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
14/*
15 * diff-files
16 */
17
18/*
19 * Has the work tree entity been removed?
20 *
21 * Return 1 if it was removed from the work tree, 0 if an entity to be
22 * compared with the cache entry ce still exists (the latter includes
23 * the case where a directory that is not a submodule repository
24 * exists for ce that is a submodule -- it is a submodule that is not
25 * checked out). Return negative for an error.
26 */
27static int check_removed(const struct cache_entry *ce, struct stat *st)
28{
29 if (lstat(ce->name, st) < 0) {
30 if (errno != ENOENT && errno != ENOTDIR)
31 return -1;
32 return 1;
33 }
34 if (has_symlink_leading_path(ce_namelen(ce), ce->name))
35 return 1;
36 if (S_ISDIR(st->st_mode)) {
37 unsigned char sub[20];
38
39 /*
40 * If ce is already a gitlink, we can have a plain
41 * directory (i.e. the submodule is not checked out),
42 * or a checked out submodule. Either case this is not
43 * a case where something was removed from the work tree,
44 * so we will return 0.
45 *
46 * Otherwise, if the directory is not a submodule
47 * repository, that means ce which was a blob turned into
48 * a directory --- the blob was removed!
49 */
50 if (!S_ISGITLINK(ce->ce_mode) &&
51 resolve_gitlink_ref(ce->name, "HEAD", sub))
52 return 1;
53 }
54 return 0;
55}
56
57int run_diff_files(struct rev_info *revs, unsigned int option)
58{
59 int entries, i;
60 int diff_unmerged_stage = revs->max_count;
61 int silent_on_removed = option & DIFF_SILENT_ON_REMOVED;
62 unsigned ce_option = ((option & DIFF_RACY_IS_MODIFIED)
63 ? CE_MATCH_RACY_IS_DIRTY : 0);
64 char symcache[PATH_MAX];
65
66 if (diff_unmerged_stage < 0)
67 diff_unmerged_stage = 2;
68 entries = active_nr;
69 symcache[0] = '\0';
70 for (i = 0; i < entries; i++) {
71 struct stat st;
72 unsigned int oldmode, newmode;
73 struct cache_entry *ce = active_cache[i];
74 int changed;
75
76 if (DIFF_OPT_TST(&revs->diffopt, QUIET) &&
77 DIFF_OPT_TST(&revs->diffopt, HAS_CHANGES))
78 break;
79
80 if (!ce_path_match(ce, revs->prune_data))
81 continue;
82
83 if (ce_stage(ce)) {
84 struct combine_diff_path *dpath;
85 int num_compare_stages = 0;
86 size_t path_len;
87
88 path_len = ce_namelen(ce);
89
90 dpath = xmalloc(combine_diff_path_size(5, path_len));
91 dpath->path = (char *) &(dpath->parent[5]);
92
93 dpath->next = NULL;
94 dpath->len = path_len;
95 memcpy(dpath->path, ce->name, path_len);
96 dpath->path[path_len] = '\0';
97 hashclr(dpath->sha1);
98 memset(&(dpath->parent[0]), 0,
99 sizeof(struct combine_diff_parent)*5);
100
101 changed = check_removed(ce, &st);
102 if (!changed)
103 dpath->mode = ce_mode_from_stat(ce, st.st_mode);
104 else {
105 if (changed < 0) {
106 perror(ce->name);
107 continue;
108 }
109 if (silent_on_removed)
110 continue;
111 }
112
113 while (i < entries) {
114 struct cache_entry *nce = active_cache[i];
115 int stage;
116
117 if (strcmp(ce->name, nce->name))
118 break;
119
120 /* Stage #2 (ours) is the first parent,
121 * stage #3 (theirs) is the second.
122 */
123 stage = ce_stage(nce);
124 if (2 <= stage) {
125 int mode = nce->ce_mode;
126 num_compare_stages++;
127 hashcpy(dpath->parent[stage-2].sha1, nce->sha1);
128 dpath->parent[stage-2].mode = ce_mode_from_stat(nce, mode);
129 dpath->parent[stage-2].status =
130 DIFF_STATUS_MODIFIED;
131 }
132
133 /* diff against the proper unmerged stage */
134 if (stage == diff_unmerged_stage)
135 ce = nce;
136 i++;
137 }
138 /*
139 * Compensate for loop update
140 */
141 i--;
142
143 if (revs->combine_merges && num_compare_stages == 2) {
144 show_combined_diff(dpath, 2,
145 revs->dense_combined_merges,
146 revs);
147 free(dpath);
148 continue;
149 }
150 free(dpath);
151 dpath = NULL;
152
153 /*
154 * Show the diff for the 'ce' if we found the one
155 * from the desired stage.
