]> git.ipfire.org Git - thirdparty/git.git/blob - grep.c
projects / thirdparty / git.git / blob
? search:


5bc0f4f32abc536c2beee3d443b69b56ca82309e
[thirdparty/git.git] / grep.c
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 struct index_state *istate);
16
17 static struct grep_opt grep_defaults;
18
19 static const char *color_grep_slots[] = {
20 [GREP_COLOR_CONTEXT] = "context",
21 [GREP_COLOR_FILENAME] = "filename",
22 [GREP_COLOR_FUNCTION] = "function",
23 [GREP_COLOR_LINENO] = "lineNumber",
24 [GREP_COLOR_COLUMNNO] = "column",
25 [GREP_COLOR_MATCH_CONTEXT] = "matchContext",
26 [GREP_COLOR_MATCH_SELECTED] = "matchSelected",
27 [GREP_COLOR_SELECTED] = "selected",
28 [GREP_COLOR_SEP] = "separator",
29 };
30
31 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
32 {
33 fwrite(buf, size, 1, stdout);
34 }
35
36 static void color_set(char *dst, const char *color_bytes)
37 {
38 xsnprintf(dst, COLOR_MAXLEN, "%s", color_bytes);
39 }
40
41 /*
42 * Initialize the grep_defaults template with hardcoded defaults.
43 * We could let the compiler do this, but without C99 initializers
44 * the code gets unwieldy and unreadable, so...
45 */
46 void init_grep_defaults(struct repository *repo)
47 {
48 struct grep_opt *opt = &grep_defaults;
49 static int run_once;
50
51 if (run_once)
52 return;
53 run_once++;
54
55 memset(opt, 0, sizeof(*opt));
56 opt->repo = repo;
57 opt->relative = 1;
58 opt->pathname = 1;
59 opt->max_depth = -1;
60 opt->pattern_type_option = GREP_PATTERN_TYPE_UNSPECIFIED;
61 color_set(opt->colors[GREP_COLOR_CONTEXT], "");
62 color_set(opt->colors[GREP_COLOR_FILENAME], "");
63 color_set(opt->colors[GREP_COLOR_FUNCTION], "");
64 color_set(opt->colors[GREP_COLOR_LINENO], "");
65 color_set(opt->colors[GREP_COLOR_COLUMNNO], "");
66 color_set(opt->colors[GREP_COLOR_MATCH_CONTEXT], GIT_COLOR_BOLD_RED);
67 color_set(opt->colors[GREP_COLOR_MATCH_SELECTED], GIT_COLOR_BOLD_RED);
68 color_set(opt->colors[GREP_COLOR_SELECTED], "");
69 color_set(opt->colors[GREP_COLOR_SEP], GIT_COLOR_CYAN);
70 opt->only_matching = 0;
71 opt->color = -1;
72 opt->output = std_output;
73 }
74
75 static int parse_pattern_type_arg(const char *opt, const char *arg)
76 {
77 if (!strcmp(arg, "default"))
78 return GREP_PATTERN_TYPE_UNSPECIFIED;
79 else if (!strcmp(arg, "basic"))
80 return GREP_PATTERN_TYPE_BRE;
81 else if (!strcmp(arg, "extended"))
82 return GREP_PATTERN_TYPE_ERE;
83 else if (!strcmp(arg, "fixed"))
84 return GREP_PATTERN_TYPE_FIXED;
85 else if (!strcmp(arg, "perl"))
86 return GREP_PATTERN_TYPE_PCRE;
87 die("bad %s argument: %s", opt, arg);
88 }
89
90 define_list_config_array_extra(color_grep_slots, {"match"});
91
92 /*
93 * Read the configuration file once and store it in
94 * the grep_defaults template.
95 */
96 int grep_config(const char *var, const char *value, void *cb)
97 {
98 struct grep_opt *opt = &grep_defaults;
99 const char *slot;
100
101 if (userdiff_config(var, value) < 0)
102 return -1;
103
104 if (!strcmp(var, "grep.extendedregexp")) {
105 opt->extended_regexp_option = git_config_bool(var, value);
106 return 0;
107 }
108
109 if (!strcmp(var, "grep.patterntype")) {
110 opt->pattern_type_option = parse_pattern_type_arg(var, value);
111 return 0;
112 }
113
114 if (!strcmp(var, "grep.linenumber")) {
115 opt->linenum = git_config_bool(var, value);
116 return 0;
117 }
118 if (!strcmp(var, "grep.column")) {
119 opt->columnnum = git_config_bool(var, value);
120 return 0;
121 }
122
123 if (!strcmp(var, "grep.fullname")) {
124 opt->relative = !git_config_bool(var, value);
125 return 0;
126 }
127
128 if (!strcmp(var, "color.grep"))
129 opt->color = git_config_colorbool(var, value);
130 if (!strcmp(var, "color.grep.match")) {
131 if (grep_config("color.grep.matchcontext", value, cb) < 0)
132 return -1;
133 if (grep_config("color.grep.matchselected", value, cb) < 0)
134 return -1;
135 } else if (skip_prefix(var, "color.grep.", &slot)) {
136 int i = LOOKUP_CONFIG(color_grep_slots, slot);
137 char *color;
138
139 if (i < 0)
140 return -1;
141 color = opt->colors[i];
142 if (!value)
143 return config_error_nonbool(var);
144 return color_parse(value, color);
145 }
146 return 0;
147 }
148
149 /*
150 * Initialize one instance of grep_opt and copy the
151 * default values from the template we read the configuration
152 * information in an earlier call to git_config(grep_config).
153 */
154 void grep_init(struct grep_opt *opt, struct repository *repo, const char *prefix)
155 {
156 struct grep_opt *def = &grep_defaults;
157 int i;
158
159 memset(opt, 0, sizeof(*opt));
160 opt->repo = repo;
161 opt->prefix = prefix;
162 opt->prefix_length = (prefix && *prefix) ? strlen(prefix) : 0;
163 opt->pattern_tail = &opt->pattern_list;
164 opt->header_tail = &opt->header_list;
165
166 opt->only_matching = def->only_matching;
167 opt->color = def->color;
168 opt->extended_regexp_option = def->extended_regexp_option;
169 opt->pattern_type_option = def->pattern_type_option;
170 opt->linenum = def->linenum;
171 opt->columnnum = def->columnnum;
172 opt->max_depth = def->max_depth;
173 opt->pathname = def->pathname;
174 opt->relative = def->relative;
175 opt->output = def->output;
176
177 for (i = 0; i < NR_GREP_COLORS; i++)
178 color_set(opt->colors[i], def->colors[i]);
179 }
180
181 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type, struct grep_opt *opt)
182 {
183 /*
184 * When committing to the pattern type by setting the relevant
185 * fields in grep_opt it's generally not necessary to zero out
186 * the fields we're not choosing, since they won't have been
187 * set by anything. The extended_regexp_option field is the
188 * only exception to this.
189 *
190 * This is because in the process of parsing grep.patternType
191 * & grep.extendedRegexp we set opt->pattern_type_option and
192 * opt->extended_regexp_option, respectively. We then
193 * internally use opt->extended_regexp_option to see if we're
194 * compiling an ERE. It must be unset if that's not actually
195 * the case.
196 */
197 if (pattern_type != GREP_PATTERN_TYPE_ERE &&
198 opt->extended_regexp_option)
199 opt->extended_regexp_option = 0;
200
201 switch (pattern_type) {
202 case GREP_PATTERN_TYPE_UNSPECIFIED:
203 /* fall through */
204
205 case GREP_PATTERN_TYPE_BRE:
206 break;
207
208 case GREP_PATTERN_TYPE_ERE:
209 opt->extended_regexp_option = 1;
210 break;
211
212 case GREP_PATTERN_TYPE_FIXED:
213 opt->fixed = 1;
214 break;
215
216 case GREP_PATTERN_TYPE_PCRE:
217 #ifdef USE_LIBPCRE2
218 opt->pcre2 = 1;
219 #else
220 /*
221 * It's important that pcre1 always be assigned to
222 * even when there's no USE_LIBPCRE* defined. We still
223 * call the PCRE stub function, it just dies with
224 * "cannot use Perl-compatible regexes[...]".
225 */
226 opt->pcre1 = 1;
227 #endif
228 break;
229 }
230 }
231
232 void grep_commit_pattern_type(enum grep_pattern_type pattern_type, struct grep_opt *opt)
233 {
234 if (pattern_type != GREP_PATTERN_TYPE_UNSPECIFIED)
235 grep_set_pattern_type_option(pattern_type, opt);
236 else if (opt->pattern_type_option != GREP_PATTERN_TYPE_UNSPECIFIED)
237 grep_set_pattern_type_option(opt->pattern_type_option, opt);
238 else if (opt->extended_regexp_option)
239 /*
240 * This branch *must* happen after setting from the
241 * opt->pattern_type_option above, we don't want
242 * grep.extendedRegexp to override grep.patternType!
