Merge branch 'maint'
[git/git-svn.git] / builtin-blame.c
blob16660221c30b0e3b1e1bd4cbc5559e3f162ae05a
1 /*
2 * Pickaxe
4 * Copyright (c) 2006, Junio C Hamano
5 */
7 #include "cache.h"
8 #include "builtin.h"
9 #include "blob.h"
10 #include "commit.h"
11 #include "tag.h"
12 #include "tree-walk.h"
13 #include "diff.h"
14 #include "diffcore.h"
15 #include "revision.h"
16 #include "xdiff-interface.h"
18 #include <time.h>
19 #include <sys/time.h>
20 #include <regex.h>
22 static char blame_usage[] =
23 "git-blame [-c] [-l] [-t] [-f] [-n] [-p] [-L n,m] [-S <revs-file>] [-M] [-C] [-C] [commit] [--] file\n"
24 " -c, --compatibility Use the same output mode as git-annotate (Default: off)\n"
25 " -l, --long Show long commit SHA1 (Default: off)\n"
26 " -t, --time Show raw timestamp (Default: off)\n"
27 " -f, --show-name Show original filename (Default: auto)\n"
28 " -n, --show-number Show original linenumber (Default: off)\n"
29 " -p, --porcelain Show in a format designed for machine consumption\n"
30 " -L n,m Process only line range n,m, counting from 1\n"
31 " -M, -C Find line movements within and across files\n"
32 " -S revs-file Use revisions from revs-file instead of calling git-rev-list\n";
34 static int longest_file;
35 static int longest_author;
36 static int max_orig_digits;
37 static int max_digits;
38 static int max_score_digits;
40 #ifndef DEBUG
41 #define DEBUG 0
42 #endif
44 /* stats */
45 static int num_read_blob;
46 static int num_get_patch;
47 static int num_commits;
49 #define PICKAXE_BLAME_MOVE 01
50 #define PICKAXE_BLAME_COPY 02
51 #define PICKAXE_BLAME_COPY_HARDER 04
54 * blame for a blame_entry with score lower than these thresholds
55 * is not passed to the parent using move/copy logic.
57 static unsigned blame_move_score;
58 static unsigned blame_copy_score;
59 #define BLAME_DEFAULT_MOVE_SCORE 20
60 #define BLAME_DEFAULT_COPY_SCORE 40
62 /* bits #0..7 in revision.h, #8..11 used for merge_bases() in commit.c */
63 #define METAINFO_SHOWN (1u<<12)
64 #define MORE_THAN_ONE_PATH (1u<<13)
67 * One blob in a commit that is being suspected
69 struct origin {
70 int refcnt;
71 struct commit *commit;
72 mmfile_t file;
73 unsigned char blob_sha1[20];
74 char path[FLEX_ARRAY];
77 static char *fill_origin_blob(struct origin *o, mmfile_t *file)
79 if (!o->file.ptr) {
80 char type[10];
81 num_read_blob++;
82 file->ptr = read_sha1_file(o->blob_sha1, type,
83 (unsigned long *)(&(file->size)));
84 o->file = *file;
86 else
87 *file = o->file;
88 return file->ptr;
91 static inline struct origin *origin_incref(struct origin *o)
93 if (o)
94 o->refcnt++;
95 return o;
98 static void origin_decref(struct origin *o)
100 if (o && --o->refcnt <= 0) {
101 if (o->file.ptr)
102 free(o->file.ptr);
103 memset(o, 0, sizeof(*o));
104 free(o);
108 struct blame_entry {
109 struct blame_entry *prev;
110 struct blame_entry *next;
112 /* the first line of this group in the final image;
113 * internally all line numbers are 0 based.
115 int lno;
117 /* how many lines this group has */
118 int num_lines;
120 /* the commit that introduced this group into the final image */
121 struct origin *suspect;
123 /* true if the suspect is truly guilty; false while we have not
124 * checked if the group came from one of its parents.
126 char guilty;
128 /* the line number of the first line of this group in the
129 * suspect's file; internally all line numbers are 0 based.
