUPPER (c)) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5749 || (class_bits & BIT_LOWER && ISLOWER (c)))))) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5750 || EQ (Qt, unified_range_table_lookup (p, c, Qnil))) |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5751 { |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5752 not_p = !not_p; |
|
3f4a234f4672
Support non-ASCII correctly in character classes, test this.
Aidan Kehoe <kehoea@parhasard.net>
parents:
5402
diff
changeset
|
5753 } |
| 428 | 5754 |
| 5755 p += unified_range_table_bytes_used (p); | |
| 5756 | |
| 458 | 5757 if (!not_p) goto fail; |
| 428 | 5758 |
| 5759 SET_REGS_MATCHED (); | |
| 867 | 5760 INC_IBYTEPTR_FMT (d, fmt); |
| 428 | 5761 break; |
| 5762 } | |
| 5763 #endif /* MULE */ | |
| 5764 | |
| 5765 | |
| 5766 /* The beginning of a group is represented by start_memory. | |
| 5767 The arguments are the register number in the next byte, and the | |
| 5768 number of groups inner to this one in the next. The text | |
| 5769 matched within the group is recorded (in the internal | |
| 5770 registers data structure) under the register number. */ | |
| 5771 case start_memory: | |
| 5041 | 5772 DEBUG_MATCH_PRINT3 ("EXECUTING start_memory %d (%d):\n", *p, p[1]); |
| 428 | 5773 |
| 5774 /* Find out if this group can match the empty string. */ | |
| 5775 p1 = p; /* To send to group_match_null_string_p. */ | |
| 5776 | |
| 5777 if (REG_MATCH_NULL_STRING_P (reg_info[*p]) == MATCH_NULL_UNSET_VALUE) | |
| 2639 | 5778 REG_MATCH_NULL_STRING_P (reg_info[*p]) |
| 5779 = group_match_null_string_p (&p1, pend, reg_info); | |
| 5780 | |
| 5041 | 5781 DEBUG_MATCH_PRINT2 (" group CAN%s match null string\n", |
| 2639 | 5782 REG_MATCH_NULL_STRING_P (reg_info[*p]) ? "NOT" : ""); |
| 428 | 5783 |
| 5784 /* Save the position in the string where we were the last time | |
| 5785 we were at this open-group operator in case the group is | |
| 5786 operated upon by a repetition operator, e.g., with `(a*)*b' | |
| 5787 against `ab'; then we want to ignore where we are now in | |
| 5788 the string in case this attempt to match fails. */ | |
| 5789 old_regstart[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p]) | |
| 5790 ? REG_UNSET (regstart[*p]) ? d : regstart[*p] | |
| 5791 : regstart[*p]; | |
| 5041 | 5792 DEBUG_MATCH_PRINT2 (" old_regstart: %d\n", |
| 428 | 5793 POINTER_TO_OFFSET (old_regstart[*p])); |
| 5794 | |
| 5795 regstart[*p] = d; | |
| 5041 | 5796 DEBUG_MATCH_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p])); |
| 428 | 5797 |
| 5798 IS_ACTIVE (reg_info[*p]) = 1; | |
| 5799 MATCHED_SOMETHING (reg_info[*p]) = 0; | |
| 5800 | |
| 5801 /* Clear this whenever we change the register activity status. */ | |
| 5802 set_regs_matched_done = 0; | |
| 5803 | |
| 5804 /* This is the new highest active register. */ | |
| 5805 highest_active_reg = *p; | |
| 5806 | |
| 5807 /* If nothing was active before, this is the new lowest active | |
| 5808 register. */ | |
| 5809 if (lowest_active_reg == NO_LOWEST_ACTIVE_REG) | |
| 5810 lowest_active_reg = *p; | |
| 5811 | |
| 5812 /* Move past the register number and inner group count. */ | |
| 5813 p += 2; | |
| 5814 just_past_start_mem = p; | |
| 5815 | |
| 5816 break; | |
| 5817 | |
| 5818 | |
| 5819 /* The stop_memory opcode represents the end of a group. Its | |
| 5820 arguments are the same as start_memory's: the register | |
| 5821 number, and the number of inner groups. */ | |
| 5822 case stop_memory: | |
| 5041 | 5823 DEBUG_MATCH_PRINT3 ("EXECUTING stop_memory %d (%d):\n", *p, p[1]); |
| 428 | 5824 |
| 5825 /* We need to save the string position the last time we were at | |
| 5826 this close-group operator in case the group is operated | |
| 5827 upon by a repetition operator, e.g., with `((a*)*(b*)*)*' | |
| 5828 against `aba'; then we want to ignore where we are now in | |
| 5829 the string in case this attempt to match fails. */ | |
| 5830 old_regend[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p]) | |
| 5831 ? REG_UNSET (regend[*p]) ? d : regend[*p] | |
| 5832 : regend[*p]; | |
| 5041 | 5833 DEBUG_MATCH_PRINT2 (" old_regend: %d\n", |
| 428 | 5834 POINTER_TO_OFFSET (old_regend[*p])); |
| 5835 | |
| 5836 regend[*p] = d; | |
| 5041 | 5837 DEBUG_MATCH_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p])); |
| 428 | 5838 |
| 5839 /* This register isn't active anymore. */ | |
| 5840 IS_ACTIVE (reg_info[*p]) = 0; | |
| 5841 | |
| 5842 /* Clear this whenever we change the register activity status. */ | |
| 5843 set_regs_matched_done = 0; | |
| 5844 | |
| 5845 /* If this was the only register active, nothing is active | |
| 5846 anymore. */ | |
| 5847 if (lowest_active_reg == highest_active_reg) | |
| 5848 { | |
| 5849 lowest_active_reg = NO_LOWEST_ACTIVE_REG; | |
| 5850 highest_active_reg = NO_HIGHEST_ACTIVE_REG; | |
| 5851 } | |
| 5852 else | |
| 5853 { /* We must scan for the new highest active register, since | |
| 5854 it isn't necessarily one less than now: consider | |
| 5855 (a(b)c(d(e)f)g). When group 3 ends, after the f), the | |
| 5856 new highest active register is 1. */ | |
| 5857 unsigned char r = *p - 1; | |
| 5858 while (r > 0 && !IS_ACTIVE (reg_info[r])) | |
| 5859 r--; | |
| 5860 | |
| 5861 /* If we end up at register zero, that means that we saved | |
| 5862 the registers as the result of an `on_failure_jump', not | |
| 5863 a `start_memory', and we jumped to past the innermost | |
| 5864 `stop_memory'. For example, in ((.)*) we save | |
| 5865 registers 1 and 2 as a result of the *, but when we pop | |
| 5866 back to the second ), we are at the stop_memory 1. | |
| 5867 Thus, nothing is active. */ | |
| 5868 if (r == 0) | |
| 5869 { | |
| 5870 lowest_active_reg = NO_LOWEST_ACTIVE_REG; | |
| 5871 highest_active_reg = NO_HIGHEST_ACTIVE_REG; | |
| 5872 } | |
| 5873 else | |
| 5874 { | |
| 5875 highest_active_reg = r; | |
| 5876 | |
| 5877 /* 98/9/21 jhod: We've also gotta set lowest_active_reg, don't we? */ | |
| 5878 r = 1; | |
| 5879 while (r < highest_active_reg && !IS_ACTIVE(reg_info[r])) | |
| 5880 r++; | |
| 5881 lowest_active_reg = r; | |
| 5882 } | |
| 5883 } | |
| 5884 | |
| 5885 /* If just failed to match something this time around with a | |
| 5886 group that's operated on by a repetition operator, try to | |
| 5887 force exit from the ``loop'', and restore the register | |
| 5888 information for this group that we had before trying this | |
| 5889 last match. */ | |
| 5890 if ((!MATCHED_SOMETHING (reg_info[*p]) | |
| 5891 || just_past_start_mem == p - 1) | |
| 5892 && (p + 2) < pend) | |
| 5893 { | |
| 460 | 5894 re_bool is_a_jump_n = false; |
| 428 | 5895 |
| 5896 p1 = p + 2; | |
| 5897 mcnt = 0; | |
| 5898 switch ((re_opcode_t) *p1++) | |
| 5899 { | |
| 5900 case jump_n: | |
| 5901 is_a_jump_n = true; | |
| 5902 case pop_failure_jump: | |
| 5903 case maybe_pop_jump: | |
| 5904 case jump: | |
| 5905 case dummy_failure_jump: | |
| 5906 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 5907 if (is_a_jump_n) | |
| 5908 p1 += 2; | |
| 5909 break; | |
| 5910 | |
| 5911 default: | |
| 5912 /* do nothing */ ; | |
| 5913 } | |
| 5914 p1 += mcnt; | |
| 5915 | |
| 5916 /* If the next operation is a jump backwards in the pattern | |
| 5917 to an on_failure_jump right before the start_memory | |
| 5918 corresponding to this stop_memory, exit from the loop | |
| 5919 by forcing a failure after pushing on the stack the | |
| 5920 on_failure_jump's jump in the pattern, and d. */ | |
| 5921 if (mcnt < 0 && (re_opcode_t) *p1 == on_failure_jump | |
| 5922 && (re_opcode_t) p1[3] == start_memory && p1[4] == *p) | |
| 5923 { | |
| 5924 /* If this group ever matched anything, then restore | |
| 5925 what its registers were before trying this last | |
