Add an (optional, since expensive) test for >2gb clones
[git/spearce.git] / diff-lib.c
blob79d06068344f5a602f6c8799f6671ccbf98cf49c
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"
15 * diff-files
19 * Has the work tree entity been removed?
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.
27 static int check_removed(const struct cache_entry *ce, struct stat *st)
29 if (lstat(ce->name, st) < 0) {
30 if (errno != ENOENT && errno != ENOTDIR)
31 return -1;
32 return 1;
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];
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.
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!
50 if (!S_ISGITLINK(ce->ce_mode) &&
51 resolve_gitlink_ref(ce->name, "HEAD", sub))
52 return 1;
54 return 0;
57 int run_diff_files(struct rev_info *revs, unsigned int option)
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);
65 diff_set_mnemonic_prefix(&revs->diffopt, "i/", "w/");
67 if (diff_unmerged_stage < 0)
68 diff_unmerged_stage = 2;
69 entries = active_nr;
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;
76 if (DIFF_OPT_TST(&revs->diffopt, QUIET) &&
77 DIFF_OPT_TST(&revs->diffopt, HAS_CHANGES))
78 break;
80 if (!ce_path_match(ce, revs->prune_data))
81 continue;
83 if (ce_stage(ce)) {
84 struct combine_diff_path *dpath;
85 int num_compare_stages = 0;
86 size_t path_len;
88 path_len = ce_namelen(ce);
90 dpath = xmalloc(combine_diff_path_size(5, path_len));
91 dpath->path = (char *) &(dpath->parent[5]);
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);
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;
109 if (silent_on_removed)
110 continue;
113 while (i < entries) {
114 struct cache_entry *nce = active_cache[i];
115 int stage;
117 if (strcmp(ce->name, nce->name))
118 break;
120 /* Stage #2 (ours) is the first parent,
121 * stage #3 (theirs) is the second.
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;
133 /* diff against the proper unmerged stage */
134 if (stage == diff_unmerged_stage)
135 ce = nce;
136 i++;
139 * Compensate for loop update
141 i--;
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;
150 free(dpath);
151 dpath = NULL;
154 * Show the diff for the 'ce' if we found the one
155 * from the desired stage.
157 diff_unmerge(&revs->diffopt, ce->name, 0, null_sha1);
158 if (ce_stage(ce) != diff_unmerged_stage)
159 continue;
162 if (ce_uptodate(ce))
163 continue;
165 changed = check_removed(ce, &st);
166 if (changed) {
167 if (changed < 0) {
168 perror(ce->name);
169 continue;
171 if (silent_on_removed)
172 continue;
173 diff_addremove(&revs->diffopt, '-', ce->ce_mode,
174 ce->sha1, ce->name);
175 continue;
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;
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);
190 diffcore_std(&revs->diffopt);
191 diff_flush(&revs->diffopt);
192 return 0;
196 * diff-index
199 /* A file entry went away or appeared */
200 static void diff_index_show_file(struct rev_info *revs,
201 const char *prefix,
202 struct cache_entry *ce,
203 const unsigned char *sha1, unsigned int mode)
205 diff_addremove(&revs->diffopt, prefix[0], mode,
206 sha1, ce->name);
209 static int get_stat_data(struct cache_entry *ce,
210 const unsigned char **sha1p,
211 unsigned int *modep,
212 int cached, int match_missing)
214 const unsigned char *sha1 = ce->sha1;
215 unsigned int mode = ce->ce_mode;
217 if (!cached) {
218 int changed;
219 struct stat st;
220 changed = check_removed(ce, &st);
221 if (changed < 0)
222 return -1;
223 else if (changed) {
224 if (match_missing) {
225 *sha1p = sha1;
226 *modep = mode;
227 return 0;
229 return -1;
231 changed = ce_match_stat(ce, &st, 0);
232 if (changed) {
233 mode = ce_mode_from_stat(ce, st.st_mode);
234 sha1 = null_sha1;
238 *sha1p = sha1;
239 *modep = mode;
240 return 0;
243 static void show_new_file(struct rev_info *revs,
244 struct cache_entry *new,
245 int cached, int match_missing)
247 const unsigned char *sha1;
248 unsigned int mode;
251 * New file in the index: it might actually be different in
252 * the working copy.
