]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Builtin "git grep" | |
3 | * | |
4 | * Copyright (c) 2006 Junio C Hamano | |
5 | */ | |
6 | #include "cache.h" | |
7 | #include "blob.h" | |
8 | #include "tree.h" | |
9 | #include "commit.h" | |
10 | #include "tag.h" | |
11 | #include "tree-walk.h" | |
12 | #include "builtin.h" | |
13 | #include "grep.h" | |
14 | ||
15 | #ifndef NO_EXTERNAL_GREP | |
16 | #ifdef __unix__ | |
17 | #define NO_EXTERNAL_GREP 0 | |
18 | #else | |
19 | #define NO_EXTERNAL_GREP 1 | |
20 | #endif | |
21 | #endif | |
22 | ||
23 | /* | |
24 | * git grep pathspecs are somewhat different from diff-tree pathspecs; | |
25 | * pathname wildcards are allowed. | |
26 | */ | |
27 | static int pathspec_matches(const char **paths, const char *name) | |
28 | { | |
29 | int namelen, i; | |
30 | if (!paths || !*paths) | |
31 | return 1; | |
32 | namelen = strlen(name); | |
33 | for (i = 0; paths[i]; i++) { | |
34 | const char *match = paths[i]; | |
35 | int matchlen = strlen(match); | |
36 | const char *cp, *meta; | |
37 | ||
38 | if (!matchlen || | |
39 | ((matchlen <= namelen) && | |
40 | !strncmp(name, match, matchlen) && | |
41 | (match[matchlen-1] == '/' || | |
42 | name[matchlen] == '\0' || name[matchlen] == '/'))) | |
43 | return 1; | |
44 | if (!fnmatch(match, name, 0)) | |
45 | return 1; | |
46 | if (name[namelen-1] != '/') | |
47 | continue; | |
48 | ||
49 | /* We are being asked if the directory ("name") is worth | |
50 | * descending into. | |
51 | * | |
52 | * Find the longest leading directory name that does | |
53 | * not have metacharacter in the pathspec; the name | |
54 | * we are looking at must overlap with that directory. | |
55 | */ | |
56 | for (cp = match, meta = NULL; cp - match < matchlen; cp++) { | |
57 | char ch = *cp; | |
58 | if (ch == '*' || ch == '[' || ch == '?') { | |
59 | meta = cp; | |
60 | break; | |
61 | } | |
62 | } | |
63 | if (!meta) | |
64 | meta = cp; /* fully literal */ | |
65 | ||
66 | if (namelen <= meta - match) { | |
67 | /* Looking at "Documentation/" and | |
68 | * the pattern says "Documentation/howto/", or | |
69 | * "Documentation/diff*.txt". The name we | |
70 | * have should match prefix. | |
71 | */ | |
72 | if (!memcmp(match, name, namelen)) | |
73 | return 1; | |
74 | continue; | |
75 | } | |
76 | ||
77 | if (meta - match < namelen) { | |
78 | /* Looking at "Documentation/howto/" and | |
79 | * the pattern says "Documentation/h*"; | |
80 | * match up to "Do.../h"; this avoids descending | |
81 | * into "Documentation/technical/". | |
82 | */ | |
83 | if (!memcmp(match, name, meta - match)) | |
84 | return 1; | |
85 | continue; | |
86 | } | |
87 | } | |
88 | return 0; | |
89 | } | |
90 | ||
91 | static int grep_sha1(struct grep_opt *opt, const unsigned char *sha1, const char *name, int tree_name_len) | |
92 | { | |
93 | unsigned long size; | |
94 | char *data; | |
95 | enum object_type type; | |
96 | char *to_free = NULL; | |
97 | int hit; | |
98 | ||
99 | data = read_sha1_file(sha1, &type, &size); | |
100 | if (!data) { | |
101 | error("'%s': unable to read %s", name, sha1_to_hex(sha1)); | |