| 5926 failed match, e.g., with `(a*)*b' against `ab' for | |
| 5927 regstart[1], and, e.g., with `((a*)*(b*)*)*' | |
| 5928 against `aba' for regend[3]. | |
| 5929 | |
| 5930 Also restore the registers for inner groups for, | |
| 5931 e.g., `((a*)(b*))*' against `aba' (register 3 would | |
| 5932 otherwise get trashed). */ | |
| 5933 | |
| 5934 if (EVER_MATCHED_SOMETHING (reg_info[*p])) | |
| 5935 { | |
| 647 | 5936 int r; |
| 428 | 5937 |
| 5938 EVER_MATCHED_SOMETHING (reg_info[*p]) = 0; | |
| 5939 | |
| 5940 /* Restore this and inner groups' (if any) registers. */ | |
| 5941 for (r = *p; r < *p + *(p + 1); r++) | |
| 5942 { | |
| 5943 regstart[r] = old_regstart[r]; | |
| 5944 | |
| 5945 /* xx why this test? */ | |
| 5946 if (old_regend[r] >= regstart[r]) | |
| 5947 regend[r] = old_regend[r]; | |
| 5948 } | |
| 5949 } | |
| 5950 p1++; | |
| 5951 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 5952 PUSH_FAILURE_POINT (p1 + mcnt, d, -2); | |
| 5953 | |
| 5954 goto fail; | |
| 5955 } | |
| 5956 } | |
| 5957 | |
| 5958 /* Move past the register number and the inner group count. */ | |
| 5959 p += 2; | |
| 5960 break; | |
| 5961 | |
| 5962 | |
| 5963 /* \<digit> has been turned into a `duplicate' command which is | |
| 502 | 5964 followed by the numeric value of <digit> as the register number. |
| 5965 (Already passed through external-to-internal-register mapping, | |
| 5966 so it refers to the actual group number, not the non-shy-only | |
| 5967 numbering used in the external world.) */ | |
| 428 | 5968 case duplicate: |
| 5969 { | |
| 446 | 5970 REGISTER re_char *d2, *dend2; |
| 502 | 5971 /* Get which register to match against. */ |
| 5972 int regno = *p++; | |
| 5041 | 5973 DEBUG_MATCH_PRINT2 ("EXECUTING duplicate %d.\n", regno); |
| 428 | 5974 |
| 5975 /* Can't back reference a group which we've never matched. */ | |
| 5976 if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) | |
| 5977 goto fail; | |
| 5978 | |
| 5979 /* Where in input to try to start matching. */ | |
| 5980 d2 = regstart[regno]; | |
| 5981 | |
| 5982 /* Where to stop matching; if both the place to start and | |
| 5983 the place to stop matching are in the same string, then | |
| 5984 set to the place to stop, otherwise, for now have to use | |
| 5985 the end of the first string. */ | |
| 5986 | |
| 5987 dend2 = ((FIRST_STRING_P (regstart[regno]) | |
| 5988 == FIRST_STRING_P (regend[regno])) | |
| 5989 ? regend[regno] : end_match_1); | |
| 5990 for (;;) | |
| 5991 { | |
| 5992 /* If necessary, advance to next segment in register | |
| 5993 contents. */ | |
| 5994 while (d2 == dend2) | |
| 5995 { | |
| 5996 if (dend2 == end_match_2) break; | |
| 5997 if (dend2 == regend[regno]) break; | |
| 5998 | |
| 5999 /* End of string1 => advance to string2. */ | |
| 6000 d2 = string2; | |
| 6001 dend2 = regend[regno]; | |
| 6002 } | |
| 6003 /* At end of register contents => success */ | |
| 6004 if (d2 == dend2) break; | |
| 6005 | |
| 6006 /* If necessary, advance to next segment in data. */ | |
| 450 | 6007 REGEX_PREFETCH (); |
| 428 | 6008 |
| 6009 /* How many characters left in this segment to match. */ | |
| 6010 mcnt = dend - d; | |
| 6011 | |
| 6012 /* Want how many consecutive characters we can match in | |
| 6013 one shot, so, if necessary, adjust the count. */ | |
| 6014 if (mcnt > dend2 - d2) | |
| 6015 mcnt = dend2 - d2; | |
| 6016 | |
| 6017 /* Compare that many; failure if mismatch, else move | |
| 6018 past them. */ | |
| 446 | 6019 if (TRANSLATE_P (translate) |
| 826 | 6020 ? bcmp_translate (d, d2, mcnt, translate |
| 6021 #ifdef emacs | |
| 6022 , fmt, lispobj | |
| 6023 #endif | |
| 6024 ) | |
| 428 | 6025 : memcmp (d, d2, mcnt)) |
| 6026 goto fail; | |
| 6027 d += mcnt, d2 += mcnt; | |
| 6028 | |
| 6029 /* Do this because we've match some characters. */ | |
| 6030 SET_REGS_MATCHED (); | |
| 6031 } | |
| 6032 } | |
| 6033 break; | |
| 6034 | |
| 6035 | |
| 6036 /* begline matches the empty string at the beginning of the string | |
| 6037 (unless `not_bol' is set in `bufp'), and, if | |
| 6038 `newline_anchor' is set, after newlines. */ | |
| 6039 case begline: | |
| 5041 | 6040 DEBUG_MATCH_PRINT1 ("EXECUTING begline.\n"); |
| 428 | 6041 |
| 6042 if (AT_STRINGS_BEG (d)) | |
| 6043 { | |
| 6044 if (!bufp->not_bol) break; | |
| 6045 } | |
| 826 | 6046 else |
| 6047 { | |
| 6048 re_char *d2 = d; | |
| 867 | 6049 DEC_IBYTEPTR (d2); |
| 6050 if (itext_ichar_ascii_fmt (d2, fmt, lispobj) == '\n' && | |
| 826 | 6051 bufp->newline_anchor) |
| 6052 break; | |
| 6053 } | |
| 428 | 6054 /* In all other cases, we fail. */ |
| 6055 goto fail; | |
| 6056 | |
| 6057 | |
| 6058 /* endline is the dual of begline. */ | |
| 6059 case endline: | |
| 5041 | 6060 DEBUG_MATCH_PRINT1 ("EXECUTING endline.\n"); |
| 428 | 6061 |
| 6062 if (AT_STRINGS_END (d)) | |
| 6063 { | |
| 6064 if (!bufp->not_eol) break; | |
| 6065 } | |
| 6066 | |
| 6067 /* We have to ``prefetch'' the next character. */ | |
| 826 | 6068 else if ((d == end1 ? |
| 867 | 6069 itext_ichar_ascii_fmt (string2, fmt, lispobj) : |
| 6070 itext_ichar_ascii_fmt (d, fmt, lispobj)) == '\n' | |
| 428 | 6071 && bufp->newline_anchor) |
| 6072 { | |
| 6073 break; | |
| 6074 } | |
| 6075 goto fail; | |
| 6076 | |
| 6077 | |
| 6078 /* Match at the very beginning of the data. */ | |
| 6079 case begbuf: | |
| 5041 | 6080 DEBUG_MATCH_PRINT1 ("EXECUTING begbuf.\n"); |
| 428 | 6081 if (AT_STRINGS_BEG (d)) |
| 6082 break; | |
| 6083 goto fail; | |
| 6084 | |
| 6085 | |
| 6086 /* Match at the very end of the data. */ | |
| 6087 case endbuf: | |
| 5041 | 6088 DEBUG_MATCH_PRINT1 ("EXECUTING endbuf.\n"); |
| 428 | 6089 if (AT_STRINGS_END (d)) |
| 6090 break; | |
| 6091 goto fail; | |
| 6092 | |
| 6093 | |
| 6094 /* on_failure_keep_string_jump is used to optimize `.*\n'. It | |
| 6095 pushes NULL as the value for the string on the stack. Then | |
| 6096 `pop_failure_point' will keep the current value for the | |
| 6097 string, instead of restoring it. To see why, consider | |
| 6098 matching `foo\nbar' against `.*\n'. The .* matches the foo; | |
| 6099 then the . fails against the \n. But the next thing we want | |
| 6100 to do is match the \n against the \n; if we restored the | |
| 6101 string value, we would be back at the foo. | |
| 6102 | |
| 6103 Because this is used only in specific cases, we don't need to | |
| 6104 check all the things that `on_failure_jump' does, to make | |
| 6105 sure the right things get saved on the stack. Hence we don't | |
| 6106 share its code. The only reason to push anything on the | |
| 6107 stack at all is that otherwise we would have to change | |
| 6108 `anychar's code to do something besides goto fail in this | |
| 6109 case; that seems worse than this. */ | |
| 6110 case on_failure_keep_string_jump: | |
| 5041 | 6111 DEBUG_MATCH_PRINT1 ("EXECUTING on_failure_keep_string_jump"); |
| 428 | 6112 |
| 6113 EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
| 5041 | 6114 DEBUG_MATCH_PRINT3 (" %d (to 0x%lx):\n", mcnt, (long) (p + mcnt)); |
| 428 | 6115 |
| 446 | 6116 PUSH_FAILURE_POINT (p + mcnt, (unsigned char *) 0, -2); |
| 428 | 6117 break; |
| 6118 | |
| 6119 | |
| 6120 /* Uses of on_failure_jump: | |
| 6121 | |
| 6122 Each alternative starts with an on_failure_jump that points | |
| 6123 to the beginning of the next alternative. Each alternative | |
| 6124 except the last ends with a jump that in effect jumps past | |
| 6125 the rest of the alternatives. (They really jump to the | |
| 6126 ending jump of the following alternative, because tensioning | |
| 6127 these jumps is a hassle.) | |
| 6128 | |
| 6129 Repeats start with an on_failure_jump that points past both | |
| 6130 the repetition text and either the following jump or | |
| 6131 pop_failure_jump back to this on_failure_jump. */ | |
