color.h: document and modernize header
[git/git.git] / grep.c
1 #include "cache.h"
2 #include "config.h"
3 #include "grep.h"
4 #include "userdiff.h"
5 #include "xdiff-interface.h"
6 #include "diff.h"
7 #include "diffcore.h"
8 #include "commit.h"
9 #include "quote.h"
10
11 static int grep_source_load(struct grep_source *gs);
12 static int grep_source_is_binary(struct grep_source *gs);
13
14 static struct grep_opt grep_defaults;
15
16 static void std_output(struct grep_opt *opt, const void *buf, size_t size)
17 {
18 fwrite(buf, size, 1, stdout);
19 }
20
21 static void color_set(char *dst, const char *color_bytes)
22 {
23 xsnprintf(dst, COLOR_MAXLEN, "%s", color_bytes);
24 }
25
26 /*
27 * Initialize the grep_defaults template with hardcoded defaults.
28 * We could let the compiler do this, but without C99 initializers
29 * the code gets unwieldy and unreadable, so...
30 */
31 void init_grep_defaults(void)
32 {
33 struct grep_opt *opt = &grep_defaults;
34 static int run_once;
35
36 if (run_once)
37 return;
38 run_once++;
39
40 memset(opt, 0, sizeof(*opt));
41 opt->relative = 1;
42 opt->pathname = 1;
43 opt->max_depth = -1;
44 opt->pattern_type_option = GREP_PATTERN_TYPE_UNSPECIFIED;
45 color_set(opt->color_context, "");
46 color_set(opt->color_filename, "");
47 color_set(opt->color_function, "");
48 color_set(opt->color_lineno, "");
49 color_set(opt->color_match_context, GIT_COLOR_BOLD_RED);
50 color_set(opt->color_match_selected, GIT_COLOR_BOLD_RED);
51 color_set(opt->color_selected, "");
52 color_set(opt->color_sep, GIT_COLOR_CYAN);
53 opt->color = -1;
54 opt->output = std_output;
55 }
56
57 static int parse_pattern_type_arg(const char *opt, const char *arg)
58 {
59 if (!strcmp(arg, "default"))
60 return GREP_PATTERN_TYPE_UNSPECIFIED;
61 else if (!strcmp(arg, "basic"))
62 return GREP_PATTERN_TYPE_BRE;
63 else if (!strcmp(arg, "extended"))
64 return GREP_PATTERN_TYPE_ERE;
65 else if (!strcmp(arg, "fixed"))
66 return GREP_PATTERN_TYPE_FIXED;
67 else if (!strcmp(arg, "perl"))
68 return GREP_PATTERN_TYPE_PCRE;
69 die("bad %s argument: %s", opt, arg);
70 }
71
72 /*
73 * Read the configuration file once and store it in
74 * the grep_defaults template.
75 */
76 int grep_config(const char *var, const char *value, void *cb)
77 {
78 struct grep_opt *opt = &grep_defaults;
79 char *color = NULL;
80
81 if (userdiff_config(var, value) < 0)
82 return -1;
83
84 if (!strcmp(var, "grep.extendedregexp")) {
85 opt->extended_regexp_option = git_config_bool(var, value);
86 return 0;
87 }
88
89 if (!strcmp(var, "grep.patterntype")) {
90 opt->pattern_type_option = parse_pattern_type_arg(var, value);
91 return 0;
92 }
93
94 if (!strcmp(var, "grep.linenumber")) {
95 opt->linenum = git_config_bool(var, value);
96 return 0;
97 }
98
99 if (!strcmp(var, "grep.fullname")) {
100 opt->relative = !git_config_bool(var, value);
101 return 0;
102 }
103
104 if (!strcmp(var, "color.grep"))
105 opt->color = git_config_colorbool(var, value);
106 else if (!strcmp(var, "color.grep.context"))
107 color = opt->color_context;
108 else if (!strcmp(var, "color.grep.filename"))
109 color = opt->color_filename;
110 else if (!strcmp(var, "color.grep.function"))
111 color = opt->color_function;
112 else if (!strcmp(var, "color.grep.linenumber"))
113 color = opt->color_lineno;
114 else if (!strcmp(var, "color.grep.matchcontext"))
115 color = opt->color_match_context;
116 else if (!strcmp(var, "color.grep.matchselected"))
117 color = opt->color_match_selected;
118 else if (!strcmp(var, "color.grep.selected"))
119 color = opt->color_selected;
120 else if (!strcmp(var, "color.grep.separator"))
121 color = opt->color_sep;
122 else if (!strcmp(var, "color.grep.match")) {
123 int rc = 0;
124 if (!value)
125 return config_error_nonbool(var);
126 rc |= color_parse(value, opt->color_match_context);
127 rc |= color_parse(value, opt->color_match_selected);
128 return rc;
129 }
130
131 if (color) {
132 if (!value)
133 return config_error_nonbool(var);
134 return color_parse(value, color);
135 }
136 return 0;
137 }
138
139 /*
140 * Initialize one instance of grep_opt and copy the
141 * default values from the template we read the configuration
142 * information in an earlier call to git_config(grep_config).
143 */
144 void grep_init(struct grep_opt *opt, const char *prefix)
145 {
146 struct grep_opt *def = &grep_defaults;
147
148 memset(opt, 0, sizeof(*opt));
149 opt->prefix = prefix;
150 opt->prefix_length = (prefix && *prefix) ? strlen(prefix) : 0;
151 opt->pattern_tail = &opt->pattern_list;
152 opt->header_tail = &opt->header_list;
153
154 opt->color = def->color;
155 opt->extended_regexp_option = def->extended_regexp_option;
156 opt->pattern_type_option = def->pattern_type_option;
157 opt->linenum = def->linenum;
158 opt->max_depth = def->max_depth;
159 opt->pathname = def->pathname;
160 opt->relative = def->relative;
161 opt->output = def->output;
162
163 color_set(opt->color_context, def->color_context);
164 color_set(opt->color_filename, def->color_filename);
165 color_set(opt->color_function, def->color_function);
166 color_set(opt->color_lineno, def->color_lineno);
167 color_set(opt->color_match_context, def->color_match_context);
168 color_set(opt->color_match_selected, def->color_match_selected);
169 color_set(opt->color_selected, def->color_selected);
170 color_set(opt->color_sep, def->color_sep);
171 }
172
173 static void grep_set_pattern_type_option(enum grep_pattern_type pattern_type, struct grep_opt *opt)
174 {
175 /*
176 * When committing to the pattern type by setting the relevant
177 * fields in grep_opt it's generally not necessary to zero out
178 * the fields we're not choosing, since they won't have been
179 * set by anything. The extended_regexp_option field is the
180 * only exception to this.
181 *
182 * This is because in the process of parsing grep.patternType
183 * & grep.extendedRegexp we set opt->pattern_type_option and
184 * opt->extended_regexp_option, respectively. We then
185 * internally use opt->extended_regexp_option to see if we're
186 * compiling an ERE. It must be unset if that's not actually
187 * the case.
188 */
189 if (pattern_type != GREP_PATTERN_TYPE_ERE &&
190 opt->extended_regexp_option)
191 opt->extended_regexp_option = 0;
192
193 switch (pattern_type) {
194 case GREP_PATTERN_TYPE_UNSPECIFIED:
195 /* fall through */
196
197 case GREP_PATTERN_TYPE_BRE:
198 break;
199
200 case GREP_PATTERN_TYPE_ERE:
201 opt->extended_regexp_option = 1;
202 break;
203
204 case GREP_PATTERN_TYPE_FIXED:
205 opt->fixed = 1;
206 break;
207
208 case GREP_PATTERN_TYPE_PCRE:
209 #ifdef USE_LIBPCRE2
210 opt->pcre2 = 1;
211 #else
212 /*
213 * It's important that pcre1 always be assigned to
214 * even when there's no USE_LIBPCRE* defined. We still
215 * call the PCRE stub function, it just dies with
216 * "cannot use Perl-compatible regexes[...]".
217 */
218 opt->pcre1 = 1;
219 #endif
220 break;
221 }
222 }
223
224 void grep_commit_pattern_type(enum grep_pattern_type pattern_type, struct grep_opt *opt)
225 {
226 if (pattern_type != GREP_PATTERN_TYPE_UNSPECIFIED)
227 grep_set_pattern_type_option(pattern_type, opt);
228 else if (opt->pattern_type_option != GREP_PATTERN_TYPE_UNSPECIFIED)
229 grep_set_pattern_type_option(opt->pattern_type_option, opt);
230 else if (opt->extended_regexp_option)
231 /*
232 * This branch *must* happen after setting from the
233 * opt->pattern_type_option above, we don't want
234 * grep.extendedRegexp to override grep.patternType!
