]> git.ipfire.org Git - thirdparty/git.git/blob - diff-lib.c
treewide: be explicit about dependence on trace.h & trace2.h
[thirdparty/git.git] / diff-lib.c
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "quote.h"
6 #include "commit.h"
7 #include "diff.h"
8 #include "diffcore.h"
9 #include "gettext.h"
10 #include "hex.h"
11 #include "revision.h"
12 #include "cache-tree.h"
13 #include "unpack-trees.h"
14 #include "refs.h"
15 #include "submodule.h"
16 #include "trace.h"
17 #include "dir.h"
18 #include "fsmonitor.h"
19 #include "commit-reach.h"
20
21 /*
22 * diff-files
23 */
24
25 /*
26 * Has the work tree entity been removed?
27 *
28 * Return 1 if it was removed from the work tree, 0 if an entity to be
29 * compared with the cache entry ce still exists (the latter includes
30 * the case where a directory that is not a submodule repository
31 * exists for ce that is a submodule -- it is a submodule that is not
32 * checked out). Return negative for an error.
33 */
34 static int check_removed(const struct index_state *istate, const struct cache_entry *ce, struct stat *st)
35 {
36 assert(is_fsmonitor_refreshed(istate));
37 if (!(ce->ce_flags & CE_FSMONITOR_VALID) && lstat(ce->name, st) < 0) {
38 if (!is_missing_file_error(errno))
39 return -1;
40 return 1;
41 }
42 if (has_symlink_leading_path(ce->name, ce_namelen(ce)))
43 return 1;
44 if (S_ISDIR(st->st_mode)) {
45 struct object_id sub;
46
47 /*
48 * If ce is already a gitlink, we can have a plain
49 * directory (i.e. the submodule is not checked out),
50 * or a checked out submodule. Either case this is not
51 * a case where something was removed from the work tree,
52 * so we will return 0.
53 *
54 * Otherwise, if the directory is not a submodule
55 * repository, that means ce which was a blob turned into
56 * a directory --- the blob was removed!
57 */
58 if (!S_ISGITLINK(ce->ce_mode) &&
59 resolve_gitlink_ref(ce->name, "HEAD", &sub))
60 return 1;
61 }
62 return 0;
63 }
64
65 /*
66 * Has a file changed or has a submodule new commits or a dirty work tree?
67 *
68 * Return 1 when changes are detected, 0 otherwise. If the DIRTY_SUBMODULES
69 * option is set, the caller does not only want to know if a submodule is
70 * modified at all but wants to know all the conditions that are met (new
71 * commits, untracked content and/or modified content).
72 */
73 static int match_stat_with_submodule(struct diff_options *diffopt,
74 const struct cache_entry *ce,
75 struct stat *st, unsigned ce_option,
76 unsigned *dirty_submodule)
77 {
78 int changed = ie_match_stat(diffopt->repo->index, ce, st, ce_option);
79 if (S_ISGITLINK(ce->ce_mode)) {
80 struct diff_flags orig_flags = diffopt->flags;
81 if (!diffopt->flags.override_submodule_config)
82 set_diffopt_flags_from_submodule_config(diffopt, ce->name);
83 if (diffopt->flags.ignore_submodules)
84 changed = 0;
85 else if (!diffopt->flags.ignore_dirty_submodules &&
86 (!changed || diffopt->flags.dirty_submodules))
87 *dirty_submodule = is_submodule_modified(ce->name,
88 diffopt->flags.ignore_untracked_in_submodules);
89 diffopt->flags = orig_flags;
90 }
91 return changed;
92 }
93
94 int run_diff_files(struct rev_info *revs, unsigned int option)
95 {
96 int entries, i;
97 int diff_unmerged_stage = revs->max_count;
98 unsigned ce_option = ((option & DIFF_RACY_IS_MODIFIED)