156 */
157 diff_unmerge(&revs->diffopt, ce->name, 0, null_sha1);
158 if (ce_stage(ce) != diff_unmerged_stage)
159 continue;
160 }
161
162 if (ce_uptodate(ce))
163 continue;
164
165 changed = check_removed(ce, &st);
166 if (changed) {
167 if (changed < 0) {
168 perror(ce->name);
169 continue;
170 }
171 if (silent_on_removed)
172 continue;
173 diff_addremove(&revs->diffopt, '-', ce->ce_mode,
174 ce->sha1, ce->name, NULL);
175 continue;
176 }
177 changed = ce_match_stat(ce, &st, ce_option);
178 if (!changed) {
179 ce_mark_uptodate(ce);
180 if (!DIFF_OPT_TST(&revs->diffopt, FIND_COPIES_HARDER))
181 continue;
182 }
183 oldmode = ce->ce_mode;
184 newmode = ce_mode_from_stat(ce, st.st_mode);
185 diff_change(&revs->diffopt, oldmode, newmode,
186 ce->sha1, (changed ? null_sha1 : ce->sha1),
187 ce->name, NULL);
188
189 }
190 diffcore_std(&revs->diffopt);
191 diff_flush(&revs->diffopt);
192 return 0;
193}
194
195/*
196 * diff-index
197 */
198
199struct oneway_unpack_data {
200 struct rev_info *revs;
201 char symcache[PATH_MAX];
202};
203
204/* A file entry went away or appeared */
205static void diff_index_show_file(struct rev_info *revs,
206 const char *prefix,
207 struct cache_entry *ce,
208 const unsigned char *sha1, unsigned int mode)
209{
210 diff_addremove(&revs->diffopt, prefix[0], mode,
211 sha1, ce->name, NULL);
212}
213
214static int get_stat_data(struct cache_entry *ce,
215 const unsigned char **sha1p,
216 unsigned int *modep,
217 int cached, int match_missing,
218 struct oneway_unpack_data *cbdata)
219{
220 const unsigned char *sha1 = ce->sha1;
221 unsigned int mode = ce->ce_mode;
222
223 if (!cached) {
224 int changed;
225 struct stat st;
226 changed = check_removed(ce, &st);
227 if (changed < 0)
228 return -1;
229 else if (changed) {
230 if (match_missing) {
231 *sha1p = sha1;
232 *modep = mode;
233 return 0;
234 }
235 return -1;
236 }
237 changed = ce_match_stat(ce, &st, 0);
238 if (changed) {
239 mode = ce_mode_from_stat(ce, st.st_mode);
240 sha1 = null_sha1;
241 }
242 }
243
244 *sha1p = sha1;
245 *modep = mode;
246 return 0;
247}
248
249static void show_new_file(struct oneway_unpack_data *cbdata,
250 struct cache_entry *new,
251 int cached, int match_missing)
252{
253 const unsigned char *sha1;
254 unsigned int mode;
255 struct rev_info *revs = cbdata->revs;
256
257 /*
258 * New file in the index: it might actually be different in
259 * the working copy.