235 */
236 grep_set_pattern_type_option(GREP_PATTERN_TYPE_ERE, opt);
237 }
238
239 static struct grep_pat *create_grep_pat(const char *pat, size_t patlen,
240 const char *origin, int no,
241 enum grep_pat_token t,
242 enum grep_header_field field)
243 {
244 struct grep_pat *p = xcalloc(1, sizeof(*p));
245 p->pattern = xmemdupz(pat, patlen);
246 p->patternlen = patlen;
247 p->origin = origin;
248 p->no = no;
249 p->token = t;
250 p->field = field;
251 return p;
252 }
253
254 static void do_append_grep_pat(struct grep_pat ***tail, struct grep_pat *p)
255 {
256 **tail = p;
257 *tail = &p->next;
258 p->next = NULL;
259
260 switch (p->token) {
261 case GREP_PATTERN: /* atom */
262 case GREP_PATTERN_HEAD:
263 case GREP_PATTERN_BODY:
264 for (;;) {
265 struct grep_pat *new_pat;
266 size_t len = 0;
267 char *cp = p->pattern + p->patternlen, *nl = NULL;
268 while (++len <= p->patternlen) {
269 if (*(--cp) == '\n') {
270 nl = cp;
271 break;
272 }
273 }
274 if (!nl)
275 break;
276 new_pat = create_grep_pat(nl + 1, len - 1, p->origin,
277 p->no, p->token, p->field);
278 new_pat->next = p->next;
279 if (!p->next)
280 *tail = &new_pat->next;
281 p->next = new_pat;
282 *nl = '\0';
283 p->patternlen -= len;
284 }
285 break;
286 default:
287 break;
288 }
289 }
290
291 void append_header_grep_pattern(struct grep_opt *opt,
292 enum grep_header_field field, const char *pat)
293 {
294 struct grep_pat *p = create_grep_pat(pat, strlen(pat), "header", 0,
295 GREP_PATTERN_HEAD, field);
296 if (field == GREP_HEADER_REFLOG)
297 opt->use_reflog_filter = 1;
298 do_append_grep_pat(&opt->header_tail, p);
299 }
300
301 void append_grep_pattern(struct grep_opt *opt, const char *pat,
302 const char *origin, int no, enum grep_pat_token t)
303 {
304 append_grep_pat(opt, pat, strlen(pat), origin, no, t);
305 }
306
307 void append_grep_pat(struct grep_opt *opt, const char *pat, size_t patlen,
308 const char *origin, int no, enum grep_pat_token t)
309 {
310 struct grep_pat *p = create_grep_pat(pat, patlen, origin, no, t, 0);
311 do_append_grep_pat(&opt->pattern_tail, p);
312 }
313
314 struct grep_opt *grep_opt_dup(const struct grep_opt *opt)
315 {
316 struct grep_pat *pat;
317 struct grep_opt *ret = xmalloc(sizeof(struct grep_opt));
318 *ret = *opt;
319
320 ret->pattern_list = NULL;
321 ret->pattern_tail = &ret->pattern_list;
322
323 for(pat = opt->pattern_list; pat != NULL; pat = pat->next)
324 {
325 if(pat->token == GREP_PATTERN_HEAD)
326 append_header_grep_pattern(ret, pat->field,
327 pat->pattern);
328 else
329 append_grep_pat(ret, pat->pattern, pat->patternlen,
330 pat->origin, pat->no, pat->token);
331 }
332
333 return ret;
334 }
335
336 static NORETURN void compile_regexp_failed(const struct grep_pat *p,
337 const char *error)
338 {
339 char where[1024];
340
341 if (p->no)
342 xsnprintf(where, sizeof(where), "In '%s' at %d, ", p->origin, p->no);
343 else if (p->origin)
344 xsnprintf(where, sizeof(where), "%s, ", p->origin);
345 else
346 where[0] = 0;
347
348 die("%s'%s': %s", where, p->pattern, error);
349 }
350
351 static int is_fixed(const char *s, size_t len)
352 {
353 size_t i;
354
355 for (i = 0; i < len; i++) {
356 if (is_regex_special(s[i]))
357 return 0;
358 }
359
360 return 1;
361 }
362
363 static int has_null(const char *s, size_t len)
364 {
365 /*
366 * regcomp cannot accept patterns with NULs so when using it
367 * we consider any pattern containing a NUL fixed.
368 */
369 if (memchr(s, 0, len))
370 return 1;
371
372 return 0;
373 }
374
375 #ifdef USE_LIBPCRE1
376 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
377 {
378 const char *error;
379 int erroffset;
380 int options = PCRE_MULTILINE;
381
382 if (opt->ignore_case) {
383 if (has_non_ascii(p->pattern))
384 p->pcre1_tables = pcre_maketables();
385 options |= PCRE_CASELESS;
386 }
387 if (is_utf8_locale() && has_non_ascii(p->pattern))
388 options |= PCRE_UTF8;
389
390 p->pcre1_regexp = pcre_compile(p->pattern, options, &error, &erroffset,
391 p->pcre1_tables);
392 if (!p->pcre1_regexp)
393 compile_regexp_failed(p, error);
394
395 p->pcre1_extra_info = pcre_study(p->pcre1_regexp, GIT_PCRE_STUDY_JIT_COMPILE, &error);
396 if (!p->pcre1_extra_info && error)
397 die("%s", error);
398
399 #ifdef GIT_PCRE1_USE_JIT
400 pcre_config(PCRE_CONFIG_JIT, &p->pcre1_jit_on);
401 if (p->pcre1_jit_on == 1) {
402 p->pcre1_jit_stack = pcre_jit_stack_alloc(1, 1024 * 1024);
403 if (!p->pcre1_jit_stack)
404 die("Couldn't allocate PCRE JIT stack");
405 pcre_assign_jit_stack(p->pcre1_extra_info, NULL, p->pcre1_jit_stack);
406 } else if (p->pcre1_jit_on != 0) {
407 die("BUG: The pcre1_jit_on variable should be 0 or 1, not %d",
408 p->pcre1_jit_on);
409 }
410 #endif
411 }
412
413 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
414 regmatch_t *match, int eflags)
415 {
416 int ovector[30], ret, flags = 0;
417
418 if (eflags & REG_NOTBOL)
419 flags |= PCRE_NOTBOL;
420
421 #ifdef GIT_PCRE1_USE_JIT
422 if (p->pcre1_jit_on) {
423 ret = pcre_jit_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
424 eol - line, 0, flags, ovector,
425 ARRAY_SIZE(ovector), p->pcre1_jit_stack);
426 } else
427 #endif
428 {
429 ret = pcre_exec(p->pcre1_regexp, p->pcre1_extra_info, line,
430 eol - line, 0, flags, ovector,
431 ARRAY_SIZE(ovector));
432 }
433
434 if (ret < 0 && ret != PCRE_ERROR_NOMATCH)
435 die("pcre_exec failed with error code %d", ret);
436 if (ret > 0) {
437 ret = 0;
438 match->rm_so = ovector[0];
439 match->rm_eo = ovector[1];
440 }
441
442 return ret;
443 }
444
445 static void free_pcre1_regexp(struct grep_pat *p)
446 {
447 pcre_free(p->pcre1_regexp);
448 #ifdef GIT_PCRE1_USE_JIT
449 if (p->pcre1_jit_on) {
450 pcre_free_study(p->pcre1_extra_info);
451 pcre_jit_stack_free(p->pcre1_jit_stack);
452 } else
453 #endif
454 {
455 pcre_free(p->pcre1_extra_info);
456 }
457 pcre_free((void *)p->pcre1_tables);
458 }
459 #else /* !USE_LIBPCRE1 */
460 static void compile_pcre1_regexp(struct grep_pat *p, const struct grep_opt *opt)
461 {
462 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
463 }
464
465 static int pcre1match(struct grep_pat *p, const char *line, const char *eol,
466 regmatch_t *match, int eflags)
467 {
468 return 1;
469 }
470
471 static void free_pcre1_regexp(struct grep_pat *p)
472 {
473 }
474 #endif /* !USE_LIBPCRE1 */
475
476 #ifdef USE_LIBPCRE2
477 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
478 {
479 int error;
480 PCRE2_UCHAR errbuf[256];
481 PCRE2_SIZE erroffset;
482 int options = PCRE2_MULTILINE;
483 const uint8_t *character_tables = NULL;
484 int jitret;
485 int patinforet;
486 size_t jitsizearg;
487
488 assert(opt->pcre2);
489
490 p->pcre2_compile_context = NULL;
491
492 if (opt->ignore_case) {
493 if (has_non_ascii(p->pattern)) {
494 character_tables = pcre2_maketables(NULL);
495 p->pcre2_compile_context = pcre2_compile_context_create(NULL);
496 pcre2_set_character_tables(p->pcre2_compile_context, character_tables);
497 }
498 options |= PCRE2_CASELESS;
499 }
500 if (is_utf8_locale() && has_non_ascii(p->pattern))
501 options |= PCRE2_UTF;
502
503 p->pcre2_pattern = pcre2_compile((PCRE2_SPTR)p->pattern,
504 p->patternlen, options, &error, &erroffset,
505 p->pcre2_compile_context);
506
507 if (p->pcre2_pattern) {
508 p->pcre2_match_data = pcre2_match_data_create_from_pattern(p->pcre2_pattern, NULL);
509 if (!p->pcre2_match_data)
510 die("Couldn't allocate PCRE2 match data");
511 } else {
512 pcre2_get_error_message(error, errbuf, sizeof(errbuf));
513 compile_regexp_failed(p, (const char *)&errbuf);
514 }
515
516 pcre2_config(PCRE2_CONFIG_JIT, &p->pcre2_jit_on);
517 if (p->pcre2_jit_on == 1) {
518 jitret = pcre2_jit_compile(p->pcre2_pattern, PCRE2_JIT_COMPLETE);
519 if (jitret)
520 die("Couldn't JIT the PCRE2 pattern '%s', got '%d'\n", p->pattern, jitret);
521
522 /*
523 * The pcre2_config(PCRE2_CONFIG_JIT, ...) call just
524 * tells us whether the library itself supports JIT,
525 * but to see whether we're going to be actually using
526 * JIT we need to extract PCRE2_INFO_JITSIZE from the
527 * pattern *after* we do pcre2_jit_compile() above.
