]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Copyright (C) 2005 Junio C Hamano | |
3 | */ | |
4 | #include "cache.h" | |
5 | #include "diff.h" | |
6 | #include "diffcore.h" | |
7 | #include "hash.h" | |
8 | #include "progress.h" | |
9 | ||
10 | /* Table of rename/copy destinations */ | |
11 | ||
12 | static struct diff_rename_dst { | |
13 | struct diff_filespec *two; | |
14 | struct diff_filepair *pair; | |
15 | } *rename_dst; | |
16 | static int rename_dst_nr, rename_dst_alloc; | |
17 | ||
18 | static struct diff_rename_dst *locate_rename_dst(struct diff_filespec *two, | |
19 | int insert_ok) | |
20 | { | |
21 | int first, last; | |
22 | ||
23 | first = 0; | |
24 | last = rename_dst_nr; | |
25 | while (last > first) { | |
26 | int next = (last + first) >> 1; | |
27 | struct diff_rename_dst *dst = &(rename_dst[next]); | |
28 | int cmp = strcmp(two->path, dst->two->path); | |
29 | if (!cmp) | |
30 | return dst; | |
31 | if (cmp < 0) { | |
32 | last = next; | |
33 | continue; | |
34 | } | |
35 | first = next+1; | |
36 | } | |
37 | /* not found */ | |
38 | if (!insert_ok) | |
39 | return NULL; | |
40 | /* insert to make it at "first" */ | |
41 | if (rename_dst_alloc <= rename_dst_nr) { | |
42 | rename_dst_alloc = alloc_nr(rename_dst_alloc); | |
43 | rename_dst = xrealloc(rename_dst, | |
44 | rename_dst_alloc * sizeof(*rename_dst)); | |
45 | } | |
46 | rename_dst_nr++; | |
47 | if (first < rename_dst_nr) | |
48 | memmove(rename_dst + first + 1, rename_dst + first, | |
49 | (rename_dst_nr - first - 1) * sizeof(*rename_dst)); | |
50 | rename_dst[first].two = alloc_filespec(two->path); | |
51 | fill_filespec(rename_dst[first].two, two->sha1, two->sha1_valid, two->mode); | |
52 | rename_dst[first].pair = NULL; | |
53 | return &(rename_dst[first]); | |
54 | } | |
55 | ||
56 | /* Table of rename/copy src files */ | |
57 | static struct diff_rename_src { | |
58 | struct diff_filepair *p; | |
59 | unsigned short score; /* to remember the break score */ | |
60 | } *rename_src; | |
61 | static int rename_src_nr, rename_src_alloc; | |
62 | ||
63 | static struct diff_rename_src *register_rename_src(struct diff_filepair *p) | |
64 | { | |
65 | int first, last; | |
66 | struct diff_filespec *one = p->one; | |
67 | unsigned short score = p->score; | |
68 | ||
69 | first = 0; | |
70 | last = rename_src_nr; | |
71 | while (last > first) { | |
72 | int next = (last + first) >> 1; | |
73 | struct diff_rename_src *src = &(rename_src[next]); | |
74 | int cmp = strcmp(one->path, src->p->one->path); | |
75 | if (!cmp) | |
76 | return src; | |
77 | if (cmp < 0) { | |
78 | last = next; | |
79 | continue; | |
80 | } | |
81 | first = next+1; | |
82 | } | |
83 | ||
84 | /* insert to make it at "first" */ | |
85 | if (rename_src_alloc <= rename_src_nr) { | |
86 | rename_src_alloc = alloc_nr(rename_src_alloc); | |
87 | rename_src = xrealloc(rename_src, | |
88 | rename_src_alloc * sizeof(*rename_src)); | |
89 | } | |
90 | rename_src_nr++; | |
91 | if (first < rename_src_nr) | |
92 | memmove(rename_src + first + 1, rename_src + first, | |
93 | (rename_src_nr - first - 1) * sizeof(*rename_src)); | |
94 | rename_src[first].p = p; | |
95 | rename_src[first].score = score; | |
96 | return &(rename_src[first]); | |
97 | } | |
98 | ||
99 | static int basename_same(struct diff_filespec *src, struct diff_filespec *dst) | |
100 | { | |
101 | int src_len = strlen(src->path), dst_len = strlen(dst->path); | |
102 | while (src_len && dst_len) { | |
103 | char c1 = src->path[--src_len]; | |