260 */
261 if (get_stat_data(new, &sha1, &mode, cached, match_missing, cbdata) < 0)
262 return;
263
264 diff_index_show_file(revs, "+", new, sha1, mode);
265}
266
267static int show_modified(struct oneway_unpack_data *cbdata,
268 struct cache_entry *old,
269 struct cache_entry *new,
270 int report_missing,
271 int cached, int match_missing)
272{
273 unsigned int mode, oldmode;
274 const unsigned char *sha1;
275 struct rev_info *revs = cbdata->revs;
276
277 if (get_stat_data(new, &sha1, &mode, cached, match_missing, cbdata) < 0) {
278 if (report_missing)
279 diff_index_show_file(revs, "-", old,
280 old->sha1, old->ce_mode);
281 return -1;
282 }
283
284 if (revs->combine_merges && !cached &&
285 (hashcmp(sha1, old->sha1) || hashcmp(old->sha1, new->sha1))) {
286 struct combine_diff_path *p;
287 int pathlen = ce_namelen(new);
288
289 p = xmalloc(combine_diff_path_size(2, pathlen));
290 p->path = (char *) &p->parent[2];
291 p->next = NULL;
292 p->len = pathlen;
293 memcpy(p->path, new->name, pathlen);
294 p->path[pathlen] = 0;
295 p->mode = mode;
296 hashclr(p->sha1);
297 memset(p->parent, 0, 2 * sizeof(struct combine_diff_parent));
298 p->parent[0].status = DIFF_STATUS_MODIFIED;
299 p->parent[0].mode = new->ce_mode;
300 hashcpy(p->parent[0].sha1, new->sha1);
301 p->parent[1].status = DIFF_STATUS_MODIFIED;
302 p->parent[1].mode = old->ce_mode;
303 hashcpy(p->parent[1].sha1, old->sha1);
304 show_combined_diff(p, 2, revs->dense_combined_merges, revs);
305 free(p);
306 return 0;
307 }
308
309 oldmode = old->ce_mode;
310 if (mode == oldmode && !hashcmp(sha1, old->sha1) &&
311 !DIFF_OPT_TST(&revs->diffopt, FIND_COPIES_HARDER))
312 return 0;
313
314 diff_change(&revs->diffopt, oldmode, mode,
315 old->sha1, sha1, old->name, NULL);
316 return 0;
317}
318
319/*
320 * This turns all merge entries into "stage 3". That guarantees that
321 * when we read in the new tree (into "stage 1"), we won't lose sight
322 * of the fact that we had unmerged entries.
323 */
324static void mark_merge_entries(void)
325{
326 int i;
327 for (i = 0; i < active_nr; i++) {
328 struct cache_entry *ce = active_cache[i];
329 if (!ce_stage(ce))
330 continue;
331 ce->ce_flags |= CE_STAGEMASK;
332 }
333}
334
335/*
336 * This gets a mix of an existing index and a tree, one pathname entry
337 * at a time. The index entry may be a single stage-0 one, but it could
338 * also be multiple unmerged entries (in which case idx_pos/idx_nr will
339 * give you the position and number of entries in the index).
340 */
341static void do_oneway_diff(struct unpack_trees_options *o,
342 struct cache_entry *idx,
343 struct cache_entry *tree)
344{
345 struct oneway_unpack_data *cbdata = o->unpack_data;
346 struct rev_info *revs = cbdata->revs;
347 int match_missing, cached;
348
349 /*
350 * Backward compatibility wart - "diff-index -m" does
351 * not mean "do not ignore merges", but "match_missing".
352 *
353 * But with the revision flag parsing, that's found in
354 * "!revs->ignore_merges".
355 */
356 cached = o->index_only;
357 match_missing = !revs->ignore_merges;
358
359 if (cached && idx && ce_stage(idx)) {
360 if (tree)
361 diff_unmerge(&revs->diffopt, idx->name, idx->ce_mode, idx->sha1);
362 return;
363 }
364
365 /*
366 * Something added to the tree?
367 */
368 if (!tree) {
369 show_new_file(cbdata, idx, cached, match_missing);
370 return;
371 }
372
373 /*
374 * Something removed from the tree?
375 */
376 if (!idx) {
377 diff_index_show_file(revs, "-", tree, tree->sha1, tree->ce_mode);
378 return;
379 }
380
381 /* Show difference between old and new */
382 show_modified(cbdata, tree, idx, 1, cached, match_missing);
383}
384
385static inline void skip_same_name(struct cache_entry *ce, struct unpack_trees_options *o)
386{
387 int len = ce_namelen(ce);
388 const struct index_state *index = o->src_index;
389
390 while (o->pos < index->cache_nr) {
391 struct cache_entry *next = index->cache[o->pos];
392 if (len != ce_namelen(next))
393 break;
394 if (memcmp(ce->name, next->name, len))
395 break;
396 o->pos++;
397 }
398}
399
400/*
401 * The unpack_trees() interface is designed for merging, so
402 * the different source entries are designed primarily for
403 * the source trees, with the old index being really mainly
404 * used for being replaced by the result.