528 *
529 * This is because if the pattern contains the
530 * (*NO_JIT) verb (see pcre2syntax(3))
531 * pcre2_jit_compile() will exit early with 0. If we
532 * then proceed to call pcre2_jit_match() further down
533 * the line instead of pcre2_match() we'll either
534 * segfault (pre PCRE 10.31) or run into a fatal error
535 * (post PCRE2 10.31)
536 */
537 patinforet = pcre2_pattern_info(p->pcre2_pattern, PCRE2_INFO_JITSIZE, &jitsizearg);
538 if (patinforet)
539 BUG("pcre2_pattern_info() failed: %d", patinforet);
540 if (jitsizearg == 0) {
541 p->pcre2_jit_on = 0;
542 return;
543 }
544
545 p->pcre2_jit_stack = pcre2_jit_stack_create(1, 1024 * 1024, NULL);
546 if (!p->pcre2_jit_stack)
547 die("Couldn't allocate PCRE2 JIT stack");
548 p->pcre2_match_context = pcre2_match_context_create(NULL);
549 if (!p->pcre2_match_context)
550 die("Couldn't allocate PCRE2 match context");
551 pcre2_jit_stack_assign(p->pcre2_match_context, NULL, p->pcre2_jit_stack);
552 } else if (p->pcre2_jit_on != 0) {
553 die("BUG: The pcre2_jit_on variable should be 0 or 1, not %d",
554 p->pcre1_jit_on);
555 }
556 }
557
558 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
559 regmatch_t *match, int eflags)
560 {
561 int ret, flags = 0;
562 PCRE2_SIZE *ovector;
563 PCRE2_UCHAR errbuf[256];
564
565 if (eflags & REG_NOTBOL)
566 flags |= PCRE2_NOTBOL;
567
568 if (p->pcre2_jit_on)
569 ret = pcre2_jit_match(p->pcre2_pattern, (unsigned char *)line,
570 eol - line, 0, flags, p->pcre2_match_data,
571 NULL);
572 else
573 ret = pcre2_match(p->pcre2_pattern, (unsigned char *)line,
574 eol - line, 0, flags, p->pcre2_match_data,
575 NULL);
576
577 if (ret < 0 && ret != PCRE2_ERROR_NOMATCH) {
578 pcre2_get_error_message(ret, errbuf, sizeof(errbuf));
579 die("%s failed with error code %d: %s",
580 (p->pcre2_jit_on ? "pcre2_jit_match" : "pcre2_match"), ret,
581 errbuf);
582 }
583 if (ret > 0) {
584 ovector = pcre2_get_ovector_pointer(p->pcre2_match_data);
585 ret = 0;
586 match->rm_so = (int)ovector[0];
587 match->rm_eo = (int)ovector[1];
588 }
589
590 return ret;
591 }
592
593 static void free_pcre2_pattern(struct grep_pat *p)
594 {
595 pcre2_compile_context_free(p->pcre2_compile_context);
596 pcre2_code_free(p->pcre2_pattern);
597 pcre2_match_data_free(p->pcre2_match_data);
598 pcre2_jit_stack_free(p->pcre2_jit_stack);
599 pcre2_match_context_free(p->pcre2_match_context);
600 }
601 #else /* !USE_LIBPCRE2 */
602 static void compile_pcre2_pattern(struct grep_pat *p, const struct grep_opt *opt)
603 {
604 /*
605 * Unreachable until USE_LIBPCRE2 becomes synonymous with
606 * USE_LIBPCRE. See the sibling comment in
607 * grep_set_pattern_type_option().
608 */
609 die("cannot use Perl-compatible regexes when not compiled with USE_LIBPCRE");
610 }
611
612 static int pcre2match(struct grep_pat *p, const char *line, const char *eol,
613 regmatch_t *match, int eflags)
614 {
615 return 1;
616 }
617
618 static void free_pcre2_pattern(struct grep_pat *p)
619 {
620 }
621 #endif /* !USE_LIBPCRE2 */
622
623 static void compile_fixed_regexp(struct grep_pat *p, struct grep_opt *opt)
624 {
625 struct strbuf sb = STRBUF_INIT;
626 int err;
627 int regflags = 0;
628
629 basic_regex_quote_buf(&sb, p->pattern);
630 if (opt->ignore_case)
631 regflags |= REG_ICASE;
632 err = regcomp(&p->regexp, sb.buf, regflags);
633 if (opt->debug)
634 fprintf(stderr, "fixed %s\n", sb.buf);
635 strbuf_release(&sb);
636 if (err) {
637 char errbuf[1024];
638 regerror(err, &p->regexp, errbuf, sizeof(errbuf));
639 regfree(&p->regexp);
640 compile_regexp_failed(p, errbuf);
641 }
642 }
643
644 static void compile_regexp(struct grep_pat *p, struct grep_opt *opt)
645 {
646 int ascii_only;
647 int err;
648 int regflags = REG_NEWLINE;
649
650 p->word_regexp = opt->word_regexp;
651 p->ignore_case = opt->ignore_case;
652 ascii_only = !has_non_ascii(p->pattern);
653
654 /*
655 * Even when -F (fixed) asks us to do a non-regexp search, we
656 * may not be able to correctly case-fold when -i
657 * (ignore-case) is asked (in which case, we'll synthesize a
658 * regexp to match the pattern that matches regexp special
659 * characters literally, while ignoring case differences). On
660 * the other hand, even without -F, if the pattern does not
661 * have any regexp special characters and there is no need for
662 * case-folding search, we can internally turn it into a
663 * simple string match using kws. p->fixed tells us if we
664 * want to use kws.
665 */
666 if (opt->fixed ||
667 has_null(p->pattern, p->patternlen) ||
668 is_fixed(p->pattern, p->patternlen))
669 p->fixed = !p->ignore_case || ascii_only;
670
671 if (p->fixed) {
672 p->kws = kwsalloc(p->ignore_case ? tolower_trans_tbl : NULL);
673 kwsincr(p->kws, p->pattern, p->patternlen);
674 kwsprep(p->kws);
675 return;
676 } else if (opt->fixed) {
677 /*
678 * We come here when the pattern has the non-ascii
679 * characters we cannot case-fold, and asked to
680 * ignore-case.