102 | return 0; | |
103 | } | |
104 | if (opt->relative && opt->prefix_length) { | |
105 | static char name_buf[PATH_MAX]; | |
106 | char *cp; | |
107 | int name_len = strlen(name) - opt->prefix_length + 1; | |
108 | ||
109 | if (!tree_name_len) | |
110 | name += opt->prefix_length; | |
111 | else { | |
112 | if (ARRAY_SIZE(name_buf) <= name_len) | |
113 | cp = to_free = xmalloc(name_len); | |
114 | else | |
115 | cp = name_buf; | |
116 | memcpy(cp, name, tree_name_len); | |
117 | strcpy(cp + tree_name_len, | |
118 | name + tree_name_len + opt->prefix_length); | |
119 | name = cp; | |
120 | } | |
121 | } | |
122 | hit = grep_buffer(opt, name, data, size); | |
123 | free(data); | |
124 | free(to_free); | |
125 | return hit; | |
126 | } | |
127 | ||
128 | static int grep_file(struct grep_opt *opt, const char *filename) | |
129 | { | |
130 | struct stat st; | |
131 | int i; | |
132 | char *data; | |
133 | size_t sz; | |
134 | ||
135 | if (lstat(filename, &st) < 0) { | |
136 | err_ret: | |
137 | if (errno != ENOENT) | |
138 | error("'%s': %s", filename, strerror(errno)); | |
139 | return 0; | |
140 | } | |
141 | if (!st.st_size) | |
142 | return 0; /* empty file -- no grep hit */ | |
143 | if (!S_ISREG(st.st_mode)) | |
144 | return 0; | |
145 | sz = xsize_t(st.st_size); | |
146 | i = open(filename, O_RDONLY); | |
147 | if (i < 0) | |
148 | goto err_ret; | |
149 | data = xmalloc(sz + 1); | |
150 | if (st.st_size != read_in_full(i, data, sz)) { | |
151 | error("'%s': short read %s", filename, strerror(errno)); | |
152 | close(i); | |
153 | free(data); | |
154 | return 0; | |
155 | } | |
156 | close(i); | |
157 | if (opt->relative && opt->prefix_length) | |
158 | filename += opt->prefix_length; | |
159 | i = grep_buffer(opt, filename, data, sz); | |
160 | free(data); | |
161 | return i; | |
162 | } | |
163 | ||
164 | #if !NO_EXTERNAL_GREP | |
165 | static int exec_grep(int argc, const char **argv) | |
166 | { | |
167 | pid_t pid; | |
168 | int status; | |
169 | ||
170 | argv[argc] = NULL; | |
171 | pid = fork(); | |
172 | if (pid < 0) | |
173 | return pid; | |
174 | if (!pid) { | |
175 | execvp("grep", (char **) argv); | |
176 | exit(255); | |
177 | } | |
178 | while (waitpid(pid, &status, 0) < 0) { | |
179 | if (errno == EINTR) | |
180 | continue; | |
181 | return -1; | |
182 | } | |
183 | if (WIFEXITED(status)) { | |
184 | if (!WEXITSTATUS(status)) | |
185 | return 1; | |
186 | return 0; | |
187 | } | |
188 | return -1; | |
189 | } | |
190 | ||
191 | #define MAXARGS 1000 | |
192 | #define ARGBUF 4096 | |
193 | #define push_arg(a) do { \ | |
194 | if (nr < MAXARGS) argv[nr++] = (a); \ | |
195 | else die("maximum number of args exceeded"); \ | |
196 | } while (0) | |
197 | ||
198 | /* | |
199 | * If you send a singleton filename to grep, it does not give | |
200 | * the name of the file. GNU grep has "-H" but we would want | |
201 | * that behaviour in a portable way. | |
202 | * | |
203 | * So we keep two pathnames in argv buffer unsent to grep in | |
204 | * the main loop if we need to do more than one grep. | |