| 6132 case on_failure_jump: | |
| 6133 on_failure: | |
| 5041 | 6134 DEBUG_MATCH_PRINT1 ("EXECUTING on_failure_jump"); |
| 428 | 6135 |
| 6136 EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
| 5041 | 6137 DEBUG_MATCH_PRINT3 (" %d (to 0x%lx)", mcnt, (long) (p + mcnt)); |
| 428 | 6138 |
| 6139 /* If this on_failure_jump comes right before a group (i.e., | |
| 6140 the original * applied to a group), save the information | |
| 6141 for that group and all inner ones, so that if we fail back | |
| 6142 to this point, the group's information will be correct. | |
| 6143 For example, in \(a*\)*\1, we need the preceding group, | |
| 6144 and in \(\(a*\)b*\)\2, we need the inner group. */ | |
| 6145 | |
| 6146 /* We can't use `p' to check ahead because we push | |
| 6147 a failure point to `p + mcnt' after we do this. */ | |
| 6148 p1 = p; | |
| 6149 | |
| 6150 /* We need to skip no_op's before we look for the | |
| 6151 start_memory in case this on_failure_jump is happening as | |
| 6152 the result of a completed succeed_n, as in \(a\)\{1,3\}b\1 | |
| 6153 against aba. */ | |
| 6154 while (p1 < pend && (re_opcode_t) *p1 == no_op) | |
| 6155 p1++; | |
| 6156 | |
| 6157 if (p1 < pend && (re_opcode_t) *p1 == start_memory) | |
| 6158 { | |
| 6159 /* We have a new highest active register now. This will | |
| 6160 get reset at the start_memory we are about to get to, | |
| 6161 but we will have saved all the registers relevant to | |
| 6162 this repetition op, as described above. */ | |
| 6163 highest_active_reg = *(p1 + 1) + *(p1 + 2); | |
| 6164 if (lowest_active_reg == NO_LOWEST_ACTIVE_REG) | |
| 6165 lowest_active_reg = *(p1 + 1); | |
| 6166 } | |
| 6167 | |
| 5041 | 6168 DEBUG_MATCH_PRINT1 (":\n"); |
| 428 | 6169 PUSH_FAILURE_POINT (p + mcnt, d, -2); |
| 6170 break; | |
| 6171 | |
| 6172 | |
| 6173 /* A smart repeat ends with `maybe_pop_jump'. | |
| 6174 We change it to either `pop_failure_jump' or `jump'. */ | |
| 6175 case maybe_pop_jump: | |
| 6176 EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
| 5041 | 6177 DEBUG_MATCH_PRINT2 ("EXECUTING maybe_pop_jump %d.\n", mcnt); |
| 428 | 6178 { |
| 6179 REGISTER unsigned char *p2 = p; | |
| 6180 | |
| 6181 /* Compare the beginning of the repeat with what in the | |
| 6182 pattern follows its end. If we can establish that there | |
| 6183 is nothing that they would both match, i.e., that we | |
| 6184 would have to backtrack because of (as in, e.g., `a*a') | |
| 6185 then we can change to pop_failure_jump, because we'll | |
| 6186 never have to backtrack. | |
| 6187 | |
| 6188 This is not true in the case of alternatives: in | |
| 6189 `(a|ab)*' we do need to backtrack to the `ab' alternative | |
| 6190 (e.g., if the string was `ab'). But instead of trying to | |
| 6191 detect that here, the alternative has put on a dummy | |
| 6192 failure point which is what we will end up popping. */ | |
| 6193 | |
| 6194 /* Skip over open/close-group commands. | |
| 6195 If what follows this loop is a ...+ construct, | |
| 6196 look at what begins its body, since we will have to | |
| 6197 match at least one of that. */ | |
| 6198 while (1) | |
| 6199 { | |
| 6200 if (p2 + 2 < pend | |
| 6201 && ((re_opcode_t) *p2 == stop_memory | |
| 6202 || (re_opcode_t) *p2 == start_memory)) | |
| 6203 p2 += 3; | |
| 6204 else if (p2 + 6 < pend | |
| 6205 && (re_opcode_t) *p2 == dummy_failure_jump) | |
| 6206 p2 += 6; | |
| 6207 else | |
| 6208 break; | |
| 6209 } | |
| 6210 | |
| 6211 p1 = p + mcnt; | |
| 6212 /* p1[0] ... p1[2] are the `on_failure_jump' corresponding | |
| 6213 to the `maybe_finalize_jump' of this case. Examine what | |
| 6214 follows. */ | |
| 6215 | |
| 6216 /* If we're at the end of the pattern, we can change. */ | |
| 6217 if (p2 == pend) | |
| 6218 { | |
| 6219 /* Consider what happens when matching ":\(.*\)" | |
| 6220 against ":/". I don't really understand this code | |
| 6221 yet. */ | |
| 6222 p[-3] = (unsigned char) pop_failure_jump; | |
| 5041 | 6223 DEBUG_MATCH_PRINT1 |
| 428 | 6224 (" End of pattern: change to `pop_failure_jump'.\n"); |
| 6225 } | |
| 6226 | |
| 6227 else if ((re_opcode_t) *p2 == exactn | |
| 6228 || (bufp->newline_anchor && (re_opcode_t) *p2 == endline)) | |
| 6229 { | |
| 6230 REGISTER unsigned char c | |
| 6231 = *p2 == (unsigned char) endline ? '\n' : p2[2]; | |
| 6232 | |
| 6233 if ((re_opcode_t) p1[3] == exactn && p1[5] != c) | |
| 6234 { | |
| 6235 p[-3] = (unsigned char) pop_failure_jump; | |
| 5041 | 6236 DEBUG_MATCH_PRINT3 (" %c != %c => pop_failure_jump.\n", |
| 428 | 6237 c, p1[5]); |
| 6238 } | |
| 6239 | |
| 6240 else if ((re_opcode_t) p1[3] == charset | |
| 6241 || (re_opcode_t) p1[3] == charset_not) | |
| 6242 { | |
| 458 | 6243 int not_p = (re_opcode_t) p1[3] == charset_not; |
| 428 | 6244 |
| 6245 if (c < (unsigned char) (p1[4] * BYTEWIDTH) | |
| 6246 && p1[5 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | |
| 458 | 6247 not_p = !not_p; |
| 6248 | |
| 6249 /* `not_p' is equal to 1 if c would match, which means | |
| 428 | 6250 that we can't change to pop_failure_jump. */ |
| 458 | 6251 if (!not_p) |
| 428 | 6252 { |
| 6253 p[-3] = (unsigned char) pop_failure_jump; | |
| 5041 | 6254 DEBUG_MATCH_PRINT1 (" No match => pop_failure_jump.\n"); |
| 428 | 6255 } |
| 6256 } | |
| 6257 } | |
| 6258 else if ((re_opcode_t) *p2 == charset) | |
| 6259 { | |
| 6260 #ifdef DEBUG | |
| 6261 REGISTER unsigned char c | |
| 6262 = *p2 == (unsigned char) endline ? '\n' : p2[2]; | |
| 6263 #endif | |
| 6264 | |
| 6265 if ((re_opcode_t) p1[3] == exactn | |
| 6266 && ! ((int) p2[1] * BYTEWIDTH > (int) p1[5] | |
| 6267 && (p2[2 + p1[5] / BYTEWIDTH] | |
| 6268 & (1 << (p1[5] % BYTEWIDTH))))) | |
| 6269 { | |
| 6270 p[-3] = (unsigned char) pop_failure_jump; | |
| 5041 | 6271 DEBUG_MATCH_PRINT3 (" %c != %c => pop_failure_jump.\n", |
| 428 | 6272 c, p1[5]); |
| 6273 } | |
| 6274 | |
| 6275 else if ((re_opcode_t) p1[3] == charset_not) | |
| 6276 { | |
| 6277 int idx; | |
| 6278 /* We win if the charset_not inside the loop | |
| 6279 lists every character listed in the charset after. */ | |
| 6280 for (idx = 0; idx < (int) p2[1]; idx++) | |
| 6281 if (! (p2[2 + idx] == 0 | |
| 6282 || (idx < (int) p1[4] | |
| 6283 && ((p2[2 + idx] & ~ p1[5 + idx]) == 0)))) | |
| 6284 break; | |
| 6285 | |
| 6286 if (idx == p2[1]) | |
| 6287 { | |
| 6288 p[-3] = (unsigned char) pop_failure_jump; | |
| 5041 | 6289 DEBUG_MATCH_PRINT1 (" No match => pop_failure_jump.\n"); |
| 428 | 6290 } |
| 6291 } | |
| 6292 else if ((re_opcode_t) p1[3] == charset) | |
| 6293 { | |
| 6294 int idx; | |
| 6295 /* We win if the charset inside the loop | |
| 6296 has no overlap with the one after the loop. */ | |
| 6297 for (idx = 0; | |
| 6298 idx < (int) p2[1] && idx < (int) p1[4]; | |
| 6299 idx++) | |
| 6300 if ((p2[2 + idx] & p1[5 + idx]) != 0) | |
| 6301 break; | |
| 6302 | |
| 6303 if (idx == p2[1] || idx == p1[4]) | |
| 6304 { | |
| 6305 p[-3] = (unsigned char) pop_failure_jump; | |
| 5041 | 6306 DEBUG_MATCH_PRINT1 (" No match => pop_failure_jump.\n"); |
| 428 | 6307 } |
| 6308 } | |
| 6309 } | |
| 6310 } | |
| 6311 p -= 2; /* Point at relative address again. */ | |
| 6312 if ((re_opcode_t) p[-1] != pop_failure_jump) | |
| 6313 { | |
| 6314 p[-1] = (unsigned char) jump; | |
| 5041 | 6315 DEBUG_MATCH_PRINT1 (" Match => jump.\n"); |
| 428 | 6316 goto unconditional_jump; |
| 6317 } | |
| 6318 /* Note fall through. */ | |
| 6319 | |
| 6320 | |
| 6321 /* The end of a simple repeat has a pop_failure_jump back to | |
| 6322 its matching on_failure_jump, where the latter will push a | |
| 6323 failure point. The pop_failure_jump takes off failure | |
| 6324 points put on by this pop_failure_jump's matching | |
| 6325 on_failure_jump; we got through the pattern to here from the | |
