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