243 */
244 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE, opt);
245 }
246
247 static struct grep_pat *create_grep_pat(const char *pat, size_t patlen,
248 const char *origin, int no,
249 enum grep_pat_token t,
250 enum grep_header_field field)
251 {
252 struct grep_pat *p = xcalloc(1, sizeof(*p));
253 p->pattern = xmemdupz(pat, patlen);
254 p->patternlen = patlen;
255 p->origin = origin;
256 p->no = no;
257 p->token = t;
258 p->field = field;
259 return p;
260 }
261
262 static void do_append_grep_pat(struct grep_pat ***tail, struct grep_pat *p)
263 {
264 **tail = p;
265 *tail = &p->next;
266 p->next = NULL;
267
268 switch (p->token) {
269 case GREP_PATTERN: /* atom */
270 case GREP_PATTERN_HEAD:
271 case GREP_PATTERN_BODY:
272 for (;;) {
273 struct grep_pat *new_pat;
274 size_t len = 0;
275 char *cp = p->pattern + p->patternlen, *nl = NULL;
276 while (++len <= p->patternlen) {
277 if (*(--cp) == '\n') {
278 nl = cp;
279 break;
280 }
281 }
282 if (!nl)
283 break;
284 new_pat = create_grep_pat(nl + 1, len - 1, p->origin,
285 p->no, p->token, p->field);
286 new_pat->next = p->next;
287 if (!p->next)
288 *tail = &new_pat->next;
289 p->next = new_pat;
290 *nl = '\0';
291 p->patternlen -= len;
292 }
293 break;
294 default:
295 break;
296 }
297 }
298
299 void append_header_grep_pattern(struct grep_opt *opt,
300 enum grep_header_field field, const char *pat)
301 {
302 struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0,
303 GREP_PATTERN_HEAD, field);
304 if (field == GREP_HEADER_REFLOG)
305 opt->use_reflog_filter = 1;
306 do_append_grep_pat(&opt->header_tail, p);
307 }
308
309 void append_grep_pattern(struct grep_opt *opt, const char *pat,
310 const char *origin, int no, enum grep_pat_token t)
311 {
312 append_grep_pat(opt, pat, strlen(pat), origin, no, t);
313 }
314
315 void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen,
316 const char *origin, int no, enum grep_pat_token t)
317 {
318 struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0);
319 do_append_grep_pat(&opt->pattern_tail, p);
320 }
321
322 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
323 {
324 struct grep_pat *pat;
325 struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
326 *ret = *opt;
327
328 ret->pattern_list = NULL;
329 ret->pattern_tail = &ret->pattern_list;
330
331 for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
332 {
333 if(pat->token == GREP_PATTERN_HEAD)
334 append_header_grep_pattern(ret, pat->field,
335 pat->pattern);
336 else
337 append_grep_pat(ret, pat->pattern, pat->patternlen,
338 pat->origin, pat->no, pat->token);
339 }
340
341 return ret;
342 }
343
344 static NORETURN void compile_regexp_failed(const struct grep_pat *p,
345 const char *error)
346 {
347 char where[1024];
348
349 if (p->no)
350 xsnprintf(where, sizeof(where), "In '%s' at %d, ", p->origin, p->no);
351 else if (p->origin)
352 xsnprintf(where, sizeof(where), "%s, ", p->origin);
353 else
354 where[0] = 0;
355
356 die("%s'%s': %s", where, p->pattern, error);
357 }
358
359 static int is_fixed(const char *s, size_t len)
360 {
361 size_t i;
362
363 for (i = 0; i < len; i++) {
364 if (is_regex_special(s[i]))
365 return 0;
366 }
367
368 return 1;
369 }
370
371 #ifdef USE_LIBPCRE1
372 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
373 {
374 const char *error;
375 int erroffset;
376 int options = PCRE_MULTILINE;
377
378 if (opt->ignore_case) {
379 if (!opt->ignore_locale && has_non_ascii(p->pattern))
380 p->pcre1_tables = pcre_maketables();
381 options |= PCRE_CASELESS;
382 }
383 if (!opt->ignore_locale && is_utf8_locale() && has_non_ascii(p->pattern))
384 options |= PCRE_UTF8;
385
386 p->pcre1_regexp = pcre_compile(p->pattern, options, &error, &erroffset,
387 p->pcre1_tables);
388 if (!p->pcre1_regexp)
389 compile_regexp_failed(p, error);
390
391 p->pcre1_extra_info = pcre_study(p->pcre1_regexp, GIT_PCRE_STUDY_JIT_COMPILE, &error);
392 if (!p->pcre1_extra_info && error)
393 die("%s", error);
394
395 #ifdef GIT_PCRE1_USE_JIT
396 pcre_config(PCRE_CONFIG_JIT, &p->pcre1_jit_on);
397 #endif
398 }
399
400 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
401 regmatch_t *match, int eflags)
402 {
403 int ovector[30], ret, flags = 0;
404
405 if (eflags & REG_NOTBOL)
406 flags |= PCRE_NOTBOL;
407
408 ret = pcre_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
409 eol - line, 0, flags, ovector,
410 ARRAY_SIZE(ovector));
411
412 if (ret < 0 && ret != PCRE_ERROR_NOMATCH)
413 die("pcre_exec failed with error code %d", ret);
414 if (ret > 0) {
415 ret = 0;
416 match->rm_so = ovector[0];
417 match->rm_eo = ovector[1];
418 }
419
420 return ret;
421 }
422
423 static void free_pcre1_regexp(struct grep_pat *p)
424 {
425 pcre_free(p->pcre1_regexp);
426 #ifdef GIT_PCRE1_USE_JIT
427 if (p->pcre1_jit_on)
428 pcre_free_study(p->pcre1_extra_info);
429 else
430 #endif
431 pcre_free(p->pcre1_extra_info);
432 pcre_free((void *)p->pcre1_tables);
433 }
434 #else /* !USE_LIBPCRE1 */
435 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
436 {
437 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
438 }
439
440 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
441 regmatch_t *match, int eflags)
442 {
443 return 1;
444 }
445
446 static void free_pcre1_regexp(struct grep_pat *p)
447 {
448 }
449 #endif /* !USE_LIBPCRE1 */
450
451 #ifdef USE_LIBPCRE2
452 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
453 {
454 int error;
455 PCRE2_UCHAR errbuf[256];
456 PCRE2_SIZE erroffset;
457 int options = PCRE2_MULTILINE;
458 const uint8_t *character_tables = NULL;
459 int jitret;
460 int patinforet;
461 size_t jitsizearg;
462
463 assert(opt->pcre2);
464
465 p->pcre2_compile_context = NULL;
466
467 if (opt->ignore_case) {
468 if (!opt->ignore_locale && has_non_ascii(p->pattern)) {
469 character_tables = pcre2_maketables(NULL);
470 p->pcre2_compile_context = pcre2_compile_context_create(NULL);
471 pcre2_set_character_tables(p->pcre2_compile_context, character_tables);
472 }
473 options |= PCRE2_CASELESS;
474 }
475 if (!opt->ignore_locale && is_utf8_locale() && has_non_ascii(p->pattern))
476 options |= PCRE2_UTF;
477
478 p->pcre2_pattern = pcre2_compile((PCRE2_SPTR)p->pattern,
479 p->patternlen, options, &error, &erroffset,
480 p->pcre2_compile_context);
481
482 if (p->pcre2_pattern) {
483 p->pcre2_match_data = pcre2_match_data_create_from_pattern(p->pcre2_pattern, NULL);
484 if (!p->pcre2_match_data)
485 die("Couldn't allocate PCRE2 match data");
486 } else {
487 pcre2_get_error_message(error, errbuf, sizeof(errbuf));
488 compile_regexp_failed(p, (const char *)&errbuf);
489 }
490
491 pcre2_config(PCRE2_CONFIG_JIT, &p->pcre2_jit_on);
492 if (p->pcre2_jit_on) {
493 jitret = pcre2_jit_compile(p->pcre2_pattern, PCRE2_JIT_COMPLETE);
494 if (jitret)
495 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p->pattern, jitret);
496
497 /*
498 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
499 * tells us whether the library itself supports JIT,
500 * but to see whether we're going to be actually using
501 * JIT we need to extract PCRE2_INFO_JITSIZE from the
502 * pattern *after* we do pcre2_jit_compile() above.