681 */
682 compile_fixed_regexp(p, opt);
683 return;
684 }
685
686 if (opt->pcre2) {
687 compile_pcre2_pattern(p, opt);
688 return;
689 }
690
691 if (opt->pcre1) {
692 compile_pcre1_regexp(p, opt);
693 return;
694 }
695
696 if (p->ignore_case)
697 regflags |= REG_ICASE;
698 if (opt->extended_regexp_option)
699 regflags |= REG_EXTENDED;
700 err = regcomp(&p->regexp, p->pattern, regflags);
701 if (err) {
702 char errbuf[1024];
703 regerror(err, &p->regexp, errbuf, 1024);
704 regfree(&p->regexp);
705 compile_regexp_failed(p, errbuf);
706 }
707 }
708
709 static struct grep_expr *compile_pattern_or(struct grep_pat **);
710 static struct grep_expr *compile_pattern_atom(struct grep_pat **list)
711 {
712 struct grep_pat *p;
713 struct grep_expr *x;
714
715 p = *list;
716 if (!p)
717 return NULL;
718 switch (p->token) {
719 case GREP_PATTERN: /* atom */
720 case GREP_PATTERN_HEAD:
721 case GREP_PATTERN_BODY:
722 x = xcalloc(1, sizeof (struct grep_expr));
723 x->node = GREP_NODE_ATOM;
724 x->u.atom = p;
725 *list = p->next;
726 return x;
727 case GREP_OPEN_PAREN:
728 *list = p->next;
729 x = compile_pattern_or(list);
730 if (!*list || (*list)->token != GREP_CLOSE_PAREN)
731 die("unmatched parenthesis");
732 *list = (*list)->next;
733 return x;
734 default:
735 return NULL;
736 }
737 }
738
739 static struct grep_expr *compile_pattern_not(struct grep_pat **list)
740 {
741 struct grep_pat *p;
742 struct grep_expr *x;
743
744 p = *list;
745 if (!p)
746 return NULL;
747 switch (p->token) {
748 case GREP_NOT:
749 if (!p->next)
750 die("--not not followed by pattern expression");
751 *list = p->next;
752 x = xcalloc(1, sizeof (struct grep_expr));
753 x->node = GREP_NODE_NOT;
754 x->u.unary = compile_pattern_not(list);
755 if (!x->u.unary)
756 die("--not followed by non pattern expression");
757 return x;
758 default:
759 return compile_pattern_atom(list);
760 }
761 }
762
763 static struct grep_expr *compile_pattern_and(struct grep_pat **list)
764 {
765 struct grep_pat *p;
766 struct grep_expr *x, *y, *z;
767
768 x = compile_pattern_not(list);
769 p = *list;
770 if (p && p->token == GREP_AND) {
771 if (!p->next)
772 die("--and not followed by pattern expression");
773 *list = p->next;
774 y = compile_pattern_and(list);
775 if (!y)
776 die("--and not followed by pattern expression");
777 z = xcalloc(1, sizeof (struct grep_expr));
778 z->node = GREP_NODE_AND;
779 z->u.binary.left = x;
780 z->u.binary.right = y;
781 return z;
782 }
783 return x;
784 }
785
786 static struct grep_expr *compile_pattern_or(struct grep_pat **list)
787 {
788 struct grep_pat *p;
789 struct grep_expr *x, *y, *z;
790
791 x = compile_pattern_and(list);
792 p = *list;
793 if (x && p && p->token != GREP_CLOSE_PAREN) {
794 y = compile_pattern_or(list);
795 if (!y)
796 die("not a pattern expression %s", p->pattern);
797 z = xcalloc(1, sizeof (struct grep_expr));
798 z->node = GREP_NODE_OR;
799 z->u.binary.left = x;
800 z->u.binary.right = y;
801 return z;
802 }
803 return x;
804 }
805
806 static struct grep_expr *compile_pattern_expr(struct grep_pat **list)
807 {
808 return compile_pattern_or(list);
809 }
810
811 static void indent(int in)
812 {
813 while (in-- > 0)
814 fputc(' ', stderr);
815 }
816
817 static void dump_grep_pat(struct grep_pat *p)
818 {
819 switch (p->token) {
820 case GREP_AND: fprintf(stderr, "*and*"); break;
821 case GREP_OPEN_PAREN: fprintf(stderr, "*(*"); break;
822 case GREP_CLOSE_PAREN: fprintf(stderr, "*)*"); break;
823 case GREP_NOT: fprintf(stderr, "*not*"); break;
824 case GREP_OR: fprintf(stderr, "*or*"); break;
825
826 case GREP_PATTERN: fprintf(stderr, "pattern"); break;
827 case GREP_PATTERN_HEAD: fprintf(stderr, "pattern_head"); break;
828 case GREP_PATTERN_BODY: fprintf(stderr, "pattern_body"); break;
829 }
830
831 switch (p->token) {
832 default: break;
833 case GREP_PATTERN_HEAD:
834 fprintf(stderr, "<head %d>", p->field); break;
835 case GREP_PATTERN_BODY:
836 fprintf(stderr, "<body>"); break;
837 }
838 switch (p->token) {
839 default: break;
840 case GREP_PATTERN_HEAD:
841 case GREP_PATTERN_BODY:
842 case GREP_PATTERN:
843 fprintf(stderr, "%.*s", (int)p->patternlen, p->pattern);
844 break;
845 }
846 fputc('\n', stderr);
847 }
848
849 static void dump_grep_expression_1(struct grep_expr *x, int in)
850 {
851 indent(in);
852 switch (x->node) {
853 case GREP_NODE_TRUE:
854 fprintf(stderr, "true\n");
855 break;
856 case GREP_NODE_ATOM:
857 dump_grep_pat(x->u.atom);
858 break;
859 case GREP_NODE_NOT:
860 fprintf(stderr, "(not\n");
861 dump_grep_expression_1(x->u.unary, in+1);
862 indent(in);
863 fprintf(stderr, ")\n");
864 break;
865 case GREP_NODE_AND:
866 fprintf(stderr, "(and\n");
867 dump_grep_expression_1(x->u.binary.left, in+1);
868 dump_grep_expression_1(x->u.binary.right, in+1);
869 indent(in);
870 fprintf(stderr, ")\n");
871 break;
872 case GREP_NODE_OR:
873 fprintf(stderr, "(or\n");
874 dump_grep_expression_1(x->u.binary.left, in+1);
875 dump_grep_expression_1(x->u.binary.right, in+1);
876 indent(in);
877 fprintf(stderr, ")\n");
878 break;
879 }
880 }
881
882 static void dump_grep_expression(struct grep_opt *opt)
883 {
884 struct grep_expr *x = opt->pattern_expression;
885
886 if (opt->all_match)
887 fprintf(stderr, "[all-match]\n");
888 dump_grep_expression_1(x, 0);
889 fflush(NULL);
890 }
891
892 static struct grep_expr *grep_true_expr(void)
893 {
894 struct grep_expr *z = xcalloc(1, sizeof(*z));
895 z->node = GREP_NODE_TRUE;
896 return z;
897 }
898
899 static struct grep_expr *grep_or_expr(struct grep_expr *left, struct grep_expr *right)
900 {
901 struct grep_expr *z = xcalloc(1, sizeof(*z));
902 z->node = GREP_NODE_OR;
903 z->u.binary.left = left;
904 z->u.binary.right = right;
905 return z;
906 }
907
908 static struct grep_expr *prep_header_patterns(struct grep_opt *opt)
909 {
910 struct grep_pat *p;
911 struct grep_expr *header_expr;
912 struct grep_expr *(header_group[GREP_HEADER_FIELD_MAX]);
913 enum grep_header_field fld;
914
915 if (!opt->header_list)
916 return NULL;
917
918 for (p = opt->header_list; p; p = p->next) {
919 if (p->token != GREP_PATTERN_HEAD)
920 die("BUG: a non-header pattern in grep header list.");
921 if (p->field < GREP_HEADER_FIELD_MIN ||
922 GREP_HEADER_FIELD_MAX <= p->field)
923 die("BUG: unknown header field %d", p->field);
924 compile_regexp(p, opt);
925 }
926
927 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++)
928 header_group[fld] = NULL;
929
930 for (p = opt->header_list; p; p = p->next) {
931 struct grep_expr *h;
932 struct grep_pat *pp = p;
933
934 h = compile_pattern_atom(&pp);
935 if (!h || pp != p->next)
936 die("BUG: malformed header expr");
937 if (!header_group[p->field]) {
938 header_group[p->field] = h;
939 continue;
940 }
941 header_group[p->field] = grep_or_expr(h, header_group[p->field]);
942 }
943
944 header_expr = NULL;
945
946 for (fld = 0; fld < GREP_HEADER_FIELD_MAX; fld++) {
947 if (!header_group[fld])
948 continue;
949 if (!header_expr)
950 header_expr = grep_true_expr();
951 header_expr = grep_or_expr(header_group[fld], header_expr);
952 }
953 return header_expr;
954 }
955
