]>
Commit | Line | Data |
---|---|---|
3b0bdc72 | 1 | /* Extended regular expression matching and search library. |
1c99f950 | 2 | Copyright (C) 2002, 2003, 2004, 2005 Free Software Foundation, Inc. |
3b0bdc72 UD |
3 | This file is part of the GNU C Library. |
4 | Contributed by Isamu Hasegawa <isamu@yamato.ibm.com>. | |
5 | ||
6 | The GNU C Library is free software; you can redistribute it and/or | |
7 | modify it under the terms of the GNU Lesser General Public | |
8 | License as published by the Free Software Foundation; either | |
9 | version 2.1 of the License, or (at your option) any later version. | |
10 | ||
11 | The GNU C Library is distributed in the hope that it will be useful, | |
12 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | Lesser General Public License for more details. | |
15 | ||
16 | You should have received a copy of the GNU Lesser General Public | |
17 | License along with the GNU C Library; if not, write to the Free | |
18 | Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA | |
19 | 02111-1307 USA. */ | |
20 | ||
a9388965 | 21 | static reg_errcode_t match_ctx_init (re_match_context_t *cache, int eflags, |
56b168be | 22 | int n) internal_function; |
3ce12656 UD |
23 | static void match_ctx_clean (re_match_context_t *mctx) internal_function; |
24 | static void match_ctx_free (re_match_context_t *cache) internal_function; | |
a9388965 | 25 | static reg_errcode_t match_ctx_add_entry (re_match_context_t *cache, int node, |
e3a87852 UD |
26 | int str_idx, int from, int to) |
27 | internal_function; | |
01ed6ceb | 28 | static int search_cur_bkref_entry (const re_match_context_t *mctx, int str_idx) |
e3a87852 | 29 | internal_function; |
6291ee3c | 30 | static reg_errcode_t match_ctx_add_subtop (re_match_context_t *mctx, int node, |
3ce12656 | 31 | int str_idx) internal_function; |
6291ee3c | 32 | static re_sub_match_last_t * match_ctx_add_sublast (re_sub_match_top_t *subtop, |
e3a87852 UD |
33 | int node, int str_idx) |
34 | internal_function; | |
0742e48e | 35 | static void sift_ctx_init (re_sift_context_t *sctx, re_dfastate_t **sifted_sts, |
15a7d175 | 36 | re_dfastate_t **limited_sts, int last_node, |
d2c38eb3 | 37 | int last_str_idx) |
e3a87852 | 38 | internal_function; |
a9388965 | 39 | static reg_errcode_t re_search_internal (const regex_t *preg, |
15a7d175 UD |
40 | const char *string, int length, |
41 | int start, int range, int stop, | |
42 | size_t nmatch, regmatch_t pmatch[], | |
3ce12656 | 43 | int eflags) internal_function; |
ac3d553b | 44 | static int re_search_2_stub (struct re_pattern_buffer *bufp, |
15a7d175 UD |
45 | const char *string1, int length1, |
46 | const char *string2, int length2, | |
47 | int start, int range, struct re_registers *regs, | |
3ce12656 | 48 | int stop, int ret_len) internal_function; |
ac3d553b | 49 | static int re_search_stub (struct re_pattern_buffer *bufp, |
15a7d175 UD |
50 | const char *string, int length, int start, |
51 | int range, int stop, struct re_registers *regs, | |
3ce12656 | 52 | int ret_len) internal_function; |
ac3d553b | 53 | static unsigned re_copy_regs (struct re_registers *regs, regmatch_t *pmatch, |
3ce12656 | 54 | int nregs, int regs_allocated) internal_function; |
a334319f UD |
55 | static inline re_dfastate_t *acquire_init_state_context |
56 | (reg_errcode_t *err, const re_match_context_t *mctx, int idx) | |
57 | __attribute ((always_inline)) internal_function; | |
e3a87852 UD |
58 | static reg_errcode_t prune_impossible_nodes (re_match_context_t *mctx) |
59 | internal_function; | |
15bad1a5 | 60 | static int check_matching (re_match_context_t *mctx, int fl_longest_match, |
76b864c8 | 61 | int *p_match_first) internal_function; |
a334319f UD |
62 | static int check_halt_node_context (const re_dfa_t *dfa, int node, |
63 | unsigned int context) internal_function; | |
e3a87852 UD |
64 | static int check_halt_state_context (const re_match_context_t *mctx, |
65 | const re_dfastate_t *state, int idx) | |
66 | internal_function; | |
76b864c8 | 67 | static void update_regs (const re_dfa_t *dfa, regmatch_t *pmatch, |
6b6557e8 | 68 | regmatch_t *prev_idx_match, int cur_node, |
3ce12656 | 69 | int cur_idx, int nmatch) internal_function; |
a334319f UD |
70 | static int proceed_next_node (const re_match_context_t *mctx, |
71 | int nregs, regmatch_t *regs, | |
72 | int *pidx, int node, re_node_set *eps_via_nodes, | |
73 | struct re_fail_stack_t *fs) internal_function; | |
1b2c2628 | 74 | static reg_errcode_t push_fail_stack (struct re_fail_stack_t *fs, |
5cf1ec52 | 75 | int str_idx, int dest_node, int nregs, |
15a7d175 | 76 | regmatch_t *regs, |
76b864c8 UD |
77 | re_node_set *eps_via_nodes) |
78 | internal_function; | |
a334319f | 79 | static int pop_fail_stack (struct re_fail_stack_t *fs, int *pidx, int nregs, |
76b864c8 UD |
80 | regmatch_t *regs, re_node_set *eps_via_nodes) |
81 | internal_function; | |
612546c6 | 82 | static reg_errcode_t set_regs (const regex_t *preg, |
15a7d175 UD |
83 | const re_match_context_t *mctx, |
84 | size_t nmatch, regmatch_t *pmatch, | |
3ce12656 | 85 | int fl_backtrack) internal_function; |
76b864c8 UD |
86 | static reg_errcode_t free_fail_stack_return (struct re_fail_stack_t *fs) |
87 | internal_function; | |
485d775d | 88 | |
434d3784 | 89 | #ifdef RE_ENABLE_I18N |
e3a87852 | 90 | static int sift_states_iter_mb (const re_match_context_t *mctx, |
15a7d175 | 91 | re_sift_context_t *sctx, |
76b864c8 UD |
92 | int node_idx, int str_idx, int max_str_idx) |
93 | internal_function; | |
434d3784 | 94 | #endif /* RE_ENABLE_I18N */ |
76b864c8 UD |
95 | static reg_errcode_t sift_states_backward (const re_match_context_t *mctx, |
96 | re_sift_context_t *sctx) | |
97 | internal_function; | |
98 | static reg_errcode_t build_sifted_states (const re_match_context_t *mctx, | |
e40a38b3 | 99 | re_sift_context_t *sctx, int str_idx, |
76b864c8 UD |
100 | re_node_set *cur_dest) |
101 | internal_function; | |
102 | static reg_errcode_t update_cur_sifted_state (const re_match_context_t *mctx, | |
15a7d175 UD |
103 | re_sift_context_t *sctx, |
104 | int str_idx, | |
76b864c8 UD |
105 | re_node_set *dest_nodes) |
106 | internal_function; | |
107 | static reg_errcode_t add_epsilon_src_nodes (const re_dfa_t *dfa, | |
15a7d175 | 108 | re_node_set *dest_nodes, |
76b864c8 UD |
109 | const re_node_set *candidates) |
110 | internal_function; | |
111 | static reg_errcode_t sub_epsilon_src_nodes (const re_dfa_t *dfa, int node, | |
a334319f | 112 | re_node_set *dest_nodes, |
76b864c8 UD |
113 | const re_node_set *and_nodes) |
114 | internal_function; | |
01ed6ceb UD |
115 | static int check_dst_limits (const re_match_context_t *mctx, |
116 | re_node_set *limits, | |
e3a87852 UD |
117 | int dst_node, int dst_idx, int src_node, |
118 | int src_idx) internal_function; | |
01ed6ceb | 119 | static int check_dst_limits_calc_pos_1 (const re_match_context_t *mctx, |
e40a38b3 | 120 | int boundaries, int subexp_idx, |
76b864c8 UD |
121 | int from_node, int bkref_idx) |
122 | internal_function; | |
01ed6ceb | 123 | static int check_dst_limits_calc_pos (const re_match_context_t *mctx, |
e40a38b3 UD |
124 | int limit, int subexp_idx, |
125 | int node, int str_idx, | |
126 | int bkref_idx) internal_function; | |
76b864c8 | 127 | static reg_errcode_t check_subexp_limits (const re_dfa_t *dfa, |
15a7d175 UD |
128 | re_node_set *dest_nodes, |
129 | const re_node_set *candidates, | |
130 | re_node_set *limits, | |
131 | struct re_backref_cache_entry *bkref_ents, | |
3ce12656 | 132 | int str_idx) internal_function; |
76b864c8 | 133 | static reg_errcode_t sift_states_bkref (const re_match_context_t *mctx, |
15a7d175 | 134 | re_sift_context_t *sctx, |
76b864c8 UD |
135 | int str_idx, const re_node_set *candidates) |
136 | internal_function; | |
a334319f | 137 | static reg_errcode_t clean_state_log_if_needed (re_match_context_t *mctx, |
76b864c8 UD |
138 | int next_state_log_idx) |
139 | internal_function; | |
140 | static reg_errcode_t merge_state_array (const re_dfa_t *dfa, | |
141 | re_dfastate_t **dst, | |
142 | re_dfastate_t **src, int num) | |
143 | internal_function; | |
4c595adb UD |
144 | static re_dfastate_t *find_recover_state (reg_errcode_t *err, |
145 | re_match_context_t *mctx) internal_function; | |
e3a87852 | 146 | static re_dfastate_t *transit_state (reg_errcode_t *err, |
15a7d175 | 147 | re_match_context_t *mctx, |
3ce12656 | 148 | re_dfastate_t *state) internal_function; |
4c595adb UD |
149 | static re_dfastate_t *merge_state_with_log (reg_errcode_t *err, |
150 | re_match_context_t *mctx, | |
76b864c8 UD |
151 | re_dfastate_t *next_state) |
152 | internal_function; | |
e3a87852 | 153 | static reg_errcode_t check_subexp_matching_top (re_match_context_t *mctx, |
6291ee3c | 154 | re_node_set *cur_nodes, |
3ce12656 | 155 | int str_idx) internal_function; |
c13c99fa | 156 | #if 0 |
e3a87852 UD |
157 | static re_dfastate_t *transit_state_sb (reg_errcode_t *err, |
158 | re_match_context_t *mctx, | |
76b864c8 UD |
159 | re_dfastate_t *pstate) |
160 | internal_function; | |
c13c99fa | 161 | #endif |
434d3784 | 162 | #ifdef RE_ENABLE_I18N |
e3a87852 | 163 | static reg_errcode_t transit_state_mb (re_match_context_t *mctx, |
76b864c8 UD |
164 | re_dfastate_t *pstate) |
165 | internal_function; | |
434d3784 | 166 | #endif /* RE_ENABLE_I18N */ |
e3a87852 | 167 | static reg_errcode_t transit_state_bkref (re_match_context_t *mctx, |
76b864c8 UD |
168 | const re_node_set *nodes) |
169 | internal_function; | |
e3a87852 | 170 | static reg_errcode_t get_subexp (re_match_context_t *mctx, |
76b864c8 UD |
171 | int bkref_node, int bkref_str_idx) |
172 | internal_function; | |
e3a87852 | 173 | static reg_errcode_t get_subexp_sub (re_match_context_t *mctx, |
fe9434bb | 174 | const re_sub_match_top_t *sub_top, |
6291ee3c | 175 | re_sub_match_last_t *sub_last, |
76b864c8 UD |
176 | int bkref_node, int bkref_str) |
177 | internal_function; | |
fe9434bb | 178 | static int find_subexp_node (const re_dfa_t *dfa, const re_node_set *nodes, |
3ce12656 | 179 | int subexp_idx, int type) internal_function; |
e3a87852 | 180 | static reg_errcode_t check_arrival (re_match_context_t *mctx, |
6291ee3c UD |
181 | state_array_t *path, int top_node, |
182 | int top_str, int last_node, int last_str, | |
3ce12656 | 183 | int type) internal_function; |
e3a87852 | 184 | static reg_errcode_t check_arrival_add_next_nodes (re_match_context_t *mctx, |
6291ee3c UD |
185 | int str_idx, |
186 | re_node_set *cur_nodes, | |
76b864c8 UD |
187 | re_node_set *next_nodes) |
188 | internal_function; | |
6efbd82c | 189 | static reg_errcode_t check_arrival_expand_ecl (const re_dfa_t *dfa, |
6291ee3c | 190 | re_node_set *cur_nodes, |
76b864c8 UD |
191 | int ex_subexp, int type) |
192 | internal_function; | |
6efbd82c | 193 | static reg_errcode_t check_arrival_expand_ecl_sub (const re_dfa_t *dfa, |
6291ee3c UD |
194 | re_node_set *dst_nodes, |
195 | int target, int ex_subexp, | |
3ce12656 | 196 | int type) internal_function; |
e3a87852 | 197 | static reg_errcode_t expand_bkref_cache (re_match_context_t *mctx, |
6291ee3c | 198 | re_node_set *cur_nodes, int cur_str, |
76b864c8 UD |
199 | int subexp_num, int type) |
200 | internal_function; | |
201 | static int build_trtable (const re_dfa_t *dfa, | |
ab4b89fe | 202 | re_dfastate_t *state) internal_function; |
434d3784 | 203 | #ifdef RE_ENABLE_I18N |
c42b4152 | 204 | static int check_node_accept_bytes (const re_dfa_t *dfa, int node_idx, |
76b864c8 UD |
205 | const re_string_t *input, int idx) |
206 | internal_function; | |
434d3784 | 207 | # ifdef _LIBC |
c202c2c5 | 208 | static unsigned int find_collation_sequence_value (const unsigned char *mbs, |
76b864c8 UD |
209 | size_t name_len) |
210 | internal_function; | |
434d3784 UD |
211 | # endif /* _LIBC */ |
212 | #endif /* RE_ENABLE_I18N */ | |
01ed6ceb | 213 | static int group_nodes_into_DFAstates (const re_dfa_t *dfa, |
15a7d175 UD |
214 | const re_dfastate_t *state, |
215 | re_node_set *states_node, | |
2c05d33f | 216 | bitset_t *states_ch) internal_function; |
e3a87852 | 217 | static int check_node_accept (const re_match_context_t *mctx, |
76b864c8 UD |
218 | const re_token_t *node, int idx) |
219 | internal_function; | |
220 | static reg_errcode_t extend_buffers (re_match_context_t *mctx) | |
221 | internal_function; | |
3b0bdc72 UD |
222 | \f |
223 | /* Entry point for POSIX code. */ | |
224 | ||
225 | /* regexec searches for a given pattern, specified by PREG, in the | |
226 | string STRING. | |
227 | ||
228 | If NMATCH is zero or REG_NOSUB was set in the cflags argument to | |
229 | `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at | |
230 | least NMATCH elements, and we set them to the offsets of the | |
231 | corresponding matched substrings. | |
232 | ||
233 | EFLAGS specifies `execution flags' which affect matching: if | |
234 | REG_NOTBOL is set, then ^ does not match at the beginning of the | |
235 | string; if REG_NOTEOL is set, then $ does not match at the end. | |
236 | ||
237 | We return 0 if we find a match and REG_NOMATCH if not. */ | |
238 | ||
239 | int | |
240 | regexec (preg, string, nmatch, pmatch, eflags) | |
92b27c74 UD |
241 | const regex_t *__restrict preg; |
242 | const char *__restrict string; | |
3b0bdc72 UD |
243 | size_t nmatch; |
244 | regmatch_t pmatch[]; | |
245 | int eflags; | |
246 | { | |
a9388965 | 247 | reg_errcode_t err; |
6fefb4e0 | 248 | int start, length; |
76b864c8 | 249 | re_dfa_t *dfa = (re_dfa_t *) preg->buffer; |
3f2fb223 UD |
250 | |
251 | if (eflags & ~(REG_NOTBOL | REG_NOTEOL | REG_STARTEND)) | |
252 | return REG_BADPAT; | |
253 | ||
6fefb4e0 UD |
254 | if (eflags & REG_STARTEND) |
255 | { | |
256 | start = pmatch[0].rm_so; | |
257 | length = pmatch[0].rm_eo; | |
258 | } | |
259 | else | |
260 | { | |
261 | start = 0; | |
262 | length = strlen (string); | |
263 | } | |
7b918993 UD |
264 | |
265 | __libc_lock_lock (dfa->lock); | |
3b0bdc72 | 266 | if (preg->no_sub) |
6fefb4e0 UD |
267 | err = re_search_internal (preg, string, length, start, length - start, |
268 | length, 0, NULL, eflags); | |
3b0bdc72 | 269 | else |
6fefb4e0 UD |
270 | err = re_search_internal (preg, string, length, start, length - start, |
271 | length, nmatch, pmatch, eflags); | |
7b918993 | 272 | __libc_lock_unlock (dfa->lock); |
a9388965 | 273 | return err != REG_NOERROR; |
3b0bdc72 | 274 | } |
3f2fb223 | 275 | |
3b0bdc72 | 276 | #ifdef _LIBC |
3f2fb223 UD |
277 | # include <shlib-compat.h> |
278 | versioned_symbol (libc, __regexec, regexec, GLIBC_2_3_4); | |
279 | ||
280 | # if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_3_4) | |
281 | __typeof__ (__regexec) __compat_regexec; | |
282 | ||
283 | int | |
78678039 | 284 | attribute_compat_text_section |
3f2fb223 UD |
285 | __compat_regexec (const regex_t *__restrict preg, |
286 | const char *__restrict string, size_t nmatch, | |
287 | regmatch_t pmatch[], int eflags) | |
288 | { | |
289 | return regexec (preg, string, nmatch, pmatch, | |
290 | eflags & (REG_NOTBOL | REG_NOTEOL)); | |
291 | } | |
292 | compat_symbol (libc, __compat_regexec, regexec, GLIBC_2_0); | |
293 | # endif | |
3b0bdc72 UD |
294 | #endif |
295 | ||
296 | /* Entry points for GNU code. */ | |
297 | ||
ac3d553b UD |
298 | /* re_match, re_search, re_match_2, re_search_2 |
299 | ||
300 | The former two functions operate on STRING with length LENGTH, | |
301 | while the later two operate on concatenation of STRING1 and STRING2 | |
302 | with lengths LENGTH1 and LENGTH2, respectively. | |
303 | ||
304 | re_match() matches the compiled pattern in BUFP against the string, | |
305 | starting at index START. | |
306 | ||
307 | re_search() first tries matching at index START, then it tries to match | |
308 | starting from index START + 1, and so on. The last start position tried | |
309 | is START + RANGE. (Thus RANGE = 0 forces re_search to operate the same | |
310 | way as re_match().) | |
311 | ||
312 | The parameter STOP of re_{match,search}_2 specifies that no match exceeding | |
313 | the first STOP characters of the concatenation of the strings should be | |
314 | concerned. | |
315 | ||
316 | If REGS is not NULL, and BUFP->no_sub is not set, the offsets of the match | |
317 | and all groups is stroed in REGS. (For the "_2" variants, the offsets are | |
318 | computed relative to the concatenation, not relative to the individual | |
319 | strings.) | |
320 | ||
321 | On success, re_match* functions return the length of the match, re_search* | |
322 | return the position of the start of the match. Return value -1 means no | |
323 | match was found and -2 indicates an internal error. */ | |
3b0bdc72 UD |
324 | |
325 | int | |
ac3d553b UD |
326 | re_match (bufp, string, length, start, regs) |
327 | struct re_pattern_buffer *bufp; | |
3b0bdc72 UD |
328 | const char *string; |
329 | int length, start; | |
330 | struct re_registers *regs; | |
331 | { | |
ac3d553b | 332 | return re_search_stub (bufp, string, length, start, 0, length, regs, 1); |
3b0bdc72 UD |
333 | } |
334 | #ifdef _LIBC | |
335 | weak_alias (__re_match, re_match) | |
336 | #endif | |
337 | ||
ac3d553b UD |
338 | int |
339 | re_search (bufp, string, length, start, range, regs) | |
340 | struct re_pattern_buffer *bufp; | |
341 | const char *string; | |
342 | int length, start, range; | |
343 | struct re_registers *regs; | |
344 | { | |
345 | return re_search_stub (bufp, string, length, start, range, length, regs, 0); | |
346 | } | |
347 | #ifdef _LIBC | |
348 | weak_alias (__re_search, re_search) | |
349 | #endif | |
3b0bdc72 UD |
350 | |
351 | int | |
ac3d553b UD |
352 | re_match_2 (bufp, string1, length1, string2, length2, start, regs, stop) |
353 | struct re_pattern_buffer *bufp; | |
354 | const char *string1, *string2; | |
355 | int length1, length2, start, stop; | |
356 | struct re_registers *regs; | |
3b0bdc72 | 357 | { |
ac3d553b | 358 | return re_search_2_stub (bufp, string1, length1, string2, length2, |
15a7d175 | 359 | start, 0, regs, stop, 1); |
3b0bdc72 UD |
360 | } |
361 | #ifdef _LIBC | |
362 | weak_alias (__re_match_2, re_match_2) | |
363 | #endif | |
364 | ||
3b0bdc72 | 365 | int |
ac3d553b UD |
366 | re_search_2 (bufp, string1, length1, string2, length2, start, range, regs, stop) |
367 | struct re_pattern_buffer *bufp; | |
368 | const char *string1, *string2; | |
369 | int length1, length2, start, range, stop; | |
370 | struct re_registers *regs; | |
371 | { | |
372 | return re_search_2_stub (bufp, string1, length1, string2, length2, | |
15a7d175 | 373 | start, range, regs, stop, 0); |
ac3d553b UD |
374 | } |
375 | #ifdef _LIBC | |
376 | weak_alias (__re_search_2, re_search_2) | |
377 | #endif | |
378 | ||
379 | static int | |
380 | re_search_2_stub (bufp, string1, length1, string2, length2, start, range, regs, | |
15a7d175 | 381 | stop, ret_len) |
ac3d553b UD |
382 | struct re_pattern_buffer *bufp; |
383 | const char *string1, *string2; | |
384 | int length1, length2, start, range, stop, ret_len; | |
385 | struct re_registers *regs; | |
386 | { | |
387 | const char *str; | |
388 | int rval; | |
389 | int len = length1 + length2; | |
390 | int free_str = 0; | |
391 | ||
392 | if (BE (length1 < 0 || length2 < 0 || stop < 0, 0)) | |
393 | return -2; | |
394 | ||
395 | /* Concatenate the strings. */ | |
396 | if (length2 > 0) | |
397 | if (length1 > 0) | |
398 | { | |
15a7d175 UD |
399 | char *s = re_malloc (char, len); |
400 | ||
401 | if (BE (s == NULL, 0)) | |
402 | return -2; | |
c42b4152 UD |
403 | #ifdef _LIBC |
404 | memcpy (__mempcpy (s, string1, length1), string2, length2); | |
405 | #else | |
15a7d175 UD |
406 | memcpy (s, string1, length1); |
407 | memcpy (s + length1, string2, length2); | |
c42b4152 | 408 | #endif |
15a7d175 UD |
409 | str = s; |
410 | free_str = 1; | |
ac3d553b UD |
411 | } |
412 | else | |
413 | str = string2; | |
414 | else | |
415 | str = string1; | |
416 | ||
417 | rval = re_search_stub (bufp, str, len, start, range, stop, regs, | |
15a7d175 | 418 | ret_len); |
ac3d553b | 419 | if (free_str) |
1b2c2628 | 420 | re_free ((char *) str); |
ac3d553b UD |
421 | return rval; |
422 | } | |
423 | ||
424 | /* The parameters have the same meaning as those of re_search. | |
425 | Additional parameters: | |
426 | If RET_LEN is nonzero the length of the match is returned (re_match style); | |
427 | otherwise the position of the match is returned. */ | |
428 | ||
429 | static int | |
430 | re_search_stub (bufp, string, length, start, range, stop, regs, ret_len) | |
431 | struct re_pattern_buffer *bufp; | |
432 | const char *string; | |
433 | int length, start, range, stop, ret_len; | |
434 | struct re_registers *regs; | |
3b0bdc72 | 435 | { |
a9388965 | 436 | reg_errcode_t result; |
3b0bdc72 | 437 | regmatch_t *pmatch; |
ac3d553b UD |
438 | int nregs, rval; |
439 | int eflags = 0; | |
76b864c8 | 440 | re_dfa_t *dfa = (re_dfa_t *) bufp->buffer; |
ac3d553b UD |
441 | |
442 | /* Check for out-of-range. */ | |
a877402c | 443 | if (BE (start < 0 || start > length, 0)) |
ac3d553b UD |
444 | return -1; |
445 | if (BE (start + range > length, 0)) | |
446 | range = length - start; | |
a877402c UD |
447 | else if (BE (start + range < 0, 0)) |
448 | range = -start; | |
3b0bdc72 | 449 | |
7b918993 UD |
450 | __libc_lock_lock (dfa->lock); |
451 | ||
3b0bdc72 UD |
452 | eflags |= (bufp->not_bol) ? REG_NOTBOL : 0; |
453 | eflags |= (bufp->not_eol) ? REG_NOTEOL : 0; | |
454 | ||
ac3d553b UD |
455 | /* Compile fastmap if we haven't yet. */ |
456 | if (range > 0 && bufp->fastmap != NULL && !bufp->fastmap_accurate) | |
457 | re_compile_fastmap (bufp); | |
458 | ||
459 | if (BE (bufp->no_sub, 0)) | |
460 | regs = NULL; | |
3b0bdc72 UD |
461 | |
462 | /* We need at least 1 register. */ | |
ac3d553b UD |
463 | if (regs == NULL) |
464 | nregs = 1; | |
465 | else if (BE (bufp->regs_allocated == REGS_FIXED && | |
15a7d175 | 466 | regs->num_regs < bufp->re_nsub + 1, 0)) |
ac3d553b UD |
467 | { |
468 | nregs = regs->num_regs; | |
469 | if (BE (nregs < 1, 0)) | |
15a7d175 UD |
470 | { |
471 | /* Nothing can be copied to regs. */ | |
472 | regs = NULL; | |
473 | nregs = 1; | |
474 | } | |
ac3d553b UD |
475 | } |
476 | else | |
477 | nregs = bufp->re_nsub + 1; | |
3b0bdc72 | 478 | pmatch = re_malloc (regmatch_t, nregs); |
bc15410e | 479 | if (BE (pmatch == NULL, 0)) |
7b918993 UD |
480 | { |
481 | rval = -2; | |
482 | goto out; | |
483 | } | |
3b0bdc72 | 484 | |