131 int s_lno;
133 /* how significant this entry is -- cached to avoid
134 * scanning the lines over and over
136 unsigned score;
139 struct scoreboard {
140 /* the final commit (i.e. where we started digging from) */
141 struct commit *final;
143 const char *path;
145 /* the contents in the final; pointed into by buf pointers of
146 * blame_entries
148 const char *final_buf;
149 unsigned long final_buf_size;
151 /* linked list of blames */
152 struct blame_entry *ent;
154 /* look-up a line in the final buffer */
155 int num_lines;
156 int *lineno;
159 static int cmp_suspect(struct origin *a, struct origin *b)
161 int cmp = hashcmp(a->commit->object.sha1, b->commit->object.sha1);
162 if (cmp)
163 return cmp;
164 return strcmp(a->path, b->path);
167 #define cmp_suspect(a, b) ( ((a)==(b)) ? 0 : cmp_suspect(a,b) )
169 static void sanity_check_refcnt(struct scoreboard *);
171 static void coalesce(struct scoreboard *sb)
173 struct blame_entry *ent, *next;
175 for (ent = sb->ent; ent && (next = ent->next); ent = next) {
176 if (!cmp_suspect(ent->suspect, next->suspect) &&
177 ent->guilty == next->guilty &&
178 ent->s_lno + ent->num_lines == next->s_lno) {
179 ent->num_lines += next->num_lines;
180 ent->next = next->next;
181 if (ent->next)
182 ent->next->prev = ent;
183 origin_decref(next->suspect);
184 free(next);
185 ent->score = 0;
186 next = ent; /* again */
190 if (DEBUG) /* sanity */
191 sanity_check_refcnt(sb);
194 static struct origin *make_origin(struct commit *commit, const char *path)
196 struct origin *o;
197 o = xcalloc(1, sizeof(*o) + strlen(path) + 1);
198 o->commit = commit;
199 o->refcnt = 1;
200 strcpy(o->path, path);
201 return o;
204 static struct origin *get_origin(struct scoreboard *sb,
205 struct commit *commit,
206 const char *path)
208 struct blame_entry *e;
210 for (e = sb->ent; e; e = e->next) {
211 if (e->suspect->commit == commit &&
212 !strcmp(e->suspect->path, path))
213 return origin_incref(e->suspect);
215 return make_origin(commit, path);
218 static int fill_blob_sha1(struct origin *origin)
220 unsigned mode;
221 char type[10];
223 if (!is_null_sha1(origin->blob_sha1))
224 return 0;
225 if (get_tree_entry(origin->commit->object.sha1,
226 origin->path,
227 origin->blob_sha1, &mode))
228 goto error_out;
229 if (sha1_object_info(origin->blob_sha1, type, NULL) ||
230 strcmp(type, blob_type))
231 goto error_out;
232 return 0;
233 error_out:
234 hashclr(origin->blob_sha1);
235 return -1;
238 static struct origin *find_origin(struct scoreboard *sb,
239 struct commit *parent,
240 struct origin *origin)
242 struct origin *porigin = NULL;
243 struct diff_options diff_opts;
244 const char *paths[2];
246 if (parent->util) {
247 /* This is a freestanding copy of origin and not
248 * refcounted.
250 struct origin *cached = parent->util;
251 if (!strcmp(cached->path, origin->path)) {
252 porigin = get_origin(sb, parent, cached->path);
253 if (porigin->refcnt == 1)
254 hashcpy(porigin->blob_sha1, cached->blob_sha1);
255 return porigin;
257 /* otherwise it was not very useful; free it */
258 free(parent->util);
259 parent->util = NULL;
262 /* See if the origin->path is different between parent
263 * and origin first. Most of the time they are the
264 * same and diff-tree is fairly efficient about this.
266 diff_setup(&diff_opts);
267 diff_opts.recursive = 1;
268 diff_opts.detect_rename = 0;
269 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
270 paths[0] = origin->path;
271 paths[1] = NULL;
273 diff_tree_setup_paths(paths, &diff_opts);
274 if (diff_setup_done(&diff_opts) < 0)
275 die("diff-setup");
276 diff_tree_sha1(parent->tree->object.sha1,
277 origin->commit->tree->object.sha1,
278 "", &diff_opts);
279 diffcore_std(&diff_opts);
281 /* It is either one entry that says "modified", or "created",
282 * or nothing.