254 if (get_stat_data(new, &sha1, &mode, cached, match_missing) < 0)
255 return;
257 diff_index_show_file(revs, "+", new, sha1, mode);
260 static int show_modified(struct rev_info *revs,
261 struct cache_entry *old,
262 struct cache_entry *new,
263 int report_missing,
264 int cached, int match_missing)
266 unsigned int mode, oldmode;
267 const unsigned char *sha1;
269 if (get_stat_data(new, &sha1, &mode, cached, match_missing) < 0) {
270 if (report_missing)
271 diff_index_show_file(revs, "-", old,
272 old->sha1, old->ce_mode);
273 return -1;
276 if (revs->combine_merges && !cached &&
277 (hashcmp(sha1, old->sha1) || hashcmp(old->sha1, new->sha1))) {
278 struct combine_diff_path *p;
279 int pathlen = ce_namelen(new);
281 p = xmalloc(combine_diff_path_size(2, pathlen));
282 p->path = (char *) &p->parent[2];
283 p->next = NULL;
284 p->len = pathlen;
285 memcpy(p->path, new->name, pathlen);
286 p->path[pathlen] = 0;
287 p->mode = mode;
288 hashclr(p->sha1);
289 memset(p->parent, 0, 2 * sizeof(struct combine_diff_parent));
290 p->parent[0].status = DIFF_STATUS_MODIFIED;
291 p->parent[0].mode = new->ce_mode;
292 hashcpy(p->parent[0].sha1, new->sha1);
293 p->parent[1].status = DIFF_STATUS_MODIFIED;
294 p->parent[1].mode = old->ce_mode;
295 hashcpy(p->parent[1].sha1, old->sha1);
296 show_combined_diff(p, 2, revs->dense_combined_merges, revs);
297 free(p);
298 return 0;
301 oldmode = old->ce_mode;
302 if (mode == oldmode && !hashcmp(sha1, old->sha1) &&
303 !DIFF_OPT_TST(&revs->diffopt, FIND_COPIES_HARDER))
304 return 0;
306 diff_change(&revs->diffopt, oldmode, mode,
307 old->sha1, sha1, old->name);
308 return 0;
312 * This turns all merge entries into "stage 3". That guarantees that
313 * when we read in the new tree (into "stage 1"), we won't lose sight
314 * of the fact that we had unmerged entries.
316 static void mark_merge_entries(void)
318 int i;
319 for (i = 0; i < active_nr; i++) {
320 struct cache_entry *ce = active_cache[i];
321 if (!ce_stage(ce))
322 continue;
323 ce->ce_flags |= CE_STAGEMASK;
328 * This gets a mix of an existing index and a tree, one pathname entry
329 * at a time. The index entry may be a single stage-0 one, but it could
330 * also be multiple unmerged entries (in which case idx_pos/idx_nr will
331 * give you the position and number of entries in the index).
333 static void do_oneway_diff(struct unpack_trees_options *o,
334 struct cache_entry *idx,
335 struct cache_entry *tree)
337 struct rev_info *revs = o->unpack_data;
338 int match_missing, cached;
341 * Backward compatibility wart - "diff-index -m" does
342 * not mean "do not ignore merges", but "match_missing".
344 * But with the revision flag parsing, that's found in
345 * "!revs->ignore_merges".
347 cached = o->index_only;
348 match_missing = !revs->ignore_merges;
350 if (cached && idx && ce_stage(idx)) {
351 if (tree)
352 diff_unmerge(&revs->diffopt, idx->name, idx->ce_mode, idx->sha1);
353 return;
357 * Something added to the tree?
359 if (!tree) {
360 show_new_file(revs, idx, cached, match_missing);
361 return;
365 * Something removed from the tree?
367 if (!idx) {
368 diff_index_show_file(revs, "-", tree, tree->sha1, tree->ce_mode);
369 return;
372 /* Show difference between old and new */
373 show_modified(revs, tree, idx, 1, cached, match_missing);
376 static inline void skip_same_name(struct cache_entry *ce, struct unpack_trees_options *o)
378 int len = ce_namelen(ce);
379 const struct index_state *index = o->src_index;
381 while (o->pos < index->cache_nr) {
382 struct cache_entry *next = index->cache[o->pos];
383 if (len != ce_namelen(next))
384 break;
385 if (memcmp(ce->name, next->name, len))
386 break;
387 o->pos++;
392 * The unpack_trees() interface is designed for merging, so
393 * the different source entries are designed primarily for
394 * the source trees, with the old index being really mainly
395 * used for being replaced by the result.