ac3d553b | 485 | result = re_search_internal (bufp, string, length, start, range, stop, |
15a7d175 | 486 | nregs, pmatch, eflags); |
3b0bdc72 | 487 | |
ac3d553b UD |
488 | rval = 0; |
489 | ||
490 | /* I hope we needn't fill ther regs with -1's when no match was found. */ | |
491 | if (result != REG_NOERROR) | |
492 | rval = -1; | |
493 | else if (regs != NULL) | |
3b0bdc72 | 494 | { |
ac3d553b UD |
495 | /* If caller wants register contents data back, copy them. */ |
496 | bufp->regs_allocated = re_copy_regs (regs, pmatch, nregs, | |
15a7d175 | 497 | bufp->regs_allocated); |
ac3d553b | 498 | if (BE (bufp->regs_allocated == REGS_UNALLOCATED, 0)) |
15a7d175 | 499 | rval = -2; |
3b0bdc72 UD |
500 | } |
501 | ||
ac3d553b | 502 | if (BE (rval == 0, 1)) |
3b0bdc72 | 503 | { |
ac3d553b | 504 | if (ret_len) |
15a7d175 UD |
505 | { |
506 | assert (pmatch[0].rm_so == start); | |
507 | rval = pmatch[0].rm_eo - start; | |
508 | } | |
ac3d553b | 509 | else |
15a7d175 | 510 | rval = pmatch[0].rm_so; |
3b0bdc72 | 511 | } |
3b0bdc72 | 512 | re_free (pmatch); |
7b918993 UD |
513 | out: |
514 | __libc_lock_unlock (dfa->lock); | |
3b0bdc72 UD |
515 | return rval; |
516 | } | |
3b0bdc72 | 517 | |
ac3d553b UD |
518 | static unsigned |
519 | re_copy_regs (regs, pmatch, nregs, regs_allocated) | |
3b0bdc72 | 520 | struct re_registers *regs; |
ac3d553b UD |
521 | regmatch_t *pmatch; |
522 | int nregs, regs_allocated; | |
3b0bdc72 | 523 | { |
ac3d553b UD |
524 | int rval = REGS_REALLOCATE; |
525 | int i; | |
526 | int need_regs = nregs + 1; | |
527 | /* We need one extra element beyond `num_regs' for the `-1' marker GNU code | |
528 | uses. */ | |
529 | ||
530 | /* Have the register data arrays been allocated? */ | |
531 | if (regs_allocated == REGS_UNALLOCATED) | |
44777771 | 532 | { /* No. So allocate them with malloc. */ |
a96c63ed UD |
533 | regs->start = re_malloc (regoff_t, need_regs); |
534 | regs->end = re_malloc (regoff_t, need_regs); | |
535 | if (BE (regs->start == NULL, 0) || BE (regs->end == NULL, 0)) | |
15a7d175 | 536 | return REGS_UNALLOCATED; |
ac3d553b UD |
537 | regs->num_regs = need_regs; |
538 | } | |
539 | else if (regs_allocated == REGS_REALLOCATE) | |
540 | { /* Yes. If we need more elements than were already | |
15a7d175 UD |
541 | allocated, reallocate them. If we need fewer, just |
542 | leave it alone. */ | |
951d6408 | 543 | if (BE (need_regs > regs->num_regs, 0)) |
15a7d175 | 544 | { |
44777771 UD |
545 | regoff_t *new_start = re_realloc (regs->start, regoff_t, need_regs); |
546 | regoff_t *new_end = re_realloc (regs->end, regoff_t, need_regs); | |
547 | if (BE (new_start == NULL, 0) || BE (new_end == NULL, 0)) | |
a96c63ed | 548 | return REGS_UNALLOCATED; |
44777771 UD |
549 | regs->start = new_start; |
550 | regs->end = new_end; | |
15a7d175 UD |
551 | regs->num_regs = need_regs; |
552 | } | |
ac3d553b UD |
553 | } |
554 | else | |
555 | { | |
556 | assert (regs_allocated == REGS_FIXED); | |
557 | /* This function may not be called with REGS_FIXED and nregs too big. */ | |
558 | assert (regs->num_regs >= nregs); | |
559 | rval = REGS_FIXED; | |
560 | } | |
561 | ||
562 | /* Copy the regs. */ | |
563 | for (i = 0; i < nregs; ++i) | |
564 | { | |
565 | regs->start[i] = pmatch[i].rm_so; | |
566 | regs->end[i] = pmatch[i].rm_eo; | |
567 | } | |
568 | for ( ; i < regs->num_regs; ++i) | |
569 | regs->start[i] = regs->end[i] = -1; | |
570 | ||
571 | return rval; | |
3b0bdc72 | 572 | } |
3b0bdc72 UD |
573 | |
574 | /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | |
575 | ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use | |
576 | this memory for recording register information. STARTS and ENDS | |
577 | must be allocated using the malloc library routine, and must each | |
578 | be at least NUM_REGS * sizeof (regoff_t) bytes long. | |
579 | ||
580 | If NUM_REGS == 0, then subsequent matches should allocate their own | |
581 | register data. | |
582 | ||
583 | Unless this function is called, the first search or match using | |
584 | PATTERN_BUFFER will allocate its own register data, without | |
585 | freeing the old data. */ | |
586 | ||
587 | void | |
588 | re_set_registers (bufp, regs, num_regs, starts, ends) | |
589 | struct re_pattern_buffer *bufp; | |
590 | struct re_registers *regs; | |
591 | unsigned num_regs; | |
592 | regoff_t *starts, *ends; | |
593 | { | |
594 | if (num_regs) | |
595 | { | |
596 | bufp->regs_allocated = REGS_REALLOCATE; | |
597 | regs->num_regs = num_regs; | |
598 | regs->start = starts; | |
599 | regs->end = ends; | |
600 | } | |
601 | else | |
602 | { | |
603 | bufp->regs_allocated = REGS_UNALLOCATED; | |
604 | regs->num_regs = 0; | |
605 | regs->start = regs->end = (regoff_t *) 0; | |
606 | } | |
607 | } | |
608 | #ifdef _LIBC | |
609 | weak_alias (__re_set_registers, re_set_registers) | |
610 | #endif | |
611 | \f | |
612 | /* Entry points compatible with 4.2 BSD regex library. We don't define | |
613 | them unless specifically requested. */ | |
614 | ||
615 | #if defined _REGEX_RE_COMP || defined _LIBC | |
616 | int | |
617 | # ifdef _LIBC | |
618 | weak_function | |
619 | # endif | |
620 | re_exec (s) | |
621 | const char *s; | |
622 | { | |
623 | return 0 == regexec (&re_comp_buf, s, 0, NULL, 0); | |
624 | } | |
625 | #endif /* _REGEX_RE_COMP */ | |
626 | \f | |
3b0bdc72 UD |
627 | /* Internal entry point. */ |
628 | ||
629 | /* Searches for a compiled pattern PREG in the string STRING, whose | |
630 | length is LENGTH. NMATCH, PMATCH, and EFLAGS have the same | |
631 | mingings with regexec. START, and RANGE have the same meanings | |
632 | with re_search. | |
a9388965 UD |
633 | Return REG_NOERROR if we find a match, and REG_NOMATCH if not, |
634 | otherwise return the error code. | |
3b0bdc72 UD |
635 | Note: We assume front end functions already check ranges. |
636 | (START + RANGE >= 0 && START + RANGE <= LENGTH) */ | |
637 | ||
a9388965 | 638 | static reg_errcode_t |
ac3d553b | 639 | re_search_internal (preg, string, length, start, range, stop, nmatch, pmatch, |
15a7d175 | 640 | eflags) |
3b0bdc72 UD |
641 | const regex_t *preg; |
642 | const char *string; | |
ac3d553b | 643 | int length, start, range, stop, eflags; |
3b0bdc72 UD |
644 | size_t nmatch; |
645 | regmatch_t pmatch[]; | |
646 | { | |
a9388965 | 647 | reg_errcode_t err; |
76b864c8 | 648 | const re_dfa_t *dfa = (const re_dfa_t *) preg->buffer; |
612546c6 | 649 | int left_lim, right_lim, incr; |
bb677c95 | 650 | int fl_longest_match, match_first, match_kind, match_last = -1; |
963d8d78 | 651 | int extra_nmatch; |
bb677c95 | 652 | int sb, ch; |
e3a87852 UD |
653 | #if defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L) |
654 | re_match_context_t mctx = { .dfa = dfa }; | |
655 | #else | |
3b0bdc72 | 656 | re_match_context_t mctx; |
e3a87852 | 657 | #endif |
bb677c95 UD |
658 | char *fastmap = (preg->fastmap != NULL && preg->fastmap_accurate |
659 | && range && !preg->can_be_null) ? preg->fastmap : NULL; | |
997470b3 | 660 | RE_TRANSLATE_TYPE t = preg->translate; |
3b0bdc72 | 661 | |
e3a87852 UD |
662 | #if !(defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L)) |
663 | memset (&mctx, '\0', sizeof (re_match_context_t)); | |
664 | mctx.dfa = dfa; | |
665 | #endif | |
666 | ||
963d8d78 UD |
667 | extra_nmatch = (nmatch > preg->re_nsub) ? nmatch - (preg->re_nsub + 1) : 0; |
668 | nmatch -= extra_nmatch; | |
669 | ||
3b0bdc72 | 670 | /* Check if the DFA haven't been compiled. */ |
bc15410e | 671 | if (BE (preg->used == 0 || dfa->init_state == NULL |
15a7d175 UD |
672 | || dfa->init_state_word == NULL || dfa->init_state_nl == NULL |
673 | || dfa->init_state_begbuf == NULL, 0)) | |
a9388965 | 674 | return REG_NOMATCH; |
3b0bdc72 | 675 | |
c70f81dd UD |
676 | #ifdef DEBUG |
677 | /* We assume front-end functions already check them. */ | |
678 | assert (start + range >= 0 && start + range <= length); | |
679 | #endif | |
680 | ||
681 | /* If initial states with non-begbuf contexts have no elements, | |
682 | the regex must be anchored. If preg->newline_anchor is set, | |
683 | we'll never use init_state_nl, so do not check it. */ | |
684 | if (dfa->init_state->nodes.nelem == 0 | |
685 | && dfa->init_state_word->nodes.nelem == 0 | |
686 | && (dfa->init_state_nl->nodes.nelem == 0 | |
687 | || !preg->newline_anchor)) | |
688 | { | |
689 | if (start != 0 && start + range != 0) | |
690 | return REG_NOMATCH; | |
691 | start = range = 0; | |
692 | } | |
693 | ||
3b0bdc72 | 694 | /* We must check the longest matching, if nmatch > 0. */ |
6291ee3c | 695 | fl_longest_match = (nmatch != 0 || dfa->nbackref); |
3b0bdc72 | 696 | |
56b168be | 697 | err = re_string_allocate (&mctx.input, string, length, dfa->nodes_len + 1, |
f0c7c524 | 698 | preg->translate, preg->syntax & RE_ICASE, dfa); |
612546c6 | 699 | if (BE (err != REG_NOERROR, 0)) |
1b2c2628 | 700 | goto free_return; |
56b168be UD |
701 | mctx.input.stop = stop; |
702 | mctx.input.raw_stop = stop; | |
703 | mctx.input.newline_anchor = preg->newline_anchor; | |
612546c6 | 704 | |
56b168be | 705 | err = match_ctx_init (&mctx, eflags, dfa->nbackref * 2); |
612546c6 | 706 | if (BE (err != REG_NOERROR, 0)) |
1b2c2628 | 707 | goto free_return; |
612546c6 | 708 | |
3b0bdc72 UD |
709 | /* We will log all the DFA states through which the dfa pass, |
710 | if nmatch > 1, or this dfa has "multibyte node", which is a | |
711 | back-reference or a node which can accept multibyte character or | |
712 | multi character collating element. */ | |
713 | if (nmatch > 1 || dfa->has_mb_node) | |
a9388965 | 714 | { |
56b168be | 715 | mctx.state_log = re_malloc (re_dfastate_t *, mctx.input.bufs_len + 1); |
612546c6 | 716 | if (BE (mctx.state_log == NULL, 0)) |
15a7d175 UD |
717 | { |
718 | err = REG_ESPACE; | |
719 | goto free_return; | |
720 | } | |
a9388965 | 721 | } |
3b0bdc72 | 722 | else |
612546c6 | 723 | mctx.state_log = NULL; |
3b0bdc72 | 724 | |
612546c6 | 725 | match_first = start; |
56b168be UD |
726 | mctx.input.tip_context = (eflags & REG_NOTBOL) ? CONTEXT_BEGBUF |
727 | : CONTEXT_NEWLINE | CONTEXT_BEGBUF; | |
612546c6 | 728 | |
3b0bdc72 | 729 | /* Check incrementally whether of not the input string match. */ |
612546c6 UD |
730 | incr = (range < 0) ? -1 : 1; |
731 | left_lim = (range < 0) ? start + range : start; | |
732 | right_lim = (range < 0) ? start : start + range; | |
3c0fb574 | 733 | sb = dfa->mb_cur_max == 1; |
bb677c95 UD |
734 | match_kind = |
735 | (fastmap | |
736 | ? ((sb || !(preg->syntax & RE_ICASE || t) ? 4 : 0) | |
737 | | (range >= 0 ? 2 : 0) | |
738 | | (t != NULL ? 1 : 0)) | |
739 | : 8); | |
740 | ||
741 | for (;; match_first += incr) | |
3b0bdc72 | 742 | { |
bb677c95 UD |
743 | err = REG_NOMATCH; |
744 | if (match_first < left_lim || right_lim < match_first) | |
745 | goto free_return; | |
746 | ||
747 | /* Advance as rapidly as possible through the string, until we | |
748 | find a plausible place to start matching. This may be done | |
749 | with varying efficiency, so there are various possibilities: | |
750 | only the most common of them are specialized, in order to | |
751 | save on code size. We use a switch statement for speed. */ | |
752 | switch (match_kind) | |
1b2c2628 | 753 | { |
bb677c95 UD |
754 | case 8: |
755 | /* No fastmap. */ | |
756 | break; | |
757 | ||
758 | case 7: | |
759 | /* Fastmap with single-byte translation, match forward. */ | |
760 | while (BE (match_first < right_lim, 1) | |
761 | && !fastmap[t[(unsigned char) string[match_first]]]) | |
762 | ++match_first; | |
763 | goto forward_match_found_start_or_reached_end; | |
764 | ||
765 | case 6: | |
766 | /* Fastmap without translation, match forward. */ | |
767 | while (BE (match_first < right_lim, 1) | |
768 | && !fastmap[(unsigned char) string[match_first]]) | |
769 | ++match_first; | |
770 | ||
771 | forward_match_found_start_or_reached_end: | |
772 | if (BE (match_first == right_lim, 0)) | |
1b2c2628 | 773 | { |
bb677c95 UD |
774 | ch = match_first >= length |
775 | ? 0 : (unsigned char) string[match_first]; | |
776 | if (!fastmap[t ? t[ch] : ch]) | |
777 | goto free_return; | |
1b2c2628 | 778 | } |
bb677c95 UD |
779 | break; |
780 | ||
781 | case 4: | |
782 | case 5: | |
783 | /* Fastmap without multi-byte translation, match backwards. */ | |
784 | while (match_first >= left_lim) | |
1b2c2628 | 785 | { |
bb677c95 UD |
786 | ch = match_first >= length |
787 | ? 0 : (unsigned char) string[match_first]; | |
788 | if (fastmap[t ? t[ch] : ch]) | |
789 | break; | |
790 | --match_first; | |
791 | } | |
792 | if (match_first < left_lim) | |
793 | goto free_return; | |
794 | break; | |
1b2c2628 | 795 | |
bb677c95 UD |
796 | default: |
797 | /* In this case, we can't determine easily the current byte, | |
798 | since it might be a component byte of a multibyte | |
799 | character. Then we use the constructed buffer instead. */ | |
800 | for (;;) | |
801 | { | |
802 | /* If MATCH_FIRST is out of the valid range, reconstruct the | |
803 | buffers. */ | |
804 | unsigned int offset = match_first - mctx.input.raw_mbs_idx; | |
805 | if (BE (offset >= (unsigned int) mctx.input.valid_raw_len, 0)) | |
1b2c2628 | 806 | { |
bb677c95 UD |
807 | err = re_string_reconstruct (&mctx.input, match_first, |
808 | eflags); | |
809 | if (BE (err != REG_NOERROR, 0)) | |
810 | goto free_return; | |
811 | ||
812 | offset = match_first - mctx.input.raw_mbs_idx; | |
1b2c2628 | 813 | } |
bb677c95 UD |
814 | /* If MATCH_FIRST is out of the buffer, leave it as '\0'. |
815 | Note that MATCH_FIRST must not be smaller than 0. */ | |
816 | ch = (match_first >= length | |
817 | ? 0 : re_string_byte_at (&mctx.input, offset)); | |
818 | if (fastmap[ch]) | |
1b2c2628 | 819 | break; |
bb677c95 UD |
820 | match_first += incr; |
821 | if (match_first < left_lim || match_first > right_lim) | |
822 | { | |
823 | err = REG_NOMATCH; | |
824 | goto free_return; | |
825 | } | |
1b2c2628 | 826 | } |
bb677c95 | 827 | break; |
1b2c2628 | 828 | } |
612546c6 | 829 | |
1b2c2628 | 830 | /* Reconstruct the buffers so that the matcher can assume that |
fe9434bb | 831 | the matching starts from the beginning of the buffer. */ |
56b168be | 832 | err = re_string_reconstruct (&mctx.input, match_first, eflags); |
1b2c2628 UD |
833 | if (BE (err != REG_NOERROR, 0)) |
834 | goto free_return; | |
bb677c95 | 835 | |
3b0bdc72 | 836 | #ifdef RE_ENABLE_I18N |
bb677c95 UD |
837 | /* Don't consider this char as a possible match start if it part, |
838 | yet isn't the head, of a multibyte character. */ | |
839 | if (!sb && !re_string_first_byte (&mctx.input, 0)) | |
840 | continue; | |
3b0bdc72 | 841 | #endif |
bb677c95 UD |
842 | |
843 | /* It seems to be appropriate one, then use the matcher. */ | |
844 | /* We assume that the matching starts from 0. */ | |
845 | mctx.state_log_top = mctx.nbkref_ents = mctx.max_mb_elem_len = 0; | |
846 | match_last = check_matching (&mctx, fl_longest_match, | |
847 | range >= 0 ? &match_first : NULL); | |
848 | if (match_last != -1) | |
1b2c2628 | 849 | { |
bb677c95 | 850 | if (BE (match_last == -2, 0)) |
1b2c2628 | 851 | { |
bb677c95 UD |
852 | err = REG_ESPACE; |
853 | goto free_return; | |
854 | } | |
855 | else | |
856 | { | |
857 | mctx.match_last = match_last; | |
858 | if ((!preg->no_sub && nmatch > 1) || dfa->nbackref) | |
1b2c2628 | 859 | { |
bb677c95 UD |
860 | re_dfastate_t *pstate = mctx.state_log[match_last]; |
861 | mctx.last_node = check_halt_state_context (&mctx, pstate, | |
862 | match_last); | |
1b2c2628 | 863 | } |
bb677c95 UD |
864 | if ((!preg->no_sub && nmatch > 1 && dfa->has_plural_match) |
865 | || dfa->nbackref) | |
6291ee3c | 866 | { |
bb677c95 UD |
867 | err = prune_impossible_nodes (&mctx); |
868 | if (err == REG_NOERROR) | |
869 | break; | |
870 | if (BE (err != REG_NOMATCH, 0)) | |
871 | goto free_return; | |
872 | match_last = -1; | |
6291ee3c | 873 | } |
bb677c95 UD |
874 | else |
875 | break; /* We found a match. */ | |
1b2c2628 UD |
876 | } |
877 | } | |
bb677c95 UD |
878 | |
879 | match_ctx_clean (&mctx); | |
3b0bdc72 UD |
880 | } |
881 | ||
bb677c95 UD |
882 | #ifdef DEBUG |
883 | assert (match_last != -1); | |
884 | assert (err == REG_NOERROR); | |
885 | #endif | |
886 | ||
3b0bdc72 | 887 | /* Set pmatch[] if we need. */ |
bb677c95 | 888 | if (nmatch > 0) |
3b0bdc72 UD |
889 | { |
890 | int reg_idx; | |
891 | ||
892 | /* Initialize registers. */ | |
97fd3a30 | 893 | for (reg_idx = 1; reg_idx < nmatch; ++reg_idx) |
15a7d175 | 894 | pmatch[reg_idx].rm_so = pmatch[reg_idx].rm_eo = -1; |
3b0bdc72 UD |
895 | |
896 | /* Set the points where matching start/end. */ | |
612546c6 | 897 | pmatch[0].rm_so = 0; |
6291ee3c | 898 | pmatch[0].rm_eo = mctx.match_last; |
3b0bdc72 UD |
899 | |
900 | if (!preg->no_sub && nmatch > 1) | |
15a7d175 | 901 | { |
15a7d175 UD |
902 | err = set_regs (preg, &mctx, nmatch, pmatch, |
903 | dfa->has_plural_match && dfa->nbackref > 0); | |
904 | if (BE (err != REG_NOERROR, 0)) | |
905 | goto free_return; | |
906 | } | |
612546c6 UD |
907 | |
908 | /* At last, add the offset to the each registers, since we slided | |
97fd3a30 UD |
909 | the buffers so that we could assume that the matching starts |
910 | from 0. */ | |
612546c6 | 911 | for (reg_idx = 0; reg_idx < nmatch; ++reg_idx) |
15a7d175 UD |
912 | if (pmatch[reg_idx].rm_so != -1) |
913 | { | |
c0d5034e | 914 | #ifdef RE_ENABLE_I18N |
56b168be | 915 | if (BE (mctx.input.offsets_needed != 0, 0)) |
bb3f4825 | 916 | { |
01ed6ceb UD |
917 | pmatch[reg_idx].rm_so = |
918 | (pmatch[reg_idx].rm_so == mctx.input.valid_len | |
919 | ? mctx.input.valid_raw_len | |
920 | : mctx.input.offsets[pmatch[reg_idx].rm_so]); | |
921 | pmatch[reg_idx].rm_eo = | |
922 | (pmatch[reg_idx].rm_eo == mctx.input.valid_len | |
923 | ? mctx.input.valid_raw_len | |
924 | : mctx.input.offsets[pmatch[reg_idx].rm_eo]); | |
bb3f4825 | 925 | } |
c0d5034e | 926 | #else |
56b168be | 927 | assert (mctx.input.offsets_needed == 0); |
c0d5034e | 928 | #endif |
15a7d175 UD |
929 | pmatch[reg_idx].rm_so += match_first; |
930 | pmatch[reg_idx].rm_eo += match_first; | |
931 | } | |
963d8d78 UD |
932 | for (reg_idx = 0; reg_idx < extra_nmatch; ++reg_idx) |
933 | { | |
934 | pmatch[nmatch + reg_idx].rm_so = -1; | |
935 | pmatch[nmatch + reg_idx].rm_eo = -1; | |
936 | } | |
c06a6956 UD |
937 | |
938 | if (dfa->subexp_map) | |
963d8d78 | 939 | for (reg_idx = 0; reg_idx + 1 < nmatch; reg_idx++) |
c06a6956 UD |
940 | if (dfa->subexp_map[reg_idx] != reg_idx) |
941 | { | |
942 | pmatch[reg_idx + 1].rm_so | |
943 | = pmatch[dfa->subexp_map[reg_idx] + 1].rm_so; | |
944 | pmatch[reg_idx + 1].rm_eo | |
945 | = pmatch[dfa->subexp_map[reg_idx] + 1].rm_eo; | |
946 | } | |
3b0bdc72 | 947 | } |
bb677c95 | 948 | |
1b2c2628 | 949 | free_return: |
612546c6 | 950 | re_free (mctx.state_log); |
3b0bdc72 UD |
951 | if (dfa->nbackref) |
952 | match_ctx_free (&mctx); | |
56b168be | 953 | re_string_destruct (&mctx.input); |
1b2c2628 | 954 | return err; |
3b0bdc72 UD |
955 | } |
956 | ||
6291ee3c | 957 | static reg_errcode_t |
e3a87852 | 958 | prune_impossible_nodes (mctx) |
6291ee3c UD |
959 | re_match_context_t *mctx; |
960 | { | |
76b864c8 | 961 | const re_dfa_t *const dfa = mctx->dfa; |
6291ee3c UD |
962 | int halt_node, match_last; |
963 | reg_errcode_t ret; | |
6291ee3c UD |
964 | re_dfastate_t **sifted_states; |
965 | re_dfastate_t **lim_states = NULL; | |
966 | re_sift_context_t sctx; | |
967 | #ifdef DEBUG | |
968 | assert (mctx->state_log != NULL); | |
969 | #endif | |
970 | match_last = mctx->match_last; | |
971 | halt_node = mctx->last_node; | |
972 | sifted_states = re_malloc (re_dfastate_t *, match_last + 1); | |
973 | if (BE (sifted_states == NULL, 0)) | |
974 | { | |
975 | ret = REG_ESPACE; | |
976 | goto free_return; | |
977 | } | |
978 | if (dfa->nbackref) | |
979 | { | |
980 | lim_states = re_malloc (re_dfastate_t *, match_last + 1); | |
981 | if (BE (lim_states == NULL, 0)) | |
982 | { | |
983 | ret = REG_ESPACE; | |
984 | goto free_return; | |
985 | } | |
986 | while (1) | |
987 | { | |
988 | memset (lim_states, '\0', | |
989 | sizeof (re_dfastate_t *) * (match_last + 1)); | |
6291ee3c | 990 | sift_ctx_init (&sctx, sifted_states, lim_states, halt_node, |
d2c38eb3 | 991 | match_last); |
e3a87852 | 992 | ret = sift_states_backward (mctx, &sctx); |
6291ee3c UD |
993 | re_node_set_free (&sctx.limits); |
994 | if (BE (ret != REG_NOERROR, 0)) | |
995 | goto free_return; | |
996 | if (sifted_states[0] != NULL || lim_states[0] != NULL) | |
997 | break; | |
998 | do | |
999 | { | |
1000 | --match_last; | |
1001 | if (match_last < 0) | |
1002 | { | |
1003 | ret = REG_NOMATCH; | |
1004 | goto free_return; | |
1005 | } | |
97fd3a30 UD |
1006 | } while (mctx->state_log[match_last] == NULL |
1007 | || !mctx->state_log[match_last]->halt); | |
e3a87852 | 1008 | halt_node = check_halt_state_context (mctx, |
6291ee3c | 1009 | mctx->state_log[match_last], |
e3a87852 | 1010 | match_last); |
6291ee3c UD |
1011 | } |
1012 | ret = merge_state_array (dfa, sifted_states, lim_states, | |
1013 | match_last + 1); | |
1014 | re_free (lim_states); | |
1015 | lim_states = NULL; | |
1016 | if (BE (ret != REG_NOERROR, 0)) | |
1017 | goto free_return; | |
1018 | } | |
1019 | else | |
1020 | { | |
d2c38eb3 | 1021 | sift_ctx_init (&sctx, sifted_states, lim_states, halt_node, match_last); |
e3a87852 | 1022 | ret = sift_states_backward (mctx, &sctx); |
6291ee3c UD |
1023 | re_node_set_free (&sctx.limits); |
1024 | if (BE (ret != REG_NOERROR, 0)) | |
1025 | goto free_return; | |
1026 | } | |
1027 | re_free (mctx->state_log); | |
1028 | mctx->state_log = sifted_states; | |
1029 | sifted_states = NULL; | |
1030 | mctx->last_node = halt_node; | |
1031 | mctx->match_last = match_last; | |
1032 | ret = REG_NOERROR; | |
1033 | free_return: | |
1034 | re_free (sifted_states); | |
1035 | re_free (lim_states); | |
1036 | return ret; | |
1037 | } | |
1038 | ||
a9388965 | 1039 | /* Acquire an initial state and return it. |
3b0bdc72 UD |
1040 | We must select appropriate initial state depending on the context, |
1041 | since initial states may have constraints like "\<", "^", etc.. */ | |
1042 | ||
bb3f4825 | 1043 | static inline re_dfastate_t * |
a334319f UD |
1044 | acquire_init_state_context (err, mctx, idx) |
1045 | reg_errcode_t *err; | |
1046 | const re_match_context_t *mctx; | |
1047 | int idx; | |
3b0bdc72 | 1048 | { |
76b864c8 | 1049 | const re_dfa_t *const dfa = mctx->dfa; |
3b0bdc72 UD |
1050 | if (dfa->init_state->has_constraint) |