284 if (!diff_queued_diff.nr) {
285 /* The path is the same as parent */
286 porigin = get_origin(sb, parent, origin->path);
287 hashcpy(porigin->blob_sha1, origin->blob_sha1);
289 else if (diff_queued_diff.nr != 1)
290 die("internal error in blame::find_origin");
291 else {
292 struct diff_filepair *p = diff_queued_diff.queue[0];
293 switch (p->status) {
294 default:
295 die("internal error in blame::find_origin (%c)",
296 p->status);
297 case 'M':
298 porigin = get_origin(sb, parent, origin->path);
299 hashcpy(porigin->blob_sha1, p->one->sha1);
300 break;
301 case 'A':
302 case 'T':
303 /* Did not exist in parent, or type changed */
304 break;
307 diff_flush(&diff_opts);
308 if (porigin) {
309 struct origin *cached;
310 cached = make_origin(porigin->commit, porigin->path);
311 hashcpy(cached->blob_sha1, porigin->blob_sha1);
312 parent->util = cached;
314 return porigin;
317 static struct origin *find_rename(struct scoreboard *sb,
318 struct commit *parent,
319 struct origin *origin)
321 struct origin *porigin = NULL;
322 struct diff_options diff_opts;
323 int i;
324 const char *paths[2];
326 diff_setup(&diff_opts);
327 diff_opts.recursive = 1;
328 diff_opts.detect_rename = DIFF_DETECT_RENAME;
329 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
330 diff_opts.single_follow = origin->path;
331 paths[0] = NULL;
332 diff_tree_setup_paths(paths, &diff_opts);
333 if (diff_setup_done(&diff_opts) < 0)
334 die("diff-setup");
335 diff_tree_sha1(parent->tree->object.sha1,
336 origin->commit->tree->object.sha1,
337 "", &diff_opts);
338 diffcore_std(&diff_opts);
340 for (i = 0; i < diff_queued_diff.nr; i++) {
341 struct diff_filepair *p = diff_queued_diff.queue[i];
342 if ((p->status == 'R' || p->status == 'C') &&
343 !strcmp(p->two->path, origin->path)) {
344 porigin = get_origin(sb, parent, p->one->path);
345 hashcpy(porigin->blob_sha1, p->one->sha1);
346 break;
349 diff_flush(&diff_opts);
350 return porigin;
353 struct chunk {
354 /* line number in postimage; up to but not including this
355 * line is the same as preimage
357 int same;
359 /* preimage line number after this chunk */
360 int p_next;
362 /* postimage line number after this chunk */
363 int t_next;
366 struct patch {
367 struct chunk *chunks;
368 int num;
371 struct blame_diff_state {
372 struct xdiff_emit_state xm;
373 struct patch *ret;
374 unsigned hunk_post_context;
375 unsigned hunk_in_pre_context : 1;
378 static void process_u_diff(void *state_, char *line, unsigned long len)
380 struct blame_diff_state *state = state_;
381 struct chunk *chunk;
382 int off1, off2, len1, len2, num;
384 num = state->ret->num;
385 if (len < 4 || line[0] != '@' || line[1] != '@') {
386 if (state->hunk_in_pre_context && line[0] == ' ')
387 state->ret->chunks[num - 1].same++;
388 else {
389 state->hunk_in_pre_context = 0;
390 if (line[0] == ' ')
391 state->hunk_post_context++;
392 else
393 state->hunk_post_context = 0;
395 return;
398 if (num && state->hunk_post_context) {
399 chunk = &state->ret->chunks[num - 1];
400 chunk->p_next -= state->hunk_post_context;
401 chunk->t_next -= state->hunk_post_context;
403 state->ret->num = ++num;