99 ? CE_MATCH_RACY_IS_DIRTY : 0);
100 uint64_t start = getnanotime();
101 struct index_state *istate = revs->diffopt.repo->index;
102
103 diff_set_mnemonic_prefix(&revs->diffopt, "i/", "w/");
104
105 refresh_fsmonitor(istate);
106
107 if (diff_unmerged_stage < 0)
108 diff_unmerged_stage = 2;
109 entries = istate->cache_nr;
110 for (i = 0; i < entries; i++) {
111 unsigned int oldmode, newmode;
112 struct cache_entry *ce = istate->cache[i];
113 int changed;
114 unsigned dirty_submodule = 0;
115 const struct object_id *old_oid, *new_oid;
116
117 if (diff_can_quit_early(&revs->diffopt))
118 break;
119
120 if (!ce_path_match(istate, ce, &revs->prune_data, NULL))
121 continue;
122
123 if (revs->diffopt.prefix &&
124 strncmp(ce->name, revs->diffopt.prefix, revs->diffopt.prefix_length))
125 continue;
126
127 if (ce_stage(ce)) {
128 struct combine_diff_path *dpath;
129 struct diff_filepair *pair;
130 unsigned int wt_mode = 0;
131 int num_compare_stages = 0;
132 size_t path_len;
133 struct stat st;
134
135 path_len = ce_namelen(ce);
136
137 dpath = xmalloc(combine_diff_path_size(5, path_len));
138 dpath->path = (char *) &(dpath->parent[5]);
139
140 dpath->next = NULL;
141 memcpy(dpath->path, ce->name, path_len);
142 dpath->path[path_len] = '\0';
143 oidclr(&dpath->oid);
144 memset(&(dpath->parent[0]), 0,
145 sizeof(struct combine_diff_parent)*5);
146
147 changed = check_removed(istate, ce, &st);
148 if (!changed)
149 wt_mode = ce_mode_from_stat(ce, st.st_mode);
150 else {
151 if (changed < 0) {
152 perror(ce->name);
153 continue;
154 }
155 wt_mode = 0;
156 }
157 dpath->mode = wt_mode;
158
159 while (i < entries) {
160 struct cache_entry *nce = istate->cache[i];
161 int stage;
162
163 if (strcmp(ce->name, nce->name))
164 break;
165
166 /* Stage #2 (ours) is the first parent,
167 * stage #3 (theirs) is the second.
168 */
169 stage = ce_stage(nce);
170 if (2 <= stage) {
171 int mode = nce->ce_mode;
172 num_compare_stages++;
173 oidcpy(&dpath->parent[stage - 2].oid,
174 &nce->oid);
175 dpath->parent[stage-2].mode = ce_mode_from_stat(nce, mode);
176 dpath->parent[stage-2].status =
177 DIFF_STATUS_MODIFIED;
178 }
179
180 /* diff against the proper unmerged stage */
181 if (stage == diff_unmerged_stage)
182 ce = nce;
183 i++;
184 }
185 /*
186 * Compensate for loop update
187 */
188 i--;
189
190 if (revs->combine_merges && num_compare_stages == 2) {
191 show_combined_diff(dpath, 2, revs);
192 free(dpath);
193 continue;
194 }
195 FREE_AND_NULL(dpath);
196
197 /*
198 * Show the diff for the 'ce' if we found the one
199 * from the desired stage.
200 */
201 pair = diff_unmerge(&revs->diffopt, ce->name);
202 if (wt_mode)
203 pair->two->mode = wt_mode;
204 if (ce_stage(ce) != diff_unmerged_stage)
205 continue;
206 }
207
208 if (ce_uptodate(ce) || ce_skip_worktree(ce))
209 continue;
210
211 /*
212 * When CE_VALID is set (via "update-index --assume-unchanged"
213 * or via adding paths while core.ignorestat is set to true),
214 * the user has promised that the working tree file for that
215 * path will not be modified. When CE_FSMONITOR_VALID is true,
216 * the fsmonitor knows that the path hasn't been modified since
217 * we refreshed the cached stat information. In either case,
218 * we do not have to stat to see if the path has been removed
219 * or modified.
