]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Blame | |
3 | * | |
4 | * Copyright (c) 2006, 2014 by its authors | |
5 | * See COPYING for licensing conditions | |
6 | */ | |
7 | ||
8 | #include "cache.h" | |
9 | #include "refs.h" | |
10 | #include "builtin.h" | |
11 | #include "commit.h" | |
12 | #include "tag.h" | |
13 | #include "tree-walk.h" | |
14 | #include "diff.h" | |
15 | #include "diffcore.h" | |
16 | #include "revision.h" | |
17 | #include "quote.h" | |
18 | #include "xdiff-interface.h" | |
19 | #include "cache-tree.h" | |
20 | #include "string-list.h" | |
21 | #include "mailmap.h" | |
22 | #include "mergesort.h" | |
23 | #include "parse-options.h" | |
24 | #include "prio-queue.h" | |
25 | #include "utf8.h" | |
26 | #include "userdiff.h" | |
27 | #include "line-range.h" | |
28 | #include "line-log.h" | |
29 | #include "dir.h" | |
30 | #include "progress.h" | |
31 | ||
32 | static char blame_usage[] = N_("git blame [<options>] [<rev-opts>] [<rev>] [--] <file>"); | |
33 | ||
34 | static const char *blame_opt_usage[] = { | |
35 | blame_usage, | |
36 | "", | |
37 | N_("<rev-opts> are documented in git-rev-list(1)"), | |
38 | NULL | |
39 | }; | |
40 | ||
41 | static int longest_file; | |
42 | static int longest_author; | |
43 | static int max_orig_digits; | |
44 | static int max_digits; | |
45 | static int max_score_digits; | |
46 | static int show_root; | |
47 | static int reverse; | |
48 | static int blank_boundary; | |
49 | static int incremental; | |
50 | static int xdl_opts; | |
51 | static int abbrev = -1; | |
52 | static int no_whole_file_rename; | |
53 | static int show_progress; | |
54 | ||
55 | static struct date_mode blame_date_mode = { DATE_ISO8601 }; | |
56 | static size_t blame_date_width; | |
57 | ||
58 | static struct string_list mailmap = STRING_LIST_INIT_NODUP; | |
59 | ||
60 | #ifndef DEBUG | |
61 | #define DEBUG 0 | |
62 | #endif | |
63 | ||
64 | /* stats */ | |
65 | static int num_read_blob; | |
66 | static int num_get_patch; | |
67 | static int num_commits; | |
68 | ||
69 | #define PICKAXE_BLAME_MOVE 01 | |
70 | #define PICKAXE_BLAME_COPY 02 | |
71 | #define PICKAXE_BLAME_COPY_HARDER 04 | |
72 | #define PICKAXE_BLAME_COPY_HARDEST 010 | |
73 | ||
74 | /* | |
75 | * blame for a blame_entry with score lower than these thresholds | |
76 | * is not passed to the parent using move/copy logic. | |
77 | */ | |
78 | static unsigned blame_move_score; | |
79 | static unsigned blame_copy_score; | |
80 | #define BLAME_DEFAULT_MOVE_SCORE 20 | |
81 | #define BLAME_DEFAULT_COPY_SCORE 40 | |
82 | ||
83 | /* Remember to update object flag allocation in object.h */ | |
84 | #define METAINFO_SHOWN (1u<<12) | |
85 | #define MORE_THAN_ONE_PATH (1u<<13) | |
86 | ||
87 | /* | |
88 | * One blob in a commit that is being suspected | |
89 | */ | |
90 | struct blame_origin { | |
91 | int refcnt; | |
92 | /* Record preceding blame record for this blob */ | |
93 | struct blame_origin *previous; | |
94 | /* origins are put in a list linked via `next' hanging off the | |
95 | * corresponding commit's util field in order to make finding | |
96 | * them fast. The presence in this chain does not count | |
97 | * towards the origin's reference count. It is tempting to | |
98 | * let it count as long as the commit is pending examination, | |
99 | * but even under circumstances where the commit will be | |
100 | * present multiple times in the priority queue of unexamined | |
101 | * commits, processing the first instance will not leave any | |
102 | * work requiring the origin data for the second instance. An | |
103 | * interspersed commit changing that would have to be | |
104 | * preexisting with a different ancestry and with the same | |
105 | * commit date in order to wedge itself between two instances | |
106 | * of the same commit in the priority queue _and_ produce | |
107 | * blame entries relevant for it. While we don't want to let | |
108 | * us get tripped up by this case, it certainly does not seem | |
109 | * worth optimizing for. | |
110 | */ | |
111 | struct blame_origin *next; | |
112 | struct commit *commit; | |
113 | /* `suspects' contains blame entries that may be attributed to | |
114 | * this origin's commit or to parent commits. When a commit | |
115 | * is being processed, all suspects will be moved, either by | |
116 | * assigning them to an origin in a different commit, or by | |
117 | * shipping them to the scoreboard's ent list because they | |
118 | * cannot be attributed to a different commit. | |
119 | */ | |
120 | struct blame_entry *suspects; | |
121 | mmfile_t file; | |
122 | struct object_id blob_oid; | |
123 | unsigned mode; | |
124 | /* guilty gets set when shipping any suspects to the final | |
125 | * blame list instead of other commits | |
126 | */ | |
127 | char guilty; | |
128 | char path[FLEX_ARRAY]; | |
129 | }; | |
130 | ||
131 | struct progress_info { | |
132 | struct progress *progress; | |
133 | int blamed_lines; | |
134 | }; | |
135 | ||
136 | static int diff_hunks(mmfile_t *file_a, mmfile_t *file_b, | |
137 | xdl_emit_hunk_consume_func_t hunk_func, void *cb_data) | |
138 | { | |
139 | xpparam_t xpp = {0}; | |
140 | xdemitconf_t xecfg = {0}; | |
141 | xdemitcb_t ecb = {NULL}; | |
142 | ||
143 | xpp.flags = xdl_opts; | |
144 | xecfg.hunk_func = hunk_func; | |
145 | ecb.priv = cb_data; | |
146 | return xdi_diff(file_a, file_b, &xpp, &xecfg, &ecb); | |
147 | } | |
148 | ||
149 | /* | |
150 | * Given an origin, prepare mmfile_t structure to be used by the | |
151 | * diff machinery | |
152 | */ | |
153 | static void fill_origin_blob(struct diff_options *opt, | |
154 | struct blame_origin *o, mmfile_t *file) | |
155 | { | |
156 | if (!o->file.ptr) { | |
157 | enum object_type type; | |
158 | unsigned long file_size; | |
159 | ||
160 | num_read_blob++; | |
161 | if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) && | |
162 | textconv_object(o->path, o->mode, &o->blob_oid, 1, &file->ptr, &file_size)) | |
163 | ; | |
164 | else | |
165 | file->ptr = read_sha1_file(o->blob_oid.hash, &type, | |
166 | &file_size); | |
167 | file->size = file_size; | |
168 | ||
169 | if (!file->ptr) | |
170 | die("Cannot read blob %s for path %s", | |
171 | oid_to_hex(&o->blob_oid), | |
172 | o->path); | |
173 | o->file = *file; | |
174 | } | |
175 | else | |
176 | *file = o->file; | |
177 | } | |
178 | ||
179 | /* | |
180 | * Origin is refcounted and usually we keep the blob contents to be | |
181 | * reused. | |
182 | */ | |
183 | static inline struct blame_origin *blame_origin_incref(struct blame_origin *o) | |
184 | { | |
185 | if (o) | |
186 | o->refcnt++; | |
187 | return o; | |
188 | } | |
189 | ||
190 | static void blame_origin_decref(struct blame_origin *o) | |
191 | { | |
192 | if (o && --o->refcnt <= 0) { | |
193 | struct blame_origin *p, *l = NULL; | |
194 | if (o->previous) | |
195 | blame_origin_decref(o->previous); | |
196 | free(o->file.ptr); | |
197 | /* Should be present exactly once in commit chain */ | |
198 | for (p = o->commit->util; p; l = p, p = p->next) { | |
199 | if (p == o) { | |
200 | if (l) | |
201 | l->next = p->next; | |
202 | else | |
203 | o->commit->util = p->next; | |
204 | free(o); | |
205 | return; | |
206 | } | |
207 | } | |
208 | die("internal error in blame_origin_decref"); | |
209 | } | |
210 | } | |
211 | ||
212 | static void drop_origin_blob(struct blame_origin *o) | |
213 | { | |
214 | if (o->file.ptr) { | |
215 | free(o->file.ptr); | |
216 | o->file.ptr = NULL; | |
217 | } | |
218 | } | |
219 | ||
220 | /* | |
221 | * Each group of lines is described by a blame_entry; it can be split | |
222 | * as we pass blame to the parents. They are arranged in linked lists | |
223 | * kept as `suspects' of some unprocessed origin, or entered (when the | |
224 | * blame origin has been finalized) into the scoreboard structure. | |
225 | * While the scoreboard structure is only sorted at the end of | |
226 | * processing (according to final image line number), the lists | |
227 | * attached to an origin are sorted by the target line number. | |
228 | */ | |
229 | struct blame_entry { | |
230 | struct blame_entry *next; | |
231 | ||
232 | /* the first line of this group in the final image; | |
233 | * internally all line numbers are 0 based. | |
234 | */ | |
235 | int lno; | |
236 | ||
237 | /* how many lines this group has */ | |
238 | int num_lines; | |
239 | ||
240 | /* the commit that introduced this group into the final image */ | |
241 | struct blame_origin *suspect; | |
242 | ||
243 | /* the line number of the first line of this group in the | |
244 | * suspect's file; internally all line numbers are 0 based. | |
245 | */ | |
246 | int s_lno; | |
247 | ||
248 | /* how significant this entry is -- cached to avoid | |
249 | * scanning the lines over and over. | |
250 | */ | |
251 | unsigned score; | |
252 | }; | |
253 | ||
254 | /* | |
255 | * Any merge of blames happens on lists of blames that arrived via | |
256 | * different parents in a single suspect. In this case, we want to | |
257 | * sort according to the suspect line numbers as opposed to the final | |
258 | * image line numbers. The function body is somewhat longish because | |
259 | * it avoids unnecessary writes. | |
260 | */ | |
261 | ||
262 | static struct blame_entry *blame_merge(struct blame_entry *list1, | |
263 | struct blame_entry *list2) | |
264 | { | |
265 | struct blame_entry *p1 = list1, *p2 = list2, | |
266 | **tail = &list1; | |
267 | ||
268 | if (!p1) | |
269 | return p2; | |
270 | if (!p2) | |
271 | return p1; | |
272 | ||
273 | if (p1->s_lno <= p2->s_lno) { | |
274 | do { | |
275 | tail = &p1->next; | |
276 | if ((p1 = *tail) == NULL) { | |
277 | *tail = p2; | |
278 | return list1; | |
279 | } | |
280 | } while (p1->s_lno <= p2->s_lno); | |
281 | } | |
282 | for (;;) { | |
283 | *tail = p2; | |
284 | do { | |
285 | tail = &p2->next; | |
286 | if ((p2 = *tail) == NULL) { | |
287 | *tail = p1; | |
288 | return list1; | |
289 | } | |
290 | } while (p1->s_lno > p2->s_lno); | |
291 | *tail = p1; | |
292 | do { | |
293 | tail = &p1->next; | |
294 | if ((p1 = *tail) == NULL) { | |
295 | *tail = p2; | |
296 | return list1; | |
297 | } | |
298 | } while (p1->s_lno <= p2->s_lno); | |
299 | } | |
300 | } | |
301 | ||
302 | static void *get_next_blame(const void *p) | |
303 | { | |
304 | return ((struct blame_entry *)p)->next; | |
305 | } | |
306 | ||
307 | static void set_next_blame(void *p1, void *p2) | |
308 | { | |
309 | ((struct blame_entry *)p1)->next = p2; | |
310 | } | |
311 | ||
312 | /* | |
313 | * Final image line numbers are all different, so we don't need a | |
314 | * three-way comparison here. | |
315 | */ | |
316 | ||
317 | static int compare_blame_final(const void *p1, const void *p2) | |
318 | { | |
319 | return ((struct blame_entry *)p1)->lno > ((struct blame_entry *)p2)->lno | |
320 | ? 1 : -1; | |
321 | } | |
322 | ||
323 | static int compare_blame_suspect(const void *p1, const void *p2) | |
324 | { | |
325 | const struct blame_entry *s1 = p1, *s2 = p2; | |
326 | /* | |
327 | * to allow for collating suspects, we sort according to the | |
328 | * respective pointer value as the primary sorting criterion. | |
329 | * The actual relation is pretty unimportant as long as it | |
330 | * establishes a total order. Comparing as integers gives us | |
331 | * that. | |
332 | */ | |
333 | if (s1->suspect != s2->suspect) | |
334 | return (intptr_t)s1->suspect > (intptr_t)s2->suspect ? 1 : -1; | |
335 | if (s1->s_lno == s2->s_lno) | |
336 | return 0; | |
337 | return s1->s_lno > s2->s_lno ? 1 : -1; | |
338 | } | |
339 | ||
340 | static struct blame_entry *blame_sort(struct blame_entry *head, | |
341 | int (*compare_fn)(const void *, const void *)) | |
342 | { | |
343 | return llist_mergesort (head, get_next_blame, set_next_blame, compare_fn); | |
344 | } | |
345 | ||
346 | static int compare_commits_by_reverse_commit_date(const void *a, | |
347 | const void *b, | |
348 | void *c) | |
349 | { | |
350 | return -compare_commits_by_commit_date(a, b, c); | |
351 | } | |
352 | ||
353 | /* | |
354 | * The current state of the blame assignment. | |
355 | */ | |
356 | struct blame_scoreboard { | |
357 | /* the final commit (i.e. where we started digging from) */ | |
358 | struct commit *final; | |
359 | /* Priority queue for commits with unassigned blame records */ | |
360 | struct prio_queue commits; | |
361 | struct rev_info *revs; | |
362 | const char *path; | |
363 | ||
364 | /* | |
365 | * The contents in the final image. | |
366 | * Used by many functions to obtain contents of the nth line, | |
367 | * indexed with scoreboard.lineno[blame_entry.lno]. | |
368 | */ | |
369 | const char *final_buf; | |
370 | unsigned long final_buf_size; | |
371 | ||
372 | /* linked list of blames */ | |
373 | struct blame_entry *ent; | |
374 | ||
375 | /* look-up a line in the final buffer */ | |
376 | int num_lines; | |
377 | int *lineno; | |
378 | }; | |
379 | ||
380 | static void sanity_check_refcnt(struct blame_scoreboard *); | |
381 | ||
382 | /* | |
383 | * If two blame entries that are next to each other came from | |
384 | * contiguous lines in the same origin (i.e. <commit, path> pair), | |
385 | * merge them together. | |
386 | */ | |