503 *
504 * This is because if the pattern contains the
505 * (*NO_JIT) verb (see pcre2syntax(3))
506 * pcre2_jit_compile() will exit early with 0. If we
507 * then proceed to call pcre2_jit_match() further down
508 * the line instead of pcre2_match() we'll either
509 * segfault (pre PCRE 10.31) or run into a fatal error
510 * (post PCRE2 10.31)
511 */
512 patinforet = pcre2_pattern_info(p->pcre2_pattern, PCRE2_INFO_JITSIZE, &jitsizearg);
513 if (patinforet)
514 BUG("pcre2_pattern_info() failed: %d", patinforet);
515 if (jitsizearg == 0) {
516 p->pcre2_jit_on = 0;
517 return;
518 }
519 }
520 }
521
522 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
523 regmatch_t *match, int eflags)
524 {
525 int ret, flags = 0;
526 PCRE2_SIZE *ovector;
527 PCRE2_UCHAR errbuf[256];
528
529 if (eflags & REG_NOTBOL)
530 flags |= PCRE2_NOTBOL;
531
532 if (p->pcre2_jit_on)
533 ret = pcre2_jit_match(p->pcre2_pattern, (unsigned char *)line,
534 eol - line, 0, flags, p->pcre2_match_data,
535 NULL);
536 else
537 ret = pcre2_match(p->pcre2_pattern, (unsigned char *)line,
538 eol - line, 0, flags, p->pcre2_match_data,
539 NULL);
540
541 if (ret < 0 && ret != PCRE2_ERROR_NOMATCH) {
542 pcre2_get_error_message(ret, errbuf, sizeof(errbuf));
543 die("%s failed with error code %d: %s",
544 (p->pcre2_jit_on ? "pcre2_jit_match" : "pcre2_match"), ret,
545 errbuf);
546 }
547 if (ret > 0) {
548 ovector = pcre2_get_ovector_pointer(p->pcre2_match_data);
549 ret = 0;
550 match->rm_so = (int)ovector[0];
551 match->rm_eo = (int)ovector[1];
552 }
553
554 return ret;
555 }
556
557 static void free_pcre2_pattern(struct grep_pat *p)
558 {
559 pcre2_compile_context_free(p->pcre2_compile_context);
560 pcre2_code_free(p->pcre2_pattern);
561 pcre2_match_data_free(p->pcre2_match_data);
562 }
563 #else /* !USE_LIBPCRE2 */
564 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
565 {
566 /*
567 * Unreachable until USE_LIBPCRE2 becomes synonymous with
568 * USE_LIBPCRE. See the sibling comment in
569 * grep_set_pattern_type_option().
570 */
571 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
572 }
573
574 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
575 regmatch_t *match, int eflags)
576 {
577 return 1;
578 }
579
580 static void free_pcre2_pattern(struct grep_pat *p)
581 {
582 }
583
584 static void compile_fixed_regexp(struct grep_pat *p, struct grep_opt *opt)
585 {
586 struct strbuf sb = STRBUF_INIT;
587 int err;
588 int regflags = 0;
589
590 basic_regex_quote_buf(&sb, p->pattern);
591 if (opt->ignore_case)
592 regflags |= REG_ICASE;
593 err = regcomp(&p->regexp, sb.buf, regflags);
594 if (opt->debug)
595 fprintf(stderr, "fixed %s\n", sb.buf);
596 strbuf_release(&sb);
597 if (err) {
598 char errbuf[1024];
599 regerror(err, &p->regexp, errbuf, sizeof(errbuf));
600 compile_regexp_failed(p, errbuf);
601 }
602 }
603 #endif /* !USE_LIBPCRE2 */
604
605 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
606 {
607 int err;
608 int regflags = REG_NEWLINE;
609
610 p->word_regexp = opt->word_regexp;
611 p->ignore_case = opt->ignore_case;
612 p->fixed = opt->fixed;
613
614 if (memchr(p->pattern, 0, p->patternlen) && !opt->pcre2)
615 die(_("given pattern contains NULL byte (via -f <file>). This is only supported with -P under PCRE v2"));
616
617 p->is_fixed = is_fixed(p->pattern, p->patternlen);
618 #ifdef USE_LIBPCRE2
619 if (!p->fixed && !p->is_fixed) {
620 const char *no_jit = "(*NO_JIT)";
621 const int no_jit_len = strlen(no_jit);
622 if (starts_with(p->pattern, no_jit) &&
623 is_fixed(p->pattern + no_jit_len,
624 p->patternlen - no_jit_len))
625 p->is_fixed = 1;
626 }
627 #endif
628 if (p->fixed || p->is_fixed) {
629 #ifdef USE_LIBPCRE2
630 opt->pcre2 = 1;
631 if (p->is_fixed) {
632 compile_pcre2_pattern(p, opt);
633 } else {
634 /*
635 * E.g. t7811-grep-open.sh relies on the
636 * pattern being restored.
637 */
638 char *old_pattern = p->pattern;
639 size_t old_patternlen = p->patternlen;
640 struct strbuf sb = STRBUF_INIT;
641
642 /*
643 * There is the PCRE2_LITERAL flag, but it's
644 * only in PCRE v2 10.30 and later. Needing to
645 * ifdef our way around that and dealing with
646 * it + PCRE2_MULTILINE being an error is more
647 * complex than just quoting this ourselves.
648 */
649 strbuf_add(&sb, "\\Q", 2);
650 strbuf_add(&sb, p->pattern, p->patternlen);
651 strbuf_add(&sb, "\\E", 2);
652
653 p->pattern = sb.buf;
654 p->patternlen = sb.len;
655 compile_pcre2_pattern(p, opt);
656 p->pattern = old_pattern;
657 p->patternlen = old_patternlen;
658 strbuf_release(&sb);
659 }
660 #else /* !USE_LIBPCRE2 */
661 compile_fixed_regexp(p, opt);
662 #endif /* !USE_LIBPCRE2 */
663 return;
664 }
665
666 if (opt->pcre2) {
667 compile_pcre2_pattern(p, opt);
668 return;
669 }
670
671 if (opt->pcre1) {
672 compile_pcre1_regexp(p, opt);
673 return;
674 }
675
676 if (p->ignore_case)
677 regflags |= REG_ICASE;
678 if (opt->extended_regexp_option)
679 regflags |= REG_EXTENDED;
680 err = regcomp(&p->regexp, p->pattern, regflags);
681 if (err) {
682 char errbuf[1024];
683 regerror(err, &p->regexp, errbuf, 1024);
684 compile_regexp_failed(p, errbuf);
685 }
686 }
687
688 static struct grep_expr *compile_pattern_or(struct grep_pat **);
689 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
690 {
691 struct grep_pat *p;
692 struct grep_expr *x;
693
694 p = *list;
695 if (!p)
696 return NULL;
697 switch (p->token) {
698 case GREP_PATTERN: /* atom */
699 case GREP_PATTERN_HEAD:
700 case GREP_PATTERN_BODY:
701 x = xcalloc(1, sizeof (struct grep_expr));
702 x->node = GREP_NODE_ATOM;
703 x->u.atom = p;
704 *list = p->next;
705 return x;
706 case GREP_OPEN_PAREN:
707 *list = p->next;
708 x = compile_pattern_or(list);
709 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
710 die("unmatched parenthesis");
711 *list = (*list)->next;
712 return x;
713 default:
714 return NULL;
715 }
716 }
717
718 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
719 {
720 struct grep_pat *p;
721 struct grep_expr *x;
722
723 p = *list;
724 if (!p)
725 return NULL;
726 switch (p->token) {
727 case GREP_NOT:
728 if (!p->next)
729 die("--not not followed by pattern expression");
730 *list = p->next;
731 x = xcalloc(1, sizeof (struct grep_expr));
732 x->node = GREP_NODE_NOT;
733 x->u.unary = compile_pattern_not(list);
734 if (!x->u.unary)
735 die("--not followed by non pattern expression");
736 return x;
737 default:
738 return compile_pattern_atom(list);
739 }
740 }
741
742 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
743 {
744 struct grep_pat *p;
745 struct grep_expr *x, *y, *z;
746
747 x = compile_pattern_not(list);
748 p = *list;
749 if (p && p->token == GREP_AND) {
750 if (!p->next)
751 die("--and not followed by pattern expression");
752 *list = p->next;
753 y = compile_pattern_and(list);
754 if (!y)
755 die("--and not followed by pattern expression");
756 z = xcalloc(1, sizeof (struct grep_expr));
757 z->node = GREP_NODE_AND;
758 z->u.binary.left = x;
759 z->u.binary.right = y;
760 return z;
761 }
762 return x;
763 }
764
765 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
766 {
767 struct grep_pat *p;
768 struct grep_expr *x, *y, *z;
769
770 x = compile_pattern_and(list);