205 | */ | |
206 | static int flush_grep(struct grep_opt *opt, | |
207 | int argc, int arg0, const char **argv, int *kept) | |
208 | { | |
209 | int status; | |
210 | int count = argc - arg0; | |
211 | const char *kept_0 = NULL; | |
212 | ||
213 | if (count <= 2) { | |
214 | /* | |
215 | * Because we keep at least 2 paths in the call from | |
216 | * the main loop (i.e. kept != NULL), and MAXARGS is | |
217 | * far greater than 2, this usually is a call to | |
218 | * conclude the grep. However, the user could attempt | |
219 | * to overflow the argv buffer by giving too many | |
220 | * options to leave very small number of real | |
221 | * arguments even for the call in the main loop. | |
222 | */ | |
223 | if (kept) | |
224 | die("insanely many options to grep"); | |
225 | ||
226 | /* | |
227 | * If we have two or more paths, we do not have to do | |
228 | * anything special, but we need to push /dev/null to | |
229 | * get "-H" behaviour of GNU grep portably but when we | |
230 | * are not doing "-l" nor "-L" nor "-c". | |
231 | */ | |
232 | if (count == 1 && | |
233 | !opt->name_only && | |
234 | !opt->unmatch_name_only && | |
235 | !opt->count) { | |
236 | argv[argc++] = "/dev/null"; | |
237 | argv[argc] = NULL; | |
238 | } | |
239 | } | |
240 | ||
241 | else if (kept) { | |
242 | /* | |
243 | * Called because we found many paths and haven't finished | |
244 | * iterating over the cache yet. We keep two paths | |
245 | * for the concluding call. argv[argc-2] and argv[argc-1] | |
246 | * has the last two paths, so save the first one away, | |
247 | * replace it with NULL while sending the list to grep, | |
248 | * and recover them after we are done. | |
249 | */ | |
250 | *kept = 2; | |
251 | kept_0 = argv[argc-2]; | |
252 | argv[argc-2] = NULL; | |
253 | argc -= 2; | |
254 | } | |
255 | ||
256 | status = exec_grep(argc, argv); | |
257 | ||
258 | if (kept_0) { | |
259 | /* | |
260 | * Then recover them. Now the last arg is beyond the | |
261 | * terminating NULL which is at argc, and the second | |
262 | * from the last is what we saved away in kept_0 | |
263 | */ | |
264 | argv[arg0++] = kept_0; | |
265 | argv[arg0] = argv[argc+1]; | |
266 | } | |
267 | return status; | |
268 | } | |
269 | ||
270 | static int external_grep(struct grep_opt *opt, const char **paths, int cached) | |
271 | { | |
272 | int i, nr, argc, hit, len, status; | |
273 | const char *argv[MAXARGS+1]; | |
274 | char randarg[ARGBUF]; | |
275 | char *argptr = randarg; | |
276 | struct grep_pat *p; | |
277 | ||
278 | if (opt->extended || (opt->relative && opt->prefix_length)) | |
279 | return -1; | |
280 | len = nr = 0; | |
281 | push_arg("grep"); | |
282 | if (opt->fixed) | |
283 | push_arg("-F"); | |
284 | if (opt->linenum) | |
285 | push_arg("-n"); | |
286 | if (!opt->pathname) | |
287 | push_arg("-h"); | |
288 | if (opt->regflags & REG_EXTENDED) | |
289 | push_arg("-E"); | |
290 | if (opt->regflags & REG_ICASE) | |
291 | push_arg("-i"); | |
292 | if (opt->word_regexp) | |
293 | push_arg("-w"); | |
294 | if (opt->name_only) | |
295 | push_arg("-l"); | |