220 */
221 if (ce->ce_flags & (CE_VALID | CE_FSMONITOR_VALID)) {
222 changed = 0;
223 newmode = ce->ce_mode;
224 } else {
225 struct stat st;
226
227 changed = check_removed(istate, ce, &st);
228 if (changed) {
229 if (changed < 0) {
230 perror(ce->name);
231 continue;
232 }
233 diff_addremove(&revs->diffopt, '-', ce->ce_mode,
234 &ce->oid,
235 !is_null_oid(&ce->oid),
236 ce->name, 0);
237 continue;
238 } else if (revs->diffopt.ita_invisible_in_index &&
239 ce_intent_to_add(ce)) {
240 newmode = ce_mode_from_stat(ce, st.st_mode);
241 diff_addremove(&revs->diffopt, '+', newmode,
242 null_oid(), 0, ce->name, 0);
243 continue;
244 }
245
246 changed = match_stat_with_submodule(&revs->diffopt, ce, &st,
247 ce_option, &dirty_submodule);
248 newmode = ce_mode_from_stat(ce, st.st_mode);
249 }
250
251 if (!changed && !dirty_submodule) {
252 ce_mark_uptodate(ce);
253 mark_fsmonitor_valid(istate, ce);
254 if (!revs->diffopt.flags.find_copies_harder)
255 continue;
256 }
257 oldmode = ce->ce_mode;
258 old_oid = &ce->oid;
259 new_oid = changed ? null_oid() : &ce->oid;
260 diff_change(&revs->diffopt, oldmode, newmode,
261 old_oid, new_oid,
262 !is_null_oid(old_oid),
263 !is_null_oid(new_oid),
264 ce->name, 0, dirty_submodule);
265
266 }
267 diffcore_std(&revs->diffopt);
268 diff_flush(&revs->diffopt);
269 trace_performance_since(start, "diff-files");
270 return 0;
271 }
272
273 /*
274 * diff-index
275 */
276
277 /* A file entry went away or appeared */
278 static void diff_index_show_file(struct rev_info *revs,
279 const char *prefix,
280 const struct cache_entry *ce,
281 const struct object_id *oid, int oid_valid,
282 unsigned int mode,
283 unsigned dirty_submodule)
284 {
285 diff_addremove(&revs->diffopt, prefix[0], mode,
286 oid, oid_valid, ce->name, dirty_submodule);
287 }
288
289 static int get_stat_data(const struct index_state *istate,
290 const struct cache_entry *ce,
291 const struct object_id **oidp,
292 unsigned int *modep,
293 int cached, int match_missing,
294 unsigned *dirty_submodule, struct diff_options *diffopt)
295 {
296 const struct object_id *oid = &ce->oid;
297 unsigned int mode = ce->ce_mode;
298
299 if (!cached && !ce_uptodate(ce)) {
300 int changed;
301 struct stat st;
302 changed = check_removed(istate, ce, &st);
303 if (changed < 0)
304 return -1;
305 else if (changed) {
306 if (match_missing) {
307 *oidp = oid;
308 *modep = mode;
309 return 0;
310 }
311 return -1;
312 }
313 changed = match_stat_with_submodule(diffopt, ce, &st,
314 0, dirty_submodule);
315 if (changed) {
316 mode = ce_mode_from_stat(ce, st.st_mode);
317 oid = null_oid();
318 }
319 }
320
321 *oidp = oid;
322 *modep = mode;
323 return 0;
324 }
325
326 static void show_new_file(struct rev_info *revs,
327 const struct cache_entry *new_file,
328 int cached, int match_missing)
329 {
330 const struct object_id *oid;
331 unsigned int mode;
332 unsigned dirty_submodule = 0;
333 struct index_state *istate = revs->diffopt.repo->index;
334
335 if (new_file && S_ISSPARSEDIR(new_file->ce_mode)) {
336 diff_tree_oid(NULL, &new_file->oid, new_file->name, &revs->diffopt);
337 return;
338 }
339
340 /*
341 * New file in the index: it might actually be different in
342 * the working tree.
343 */
344 if (get_stat_data(istate, new_file, &oid, &mode, cached, match_missing,
345 &dirty_submodule, &revs->diffopt) < 0)
346 return;
347
348 diff_index_show_file(revs, "+", new_file, oid, !is_null_oid(oid), mode, dirty_submodule);
349 }
350
351 static int show_modified(struct rev_info *revs,
352 const struct cache_entry *old_entry,
353 const struct cache_entry *new_entry,
354 int report_missing,
355 int cached, int match_missing)
356 {
357 unsigned int mode, oldmode;
358 const struct object_id *oid;
359 unsigned dirty_submodule = 0;
360 struct index_state *istate = revs->diffopt.repo->index;
361
362 assert(S_ISSPARSEDIR(old_entry->ce_mode) ==
363 S_ISSPARSEDIR(new_entry->ce_mode));
364
365 /*
366 * If both are sparse directory entries, then expand the
367 * modifications to the file level. If only one was a sparse
368 * directory, then they appear as an add and delete instead of
369 * a modification.