1051 | { | |
1052 | unsigned int context; | |
56b168be | 1053 | context = re_string_context_at (&mctx->input, idx - 1, mctx->eflags); |
3b0bdc72 | 1054 | if (IS_WORD_CONTEXT (context)) |
15a7d175 | 1055 | return dfa->init_state_word; |
3b0bdc72 | 1056 | else if (IS_ORDINARY_CONTEXT (context)) |
15a7d175 | 1057 | return dfa->init_state; |
3b0bdc72 | 1058 | else if (IS_BEGBUF_CONTEXT (context) && IS_NEWLINE_CONTEXT (context)) |
15a7d175 | 1059 | return dfa->init_state_begbuf; |
3b0bdc72 | 1060 | else if (IS_NEWLINE_CONTEXT (context)) |
15a7d175 | 1061 | return dfa->init_state_nl; |
3b0bdc72 | 1062 | else if (IS_BEGBUF_CONTEXT (context)) |
15a7d175 UD |
1063 | { |
1064 | /* It is relatively rare case, then calculate on demand. */ | |
4c595adb UD |
1065 | return re_acquire_state_context (err, dfa, |
1066 | dfa->init_state->entrance_nodes, | |
1067 | context); | |
15a7d175 | 1068 | } |
3b0bdc72 | 1069 | else |
15a7d175 UD |
1070 | /* Must not happen? */ |
1071 | return dfa->init_state; | |
3b0bdc72 UD |
1072 | } |
1073 | else | |
1074 | return dfa->init_state; | |
1075 | } | |
1076 | ||
1077 | /* Check whether the regular expression match input string INPUT or not, | |
a9388965 UD |
1078 | and return the index where the matching end, return -1 if not match, |
1079 | or return -2 in case of an error. | |
612546c6 | 1080 | FL_LONGEST_MATCH means we want the POSIX longest matching. |
15bad1a5 UD |
1081 | If P_MATCH_FIRST is not NULL, and the match fails, it is set to the |
1082 | next place where we may want to try matching. | |
612546c6 UD |
1083 | Note that the matcher assume that the maching starts from the current |
1084 | index of the buffer. */ | |
3b0bdc72 UD |
1085 | |
1086 | static int | |
a334319f | 1087 | check_matching (mctx, fl_longest_match, p_match_first) |
76b864c8 UD |
1088 | re_match_context_t *mctx; |
1089 | int fl_longest_match; | |
1090 | int *p_match_first; | |
3b0bdc72 | 1091 | { |
76b864c8 | 1092 | const re_dfa_t *const dfa = mctx->dfa; |
a9388965 | 1093 | reg_errcode_t err; |
612546c6 UD |
1094 | int match = 0; |
1095 | int match_last = -1; | |
56b168be | 1096 | int cur_str_idx = re_string_cur_idx (&mctx->input); |
3b0bdc72 | 1097 | re_dfastate_t *cur_state; |
4c595adb UD |
1098 | int at_init_state = p_match_first != NULL; |
1099 | int next_start_idx = cur_str_idx; | |
3b0bdc72 | 1100 | |
4c595adb | 1101 | err = REG_NOERROR; |
e3a87852 | 1102 | cur_state = acquire_init_state_context (&err, mctx, cur_str_idx); |
4c595adb | 1103 | /* An initial state must not be NULL (invalid). */ |
bc15410e | 1104 | if (BE (cur_state == NULL, 0)) |
4c595adb UD |
1105 | { |
1106 | assert (err == REG_ESPACE); | |
1107 | return -2; | |
1108 | } | |
0742e48e | 1109 | |
4c595adb | 1110 | if (mctx->state_log != NULL) |
6291ee3c | 1111 | { |
4c595adb | 1112 | mctx->state_log[cur_str_idx] = cur_state; |
6291ee3c | 1113 | |
4c595adb UD |
1114 | /* Check OP_OPEN_SUBEXP in the initial state in case that we use them |
1115 | later. E.g. Processing back references. */ | |
1116 | if (BE (dfa->nbackref, 0)) | |
bb3f4825 | 1117 | { |
4c595adb UD |
1118 | at_init_state = 0; |
1119 | err = check_subexp_matching_top (mctx, &cur_state->nodes, 0); | |
bb3f4825 UD |
1120 | if (BE (err != REG_NOERROR, 0)) |
1121 | return err; | |
4c595adb UD |
1122 | |
1123 | if (cur_state->has_backref) | |
1124 | { | |
1125 | err = transit_state_bkref (mctx, &cur_state->nodes); | |
1126 | if (BE (err != REG_NOERROR, 0)) | |
1127 | return err; | |
1128 | } | |
bb3f4825 | 1129 | } |
0742e48e UD |
1130 | } |
1131 | ||
3b0bdc72 | 1132 | /* If the RE accepts NULL string. */ |
bb3f4825 | 1133 | if (BE (cur_state->halt, 0)) |
3b0bdc72 UD |
1134 | { |
1135 | if (!cur_state->has_constraint | |
e3a87852 | 1136 | || check_halt_state_context (mctx, cur_state, cur_str_idx)) |
15a7d175 UD |
1137 | { |
1138 | if (!fl_longest_match) | |
1139 | return cur_str_idx; | |
1140 | else | |
1141 | { | |
1142 | match_last = cur_str_idx; | |
1143 | match = 1; | |
1144 | } | |
1145 | } | |
3b0bdc72 UD |
1146 | } |
1147 | ||
56b168be | 1148 | while (!re_string_eoi (&mctx->input)) |
3b0bdc72 | 1149 | { |
15bad1a5 | 1150 | re_dfastate_t *old_state = cur_state; |
bb677c95 UD |
1151 | int next_char_idx = re_string_cur_idx (&mctx->input) + 1; |
1152 | ||
1153 | if (BE (next_char_idx >= mctx->input.bufs_len, 0) | |
1154 | || (BE (next_char_idx >= mctx->input.valid_len, 0) | |
1155 | && mctx->input.valid_len < mctx->input.len)) | |
1156 | { | |
1157 | err = extend_buffers (mctx); | |
1158 | if (BE (err != REG_NOERROR, 0)) | |
1159 | { | |
1160 | assert (err == REG_ESPACE); | |
1161 | return -2; | |
1162 | } | |
1163 | } | |
1164 | ||
e3a87852 | 1165 | cur_state = transit_state (&err, mctx, cur_state); |
4c595adb UD |
1166 | if (mctx->state_log != NULL) |
1167 | cur_state = merge_state_with_log (&err, mctx, cur_state); | |
15bad1a5 | 1168 | |
4c595adb | 1169 | if (cur_state == NULL) |
15a7d175 | 1170 | { |
4c595adb UD |
1171 | /* Reached the invalid state or an error. Try to recover a valid |
1172 | state using the state log, if available and if we have not | |
1173 | already found a valid (even if not the longest) match. */ | |
15a7d175 UD |
1174 | if (BE (err != REG_NOERROR, 0)) |
1175 | return -2; | |
4c595adb UD |
1176 | |
1177 | if (mctx->state_log == NULL | |
1178 | || (match && !fl_longest_match) | |
1179 | || (cur_state = find_recover_state (&err, mctx)) == NULL) | |
15a7d175 | 1180 | break; |
4c595adb UD |
1181 | } |
1182 | ||
bb677c95 | 1183 | if (BE (at_init_state, 0)) |
4c595adb UD |
1184 | { |
1185 | if (old_state == cur_state) | |
bb677c95 | 1186 | next_start_idx = next_char_idx; |
c13c99fa | 1187 | else |
4c595adb | 1188 | at_init_state = 0; |
15a7d175 | 1189 | } |
3b0bdc72 | 1190 | |
4c595adb | 1191 | if (cur_state->halt) |
15a7d175 | 1192 | { |
4c595adb | 1193 | /* Reached a halt state. |
15a7d175 UD |
1194 | Check the halt state can satisfy the current context. */ |
1195 | if (!cur_state->has_constraint | |
e3a87852 | 1196 | || check_halt_state_context (mctx, cur_state, |
56b168be | 1197 | re_string_cur_idx (&mctx->input))) |
15a7d175 UD |
1198 | { |
1199 | /* We found an appropriate halt state. */ | |
56b168be | 1200 | match_last = re_string_cur_idx (&mctx->input); |
15a7d175 | 1201 | match = 1; |
bb677c95 UD |
1202 | |
1203 | /* We found a match, do not modify match_first below. */ | |
1204 | p_match_first = NULL; | |
15a7d175 UD |
1205 | if (!fl_longest_match) |
1206 | break; | |
1207 | } | |
1208 | } | |
bb677c95 | 1209 | } |
15bad1a5 | 1210 | |
bb677c95 | 1211 | if (p_match_first) |
4c595adb | 1212 | *p_match_first += next_start_idx; |
15bad1a5 | 1213 | |
3b0bdc72 UD |
1214 | return match_last; |
1215 | } | |
1216 | ||
1217 | /* Check NODE match the current context. */ | |
1218 | ||
a334319f UD |
1219 | static int check_halt_node_context (dfa, node, context) |
1220 | const re_dfa_t *dfa; | |
1221 | int node; | |
1222 | unsigned int context; | |
3b0bdc72 | 1223 | { |
3b0bdc72 | 1224 | re_token_type_t type = dfa->nodes[node].type; |
485d775d UD |
1225 | unsigned int constraint = dfa->nodes[node].constraint; |
1226 | if (type != END_OF_RE) | |
3b0bdc72 | 1227 | return 0; |
485d775d UD |
1228 | if (!constraint) |
1229 | return 1; | |
1230 | if (NOT_SATISFY_NEXT_CONSTRAINT (constraint, context)) | |
3b0bdc72 UD |
1231 | return 0; |
1232 | return 1; | |
1233 | } | |
1234 | ||
1235 | /* Check the halt state STATE match the current context. | |
1236 | Return 0 if not match, if the node, STATE has, is a halt node and | |
1237 | match the context, return the node. */ | |
1238 | ||
1239 | static int | |
a334319f UD |
1240 | check_halt_state_context (mctx, state, idx) |
1241 | const re_match_context_t *mctx; | |
1242 | const re_dfastate_t *state; | |
1243 | int idx; | |
3b0bdc72 | 1244 | { |
3b0bdc72 UD |
1245 | int i; |
1246 | unsigned int context; | |
1247 | #ifdef DEBUG | |
1248 | assert (state->halt); | |
1249 | #endif | |
56b168be | 1250 | context = re_string_context_at (&mctx->input, idx, mctx->eflags); |
3b0bdc72 | 1251 | for (i = 0; i < state->nodes.nelem; ++i) |
e3a87852 | 1252 | if (check_halt_node_context (mctx->dfa, state->nodes.elems[i], context)) |
3b0bdc72 UD |
1253 | return state->nodes.elems[i]; |
1254 | return 0; | |
1255 | } | |
1256 | ||
a9388965 UD |
1257 | /* Compute the next node to which "NFA" transit from NODE("NFA" is a NFA |
1258 | corresponding to the DFA). | |
1259 | Return the destination node, and update EPS_VIA_NODES, return -1 in case | |
1260 | of errors. */ | |
3b0bdc72 UD |
1261 | |
1262 | static int | |
a334319f UD |
1263 | proceed_next_node (mctx, nregs, regs, pidx, node, eps_via_nodes, fs) |
1264 | const re_match_context_t *mctx; | |
1265 | regmatch_t *regs; | |
1266 | int nregs, *pidx, node; | |
1267 | re_node_set *eps_via_nodes; | |
1268 | struct re_fail_stack_t *fs; | |
3b0bdc72 | 1269 | { |
76b864c8 | 1270 | const re_dfa_t *const dfa = mctx->dfa; |
2d87db5b | 1271 | int i, err; |
3b0bdc72 UD |
1272 | if (IS_EPSILON_NODE (dfa->nodes[node].type)) |
1273 | { | |
485d775d | 1274 | re_node_set *cur_nodes = &mctx->state_log[*pidx]->nodes; |
5cf1ec52 UD |
1275 | re_node_set *edests = &dfa->edests[node]; |
1276 | int dest_node; | |
a9388965 | 1277 | err = re_node_set_insert (eps_via_nodes, node); |
bc15410e | 1278 | if (BE (err < 0, 0)) |
44777771 | 1279 | return -2; |
5cf1ec52 UD |
1280 | /* Pick up a valid destination, or return -1 if none is found. */ |
1281 | for (dest_node = -1, i = 0; i < edests->nelem; ++i) | |
15a7d175 | 1282 | { |
5cf1ec52 | 1283 | int candidate = edests->elems[i]; |
15a7d175 UD |
1284 | if (!re_node_set_contains (cur_nodes, candidate)) |
1285 | continue; | |
5cf1ec52 UD |
1286 | if (dest_node == -1) |
1287 | dest_node = candidate; | |
1288 | ||
1289 | else | |
1290 | { | |
1291 | /* In order to avoid infinite loop like "(a*)*", return the second | |
1292 | epsilon-transition if the first was already considered. */ | |
1293 | if (re_node_set_contains (eps_via_nodes, dest_node)) | |
1294 | return candidate; | |
1295 | ||
1296 | /* Otherwise, push the second epsilon-transition on the fail stack. */ | |
1297 | else if (fs != NULL | |
1298 | && push_fail_stack (fs, *pidx, candidate, nregs, regs, | |
1299 | eps_via_nodes)) | |
1300 | return -2; | |
1301 | ||
1302 | /* We know we are going to exit. */ | |
1303 | break; | |
1304 | } | |
1305 | } | |
1306 | return dest_node; | |
3b0bdc72 UD |
1307 | } |
1308 | else | |
1309 | { | |
485d775d | 1310 | int naccepted = 0; |
3b0bdc72 | 1311 | re_token_type_t type = dfa->nodes[node].type; |
3b0bdc72 | 1312 | |
434d3784 | 1313 | #ifdef RE_ENABLE_I18N |
02f3550c | 1314 | if (dfa->nodes[node].accept_mb) |
56b168be | 1315 | naccepted = check_node_accept_bytes (dfa, node, &mctx->input, *pidx); |
434d3784 UD |
1316 | else |
1317 | #endif /* RE_ENABLE_I18N */ | |
1318 | if (type == OP_BACK_REF) | |
15a7d175 | 1319 | { |
d8f73de8 | 1320 | int subexp_idx = dfa->nodes[node].opr.idx + 1; |
15a7d175 UD |
1321 | naccepted = regs[subexp_idx].rm_eo - regs[subexp_idx].rm_so; |
1322 | if (fs != NULL) | |
1323 | { | |
1324 | if (regs[subexp_idx].rm_so == -1 || regs[subexp_idx].rm_eo == -1) | |
1325 | return -1; | |
1326 | else if (naccepted) | |
1327 | { | |
56b168be | 1328 | char *buf = (char *) re_string_get_buffer (&mctx->input); |
6291ee3c UD |
1329 | if (memcmp (buf + regs[subexp_idx].rm_so, buf + *pidx, |
1330 | naccepted) != 0) | |
15a7d175 UD |
1331 | return -1; |
1332 | } | |
1333 | } | |
1334 | ||
1335 | if (naccepted == 0) | |
1336 | { | |
2d87db5b | 1337 | int dest_node; |
15a7d175 UD |
1338 | err = re_node_set_insert (eps_via_nodes, node); |
1339 | if (BE (err < 0, 0)) | |
1340 | return -2; | |
1341 | dest_node = dfa->edests[node].elems[0]; | |
1342 | if (re_node_set_contains (&mctx->state_log[*pidx]->nodes, | |
1343 | dest_node)) | |
1344 | return dest_node; | |
1345 | } | |
1346 | } | |
3b0bdc72 UD |
1347 | |
1348 | if (naccepted != 0 | |
e3a87852 | 1349 | || check_node_accept (mctx, dfa->nodes + node, *pidx)) |
15a7d175 | 1350 | { |
2d87db5b | 1351 | int dest_node = dfa->nexts[node]; |
15a7d175 UD |
1352 | *pidx = (naccepted == 0) ? *pidx + 1 : *pidx + naccepted; |
1353 | if (fs && (*pidx > mctx->match_last || mctx->state_log[*pidx] == NULL | |
1354 | || !re_node_set_contains (&mctx->state_log[*pidx]->nodes, | |
1355 | dest_node))) | |
1356 | return -1; | |
1357 | re_node_set_empty (eps_via_nodes); | |
1358 | return dest_node; | |
1359 | } | |
3b0bdc72 | 1360 | } |
a3022b82 UD |
1361 | return -1; |
1362 | } | |
1363 | ||
1364 | static reg_errcode_t | |
a334319f UD |
1365 | push_fail_stack (fs, str_idx, dest_node, nregs, regs, eps_via_nodes) |
1366 | struct re_fail_stack_t *fs; | |
1367 | int str_idx, dest_node, nregs; | |
1368 | regmatch_t *regs; | |
1369 | re_node_set *eps_via_nodes; | |
a3022b82 UD |
1370 | { |
1371 | reg_errcode_t err; | |
1372 | int num = fs->num++; | |
1373 | if (fs->num == fs->alloc) | |
1374 | { | |
1b2c2628 | 1375 | struct re_fail_stack_ent_t *new_array; |
1b2c2628 | 1376 | new_array = realloc (fs->stack, (sizeof (struct re_fail_stack_ent_t) |
44777771 | 1377 | * fs->alloc * 2)); |
1b2c2628 | 1378 | if (new_array == NULL) |
15a7d175 | 1379 | return REG_ESPACE; |
44777771 | 1380 | fs->alloc *= 2; |
1b2c2628 | 1381 | fs->stack = new_array; |
a3022b82 UD |
1382 | } |
1383 | fs->stack[num].idx = str_idx; | |
5cf1ec52 | 1384 | fs->stack[num].node = dest_node; |
a3022b82 | 1385 | fs->stack[num].regs = re_malloc (regmatch_t, nregs); |
5a299c96 UD |
1386 | if (fs->stack[num].regs == NULL) |
1387 | return REG_ESPACE; | |
a3022b82 UD |
1388 | memcpy (fs->stack[num].regs, regs, sizeof (regmatch_t) * nregs); |
1389 | err = re_node_set_init_copy (&fs->stack[num].eps_via_nodes, eps_via_nodes); | |
1390 | return err; | |
1391 | } | |
1b2c2628 | 1392 | |
a3022b82 | 1393 | static int |
a334319f UD |
1394 | pop_fail_stack (fs, pidx, nregs, regs, eps_via_nodes) |
1395 | struct re_fail_stack_t *fs; | |
1396 | int *pidx, nregs; | |
1397 | regmatch_t *regs; | |
1398 | re_node_set *eps_via_nodes; | |
a3022b82 UD |
1399 | { |
1400 | int num = --fs->num; | |
1401 | assert (num >= 0); | |
44777771 | 1402 | *pidx = fs->stack[num].idx; |
a3022b82 UD |
1403 | memcpy (regs, fs->stack[num].regs, sizeof (regmatch_t) * nregs); |
1404 | re_node_set_free (eps_via_nodes); | |
485d775d | 1405 | re_free (fs->stack[num].regs); |
a3022b82 UD |
1406 | *eps_via_nodes = fs->stack[num].eps_via_nodes; |
1407 | return fs->stack[num].node; | |
3b0bdc72 UD |
1408 | } |
1409 | ||
1410 | /* Set the positions where the subexpressions are starts/ends to registers | |
1411 | PMATCH. | |
1412 | Note: We assume that pmatch[0] is already set, and | |
97fd3a30 | 1413 | pmatch[i].rm_so == pmatch[i].rm_eo == -1 for 0 < i < nmatch. */ |
3b0bdc72 | 1414 | |
a9388965 | 1415 | static reg_errcode_t |
a334319f UD |
1416 | set_regs (preg, mctx, nmatch, pmatch, fl_backtrack) |
1417 | const regex_t *preg; | |
1418 | const re_match_context_t *mctx; | |
1419 | size_t nmatch; | |
1420 | regmatch_t *pmatch; | |
1421 | int fl_backtrack; | |
3b0bdc72 | 1422 | { |
76b864c8 | 1423 | const re_dfa_t *dfa = (const re_dfa_t *) preg->buffer; |
963d8d78 | 1424 | int idx, cur_node; |
3b0bdc72 | 1425 | re_node_set eps_via_nodes; |
a3022b82 | 1426 | struct re_fail_stack_t *fs; |
6b6557e8 UD |
1427 | struct re_fail_stack_t fs_body = { 0, 2, NULL }; |
1428 | regmatch_t *prev_idx_match; | |
2d87db5b | 1429 | int prev_idx_match_malloced = 0; |
6b6557e8 | 1430 | |
3b0bdc72 UD |
1431 | #ifdef DEBUG |
1432 | assert (nmatch > 1); | |
612546c6 | 1433 | assert (mctx->state_log != NULL); |
3b0bdc72 | 1434 | #endif |
a3022b82 UD |
1435 | if (fl_backtrack) |
1436 | { | |
1437 | fs = &fs_body; | |
1438 | fs->stack = re_malloc (struct re_fail_stack_ent_t, fs->alloc); | |
6b6557e8 UD |
1439 | if (fs->stack == NULL) |
1440 | return REG_ESPACE; | |
a3022b82 UD |
1441 | } |
1442 | else | |
1443 | fs = NULL; | |
6b6557e8 | 1444 | |
3b0bdc72 | 1445 | cur_node = dfa->init_node; |
3b0bdc72 | 1446 | re_node_set_init_empty (&eps_via_nodes); |
6b6557e8 | 1447 | |
2d87db5b UD |
1448 | if (__libc_use_alloca (nmatch * sizeof (regmatch_t))) |
1449 | prev_idx_match = (regmatch_t *) alloca (nmatch * sizeof (regmatch_t)); | |
1450 | else | |
1451 | { | |
1452 | prev_idx_match = re_malloc (regmatch_t, nmatch); | |
1453 | if (prev_idx_match == NULL) | |
1454 | { | |
1455 | free_fail_stack_return (fs); | |
1456 | return REG_ESPACE; | |
1457 | } | |
1458 | prev_idx_match_malloced = 1; | |
1459 | } | |
963d8d78 | 1460 | memcpy (prev_idx_match, pmatch, sizeof (regmatch_t) * nmatch); |
6b6557e8 | 1461 | |
3b0bdc72 UD |
1462 | for (idx = pmatch[0].rm_so; idx <= pmatch[0].rm_eo ;) |
1463 | { | |
963d8d78 | 1464 | update_regs (dfa, pmatch, prev_idx_match, cur_node, idx, nmatch); |
6b6557e8 | 1465 | |
a3022b82 | 1466 | if (idx == pmatch[0].rm_eo && cur_node == mctx->last_node) |
15a7d175 UD |
1467 | { |
1468 | int reg_idx; | |
1469 | if (fs) | |
1470 | { | |
1471 | for (reg_idx = 0; reg_idx < nmatch; ++reg_idx) | |
1472 | if (pmatch[reg_idx].rm_so > -1 && pmatch[reg_idx].rm_eo == -1) | |
1473 | break; | |
1474 | if (reg_idx == nmatch) | |
1475 | { | |
1476 | re_node_set_free (&eps_via_nodes); | |
2d87db5b UD |
1477 | if (prev_idx_match_malloced) |
1478 | re_free (prev_idx_match); | |
15a7d175 UD |
1479 | return free_fail_stack_return (fs); |
1480 | } | |
1481 | cur_node = pop_fail_stack (fs, &idx, nmatch, pmatch, | |
1482 | &eps_via_nodes); | |
1483 | } | |
1484 | else | |
1485 | { | |
1486 | re_node_set_free (&eps_via_nodes); | |
2d87db5b UD |
1487 | if (prev_idx_match_malloced) |
1488 | re_free (prev_idx_match); | |
15a7d175 UD |
1489 | return REG_NOERROR; |
1490 | } | |
1491 | } | |
3b0bdc72 UD |
1492 | |
1493 | /* Proceed to next node. */ | |
e3a87852 | 1494 | cur_node = proceed_next_node (mctx, nmatch, pmatch, &idx, cur_node, |
15a7d175 | 1495 | &eps_via_nodes, fs); |
a3022b82 | 1496 | |
bc15410e | 1497 | if (BE (cur_node < 0, 0)) |
15a7d175 | 1498 | { |
44777771 UD |
1499 | if (BE (cur_node == -2, 0)) |
1500 | { | |
1501 | re_node_set_free (&eps_via_nodes); | |
2d87db5b UD |
1502 | if (prev_idx_match_malloced) |
1503 | re_free (prev_idx_match); | |
44777771 UD |
1504 | free_fail_stack_return (fs); |
1505 | return REG_ESPACE; | |
1506 | } | |
15a7d175 UD |
1507 | if (fs) |
1508 | cur_node = pop_fail_stack (fs, &idx, nmatch, pmatch, | |
1509 | &eps_via_nodes); | |
1510 | else | |
1511 | { | |
1512 | re_node_set_free (&eps_via_nodes); | |
2d87db5b UD |
1513 | if (prev_idx_match_malloced) |
1514 | re_free (prev_idx_match); | |
15a7d175 UD |
1515 | return REG_NOMATCH; |
1516 | } | |
1517 | } | |
3b0bdc72 UD |
1518 | } |
1519 | re_node_set_free (&eps_via_nodes); | |
2d87db5b UD |
1520 | if (prev_idx_match_malloced) |
1521 | re_free (prev_idx_match); | |
485d775d UD |
1522 | return free_fail_stack_return (fs); |
1523 | } | |
1524 | ||
1525 | static reg_errcode_t | |
a334319f UD |
1526 | free_fail_stack_return (fs) |
1527 | struct re_fail_stack_t *fs; | |
485d775d UD |
1528 | { |
1529 | if (fs) | |
1530 | { | |
1531 | int fs_idx; | |
1532 | for (fs_idx = 0; fs_idx < fs->num; ++fs_idx) | |
15a7d175 UD |
1533 | { |
1534 | re_node_set_free (&fs->stack[fs_idx].eps_via_nodes); | |
1535 | re_free (fs->stack[fs_idx].regs); | |
1536 | } | |
485d775d UD |
1537 | re_free (fs->stack); |
1538 | } | |
a9388965 | 1539 | return REG_NOERROR; |
3b0bdc72 UD |
1540 | } |
1541 | ||
81c64d40 | 1542 | static void |
a334319f | 1543 | update_regs (dfa, pmatch, prev_idx_match, cur_node, cur_idx, nmatch) |
76b864c8 | 1544 | const re_dfa_t *dfa; |
a334319f UD |
1545 | regmatch_t *pmatch, *prev_idx_match; |
1546 | int cur_node, cur_idx, nmatch; | |
81c64d40 UD |
1547 | { |
1548 | int type = dfa->nodes[cur_node].type; | |
81c64d40 UD |
1549 | if (type == OP_OPEN_SUBEXP) |
1550 | { | |
6b6557e8 UD |
1551 | int reg_num = dfa->nodes[cur_node].opr.idx + 1; |
1552 | ||
81c64d40 | 1553 | /* We are at the first node of this sub expression. */ |
6b6557e8 UD |
1554 | if (reg_num < nmatch) |
1555 | { | |
1556 | pmatch[reg_num].rm_so = cur_idx; | |
1557 | pmatch[reg_num].rm_eo = -1; | |
1558 | } | |
81c64d40 UD |
1559 | } |
1560 | else if (type == OP_CLOSE_SUBEXP) | |
6b6557e8 UD |
1561 | { |
1562 | int reg_num = dfa->nodes[cur_node].opr.idx + 1; | |
1563 | if (reg_num < nmatch) | |
1564 | { | |
1565 | /* We are at the last node of this sub expression. */ | |
1566 | if (pmatch[reg_num].rm_so < cur_idx) | |
1567 | { | |
1568 | pmatch[reg_num].rm_eo = cur_idx; | |
1569 | /* This is a non-empty match or we are not inside an optional | |
1570 | subexpression. Accept this right away. */ | |
1571 | memcpy (prev_idx_match, pmatch, sizeof (regmatch_t) * nmatch); | |
1572 | } | |
1573 | else | |
1574 | { | |
1575 | if (dfa->nodes[cur_node].opt_subexp | |
1576 | && prev_idx_match[reg_num].rm_so != -1) | |
1577 | /* We transited through an empty match for an optional | |
1578 | subexpression, like (a?)*, and this is not the subexp's | |
1579 | first match. Copy back the old content of the registers | |
1580 | so that matches of an inner subexpression are undone as | |
1581 | well, like in ((a?))*. */ | |
1582 | memcpy (pmatch, prev_idx_match, sizeof (regmatch_t) * nmatch); | |
1583 | else | |
1584 | /* We completed a subexpression, but it may be part of | |
1585 | an optional one, so do not update PREV_IDX_MATCH. */ | |
1586 | pmatch[reg_num].rm_eo = cur_idx; | |
1587 | } | |
1588 | } | |
1589 | } | |
0742e48e | 1590 | } |
81c64d40 | 1591 | |
0742e48e | 1592 | /* This function checks the STATE_LOG from the SCTX->last_str_idx to 0 |
612546c6 UD |
1593 | and sift the nodes in each states according to the following rules. |
1594 | Updated state_log will be wrote to STATE_LOG. | |
3b0bdc72 UD |
1595 | |
1596 | Rules: We throw away the Node `a' in the STATE_LOG[STR_IDX] if... | |
1597 | 1. When STR_IDX == MATCH_LAST(the last index in the state_log): | |
15a7d175 UD |
1598 | If `a' isn't the LAST_NODE and `a' can't epsilon transit to |
1599 | the LAST_NODE, we throw away the node `a'. | |
612546c6 | 1600 | 2. When 0 <= STR_IDX < MATCH_LAST and `a' accepts |
15a7d175 UD |
1601 | string `s' and transit to `b': |
1602 | i. If 'b' isn't in the STATE_LOG[STR_IDX+strlen('s')], we throw | |
1603 | away the node `a'. | |
1604 | ii. If 'b' is in the STATE_LOG[STR_IDX+strlen('s')] but 'b' is | |
bb3f4825 | 1605 | thrown away, we throw away the node `a'. |