296 | if (opt->unmatch_name_only) | |
297 | push_arg("-L"); | |
298 | if (opt->count) | |
299 | push_arg("-c"); | |
300 | if (opt->post_context || opt->pre_context) { | |
301 | if (opt->post_context != opt->pre_context) { | |
302 | if (opt->pre_context) { | |
303 | push_arg("-B"); | |
304 | len += snprintf(argptr, sizeof(randarg)-len, | |
305 | "%u", opt->pre_context) + 1; | |
306 | if (sizeof(randarg) <= len) | |
307 | die("maximum length of args exceeded"); | |
308 | push_arg(argptr); | |
309 | argptr += len; | |
310 | } | |
311 | if (opt->post_context) { | |
312 | push_arg("-A"); | |
313 | len += snprintf(argptr, sizeof(randarg)-len, | |
314 | "%u", opt->post_context) + 1; | |
315 | if (sizeof(randarg) <= len) | |
316 | die("maximum length of args exceeded"); | |
317 | push_arg(argptr); | |
318 | argptr += len; | |
319 | } | |
320 | } | |
321 | else { | |
322 | push_arg("-C"); | |
323 | len += snprintf(argptr, sizeof(randarg)-len, | |
324 | "%u", opt->post_context) + 1; | |
325 | if (sizeof(randarg) <= len) | |
326 | die("maximum length of args exceeded"); | |
327 | push_arg(argptr); | |
328 | argptr += len; | |
329 | } | |
330 | } | |
331 | for (p = opt->pattern_list; p; p = p->next) { | |
332 | push_arg("-e"); | |
333 | push_arg(p->pattern); | |
334 | } | |
335 | ||
336 | hit = 0; | |
337 | argc = nr; | |
338 | for (i = 0; i < active_nr; i++) { | |
339 | struct cache_entry *ce = active_cache[i]; | |
340 | char *name; | |
341 | int kept; | |
342 | if (!S_ISREG(ce->ce_mode)) | |
343 | continue; | |
344 | if (!pathspec_matches(paths, ce->name)) | |
345 | continue; | |
346 | name = ce->name; | |
347 | if (name[0] == '-') { | |
348 | int len = ce_namelen(ce); | |
349 | name = xmalloc(len + 3); | |
350 | memcpy(name, "./", 2); | |
351 | memcpy(name + 2, ce->name, len + 1); | |
352 | } | |
353 | argv[argc++] = name; | |
354 | if (MAXARGS <= argc) { | |
355 | status = flush_grep(opt, argc, nr, argv, &kept); | |
356 | if (0 < status) | |
357 | hit = 1; | |
358 | argc = nr + kept; | |
359 | } | |
360 | if (ce_stage(ce)) { | |
361 | do { | |
362 | i++; | |
363 | } while (i < active_nr && | |
364 | !strcmp(ce->name, active_cache[i]->name)); | |
365 | i--; /* compensate for loop control */ | |
366 | } | |
367 | } | |
368 | if (argc > nr) { | |
369 | status = flush_grep(opt, argc, nr, argv, NULL); | |
370 | if (0 < status) | |
371 | hit = 1; | |
372 | } | |
373 | return hit; | |
374 | } | |
375 | #endif | |
376 | ||
377 | static int grep_cache(struct grep_opt *opt, const char **paths, int cached) | |
378 | { | |
379 | int hit = 0; | |
380 | int nr; | |
381 | read_cache(); | |
382 | ||
383 | #if !NO_EXTERNAL_GREP | |
384 | /* | |
385 | * Use the external "grep" command for the case where | |
386 | * we grep through the checked-out files. It tends to | |
387 | * be a lot more optimized | |
388 | */ | |
389 | if (!cached) { | |
390 | hit = external_grep(opt, paths, cached); | |
391 | if (hit >= 0) | |
392 | return hit; | |
393 | } | |
394 | #endif | |
395 | ||
396 | for (nr = 0; nr < active_nr; nr++) { | |
397 | struct cache_entry *ce = active_cache[nr]; | |
