]>
Commit | Line | Data |
---|---|---|
1 | #include "cache.h" | |
2 | #include "config.h" | |
3 | #include "grep.h" | |
4 | #include "object-store.h" | |
5 | #include "userdiff.h" | |
6 | #include "xdiff-interface.h" | |
7 | #include "diff.h" | |
8 | #include "diffcore.h" | |
9 | #include "commit.h" | |
10 | #include "quote.h" | |
11 | #include "help.h" | |
12 | ||
13 | static int grep_source_load(struct grep_source *gs); | |
14 | static int grep_source_is_binary(struct grep_source *gs); | |
15 | ||
16 | static struct grep_opt grep_defaults; | |
17 | ||
18 | static const char *color_grep_slots[] = { | |
19 | [GREP_COLOR_CONTEXT] = "context", | |
20 | [GREP_COLOR_FILENAME] = "filename", | |
21 | [GREP_COLOR_FUNCTION] = "function", | |
22 | [GREP_COLOR_LINENO] = "lineNumber", | |
23 | [GREP_COLOR_COLUMNNO] = "column", | |
24 | [GREP_COLOR_MATCH_CONTEXT] = "matchContext", | |
25 | [GREP_COLOR_MATCH_SELECTED] = "matchSelected", | |
26 | [GREP_COLOR_SELECTED] = "selected", | |
27 | [GREP_COLOR_SEP] = "separator", | |
28 | }; | |
29 | ||
30 | static void std_output(struct grep_opt *opt, const void *buf, size_t size) | |
31 | { | |
32 | fwrite(buf, size, 1, stdout); | |
33 | } | |
34 | ||
35 | static void color_set(char *dst, const char *color_bytes) | |
36 | { | |
37 | xsnprintf(dst, COLOR_MAXLEN, "%s", color_bytes); | |
38 | } | |
39 | ||
40 | /* | |
41 | * Initialize the grep_defaults template with hardcoded defaults. | |
42 | * We could let the compiler do this, but without C99 initializers | |
43 | * the code gets unwieldy and unreadable, so... | |
44 | */ | |
45 | void init_grep_defaults(struct repository *repo) | |
46 | { | |
47 | struct grep_opt *opt = &grep_defaults; | |
48 | static int run_once; | |
49 | ||
50 | if (run_once) | |
51 | return; | |
52 | run_once++; | |
53 | ||
54 | memset(opt, 0, sizeof(*opt)); | |
55 | opt->repo = repo; | |
56 | opt->relative = 1; | |
57 | opt->pathname = 1; | |
58 | opt->max_depth = -1; | |
59 | opt->pattern_type_option = GREP_PATTERN_TYPE_UNSPECIFIED; | |
60 | color_set(opt->colors[GREP_COLOR_CONTEXT], ""); | |
61 | color_set(opt->colors[GREP_COLOR_FILENAME], ""); | |
62 | color_set(opt->colors[GREP_COLOR_FUNCTION], ""); | |
63 | color_set(opt->colors[GREP_COLOR_LINENO], ""); | |
64 | color_set(opt->colors[GREP_COLOR_COLUMNNO], ""); | |
65 | color_set(opt->colors[GREP_COLOR_MATCH_CONTEXT], GIT_COLOR_BOLD_RED); | |
66 | color_set(opt->colors[GREP_COLOR_MATCH_SELECTED], GIT_COLOR_BOLD_RED); | |
67 | color_set(opt->colors[GREP_COLOR_SELECTED], ""); | |
68 | color_set(opt->colors[GREP_COLOR_SEP], GIT_COLOR_CYAN); | |
69 | opt->only_matching = 0; | |
70 | opt->color = -1; | |
71 | opt->output = std_output; | |
72 | } | |
73 | ||
74 | static int parse_pattern_type_arg(const char *opt, const char *arg) | |
75 | { | |
76 | if (!strcmp(arg, "default")) | |
77 | return GREP_PATTERN_TYPE_UNSPECIFIED; | |
78 | else if (!strcmp(arg, "basic")) | |
79 | return GREP_PATTERN_TYPE_BRE; | |
80 | else if (!strcmp(arg, "extended")) | |
81 | return GREP_PATTERN_TYPE_ERE; | |
82 | else if (!strcmp(arg, "fixed")) | |
83 | return GREP_PATTERN_TYPE_FIXED; | |
84 | else if (!strcmp(arg, "perl")) | |
85 | return GREP_PATTERN_TYPE_PCRE; | |
86 | die("bad %s argument: %s", opt, arg); | |
87 | } | |
88 | ||
89 | define_list_config_array_extra(color_grep_slots, {"match"}); | |
90 | ||
91 | /* | |
92 | * Read the configuration file once and store it in | |
93 | * the grep_defaults template. | |
94 | */ | |
95 | int grep_config(const char *var, const char *value, void *cb) | |
96 | { | |
97 | struct grep_opt *opt = &grep_defaults; | |
98 | const char *slot; | |
99 | ||
100 | if (userdiff_config(var, value) < 0) | |
101 | return -1; | |
102 | ||
103 | if (!strcmp(var, "grep.extendedregexp")) { | |
104 | opt->extended_regexp_option = git_config_bool(var, value); | |
105 | return 0; | |
106 | } | |
107 | ||
108 | if (!strcmp(var, "grep.patterntype")) { | |
109 | opt->pattern_type_option = parse_pattern_type_arg(var, value); | |
110 | return 0; | |
111 | } | |
112 | ||
113 | if (!strcmp(var, "grep.linenumber")) { | |
114 | opt->linenum = git_config_bool(var, value); | |
115 | return 0; | |
116 | } | |
117 | if (!strcmp(var, "grep.column")) { | |
118 | opt->columnnum = git_config_bool(var, value); | |
119 | return 0; | |
120 | } | |
121 | ||
122 | if (!strcmp(var, "grep.fullname")) { | |
123 | opt->relative = !git_config_bool(var, value); | |
124 | return 0; | |
125 | } | |
126 | ||
127 | if (!strcmp(var, "color.grep")) | |
128 | opt->color = git_config_colorbool(var, value); | |
129 | if (!strcmp(var, "color.grep.match")) { | |
130 | if (grep_config("color.grep.matchcontext", value, cb) < 0) | |
131 | return -1; | |
132 | if (grep_config("color.grep.matchselected", value, cb) < 0) | |
133 | return -1; | |
134 | } else if (skip_prefix(var, "color.grep.", &slot)) { | |
135 | int i = LOOKUP_CONFIG(color_grep_slots, slot); | |
136 | char *color; | |
137 | ||
138 | if (i < 0) | |
139 | return -1; | |
140 | color = opt->colors[i]; | |
141 | if (!value) | |
142 | return config_error_nonbool(var); | |
143 | return color_parse(value, color); | |
144 | } | |
145 | return 0; | |
146 | } | |
147 | ||
148 | /* | |
149 | * Initialize one instance of grep_opt and copy the | |
150 | * default values from the template we read the configuration | |
151 | * information in an earlier call to git_config(grep_config). | |
152 | */ | |
153 | void grep_init(struct grep_opt *opt, struct repository *repo, const char *prefix) | |
154 | { | |
155 | struct grep_opt *def = &grep_defaults; | |
156 | int i; | |
157 | ||
158 | memset(opt, 0, sizeof(*opt)); | |
159 | opt->repo = repo; | |
160 | opt->prefix = prefix; | |
161 | opt->prefix_length = (prefix && *prefix) ? strlen(prefix) : 0; | |
162 | opt->pattern_tail = &opt->pattern_list; | |
163 | opt->header_tail = &opt->header_list; | |
164 | ||
165 | opt->only_matching = def->only_matching; | |
166 | opt->color = def->color; | |
167 | opt->extended_regexp_option = def->extended_regexp_option; | |
168 | opt->pattern_type_option = def->pattern_type_option; | |
169 | opt->linenum = def->linenum; | |
170 | opt->columnnum = def->columnnum; | |
171 | opt->max_depth = def->max_depth; | |
172 | opt->pathname = def->pathname; | |
173 | opt->relative = def->relative; | |
174 | opt->output = def->output; | |
175 | ||
176 | for (i = 0; i < NR_GREP_COLORS; i++) | |
177 | color_set(opt->colors[i], def->colors[i]); | |
178 | } | |
179 | ||
180 | static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type, struct grep_opt *opt) | |
181 | { | |
182 | /* | |
183 | * When committing to the pattern type by setting the relevant | |
184 | * fields in grep_opt it's generally not necessary to zero out | |
185 | * the fields we're not choosing, since they won't have been | |
186 | * set by anything. The extended_regexp_option field is the | |
187 | * only exception to this. | |
188 | * | |
189 | * This is because in the process of parsing grep.patternType | |
190 | * & grep.extendedRegexp we set opt->pattern_type_option and | |
191 | * opt->extended_regexp_option, respectively. We then | |
192 | * internally use opt->extended_regexp_option to see if we're | |
193 | * compiling an ERE. It must be unset if that's not actually | |
194 | * the case. | |
195 | */ | |
196 | if (pattern_type != GREP_PATTERN_TYPE_ERE && | |
197 | opt->extended_regexp_option) | |
198 | opt->extended_regexp_option = 0; | |
199 | ||
200 | switch (pattern_type) { | |
201 | case GREP_PATTERN_TYPE_UNSPECIFIED: | |
202 | /* fall through */ | |
203 | ||
204 | case GREP_PATTERN_TYPE_BRE: | |
205 | break; | |
206 | ||
207 | case GREP_PATTERN_TYPE_ERE: | |
208 | opt->extended_regexp_option = 1; | |
209 | break; | |
210 | ||
211 | case GREP_PATTERN_TYPE_FIXED: | |
212 | opt->fixed = 1; | |
213 | break; | |
214 | ||
215 | case GREP_PATTERN_TYPE_PCRE: | |
216 | #ifdef USE_LIBPCRE2 | |
217 | opt->pcre2 = 1; | |
218 | #else | |
219 | /* | |
220 | * It's important that pcre1 always be assigned to | |
221 | * even when there's no USE_LIBPCRE* defined. We still | |
222 | * call the PCRE stub function, it just dies with | |
223 | * "cannot use Perl-compatible regexes[...]". | |
224 | */ | |
225 | opt->pcre1 = 1; | |
226 | #endif | |
227 | break; | |
228 | } | |
229 | } | |
230 | ||
231 | void grep_commit_pattern_type(enum grep_pattern_type pattern_type, struct grep_opt *opt) | |
232 | { | |
233 | if (pattern_type != GREP_PATTERN_TYPE_UNSPECIFIED) | |
234 | grep_set_pattern_type_option(pattern_type, opt); | |
235 | else if (opt->pattern_type_option != GREP_PATTERN_TYPE_UNSPECIFIED) | |
236 | grep_set_pattern_type_option(opt->pattern_type_option, opt); | |
237 | else if (opt->extended_regexp_option) | |
238 | /* | |
239 | * This branch *must* happen after setting from the | |
240 | * opt->pattern_type_option above, we don't want | |
241 | * grep.extendedRegexp to override grep.patternType! | |
242 | */ | |
243 | grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE, opt); | |
244 | } | |
245 | ||
246 | static struct grep_pat *create_grep_pat(const char *pat, size_t patlen, | |
247 | const char *origin, int no, | |