d0b96fc4 | 1606 | 3. When 0 <= STR_IDX < MATCH_LAST and 'a' epsilon transit to 'b': |
15a7d175 UD |
1607 | i. If 'b' isn't in the STATE_LOG[STR_IDX], we throw away the |
1608 | node `a'. | |
bb3f4825 | 1609 | ii. If 'b' is in the STATE_LOG[STR_IDX] but 'b' is thrown away, |
15a7d175 | 1610 | we throw away the node `a'. */ |
3b0bdc72 UD |
1611 | |
1612 | #define STATE_NODE_CONTAINS(state,node) \ | |
1613 | ((state) != NULL && re_node_set_contains (&(state)->nodes, node)) | |
1614 | ||
a9388965 | 1615 | static reg_errcode_t |
a334319f | 1616 | sift_states_backward (mctx, sctx) |
76b864c8 | 1617 | const re_match_context_t *mctx; |
a334319f | 1618 | re_sift_context_t *sctx; |
3b0bdc72 | 1619 | { |
a9388965 | 1620 | reg_errcode_t err; |
0742e48e UD |
1621 | int null_cnt = 0; |
1622 | int str_idx = sctx->last_str_idx; | |
1623 | re_node_set cur_dest; | |
3b0bdc72 UD |
1624 | |
1625 | #ifdef DEBUG | |
612546c6 | 1626 | assert (mctx->state_log != NULL && mctx->state_log[str_idx] != NULL); |
3b0bdc72 | 1627 | #endif |
3b0bdc72 UD |
1628 | |
1629 | /* Build sifted state_log[str_idx]. It has the nodes which can epsilon | |
1630 | transit to the last_node and the last_node itself. */ | |
0742e48e | 1631 | err = re_node_set_init_1 (&cur_dest, sctx->last_node); |
bc15410e | 1632 | if (BE (err != REG_NOERROR, 0)) |
a9388965 | 1633 | return err; |
e3a87852 | 1634 | err = update_cur_sifted_state (mctx, sctx, str_idx, &cur_dest); |
0742e48e | 1635 | if (BE (err != REG_NOERROR, 0)) |
1b2c2628 | 1636 | goto free_return; |
3b0bdc72 UD |
1637 | |
1638 | /* Then check each states in the state_log. */ | |
612546c6 | 1639 | while (str_idx > 0) |
3b0bdc72 | 1640 | { |
3b0bdc72 | 1641 | /* Update counters. */ |
0742e48e UD |
1642 | null_cnt = (sctx->sifted_states[str_idx] == NULL) ? null_cnt + 1 : 0; |
1643 | if (null_cnt > mctx->max_mb_elem_len) | |
15a7d175 UD |
1644 | { |
1645 | memset (sctx->sifted_states, '\0', | |
1646 | sizeof (re_dfastate_t *) * str_idx); | |
1647 | re_node_set_free (&cur_dest); | |
1648 | return REG_NOERROR; | |
1649 | } | |
0742e48e | 1650 | re_node_set_empty (&cur_dest); |
3b0bdc72 | 1651 | --str_idx; |
15a7d175 | 1652 | |
e40a38b3 UD |
1653 | if (mctx->state_log[str_idx]) |
1654 | { | |
1655 | err = build_sifted_states (mctx, sctx, str_idx, &cur_dest); | |
1656 | if (BE (err != REG_NOERROR, 0)) | |
1657 | goto free_return; | |
15a7d175 | 1658 | } |
3b0bdc72 | 1659 | |
0742e48e | 1660 | /* Add all the nodes which satisfy the following conditions: |
15a7d175 UD |
1661 | - It can epsilon transit to a node in CUR_DEST. |
1662 | - It is in CUR_SRC. | |
1663 | And update state_log. */ | |
e3a87852 | 1664 | err = update_cur_sifted_state (mctx, sctx, str_idx, &cur_dest); |
0742e48e | 1665 | if (BE (err != REG_NOERROR, 0)) |
15a7d175 | 1666 | goto free_return; |
3b0bdc72 | 1667 | } |
1b2c2628 UD |
1668 | err = REG_NOERROR; |
1669 | free_return: | |
0742e48e | 1670 | re_node_set_free (&cur_dest); |
1b2c2628 | 1671 | return err; |
3b0bdc72 UD |
1672 | } |
1673 | ||
e40a38b3 | 1674 | static reg_errcode_t |
a334319f | 1675 | build_sifted_states (mctx, sctx, str_idx, cur_dest) |
76b864c8 | 1676 | const re_match_context_t *mctx; |
a334319f UD |
1677 | re_sift_context_t *sctx; |
1678 | int str_idx; | |
1679 | re_node_set *cur_dest; | |
e40a38b3 | 1680 | { |
76b864c8 UD |
1681 | const re_dfa_t *const dfa = mctx->dfa; |
1682 | const re_node_set *cur_src = &mctx->state_log[str_idx]->non_eps_nodes; | |
e40a38b3 UD |
1683 | int i; |
1684 | ||
1685 | /* Then build the next sifted state. | |
1686 | We build the next sifted state on `cur_dest', and update | |
1687 | `sifted_states[str_idx]' with `cur_dest'. | |
1688 | Note: | |
1689 | `cur_dest' is the sifted state from `state_log[str_idx + 1]'. | |
5cf1ec52 UD |
1690 | `cur_src' points the node_set of the old `state_log[str_idx]' |
1691 | (with the epsilon nodes pre-filtered out). */ | |
e40a38b3 UD |
1692 | for (i = 0; i < cur_src->nelem; i++) |
1693 | { | |
1694 | int prev_node = cur_src->elems[i]; | |
1695 | int naccepted = 0; | |
e40a38b3 UD |
1696 | int ret; |
1697 | ||
a334319f | 1698 | #ifdef DEBUG |
02f3550c | 1699 | re_token_type_t type = dfa->nodes[prev_node].type; |
5cf1ec52 UD |
1700 | assert (!IS_EPSILON_NODE (type)); |
1701 | #endif | |
e40a38b3 UD |
1702 | #ifdef RE_ENABLE_I18N |
1703 | /* If the node may accept `multi byte'. */ | |
02f3550c | 1704 | if (dfa->nodes[prev_node].accept_mb) |
e40a38b3 UD |
1705 | naccepted = sift_states_iter_mb (mctx, sctx, prev_node, |
1706 | str_idx, sctx->last_str_idx); | |
1707 | #endif /* RE_ENABLE_I18N */ | |
1708 | ||
1709 | /* We don't check backreferences here. | |
1710 | See update_cur_sifted_state(). */ | |
1711 | if (!naccepted | |
1712 | && check_node_accept (mctx, dfa->nodes + prev_node, str_idx) | |
1713 | && STATE_NODE_CONTAINS (sctx->sifted_states[str_idx + 1], | |
1714 | dfa->nexts[prev_node])) | |
1715 | naccepted = 1; | |
1716 | ||
1717 | if (naccepted == 0) | |
1718 | continue; | |
1719 | ||
1720 | if (sctx->limits.nelem) | |
1721 | { | |
1722 | int to_idx = str_idx + naccepted; | |
1723 | if (check_dst_limits (mctx, &sctx->limits, | |
1724 | dfa->nexts[prev_node], to_idx, | |
1725 | prev_node, str_idx)) | |
1726 | continue; | |
1727 | } | |
1728 | ret = re_node_set_insert (cur_dest, prev_node); | |
1729 | if (BE (ret == -1, 0)) | |
1730 | return REG_ESPACE; | |
1731 | } | |
1732 | ||
1733 | return REG_NOERROR; | |
1734 | } | |
1735 | ||
3b0bdc72 UD |
1736 | /* Helper functions. */ |
1737 | ||
25337753 | 1738 | static reg_errcode_t |
a334319f UD |
1739 | clean_state_log_if_needed (mctx, next_state_log_idx) |
1740 | re_match_context_t *mctx; | |
1741 | int next_state_log_idx; | |
3b0bdc72 UD |
1742 | { |
1743 | int top = mctx->state_log_top; | |
612546c6 | 1744 | |
56b168be UD |
1745 | if (next_state_log_idx >= mctx->input.bufs_len |
1746 | || (next_state_log_idx >= mctx->input.valid_len | |
1747 | && mctx->input.valid_len < mctx->input.len)) | |
612546c6 UD |
1748 | { |
1749 | reg_errcode_t err; | |
1750 | err = extend_buffers (mctx); | |
1751 | if (BE (err != REG_NOERROR, 0)) | |
15a7d175 | 1752 | return err; |
612546c6 UD |
1753 | } |
1754 | ||
3b0bdc72 UD |
1755 | if (top < next_state_log_idx) |
1756 | { | |
612546c6 | 1757 | memset (mctx->state_log + top + 1, '\0', |
15a7d175 | 1758 | sizeof (re_dfastate_t *) * (next_state_log_idx - top)); |
3b0bdc72 UD |
1759 | mctx->state_log_top = next_state_log_idx; |
1760 | } | |
612546c6 | 1761 | return REG_NOERROR; |
3b0bdc72 UD |
1762 | } |
1763 | ||
1b2c2628 | 1764 | static reg_errcode_t |
a334319f | 1765 | merge_state_array (dfa, dst, src, num) |
76b864c8 | 1766 | const re_dfa_t *dfa; |
a334319f UD |
1767 | re_dfastate_t **dst; |
1768 | re_dfastate_t **src; | |
1769 | int num; | |
3b0bdc72 | 1770 | { |
0742e48e UD |
1771 | int st_idx; |
1772 | reg_errcode_t err; | |
1773 | for (st_idx = 0; st_idx < num; ++st_idx) | |
1774 | { | |
1775 | if (dst[st_idx] == NULL) | |
15a7d175 | 1776 | dst[st_idx] = src[st_idx]; |
0742e48e | 1777 | else if (src[st_idx] != NULL) |
15a7d175 UD |
1778 | { |
1779 | re_node_set merged_set; | |
1780 | err = re_node_set_init_union (&merged_set, &dst[st_idx]->nodes, | |
1781 | &src[st_idx]->nodes); | |
1782 | if (BE (err != REG_NOERROR, 0)) | |
1783 | return err; | |
1784 | dst[st_idx] = re_acquire_state (&err, dfa, &merged_set); | |
1785 | re_node_set_free (&merged_set); | |
1786 | if (BE (err != REG_NOERROR, 0)) | |
1787 | return err; | |
1788 | } | |
0742e48e UD |
1789 | } |
1790 | return REG_NOERROR; | |
3b0bdc72 UD |
1791 | } |
1792 | ||
0742e48e | 1793 | static reg_errcode_t |
a334319f | 1794 | update_cur_sifted_state (mctx, sctx, str_idx, dest_nodes) |
76b864c8 | 1795 | const re_match_context_t *mctx; |
a334319f UD |
1796 | re_sift_context_t *sctx; |
1797 | int str_idx; | |
1798 | re_node_set *dest_nodes; | |
3b0bdc72 | 1799 | { |
76b864c8 | 1800 | const re_dfa_t *const dfa = mctx->dfa; |
c0c9615c | 1801 | reg_errcode_t err = REG_NOERROR; |
0742e48e | 1802 | const re_node_set *candidates; |
e40a38b3 | 1803 | candidates = ((mctx->state_log[str_idx] == NULL) ? NULL |
15a7d175 | 1804 | : &mctx->state_log[str_idx]->nodes); |
0742e48e | 1805 | |
e40a38b3 UD |
1806 | if (dest_nodes->nelem == 0) |
1807 | sctx->sifted_states[str_idx] = NULL; | |
1808 | else | |
485d775d | 1809 | { |
e40a38b3 UD |
1810 | if (candidates) |
1811 | { | |
1812 | /* At first, add the nodes which can epsilon transit to a node in | |
1813 | DEST_NODE. */ | |
1814 | err = add_epsilon_src_nodes (dfa, dest_nodes, candidates); | |
1815 | if (BE (err != REG_NOERROR, 0)) | |
1816 | return err; | |
0742e48e | 1817 | |
e40a38b3 UD |
1818 | /* Then, check the limitations in the current sift_context. */ |
1819 | if (sctx->limits.nelem) | |
1820 | { | |
1821 | err = check_subexp_limits (dfa, dest_nodes, candidates, &sctx->limits, | |
1822 | mctx->bkref_ents, str_idx); | |
1823 | if (BE (err != REG_NOERROR, 0)) | |
1824 | return err; | |
1825 | } | |
1826 | } | |
1827 | ||
1828 | sctx->sifted_states[str_idx] = re_acquire_state (&err, dfa, dest_nodes); | |
0742e48e | 1829 | if (BE (err != REG_NOERROR, 0)) |
15a7d175 | 1830 | return err; |
0742e48e UD |
1831 | } |
1832 | ||
e40a38b3 | 1833 | if (candidates && mctx->state_log[str_idx]->has_backref) |
0742e48e | 1834 | { |
d2c38eb3 | 1835 | err = sift_states_bkref (mctx, sctx, str_idx, candidates); |
0742e48e | 1836 | if (BE (err != REG_NOERROR, 0)) |
15a7d175 | 1837 | return err; |
0742e48e UD |
1838 | } |
1839 | return REG_NOERROR; | |
3b0bdc72 UD |
1840 | } |
1841 | ||
a9388965 | 1842 | static reg_errcode_t |
a334319f | 1843 | add_epsilon_src_nodes (dfa, dest_nodes, candidates) |
76b864c8 | 1844 | const re_dfa_t *dfa; |
a334319f UD |
1845 | re_node_set *dest_nodes; |
1846 | const re_node_set *candidates; | |
3b0bdc72 | 1847 | { |
5cf1ec52 UD |
1848 | reg_errcode_t err = REG_NOERROR; |
1849 | int i; | |
0742e48e | 1850 | |
5cf1ec52 | 1851 | re_dfastate_t *state = re_acquire_state (&err, dfa, dest_nodes); |
0742e48e UD |
1852 | if (BE (err != REG_NOERROR, 0)) |
1853 | return err; | |
5cf1ec52 UD |
1854 | |
1855 | if (!state->inveclosure.alloc) | |
3b0bdc72 | 1856 | { |
5cf1ec52 | 1857 | err = re_node_set_alloc (&state->inveclosure, dest_nodes->nelem); |
0742e48e | 1858 | if (BE (err != REG_NOERROR, 0)) |
5cf1ec52 UD |
1859 | return REG_ESPACE; |
1860 | for (i = 0; i < dest_nodes->nelem; i++) | |
1861 | re_node_set_merge (&state->inveclosure, | |
1862 | dfa->inveclosures + dest_nodes->elems[i]); | |
0742e48e | 1863 | } |
5cf1ec52 UD |
1864 | return re_node_set_add_intersect (dest_nodes, candidates, |
1865 | &state->inveclosure); | |
0742e48e | 1866 | } |
3b0bdc72 | 1867 | |
0742e48e | 1868 | static reg_errcode_t |
a334319f | 1869 | sub_epsilon_src_nodes (dfa, node, dest_nodes, candidates) |
76b864c8 | 1870 | const re_dfa_t *dfa; |
a334319f UD |
1871 | int node; |
1872 | re_node_set *dest_nodes; | |
1873 | const re_node_set *candidates; | |
0742e48e UD |
1874 | { |
1875 | int ecl_idx; | |
1876 | reg_errcode_t err; | |
1877 | re_node_set *inv_eclosure = dfa->inveclosures + node; | |
1878 | re_node_set except_nodes; | |
1879 | re_node_set_init_empty (&except_nodes); | |
1880 | for (ecl_idx = 0; ecl_idx < inv_eclosure->nelem; ++ecl_idx) | |
1881 | { | |
15a7d175 UD |
1882 | int cur_node = inv_eclosure->elems[ecl_idx]; |
1883 | if (cur_node == node) | |
1884 | continue; | |
1885 | if (IS_EPSILON_NODE (dfa->nodes[cur_node].type)) | |
1886 | { | |
1887 | int edst1 = dfa->edests[cur_node].elems[0]; | |
1888 | int edst2 = ((dfa->edests[cur_node].nelem > 1) | |
1889 | ? dfa->edests[cur_node].elems[1] : -1); | |
1890 | if ((!re_node_set_contains (inv_eclosure, edst1) | |
1891 | && re_node_set_contains (dest_nodes, edst1)) | |
1892 | || (edst2 > 0 | |
1893 | && !re_node_set_contains (inv_eclosure, edst2) | |
1894 | && re_node_set_contains (dest_nodes, edst2))) | |
1895 | { | |
1896 | err = re_node_set_add_intersect (&except_nodes, candidates, | |
1897 | dfa->inveclosures + cur_node); | |
1898 | if (BE (err != REG_NOERROR, 0)) | |
1899 | { | |
1900 | re_node_set_free (&except_nodes); | |
1901 | return err; | |
1902 | } | |
1903 | } | |
1904 | } | |
0742e48e UD |
1905 | } |
1906 | for (ecl_idx = 0; ecl_idx < inv_eclosure->nelem; ++ecl_idx) | |
1907 | { | |
15a7d175 UD |
1908 | int cur_node = inv_eclosure->elems[ecl_idx]; |
1909 | if (!re_node_set_contains (&except_nodes, cur_node)) | |
1910 | { | |
1911 | int idx = re_node_set_contains (dest_nodes, cur_node) - 1; | |
1912 | re_node_set_remove_at (dest_nodes, idx); | |
1913 | } | |
0742e48e UD |
1914 | } |
1915 | re_node_set_free (&except_nodes); | |
1916 | return REG_NOERROR; | |
1917 | } | |
1918 | ||
a3022b82 | 1919 | static int |
a334319f | 1920 | check_dst_limits (mctx, limits, dst_node, dst_idx, src_node, src_idx) |
01ed6ceb | 1921 | const re_match_context_t *mctx; |
a334319f UD |
1922 | re_node_set *limits; |
1923 | int dst_node, dst_idx, src_node, src_idx; | |
a3022b82 | 1924 | { |
76b864c8 | 1925 | const re_dfa_t *const dfa = mctx->dfa; |
a3022b82 UD |
1926 | int lim_idx, src_pos, dst_pos; |
1927 | ||
e40a38b3 UD |
1928 | int dst_bkref_idx = search_cur_bkref_entry (mctx, dst_idx); |
1929 | int src_bkref_idx = search_cur_bkref_entry (mctx, src_idx); | |
a3022b82 UD |
1930 | for (lim_idx = 0; lim_idx < limits->nelem; ++lim_idx) |
1931 | { | |
485d775d | 1932 | int subexp_idx; |
a3022b82 UD |
1933 | struct re_backref_cache_entry *ent; |
1934 | ent = mctx->bkref_ents + limits->elems[lim_idx]; | |
d8f73de8 | 1935 | subexp_idx = dfa->nodes[ent->node].opr.idx; |
a3022b82 | 1936 | |
e3a87852 | 1937 | dst_pos = check_dst_limits_calc_pos (mctx, limits->elems[lim_idx], |
e40a38b3 UD |
1938 | subexp_idx, dst_node, dst_idx, |
1939 | dst_bkref_idx); | |
e3a87852 | 1940 | src_pos = check_dst_limits_calc_pos (mctx, limits->elems[lim_idx], |
e40a38b3 UD |
1941 | subexp_idx, src_node, src_idx, |
1942 | src_bkref_idx); | |
a3022b82 UD |
1943 | |
1944 | /* In case of: | |
15a7d175 UD |
1945 | <src> <dst> ( <subexp> ) |
1946 | ( <subexp> ) <src> <dst> | |
1947 | ( <subexp1> <src> <subexp2> <dst> <subexp3> ) */ | |
a3022b82 | 1948 | if (src_pos == dst_pos) |
15a7d175 | 1949 | continue; /* This is unrelated limitation. */ |
a3022b82 | 1950 | else |
15a7d175 | 1951 | return 1; |
a3022b82 UD |
1952 | } |
1953 | return 0; | |
1954 | } | |
1955 | ||
1956 | static int | |
a334319f | 1957 | check_dst_limits_calc_pos_1 (mctx, boundaries, subexp_idx, from_node, bkref_idx) |
01ed6ceb | 1958 | const re_match_context_t *mctx; |
a334319f | 1959 | int boundaries, subexp_idx, from_node, bkref_idx; |
a3022b82 | 1960 | { |
76b864c8 UD |
1961 | const re_dfa_t *const dfa = mctx->dfa; |
1962 | const re_node_set *eclosures = dfa->eclosures + from_node; | |
1cef7b3c | 1963 | int node_idx; |
d0b96fc4 | 1964 | |
d0b96fc4 UD |
1965 | /* Else, we are on the boundary: examine the nodes on the epsilon |
1966 | closure. */ | |
1cef7b3c UD |
1967 | for (node_idx = 0; node_idx < eclosures->nelem; ++node_idx) |
1968 | { | |
1969 | int node = eclosures->elems[node_idx]; | |
d0b96fc4 UD |
1970 | switch (dfa->nodes[node].type) |
1971 | { | |
1972 | case OP_BACK_REF: | |
d8f73de8 UD |
1973 | if (bkref_idx != -1) |
1974 | { | |
1975 | struct re_backref_cache_entry *ent = mctx->bkref_ents + bkref_idx; | |
1976 | do | |
1977 | { | |
1978 | int dst, cpos; | |
d0b96fc4 | 1979 | |
d8f73de8 UD |
1980 | if (ent->node != node) |
1981 | continue; | |
d0b96fc4 | 1982 | |
2c05d33f UD |
1983 | if (subexp_idx < BITSET_WORD_BITS |
1984 | && !(ent->eps_reachable_subexps_map | |
1985 | & ((bitset_word_t) 1 << subexp_idx))) | |
d8f73de8 | 1986 | continue; |
d0b96fc4 | 1987 | |
d8f73de8 UD |
1988 | /* Recurse trying to reach the OP_OPEN_SUBEXP and |
1989 | OP_CLOSE_SUBEXP cases below. But, if the | |
1990 | destination node is the same node as the source | |
1991 | node, don't recurse because it would cause an | |
1992 | infinite loop: a regex that exhibits this behavior | |
1993 | is ()\1*\1* */ | |
1994 | dst = dfa->edests[node].elems[0]; | |
1995 | if (dst == from_node) | |
1996 | { | |
1997 | if (boundaries & 1) | |
1998 | return -1; | |
1999 | else /* if (boundaries & 2) */ | |
2000 | return 0; | |
2001 | } | |
7db61208 | 2002 | |
d8f73de8 UD |
2003 | cpos = |
2004 | check_dst_limits_calc_pos_1 (mctx, boundaries, subexp_idx, | |
2005 | dst, bkref_idx); | |
2006 | if (cpos == -1 /* && (boundaries & 1) */) | |
2007 | return -1; | |
2008 | if (cpos == 0 && (boundaries & 2)) | |
2009 | return 0; | |
2010 | ||
2c05d33f UD |
2011 | if (subexp_idx < BITSET_WORD_BITS) |
2012 | ent->eps_reachable_subexps_map | |
2013 | &= ~((bitset_word_t) 1 << subexp_idx); | |
d8f73de8 UD |
2014 | } |
2015 | while (ent++->more); | |
2016 | } | |
2017 | break; | |
f0c7c524 | 2018 | |
d0b96fc4 | 2019 | case OP_OPEN_SUBEXP: |
e40a38b3 | 2020 | if ((boundaries & 1) && subexp_idx == dfa->nodes[node].opr.idx) |
d0b96fc4 UD |
2021 | return -1; |
2022 | break; | |
f0c7c524 | 2023 | |
d0b96fc4 | 2024 | case OP_CLOSE_SUBEXP: |
e40a38b3 | 2025 | if ((boundaries & 2) && subexp_idx == dfa->nodes[node].opr.idx) |
d0b96fc4 UD |
2026 | return 0; |
2027 | break; | |
2028 | ||
2029 | default: | |
15a7d175 UD |
2030 | break; |
2031 | } | |
a3022b82 | 2032 | } |
d0b96fc4 | 2033 | |
e40a38b3 UD |
2034 | return (boundaries & 2) ? 1 : 0; |
2035 | } | |
2036 | ||
2037 | static int | |
a334319f | 2038 | check_dst_limits_calc_pos (mctx, limit, subexp_idx, from_node, str_idx, bkref_idx) |
01ed6ceb | 2039 | const re_match_context_t *mctx; |
a334319f | 2040 | int limit, subexp_idx, from_node, str_idx, bkref_idx; |
e40a38b3 UD |
2041 | { |
2042 | struct re_backref_cache_entry *lim = mctx->bkref_ents + limit; | |
2043 | int boundaries; | |
2044 | ||
2045 | /* If we are outside the range of the subexpression, return -1 or 1. */ | |
2046 | if (str_idx < lim->subexp_from) | |
2047 | return -1; | |
2048 | ||
2049 | if (lim->subexp_to < str_idx) | |
d0b96fc4 | 2050 | return 1; |
e40a38b3 UD |
2051 | |
2052 | /* If we are within the subexpression, return 0. */ | |
2053 | boundaries = (str_idx == lim->subexp_from); | |
2054 | boundaries |= (str_idx == lim->subexp_to) << 1; | |
2055 | if (boundaries == 0) | |
d0b96fc4 | 2056 | return 0; |
e40a38b3 UD |
2057 | |
2058 | /* Else, examine epsilon closure. */ | |
2059 | return check_dst_limits_calc_pos_1 (mctx, boundaries, subexp_idx, | |
2060 | from_node, bkref_idx); | |
a3022b82 UD |
2061 | } |
2062 | ||
0742e48e UD |
2063 | /* Check the limitations of sub expressions LIMITS, and remove the nodes |
2064 | which are against limitations from DEST_NODES. */ | |
2065 | ||
2066 | static reg_errcode_t | |
a334319f | 2067 | check_subexp_limits (dfa, dest_nodes, candidates, limits, bkref_ents, str_idx) |
76b864c8 | 2068 | const re_dfa_t *dfa; |
a334319f UD |
2069 | re_node_set *dest_nodes; |
2070 | const re_node_set *candidates; | |
2071 | re_node_set *limits; | |
2072 | struct re_backref_cache_entry *bkref_ents; | |
2073 | int str_idx; | |
0742e48e UD |
2074 | { |
2075 | reg_errcode_t err; | |
2076 | int node_idx, lim_idx; | |
2077 | ||
2078 | for (lim_idx = 0; lim_idx < limits->nelem; ++lim_idx) | |
2079 | { | |
485d775d | 2080 | int subexp_idx; |
0742e48e UD |
2081 | struct re_backref_cache_entry *ent; |
2082 | ent = bkref_ents + limits->elems[lim_idx]; | |
2083 | ||
2084 | if (str_idx <= ent->subexp_from || ent->str_idx < str_idx) | |
15a7d175 | 2085 | continue; /* This is unrelated limitation. */ |
0742e48e | 2086 | |
d8f73de8 | 2087 | subexp_idx = dfa->nodes[ent->node].opr.idx; |
0742e48e | 2088 | if (ent->subexp_to == str_idx) |
15a7d175 UD |
2089 | { |
2090 | int ops_node = -1; | |
2091 | int cls_node = -1; | |
2092 | for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx) | |
2093 | { | |
2094 | int node = dest_nodes->elems[node_idx]; | |
46bf9de7 | 2095 | re_token_type_t type = dfa->nodes[node].type; |
15a7d175 UD |
2096 | if (type == OP_OPEN_SUBEXP |
2097 | && subexp_idx == dfa->nodes[node].opr.idx) | |
2098 | ops_node = node; | |
2099 | else if (type == OP_CLOSE_SUBEXP | |
2100 | && subexp_idx == dfa->nodes[node].opr.idx) | |
2101 | cls_node = node; | |
2102 | } | |
2103 | ||
2104 | /* Check the limitation of the open subexpression. */ | |
2105 | /* Note that (ent->subexp_to = str_idx != ent->subexp_from). */ | |
2106 | if (ops_node >= 0) | |
2107 | { | |
46bf9de7 UD |
2108 | err = sub_epsilon_src_nodes (dfa, ops_node, dest_nodes, |
2109 | candidates); | |
15a7d175 UD |
2110 | if (BE (err != REG_NOERROR, 0)) |
2111 | return err; | |
2112 | } | |
46bf9de7 | 2113 | |
15a7d175 | 2114 | /* Check the limitation of the close subexpression. */ |
46bf9de7 UD |
2115 | if (cls_node >= 0) |
2116 | for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx) | |
2117 | { | |
2118 | int node = dest_nodes->elems[node_idx]; | |
2119 | if (!re_node_set_contains (dfa->inveclosures + node, | |
2120 | cls_node) | |
2121 | && !re_node_set_contains (dfa->eclosures + node, | |
2122 | cls_node)) | |
2123 | { | |
2124 | /* It is against this limitation. | |
2125 | Remove it form the current sifted state. */ | |
2126 | err = sub_epsilon_src_nodes (dfa, node, dest_nodes, | |
2127 | candidates); | |
2128 | if (BE (err != REG_NOERROR, 0)) | |
2129 | return err; | |
2130 | --node_idx; | |
2131 | } | |
2132 | } | |
15a7d175 | 2133 | } |
0742e48e | 2134 | else /* (ent->subexp_to != str_idx) */ |
15a7d175 UD |
2135 | { |
2136 | for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx) | |
2137 | { | |
2138 | int node = dest_nodes->elems[node_idx]; | |
46bf9de7 | 2139 | re_token_type_t type = dfa->nodes[node].type; |
15a7d175 UD |
2140 | if (type == OP_CLOSE_SUBEXP || type == OP_OPEN_SUBEXP) |
2141 | { | |
2142 | if (subexp_idx != dfa->nodes[node].opr.idx) | |
2143 | continue; | |
d8f73de8 UD |
2144 | /* It is against this limitation. |
2145 | Remove it form the current sifted state. */ | |
2146 | err = sub_epsilon_src_nodes (dfa, node, dest_nodes, | |
2147 | candidates); | |
2148 | if (BE (err != REG_NOERROR, 0)) | |
2149 | return err; | |
15a7d175 UD |
2150 | } |
2151 | } | |
2152 | } | |
0742e48e UD |
2153 | } |
2154 | return REG_NOERROR; | |
2155 | } | |
2156 | ||
0742e48e | 2157 | static reg_errcode_t |
a334319f | 2158 | sift_states_bkref (mctx, sctx, str_idx, candidates) |
76b864c8 | 2159 | const re_match_context_t *mctx; |
a334319f UD |
2160 | re_sift_context_t *sctx; |
2161 | int str_idx; | |