104 | char c2 = dst->path[--dst_len]; | |
105 | if (c1 != c2) | |
106 | return 0; | |
107 | if (c1 == '/') | |
108 | return 1; | |
109 | } | |
110 | return (!src_len || src->path[src_len - 1] == '/') && | |
111 | (!dst_len || dst->path[dst_len - 1] == '/'); | |
112 | } | |
113 | ||
114 | struct diff_score { | |
115 | int src; /* index in rename_src */ | |
116 | int dst; /* index in rename_dst */ | |
117 | unsigned short score; | |
118 | short name_score; | |
119 | }; | |
120 | ||
121 | static int estimate_similarity(struct diff_filespec *src, | |
122 | struct diff_filespec *dst, | |
123 | int minimum_score) | |
124 | { | |
125 | /* src points at a file that existed in the original tree (or | |
126 | * optionally a file in the destination tree) and dst points | |
127 | * at a newly created file. They may be quite similar, in which | |
128 | * case we want to say src is renamed to dst or src is copied into | |
129 | * dst, and then some edit has been applied to dst. | |
130 | * | |
131 | * Compare them and return how similar they are, representing | |
132 | * the score as an integer between 0 and MAX_SCORE. | |
133 | * | |
134 | * When there is an exact match, it is considered a better | |
135 | * match than anything else; the destination does not even | |
136 | * call into this function in that case. | |
137 | */ | |
138 | unsigned long max_size, delta_size, base_size, src_copied, literal_added; | |
139 | unsigned long delta_limit; | |
140 | int score; | |
141 | ||
142 | /* We deal only with regular files. Symlink renames are handled | |
143 | * only when they are exact matches --- in other words, no edits | |
144 | * after renaming. | |
145 | */ | |
146 | if (!S_ISREG(src->mode) || !S_ISREG(dst->mode)) | |
147 | return 0; | |
148 | ||
149 | /* | |
150 | * Need to check that source and destination sizes are | |
151 | * filled in before comparing them. | |
152 | * | |
153 | * If we already have "cnt_data" filled in, we know it's | |
154 | * all good (avoid checking the size for zero, as that | |
155 | * is a possible size - we really should have a flag to | |
156 | * say whether the size is valid or not!) | |
157 | */ | |
158 | if (!src->cnt_data && diff_populate_filespec(src, 1)) | |
159 | return 0; | |
160 | if (!dst->cnt_data && diff_populate_filespec(dst, 1)) | |
161 | return 0; | |
162 | ||
163 | max_size = ((src->size > dst->size) ? src->size : dst->size); | |
164 | base_size = ((src->size < dst->size) ? src->size : dst->size); | |
165 | delta_size = max_size - base_size; | |
166 | ||
167 | /* We would not consider edits that change the file size so | |
168 | * drastically. delta_size must be smaller than | |
169 | * (MAX_SCORE-minimum_score)/MAX_SCORE * min(src->size, dst->size). | |
170 | * | |
171 | * Note that base_size == 0 case is handled here already | |
172 | * and the final score computation below would not have a | |
173 | * divide-by-zero issue. | |
174 | */ | |
175 | if (max_size * (MAX_SCORE-minimum_score) < delta_size * MAX_SCORE) | |
176 | return 0; | |
177 | ||
178 | if (!src->cnt_data && diff_populate_filespec(src, 0)) | |
179 | return 0; | |
180 | if (!dst->cnt_data && diff_populate_filespec(dst, 0)) | |
181 | return 0; | |
182 | ||
183 | delta_limit = (unsigned long) | |
184 | (base_size * (MAX_SCORE-minimum_score) / MAX_SCORE); | |
185 | if (diffcore_count_changes(src, dst, | |
186 | &src->cnt_data, &dst->cnt_data, | |
187 | delta_limit, | |
188 | &src_copied, &literal_added)) | |
189 | return 0; | |
190 | ||
191 | /* How similar are they? | |