370 */
371 if (S_ISSPARSEDIR(new_entry->ce_mode)) {
372 diff_tree_oid(&old_entry->oid, &new_entry->oid, new_entry->name, &revs->diffopt);
373 return 0;
374 }
375
376 if (get_stat_data(istate, new_entry, &oid, &mode, cached, match_missing,
377 &dirty_submodule, &revs->diffopt) < 0) {
378 if (report_missing)
379 diff_index_show_file(revs, "-", old_entry,
380 &old_entry->oid, 1, old_entry->ce_mode,
381 0);
382 return -1;
383 }
384
385 if (revs->combine_merges && !cached &&
386 (!oideq(oid, &old_entry->oid) || !oideq(&old_entry->oid, &new_entry->oid))) {
387 struct combine_diff_path *p;
388 int pathlen = ce_namelen(new_entry);
389
390 p = xmalloc(combine_diff_path_size(2, pathlen));
391 p->path = (char *) &p->parent[2];
392 p->next = NULL;
393 memcpy(p->path, new_entry->name, pathlen);
394 p->path[pathlen] = 0;
395 p->mode = mode;
396 oidclr(&p->oid);
397 memset(p->parent, 0, 2 * sizeof(struct combine_diff_parent));
398 p->parent[0].status = DIFF_STATUS_MODIFIED;
399 p->parent[0].mode = new_entry->ce_mode;
400 oidcpy(&p->parent[0].oid, &new_entry->oid);
401 p->parent[1].status = DIFF_STATUS_MODIFIED;
402 p->parent[1].mode = old_entry->ce_mode;
403 oidcpy(&p->parent[1].oid, &old_entry->oid);
404 show_combined_diff(p, 2, revs);
405 free(p);
406 return 0;
407 }
408
409 oldmode = old_entry->ce_mode;
410 if (mode == oldmode && oideq(oid, &old_entry->oid) && !dirty_submodule &&
411 !revs->diffopt.flags.find_copies_harder)
412 return 0;
413
414 diff_change(&revs->diffopt, oldmode, mode,
415 &old_entry->oid, oid, 1, !is_null_oid(oid),
416 old_entry->name, 0, dirty_submodule);
417 return 0;
418 }
419
420 /*
421 * This gets a mix of an existing index and a tree, one pathname entry
422 * at a time. The index entry may be a single stage-0 one, but it could
423 * also be multiple unmerged entries (in which case idx_pos/idx_nr will
424 * give you the position and number of entries in the index).
425 */
426 static void do_oneway_diff(struct unpack_trees_options *o,
427 const struct cache_entry *idx,
428 const struct cache_entry *tree)
429 {
430 struct rev_info *revs = o->unpack_data;
431 int match_missing, cached;
432
433 /*
434 * i-t-a entries do not actually exist in the index (if we're
435 * looking at its content)
436 */
437 if (o->index_only &&
438 revs->diffopt.ita_invisible_in_index &&
439 idx && ce_intent_to_add(idx)) {
440 idx = NULL;
441 if (!tree)
442 return; /* nothing to diff.. */
443 }
444
445 /* if the entry is not checked out, don't examine work tree */
446 cached = o->index_only ||
447 (idx && ((idx->ce_flags & CE_VALID) || ce_skip_worktree(idx)));
448
449 match_missing = revs->match_missing;
450
451 if (cached && idx && ce_stage(idx)) {
452 struct diff_filepair *pair;
453 pair = diff_unmerge(&revs->diffopt, idx->name);
454 if (tree)
455 fill_filespec(pair->one, &tree->oid, 1,
456 tree->ce_mode);
457 return;
458 }
459
460 /*
461 * Something added to the tree?
462 */
463 if (!tree) {
464 show_new_file(revs, idx, cached, match_missing);
465 return;
466 }
467
468 /*
469 * Something removed from the tree?