387 | static void blame_coalesce(struct blame_scoreboard *sb) | |
388 | { | |
389 | struct blame_entry *ent, *next; | |
390 | ||
391 | for (ent = sb->ent; ent && (next = ent->next); ent = next) { | |
392 | if (ent->suspect == next->suspect && | |
393 | ent->s_lno + ent->num_lines == next->s_lno) { | |
394 | ent->num_lines += next->num_lines; | |
395 | ent->next = next->next; | |
396 | blame_origin_decref(next->suspect); | |
397 | free(next); | |
398 | ent->score = 0; | |
399 | next = ent; /* again */ | |
400 | } | |
401 | } | |
402 | ||
403 | if (DEBUG) /* sanity */ | |
404 | sanity_check_refcnt(sb); | |
405 | } | |
406 | ||
407 | /* | |
408 | * Merge the given sorted list of blames into a preexisting origin. | |
409 | * If there were no previous blames to that commit, it is entered into | |
410 | * the commit priority queue of the score board. | |
411 | */ | |
412 | ||
413 | static void queue_blames(struct blame_scoreboard *sb, struct blame_origin *porigin, | |
414 | struct blame_entry *sorted) | |
415 | { | |
416 | if (porigin->suspects) | |
417 | porigin->suspects = blame_merge(porigin->suspects, sorted); | |
418 | else { | |
419 | struct blame_origin *o; | |
420 | for (o = porigin->commit->util; o; o = o->next) { | |
421 | if (o->suspects) { | |
422 | porigin->suspects = sorted; | |
423 | return; | |
424 | } | |
425 | } | |
426 | porigin->suspects = sorted; | |
427 | prio_queue_put(&sb->commits, porigin->commit); | |
428 | } | |
429 | } | |
430 | ||
431 | /* | |
432 | * Given a commit and a path in it, create a new origin structure. | |
433 | * The callers that add blame to the scoreboard should use | |
434 | * get_origin() to obtain shared, refcounted copy instead of calling | |
435 | * this function directly. | |
436 | */ | |
437 | static struct blame_origin *make_origin(struct commit *commit, const char *path) | |
438 | { | |
439 | struct blame_origin *o; | |
440 | FLEX_ALLOC_STR(o, path, path); | |
441 | o->commit = commit; | |
442 | o->refcnt = 1; | |
443 | o->next = commit->util; | |
444 | commit->util = o; | |
445 | return o; | |
446 | } | |
447 | ||
448 | /* | |
449 | * Locate an existing origin or create a new one. | |
450 | * This moves the origin to front position in the commit util list. | |
451 | */ | |
452 | static struct blame_origin *get_origin(struct commit *commit, const char *path) | |
453 | { | |
454 | struct blame_origin *o, *l; | |
455 | ||
456 | for (o = commit->util, l = NULL; o; l = o, o = o->next) { | |
457 | if (!strcmp(o->path, path)) { | |
458 | /* bump to front */ | |
459 | if (l) { | |
460 | l->next = o->next; | |
461 | o->next = commit->util; | |
462 | commit->util = o; | |
463 | } | |
464 | return blame_origin_incref(o); | |
465 | } | |
466 | } | |
467 | return make_origin(commit, path); | |
468 | } | |
469 | ||
470 | /* | |
471 | * Fill the blob_sha1 field of an origin if it hasn't, so that later | |
472 | * call to fill_origin_blob() can use it to locate the data. blob_sha1 | |
473 | * for an origin is also used to pass the blame for the entire file to | |
474 | * the parent to detect the case where a child's blob is identical to | |
475 | * that of its parent's. | |
476 | * | |
477 | * This also fills origin->mode for corresponding tree path. | |
478 | */ | |
479 | static int fill_blob_sha1_and_mode(struct blame_origin *origin) | |
480 | { | |
481 | if (!is_null_oid(&origin->blob_oid)) | |
482 | return 0; | |
483 | if (get_tree_entry(origin->commit->object.oid.hash, | |
484 | origin->path, | |
485 | origin->blob_oid.hash, &origin->mode)) | |
486 | goto error_out; | |
487 | if (sha1_object_info(origin->blob_oid.hash, NULL) != OBJ_BLOB) | |
488 | goto error_out; | |
489 | return 0; | |
490 | error_out: | |
491 | oidclr(&origin->blob_oid); | |
492 | origin->mode = S_IFINVALID; | |
493 | return -1; | |
494 | } | |
495 | ||
496 | /* | |
497 | * We have an origin -- check if the same path exists in the | |
498 | * parent and return an origin structure to represent it. | |
499 | */ | |
500 | static struct blame_origin *find_origin(struct commit *parent, | |
501 | struct blame_origin *origin) | |
502 | { | |
503 | struct blame_origin *porigin; | |
504 | struct diff_options diff_opts; | |
505 | const char *paths[2]; | |
506 | ||
507 | /* First check any existing origins */ | |
508 | for (porigin = parent->util; porigin; porigin = porigin->next) | |
509 | if (!strcmp(porigin->path, origin->path)) { | |
510 | /* | |
511 | * The same path between origin and its parent | |
512 | * without renaming -- the most common case. | |
513 | */ | |
514 | return blame_origin_incref (porigin); | |
515 | } | |
516 | ||
517 | /* See if the origin->path is different between parent | |
518 | * and origin first. Most of the time they are the | |
519 | * same and diff-tree is fairly efficient about this. | |
520 | */ | |
521 | diff_setup(&diff_opts); | |
522 | DIFF_OPT_SET(&diff_opts, RECURSIVE); | |
523 | diff_opts.detect_rename = 0; | |
524 | diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT; | |
525 | paths[0] = origin->path; | |
526 | paths[1] = NULL; | |
527 | ||
528 | parse_pathspec(&diff_opts.pathspec, | |
529 | PATHSPEC_ALL_MAGIC & ~PATHSPEC_LITERAL, | |
530 | PATHSPEC_LITERAL_PATH, "", paths); | |
531 | diff_setup_done(&diff_opts); | |
532 | ||
533 | if (is_null_oid(&origin->commit->object.oid)) | |
534 | do_diff_cache(parent->tree->object.oid.hash, &diff_opts); | |
535 | else | |
536 | diff_tree_sha1(parent->tree->object.oid.hash, | |
537 | origin->commit->tree->object.oid.hash, | |
538 | "", &diff_opts); | |
539 | diffcore_std(&diff_opts); | |
540 | ||
541 | if (!diff_queued_diff.nr) { | |
542 | /* The path is the same as parent */ | |
543 | porigin = get_origin(parent, origin->path); | |
544 | oidcpy(&porigin->blob_oid, &origin->blob_oid); | |
545 | porigin->mode = origin->mode; | |
546 | } else { | |
547 | /* | |
548 | * Since origin->path is a pathspec, if the parent | |
549 | * commit had it as a directory, we will see a whole | |
550 | * bunch of deletion of files in the directory that we | |
551 | * do not care about. | |
552 | */ | |
553 | int i; | |
554 | struct diff_filepair *p = NULL; | |
555 | for (i = 0; i < diff_queued_diff.nr; i++) { | |
556 | const char *name; | |
557 | p = diff_queued_diff.queue[i]; | |
558 | name = p->one->path ? p->one->path : p->two->path; | |
559 | if (!strcmp(name, origin->path)) | |
560 | break; | |
561 | } | |
562 | if (!p) | |
563 | die("internal error in blame::find_origin"); | |
564 | switch (p->status) { | |
565 | default: | |
566 | die("internal error in blame::find_origin (%c)", | |
567 | p->status); | |
568 | case 'M': | |
569 | porigin = get_origin(parent, origin->path); | |
570 | oidcpy(&porigin->blob_oid, &p->one->oid); | |
571 | porigin->mode = p->one->mode; | |
572 | break; | |
573 | case 'A': | |
574 | case 'T': | |
575 | /* Did not exist in parent, or type changed */ | |
576 | break; | |
577 | } | |
578 | } | |
579 | diff_flush(&diff_opts); | |
580 | clear_pathspec(&diff_opts.pathspec); | |
581 | return porigin; | |
582 | } | |
583 | ||
584 | /* | |
585 | * We have an origin -- find the path that corresponds to it in its | |
586 | * parent and return an origin structure to represent it. | |
587 | */ | |
588 | static struct blame_origin *find_rename(struct commit *parent, | |
589 | struct blame_origin *origin) | |
590 | { | |
591 | struct blame_origin *porigin = NULL; | |
592 | struct diff_options diff_opts; | |
593 | int i; | |
594 | ||
595 | diff_setup(&diff_opts); | |
596 | DIFF_OPT_SET(&diff_opts, RECURSIVE); | |
597 | diff_opts.detect_rename = DIFF_DETECT_RENAME; | |
598 | diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT; | |
599 | diff_opts.single_follow = origin->path; | |
600 | diff_setup_done(&diff_opts); | |
601 | ||
602 | if (is_null_oid(&origin->commit->object.oid)) | |
603 | do_diff_cache(parent->tree->object.oid.hash, &diff_opts); | |
604 | else | |
605 | diff_tree_sha1(parent->tree->object.oid.hash, | |
606 | origin->commit->tree->object.oid.hash, | |
607 | "", &diff_opts); | |
608 | diffcore_std(&diff_opts); | |
609 | ||
610 | for (i = 0; i < diff_queued_diff.nr; i++) { | |
611 | struct diff_filepair *p = diff_queued_diff.queue[i]; | |
612 | if ((p->status == 'R' || p->status == 'C') && | |
613 | !strcmp(p->two->path, origin->path)) { | |
614 | porigin = get_origin(parent, p->one->path); | |
615 | oidcpy(&porigin->blob_oid, &p->one->oid); | |
616 | porigin->mode = p->one->mode; | |
617 | break; | |
618 | } | |
619 | } | |
620 | diff_flush(&diff_opts); | |
621 | clear_pathspec(&diff_opts.pathspec); | |
622 | return porigin; | |
623 | } | |
624 | ||
625 | /* | |
626 | * Append a new blame entry to a given output queue. | |
627 | */ | |
628 | static void add_blame_entry(struct blame_entry ***queue, | |
629 | const struct blame_entry *src) | |
630 | { | |
631 | struct blame_entry *e = xmalloc(sizeof(*e)); | |
632 | memcpy(e, src, sizeof(*e)); | |
633 | blame_origin_incref(e->suspect); | |
634 | ||
635 | e->next = **queue; | |
636 | **queue = e; | |
637 | *queue = &e->next; | |
638 | } | |
639 | ||
640 | /* | |
641 | * src typically is on-stack; we want to copy the information in it to | |
642 | * a malloced blame_entry that gets added to the given queue. The | |
643 | * origin of dst loses a refcnt. | |
644 | */ | |
645 | static void dup_entry(struct blame_entry ***queue, | |
646 | struct blame_entry *dst, struct blame_entry *src) | |
647 | { | |
648 | blame_origin_incref(src->suspect); | |
649 | blame_origin_decref(dst->suspect); | |
650 | memcpy(dst, src, sizeof(*src)); | |
651 | dst->next = **queue; | |
652 | **queue = dst; | |
653 | *queue = &dst->next; | |
654 | } | |
655 | ||
656 | static const char *nth_line(struct blame_scoreboard *sb, long lno) | |
657 | { | |
658 | return sb->final_buf + sb->lineno[lno]; | |
659 | } | |
660 | ||
661 | static const char *nth_line_cb(void *data, long lno) | |
662 | { | |
663 | return nth_line((struct blame_scoreboard *)data, lno); | |
664 | } | |
665 | ||
666 | /* | |
667 | * It is known that lines between tlno to same came from parent, and e | |
668 | * has an overlap with that range. it also is known that parent's | |
669 | * line plno corresponds to e's line tlno. | |
670 | * | |
671 | * <---- e -----> | |
672 | * <------> | |
673 | * <------------> | |
674 | * <------------> | |
675 | * <------------------> | |
676 | * | |
677 | * Split e into potentially three parts; before this chunk, the chunk | |
678 | * to be blamed for the parent, and after that portion. | |
679 | */ | |
680 | static void split_overlap(struct blame_entry *split, | |
681 | struct blame_entry *e, | |
682 | int tlno, int plno, int same, | |
683 | struct blame_origin *parent) | |
684 | { | |
685 | int chunk_end_lno; | |
686 | memset(split, 0, sizeof(struct blame_entry [3])); | |
687 | ||
688 | if (e->s_lno < tlno) { | |
689 | /* there is a pre-chunk part not blamed on parent */ | |
690 | split[0].suspect = blame_origin_incref(e->suspect); | |
691 | split[0].lno = e->lno; | |
692 | split[0].s_lno = e->s_lno; | |
693 | split[0].num_lines = tlno - e->s_lno; | |
694 | split[1].lno = e->lno + tlno - e->s_lno; | |
695 | split[1].s_lno = plno; | |
696 | } | |
697 | else { | |
698 | split[1].lno = e->lno; | |
699 | split[1].s_lno = plno + (e->s_lno - tlno); | |
700 | } | |
701 | ||
702 | if (same < e->s_lno + e->num_lines) { | |
703 | /* there is a post-chunk part not blamed on parent */ | |
704 | split[2].suspect = blame_origin_incref(e->suspect); | |
705 | split[2].lno = e->lno + (same - e->s_lno); | |
706 | split[2].s_lno = e->s_lno + (same - e->s_lno); | |
707 | split[2].num_lines = e->s_lno + e->num_lines - same; | |
708 | chunk_end_lno = split[2].lno; | |
709 | } | |
710 | else | |
711 | chunk_end_lno = e->lno + e->num_lines; | |
712 | split[1].num_lines = chunk_end_lno - split[1].lno; | |
713 | ||
714 | /* | |
715 | * if it turns out there is nothing to blame the parent for, | |
716 | * forget about the splitting. !split[1].suspect signals this. | |
717 | */ | |
718 | if (split[1].num_lines < 1) | |
719 | return; | |
720 | split[1].suspect = blame_origin_incref(parent); | |
721 | } | |
722 | ||
723 | /* | |
724 | * split_overlap() divided an existing blame e into up to three parts | |
725 | * in split. Any assigned blame is moved to queue to | |
726 | * reflect the split. | |
727 | */ | |
728 | static void split_blame(struct blame_entry ***blamed, | |
729 | struct blame_entry ***unblamed, | |
730 | struct blame_entry *split, | |
731 | struct blame_entry *e) | |
732 | { | |
733 | if (split[0].suspect && split[2].suspect) { | |
734 | /* The first part (reuse storage for the existing entry e) */ | |
735 | dup_entry(unblamed, e, &split[0]); | |
736 | ||
737 | /* The last part -- me */ | |
738 | add_blame_entry(unblamed, &split[2]); | |
739 | ||
740 | /* ... and the middle part -- parent */ | |
741 | add_blame_entry(blamed, &split[1]); | |
742 | } | |
743 | else if (!split[0].suspect && !split[2].suspect) | |
744 | /* | |
745 | * The parent covers the entire area; reuse storage for | |
746 | * e and replace it with the parent. | |
747 | */ | |
748 | dup_entry(blamed, e, &split[1]); | |
749 | else if (split[0].suspect) { | |
750 | /* me and then parent */ | |