| 6326 matching on_failure_jump, so didn't fail. */ | |
| 6327 case pop_failure_jump: | |
| 6328 { | |
| 6329 /* We need to pass separate storage for the lowest and | |
| 6330 highest registers, even though we don't care about the | |
| 6331 actual values. Otherwise, we will restore only one | |
| 6332 register from the stack, since lowest will == highest in | |
| 6333 `pop_failure_point'. */ | |
| 647 | 6334 int dummy_low_reg, dummy_high_reg; |
| 428 | 6335 unsigned char *pdummy; |
| 446 | 6336 re_char *sdummy = NULL; |
| 428 | 6337 |
| 5041 | 6338 DEBUG_MATCH_PRINT1 ("EXECUTING pop_failure_jump.\n"); |
| 428 | 6339 POP_FAILURE_POINT (sdummy, pdummy, |
| 6340 dummy_low_reg, dummy_high_reg, | |
| 6341 reg_dummy, reg_dummy, reg_info_dummy); | |
| 6342 } | |
| 6343 /* Note fall through. */ | |
| 6344 | |
| 6345 | |
| 6346 /* Unconditionally jump (without popping any failure points). */ | |
| 6347 case jump: | |
| 6348 unconditional_jump: | |
| 6349 EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ | |
| 5041 | 6350 DEBUG_MATCH_PRINT2 ("EXECUTING jump %d ", mcnt); |
| 428 | 6351 p += mcnt; /* Do the jump. */ |
| 5041 | 6352 DEBUG_MATCH_PRINT2 ("(to 0x%lx).\n", (long) p); |
| 428 | 6353 break; |
| 6354 | |
| 6355 | |
| 6356 /* We need this opcode so we can detect where alternatives end | |
| 6357 in `group_match_null_string_p' et al. */ | |
| 6358 case jump_past_alt: | |
| 5041 | 6359 DEBUG_MATCH_PRINT1 ("EXECUTING jump_past_alt.\n"); |
| 428 | 6360 goto unconditional_jump; |
| 6361 | |
| 6362 | |
| 6363 /* Normally, the on_failure_jump pushes a failure point, which | |
| 6364 then gets popped at pop_failure_jump. We will end up at | |
| 6365 pop_failure_jump, also, and with a pattern of, say, `a+', we | |
| 6366 are skipping over the on_failure_jump, so we have to push | |
| 6367 something meaningless for pop_failure_jump to pop. */ | |
| 6368 case dummy_failure_jump: | |
| 5041 | 6369 DEBUG_MATCH_PRINT1 ("EXECUTING dummy_failure_jump.\n"); |
| 428 | 6370 /* It doesn't matter what we push for the string here. What |
| 6371 the code at `fail' tests is the value for the pattern. */ | |
| 446 | 6372 PUSH_FAILURE_POINT ((unsigned char *) 0, (unsigned char *) 0, -2); |
| 428 | 6373 goto unconditional_jump; |
| 6374 | |
| 6375 | |
| 6376 /* At the end of an alternative, we need to push a dummy failure | |
| 6377 point in case we are followed by a `pop_failure_jump', because | |
| 6378 we don't want the failure point for the alternative to be | |
| 6379 popped. For example, matching `(a|ab)*' against `aab' | |
| 6380 requires that we match the `ab' alternative. */ | |
| 6381 case push_dummy_failure: | |
| 5041 | 6382 DEBUG_MATCH_PRINT1 ("EXECUTING push_dummy_failure.\n"); |
| 428 | 6383 /* See comments just above at `dummy_failure_jump' about the |
| 6384 two zeroes. */ | |
| 446 | 6385 PUSH_FAILURE_POINT ((unsigned char *) 0, (unsigned char *) 0, -2); |
| 428 | 6386 break; |
| 6387 | |
| 6388 /* Have to succeed matching what follows at least n times. | |
| 6389 After that, handle like `on_failure_jump'. */ | |
| 6390 case succeed_n: | |
| 6391 EXTRACT_NUMBER (mcnt, p + 2); | |
| 5041 | 6392 DEBUG_MATCH_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt); |
| 428 | 6393 |
| 6394 assert (mcnt >= 0); | |
| 6395 /* Originally, this is how many times we HAVE to succeed. */ | |
| 6396 if (mcnt > 0) | |
| 6397 { | |
| 6398 mcnt--; | |
| 6399 p += 2; | |
| 6400 STORE_NUMBER_AND_INCR (p, mcnt); | |
| 5041 | 6401 DEBUG_MATCH_PRINT3 (" Setting 0x%lx to %d.\n", (long) p, mcnt); |
| 428 | 6402 } |
| 6403 else if (mcnt == 0) | |
| 6404 { | |
| 5041 | 6405 DEBUG_MATCH_PRINT2 (" Setting two bytes from 0x%lx to no_op.\n", |
| 428 | 6406 (long) (p+2)); |
| 6407 p[2] = (unsigned char) no_op; | |
| 6408 p[3] = (unsigned char) no_op; | |
| 6409 goto on_failure; | |
| 6410 } | |
| 6411 break; | |
| 6412 | |
| 6413 case jump_n: | |
| 6414 EXTRACT_NUMBER (mcnt, p + 2); | |
| 5041 | 6415 DEBUG_MATCH_PRINT2 ("EXECUTING jump_n %d.\n", mcnt); |
| 428 | 6416 |
| 6417 /* Originally, this is how many times we CAN jump. */ | |
| 6418 if (mcnt) | |
| 6419 { | |
| 6420 mcnt--; | |
| 6421 STORE_NUMBER (p + 2, mcnt); | |
| 6422 goto unconditional_jump; | |
| 6423 } | |
| 6424 /* If don't have to jump any more, skip over the rest of command. */ | |
| 6425 else | |
| 6426 p += 4; | |
| 6427 break; | |
| 6428 | |
| 6429 case set_number_at: | |
| 6430 { | |
| 5041 | 6431 DEBUG_MATCH_PRINT1 ("EXECUTING set_number_at.\n"); |
| 428 | 6432 |
| 6433 EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
| 6434 p1 = p + mcnt; | |
| 6435 EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
| 5041 | 6436 DEBUG_MATCH_PRINT3 (" Setting 0x%lx to %d.\n", (long) p1, mcnt); |
| 428 | 6437 STORE_NUMBER (p1, mcnt); |
| 6438 break; | |
| 6439 } | |
| 6440 | |
| 6441 case wordbound: | |
| 5041 | 6442 DEBUG_MATCH_PRINT1 ("EXECUTING wordbound.\n"); |
| 428 | 6443 should_succeed = 1; |
| 6444 matchwordbound: | |
| 6445 { | |
| 6446 /* XEmacs change */ | |
| 1377 | 6447 /* Straightforward and (I hope) correct implementation. |
| 6448 Probably should be optimized by arranging to compute | |
| 1497 | 6449 charpos only once. */ |
| 1377 | 6450 /* emch1 is the character before d, syn1 is the syntax of |
| 6451 emch1, emch2 is the character at d, and syn2 is the | |
| 6452 syntax of emch2. */ | |
| 6453 Ichar emch1, emch2; | |
| 1468 | 6454 int syn1 = 0, |
| 6455 syn2 = 0; | |
| 1377 | 6456 re_char *d_before, *d_after; |
| 6457 int result, | |
| 6458 at_beg = AT_STRINGS_BEG (d), | |
| 6459 at_end = AT_STRINGS_END (d); | |
| 6460 #ifdef emacs | |
| 1497 | 6461 Charxpos charpos; |
| 1377 | 6462 #endif |
| 6463 | |
| 6464 if (at_beg && at_end) | |
| 6465 { | |
| 6466 result = 0; | |
| 6467 } | |
| 428 | 6468 else |
| 6469 { | |
| 1377 | 6470 if (!at_beg) |
| 6471 { | |
| 6472 d_before = POS_BEFORE_GAP_UNSAFE (d); | |
| 6473 DEC_IBYTEPTR_FMT (d_before, fmt); | |
| 6474 emch1 = itext_ichar_fmt (d_before, fmt, lispobj); | |
| 460 | 6475 #ifdef emacs |
| 1497 | 6476 charpos = offset_to_charxpos (lispobj, |
| 6477 PTR_TO_OFFSET (d)) - 1; | |
| 1377 | 6478 BEGIN_REGEX_MALLOC_OK (); |
| 1497 | 6479 UPDATE_SYNTAX_CACHE (scache, charpos); |
| 460 | 6480 #endif |
| 1377 | 6481 syn1 = SYNTAX_FROM_CACHE (scache, emch1); |
| 6482 END_REGEX_MALLOC_OK (); | |
| 6483 } | |
| 6484 if (!at_end) | |
| 6485 { | |
| 6486 d_after = POS_AFTER_GAP_UNSAFE (d); | |
| 6487 emch2 = itext_ichar_fmt (d_after, fmt, lispobj); | |
| 460 | 6488 #ifdef emacs |
| 1497 | 6489 charpos = offset_to_charxpos (lispobj, PTR_TO_OFFSET (d)); |
| 1377 | 6490 BEGIN_REGEX_MALLOC_OK (); |
| 1497 | 6491 UPDATE_SYNTAX_CACHE_FORWARD (scache, charpos); |
| 460 | 6492 #endif |
| 1377 | 6493 syn2 = SYNTAX_FROM_CACHE (scache, emch2); |
| 6494 END_REGEX_MALLOC_OK (); | |
| 6495 } | |
| 1333 | 6496 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); |
| 1377 | 6497 |
| 6498 if (at_beg) | |
| 6499 result = (syn2 == Sword); | |
| 6500 else if (at_end) | |
| 6501 result = (syn1 == Sword); | |
| 6502 else | |
| 6503 result = ((syn1 == Sword) != (syn2 == Sword)); | |
| 428 | 6504 } |
| 1377 | 6505 |
| 428 | 6506 if (result == should_succeed) |
| 6507 break; | |
| 6508 goto fail; | |
| 6509 } | |
| 6510 | |
| 6511 case notwordbound: | |
| 5041 | 6512 DEBUG_MATCH_PRINT1 ("EXECUTING notwordbound.\n"); |
| 428 | 6513 should_succeed = 0; |
| 6514 goto matchwordbound; | |
| 6515 | |
| 6516 case wordbeg: | |
| 5041 | 6517 DEBUG_MATCH_PRINT1 ("EXECUTING wordbeg.\n"); |
| 460 | 6518 if (AT_STRINGS_END (d)) |
| 6519 goto fail; | |
| 428 | 6520 { |
| 6521 /* XEmacs: this originally read: | |
| 6522 | |
| 6523 if (WORDCHAR_P (d) && (AT_STRINGS_BEG (d) || !WORDCHAR_P (d - 1))) | |
| 6524 break; | |
| 6525 | |
| 6526 */ | |