248 | enum grep_pat_token t, | |
249 | enum grep_header_field field) | |
250 | { | |
251 | struct grep_pat *p = xcalloc(1, sizeof(*p)); | |
252 | p->pattern = xmemdupz(pat, patlen); | |
253 | p->patternlen = patlen; | |
254 | p->origin = origin; | |
255 | p->no = no; | |
256 | p->token = t; | |
257 | p->field = field; | |
258 | return p; | |
259 | } | |
260 | ||
261 | static void do_append_grep_pat(struct grep_pat ***tail, struct grep_pat *p) | |
262 | { | |
263 | **tail = p; | |
264 | *tail = &p->next; | |
265 | p->next = NULL; | |
266 | ||
267 | switch (p->token) { | |
268 | case GREP_PATTERN: /* atom */ | |
269 | case GREP_PATTERN_HEAD: | |
270 | case GREP_PATTERN_BODY: | |
271 | for (;;) { | |
272 | struct grep_pat *new_pat; | |
273 | size_t len = 0; | |
274 | char *cp = p->pattern + p->patternlen, *nl = NULL; | |
275 | while (++len <= p->patternlen) { | |
276 | if (*(--cp) == '\n') { | |
277 | nl = cp; | |
278 | break; | |
279 | } | |
280 | } | |
281 | if (!nl) | |
282 | break; | |
283 | new_pat = create_grep_pat(nl + 1, len - 1, p->origin, | |
284 | p->no, p->token, p->field); | |
285 | new_pat->next = p->next; | |
286 | if (!p->next) | |
287 | *tail = &new_pat->next; | |
288 | p->next = new_pat; | |
289 | *nl = '\0'; | |
290 | p->patternlen -= len; | |
291 | } | |
292 | break; | |
293 | default: | |
294 | break; | |
295 | } | |
296 | } | |
297 | ||
298 | void append_header_grep_pattern(struct grep_opt *opt, | |
299 | enum grep_header_field field, const char *pat) | |
300 | { | |
301 | struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0, | |
302 | GREP_PATTERN_HEAD, field); | |
303 | if (field == GREP_HEADER_REFLOG) | |
304 | opt->use_reflog_filter = 1; | |
305 | do_append_grep_pat(&opt->header_tail, p); | |
306 | } | |
307 | ||
308 | void append_grep_pattern(struct grep_opt *opt, const char *pat, | |
309 | const char *origin, int no, enum grep_pat_token t) | |
310 | { | |
311 | append_grep_pat(opt, pat, strlen(pat), origin, no, t); | |
312 | } | |
313 | ||
314 | void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen, | |
315 | const char *origin, int no, enum grep_pat_token t) | |
316 | { | |
317 | struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0); | |
318 | do_append_grep_pat(&opt->pattern_tail, p); | |
319 | } | |
320 | ||
321 | struct grep_opt *grep_opt_dup(const struct grep_opt *opt) | |
322 | { | |
323 | struct grep_pat *pat; | |
324 | struct grep_opt *ret = xmalloc(sizeof(struct grep_opt)); | |
325 | *ret = *opt; | |
326 | ||
327 | ret->pattern_list = NULL; | |
328 | ret->pattern_tail = &ret->pattern_list; | |
329 | ||
330 | for(pat = opt->pattern_list; pat != NULL; pat = pat->next) | |
331 | { | |
332 | if(pat->token == GREP_PATTERN_HEAD) | |
333 | append_header_grep_pattern(ret, pat->field, | |
334 | pat->pattern); | |
335 | else | |
336 | append_grep_pat(ret, pat->pattern, pat->patternlen, | |
337 | pat->origin, pat->no, pat->token); | |
338 | } | |
339 | ||
340 | return ret; | |
341 | } | |
342 | ||
343 | static NORETURN void compile_regexp_failed(const struct grep_pat *p, | |
344 | const char *error) | |
345 | { | |
346 | char where[1024]; | |
347 | ||
348 | if (p->no) | |
349 | xsnprintf(where, sizeof(where), "In '%s' at %d, ", p->origin, p->no); | |
350 | else if (p->origin) | |
351 | xsnprintf(where, sizeof(where), "%s, ", p->origin); | |
352 | else | |
353 | where[0] = 0; | |
354 | ||
355 | die("%s'%s': %s", where, p->pattern, error); | |
356 | } | |
357 | ||
358 | static int is_fixed(const char *s, size_t len) | |
359 | { | |
360 | size_t i; | |
361 | ||
362 | for (i = 0; i < len; i++) { | |
363 | if (is_regex_special(s[i])) | |
364 | return 0; | |
365 | } | |
366 | ||
367 | return 1; | |
368 | } | |
369 | ||
370 | static int has_null(const char *s, size_t len) | |
371 | { | |
372 | /* | |
373 | * regcomp cannot accept patterns with NULs so when using it | |
374 | * we consider any pattern containing a NUL fixed. | |
375 | */ | |
376 | if (memchr(s, 0, len)) | |
377 | return 1; | |
378 | ||
379 | return 0; | |
380 | } | |
381 | ||
382 | #ifdef USE_LIBPCRE1 | |
383 | static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt) | |
384 | { | |
385 | const char *error; | |
386 | int erroffset; | |
387 | int options = PCRE_MULTILINE; | |
388 | ||
389 | if (opt->ignore_case) { | |
390 | if (has_non_ascii(p->pattern)) | |
391 | p->pcre1_tables = pcre_maketables(); | |
392 | options |= PCRE_CASELESS; | |
393 | } | |
394 | if (is_utf8_locale() && has_non_ascii(p->pattern)) | |
395 | options |= PCRE_UTF8; | |
396 | ||
397 | p->pcre1_regexp = pcre_compile(p->pattern, options, &error, &erroffset, | |
398 | p->pcre1_tables); | |
399 | if (!p->pcre1_regexp) | |
400 | compile_regexp_failed(p, error); | |
401 | ||
402 | p->pcre1_extra_info = pcre_study(p->pcre1_regexp, GIT_PCRE_STUDY_JIT_COMPILE, &error); | |
403 | if (!p->pcre1_extra_info && error) | |
404 | die("%s", error); | |
405 | ||
406 | #ifdef GIT_PCRE1_USE_JIT | |
407 | pcre_config(PCRE_CONFIG_JIT, &p->pcre1_jit_on); | |
408 | if (p->pcre1_jit_on == 1) { | |
409 | p->pcre1_jit_stack = pcre_jit_stack_alloc(1, 1024 * 1024); | |
410 | if (!p->pcre1_jit_stack) | |
411 | die("Couldn't allocate PCRE JIT stack"); | |
412 | pcre_assign_jit_stack(p->pcre1_extra_info, NULL, p->pcre1_jit_stack); | |
413 | } else if (p->pcre1_jit_on != 0) { | |
414 | BUG("The pcre1_jit_on variable should be 0 or 1, not %d", | |
415 | p->pcre1_jit_on); | |
416 | } | |
417 | #endif | |
418 | } | |
419 | ||
420 | static int pcre1match(struct grep_pat *p, const char *line, const char *eol, | |
421 | regmatch_t *match, int eflags) | |
422 | { | |
423 | int ovector[30], ret, flags = 0; | |
424 | ||
425 | if (eflags & REG_NOTBOL) | |
426 | flags |= PCRE_NOTBOL; | |
427 | ||
428 | #ifdef GIT_PCRE1_USE_JIT | |
429 | if (p->pcre1_jit_on) { | |
430 | ret = pcre_jit_exec(p->pcre1_regexp, p->pcre1_extra_info, line, | |
431 | eol - line, 0, flags, ovector, | |
432 | ARRAY_SIZE(ovector), p->pcre1_jit_stack); | |
433 | } else | |
434 | #endif | |
435 | { | |
436 | ret = pcre_exec(p->pcre1_regexp, p->pcre1_extra_info, line, | |
437 | eol - line, 0, flags, ovector, | |
438 | ARRAY_SIZE(ovector)); | |
439 | } | |
440 | ||
441 | if (ret < 0 && ret != PCRE_ERROR_NOMATCH) | |
442 | die("pcre_exec failed with error code %d", ret); | |
443 | if (ret > 0) { | |
444 | ret = 0; | |
445 | match->rm_so = ovector[0]; | |
446 | match->rm_eo = ovector[1]; | |
447 | } | |
448 | ||
449 | return ret; | |
450 | } | |
451 | ||
452 | static void free_pcre1_regexp(struct grep_pat *p) | |
453 | { | |
454 | pcre_free(p->pcre1_regexp); | |
455 | #ifdef GIT_PCRE1_USE_JIT | |
456 | if (p->pcre1_jit_on) { | |
457 | pcre_free_study(p->pcre1_extra_info); | |
458 | pcre_jit_stack_free(p->pcre1_jit_stack); | |
459 | } else | |
460 | #endif | |
461 | { | |
462 | pcre_free(p->pcre1_extra_info); | |
463 | } | |
464 | pcre_free((void *)p->pcre1_tables); | |
465 | } | |
466 | #else /* !USE_LIBPCRE1 */ | |
467 | static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt) | |
468 | { | |
469 | die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE"); | |
470 | } | |
471 | ||
472 | static int pcre1match(struct grep_pat *p, const char *line, const char *eol, | |
473 | regmatch_t *match, int eflags) | |
474 | { | |
475 | return 1; | |
476 | } | |
477 | ||
478 | static void free_pcre1_regexp(struct grep_pat *p) | |
479 | { | |
480 | } | |
481 | #endif /* !USE_LIBPCRE1 */ | |
482 | ||
483 | #ifdef USE_LIBPCRE2 | |
484 | static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt) | |
485 | { | |
486 | int error; | |
487 | PCRE2_UCHAR errbuf[256]; | |
488 | PCRE2_SIZE erroffset; | |
489 | int options = PCRE2_MULTILINE; | |
490 | const uint8_t *character_tables = NULL; | |
491 | int jitret; | |
492 | int patinforet; | |
493 | size_t jitsizearg; | |
494 | ||
495 | assert(opt->pcre2); | |
496 | ||
497 | p->pcre2_compile_context = NULL; | |
498 | ||
499 | if (opt->ignore_case) { | |
500 | if (has_non_ascii(p->pattern)) { | |
501 | character_tables = pcre2_maketables(NULL); | |
502 | p->pcre2_compile_context = pcre2_compile_context_create(NULL); | |
503 | pcre2_set_character_tables(p->pcre2_compile_context, character_tables); | |
504 | } | |
505 | options |= PCRE2_CASELESS; | |
506 | } | |
507 | if (is_utf8_locale() && has_non_ascii(p->pattern)) | |
508 | options |= PCRE2_UTF; | |
509 | ||
510 | p->pcre2_pattern = pcre2_compile((PCRE2_SPTR)p->pattern, | |
511 | p->patternlen, options, &error, &erroffset, | |
512 | p->pcre2_compile_context); | |
513 | ||
514 | if (p->pcre2_pattern) { | |
515 | p->pcre2_match_data = pcre2_match_data_create_from_pattern(p->pcre2_pattern, NULL); | |
516 | if (!p->pcre2_match_data) | |
517 | die("Couldn't allocate PCRE2 match data"); | |
518 | } else { | |
519 | pcre2_get_error_message(error, errbuf, sizeof(errbuf)); | |
520 | compile_regexp_failed(p, (const char *)&errbuf); | |
521 | } | |
522 | ||
523 | pcre2_config(PCRE2_CONFIG_JIT, &p->pcre2_jit_on); | |
524 | if (p->pcre2_jit_on == 1) { | |
525 | jitret = pcre2_jit_compile(p->pcre2_pattern, PCRE2_JIT_COMPLETE); | |
526 | if (jitret) | |
527 | die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p->pattern, jitret); | |