470 */
471 if (!idx) {
472 if (S_ISSPARSEDIR(tree->ce_mode)) {
473 diff_tree_oid(&tree->oid, NULL, tree->name, &revs->diffopt);
474 return;
475 }
476
477 diff_index_show_file(revs, "-", tree, &tree->oid, 1,
478 tree->ce_mode, 0);
479 return;
480 }
481
482 /* Show difference between old and new */
483 show_modified(revs, tree, idx, 1, cached, match_missing);
484 }
485
486 /*
487 * The unpack_trees() interface is designed for merging, so
488 * the different source entries are designed primarily for
489 * the source trees, with the old index being really mainly
490 * used for being replaced by the result.
491 *
492 * For diffing, the index is more important, and we only have a
493 * single tree.
494 *
495 * We're supposed to advance o->pos to skip what we have already processed.
496 *
497 * This wrapper makes it all more readable, and takes care of all
498 * the fairly complex unpack_trees() semantic requirements, including
499 * the skipping, the path matching, the type conflict cases etc.
500 */
501 static int oneway_diff(const struct cache_entry * const *src,
502 struct unpack_trees_options *o)
503 {
504 const struct cache_entry *idx = src[0];
505 const struct cache_entry *tree = src[1];
506 struct rev_info *revs = o->unpack_data;
507
508 /*
509 * Unpack-trees generates a DF/conflict entry if
510 * there was a directory in the index and a tree
511 * in the tree. From a diff standpoint, that's a
512 * delete of the tree and a create of the file.
513 */
514 if (tree == o->df_conflict_entry)
515 tree = NULL;
516
517 if (ce_path_match(revs->diffopt.repo->index,
518 idx ? idx : tree,
519 &revs->prune_data, NULL)) {
520 do_oneway_diff(o, idx, tree);
521 if (diff_can_quit_early(&revs->diffopt)) {
522 o->exiting_early = 1;
523 return -1;
524 }
525 }
526
527 return 0;
528 }
529
530 static int diff_cache(struct rev_info *revs,
531 const struct object_id *tree_oid,
532 const char *tree_name,
533 int cached)
534 {
535 struct tree *tree;
536 struct tree_desc t;
537 struct unpack_trees_options opts;
538
539 tree = parse_tree_indirect(tree_oid);
540 if (!tree)
541 return error("bad tree object %s",
542 tree_name ? tree_name : oid_to_hex(tree_oid));
543 memset(&opts, 0, sizeof(opts));
544 opts.head_idx = 1;
545 opts.index_only = cached;
546 opts.diff_index_cached = (cached &&
547 !revs->diffopt.flags.find_copies_harder);
548 opts.merge = 1;
549 opts.fn = oneway_diff;
550 opts.unpack_data = revs;
551 opts.src_index = revs->diffopt.repo->index;
552 opts.dst_index = NULL;
553 opts.pathspec = &revs->diffopt.pathspec;
554 opts.pathspec->recursive = 1;
555
556 init_tree_desc(&t, tree->buffer, tree->size);
557 return unpack_trees(1, &t, &opts);
558 }
559
560 void diff_get_merge_base(const struct rev_info *revs, struct object_id *mb)
561 {
562 int i;
563 struct commit *mb_child[2] = {0};
564 struct commit_list *merge_bases;
565
566 for (i = 0; i < revs->pending.nr; i++) {
567 struct object *obj = revs->pending.objects[i].item;
568 if (obj->flags)
569 die(_("--merge-base does not work with ranges"));
570 if (obj->type != OBJ_COMMIT)
571 die(_("--merge-base only works with commits"));
572 }
573
574 /*
575 * This check must go after the for loop above because A...B
576 * ranges produce three pending commits, resulting in a
577 * misleading error message.
