Merge branch 'sb/diff-blobfind-pickaxe'
[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 "revision.h"
10 #include "cache-tree.h"
11 #include "unpack-trees.h"
12 #include "refs.h"
13 #include "submodule.h"
14 #include "dir.h"
15 #include "fsmonitor.h"
16
17 /*
18  * diff-files
19  */
20
21 /*
22  * Has the work tree entity been removed?
23  *
24  * Return 1 if it was removed from the work tree, 0 if an entity to be
25  * compared with the cache entry ce still exists (the latter includes
26  * the case where a directory that is not a submodule repository
27  * exists for ce that is a submodule -- it is a submodule that is not
28  * checked out).  Return negative for an error.
29  */
30 static int check_removed(const struct cache_entry *ce, struct stat *st)
31 {
32         if (lstat(ce->name, st) < 0) {
33                 if (!is_missing_file_error(errno))
34                         return -1;
35                 return 1;
36         }
37         if (has_symlink_leading_path(ce->name, ce_namelen(ce)))
38                 return 1;
39         if (S_ISDIR(st->st_mode)) {
40                 struct object_id sub;
41
42                 /*
43                  * If ce is already a gitlink, we can have a plain
44                  * directory (i.e. the submodule is not checked out),
45                  * or a checked out submodule.  Either case this is not
46                  * a case where something was removed from the work tree,
47                  * so we will return 0.
48                  *
49                  * Otherwise, if the directory is not a submodule
50                  * repository, that means ce which was a blob turned into
51                  * a directory --- the blob was removed!
52                  */
53                 if (!S_ISGITLINK(ce->ce_mode) &&
54                     resolve_gitlink_ref(ce->name, "HEAD", &sub))
55                         return 1;
56         }
57         return 0;
58 }
59
60 /*
61  * Has a file changed or has a submodule new commits or a dirty work tree?
62  *
63  * Return 1 when changes are detected, 0 otherwise. If the DIRTY_SUBMODULES
64  * option is set, the caller does not only want to know if a submodule is
65  * modified at all but wants to know all the conditions that are met (new
66  * commits, untracked content and/or modified content).
67  */
68 static int match_stat_with_submodule(struct diff_options *diffopt,
69                                      const struct cache_entry *ce,
70                                      struct stat *st, unsigned ce_option,
71                                      unsigned *dirty_submodule)
72 {
73         int changed = ce_match_stat(ce, st, ce_option);
74         if (S_ISGITLINK(ce->ce_mode)) {
75                 struct diff_flags orig_flags = diffopt->flags;
76                 if (!diffopt->flags.override_submodule_config)
77                         set_diffopt_flags_from_submodule_config(diffopt, ce->name);
78                 if (diffopt->flags.ignore_submodules)
79                         changed = 0;
80                 else if (!diffopt->flags.ignore_dirty_submodules &&
81                          (!changed || diffopt->flags.dirty_submodules))
82                         *dirty_submodule = is_submodule_modified(ce->name,
83                                                                  diffopt->flags.ignore_untracked_in_submodules);
84                 diffopt->flags = orig_flags;
85         }
86         return changed;
87 }
88
89 int run_diff_files(struct rev_info *revs, unsigned int option)
90 {
91         int entries, i;
92         int diff_unmerged_stage = revs->max_count;
93         unsigned ce_option = ((option & DIFF_RACY_IS_MODIFIED)
94                               ? CE_MATCH_RACY_IS_DIRTY : 0);
95
96         diff_set_mnemonic_prefix(&revs->diffopt, "i/", "w/");
97
98         if (diff_unmerged_stage < 0)
99                 diff_unmerged_stage = 2;
100         entries = active_nr;
101         for (i = 0; i < entries; i++) {
102                 unsigned int oldmode, newmode;
103                 struct cache_entry *ce = active_cache[i];
104                 int changed;
105                 unsigned dirty_submodule = 0;
106                 const struct object_id *old_oid, *new_oid;
107
108                 if (diff_can_quit_early(&revs->diffopt))
109                         break;
110
111                 if (!ce_path_match(ce, &revs->prune_data, NULL))
112                         continue;
113
114                 if (ce_stage(ce)) {
115                         struct combine_diff_path *dpath;