771 p = *list;
772 if (x && p && p->token != GREP_CLOSE_PAREN) {
773 y = compile_pattern_or(list);
774 if (!y)
775 die("not a pattern expression %s", p->pattern);
776 z = xcalloc(1, sizeof (struct grep_expr));
777 z->node = GREP_NODE_OR;
778 z->u.binary.left = x;
779 z->u.binary.right = y;
780 return z;
781 }
782 return x;
783 }
784
785 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
786 {
787 return compile_pattern_or(list);
788 }
789
790 static void indent(int in)
791 {
792 while (in-- > 0)
793 fputc(' ', stderr);
794 }
795
796 static void dump_grep_pat(struct grep_pat *p)
797 {
798 switch (p->token) {
799 case GREP_AND: fprintf(stderr, "*and*"); break;
800 case GREP_OPEN_PAREN: fprintf(stderr, "*(*"); break;
801 case GREP_CLOSE_PAREN: fprintf(stderr, "*)*"); break;
802 case GREP_NOT: fprintf(stderr, "*not*"); break;
803 case GREP_OR: fprintf(stderr, "*or*"); break;
804
805 case GREP_PATTERN: fprintf(stderr, "pattern"); break;
806 case GREP_PATTERN_HEAD: fprintf(stderr, "pattern_head"); break;
807 case GREP_PATTERN_BODY: fprintf(stderr, "pattern_body"); break;
808 }
809
810 switch (p->token) {
811 default: break;
812 case GREP_PATTERN_HEAD:
813 fprintf(stderr, "<head %d>", p->field); break;
814 case GREP_PATTERN_BODY:
815 fprintf(stderr, "<body>"); break;
816 }
817 switch (p->token) {
818 default: break;
819 case GREP_PATTERN_HEAD:
820 case GREP_PATTERN_BODY:
821 case GREP_PATTERN:
822 fprintf(stderr, "%.*s", (int)p->patternlen, p->pattern);
823 break;
824 }
825 fputc('\n', stderr);
826 }
827
828 static void dump_grep_expression_1(struct grep_expr *x, int in)
829 {
830 indent(in);
831 switch (x->node) {
832 case GREP_NODE_TRUE:
833 fprintf(stderr, "true\n");
834 break;
835 case GREP_NODE_ATOM:
836 dump_grep_pat(x->u.atom);
837 break;
838 case GREP_NODE_NOT:
839 fprintf(stderr, "(not\n");
840 dump_grep_expression_1(x->u.unary, in+1);
841 indent(in);
842 fprintf(stderr, ")\n");
843 break;
844 case GREP_NODE_AND:
845 fprintf(stderr, "(and\n");
846 dump_grep_expression_1(x->u.binary.left, in+1);
847 dump_grep_expression_1(x->u.binary.right, in+1);
848 indent(in);
849 fprintf(stderr, ")\n");
850 break;
851 case GREP_NODE_OR:
852 fprintf(stderr, "(or\n");
853 dump_grep_expression_1(x->u.binary.left, in+1);
854 dump_grep_expression_1(x->u.binary.right, in+1);
855 indent(in);
856 fprintf(stderr, ")\n");
857 break;
858 }
859 }
860
861 static void dump_grep_expression(struct grep_opt *opt)
862 {
863 struct grep_expr *x = opt->pattern_expression;
864
865 if (opt->all_match)
866 fprintf(stderr, "[all-match]\n");
867 dump_grep_expression_1(x, 0);
868 fflush(NULL);
869 }
870
871 static struct grep_expr *grep_true_expr(void)
872 {
873 struct grep_expr *z = xcalloc(1, sizeof(*z));
874 z->node = GREP_NODE_TRUE;
875 return z;
876 }
877
878 static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right)
879 {
880 struct grep_expr *z = xcalloc(1, sizeof(*z));
881 z->node = GREP_NODE_OR;
882 z->u.binary.left = left;
883 z->u.binary.right = right;
884 return z;
885 }
886
887 static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
888 {
889 struct grep_pat *p;
890 struct grep_expr *header_expr;
891 struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]);
892 enum grep_header_field fld;
893
894 if (!opt->header_list)
895 return NULL;
896
897 for (p = opt->header_list; p; p = p->next) {
898 if (p->token != GREP_PATTERN_HEAD)
899 BUG("a non-header pattern in grep header list.");
900 if (p->field < GREP_HEADER_FIELD_MIN ||
901 GREP_HEADER_FIELD_MAX <= p->field)
902 BUG("unknown header field %d", p->field);
903 compile_regexp(p, opt);
904 }
905
906 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++)
907 header_group[fld] = NULL;
908
909 for (p = opt->header_list; p; p = p->next) {
910 struct grep_expr *h;
911 struct grep_pat *pp = p;
912
913 h = compile_pattern_atom(&pp);
914 if (!h || pp != p->next)
915 BUG("malformed header expr");
916 if (!header_group[p->field]) {
917 header_group[p->field] = h;
918 continue;
919 }
920 header_group[p->field] = grep_or_expr(h, header_group[p->field]);
921 }
922
923 header_expr = NULL;
924
925 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
926 if (!header_group[fld])
927 continue;
928 if (!header_expr)
929 header_expr = grep_true_expr();
930 header_expr = grep_or_expr(header_group[fld], header_expr);
931 }
932 return header_expr;
933 }
934
935 static struct grep_expr *grep_splice_or(struct grep_expr *x, struct grep_expr *y)
936 {
937 struct grep_expr *z = x;
938
939 while (x) {
940 assert(x->node == GREP_NODE_OR);
941 if (x->u.binary.right &&
942 x->u.binary.right->node == GREP_NODE_TRUE) {
943 x->u.binary.right = y;
944 break;
945 }
946 x = x->u.binary.right;
947 }
948 return z;
949 }
950
951 static void compile_grep_patterns_real(struct grep_opt *opt)
952 {
953 struct grep_pat *p;
954 struct grep_expr *header_expr = prep_header_patterns(opt);
955
956 for (p = opt->pattern_list; p; p = p->next) {
957 switch (p->token) {
958 case GREP_PATTERN: /* atom */
959 case GREP_PATTERN_HEAD:
960 case GREP_PATTERN_BODY:
961 compile_regexp(p, opt);
962 break;
963 default:
964 opt->extended = 1;
965 break;
966 }
967 }
968
969 if (opt->all_match || header_expr)
970 opt->extended = 1;
971 else if (!opt->extended && !opt->debug)
972 return;
973
974 p = opt->pattern_list;
975 if (p)
976 opt->pattern_expression = compile_pattern_expr(&p);
977 if (p)
978 die("incomplete pattern expression: %s", p->pattern);
979
980 if (!header_expr)
981 return;
982
983 if (!opt->pattern_expression)
984 opt->pattern_expression = header_expr;
985 else if (opt->all_match)
986 opt->pattern_expression = grep_splice_or(header_expr,
987 opt->pattern_expression);
988 else
989 opt->pattern_expression = grep_or_expr(opt->pattern_expression,
990 header_expr);
991 opt->all_match = 1;
992 }
993
994 void compile_grep_patterns(struct grep_opt *opt)
995 {
996 compile_grep_patterns_real(opt);
997 if (opt->debug)
998 dump_grep_expression(opt);
999 }
1000
1001 static void free_pattern_expr(struct grep_expr *x)
1002 {
1003 switch (x->node) {
1004 case GREP_NODE_TRUE:
1005 case GREP_NODE_ATOM:
1006 break;
1007 case GREP_NODE_NOT:
1008 free_pattern_expr(x->u.unary);
1009 break;
1010 case GREP_NODE_AND:
1011 case GREP_NODE_OR:
1012 free_pattern_expr(x->u.binary.left);
1013 free_pattern_expr(x->u.binary.right);
1014 break;
1015 }
1016 free(x);
1017 }
1018
1019 void free_grep_patterns(struct grep_opt *opt)
1020 {
1021 struct grep_pat *p, *n;
1022
1023 for (p = opt->pattern_list; p; p = n) {
1024 n = p->next;
1025 switch (p->token) {
1026 case GREP_PATTERN: /* atom */
1027 case GREP_PATTERN_HEAD:
1028 case GREP_PATTERN_BODY:
1029 if (p->pcre1_regexp)
1030 free_pcre1_regexp(p);
1031 else if (p->pcre2_pattern)
1032 free_pcre2_pattern(p);
1033 else
1034 regfree(&p->regexp);
1035 free(p->pattern);
1036 break;
1037 default:
1038 break;
1039 }
1040 free(p);
1041 }
1042
1043 if (!opt->extended)
1044 return;
1045 free_pattern_expr(opt->pattern_expression);
1046 }
1047
1048 static char *end_of_line(char *cp, unsigned long *left)
1049 {
1050 unsigned long l = *left;
1051 while (l && *cp != '\n') {
1052 l--;
1053 cp++;
1054 }
1055 *left = l;