751 | dup_entry(unblamed, e, &split[0]); | |
752 | add_blame_entry(blamed, &split[1]); | |
753 | } | |
754 | else { | |
755 | /* parent and then me */ | |
756 | dup_entry(blamed, e, &split[1]); | |
757 | add_blame_entry(unblamed, &split[2]); | |
758 | } | |
759 | } | |
760 | ||
761 | /* | |
762 | * After splitting the blame, the origins used by the | |
763 | * on-stack blame_entry should lose one refcnt each. | |
764 | */ | |
765 | static void decref_split(struct blame_entry *split) | |
766 | { | |
767 | int i; | |
768 | ||
769 | for (i = 0; i < 3; i++) | |
770 | blame_origin_decref(split[i].suspect); | |
771 | } | |
772 | ||
773 | /* | |
774 | * reverse_blame reverses the list given in head, appending tail. | |
775 | * That allows us to build lists in reverse order, then reverse them | |
776 | * afterwards. This can be faster than building the list in proper | |
777 | * order right away. The reason is that building in proper order | |
778 | * requires writing a link in the _previous_ element, while building | |
779 | * in reverse order just requires placing the list head into the | |
780 | * _current_ element. | |
781 | */ | |
782 | ||
783 | static struct blame_entry *reverse_blame(struct blame_entry *head, | |
784 | struct blame_entry *tail) | |
785 | { | |
786 | while (head) { | |
787 | struct blame_entry *next = head->next; | |
788 | head->next = tail; | |
789 | tail = head; | |
790 | head = next; | |
791 | } | |
792 | return tail; | |
793 | } | |
794 | ||
795 | /* | |
796 | * Process one hunk from the patch between the current suspect for | |
797 | * blame_entry e and its parent. This first blames any unfinished | |
798 | * entries before the chunk (which is where target and parent start | |
799 | * differing) on the parent, and then splits blame entries at the | |
800 | * start and at the end of the difference region. Since use of -M and | |
801 | * -C options may lead to overlapping/duplicate source line number | |
802 | * ranges, all we can rely on from sorting/merging is the order of the | |
803 | * first suspect line number. | |
804 | */ | |
805 | static void blame_chunk(struct blame_entry ***dstq, struct blame_entry ***srcq, | |
806 | int tlno, int offset, int same, | |
807 | struct blame_origin *parent) | |
808 | { | |
809 | struct blame_entry *e = **srcq; | |
810 | struct blame_entry *samep = NULL, *diffp = NULL; | |
811 | ||
812 | while (e && e->s_lno < tlno) { | |
813 | struct blame_entry *next = e->next; | |
814 | /* | |
815 | * current record starts before differing portion. If | |
816 | * it reaches into it, we need to split it up and | |
817 | * examine the second part separately. | |
818 | */ | |
819 | if (e->s_lno + e->num_lines > tlno) { | |
820 | /* Move second half to a new record */ | |
821 | int len = tlno - e->s_lno; | |
822 | struct blame_entry *n = xcalloc(1, sizeof (struct blame_entry)); | |
823 | n->suspect = e->suspect; | |
824 | n->lno = e->lno + len; | |
825 | n->s_lno = e->s_lno + len; | |
826 | n->num_lines = e->num_lines - len; | |
827 | e->num_lines = len; | |
828 | e->score = 0; | |
829 | /* Push new record to diffp */ | |
830 | n->next = diffp; | |
831 | diffp = n; | |
832 | } else | |
833 | blame_origin_decref(e->suspect); | |
834 | /* Pass blame for everything before the differing | |
835 | * chunk to the parent */ | |
836 | e->suspect = blame_origin_incref(parent); | |
837 | e->s_lno += offset; | |
838 | e->next = samep; | |
839 | samep = e; | |
840 | e = next; | |
841 | } | |
842 | /* | |
843 | * As we don't know how much of a common stretch after this | |
844 | * diff will occur, the currently blamed parts are all that we | |
845 | * can assign to the parent for now. | |
846 | */ | |
847 | ||
848 | if (samep) { | |
849 | **dstq = reverse_blame(samep, **dstq); | |
850 | *dstq = &samep->next; | |
851 | } | |
852 | /* | |
853 | * Prepend the split off portions: everything after e starts | |
854 | * after the blameable portion. | |
855 | */ | |
856 | e = reverse_blame(diffp, e); | |
857 | ||
858 | /* | |
859 | * Now retain records on the target while parts are different | |
860 | * from the parent. | |
861 | */ | |
862 | samep = NULL; | |
863 | diffp = NULL; | |
864 | while (e && e->s_lno < same) { | |
865 | struct blame_entry *next = e->next; | |
866 | ||
867 | /* | |
868 | * If current record extends into sameness, need to split. | |
869 | */ | |
870 | if (e->s_lno + e->num_lines > same) { | |
871 | /* | |
872 | * Move second half to a new record to be | |
873 | * processed by later chunks | |
874 | */ | |
875 | int len = same - e->s_lno; | |
876 | struct blame_entry *n = xcalloc(1, sizeof (struct blame_entry)); | |
877 | n->suspect = blame_origin_incref(e->suspect); | |
878 | n->lno = e->lno + len; | |
879 | n->s_lno = e->s_lno + len; | |
880 | n->num_lines = e->num_lines - len; | |
881 | e->num_lines = len; | |
882 | e->score = 0; | |
883 | /* Push new record to samep */ | |
884 | n->next = samep; | |
885 | samep = n; | |
886 | } | |
887 | e->next = diffp; | |
888 | diffp = e; | |
889 | e = next; | |
890 | } | |
891 | **srcq = reverse_blame(diffp, reverse_blame(samep, e)); | |
892 | /* Move across elements that are in the unblamable portion */ | |
893 | if (diffp) | |
894 | *srcq = &diffp->next; | |
895 | } | |
896 | ||
897 | struct blame_chunk_cb_data { | |
898 | struct blame_origin *parent; | |
899 | long offset; | |
900 | struct blame_entry **dstq; | |
901 | struct blame_entry **srcq; | |
902 | }; | |
903 | ||
904 | /* diff chunks are from parent to target */ | |
905 | static int blame_chunk_cb(long start_a, long count_a, | |
906 | long start_b, long count_b, void *data) | |
907 | { | |
908 | struct blame_chunk_cb_data *d = data; | |
909 | if (start_a - start_b != d->offset) | |
910 | die("internal error in blame::blame_chunk_cb"); | |
911 | blame_chunk(&d->dstq, &d->srcq, start_b, start_a - start_b, | |
912 | start_b + count_b, d->parent); | |
913 | d->offset = start_a + count_a - (start_b + count_b); | |
914 | return 0; | |
915 | } | |
916 | ||
917 | /* | |
918 | * We are looking at the origin 'target' and aiming to pass blame | |
919 | * for the lines it is suspected to its parent. Run diff to find | |
920 | * which lines came from parent and pass blame for them. | |
921 | */ | |
922 | static void pass_blame_to_parent(struct blame_scoreboard *sb, | |
923 | struct blame_origin *target, | |
924 | struct blame_origin *parent) | |
925 | { | |
926 | mmfile_t file_p, file_o; | |
927 | struct blame_chunk_cb_data d; | |
928 | struct blame_entry *newdest = NULL; | |
929 | ||
930 | if (!target->suspects) | |
931 | return; /* nothing remains for this target */ | |
932 | ||
933 | d.parent = parent; | |
934 | d.offset = 0; | |
935 | d.dstq = &newdest; d.srcq = &target->suspects; | |
936 | ||
937 | fill_origin_blob(&sb->revs->diffopt, parent, &file_p); | |
938 | fill_origin_blob(&sb->revs->diffopt, target, &file_o); | |
939 | num_get_patch++; | |
940 | ||
941 | if (diff_hunks(&file_p, &file_o, blame_chunk_cb, &d)) | |
942 | die("unable to generate diff (%s -> %s)", | |
943 | oid_to_hex(&parent->commit->object.oid), | |
944 | oid_to_hex(&target->commit->object.oid)); | |
945 | /* The rest are the same as the parent */ | |
946 | blame_chunk(&d.dstq, &d.srcq, INT_MAX, d.offset, INT_MAX, parent); | |
947 | *d.dstq = NULL; | |
948 | queue_blames(sb, parent, newdest); | |
949 | ||
950 | return; | |
951 | } | |
952 | ||
953 | /* | |
954 | * The lines in blame_entry after splitting blames many times can become | |
955 | * very small and trivial, and at some point it becomes pointless to | |
956 | * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any | |
957 | * ordinary C program, and it is not worth to say it was copied from | |
958 | * totally unrelated file in the parent. | |
959 | * | |
960 | * Compute how trivial the lines in the blame_entry are. | |
961 | */ | |
962 | static unsigned ent_score(struct blame_scoreboard *sb, struct blame_entry *e) | |
963 | { | |
964 | unsigned score; | |
965 | const char *cp, *ep; | |
966 | ||
967 | if (e->score) | |
968 | return e->score; | |
969 | ||
970 | score = 1; | |
971 | cp = nth_line(sb, e->lno); | |
972 | ep = nth_line(sb, e->lno + e->num_lines); | |
973 | while (cp < ep) { | |
974 | unsigned ch = *((unsigned char *)cp); | |
975 | if (isalnum(ch)) | |
976 | score++; | |
977 | cp++; | |
978 | } | |
979 | e->score = score; | |
980 | return score; | |
981 | } | |
982 | ||
983 | /* | |
984 | * best_so_far[] and this[] are both a split of an existing blame_entry | |
985 | * that passes blame to the parent. Maintain best_so_far the best split | |
986 | * so far, by comparing this and best_so_far and copying this into | |
987 | * bst_so_far as needed. | |
988 | */ | |
989 | static void copy_split_if_better(struct blame_scoreboard *sb, | |
990 | struct blame_entry *best_so_far, | |
991 | struct blame_entry *this) | |
992 | { | |
993 | int i; | |
994 | ||
995 | if (!this[1].suspect) | |
996 | return; | |
997 | if (best_so_far[1].suspect) { | |
998 | if (ent_score(sb, &this[1]) < ent_score(sb, &best_so_far[1])) | |
999 | return; | |
1000 | } | |
1001 | ||
1002 | for (i = 0; i < 3; i++) | |
1003 | blame_origin_incref(this[i].suspect); | |
1004 | decref_split(best_so_far); | |
1005 | memcpy(best_so_far, this, sizeof(struct blame_entry [3])); | |
1006 | } | |
1007 | ||
1008 | /* | |
1009 | * We are looking at a part of the final image represented by | |
1010 | * ent (tlno and same are offset by ent->s_lno). | |
1011 | * tlno is where we are looking at in the final image. | |
1012 | * up to (but not including) same match preimage. | |
1013 | * plno is where we are looking at in the preimage. | |
1014 | * | |
1015 | * <-------------- final image ----------------------> | |
1016 | * <------ent------> | |
1017 | * ^tlno ^same | |
1018 | * <---------preimage-----> | |
1019 | * ^plno | |
1020 | * | |
1021 | * All line numbers are 0-based. | |
1022 | */ | |
1023 | static void handle_split(struct blame_scoreboard *sb, | |
1024 | struct blame_entry *ent, | |
1025 | int tlno, int plno, int same, | |
1026 | struct blame_origin *parent, | |
1027 | struct blame_entry *split) | |
1028 | { | |
1029 | if (ent->num_lines <= tlno) | |
1030 | return; | |
1031 | if (tlno < same) { | |
1032 | struct blame_entry this[3]; | |
1033 | tlno += ent->s_lno; | |
1034 | same += ent->s_lno; | |
1035 | split_overlap(this, ent, tlno, plno, same, parent); | |
1036 | copy_split_if_better(sb, split, this); | |
1037 | decref_split(this); | |
1038 | } | |
1039 | } | |
1040 | ||
1041 | struct handle_split_cb_data { | |
1042 | struct blame_scoreboard *sb; | |
1043 | struct blame_entry *ent; | |
1044 | struct blame_origin *parent; | |
1045 | struct blame_entry *split; | |
1046 | long plno; | |
1047 | long tlno; | |
1048 | }; | |
1049 | ||
1050 | static int handle_split_cb(long start_a, long count_a, | |
1051 | long start_b, long count_b, void *data) | |
1052 | { | |
1053 | struct handle_split_cb_data *d = data; | |
1054 | handle_split(d->sb, d->ent, d->tlno, d->plno, start_b, d->parent, | |
1055 | d->split); | |
1056 | d->plno = start_a + count_a; | |
1057 | d->tlno = start_b + count_b; | |
1058 | return 0; | |
1059 | } | |
1060 | ||
1061 | /* | |
1062 | * Find the lines from parent that are the same as ent so that | |
1063 | * we can pass blames to it. file_p has the blob contents for | |
1064 | * the parent. | |
1065 | */ | |
1066 | static void find_copy_in_blob(struct blame_scoreboard *sb, | |
1067 | struct blame_entry *ent, | |
1068 | struct blame_origin *parent, | |
1069 | struct blame_entry *split, | |
1070 | mmfile_t *file_p) | |
1071 | { | |
1072 | const char *cp; | |
1073 | mmfile_t file_o; | |
1074 | struct handle_split_cb_data d; | |
1075 | ||
1076 | memset(&d, 0, sizeof(d)); | |
1077 | d.sb = sb; d.ent = ent; d.parent = parent; d.split = split; | |
1078 | /* | |
1079 | * Prepare mmfile that contains only the lines in ent. | |
1080 | */ | |
1081 | cp = nth_line(sb, ent->lno); | |
1082 | file_o.ptr = (char *) cp; | |
1083 | file_o.size = nth_line(sb, ent->lno + ent->num_lines) - cp; | |
1084 | ||
1085 | /* | |
1086 | * file_o is a part of final image we are annotating. | |
1087 | * file_p partially may match that image. | |
1088 | */ | |
1089 | memset(split, 0, sizeof(struct blame_entry [3])); | |
1090 | if (diff_hunks(file_p, &file_o, handle_split_cb, &d)) | |
1091 | die("unable to generate diff (%s)", | |
1092 | oid_to_hex(&parent->commit->object.oid)); | |
1093 | /* remainder, if any, all match the preimage */ | |
1094 | handle_split(sb, ent, d.tlno, d.plno, ent->num_lines, parent, split); | |
1095 | } | |
1096 | ||
1097 | /* Move all blame entries from list *source that have a score smaller | |
1098 | * than score_min to the front of list *small. | |
1099 | * Returns a pointer to the link pointing to the old head of the small list. | |
1100 | */ | |
1101 | ||
1102 | static struct blame_entry **filter_small(struct blame_scoreboard *sb, | |
1103 | struct blame_entry **small, | |
1104 | struct blame_entry **source, | |
1105 | unsigned score_min) | |
1106 | { | |
1107 | struct blame_entry *p = *source; | |
1108 | struct blame_entry *oldsmall = *small; | |
1109 | while (p) { | |
1110 | if (ent_score(sb, p) <= score_min) { | |
1111 | *small = p; | |