| 460 | 6527 re_char *dtmp = POS_AFTER_GAP_UNSAFE (d); |
| 867 | 6528 Ichar emch = itext_ichar_fmt (dtmp, fmt, lispobj); |
| 1333 | 6529 int tempres; |
| 1347 | 6530 #ifdef emacs |
| 6531 Charxpos charpos = offset_to_charxpos (lispobj, PTR_TO_OFFSET (d)); | |
| 6532 #endif | |
| 1333 | 6533 BEGIN_REGEX_MALLOC_OK (); |
| 460 | 6534 #ifdef emacs |
| 826 | 6535 UPDATE_SYNTAX_CACHE (scache, charpos); |
| 460 | 6536 #endif |
| 1333 | 6537 tempres = (SYNTAX_FROM_CACHE (scache, emch) != Sword); |
| 6538 END_REGEX_MALLOC_OK (); | |
| 6539 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 6540 if (tempres) | |
| 428 | 6541 goto fail; |
| 6542 if (AT_STRINGS_BEG (d)) | |
| 6543 break; | |
| 460 | 6544 dtmp = POS_BEFORE_GAP_UNSAFE (d); |
| 867 | 6545 DEC_IBYTEPTR_FMT (dtmp, fmt); |
| 6546 emch = itext_ichar_fmt (dtmp, fmt, lispobj); | |
| 1333 | 6547 BEGIN_REGEX_MALLOC_OK (); |
| 460 | 6548 #ifdef emacs |
| 826 | 6549 UPDATE_SYNTAX_CACHE_BACKWARD (scache, charpos - 1); |
| 460 | 6550 #endif |
| 1333 | 6551 tempres = (SYNTAX_FROM_CACHE (scache, emch) != Sword); |
| 6552 END_REGEX_MALLOC_OK (); | |
| 6553 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 6554 if (tempres) | |
| 428 | 6555 break; |
| 6556 goto fail; | |
| 6557 } | |
| 6558 | |
| 6559 case wordend: | |
| 5041 | 6560 DEBUG_MATCH_PRINT1 ("EXECUTING wordend.\n"); |
| 460 | 6561 if (AT_STRINGS_BEG (d)) |
| 6562 goto fail; | |
| 428 | 6563 { |
| 6564 /* XEmacs: this originally read: | |
| 6565 | |
| 6566 if (!AT_STRINGS_BEG (d) && WORDCHAR_P (d - 1) | |
| 6567 && (!WORDCHAR_P (d) || AT_STRINGS_END (d))) | |
| 6568 break; | |
| 6569 | |
| 6570 The or condition is incorrect (reversed). | |
| 6571 */ | |
| 460 | 6572 re_char *dtmp; |
| 867 | 6573 Ichar emch; |
| 1333 | 6574 int tempres; |
| 460 | 6575 #ifdef emacs |
| 826 | 6576 Charxpos charpos = offset_to_charxpos (lispobj, PTR_TO_OFFSET (d)); |
| 1347 | 6577 BEGIN_REGEX_MALLOC_OK (); |
| 826 | 6578 UPDATE_SYNTAX_CACHE (scache, charpos); |
| 1333 | 6579 END_REGEX_MALLOC_OK (); |
| 6580 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 1347 | 6581 #endif |
| 460 | 6582 dtmp = POS_BEFORE_GAP_UNSAFE (d); |
| 867 | 6583 DEC_IBYTEPTR_FMT (dtmp, fmt); |
| 6584 emch = itext_ichar_fmt (dtmp, fmt, lispobj); | |
| 1333 | 6585 BEGIN_REGEX_MALLOC_OK (); |
| 6586 tempres = (SYNTAX_FROM_CACHE (scache, emch) != Sword); | |
| 6587 END_REGEX_MALLOC_OK (); | |
| 6588 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 6589 if (tempres) | |
| 428 | 6590 goto fail; |
| 6591 if (AT_STRINGS_END (d)) | |
| 6592 break; | |
| 460 | 6593 dtmp = POS_AFTER_GAP_UNSAFE (d); |
| 867 | 6594 emch = itext_ichar_fmt (dtmp, fmt, lispobj); |
| 1333 | 6595 BEGIN_REGEX_MALLOC_OK (); |
| 460 | 6596 #ifdef emacs |
| 826 | 6597 UPDATE_SYNTAX_CACHE_FORWARD (scache, charpos + 1); |
| 460 | 6598 #endif |
| 1333 | 6599 tempres = (SYNTAX_FROM_CACHE (scache, emch) != Sword); |
| 6600 END_REGEX_MALLOC_OK (); | |
| 6601 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 6602 if (tempres) | |
| 428 | 6603 break; |
| 6604 goto fail; | |
| 6605 } | |
| 6606 | |
| 6607 #ifdef emacs | |
| 6608 case before_dot: | |
| 5041 | 6609 DEBUG_MATCH_PRINT1 ("EXECUTING before_dot.\n"); |
| 826 | 6610 if (!BUFFERP (lispobj) |
| 6611 || (BUF_PTR_BYTE_POS (XBUFFER (lispobj), (unsigned char *) d) | |
| 6612 >= BUF_PT (XBUFFER (lispobj)))) | |
| 428 | 6613 goto fail; |
| 6614 break; | |
| 6615 | |
| 6616 case at_dot: | |
| 5041 | 6617 DEBUG_MATCH_PRINT1 ("EXECUTING at_dot.\n"); |
| 826 | 6618 if (!BUFFERP (lispobj) |
| 6619 || (BUF_PTR_BYTE_POS (XBUFFER (lispobj), (unsigned char *) d) | |
| 6620 != BUF_PT (XBUFFER (lispobj)))) | |
| 428 | 6621 goto fail; |
| 6622 break; | |
| 6623 | |
| 6624 case after_dot: | |
| 5041 | 6625 DEBUG_MATCH_PRINT1 ("EXECUTING after_dot.\n"); |
| 826 | 6626 if (!BUFFERP (lispobj) |
| 6627 || (BUF_PTR_BYTE_POS (XBUFFER (lispobj), (unsigned char *) d) | |
| 6628 <= BUF_PT (XBUFFER (lispobj)))) | |
| 428 | 6629 goto fail; |
| 6630 break; | |
| 6631 | |
| 6632 case syntaxspec: | |
| 5041 | 6633 DEBUG_MATCH_PRINT2 ("EXECUTING syntaxspec %d.\n", mcnt); |
| 428 | 6634 mcnt = *p++; |
| 6635 goto matchsyntax; | |
| 6636 | |
| 6637 case wordchar: | |
| 5041 | 6638 DEBUG_MATCH_PRINT1 ("EXECUTING Emacs wordchar.\n"); |
| 428 | 6639 mcnt = (int) Sword; |
| 6640 matchsyntax: | |
| 6641 should_succeed = 1; | |
| 6642 matchornotsyntax: | |
| 6643 { | |
| 6644 int matches; | |
| 867 | 6645 Ichar emch; |
| 428 | 6646 |
| 450 | 6647 REGEX_PREFETCH (); |
| 1333 | 6648 BEGIN_REGEX_MALLOC_OK (); |
| 826 | 6649 UPDATE_SYNTAX_CACHE |
| 6650 (scache, offset_to_charxpos (lispobj, PTR_TO_OFFSET (d))); | |
| 1333 | 6651 END_REGEX_MALLOC_OK (); |
| 6652 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 826 | 6653 |
| 867 | 6654 emch = itext_ichar_fmt (d, fmt, lispobj); |
| 1333 | 6655 BEGIN_REGEX_MALLOC_OK (); |
| 826 | 6656 matches = (SYNTAX_FROM_CACHE (scache, emch) == |
| 6657 (enum syntaxcode) mcnt); | |
| 1333 | 6658 END_REGEX_MALLOC_OK (); |
| 6659 RE_MATCH_RELOCATE_MOVEABLE_DATA_POINTERS (); | |
| 867 | 6660 INC_IBYTEPTR_FMT (d, fmt); |
| 428 | 6661 if (matches != should_succeed) |
| 6662 goto fail; | |
| 6663 SET_REGS_MATCHED (); | |
| 6664 } | |
| 6665 break; | |
| 6666 | |
| 6667 case notsyntaxspec: | |
| 5041 | 6668 DEBUG_MATCH_PRINT2 ("EXECUTING notsyntaxspec %d.\n", mcnt); |
| 428 | 6669 mcnt = *p++; |
| 6670 goto matchnotsyntax; | |
| 6671 | |
| 6672 case notwordchar: | |
| 5041 | 6673 DEBUG_MATCH_PRINT1 ("EXECUTING Emacs notwordchar.\n"); |
| 428 | 6674 mcnt = (int) Sword; |
| 6675 matchnotsyntax: | |
| 6676 should_succeed = 0; | |
| 6677 goto matchornotsyntax; | |
| 6678 | |
| 6679 #ifdef MULE | |
| 6680 /* 97/2/17 jhod Mule category code patch */ | |
| 6681 case categoryspec: | |
| 6682 should_succeed = 1; | |
| 6683 matchornotcategory: | |
| 6684 { | |
| 867 | 6685 Ichar emch; |
| 428 | 6686 |
| 6687 mcnt = *p++; | |
| 450 | 6688 REGEX_PREFETCH (); |
| 867 | 6689 emch = itext_ichar_fmt (d, fmt, lispobj); |
| 6690 INC_IBYTEPTR_FMT (d, fmt); | |
| 826 | 6691 if (check_category_char (emch, BUFFER_CATEGORY_TABLE (lispbuf), |
| 6692 mcnt, should_succeed)) | |
| 428 | 6693 goto fail; |
| 6694 SET_REGS_MATCHED (); | |
| 6695 } | |
| 6696 break; | |
| 6697 | |
| 6698 case notcategoryspec: | |
| 6699 should_succeed = 0; | |
| 6700 goto matchornotcategory; | |
| 6701 /* end of category patch */ | |
| 6702 #endif /* MULE */ | |
| 6703 #else /* not emacs */ | |
| 6704 case wordchar: | |
| 5041 | 6705 DEBUG_MATCH_PRINT1 ("EXECUTING non-Emacs wordchar.\n"); |
| 450 | 6706 REGEX_PREFETCH (); |
| 826 | 6707 if (!WORDCHAR_P ((int) (*d))) |
| 428 | 6708 goto fail; |
| 6709 SET_REGS_MATCHED (); | |
| 6710 d++; | |
| 6711 break; | |
| 6712 | |
| 6713 case notwordchar: | |
| 5041 | 6714 DEBUG_MATCH_PRINT1 ("EXECUTING non-Emacs notwordchar.\n"); |
| 450 | 6715 REGEX_PREFETCH (); |
| 826 | 6716 if (!WORDCHAR_P ((int) (*d))) |
| 428 | 6717 goto fail; |
| 6718 SET_REGS_MATCHED (); | |
| 6719 d++; | |
| 6720 break; | |
| 446 | 6721 #endif /* emacs */ |
| 428 | 6722 |
| 6723 default: | |
| 2500 | 6724 ABORT (); |
| 428 | 6725 } |
| 6726 continue; /* Successfully executed one pattern command; keep going. */ | |
| 6727 | |
| 6728 | |
| 6729 /* We goto here if a matching operation fails. */ | |
| 6730 fail: | |
| 6731 if (!FAIL_STACK_EMPTY ()) | |
| 6732 { /* A restart point is known. Restore to that state. */ | |
| 5041 | 6733 DEBUG_MATCH_PRINT1 ("\nFAIL:\n"); |
| 428 | 6734 POP_FAILURE_POINT (d, p, |
| 6735 lowest_active_reg, highest_active_reg, | |
| 6736 regstart, regend, reg_info); | |
| 6737 | |
| 6738 /* If this failure point is a dummy, try the next one. */ | |
| 6739 if (!p) | |
| 6740 goto fail; | |
| 6741 | |
| 6742 /* If we failed to the end of the pattern, don't examine *p. */ | |