1056 return cp;
1057 }
1058
1059 static int word_char(char ch)
1060 {
1061 return isalnum(ch) || ch == '_';
1062 }
1063
1064 static void output_color(struct grep_opt *opt, const void *data, size_t size,
1065 const char *color)
1066 {
1067 if (want_color(opt->color) && color && color[0]) {
1068 opt->output(opt, color, strlen(color));
1069 opt->output(opt, data, size);
1070 opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
1071 } else
1072 opt->output(opt, data, size);
1073 }
1074
1075 static void output_sep(struct grep_opt *opt, char sign)
1076 {
1077 if (opt->null_following_name)
1078 opt->output(opt, "\0", 1);
1079 else
1080 output_color(opt, &sign, 1, opt->colors[GREP_COLOR_SEP]);
1081 }
1082
1083 static void show_name(struct grep_opt *opt, const char *name)
1084 {
1085 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1086 opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
1087 }
1088
1089 static int patmatch(struct grep_pat *p, char *line, char *eol,
1090 regmatch_t *match, int eflags)
1091 {
1092 int hit;
1093
1094 if (p->pcre1_regexp)
1095 hit = !pcre1match(p, line, eol, match, eflags);
1096 else if (p->pcre2_pattern)
1097 hit = !pcre2match(p, line, eol, match, eflags);
1098 else
1099 hit = !regexec_buf(&p->regexp, line, eol - line, 1, match,
1100 eflags);
1101
1102 return hit;
1103 }
1104
1105 static int strip_timestamp(char *bol, char **eol_p)
1106 {
1107 char *eol = *eol_p;
1108 int ch;
1109
1110 while (bol < --eol) {
1111 if (*eol != '>')
1112 continue;
1113 *eol_p = ++eol;
1114 ch = *eol;
1115 *eol = '\0';
1116 return ch;
1117 }
1118 return 0;
1119 }
1120
1121 static struct {
1122 const char *field;
1123 size_t len;
1124 } header_field[] = {
1125 { "author ", 7 },
1126 { "committer ", 10 },
1127 { "reflog ", 7 },
1128 };
1129
1130 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
1131 enum grep_context ctx,
1132 regmatch_t *pmatch, int eflags)
1133 {
1134 int hit = 0;
1135 int saved_ch = 0;
1136 const char *start = bol;
1137
1138 if ((p->token != GREP_PATTERN) &&
1139 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
1140 return 0;
1141
1142 if (p->token == GREP_PATTERN_HEAD) {
1143 const char *field;
1144 size_t len;
1145 assert(p->field < ARRAY_SIZE(header_field));
1146 field = header_field[p->field].field;
1147 len = header_field[p->field].len;
1148 if (strncmp(bol, field, len))
1149 return 0;
1150 bol += len;
1151 switch (p->field) {
1152 case GREP_HEADER_AUTHOR:
1153 case GREP_HEADER_COMMITTER:
1154 saved_ch = strip_timestamp(bol, &eol);
1155 break;
1156 default:
1157 break;
1158 }
1159 }
1160
1161 again:
1162 hit = patmatch(p, bol, eol, pmatch, eflags);
1163
1164 if (hit && p->word_regexp) {
1165 if ((pmatch[0].rm_so < 0) ||
1166 (eol - bol) < pmatch[0].rm_so ||
1167 (pmatch[0].rm_eo < 0) ||
1168 (eol - bol) < pmatch[0].rm_eo)
1169 die("regexp returned nonsense");
1170
1171 /* Match beginning must be either beginning of the
1172 * line, or at word boundary (i.e. the last char must
1173 * not be a word char). Similarly, match end must be
1174 * either end of the line, or at word boundary
1175 * (i.e. the next char must not be a word char).
1176 */
1177 if ( ((pmatch[0].rm_so == 0) ||
1178 !word_char(bol[pmatch[0].rm_so-1])) &&
1179 ((pmatch[0].rm_eo == (eol-bol)) ||
1180 !word_char(bol[pmatch[0].rm_eo])) )
1181 ;
1182 else
1183 hit = 0;
1184
1185 /* Words consist of at least one character. */
1186 if (pmatch->rm_so == pmatch->rm_eo)
1187 hit = 0;
1188
1189 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
1190 /* There could be more than one match on the
1191 * line, and the first match might not be
1192 * strict word match. But later ones could be!
1193 * Forward to the next possible start, i.e. the
1194 * next position following a non-word char.
1195 */
1196 bol = pmatch[0].rm_so + bol + 1;
1197 while (word_char(bol[-1]) && bol < eol)
1198 bol++;
1199 eflags |= REG_NOTBOL;
1200 if (bol < eol)
1201 goto again;
1202 }
1203 }
1204 if (p->token == GREP_PATTERN_HEAD && saved_ch)
1205 *eol = saved_ch;
1206 if (hit) {
1207 pmatch[0].rm_so += bol - start;
1208 pmatch[0].rm_eo += bol - start;
1209 }
1210 return hit;
1211 }
1212
1213 static int match_expr_eval(struct grep_opt *opt, struct grep_expr *x, char *bol,
1214 char *eol, enum grep_context ctx, ssize_t *col,
1215 ssize_t *icol, int collect_hits)
1216 {
1217 int h = 0;
1218
1219 if (!x)
1220 die("Not a valid grep expression");
1221 switch (x->node) {
1222 case GREP_NODE_TRUE:
1223 h = 1;
1224 break;
1225 case GREP_NODE_ATOM:
1226 {
1227 regmatch_t tmp;
1228 h = match_one_pattern(x->u.atom, bol, eol, ctx,
1229 &tmp, 0);
1230 if (h && (*col < 0 || tmp.rm_so < *col))
1231 *col = tmp.rm_so;
1232 }
1233 break;
1234 case GREP_NODE_NOT:
1235 /*
1236 * Upon visiting a GREP_NODE_NOT, col and icol become swapped.
1237 */
1238 h = !match_expr_eval(opt, x->u.unary, bol, eol, ctx, icol, col,
1239 0);
1240 break;
1241 case GREP_NODE_AND:
1242 h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col,
1243 icol, 0);
1244 if (h || opt->columnnum) {
1245 /*
1246 * Don't short-circuit AND when given --column, since a
1247 * NOT earlier in the tree may turn this into an OR. In
1248 * this case, see the below comment.
1249 */
1250 h &= match_expr_eval(opt, x->u.binary.right, bol, eol,
1251 ctx, col, icol, 0);
1252 }
1253 break;
1254 case GREP_NODE_OR:
1255 if (!(collect_hits || opt->columnnum)) {
1256 /*
1257 * Don't short-circuit OR when given --column (or
1258 * collecting hits) to ensure we don't skip a later
1259 * child that would produce an earlier match.
1260 */
1261 return (match_expr_eval(opt, x->u.binary.left, bol, eol,
1262 ctx, col, icol, 0) ||
1263 match_expr_eval(opt, x->u.binary.right, bol,
1264 eol, ctx, col, icol, 0));
1265 }
1266 h = match_expr_eval(opt, x->u.binary.left, bol, eol, ctx, col,
1267 icol, 0);
1268 if (collect_hits)
1269 x->u.binary.left->hit |= h;
1270 h |= match_expr_eval(opt, x->u.binary.right, bol, eol, ctx, col,
1271 icol, collect_hits);
1272 break;
1273 default:
1274 die("Unexpected node type (internal error) %d", x->node);
1275 }
1276 if (collect_hits)
1277 x->hit |= h;
1278 return h;
1279 }
1280
1281 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
1282 enum grep_context ctx, ssize_t *col,
1283 ssize_t *icol, int collect_hits)
1284 {
1285 struct grep_expr *x = opt->pattern_expression;
1286 return match_expr_eval(opt, x, bol, eol, ctx, col, icol, collect_hits);
1287 }
1288
1289 static int match_line(struct grep_opt *opt, char *bol, char *eol,
1290 ssize_t *col, ssize_t *icol,
1291 enum grep_context ctx, int collect_hits)
1292 {
1293 struct grep_pat *p;
1294 int hit = 0;
1295
1296 if (opt->extended)
1297 return match_expr(opt, bol, eol, ctx, col, icol,
1298 collect_hits);
1299
1300 /* we do not call with collect_hits without being extended */
1301 for (p = opt->pattern_list; p; p = p->next) {
1302 regmatch_t tmp;
1303 if (match_one_pattern(p, bol, eol, ctx, &tmp, 0)) {
1304 hit |= 1;
1305 if (!opt->columnnum) {
1306 /*
1307 * Without --column, any single match on a line
1308 * is enough to know that it needs to be
1309 * printed. With --column, scan _all_ patterns
1310 * to find the earliest.