528 | ||
529 | /* | |
530 | * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just | |
531 | * tells us whether the library itself supports JIT, | |
532 | * but to see whether we're going to be actually using | |
533 | * JIT we need to extract PCRE2_INFO_JITSIZE from the | |
534 | * pattern *after* we do pcre2_jit_compile() above. | |
535 | * | |
536 | * This is because if the pattern contains the | |
537 | * (*NO_JIT) verb (see pcre2syntax(3)) | |
538 | * pcre2_jit_compile() will exit early with 0. If we | |
539 | * then proceed to call pcre2_jit_match() further down | |
540 | * the line instead of pcre2_match() we'll either | |
541 | * segfault (pre PCRE 10.31) or run into a fatal error | |
542 | * (post PCRE2 10.31) | |
543 | */ | |
544 | patinforet = pcre2_pattern_info(p->pcre2_pattern, PCRE2_INFO_JITSIZE, &jitsizearg); | |
545 | if (patinforet) | |
546 | BUG("pcre2_pattern_info() failed: %d", patinforet); | |
547 | if (jitsizearg == 0) { | |
548 | p->pcre2_jit_on = 0; | |
549 | return; | |
550 | } | |
551 | ||
552 | p->pcre2_jit_stack = pcre2_jit_stack_create(1, 1024 * 1024, NULL); | |
553 | if (!p->pcre2_jit_stack) | |
554 | die("Couldn't allocate PCRE2 JIT stack"); | |
555 | p->pcre2_match_context = pcre2_match_context_create(NULL); | |
556 | if (!p->pcre2_match_context) | |
557 | die("Couldn't allocate PCRE2 match context"); | |
558 | pcre2_jit_stack_assign(p->pcre2_match_context, NULL, p->pcre2_jit_stack); | |
559 | } else if (p->pcre2_jit_on != 0) { | |
560 | BUG("The pcre2_jit_on variable should be 0 or 1, not %d", | |
561 | p->pcre1_jit_on); | |
562 | } | |
563 | } | |
564 | ||
565 | static int pcre2match(struct grep_pat *p, const char *line, const char *eol, | |
566 | regmatch_t *match, int eflags) | |
567 | { | |
568 | int ret, flags = 0; | |
569 | PCRE2_SIZE *ovector; | |
570 | PCRE2_UCHAR errbuf[256]; | |
571 | ||
572 | if (eflags & REG_NOTBOL) | |
573 | flags |= PCRE2_NOTBOL; | |
574 | ||
575 | if (p->pcre2_jit_on) | |
576 | ret = pcre2_jit_match(p->pcre2_pattern, (unsigned char *)line, | |
577 | eol - line, 0, flags, p->pcre2_match_data, | |
578 | NULL); | |
579 | else | |
580 | ret = pcre2_match(p->pcre2_pattern, (unsigned char *)line, | |
581 | eol - line, 0, flags, p->pcre2_match_data, | |
582 | NULL); | |
583 | ||
584 | if (ret < 0 && ret != PCRE2_ERROR_NOMATCH) { | |
585 | pcre2_get_error_message(ret, errbuf, sizeof(errbuf)); | |
586 | die("%s failed with error code %d: %s", | |
587 | (p->pcre2_jit_on ? "pcre2_jit_match" : "pcre2_match"), ret, | |
588 | errbuf); | |
589 | } | |
590 | if (ret > 0) { | |
591 | ovector = pcre2_get_ovector_pointer(p->pcre2_match_data); | |
592 | ret = 0; | |
593 | match->rm_so = (int)ovector[0]; | |
594 | match->rm_eo = (int)ovector[1]; | |
595 | } | |
596 | ||
597 | return ret; | |
598 | } | |
599 | ||
600 | static void free_pcre2_pattern(struct grep_pat *p) | |
601 | { | |
602 | pcre2_compile_context_free(p->pcre2_compile_context); | |
603 | pcre2_code_free(p->pcre2_pattern); | |
604 | pcre2_match_data_free(p->pcre2_match_data); | |
605 | pcre2_jit_stack_free(p->pcre2_jit_stack); | |
606 | pcre2_match_context_free(p->pcre2_match_context); | |
607 | } | |
608 | #else /* !USE_LIBPCRE2 */ | |
609 | static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt) | |
610 | { | |
611 | /* | |
612 | * Unreachable until USE_LIBPCRE2 becomes synonymous with | |
613 | * USE_LIBPCRE. See the sibling comment in | |
614 | * grep_set_pattern_type_option(). | |
615 | */ | |
616 | die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE"); | |
617 | } | |
618 | ||
619 | static int pcre2match(struct grep_pat *p, const char *line, const char *eol, | |
620 | regmatch_t *match, int eflags) | |
621 | { | |
622 | return 1; | |
623 | } | |
624 | ||
625 | static void free_pcre2_pattern(struct grep_pat *p) | |
626 | { | |
627 | } | |
628 | #endif /* !USE_LIBPCRE2 */ | |
629 | ||
630 | static void compile_fixed_regexp(struct grep_pat *p, struct grep_opt *opt) | |
631 | { | |
632 | struct strbuf sb = STRBUF_INIT; | |
633 | int err; | |
634 | int regflags = 0; | |
635 | ||
636 | basic_regex_quote_buf(&sb, p->pattern); | |
637 | if (opt->ignore_case) | |
638 | regflags |= REG_ICASE; | |
639 | err = regcomp(&p->regexp, sb.buf, regflags); | |
640 | if (opt->debug) | |
641 | fprintf(stderr, "fixed %s\n", sb.buf); | |
642 | strbuf_release(&sb); | |
643 | if (err) { | |
644 | char errbuf[1024]; | |
645 | regerror(err, &p->regexp, errbuf, sizeof(errbuf)); | |
646 | compile_regexp_failed(p, errbuf); | |
647 | } | |
648 | } | |
649 | ||
650 | static void compile_regexp(struct grep_pat *p, struct grep_opt *opt) | |
651 | { | |
652 | int ascii_only; | |
653 | int err; | |
654 | int regflags = REG_NEWLINE; | |
655 | ||
656 | p->word_regexp = opt->word_regexp; | |
657 | p->ignore_case = opt->ignore_case; | |
658 | ascii_only = !has_non_ascii(p->pattern); | |
659 | ||
660 | /* | |
661 | * Even when -F (fixed) asks us to do a non-regexp search, we | |
662 | * may not be able to correctly case-fold when -i | |
663 | * (ignore-case) is asked (in which case, we'll synthesize a | |
664 | * regexp to match the pattern that matches regexp special | |
665 | * characters literally, while ignoring case differences). On | |
666 | * the other hand, even without -F, if the pattern does not | |
667 | * have any regexp special characters and there is no need for | |
668 | * case-folding search, we can internally turn it into a | |
669 | * simple string match using kws. p->fixed tells us if we | |
670 | * want to use kws. | |
671 | */ | |
672 | if (opt->fixed || | |
673 | has_null(p->pattern, p->patternlen) || | |
674 | is_fixed(p->pattern, p->patternlen)) | |
675 | p->fixed = !p->ignore_case || ascii_only; | |
676 | ||
677 | if (p->fixed) { | |
678 | p->kws = kwsalloc(p->ignore_case ? tolower_trans_tbl : NULL); | |
679 | kwsincr(p->kws, p->pattern, p->patternlen); | |
680 | kwsprep(p->kws); | |
681 | return; | |
682 | } else if (opt->fixed) { | |
683 | /* | |
684 | * We come here when the pattern has the non-ascii | |
685 | * characters we cannot case-fold, and asked to | |
686 | * ignore-case. | |
687 | */ | |
688 | compile_fixed_regexp(p, opt); | |
689 | return; | |
690 | } | |
691 | ||
692 | if (opt->pcre2) { | |
693 | compile_pcre2_pattern(p, opt); | |
694 | return; | |
695 | } | |
696 | ||
697 | if (opt->pcre1) { | |
698 | compile_pcre1_regexp(p, opt); | |
699 | return; | |
700 | } | |
701 | ||
702 | if (p->ignore_case) | |
703 | regflags |= REG_ICASE; | |
704 | if (opt->extended_regexp_option) | |
705 | regflags |= REG_EXTENDED; | |
706 | err = regcomp(&p->regexp, p->pattern, regflags); | |
707 | if (err) { | |
708 | char errbuf[1024]; | |
709 | regerror(err, &p->regexp, errbuf, 1024); | |
710 | compile_regexp_failed(p, errbuf); | |
711 | } | |
712 | } | |
713 | ||
714 | static struct grep_expr *compile_pattern_or(struct grep_pat **); | |
715 | static struct grep_expr *compile_pattern_atom(struct grep_pat **list) | |
716 | { | |
717 | struct grep_pat *p; | |
718 | struct grep_expr *x; | |
719 | ||
720 | p = *list; | |
721 | if (!p) | |
722 | return NULL; | |
723 | switch (p->token) { | |
724 | case GREP_PATTERN: /* atom */ | |
725 | case GREP_PATTERN_HEAD: | |
726 | case GREP_PATTERN_BODY: | |
727 | x = xcalloc(1, sizeof (struct grep_expr)); | |
728 | x->node = GREP_NODE_ATOM; | |
729 | x->u.atom = p; | |
730 | *list = p->next; | |
731 | return x; | |
732 | case GREP_OPEN_PAREN: | |
733 | *list = p->next; | |
734 | x = compile_pattern_or(list); | |
735 | if (!*list || (*list)->token != GREP_CLOSE_PAREN) | |
736 | die("unmatched parenthesis"); | |
737 | *list = (*list)->next; | |
738 | return x; | |
739 | default: | |
740 | return NULL; | |
741 | } | |
742 | } | |
743 | ||
744 | static struct grep_expr *compile_pattern_not(struct grep_pat **list) | |
745 | { | |
746 | struct grep_pat *p; | |
747 | struct grep_expr *x; | |
748 | ||
749 | p = *list; | |
750 | if (!p) | |
751 | return NULL; | |
752 | switch (p->token) { | |
753 | case GREP_NOT: | |
754 | if (!p->next) | |
755 | die("--not not followed by pattern expression"); | |
756 | *list = p->next; | |
757 | x = xcalloc(1, sizeof (struct grep_expr)); | |
758 | x->node = GREP_NODE_NOT; | |
759 | x->u.unary = compile_pattern_not(list); | |
760 | if (!x->u.unary) | |
761 | die("--not followed by non pattern expression"); | |
762 | return x; | |
763 | default: | |
764 | return compile_pattern_atom(list); | |
765 | } | |
766 | } | |
767 | ||
768 | static struct grep_expr *compile_pattern_and(struct grep_pat **list) | |
769 | { | |
770 | struct grep_pat *p; | |
771 | struct grep_expr *x, *y, *z; | |
772 | ||
773 | x = compile_pattern_not(list); | |
774 | p = *list; | |
775 | if (p && p->token == GREP_AND) { | |
776 | if (!p->next) | |
777 | die("--and not followed by pattern expression"); | |
778 | *list = p->next; | |
779 | y = compile_pattern_and(list); | |
780 | if (!y) | |
781 | die("--and not followed by pattern expression"); | |
782 | z = xcalloc(1, sizeof (struct grep_expr)); | |
783 | z->node = GREP_NODE_AND; | |
784 | z->u.binary.left = x; | |
785 | z->u.binary.right = y; | |
786 | return z; | |