1112 | small = &p->next; | |
1113 | p = *small; | |
1114 | } else { | |
1115 | *source = p; | |
1116 | source = &p->next; | |
1117 | p = *source; | |
1118 | } | |
1119 | } | |
1120 | *small = oldsmall; | |
1121 | *source = NULL; | |
1122 | return small; | |
1123 | } | |
1124 | ||
1125 | /* | |
1126 | * See if lines currently target is suspected for can be attributed to | |
1127 | * parent. | |
1128 | */ | |
1129 | static void find_move_in_parent(struct blame_scoreboard *sb, | |
1130 | struct blame_entry ***blamed, | |
1131 | struct blame_entry **toosmall, | |
1132 | struct blame_origin *target, | |
1133 | struct blame_origin *parent) | |
1134 | { | |
1135 | struct blame_entry *e, split[3]; | |
1136 | struct blame_entry *unblamed = target->suspects; | |
1137 | struct blame_entry *leftover = NULL; | |
1138 | mmfile_t file_p; | |
1139 | ||
1140 | if (!unblamed) | |
1141 | return; /* nothing remains for this target */ | |
1142 | ||
1143 | fill_origin_blob(&sb->revs->diffopt, parent, &file_p); | |
1144 | if (!file_p.ptr) | |
1145 | return; | |
1146 | ||
1147 | /* At each iteration, unblamed has a NULL-terminated list of | |
1148 | * entries that have not yet been tested for blame. leftover | |
1149 | * contains the reversed list of entries that have been tested | |
1150 | * without being assignable to the parent. | |
1151 | */ | |
1152 | do { | |
1153 | struct blame_entry **unblamedtail = &unblamed; | |
1154 | struct blame_entry *next; | |
1155 | for (e = unblamed; e; e = next) { | |
1156 | next = e->next; | |
1157 | find_copy_in_blob(sb, e, parent, split, &file_p); | |
1158 | if (split[1].suspect && | |
1159 | blame_move_score < ent_score(sb, &split[1])) { | |
1160 | split_blame(blamed, &unblamedtail, split, e); | |
1161 | } else { | |
1162 | e->next = leftover; | |
1163 | leftover = e; | |
1164 | } | |
1165 | decref_split(split); | |
1166 | } | |
1167 | *unblamedtail = NULL; | |
1168 | toosmall = filter_small(sb, toosmall, &unblamed, blame_move_score); | |
1169 | } while (unblamed); | |
1170 | target->suspects = reverse_blame(leftover, NULL); | |
1171 | } | |
1172 | ||
1173 | struct blame_list { | |
1174 | struct blame_entry *ent; | |
1175 | struct blame_entry split[3]; | |
1176 | }; | |
1177 | ||
1178 | /* | |
1179 | * Count the number of entries the target is suspected for, | |
1180 | * and prepare a list of entry and the best split. | |
1181 | */ | |
1182 | static struct blame_list *setup_blame_list(struct blame_entry *unblamed, | |
1183 | int *num_ents_p) | |
1184 | { | |
1185 | struct blame_entry *e; | |
1186 | int num_ents, i; | |
1187 | struct blame_list *blame_list = NULL; | |
1188 | ||
1189 | for (e = unblamed, num_ents = 0; e; e = e->next) | |
1190 | num_ents++; | |
1191 | if (num_ents) { | |
1192 | blame_list = xcalloc(num_ents, sizeof(struct blame_list)); | |
1193 | for (e = unblamed, i = 0; e; e = e->next) | |
1194 | blame_list[i++].ent = e; | |
1195 | } | |
1196 | *num_ents_p = num_ents; | |
1197 | return blame_list; | |
1198 | } | |
1199 | ||
1200 | /* | |
1201 | * For lines target is suspected for, see if we can find code movement | |
1202 | * across file boundary from the parent commit. porigin is the path | |
1203 | * in the parent we already tried. | |
1204 | */ | |
1205 | static void find_copy_in_parent(struct blame_scoreboard *sb, | |
1206 | struct blame_entry ***blamed, | |
1207 | struct blame_entry **toosmall, | |
1208 | struct blame_origin *target, | |
1209 | struct commit *parent, | |
1210 | struct blame_origin *porigin, | |
1211 | int opt) | |
1212 | { | |
1213 | struct diff_options diff_opts; | |
1214 | int i, j; | |
1215 | struct blame_list *blame_list; | |
1216 | int num_ents; | |
1217 | struct blame_entry *unblamed = target->suspects; | |
1218 | struct blame_entry *leftover = NULL; | |
1219 | ||
1220 | if (!unblamed) | |
1221 | return; /* nothing remains for this target */ | |
1222 | ||
1223 | diff_setup(&diff_opts); | |
1224 | DIFF_OPT_SET(&diff_opts, RECURSIVE); | |
1225 | diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT; | |
1226 | ||
1227 | diff_setup_done(&diff_opts); | |
1228 | ||
1229 | /* Try "find copies harder" on new path if requested; | |
1230 | * we do not want to use diffcore_rename() actually to | |
1231 | * match things up; find_copies_harder is set only to | |
1232 | * force diff_tree_sha1() to feed all filepairs to diff_queue, | |
1233 | * and this code needs to be after diff_setup_done(), which | |
1234 | * usually makes find-copies-harder imply copy detection. | |
1235 | */ | |
1236 | if ((opt & PICKAXE_BLAME_COPY_HARDEST) | |
1237 | || ((opt & PICKAXE_BLAME_COPY_HARDER) | |
1238 | && (!porigin || strcmp(target->path, porigin->path)))) | |
1239 | DIFF_OPT_SET(&diff_opts, FIND_COPIES_HARDER); | |
1240 | ||
1241 | if (is_null_oid(&target->commit->object.oid)) | |
1242 | do_diff_cache(parent->tree->object.oid.hash, &diff_opts); | |
1243 | else | |
1244 | diff_tree_sha1(parent->tree->object.oid.hash, | |
1245 | target->commit->tree->object.oid.hash, | |
1246 | "", &diff_opts); | |
1247 | ||
1248 | if (!DIFF_OPT_TST(&diff_opts, FIND_COPIES_HARDER)) | |
1249 | diffcore_std(&diff_opts); | |
1250 | ||
1251 | do { | |
1252 | struct blame_entry **unblamedtail = &unblamed; | |
1253 | blame_list = setup_blame_list(unblamed, &num_ents); | |
1254 | ||
1255 | for (i = 0; i < diff_queued_diff.nr; i++) { | |
1256 | struct diff_filepair *p = diff_queued_diff.queue[i]; | |
1257 | struct blame_origin *norigin; | |
1258 | mmfile_t file_p; | |
1259 | struct blame_entry this[3]; | |
1260 | ||
1261 | if (!DIFF_FILE_VALID(p->one)) | |
1262 | continue; /* does not exist in parent */ | |
1263 | if (S_ISGITLINK(p->one->mode)) | |
1264 | continue; /* ignore git links */ | |
1265 | if (porigin && !strcmp(p->one->path, porigin->path)) | |
1266 | /* find_move already dealt with this path */ | |
1267 | continue; | |
1268 | ||
1269 | norigin = get_origin(parent, p->one->path); | |
1270 | oidcpy(&norigin->blob_oid, &p->one->oid); | |
1271 | norigin->mode = p->one->mode; | |
1272 | fill_origin_blob(&sb->revs->diffopt, norigin, &file_p); | |
1273 | if (!file_p.ptr) | |
1274 | continue; | |
1275 | ||
1276 | for (j = 0; j < num_ents; j++) { | |
1277 | find_copy_in_blob(sb, blame_list[j].ent, | |
1278 | norigin, this, &file_p); | |
1279 | copy_split_if_better(sb, blame_list[j].split, | |
1280 | this); | |
1281 | decref_split(this); | |
1282 | } | |
1283 | blame_origin_decref(norigin); | |
1284 | } | |
1285 | ||
1286 | for (j = 0; j < num_ents; j++) { | |
1287 | struct blame_entry *split = blame_list[j].split; | |
1288 | if (split[1].suspect && | |
1289 | blame_copy_score < ent_score(sb, &split[1])) { | |
1290 | split_blame(blamed, &unblamedtail, split, | |
1291 | blame_list[j].ent); | |
1292 | } else { | |
1293 | blame_list[j].ent->next = leftover; | |
1294 | leftover = blame_list[j].ent; | |
1295 | } | |
1296 | decref_split(split); | |
1297 | } | |
1298 | free(blame_list); | |
1299 | *unblamedtail = NULL; | |
1300 | toosmall = filter_small(sb, toosmall, &unblamed, blame_copy_score); | |
1301 | } while (unblamed); | |
1302 | target->suspects = reverse_blame(leftover, NULL); | |
1303 | diff_flush(&diff_opts); | |
1304 | clear_pathspec(&diff_opts.pathspec); | |
1305 | } | |
1306 | ||
1307 | /* | |
1308 | * The blobs of origin and porigin exactly match, so everything | |
1309 | * origin is suspected for can be blamed on the parent. | |
1310 | */ | |
1311 | static void pass_whole_blame(struct blame_scoreboard *sb, | |
1312 | struct blame_origin *origin, struct blame_origin *porigin) | |
1313 | { | |
1314 | struct blame_entry *e, *suspects; | |
1315 | ||
1316 | if (!porigin->file.ptr && origin->file.ptr) { | |
1317 | /* Steal its file */ | |
1318 | porigin->file = origin->file; | |
1319 | origin->file.ptr = NULL; | |
1320 | } | |
1321 | suspects = origin->suspects; | |
1322 | origin->suspects = NULL; | |
1323 | for (e = suspects; e; e = e->next) { | |
1324 | blame_origin_incref(porigin); | |
1325 | blame_origin_decref(e->suspect); | |
1326 | e->suspect = porigin; | |
1327 | } | |
1328 | queue_blames(sb, porigin, suspects); | |
1329 | } | |
1330 | ||
1331 | /* | |
1332 | * We pass blame from the current commit to its parents. We keep saying | |
1333 | * "parent" (and "porigin"), but what we mean is to find scapegoat to | |
1334 | * exonerate ourselves. | |
1335 | */ | |
1336 | static struct commit_list *first_scapegoat(struct rev_info *revs, struct commit *commit) | |
1337 | { | |
1338 | if (!reverse) { | |
1339 | if (revs->first_parent_only && | |
1340 | commit->parents && | |
1341 | commit->parents->next) { | |
1342 | free_commit_list(commit->parents->next); | |
1343 | commit->parents->next = NULL; | |
1344 | } | |
1345 | return commit->parents; | |
1346 | } | |
1347 | return lookup_decoration(&revs->children, &commit->object); | |
1348 | } | |
1349 | ||
1350 | static int num_scapegoats(struct rev_info *revs, struct commit *commit) | |
1351 | { | |
1352 | struct commit_list *l = first_scapegoat(revs, commit); | |
1353 | return commit_list_count(l); | |
1354 | } | |
1355 | ||
1356 | /* Distribute collected unsorted blames to the respected sorted lists | |
1357 | * in the various origins. | |
1358 | */ | |
1359 | static void distribute_blame(struct blame_scoreboard *sb, struct blame_entry *blamed) | |
1360 | { | |
1361 | blamed = blame_sort(blamed, compare_blame_suspect); | |
1362 | while (blamed) | |
1363 | { | |
1364 | struct blame_origin *porigin = blamed->suspect; | |
1365 | struct blame_entry *suspects = NULL; | |
1366 | do { | |
1367 | struct blame_entry *next = blamed->next; | |
1368 | blamed->next = suspects; | |
1369 | suspects = blamed; | |
1370 | blamed = next; | |
1371 | } while (blamed && blamed->suspect == porigin); | |
1372 | suspects = reverse_blame(suspects, NULL); | |
1373 | queue_blames(sb, porigin, suspects); | |
1374 | } | |
1375 | } | |
1376 | ||
1377 | #define MAXSG 16 | |
1378 | ||
1379 | static void pass_blame(struct blame_scoreboard *sb, struct blame_origin *origin, int opt) | |
1380 | { | |
1381 | struct rev_info *revs = sb->revs; | |
1382 | int i, pass, num_sg; | |
1383 | struct commit *commit = origin->commit; | |
1384 | struct commit_list *sg; | |
1385 | struct blame_origin *sg_buf[MAXSG]; | |
1386 | struct blame_origin *porigin, **sg_origin = sg_buf; | |
1387 | struct blame_entry *toosmall = NULL; | |
1388 | struct blame_entry *blames, **blametail = &blames; | |
1389 | ||
1390 | num_sg = num_scapegoats(revs, commit); | |
1391 | if (!num_sg) | |
1392 | goto finish; | |
1393 | else if (num_sg < ARRAY_SIZE(sg_buf)) | |
1394 | memset(sg_buf, 0, sizeof(sg_buf)); | |
1395 | else | |
1396 | sg_origin = xcalloc(num_sg, sizeof(*sg_origin)); | |
1397 | ||
1398 | /* | |
1399 | * The first pass looks for unrenamed path to optimize for | |
1400 | * common cases, then we look for renames in the second pass. | |
1401 | */ | |
1402 | for (pass = 0; pass < 2 - no_whole_file_rename; pass++) { | |
1403 | struct blame_origin *(*find)(struct commit *, struct blame_origin *); | |
1404 | find = pass ? find_rename : find_origin; | |
1405 | ||
1406 | for (i = 0, sg = first_scapegoat(revs, commit); | |
1407 | i < num_sg && sg; | |
1408 | sg = sg->next, i++) { | |
1409 | struct commit *p = sg->item; | |
1410 | int j, same; | |
1411 | ||
1412 | if (sg_origin[i]) | |
1413 | continue; | |
1414 | if (parse_commit(p)) | |
1415 | continue; | |
1416 | porigin = find(p, origin); | |
1417 | if (!porigin) | |
1418 | continue; | |
1419 | if (!oidcmp(&porigin->blob_oid, &origin->blob_oid)) { | |
1420 | pass_whole_blame(sb, origin, porigin); | |
1421 | blame_origin_decref(porigin); | |
1422 | goto finish; | |
1423 | } | |
1424 | for (j = same = 0; j < i; j++) | |
1425 | if (sg_origin[j] && | |
1426 | !oidcmp(&sg_origin[j]->blob_oid, &porigin->blob_oid)) { | |
1427 | same = 1; | |
1428 | break; | |
1429 | } | |
1430 | if (!same) | |
1431 | sg_origin[i] = porigin; | |
1432 | else | |
1433 | blame_origin_decref(porigin); | |
1434 | } | |
1435 | } | |
1436 | ||
1437 | num_commits++; | |
1438 | for (i = 0, sg = first_scapegoat(revs, commit); | |
1439 | i < num_sg && sg; | |
1440 | sg = sg->next, i++) { | |
1441 | struct blame_origin *porigin = sg_origin[i]; | |
1442 | if (!porigin) | |
1443 | continue; | |
1444 | if (!origin->previous) { | |
1445 | blame_origin_incref(porigin); | |
1446 | origin->previous = porigin; | |
1447 | } | |
1448 | pass_blame_to_parent(sb, origin, porigin); | |
1449 | if (!origin->suspects) | |
1450 | goto finish; | |
1451 | } | |
1452 | ||
1453 | /* | |
1454 | * Optionally find moves in parents' files. | |
1455 | */ | |
1456 | if (opt & PICKAXE_BLAME_MOVE) { | |
1457 | filter_small(sb, &toosmall, &origin->suspects, blame_move_score); | |
1458 | if (origin->suspects) { | |
1459 | for (i = 0, sg = first_scapegoat(revs, commit); | |
1460 | i < num_sg && sg; | |
1461 | sg = sg->next, i++) { | |
1462 | struct blame_origin *porigin = sg_origin[i]; | |
1463 | if (!porigin) | |
1464 | continue; | |
1465 | find_move_in_parent(sb, &blametail, &toosmall, origin, porigin); | |
1466 | if (!origin->suspects) | |
1467 | break; | |
1468 | } | |
1469 | } | |
1470 | } | |
1471 | ||
1472 | /* | |
1473 | * Optionally find copies from parents' files. | |
1474 | */ | |