116                         struct diff_filepair *pair;
117                         unsigned int wt_mode = 0;
118                         int num_compare_stages = 0;
119                         size_t path_len;
120                         struct stat st;
121
122                         path_len = ce_namelen(ce);
123
124                         dpath = xmalloc(combine_diff_path_size(5, path_len));
125                         dpath->path = (char *) &(dpath->parent[5]);
126
127                         dpath->next = NULL;
128                         memcpy(dpath->path, ce->name, path_len);
129                         dpath->path[path_len] = '\0';
130                         oidclr(&dpath->oid);
131                         memset(&(dpath->parent[0]), 0,
132                                sizeof(struct combine_diff_parent)*5);
133
134                         changed = check_removed(ce, &st);
135                         if (!changed)
136                                 wt_mode = ce_mode_from_stat(ce, st.st_mode);
137                         else {
138                                 if (changed < 0) {
139                                         perror(ce->name);
140                                         continue;
141                                 }
142                                 wt_mode = 0;
143                         }
144                         dpath->mode = wt_mode;
145
146                         while (i < entries) {
147                                 struct cache_entry *nce = active_cache[i];
148                                 int stage;
149
150                                 if (strcmp(ce->name, nce->name))
151                                         break;
152
153                                 /* Stage #2 (ours) is the first parent,
154                                  * stage #3 (theirs) is the second.
155                                  */
156                                 stage = ce_stage(nce);
157                                 if (2 <= stage) {
158                                         int mode = nce->ce_mode;
159                                         num_compare_stages++;
160                                         oidcpy(&dpath->parent[stage - 2].oid,
161                                                &nce->oid);
162                                         dpath->parent[stage-2].mode = ce_mode_from_stat(nce, mode);
163                                         dpath->parent[stage-2].status =
164                                                 DIFF_STATUS_MODIFIED;
165                                 }
166
167                                 /* diff against the proper unmerged stage */
168                                 if (stage == diff_unmerged_stage)
169                                         ce = nce;
170                                 i++;
171                         }
172                         /*
173                          * Compensate for loop update
174                          */
175                         i--;
176
177                         if (revs->combine_merges && num_compare_stages == 2) {
178                                 show_combined_diff(dpath, 2,
179                                                    revs->dense_combined_merges,
180                                                    revs);
181                                 free(dpath);
182                                 continue;
183                         }
184                         FREE_AND_NULL(dpath);
185
186                         /*
187                          * Show the diff for the 'ce' if we found the one
188                          * from the desired stage.
189                          */
190                         pair = diff_unmerge(&revs->diffopt, ce->name);
191                         if (wt_mode)
192                                 pair->two->mode = wt_mode;
193                         if (ce_stage(ce) != diff_unmerged_stage)
194                                 continue;
195                 }
196
197                 if (ce_uptodate(ce) || ce_skip_worktree(ce))
198                         continue;
199
200                 /* If CE_VALID is set, don't look at workdir for file removal */
201                 if (ce->ce_flags & CE_VALID) {
202                         changed = 0;
203                         newmode = ce->ce_mode;
204                 } else {
205                         struct stat st;
206
207                         changed = check_removed(ce, &st);
208                         if (changed) {
209                                 if (changed < 0) {