787 | } | |
788 | return x; | |
789 | } | |
790 | ||
791 | static struct grep_expr *compile_pattern_or(struct grep_pat **list) | |
792 | { | |
793 | struct grep_pat *p; | |
794 | struct grep_expr *x, *y, *z; | |
795 | ||
796 | x = compile_pattern_and(list); | |
797 | p = *list; | |
798 | if (x && p && p->token != GREP_CLOSE_PAREN) { | |
799 | y = compile_pattern_or(list); | |
800 | if (!y) | |
801 | die("not a pattern expression %s", p->pattern); | |
802 | z = xcalloc(1, sizeof (struct grep_expr)); | |
803 | z->node = GREP_NODE_OR; | |
804 | z->u.binary.left = x; | |
805 | z->u.binary.right = y; | |
806 | return z; | |
807 | } | |
808 | return x; | |
809 | } | |
810 | ||
811 | static struct grep_expr *compile_pattern_expr(struct grep_pat **list) | |
812 | { | |
813 | return compile_pattern_or(list); | |
814 | } | |
815 | ||
816 | static void indent(int in) | |
817 | { | |
818 | while (in-- > 0) | |
819 | fputc(' ', stderr); | |
820 | } | |
821 | ||
822 | static void dump_grep_pat(struct grep_pat *p) | |
823 | { | |
824 | switch (p->token) { | |
825 | case GREP_AND: fprintf(stderr, "*and*"); break; | |
826 | case GREP_OPEN_PAREN: fprintf(stderr, "*(*"); break; | |
827 | case GREP_CLOSE_PAREN: fprintf(stderr, "*)*"); break; | |
828 | case GREP_NOT: fprintf(stderr, "*not*"); break; | |
829 | case GREP_OR: fprintf(stderr, "*or*"); break; | |
830 | ||
831 | case GREP_PATTERN: fprintf(stderr, "pattern"); break; | |
832 | case GREP_PATTERN_HEAD: fprintf(stderr, "pattern_head"); break; | |
833 | case GREP_PATTERN_BODY: fprintf(stderr, "pattern_body"); break; | |
834 | } | |
835 | ||
836 | switch (p->token) { | |
837 | default: break; | |
838 | case GREP_PATTERN_HEAD: | |
839 | fprintf(stderr, "<head %d>", p->field); break; | |
840 | case GREP_PATTERN_BODY: | |
841 | fprintf(stderr, "<body>"); break; | |
842 | } | |
843 | switch (p->token) { | |
844 | default: break; | |
845 | case GREP_PATTERN_HEAD: | |
846 | case GREP_PATTERN_BODY: | |
847 | case GREP_PATTERN: | |
848 | fprintf(stderr, "%.*s", (int)p->patternlen, p->pattern); | |
849 | break; | |
850 | } | |
851 | fputc('\n', stderr); | |
852 | } | |
853 | ||
854 | static void dump_grep_expression_1(struct grep_expr *x, int in) | |
855 | { | |
856 | indent(in); | |
857 | switch (x->node) { | |
858 | case GREP_NODE_TRUE: | |
859 | fprintf(stderr, "true\n"); | |
860 | break; | |
861 | case GREP_NODE_ATOM: | |
862 | dump_grep_pat(x->u.atom); | |
863 | break; | |
864 | case GREP_NODE_NOT: | |
865 | fprintf(stderr, "(not\n"); | |
866 | dump_grep_expression_1(x->u.unary, in+1); | |
867 | indent(in); | |
868 | fprintf(stderr, ")\n"); | |
869 | break; | |
870 | case GREP_NODE_AND: | |
871 | fprintf(stderr, "(and\n"); | |
872 | dump_grep_expression_1(x->u.binary.left, in+1); | |
873 | dump_grep_expression_1(x->u.binary.right, in+1); | |
874 | indent(in); | |
875 | fprintf(stderr, ")\n"); | |
876 | break; | |
877 | case GREP_NODE_OR: | |
878 | fprintf(stderr, "(or\n"); | |
879 | dump_grep_expression_1(x->u.binary.left, in+1); | |
880 | dump_grep_expression_1(x->u.binary.right, in+1); | |
881 | indent(in); | |
882 | fprintf(stderr, ")\n"); | |
883 | break; | |
884 | } | |
885 | } | |
886 | ||
887 | static void dump_grep_expression(struct grep_opt *opt) | |
888 | { | |
889 | struct grep_expr *x = opt->pattern_expression; | |
890 | ||
891 | if (opt->all_match) | |
892 | fprintf(stderr, "[all-match]\n"); | |
893 | dump_grep_expression_1(x, 0); | |
894 | fflush(NULL); | |
895 | } | |
896 | ||
897 | static struct grep_expr *grep_true_expr(void) | |
898 | { | |
899 | struct grep_expr *z = xcalloc(1, sizeof(*z)); | |
900 | z->node = GREP_NODE_TRUE; | |
901 | return z; | |
902 | } | |
903 | ||
904 | static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right) | |
905 | { | |
906 | struct grep_expr *z = xcalloc(1, sizeof(*z)); | |
907 | z->node = GREP_NODE_OR; | |
908 | z->u.binary.left = left; | |
909 | z->u.binary.right = right; | |
910 | return z; | |
911 | } | |
912 | ||
913 | static struct grep_expr *prep_header_patterns(struct grep_opt *opt) | |
914 | { | |
915 | struct grep_pat *p; | |
916 | struct grep_expr *header_expr; | |
917 | struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]); | |
918 | enum grep_header_field fld; | |
919 | ||
920 | if (!opt->header_list) | |
921 | return NULL; | |
922 | ||
923 | for (p = opt->header_list; p; p = p->next) { | |
924 | if (p->token != GREP_PATTERN_HEAD) | |
925 | BUG("a non-header pattern in grep header list."); | |
926 | if (p->field < GREP_HEADER_FIELD_MIN || | |
927 | GREP_HEADER_FIELD_MAX <= p->field) | |
928 | BUG("unknown header field %d", p->field); | |
929 | compile_regexp(p, opt); | |
930 | } | |
931 | ||
932 | for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) | |
933 | header_group[fld] = NULL; | |
934 | ||
935 | for (p = opt->header_list; p; p = p->next) { | |
936 | struct grep_expr *h; | |
937 | struct grep_pat *pp = p; | |
938 | ||
939 | h = compile_pattern_atom(&pp); | |
940 | if (!h || pp != p->next) | |
941 | BUG("malformed header expr"); | |
942 | if (!header_group[p->field]) { | |
943 | header_group[p->field] = h; | |
944 | continue; | |
945 | } | |
946 | header_group[p->field] = grep_or_expr(h, header_group[p->field]); | |
947 | } | |
948 | ||
949 | header_expr = NULL; | |
950 | ||
951 | for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) { | |
952 | if (!header_group[fld]) | |
953 | continue; | |
954 | if (!header_expr) | |
955 | header_expr = grep_true_expr(); | |
956 | header_expr = grep_or_expr(header_group[fld], header_expr); | |
957 | } | |
958 | return header_expr; | |
959 | } | |
960 | ||
961 | static struct grep_expr *grep_splice_or(struct grep_expr *x, struct grep_expr *y) | |
962 | { | |
963 | struct grep_expr *z = x; | |
964 | ||
965 | while (x) { | |
966 | assert(x->node == GREP_NODE_OR); | |
967 | if (x->u.binary.right && | |
968 | x->u.binary.right->node == GREP_NODE_TRUE) { | |
969 | x->u.binary.right = y; | |
970 | break; | |
971 | } | |
972 | x = x->u.binary.right; | |
973 | } | |
974 | return z; | |
975 | } | |
976 | ||
977 | static void compile_grep_patterns_real(struct grep_opt *opt) | |
978 | { | |
979 | struct grep_pat *p; | |
980 | struct grep_expr *header_expr = prep_header_patterns(opt); | |
981 | ||
982 | for (p = opt->pattern_list; p; p = p->next) { | |
983 | switch (p->token) { | |
984 | case GREP_PATTERN: /* atom */ | |
985 | case GREP_PATTERN_HEAD: | |
986 | case GREP_PATTERN_BODY: | |
987 | compile_regexp(p, opt); | |
988 | break; | |
989 | default: | |
990 | opt->extended = 1; | |
991 | break; | |
992 | } | |
993 | } | |
994 | ||
995 | if (opt->all_match || header_expr) | |
996 | opt->extended = 1; | |
997 | else if (!opt->extended && !opt->debug) | |
998 | return; | |
999 | ||
1000 | p = opt->pattern_list; | |
1001 | if (p) | |
1002 | opt->pattern_expression = compile_pattern_expr(&p); | |
1003 | if (p) | |
1004 | die("incomplete pattern expression: %s", p->pattern); | |
1005 | ||
1006 | if (!header_expr) | |
1007 | return; | |
1008 | ||
1009 | if (!opt->pattern_expression) | |
1010 | opt->pattern_expression = header_expr; | |
1011 | else if (opt->all_match) | |
1012 | opt->pattern_expression = grep_splice_or(header_expr, | |
1013 | opt->pattern_expression); | |
1014 | else | |
1015 | opt->pattern_expression = grep_or_expr(opt->pattern_expression, | |
1016 | header_expr); | |
1017 | opt->all_match = 1; | |
1018 | } | |
1019 | ||
1020 | void compile_grep_patterns(struct grep_opt *opt) | |
1021 | { | |
1022 | compile_grep_patterns_real(opt); | |
1023 | if (opt->debug) | |
1024 | dump_grep_expression(opt); | |
1025 | } | |
1026 | ||
1027 | static void free_pattern_expr(struct grep_expr *x) | |
1028 | { | |
1029 | switch (x->node) { | |
1030 | case GREP_NODE_TRUE: | |
1031 | case GREP_NODE_ATOM: | |
1032 | break; | |
1033 | case GREP_NODE_NOT: | |
1034 | free_pattern_expr(x->u.unary); | |
1035 | break; | |
1036 | case GREP_NODE_AND: | |
1037 | case GREP_NODE_OR: | |
1038 | free_pattern_expr(x->u.binary.left); | |
1039 | free_pattern_expr(x->u.binary.right); | |
1040 | break; | |
1041 | } | |
1042 | free(x); | |
1043 | } | |
1044 | ||
1045 | void free_grep_patterns(struct grep_opt *opt) | |
1046 | { | |
1047 | struct grep_pat *p, *n; | |
1048 | ||
1049 | for (p = opt->pattern_list; p; p = n) { | |
1050 | n = p->next; | |
1051 | switch (p->token) { | |
1052 | case GREP_PATTERN: /* atom */ | |
1053 | case GREP_PATTERN_HEAD: | |
1054 | case GREP_PATTERN_BODY: | |
1055 | if (p->kws) | |
1056 | kwsfree(p->kws); | |
1057 | else if (p->pcre1_regexp) | |
1058 | free_pcre1_regexp(p); | |
1059 | else if (p->pcre2_pattern) | |
1060 | free_pcre2_pattern(p); | |
1061 | else | |
1062 | regfree(&p->regexp); | |
1063 | free(p->pattern); | |
1064 | break; | |
1065 | default: | |
1066 | break; | |
1067 | } | |
1068 | free(p); | |
1069 | } | |
1070 | ||
1071 | if (!opt->extended) | |
1072 | return; | |
1073 | free_pattern_expr(opt->pattern_expression); | |
1074 | } | |
1075 | ||
1076 | static char *end_of_line(char *cp, unsigned long *left) | |
1077 | { | |
1078 | unsigned long l = *left; | |
1079 | while (l && *cp != '\n') { | |
1080 | l--; | |