1311 */
1312 break;
1313 }
1314 if (*col < 0 || tmp.rm_so < *col)
1315 *col = tmp.rm_so;
1316 }
1317 }
1318 return hit;
1319 }
1320
1321 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
1322 enum grep_context ctx,
1323 regmatch_t *pmatch, int eflags)
1324 {
1325 regmatch_t match;
1326
1327 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
1328 return 0;
1329 if (match.rm_so < 0 || match.rm_eo < 0)
1330 return 0;
1331 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
1332 if (match.rm_so > pmatch->rm_so)
1333 return 1;
1334 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
1335 return 1;
1336 }
1337 pmatch->rm_so = match.rm_so;
1338 pmatch->rm_eo = match.rm_eo;
1339 return 1;
1340 }
1341
1342 static int next_match(struct grep_opt *opt, char *bol, char *eol,
1343 enum grep_context ctx, regmatch_t *pmatch, int eflags)
1344 {
1345 struct grep_pat *p;
1346 int hit = 0;
1347
1348 pmatch->rm_so = pmatch->rm_eo = -1;
1349 if (bol < eol) {
1350 for (p = opt->pattern_list; p; p = p->next) {
1351 switch (p->token) {
1352 case GREP_PATTERN: /* atom */
1353 case GREP_PATTERN_HEAD:
1354 case GREP_PATTERN_BODY:
1355 hit |= match_next_pattern(p, bol, eol, ctx,
1356 pmatch, eflags);
1357 break;
1358 default:
1359 break;
1360 }
1361 }
1362 }
1363 return hit;
1364 }
1365
1366 static void show_line_header(struct grep_opt *opt, const char *name,
1367 unsigned lno, ssize_t cno, char sign)
1368 {
1369 if (opt->heading && opt->last_shown == 0) {
1370 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1371 opt->output(opt, "\n", 1);
1372 }
1373 opt->last_shown = lno;
1374
1375 if (!opt->heading && opt->pathname) {
1376 output_color(opt, name, strlen(name), opt->colors[GREP_COLOR_FILENAME]);
1377 output_sep(opt, sign);
1378 }
1379 if (opt->linenum) {
1380 char buf[32];
1381 xsnprintf(buf, sizeof(buf), "%d", lno);
1382 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_LINENO]);
1383 output_sep(opt, sign);
1384 }
1385 /*
1386 * Treat 'cno' as the 1-indexed offset from the start of a non-context
1387 * line to its first match. Otherwise, 'cno' is 0 indicating that we are
1388 * being called with a context line.
1389 */
1390 if (opt->columnnum && cno) {
1391 char buf[32];
1392 xsnprintf(buf, sizeof(buf), "%"PRIuMAX, (uintmax_t)cno);
1393 output_color(opt, buf, strlen(buf), opt->colors[GREP_COLOR_COLUMNNO]);
1394 output_sep(opt, sign);
1395 }
1396 }
1397
1398 static void show_line(struct grep_opt *opt, char *bol, char *eol,
1399 const char *name, unsigned lno, ssize_t cno, char sign)
1400 {
1401 int rest = eol - bol;
1402 const char *match_color = NULL;
1403 const char *line_color = NULL;
1404
1405 if (opt->file_break && opt->last_shown == 0) {
1406 if (opt->show_hunk_mark)
1407 opt->output(opt, "\n", 1);
1408 } else if (opt->pre_context || opt->post_context || opt->funcbody) {
1409 if (opt->last_shown == 0) {
1410 if (opt->show_hunk_mark) {
1411 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1412 opt->output(opt, "\n", 1);
1413 }
1414 } else if (lno > opt->last_shown + 1) {
1415 output_color(opt, "--", 2, opt->colors[GREP_COLOR_SEP]);
1416 opt->output(opt, "\n", 1);
1417 }
1418 }
1419 if (!opt->only_matching) {
1420 /*
1421 * In case the line we're being called with contains more than
1422 * one match, leave printing each header to the loop below.
1423 */
1424 show_line_header(opt, name, lno, cno, sign);
1425 }
1426 if (opt->color || opt->only_matching) {
1427 regmatch_t match;
1428 enum grep_context ctx = GREP_CONTEXT_BODY;
1429 int ch = *eol;
1430 int eflags = 0;
1431
1432 if (opt->color) {
1433 if (sign == ':')
1434 match_color = opt->colors[GREP_COLOR_MATCH_SELECTED];
1435 else
1436 match_color = opt->colors[GREP_COLOR_MATCH_CONTEXT];
1437 if (sign == ':')
1438 line_color = opt->colors[GREP_COLOR_SELECTED];
1439 else if (sign == '-')
1440 line_color = opt->colors[GREP_COLOR_CONTEXT];
1441 else if (sign == '=')
1442 line_color = opt->colors[GREP_COLOR_FUNCTION];
1443 }
1444 *eol = '\0';
1445 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
1446 if (match.rm_so == match.rm_eo)
1447 break;
1448
1449 if (opt->only_matching)
1450 show_line_header(opt, name, lno, cno, sign);
1451 else
1452 output_color(opt, bol, match.rm_so, line_color);
1453 output_color(opt, bol + match.rm_so,
1454 match.rm_eo - match.rm_so, match_color);
1455 if (opt->only_matching)
1456 opt->output(opt, "\n", 1);
1457 bol += match.rm_eo;
1458 cno += match.rm_eo;
1459 rest -= match.rm_eo;
1460 eflags = REG_NOTBOL;
1461 }
1462 *eol = ch;
1463 }
1464 if (!opt->only_matching) {
1465 output_color(opt, bol, rest, line_color);
1466 opt->output(opt, "\n", 1);
1467 }
1468 }
1469
1470 int grep_use_locks;
1471
1472 /*
1473 * This lock protects access to the gitattributes machinery, which is
1474 * not thread-safe.
1475 */
1476 pthread_mutex_t grep_attr_mutex;
1477
1478 static inline void grep_attr_lock(void)
1479 {
1480 if (grep_use_locks)
1481 pthread_mutex_lock(&grep_attr_mutex);
1482 }
1483
1484 static inline void grep_attr_unlock(void)
1485 {
1486 if (grep_use_locks)
1487 pthread_mutex_unlock(&grep_attr_mutex);
1488 }
1489
1490 /*
1491 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1492 */
1493 pthread_mutex_t grep_read_mutex;
1494
1495 static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol)
1496 {
1497 xdemitconf_t *xecfg = opt->priv;
1498 if (xecfg && !xecfg->find_func) {
1499 grep_source_load_driver(gs, opt->repo->index);
1500 if (gs->driver->funcname.pattern) {
1501 const struct userdiff_funcname *pe = &gs->driver->funcname;
1502 xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
1503 } else {
1504 xecfg = opt->priv = NULL;
1505 }
1506 }
1507
1508 if (xecfg) {
1509 char buf[1];
1510 return xecfg->find_func(bol, eol - bol, buf, 1,
1511 xecfg->find_func_priv) >= 0;
1512 }
1513
1514 if (bol == eol)
1515 return 0;
1516 if (isalpha(*bol) || *bol == '_' || *bol == '$')
1517 return 1;
1518 return 0;
1519 }
1520
1521 static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs,
1522 char *bol, unsigned lno)
1523 {
1524 while (bol > gs->buf) {
1525 char *eol = --bol;
1526
1527 while (bol > gs->buf && bol[-1] != '\n')
1528 bol--;
1529 lno--;
1530
1531 if (lno <= opt->last_shown)
1532 break;
1533
1534 if (match_funcname(opt, gs, bol, eol)) {
1535 show_line(opt, bol, eol, gs->name, lno, 0, '=');
1536 break;
1537 }
1538 }
1539 }
1540
1541 static int is_empty_line(const char *bol, const char *eol);
1542
1543 static void show_pre_context(struct grep_opt *opt, struct grep_source *gs,
1544 char *bol, char *end, unsigned lno)
1545 {
1546 unsigned cur = lno, from = 1, funcname_lno = 0, orig_from;
1547 int funcname_needed = !!opt->funcname, comment_needed = 0;
1548
1549 if (opt->pre_context < lno)
1550 from = lno - opt->pre_context;
1551 if (from <= opt->last_shown)
1552 from = opt->last_shown + 1;
1553 orig_from = from;
1554 if (opt->funcbody) {
1555 if (match_funcname(opt, gs, bol, end))
1556 comment_needed = 1;
1557 else
1558 funcname_needed = 1;
1559 from = opt->last_shown + 1;
1560 }
1561
1562 /* Rewind. */
1563 while (bol > gs->buf && cur > from) {
1564 char *next_bol = bol;
1565 char *eol = --bol;
1566
1567 while (bol > gs->buf && bol[-1] != '\n')
1568 bol--;
1569 cur--;
1570 if (comment_needed && (is_empty_line(bol, eol) ||
1571 match_funcname(opt, gs, bol, eol))) {
1572 comment_needed = 0;
1573 from = orig_from;
1574 if (cur < from) {
1575 cur++;
1576 bol = next_bol;
1577 break;
1578 }
1579 }
1580 if (funcname_needed && match_funcname(opt, gs, bol, eol)) {
1581 funcname_lno = cur;
1582 funcname_needed = 0;
1583 if (opt->funcbody)
1584 comment_needed = 1;
1585 else
1586 from = orig_from;
1587 }
1588 }
1589
1590 /* We need to look even further back to find a function signature. */
1591 if (opt->funcname && funcname_needed)
1592 show_funcname_line(opt, gs, bol, cur);
1593
1594 /* Back forward. */
1595 while (cur < lno) {
1596 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
1597
1598 while (*eol != '\n')
1599 eol++;
1600 show_line(opt, bol, eol, gs->name, cur, 0, sign);
1601 bol = eol + 1;
1602 cur++;
1603 }
1604 }
1605
1606 static int should_lookahead(struct grep_opt *opt)
1607 {
1608 struct grep_pat *p;
1609
1610 if (opt->extended)
1611 return 0; /* punt for too complex stuff */
1612 if (opt->invert)
1613 return 0;
1614 for (p = opt->pattern_list; p; p = p->next) {
1615 if (p->token != GREP_PATTERN)
1616 return 0; /* punt for "header only" and stuff */
1617 }
1618 return 1;
1619 }
1620
1621 static int look_ahead(struct grep_opt *opt,
1622 unsigned long *left_p,
1623 unsigned *lno_p,
1624 char **bol_p)
1625 {
1626 unsigned lno = *lno_p;
1627 char *bol = *bol_p;
1628 struct grep_pat *p;
1629 char *sp, *last_bol;
1630 regoff_t earliest = -1;
1631
1632 for (p = opt->pattern_list; p; p = p->next) {
1633 int hit;
1634 regmatch_t m;
1635
1636 hit = patmatch(p, bol, bol + *left_p, &m, 0);
1637 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
1638 continue;
1639 if (earliest < 0 || m.rm_so < earliest)
1640 earliest = m.rm_so;
1641 }
1642
1643 if (earliest < 0) {
1644 *bol_p = bol + *left_p;
1645 *left_p = 0;
1646 return 1;
1647 }
1648 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
1649 ; /* find the beginning of the line */
1650 last_bol = sp;
1651
1652 for (sp = bol; sp < last_bol; sp++) {
1653 if (*sp == '\n')
1654 lno++;
1655 }
1656 *left_p -= last_bol - bol;
1657 *bol_p = last_bol;
1658 *lno_p = lno;
1659 return 0;
1660 }
1661
1662 static int fill_textconv_grep(struct repository *r,
1663 struct userdiff_driver *driver,
1664 struct grep_source *gs)
1665 {
1666 struct diff_filespec *df;
1667 char *buf;
1668 size_t size;
1669
1670 if (!driver || !driver->textconv)
1671 return grep_source_load(gs);
1672
1673 /*
1674 * The textconv interface is intimately tied to diff_filespecs, so we
1675 * have to pretend to be one. If we could unify the grep_source
1676 * and diff_filespec structs, this mess could just go away.