210                                         perror(ce->name);
211                                         continue;
212                                 }
213                                 diff_addremove(&revs->diffopt, '-', ce->ce_mode,
214                                                &ce->oid,
215                                                !is_null_oid(&ce->oid),
216                                                ce->name, 0);
217                                 continue;
218                         } else if (revs->diffopt.ita_invisible_in_index &&
219                                    ce_intent_to_add(ce)) {
220                                 diff_addremove(&revs->diffopt, '+', ce->ce_mode,
221                                                the_hash_algo->empty_tree, 0,
222                                                ce->name, 0);
223                                 continue;
224                         }
225
226                         changed = match_stat_with_submodule(&revs->diffopt, ce, &st,
227                                                             ce_option, &dirty_submodule);
228                         newmode = ce_mode_from_stat(ce, st.st_mode);
229                 }
230
231                 if (!changed && !dirty_submodule) {
232                         ce_mark_uptodate(ce);
233                         mark_fsmonitor_valid(ce);
234                         if (!revs->diffopt.flags.find_copies_harder)
235                                 continue;
236                 }
237                 oldmode = ce->ce_mode;
238                 old_oid = &ce->oid;
239                 new_oid = changed ? &null_oid : &ce->oid;
240                 diff_change(&revs->diffopt, oldmode, newmode,
241                             old_oid, new_oid,
242                             !is_null_oid(old_oid),
243                             !is_null_oid(new_oid),
244                             ce->name, 0, dirty_submodule);
245
246         }
247         diffcore_std(&revs->diffopt);
248         diff_flush(&revs->diffopt);
249         return 0;
250 }
251
252 /*
253  * diff-index
254  */
255
256 /* A file entry went away or appeared */
257 static void diff_index_show_file(struct rev_info *revs,
258                                  const char *prefix,
259                                  const struct cache_entry *ce,
260                                  const struct object_id *oid, int oid_valid,
261                                  unsigned int mode,
262                                  unsigned dirty_submodule)
263 {
264         diff_addremove(&revs->diffopt, prefix[0], mode,
265                        oid, oid_valid, ce->name, dirty_submodule);
266 }
267
268 static int get_stat_data(const struct cache_entry *ce,
269                          const struct object_id **oidp,
270                          unsigned int *modep,
271                          int cached, int match_missing,
272                          unsigned *dirty_submodule, struct diff_options *diffopt)
273 {
274         const struct object_id *oid = &ce->oid;
275         unsigned int mode = ce->ce_mode;
276
277         if (!cached && !ce_uptodate(ce)) {
278                 int changed;
279                 struct stat st;
280                 changed = check_removed(ce, &st);
281                 if (changed < 0)
282                         return -1;
283                 else if (changed) {
284                         if (match_missing) {
285                                 *oidp = oid;
286                                 *modep = mode;
287                                 return 0;
288                         }
289                         return -1;
290                 }
291                 changed = match_stat_with_submodule(diffopt, ce, &st,
292                                                     0, dirty_submodule);
293                 if (changed) {
294                         mode = ce_mode_from_stat(ce, st.st_mode);
295                         oid = &null_oid;
296                 }
297         }
298
299         *oidp = oid;
300         *modep = mode;
301         return 0;
302 }
303
304 static void show_new_file(struct rev_info *revs,
305                           const struct cache_entry *new,
306                           int cached, int match_missing)
307 {
308         const struct object_id *oid;
309         unsigned int mode;
310         unsigned dirty_submodule = 0;
311
312         /*
313          * New file in the index: it might actually be different in
314          * the working tree.