192 | * what percentage of material in dst are from source? | |
193 | */ | |
194 | if (!dst->size) | |
195 | score = 0; /* should not happen */ | |
196 | else | |
197 | score = (int)(src_copied * MAX_SCORE / max_size); | |
198 | return score; | |
199 | } | |
200 | ||
201 | static void record_rename_pair(int dst_index, int src_index, int score) | |
202 | { | |
203 | struct diff_filespec *src, *dst; | |
204 | struct diff_filepair *dp; | |
205 | ||
206 | if (rename_dst[dst_index].pair) | |
207 | die("internal error: dst already matched."); | |
208 | ||
209 | src = rename_src[src_index].p->one; | |
210 | src->rename_used++; | |
211 | src->count++; | |
212 | ||
213 | dst = rename_dst[dst_index].two; | |
214 | dst->count++; | |
215 | ||
216 | dp = diff_queue(NULL, src, dst); | |
217 | dp->renamed_pair = 1; | |
218 | if (!strcmp(src->path, dst->path)) | |
219 | dp->score = rename_src[src_index].score; | |
220 | else | |
221 | dp->score = score; | |
222 | rename_dst[dst_index].pair = dp; | |
223 | } | |
224 | ||
225 | /* | |
226 | * We sort the rename similarity matrix with the score, in descending | |
227 | * order (the most similar first). | |
228 | */ | |
229 | static int score_compare(const void *a_, const void *b_) | |
230 | { | |
231 | const struct diff_score *a = a_, *b = b_; | |
232 | ||
233 | /* sink the unused ones to the bottom */ | |
234 | if (a->dst < 0) | |
235 | return (0 <= b->dst); | |
236 | else if (b->dst < 0) | |
237 | return -1; | |
238 | ||
239 | if (a->score == b->score) | |
240 | return b->name_score - a->name_score; | |
241 | ||
242 | return b->score - a->score; | |
243 | } | |
244 | ||
245 | struct file_similarity { | |
246 | int src_dst, index; | |
247 | struct diff_filespec *filespec; | |
248 | struct file_similarity *next; | |
249 | }; | |
250 | ||
251 | static int find_identical_files(struct file_similarity *src, | |
252 | struct file_similarity *dst, | |
253 | struct diff_options *options) | |
254 | { | |
255 | int renames = 0; | |
256 | ||
257 | /* | |
258 | * Walk over all the destinations ... | |
259 | */ | |
260 | do { | |
261 | struct diff_filespec *target = dst->filespec; | |
262 | struct file_similarity *p, *best; | |
263 | int i = 100, best_score = -1; | |
264 | ||
265 | /* | |
266 | * .. to find the best source match | |
267 | */ | |
268 | best = NULL; | |
269 | for (p = src; p; p = p->next) { | |
270 | int score; | |
271 | struct diff_filespec *source = p->filespec; | |
272 | ||
273 | /* False hash collision? */ | |
274 | if (hashcmp(source->sha1, target->sha1)) | |
275 | continue; | |
276 | /* Non-regular files? If so, the modes must match! */ | |
277 | if (!S_ISREG(source->mode) || !S_ISREG(target->mode)) { | |
278 | if (source->mode != target->mode) | |
279 | continue; | |
280 | } | |
281 | /* Give higher scores to sources that haven't been used already */ | |
282 | score = !source->rename_used; | |
283 | if (source->rename_used && options->detect_rename != DIFF_DETECT_COPY) | |
284 | continue; | |
285 | score += basename_same(source, target); | |
286 | if (score > best_score) { | |
287 | best = p; | |
288 | best_score = score; | |
289 | if (score == 2) | |
290 | break; | |
291 | } | |
292 | ||
293 | /* Too many identical alternatives? Pick one */ | |
294 | if (!--i) | |
295 | break; | |
296 | } | |
297 | if (best) { | |
298 | record_rename_pair(dst->index, best->index, MAX_SCORE); | |
299 | renames++; | |
300 | } | |
301 | } while ((dst = dst->next) != NULL); | |
302 | return renames; | |
303 | } | |
304 | ||
305 | static void free_similarity_list(struct file_similarity *p) | |