1475 | if (opt & PICKAXE_BLAME_COPY) { | |
1476 | if (blame_copy_score > blame_move_score) | |
1477 | filter_small(sb, &toosmall, &origin->suspects, blame_copy_score); | |
1478 | else if (blame_copy_score < blame_move_score) { | |
1479 | origin->suspects = blame_merge(origin->suspects, toosmall); | |
1480 | toosmall = NULL; | |
1481 | filter_small(sb, &toosmall, &origin->suspects, blame_copy_score); | |
1482 | } | |
1483 | if (!origin->suspects) | |
1484 | goto finish; | |
1485 | ||
1486 | for (i = 0, sg = first_scapegoat(revs, commit); | |
1487 | i < num_sg && sg; | |
1488 | sg = sg->next, i++) { | |
1489 | struct blame_origin *porigin = sg_origin[i]; | |
1490 | find_copy_in_parent(sb, &blametail, &toosmall, | |
1491 | origin, sg->item, porigin, opt); | |
1492 | if (!origin->suspects) | |
1493 | goto finish; | |
1494 | } | |
1495 | } | |
1496 | ||
1497 | finish: | |
1498 | *blametail = NULL; | |
1499 | distribute_blame(sb, blames); | |
1500 | /* | |
1501 | * prepend toosmall to origin->suspects | |
1502 | * | |
1503 | * There is no point in sorting: this ends up on a big | |
1504 | * unsorted list in the caller anyway. | |
1505 | */ | |
1506 | if (toosmall) { | |
1507 | struct blame_entry **tail = &toosmall; | |
1508 | while (*tail) | |
1509 | tail = &(*tail)->next; | |
1510 | *tail = origin->suspects; | |
1511 | origin->suspects = toosmall; | |
1512 | } | |
1513 | for (i = 0; i < num_sg; i++) { | |
1514 | if (sg_origin[i]) { | |
1515 | drop_origin_blob(sg_origin[i]); | |
1516 | blame_origin_decref(sg_origin[i]); | |
1517 | } | |
1518 | } | |
1519 | drop_origin_blob(origin); | |
1520 | if (sg_buf != sg_origin) | |
1521 | free(sg_origin); | |
1522 | } | |
1523 | ||
1524 | /* | |
1525 | * Information on commits, used for output. | |
1526 | */ | |
1527 | struct commit_info { | |
1528 | struct strbuf author; | |
1529 | struct strbuf author_mail; | |
1530 | timestamp_t author_time; | |
1531 | struct strbuf author_tz; | |
1532 | ||
1533 | /* filled only when asked for details */ | |
1534 | struct strbuf committer; | |
1535 | struct strbuf committer_mail; | |
1536 | timestamp_t committer_time; | |
1537 | struct strbuf committer_tz; | |
1538 | ||
1539 | struct strbuf summary; | |
1540 | }; | |
1541 | ||
1542 | /* | |
1543 | * Parse author/committer line in the commit object buffer | |
1544 | */ | |
1545 | static void get_ac_line(const char *inbuf, const char *what, | |
1546 | struct strbuf *name, struct strbuf *mail, | |
1547 | timestamp_t *time, struct strbuf *tz) | |
1548 | { | |
1549 | struct ident_split ident; | |
1550 | size_t len, maillen, namelen; | |
1551 | char *tmp, *endp; | |
1552 | const char *namebuf, *mailbuf; | |
1553 | ||
1554 | tmp = strstr(inbuf, what); | |
1555 | if (!tmp) | |
1556 | goto error_out; | |
1557 | tmp += strlen(what); | |
1558 | endp = strchr(tmp, '\n'); | |
1559 | if (!endp) | |
1560 | len = strlen(tmp); | |
1561 | else | |
1562 | len = endp - tmp; | |
1563 | ||
1564 | if (split_ident_line(&ident, tmp, len)) { | |
1565 | error_out: | |
1566 | /* Ugh */ | |
1567 | tmp = "(unknown)"; | |
1568 | strbuf_addstr(name, tmp); | |
1569 | strbuf_addstr(mail, tmp); | |
1570 | strbuf_addstr(tz, tmp); | |
1571 | *time = 0; | |
1572 | return; | |
1573 | } | |
1574 | ||
1575 | namelen = ident.name_end - ident.name_begin; | |
1576 | namebuf = ident.name_begin; | |
1577 | ||
1578 | maillen = ident.mail_end - ident.mail_begin; | |
1579 | mailbuf = ident.mail_begin; | |
1580 | ||
1581 | if (ident.date_begin && ident.date_end) | |
1582 | *time = strtoul(ident.date_begin, NULL, 10); | |
1583 | else | |
1584 | *time = 0; | |
1585 | ||
1586 | if (ident.tz_begin && ident.tz_end) | |
1587 | strbuf_add(tz, ident.tz_begin, ident.tz_end - ident.tz_begin); | |
1588 | else | |
1589 | strbuf_addstr(tz, "(unknown)"); | |
1590 | ||
1591 | /* | |
1592 | * Now, convert both name and e-mail using mailmap | |
1593 | */ | |
1594 | map_user(&mailmap, &mailbuf, &maillen, | |
1595 | &namebuf, &namelen); | |
1596 | ||
1597 | strbuf_addf(mail, "<%.*s>", (int)maillen, mailbuf); | |
1598 | strbuf_add(name, namebuf, namelen); | |
1599 | } | |
1600 | ||
1601 | static void commit_info_init(struct commit_info *ci) | |
1602 | { | |
1603 | ||
1604 | strbuf_init(&ci->author, 0); | |
1605 | strbuf_init(&ci->author_mail, 0); | |
1606 | strbuf_init(&ci->author_tz, 0); | |
1607 | strbuf_init(&ci->committer, 0); | |
1608 | strbuf_init(&ci->committer_mail, 0); | |
1609 | strbuf_init(&ci->committer_tz, 0); | |
1610 | strbuf_init(&ci->summary, 0); | |
1611 | } | |
1612 | ||
1613 | static void commit_info_destroy(struct commit_info *ci) | |
1614 | { | |
1615 | ||
1616 | strbuf_release(&ci->author); | |
1617 | strbuf_release(&ci->author_mail); | |
1618 | strbuf_release(&ci->author_tz); | |
1619 | strbuf_release(&ci->committer); | |
1620 | strbuf_release(&ci->committer_mail); | |
1621 | strbuf_release(&ci->committer_tz); | |
1622 | strbuf_release(&ci->summary); | |
1623 | } | |
1624 | ||
1625 | static void get_commit_info(struct commit *commit, | |
1626 | struct commit_info *ret, | |
1627 | int detailed) | |
1628 | { | |
1629 | int len; | |
1630 | const char *subject, *encoding; | |
1631 | const char *message; | |
1632 | ||
1633 | commit_info_init(ret); | |
1634 | ||
1635 | encoding = get_log_output_encoding(); | |
1636 | message = logmsg_reencode(commit, NULL, encoding); | |
1637 | get_ac_line(message, "\nauthor ", | |
1638 | &ret->author, &ret->author_mail, | |
1639 | &ret->author_time, &ret->author_tz); | |
1640 | ||
1641 | if (!detailed) { | |
1642 | unuse_commit_buffer(commit, message); | |
1643 | return; | |
1644 | } | |
1645 | ||
1646 | get_ac_line(message, "\ncommitter ", | |
1647 | &ret->committer, &ret->committer_mail, | |
1648 | &ret->committer_time, &ret->committer_tz); | |
1649 | ||
1650 | len = find_commit_subject(message, &subject); | |
1651 | if (len) | |
1652 | strbuf_add(&ret->summary, subject, len); | |
1653 | else | |
1654 | strbuf_addf(&ret->summary, "(%s)", oid_to_hex(&commit->object.oid)); | |
1655 | ||
1656 | unuse_commit_buffer(commit, message); | |
1657 | } | |
1658 | ||
1659 | /* | |
1660 | * Write out any suspect information which depends on the path. This must be | |
1661 | * handled separately from emit_one_suspect_detail(), because a given commit | |
1662 | * may have changes in multiple paths. So this needs to appear each time | |
1663 | * we mention a new group. | |
1664 | * | |
1665 | * To allow LF and other nonportable characters in pathnames, | |
1666 | * they are c-style quoted as needed. | |
1667 | */ | |
1668 | static void write_filename_info(struct blame_origin *suspect) | |
1669 | { | |
1670 | if (suspect->previous) { | |
1671 | struct blame_origin *prev = suspect->previous; | |
1672 | printf("previous %s ", oid_to_hex(&prev->commit->object.oid)); | |
1673 | write_name_quoted(prev->path, stdout, '\n'); | |
1674 | } | |
1675 | printf("filename "); | |
1676 | write_name_quoted(suspect->path, stdout, '\n'); | |
1677 | } | |
1678 | ||
1679 | /* | |
1680 | * Porcelain/Incremental format wants to show a lot of details per | |
1681 | * commit. Instead of repeating this every line, emit it only once, | |
1682 | * the first time each commit appears in the output (unless the | |
1683 | * user has specifically asked for us to repeat). | |
1684 | */ | |
1685 | static int emit_one_suspect_detail(struct blame_origin *suspect, int repeat) | |
1686 | { | |
1687 | struct commit_info ci; | |
1688 | ||
1689 | if (!repeat && (suspect->commit->object.flags & METAINFO_SHOWN)) | |
1690 | return 0; | |
1691 | ||
1692 | suspect->commit->object.flags |= METAINFO_SHOWN; | |
1693 | get_commit_info(suspect->commit, &ci, 1); | |
1694 | printf("author %s\n", ci.author.buf); | |
1695 | printf("author-mail %s\n", ci.author_mail.buf); | |
1696 | printf("author-time %"PRItime"\n", ci.author_time); | |
1697 | printf("author-tz %s\n", ci.author_tz.buf); | |
1698 | printf("committer %s\n", ci.committer.buf); | |
1699 | printf("committer-mail %s\n", ci.committer_mail.buf); | |
1700 | printf("committer-time %"PRItime"\n", ci.committer_time); | |
1701 | printf("committer-tz %s\n", ci.committer_tz.buf); | |
1702 | printf("summary %s\n", ci.summary.buf); | |
1703 | if (suspect->commit->object.flags & UNINTERESTING) | |
1704 | printf("boundary\n"); | |
1705 | ||
1706 | commit_info_destroy(&ci); | |
1707 | ||
1708 | return 1; | |
1709 | } | |
1710 | ||
1711 | /* | |
1712 | * The blame_entry is found to be guilty for the range. | |
1713 | * Show it in incremental output. | |
1714 | */ | |
1715 | static void found_guilty_entry(struct blame_entry *ent, | |
1716 | struct progress_info *pi) | |
1717 | { | |
1718 | if (incremental) { | |
1719 | struct blame_origin *suspect = ent->suspect; | |
1720 | ||
1721 | printf("%s %d %d %d\n", | |
1722 | oid_to_hex(&suspect->commit->object.oid), | |
1723 | ent->s_lno + 1, ent->lno + 1, ent->num_lines); | |
1724 | emit_one_suspect_detail(suspect, 0); | |
1725 | write_filename_info(suspect); | |
1726 | maybe_flush_or_die(stdout, "stdout"); | |
1727 | } | |
1728 | pi->blamed_lines += ent->num_lines; | |
1729 | display_progress(pi->progress, pi->blamed_lines); | |
1730 | } | |
1731 | ||
1732 | /* | |
1733 | * The main loop -- while we have blobs with lines whose true origin | |
1734 | * is still unknown, pick one blob, and allow its lines to pass blames | |
1735 | * to its parents. */ | |
1736 | static void assign_blame(struct blame_scoreboard *sb, int opt) | |
1737 | { | |
1738 | struct rev_info *revs = sb->revs; | |
1739 | struct commit *commit = prio_queue_get(&sb->commits); | |
1740 | struct progress_info pi = { NULL, 0 }; | |
1741 | ||
1742 | if (show_progress) | |
1743 | pi.progress = start_progress_delay(_("Blaming lines"), | |
1744 | sb->num_lines, 50, 1); | |
1745 | ||
1746 | while (commit) { | |
1747 | struct blame_entry *ent; | |
1748 | struct blame_origin *suspect = commit->util; | |
1749 | ||
1750 | /* find one suspect to break down */ | |
1751 | while (suspect && !suspect->suspects) | |
1752 | suspect = suspect->next; | |
1753 | ||
1754 | if (!suspect) { | |
1755 | commit = prio_queue_get(&sb->commits); | |
1756 | continue; | |
1757 | } | |
1758 | ||
1759 | assert(commit == suspect->commit); | |
1760 | ||
1761 | /* | |
1762 | * We will use this suspect later in the loop, | |
1763 | * so hold onto it in the meantime. | |
1764 | */ | |
1765 | blame_origin_incref(suspect); | |
1766 | parse_commit(commit); | |
1767 | if (reverse || | |
1768 | (!(commit->object.flags & UNINTERESTING) && | |
1769 | !(revs->max_age != -1 && commit->date < revs->max_age))) | |
1770 | pass_blame(sb, suspect, opt); | |
1771 | else { | |
1772 | commit->object.flags |= UNINTERESTING; | |
1773 | if (commit->object.parsed) | |
1774 | mark_parents_uninteresting(commit); | |
1775 | } | |
1776 | /* treat root commit as boundary */ | |
1777 | if (!commit->parents && !show_root) | |
1778 | commit->object.flags |= UNINTERESTING; | |
1779 | ||
1780 | /* Take responsibility for the remaining entries */ | |
1781 | ent = suspect->suspects; | |
1782 | if (ent) { | |
1783 | suspect->guilty = 1; | |
1784 | for (;;) { | |
1785 | struct blame_entry *next = ent->next; | |
1786 | found_guilty_entry(ent, &pi); | |
1787 | if (next) { | |
1788 | ent = next; | |
1789 | continue; | |
1790 | } | |
1791 | ent->next = sb->ent; | |
1792 | sb->ent = suspect->suspects; | |
1793 | suspect->suspects = NULL; | |
1794 | break; | |
1795 | } | |
1796 | } | |
1797 | blame_origin_decref(suspect); | |
1798 | ||
1799 | if (DEBUG) /* sanity */ | |
1800 | sanity_check_refcnt(sb); | |
1801 | } | |
1802 | ||
1803 | stop_progress(&pi.progress); | |
1804 | } | |
1805 | ||
1806 | static const char *format_time(timestamp_t time, const char *tz_str, | |
1807 | int show_raw_time) | |
1808 | { | |
1809 | static struct strbuf time_buf = STRBUF_INIT; | |
1810 | ||
1811 | strbuf_reset(&time_buf); | |
1812 | if (show_raw_time) { | |
1813 | strbuf_addf(&time_buf, "%"PRItime" %s", time, tz_str); | |
1814 | } | |
1815 | else { | |
1816 | const char *time_str; | |
1817 | size_t time_width; | |
1818 | int tz; | |
1819 | tz = atoi(tz_str); | |
1820 | time_str = show_date(time, tz, &blame_date_mode); | |
1821 | strbuf_addstr(&time_buf, time_str); | |
1822 | /* | |
1823 | * Add space paddings to time_buf to display a fixed width | |
1824 | * string, and use time_width for display width calibration. | |
1825 | */ | |
1826 | for (time_width = utf8_strwidth(time_str); | |
1827 | time_width < blame_date_width; | |
1828 | time_width++) | |
1829 | strbuf_addch(&time_buf, ' '); | |
1830 | } | |
1831 | return time_buf.buf; | |
1832 | } | |
1833 | ||
1834 | #define OUTPUT_ANNOTATE_COMPAT 001 | |
1835 | #define OUTPUT_LONG_OBJECT_NAME 002 | |
1836 | #define OUTPUT_RAW_TIMESTAMP 004 | |
1837 | #define OUTPUT_PORCELAIN 010 | |
1838 | #define OUTPUT_SHOW_NAME 020 | |
1839 | #define OUTPUT_SHOW_NUMBER 040 | |
1840 | #define OUTPUT_SHOW_SCORE 0100 | |
1841 | #define OUTPUT_NO_AUTHOR 0200 | |
1842 | #define OUTPUT_SHOW_EMAIL 0400 | |
1843 | #define OUTPUT_LINE_PORCELAIN 01000 | |
1844 | ||
1845 | static void emit_porcelain_details(struct blame_origin *suspect, int repeat) | |
1846 | { | |
1847 | if (emit_one_suspect_detail(suspect, repeat) || | |