956 static struct grep_expr *grep_splice_or(struct grep_expr *x, struct grep_expr *y)
957 {
958 struct grep_expr *z = x;
959
960 while (x) {
961 assert(x->node == GREP_NODE_OR);
962 if (x->u.binary.right &&
963 x->u.binary.right->node == GREP_NODE_TRUE) {
964 x->u.binary.right = y;
965 break;
966 }
967 x = x->u.binary.right;
968 }
969 return z;
970 }
971
972 static void compile_grep_patterns_real(struct grep_opt *opt)
973 {
974 struct grep_pat *p;
975 struct grep_expr *header_expr = prep_header_patterns(opt);
976
977 for (p = opt->pattern_list; p; p = p->next) {
978 switch (p->token) {
979 case GREP_PATTERN: /* atom */
980 case GREP_PATTERN_HEAD:
981 case GREP_PATTERN_BODY:
982 compile_regexp(p, opt);
983 break;
984 default:
985 opt->extended = 1;
986 break;
987 }
988 }
989
990 if (opt->all_match || header_expr)
991 opt->extended = 1;
992 else if (!opt->extended && !opt->debug)
993 return;
994
995 p = opt->pattern_list;
996 if (p)
997 opt->pattern_expression = compile_pattern_expr(&p);
998 if (p)
999 die("incomplete pattern expression: %s", p->pattern);
1000
1001 if (!header_expr)
1002 return;
1003
1004 if (!opt->pattern_expression)
1005 opt->pattern_expression = header_expr;
1006 else if (opt->all_match)
1007 opt->pattern_expression = grep_splice_or(header_expr,
1008 opt->pattern_expression);
1009 else
1010 opt->pattern_expression = grep_or_expr(opt->pattern_expression,
1011 header_expr);
1012 opt->all_match = 1;
1013 }
1014
1015 void compile_grep_patterns(struct grep_opt *opt)
1016 {
1017 compile_grep_patterns_real(opt);
1018 if (opt->debug)
1019 dump_grep_expression(opt);
1020 }
1021
1022 static void free_pattern_expr(struct grep_expr *x)
1023 {
1024 switch (x->node) {
1025 case GREP_NODE_TRUE:
1026 case GREP_NODE_ATOM:
1027 break;
1028 case GREP_NODE_NOT:
1029 free_pattern_expr(x->u.unary);
1030 break;
1031 case GREP_NODE_AND:
1032 case GREP_NODE_OR:
1033 free_pattern_expr(x->u.binary.left);
1034 free_pattern_expr(x->u.binary.right);
1035 break;
1036 }
1037 free(x);
1038 }
1039
1040 void free_grep_patterns(struct grep_opt *opt)
1041 {
1042 struct grep_pat *p, *n;
1043
1044 for (p = opt->pattern_list; p; p = n) {
1045 n = p->next;
1046 switch (p->token) {
1047 case GREP_PATTERN: /* atom */
1048 case GREP_PATTERN_HEAD:
1049 case GREP_PATTERN_BODY:
1050 if (p->kws)
1051 kwsfree(p->kws);
1052 else if (p->pcre1_regexp)
1053 free_pcre1_regexp(p);
1054 else if (p->pcre2_pattern)
1055 free_pcre2_pattern(p);
1056 else
1057 regfree(&p->regexp);
1058 free(p->pattern);
1059 break;
1060 default:
1061 break;
1062 }
1063 free(p);
1064 }
1065
1066 if (!opt->extended)
1067 return;
1068 free_pattern_expr(opt->pattern_expression);
1069 }
1070
1071 static char *end_of_line(char *cp, unsigned long *left)
1072 {
1073 unsigned long l = *left;
1074 while (l && *cp != '\n') {
1075 l--;
1076 cp++;
1077 }
1078 *left = l;
1079 return cp;
1080 }
1081
1082 static int word_char(char ch)
1083 {
1084 return isalnum(ch) || ch == '_';
1085 }
1086
1087 static void output_color(struct grep_opt *opt, const void *data, size_t size,
1088 const char *color)
1089 {
1090 if (want_color(opt->color) && color && color[0]) {
1091 opt->output(opt, color, strlen(color));
1092 opt->output(opt, data, size);
1093 opt->output(opt, GIT_COLOR_RESET, strlen(GIT_COLOR_RESET));
1094 } else
1095 opt->output(opt, data, size);
1096 }
1097
1098 static void output_sep(struct grep_opt *opt, char sign)
1099 {
1100 if (opt->null_following_name)
1101 opt->output(opt, "\0", 1);
1102 else
1103 output_color(opt, &sign, 1, opt->color_sep);
1104 }
1105
1106 static void show_name(struct grep_opt *opt, const char *name)
1107 {
1108 output_color(opt, name, strlen(name), opt->color_filename);
1109 opt->output(opt, opt->null_following_name ? "\0" : "\n", 1);
1110 }
1111
1112 static int fixmatch(struct grep_pat *p, char *line, char *eol,
1113 regmatch_t *match)
1114 {
1115 struct kwsmatch kwsm;
1116 size_t offset = kwsexec(p->kws, line, eol - line, &kwsm);
1117 if (offset == -1) {
1118 match->rm_so = match->rm_eo = -1;
1119 return REG_NOMATCH;
1120 } else {
1121 match->rm_so = offset;
1122 match->rm_eo = match->rm_so + kwsm.size[0];
1123 return 0;
1124 }
1125 }
1126
1127 static int patmatch(struct grep_pat *p, char *line, char *eol,
1128 regmatch_t *match, int eflags)
1129 {
1130 int hit;
1131
1132 if (p->fixed)
1133 hit = !fixmatch(p, line, eol, match);
1134 else if (p->pcre1_regexp)
1135 hit = !pcre1match(p, line, eol, match, eflags);
1136 else if (p->pcre2_pattern)
1137 hit = !pcre2match(p, line, eol, match, eflags);
1138 else
1139 hit = !regexec_buf(&p->regexp, line, eol - line, 1, match,
1140 eflags);
1141
1142 return hit;
1143 }
1144
1145 static int strip_timestamp(char *bol, char **eol_p)
1146 {
1147 char *eol = *eol_p;
1148 int ch;
1149
1150 while (bol < --eol) {
1151 if (*eol != '>')
1152 continue;
1153 *eol_p = ++eol;
1154 ch = *eol;
1155 *eol = '\0';
1156 return ch;
1157 }
1158 return 0;
1159 }
1160
1161 static struct {
1162 const char *field;
1163 size_t len;
1164 } header_field[] = {
1165 { "author ", 7 },
1166 { "committer ", 10 },
1167 { "reflog ", 7 },
1168 };
1169
1170 static int match_one_pattern(struct grep_pat *p, char *bol, char *eol,
1171 enum grep_context ctx,
1172 regmatch_t *pmatch, int eflags)
1173 {
1174 int hit = 0;
1175 int saved_ch = 0;
1176 const char *start = bol;
1177
1178 if ((p->token != GREP_PATTERN) &&
1179 ((p->token == GREP_PATTERN_HEAD) != (ctx == GREP_CONTEXT_HEAD)))
1180 return 0;
1181
1182 if (p->token == GREP_PATTERN_HEAD) {
1183 const char *field;
1184 size_t len;
1185 assert(p->field < ARRAY_SIZE(header_field));
1186 field = header_field[p->field].field;
1187 len = header_field[p->field].len;
1188 if (strncmp(bol, field, len))
1189 return 0;
1190 bol += len;
1191 switch (p->field) {
1192 case GREP_HEADER_AUTHOR:
1193 case GREP_HEADER_COMMITTER:
1194 saved_ch = strip_timestamp(bol, &eol);
1195 break;
1196 default:
1197 break;
1198 }
1199 }
1200
1201 again:
1202 hit = patmatch(p, bol, eol, pmatch, eflags);
1203
1204 if (hit && p->word_regexp) {
1205 if ((pmatch[0].rm_so < 0) ||
1206 (eol - bol) < pmatch[0].rm_so ||
1207 (pmatch[0].rm_eo < 0) ||
1208 (eol - bol) < pmatch[0].rm_eo)
1209 die("regexp returned nonsense");
1210
1211 /* Match beginning must be either beginning of the
1212 * line, or at word boundary (i.e. the last char must
1213 * not be a word char). Similarly, match end must be
1214 * either end of the line, or at word boundary
1215 * (i.e. the next char must not be a word char).
1216 */
1217 if ( ((pmatch[0].rm_so == 0) ||
1218 !word_char(bol[pmatch[0].rm_so-1])) &&
1219 ((pmatch[0].rm_eo == (eol-bol)) ||
1220 !word_char(bol[pmatch[0].rm_eo])) )
1221 ;
1222 else
1223 hit = 0;
1224
1225 /* Words consist of at least one character. */
1226 if (pmatch->rm_so == pmatch->rm_eo)
1227 hit = 0;
1228
1229 if (!hit && pmatch[0].rm_so + bol + 1 < eol) {
1230 /* There could be more than one match on the
1231 * line, and the first match might not be
1232 * strict word match. But later ones could be!
1233 * Forward to the next possible start, i.e. the
1234 * next position following a non-word char.