398 | if (!S_ISREG(ce->ce_mode)) | |
399 | continue; | |
400 | if (!pathspec_matches(paths, ce->name)) | |
401 | continue; | |
402 | if (cached) { | |
403 | if (ce_stage(ce)) | |
404 | continue; | |
405 | hit |= grep_sha1(opt, ce->sha1, ce->name, 0); | |
406 | } | |
407 | else | |
408 | hit |= grep_file(opt, ce->name); | |
409 | if (ce_stage(ce)) { | |
410 | do { | |
411 | nr++; | |
412 | } while (nr < active_nr && | |
413 | !strcmp(ce->name, active_cache[nr]->name)); | |
414 | nr--; /* compensate for loop control */ | |
415 | } | |
416 | } | |
417 | free_grep_patterns(opt); | |
418 | return hit; | |
419 | } | |
420 | ||
421 | static int grep_tree(struct grep_opt *opt, const char **paths, | |
422 | struct tree_desc *tree, | |
423 | const char *tree_name, const char *base) | |
424 | { | |
425 | int len; | |
426 | int hit = 0; | |
427 | struct name_entry entry; | |
428 | char *down; | |
429 | int tn_len = strlen(tree_name); | |
430 | char *path_buf = xmalloc(PATH_MAX + tn_len + 100); | |
431 | ||
432 | if (tn_len) { | |
433 | tn_len = sprintf(path_buf, "%s:", tree_name); | |
434 | down = path_buf + tn_len; | |
435 | strcat(down, base); | |
436 | } | |
437 | else { | |
438 | down = path_buf; | |
439 | strcpy(down, base); | |
440 | } | |
441 | len = strlen(path_buf); | |
442 | ||
443 | while (tree_entry(tree, &entry)) { | |
444 | strcpy(path_buf + len, entry.path); | |
445 | ||
446 | if (S_ISDIR(entry.mode)) | |
447 | /* Match "abc/" against pathspec to | |
448 | * decide if we want to descend into "abc" | |
449 | * directory. | |
450 | */ | |
451 | strcpy(path_buf + len + tree_entry_len(entry.path, entry.sha1), "/"); | |
452 | ||
453 | if (!pathspec_matches(paths, down)) | |
454 | ; | |
455 | else if (S_ISREG(entry.mode)) | |
456 | hit |= grep_sha1(opt, entry.sha1, path_buf, tn_len); | |
457 | else if (S_ISDIR(entry.mode)) { | |
458 | enum object_type type; | |
459 | struct tree_desc sub; | |
460 | void *data; | |
461 | unsigned long size; | |
462 | ||
463 | data = read_sha1_file(entry.sha1, &type, &size); | |
464 | if (!data) | |
465 | die("unable to read tree (%s)", | |
466 | sha1_to_hex(entry.sha1)); | |
467 | init_tree_desc(&sub, data, size); | |
468 | hit |= grep_tree(opt, paths, &sub, tree_name, down); | |
469 | free(data); | |
470 | } | |
471 | } | |
472 | return hit; | |
473 | } | |
474 | ||
475 | static int grep_object(struct grep_opt *opt, const char **paths, | |
476 | struct object *obj, const char *name) | |
477 | { | |
478 | if (obj->type == OBJ_BLOB) | |
479 | return grep_sha1(opt, obj->sha1, name, 0); | |
480 | if (obj->type == OBJ_COMMIT || obj->type == OBJ_TREE) { | |
481 | struct tree_desc tree; | |
482 | void *data; | |
483 | unsigned long size; | |
484 | int hit; | |
485 | data = read_object_with_reference(obj->sha1, tree_type, | |
486 | &size, NULL); | |
487 | if (!data) | |
488 | die("unable to read tree (%s)", sha1_to_hex(obj->sha1)); | |
489 | init_tree_desc(&tree, data, size); | |
490 | hit = grep_tree(opt, paths, &tree, name, ""); | |
491 | free(data); | |
492 | return hit; | |
493 | } | |