1677 */
1678 df = alloc_filespec(gs->path);
1679 switch (gs->type) {
1680 case GREP_SOURCE_OID:
1681 fill_filespec(df, gs->identifier, 1, 0100644);
1682 break;
1683 case GREP_SOURCE_FILE:
1684 fill_filespec(df, &null_oid, 0, 0100644);
1685 break;
1686 default:
1687 BUG("attempt to textconv something without a path?");
1688 }
1689
1690 /*
1691 * fill_textconv is not remotely thread-safe; it may load objects
1692 * behind the scenes, and it modifies the global diff tempfile
1693 * structure.
1694 */
1695 grep_read_lock();
1696 size = fill_textconv(r, driver, df, &buf);
1697 grep_read_unlock();
1698 free_filespec(df);
1699
1700 /*
1701 * The normal fill_textconv usage by the diff machinery would just keep
1702 * the textconv'd buf separate from the diff_filespec. But much of the
1703 * grep code passes around a grep_source and assumes that its "buf"
1704 * pointer is the beginning of the thing we are searching. So let's
1705 * install our textconv'd version into the grep_source, taking care not
1706 * to leak any existing buffer.
1707 */
1708 grep_source_clear_data(gs);
1709 gs->buf = buf;
1710 gs->size = size;
1711
1712 return 0;
1713 }
1714
1715 static int is_empty_line(const char *bol, const char *eol)
1716 {
1717 while (bol < eol && isspace(*bol))
1718 bol++;
1719 return bol == eol;
1720 }
1721
1722 static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits)
1723 {
1724 char *bol;
1725 char *peek_bol = NULL;
1726 unsigned long left;
1727 unsigned lno = 1;
1728 unsigned last_hit = 0;
1729 int binary_match_only = 0;
1730 unsigned count = 0;
1731 int try_lookahead = 0;
1732 int show_function = 0;
1733 struct userdiff_driver *textconv = NULL;
1734 enum grep_context ctx = GREP_CONTEXT_HEAD;
1735 xdemitconf_t xecfg;
1736
1737 if (!opt->status_only && gs->name == NULL)
1738 BUG("grep call which could print a name requires "
1739 "grep_source.name be non-NULL");
1740
1741 if (!opt->output)
1742 opt->output = std_output;
1743
1744 if (opt->pre_context || opt->post_context || opt->file_break ||
1745 opt->funcbody) {
1746 /* Show hunk marks, except for the first file. */
1747 if (opt->last_shown)
1748 opt->show_hunk_mark = 1;
1749 /*
1750 * If we're using threads then we can't easily identify
1751 * the first file. Always put hunk marks in that case
1752 * and skip the very first one later in work_done().
1753 */
1754 if (opt->output != std_output)
1755 opt->show_hunk_mark = 1;
1756 }
1757 opt->last_shown = 0;
1758
1759 if (opt->allow_textconv) {
1760 grep_source_load_driver(gs, opt->repo->index);
1761 /*
1762 * We might set up the shared textconv cache data here, which
1763 * is not thread-safe.
1764 */
1765 grep_attr_lock();
1766 textconv = userdiff_get_textconv(opt->repo, gs->driver);
1767 grep_attr_unlock();
1768 }
1769
1770 /*
1771 * We know the result of a textconv is text, so we only have to care
1772 * about binary handling if we are not using it.
1773 */
1774 if (!textconv) {
1775 switch (opt->binary) {
1776 case GREP_BINARY_DEFAULT:
1777 if (grep_source_is_binary(gs, opt->repo->index))
1778 binary_match_only = 1;
1779 break;
1780 case GREP_BINARY_NOMATCH:
1781 if (grep_source_is_binary(gs, opt->repo->index))
1782 return 0; /* Assume unmatch */
1783 break;
1784 case GREP_BINARY_TEXT:
1785 break;
1786 default:
1787 BUG("unknown binary handling mode");
1788 }
1789 }
1790
1791 memset(&xecfg, 0, sizeof(xecfg));
1792 opt->priv = &xecfg;
1793
1794 try_lookahead = should_lookahead(opt);
1795
1796 if (fill_textconv_grep(opt->repo, textconv, gs) < 0)
1797 return 0;
1798
1799 bol = gs->buf;
1800 left = gs->size;
1801 while (left) {
1802 char *eol, ch;
1803 int hit;
1804 ssize_t cno;
1805 ssize_t col = -1, icol = -1;
1806
1807 /*
1808 * look_ahead() skips quickly to the line that possibly
1809 * has the next hit; don't call it if we need to do
1810 * something more than just skipping the current line
1811 * in response to an unmatch for the current line. E.g.
1812 * inside a post-context window, we will show the current
1813 * line as a context around the previous hit when it
1814 * doesn't hit.
1815 */
1816 if (try_lookahead
1817 && !(last_hit
1818 && (show_function ||
1819 lno <= last_hit + opt->post_context))
1820 && look_ahead(opt, &left, &lno, &bol))
1821 break;
1822 eol = end_of_line(bol, &left);
1823 ch = *eol;
1824 *eol = 0;
1825
1826 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
1827 ctx = GREP_CONTEXT_BODY;
1828
1829 hit = match_line(opt, bol, eol, &col, &icol, ctx, collect_hits);
1830 *eol = ch;
1831
1832 if (collect_hits)
1833 goto next_line;
1834
1835 /* "grep -v -e foo -e bla" should list lines
1836 * that do not have either, so inversion should
1837 * be done outside.
1838 */
1839 if (opt->invert)
1840 hit = !hit;
1841 if (opt->unmatch_name_only) {
1842 if (hit)
1843 return 0;
1844 goto next_line;
1845 }
1846 if (hit) {
1847 count++;
1848 if (opt->status_only)
1849 return 1;
1850 if (opt->name_only) {
1851 show_name(opt, gs->name);
1852 return 1;
1853 }
1854 if (opt->count)
1855 goto next_line;
1856 if (binary_match_only) {
1857 opt->output(opt, "Binary file ", 12);
1858 output_color(opt, gs->name, strlen(gs->name),
1859 opt->colors[GREP_COLOR_FILENAME]);
1860 opt->output(opt, " matches\n", 9);
1861 return 1;
1862 }
1863 /* Hit at this line. If we haven't shown the
1864 * pre-context lines, we would need to show them.