315          */
316         if (get_stat_data(new, &oid, &mode, cached, match_missing,
317             &dirty_submodule, &revs->diffopt) < 0)
318                 return;
319
320         diff_index_show_file(revs, "+", new, oid, !is_null_oid(oid), mode, dirty_submodule);
321 }
322
323 static int show_modified(struct rev_info *revs,
324                          const struct cache_entry *old,
325                          const struct cache_entry *new,
326                          int report_missing,
327                          int cached, int match_missing)
328 {
329         unsigned int mode, oldmode;
330         const struct object_id *oid;
331         unsigned dirty_submodule = 0;
332
333         if (get_stat_data(new, &oid, &mode, cached, match_missing,
334                           &dirty_submodule, &revs->diffopt) < 0) {
335                 if (report_missing)
336                         diff_index_show_file(revs, "-", old,
337                                              &old->oid, 1, old->ce_mode,
338                                              0);
339                 return -1;
340         }
341
342         if (revs->combine_merges && !cached &&
343             (oidcmp(oid, &old->oid) || oidcmp(&old->oid, &new->oid))) {
344                 struct combine_diff_path *p;
345                 int pathlen = ce_namelen(new);
346
347                 p = xmalloc(combine_diff_path_size(2, pathlen));
348                 p->path = (char *) &p->parent[2];
349                 p->next = NULL;
350                 memcpy(p->path, new->name, pathlen);
351                 p->path[pathlen] = 0;
352                 p->mode = mode;
353                 oidclr(&p->oid);
354                 memset(p->parent, 0, 2 * sizeof(struct combine_diff_parent));
355                 p->parent[0].status = DIFF_STATUS_MODIFIED;
356                 p->parent[0].mode = new->ce_mode;
357                 oidcpy(&p->parent[0].oid, &new->oid);
358                 p->parent[1].status = DIFF_STATUS_MODIFIED;
359                 p->parent[1].mode = old->ce_mode;
360                 oidcpy(&p->parent[1].oid, &old->oid);
361                 show_combined_diff(p, 2, revs->dense_combined_merges, revs);
362                 free(p);
363                 return 0;
364         }
365
366         oldmode = old->ce_mode;
367         if (mode == oldmode && !oidcmp(oid, &old->oid) && !dirty_submodule &&
368             !revs->diffopt.flags.find_copies_harder)
369                 return 0;
370
371         diff_change(&revs->diffopt, oldmode, mode,
372                     &old->oid, oid, 1, !is_null_oid(oid),
373                     old->name, 0, dirty_submodule);
374         return 0;
375 }
376
377 /*
378  * This gets a mix of an existing index and a tree, one pathname entry
379  * at a time. The index entry may be a single stage-0 one, but it could
380  * also be multiple unmerged entries (in which case idx_pos/idx_nr will
381  * give you the position and number of entries in the index).
382  */
383 static void do_oneway_diff(struct unpack_trees_options *o,
384                            const struct cache_entry *idx,
385                            const struct cache_entry *tree)
386 {
387         struct rev_info *revs = o->unpack_data;
388         int match_missing, cached;
389
390         /* i-t-a entries do not actually exist in the index */
391         if (revs->diffopt.ita_invisible_in_index &&
392             idx && ce_intent_to_add(idx)) {
393                 idx = NULL;
394                 if (!tree)
395                         return; /* nothing to diff.. */
396         }
397
398         /* if the entry is not checked out, don't examine work tree */
399         cached = o->index_only ||
400                 (idx && ((idx->ce_flags & CE_VALID) || ce_skip_worktree(idx)));
401         /*
402          * Backward compatibility wart - "diff-index -m" does
403          * not mean "do not ignore merges", but "match_missing".
404          *
405          * But with the revision flag parsing, that's found in
406          * "!revs->ignore_merges".
407          */
408         match_missing = !revs->ignore_merges;
409
410         if (cached && idx && ce_stage(idx)) {
411                 struct diff_filepair *pair;
412                 pair = diff_unmerge(&revs->diffopt, idx->name);
413                 if (tree)
414                         fill_filespec(pair->one, &tree->oid, 1,
415                                       tree->ce_mode);
416                 return;
417         }
418
419         /*
420          * Something added to the tree?
421          */
422         if (!tree) {
423                 show_new_file(revs, idx, cached, match_missing);
424                 return;
425         }
426
427         /*
428          * Something removed from the tree?
429          */
430         if (!idx) {
431                 diff_index_show_file(revs, "-", tree, &tree->oid, 1,
432                                      tree->ce_mode, 0);
433                 return;
434         }
435
436         /* Show difference between old and new */
437         show_modified(revs, tree, idx, 1, cached, match_missing);
438 }
439
440 /*
441  * The unpack_trees() interface is designed for merging, so
442  * the different source entries are designed primarily for
443  * the source trees, with the old index being really mainly
444  * used for being replaced by the result.
445  *
446  * For diffing, the index is more important, and we only have a
447  * single tree.
448  *
449  * We're supposed to advance o->pos to skip what we have already processed.