1848 | (suspect->commit->object.flags & MORE_THAN_ONE_PATH)) | |
1849 | write_filename_info(suspect); | |
1850 | } | |
1851 | ||
1852 | static void emit_porcelain(struct blame_scoreboard *sb, struct blame_entry *ent, | |
1853 | int opt) | |
1854 | { | |
1855 | int repeat = opt & OUTPUT_LINE_PORCELAIN; | |
1856 | int cnt; | |
1857 | const char *cp; | |
1858 | struct blame_origin *suspect = ent->suspect; | |
1859 | char hex[GIT_MAX_HEXSZ + 1]; | |
1860 | ||
1861 | oid_to_hex_r(hex, &suspect->commit->object.oid); | |
1862 | printf("%s %d %d %d\n", | |
1863 | hex, | |
1864 | ent->s_lno + 1, | |
1865 | ent->lno + 1, | |
1866 | ent->num_lines); | |
1867 | emit_porcelain_details(suspect, repeat); | |
1868 | ||
1869 | cp = nth_line(sb, ent->lno); | |
1870 | for (cnt = 0; cnt < ent->num_lines; cnt++) { | |
1871 | char ch; | |
1872 | if (cnt) { | |
1873 | printf("%s %d %d\n", hex, | |
1874 | ent->s_lno + 1 + cnt, | |
1875 | ent->lno + 1 + cnt); | |
1876 | if (repeat) | |
1877 | emit_porcelain_details(suspect, 1); | |
1878 | } | |
1879 | putchar('\t'); | |
1880 | do { | |
1881 | ch = *cp++; | |
1882 | putchar(ch); | |
1883 | } while (ch != '\n' && | |
1884 | cp < sb->final_buf + sb->final_buf_size); | |
1885 | } | |
1886 | ||
1887 | if (sb->final_buf_size && cp[-1] != '\n') | |
1888 | putchar('\n'); | |
1889 | } | |
1890 | ||
1891 | static void emit_other(struct blame_scoreboard *sb, struct blame_entry *ent, int opt) | |
1892 | { | |
1893 | int cnt; | |
1894 | const char *cp; | |
1895 | struct blame_origin *suspect = ent->suspect; | |
1896 | struct commit_info ci; | |
1897 | char hex[GIT_MAX_HEXSZ + 1]; | |
1898 | int show_raw_time = !!(opt & OUTPUT_RAW_TIMESTAMP); | |
1899 | ||
1900 | get_commit_info(suspect->commit, &ci, 1); | |
1901 | oid_to_hex_r(hex, &suspect->commit->object.oid); | |
1902 | ||
1903 | cp = nth_line(sb, ent->lno); | |
1904 | for (cnt = 0; cnt < ent->num_lines; cnt++) { | |
1905 | char ch; | |
1906 | int length = (opt & OUTPUT_LONG_OBJECT_NAME) ? GIT_SHA1_HEXSZ : abbrev; | |
1907 | ||
1908 | if (suspect->commit->object.flags & UNINTERESTING) { | |
1909 | if (blank_boundary) | |
1910 | memset(hex, ' ', length); | |
1911 | else if (!(opt & OUTPUT_ANNOTATE_COMPAT)) { | |
1912 | length--; | |
1913 | putchar('^'); | |
1914 | } | |
1915 | } | |
1916 | ||
1917 | printf("%.*s", length, hex); | |
1918 | if (opt & OUTPUT_ANNOTATE_COMPAT) { | |
1919 | const char *name; | |
1920 | if (opt & OUTPUT_SHOW_EMAIL) | |
1921 | name = ci.author_mail.buf; | |
1922 | else | |
1923 | name = ci.author.buf; | |
1924 | printf("\t(%10s\t%10s\t%d)", name, | |
1925 | format_time(ci.author_time, ci.author_tz.buf, | |
1926 | show_raw_time), | |
1927 | ent->lno + 1 + cnt); | |
1928 | } else { | |
1929 | if (opt & OUTPUT_SHOW_SCORE) | |
1930 | printf(" %*d %02d", | |
1931 | max_score_digits, ent->score, | |
1932 | ent->suspect->refcnt); | |
1933 | if (opt & OUTPUT_SHOW_NAME) | |
1934 | printf(" %-*.*s", longest_file, longest_file, | |
1935 | suspect->path); | |
1936 | if (opt & OUTPUT_SHOW_NUMBER) | |
1937 | printf(" %*d", max_orig_digits, | |
1938 | ent->s_lno + 1 + cnt); | |
1939 | ||
1940 | if (!(opt & OUTPUT_NO_AUTHOR)) { | |
1941 | const char *name; | |
1942 | int pad; | |
1943 | if (opt & OUTPUT_SHOW_EMAIL) | |
1944 | name = ci.author_mail.buf; | |
1945 | else | |
1946 | name = ci.author.buf; | |
1947 | pad = longest_author - utf8_strwidth(name); | |
1948 | printf(" (%s%*s %10s", | |
1949 | name, pad, "", | |
1950 | format_time(ci.author_time, | |
1951 | ci.author_tz.buf, | |
1952 | show_raw_time)); | |
1953 | } | |
1954 | printf(" %*d) ", | |
1955 | max_digits, ent->lno + 1 + cnt); | |
1956 | } | |
1957 | do { | |
1958 | ch = *cp++; | |
1959 | putchar(ch); | |
1960 | } while (ch != '\n' && | |
1961 | cp < sb->final_buf + sb->final_buf_size); | |
1962 | } | |
1963 | ||
1964 | if (sb->final_buf_size && cp[-1] != '\n') | |
1965 | putchar('\n'); | |
1966 | ||
1967 | commit_info_destroy(&ci); | |
1968 | } | |
1969 | ||
1970 | static void output(struct blame_scoreboard *sb, int option) | |
1971 | { | |
1972 | struct blame_entry *ent; | |
1973 | ||
1974 | if (option & OUTPUT_PORCELAIN) { | |
1975 | for (ent = sb->ent; ent; ent = ent->next) { | |
1976 | int count = 0; | |
1977 | struct blame_origin *suspect; | |
1978 | struct commit *commit = ent->suspect->commit; | |
1979 | if (commit->object.flags & MORE_THAN_ONE_PATH) | |
1980 | continue; | |
1981 | for (suspect = commit->util; suspect; suspect = suspect->next) { | |
1982 | if (suspect->guilty && count++) { | |
1983 | commit->object.flags |= MORE_THAN_ONE_PATH; | |
1984 | break; | |
1985 | } | |
1986 | } | |
1987 | } | |
1988 | } | |
1989 | ||
1990 | for (ent = sb->ent; ent; ent = ent->next) { | |
1991 | if (option & OUTPUT_PORCELAIN) | |
1992 | emit_porcelain(sb, ent, option); | |
1993 | else { | |
1994 | emit_other(sb, ent, option); | |
1995 | } | |
1996 | } | |
1997 | } | |
1998 | ||
1999 | static const char *get_next_line(const char *start, const char *end) | |
2000 | { | |
2001 | const char *nl = memchr(start, '\n', end - start); | |
2002 | return nl ? nl + 1 : end; | |
2003 | } | |
2004 | ||
2005 | /* | |
2006 | * To allow quick access to the contents of nth line in the | |
2007 | * final image, prepare an index in the scoreboard. | |
2008 | */ | |
2009 | static int prepare_lines(struct blame_scoreboard *sb) | |
2010 | { | |
2011 | const char *buf = sb->final_buf; | |
2012 | unsigned long len = sb->final_buf_size; | |
2013 | const char *end = buf + len; | |
2014 | const char *p; | |
2015 | int *lineno; | |
2016 | int num = 0; | |
2017 | ||
2018 | for (p = buf; p < end; p = get_next_line(p, end)) | |
2019 | num++; | |
2020 | ||
2021 | ALLOC_ARRAY(sb->lineno, num + 1); | |
2022 | lineno = sb->lineno; | |
2023 | ||
2024 | for (p = buf; p < end; p = get_next_line(p, end)) | |
2025 | *lineno++ = p - buf; | |
2026 | ||
2027 | *lineno = len; | |
2028 | ||
2029 | sb->num_lines = num; | |
2030 | return sb->num_lines; | |
2031 | } | |
2032 | ||
2033 | /* | |
2034 | * Add phony grafts for use with -S; this is primarily to | |
2035 | * support git's cvsserver that wants to give a linear history | |
2036 | * to its clients. | |
2037 | */ | |
2038 | static int read_ancestry(const char *graft_file) | |
2039 | { | |
2040 | FILE *fp = fopen(graft_file, "r"); | |
2041 | struct strbuf buf = STRBUF_INIT; | |
2042 | if (!fp) | |
2043 | return -1; | |
2044 | while (!strbuf_getwholeline(&buf, fp, '\n')) { | |
2045 | /* The format is just "Commit Parent1 Parent2 ...\n" */ | |
2046 | struct commit_graft *graft = read_graft_line(buf.buf, buf.len); | |
2047 | if (graft) | |
2048 | register_commit_graft(graft, 0); | |
2049 | } | |
2050 | fclose(fp); | |
2051 | strbuf_release(&buf); | |
2052 | return 0; | |
2053 | } | |
2054 | ||
2055 | static int update_auto_abbrev(int auto_abbrev, struct blame_origin *suspect) | |
2056 | { | |
2057 | const char *uniq = find_unique_abbrev(suspect->commit->object.oid.hash, | |
2058 | auto_abbrev); | |
2059 | int len = strlen(uniq); | |
2060 | if (auto_abbrev < len) | |
2061 | return len; | |
2062 | return auto_abbrev; | |
2063 | } | |
2064 | ||
2065 | /* | |
2066 | * How many columns do we need to show line numbers, authors, | |
2067 | * and filenames? | |
2068 | */ | |
2069 | static void find_alignment(struct blame_scoreboard *sb, int *option) | |
2070 | { | |
2071 | int longest_src_lines = 0; | |
2072 | int longest_dst_lines = 0; | |
2073 | unsigned largest_score = 0; | |
2074 | struct blame_entry *e; | |
2075 | int compute_auto_abbrev = (abbrev < 0); | |
2076 | int auto_abbrev = DEFAULT_ABBREV; | |
2077 | ||
2078 | for (e = sb->ent; e; e = e->next) { | |
2079 | struct blame_origin *suspect = e->suspect; | |
2080 | int num; | |
2081 | ||
2082 | if (compute_auto_abbrev) | |
2083 | auto_abbrev = update_auto_abbrev(auto_abbrev, suspect); | |
2084 | if (strcmp(suspect->path, sb->path)) | |
2085 | *option |= OUTPUT_SHOW_NAME; | |
2086 | num = strlen(suspect->path); | |
2087 | if (longest_file < num) | |
2088 | longest_file = num; | |
2089 | if (!(suspect->commit->object.flags & METAINFO_SHOWN)) { | |
2090 | struct commit_info ci; | |
2091 | suspect->commit->object.flags |= METAINFO_SHOWN; | |
2092 | get_commit_info(suspect->commit, &ci, 1); | |
2093 | if (*option & OUTPUT_SHOW_EMAIL) | |
2094 | num = utf8_strwidth(ci.author_mail.buf); | |
2095 | else | |
2096 | num = utf8_strwidth(ci.author.buf); | |
2097 | if (longest_author < num) | |
2098 | longest_author = num; | |
2099 | commit_info_destroy(&ci); | |
2100 | } | |
2101 | num = e->s_lno + e->num_lines; | |
2102 | if (longest_src_lines < num) | |
2103 | longest_src_lines = num; | |
2104 | num = e->lno + e->num_lines; | |
2105 | if (longest_dst_lines < num) | |
2106 | longest_dst_lines = num; | |
2107 | if (largest_score < ent_score(sb, e)) | |
2108 | largest_score = ent_score(sb, e); | |
2109 | } | |
2110 | max_orig_digits = decimal_width(longest_src_lines); | |
2111 | max_digits = decimal_width(longest_dst_lines); | |
2112 | max_score_digits = decimal_width(largest_score); | |
2113 | ||
2114 | if (compute_auto_abbrev) | |
2115 | /* one more abbrev length is needed for the boundary commit */ | |
2116 | abbrev = auto_abbrev + 1; | |
2117 | } | |
2118 | ||
2119 | /* | |
2120 | * For debugging -- origin is refcounted, and this asserts that | |
2121 | * we do not underflow. | |
2122 | */ | |
2123 | static void sanity_check_refcnt(struct blame_scoreboard *sb) | |
2124 | { | |
2125 | int baa = 0; | |
2126 | struct blame_entry *ent; | |
2127 | ||
2128 | for (ent = sb->ent; ent; ent = ent->next) { | |
2129 | /* Nobody should have zero or negative refcnt */ | |
2130 | if (ent->suspect->refcnt <= 0) { | |
2131 | fprintf(stderr, "%s in %s has negative refcnt %d\n", | |
2132 | ent->suspect->path, | |
2133 | oid_to_hex(&ent->suspect->commit->object.oid), | |
2134 | ent->suspect->refcnt); | |
2135 | baa = 1; | |
2136 | } | |
2137 | } | |
2138 | if (baa) { | |
2139 | int opt = 0160; | |
2140 | find_alignment(sb, &opt); | |
2141 | output(sb, opt); | |
2142 | die("Baa %d!", baa); | |
2143 | } | |
2144 | } | |
2145 | ||
2146 | static unsigned parse_score(const char *arg) | |
2147 | { | |
2148 | char *end; | |
2149 | unsigned long score = strtoul(arg, &end, 10); | |
2150 | if (*end) | |
2151 | return 0; | |
2152 | return score; | |
2153 | } | |
2154 | ||
2155 | static const char *add_prefix(const char *prefix, const char *path) | |
2156 | { | |
2157 | return prefix_path(prefix, prefix ? strlen(prefix) : 0, path); | |
2158 | } | |
2159 | ||
2160 | static int git_blame_config(const char *var, const char *value, void *cb) | |
2161 | { | |
2162 | if (!strcmp(var, "blame.showroot")) { | |
2163 | show_root = git_config_bool(var, value); | |
2164 | return 0; | |
2165 | } | |
2166 | if (!strcmp(var, "blame.blankboundary")) { | |
2167 | blank_boundary = git_config_bool(var, value); | |
2168 | return 0; | |
2169 | } | |
2170 | if (!strcmp(var, "blame.showemail")) { | |
2171 | int *output_option = cb; | |
2172 | if (git_config_bool(var, value)) | |
2173 | *output_option |= OUTPUT_SHOW_EMAIL; | |
2174 | else | |
2175 | *output_option &= ~OUTPUT_SHOW_EMAIL; | |
2176 | return 0; | |
2177 | } | |
2178 | if (!strcmp(var, "blame.date")) { | |
2179 | if (!value) | |
2180 | return config_error_nonbool(var); | |
2181 | parse_date_format(value, &blame_date_mode); | |
2182 | return 0; | |
2183 | } | |
2184 | ||
2185 | if (git_diff_heuristic_config(var, value, cb) < 0) | |
2186 | return -1; | |
2187 | if (userdiff_config(var, value) < 0) | |
2188 | return -1; | |
2189 | ||
2190 | return git_default_config(var, value, cb); | |
2191 | } | |
2192 | ||
2193 | static void verify_working_tree_path(struct commit *work_tree, const char *path) | |
2194 | { | |
2195 | struct commit_list *parents; | |
2196 | int pos; | |
2197 | ||
2198 | for (parents = work_tree->parents; parents; parents = parents->next) { | |
2199 | const struct object_id *commit_oid = &parents->item->object.oid; | |
2200 | struct object_id blob_oid; | |
2201 | unsigned mode; | |
2202 | ||
2203 | if (!get_tree_entry(commit_oid->hash, path, blob_oid.hash, &mode) && | |
2204 | sha1_object_info(blob_oid.hash, NULL) == OBJ_BLOB) | |
2205 | return; | |
2206 | } | |
2207 | ||
2208 | pos = cache_name_pos(path, strlen(path)); | |
2209 | if (pos >= 0) | |
2210 | ; /* path is in the index */ | |
2211 | else if (-1 - pos < active_nr && | |
2212 | !strcmp(active_cache[-1 - pos]->name, path)) | |
2213 | ; /* path is in the index, unmerged */ | |
2214 | else | |
2215 | die("no such path '%s' in HEAD", path); | |
2216 | } | |
2217 | ||
2218 | static struct commit_list **append_parent(struct commit_list **tail, const struct object_id *oid) | |
2219 | { | |
2220 | struct commit *parent; | |
2221 | ||
2222 | parent = lookup_commit_reference(oid->hash); | |
2223 | if (!parent) | |
2224 | die("no such commit %s", oid_to_hex(oid)); | |
2225 | return &commit_list_insert(parent, tail)->next; | |
2226 | } | |
2227 | ||
2228 | static void append_merge_parents(struct commit_list **tail) | |
2229 | { | |
2230 | int merge_head; | |
2231 | struct strbuf line = STRBUF_INIT; | |
2232 | ||
2233 | merge_head = open(git_path_merge_head(), O_RDONLY); | |
2234 | if (merge_head < 0) { | |
2235 | if (errno == ENOENT) | |
2236 | return; | |