1081 | cp++; | |
1082 | } | |
1083 | *left = l; | |
1084 | return cp; | |
1085 | } | |
1086 | ||
1087 | static int word_char(char ch) | |
1088 | { | |
1089 | return isalnum(ch) || ch == '_'; | |
1090 | } | |
1091 | ||
1092 | static void output_color(struct grep_opt *opt, const void *data, size_t size, | |
1093 | const char *color) | |
1094 | { | |
1095 | if (want_color(opt->color) && color && color[0]) { | |
1096 | opt->output(opt, color, strlen(color)); | |
1097 | opt->output(opt, data, size); | |
1098 | opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET)); | |
1099 | } else | |
1100 | opt->output(opt, data, size); | |
1101 | } | |
1102 | ||
1103 | static void output_sep(struct grep_opt *opt, char sign) | |
1104 | { | |
1105 | if (opt->null_following_name) | |
1106 | opt->output(opt, "\0", 1); | |
1107 | else | |
1108 | output_color(opt, &sign, 1, opt->colors[GREP_COLOR_SEP]); | |
1109 | } | |
1110 | ||
1111 | static void show_name(struct grep_opt *opt, const char *name) | |
1112 | { | |
1113 | output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]); | |
1114 | opt->output(opt, opt->null_following_name ? "\0" : "\n", 1); | |
1115 | } | |
1116 | ||
1117 | static int fixmatch(struct grep_pat *p, char *line, char *eol, | |
1118 | regmatch_t *match) | |
1119 | { | |
1120 | struct kwsmatch kwsm; | |
1121 | size_t offset = kwsexec(p->kws, line, eol - line, &kwsm); | |
1122 | if (offset == -1) { | |
1123 | match->rm_so = match->rm_eo = -1; | |
1124 | return REG_NOMATCH; | |
1125 | } else { | |
1126 | match->rm_so = offset; | |
1127 | match->rm_eo = match->rm_so + kwsm.size[0]; | |
1128 | return 0; | |
1129 | } | |
1130 | } | |
1131 | ||
1132 | static int patmatch(struct grep_pat *p, char *line, char *eol, | |
1133 | regmatch_t *match, int eflags) | |
1134 | { | |
1135 | int hit; | |
1136 | ||
1137 | if (p->fixed) | |
1138 | hit = !fixmatch(p, line, eol, match); | |
1139 | else if (p->pcre1_regexp) | |
1140 | hit = !pcre1match(p, line, eol, match, eflags); | |
1141 | else if (p->pcre2_pattern) | |
1142 | hit = !pcre2match(p, line, eol, match, eflags); | |
1143 | else | |
1144 | hit = !regexec_buf(&p->regexp, line, eol - line, 1, match, | |
1145 | eflags); | |
1146 | ||
1147 | return hit; | |
1148 | } | |
1149 | ||
1150 | static int strip_timestamp(char *bol, char **eol_p) | |
1151 | { | |
1152 | char *eol = *eol_p; | |
1153 | int ch; | |
1154 | ||
1155 | while (bol < --eol) { | |
1156 | if (*eol != '>') | |
1157 | continue; | |
1158 | *eol_p = ++eol; | |
1159 | ch = *eol; | |
1160 | *eol = '\0'; | |
1161 | return ch; | |
1162 | } | |
1163 | return 0; | |
1164 | } | |
1165 | ||
1166 | static struct { | |
1167 | const char *field; | |
1168 | size_t len; | |
1169 | } header_field[] = { | |
1170 | { "author ", 7 }, | |
1171 | { "committer ", 10 }, | |
1172 | { "reflog ", 7 }, | |
1173 | }; | |
1174 | ||
1175 | static int match_one_pattern(struct grep_pat *p, char *bol, char *eol, | |
1176 | enum grep_context ctx, | |
1177 | regmatch_t *pmatch, int eflags) | |
1178 | { | |
1179 | int hit = 0; | |
1180 | int saved_ch = 0; | |
1181 | const char *start = bol; | |
1182 | ||
1183 | if ((p->token != GREP_PATTERN) && | |
1184 | ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD))) | |
1185 | return 0; | |
1186 | ||
1187 | if (p->token == GREP_PATTERN_HEAD) { | |
1188 | const char *field; | |
1189 | size_t len; | |
1190 | assert(p->field < ARRAY_SIZE(header_field)); | |
1191 | field = header_field[p->field].field; | |
1192 | len = header_field[p->field].len; | |
1193 | if (strncmp(bol, field, len)) | |
1194 | return 0; | |
1195 | bol += len; | |
1196 | switch (p->field) { | |
1197 | case GREP_HEADER_AUTHOR: | |
1198 | case GREP_HEADER_COMMITTER: | |
1199 | saved_ch = strip_timestamp(bol, &eol); | |
1200 | break; | |
1201 | default: | |
1202 | break; | |
1203 | } | |
1204 | } | |
1205 | ||
1206 | again: | |
1207 | hit = patmatch(p, bol, eol, pmatch, eflags); | |
1208 | ||
1209 | if (hit && p->word_regexp) { | |
1210 | if ((pmatch[0].rm_so < 0) || | |
1211 | (eol - bol) < pmatch[0].rm_so || | |
1212 | (pmatch[0].rm_eo < 0) || | |
1213 | (eol - bol) < pmatch[0].rm_eo) | |
1214 | die("regexp returned nonsense"); | |
1215 | ||
1216 | /* Match beginning must be either beginning of the | |
1217 | * line, or at word boundary (i.e. the last char must | |
1218 | * not be a word char). Similarly, match end must be | |
1219 | * either end of the line, or at word boundary | |
1220 | * (i.e. the next char must not be a word char). | |
1221 | */ | |
1222 | if ( ((pmatch[0].rm_so == 0) || | |
1223 | !word_char(bol[pmatch[0].rm_so-1])) && | |
1224 | ((pmatch[0].rm_eo == (eol-bol)) || | |
1225 | !word_char(bol[pmatch[0].rm_eo])) ) | |
1226 | ; | |
1227 | else | |
1228 | hit = 0; | |
1229 | ||
1230 | /* Words consist of at least one character. */ | |
1231 | if (pmatch->rm_so == pmatch->rm_eo) | |
1232 | hit = 0; | |
1233 | ||
1234 | if (!hit && pmatch[0].rm_so + bol + 1 < eol) { | |
1235 | /* There could be more than one match on the | |
1236 | * line, and the first match might not be | |
1237 | * strict word match. But later ones could be! | |
1238 | * Forward to the next possible start, i.e. the | |
1239 | * next position following a non-word char. | |
1240 | */ | |
1241 | bol = pmatch[0].rm_so + bol + 1; | |
1242 | while (word_char(bol[-1]) && bol < eol) | |
1243 | bol++; | |
1244 | eflags |= REG_NOTBOL; | |
1245 | if (bol < eol) | |
1246 | goto again; | |
1247 | } | |
1248 | } | |
1249 | if (p->token == GREP_PATTERN_HEAD && saved_ch) | |
1250 | *eol = saved_ch; | |
1251 | if (hit) { | |
1252 | pmatch[0].rm_so += bol - start; | |
1253 | pmatch[0].rm_eo += bol - start; | |
1254 | } | |
1255 | return hit; | |
1256 | } | |
1257 | ||
1258 | static int match_expr_eval(struct grep_opt *opt, struct grep_expr *x, char *bol, | |
1259 | char *eol, enum grep_context ctx, ssize_t *col, | |
1260 | ssize_t *icol, int collect_hits) | |
1261 | { | |
1262 | int h = 0; | |
1263 | ||
1264 | if (!x) | |
1265 | die("Not a valid grep expression"); | |
1266 | switch (x->node) { | |
1267 | case GREP_NODE_TRUE: | |
1268 | h = 1; | |
1269 | break; | |
1270 | case GREP_NODE_ATOM: | |
1271 | { | |
1272 | regmatch_t tmp; | |
1273 | h = match_one_pattern(x->u.atom, bol, eol, ctx, | |
1274 | &tmp, 0); | |
1275 | if (h && (*col < 0 || tmp.rm_so < *col)) | |
1276 | *col = tmp.rm_so; | |
1277 | } | |
1278 | break; | |
1279 | case GREP_NODE_NOT: | |
1280 | /* | |
1281 | * Upon visiting a GREP_NODE_NOT, col and icol become swapped. | |
1282 | */ | |
1283 | h = !match_expr_eval(opt, x->u.unary, bol, eol, ctx, icol, col, | |
1284 | 0); | |
1285 | break; | |
1286 | case GREP_NODE_AND: | |
1287 | h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col, | |
1288 | icol, 0); | |
1289 | if (h || opt->columnnum) { | |
1290 | /* | |
1291 | * Don't short-circuit AND when given --column, since a | |
1292 | * NOT earlier in the tree may turn this into an OR. In | |
1293 | * this case, see the below comment. | |
1294 | */ | |
1295 | h &= match_expr_eval(opt, x->u.binary.right, bol, eol, | |
1296 | ctx, col, icol, 0); | |
1297 | } | |
1298 | break; | |
1299 | case GREP_NODE_OR: | |
1300 | if (!(collect_hits || opt->columnnum)) { | |
1301 | /* | |
1302 | * Don't short-circuit OR when given --column (or | |
1303 | * collecting hits) to ensure we don't skip a later | |
1304 | * child that would produce an earlier match. | |
1305 | */ | |
1306 | return (match_expr_eval(opt, x->u.binary.left, bol, eol, | |
1307 | ctx, col, icol, 0) || | |
1308 | match_expr_eval(opt, x->u.binary.right, bol, | |
1309 | eol, ctx, col, icol, 0)); | |
1310 | } | |
1311 | h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col, | |
1312 | icol, 0); | |
1313 | if (collect_hits) | |
1314 | x->u.binary.left->hit |= h; | |
1315 | h |= match_expr_eval(opt, x->u.binary.right, bol, eol, ctx, col, | |
1316 | icol, collect_hits); | |
1317 | break; | |
1318 | default: | |
1319 | die("Unexpected node type (internal error) %d", x->node); | |
1320 | } | |
1321 | if (collect_hits) | |
1322 | x->hit |= h; | |
1323 | return h; | |
1324 | } | |
1325 | ||
1326 | static int match_expr(struct grep_opt *opt, char *bol, char *eol, | |
1327 | enum grep_context ctx, ssize_t *col, | |
1328 | ssize_t *icol, int collect_hits) | |
1329 | { | |
1330 | struct grep_expr *x = opt->pattern_expression; | |
1331 | return match_expr_eval(opt, x, bol, eol, ctx, col, icol, collect_hits); | |
1332 | } | |
1333 | ||
1334 | static int match_line(struct grep_opt *opt, char *bol, char *eol, | |
1335 | ssize_t *col, ssize_t *icol, | |
1336 | enum grep_context ctx, int collect_hits) | |
1337 | { | |
1338 | struct grep_pat *p; | |
1339 | int hit = 0; | |
1340 | ||
1341 | if (opt->extended) | |
1342 | return match_expr(opt, bol, eol, ctx, col, icol, | |
1343 | collect_hits); | |
1344 | ||
1345 | /* we do not call with collect_hits without being extended */ | |
1346 | for (p = opt->pattern_list; p; p = p->next) { | |
1347 | regmatch_t tmp; | |
1348 | if (match_one_pattern(p, bol, eol, ctx, &tmp, 0)) { | |
1349 | hit |= 1; | |
1350 | if (!opt->columnnum) { | |
1351 | /* | |
1352 | * Without --column, any single match on a line | |