404 state->ret->chunks = xrealloc(state->ret->chunks,
405 sizeof(struct chunk) * num);
406 chunk = &state->ret->chunks[num - 1];
407 if (parse_hunk_header(line, len, &off1, &len1, &off2, &len2)) {
408 state->ret->num--;
409 return;
412 /* Line numbers in patch output are one based. */
413 off1--;
414 off2--;
416 chunk->same = len2 ? off2 : (off2 + 1);
418 chunk->p_next = off1 + (len1 ? len1 : 1);
419 chunk->t_next = chunk->same + len2;
420 state->hunk_in_pre_context = 1;
421 state->hunk_post_context = 0;
424 static struct patch *compare_buffer(mmfile_t *file_p, mmfile_t *file_o,
425 int context)
427 struct blame_diff_state state;
428 xpparam_t xpp;
429 xdemitconf_t xecfg;
430 xdemitcb_t ecb;
432 xpp.flags = XDF_NEED_MINIMAL;
433 xecfg.ctxlen = context;
434 xecfg.flags = 0;
435 ecb.outf = xdiff_outf;
436 ecb.priv = &state;
437 memset(&state, 0, sizeof(state));
438 state.xm.consume = process_u_diff;
439 state.ret = xmalloc(sizeof(struct patch));
440 state.ret->chunks = NULL;
441 state.ret->num = 0;
443 xdl_diff(file_p, file_o, &xpp, &xecfg, &ecb);
445 if (state.ret->num) {
446 struct chunk *chunk;
447 chunk = &state.ret->chunks[state.ret->num - 1];
448 chunk->p_next -= state.hunk_post_context;
449 chunk->t_next -= state.hunk_post_context;
451 return state.ret;
454 static struct patch *get_patch(struct origin *parent, struct origin *origin)
456 mmfile_t file_p, file_o;
457 struct patch *patch;
459 fill_origin_blob(parent, &file_p);
460 fill_origin_blob(origin, &file_o);
461 if (!file_p.ptr || !file_o.ptr)
462 return NULL;
463 patch = compare_buffer(&file_p, &file_o, 0);
464 num_get_patch++;
465 return patch;
468 static void free_patch(struct patch *p)
470 free(p->chunks);
471 free(p);
474 static void add_blame_entry(struct scoreboard *sb, struct blame_entry *e)
476 struct blame_entry *ent, *prev = NULL;
478 origin_incref(e->suspect);
480 for (ent = sb->ent; ent && ent->lno < e->lno; ent = ent->next)
481 prev = ent;
483 /* prev, if not NULL, is the last one that is below e */
484 e->prev = prev;
485 if (prev) {
486 e->next = prev->next;
487 prev->next = e;
489 else {
490 e->next = sb->ent;
491 sb->ent = e;
493 if (e->next)
494 e->next->prev = e;
497 static void dup_entry(struct blame_entry *dst, struct blame_entry *src)
499 struct blame_entry *p, *n;
501 p = dst->prev;
502 n = dst->next;
503 origin_incref(src->suspect);
504 origin_decref(dst->suspect);
505 memcpy(dst, src, sizeof(*src));
506 dst->prev = p;
507 dst->next = n;
508 dst->score = 0;
511 static const char *nth_line(struct scoreboard *sb, int lno)
513 return sb->final_buf + sb->lineno[lno];
516 static void split_overlap(struct blame_entry *split,
517 struct blame_entry *e,
518 int tlno, int plno, int same,
519 struct origin *parent)
521 /* it is known that lines between tlno to same came from
522 * parent, and e has an overlap with that range. it also is
523 * known that parent's line plno corresponds to e's line tlno.
525 * <---- e ----->
526 * <------>
527 * <------------>
528 * <------------>
529 * <------------------>
531 * Potentially we need to split e into three parts; before
532 * this chunk, the chunk to be blamed for parent, and after
533 * that portion.