2162 | const re_node_set *candidates; | |
0742e48e | 2163 | { |
76b864c8 | 2164 | const re_dfa_t *const dfa = mctx->dfa; |
0742e48e | 2165 | reg_errcode_t err; |
0742e48e UD |
2166 | int node_idx, node; |
2167 | re_sift_context_t local_sctx; | |
e40a38b3 UD |
2168 | int first_idx = search_cur_bkref_entry (mctx, str_idx); |
2169 | ||
2170 | if (first_idx == -1) | |
2171 | return REG_NOERROR; | |
2172 | ||
0742e48e UD |
2173 | local_sctx.sifted_states = NULL; /* Mark that it hasn't been initialized. */ |
2174 | ||
2175 | for (node_idx = 0; node_idx < candidates->nelem; ++node_idx) | |
2176 | { | |
e40a38b3 | 2177 | int enabled_idx; |
0742e48e | 2178 | re_token_type_t type; |
e40a38b3 | 2179 | struct re_backref_cache_entry *entry; |
0742e48e UD |
2180 | node = candidates->elems[node_idx]; |
2181 | type = dfa->nodes[node].type; | |
0742e48e UD |
2182 | /* Avoid infinite loop for the REs like "()\1+". */ |
2183 | if (node == sctx->last_node && str_idx == sctx->last_str_idx) | |
15a7d175 | 2184 | continue; |
e40a38b3 UD |
2185 | if (type != OP_BACK_REF) |
2186 | continue; | |
2187 | ||
2188 | entry = mctx->bkref_ents + first_idx; | |
2189 | enabled_idx = first_idx; | |
2190 | do | |
15a7d175 | 2191 | { |
2d87db5b UD |
2192 | int subexp_len; |
2193 | int to_idx; | |
2194 | int dst_node; | |
2195 | int ret; | |
e40a38b3 UD |
2196 | re_dfastate_t *cur_state; |
2197 | ||
2198 | if (entry->node != node) | |
2199 | continue; | |
2200 | subexp_len = entry->subexp_to - entry->subexp_from; | |
2201 | to_idx = str_idx + subexp_len; | |
2202 | dst_node = (subexp_len ? dfa->nexts[node] | |
2203 | : dfa->edests[node].elems[0]); | |
2204 | ||
2205 | if (to_idx > sctx->last_str_idx | |
2206 | || sctx->sifted_states[to_idx] == NULL | |
2207 | || !STATE_NODE_CONTAINS (sctx->sifted_states[to_idx], dst_node) | |
2208 | || check_dst_limits (mctx, &sctx->limits, node, | |
2209 | str_idx, dst_node, to_idx)) | |
2210 | continue; | |
2211 | ||
2212 | if (local_sctx.sifted_states == NULL) | |
15a7d175 | 2213 | { |
e40a38b3 UD |
2214 | local_sctx = *sctx; |
2215 | err = re_node_set_init_copy (&local_sctx.limits, &sctx->limits); | |
2216 | if (BE (err != REG_NOERROR, 0)) | |
2217 | goto free_return; | |
2218 | } | |
2219 | local_sctx.last_node = node; | |
2220 | local_sctx.last_str_idx = str_idx; | |
2d87db5b UD |
2221 | ret = re_node_set_insert (&local_sctx.limits, enabled_idx); |
2222 | if (BE (ret < 0, 0)) | |
e40a38b3 UD |
2223 | { |
2224 | err = REG_ESPACE; | |
2225 | goto free_return; | |
15a7d175 | 2226 | } |
e40a38b3 UD |
2227 | cur_state = local_sctx.sifted_states[str_idx]; |
2228 | err = sift_states_backward (mctx, &local_sctx); | |
2229 | if (BE (err != REG_NOERROR, 0)) | |
2230 | goto free_return; | |
2231 | if (sctx->limited_states != NULL) | |
2232 | { | |
2233 | err = merge_state_array (dfa, sctx->limited_states, | |
2234 | local_sctx.sifted_states, | |
2235 | str_idx + 1); | |
2236 | if (BE (err != REG_NOERROR, 0)) | |
2237 | goto free_return; | |
2238 | } | |
2239 | local_sctx.sifted_states[str_idx] = cur_state; | |
2240 | re_node_set_remove (&local_sctx.limits, enabled_idx); | |
2241 | ||
2242 | /* mctx->bkref_ents may have changed, reload the pointer. */ | |
2243 | entry = mctx->bkref_ents + enabled_idx; | |
15a7d175 | 2244 | } |
e40a38b3 | 2245 | while (enabled_idx++, entry++->more); |
0742e48e | 2246 | } |
1b2c2628 UD |
2247 | err = REG_NOERROR; |
2248 | free_return: | |
0742e48e UD |
2249 | if (local_sctx.sifted_states != NULL) |
2250 | { | |
0742e48e UD |
2251 | re_node_set_free (&local_sctx.limits); |
2252 | } | |
2253 | ||
1b2c2628 | 2254 | return err; |
0742e48e UD |
2255 | } |
2256 | ||
2257 | ||
2258 | #ifdef RE_ENABLE_I18N | |
2259 | static int | |
a334319f UD |
2260 | sift_states_iter_mb (mctx, sctx, node_idx, str_idx, max_str_idx) |
2261 | const re_match_context_t *mctx; | |
2262 | re_sift_context_t *sctx; | |
2263 | int node_idx, str_idx, max_str_idx; | |
0742e48e | 2264 | { |
c42b4152 | 2265 | const re_dfa_t *const dfa = mctx->dfa; |
0742e48e UD |
2266 | int naccepted; |
2267 | /* Check the node can accept `multi byte'. */ | |
56b168be | 2268 | naccepted = check_node_accept_bytes (dfa, node_idx, &mctx->input, str_idx); |
0742e48e UD |
2269 | if (naccepted > 0 && str_idx + naccepted <= max_str_idx && |
2270 | !STATE_NODE_CONTAINS (sctx->sifted_states[str_idx + naccepted], | |
15a7d175 | 2271 | dfa->nexts[node_idx])) |
0742e48e | 2272 | /* The node can't accept the `multi byte', or the |
bb3f4825 | 2273 | destination was already thrown away, then the node |
0742e48e UD |
2274 | could't accept the current input `multi byte'. */ |
2275 | naccepted = 0; | |
2276 | /* Otherwise, it is sure that the node could accept | |
2277 | `naccepted' bytes input. */ | |
2278 | return naccepted; | |
2279 | } | |
2280 | #endif /* RE_ENABLE_I18N */ | |
2281 | ||
3b0bdc72 UD |
2282 | \f |
2283 | /* Functions for state transition. */ | |
2284 | ||
2285 | /* Return the next state to which the current state STATE will transit by | |
2286 | accepting the current input byte, and update STATE_LOG if necessary. | |
2287 | If STATE can accept a multibyte char/collating element/back reference | |
2288 | update the destination of STATE_LOG. */ | |
2289 | ||
2290 | static re_dfastate_t * | |
a334319f UD |
2291 | transit_state (err, mctx, state) |
2292 | reg_errcode_t *err; | |
2293 | re_match_context_t *mctx; | |
2294 | re_dfastate_t *state; | |
3b0bdc72 | 2295 | { |
58845a70 | 2296 | re_dfastate_t **trtable; |
3b0bdc72 UD |
2297 | unsigned char ch; |
2298 | ||
bb677c95 UD |
2299 | #ifdef RE_ENABLE_I18N |
2300 | /* If the current state can accept multibyte. */ | |
2301 | if (BE (state->accept_mb, 0)) | |
612546c6 | 2302 | { |
bb677c95 | 2303 | *err = transit_state_mb (mctx, state); |
612546c6 | 2304 | if (BE (*err != REG_NOERROR, 0)) |
15a7d175 | 2305 | return NULL; |
612546c6 | 2306 | } |
434d3784 | 2307 | #endif /* RE_ENABLE_I18N */ |
3b0bdc72 | 2308 | |
4c595adb | 2309 | /* Then decide the next state with the single byte. */ |
ab4b89fe UD |
2310 | #if 0 |
2311 | if (0) | |
2312 | /* don't use transition table */ | |
2313 | return transit_state_sb (err, mctx, state); | |
2314 | #endif | |
2315 | ||
2316 | /* Use transition table */ | |
2317 | ch = re_string_fetch_byte (&mctx->input); | |
2318 | for (;;) | |
4c595adb | 2319 | { |
4c595adb | 2320 | trtable = state->trtable; |
ab4b89fe UD |
2321 | if (BE (trtable != NULL, 1)) |
2322 | return trtable[ch]; | |
2323 | ||
2324 | trtable = state->word_trtable; | |
2325 | if (BE (trtable != NULL, 1)) | |
4c595adb UD |
2326 | { |
2327 | unsigned int context; | |
2328 | context | |
2329 | = re_string_context_at (&mctx->input, | |
2330 | re_string_cur_idx (&mctx->input) - 1, | |
2331 | mctx->eflags); | |
2332 | if (IS_WORD_CONTEXT (context)) | |
2333 | return trtable[ch + SBC_MAX]; | |
c13c99fa | 2334 | else |
4c595adb | 2335 | return trtable[ch]; |
15a7d175 | 2336 | } |
ab4b89fe UD |
2337 | |
2338 | if (!build_trtable (mctx->dfa, state)) | |
2339 | { | |
2340 | *err = REG_ESPACE; | |
2341 | return NULL; | |
2342 | } | |
2343 | ||
2344 | /* Retry, we now have a transition table. */ | |
3b0bdc72 | 2345 | } |
4c595adb | 2346 | } |
3b0bdc72 | 2347 | |
4c595adb | 2348 | /* Update the state_log if we need */ |
58845a70 | 2349 | re_dfastate_t * |
a334319f UD |
2350 | merge_state_with_log (err, mctx, next_state) |
2351 | reg_errcode_t *err; | |
2352 | re_match_context_t *mctx; | |
2353 | re_dfastate_t *next_state; | |
58845a70 | 2354 | { |
c42b4152 | 2355 | const re_dfa_t *const dfa = mctx->dfa; |
4c595adb UD |
2356 | int cur_idx = re_string_cur_idx (&mctx->input); |
2357 | ||
2358 | if (cur_idx > mctx->state_log_top) | |
3b0bdc72 | 2359 | { |
4c595adb UD |
2360 | mctx->state_log[cur_idx] = next_state; |
2361 | mctx->state_log_top = cur_idx; | |
2362 | } | |
2363 | else if (mctx->state_log[cur_idx] == 0) | |
2364 | { | |
2365 | mctx->state_log[cur_idx] = next_state; | |
2366 | } | |
2367 | else | |
2368 | { | |
2369 | re_dfastate_t *pstate; | |
2370 | unsigned int context; | |
2371 | re_node_set next_nodes, *log_nodes, *table_nodes = NULL; | |
2372 | /* If (state_log[cur_idx] != 0), it implies that cur_idx is | |
2373 | the destination of a multibyte char/collating element/ | |
2374 | back reference. Then the next state is the union set of | |
2375 | these destinations and the results of the transition table. */ | |
2376 | pstate = mctx->state_log[cur_idx]; | |
2377 | log_nodes = pstate->entrance_nodes; | |
2378 | if (next_state != NULL) | |
2379 | { | |
2380 | table_nodes = next_state->entrance_nodes; | |
2381 | *err = re_node_set_init_union (&next_nodes, table_nodes, | |
15a7d175 | 2382 | log_nodes); |
4c595adb UD |
2383 | if (BE (*err != REG_NOERROR, 0)) |
2384 | return NULL; | |
2385 | } | |
2386 | else | |
2387 | next_nodes = *log_nodes; | |
2388 | /* Note: We already add the nodes of the initial state, | |
2389 | then we don't need to add them here. */ | |
2390 | ||
2391 | context = re_string_context_at (&mctx->input, | |
2392 | re_string_cur_idx (&mctx->input) - 1, | |
2393 | mctx->eflags); | |
2394 | next_state = mctx->state_log[cur_idx] | |
2395 | = re_acquire_state_context (err, dfa, &next_nodes, context); | |
2396 | /* We don't need to check errors here, since the return value of | |
2397 | this function is next_state and ERR is already set. */ | |
2398 | ||
2399 | if (table_nodes != NULL) | |
2400 | re_node_set_free (&next_nodes); | |
6291ee3c UD |
2401 | } |
2402 | ||
bb3f4825 | 2403 | if (BE (dfa->nbackref, 0) && next_state != NULL) |
6291ee3c | 2404 | { |
bb3f4825 UD |
2405 | /* Check OP_OPEN_SUBEXP in the current state in case that we use them |
2406 | later. We must check them here, since the back references in the | |
2407 | next state might use them. */ | |
e3a87852 | 2408 | *err = check_subexp_matching_top (mctx, &next_state->nodes, |
6291ee3c UD |
2409 | cur_idx); |
2410 | if (BE (*err != REG_NOERROR, 0)) | |
2411 | return NULL; | |
6291ee3c | 2412 | |
bb3f4825 UD |
2413 | /* If the next state has back references. */ |
2414 | if (next_state->has_backref) | |
2415 | { | |
e3a87852 | 2416 | *err = transit_state_bkref (mctx, &next_state->nodes); |
bb3f4825 UD |
2417 | if (BE (*err != REG_NOERROR, 0)) |
2418 | return NULL; | |
2419 | next_state = mctx->state_log[cur_idx]; | |
2420 | } | |
3b0bdc72 | 2421 | } |
4c595adb | 2422 | |
3b0bdc72 UD |
2423 | return next_state; |
2424 | } | |
2425 | ||
4c595adb UD |
2426 | /* Skip bytes in the input that correspond to part of a |
2427 | multi-byte match, then look in the log for a state | |
2428 | from which to restart matching. */ | |
2429 | re_dfastate_t * | |
a334319f UD |
2430 | find_recover_state (err, mctx) |
2431 | reg_errcode_t *err; | |
2432 | re_match_context_t *mctx; | |
4c595adb | 2433 | { |
1878e9af | 2434 | re_dfastate_t *cur_state; |
4c595adb UD |
2435 | do |
2436 | { | |
2437 | int max = mctx->state_log_top; | |
2438 | int cur_str_idx = re_string_cur_idx (&mctx->input); | |
2439 | ||
2440 | do | |
2441 | { | |
2442 | if (++cur_str_idx > max) | |
2443 | return NULL; | |
2444 | re_string_skip_bytes (&mctx->input, 1); | |
2445 | } | |
2446 | while (mctx->state_log[cur_str_idx] == NULL); | |
2447 | ||
2448 | cur_state = merge_state_with_log (err, mctx, NULL); | |
2449 | } | |
2d87db5b | 2450 | while (*err == REG_NOERROR && cur_state == NULL); |
4c595adb UD |
2451 | return cur_state; |
2452 | } | |
2453 | ||
3b0bdc72 UD |
2454 | /* Helper functions for transit_state. */ |
2455 | ||
6291ee3c UD |
2456 | /* From the node set CUR_NODES, pick up the nodes whose types are |
2457 | OP_OPEN_SUBEXP and which have corresponding back references in the regular | |
2458 | expression. And register them to use them later for evaluating the | |
2459 | correspoding back references. */ | |
2460 | ||
2461 | static reg_errcode_t | |
a334319f UD |
2462 | check_subexp_matching_top (mctx, cur_nodes, str_idx) |
2463 | re_match_context_t *mctx; | |
2464 | re_node_set *cur_nodes; | |
2465 | int str_idx; | |
6291ee3c | 2466 | { |
76b864c8 | 2467 | const re_dfa_t *const dfa = mctx->dfa; |
6291ee3c UD |
2468 | int node_idx; |
2469 | reg_errcode_t err; | |
2470 | ||
2471 | /* TODO: This isn't efficient. | |
2472 | Because there might be more than one nodes whose types are | |
2473 | OP_OPEN_SUBEXP and whose index is SUBEXP_IDX, we must check all | |
2474 | nodes. | |
2475 | E.g. RE: (a){2} */ | |
2476 | for (node_idx = 0; node_idx < cur_nodes->nelem; ++node_idx) | |
2477 | { | |
2478 | int node = cur_nodes->elems[node_idx]; | |
2479 | if (dfa->nodes[node].type == OP_OPEN_SUBEXP | |
2c05d33f UD |
2480 | && dfa->nodes[node].opr.idx < BITSET_WORD_BITS |
2481 | && (dfa->used_bkref_map | |
2482 | & ((bitset_word_t) 1 << dfa->nodes[node].opr.idx))) | |
6291ee3c UD |
2483 | { |
2484 | err = match_ctx_add_subtop (mctx, node, str_idx); | |
2485 | if (BE (err != REG_NOERROR, 0)) | |
2486 | return err; | |
2487 | } | |
2488 | } | |
2489 | return REG_NOERROR; | |
2490 | } | |
2491 | ||
c13c99fa | 2492 | #if 0 |
3b0bdc72 UD |
2493 | /* Return the next state to which the current state STATE will transit by |
2494 | accepting the current input byte. */ | |
2495 | ||
2496 | static re_dfastate_t * | |
a334319f UD |
2497 | transit_state_sb (err, mctx, state) |
2498 | reg_errcode_t *err; | |
2499 | re_match_context_t *mctx; | |
2500 | re_dfastate_t *state; | |
3b0bdc72 | 2501 | { |
c42b4152 | 2502 | const re_dfa_t *const dfa = mctx->dfa; |
3b0bdc72 UD |
2503 | re_node_set next_nodes; |
2504 | re_dfastate_t *next_state; | |
56b168be | 2505 | int node_cnt, cur_str_idx = re_string_cur_idx (&mctx->input); |
3b0bdc72 UD |
2506 | unsigned int context; |
2507 | ||
a9388965 | 2508 | *err = re_node_set_alloc (&next_nodes, state->nodes.nelem + 1); |
bc15410e | 2509 | if (BE (*err != REG_NOERROR, 0)) |
a9388965 | 2510 | return NULL; |
3b0bdc72 UD |
2511 | for (node_cnt = 0; node_cnt < state->nodes.nelem; ++node_cnt) |
2512 | { | |
2513 | int cur_node = state->nodes.elems[node_cnt]; | |
e3a87852 | 2514 | if (check_node_accept (mctx, dfa->nodes + cur_node, cur_str_idx)) |
15a7d175 UD |
2515 | { |
2516 | *err = re_node_set_merge (&next_nodes, | |
2517 | dfa->eclosures + dfa->nexts[cur_node]); | |
2518 | if (BE (*err != REG_NOERROR, 0)) | |
2519 | { | |
2520 | re_node_set_free (&next_nodes); | |
2521 | return NULL; | |
2522 | } | |
2523 | } | |
3b0bdc72 | 2524 | } |
56b168be | 2525 | context = re_string_context_at (&mctx->input, cur_str_idx, mctx->eflags); |
a9388965 UD |
2526 | next_state = re_acquire_state_context (err, dfa, &next_nodes, context); |
2527 | /* We don't need to check errors here, since the return value of | |
2528 | this function is next_state and ERR is already set. */ | |
2529 | ||
3b0bdc72 | 2530 | re_node_set_free (&next_nodes); |
56b168be | 2531 | re_string_skip_bytes (&mctx->input, 1); |
3b0bdc72 UD |
2532 | return next_state; |
2533 | } | |
c13c99fa | 2534 | #endif |
3b0bdc72 | 2535 | |
434d3784 | 2536 | #ifdef RE_ENABLE_I18N |
a9388965 | 2537 | static reg_errcode_t |
a334319f UD |
2538 | transit_state_mb (mctx, pstate) |
2539 | re_match_context_t *mctx; | |
2540 | re_dfastate_t *pstate; | |
3b0bdc72 | 2541 | { |
c42b4152 | 2542 | const re_dfa_t *const dfa = mctx->dfa; |
a9388965 | 2543 | reg_errcode_t err; |
3b0bdc72 UD |
2544 | int i; |
2545 | ||
2546 | for (i = 0; i < pstate->nodes.nelem; ++i) | |
2547 | { | |
2548 | re_node_set dest_nodes, *new_nodes; | |
2549 | int cur_node_idx = pstate->nodes.elems[i]; | |
963d8d78 | 2550 | int naccepted, dest_idx; |
3b0bdc72 UD |
2551 | unsigned int context; |
2552 | re_dfastate_t *dest_state; | |
2553 | ||
963d8d78 UD |
2554 | if (!dfa->nodes[cur_node_idx].accept_mb) |
2555 | continue; | |
2556 | ||
485d775d | 2557 | if (dfa->nodes[cur_node_idx].constraint) |
15a7d175 | 2558 | { |
56b168be UD |
2559 | context = re_string_context_at (&mctx->input, |
2560 | re_string_cur_idx (&mctx->input), | |
2561 | mctx->eflags); | |
15a7d175 UD |
2562 | if (NOT_SATISFY_NEXT_CONSTRAINT (dfa->nodes[cur_node_idx].constraint, |
2563 | context)) | |
2564 | continue; | |
2565 | } | |
3b0bdc72 | 2566 | |
ad7f28c2 | 2567 | /* How many bytes the node can accept? */ |
963d8d78 UD |
2568 | naccepted = check_node_accept_bytes (dfa, cur_node_idx, &mctx->input, |
2569 | re_string_cur_idx (&mctx->input)); | |
3b0bdc72 | 2570 | if (naccepted == 0) |
15a7d175 | 2571 | continue; |
3b0bdc72 UD |
2572 | |
2573 | /* The node can accepts `naccepted' bytes. */ | |
56b168be | 2574 | dest_idx = re_string_cur_idx (&mctx->input) + naccepted; |
0742e48e | 2575 | mctx->max_mb_elem_len = ((mctx->max_mb_elem_len < naccepted) ? naccepted |
15a7d175 | 2576 | : mctx->max_mb_elem_len); |
7c1be3ec | 2577 | err = clean_state_log_if_needed (mctx, dest_idx); |
612546c6 | 2578 | if (BE (err != REG_NOERROR, 0)) |
15a7d175 | 2579 | return err; |
3b0bdc72 UD |
2580 | #ifdef DEBUG |
2581 | assert (dfa->nexts[cur_node_idx] != -1); | |
2582 | #endif | |
963d8d78 | 2583 | new_nodes = dfa->eclosures + dfa->nexts[cur_node_idx]; |
3b0bdc72 | 2584 | |
612546c6 | 2585 | dest_state = mctx->state_log[dest_idx]; |
3b0bdc72 | 2586 | if (dest_state == NULL) |
15a7d175 | 2587 | dest_nodes = *new_nodes; |
3b0bdc72 | 2588 | else |
15a7d175 UD |
2589 | { |
2590 | err = re_node_set_init_union (&dest_nodes, | |
2591 | dest_state->entrance_nodes, new_nodes); | |
2592 | if (BE (err != REG_NOERROR, 0)) | |
2593 | return err; | |
2594 | } | |
01ed6ceb UD |
2595 | context = re_string_context_at (&mctx->input, dest_idx - 1, |
2596 | mctx->eflags); | |
612546c6 | 2597 | mctx->state_log[dest_idx] |
15a7d175 | 2598 | = re_acquire_state_context (&err, dfa, &dest_nodes, context); |
3b0bdc72 | 2599 | if (dest_state != NULL) |
15a7d175 | 2600 | re_node_set_free (&dest_nodes); |
1b2c2628 | 2601 | if (BE (mctx->state_log[dest_idx] == NULL && err != REG_NOERROR, 0)) |
15a7d175 | 2602 | return err; |
3b0bdc72 | 2603 | } |
a9388965 | 2604 | return REG_NOERROR; |
3b0bdc72 | 2605 | } |
434d3784 | 2606 | #endif /* RE_ENABLE_I18N */ |
3b0bdc72 | 2607 | |
a9388965 | 2608 | static reg_errcode_t |
a334319f UD |
2609 | transit_state_bkref (mctx, nodes) |
2610 | re_match_context_t *mctx; | |
2611 | const re_node_set *nodes; | |
3b0bdc72 | 2612 | { |
6efbd82c | 2613 | const re_dfa_t *const dfa = mctx->dfa; |
a9388965 | 2614 | reg_errcode_t err; |
0742e48e | 2615 | int i; |
56b168be | 2616 | int cur_str_idx = re_string_cur_idx (&mctx->input); |
3b0bdc72 UD |
2617 | |
2618 | for (i = 0; i < nodes->nelem; ++i) | |
2619 | { | |
6291ee3c | 2620 | int dest_str_idx, prev_nelem, bkc_idx; |
3b0bdc72 UD |
2621 | int node_idx = nodes->elems[i]; |
2622 | unsigned int context; | |
fe9434bb | 2623 | const re_token_t *node = dfa->nodes + node_idx; |
3b0bdc72 UD |
2624 | re_node_set *new_dest_nodes; |
2625 | ||
2626 | /* Check whether `node' is a backreference or not. */ | |
6291ee3c | 2627 | if (node->type != OP_BACK_REF) |
15a7d175 | 2628 | continue; |
485d775d UD |
2629 | |
2630 | if (node->constraint) | |
15a7d175 | 2631 | { |
56b168be UD |
2632 | context = re_string_context_at (&mctx->input, cur_str_idx, |
2633 | mctx->eflags); | |
15a7d175 UD |
2634 | if (NOT_SATISFY_NEXT_CONSTRAINT (node->constraint, context)) |
2635 | continue; | |
2636 | } | |
3b0bdc72 UD |
2637 | |
2638 | /* `node' is a backreference. | |
15a7d175 | 2639 | Check the substring which the substring matched. */ |
6291ee3c | 2640 | bkc_idx = mctx->nbkref_ents; |
e3a87852 | 2641 | err = get_subexp (mctx, node_idx, cur_str_idx); |
612546c6 | 2642 | if (BE (err != REG_NOERROR, 0)) |
15a7d175 | 2643 | goto free_return; |
3b0bdc72 UD |
2644 | |
2645 | /* And add the epsilon closures (which is `new_dest_nodes') of | |
15a7d175 | 2646 | the backreference to appropriate state_log. */ |
3b0bdc72 UD |
2647 | #ifdef DEBUG |
2648 | assert (dfa->nexts[node_idx] != -1); | |
2649 | #endif | |
6291ee3c | 2650 | for (; bkc_idx < mctx->nbkref_ents; ++bkc_idx) |
15a7d175 UD |
2651 | { |
2652 | int subexp_len; | |
2653 | re_dfastate_t *dest_state; | |
2654 | struct re_backref_cache_entry *bkref_ent; | |
2655 | bkref_ent = mctx->bkref_ents + bkc_idx; | |
2656 | if (bkref_ent->node != node_idx || bkref_ent->str_idx != cur_str_idx) | |
2657 | continue; | |
2658 | subexp_len = bkref_ent->subexp_to - bkref_ent->subexp_from; | |
2659 | new_dest_nodes = (subexp_len == 0 | |
2660 | ? dfa->eclosures + dfa->edests[node_idx].elems[0] | |
2661 | : dfa->eclosures + dfa->nexts[node_idx]); | |
2662 | dest_str_idx = (cur_str_idx + bkref_ent->subexp_to | |
2663 | - bkref_ent->subexp_from); | |
56b168be UD |
2664 | context = re_string_context_at (&mctx->input, dest_str_idx - 1, |
2665 | mctx->eflags); | |
15a7d175 UD |
2666 | dest_state = mctx->state_log[dest_str_idx]; |
2667 | prev_nelem = ((mctx->state_log[cur_str_idx] == NULL) ? 0 | |
2668 | : mctx->state_log[cur_str_idx]->nodes.nelem); | |
2669 | /* Add `new_dest_node' to state_log. */ | |
2670 | if (dest_state == NULL) | |
2671 | { | |
2672 | mctx->state_log[dest_str_idx] | |
2673 | = re_acquire_state_context (&err, dfa, new_dest_nodes, | |
2674 | context); | |
2675 | if (BE (mctx->state_log[dest_str_idx] == NULL | |
2676 | && err != REG_NOERROR, 0)) | |
2677 | goto free_return; | |
2678 | } | |
2679 | else | |
2680 | { | |
2681 | re_node_set dest_nodes; | |
2682 | err = re_node_set_init_union (&dest_nodes, | |
2683 | dest_state->entrance_nodes, | |
2684 | new_dest_nodes); | |
2685 | if (BE (err != REG_NOERROR, 0)) | |
2686 | { | |
2687 | re_node_set_free (&dest_nodes); | |
2688 | goto free_return; | |
2689 | } | |
2690 | mctx->state_log[dest_str_idx] | |
2691 | = re_acquire_state_context (&err, dfa, &dest_nodes, context); | |
2692 | re_node_set_free (&dest_nodes); | |
2693 | if (BE (mctx->state_log[dest_str_idx] == NULL | |
2694 | && err != REG_NOERROR, 0)) | |
2695 | goto free_return; | |
2696 | } | |
2697 | /* We need to check recursively if the backreference can epsilon | |
2698 | transit. */ | |
2699 | if (subexp_len == 0 | |
2700 | && mctx->state_log[cur_str_idx]->nodes.nelem > prev_nelem) | |
2701 | { | |
e3a87852 | 2702 | err = check_subexp_matching_top (mctx, new_dest_nodes, |
6291ee3c UD |
2703 | cur_str_idx); |
2704 | if (BE (err != REG_NOERROR, 0)) | |
2705 | goto free_return; | |
e3a87852 | 2706 | err = transit_state_bkref (mctx, new_dest_nodes); |
15a7d175 UD |
2707 | if (BE (err != REG_NOERROR, 0)) |
2708 | goto free_return; | |
2709 | } | |
2710 | } | |