1353 | * is enough to know that it needs to be | |
1354 | * printed. With --column, scan _all_ patterns | |
1355 | * to find the earliest. | |
1356 | */ | |
1357 | break; | |
1358 | } | |
1359 | if (*col < 0 || tmp.rm_so < *col) | |
1360 | *col = tmp.rm_so; | |
1361 | } | |
1362 | } | |
1363 | return hit; | |
1364 | } | |
1365 | ||
1366 | static int match_next_pattern(struct grep_pat *p, char *bol, char *eol, | |
1367 | enum grep_context ctx, | |
1368 | regmatch_t *pmatch, int eflags) | |
1369 | { | |
1370 | regmatch_t match; | |
1371 | ||
1372 | if (!match_one_pattern(p, bol, eol, ctx, &match, eflags)) | |
1373 | return 0; | |
1374 | if (match.rm_so < 0 || match.rm_eo < 0) | |
1375 | return 0; | |
1376 | if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) { | |
1377 | if (match.rm_so > pmatch->rm_so) | |
1378 | return 1; | |
1379 | if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo) | |
1380 | return 1; | |
1381 | } | |
1382 | pmatch->rm_so = match.rm_so; | |
1383 | pmatch->rm_eo = match.rm_eo; | |
1384 | return 1; | |
1385 | } | |
1386 | ||
1387 | static int next_match(struct grep_opt *opt, char *bol, char *eol, | |
1388 | enum grep_context ctx, regmatch_t *pmatch, int eflags) | |
1389 | { | |
1390 | struct grep_pat *p; | |
1391 | int hit = 0; | |
1392 | ||
1393 | pmatch->rm_so = pmatch->rm_eo = -1; | |
1394 | if (bol < eol) { | |
1395 | for (p = opt->pattern_list; p; p = p->next) { | |
1396 | switch (p->token) { | |
1397 | case GREP_PATTERN: /* atom */ | |
1398 | case GREP_PATTERN_HEAD: | |
1399 | case GREP_PATTERN_BODY: | |
1400 | hit |= match_next_pattern(p, bol, eol, ctx, | |
1401 | pmatch, eflags); | |
1402 | break; | |
1403 | default: | |
1404 | break; | |
1405 | } | |
1406 | } | |
1407 | } | |
1408 | return hit; | |
1409 | } | |
1410 | ||
1411 | static void show_line_header(struct grep_opt *opt, const char *name, | |
1412 | unsigned lno, ssize_t cno, char sign) | |
1413 | { | |
1414 | if (opt->heading && opt->last_shown == 0) { | |
1415 | output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]); | |
1416 | opt->output(opt, "\n", 1); | |
1417 | } | |
1418 | opt->last_shown = lno; | |
1419 | ||
1420 | if (!opt->heading && opt->pathname) { | |
1421 | output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]); | |
1422 | output_sep(opt, sign); | |
1423 | } | |
1424 | if (opt->linenum) { | |
1425 | char buf[32]; | |
1426 | xsnprintf(buf, sizeof(buf), "%d", lno); | |
1427 | output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_LINENO]); | |
1428 | output_sep(opt, sign); | |
1429 | } | |
1430 | /* | |
1431 | * Treat 'cno' as the 1-indexed offset from the start of a non-context | |
1432 | * line to its first match. Otherwise, 'cno' is 0 indicating that we are | |
1433 | * being called with a context line. | |
1434 | */ | |
1435 | if (opt->columnnum && cno) { | |
1436 | char buf[32]; | |
1437 | xsnprintf(buf, sizeof(buf), "%"PRIuMAX, (uintmax_t)cno); | |
1438 | output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_COLUMNNO]); | |
1439 | output_sep(opt, sign); | |
1440 | } | |
1441 | } | |
1442 | ||
1443 | static void show_line(struct grep_opt *opt, char *bol, char *eol, | |
1444 | const char *name, unsigned lno, ssize_t cno, char sign) | |
1445 | { | |
1446 | int rest = eol - bol; | |
1447 | const char *match_color = NULL; | |
1448 | const char *line_color = NULL; | |
1449 | ||
1450 | if (opt->file_break && opt->last_shown == 0) { | |
1451 | if (opt->show_hunk_mark) | |
1452 | opt->output(opt, "\n", 1); | |
1453 | } else if (opt->pre_context || opt->post_context || opt->funcbody) { | |
1454 | if (opt->last_shown == 0) { | |
1455 | if (opt->show_hunk_mark) { | |
1456 | output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]); | |
1457 | opt->output(opt, "\n", 1); | |
1458 | } | |
1459 | } else if (lno > opt->last_shown + 1) { | |
1460 | output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]); | |
1461 | opt->output(opt, "\n", 1); | |
1462 | } | |
1463 | } | |
1464 | if (!opt->only_matching) { | |
1465 | /* | |
1466 | * In case the line we're being called with contains more than | |
1467 | * one match, leave printing each header to the loop below. | |
1468 | */ | |
1469 | show_line_header(opt, name, lno, cno, sign); | |
1470 | } | |
1471 | if (opt->color || opt->only_matching) { | |
1472 | regmatch_t match; | |
1473 | enum grep_context ctx = GREP_CONTEXT_BODY; | |
1474 | int ch = *eol; | |
1475 | int eflags = 0; | |
1476 | ||
1477 | if (opt->color) { | |
1478 | if (sign == ':') | |
1479 | match_color = opt->colors[GREP_COLOR_MATCH_SELECTED]; | |
1480 | else | |
1481 | match_color = opt->colors[GREP_COLOR_MATCH_CONTEXT]; | |
1482 | if (sign == ':') | |
1483 | line_color = opt->colors[GREP_COLOR_SELECTED]; | |
1484 | else if (sign == '-') | |
1485 | line_color = opt->colors[GREP_COLOR_CONTEXT]; | |
1486 | else if (sign == '=') | |
1487 | line_color = opt->colors[GREP_COLOR_FUNCTION]; | |
1488 | } | |
1489 | *eol = '\0'; | |
1490 | while (next_match(opt, bol, eol, ctx, &match, eflags)) { | |
1491 | if (match.rm_so == match.rm_eo) | |
1492 | break; | |
1493 | ||
1494 | if (opt->only_matching) | |
1495 | show_line_header(opt, name, lno, cno, sign); | |
1496 | else | |
1497 | output_color(opt, bol, match.rm_so, line_color); | |
1498 | output_color(opt, bol + match.rm_so, | |
1499 | match.rm_eo - match.rm_so, match_color); | |
1500 | if (opt->only_matching) | |
1501 | opt->output(opt, "\n", 1); | |
1502 | bol += match.rm_eo; | |
1503 | cno += match.rm_eo; | |
1504 | rest -= match.rm_eo; | |
1505 | eflags = REG_NOTBOL; | |
1506 | } | |
1507 | *eol = ch; | |
1508 | } | |
1509 | if (!opt->only_matching) { | |
1510 | output_color(opt, bol, rest, line_color); | |
1511 | opt->output(opt, "\n", 1); | |
1512 | } | |
1513 | } | |
1514 | ||
1515 | #ifndef NO_PTHREADS | |
1516 | int grep_use_locks; | |
1517 | ||
1518 | /* | |
1519 | * This lock protects access to the gitattributes machinery, which is | |
1520 | * not thread-safe. | |
1521 | */ | |
1522 | pthread_mutex_t grep_attr_mutex; | |
1523 | ||
1524 | static inline void grep_attr_lock(void) | |
1525 | { | |
1526 | if (grep_use_locks) | |
1527 | pthread_mutex_lock(&grep_attr_mutex); | |
1528 | } | |
1529 | ||
1530 | static inline void grep_attr_unlock(void) | |
1531 | { | |
1532 | if (grep_use_locks) | |
1533 | pthread_mutex_unlock(&grep_attr_mutex); | |
1534 | } | |
1535 | ||
1536 | /* | |
1537 | * Same as git_attr_mutex, but protecting the thread-unsafe object db access. | |
1538 | */ | |
1539 | pthread_mutex_t grep_read_mutex; | |
1540 | ||
1541 | #else | |
1542 | #define grep_attr_lock() | |
1543 | #define grep_attr_unlock() | |
1544 | #endif | |
1545 | ||
1546 | static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol) | |
1547 | { | |
1548 | xdemitconf_t *xecfg = opt->priv; | |
1549 | if (xecfg && !xecfg->find_func) { | |
1550 | grep_source_load_driver(gs); | |
1551 | if (gs->driver->funcname.pattern) { | |
1552 | const struct userdiff_funcname *pe = &gs->driver->funcname; | |
1553 | xdiff_set_find_func(xecfg, pe->pattern, pe->cflags); | |
1554 | } else { | |
1555 | xecfg = opt->priv = NULL; | |
1556 | } | |
1557 | } | |
1558 | ||
1559 | if (xecfg) { | |
1560 | char buf[1]; | |
1561 | return xecfg->find_func(bol, eol - bol, buf, 1, | |
1562 | xecfg->find_func_priv) >= 0; | |
1563 | } | |
1564 | ||
1565 | if (bol == eol) | |
1566 | return 0; | |
1567 | if (isalpha(*bol) || *bol == '_' || *bol == '$') | |
1568 | return 1; | |
1569 | return 0; | |
1570 | } | |
1571 | ||
1572 | static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs, | |
1573 | char *bol, unsigned lno) | |
1574 | { | |
1575 | while (bol > gs->buf) { | |
1576 | char *eol = --bol; | |
1577 | ||
1578 | while (bol > gs->buf && bol[-1] != '\n') | |
1579 | bol--; | |
1580 | lno--; | |
1581 | ||
1582 | if (lno <= opt->last_shown) | |
1583 | break; | |
1584 | ||
1585 | if (match_funcname(opt, gs, bol, eol)) { | |
1586 | show_line(opt, bol, eol, gs->name, lno, 0, '='); | |
1587 | break; | |
1588 | } | |
1589 | } | |
1590 | } | |
1591 | ||
1592 | static int is_empty_line(const char *bol, const char *eol); | |
1593 | ||
1594 | static void show_pre_context(struct grep_opt *opt, struct grep_source *gs, | |
1595 | char *bol, char *end, unsigned lno) | |
1596 | { | |
1597 | unsigned cur = lno, from = 1, funcname_lno = 0, orig_from; | |
1598 | int funcname_needed = !!opt->funcname, comment_needed = 0; | |
1599 | ||
1600 | if (opt->pre_context < lno) | |
1601 | from = lno - opt->pre_context; | |
1602 | if (from <= opt->last_shown) | |
1603 | from = opt->last_shown + 1; | |
1604 | orig_from = from; | |
1605 | if (opt->funcbody) { | |
1606 | if (match_funcname(opt, gs, bol, end)) | |
1607 | comment_needed = 1; | |
1608 | else | |
1609 | funcname_needed = 1; | |
1610 | from = opt->last_shown + 1; | |
1611 | } | |
1612 | ||
1613 | /* Rewind. */ | |
1614 | while (bol > gs->buf && cur > from) { | |
1615 | char *next_bol = bol; | |
1616 | char *eol = --bol; | |
1617 | ||
1618 | while (bol > gs->buf && bol[-1] != '\n') | |
1619 | bol--; | |
1620 | cur--; | |
1621 | if (comment_needed && (is_empty_line(bol, eol) || | |