535 int chunk_end_lno;
536 memset(split, 0, sizeof(struct blame_entry [3]));
538 if (e->s_lno < tlno) {
539 /* there is a pre-chunk part not blamed on parent */
540 split[0].suspect = origin_incref(e->suspect);
541 split[0].lno = e->lno;
542 split[0].s_lno = e->s_lno;
543 split[0].num_lines = tlno - e->s_lno;
544 split[1].lno = e->lno + tlno - e->s_lno;
545 split[1].s_lno = plno;
547 else {
548 split[1].lno = e->lno;
549 split[1].s_lno = plno + (e->s_lno - tlno);
552 if (same < e->s_lno + e->num_lines) {
553 /* there is a post-chunk part not blamed on parent */
554 split[2].suspect = origin_incref(e->suspect);
555 split[2].lno = e->lno + (same - e->s_lno);
556 split[2].s_lno = e->s_lno + (same - e->s_lno);
557 split[2].num_lines = e->s_lno + e->num_lines - same;
558 chunk_end_lno = split[2].lno;
560 else
561 chunk_end_lno = e->lno + e->num_lines;
562 split[1].num_lines = chunk_end_lno - split[1].lno;
564 if (split[1].num_lines < 1)
565 return;
566 split[1].suspect = origin_incref(parent);
569 static void split_blame(struct scoreboard *sb,
570 struct blame_entry *split,
571 struct blame_entry *e)
573 struct blame_entry *new_entry;
575 if (split[0].suspect && split[2].suspect) {
576 /* we need to split e into two and add another for parent */
577 dup_entry(e, &split[0]);
579 new_entry = xmalloc(sizeof(*new_entry));
580 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
581 add_blame_entry(sb, new_entry);
583 new_entry = xmalloc(sizeof(*new_entry));
584 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
585 add_blame_entry(sb, new_entry);
587 else if (!split[0].suspect && !split[2].suspect)
588 /* parent covers the entire area */
589 dup_entry(e, &split[1]);
590 else if (split[0].suspect) {
591 dup_entry(e, &split[0]);
593 new_entry = xmalloc(sizeof(*new_entry));
594 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
595 add_blame_entry(sb, new_entry);
597 else {
598 dup_entry(e, &split[1]);
600 new_entry = xmalloc(sizeof(*new_entry));
601 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
602 add_blame_entry(sb, new_entry);
605 if (DEBUG) { /* sanity */
606 struct blame_entry *ent;
607 int lno = sb->ent->lno, corrupt = 0;
609 for (ent = sb->ent; ent; ent = ent->next) {
610 if (lno != ent->lno)
611 corrupt = 1;
612 if (ent->s_lno < 0)
613 corrupt = 1;
614 lno += ent->num_lines;
616 if (corrupt) {
617 lno = sb->ent->lno;
618 for (ent = sb->ent; ent; ent = ent->next) {
619 printf("L %8d l %8d n %8d\n",
620 lno, ent->lno, ent->num_lines);
621 lno = ent->lno + ent->num_lines;
623 die("oops");
628 static void decref_split(struct blame_entry *split)
630 int i;
632 for (i = 0; i < 3; i++)
633 origin_decref(split[i].suspect);
636 static void blame_overlap(struct scoreboard *sb, struct blame_entry *e,
637 int tlno, int plno, int same,
638 struct origin *parent)
640 struct blame_entry split[3];
642 split_overlap(split, e, tlno, plno, same, parent);
643 if (split[1].suspect)
644 split_blame(sb, split, e);
645 decref_split(split);
648 static int find_last_in_target(struct scoreboard *sb, struct origin *target)
650 struct blame_entry *e;
651 int last_in_target = -1;
653 for (e = sb->ent; e; e = e->next) {
654 if (e->guilty || cmp_suspect(e->suspect, target))
655 continue;
656 if (last_in_target < e->s_lno + e->num_lines)
657 last_in_target = e->s_lno + e->num_lines;
659 return last_in_target;
662 static void blame_chunk(struct scoreboard *sb,
663 int tlno, int plno, int same,
664 struct origin *target, struct origin *parent)
666 struct blame_entry *e;
668 for (e = sb->ent; e; e = e->next) {
669 if (e->guilty || cmp_suspect(e->suspect, target))
670 continue;
671 if (same <= e->s_lno)
672 continue;
673 if (tlno < e->s_lno + e->num_lines)
674 blame_overlap(sb, e, tlno, plno, same, parent);
678 static int pass_blame_to_parent(struct scoreboard *sb,
679 struct origin *target,