3b0bdc72 | 2711 | } |
1b2c2628 UD |
2712 | err = REG_NOERROR; |
2713 | free_return: | |
1b2c2628 | 2714 | return err; |
3b0bdc72 UD |
2715 | } |
2716 | ||
6291ee3c UD |
2717 | /* Enumerate all the candidates which the backreference BKREF_NODE can match |
2718 | at BKREF_STR_IDX, and register them by match_ctx_add_entry(). | |
2719 | Note that we might collect inappropriate candidates here. | |
2720 | However, the cost of checking them strictly here is too high, then we | |
2721 | delay these checking for prune_impossible_nodes(). */ | |
3b0bdc72 | 2722 | |
6291ee3c | 2723 | static reg_errcode_t |
a334319f UD |
2724 | get_subexp (mctx, bkref_node, bkref_str_idx) |
2725 | re_match_context_t *mctx; | |
2726 | int bkref_node, bkref_str_idx; | |
3b0bdc72 | 2727 | { |
6efbd82c | 2728 | const re_dfa_t *const dfa = mctx->dfa; |
6291ee3c | 2729 | int subexp_num, sub_top_idx; |
56b168be | 2730 | const char *buf = (const char *) re_string_get_buffer (&mctx->input); |
6291ee3c UD |
2731 | /* Return if we have already checked BKREF_NODE at BKREF_STR_IDX. */ |
2732 | int cache_idx = search_cur_bkref_entry (mctx, bkref_str_idx); | |
e40a38b3 | 2733 | if (cache_idx != -1) |
6291ee3c | 2734 | { |
01ed6ceb UD |
2735 | const struct re_backref_cache_entry *entry |
2736 | = mctx->bkref_ents + cache_idx; | |
e40a38b3 UD |
2737 | do |
2738 | if (entry->node == bkref_node) | |
2739 | return REG_NOERROR; /* We already checked it. */ | |
2740 | while (entry++->more); | |
6291ee3c | 2741 | } |
e40a38b3 | 2742 | |
d8f73de8 | 2743 | subexp_num = dfa->nodes[bkref_node].opr.idx; |
6291ee3c UD |
2744 | |
2745 | /* For each sub expression */ | |
2746 | for (sub_top_idx = 0; sub_top_idx < mctx->nsub_tops; ++sub_top_idx) | |
2747 | { | |
2748 | reg_errcode_t err; | |
2749 | re_sub_match_top_t *sub_top = mctx->sub_tops[sub_top_idx]; | |
2750 | re_sub_match_last_t *sub_last; | |
7c1be3ec | 2751 | int sub_last_idx, sl_str, bkref_str_off; |
6291ee3c UD |
2752 | |
2753 | if (dfa->nodes[sub_top->node].opr.idx != subexp_num) | |
2754 | continue; /* It isn't related. */ | |
2755 | ||
2756 | sl_str = sub_top->str_idx; | |
7c1be3ec | 2757 | bkref_str_off = bkref_str_idx; |
6291ee3c UD |
2758 | /* At first, check the last node of sub expressions we already |
2759 | evaluated. */ | |
2760 | for (sub_last_idx = 0; sub_last_idx < sub_top->nlasts; ++sub_last_idx) | |
2761 | { | |
2762 | int sl_str_diff; | |
2763 | sub_last = sub_top->lasts[sub_last_idx]; | |
2764 | sl_str_diff = sub_last->str_idx - sl_str; | |
2765 | /* The matched string by the sub expression match with the substring | |
2766 | at the back reference? */ | |
c1baba0f UD |
2767 | if (sl_str_diff > 0) |
2768 | { | |
2769 | if (BE (bkref_str_off + sl_str_diff > mctx->input.valid_len, 0)) | |
2770 | { | |
2771 | /* Not enough chars for a successful match. */ | |
2772 | if (bkref_str_off + sl_str_diff > mctx->input.len) | |
2773 | break; | |
2774 | ||
2775 | err = clean_state_log_if_needed (mctx, | |
2776 | bkref_str_off | |
2777 | + sl_str_diff); | |
2778 | if (BE (err != REG_NOERROR, 0)) | |
2779 | return err; | |
2780 | buf = (const char *) re_string_get_buffer (&mctx->input); | |
2781 | } | |
2782 | if (memcmp (buf + bkref_str_off, buf + sl_str, sl_str_diff) != 0) | |
01ed6ceb UD |
2783 | /* We don't need to search this sub expression any more. */ |
2784 | break; | |
c1baba0f | 2785 | } |
7c1be3ec | 2786 | bkref_str_off += sl_str_diff; |
6291ee3c | 2787 | sl_str += sl_str_diff; |
e3a87852 | 2788 | err = get_subexp_sub (mctx, sub_top, sub_last, bkref_node, |
6291ee3c | 2789 | bkref_str_idx); |
3ee363e2 | 2790 | |
7c1be3ec UD |
2791 | /* Reload buf, since the preceding call might have reallocated |
2792 | the buffer. */ | |
56b168be | 2793 | buf = (const char *) re_string_get_buffer (&mctx->input); |
3ee363e2 | 2794 | |
6291ee3c UD |
2795 | if (err == REG_NOMATCH) |
2796 | continue; | |
2797 | if (BE (err != REG_NOERROR, 0)) | |
2798 | return err; | |
2799 | } | |
7c1be3ec | 2800 | |
6291ee3c UD |
2801 | if (sub_last_idx < sub_top->nlasts) |
2802 | continue; | |
2803 | if (sub_last_idx > 0) | |
2804 | ++sl_str; | |
2805 | /* Then, search for the other last nodes of the sub expression. */ | |
2806 | for (; sl_str <= bkref_str_idx; ++sl_str) | |
2807 | { | |
2808 | int cls_node, sl_str_off; | |
fe9434bb | 2809 | const re_node_set *nodes; |
6291ee3c UD |
2810 | sl_str_off = sl_str - sub_top->str_idx; |
2811 | /* The matched string by the sub expression match with the substring | |
2812 | at the back reference? */ | |
c1baba0f UD |
2813 | if (sl_str_off > 0) |
2814 | { | |
2815 | if (BE (bkref_str_off >= mctx->input.valid_len, 0)) | |
2816 | { | |
2817 | /* If we are at the end of the input, we cannot match. */ | |
2818 | if (bkref_str_off >= mctx->input.len) | |
2819 | break; | |
2820 | ||
2821 | err = extend_buffers (mctx); | |
2822 | if (BE (err != REG_NOERROR, 0)) | |
2823 | return err; | |
2824 | ||
2825 | buf = (const char *) re_string_get_buffer (&mctx->input); | |
2826 | } | |
2827 | if (buf [bkref_str_off++] != buf[sl_str - 1]) | |
2828 | break; /* We don't need to search this sub expression | |
2829 | any more. */ | |
2830 | } | |
6291ee3c UD |
2831 | if (mctx->state_log[sl_str] == NULL) |
2832 | continue; | |
2833 | /* Does this state have a ')' of the sub expression? */ | |
2834 | nodes = &mctx->state_log[sl_str]->nodes; | |
01ed6ceb UD |
2835 | cls_node = find_subexp_node (dfa, nodes, subexp_num, |
2836 | OP_CLOSE_SUBEXP); | |
6291ee3c UD |
2837 | if (cls_node == -1) |
2838 | continue; /* No. */ | |
2839 | if (sub_top->path == NULL) | |
2840 | { | |
2841 | sub_top->path = calloc (sizeof (state_array_t), | |
2842 | sl_str - sub_top->str_idx + 1); | |
2843 | if (sub_top->path == NULL) | |
2844 | return REG_ESPACE; | |
2845 | } | |
2846 | /* Can the OP_OPEN_SUBEXP node arrive the OP_CLOSE_SUBEXP node | |
2847 | in the current context? */ | |
e3a87852 | 2848 | err = check_arrival (mctx, sub_top->path, sub_top->node, |
01ed6ceb UD |
2849 | sub_top->str_idx, cls_node, sl_str, |
2850 | OP_CLOSE_SUBEXP); | |
6291ee3c UD |
2851 | if (err == REG_NOMATCH) |
2852 | continue; | |
2853 | if (BE (err != REG_NOERROR, 0)) | |
2854 | return err; | |
2855 | sub_last = match_ctx_add_sublast (sub_top, cls_node, sl_str); | |
2856 | if (BE (sub_last == NULL, 0)) | |
2857 | return REG_ESPACE; | |
e3a87852 | 2858 | err = get_subexp_sub (mctx, sub_top, sub_last, bkref_node, |
6291ee3c UD |
2859 | bkref_str_idx); |
2860 | if (err == REG_NOMATCH) | |
2861 | continue; | |
2862 | } | |
2863 | } | |
2864 | return REG_NOERROR; | |
2865 | } | |
2866 | ||
2867 | /* Helper functions for get_subexp(). */ | |
2868 | ||
2869 | /* Check SUB_LAST can arrive to the back reference BKREF_NODE at BKREF_STR. | |
2870 | If it can arrive, register the sub expression expressed with SUB_TOP | |
2871 | and SUB_LAST. */ | |
2872 | ||
2873 | static reg_errcode_t | |
a334319f UD |
2874 | get_subexp_sub (mctx, sub_top, sub_last, bkref_node, bkref_str) |
2875 | re_match_context_t *mctx; | |
2876 | const re_sub_match_top_t *sub_top; | |
2877 | re_sub_match_last_t *sub_last; | |
2878 | int bkref_node, bkref_str; | |
6291ee3c UD |
2879 | { |
2880 | reg_errcode_t err; | |
2881 | int to_idx; | |
2882 | /* Can the subexpression arrive the back reference? */ | |
e3a87852 | 2883 | err = check_arrival (mctx, &sub_last->path, sub_last->node, |
01ed6ceb UD |
2884 | sub_last->str_idx, bkref_node, bkref_str, |
2885 | OP_OPEN_SUBEXP); | |
6291ee3c UD |
2886 | if (err != REG_NOERROR) |
2887 | return err; | |
2888 | err = match_ctx_add_entry (mctx, bkref_node, bkref_str, sub_top->str_idx, | |
2889 | sub_last->str_idx); | |
2890 | if (BE (err != REG_NOERROR, 0)) | |
2891 | return err; | |
2892 | to_idx = bkref_str + sub_last->str_idx - sub_top->str_idx; | |
c1baba0f | 2893 | return clean_state_log_if_needed (mctx, to_idx); |
6291ee3c UD |
2894 | } |
2895 | ||
2896 | /* Find the first node which is '(' or ')' and whose index is SUBEXP_IDX. | |
2897 | Search '(' if FL_OPEN, or search ')' otherwise. | |
2898 | TODO: This function isn't efficient... | |
2899 | Because there might be more than one nodes whose types are | |
2900 | OP_OPEN_SUBEXP and whose index is SUBEXP_IDX, we must check all | |
2901 | nodes. | |
2902 | E.g. RE: (a){2} */ | |
2903 | ||
2904 | static int | |
a334319f UD |
2905 | find_subexp_node (dfa, nodes, subexp_idx, type) |
2906 | const re_dfa_t *dfa; | |
2907 | const re_node_set *nodes; | |
2908 | int subexp_idx, type; | |
6291ee3c UD |
2909 | { |
2910 | int cls_idx; | |
2911 | for (cls_idx = 0; cls_idx < nodes->nelem; ++cls_idx) | |
2912 | { | |
2913 | int cls_node = nodes->elems[cls_idx]; | |
fe9434bb | 2914 | const re_token_t *node = dfa->nodes + cls_node; |
0ce7f49c | 2915 | if (node->type == type |
6291ee3c UD |
2916 | && node->opr.idx == subexp_idx) |
2917 | return cls_node; | |
2918 | } | |
2919 | return -1; | |
2920 | } | |
2921 | ||
2922 | /* Check whether the node TOP_NODE at TOP_STR can arrive to the node | |
2923 | LAST_NODE at LAST_STR. We record the path onto PATH since it will be | |
2924 | heavily reused. | |
2925 | Return REG_NOERROR if it can arrive, or REG_NOMATCH otherwise. */ | |
2926 | ||
2927 | static reg_errcode_t | |
a334319f UD |
2928 | check_arrival (mctx, path, top_node, top_str, last_node, last_str, |
2929 | type) | |
2930 | re_match_context_t *mctx; | |
2931 | state_array_t *path; | |
2932 | int top_node, top_str, last_node, last_str, type; | |
6291ee3c | 2933 | { |
6efbd82c | 2934 | const re_dfa_t *const dfa = mctx->dfa; |
c0c9615c | 2935 | reg_errcode_t err = REG_NOERROR; |
6291ee3c UD |
2936 | int subexp_num, backup_cur_idx, str_idx, null_cnt; |
2937 | re_dfastate_t *cur_state = NULL; | |
2938 | re_node_set *cur_nodes, next_nodes; | |
2939 | re_dfastate_t **backup_state_log; | |
2940 | unsigned int context; | |
2941 | ||
2942 | subexp_num = dfa->nodes[top_node].opr.idx; | |
2943 | /* Extend the buffer if we need. */ | |
951d6408 | 2944 | if (BE (path->alloc < last_str + mctx->max_mb_elem_len + 1, 0)) |
6291ee3c UD |
2945 | { |
2946 | re_dfastate_t **new_array; | |
2947 | int old_alloc = path->alloc; | |
2948 | path->alloc += last_str + mctx->max_mb_elem_len + 1; | |
2949 | new_array = re_realloc (path->array, re_dfastate_t *, path->alloc); | |
6efbd82c | 2950 | if (BE (new_array == NULL, 0)) |
951d6408 UD |
2951 | { |
2952 | path->alloc = old_alloc; | |
2953 | return REG_ESPACE; | |
2954 | } | |
6291ee3c UD |
2955 | path->array = new_array; |
2956 | memset (new_array + old_alloc, '\0', | |
2957 | sizeof (re_dfastate_t *) * (path->alloc - old_alloc)); | |
2958 | } | |
2959 | ||
a334319f | 2960 | str_idx = path->next_idx == 0 ? top_str : path->next_idx; |
6291ee3c UD |
2961 | |
2962 | /* Temporary modify MCTX. */ | |
2963 | backup_state_log = mctx->state_log; | |
56b168be | 2964 | backup_cur_idx = mctx->input.cur_idx; |
6291ee3c | 2965 | mctx->state_log = path->array; |
56b168be | 2966 | mctx->input.cur_idx = str_idx; |
6291ee3c UD |
2967 | |
2968 | /* Setup initial node set. */ | |
56b168be | 2969 | context = re_string_context_at (&mctx->input, str_idx - 1, mctx->eflags); |
6291ee3c UD |
2970 | if (str_idx == top_str) |
2971 | { | |
2972 | err = re_node_set_init_1 (&next_nodes, top_node); | |
2973 | if (BE (err != REG_NOERROR, 0)) | |
2974 | return err; | |
0ce7f49c | 2975 | err = check_arrival_expand_ecl (dfa, &next_nodes, subexp_num, type); |
6291ee3c UD |
2976 | if (BE (err != REG_NOERROR, 0)) |
2977 | { | |
2978 | re_node_set_free (&next_nodes); | |
2979 | return err; | |
2980 | } | |
2981 | } | |
2982 | else | |
2983 | { | |
2984 | cur_state = mctx->state_log[str_idx]; | |
2985 | if (cur_state && cur_state->has_backref) | |
2986 | { | |
2987 | err = re_node_set_init_copy (&next_nodes, &cur_state->nodes); | |
6efbd82c | 2988 | if (BE (err != REG_NOERROR, 0)) |
6291ee3c UD |
2989 | return err; |
2990 | } | |
2991 | else | |
2992 | re_node_set_init_empty (&next_nodes); | |
2993 | } | |
2994 | if (str_idx == top_str || (cur_state && cur_state->has_backref)) | |
2995 | { | |
2996 | if (next_nodes.nelem) | |
2997 | { | |
d2c38eb3 | 2998 | err = expand_bkref_cache (mctx, &next_nodes, str_idx, |
0ce7f49c | 2999 | subexp_num, type); |
6efbd82c | 3000 | if (BE (err != REG_NOERROR, 0)) |
6291ee3c UD |
3001 | { |
3002 | re_node_set_free (&next_nodes); | |
3003 | return err; | |
3004 | } | |
3005 | } | |
3006 | cur_state = re_acquire_state_context (&err, dfa, &next_nodes, context); | |
3007 | if (BE (cur_state == NULL && err != REG_NOERROR, 0)) | |
3008 | { | |
3009 | re_node_set_free (&next_nodes); | |
3010 | return err; | |
3011 | } | |
3012 | mctx->state_log[str_idx] = cur_state; | |
3013 | } | |
3014 | ||
3015 | for (null_cnt = 0; str_idx < last_str && null_cnt <= mctx->max_mb_elem_len;) | |
3016 | { | |
3017 | re_node_set_empty (&next_nodes); | |
3018 | if (mctx->state_log[str_idx + 1]) | |
3019 | { | |
3020 | err = re_node_set_merge (&next_nodes, | |
3021 | &mctx->state_log[str_idx + 1]->nodes); | |
3022 | if (BE (err != REG_NOERROR, 0)) | |
3023 | { | |
3024 | re_node_set_free (&next_nodes); | |
3025 | return err; | |
3026 | } | |
3027 | } | |
3028 | if (cur_state) | |
3029 | { | |
e3a87852 | 3030 | err = check_arrival_add_next_nodes (mctx, str_idx, |
6efbd82c UD |
3031 | &cur_state->non_eps_nodes, |
3032 | &next_nodes); | |
6291ee3c UD |
3033 | if (BE (err != REG_NOERROR, 0)) |
3034 | { | |
3035 | re_node_set_free (&next_nodes); | |
3036 | return err; | |
3037 | } | |
3038 | } | |
3039 | ++str_idx; | |
3040 | if (next_nodes.nelem) | |
3041 | { | |
0ce7f49c | 3042 | err = check_arrival_expand_ecl (dfa, &next_nodes, subexp_num, type); |
6291ee3c UD |
3043 | if (BE (err != REG_NOERROR, 0)) |
3044 | { | |
3045 | re_node_set_free (&next_nodes); | |
3046 | return err; | |
3047 | } | |
d2c38eb3 | 3048 | err = expand_bkref_cache (mctx, &next_nodes, str_idx, |
0ce7f49c | 3049 | subexp_num, type); |
6efbd82c | 3050 | if (BE (err != REG_NOERROR, 0)) |
6291ee3c UD |
3051 | { |
3052 | re_node_set_free (&next_nodes); | |
3053 | return err; | |
3054 | } | |
3055 | } | |
56b168be | 3056 | context = re_string_context_at (&mctx->input, str_idx - 1, mctx->eflags); |
6291ee3c UD |
3057 | cur_state = re_acquire_state_context (&err, dfa, &next_nodes, context); |
3058 | if (BE (cur_state == NULL && err != REG_NOERROR, 0)) | |
3059 | { | |
3060 | re_node_set_free (&next_nodes); | |
3061 | return err; | |
3062 | } | |
3063 | mctx->state_log[str_idx] = cur_state; | |
3064 | null_cnt = cur_state == NULL ? null_cnt + 1 : 0; | |
3065 | } | |
3066 | re_node_set_free (&next_nodes); | |
3067 | cur_nodes = (mctx->state_log[last_str] == NULL ? NULL | |
3068 | : &mctx->state_log[last_str]->nodes); | |
3069 | path->next_idx = str_idx; | |
3070 | ||
3071 | /* Fix MCTX. */ | |
3072 | mctx->state_log = backup_state_log; | |
56b168be | 3073 | mctx->input.cur_idx = backup_cur_idx; |
6291ee3c | 3074 | |
6291ee3c | 3075 | /* Then check the current node set has the node LAST_NODE. */ |
c0d5034e UD |
3076 | if (cur_nodes != NULL && re_node_set_contains (cur_nodes, last_node)) |
3077 | return REG_NOERROR; | |
3078 | ||
3079 | return REG_NOMATCH; | |
6291ee3c UD |
3080 | } |
3081 | ||
3082 | /* Helper functions for check_arrival. */ | |
3083 | ||
3084 | /* Calculate the destination nodes of CUR_NODES at STR_IDX, and append them | |
3085 | to NEXT_NODES. | |
3086 | TODO: This function is similar to the functions transit_state*(), | |
3087 | however this function has many additional works. | |
3088 | Can't we unify them? */ | |
3089 | ||
3090 | static reg_errcode_t | |
a334319f UD |
3091 | check_arrival_add_next_nodes (mctx, str_idx, cur_nodes, next_nodes) |
3092 | re_match_context_t *mctx; | |
3093 | int str_idx; | |
3094 | re_node_set *cur_nodes, *next_nodes; | |
6291ee3c | 3095 | { |
c42b4152 | 3096 | const re_dfa_t *const dfa = mctx->dfa; |
d2c38eb3 | 3097 | int result; |
6291ee3c | 3098 | int cur_idx; |
c0c9615c | 3099 | reg_errcode_t err = REG_NOERROR; |
6291ee3c UD |
3100 | re_node_set union_set; |
3101 | re_node_set_init_empty (&union_set); | |
3102 | for (cur_idx = 0; cur_idx < cur_nodes->nelem; ++cur_idx) | |
3103 | { | |
3104 | int naccepted = 0; | |
3105 | int cur_node = cur_nodes->elems[cur_idx]; | |
a334319f | 3106 | #ifdef DEBUG |
02f3550c | 3107 | re_token_type_t type = dfa->nodes[cur_node].type; |
5cf1ec52 UD |
3108 | assert (!IS_EPSILON_NODE (type)); |
3109 | #endif | |
6291ee3c UD |
3110 | #ifdef RE_ENABLE_I18N |
3111 | /* If the node may accept `multi byte'. */ | |
02f3550c | 3112 | if (dfa->nodes[cur_node].accept_mb) |
6291ee3c | 3113 | { |
56b168be | 3114 | naccepted = check_node_accept_bytes (dfa, cur_node, &mctx->input, |
6291ee3c UD |
3115 | str_idx); |
3116 | if (naccepted > 1) | |
3117 | { | |
3118 | re_dfastate_t *dest_state; | |
3119 | int next_node = dfa->nexts[cur_node]; | |
3120 | int next_idx = str_idx + naccepted; | |
3121 | dest_state = mctx->state_log[next_idx]; | |
3122 | re_node_set_empty (&union_set); | |
3123 | if (dest_state) | |
3124 | { | |
3125 | err = re_node_set_merge (&union_set, &dest_state->nodes); | |
3126 | if (BE (err != REG_NOERROR, 0)) | |
3127 | { | |
3128 | re_node_set_free (&union_set); | |
3129 | return err; | |
3130 | } | |
6291ee3c | 3131 | } |
d2c38eb3 UD |
3132 | result = re_node_set_insert (&union_set, next_node); |
3133 | if (BE (result < 0, 0)) | |
6291ee3c | 3134 | { |
44777771 UD |
3135 | re_node_set_free (&union_set); |
3136 | return REG_ESPACE; | |
6291ee3c UD |
3137 | } |
3138 | mctx->state_log[next_idx] = re_acquire_state (&err, dfa, | |
3139 | &union_set); | |
3140 | if (BE (mctx->state_log[next_idx] == NULL | |
3141 | && err != REG_NOERROR, 0)) | |
3142 | { | |
3143 | re_node_set_free (&union_set); | |
3144 | return err; | |
3145 | } | |
3146 | } | |
3147 | } | |
3148 | #endif /* RE_ENABLE_I18N */ | |
3149 | if (naccepted | |
e3a87852 | 3150 | || check_node_accept (mctx, dfa->nodes + cur_node, str_idx)) |
6291ee3c | 3151 | { |
d2c38eb3 UD |
3152 | result = re_node_set_insert (next_nodes, dfa->nexts[cur_node]); |
3153 | if (BE (result < 0, 0)) | |
6291ee3c UD |
3154 | { |
3155 | re_node_set_free (&union_set); | |
3156 | return REG_ESPACE; | |
3157 | } | |
3158 | } | |
3159 | } | |
3160 | re_node_set_free (&union_set); | |
3161 | return REG_NOERROR; | |
3162 | } | |
3163 | ||
3164 | /* For all the nodes in CUR_NODES, add the epsilon closures of them to | |
3165 | CUR_NODES, however exclude the nodes which are: | |
3166 | - inside the sub expression whose number is EX_SUBEXP, if FL_OPEN. | |
3167 | - out of the sub expression whose number is EX_SUBEXP, if !FL_OPEN. | |
3168 | */ | |
3169 | ||
3170 | static reg_errcode_t | |
a334319f | 3171 | check_arrival_expand_ecl (dfa, cur_nodes, ex_subexp, type) |
6efbd82c | 3172 | const re_dfa_t *dfa; |
a334319f UD |
3173 | re_node_set *cur_nodes; |
3174 | int ex_subexp, type; | |
6291ee3c UD |
3175 | { |
3176 | reg_errcode_t err; | |
3177 | int idx, outside_node; | |
3178 | re_node_set new_nodes; | |
3179 | #ifdef DEBUG | |
3180 | assert (cur_nodes->nelem); | |
3181 | #endif | |
3182 | err = re_node_set_alloc (&new_nodes, cur_nodes->nelem); | |
3183 | if (BE (err != REG_NOERROR, 0)) | |
3184 | return err; | |
3185 | /* Create a new node set NEW_NODES with the nodes which are epsilon | |
3186 | closures of the node in CUR_NODES. */ | |
3187 | ||
3188 | for (idx = 0; idx < cur_nodes->nelem; ++idx) | |
3189 | { | |
3190 | int cur_node = cur_nodes->elems[idx]; | |
6efbd82c | 3191 | const re_node_set *eclosure = dfa->eclosures + cur_node; |
0ce7f49c | 3192 | outside_node = find_subexp_node (dfa, eclosure, ex_subexp, type); |
6291ee3c UD |
3193 | if (outside_node == -1) |
3194 | { | |
3195 | /* There are no problematic nodes, just merge them. */ | |
3196 | err = re_node_set_merge (&new_nodes, eclosure); | |
3197 | if (BE (err != REG_NOERROR, 0)) | |
3198 | { | |
3199 | re_node_set_free (&new_nodes); | |
3200 | return err; | |
3201 | } | |
3202 | } | |
3203 | else | |
3204 | { | |
3205 | /* There are problematic nodes, re-calculate incrementally. */ | |
3206 | err = check_arrival_expand_ecl_sub (dfa, &new_nodes, cur_node, | |
0ce7f49c | 3207 | ex_subexp, type); |
6291ee3c UD |
3208 | if (BE (err != REG_NOERROR, 0)) |
3209 | { | |
3210 | re_node_set_free (&new_nodes); | |
3211 | return err; | |
3212 | } | |
3213 | } | |
3214 | } | |
3215 | re_node_set_free (cur_nodes); | |
3216 | *cur_nodes = new_nodes; | |
3217 | return REG_NOERROR; | |
3218 | } | |
3219 | ||
3220 | /* Helper function for check_arrival_expand_ecl. | |
3221 | Check incrementally the epsilon closure of TARGET, and if it isn't | |
3222 | problematic append it to DST_NODES. */ | |
3223 | ||
3224 | static reg_errcode_t | |
a334319f | 3225 | check_arrival_expand_ecl_sub (dfa, dst_nodes, target, ex_subexp, type) |
6efbd82c | 3226 | const re_dfa_t *dfa; |
a334319f UD |
3227 | int target, ex_subexp, type; |
3228 | re_node_set *dst_nodes; | |
6291ee3c | 3229 | { |
0ce7f49c | 3230 | int cur_node; |
6291ee3c UD |
3231 | for (cur_node = target; !re_node_set_contains (dst_nodes, cur_node);) |
3232 | { | |
3233 | int err; | |
6291ee3c | 3234 | |
0ce7f49c | 3235 | if (dfa->nodes[cur_node].type == type |
6291ee3c UD |
3236 | && dfa->nodes[cur_node].opr.idx == ex_subexp) |
3237 | { | |
0ce7f49c | 3238 | if (type == OP_CLOSE_SUBEXP) |
6291ee3c UD |
3239 | { |
3240 | err = re_node_set_insert (dst_nodes, cur_node); | |
3241 | if (BE (err == -1, 0)) | |
3242 | return REG_ESPACE; | |
3243 | } | |
3244 | break; | |
3245 | } | |
3246 | err = re_node_set_insert (dst_nodes, cur_node); | |
3247 | if (BE (err == -1, 0)) | |
3248 | return REG_ESPACE; | |
3249 | if (dfa->edests[cur_node].nelem == 0) | |
3250 | break; | |
3251 | if (dfa->edests[cur_node].nelem == 2) | |
3252 | { | |
3253 | err = check_arrival_expand_ecl_sub (dfa, dst_nodes, | |
3254 | dfa->edests[cur_node].elems[1], | |
0ce7f49c | 3255 | ex_subexp, type); |
6291ee3c UD |
3256 | if (BE (err != REG_NOERROR, 0)) |
3257 | return err; | |
3258 | } | |
3259 | cur_node = dfa->edests[cur_node].elems[0]; | |
3260 | } | |
3261 | return REG_NOERROR; | |
3262 | } | |
3263 | ||
3264 | ||
3265 | /* For all the back references in the current state, calculate the | |
3266 | destination of the back references by the appropriate entry | |
3267 | in MCTX->BKREF_ENTS. */ | |
3268 | ||
3269 | static reg_errcode_t | |
a334319f UD |
3270 | expand_bkref_cache (mctx, cur_nodes, cur_str, subexp_num, |