1865 */
1866 if (opt->pre_context || opt->funcbody)
1867 show_pre_context(opt, gs, bol, eol, lno);
1868 else if (opt->funcname)
1869 show_funcname_line(opt, gs, bol, lno);
1870 cno = opt->invert ? icol : col;
1871 if (cno < 0) {
1872 /*
1873 * A negative cno indicates that there was no
1874 * match on the line. We are thus inverted and
1875 * being asked to show all lines that _don't_
1876 * match a given expression. Therefore, set cno
1877 * to 0 to suggest the whole line matches.
1878 */
1879 cno = 0;
1880 }
1881 show_line(opt, bol, eol, gs->name, lno, cno + 1, ':');
1882 last_hit = lno;
1883 if (opt->funcbody)
1884 show_function = 1;
1885 goto next_line;
1886 }
1887 if (show_function && (!peek_bol || peek_bol < bol)) {
1888 unsigned long peek_left = left;
1889 char *peek_eol = eol;
1890
1891 /*
1892 * Trailing empty lines are not interesting.
1893 * Peek past them to see if they belong to the
1894 * body of the current function.
1895 */
1896 peek_bol = bol;
1897 while (is_empty_line(peek_bol, peek_eol)) {
1898 peek_bol = peek_eol + 1;
1899 peek_eol = end_of_line(peek_bol, &peek_left);
1900 }
1901
1902 if (match_funcname(opt, gs, peek_bol, peek_eol))
1903 show_function = 0;
1904 }
1905 if (show_function ||
1906 (last_hit && lno <= last_hit + opt->post_context)) {
1907 /* If the last hit is within the post context,
1908 * we need to show this line.
1909 */
1910 show_line(opt, bol, eol, gs->name, lno, col + 1, '-');
1911 }
1912
1913 next_line:
1914 bol = eol + 1;
1915 if (!left)
1916 break;
1917 left--;
1918 lno++;
1919 }
1920
1921 if (collect_hits)
1922 return 0;
1923
1924 if (opt->status_only)
1925 return opt->unmatch_name_only;
1926 if (opt->unmatch_name_only) {
1927 /* We did not see any hit, so we want to show this */
1928 show_name(opt, gs->name);
1929 return 1;
1930 }
1931
1932 xdiff_clear_find_func(&xecfg);
1933 opt->priv = NULL;
1934
1935 /* NEEDSWORK:
1936 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1937 * which feels mostly useless but sometimes useful. Maybe
1938 * make it another option? For now suppress them.
1939 */
1940 if (opt->count && count) {
1941 char buf[32];
1942 if (opt->pathname) {
1943 output_color(opt, gs->name, strlen(gs->name),
1944 opt->colors[GREP_COLOR_FILENAME]);
1945 output_sep(opt, ':');
1946 }
1947 xsnprintf(buf, sizeof(buf), "%u\n", count);
1948 opt->output(opt, buf, strlen(buf));
1949 return 1;
1950 }
1951 return !!last_hit;
1952 }
1953
1954 static void clr_hit_marker(struct grep_expr *x)
1955 {
1956 /* All-hit markers are meaningful only at the very top level
1957 * OR node.
1958 */
1959 while (1) {
1960 x->hit = 0;
1961 if (x->node != GREP_NODE_OR)
1962 return;
1963 x->u.binary.left->hit = 0;
1964 x = x->u.binary.right;
1965 }
1966 }
1967
1968 static int chk_hit_marker(struct grep_expr *x)
1969 {
1970 /* Top level nodes have hit markers. See if they all are hits */
1971 while (1) {
1972 if (x->node != GREP_NODE_OR)
1973 return x->hit;
1974 if (!x->u.binary.left->hit)
1975 return 0;
1976 x = x->u.binary.right;
1977 }
1978 }
1979
1980 int grep_source(struct grep_opt *opt, struct grep_source *gs)
1981 {
1982 /*
1983 * we do not have to do the two-pass grep when we do not check
1984 * buffer-wide "all-match".
1985 */
1986 if (!opt->all_match)
1987 return grep_source_1(opt, gs, 0);
1988
1989 /* Otherwise the toplevel "or" terms hit a bit differently.
1990 * We first clear hit markers from them.
1991 */
1992 clr_hit_marker(opt->pattern_expression);
1993 grep_source_1(opt, gs, 1);
1994
1995 if (!chk_hit_marker(opt->pattern_expression))
1996 return 0;
1997
1998 return grep_source_1(opt, gs, 0);
1999 }
2000
2001 int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size)
2002 {
2003 struct grep_source gs;
2004 int r;
2005
2006 grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL, NULL);
2007 gs.buf = buf;
2008 gs.size = size;
2009
2010 r = grep_source(opt, &gs);
2011
2012 grep_source_clear(&gs);
2013 return r;
2014 }
2015
2016 void grep_source_init(struct grep_source *gs, enum grep_source_type type,
2017 const char *name, const char *path,
2018 const void *identifier)
2019 {
2020 gs->type = type;
2021 gs->name = xstrdup_or_null(name);
2022 gs->path = xstrdup_or_null(path);
2023 gs->buf = NULL;
2024 gs->size = 0;
2025 gs->driver = NULL;
2026
2027 switch (type) {
2028 case GREP_SOURCE_FILE:
2029 gs->identifier = xstrdup(identifier);
2030 break;
2031 case GREP_SOURCE_OID:
2032 gs->identifier = oiddup(identifier);
2033 break;
2034 case GREP_SOURCE_BUF:
2035 gs->identifier = NULL;
2036 break;
2037 }
2038 }
2039
2040 void grep_source_clear(struct grep_source *gs)
2041 {
2042 FREE_AND_NULL(gs->name);
2043 FREE_AND_NULL(gs->path);
2044 FREE_AND_NULL(gs->identifier);
2045 grep_source_clear_data(gs);
2046 }
2047
2048 void grep_source_clear_data(struct grep_source *gs)
2049 {
2050 switch (gs->type) {
2051 case GREP_SOURCE_FILE:
2052 case GREP_SOURCE_OID:
2053 FREE_AND_NULL(gs->buf);
2054 gs->size = 0;
2055 break;
2056 case GREP_SOURCE_BUF:
2057 /* leave user-provided buf intact */
2058 break;
2059 }
2060 }
2061
2062 static int grep_source_load_oid(struct grep_source *gs)
2063 {
2064 enum object_type type;
2065
2066 grep_read_lock();
2067 gs->buf = read_object_file(gs->identifier, &type, &gs->size);
2068 grep_read_unlock();
2069
2070 if (!gs->buf)
2071 return error(_("'%s': unable to read %s"),
2072 gs->name,
2073 oid_to_hex(gs->identifier));
2074 return 0;
2075 }
2076
2077 static int grep_source_load_file(struct grep_source *gs)
2078 {
2079 const char *filename = gs->identifier;
2080 struct stat st;
2081 char *data;
2082 size_t size;
2083 int i;
2084
2085 if (lstat(filename, &st) < 0) {
2086 err_ret:
2087 if (errno != ENOENT)
2088 error_errno(_("failed to stat '%s'"), filename);
2089 return -1;
2090 }
2091 if (!S_ISREG(st.st_mode))
2092 return -1;
2093 size = xsize_t(st.st_size);
2094 i = open(filename, O_RDONLY);
2095 if (i < 0)
2096 goto err_ret;
2097 data = xmallocz(size);
2098 if (st.st_size != read_in_full(i, data, size)) {
2099 error_errno(_("'%s': short read"), filename);
2100 close(i);
2101 free(data);
2102 return -1;
2103 }
2104 close(i);
2105
2106 gs->buf = data;
2107 gs->size = size;
2108 return 0;
2109 }
2110
2111 static int grep_source_load(struct grep_source *gs)
2112 {
2113 if (gs->buf)
2114 return 0;
2115
2116 switch (gs->type) {
2117 case GREP_SOURCE_FILE:
2118 return grep_source_load_file(gs);
2119 case GREP_SOURCE_OID:
2120 return grep_source_load_oid(gs);
2121 case GREP_SOURCE_BUF:
2122 return gs->buf ? 0 : -1;
2123 }
2124 BUG("invalid grep_source type to load");
2125 }
2126
2127 void grep_source_load_driver(struct grep_source *gs,
2128 struct index_state *istate)
2129 {
2130 if (gs->driver)
2131 return;
2132
2133 grep_attr_lock();
2134 if (gs->path)
2135 gs->driver = userdiff_find_by_path(istate, gs->path);
2136 if (!gs->driver)
2137 gs->driver = userdiff_find_by_name("default");
2138 grep_attr_unlock();
2139 }
2140
2141 static int grep_source_is_binary(struct grep_source *gs,
2142 struct index_state *istate)
2143 {
2144 grep_source_load_driver(gs, istate);
2145 if (gs->driver->binary != -1)
2146 return gs->driver->binary;
2147
2148 if (!grep_source_load(gs))
2149 return buffer_is_binary(gs->buf, gs->size);
2150
2151 return 0;
2152 }
Unnamed repository; edit this file 'description' to name the repository.