397 * For diffing, the index is more important, and we only have a
398 * single tree.
400 * We're supposed to return how many index entries we want to skip.
402 * This wrapper makes it all more readable, and takes care of all
403 * the fairly complex unpack_trees() semantic requirements, including
404 * the skipping, the path matching, the type conflict cases etc.
406 static int oneway_diff(struct cache_entry **src, struct unpack_trees_options *o)
408 struct cache_entry *idx = src[0];
409 struct cache_entry *tree = src[1];
410 struct rev_info *revs = o->unpack_data;
412 if (idx && ce_stage(idx))
413 skip_same_name(idx, o);
416 * Unpack-trees generates a DF/conflict entry if
417 * there was a directory in the index and a tree
418 * in the tree. From a diff standpoint, that's a
419 * delete of the tree and a create of the file.
421 if (tree == o->df_conflict_entry)
422 tree = NULL;
424 if (ce_path_match(idx ? idx : tree, revs->prune_data))
425 do_oneway_diff(o, idx, tree);
427 return 0;
430 int run_diff_index(struct rev_info *revs, int cached)
432 struct object *ent;
433 struct tree *tree;
434 const char *tree_name;
435 struct unpack_trees_options opts;
436 struct tree_desc t;
438 mark_merge_entries();
440 ent = revs->pending.objects[0].item;
441 tree_name = revs->pending.objects[0].name;
442 tree = parse_tree_indirect(ent->sha1);
443 if (!tree)
444 return error("bad tree object %s", tree_name);
446 memset(&opts, 0, sizeof(opts));
447 opts.head_idx = 1;
448 opts.index_only = cached;
449 opts.merge = 1;
450 opts.fn = oneway_diff;
451 opts.unpack_data = revs;
452 opts.src_index = &the_index;
453 opts.dst_index = NULL;
455 init_tree_desc(&t, tree->buffer, tree->size);
456 if (unpack_trees(1, &t, &opts))
457 exit(128);
459 diff_set_mnemonic_prefix(&revs->diffopt, "c/", cached ? "i/" : "w/");
460 diffcore_std(&revs->diffopt);
461 diff_flush(&revs->diffopt);
462 return 0;
465 int do_diff_cache(const unsigned char *tree_sha1, struct diff_options *opt)
467 struct tree *tree;
468 struct rev_info revs;
469 int i;
470 struct cache_entry **dst;
471 struct cache_entry *last = NULL;
472 struct unpack_trees_options opts;
473 struct tree_desc t;
476 * This is used by git-blame to run diff-cache internally;
477 * it potentially needs to repeatedly run this, so we will
478 * start by removing the higher order entries the last round
479 * left behind.
481 dst = active_cache;
482 for (i = 0; i < active_nr; i++) {
483 struct cache_entry *ce = active_cache[i];
484 if (ce_stage(ce)) {
485 if (last && !strcmp(ce->name, last->name))
486 continue;
487 cache_tree_invalidate_path(active_cache_tree,
488 ce->name);
489 last = ce;
490 ce->ce_flags |= CE_REMOVE;
492 *dst++ = ce;
494 active_nr = dst - active_cache;
496 init_revisions(&revs, NULL);
497 revs.prune_data = opt->paths;
498 tree = parse_tree_indirect(tree_sha1);
499 if (!tree)
500 die("bad tree object %s", sha1_to_hex(tree_sha1));
502 memset(&opts, 0, sizeof(opts));
503 opts.head_idx = 1;
504 opts.index_only = 1;
505 opts.merge = 1;
506 opts.fn = oneway_diff;
507 opts.unpack_data = &revs;
508 opts.src_index = &the_index;
509 opts.dst_index = &the_index;
511 init_tree_desc(&t, tree->buffer, tree->size);
512 if (unpack_trees(1, &t, &opts))
513 exit(128);
514 return 0;
517 int index_differs_from(const char *def, int diff_flags)
519 struct rev_info rev;
521 init_revisions(&rev, NULL);
522 setup_revisions(0, NULL, &rev, def);
523 DIFF_OPT_SET(&rev.diffopt, QUIET);
524 DIFF_OPT_SET(&rev.diffopt, EXIT_WITH_STATUS);
525 rev.diffopt.flags |= diff_flags;
526 run_diff_index(&rev, 1);
527 if (rev.pending.alloc)
528 free(rev.pending.objects);
529 return (DIFF_OPT_TST(&rev.diffopt, HAS_CHANGES) != 0);