1235 */
1236 bol = pmatch[0].rm_so + bol + 1;
1237 while (word_char(bol[-1]) && bol < eol)
1238 bol++;
1239 eflags |= REG_NOTBOL;
1240 if (bol < eol)
1241 goto again;
1242 }
1243 }
1244 if (p->token == GREP_PATTERN_HEAD && saved_ch)
1245 *eol = saved_ch;
1246 if (hit) {
1247 pmatch[0].rm_so += bol - start;
1248 pmatch[0].rm_eo += bol - start;
1249 }
1250 return hit;
1251 }
1252
1253 static int match_expr_eval(struct grep_expr *x, char *bol, char *eol,
1254 enum grep_context ctx, int collect_hits)
1255 {
1256 int h = 0;
1257 regmatch_t match;
1258
1259 if (!x)
1260 die("Not a valid grep expression");
1261 switch (x->node) {
1262 case GREP_NODE_TRUE:
1263 h = 1;
1264 break;
1265 case GREP_NODE_ATOM:
1266 h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0);
1267 break;
1268 case GREP_NODE_NOT:
1269 h = !match_expr_eval(x->u.unary, bol, eol, ctx, 0);
1270 break;
1271 case GREP_NODE_AND:
1272 if (!match_expr_eval(x->u.binary.left, bol, eol, ctx, 0))
1273 return 0;
1274 h = match_expr_eval(x->u.binary.right, bol, eol, ctx, 0);
1275 break;
1276 case GREP_NODE_OR:
1277 if (!collect_hits)
1278 return (match_expr_eval(x->u.binary.left,
1279 bol, eol, ctx, 0) ||
1280 match_expr_eval(x->u.binary.right,
1281 bol, eol, ctx, 0));
1282 h = match_expr_eval(x->u.binary.left, bol, eol, ctx, 0);
1283 x->u.binary.left->hit |= h;
1284 h |= match_expr_eval(x->u.binary.right, bol, eol, ctx, 1);
1285 break;
1286 default:
1287 die("Unexpected node type (internal error) %d", x->node);
1288 }
1289 if (collect_hits)
1290 x->hit |= h;
1291 return h;
1292 }
1293
1294 static int match_expr(struct grep_opt *opt, char *bol, char *eol,
1295 enum grep_context ctx, int collect_hits)
1296 {
1297 struct grep_expr *x = opt->pattern_expression;
1298 return match_expr_eval(x, bol, eol, ctx, collect_hits);
1299 }
1300
1301 static int match_line(struct grep_opt *opt, char *bol, char *eol,
1302 enum grep_context ctx, int collect_hits)
1303 {
1304 struct grep_pat *p;
1305 regmatch_t match;
1306
1307 if (opt->extended)
1308 return match_expr(opt, bol, eol, ctx, collect_hits);
1309
1310 /* we do not call with collect_hits without being extended */
1311 for (p = opt->pattern_list; p; p = p->next) {
1312 if (match_one_pattern(p, bol, eol, ctx, &match, 0))
1313 return 1;
1314 }
1315 return 0;
1316 }
1317
1318 static int match_next_pattern(struct grep_pat *p, char *bol, char *eol,
1319 enum grep_context ctx,
1320 regmatch_t *pmatch, int eflags)
1321 {
1322 regmatch_t match;
1323
1324 if (!match_one_pattern(p, bol, eol, ctx, &match, eflags))
1325 return 0;
1326 if (match.rm_so < 0 || match.rm_eo < 0)
1327 return 0;
1328 if (pmatch->rm_so >= 0 && pmatch->rm_eo >= 0) {
1329 if (match.rm_so > pmatch->rm_so)
1330 return 1;
1331 if (match.rm_so == pmatch->rm_so && match.rm_eo < pmatch->rm_eo)
1332 return 1;
1333 }
1334 pmatch->rm_so = match.rm_so;
1335 pmatch->rm_eo = match.rm_eo;
1336 return 1;
1337 }
1338
1339 static int next_match(struct grep_opt *opt, char *bol, char *eol,
1340 enum grep_context ctx, regmatch_t *pmatch, int eflags)
1341 {
1342 struct grep_pat *p;
1343 int hit = 0;
1344
1345 pmatch->rm_so = pmatch->rm_eo = -1;
1346 if (bol < eol) {
1347 for (p = opt->pattern_list; p; p = p->next) {
1348 switch (p->token) {
1349 case GREP_PATTERN: /* atom */
1350 case GREP_PATTERN_HEAD:
1351 case GREP_PATTERN_BODY:
1352 hit |= match_next_pattern(p, bol, eol, ctx,
1353 pmatch, eflags);
1354 break;
1355 default:
1356 break;
1357 }
1358 }
1359 }
1360 return hit;
1361 }
1362
1363 static void show_line(struct grep_opt *opt, char *bol, char *eol,
1364 const char *name, unsigned lno, char sign)
1365 {
1366 int rest = eol - bol;
1367 const char *match_color, *line_color = NULL;
1368
1369 if (opt->file_break && opt->last_shown == 0) {
1370 if (opt->show_hunk_mark)
1371 opt->output(opt, "\n", 1);
1372 } else if (opt->pre_context || opt->post_context || opt->funcbody) {
1373 if (opt->last_shown == 0) {
1374 if (opt->show_hunk_mark) {
1375 output_color(opt, "--", 2, opt->color_sep);
1376 opt->output(opt, "\n", 1);
1377 }
1378 } else if (lno > opt->last_shown + 1) {
1379 output_color(opt, "--", 2, opt->color_sep);
1380 opt->output(opt, "\n", 1);
1381 }
1382 }
1383 if (opt->heading && opt->last_shown == 0) {
1384 output_color(opt, name, strlen(name), opt->color_filename);
1385 opt->output(opt, "\n", 1);
1386 }
1387 opt->last_shown = lno;
1388
1389 if (!opt->heading && opt->pathname) {
1390 output_color(opt, name, strlen(name), opt->color_filename);
1391 output_sep(opt, sign);
1392 }
1393 if (opt->linenum) {
1394 char buf[32];
1395 xsnprintf(buf, sizeof(buf), "%d", lno);
1396 output_color(opt, buf, strlen(buf), opt->color_lineno);
1397 output_sep(opt, sign);
1398 }
1399 if (opt->color) {
1400 regmatch_t match;
1401 enum grep_context ctx = GREP_CONTEXT_BODY;
1402 int ch = *eol;
1403 int eflags = 0;
1404
1405 if (sign == ':')
1406 match_color = opt->color_match_selected;
1407 else
1408 match_color = opt->color_match_context;
1409 if (sign == ':')
1410 line_color = opt->color_selected;
1411 else if (sign == '-')
1412 line_color = opt->color_context;
1413 else if (sign == '=')
1414 line_color = opt->color_function;
1415 *eol = '\0';
1416 while (next_match(opt, bol, eol, ctx, &match, eflags)) {
1417 if (match.rm_so == match.rm_eo)
1418 break;
1419
1420 output_color(opt, bol, match.rm_so, line_color);
1421 output_color(opt, bol + match.rm_so,
1422 match.rm_eo - match.rm_so, match_color);
1423 bol += match.rm_eo;
1424 rest -= match.rm_eo;
1425 eflags = REG_NOTBOL;
1426 }
1427 *eol = ch;
1428 }
1429 output_color(opt, bol, rest, line_color);
1430 opt->output(opt, "\n", 1);
1431 }
1432
1433 #ifndef NO_PTHREADS
1434 int grep_use_locks;
1435
1436 /*
1437 * This lock protects access to the gitattributes machinery, which is
1438 * not thread-safe.
1439 */
1440 pthread_mutex_t grep_attr_mutex;
1441
1442 static inline void grep_attr_lock(void)
1443 {
1444 if (grep_use_locks)
1445 pthread_mutex_lock(&grep_attr_mutex);
1446 }
1447
1448 static inline void grep_attr_unlock(void)
1449 {
1450 if (grep_use_locks)
1451 pthread_mutex_unlock(&grep_attr_mutex);
1452 }
1453
1454 /*
1455 * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
1456 */
1457 pthread_mutex_t grep_read_mutex;
1458
1459 #else
1460 #define grep_attr_lock()
1461 #define grep_attr_unlock()
1462 #endif
1463
1464 static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol)
1465 {
1466 xdemitconf_t *xecfg = opt->priv;
1467 if (xecfg && !xecfg->find_func) {
1468 grep_source_load_driver(gs);
1469 if (gs->driver->funcname.pattern) {
1470 const struct userdiff_funcname *pe = &gs->driver->funcname;
1471 xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
1472 } else {
1473 xecfg = opt->priv = NULL;
1474 }
1475 }
1476
1477 if (xecfg) {
1478 char buf[1];
1479 return xecfg->find_func(bol, eol - bol, buf, 1,
1480 xecfg->find_func_priv) >= 0;
1481 }
1482
1483 if (bol == eol)
1484 return 0;
1485 if (isalpha(*bol) || *bol == '_' || *bol == '$')
1486 return 1;
1487 return 0;
1488 }
1489
1490 static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs,
1491 char *bol, unsigned lno)
1492 {
1493 while (bol > gs->buf) {
1494 char *eol = --bol;
1495
1496 while (bol > gs->buf && bol[-1] != '\n')
1497 bol--;
1498 lno--;
1499
1500 if (lno <= opt->last_shown)
1501 break;
1502
1503 if (match_funcname(opt, gs, bol, eol)) {
1504 show_line(opt, bol, eol, gs->name, lno, '=');
1505 break;
1506 }
1507 }
1508 }
1509
1510 static int is_empty_line(const char *bol, const char *eol);
1511
1512 static void show_pre_context(struct grep_opt *opt, struct grep_source *gs,
1513 char *bol, char *end, unsigned lno)
1514 {
1515 unsigned cur = lno, from = 1, funcname_lno = 0, orig_from;