405 *
406 * For diffing, the index is more important, and we only have a
407 * single tree.
408 *
409 * We're supposed to return how many index entries we want to skip.
410 *
411 * This wrapper makes it all more readable, and takes care of all
412 * the fairly complex unpack_trees() semantic requirements, including
413 * the skipping, the path matching, the type conflict cases etc.
414 */
415static int oneway_diff(struct cache_entry **src, struct unpack_trees_options *o)
416{
417 struct cache_entry *idx = src[0];
418 struct cache_entry *tree = src[1];
419 struct oneway_unpack_data *cbdata = o->unpack_data;
420 struct rev_info *revs = cbdata->revs;
421
422 if (idx && ce_stage(idx))
423 skip_same_name(idx, o);
424
425 /*
426 * Unpack-trees generates a DF/conflict entry if
427 * there was a directory in the index and a tree
428 * in the tree. From a diff standpoint, that's a
429 * delete of the tree and a create of the file.
430 */
431 if (tree == o->df_conflict_entry)
432 tree = NULL;
433
434 if (ce_path_match(idx ? idx : tree, revs->prune_data))
435 do_oneway_diff(o, idx, tree);
436
437 return 0;
438}
439
440int run_diff_index(struct rev_info *revs, int cached)
441{
442 struct object *ent;
443 struct tree *tree;
444 const char *tree_name;
445 struct unpack_trees_options opts;
446 struct tree_desc t;
447 struct oneway_unpack_data unpack_cb;
448
449 mark_merge_entries();
450
451 ent = revs->pending.objects[0].item;
452 tree_name = revs->pending.objects[0].name;
453 tree = parse_tree_indirect(ent->sha1);
454 if (!tree)
455 return error("bad tree object %s", tree_name);
456
457 unpack_cb.revs = revs;
458 unpack_cb.symcache[0] = '\0';
459 memset(&opts, 0, sizeof(opts));
460 opts.head_idx = 1;
461 opts.index_only = cached;
462 opts.merge = 1;
463 opts.fn = oneway_diff;
464 opts.unpack_data = &unpack_cb;
465 opts.src_index = &the_index;
466 opts.dst_index = NULL;
467
468 init_tree_desc(&t, tree->buffer, tree->size);
469 if (unpack_trees(1, &t, &opts))
470 exit(128);
471
472 diffcore_std(&revs->diffopt);
473 diff_flush(&revs->diffopt);
474 return 0;
475}
476
477int do_diff_cache(const unsigned char *tree_sha1, struct diff_options *opt)
478{
479 struct tree *tree;
480 struct rev_info revs;
481 int i;
482 struct cache_entry **dst;
483 struct cache_entry *last = NULL;
484 struct unpack_trees_options opts;
485 struct tree_desc t;
486 struct oneway_unpack_data unpack_cb;
487
488 /*
489 * This is used by git-blame to run diff-cache internally;
490 * it potentially needs to repeatedly run this, so we will
491 * start by removing the higher order entries the last round
492 * left behind.
493 */
494 dst = active_cache;
495 for (i = 0; i < active_nr; i++) {
496 struct cache_entry *ce = active_cache[i];
497 if (ce_stage(ce)) {
498 if (last && !strcmp(ce->name, last->name))
499 continue;
500 cache_tree_invalidate_path(active_cache_tree,
501 ce->name);
502 last = ce;
503 ce->ce_flags |= CE_REMOVE;
504 }
505 *dst++ = ce;
506 }
507 active_nr = dst - active_cache;
508
509 init_revisions(&revs, NULL);
510 revs.prune_data = opt->paths;
511 tree = parse_tree_indirect(tree_sha1);
512 if (!tree)
513 die("bad tree object %s", sha1_to_hex(tree_sha1));
514
515 unpack_cb.revs = &revs;
516 unpack_cb.symcache[0] = '\0';
517 memset(&opts, 0, sizeof(opts));
518 opts.head_idx = 1;
519 opts.index_only = 1;
520 opts.merge = 1;
521 opts.fn = oneway_diff;
522 opts.unpack_data = &unpack_cb;
523 opts.src_index = &the_index;
524 opts.dst_index = &the_index;
525
526 init_tree_desc(&t, tree->buffer, tree->size);
527 if (unpack_trees(1, &t, &opts))
528 exit(128);
529 return 0;
530}