3271 | type) | |
3272 | re_match_context_t *mctx; | |
3273 | int cur_str, subexp_num, type; | |
3274 | re_node_set *cur_nodes; | |
6291ee3c | 3275 | { |
76b864c8 | 3276 | const re_dfa_t *const dfa = mctx->dfa; |
6291ee3c | 3277 | reg_errcode_t err; |
e40a38b3 UD |
3278 | int cache_idx_start = search_cur_bkref_entry (mctx, cur_str); |
3279 | struct re_backref_cache_entry *ent; | |
6291ee3c | 3280 | |
e40a38b3 UD |
3281 | if (cache_idx_start == -1) |
3282 | return REG_NOERROR; | |
3283 | ||
3284 | restart: | |
3285 | ent = mctx->bkref_ents + cache_idx_start; | |
3286 | do | |
6291ee3c UD |
3287 | { |
3288 | int to_idx, next_node; | |
e40a38b3 | 3289 | |
6291ee3c UD |
3290 | /* Is this entry ENT is appropriate? */ |
3291 | if (!re_node_set_contains (cur_nodes, ent->node)) | |
3292 | continue; /* No. */ | |
3293 | ||
3294 | to_idx = cur_str + ent->subexp_to - ent->subexp_from; | |
3295 | /* Calculate the destination of the back reference, and append it | |
3296 | to MCTX->STATE_LOG. */ | |
3297 | if (to_idx == cur_str) | |
3298 | { | |
3299 | /* The backreference did epsilon transit, we must re-check all the | |
3300 | node in the current state. */ | |
3301 | re_node_set new_dests; | |
3302 | reg_errcode_t err2, err3; | |
3303 | next_node = dfa->edests[ent->node].elems[0]; | |
3304 | if (re_node_set_contains (cur_nodes, next_node)) | |
3305 | continue; | |
3306 | err = re_node_set_init_1 (&new_dests, next_node); | |
0ce7f49c | 3307 | err2 = check_arrival_expand_ecl (dfa, &new_dests, subexp_num, type); |
6291ee3c UD |
3308 | err3 = re_node_set_merge (cur_nodes, &new_dests); |
3309 | re_node_set_free (&new_dests); | |
3310 | if (BE (err != REG_NOERROR || err2 != REG_NOERROR | |
3311 | || err3 != REG_NOERROR, 0)) | |
3312 | { | |
3313 | err = (err != REG_NOERROR ? err | |
3314 | : (err2 != REG_NOERROR ? err2 : err3)); | |
3315 | return err; | |
3316 | } | |
3317 | /* TODO: It is still inefficient... */ | |
e40a38b3 | 3318 | goto restart; |
6291ee3c UD |
3319 | } |
3320 | else | |
3321 | { | |
3322 | re_node_set union_set; | |
3323 | next_node = dfa->nexts[ent->node]; | |
3324 | if (mctx->state_log[to_idx]) | |
3325 | { | |
3326 | int ret; | |
3327 | if (re_node_set_contains (&mctx->state_log[to_idx]->nodes, | |
3328 | next_node)) | |
3329 | continue; | |
3330 | err = re_node_set_init_copy (&union_set, | |
3331 | &mctx->state_log[to_idx]->nodes); | |
3332 | ret = re_node_set_insert (&union_set, next_node); | |
3333 | if (BE (err != REG_NOERROR || ret < 0, 0)) | |
3334 | { | |
3335 | re_node_set_free (&union_set); | |
3336 | err = err != REG_NOERROR ? err : REG_ESPACE; | |
3337 | return err; | |
3338 | } | |
3339 | } | |
3340 | else | |
3341 | { | |
3342 | err = re_node_set_init_1 (&union_set, next_node); | |
3343 | if (BE (err != REG_NOERROR, 0)) | |
3344 | return err; | |
3345 | } | |
3346 | mctx->state_log[to_idx] = re_acquire_state (&err, dfa, &union_set); | |
3347 | re_node_set_free (&union_set); | |
3348 | if (BE (mctx->state_log[to_idx] == NULL | |
3349 | && err != REG_NOERROR, 0)) | |
3350 | return err; | |
3351 | } | |
3352 | } | |
e40a38b3 | 3353 | while (ent++->more); |
6291ee3c UD |
3354 | return REG_NOERROR; |
3355 | } | |
3356 | ||
3357 | /* Build transition table for the state. | |
ab4b89fe | 3358 | Return 1 if succeeded, otherwise return NULL. */ |
6291ee3c | 3359 | |
ab4b89fe | 3360 | static int |
a334319f | 3361 | build_trtable (dfa, state) |
76b864c8 UD |
3362 | const re_dfa_t *dfa; |
3363 | re_dfastate_t *state; | |
6291ee3c UD |
3364 | { |
3365 | reg_errcode_t err; | |
ab4b89fe | 3366 | int i, j, ch, need_word_trtable = 0; |
2c05d33f UD |
3367 | bitset_word_t elem, mask; |
3368 | bool dests_node_malloced = false; | |
3369 | bool dest_states_malloced = false; | |
6291ee3c UD |
3370 | int ndests; /* Number of the destination states from `state'. */ |
3371 | re_dfastate_t **trtable; | |
3372 | re_dfastate_t **dest_states = NULL, **dest_states_word, **dest_states_nl; | |
3373 | re_node_set follows, *dests_node; | |
2c05d33f UD |
3374 | bitset_t *dests_ch; |
3375 | bitset_t acceptable; | |
3376 | ||
3377 | struct dests_alloc | |
3378 | { | |
3379 | re_node_set dests_node[SBC_MAX]; | |
3380 | bitset_t dests_ch[SBC_MAX]; | |
3381 | } *dests_alloc; | |
3b0bdc72 UD |
3382 | |
3383 | /* We build DFA states which corresponds to the destination nodes | |
3384 | from `state'. `dests_node[i]' represents the nodes which i-th | |
3385 | destination state contains, and `dests_ch[i]' represents the | |
3386 | characters which i-th destination state accepts. */ | |
2c05d33f UD |
3387 | if (__libc_use_alloca (sizeof (struct dests_alloc))) |
3388 | dests_alloc = (struct dests_alloc *) alloca (sizeof (struct dests_alloc)); | |
05dab910 | 3389 | else |
05dab910 | 3390 | { |
2c05d33f UD |
3391 | dests_alloc = re_malloc (struct dests_alloc, 1); |
3392 | if (BE (dests_alloc == NULL, 0)) | |
ab4b89fe | 3393 | return 0; |
2c05d33f | 3394 | dests_node_malloced = true; |
05dab910 | 3395 | } |
2c05d33f UD |
3396 | dests_node = dests_alloc->dests_node; |
3397 | dests_ch = dests_alloc->dests_ch; | |
3b0bdc72 UD |
3398 | |
3399 | /* Initialize transiton table. */ | |
ab4b89fe | 3400 | state->word_trtable = state->trtable = NULL; |
3b0bdc72 UD |
3401 | |
3402 | /* At first, group all nodes belonging to `state' into several | |
3403 | destinations. */ | |
56b168be | 3404 | ndests = group_nodes_into_DFAstates (dfa, state, dests_node, dests_ch); |
bc15410e | 3405 | if (BE (ndests <= 0, 0)) |
3b0bdc72 | 3406 | { |
05dab910 | 3407 | if (dests_node_malloced) |
2c05d33f | 3408 | free (dests_alloc); |
ab4b89fe | 3409 | /* Return 0 in case of an error, 1 otherwise. */ |
05dab910 | 3410 | if (ndests == 0) |
c13c99fa | 3411 | { |
3ce12656 | 3412 | state->trtable = (re_dfastate_t **) |
ab4b89fe UD |
3413 | calloc (sizeof (re_dfastate_t *), SBC_MAX); |
3414 | return 1; | |
c13c99fa | 3415 | } |
ab4b89fe | 3416 | return 0; |
3b0bdc72 UD |
3417 | } |
3418 | ||
a9388965 | 3419 | err = re_node_set_alloc (&follows, ndests + 1); |
05dab910 RM |
3420 | if (BE (err != REG_NOERROR, 0)) |
3421 | goto out_free; | |
3422 | ||
2c05d33f | 3423 | if (__libc_use_alloca ((sizeof (re_node_set) + sizeof (bitset_t)) * SBC_MAX |
05dab910 RM |
3424 | + ndests * 3 * sizeof (re_dfastate_t *))) |
3425 | dest_states = (re_dfastate_t **) | |
ab4b89fe | 3426 | alloca (ndests * 3 * sizeof (re_dfastate_t *)); |
05dab910 | 3427 | else |
05dab910 RM |
3428 | { |
3429 | dest_states = (re_dfastate_t **) | |
ab4b89fe | 3430 | malloc (ndests * 3 * sizeof (re_dfastate_t *)); |
05dab910 RM |
3431 | if (BE (dest_states == NULL, 0)) |
3432 | { | |
3433 | out_free: | |
3434 | if (dest_states_malloced) | |
3435 | free (dest_states); | |
3436 | re_node_set_free (&follows); | |
3437 | for (i = 0; i < ndests; ++i) | |
3438 | re_node_set_free (dests_node + i); | |
05dab910 | 3439 | if (dests_node_malloced) |
2c05d33f | 3440 | free (dests_alloc); |
ab4b89fe | 3441 | return 0; |
05dab910 | 3442 | } |
2c05d33f | 3443 | dest_states_malloced = true; |
05dab910 RM |
3444 | } |
3445 | dest_states_word = dest_states + ndests; | |
3446 | dest_states_nl = dest_states_word + ndests; | |
3447 | bitset_empty (acceptable); | |
a9388965 | 3448 | |
3b0bdc72 UD |
3449 | /* Then build the states for all destinations. */ |
3450 | for (i = 0; i < ndests; ++i) | |
3451 | { | |
3452 | int next_node; | |
3453 | re_node_set_empty (&follows); | |
3454 | /* Merge the follows of this destination states. */ | |
3455 | for (j = 0; j < dests_node[i].nelem; ++j) | |
15a7d175 UD |
3456 | { |
3457 | next_node = dfa->nexts[dests_node[i].elems[j]]; | |
3458 | if (next_node != -1) | |
3459 | { | |
3460 | err = re_node_set_merge (&follows, dfa->eclosures + next_node); | |
3461 | if (BE (err != REG_NOERROR, 0)) | |
05dab910 | 3462 | goto out_free; |
15a7d175 UD |
3463 | } |
3464 | } | |
a9388965 | 3465 | dest_states[i] = re_acquire_state_context (&err, dfa, &follows, 0); |
bc15410e | 3466 | if (BE (dest_states[i] == NULL && err != REG_NOERROR, 0)) |
15a7d175 | 3467 | goto out_free; |
3b0bdc72 | 3468 | /* If the new state has context constraint, |
15a7d175 | 3469 | build appropriate states for these contexts. */ |
3b0bdc72 | 3470 | if (dest_states[i]->has_constraint) |
15a7d175 UD |
3471 | { |
3472 | dest_states_word[i] = re_acquire_state_context (&err, dfa, &follows, | |
3473 | CONTEXT_WORD); | |
3474 | if (BE (dest_states_word[i] == NULL && err != REG_NOERROR, 0)) | |
3475 | goto out_free; | |
3ce12656 | 3476 | |
ab4b89fe UD |
3477 | if (dest_states[i] != dest_states_word[i] && dfa->mb_cur_max > 1) |
3478 | need_word_trtable = 1; | |
3ce12656 | 3479 | |
15a7d175 UD |
3480 | dest_states_nl[i] = re_acquire_state_context (&err, dfa, &follows, |
3481 | CONTEXT_NEWLINE); | |
3482 | if (BE (dest_states_nl[i] == NULL && err != REG_NOERROR, 0)) | |
3483 | goto out_free; | |
3ce12656 | 3484 | } |
3b0bdc72 | 3485 | else |
15a7d175 UD |
3486 | { |
3487 | dest_states_word[i] = dest_states[i]; | |
3488 | dest_states_nl[i] = dest_states[i]; | |
3489 | } | |
3b0bdc72 UD |
3490 | bitset_merge (acceptable, dests_ch[i]); |
3491 | } | |
3492 | ||
ab4b89fe | 3493 | if (!BE (need_word_trtable, 0)) |
3ce12656 UD |
3494 | { |
3495 | /* We don't care about whether the following character is a word | |
3496 | character, or we are in a single-byte character set so we can | |
3497 | discern by looking at the character code: allocate a | |
3498 | 256-entry transition table. */ | |
ab4b89fe UD |
3499 | trtable = state->trtable = |
3500 | (re_dfastate_t **) calloc (sizeof (re_dfastate_t *), SBC_MAX); | |
3ce12656 UD |
3501 | if (BE (trtable == NULL, 0)) |
3502 | goto out_free; | |
3503 | ||
3504 | /* For all characters ch...: */ | |
2c05d33f UD |
3505 | for (i = 0; i < BITSET_WORDS; ++i) |
3506 | for (ch = i * BITSET_WORD_BITS, elem = acceptable[i], mask = 1; | |
3ce12656 UD |
3507 | elem; |
3508 | mask <<= 1, elem >>= 1, ++ch) | |
3509 | if (BE (elem & 1, 0)) | |
3510 | { | |
3511 | /* There must be exactly one destination which accepts | |
3512 | character ch. See group_nodes_into_DFAstates. */ | |
3513 | for (j = 0; (dests_ch[j][i] & mask) == 0; ++j) | |
3514 | ; | |
6b6557e8 | 3515 | |
3ce12656 | 3516 | /* j-th destination accepts the word character ch. */ |
56b168be | 3517 | if (dfa->word_char[i] & mask) |
3ce12656 UD |
3518 | trtable[ch] = dest_states_word[j]; |
3519 | else | |
3520 | trtable[ch] = dest_states[j]; | |
3521 | } | |
3522 | } | |
3523 | else | |
3524 | { | |
3525 | /* We care about whether the following character is a word | |
3526 | character, and we are in a multi-byte character set: discern | |
3527 | by looking at the character code: build two 256-entry | |
3528 | transition tables, one starting at trtable[0] and one | |
3529 | starting at trtable[SBC_MAX]. */ | |
ab4b89fe UD |
3530 | trtable = state->word_trtable = |
3531 | (re_dfastate_t **) calloc (sizeof (re_dfastate_t *), 2 * SBC_MAX); | |
3ce12656 UD |
3532 | if (BE (trtable == NULL, 0)) |
3533 | goto out_free; | |
6b6557e8 | 3534 | |
3ce12656 | 3535 | /* For all characters ch...: */ |
2c05d33f UD |
3536 | for (i = 0; i < BITSET_WORDS; ++i) |
3537 | for (ch = i * BITSET_WORD_BITS, elem = acceptable[i], mask = 1; | |
3ce12656 UD |
3538 | elem; |
3539 | mask <<= 1, elem >>= 1, ++ch) | |
3540 | if (BE (elem & 1, 0)) | |
3541 | { | |
3542 | /* There must be exactly one destination which accepts | |
3543 | character ch. See group_nodes_into_DFAstates. */ | |
3544 | for (j = 0; (dests_ch[j][i] & mask) == 0; ++j) | |
3545 | ; | |
6b6557e8 | 3546 | |
3ce12656 UD |
3547 | /* j-th destination accepts the word character ch. */ |
3548 | trtable[ch] = dest_states[j]; | |
3549 | trtable[ch + SBC_MAX] = dest_states_word[j]; | |
3550 | } | |
3551 | } | |
6b6557e8 | 3552 | |
3b0bdc72 | 3553 | /* new line */ |
c202c2c5 UD |
3554 | if (bitset_contain (acceptable, NEWLINE_CHAR)) |
3555 | { | |
3556 | /* The current state accepts newline character. */ | |
3ce12656 UD |
3557 | for (j = 0; j < ndests; ++j) |
3558 | if (bitset_contain (dests_ch[j], NEWLINE_CHAR)) | |
15a7d175 UD |
3559 | { |
3560 | /* k-th destination accepts newline character. */ | |
3ce12656 | 3561 | trtable[NEWLINE_CHAR] = dest_states_nl[j]; |
ab4b89fe | 3562 | if (need_word_trtable) |
3ce12656 | 3563 | trtable[NEWLINE_CHAR + SBC_MAX] = dest_states_nl[j]; |
15a7d175 UD |
3564 | /* There must be only one destination which accepts |
3565 | newline. See group_nodes_into_DFAstates. */ | |
3566 | break; | |
3567 | } | |
c202c2c5 | 3568 | } |
3b0bdc72 | 3569 | |
05dab910 RM |
3570 | if (dest_states_malloced) |
3571 | free (dest_states); | |
3b0bdc72 UD |
3572 | |
3573 | re_node_set_free (&follows); | |
3574 | for (i = 0; i < ndests; ++i) | |
3575 | re_node_set_free (dests_node + i); | |
3576 | ||
05dab910 | 3577 | if (dests_node_malloced) |
2c05d33f | 3578 | free (dests_alloc); |
3b0bdc72 | 3579 | |
ab4b89fe | 3580 | return 1; |
3b0bdc72 UD |
3581 | } |
3582 | ||
3583 | /* Group all nodes belonging to STATE into several destinations. | |
3584 | Then for all destinations, set the nodes belonging to the destination | |
3585 | to DESTS_NODE[i] and set the characters accepted by the destination | |
3586 | to DEST_CH[i]. This function return the number of destinations. */ | |
3587 | ||
3588 | static int | |
a334319f | 3589 | group_nodes_into_DFAstates (dfa, state, dests_node, dests_ch) |
01ed6ceb UD |
3590 | const re_dfa_t *dfa; |
3591 | const re_dfastate_t *state; | |
3592 | re_node_set *dests_node; | |
2c05d33f | 3593 | bitset_t *dests_ch; |
3b0bdc72 | 3594 | { |
a9388965 | 3595 | reg_errcode_t err; |
d2c38eb3 | 3596 | int result; |
3b0bdc72 UD |
3597 | int i, j, k; |
3598 | int ndests; /* Number of the destinations from `state'. */ | |
2c05d33f | 3599 | bitset_t accepts; /* Characters a node can accept. */ |
3b0bdc72 UD |
3600 | const re_node_set *cur_nodes = &state->nodes; |
3601 | bitset_empty (accepts); | |
3602 | ndests = 0; | |
3603 | ||
3604 | /* For all the nodes belonging to `state', */ | |
3605 | for (i = 0; i < cur_nodes->nelem; ++i) | |
3606 | { | |
3b0bdc72 UD |
3607 | re_token_t *node = &dfa->nodes[cur_nodes->elems[i]]; |
3608 | re_token_type_t type = node->type; | |
485d775d | 3609 | unsigned int constraint = node->constraint; |
3b0bdc72 UD |
3610 | |
3611 | /* Enumerate all single byte character this node can accept. */ | |
3612 | if (type == CHARACTER) | |
15a7d175 | 3613 | bitset_set (accepts, node->opr.c); |
3b0bdc72 | 3614 | else if (type == SIMPLE_BRACKET) |
15a7d175 UD |
3615 | { |
3616 | bitset_merge (accepts, node->opr.sbcset); | |
3617 | } | |
3b0bdc72 | 3618 | else if (type == OP_PERIOD) |
15a7d175 | 3619 | { |
65e6becf UD |
3620 | #ifdef RE_ENABLE_I18N |
3621 | if (dfa->mb_cur_max > 1) | |
3622 | bitset_merge (accepts, dfa->sb_char); | |
3623 | else | |
0ce7f49c | 3624 | #endif |
65e6becf | 3625 | bitset_set_all (accepts); |
56b168be | 3626 | if (!(dfa->syntax & RE_DOT_NEWLINE)) |
15a7d175 | 3627 | bitset_clear (accepts, '\n'); |
56b168be | 3628 | if (dfa->syntax & RE_DOT_NOT_NULL) |
15a7d175 UD |
3629 | bitset_clear (accepts, '\0'); |
3630 | } | |
c0d5034e | 3631 | #ifdef RE_ENABLE_I18N |
ad7f28c2 UD |
3632 | else if (type == OP_UTF8_PERIOD) |
3633 | { | |
2c05d33f | 3634 | memset (accepts, '\xff', sizeof (bitset_t) / 2); |
56b168be | 3635 | if (!(dfa->syntax & RE_DOT_NEWLINE)) |
ad7f28c2 | 3636 | bitset_clear (accepts, '\n'); |
56b168be | 3637 | if (dfa->syntax & RE_DOT_NOT_NULL) |
ad7f28c2 UD |
3638 | bitset_clear (accepts, '\0'); |
3639 | } | |
c0d5034e | 3640 | #endif |
3b0bdc72 | 3641 | else |
15a7d175 | 3642 | continue; |
3b0bdc72 UD |
3643 | |
3644 | /* Check the `accepts' and sift the characters which are not | |
15a7d175 | 3645 | match it the context. */ |
3b0bdc72 | 3646 | if (constraint) |
15a7d175 | 3647 | { |
15a7d175 UD |
3648 | if (constraint & NEXT_NEWLINE_CONSTRAINT) |
3649 | { | |
2c05d33f | 3650 | bool accepts_newline = bitset_contain (accepts, NEWLINE_CHAR); |
15a7d175 UD |
3651 | bitset_empty (accepts); |
3652 | if (accepts_newline) | |
3653 | bitset_set (accepts, NEWLINE_CHAR); | |
3654 | else | |
3655 | continue; | |
3656 | } | |
66b110e8 UD |
3657 | if (constraint & NEXT_ENDBUF_CONSTRAINT) |
3658 | { | |
3659 | bitset_empty (accepts); | |
3660 | continue; | |
3661 | } | |
c13c99fa | 3662 | |
66b110e8 | 3663 | if (constraint & NEXT_WORD_CONSTRAINT) |
65e6becf | 3664 | { |
2c05d33f | 3665 | bitset_word_t any_set = 0; |
1cef7b3c UD |
3666 | if (type == CHARACTER && !node->word_char) |
3667 | { | |
3668 | bitset_empty (accepts); | |
3669 | continue; | |
3670 | } | |
65e6becf UD |
3671 | #ifdef RE_ENABLE_I18N |
3672 | if (dfa->mb_cur_max > 1) | |
2c05d33f | 3673 | for (j = 0; j < BITSET_WORDS; ++j) |
457beec8 | 3674 | any_set |= (accepts[j] &= (dfa->word_char[j] | ~dfa->sb_char[j])); |
65e6becf UD |
3675 | else |
3676 | #endif | |
2c05d33f | 3677 | for (j = 0; j < BITSET_WORDS; ++j) |
457beec8 UD |
3678 | any_set |= (accepts[j] &= dfa->word_char[j]); |
3679 | if (!any_set) | |
3680 | continue; | |
65e6becf | 3681 | } |
66b110e8 | 3682 | if (constraint & NEXT_NOTWORD_CONSTRAINT) |
65e6becf | 3683 | { |
2c05d33f | 3684 | bitset_word_t any_set = 0; |
1cef7b3c UD |
3685 | if (type == CHARACTER && node->word_char) |
3686 | { | |
3687 | bitset_empty (accepts); | |
3688 | continue; | |
3689 | } | |
65e6becf UD |
3690 | #ifdef RE_ENABLE_I18N |
3691 | if (dfa->mb_cur_max > 1) | |
2c05d33f | 3692 | for (j = 0; j < BITSET_WORDS; ++j) |
457beec8 | 3693 | any_set |= (accepts[j] &= ~(dfa->word_char[j] & dfa->sb_char[j])); |
65e6becf UD |
3694 | else |
3695 | #endif | |
2c05d33f | 3696 | for (j = 0; j < BITSET_WORDS; ++j) |
457beec8 UD |
3697 | any_set |= (accepts[j] &= ~dfa->word_char[j]); |
3698 | if (!any_set) | |
3699 | continue; | |
65e6becf | 3700 | } |
15a7d175 | 3701 | } |
3b0bdc72 UD |
3702 | |
3703 | /* Then divide `accepts' into DFA states, or create a new | |
457beec8 | 3704 | state. Above, we make sure that accepts is not empty. */ |
3b0bdc72 | 3705 | for (j = 0; j < ndests; ++j) |
15a7d175 | 3706 | { |
2c05d33f UD |
3707 | bitset_t intersec; /* Intersection sets, see below. */ |
3708 | bitset_t remains; | |
15a7d175 | 3709 | /* Flags, see below. */ |
2c05d33f | 3710 | bitset_word_t has_intersec, not_subset, not_consumed; |
15a7d175 UD |
3711 | |
3712 | /* Optimization, skip if this state doesn't accept the character. */ | |
3713 | if (type == CHARACTER && !bitset_contain (dests_ch[j], node->opr.c)) | |
3714 | continue; | |
3715 | ||
3716 | /* Enumerate the intersection set of this state and `accepts'. */ | |
3717 | has_intersec = 0; | |
2c05d33f | 3718 | for (k = 0; k < BITSET_WORDS; ++k) |
15a7d175 UD |
3719 | has_intersec |= intersec[k] = accepts[k] & dests_ch[j][k]; |
3720 | /* And skip if the intersection set is empty. */ | |
3721 | if (!has_intersec) | |
3722 | continue; | |
3723 | ||
3724 | /* Then check if this state is a subset of `accepts'. */ | |
3725 | not_subset = not_consumed = 0; | |
2c05d33f | 3726 | for (k = 0; k < BITSET_WORDS; ++k) |
15a7d175 UD |
3727 | { |
3728 | not_subset |= remains[k] = ~accepts[k] & dests_ch[j][k]; | |
3729 | not_consumed |= accepts[k] = accepts[k] & ~dests_ch[j][k]; | |
3730 | } | |
3731 | ||
3732 | /* If this state isn't a subset of `accepts', create a | |
3733 | new group state, which has the `remains'. */ | |
3734 | if (not_subset) | |
3735 | { | |
3736 | bitset_copy (dests_ch[ndests], remains); | |
3737 | bitset_copy (dests_ch[j], intersec); | |
3738 | err = re_node_set_init_copy (dests_node + ndests, &dests_node[j]); | |
3739 | if (BE (err != REG_NOERROR, 0)) | |
3740 | goto error_return; | |
3741 | ++ndests; | |
3742 | } | |
3743 | ||
3744 | /* Put the position in the current group. */ | |
d2c38eb3 UD |
3745 | result = re_node_set_insert (&dests_node[j], cur_nodes->elems[i]); |
3746 | if (BE (result < 0, 0)) | |
15a7d175 UD |
3747 | goto error_return; |
3748 | ||
3749 | /* If all characters are consumed, go to next node. */ | |
3750 | if (!not_consumed) | |
3751 | break; | |
3752 | } | |
3b0bdc72 UD |
3753 | /* Some characters remain, create a new group. */ |
3754 | if (j == ndests) | |
15a7d175 UD |
3755 | { |
3756 | bitset_copy (dests_ch[ndests], accepts); | |
3757 | err = re_node_set_init_1 (dests_node + ndests, cur_nodes->elems[i]); | |
3758 | if (BE (err != REG_NOERROR, 0)) | |
3759 | goto error_return; | |
3760 | ++ndests; | |
3761 | bitset_empty (accepts); | |
3762 | } | |
3b0bdc72 UD |
3763 | } |
3764 | return ndests; | |
1b2c2628 UD |
3765 | error_return: |
3766 | for (j = 0; j < ndests; ++j) | |
3767 | re_node_set_free (dests_node + j); | |
3768 | return -1; | |
3b0bdc72 UD |
3769 | } |
3770 | ||
434d3784 UD |
3771 | #ifdef RE_ENABLE_I18N |
3772 | /* Check how many bytes the node `dfa->nodes[node_idx]' accepts. | |
3773 | Return the number of the bytes the node accepts. | |
3774 | STR_IDX is the current index of the input string. | |
3775 | ||
3776 | This function handles the nodes which can accept one character, or | |
3777 | one collating element like '.', '[a-z]', opposite to the other nodes | |
3778 | can only accept one byte. */ | |
3b0bdc72 UD |
3779 | |
3780 | static int | |
a334319f | 3781 | check_node_accept_bytes (dfa, node_idx, input, str_idx) |
c42b4152 UD |
3782 | const re_dfa_t *dfa; |
3783 | int node_idx, str_idx; | |
3784 | const re_string_t *input; | |
3b0bdc72 | 3785 | { |
3b0bdc72 | 3786 | const re_token_t *node = dfa->nodes + node_idx; |
ad7f28c2 | 3787 | int char_len, elem_len; |
434d3784 | 3788 | int i; |
ad7f28c2 UD |
3789 | |
3790 | if (BE (node->type == OP_UTF8_PERIOD, 0)) | |
3791 | { | |
3792 | unsigned char c = re_string_byte_at (input, str_idx), d; | |
3793 | if (BE (c < 0xc2, 1)) | |
3794 | return 0; | |
3795 | ||
3796 | if (str_idx + 2 > input->len) | |
3797 | return 0; | |
3798 | ||
3799 | d = re_string_byte_at (input, str_idx + 1); | |
3800 | if (c < 0xe0) | |
3801 | return (d < 0x80 || d > 0xbf) ? 0 : 2; | |
3802 | else if (c < 0xf0) | |
3803 | { | |
3804 | char_len = 3; | |
3805 | if (c == 0xe0 && d < 0xa0) | |
3806 | return 0; | |
3807 | } | |
3808 | else if (c < 0xf8) | |
3809 | { | |
3810 | char_len = 4; | |
3811 | if (c == 0xf0 && d < 0x90) | |
3812 | return 0; | |
3813 | } | |
3814 | else if (c < 0xfc) | |
3815 | { | |
3816 | char_len = 5; | |
3817 | if (c == 0xf8 && d < 0x88) | |
3818 | return 0; | |
3819 | } | |
3820 | else if (c < 0xfe) | |
3821 | { | |
3822 | char_len = 6; | |