1516 int funcname_needed = !!opt->funcname, comment_needed = 0;
1517
1518 if (opt->pre_context < lno)
1519 from = lno - opt->pre_context;
1520 if (from <= opt->last_shown)
1521 from = opt->last_shown + 1;
1522 orig_from = from;
1523 if (opt->funcbody) {
1524 if (match_funcname(opt, gs, bol, end))
1525 comment_needed = 1;
1526 else
1527 funcname_needed = 1;
1528 from = opt->last_shown + 1;
1529 }
1530
1531 /* Rewind. */
1532 while (bol > gs->buf && cur > from) {
1533 char *next_bol = bol;
1534 char *eol = --bol;
1535
1536 while (bol > gs->buf && bol[-1] != '\n')
1537 bol--;
1538 cur--;
1539 if (comment_needed && (is_empty_line(bol, eol) ||
1540 match_funcname(opt, gs, bol, eol))) {
1541 comment_needed = 0;
1542 from = orig_from;
1543 if (cur < from) {
1544 cur++;
1545 bol = next_bol;
1546 break;
1547 }
1548 }
1549 if (funcname_needed && match_funcname(opt, gs, bol, eol)) {
1550 funcname_lno = cur;
1551 funcname_needed = 0;
1552 if (opt->funcbody)
1553 comment_needed = 1;
1554 else
1555 from = orig_from;
1556 }
1557 }
1558
1559 /* We need to look even further back to find a function signature. */
1560 if (opt->funcname && funcname_needed)
1561 show_funcname_line(opt, gs, bol, cur);
1562
1563 /* Back forward. */
1564 while (cur < lno) {
1565 char *eol = bol, sign = (cur == funcname_lno) ? '=' : '-';
1566
1567 while (*eol != '\n')
1568 eol++;
1569 show_line(opt, bol, eol, gs->name, cur, sign);
1570 bol = eol + 1;
1571 cur++;
1572 }
1573 }
1574
1575 static int should_lookahead(struct grep_opt *opt)
1576 {
1577 struct grep_pat *p;
1578
1579 if (opt->extended)
1580 return 0; /* punt for too complex stuff */
1581 if (opt->invert)
1582 return 0;
1583 for (p = opt->pattern_list; p; p = p->next) {
1584 if (p->token != GREP_PATTERN)
1585 return 0; /* punt for "header only" and stuff */
1586 }
1587 return 1;
1588 }
1589
1590 static int look_ahead(struct grep_opt *opt,
1591 unsigned long *left_p,
1592 unsigned *lno_p,
1593 char **bol_p)
1594 {
1595 unsigned lno = *lno_p;
1596 char *bol = *bol_p;
1597 struct grep_pat *p;
1598 char *sp, *last_bol;
1599 regoff_t earliest = -1;
1600
1601 for (p = opt->pattern_list; p; p = p->next) {
1602 int hit;
1603 regmatch_t m;
1604
1605 hit = patmatch(p, bol, bol + *left_p, &m, 0);
1606 if (!hit || m.rm_so < 0 || m.rm_eo < 0)
1607 continue;
1608 if (earliest < 0 || m.rm_so < earliest)
1609 earliest = m.rm_so;
1610 }
1611
1612 if (earliest < 0) {
1613 *bol_p = bol + *left_p;
1614 *left_p = 0;
1615 return 1;
1616 }
1617 for (sp = bol + earliest; bol < sp && sp[-1] != '\n'; sp--)
1618 ; /* find the beginning of the line */
1619 last_bol = sp;
1620
1621 for (sp = bol; sp < last_bol; sp++) {
1622 if (*sp == '\n')
1623 lno++;
1624 }
1625 *left_p -= last_bol - bol;
1626 *bol_p = last_bol;
1627 *lno_p = lno;
1628 return 0;
1629 }
1630
1631 static int fill_textconv_grep(struct userdiff_driver *driver,
1632 struct grep_source *gs)
1633 {
1634 struct diff_filespec *df;
1635 char *buf;
1636 size_t size;
1637
1638 if (!driver || !driver->textconv)
1639 return grep_source_load(gs);
1640
1641 /*
1642 * The textconv interface is intimately tied to diff_filespecs, so we
1643 * have to pretend to be one. If we could unify the grep_source
1644 * and diff_filespec structs, this mess could just go away.
1645 */
1646 df = alloc_filespec(gs->path);
1647 switch (gs->type) {
1648 case GREP_SOURCE_OID:
1649 fill_filespec(df, gs->identifier, 1, 0100644);
1650 break;
1651 case GREP_SOURCE_FILE:
1652 fill_filespec(df, &null_oid, 0, 0100644);
1653 break;
1654 default:
1655 die("BUG: attempt to textconv something without a path?");
1656 }
1657
1658 /*
1659 * fill_textconv is not remotely thread-safe; it may load objects
1660 * behind the scenes, and it modifies the global diff tempfile
1661 * structure.
1662 */
1663 grep_read_lock();
1664 size = fill_textconv(driver, df, &buf);
1665 grep_read_unlock();
1666 free_filespec(df);
1667
1668 /*
1669 * The normal fill_textconv usage by the diff machinery would just keep
1670 * the textconv'd buf separate from the diff_filespec. But much of the
1671 * grep code passes around a grep_source and assumes that its "buf"
1672 * pointer is the beginning of the thing we are searching. So let's
1673 * install our textconv'd version into the grep_source, taking care not
1674 * to leak any existing buffer.
1675 */
1676 grep_source_clear_data(gs);
1677 gs->buf = buf;
1678 gs->size = size;
1679
1680 return 0;
1681 }
1682
1683 static int is_empty_line(const char *bol, const char *eol)
1684 {
1685 while (bol < eol && isspace(*bol))
1686 bol++;
1687 return bol == eol;
1688 }
1689
1690 static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits)
1691 {
1692 char *bol;
1693 char *peek_bol = NULL;
1694 unsigned long left;
1695 unsigned lno = 1;
1696 unsigned last_hit = 0;
1697 int binary_match_only = 0;
1698 unsigned count = 0;
1699 int try_lookahead = 0;
1700 int show_function = 0;
1701 struct userdiff_driver *textconv = NULL;
1702 enum grep_context ctx = GREP_CONTEXT_HEAD;
1703 xdemitconf_t xecfg;
1704
1705 if (!opt->output)
1706 opt->output = std_output;
1707
1708 if (opt->pre_context || opt->post_context || opt->file_break ||
1709 opt->funcbody) {
1710 /* Show hunk marks, except for the first file. */
1711 if (opt->last_shown)
1712 opt->show_hunk_mark = 1;
1713 /*
1714 * If we're using threads then we can't easily identify
1715 * the first file. Always put hunk marks in that case
1716 * and skip the very first one later in work_done().
1717 */
1718 if (opt->output != std_output)
1719 opt->show_hunk_mark = 1;
1720 }
1721 opt->last_shown = 0;
1722
1723 if (opt->allow_textconv) {
1724 grep_source_load_driver(gs);
1725 /*
1726 * We might set up the shared textconv cache data here, which
1727 * is not thread-safe.
1728 */
1729 grep_attr_lock();
1730 textconv = userdiff_get_textconv(gs->driver);
1731 grep_attr_unlock();
1732 }
1733
1734 /*
1735 * We know the result of a textconv is text, so we only have to care
1736 * about binary handling if we are not using it.
1737 */
1738 if (!textconv) {
1739 switch (opt->binary) {
1740 case GREP_BINARY_DEFAULT:
1741 if (grep_source_is_binary(gs))
1742 binary_match_only = 1;
1743 break;
1744 case GREP_BINARY_NOMATCH:
1745 if (grep_source_is_binary(gs))
1746 return 0; /* Assume unmatch */
1747 break;
1748 case GREP_BINARY_TEXT:
1749 break;
1750 default:
1751 die("BUG: unknown binary handling mode");
1752 }
1753 }
1754
1755 memset(&xecfg, 0, sizeof(xecfg));
1756 opt->priv = &xecfg;
1757
1758 try_lookahead = should_lookahead(opt);
1759
1760 if (fill_textconv_grep(textconv, gs) < 0)
1761 return 0;
1762
1763 bol = gs->buf;
1764 left = gs->size;
1765 while (left) {
1766 char *eol, ch;
1767 int hit;
1768
1769 /*
1770 * look_ahead() skips quickly to the line that possibly
1771 * has the next hit; don't call it if we need to do
1772 * something more than just skipping the current line
1773 * in response to an unmatch for the current line. E.g.
1774 * inside a post-context window, we will show the current
1775 * line as a context around the previous hit when it
1776 * doesn't hit.
1777 */
1778 if (try_lookahead
1779 && !(last_hit
1780 && (show_function ||
1781 lno <= last_hit + opt->post_context))
1782 && look_ahead(opt, &left, &lno, &bol))
1783 break;
1784 eol = end_of_line(bol, &left);
1785 ch = *eol;
1786 *eol = 0;
1787
1788 if ((ctx == GREP_CONTEXT_HEAD) && (eol == bol))
1789 ctx = GREP_CONTEXT_BODY;
1790
1791 hit = match_line(opt, bol, eol, ctx, collect_hits);
1792 *eol = ch;
1793
1794 if (collect_hits)
1795 goto next_line;
1796
1797 /* "grep -v -e foo -e bla" should list lines
1798 * that do not have either, so inversion should
1799 * be done outside.