306 | { | |
307 | while (p) { | |
308 | struct file_similarity *entry = p; | |
309 | p = p->next; | |
310 | free(entry); | |
311 | } | |
312 | } | |
313 | ||
314 | static int find_same_files(void *ptr, void *data) | |
315 | { | |
316 | int ret; | |
317 | struct file_similarity *p = ptr; | |
318 | struct file_similarity *src = NULL, *dst = NULL; | |
319 | struct diff_options *options = data; | |
320 | ||
321 | /* Split the hash list up into sources and destinations */ | |
322 | do { | |
323 | struct file_similarity *entry = p; | |
324 | p = p->next; | |
325 | if (entry->src_dst < 0) { | |
326 | entry->next = src; | |
327 | src = entry; | |
328 | } else { | |
329 | entry->next = dst; | |
330 | dst = entry; | |
331 | } | |
332 | } while (p); | |
333 | ||
334 | /* | |
335 | * If we have both sources *and* destinations, see if | |
336 | * we can match them up | |
337 | */ | |
338 | ret = (src && dst) ? find_identical_files(src, dst, options) : 0; | |
339 | ||
340 | /* Free the hashes and return the number of renames found */ | |
341 | free_similarity_list(src); | |
342 | free_similarity_list(dst); | |
343 | return ret; | |
344 | } | |
345 | ||
346 | static unsigned int hash_filespec(struct diff_filespec *filespec) | |
347 | { | |
348 | unsigned int hash; | |
349 | if (!filespec->sha1_valid) { | |
350 | if (diff_populate_filespec(filespec, 0)) | |
351 | return 0; | |
352 | hash_sha1_file(filespec->data, filespec->size, "blob", filespec->sha1); | |
353 | } | |
354 | memcpy(&hash, filespec->sha1, sizeof(hash)); | |
355 | return hash; | |
356 | } | |
357 | ||
358 | static void insert_file_table(struct hash_table *table, int src_dst, int index, struct diff_filespec *filespec) | |
359 | { | |
360 | void **pos; | |
361 | unsigned int hash; | |
362 | struct file_similarity *entry = xmalloc(sizeof(*entry)); | |
363 | ||
364 | entry->src_dst = src_dst; | |
365 | entry->index = index; | |
366 | entry->filespec = filespec; | |
367 | entry->next = NULL; | |
368 | ||
369 | hash = hash_filespec(filespec); | |
370 | pos = insert_hash(hash, entry, table); | |
371 | ||
372 | /* We already had an entry there? */ | |
373 | if (pos) { | |
374 | entry->next = *pos; | |
375 | *pos = entry; | |
376 | } | |
377 | } | |
378 | ||
379 | /* | |
380 | * Find exact renames first. | |
381 | * | |
382 | * The first round matches up the up-to-date entries, | |
383 | * and then during the second round we try to match | |
384 | * cache-dirty entries as well. | |
385 | */ | |
386 | static int find_exact_renames(struct diff_options *options) | |
387 | { | |
388 | int i; | |
389 | struct hash_table file_table; | |
390 | ||
391 | init_hash(&file_table); | |
392 | for (i = 0; i < rename_src_nr; i++) | |
393 | insert_file_table(&file_table, -1, i, rename_src[i].p->one); | |
394 | ||
395 | for (i = 0; i < rename_dst_nr; i++) | |
396 | insert_file_table(&file_table, 1, i, rename_dst[i].two); | |
397 | ||
398 | /* Find the renames */ | |
399 | i = for_each_hash(&file_table, find_same_files, options); | |
400 | ||
401 | /* .. and free the hash data structure */ | |
402 | free_hash(&file_table); | |
403 | ||
404 | return i; | |
405 | } | |
406 | ||
407 | #define NUM_CANDIDATE_PER_DST 4 | |
408 | static void record_if_better(struct diff_score m[], struct diff_score *o) | |
409 | { | |
410 | int i, worst; | |
411 | ||
412 | /* find the worst one */ | |
413 | worst = 0; | |
414 | for (i = 1; i < NUM_CANDIDATE_PER_DST; i++) | |
415 | if (score_compare(&m[i], &m[worst]) > 0) | |
416 | worst = i; | |