1622 | match_funcname(opt, gs, bol, eol))) { | |
1623 | comment_needed = 0; | |
1624 | from = orig_from; | |
1625 | if (cur < from) { | |
1626 | cur++; | |
1627 | bol = next_bol; | |
1628 | break; | |
1629 | } | |
1630 | } | |
1631 | if (funcname_needed && match_funcname(opt, gs, bol, eol)) { | |
1632 | funcname_lno = cur; | |
1633 | funcname_needed = 0; | |
1634 | if (opt->funcbody) | |
1635 | comment_needed = 1; | |
1636 | else | |
1637 | from = orig_from; | |
1638 | } | |
1639 | } | |
1640 | ||
1641 | /* We need to look even further back to find a function signature. */ | |
1642 | if (opt->funcname && funcname_needed) | |
1643 | show_funcname_line(opt, gs, bol, cur); | |
1644 | ||
1645 | /* Back forward. */ | |
1646 | while (cur < lno) { | |
1647 | char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-'; | |
1648 | ||
1649 | while (*eol != '\n') | |
1650 | eol++; | |
1651 | show_line(opt, bol, eol, gs->name, cur, 0, sign); | |
1652 | bol = eol + 1; | |
1653 | cur++; | |
1654 | } | |
1655 | } | |
1656 | ||
1657 | static int should_lookahead(struct grep_opt *opt) | |
1658 | { | |
1659 | struct grep_pat *p; | |
1660 | ||
1661 | if (opt->extended) | |
1662 | return 0; /* punt for too complex stuff */ | |
1663 | if (opt->invert) | |
1664 | return 0; | |
1665 | for (p = opt->pattern_list; p; p = p->next) { | |
1666 | if (p->token != GREP_PATTERN) | |
1667 | return 0; /* punt for "header only" and stuff */ | |
1668 | } | |
1669 | return 1; | |
1670 | } | |
1671 | ||
1672 | static int look_ahead(struct grep_opt *opt, | |
1673 | unsigned long *left_p, | |
1674 | unsigned *lno_p, | |
1675 | char **bol_p) | |
1676 | { | |
1677 | unsigned lno = *lno_p; | |
1678 | char *bol = *bol_p; | |
1679 | struct grep_pat *p; | |
1680 | char *sp, *last_bol; | |
1681 | regoff_t earliest = -1; | |
1682 | ||
1683 | for (p = opt->pattern_list; p; p = p->next) { | |
1684 | int hit; | |
1685 | regmatch_t m; | |
1686 | ||
1687 | hit = patmatch(p, bol, bol + *left_p, &m, 0); | |
1688 | if (!hit || m.rm_so < 0 || m.rm_eo < 0) | |
1689 | continue; | |
1690 | if (earliest < 0 || m.rm_so < earliest) | |
1691 | earliest = m.rm_so; | |
1692 | } | |
1693 | ||
1694 | if (earliest < 0) { | |
1695 | *bol_p = bol + *left_p; | |
1696 | *left_p = 0; | |
1697 | return 1; | |
1698 | } | |
1699 | for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--) | |
1700 | ; /* find the beginning of the line */ | |
1701 | last_bol = sp; | |
1702 | ||
1703 | for (sp = bol; sp < last_bol; sp++) { | |
1704 | if (*sp == '\n') | |
1705 | lno++; | |
1706 | } | |
1707 | *left_p -= last_bol - bol; | |
1708 | *bol_p = last_bol; | |
1709 | *lno_p = lno; | |
1710 | return 0; | |
1711 | } | |
1712 | ||
1713 | static int fill_textconv_grep(struct repository *r, | |
1714 | struct userdiff_driver *driver, | |
1715 | struct grep_source *gs) | |
1716 | { | |
1717 | struct diff_filespec *df; | |
1718 | char *buf; | |
1719 | size_t size; | |
1720 | ||
1721 | if (!driver || !driver->textconv) | |
1722 | return grep_source_load(gs); | |
1723 | ||
1724 | /* | |
1725 | * The textconv interface is intimately tied to diff_filespecs, so we | |
1726 | * have to pretend to be one. If we could unify the grep_source | |
1727 | * and diff_filespec structs, this mess could just go away. | |
1728 | */ | |
1729 | df = alloc_filespec(gs->path); | |
1730 | switch (gs->type) { | |
1731 | case GREP_SOURCE_OID: | |
1732 | fill_filespec(df, gs->identifier, 1, 0100644); | |
1733 | break; | |
1734 | case GREP_SOURCE_FILE: | |
1735 | fill_filespec(df, &null_oid, 0, 0100644); | |
1736 | break; | |
1737 | default: | |
1738 | BUG("attempt to textconv something without a path?"); | |
1739 | } | |
1740 | ||
1741 | /* | |
1742 | * fill_textconv is not remotely thread-safe; it may load objects | |
1743 | * behind the scenes, and it modifies the global diff tempfile | |
1744 | * structure. | |
1745 | */ | |
1746 | grep_read_lock(); | |
1747 | size = fill_textconv(r, driver, df, &buf); | |
1748 | grep_read_unlock(); | |
1749 | free_filespec(df); | |
1750 | ||
1751 | /* | |
1752 | * The normal fill_textconv usage by the diff machinery would just keep | |
1753 | * the textconv'd buf separate from the diff_filespec. But much of the | |
1754 | * grep code passes around a grep_source and assumes that its "buf" | |
1755 | * pointer is the beginning of the thing we are searching. So let's | |
1756 | * install our textconv'd version into the grep_source, taking care not | |
1757 | * to leak any existing buffer. | |
1758 | */ | |
1759 | grep_source_clear_data(gs); | |
1760 | gs->buf = buf; | |
1761 | gs->size = size; | |
1762 | ||
1763 | return 0; | |
1764 | } | |
1765 | ||
1766 | static int is_empty_line(const char *bol, const char *eol) | |
1767 | { | |
1768 | while (bol < eol && isspace(*bol)) | |
1769 | bol++; | |
1770 | return bol == eol; | |
1771 | } | |
1772 | ||
1773 | static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits) | |
1774 | { | |
1775 | char *bol; | |
1776 | char *peek_bol = NULL; | |
1777 | unsigned long left; | |
1778 | unsigned lno = 1; | |
1779 | unsigned last_hit = 0; | |
1780 | int binary_match_only = 0; | |
1781 | unsigned count = 0; | |
1782 | int try_lookahead = 0; | |
1783 | int show_function = 0; | |
1784 | struct userdiff_driver *textconv = NULL; | |
1785 | enum grep_context ctx = GREP_CONTEXT_HEAD; | |
1786 | xdemitconf_t xecfg; | |
1787 | ||
1788 | if (!opt->output) | |
1789 | opt->output = std_output; | |
1790 | ||
1791 | if (opt->pre_context || opt->post_context || opt->file_break || | |
1792 | opt->funcbody) { | |
1793 | /* Show hunk marks, except for the first file. */ | |
1794 | if (opt->last_shown) | |
1795 | opt->show_hunk_mark = 1; | |
1796 | /* | |
1797 | * If we're using threads then we can't easily identify | |
1798 | * the first file. Always put hunk marks in that case | |
1799 | * and skip the very first one later in work_done(). | |
1800 | */ | |
1801 | if (opt->output != std_output) | |
1802 | opt->show_hunk_mark = 1; | |
1803 | } | |
1804 | opt->last_shown = 0; | |
1805 | ||
1806 | if (opt->allow_textconv) { | |
1807 | grep_source_load_driver(gs); | |
1808 | /* | |
1809 | * We might set up the shared textconv cache data here, which | |
1810 | * is not thread-safe. | |
1811 | */ | |
1812 | grep_attr_lock(); | |
1813 | textconv = userdiff_get_textconv(gs->driver); | |
1814 | grep_attr_unlock(); | |
1815 | } | |
1816 | ||
1817 | /* | |
1818 | * We know the result of a textconv is text, so we only have to care | |
1819 | * about binary handling if we are not using it. | |
1820 | */ | |
1821 | if (!textconv) { | |
1822 | switch (opt->binary) { | |
1823 | case GREP_BINARY_DEFAULT: | |
1824 | if (grep_source_is_binary(gs)) | |
1825 | binary_match_only = 1; | |
1826 | break; | |
1827 | case GREP_BINARY_NOMATCH: | |
1828 | if (grep_source_is_binary(gs)) | |
1829 | return 0; /* Assume unmatch */ | |
1830 | break; | |
1831 | case GREP_BINARY_TEXT: | |
1832 | break; | |
1833 | default: | |
1834 | BUG("unknown binary handling mode"); | |
1835 | } | |
1836 | } | |
1837 | ||
1838 | memset(&xecfg, 0, sizeof(xecfg)); | |
1839 | opt->priv = &xecfg; | |
1840 | ||
1841 | try_lookahead = should_lookahead(opt); | |
1842 | ||
1843 | if (fill_textconv_grep(opt->repo, textconv, gs) < 0) | |
1844 | return 0; | |
1845 | ||
1846 | bol = gs->buf; | |
1847 | left = gs->size; | |
1848 | while (left) { | |
1849 | char *eol, ch; | |
1850 | int hit; | |
1851 | ssize_t cno; | |
1852 | ssize_t col = -1, icol = -1; | |
1853 | ||
1854 | /* | |
1855 | * look_ahead() skips quickly to the line that possibly | |
1856 | * has the next hit; don't call it if we need to do | |
1857 | * something more than just skipping the current line | |
1858 | * in response to an unmatch for the current line. E.g. | |
1859 | * inside a post-context window, we will show the current | |
1860 | * line as a context around the previous hit when it | |
1861 | * doesn't hit. | |
1862 | */ | |
1863 | if (try_lookahead | |
1864 | && !(last_hit | |
1865 | && (show_function || | |
1866 | lno <= last_hit + opt->post_context)) | |
1867 | && look_ahead(opt, &left, &lno, &bol)) | |
1868 | break; | |
1869 | eol = end_of_line(bol, &left); | |
1870 | ch = *eol; | |
1871 | *eol = 0; | |
1872 | ||
1873 | if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol)) | |
1874 | ctx = GREP_CONTEXT_BODY; | |
1875 | ||
1876 | hit = match_line(opt, bol, eol, &col, &icol, ctx, collect_hits); | |
1877 | *eol = ch; | |
1878 | ||
1879 | if (collect_hits) | |
1880 | goto next_line; | |
1881 | ||
1882 | /* "grep -v -e foo -e bla" should list lines | |
1883 | * that do not have either, so inversion should | |
1884 | * be done outside. | |
1885 | */ | |
1886 | if (opt->invert) | |
1887 | hit = !hit; | |
1888 | if (opt->unmatch_name_only) { | |
1889 | if (hit) | |
1890 | return 0; | |
1891 | goto next_line; | |
1892 | } | |
1893 | if (hit) { | |
1894 | count++; | |
1895 | if (opt->status_only) | |
1896 | return 1; | |
1897 | if (opt->name_only) { | |
1898 | show_name(opt, gs->name); | |
1899 | return 1; | |
1900 | } | |
1901 | if (opt->count) | |
1902 | goto next_line; | |
1903 | if (binary_match_only) { | |
1904 | opt->output(opt, "Binary file ", 12); | |