578 */
579 if (revs->pending.nr < 1 || revs->pending.nr > 2)
580 BUG("unexpected revs->pending.nr: %d", revs->pending.nr);
581
582 for (i = 0; i < revs->pending.nr; i++)
583 mb_child[i] = lookup_commit_reference(the_repository, &revs->pending.objects[i].item->oid);
584 if (revs->pending.nr == 1) {
585 struct object_id oid;
586
587 if (repo_get_oid(the_repository, "HEAD", &oid))
588 die(_("unable to get HEAD"));
589
590 mb_child[1] = lookup_commit_reference(the_repository, &oid);
591 }
592
593 merge_bases = repo_get_merge_bases(the_repository, mb_child[0], mb_child[1]);
594 if (!merge_bases)
595 die(_("no merge base found"));
596 if (merge_bases->next)
597 die(_("multiple merge bases found"));
598
599 oidcpy(mb, &merge_bases->item->object.oid);
600
601 free_commit_list(merge_bases);
602 }
603
604 int run_diff_index(struct rev_info *revs, unsigned int option)
605 {
606 struct object_array_entry *ent;
607 int cached = !!(option & DIFF_INDEX_CACHED);
608 int merge_base = !!(option & DIFF_INDEX_MERGE_BASE);
609 struct object_id oid;
610 const char *name;
611 char merge_base_hex[GIT_MAX_HEXSZ + 1];
612 struct index_state *istate = revs->diffopt.repo->index;
613
614 if (revs->pending.nr != 1)
615 BUG("run_diff_index must be passed exactly one tree");
616
617 trace_performance_enter();
618 ent = revs->pending.objects;
619
620 refresh_fsmonitor(istate);
621
622 if (merge_base) {
623 diff_get_merge_base(revs, &oid);
624 name = oid_to_hex_r(merge_base_hex, &oid);
625 } else {
626 oidcpy(&oid, &ent->item->oid);
627 name = ent->name;
628 }
629
630 if (diff_cache(revs, &oid, name, cached))
631 exit(128);
632
633 diff_set_mnemonic_prefix(&revs->diffopt, "c/", cached ? "i/" : "w/");
634 diffcore_fix_diff_index();
635 diffcore_std(&revs->diffopt);
636 diff_flush(&revs->diffopt);
637 trace_performance_leave("diff-index");
638 return 0;
639 }
640
641 int do_diff_cache(const struct object_id *tree_oid, struct diff_options *opt)
642 {
643 struct rev_info revs;
644
645 repo_init_revisions(opt->repo, &revs, NULL);
646 copy_pathspec(&revs.prune_data, &opt->pathspec);
647 diff_setup_done(&revs.diffopt);
648 revs.diffopt = *opt;
649
650 if (diff_cache(&revs, tree_oid, NULL, 1))
651 exit(128);
652 release_revisions(&revs);
653 return 0;
654 }
655
656 int index_differs_from(struct repository *r,
657 const char *def, const struct diff_flags *flags,
658 int ita_invisible_in_index)
659 {
660 struct rev_info rev;
661 struct setup_revision_opt opt;
662 unsigned has_changes;
663
664 repo_init_revisions(r, &rev, NULL);
665 memset(&opt, 0, sizeof(opt));
666 opt.def = def;
667 setup_revisions(0, NULL, &rev, &opt);
668 rev.diffopt.flags.quick = 1;
669 rev.diffopt.flags.exit_with_status = 1;
670 if (flags)
671 diff_flags_or(&rev.diffopt.flags, flags);
672 rev.diffopt.ita_invisible_in_index = ita_invisible_in_index;
673 run_diff_index(&rev, 1);
674 has_changes = rev.diffopt.flags.has_changes;
675 release_revisions(&rev);
676 return (has_changes != 0);
677 }
678
679 static struct strbuf *idiff_prefix_cb(struct diff_options *opt UNUSED, void *data)
680 {
681 return data;
682 }
683
684 void show_interdiff(const struct object_id *oid1, const struct object_id *oid2,
685 int indent, struct diff_options *diffopt)
686 {
687 struct diff_options opts;
688 struct strbuf prefix = STRBUF_INIT;
689
690 memcpy(&opts, diffopt, sizeof(opts));
691 opts.output_format = DIFF_FORMAT_PATCH;
692 opts.output_prefix = idiff_prefix_cb;
693 strbuf_addchars(&prefix, ' ', indent);
694 opts.output_prefix_data = &prefix;
695 diff_setup_done(&opts);
696
697 diff_tree_oid(oid1, oid2, "", &opts);
698 diffcore_std(&opts);
699 diff_flush(&opts);
700
701 strbuf_release(&prefix);
702 }