417 | ||
418 | /* is it better than the worst one? */ | |
419 | if (score_compare(&m[worst], o) > 0) | |
420 | m[worst] = *o; | |
421 | } | |
422 | ||
423 | /* | |
424 | * Returns: | |
425 | * 0 if we are under the limit; | |
426 | * 1 if we need to disable inexact rename detection; | |
427 | * 2 if we would be under the limit if we were given -C instead of -C -C. | |
428 | */ | |
429 | static int too_many_rename_candidates(int num_create, | |
430 | struct diff_options *options) | |
431 | { | |
432 | int rename_limit = options->rename_limit; | |
433 | int num_src = rename_src_nr; | |
434 | int i; | |
435 | ||
436 | options->needed_rename_limit = 0; | |
437 | ||
438 | /* | |
439 | * This basically does a test for the rename matrix not | |
440 | * growing larger than a "rename_limit" square matrix, ie: | |
441 | * | |
442 | * num_create * num_src > rename_limit * rename_limit | |
443 | * | |
444 | * but handles the potential overflow case specially (and we | |
445 | * assume at least 32-bit integers) | |
446 | */ | |
447 | if (rename_limit <= 0 || rename_limit > 32767) | |
448 | rename_limit = 32767; | |
449 | if ((num_create <= rename_limit || num_src <= rename_limit) && | |
450 | (num_create * num_src <= rename_limit * rename_limit)) | |
451 | return 0; | |
452 | ||
453 | options->needed_rename_limit = | |
454 | num_src > num_create ? num_src : num_create; | |
455 | ||
456 | /* Are we running under -C -C? */ | |
457 | if (!DIFF_OPT_TST(options, FIND_COPIES_HARDER)) | |
458 | return 1; | |
459 | ||
460 | /* Would we bust the limit if we were running under -C? */ | |
461 | for (num_src = i = 0; i < rename_src_nr; i++) { | |
462 | if (diff_unmodified_pair(rename_src[i].p)) | |
463 | continue; | |
464 | num_src++; | |
465 | } | |
466 | if ((num_create <= rename_limit || num_src <= rename_limit) && | |
467 | (num_create * num_src <= rename_limit * rename_limit)) | |
468 | return 2; | |
469 | return 1; | |
470 | } | |
471 | ||
472 | static int find_renames(struct diff_score *mx, int dst_cnt, int minimum_score, int copies) | |
473 | { | |
474 | int count = 0, i; | |
475 | ||
476 | for (i = 0; i < dst_cnt * NUM_CANDIDATE_PER_DST; i++) { | |
477 | struct diff_rename_dst *dst; | |
478 | ||
479 | if ((mx[i].dst < 0) || | |
480 | (mx[i].score < minimum_score)) | |
481 | break; /* there is no more usable pair. */ | |
482 | dst = &rename_dst[mx[i].dst]; | |
483 | if (dst->pair) | |
484 | continue; /* already done, either exact or fuzzy. */ | |
485 | if (!copies && rename_src[mx[i].src].p->one->rename_used) | |
486 | continue; | |
487 | record_rename_pair(mx[i].dst, mx[i].src, mx[i].score); | |
488 | count++; | |
489 | } | |
490 | return count; | |
491 | } | |
492 | ||
493 | void diffcore_rename(struct diff_options *options) | |
494 | { | |
495 | int detect_rename = options->detect_rename; | |
496 | int minimum_score = options->rename_score; | |
497 | struct diff_queue_struct *q = &diff_queued_diff; | |
498 | struct diff_queue_struct outq; | |
499 | struct diff_score *mx; | |
500 | int i, j, rename_count, skip_unmodified = 0; | |
501 | int num_create, dst_cnt; | |
502 | struct progress *progress = NULL; | |
503 | ||
504 | if (!minimum_score) | |
505 | minimum_score = DEFAULT_RENAME_SCORE; | |
506 | ||
507 | for (i = 0; i < q->nr; i++) { | |
508 | struct diff_filepair *p = q->queue[i]; | |
509 | if (!DIFF_FILE_VALID(p->one)) { | |
510 | if (!DIFF_FILE_VALID(p->two)) | |
511 | continue; /* unmerged */ | |
512 | else if (options->single_follow && | |
513 | strcmp(options->single_follow, p->two->path)) | |