1800 */
1801 if (opt->invert)
1802 hit = !hit;
1803 if (opt->unmatch_name_only) {
1804 if (hit)
1805 return 0;
1806 goto next_line;
1807 }
1808 if (hit) {
1809 count++;
1810 if (opt->status_only)
1811 return 1;
1812 if (opt->name_only) {
1813 show_name(opt, gs->name);
1814 return 1;
1815 }
1816 if (opt->count)
1817 goto next_line;
1818 if (binary_match_only) {
1819 opt->output(opt, "Binary file ", 12);
1820 output_color(opt, gs->name, strlen(gs->name),
1821 opt->color_filename);
1822 opt->output(opt, " matches\n", 9);
1823 return 1;
1824 }
1825 /* Hit at this line. If we haven't shown the
1826 * pre-context lines, we would need to show them.
1827 */
1828 if (opt->pre_context || opt->funcbody)
1829 show_pre_context(opt, gs, bol, eol, lno);
1830 else if (opt->funcname)
1831 show_funcname_line(opt, gs, bol, lno);
1832 show_line(opt, bol, eol, gs->name, lno, ':');
1833 last_hit = lno;
1834 if (opt->funcbody)
1835 show_function = 1;
1836 goto next_line;
1837 }
1838 if (show_function && (!peek_bol || peek_bol < bol)) {
1839 unsigned long peek_left = left;
1840 char *peek_eol = eol;
1841
1842 /*
1843 * Trailing empty lines are not interesting.
1844 * Peek past them to see if they belong to the
1845 * body of the current function.
1846 */
1847 peek_bol = bol;
1848 while (is_empty_line(peek_bol, peek_eol)) {
1849 peek_bol = peek_eol + 1;
1850 peek_eol = end_of_line(peek_bol, &peek_left);
1851 }
1852
1853 if (match_funcname(opt, gs, peek_bol, peek_eol))
1854 show_function = 0;
1855 }
1856 if (show_function ||
1857 (last_hit && lno <= last_hit + opt->post_context)) {
1858 /* If the last hit is within the post context,
1859 * we need to show this line.
1860 */
1861 show_line(opt, bol, eol, gs->name, lno, '-');
1862 }
1863
1864 next_line:
1865 bol = eol + 1;
1866 if (!left)
1867 break;
1868 left--;
1869 lno++;
1870 }
1871
1872 if (collect_hits)
1873 return 0;
1874
1875 if (opt->status_only)
1876 return opt->unmatch_name_only;
1877 if (opt->unmatch_name_only) {
1878 /* We did not see any hit, so we want to show this */
1879 show_name(opt, gs->name);
1880 return 1;
1881 }
1882
1883 xdiff_clear_find_func(&xecfg);
1884 opt->priv = NULL;
1885
1886 /* NEEDSWORK:
1887 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
1888 * which feels mostly useless but sometimes useful. Maybe
1889 * make it another option? For now suppress them.
1890 */
1891 if (opt->count && count) {
1892 char buf[32];
1893 if (opt->pathname) {
1894 output_color(opt, gs->name, strlen(gs->name),
1895 opt->color_filename);
1896 output_sep(opt, ':');
1897 }
1898 xsnprintf(buf, sizeof(buf), "%u\n", count);
1899 opt->output(opt, buf, strlen(buf));
1900 return 1;
1901 }
1902 return !!last_hit;
1903 }
1904
1905 static void clr_hit_marker(struct grep_expr *x)
1906 {
1907 /* All-hit markers are meaningful only at the very top level
1908 * OR node.
1909 */
1910 while (1) {
1911 x->hit = 0;
1912 if (x->node != GREP_NODE_OR)
1913 return;
1914 x->u.binary.left->hit = 0;
1915 x = x->u.binary.right;
1916 }
1917 }
1918
1919 static int chk_hit_marker(struct grep_expr *x)
1920 {
1921 /* Top level nodes have hit markers. See if they all are hits */
1922 while (1) {
1923 if (x->node != GREP_NODE_OR)
1924 return x->hit;
1925 if (!x->u.binary.left->hit)
1926 return 0;
1927 x = x->u.binary.right;
1928 }
1929 }
1930
1931 int grep_source(struct grep_opt *opt, struct grep_source *gs)
1932 {
1933 /*
1934 * we do not have to do the two-pass grep when we do not check
1935 * buffer-wide "all-match".
1936 */
1937 if (!opt->all_match)
1938 return grep_source_1(opt, gs, 0);
1939
1940 /* Otherwise the toplevel "or" terms hit a bit differently.
1941 * We first clear hit markers from them.
1942 */
1943 clr_hit_marker(opt->pattern_expression);
1944 grep_source_1(opt, gs, 1);
1945
1946 if (!chk_hit_marker(opt->pattern_expression))
1947 return 0;
1948
1949 return grep_source_1(opt, gs, 0);
1950 }
1951
1952 int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size)
1953 {
1954 struct grep_source gs;
1955 int r;
1956
1957 grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL, NULL);
1958 gs.buf = buf;
1959 gs.size = size;
1960
1961 r = grep_source(opt, &gs);
1962
1963 grep_source_clear(&gs);
1964 return r;
1965 }
1966
1967 void grep_source_init(struct grep_source *gs, enum grep_source_type type,
1968 const char *name, const char *path,
1969 const void *identifier)
1970 {
1971 gs->type = type;
1972 gs->name = xstrdup_or_null(name);
1973 gs->path = xstrdup_or_null(path);
1974 gs->buf = NULL;
1975 gs->size = 0;
1976 gs->driver = NULL;
1977
1978 switch (type) {
1979 case GREP_SOURCE_FILE:
1980 gs->identifier = xstrdup(identifier);
1981 break;
1982 case GREP_SOURCE_OID:
1983 gs->identifier = oiddup(identifier);
1984 break;
1985 case GREP_SOURCE_BUF:
1986 gs->identifier = NULL;
1987 break;
1988 }
1989 }
1990
1991 void grep_source_clear(struct grep_source *gs)
1992 {
1993 FREE_AND_NULL(gs->name);
1994 FREE_AND_NULL(gs->path);
1995 FREE_AND_NULL(gs->identifier);
1996 grep_source_clear_data(gs);
1997 }
1998
1999 void grep_source_clear_data(struct grep_source *gs)
2000 {
2001 switch (gs->type) {
2002 case GREP_SOURCE_FILE:
2003 case GREP_SOURCE_OID:
2004 FREE_AND_NULL(gs->buf);
2005 gs->size = 0;
2006 break;
2007 case GREP_SOURCE_BUF:
2008 /* leave user-provided buf intact */
2009 break;
2010 }
2011 }
2012
2013 static int grep_source_load_oid(struct grep_source *gs)
2014 {
2015 enum object_type type;
2016
2017 grep_read_lock();
2018 gs->buf = read_sha1_file(gs->identifier, &type, &gs->size);
2019 grep_read_unlock();
2020
2021 if (!gs->buf)
2022 return error(_("'%s': unable to read %s"),
2023 gs->name,
2024 oid_to_hex(gs->identifier));
2025 return 0;
2026 }
2027
2028 static int grep_source_load_file(struct grep_source *gs)
2029 {
2030 const char *filename = gs->identifier;
2031 struct stat st;
2032 char *data;
2033 size_t size;
2034 int i;
2035
2036 if (lstat(filename, &st) < 0) {
2037 err_ret:
2038 if (errno != ENOENT)
2039 error_errno(_("failed to stat '%s'"), filename);
2040 return -1;
2041 }
2042 if (!S_ISREG(st.st_mode))
2043 return -1;
2044 size = xsize_t(st.st_size);
2045 i = open(filename, O_RDONLY);
2046 if (i < 0)
2047 goto err_ret;
2048 data = xmallocz(size);
2049 if (st.st_size != read_in_full(i, data, size)) {
2050 error_errno(_("'%s': short read"), filename);
2051 close(i);
2052 free(data);
2053 return -1;
2054 }
2055 close(i);
2056
2057 gs->buf = data;
2058 gs->size = size;
2059 return 0;
2060 }
2061
2062 static int grep_source_load(struct grep_source *gs)
2063 {
2064 if (gs->buf)
2065 return 0;
2066
2067 switch (gs->type) {
2068 case GREP_SOURCE_FILE:
2069 return grep_source_load_file(gs);
2070 case GREP_SOURCE_OID:
2071 return grep_source_load_oid(gs);
2072 case GREP_SOURCE_BUF:
2073 return gs->buf ? 0 : -1;
2074 }
2075 die("BUG: invalid grep_source type to load");
2076 }
2077
2078 void grep_source_load_driver(struct grep_source *gs)
2079 {
2080 if (gs->driver)
2081 return;
2082
2083 grep_attr_lock();
2084 if (gs->path)
2085 gs->driver = userdiff_find_by_path(gs->path);
2086 if (!gs->driver)
2087 gs->driver = userdiff_find_by_name("default");
2088 grep_attr_unlock();
2089 }
2090
2091 static int grep_source_is_binary(struct grep_source *gs)
2092 {
2093 grep_source_load_driver(gs);
2094 if (gs->driver->binary != -1)
2095 return gs->driver->binary;
2096
2097 if (!grep_source_load(gs))
2098 return buffer_is_binary(gs->buf, gs->size);
2099
2100 return 0;
2101 }