3823 | if (c == 0xfc && d < 0x84) | |
3824 | return 0; | |
3825 | } | |
3826 | else | |
3827 | return 0; | |
3828 | ||
3829 | if (str_idx + char_len > input->len) | |
3830 | return 0; | |
3831 | ||
3832 | for (i = 1; i < char_len; ++i) | |
3833 | { | |
3834 | d = re_string_byte_at (input, str_idx + i); | |
3835 | if (d < 0x80 || d > 0xbf) | |
3836 | return 0; | |
3837 | } | |
3838 | return char_len; | |
3839 | } | |
3840 | ||
3841 | char_len = re_string_char_size_at (input, str_idx); | |
3b0bdc72 UD |
3842 | if (node->type == OP_PERIOD) |
3843 | { | |
ad7f28c2 UD |
3844 | if (char_len <= 1) |
3845 | return 0; | |
3846 | /* FIXME: I don't think this if is needed, as both '\n' | |
3847 | and '\0' are char_len == 1. */ | |
434d3784 | 3848 | /* '.' accepts any one character except the following two cases. */ |
56b168be | 3849 | if ((!(dfa->syntax & RE_DOT_NEWLINE) && |
15a7d175 | 3850 | re_string_byte_at (input, str_idx) == '\n') || |
56b168be | 3851 | ((dfa->syntax & RE_DOT_NOT_NULL) && |
15a7d175 UD |
3852 | re_string_byte_at (input, str_idx) == '\0')) |
3853 | return 0; | |
3b0bdc72 UD |
3854 | return char_len; |
3855 | } | |
ad7f28c2 UD |
3856 | |
3857 | elem_len = re_string_elem_size_at (input, str_idx); | |
6c2a04a7 | 3858 | if ((elem_len <= 1 && char_len <= 1) || char_len == 0) |
ad7f28c2 UD |
3859 | return 0; |
3860 | ||
3861 | if (node->type == COMPLEX_BRACKET) | |
3b0bdc72 UD |
3862 | { |
3863 | const re_charset_t *cset = node->opr.mbcset; | |
434d3784 | 3864 | # ifdef _LIBC |
1c99f950 UD |
3865 | const unsigned char *pin |
3866 | = ((const unsigned char *) re_string_get_buffer (input) + str_idx); | |
5f93cd52 UD |
3867 | int j; |
3868 | uint32_t nrules; | |
434d3784 UD |
3869 | # endif /* _LIBC */ |
3870 | int match_len = 0; | |
3871 | wchar_t wc = ((cset->nranges || cset->nchar_classes || cset->nmbchars) | |
15a7d175 | 3872 | ? re_string_wchar_at (input, str_idx) : 0); |
434d3784 UD |
3873 | |
3874 | /* match with multibyte character? */ | |
3875 | for (i = 0; i < cset->nmbchars; ++i) | |
15a7d175 UD |
3876 | if (wc == cset->mbchars[i]) |
3877 | { | |
3878 | match_len = char_len; | |
3879 | goto check_node_accept_bytes_match; | |
3880 | } | |
434d3784 UD |
3881 | /* match with character_class? */ |
3882 | for (i = 0; i < cset->nchar_classes; ++i) | |
15a7d175 UD |
3883 | { |
3884 | wctype_t wt = cset->char_classes[i]; | |
3885 | if (__iswctype (wc, wt)) | |
3886 | { | |
3887 | match_len = char_len; | |
3888 | goto check_node_accept_bytes_match; | |
3889 | } | |
3890 | } | |
434d3784 UD |
3891 | |
3892 | # ifdef _LIBC | |
5f93cd52 | 3893 | nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); |
3b0bdc72 | 3894 | if (nrules != 0) |
15a7d175 UD |
3895 | { |
3896 | unsigned int in_collseq = 0; | |
3897 | const int32_t *table, *indirect; | |
3898 | const unsigned char *weights, *extra; | |
3899 | const char *collseqwc; | |
3900 | int32_t idx; | |
3901 | /* This #include defines a local function! */ | |
434d3784 | 3902 | # include <locale/weight.h> |
3b0bdc72 | 3903 | |
15a7d175 UD |
3904 | /* match with collating_symbol? */ |
3905 | if (cset->ncoll_syms) | |
3906 | extra = (const unsigned char *) | |
3907 | _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB); | |
3908 | for (i = 0; i < cset->ncoll_syms; ++i) | |
3909 | { | |
3910 | const unsigned char *coll_sym = extra + cset->coll_syms[i]; | |
3911 | /* Compare the length of input collating element and | |
3912 | the length of current collating element. */ | |
3913 | if (*coll_sym != elem_len) | |
3914 | continue; | |
3915 | /* Compare each bytes. */ | |
3916 | for (j = 0; j < *coll_sym; j++) | |
3917 | if (pin[j] != coll_sym[1 + j]) | |
3918 | break; | |
3919 | if (j == *coll_sym) | |
3920 | { | |
3921 | /* Match if every bytes is equal. */ | |
3922 | match_len = j; | |
3923 | goto check_node_accept_bytes_match; | |
3924 | } | |
3925 | } | |
3926 | ||
3927 | if (cset->nranges) | |
3928 | { | |
3929 | if (elem_len <= char_len) | |
3930 | { | |
3931 | collseqwc = _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQWC); | |
25337753 | 3932 | in_collseq = __collseq_table_lookup (collseqwc, wc); |
15a7d175 UD |
3933 | } |
3934 | else | |
3935 | in_collseq = find_collation_sequence_value (pin, elem_len); | |
3936 | } | |
3937 | /* match with range expression? */ | |
3938 | for (i = 0; i < cset->nranges; ++i) | |
3939 | if (cset->range_starts[i] <= in_collseq | |
3940 | && in_collseq <= cset->range_ends[i]) | |
3941 | { | |
3942 | match_len = elem_len; | |
3943 | goto check_node_accept_bytes_match; | |
3944 | } | |
3945 | ||
3946 | /* match with equivalence_class? */ | |
3947 | if (cset->nequiv_classes) | |
3948 | { | |
3949 | const unsigned char *cp = pin; | |
3950 | table = (const int32_t *) | |
3951 | _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB); | |
3952 | weights = (const unsigned char *) | |
3953 | _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTMB); | |
3954 | extra = (const unsigned char *) | |
3955 | _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB); | |
3956 | indirect = (const int32_t *) | |
3957 | _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTMB); | |
3958 | idx = findidx (&cp); | |
3959 | if (idx > 0) | |
3960 | for (i = 0; i < cset->nequiv_classes; ++i) | |
3961 | { | |
3962 | int32_t equiv_class_idx = cset->equiv_classes[i]; | |
3963 | size_t weight_len = weights[idx]; | |
3964 | if (weight_len == weights[equiv_class_idx]) | |
3965 | { | |
3966 | int cnt = 0; | |
3967 | while (cnt <= weight_len | |
3968 | && (weights[equiv_class_idx + 1 + cnt] | |
3969 | == weights[idx + 1 + cnt])) | |
3970 | ++cnt; | |
3971 | if (cnt > weight_len) | |
3972 | { | |
3973 | match_len = elem_len; | |
3974 | goto check_node_accept_bytes_match; | |
3975 | } | |
3976 | } | |
3977 | } | |
3978 | } | |
3979 | } | |
434d3784 UD |
3980 | else |
3981 | # endif /* _LIBC */ | |
15a7d175 UD |
3982 | { |
3983 | /* match with range expression? */ | |
92b27c74 | 3984 | #if __GNUC__ >= 2 |
15a7d175 | 3985 | wchar_t cmp_buf[] = {L'\0', L'\0', wc, L'\0', L'\0', L'\0'}; |
92b27c74 | 3986 | #else |
15a7d175 UD |
3987 | wchar_t cmp_buf[] = {L'\0', L'\0', L'\0', L'\0', L'\0', L'\0'}; |
3988 | cmp_buf[2] = wc; | |
92b27c74 | 3989 | #endif |
15a7d175 UD |
3990 | for (i = 0; i < cset->nranges; ++i) |
3991 | { | |
3992 | cmp_buf[0] = cset->range_starts[i]; | |
3993 | cmp_buf[4] = cset->range_ends[i]; | |
3994 | if (wcscoll (cmp_buf, cmp_buf + 2) <= 0 | |
3995 | && wcscoll (cmp_buf + 2, cmp_buf + 4) <= 0) | |
3996 | { | |
3997 | match_len = char_len; | |
3998 | goto check_node_accept_bytes_match; | |
3999 | } | |
4000 | } | |
4001 | } | |
434d3784 UD |
4002 | check_node_accept_bytes_match: |
4003 | if (!cset->non_match) | |
15a7d175 | 4004 | return match_len; |
434d3784 | 4005 | else |
15a7d175 UD |
4006 | { |
4007 | if (match_len > 0) | |
4008 | return 0; | |
4009 | else | |
4010 | return (elem_len > char_len) ? elem_len : char_len; | |
4011 | } | |
3b0bdc72 UD |
4012 | } |
4013 | return 0; | |
4014 | } | |
4015 | ||
434d3784 | 4016 | # ifdef _LIBC |
3b0bdc72 | 4017 | static unsigned int |
a334319f UD |
4018 | find_collation_sequence_value (mbs, mbs_len) |
4019 | const unsigned char *mbs; | |
4020 | size_t mbs_len; | |
3b0bdc72 UD |
4021 | { |
4022 | uint32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); | |
4023 | if (nrules == 0) | |
4024 | { | |
4025 | if (mbs_len == 1) | |
15a7d175 UD |
4026 | { |
4027 | /* No valid character. Match it as a single byte character. */ | |
4028 | const unsigned char *collseq = (const unsigned char *) | |
4029 | _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQMB); | |
4030 | return collseq[mbs[0]]; | |
4031 | } | |
3b0bdc72 UD |
4032 | return UINT_MAX; |
4033 | } | |
4034 | else | |
4035 | { | |
4036 | int32_t idx; | |
4037 | const unsigned char *extra = (const unsigned char *) | |
15a7d175 | 4038 | _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB); |
6c2a04a7 UD |
4039 | int32_t extrasize = (const unsigned char *) |
4040 | _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB + 1) - extra; | |
3b0bdc72 | 4041 | |
6c2a04a7 | 4042 | for (idx = 0; idx < extrasize;) |
15a7d175 UD |
4043 | { |
4044 | int mbs_cnt, found = 0; | |
4045 | int32_t elem_mbs_len; | |
4046 | /* Skip the name of collating element name. */ | |
4047 | idx = idx + extra[idx] + 1; | |
4048 | elem_mbs_len = extra[idx++]; | |
4049 | if (mbs_len == elem_mbs_len) | |
4050 | { | |
4051 | for (mbs_cnt = 0; mbs_cnt < elem_mbs_len; ++mbs_cnt) | |
4052 | if (extra[idx + mbs_cnt] != mbs[mbs_cnt]) | |
4053 | break; | |
4054 | if (mbs_cnt == elem_mbs_len) | |
4055 | /* Found the entry. */ | |
4056 | found = 1; | |
4057 | } | |
4058 | /* Skip the byte sequence of the collating element. */ | |
4059 | idx += elem_mbs_len; | |
4060 | /* Adjust for the alignment. */ | |
4061 | idx = (idx + 3) & ~3; | |
4062 | /* Skip the collation sequence value. */ | |
4063 | idx += sizeof (uint32_t); | |
4064 | /* Skip the wide char sequence of the collating element. */ | |
4065 | idx = idx + sizeof (uint32_t) * (extra[idx] + 1); | |
4066 | /* If we found the entry, return the sequence value. */ | |
4067 | if (found) | |
4068 | return *(uint32_t *) (extra + idx); | |
4069 | /* Skip the collation sequence value. */ | |
4070 | idx += sizeof (uint32_t); | |
4071 | } | |
6c2a04a7 | 4072 | return UINT_MAX; |
3b0bdc72 UD |
4073 | } |
4074 | } | |
434d3784 UD |
4075 | # endif /* _LIBC */ |
4076 | #endif /* RE_ENABLE_I18N */ | |
3b0bdc72 UD |
4077 | |
4078 | /* Check whether the node accepts the byte which is IDX-th | |
4079 | byte of the INPUT. */ | |
4080 | ||
4081 | static int | |
a334319f UD |
4082 | check_node_accept (mctx, node, idx) |
4083 | const re_match_context_t *mctx; | |
4084 | const re_token_t *node; | |
4085 | int idx; | |
3b0bdc72 | 4086 | { |
3b0bdc72 | 4087 | unsigned char ch; |
56b168be | 4088 | ch = re_string_byte_at (&mctx->input, idx); |
ad7f28c2 UD |
4089 | switch (node->type) |
4090 | { | |
4091 | case CHARACTER: | |
5cf1ec52 UD |
4092 | if (node->opr.c != ch) |
4093 | return 0; | |
4094 | break; | |
4095 | ||
ad7f28c2 | 4096 | case SIMPLE_BRACKET: |
5cf1ec52 UD |
4097 | if (!bitset_contain (node->opr.sbcset, ch)) |
4098 | return 0; | |
4099 | break; | |
4100 | ||
c0d5034e | 4101 | #ifdef RE_ENABLE_I18N |
ad7f28c2 UD |
4102 | case OP_UTF8_PERIOD: |
4103 | if (ch >= 0x80) | |
4104 | return 0; | |
4105 | /* FALLTHROUGH */ | |
c0d5034e | 4106 | #endif |
ad7f28c2 | 4107 | case OP_PERIOD: |
5cf1ec52 UD |
4108 | if ((ch == '\n' && !(mctx->dfa->syntax & RE_DOT_NEWLINE)) |
4109 | || (ch == '\0' && (mctx->dfa->syntax & RE_DOT_NOT_NULL))) | |
4110 | return 0; | |
4111 | break; | |
4112 | ||
ad7f28c2 UD |
4113 | default: |
4114 | return 0; | |
4115 | } | |
5cf1ec52 UD |
4116 | |
4117 | if (node->constraint) | |
4118 | { | |
4119 | /* The node has constraints. Check whether the current context | |
4120 | satisfies the constraints. */ | |
4121 | unsigned int context = re_string_context_at (&mctx->input, idx, | |
4122 | mctx->eflags); | |
4123 | if (NOT_SATISFY_NEXT_CONSTRAINT (node->constraint, context)) | |
4124 | return 0; | |
4125 | } | |
4126 | ||
4127 | return 1; | |
3b0bdc72 | 4128 | } |
612546c6 UD |
4129 | |
4130 | /* Extend the buffers, if the buffers have run out. */ | |
4131 | ||
4132 | static reg_errcode_t | |
a334319f UD |
4133 | extend_buffers (mctx) |
4134 | re_match_context_t *mctx; | |
612546c6 UD |
4135 | { |
4136 | reg_errcode_t ret; | |
56b168be | 4137 | re_string_t *pstr = &mctx->input; |
612546c6 UD |
4138 | |
4139 | /* Double the lengthes of the buffers. */ | |
4140 | ret = re_string_realloc_buffers (pstr, pstr->bufs_len * 2); | |
4141 | if (BE (ret != REG_NOERROR, 0)) | |
4142 | return ret; | |
4143 | ||
4144 | if (mctx->state_log != NULL) | |
4145 | { | |
4146 | /* And double the length of state_log. */ | |
951d6408 UD |
4147 | /* XXX We have no indication of the size of this buffer. If this |
4148 | allocation fail we have no indication that the state_log array | |
4149 | does not have the right size. */ | |
4150 | re_dfastate_t **new_array = re_realloc (mctx->state_log, re_dfastate_t *, | |
4151 | pstr->bufs_len + 1); | |
1b2c2628 | 4152 | if (BE (new_array == NULL, 0)) |
15a7d175 | 4153 | return REG_ESPACE; |
1b2c2628 | 4154 | mctx->state_log = new_array; |
612546c6 UD |
4155 | } |
4156 | ||
4157 | /* Then reconstruct the buffers. */ | |
4158 | if (pstr->icase) | |
4159 | { | |
4160 | #ifdef RE_ENABLE_I18N | |
3c0fb574 | 4161 | if (pstr->mb_cur_max > 1) |
bb3f4825 UD |
4162 | { |
4163 | ret = build_wcs_upper_buffer (pstr); | |
4164 | if (BE (ret != REG_NOERROR, 0)) | |
4165 | return ret; | |
4166 | } | |
612546c6 UD |
4167 | else |
4168 | #endif /* RE_ENABLE_I18N */ | |
15a7d175 | 4169 | build_upper_buffer (pstr); |
612546c6 UD |
4170 | } |
4171 | else | |
4172 | { | |
4173 | #ifdef RE_ENABLE_I18N | |
3c0fb574 | 4174 | if (pstr->mb_cur_max > 1) |
15a7d175 | 4175 | build_wcs_buffer (pstr); |
612546c6 UD |
4176 | else |
4177 | #endif /* RE_ENABLE_I18N */ | |
15a7d175 UD |
4178 | { |
4179 | if (pstr->trans != NULL) | |
4180 | re_string_translate_buffer (pstr); | |
15a7d175 | 4181 | } |
612546c6 UD |
4182 | } |
4183 | return REG_NOERROR; | |
4184 | } | |
4185 | ||
3b0bdc72 UD |
4186 | \f |
4187 | /* Functions for matching context. */ | |
4188 | ||
6291ee3c UD |
4189 | /* Initialize MCTX. */ |
4190 | ||
a9388965 | 4191 | static reg_errcode_t |
a334319f UD |
4192 | match_ctx_init (mctx, eflags, n) |
4193 | re_match_context_t *mctx; | |
4194 | int eflags, n; | |
3b0bdc72 UD |
4195 | { |
4196 | mctx->eflags = eflags; | |
612546c6 | 4197 | mctx->match_last = -1; |
3b0bdc72 | 4198 | if (n > 0) |
a9388965 UD |
4199 | { |
4200 | mctx->bkref_ents = re_malloc (struct re_backref_cache_entry, n); | |
6291ee3c UD |
4201 | mctx->sub_tops = re_malloc (re_sub_match_top_t *, n); |
4202 | if (BE (mctx->bkref_ents == NULL || mctx->sub_tops == NULL, 0)) | |
15a7d175 | 4203 | return REG_ESPACE; |
a9388965 | 4204 | } |
56b168be UD |
4205 | /* Already zero-ed by the caller. |
4206 | else | |
4207 | mctx->bkref_ents = NULL; | |
4208 | mctx->nbkref_ents = 0; | |
4209 | mctx->nsub_tops = 0; */ | |
3b0bdc72 | 4210 | mctx->abkref_ents = n; |
6291ee3c | 4211 | mctx->max_mb_elem_len = 1; |
6291ee3c | 4212 | mctx->asub_tops = n; |
a9388965 | 4213 | return REG_NOERROR; |
3b0bdc72 UD |
4214 | } |
4215 | ||
6291ee3c UD |
4216 | /* Clean the entries which depend on the current input in MCTX. |
4217 | This function must be invoked when the matcher changes the start index | |
4218 | of the input, or changes the input string. */ | |
4219 | ||
4220 | static void | |
a334319f UD |
4221 | match_ctx_clean (mctx) |
4222 | re_match_context_t *mctx; | |
6291ee3c UD |
4223 | { |
4224 | int st_idx; | |
4225 | for (st_idx = 0; st_idx < mctx->nsub_tops; ++st_idx) | |
4226 | { | |
4227 | int sl_idx; | |
4228 | re_sub_match_top_t *top = mctx->sub_tops[st_idx]; | |
4229 | for (sl_idx = 0; sl_idx < top->nlasts; ++sl_idx) | |
4230 | { | |
4231 | re_sub_match_last_t *last = top->lasts[sl_idx]; | |
4232 | re_free (last->path.array); | |
4233 | re_free (last); | |
4234 | } | |
4235 | re_free (top->lasts); | |
4236 | if (top->path) | |
4237 | { | |
4238 | re_free (top->path->array); | |
4239 | re_free (top->path); | |
4240 | } | |
4241 | free (top); | |
4242 | } | |
cb265fec UD |
4243 | |
4244 | mctx->nsub_tops = 0; | |
4245 | mctx->nbkref_ents = 0; | |
4246 | } | |
4247 | ||
4248 | /* Free all the memory associated with MCTX. */ | |
4249 | ||
4250 | static void | |
a334319f UD |
4251 | match_ctx_free (mctx) |
4252 | re_match_context_t *mctx; | |
cb265fec UD |
4253 | { |
4254 | /* First, free all the memory associated with MCTX->SUB_TOPS. */ | |
4255 | match_ctx_clean (mctx); | |
4256 | re_free (mctx->sub_tops); | |
4257 | re_free (mctx->bkref_ents); | |
6291ee3c UD |
4258 | } |
4259 | ||
4260 | /* Add a new backreference entry to MCTX. | |
4261 | Note that we assume that caller never call this function with duplicate | |
4262 | entry, and call with STR_IDX which isn't smaller than any existing entry. | |
4263 | */ | |
3b0bdc72 | 4264 | |
a9388965 | 4265 | static reg_errcode_t |
a334319f UD |
4266 | match_ctx_add_entry (mctx, node, str_idx, from, to) |
4267 | re_match_context_t *mctx; | |
4268 | int node, str_idx, from, to; | |
3b0bdc72 UD |
4269 | { |
4270 | if (mctx->nbkref_ents >= mctx->abkref_ents) | |
4271 | { | |
1b2c2628 UD |
4272 | struct re_backref_cache_entry* new_entry; |
4273 | new_entry = re_realloc (mctx->bkref_ents, struct re_backref_cache_entry, | |
15a7d175 | 4274 | mctx->abkref_ents * 2); |
1b2c2628 | 4275 | if (BE (new_entry == NULL, 0)) |
15a7d175 UD |
4276 | { |
4277 | re_free (mctx->bkref_ents); | |
4278 | return REG_ESPACE; | |
4279 | } | |
1b2c2628 | 4280 | mctx->bkref_ents = new_entry; |
3b0bdc72 | 4281 | memset (mctx->bkref_ents + mctx->nbkref_ents, '\0', |
15a7d175 | 4282 | sizeof (struct re_backref_cache_entry) * mctx->abkref_ents); |
3b0bdc72 UD |
4283 | mctx->abkref_ents *= 2; |
4284 | } | |
e40a38b3 UD |
4285 | if (mctx->nbkref_ents > 0 |
4286 | && mctx->bkref_ents[mctx->nbkref_ents - 1].str_idx == str_idx) | |
4287 | mctx->bkref_ents[mctx->nbkref_ents - 1].more = 1; | |
4288 | ||
3b0bdc72 | 4289 | mctx->bkref_ents[mctx->nbkref_ents].node = node; |
0742e48e UD |
4290 | mctx->bkref_ents[mctx->nbkref_ents].str_idx = str_idx; |
4291 | mctx->bkref_ents[mctx->nbkref_ents].subexp_from = from; | |
e40a38b3 | 4292 | mctx->bkref_ents[mctx->nbkref_ents].subexp_to = to; |
7db61208 UD |
4293 | |
4294 | /* This is a cache that saves negative results of check_dst_limits_calc_pos. | |
4295 | If bit N is clear, means that this entry won't epsilon-transition to | |
4296 | an OP_OPEN_SUBEXP or OP_CLOSE_SUBEXP for the N+1-th subexpression. If | |
4297 | it is set, check_dst_limits_calc_pos_1 will recurse and try to find one | |
4298 | such node. | |
4299 | ||
4300 | A backreference does not epsilon-transition unless it is empty, so set | |
4301 | to all zeros if FROM != TO. */ | |
4302 | mctx->bkref_ents[mctx->nbkref_ents].eps_reachable_subexps_map | |
4303 | = (from == to ? ~0 : 0); | |
4304 | ||
e40a38b3 | 4305 | mctx->bkref_ents[mctx->nbkref_ents++].more = 0; |
0742e48e UD |
4306 | if (mctx->max_mb_elem_len < to - from) |
4307 | mctx->max_mb_elem_len = to - from; | |
a9388965 | 4308 | return REG_NOERROR; |
3b0bdc72 | 4309 | } |
0742e48e | 4310 | |
e40a38b3 UD |
4311 | /* Search for the first entry which has the same str_idx, or -1 if none is |
4312 | found. Note that MCTX->BKREF_ENTS is already sorted by MCTX->STR_IDX. */ | |
6291ee3c UD |
4313 | |
4314 | static int | |
a334319f | 4315 | search_cur_bkref_entry (mctx, str_idx) |
01ed6ceb | 4316 | const re_match_context_t *mctx; |
a334319f | 4317 | int str_idx; |
6291ee3c | 4318 | { |
e40a38b3 UD |
4319 | int left, right, mid, last; |
4320 | last = right = mctx->nbkref_ents; | |
6291ee3c UD |
4321 | for (left = 0; left < right;) |
4322 | { | |
4323 | mid = (left + right) / 2; | |
4324 | if (mctx->bkref_ents[mid].str_idx < str_idx) | |
4325 | left = mid + 1; | |
4326 | else | |
4327 | right = mid; | |
4328 | } | |
e40a38b3 UD |
4329 | if (left < last && mctx->bkref_ents[left].str_idx == str_idx) |
4330 | return left; | |
4331 | else | |
4332 | return -1; | |
6291ee3c UD |
4333 | } |
4334 | ||
6291ee3c UD |
4335 | /* Register the node NODE, whose type is OP_OPEN_SUBEXP, and which matches |
4336 | at STR_IDX. */ | |
4337 | ||
4338 | static reg_errcode_t | |
a334319f UD |
4339 | match_ctx_add_subtop (mctx, node, str_idx) |
4340 | re_match_context_t *mctx; | |
4341 | int node, str_idx; | |
6291ee3c UD |
4342 | { |
4343 | #ifdef DEBUG | |
4344 | assert (mctx->sub_tops != NULL); | |
4345 | assert (mctx->asub_tops > 0); | |
4346 | #endif | |
951d6408 | 4347 | if (BE (mctx->nsub_tops == mctx->asub_tops, 0)) |
6291ee3c | 4348 | { |
951d6408 UD |
4349 | int new_asub_tops = mctx->asub_tops * 2; |
4350 | re_sub_match_top_t **new_array = re_realloc (mctx->sub_tops, | |
4351 | re_sub_match_top_t *, | |
4352 | new_asub_tops); | |
6291ee3c UD |
4353 | if (BE (new_array == NULL, 0)) |
4354 | return REG_ESPACE; | |
4355 | mctx->sub_tops = new_array; | |
951d6408 | 4356 | mctx->asub_tops = new_asub_tops; |
6291ee3c UD |
4357 | } |
4358 | mctx->sub_tops[mctx->nsub_tops] = calloc (1, sizeof (re_sub_match_top_t)); | |
951d6408 | 4359 | if (BE (mctx->sub_tops[mctx->nsub_tops] == NULL, 0)) |
6291ee3c UD |
4360 | return REG_ESPACE; |
4361 | mctx->sub_tops[mctx->nsub_tops]->node = node; | |
4362 | mctx->sub_tops[mctx->nsub_tops++]->str_idx = str_idx; | |
4363 | return REG_NOERROR; | |
4364 | } | |
4365 | ||
4366 | /* Register the node NODE, whose type is OP_CLOSE_SUBEXP, and which matches | |
4367 | at STR_IDX, whose corresponding OP_OPEN_SUBEXP is SUB_TOP. */ | |
4368 | ||
4369 | static re_sub_match_last_t * | |
a334319f UD |
4370 | match_ctx_add_sublast (subtop, node, str_idx) |
4371 | re_sub_match_top_t *subtop; | |
4372 | int node, str_idx; | |
6291ee3c UD |
4373 | { |
4374 | re_sub_match_last_t *new_entry; | |
951d6408 | 4375 | if (BE (subtop->nlasts == subtop->alasts, 0)) |
6291ee3c | 4376 | { |
951d6408 UD |
4377 | int new_alasts = 2 * subtop->alasts + 1; |
4378 | re_sub_match_last_t **new_array = re_realloc (subtop->lasts, | |
4379 | re_sub_match_last_t *, | |
4380 | new_alasts); | |
6291ee3c UD |
4381 | if (BE (new_array == NULL, 0)) |
4382 | return NULL; | |
4383 | subtop->lasts = new_array; | |
951d6408 | 4384 | subtop->alasts = new_alasts; |
6291ee3c UD |
4385 | } |
4386 | new_entry = calloc (1, sizeof (re_sub_match_last_t)); | |
951d6408 UD |
4387 | if (BE (new_entry != NULL, 1)) |
4388 | { | |
4389 | subtop->lasts[subtop->nlasts] = new_entry; | |
4390 | new_entry->node = node; | |
4391 | new_entry->str_idx = str_idx; | |
4392 | ++subtop->nlasts; | |
4393 | } | |
6291ee3c UD |
4394 | return new_entry; |
4395 | } | |
4396 | ||
0742e48e | 4397 | static void |
a334319f UD |
4398 | sift_ctx_init (sctx, sifted_sts, limited_sts, last_node, last_str_idx) |
4399 | re_sift_context_t *sctx; | |
4400 | re_dfastate_t **sifted_sts, **limited_sts; | |
4401 | int last_node, last_str_idx; | |
0742e48e UD |
4402 | { |
4403 | sctx->sifted_states = sifted_sts; | |
4404 | sctx->limited_states = limited_sts; | |
4405 | sctx->last_node = last_node; | |
4406 | sctx->last_str_idx = last_str_idx; | |
0742e48e UD |
4407 | re_node_set_init_empty (&sctx->limits); |
4408 | } |