514 | continue; /* not interested */ | |
515 | else if (!DIFF_OPT_TST(options, RENAME_EMPTY) && | |
516 | is_empty_blob_sha1(p->two->sha1)) | |
517 | continue; | |
518 | else | |
519 | locate_rename_dst(p->two, 1); | |
520 | } | |
521 | else if (!DIFF_OPT_TST(options, RENAME_EMPTY) && | |
522 | is_empty_blob_sha1(p->one->sha1)) | |
523 | continue; | |
524 | else if (!DIFF_PAIR_UNMERGED(p) && !DIFF_FILE_VALID(p->two)) { | |
525 | /* | |
526 | * If the source is a broken "delete", and | |
527 | * they did not really want to get broken, | |
528 | * that means the source actually stays. | |
529 | * So we increment the "rename_used" score | |
530 | * by one, to indicate ourselves as a user | |
531 | */ | |
532 | if (p->broken_pair && !p->score) | |
533 | p->one->rename_used++; | |
534 | register_rename_src(p); | |
535 | } | |
536 | else if (detect_rename == DIFF_DETECT_COPY) { | |
537 | /* | |
538 | * Increment the "rename_used" score by | |
539 | * one, to indicate ourselves as a user. | |
540 | */ | |
541 | p->one->rename_used++; | |
542 | register_rename_src(p); | |
543 | } | |
544 | } | |
545 | if (rename_dst_nr == 0 || rename_src_nr == 0) | |
546 | goto cleanup; /* nothing to do */ | |
547 | ||
548 | /* | |
549 | * We really want to cull the candidates list early | |
550 | * with cheap tests in order to avoid doing deltas. | |
551 | */ | |
552 | rename_count = find_exact_renames(options); | |
553 | ||
554 | /* Did we only want exact renames? */ | |
555 | if (minimum_score == MAX_SCORE) | |
556 | goto cleanup; | |
557 | ||
558 | /* | |
559 | * Calculate how many renames are left (but all the source | |
560 | * files still remain as options for rename/copies!) | |
561 | */ | |
562 | num_create = (rename_dst_nr - rename_count); | |
563 | ||
564 | /* All done? */ | |
565 | if (!num_create) | |
566 | goto cleanup; | |
567 | ||
568 | switch (too_many_rename_candidates(num_create, options)) { | |
569 | case 1: | |
570 | goto cleanup; | |
571 | case 2: | |
572 | options->degraded_cc_to_c = 1; | |
573 | skip_unmodified = 1; | |
574 | break; | |
575 | default: | |
576 | break; | |
577 | } | |
578 | ||
579 | if (options->show_rename_progress) { | |
580 | progress = start_progress_delay( | |
581 | "Performing inexact rename detection", | |
582 | rename_dst_nr * rename_src_nr, 50, 1); | |
583 | } | |
584 | ||
585 | mx = xcalloc(num_create * NUM_CANDIDATE_PER_DST, sizeof(*mx)); | |
586 | for (dst_cnt = i = 0; i < rename_dst_nr; i++) { | |
587 | struct diff_filespec *two = rename_dst[i].two; | |
588 | struct diff_score *m; | |
589 | ||
590 | if (rename_dst[i].pair) | |
591 | continue; /* dealt with exact match already. */ | |
592 | ||
593 | m = &mx[dst_cnt * NUM_CANDIDATE_PER_DST]; | |
594 | for (j = 0; j < NUM_CANDIDATE_PER_DST; j++) | |
595 | m[j].dst = -1; | |
596 | ||
597 | for (j = 0; j < rename_src_nr; j++) { | |
598 | struct diff_filespec *one = rename_src[j].p->one; | |
599 | struct diff_score this_src; | |
600 | ||
601 | if (skip_unmodified && | |
602 | diff_unmodified_pair(rename_src[j].p)) | |
603 | continue; | |
604 | ||
605 | this_src.score = estimate_similarity(one, two, | |
606 | minimum_score); | |
607 | this_src.name_score = basename_same(one, two); | |
608 | this_src.dst = i; | |
609 | this_src.src = j; | |
610 | record_if_better(m, &this_src); | |
611 | /* | |
612 | * Once we run estimate_similarity, | |
613 | * We do not need the text anymore. | |
614 | */ | |
615 | diff_free_filespec_blob(one); | |
616 | diff_free_filespec_blob(two); | |