| 6743 assert (p <= pend); | |
| 6744 if (p < pend) | |
| 6745 { | |
| 460 | 6746 re_bool is_a_jump_n = false; |
| 428 | 6747 |
| 6748 /* If failed to a backwards jump that's part of a repetition | |
| 6749 loop, need to pop this failure point and use the next one. */ | |
| 6750 switch ((re_opcode_t) *p) | |
| 6751 { | |
| 6752 case jump_n: | |
| 6753 is_a_jump_n = true; | |
| 6754 case maybe_pop_jump: | |
| 6755 case pop_failure_jump: | |
| 6756 case jump: | |
| 6757 p1 = p + 1; | |
| 6758 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 6759 p1 += mcnt; | |
| 6760 | |
| 6761 if ((is_a_jump_n && (re_opcode_t) *p1 == succeed_n) | |
| 6762 || (!is_a_jump_n | |
| 6763 && (re_opcode_t) *p1 == on_failure_jump)) | |
| 6764 goto fail; | |
| 6765 break; | |
| 6766 default: | |
| 6767 /* do nothing */ ; | |
| 6768 } | |
| 6769 } | |
| 6770 | |
| 6771 if (d >= string1 && d <= end1) | |
| 6772 dend = end_match_1; | |
| 6773 } | |
| 6774 else | |
| 6775 break; /* Matching at this starting point really fails. */ | |
| 6776 } /* for (;;) */ | |
| 6777 | |
| 6778 if (best_regs_set) | |
| 6779 goto restore_best_regs; | |
| 6780 | |
| 6781 FREE_VARIABLES (); | |
| 6782 | |
| 6783 return -1; /* Failure to match. */ | |
| 1333 | 6784 } /* re_match_2_internal */ |
| 428 | 6785 |
| 6786 /* Subroutine definitions for re_match_2. */ | |
| 6787 | |
| 6788 | |
| 6789 /* We are passed P pointing to a register number after a start_memory. | |
| 6790 | |
| 6791 Return true if the pattern up to the corresponding stop_memory can | |
| 6792 match the empty string, and false otherwise. | |
| 6793 | |
| 6794 If we find the matching stop_memory, sets P to point to one past its number. | |
| 6795 Otherwise, sets P to an undefined byte less than or equal to END. | |
| 6796 | |
| 6797 We don't handle duplicates properly (yet). */ | |
| 6798 | |
| 460 | 6799 static re_bool |
| 428 | 6800 group_match_null_string_p (unsigned char **p, unsigned char *end, |
| 6801 register_info_type *reg_info) | |
| 6802 { | |
| 6803 int mcnt; | |
| 6804 /* Point to after the args to the start_memory. */ | |
| 6805 unsigned char *p1 = *p + 2; | |
| 6806 | |
| 6807 while (p1 < end) | |
| 6808 { | |
| 6809 /* Skip over opcodes that can match nothing, and return true or | |
| 6810 false, as appropriate, when we get to one that can't, or to the | |
| 6811 matching stop_memory. */ | |
| 6812 | |
| 6813 switch ((re_opcode_t) *p1) | |
| 6814 { | |
| 6815 /* Could be either a loop or a series of alternatives. */ | |
| 6816 case on_failure_jump: | |
| 6817 p1++; | |
| 6818 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 6819 | |
| 6820 /* If the next operation is not a jump backwards in the | |
| 6821 pattern. */ | |
| 6822 | |
| 6823 if (mcnt >= 0) | |
| 6824 { | |
| 6825 /* Go through the on_failure_jumps of the alternatives, | |
| 6826 seeing if any of the alternatives cannot match nothing. | |
| 6827 The last alternative starts with only a jump, | |
| 6828 whereas the rest start with on_failure_jump and end | |
| 6829 with a jump, e.g., here is the pattern for `a|b|c': | |
| 6830 | |
| 6831 /on_failure_jump/0/6/exactn/1/a/jump_past_alt/0/6 | |
| 6832 /on_failure_jump/0/6/exactn/1/b/jump_past_alt/0/3 | |
| 6833 /exactn/1/c | |
| 6834 | |
| 6835 So, we have to first go through the first (n-1) | |
| 6836 alternatives and then deal with the last one separately. */ | |
| 6837 | |
| 6838 | |
| 6839 /* Deal with the first (n-1) alternatives, which start | |
| 6840 with an on_failure_jump (see above) that jumps to right | |
| 6841 past a jump_past_alt. */ | |
| 6842 | |
| 6843 while ((re_opcode_t) p1[mcnt-3] == jump_past_alt) | |
| 6844 { | |
| 6845 /* `mcnt' holds how many bytes long the alternative | |
| 6846 is, including the ending `jump_past_alt' and | |
| 6847 its number. */ | |
| 6848 | |
| 6849 if (!alt_match_null_string_p (p1, p1 + mcnt - 3, | |
| 6850 reg_info)) | |
| 6851 return false; | |
| 6852 | |
| 6853 /* Move to right after this alternative, including the | |
| 6854 jump_past_alt. */ | |
| 6855 p1 += mcnt; | |
| 6856 | |
| 6857 /* Break if it's the beginning of an n-th alternative | |
| 6858 that doesn't begin with an on_failure_jump. */ | |
| 6859 if ((re_opcode_t) *p1 != on_failure_jump) | |
| 6860 break; | |
| 6861 | |
| 6862 /* Still have to check that it's not an n-th | |
| 6863 alternative that starts with an on_failure_jump. */ | |
| 6864 p1++; | |
| 6865 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 6866 if ((re_opcode_t) p1[mcnt-3] != jump_past_alt) | |
| 6867 { | |
| 6868 /* Get to the beginning of the n-th alternative. */ | |
| 6869 p1 -= 3; | |
| 6870 break; | |
| 6871 } | |
| 6872 } | |
| 6873 | |
| 6874 /* Deal with the last alternative: go back and get number | |
| 6875 of the `jump_past_alt' just before it. `mcnt' contains | |
| 6876 the length of the alternative. */ | |
| 6877 EXTRACT_NUMBER (mcnt, p1 - 2); | |
| 6878 | |
| 6879 if (!alt_match_null_string_p (p1, p1 + mcnt, reg_info)) | |
| 6880 return false; | |
| 6881 | |
| 6882 p1 += mcnt; /* Get past the n-th alternative. */ | |
| 6883 } /* if mcnt > 0 */ | |
| 6884 break; | |
| 6885 | |
| 6886 | |
| 6887 case stop_memory: | |
| 6888 assert (p1[1] == **p); | |
| 6889 *p = p1 + 2; | |
| 6890 return true; | |
| 6891 | |
| 6892 | |
| 6893 default: | |
| 6894 if (!common_op_match_null_string_p (&p1, end, reg_info)) | |
| 6895 return false; | |
| 6896 } | |
| 6897 } /* while p1 < end */ | |
| 6898 | |
| 6899 return false; | |
| 6900 } /* group_match_null_string_p */ | |
| 6901 | |
| 6902 | |
| 6903 /* Similar to group_match_null_string_p, but doesn't deal with alternatives: | |
| 6904 It expects P to be the first byte of a single alternative and END one | |
| 6905 byte past the last. The alternative can contain groups. */ | |
| 6906 | |
| 460 | 6907 static re_bool |
| 428 | 6908 alt_match_null_string_p (unsigned char *p, unsigned char *end, |
| 6909 register_info_type *reg_info) | |
| 6910 { | |
| 6911 int mcnt; | |
| 6912 unsigned char *p1 = p; | |
| 6913 | |
| 6914 while (p1 < end) | |
| 6915 { | |
| 6916 /* Skip over opcodes that can match nothing, and break when we get | |
| 6917 to one that can't. */ | |
| 6918 | |
| 6919 switch ((re_opcode_t) *p1) | |
| 6920 { | |
| 6921 /* It's a loop. */ | |
| 6922 case on_failure_jump: | |
| 6923 p1++; | |
| 6924 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 6925 p1 += mcnt; | |
| 6926 break; | |
| 6927 | |
| 6928 default: | |
| 6929 if (!common_op_match_null_string_p (&p1, end, reg_info)) | |
| 6930 return false; | |
| 6931 } | |
| 6932 } /* while p1 < end */ | |
| 6933 | |
| 6934 return true; | |
| 6935 } /* alt_match_null_string_p */ | |
| 6936 | |
| 6937 | |
| 6938 /* Deals with the ops common to group_match_null_string_p and | |
| 6939 alt_match_null_string_p. | |
| 6940 | |
| 6941 Sets P to one after the op and its arguments, if any. */ | |
| 6942 | |
| 460 | 6943 static re_bool |
| 428 | 6944 common_op_match_null_string_p (unsigned char **p, unsigned char *end, |
| 6945 register_info_type *reg_info) | |
| 6946 { | |
| 6947 int mcnt; | |
| 460 | 6948 re_bool ret; |
| 428 | 6949 int reg_no; |
| 6950 unsigned char *p1 = *p; | |
| 6951 | |
| 6952 switch ((re_opcode_t) *p1++) | |
| 6953 { | |
| 6954 case no_op: | |
| 6955 case begline: | |
| 6956 case endline: | |
| 6957 case begbuf: | |
| 6958 case endbuf: | |
| 6959 case wordbeg: | |
| 6960 case wordend: | |
| 6961 case wordbound: | |
| 6962 case notwordbound: | |
| 6963 #ifdef emacs | |
| 6964 case before_dot: | |
| 6965 case at_dot: | |
| 6966 case after_dot: | |
| 6967 #endif | |
| 6968 break; | |
| 6969 | |
| 6970 case start_memory: | |
| 6971 reg_no = *p1; | |
| 6972 assert (reg_no > 0 && reg_no <= MAX_REGNUM); | |