494 | die("unable to grep from object of type %s", typename(obj->type)); | |
495 | } | |
496 | ||
497 | static const char builtin_grep_usage[] = | |
498 | "git-grep <option>* <rev>* [-e] <pattern> [<path>...]"; | |
499 | ||
500 | static const char emsg_invalid_context_len[] = | |
501 | "%s: invalid context length argument"; | |
502 | static const char emsg_missing_context_len[] = | |
503 | "missing context length argument"; | |
504 | static const char emsg_missing_argument[] = | |
505 | "option requires an argument -%s"; | |
506 | ||
507 | int cmd_grep(int argc, const char **argv, const char *prefix) | |
508 | { | |
509 | int hit = 0; | |
510 | int cached = 0; | |
511 | int seen_dashdash = 0; | |
512 | struct grep_opt opt; | |
513 | struct object_array list = { 0, 0, NULL }; | |
514 | const char **paths = NULL; | |
515 | int i; | |
516 | ||
517 | memset(&opt, 0, sizeof(opt)); | |
518 | opt.prefix_length = (prefix && *prefix) ? strlen(prefix) : 0; | |
519 | opt.relative = 1; | |
520 | opt.pathname = 1; | |
521 | opt.pattern_tail = &opt.pattern_list; | |
522 | opt.regflags = REG_NEWLINE; | |
523 | ||
524 | /* | |
525 | * If there is no -- then the paths must exist in the working | |
526 | * tree. If there is no explicit pattern specified with -e or | |
527 | * -f, we take the first unrecognized non option to be the | |
528 | * pattern, but then what follows it must be zero or more | |
529 | * valid refs up to the -- (if exists), and then existing | |
530 | * paths. If there is an explicit pattern, then the first | |
531 | * unrecognized non option is the beginning of the refs list | |
532 | * that continues up to the -- (if exists), and then paths. | |
533 | */ | |
534 | ||
535 | while (1 < argc) { | |
536 | const char *arg = argv[1]; | |
537 | argc--; argv++; | |
538 | if (!strcmp("--cached", arg)) { | |
539 | cached = 1; | |
540 | continue; | |
541 | } | |
542 | if (!strcmp("-a", arg) || | |
543 | !strcmp("--text", arg)) { | |
544 | opt.binary = GREP_BINARY_TEXT; | |
545 | continue; | |
546 | } | |
547 | if (!strcmp("-i", arg) || | |
548 | !strcmp("--ignore-case", arg)) { | |
549 | opt.regflags |= REG_ICASE; | |
550 | continue; | |
551 | } | |
552 | if (!strcmp("-I", arg)) { | |
553 | opt.binary = GREP_BINARY_NOMATCH; | |
554 | continue; | |
555 | } | |
556 | if (!strcmp("-v", arg) || | |
557 | !strcmp("--invert-match", arg)) { | |
558 | opt.invert = 1; | |
559 | continue; | |
560 | } | |
561 | if (!strcmp("-E", arg) || | |
562 | !strcmp("--extended-regexp", arg)) { | |
563 | opt.regflags |= REG_EXTENDED; | |
564 | continue; | |
565 | } | |
566 | if (!strcmp("-F", arg) || | |
567 | !strcmp("--fixed-strings", arg)) { | |
568 | opt.fixed = 1; | |
569 | continue; | |
570 | } | |
571 | if (!strcmp("-G", arg) || | |
572 | !strcmp("--basic-regexp", arg)) { | |
573 | opt.regflags &= ~REG_EXTENDED; | |
574 | continue; | |
575 | } | |
576 | if (!strcmp("-n", arg)) { | |
577 | opt.linenum = 1; | |
578 | continue; | |
579 | } | |
580 | if (!strcmp("-h", arg)) { | |
581 | opt.pathname = 0; | |
582 | continue; | |
583 | } | |
584 | if (!strcmp("-H", arg)) { | |
585 | opt.pathname = 1; | |