2237 | die("cannot open '%s' for reading", git_path_merge_head()); | |
2238 | } | |
2239 | ||
2240 | while (!strbuf_getwholeline_fd(&line, merge_head, '\n')) { | |
2241 | struct object_id oid; | |
2242 | if (line.len < GIT_SHA1_HEXSZ || get_oid_hex(line.buf, &oid)) | |
2243 | die("unknown line in '%s': %s", git_path_merge_head(), line.buf); | |
2244 | tail = append_parent(tail, &oid); | |
2245 | } | |
2246 | close(merge_head); | |
2247 | strbuf_release(&line); | |
2248 | } | |
2249 | ||
2250 | /* | |
2251 | * This isn't as simple as passing sb->buf and sb->len, because we | |
2252 | * want to transfer ownership of the buffer to the commit (so we | |
2253 | * must use detach). | |
2254 | */ | |
2255 | static void set_commit_buffer_from_strbuf(struct commit *c, struct strbuf *sb) | |
2256 | { | |
2257 | size_t len; | |
2258 | void *buf = strbuf_detach(sb, &len); | |
2259 | set_commit_buffer(c, buf, len); | |
2260 | } | |
2261 | ||
2262 | /* | |
2263 | * Prepare a dummy commit that represents the work tree (or staged) item. | |
2264 | * Note that annotating work tree item never works in the reverse. | |
2265 | */ | |
2266 | static struct commit *fake_working_tree_commit(struct diff_options *opt, | |
2267 | const char *path, | |
2268 | const char *contents_from) | |
2269 | { | |
2270 | struct commit *commit; | |
2271 | struct blame_origin *origin; | |
2272 | struct commit_list **parent_tail, *parent; | |
2273 | struct object_id head_oid; | |
2274 | struct strbuf buf = STRBUF_INIT; | |
2275 | const char *ident; | |
2276 | time_t now; | |
2277 | int size, len; | |
2278 | struct cache_entry *ce; | |
2279 | unsigned mode; | |
2280 | struct strbuf msg = STRBUF_INIT; | |
2281 | ||
2282 | read_cache(); | |
2283 | time(&now); | |
2284 | commit = alloc_commit_node(); | |
2285 | commit->object.parsed = 1; | |
2286 | commit->date = now; | |
2287 | parent_tail = &commit->parents; | |
2288 | ||
2289 | if (!resolve_ref_unsafe("HEAD", RESOLVE_REF_READING, head_oid.hash, NULL)) | |
2290 | die("no such ref: HEAD"); | |
2291 | ||
2292 | parent_tail = append_parent(parent_tail, &head_oid); | |
2293 | append_merge_parents(parent_tail); | |
2294 | verify_working_tree_path(commit, path); | |
2295 | ||
2296 | origin = make_origin(commit, path); | |
2297 | ||
2298 | ident = fmt_ident("Not Committed Yet", "not.committed.yet", NULL, 0); | |
2299 | strbuf_addstr(&msg, "tree 0000000000000000000000000000000000000000\n"); | |
2300 | for (parent = commit->parents; parent; parent = parent->next) | |
2301 | strbuf_addf(&msg, "parent %s\n", | |
2302 | oid_to_hex(&parent->item->object.oid)); | |
2303 | strbuf_addf(&msg, | |
2304 | "author %s\n" | |
2305 | "committer %s\n\n" | |
2306 | "Version of %s from %s\n", | |
2307 | ident, ident, path, | |
2308 | (!contents_from ? path : | |
2309 | (!strcmp(contents_from, "-") ? "standard input" : contents_from))); | |
2310 | set_commit_buffer_from_strbuf(commit, &msg); | |
2311 | ||
2312 | if (!contents_from || strcmp("-", contents_from)) { | |
2313 | struct stat st; | |
2314 | const char *read_from; | |
2315 | char *buf_ptr; | |
2316 | unsigned long buf_len; | |
2317 | ||
2318 | if (contents_from) { | |
2319 | if (stat(contents_from, &st) < 0) | |
2320 | die_errno("Cannot stat '%s'", contents_from); | |
2321 | read_from = contents_from; | |
2322 | } | |
2323 | else { | |
2324 | if (lstat(path, &st) < 0) | |
2325 | die_errno("Cannot lstat '%s'", path); | |
2326 | read_from = path; | |
2327 | } | |
2328 | mode = canon_mode(st.st_mode); | |
2329 | ||
2330 | switch (st.st_mode & S_IFMT) { | |
2331 | case S_IFREG: | |
2332 | if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) && | |
2333 | textconv_object(read_from, mode, &null_oid, 0, &buf_ptr, &buf_len)) | |
2334 | strbuf_attach(&buf, buf_ptr, buf_len, buf_len + 1); | |
2335 | else if (strbuf_read_file(&buf, read_from, st.st_size) != st.st_size) | |
2336 | die_errno("cannot open or read '%s'", read_from); | |
2337 | break; | |
2338 | case S_IFLNK: | |
2339 | if (strbuf_readlink(&buf, read_from, st.st_size) < 0) | |
2340 | die_errno("cannot readlink '%s'", read_from); | |
2341 | break; | |
2342 | default: | |
2343 | die("unsupported file type %s", read_from); | |
2344 | } | |
2345 | } | |
2346 | else { | |
2347 | /* Reading from stdin */ | |
2348 | mode = 0; | |
2349 | if (strbuf_read(&buf, 0, 0) < 0) | |
2350 | die_errno("failed to read from stdin"); | |
2351 | } | |
2352 | convert_to_git(path, buf.buf, buf.len, &buf, 0); | |
2353 | origin->file.ptr = buf.buf; | |
2354 | origin->file.size = buf.len; | |
2355 | pretend_sha1_file(buf.buf, buf.len, OBJ_BLOB, origin->blob_oid.hash); | |
2356 | ||
2357 | /* | |
2358 | * Read the current index, replace the path entry with | |
2359 | * origin->blob_sha1 without mucking with its mode or type | |
2360 | * bits; we are not going to write this index out -- we just | |
2361 | * want to run "diff-index --cached". | |
2362 | */ | |
2363 | discard_cache(); | |
2364 | read_cache(); | |
2365 | ||
2366 | len = strlen(path); | |
2367 | if (!mode) { | |
2368 | int pos = cache_name_pos(path, len); | |
2369 | if (0 <= pos) | |
2370 | mode = active_cache[pos]->ce_mode; | |
2371 | else | |
2372 | /* Let's not bother reading from HEAD tree */ | |
2373 | mode = S_IFREG | 0644; | |
2374 | } | |
2375 | size = cache_entry_size(len); | |
2376 | ce = xcalloc(1, size); | |
2377 | oidcpy(&ce->oid, &origin->blob_oid); | |
2378 | memcpy(ce->name, path, len); | |
2379 | ce->ce_flags = create_ce_flags(0); | |
2380 | ce->ce_namelen = len; | |
2381 | ce->ce_mode = create_ce_mode(mode); | |
2382 | add_cache_entry(ce, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
2383 | ||
2384 | cache_tree_invalidate_path(&the_index, path); | |
2385 | ||
2386 | return commit; | |
2387 | } | |
2388 | ||
2389 | static struct commit *find_single_final(struct rev_info *revs, | |
2390 | const char **name_p) | |
2391 | { | |
2392 | int i; | |
2393 | struct commit *found = NULL; | |
2394 | const char *name = NULL; | |
2395 | ||
2396 | for (i = 0; i < revs->pending.nr; i++) { | |
2397 | struct object *obj = revs->pending.objects[i].item; | |
2398 | if (obj->flags & UNINTERESTING) | |
2399 | continue; | |
2400 | obj = deref_tag(obj, NULL, 0); | |
2401 | if (obj->type != OBJ_COMMIT) | |
2402 | die("Non commit %s?", revs->pending.objects[i].name); | |
2403 | if (found) | |
2404 | die("More than one commit to dig from %s and %s?", | |
2405 | revs->pending.objects[i].name, name); | |
2406 | found = (struct commit *)obj; | |
2407 | name = revs->pending.objects[i].name; | |
2408 | } | |
2409 | if (name_p) | |
2410 | *name_p = name; | |
2411 | return found; | |
2412 | } | |
2413 | ||
2414 | static char *prepare_final(struct blame_scoreboard *sb) | |
2415 | { | |
2416 | const char *name; | |
2417 | sb->final = find_single_final(sb->revs, &name); | |
2418 | return xstrdup_or_null(name); | |
2419 | } | |
2420 | ||
2421 | static const char *dwim_reverse_initial(struct blame_scoreboard *sb) | |
2422 | { | |
2423 | /* | |
2424 | * DWIM "git blame --reverse ONE -- PATH" as | |
2425 | * "git blame --reverse ONE..HEAD -- PATH" but only do so | |
2426 | * when it makes sense. | |
2427 | */ | |
2428 | struct object *obj; | |
2429 | struct commit *head_commit; | |
2430 | unsigned char head_sha1[20]; | |
2431 | ||
2432 | if (sb->revs->pending.nr != 1) | |
2433 | return NULL; | |
2434 | ||
2435 | /* Is that sole rev a committish? */ | |
2436 | obj = sb->revs->pending.objects[0].item; | |
2437 | obj = deref_tag(obj, NULL, 0); | |
2438 | if (obj->type != OBJ_COMMIT) | |
2439 | return NULL; | |
2440 | ||
2441 | /* Do we have HEAD? */ | |
2442 | if (!resolve_ref_unsafe("HEAD", RESOLVE_REF_READING, head_sha1, NULL)) | |
2443 | return NULL; | |
2444 | head_commit = lookup_commit_reference_gently(head_sha1, 1); | |
2445 | if (!head_commit) | |
2446 | return NULL; | |
2447 | ||
2448 | /* Turn "ONE" into "ONE..HEAD" then */ | |
2449 | obj->flags |= UNINTERESTING; | |
2450 | add_pending_object(sb->revs, &head_commit->object, "HEAD"); | |
2451 | ||
2452 | sb->final = (struct commit *)obj; | |
2453 | return sb->revs->pending.objects[0].name; | |
2454 | } | |
2455 | ||
2456 | static char *prepare_initial(struct blame_scoreboard *sb) | |
2457 | { | |
2458 | int i; | |
2459 | const char *final_commit_name = NULL; | |
2460 | struct rev_info *revs = sb->revs; | |
2461 | ||
2462 | /* | |
2463 | * There must be one and only one negative commit, and it must be | |
2464 | * the boundary. | |
2465 | */ | |
2466 | for (i = 0; i < revs->pending.nr; i++) { | |
2467 | struct object *obj = revs->pending.objects[i].item; | |
2468 | if (!(obj->flags & UNINTERESTING)) | |
2469 | continue; | |
2470 | obj = deref_tag(obj, NULL, 0); | |
2471 | if (obj->type != OBJ_COMMIT) | |
2472 | die("Non commit %s?", revs->pending.objects[i].name); | |
2473 | if (sb->final) | |
2474 | die("More than one commit to dig up from, %s and %s?", | |
2475 | revs->pending.objects[i].name, | |
2476 | final_commit_name); | |
2477 | sb->final = (struct commit *) obj; | |
2478 | final_commit_name = revs->pending.objects[i].name; | |
2479 | } | |
2480 | ||
2481 | if (!final_commit_name) | |
2482 | final_commit_name = dwim_reverse_initial(sb); | |
2483 | if (!final_commit_name) | |
2484 | die("No commit to dig up from?"); | |
2485 | return xstrdup(final_commit_name); | |
2486 | } | |
2487 | ||
2488 | static int blame_copy_callback(const struct option *option, const char *arg, int unset) | |
2489 | { | |
2490 | int *opt = option->value; | |
2491 | ||
2492 | /* | |
2493 | * -C enables copy from removed files; | |
2494 | * -C -C enables copy from existing files, but only | |
2495 | * when blaming a new file; | |
2496 | * -C -C -C enables copy from existing files for | |
2497 | * everybody | |
2498 | */ | |
2499 | if (*opt & PICKAXE_BLAME_COPY_HARDER) | |
2500 | *opt |= PICKAXE_BLAME_COPY_HARDEST; | |
2501 | if (*opt & PICKAXE_BLAME_COPY) | |
2502 | *opt |= PICKAXE_BLAME_COPY_HARDER; | |
2503 | *opt |= PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE; | |
2504 | ||
2505 | if (arg) | |
2506 | blame_copy_score = parse_score(arg); | |
2507 | return 0; | |
2508 | } | |
2509 | ||
2510 | static int blame_move_callback(const struct option *option, const char *arg, int unset) | |
2511 | { | |
2512 | int *opt = option->value; | |
2513 | ||
2514 | *opt |= PICKAXE_BLAME_MOVE; | |
2515 | ||
2516 | if (arg) | |
2517 | blame_move_score = parse_score(arg); | |
2518 | return 0; | |
2519 | } | |
2520 | ||
2521 | int cmd_blame(int argc, const char **argv, const char *prefix) | |
2522 | { | |
2523 | struct rev_info revs; | |
2524 | const char *path; | |
2525 | struct blame_scoreboard sb; | |
2526 | struct blame_origin *o; | |
2527 | struct blame_entry *ent = NULL; | |
2528 | long dashdash_pos, lno; | |
2529 | char *final_commit_name = NULL; | |
2530 | enum object_type type; | |
2531 | struct commit *final_commit = NULL; | |
2532 | ||
2533 | struct string_list range_list = STRING_LIST_INIT_NODUP; | |
2534 | int output_option = 0, opt = 0; | |
2535 | int show_stats = 0; | |
2536 | const char *revs_file = NULL; | |
2537 | const char *contents_from = NULL; | |
2538 | const struct option options[] = { | |
2539 | OPT_BOOL(0, "incremental", &incremental, N_("Show blame entries as we find them, incrementally")), | |
2540 | OPT_BOOL('b', NULL, &blank_boundary, N_("Show blank SHA-1 for boundary commits (Default: off)")), | |
2541 | OPT_BOOL(0, "root", &show_root, N_("Do not treat root commits as boundaries (Default: off)")), | |
2542 | OPT_BOOL(0, "show-stats", &show_stats, N_("Show work cost statistics")), | |
2543 | OPT_BOOL(0, "progress", &show_progress, N_("Force progress reporting")), | |
2544 | OPT_BIT(0, "score-debug", &output_option, N_("Show output score for blame entries"), OUTPUT_SHOW_SCORE), | |
2545 | OPT_BIT('f', "show-name", &output_option, N_("Show original filename (Default: auto)"), OUTPUT_SHOW_NAME), | |
2546 | OPT_BIT('n', "show-number", &output_option, N_("Show original linenumber (Default: off)"), OUTPUT_SHOW_NUMBER), | |
2547 | OPT_BIT('p', "porcelain", &output_option, N_("Show in a format designed for machine consumption"), OUTPUT_PORCELAIN), | |
2548 | OPT_BIT(0, "line-porcelain", &output_option, N_("Show porcelain format with per-line commit information"), OUTPUT_PORCELAIN|OUTPUT_LINE_PORCELAIN), | |
2549 | OPT_BIT('c', NULL, &output_option, N_("Use the same output mode as git-annotate (Default: off)"), OUTPUT_ANNOTATE_COMPAT), | |
2550 | OPT_BIT('t', NULL, &output_option, N_("Show raw timestamp (Default: off)"), OUTPUT_RAW_TIMESTAMP), | |
2551 | OPT_BIT('l', NULL, &output_option, N_("Show long commit SHA1 (Default: off)"), OUTPUT_LONG_OBJECT_NAME), | |
2552 | OPT_BIT('s', NULL, &output_option, N_("Suppress author name and timestamp (Default: off)"), OUTPUT_NO_AUTHOR), | |
2553 | OPT_BIT('e', "show-email", &output_option, N_("Show author email instead of name (Default: off)"), OUTPUT_SHOW_EMAIL), | |
2554 | OPT_BIT('w', NULL, &xdl_opts, N_("Ignore whitespace differences"), XDF_IGNORE_WHITESPACE), | |
2555 | ||
2556 | /* | |
2557 | * The following two options are parsed by parse_revision_opt() | |
2558 | * and are only included here to get included in the "-h" | |
2559 | * output: | |
2560 | */ | |