| 6973 ret = group_match_null_string_p (&p1, end, reg_info); | |
| 6974 | |
| 6975 /* Have to set this here in case we're checking a group which | |
| 6976 contains a group and a back reference to it. */ | |
| 6977 | |
| 6978 if (REG_MATCH_NULL_STRING_P (reg_info[reg_no]) == MATCH_NULL_UNSET_VALUE) | |
| 6979 REG_MATCH_NULL_STRING_P (reg_info[reg_no]) = ret; | |
| 6980 | |
| 6981 if (!ret) | |
| 6982 return false; | |
| 6983 break; | |
| 6984 | |
| 6985 /* If this is an optimized succeed_n for zero times, make the jump. */ | |
| 6986 case jump: | |
| 6987 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 6988 if (mcnt >= 0) | |
| 6989 p1 += mcnt; | |
| 6990 else | |
| 6991 return false; | |
| 6992 break; | |
| 6993 | |
| 6994 case succeed_n: | |
| 6995 /* Get to the number of times to succeed. */ | |
| 6996 p1 += 2; | |
| 6997 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 6998 | |
| 6999 if (mcnt == 0) | |
| 7000 { | |
| 7001 p1 -= 4; | |
| 7002 EXTRACT_NUMBER_AND_INCR (mcnt, p1); | |
| 7003 p1 += mcnt; | |
| 7004 } | |
| 7005 else | |
| 7006 return false; | |
| 7007 break; | |
| 7008 | |
| 7009 case duplicate: | |
| 7010 if (!REG_MATCH_NULL_STRING_P (reg_info[*p1])) | |
| 7011 return false; | |
| 7012 break; | |
| 7013 | |
| 7014 case set_number_at: | |
| 7015 p1 += 4; | |
| 7016 | |
| 7017 default: | |
| 7018 /* All other opcodes mean we cannot match the empty string. */ | |
| 7019 return false; | |
| 7020 } | |
| 7021 | |
| 7022 *p = p1; | |
| 7023 return true; | |
| 7024 } /* common_op_match_null_string_p */ | |
| 7025 | |
| 7026 | |
| 7027 /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN | |
| 7028 bytes; nonzero otherwise. */ | |
| 7029 | |
| 7030 static int | |
| 446 | 7031 bcmp_translate (re_char *s1, re_char *s2, |
| 826 | 7032 REGISTER int len, RE_TRANSLATE_TYPE translate |
| 7033 #ifdef emacs | |
| 2333 | 7034 , Internal_Format USED_IF_MULE (fmt), |
| 7035 Lisp_Object USED_IF_MULE (lispobj) | |
| 826 | 7036 #endif |
| 7037 ) | |
| 428 | 7038 { |
| 826 | 7039 REGISTER re_char *p1 = s1, *p2 = s2; |
| 446 | 7040 #ifdef MULE |
| 826 | 7041 re_char *p1_end = s1 + len; |
| 7042 re_char *p2_end = s2 + len; | |
| 446 | 7043 |
| 7044 while (p1 != p1_end && p2 != p2_end) | |
| 7045 { | |
| 867 | 7046 Ichar p1_ch, p2_ch; |
| 7047 | |
| 7048 p1_ch = itext_ichar_fmt (p1, fmt, lispobj); | |
| 7049 p2_ch = itext_ichar_fmt (p2, fmt, lispobj); | |
| 826 | 7050 |
| 7051 if (RE_TRANSLATE_1 (p1_ch) | |
| 7052 != RE_TRANSLATE_1 (p2_ch)) | |
| 446 | 7053 return 1; |
| 867 | 7054 INC_IBYTEPTR_FMT (p1, fmt); |
| 7055 INC_IBYTEPTR_FMT (p2, fmt); | |
| 446 | 7056 } |
| 7057 #else /* not MULE */ | |
| 428 | 7058 while (len) |
| 7059 { | |
| 826 | 7060 if (RE_TRANSLATE_1 (*p1++) != RE_TRANSLATE_1 (*p2++)) return 1; |
| 428 | 7061 len--; |
| 7062 } | |
| 446 | 7063 #endif /* MULE */ |
| 428 | 7064 return 0; |
| 7065 } | |
| 7066 | |
| 7067 /* Entry points for GNU code. */ | |
| 7068 | |
| 7069 /* re_compile_pattern is the GNU regular expression compiler: it | |
| 7070 compiles PATTERN (of length SIZE) and puts the result in BUFP. | |
| 7071 Returns 0 if the pattern was valid, otherwise an error string. | |
| 7072 | |
| 7073 Assumes the `allocated' (and perhaps `buffer') and `translate' fields | |
| 7074 are set in BUFP on entry. | |
| 7075 | |
| 7076 We call regex_compile to do the actual compilation. */ | |
| 7077 | |
| 442 | 7078 const char * |
| 7079 re_compile_pattern (const char *pattern, int length, | |
| 428 | 7080 struct re_pattern_buffer *bufp) |
| 7081 { | |
| 7082 reg_errcode_t ret; | |
| 7083 | |
| 7084 /* GNU code is written to assume at least RE_NREGS registers will be set | |
| 7085 (and at least one extra will be -1). */ | |
| 7086 bufp->regs_allocated = REGS_UNALLOCATED; | |
| 7087 | |
| 7088 /* And GNU code determines whether or not to get register information | |
| 7089 by passing null for the REGS argument to re_match, etc., not by | |
| 7090 setting no_sub. */ | |
| 7091 bufp->no_sub = 0; | |
| 7092 | |
| 7093 /* Match anchors at newline. */ | |
| 7094 bufp->newline_anchor = 1; | |
| 7095 | |
| 826 | 7096 ret = regex_compile ((unsigned char *) pattern, length, re_syntax_options, |
| 7097 bufp); | |
| 428 | 7098 |
| 7099 if (!ret) | |
| 7100 return NULL; | |
| 7101 return gettext (re_error_msgid[(int) ret]); | |
| 7102 } | |
| 7103 | |
| 7104 /* Entry points compatible with 4.2 BSD regex library. We don't define | |
| 7105 them unless specifically requested. */ | |
| 7106 | |
| 7107 #ifdef _REGEX_RE_COMP | |
| 7108 | |
| 7109 /* BSD has one and only one pattern buffer. */ | |
| 7110 static struct re_pattern_buffer re_comp_buf; | |
| 7111 | |
| 7112 char * | |
| 442 | 7113 re_comp (const char *s) |
| 428 | 7114 { |
| 7115 reg_errcode_t ret; | |
| 7116 | |
| 7117 if (!s) | |
| 7118 { | |
| 7119 if (!re_comp_buf.buffer) | |
| 7120 return gettext ("No previous regular expression"); | |
| 7121 return 0; | |
| 7122 } | |
| 7123 | |
| 7124 if (!re_comp_buf.buffer) | |
| 7125 { | |
| 1333 | 7126 re_comp_buf.buffer = (unsigned char *) xmalloc (200); |
| 428 | 7127 if (re_comp_buf.buffer == NULL) |
| 7128 return gettext (re_error_msgid[(int) REG_ESPACE]); | |
| 7129 re_comp_buf.allocated = 200; | |
| 7130 | |
| 1333 | 7131 re_comp_buf.fastmap = (char *) xmalloc (1 << BYTEWIDTH); |
| 428 | 7132 if (re_comp_buf.fastmap == NULL) |
| 7133 return gettext (re_error_msgid[(int) REG_ESPACE]); | |
| 7134 } | |
| 7135 | |
| 7136 /* Since `re_exec' always passes NULL for the `regs' argument, we | |
| 7137 don't need to initialize the pattern buffer fields which affect it. */ | |
| 7138 | |
| 7139 /* Match anchors at newlines. */ | |
| 7140 re_comp_buf.newline_anchor = 1; | |
| 7141 | |
| 826 | 7142 ret = regex_compile ((unsigned char *)s, strlen (s), re_syntax_options, |
| 7143 &re_comp_buf); | |
| 428 | 7144 |
| 7145 if (!ret) | |
| 7146 return NULL; | |
| 7147 | |
| 442 | 7148 /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
| 428 | 7149 return (char *) gettext (re_error_msgid[(int) ret]); |
| 7150 } | |
| 7151 | |
| 7152 | |
| 7153 int | |
| 442 | 7154 re_exec (const char *s) |
| 428 | 7155 { |
| 442 | 7156 const int len = strlen (s); |
| 428 | 7157 return |
| 7158 0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0); | |
| 7159 } | |
| 7160 #endif /* _REGEX_RE_COMP */ | |
| 7161 | |
| 7162 /* POSIX.2 functions. Don't define these for Emacs. */ | |
| 7163 | |
| 7164 #ifndef emacs | |
| 7165 | |
| 7166 /* regcomp takes a regular expression as a string and compiles it. | |
| 7167 | |
| 7168 PREG is a regex_t *. We do not expect any fields to be initialized, | |
| 7169 since POSIX says we shouldn't. Thus, we set | |
| 7170 | |
| 7171 `buffer' to the compiled pattern; | |
| 7172 `used' to the length of the compiled pattern; | |
| 7173 `syntax' to RE_SYNTAX_POSIX_EXTENDED if the | |
| 7174 REG_EXTENDED bit in CFLAGS is set; otherwise, to | |
| 7175 RE_SYNTAX_POSIX_BASIC; | |
| 7176 `newline_anchor' to REG_NEWLINE being set in CFLAGS; | |
| 7177 `fastmap' and `fastmap_accurate' to zero; | |
| 7178 `re_nsub' to the number of subexpressions in PATTERN. | |
| 502 | 7179 (non-shy of course. POSIX probably doesn't know about |
| 7180 shy ones, and in any case they should be invisible.) | |
| 428 | 7181 |
| 7182 PATTERN is the address of the pattern string. | |
| 7183 | |
| 7184 CFLAGS is a series of bits which affect compilation. | |
| 7185 | |
| 7186 If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we | |
| 7187 use POSIX basic syntax. | |
| 7188 | |
| 7189 If REG_NEWLINE is set, then . and [^...] don't match newline. | |
| 7190 Also, regexec will try a match beginning after every newline. | |
| 7191 | |
| 7192 If REG_ICASE is set, then we considers upper- and lowercase | |
| 7193 versions of letters to be equivalent when matching. | |