586 | continue; | |
587 | } | |
588 | if (!strcmp("-l", arg) || | |
589 | !strcmp("--name-only", arg) || | |
590 | !strcmp("--files-with-matches", arg)) { | |
591 | opt.name_only = 1; | |
592 | continue; | |
593 | } | |
594 | if (!strcmp("-L", arg) || | |
595 | !strcmp("--files-without-match", arg)) { | |
596 | opt.unmatch_name_only = 1; | |
597 | continue; | |
598 | } | |
599 | if (!strcmp("-c", arg) || | |
600 | !strcmp("--count", arg)) { | |
601 | opt.count = 1; | |
602 | continue; | |
603 | } | |
604 | if (!strcmp("-w", arg) || | |
605 | !strcmp("--word-regexp", arg)) { | |
606 | opt.word_regexp = 1; | |
607 | continue; | |
608 | } | |
609 | if (!prefixcmp(arg, "-A") || | |
610 | !prefixcmp(arg, "-B") || | |
611 | !prefixcmp(arg, "-C") || | |
612 | (arg[0] == '-' && '1' <= arg[1] && arg[1] <= '9')) { | |
613 | unsigned num; | |
614 | const char *scan; | |
615 | switch (arg[1]) { | |
616 | case 'A': case 'B': case 'C': | |
617 | if (!arg[2]) { | |
618 | if (argc <= 1) | |
619 | die(emsg_missing_context_len); | |
620 | scan = *++argv; | |
621 | argc--; | |
622 | } | |
623 | else | |
624 | scan = arg + 2; | |
625 | break; | |
626 | default: | |
627 | scan = arg + 1; | |
628 | break; | |
629 | } | |
630 | if (strtoul_ui(scan, 10, &num)) | |
631 | die(emsg_invalid_context_len, scan); | |
632 | switch (arg[1]) { | |
633 | case 'A': | |
634 | opt.post_context = num; | |
635 | break; | |
636 | default: | |
637 | case 'C': | |
638 | opt.post_context = num; | |
639 | case 'B': | |
640 | opt.pre_context = num; | |
641 | break; | |
642 | } | |
643 | continue; | |
644 | } | |
645 | if (!strcmp("-f", arg)) { | |
646 | FILE *patterns; | |
647 | int lno = 0; | |
648 | char buf[1024]; | |
649 | if (argc <= 1) | |
650 | die(emsg_missing_argument, arg); | |
651 | patterns = fopen(argv[1], "r"); | |
652 | if (!patterns) | |
653 | die("'%s': %s", argv[1], strerror(errno)); | |
654 | while (fgets(buf, sizeof(buf), patterns)) { | |
655 | int len = strlen(buf); | |
656 | if (len && buf[len-1] == '\n') | |
657 | buf[len-1] = 0; | |
658 | /* ignore empty line like grep does */ | |
659 | if (!buf[0]) | |
660 | continue; | |
661 | append_grep_pattern(&opt, xstrdup(buf), | |
662 | argv[1], ++lno, | |
663 | GREP_PATTERN); | |
664 | } | |
665 | fclose(patterns); | |
666 | argv++; | |
667 | argc--; | |
668 | continue; | |
669 | } | |
670 | if (!strcmp("--not", arg)) { | |
671 | append_grep_pattern(&opt, arg, "command line", 0, | |
672 | GREP_NOT); | |
673 | continue; | |
674 | } | |
675 | if (!strcmp("--and", arg)) { | |
676 | append_grep_pattern(&opt, arg, "command line", 0, | |
677 | GREP_AND); | |
678 | continue; | |
679 | } | |
680 | if (!strcmp("--or", arg)) | |
681 | continue; /* no-op */ | |
682 | if (!strcmp("(", arg)) { | |
683 | append_grep_pattern(&opt, arg, "command line", 0, | |
684 | GREP_OPEN_PAREN); | |
685 | continue; | |
686 | } | |
687 | if (!strcmp(")", arg)) { | |
688 | append_grep_pattern(&opt, arg, "command line", 0, | |
689 | GREP_CLOSE_PAREN); | |
690 | continue; | |
691 | } | |
692 | if (!strcmp("--all-match", arg)) { | |