450  *
451  * This wrapper makes it all more readable, and takes care of all
452  * the fairly complex unpack_trees() semantic requirements, including
453  * the skipping, the path matching, the type conflict cases etc.
454  */
455 static int oneway_diff(const struct cache_entry * const *src,
456                        struct unpack_trees_options *o)
457 {
458         const struct cache_entry *idx = src[0];
459         const struct cache_entry *tree = src[1];
460         struct rev_info *revs = o->unpack_data;
461
462         /*
463          * Unpack-trees generates a DF/conflict entry if
464          * there was a directory in the index and a tree
465          * in the tree. From a diff standpoint, that's a
466          * delete of the tree and a create of the file.
467          */
468         if (tree == o->df_conflict_entry)
469                 tree = NULL;
470
471         if (ce_path_match(idx ? idx : tree, &revs->prune_data, NULL)) {
472                 do_oneway_diff(o, idx, tree);
473                 if (diff_can_quit_early(&revs->diffopt)) {
474                         o->exiting_early = 1;
475                         return -1;
476                 }
477         }
478
479         return 0;
480 }
481
482 static int diff_cache(struct rev_info *revs,
483                       const struct object_id *tree_oid,
484                       const char *tree_name,
485                       int cached)
486 {
487         struct tree *tree;
488         struct tree_desc t;
489         struct unpack_trees_options opts;
490
491         tree = parse_tree_indirect(tree_oid);
492         if (!tree)
493                 return error("bad tree object %s",
494                              tree_name ? tree_name : oid_to_hex(tree_oid));
495         memset(&opts, 0, sizeof(opts));
496         opts.head_idx = 1;
497         opts.index_only = cached;
498         opts.diff_index_cached = (cached &&
499                                   !revs->diffopt.flags.find_copies_harder);
500         opts.merge = 1;
501         opts.fn = oneway_diff;
502         opts.unpack_data = revs;
503         opts.src_index = &the_index;
504         opts.dst_index = NULL;
505         opts.pathspec = &revs->diffopt.pathspec;
506         opts.pathspec->recursive = 1;
507
508         init_tree_desc(&t, tree->buffer, tree->size);
509         return unpack_trees(1, &t, &opts);
510 }
511
512 int run_diff_index(struct rev_info *revs, int cached)
513 {
514         struct object_array_entry *ent;
515
516         ent = revs->pending.objects;
517         if (diff_cache(revs, &ent->item->oid, ent->name, cached))
518                 exit(128);
519
520         diff_set_mnemonic_prefix(&revs->diffopt, "c/", cached ? "i/" : "w/");
521         diffcore_fix_diff_index(&revs->diffopt);
522         diffcore_std(&revs->diffopt);
523         diff_flush(&revs->diffopt);
524         return 0;
525 }
526
527 int do_diff_cache(const struct object_id *tree_oid, struct diff_options *opt)
528 {
529         struct rev_info revs;
530
531         init_revisions(&revs, NULL);
532         copy_pathspec(&revs.prune_data, &opt->pathspec);
533         revs.diffopt = *opt;
534
535         if (diff_cache(&revs, tree_oid, NULL, 1))
536                 exit(128);
537         return 0;
538 }
539
540 int index_differs_from(const char *def, const struct diff_flags *flags,
541                        int ita_invisible_in_index)
542 {
543         struct rev_info rev;
544         struct setup_revision_opt opt;
545
546         init_revisions(&rev, NULL);
547         memset(&opt, 0, sizeof(opt));
548         opt.def = def;
549         setup_revisions(0, NULL, &rev, &opt);
550         rev.diffopt.flags.quick = 1;
551         rev.diffopt.flags.exit_with_status = 1;
552         if (flags)
553                 diff_flags_or(&rev.diffopt.flags, flags);
554         rev.diffopt.ita_invisible_in_index = ita_invisible_in_index;
555         run_diff_index(&rev, 1);
556         object_array_clear(&rev.pending);
557         return (rev.diffopt.flags.has_changes != 0);
558 }