2561 | { OPTION_LOWLEVEL_CALLBACK, 0, "indent-heuristic", NULL, NULL, N_("Use an experimental heuristic to improve diffs"), PARSE_OPT_NOARG, parse_opt_unknown_cb }, | |
2562 | ||
2563 | OPT_BIT(0, "minimal", &xdl_opts, N_("Spend extra cycles to find better match"), XDF_NEED_MINIMAL), | |
2564 | OPT_STRING('S', NULL, &revs_file, N_("file"), N_("Use revisions from <file> instead of calling git-rev-list")), | |
2565 | OPT_STRING(0, "contents", &contents_from, N_("file"), N_("Use <file>'s contents as the final image")), | |
2566 | { OPTION_CALLBACK, 'C', NULL, &opt, N_("score"), N_("Find line copies within and across files"), PARSE_OPT_OPTARG, blame_copy_callback }, | |
2567 | { OPTION_CALLBACK, 'M', NULL, &opt, N_("score"), N_("Find line movements within and across files"), PARSE_OPT_OPTARG, blame_move_callback }, | |
2568 | OPT_STRING_LIST('L', NULL, &range_list, N_("n,m"), N_("Process only line range n,m, counting from 1")), | |
2569 | OPT__ABBREV(&abbrev), | |
2570 | OPT_END() | |
2571 | }; | |
2572 | ||
2573 | struct parse_opt_ctx_t ctx; | |
2574 | int cmd_is_annotate = !strcmp(argv[0], "annotate"); | |
2575 | struct range_set ranges; | |
2576 | unsigned int range_i; | |
2577 | long anchor; | |
2578 | ||
2579 | git_config(git_blame_config, &output_option); | |
2580 | init_revisions(&revs, NULL); | |
2581 | revs.date_mode = blame_date_mode; | |
2582 | DIFF_OPT_SET(&revs.diffopt, ALLOW_TEXTCONV); | |
2583 | DIFF_OPT_SET(&revs.diffopt, FOLLOW_RENAMES); | |
2584 | ||
2585 | save_commit_buffer = 0; | |
2586 | dashdash_pos = 0; | |
2587 | show_progress = -1; | |
2588 | ||
2589 | parse_options_start(&ctx, argc, argv, prefix, options, | |
2590 | PARSE_OPT_KEEP_DASHDASH | PARSE_OPT_KEEP_ARGV0); | |
2591 | for (;;) { | |
2592 | switch (parse_options_step(&ctx, options, blame_opt_usage)) { | |
2593 | case PARSE_OPT_HELP: | |
2594 | exit(129); | |
2595 | case PARSE_OPT_DONE: | |
2596 | if (ctx.argv[0]) | |
2597 | dashdash_pos = ctx.cpidx; | |
2598 | goto parse_done; | |
2599 | } | |
2600 | ||
2601 | if (!strcmp(ctx.argv[0], "--reverse")) { | |
2602 | ctx.argv[0] = "--children"; | |
2603 | reverse = 1; | |
2604 | } | |
2605 | parse_revision_opt(&revs, &ctx, options, blame_opt_usage); | |
2606 | } | |
2607 | parse_done: | |
2608 | no_whole_file_rename = !DIFF_OPT_TST(&revs.diffopt, FOLLOW_RENAMES); | |
2609 | xdl_opts |= revs.diffopt.xdl_opts & XDF_INDENT_HEURISTIC; | |
2610 | DIFF_OPT_CLR(&revs.diffopt, FOLLOW_RENAMES); | |
2611 | argc = parse_options_end(&ctx); | |
2612 | ||
2613 | if (incremental || (output_option & OUTPUT_PORCELAIN)) { | |
2614 | if (show_progress > 0) | |
2615 | die(_("--progress can't be used with --incremental or porcelain formats")); | |
2616 | show_progress = 0; | |
2617 | } else if (show_progress < 0) | |
2618 | show_progress = isatty(2); | |
2619 | ||
2620 | if (0 < abbrev && abbrev < GIT_SHA1_HEXSZ) | |
2621 | /* one more abbrev length is needed for the boundary commit */ | |
2622 | abbrev++; | |
2623 | else if (!abbrev) | |
2624 | abbrev = GIT_SHA1_HEXSZ; | |
2625 | ||
2626 | if (revs_file && read_ancestry(revs_file)) | |
2627 | die_errno("reading graft file '%s' failed", revs_file); | |
2628 | ||
2629 | if (cmd_is_annotate) { | |
2630 | output_option |= OUTPUT_ANNOTATE_COMPAT; | |
2631 | blame_date_mode.type = DATE_ISO8601; | |
2632 | } else { | |
2633 | blame_date_mode = revs.date_mode; | |
2634 | } | |
2635 | ||
2636 | /* The maximum width used to show the dates */ | |
2637 | switch (blame_date_mode.type) { | |
2638 | case DATE_RFC2822: | |
2639 | blame_date_width = sizeof("Thu, 19 Oct 2006 16:00:04 -0700"); | |
2640 | break; | |
2641 | case DATE_ISO8601_STRICT: | |
2642 | blame_date_width = sizeof("2006-10-19T16:00:04-07:00"); | |
2643 | break; | |
2644 | case DATE_ISO8601: | |
2645 | blame_date_width = sizeof("2006-10-19 16:00:04 -0700"); | |
2646 | break; | |
2647 | case DATE_RAW: | |
2648 | blame_date_width = sizeof("1161298804 -0700"); | |
2649 | break; | |
2650 | case DATE_UNIX: | |
2651 | blame_date_width = sizeof("1161298804"); | |
2652 | break; | |
2653 | case DATE_SHORT: | |
2654 | blame_date_width = sizeof("2006-10-19"); | |
2655 | break; | |
2656 | case DATE_RELATIVE: | |
2657 | /* TRANSLATORS: This string is used to tell us the maximum | |
2658 | display width for a relative timestamp in "git blame" | |
2659 | output. For C locale, "4 years, 11 months ago", which | |
2660 | takes 22 places, is the longest among various forms of | |
2661 | relative timestamps, but your language may need more or | |
2662 | fewer display columns. */ | |
2663 | blame_date_width = utf8_strwidth(_("4 years, 11 months ago")) + 1; /* add the null */ | |
2664 | break; | |
2665 | case DATE_NORMAL: | |
2666 | blame_date_width = sizeof("Thu Oct 19 16:00:04 2006 -0700"); | |
2667 | break; | |
2668 | case DATE_STRFTIME: | |
2669 | blame_date_width = strlen(show_date(0, 0, &blame_date_mode)) + 1; /* add the null */ | |
2670 | break; | |
2671 | } | |
2672 | blame_date_width -= 1; /* strip the null */ | |
2673 | ||
2674 | if (DIFF_OPT_TST(&revs.diffopt, FIND_COPIES_HARDER)) | |
2675 | opt |= (PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE | | |
2676 | PICKAXE_BLAME_COPY_HARDER); | |
2677 | ||
2678 | if (!blame_move_score) | |
2679 | blame_move_score = BLAME_DEFAULT_MOVE_SCORE; | |
2680 | if (!blame_copy_score) | |
2681 | blame_copy_score = BLAME_DEFAULT_COPY_SCORE; | |
2682 | ||
2683 | /* | |
2684 | * We have collected options unknown to us in argv[1..unk] | |
2685 | * which are to be passed to revision machinery if we are | |
2686 | * going to do the "bottom" processing. | |
2687 | * | |
2688 | * The remaining are: | |
2689 | * | |
2690 | * (1) if dashdash_pos != 0, it is either | |
2691 | * "blame [revisions] -- <path>" or | |
2692 | * "blame -- <path> <rev>" | |
2693 | * | |
2694 | * (2) otherwise, it is one of the two: | |
2695 | * "blame [revisions] <path>" | |
2696 | * "blame <path> <rev>" | |
2697 | * | |
2698 | * Note that we must strip out <path> from the arguments: we do not | |
2699 | * want the path pruning but we may want "bottom" processing. | |
2700 | */ | |
2701 | if (dashdash_pos) { | |
2702 | switch (argc - dashdash_pos - 1) { | |
2703 | case 2: /* (1b) */ | |
2704 | if (argc != 4) | |
2705 | usage_with_options(blame_opt_usage, options); | |
2706 | /* reorder for the new way: <rev> -- <path> */ | |
2707 | argv[1] = argv[3]; | |
2708 | argv[3] = argv[2]; | |
2709 | argv[2] = "--"; | |
2710 | /* FALLTHROUGH */ | |
2711 | case 1: /* (1a) */ | |
2712 | path = add_prefix(prefix, argv[--argc]); | |
2713 | argv[argc] = NULL; | |
2714 | break; | |
2715 | default: | |
2716 | usage_with_options(blame_opt_usage, options); | |
2717 | } | |
2718 | } else { | |
2719 | if (argc < 2) | |
2720 | usage_with_options(blame_opt_usage, options); | |
2721 | path = add_prefix(prefix, argv[argc - 1]); | |
2722 | if (argc == 3 && !file_exists(path)) { /* (2b) */ | |
2723 | path = add_prefix(prefix, argv[1]); | |
2724 | argv[1] = argv[2]; | |
2725 | } | |
2726 | argv[argc - 1] = "--"; | |
2727 | ||
2728 | setup_work_tree(); | |
2729 | if (!file_exists(path)) | |
2730 | die_errno("cannot stat path '%s'", path); | |
2731 | } | |
2732 | ||
2733 | revs.disable_stdin = 1; | |
2734 | setup_revisions(argc, argv, &revs, NULL); | |
2735 | memset(&sb, 0, sizeof(sb)); | |
2736 | ||
2737 | sb.revs = &revs; | |
2738 | if (!reverse) { | |
2739 | final_commit_name = prepare_final(&sb); | |
2740 | sb.commits.compare = compare_commits_by_commit_date; | |
2741 | } | |
2742 | else if (contents_from) | |
2743 | die(_("--contents and --reverse do not blend well.")); | |
2744 | else { | |
2745 | final_commit_name = prepare_initial(&sb); | |
2746 | sb.commits.compare = compare_commits_by_reverse_commit_date; | |
2747 | if (revs.first_parent_only) | |
2748 | revs.children.name = NULL; | |
2749 | } | |
2750 | ||
2751 | if (!sb.final) { | |
2752 | /* | |
2753 | * "--not A B -- path" without anything positive; | |
2754 | * do not default to HEAD, but use the working tree | |
2755 | * or "--contents". | |
2756 | */ | |
2757 | setup_work_tree(); | |
2758 | sb.final = fake_working_tree_commit(&sb.revs->diffopt, | |
2759 | path, contents_from); | |
2760 | add_pending_object(&revs, &(sb.final->object), ":"); | |
2761 | } | |
2762 | else if (contents_from) | |
2763 | die(_("cannot use --contents with final commit object name")); | |
2764 | ||
2765 | if (reverse && revs.first_parent_only) { | |
2766 | final_commit = find_single_final(sb.revs, NULL); | |
2767 | if (!final_commit) | |
2768 | die(_("--reverse and --first-parent together require specified latest commit")); | |
2769 | } | |
2770 | ||
2771 | /* | |
2772 | * If we have bottom, this will mark the ancestors of the | |
2773 | * bottom commits we would reach while traversing as | |
2774 | * uninteresting. | |
2775 | */ | |
2776 | if (prepare_revision_walk(&revs)) | |
2777 | die(_("revision walk setup failed")); | |
2778 | ||
2779 | if (reverse && revs.first_parent_only) { | |
2780 | struct commit *c = final_commit; | |
2781 | ||
2782 | sb.revs->children.name = "children"; | |
2783 | while (c->parents && | |
2784 | oidcmp(&c->object.oid, &sb.final->object.oid)) { | |
2785 | struct commit_list *l = xcalloc(1, sizeof(*l)); | |
2786 | ||
2787 | l->item = c; | |
2788 | if (add_decoration(&sb.revs->children, | |
2789 | &c->parents->item->object, l)) | |
2790 | die("BUG: not unique item in first-parent chain"); | |
2791 | c = c->parents->item; | |
2792 | } | |
2793 | ||
2794 | if (oidcmp(&c->object.oid, &sb.final->object.oid)) | |
2795 | die(_("--reverse --first-parent together require range along first-parent chain")); | |
2796 | } | |
2797 | ||
2798 | if (is_null_oid(&sb.final->object.oid)) { | |
2799 | o = sb.final->util; | |
2800 | sb.final_buf = xmemdupz(o->file.ptr, o->file.size); | |
2801 | sb.final_buf_size = o->file.size; | |
2802 | } | |
2803 | else { | |
2804 | o = get_origin(sb.final, path); | |
2805 | if (fill_blob_sha1_and_mode(o)) | |
2806 | die(_("no such path %s in %s"), path, final_commit_name); | |
2807 | ||
2808 | if (DIFF_OPT_TST(&sb.revs->diffopt, ALLOW_TEXTCONV) && | |
2809 | textconv_object(path, o->mode, &o->blob_oid, 1, (char **) &sb.final_buf, | |
2810 | &sb.final_buf_size)) | |
2811 | ; | |
2812 | else | |
2813 | sb.final_buf = read_sha1_file(o->blob_oid.hash, &type, | |
2814 | &sb.final_buf_size); | |
2815 | ||
2816 | if (!sb.final_buf) | |
2817 | die(_("cannot read blob %s for path %s"), | |
2818 | oid_to_hex(&o->blob_oid), | |
2819 | path); | |
2820 | } | |
2821 | num_read_blob++; | |
2822 | lno = prepare_lines(&sb); | |
2823 | ||
2824 | if (lno && !range_list.nr) | |
2825 | string_list_append(&range_list, "1"); | |
2826 | ||
2827 | anchor = 1; | |
2828 | range_set_init(&ranges, range_list.nr); | |
2829 | for (range_i = 0; range_i < range_list.nr; ++range_i) { | |
2830 | long bottom, top; | |
2831 | if (parse_range_arg(range_list.items[range_i].string, | |
2832 | nth_line_cb, &sb, lno, anchor, | |
2833 | &bottom, &top, sb.path)) | |
2834 | usage(blame_usage); | |
2835 | if (lno < top || ((lno || bottom) && lno < bottom)) | |
2836 | die(Q_("file %s has only %lu line", | |
2837 | "file %s has only %lu lines", | |
2838 | lno), path, lno); | |
2839 | if (bottom < 1) | |
2840 | bottom = 1; | |
2841 | if (top < 1) | |
2842 | top = lno; | |
2843 | bottom--; | |
2844 | range_set_append_unsafe(&ranges, bottom, top); | |
2845 | anchor = top + 1; | |
2846 | } | |
2847 | sort_and_merge_range_set(&ranges); | |
2848 | ||
2849 | for (range_i = ranges.nr; range_i > 0; --range_i) { | |
2850 | const struct range *r = &ranges.ranges[range_i - 1]; | |
2851 | long bottom = r->start; | |
2852 | long top = r->end; | |
2853 | struct blame_entry *next = ent; | |
2854 | ent = xcalloc(1, sizeof(*ent)); | |
2855 | ent->lno = bottom; | |
2856 | ent->num_lines = top - bottom; | |
2857 | ent->suspect = o; | |
2858 | ent->s_lno = bottom; | |
2859 | ent->next = next; | |
2860 | blame_origin_incref(o); | |
2861 | } | |
2862 | ||
2863 | o->suspects = ent; | |
2864 | prio_queue_put(&sb.commits, o->commit); | |
2865 | ||
2866 | blame_origin_decref(o); | |
2867 | ||
2868 | range_set_release(&ranges); | |
2869 | string_list_clear(&range_list, 0); | |
2870 | ||
2871 | sb.ent = NULL; | |
2872 | sb.path = path; | |
2873 | ||
2874 | read_mailmap(&mailmap, NULL); | |
2875 | ||
2876 | assign_blame(&sb, opt); | |
2877 | ||
2878 | if (!incremental) | |
2879 | setup_pager(); | |
2880 | ||
2881 | free(final_commit_name); | |
2882 | ||
2883 | if (incremental) | |
2884 | return 0; | |
2885 | ||
2886 | sb.ent = blame_sort(sb.ent, compare_blame_final); | |
2887 | ||
2888 | blame_coalesce(&sb); | |
2889 | ||
2890 | if (!(output_option & OUTPUT_PORCELAIN)) | |
2891 | find_alignment(&sb, &output_option); | |
2892 | ||
2893 | output(&sb, output_option); | |
2894 | free((void *)sb.final_buf); | |
2895 | for (ent = sb.ent; ent; ) { | |
2896 | struct blame_entry *e = ent->next; | |
2897 | free(ent); | |
2898 | ent = e; | |
2899 | } | |
2900 | ||
2901 | if (show_stats) { | |
2902 | printf("num read blob: %d\n", num_read_blob); | |
2903 | printf("num get patch: %d\n", num_get_patch); | |
2904 | printf("num commits: %d\n", num_commits); | |
2905 | } | |
2906 | return 0; | |
2907 | } |