693 | opt.all_match = 1; | |
694 | continue; | |
695 | } | |
696 | if (!strcmp("-e", arg)) { | |
697 | if (1 < argc) { | |
698 | append_grep_pattern(&opt, argv[1], | |
699 | "-e option", 0, | |
700 | GREP_PATTERN); | |
701 | argv++; | |
702 | argc--; | |
703 | continue; | |
704 | } | |
705 | die(emsg_missing_argument, arg); | |
706 | } | |
707 | if (!strcmp("--full-name", arg)) { | |
708 | opt.relative = 0; | |
709 | continue; | |
710 | } | |
711 | if (!strcmp("--", arg)) { | |
712 | /* later processing wants to have this at argv[1] */ | |
713 | argv--; | |
714 | argc++; | |
715 | break; | |
716 | } | |
717 | if (*arg == '-') | |
718 | usage(builtin_grep_usage); | |
719 | ||
720 | /* First unrecognized non-option token */ | |
721 | if (!opt.pattern_list) { | |
722 | append_grep_pattern(&opt, arg, "command line", 0, | |
723 | GREP_PATTERN); | |
724 | break; | |
725 | } | |
726 | else { | |
727 | /* We are looking at the first path or rev; | |
728 | * it is found at argv[1] after leaving the | |
729 | * loop. | |
730 | */ | |
731 | argc++; argv--; | |
732 | break; | |
733 | } | |
734 | } | |
735 | ||
736 | if (!opt.pattern_list) | |
737 | die("no pattern given."); | |
738 | if ((opt.regflags != REG_NEWLINE) && opt.fixed) | |
739 | die("cannot mix --fixed-strings and regexp"); | |
740 | compile_grep_patterns(&opt); | |
741 | ||
742 | /* Check revs and then paths */ | |
743 | for (i = 1; i < argc; i++) { | |
744 | const char *arg = argv[i]; | |
745 | unsigned char sha1[20]; | |
746 | /* Is it a rev? */ | |
747 | if (!get_sha1(arg, sha1)) { | |
748 | struct object *object = parse_object(sha1); | |
749 | if (!object) | |
750 | die("bad object %s", arg); | |
751 | add_object_array(object, arg, &list); | |
752 | continue; | |
753 | } | |
754 | if (!strcmp(arg, "--")) { | |
755 | i++; | |
756 | seen_dashdash = 1; | |
757 | } | |
758 | break; | |
759 | } | |
760 | ||
761 | /* The rest are paths */ | |
762 | if (!seen_dashdash) { | |
763 | int j; | |
764 | for (j = i; j < argc; j++) | |
765 | verify_filename(prefix, argv[j]); | |
766 | } | |
767 | ||
768 | if (i < argc) { | |
769 | paths = get_pathspec(prefix, argv + i); | |
770 | if (opt.prefix_length && opt.relative) { | |
771 | /* Make sure we do not get outside of paths */ | |
772 | for (i = 0; paths[i]; i++) | |
773 | if (strncmp(prefix, paths[i], opt.prefix_length)) | |
774 | die("git-grep: cannot generate relative filenames containing '..'"); | |
775 | } | |
776 | } | |
777 | else if (prefix) { | |
778 | paths = xcalloc(2, sizeof(const char *)); | |
779 | paths[0] = prefix; | |
780 | paths[1] = NULL; | |
781 | } | |
782 | ||
783 | if (!list.nr) | |
784 | return !grep_cache(&opt, paths, cached); | |
785 | ||
786 | if (cached) | |
787 | die("both --cached and trees are given."); | |
788 | ||
789 | for (i = 0; i < list.nr; i++) { | |
790 | struct object *real_obj; | |
791 | real_obj = deref_tag(list.objects[i].item, NULL, 0); | |
792 | if (grep_object(&opt, paths, real_obj, list.objects[i].name)) | |
793 | hit = 1; | |
794 | } | |
795 | free_grep_patterns(&opt); | |
796 | return !hit; | |
797 | } |