1905 | output_color(opt, gs->name, strlen(gs->name), | |
1906 | opt->colors[GREP_COLOR_FILENAME]); | |
1907 | opt->output(opt, " matches\n", 9); | |
1908 | return 1; | |
1909 | } | |
1910 | /* Hit at this line. If we haven't shown the | |
1911 | * pre-context lines, we would need to show them. | |
1912 | */ | |
1913 | if (opt->pre_context || opt->funcbody) | |
1914 | show_pre_context(opt, gs, bol, eol, lno); | |
1915 | else if (opt->funcname) | |
1916 | show_funcname_line(opt, gs, bol, lno); | |
1917 | cno = opt->invert ? icol : col; | |
1918 | if (cno < 0) { | |
1919 | /* | |
1920 | * A negative cno indicates that there was no | |
1921 | * match on the line. We are thus inverted and | |
1922 | * being asked to show all lines that _don't_ | |
1923 | * match a given expression. Therefore, set cno | |
1924 | * to 0 to suggest the whole line matches. | |
1925 | */ | |
1926 | cno = 0; | |
1927 | } | |
1928 | show_line(opt, bol, eol, gs->name, lno, cno + 1, ':'); | |
1929 | last_hit = lno; | |
1930 | if (opt->funcbody) | |
1931 | show_function = 1; | |
1932 | goto next_line; | |
1933 | } | |
1934 | if (show_function && (!peek_bol || peek_bol < bol)) { | |
1935 | unsigned long peek_left = left; | |
1936 | char *peek_eol = eol; | |
1937 | ||
1938 | /* | |
1939 | * Trailing empty lines are not interesting. | |
1940 | * Peek past them to see if they belong to the | |
1941 | * body of the current function. | |
1942 | */ | |
1943 | peek_bol = bol; | |
1944 | while (is_empty_line(peek_bol, peek_eol)) { | |
1945 | peek_bol = peek_eol + 1; | |
1946 | peek_eol = end_of_line(peek_bol, &peek_left); | |
1947 | } | |
1948 | ||
1949 | if (match_funcname(opt, gs, peek_bol, peek_eol)) | |
1950 | show_function = 0; | |
1951 | } | |
1952 | if (show_function || | |
1953 | (last_hit && lno <= last_hit + opt->post_context)) { | |
1954 | /* If the last hit is within the post context, | |
1955 | * we need to show this line. | |
1956 | */ | |
1957 | show_line(opt, bol, eol, gs->name, lno, col + 1, '-'); | |
1958 | } | |
1959 | ||
1960 | next_line: | |
1961 | bol = eol + 1; | |
1962 | if (!left) | |
1963 | break; | |
1964 | left--; | |
1965 | lno++; | |
1966 | } | |
1967 | ||
1968 | if (collect_hits) | |
1969 | return 0; | |
1970 | ||
1971 | if (opt->status_only) | |
1972 | return opt->unmatch_name_only; | |
1973 | if (opt->unmatch_name_only) { | |
1974 | /* We did not see any hit, so we want to show this */ | |
1975 | show_name(opt, gs->name); | |
1976 | return 1; | |
1977 | } | |
1978 | ||
1979 | xdiff_clear_find_func(&xecfg); | |
1980 | opt->priv = NULL; | |
1981 | ||
1982 | /* NEEDSWORK: | |
1983 | * The real "grep -c foo *.c" gives many "bar.c:0" lines, | |
1984 | * which feels mostly useless but sometimes useful. Maybe | |
1985 | * make it another option? For now suppress them. | |
1986 | */ | |
1987 | if (opt->count && count) { | |
1988 | char buf[32]; | |
1989 | if (opt->pathname) { | |
1990 | output_color(opt, gs->name, strlen(gs->name), | |
1991 | opt->colors[GREP_COLOR_FILENAME]); | |
1992 | output_sep(opt, ':'); | |
1993 | } | |
1994 | xsnprintf(buf, sizeof(buf), "%u\n", count); | |
1995 | opt->output(opt, buf, strlen(buf)); | |
1996 | return 1; | |
1997 | } | |
1998 | return !!last_hit; | |
1999 | } | |
2000 | ||
2001 | static void clr_hit_marker(struct grep_expr *x) | |
2002 | { | |
2003 | /* All-hit markers are meaningful only at the very top level | |
2004 | * OR node. | |
2005 | */ | |
2006 | while (1) { | |
2007 | x->hit = 0; | |
2008 | if (x->node != GREP_NODE_OR) | |
2009 | return; | |
2010 | x->u.binary.left->hit = 0; | |
2011 | x = x->u.binary.right; | |
2012 | } | |
2013 | } | |
2014 | ||
2015 | static int chk_hit_marker(struct grep_expr *x) | |
2016 | { | |
2017 | /* Top level nodes have hit markers. See if they all are hits */ | |
2018 | while (1) { | |
2019 | if (x->node != GREP_NODE_OR) | |
2020 | return x->hit; | |
2021 | if (!x->u.binary.left->hit) | |
2022 | return 0; | |
2023 | x = x->u.binary.right; | |
2024 | } | |
2025 | } | |
2026 | ||
2027 | int grep_source(struct grep_opt *opt, struct grep_source *gs) | |
2028 | { | |
2029 | /* | |
2030 | * we do not have to do the two-pass grep when we do not check | |
2031 | * buffer-wide "all-match". | |
2032 | */ | |
2033 | if (!opt->all_match) | |
2034 | return grep_source_1(opt, gs, 0); | |
2035 | ||
2036 | /* Otherwise the toplevel "or" terms hit a bit differently. | |
2037 | * We first clear hit markers from them. | |
2038 | */ | |
2039 | clr_hit_marker(opt->pattern_expression); | |
2040 | grep_source_1(opt, gs, 1); | |
2041 | ||
2042 | if (!chk_hit_marker(opt->pattern_expression)) | |
2043 | return 0; | |
2044 | ||
2045 | return grep_source_1(opt, gs, 0); | |
2046 | } | |
2047 | ||
2048 | int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size) | |
2049 | { | |
2050 | struct grep_source gs; | |
2051 | int r; | |
2052 | ||
2053 | grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL, NULL); | |
2054 | gs.buf = buf; | |
2055 | gs.size = size; | |
2056 | ||
2057 | r = grep_source(opt, &gs); | |
2058 | ||
2059 | grep_source_clear(&gs); | |
2060 | return r; | |
2061 | } | |
2062 | ||
2063 | void grep_source_init(struct grep_source *gs, enum grep_source_type type, | |
2064 | const char *name, const char *path, | |
2065 | const void *identifier) | |
2066 | { | |
2067 | gs->type = type; | |
2068 | gs->name = xstrdup_or_null(name); | |
2069 | gs->path = xstrdup_or_null(path); | |
2070 | gs->buf = NULL; | |
2071 | gs->size = 0; | |
2072 | gs->driver = NULL; | |
2073 | ||
2074 | switch (type) { | |
2075 | case GREP_SOURCE_FILE: | |
2076 | gs->identifier = xstrdup(identifier); | |
2077 | break; | |
2078 | case GREP_SOURCE_OID: | |
2079 | gs->identifier = oiddup(identifier); | |
2080 | break; | |
2081 | case GREP_SOURCE_BUF: | |
2082 | gs->identifier = NULL; | |
2083 | break; | |
2084 | } | |
2085 | } | |
2086 | ||
2087 | void grep_source_clear(struct grep_source *gs) | |
2088 | { | |
2089 | FREE_AND_NULL(gs->name); | |
2090 | FREE_AND_NULL(gs->path); | |
2091 | FREE_AND_NULL(gs->identifier); | |
2092 | grep_source_clear_data(gs); | |
2093 | } | |
2094 | ||
2095 | void grep_source_clear_data(struct grep_source *gs) | |
2096 | { | |
2097 | switch (gs->type) { | |
2098 | case GREP_SOURCE_FILE: | |
2099 | case GREP_SOURCE_OID: | |
2100 | FREE_AND_NULL(gs->buf); | |
2101 | gs->size = 0; | |
2102 | break; | |
2103 | case GREP_SOURCE_BUF: | |
2104 | /* leave user-provided buf intact */ | |
2105 | break; | |
2106 | } | |
2107 | } | |
2108 | ||
2109 | static int grep_source_load_oid(struct grep_source *gs) | |
2110 | { | |
2111 | enum object_type type; | |
2112 | ||
2113 | grep_read_lock(); | |
2114 | gs->buf = read_object_file(gs->identifier, &type, &gs->size); | |
2115 | grep_read_unlock(); | |
2116 | ||
2117 | if (!gs->buf) | |
2118 | return error(_("'%s': unable to read %s"), | |
2119 | gs->name, | |
2120 | oid_to_hex(gs->identifier)); | |
2121 | return 0; | |
2122 | } | |
2123 | ||
2124 | static int grep_source_load_file(struct grep_source *gs) | |
2125 | { | |
2126 | const char *filename = gs->identifier; | |
2127 | struct stat st; | |
2128 | char *data; | |
2129 | size_t size; | |
2130 | int i; | |
2131 | ||
2132 | if (lstat(filename, &st) < 0) { | |
2133 | err_ret: | |
2134 | if (errno != ENOENT) | |
2135 | error_errno(_("failed to stat '%s'"), filename); | |
2136 | return -1; | |
2137 | } | |
2138 | if (!S_ISREG(st.st_mode)) | |
2139 | return -1; | |
2140 | size = xsize_t(st.st_size); | |
2141 | i = open(filename, O_RDONLY); | |
2142 | if (i < 0) | |
2143 | goto err_ret; | |
2144 | data = xmallocz(size); | |
2145 | if (st.st_size != read_in_full(i, data, size)) { | |
2146 | error_errno(_("'%s': short read"), filename); | |
2147 | close(i); | |
2148 | free(data); | |
2149 | return -1; | |
2150 | } | |
2151 | close(i); | |
2152 | ||
2153 | gs->buf = data; | |
2154 | gs->size = size; | |
2155 | return 0; | |
2156 | } | |
2157 | ||
2158 | static int grep_source_load(struct grep_source *gs) | |
2159 | { | |
2160 | if (gs->buf) | |
2161 | return 0; | |
2162 | ||
2163 | switch (gs->type) { | |
2164 | case GREP_SOURCE_FILE: | |
2165 | return grep_source_load_file(gs); | |
2166 | case GREP_SOURCE_OID: | |
2167 | return grep_source_load_oid(gs); | |
2168 | case GREP_SOURCE_BUF: | |
2169 | return gs->buf ? 0 : -1; | |
2170 | } | |
2171 | BUG("invalid grep_source type to load"); | |
2172 | } | |
2173 | ||
2174 | void grep_source_load_driver(struct grep_source *gs) | |
2175 | { | |
2176 | if (gs->driver) | |
2177 | return; | |
2178 | ||
2179 | grep_attr_lock(); | |
2180 | if (gs->path) | |
2181 | gs->driver = userdiff_find_by_path(gs->path); | |
2182 | if (!gs->driver) | |
2183 | gs->driver = userdiff_find_by_name("default"); | |
2184 | grep_attr_unlock(); | |
2185 | } | |
2186 | ||
2187 | static int grep_source_is_binary(struct grep_source *gs) | |
2188 | { | |
2189 | grep_source_load_driver(gs); | |
2190 | if (gs->driver->binary != -1) | |
2191 | return gs->driver->binary; | |
2192 | ||
2193 | if (!grep_source_load(gs)) | |
2194 | return buffer_is_binary(gs->buf, gs->size); | |
2195 | ||
2196 | return 0; | |
2197 | } |