| 7194 | |
| 7195 If REG_NOSUB is set, then when PREG is passed to regexec, that | |
| 7196 routine will report only success or failure, and nothing about the | |
| 7197 registers. | |
| 7198 | |
| 7199 It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for | |
| 7200 the return codes and their meanings.) */ | |
| 7201 | |
| 7202 int | |
| 442 | 7203 regcomp (regex_t *preg, const char *pattern, int cflags) |
| 428 | 7204 { |
| 7205 reg_errcode_t ret; | |
| 647 | 7206 unsigned int syntax |
| 428 | 7207 = (cflags & REG_EXTENDED) ? |
| 7208 RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; | |
| 7209 | |
| 7210 /* regex_compile will allocate the space for the compiled pattern. */ | |
| 7211 preg->buffer = 0; | |
| 7212 preg->allocated = 0; | |
| 7213 preg->used = 0; | |
| 7214 | |
| 7215 /* Don't bother to use a fastmap when searching. This simplifies the | |
| 7216 REG_NEWLINE case: if we used a fastmap, we'd have to put all the | |
| 7217 characters after newlines into the fastmap. This way, we just try | |
| 7218 every character. */ | |
| 7219 preg->fastmap = 0; | |
| 7220 | |
| 7221 if (cflags & REG_ICASE) | |
| 7222 { | |
| 647 | 7223 int i; |
| 428 | 7224 |
| 1333 | 7225 preg->translate = (char *) xmalloc (CHAR_SET_SIZE); |
| 428 | 7226 if (preg->translate == NULL) |
| 7227 return (int) REG_ESPACE; | |
| 7228 | |
| 7229 /* Map uppercase characters to corresponding lowercase ones. */ | |
| 7230 for (i = 0; i < CHAR_SET_SIZE; i++) | |
| 7231 preg->translate[i] = ISUPPER (i) ? tolower (i) : i; | |
| 7232 } | |
| 7233 else | |
| 7234 preg->translate = NULL; | |
| 7235 | |
| 7236 /* If REG_NEWLINE is set, newlines are treated differently. */ | |
| 7237 if (cflags & REG_NEWLINE) | |
| 7238 { /* REG_NEWLINE implies neither . nor [^...] match newline. */ | |
| 7239 syntax &= ~RE_DOT_NEWLINE; | |
| 7240 syntax |= RE_HAT_LISTS_NOT_NEWLINE; | |
| 7241 /* It also changes the matching behavior. */ | |
| 7242 preg->newline_anchor = 1; | |
| 7243 } | |
| 7244 else | |
| 7245 preg->newline_anchor = 0; | |
| 7246 | |
| 7247 preg->no_sub = !!(cflags & REG_NOSUB); | |
| 7248 | |
| 7249 /* POSIX says a null character in the pattern terminates it, so we | |
| 7250 can use strlen here in compiling the pattern. */ | |
| 446 | 7251 ret = regex_compile ((unsigned char *) pattern, strlen (pattern), syntax, preg); |
| 428 | 7252 |
| 7253 /* POSIX doesn't distinguish between an unmatched open-group and an | |
| 7254 unmatched close-group: both are REG_EPAREN. */ | |
| 7255 if (ret == REG_ERPAREN) ret = REG_EPAREN; | |
| 7256 | |
| 7257 return (int) ret; | |
| 7258 } | |
| 7259 | |
| 7260 | |
| 7261 /* regexec searches for a given pattern, specified by PREG, in the | |
| 7262 string STRING. | |
| 7263 | |
| 7264 If NMATCH is zero or REG_NOSUB was set in the cflags argument to | |
| 7265 `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at | |
| 7266 least NMATCH elements, and we set them to the offsets of the | |
| 7267 corresponding matched substrings. | |
| 7268 | |
| 7269 EFLAGS specifies `execution flags' which affect matching: if | |
| 7270 REG_NOTBOL is set, then ^ does not match at the beginning of the | |
| 7271 string; if REG_NOTEOL is set, then $ does not match at the end. | |
| 7272 | |
| 7273 We return 0 if we find a match and REG_NOMATCH if not. */ | |
| 7274 | |
| 7275 int | |
| 442 | 7276 regexec (const regex_t *preg, const char *string, size_t nmatch, |
| 428 | 7277 regmatch_t pmatch[], int eflags) |
| 7278 { | |
| 7279 int ret; | |
| 7280 struct re_registers regs; | |
| 7281 regex_t private_preg; | |
| 7282 int len = strlen (string); | |
| 460 | 7283 re_bool want_reg_info = !preg->no_sub && nmatch > 0; |
| 428 | 7284 |
| 7285 private_preg = *preg; | |
| 7286 | |
| 7287 private_preg.not_bol = !!(eflags & REG_NOTBOL); | |
| 7288 private_preg.not_eol = !!(eflags & REG_NOTEOL); | |
| 7289 | |
| 7290 /* The user has told us exactly how many registers to return | |
| 7291 information about, via `nmatch'. We have to pass that on to the | |
| 7292 matching routines. */ | |
| 7293 private_preg.regs_allocated = REGS_FIXED; | |
| 7294 | |
| 7295 if (want_reg_info) | |
| 7296 { | |
| 647 | 7297 regs.num_regs = (int) nmatch; |
| 7298 regs.start = TALLOC ((int) nmatch, regoff_t); | |
| 7299 regs.end = TALLOC ((int) nmatch, regoff_t); | |
| 428 | 7300 if (regs.start == NULL || regs.end == NULL) |
| 7301 return (int) REG_NOMATCH; | |
| 7302 } | |
| 7303 | |
| 7304 /* Perform the searching operation. */ | |
| 7305 ret = re_search (&private_preg, string, len, | |
| 7306 /* start: */ 0, /* range: */ len, | |
| 7307 want_reg_info ? ®s : (struct re_registers *) 0); | |
| 7308 | |
| 7309 /* Copy the register information to the POSIX structure. */ | |
| 7310 if (want_reg_info) | |
| 7311 { | |
| 7312 if (ret >= 0) | |
| 7313 { | |
| 647 | 7314 int r; |
| 7315 | |
| 7316 for (r = 0; r < (int) nmatch; r++) | |
| 428 | 7317 { |
| 7318 pmatch[r].rm_so = regs.start[r]; | |
| 7319 pmatch[r].rm_eo = regs.end[r]; | |
| 7320 } | |
| 7321 } | |
| 7322 | |
| 7323 /* If we needed the temporary register info, free the space now. */ | |
|
4976
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
7324 xfree (regs.start); |
|
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
7325 xfree (regs.end); |
| 428 | 7326 } |
| 7327 | |
| 7328 /* We want zero return to mean success, unlike `re_search'. */ | |
| 7329 return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH; | |
| 7330 } | |
| 7331 | |
| 7332 | |
| 7333 /* Returns a message corresponding to an error code, ERRCODE, returned | |
| 7334 from either regcomp or regexec. We don't use PREG here. */ | |
| 7335 | |
| 7336 size_t | |
| 2286 | 7337 regerror (int errcode, const regex_t *UNUSED (preg), char *errbuf, |
| 647 | 7338 size_t errbuf_size) |
| 428 | 7339 { |
| 442 | 7340 const char *msg; |
| 665 | 7341 Bytecount msg_size; |
| 428 | 7342 |
| 7343 if (errcode < 0 | |
| 647 | 7344 || errcode >= (int) (sizeof (re_error_msgid) / |
| 7345 sizeof (re_error_msgid[0]))) | |
| 428 | 7346 /* Only error codes returned by the rest of the code should be passed |
| 7347 to this routine. If we are given anything else, or if other regex | |
| 7348 code generates an invalid error code, then the program has a bug. | |
| 7349 Dump core so we can fix it. */ | |
| 2500 | 7350 ABORT (); |
| 428 | 7351 |
| 7352 msg = gettext (re_error_msgid[errcode]); | |
| 7353 | |
| 7354 msg_size = strlen (msg) + 1; /* Includes the null. */ | |
| 7355 | |
| 7356 if (errbuf_size != 0) | |
| 7357 { | |
| 665 | 7358 if (msg_size > (Bytecount) errbuf_size) |
| 428 | 7359 { |
| 7360 strncpy (errbuf, msg, errbuf_size - 1); | |
| 7361 errbuf[errbuf_size - 1] = 0; | |
| 7362 } | |
| 7363 else | |
| 7364 strcpy (errbuf, msg); | |
| 7365 } | |
| 7366 | |
| 647 | 7367 return (size_t) msg_size; |
| 428 | 7368 } |
| 7369 | |
| 7370 | |
| 7371 /* Free dynamically allocated space used by PREG. */ | |
| 7372 | |
| 7373 void | |
| 7374 regfree (regex_t *preg) | |
| 7375 { | |
| 7376 if (preg->buffer != NULL) | |
|
4976
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
7377 xfree (preg->buffer); |
| 428 | 7378 preg->buffer = NULL; |
| 7379 | |
| 7380 preg->allocated = 0; | |
| 7381 preg->used = 0; | |
| 7382 | |
| 7383 if (preg->fastmap != NULL) | |
|
4976
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
7384 xfree (preg->fastmap); |
| 428 | 7385 preg->fastmap = NULL; |
| 7386 preg->fastmap_accurate = 0; | |
| 7387 | |
| 7388 if (preg->translate != NULL) | |
|
4976
16112448d484
Rename xfree(FOO, TYPE) -> xfree(FOO)
Ben Wing <ben@xemacs.org>
parents:
4832
diff
changeset
|
7389 xfree (preg->translate); |
| 428 | 7390 preg->translate = NULL; |
| 7391 } | |
| 7392 | |
| 7393 #endif /* not emacs */ | |
| 7394 |