617 | } | |
618 | dst_cnt++; | |
619 | display_progress(progress, (i+1)*rename_src_nr); | |
620 | } | |
621 | stop_progress(&progress); | |
622 | ||
623 | /* cost matrix sorted by most to least similar pair */ | |
624 | qsort(mx, dst_cnt * NUM_CANDIDATE_PER_DST, sizeof(*mx), score_compare); | |
625 | ||
626 | rename_count += find_renames(mx, dst_cnt, minimum_score, 0); | |
627 | if (detect_rename == DIFF_DETECT_COPY) | |
628 | rename_count += find_renames(mx, dst_cnt, minimum_score, 1); | |
629 | free(mx); | |
630 | ||
631 | cleanup: | |
632 | /* At this point, we have found some renames and copies and they | |
633 | * are recorded in rename_dst. The original list is still in *q. | |
634 | */ | |
635 | DIFF_QUEUE_CLEAR(&outq); | |
636 | for (i = 0; i < q->nr; i++) { | |
637 | struct diff_filepair *p = q->queue[i]; | |
638 | struct diff_filepair *pair_to_free = NULL; | |
639 | ||
640 | if (DIFF_PAIR_UNMERGED(p)) { | |
641 | diff_q(&outq, p); | |
642 | } | |
643 | else if (!DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) { | |
644 | /* | |
645 | * Creation | |
646 | * | |
647 | * We would output this create record if it has | |
648 | * not been turned into a rename/copy already. | |
649 | */ | |
650 | struct diff_rename_dst *dst = | |
651 | locate_rename_dst(p->two, 0); | |
652 | if (dst && dst->pair) { | |
653 | diff_q(&outq, dst->pair); | |
654 | pair_to_free = p; | |
655 | } | |
656 | else | |
657 | /* no matching rename/copy source, so | |
658 | * record this as a creation. | |
659 | */ | |
660 | diff_q(&outq, p); | |
661 | } | |
662 | else if (DIFF_FILE_VALID(p->one) && !DIFF_FILE_VALID(p->two)) { | |
663 | /* | |
664 | * Deletion | |
665 | * | |
666 | * We would output this delete record if: | |
667 | * | |
668 | * (1) this is a broken delete and the counterpart | |
669 | * broken create remains in the output; or | |
670 | * (2) this is not a broken delete, and rename_dst | |
671 | * does not have a rename/copy to move p->one->path | |
672 | * out of existence. | |
673 | * | |
674 | * Otherwise, the counterpart broken create | |
675 | * has been turned into a rename-edit; or | |
676 | * delete did not have a matching create to | |
677 | * begin with. | |
678 | */ | |
679 | if (DIFF_PAIR_BROKEN(p)) { | |
680 | /* broken delete */ | |
681 | struct diff_rename_dst *dst = | |
682 | locate_rename_dst(p->one, 0); | |
683 | if (dst && dst->pair) | |
684 | /* counterpart is now rename/copy */ | |
685 | pair_to_free = p; | |
686 | } | |
687 | else { | |
688 | if (p->one->rename_used) | |
689 | /* this path remains */ | |
690 | pair_to_free = p; | |
691 | } | |
692 | ||
693 | if (pair_to_free) | |
694 | ; | |
695 | else | |
696 | diff_q(&outq, p); | |
697 | } | |
698 | else if (!diff_unmodified_pair(p)) | |
699 | /* all the usual ones need to be kept */ | |
700 | diff_q(&outq, p); | |
701 | else | |
702 | /* no need to keep unmodified pairs */ | |
703 | pair_to_free = p; | |
704 | ||
705 | if (pair_to_free) | |
706 | diff_free_filepair(pair_to_free); | |
707 | } | |
708 | diff_debug_queue("done copying original", &outq); | |
709 | ||
710 | free(q->queue); | |
711 | *q = outq; | |
712 | diff_debug_queue("done collapsing", q); | |
713 | ||
714 | for (i = 0; i < rename_dst_nr; i++) | |
715 | free_filespec(rename_dst[i].two); | |
716 | ||
717 | free(rename_dst); | |
718 | rename_dst = NULL; | |
719 | rename_dst_nr = rename_dst_alloc = 0; | |
720 | free(rename_src); | |
721 | rename_src = NULL; | |
722 | rename_src_nr = rename_src_alloc = 0; | |
723 | return; | |
724 | } |