680 struct origin *parent)
682 int i, last_in_target, plno, tlno;
683 struct patch *patch;
685 last_in_target = find_last_in_target(sb, target);
686 if (last_in_target < 0)
687 return 1; /* nothing remains for this target */
689 patch = get_patch(parent, target);
690 plno = tlno = 0;
691 for (i = 0; i < patch->num; i++) {
692 struct chunk *chunk = &patch->chunks[i];
694 blame_chunk(sb, tlno, plno, chunk->same, target, parent);
695 plno = chunk->p_next;
696 tlno = chunk->t_next;
698 /* rest (i.e. anything above tlno) are the same as parent */
699 blame_chunk(sb, tlno, plno, last_in_target, target, parent);
701 free_patch(patch);
702 return 0;
705 static unsigned ent_score(struct scoreboard *sb, struct blame_entry *e)
707 unsigned score;
708 const char *cp, *ep;
710 if (e->score)
711 return e->score;
713 score = 1;
714 cp = nth_line(sb, e->lno);
715 ep = nth_line(sb, e->lno + e->num_lines);
716 while (cp < ep) {
717 unsigned ch = *((unsigned char *)cp);
718 if (isalnum(ch))
719 score++;
720 cp++;
722 e->score = score;
723 return score;
726 static void copy_split_if_better(struct scoreboard *sb,
727 struct blame_entry *best_so_far,
728 struct blame_entry *this)
730 int i;
732 if (!this[1].suspect)
733 return;
734 if (best_so_far[1].suspect) {
735 if (ent_score(sb, &this[1]) < ent_score(sb, &best_so_far[1]))
736 return;
739 for (i = 0; i < 3; i++)
740 origin_incref(this[i].suspect);
741 decref_split(best_so_far);
742 memcpy(best_so_far, this, sizeof(struct blame_entry [3]));
745 static void find_copy_in_blob(struct scoreboard *sb,
746 struct blame_entry *ent,
747 struct origin *parent,
748 struct blame_entry *split,
749 mmfile_t *file_p)
751 const char *cp;
752 int cnt;
753 mmfile_t file_o;
754 struct patch *patch;
755 int i, plno, tlno;
757 cp = nth_line(sb, ent->lno);
758 file_o.ptr = (char*) cp;
759 cnt = ent->num_lines;
761 while (cnt && cp < sb->final_buf + sb->final_buf_size) {
762 if (*cp++ == '\n')
763 cnt--;
765 file_o.size = cp - file_o.ptr;
767 patch = compare_buffer(file_p, &file_o, 1);
769 memset(split, 0, sizeof(struct blame_entry [3]));
770 plno = tlno = 0;
771 for (i = 0; i < patch->num; i++) {
772 struct chunk *chunk = &patch->chunks[i];
774 /* tlno to chunk->same are the same as ent */
775 if (ent->num_lines <= tlno)
776 break;
777 if (tlno < chunk->same) {
778 struct blame_entry this[3];
779 split_overlap(this, ent,
780 tlno + ent->s_lno, plno,
781 chunk->same + ent->s_lno,
782 parent);
783 copy_split_if_better(sb, split, this);
784 decref_split(this);
786 plno = chunk->p_next;
787 tlno = chunk->t_next;
789 free_patch(patch);
792 static int find_move_in_parent(struct scoreboard *sb,
793 struct origin *target,
794 struct origin *parent)
796 int last_in_target, made_progress;
797 struct blame_entry *e, split[3];
798 mmfile_t file_p;
800 last_in_target = find_last_in_target(sb, target);
801 if (last_in_target < 0)
802 return 1; /* nothing remains for this target */
804 fill_origin_blob(parent, &file_p);
805 if (!file_p.ptr)
806 return 0;
808 made_progress = 1;
809 while (made_progress) {
810 made_progress = 0;
811 for (e = sb->ent; e; e = e->next) {
812 if (e->guilty || cmp_suspect(e->suspect, target))
813 continue;
814 find_copy_in_blob(sb, e, parent, split, &file_p);
815 if (split[1].suspect &&
816 blame_move_score < ent_score(sb, &split[1])) {
817 split_blame(sb, split, e);
818 made_progress = 1;
820 decref_split(split);
823 return 0;
827 struct blame_list {
828 struct blame_entry *ent;
829 struct blame_entry split[3];
832 static struct blame_list *setup_blame_list(struct scoreboard *sb,
833 struct origin *target,
834 int *num_ents_p)
836 struct blame_entry *e;
837 int num_ents, i;
838 struct blame_list *blame_list = NULL;
840 /* Count the number of entries the target is suspected for,
841 * and prepare a list of entry and the best split.
843 for (e = sb->ent, num_ents = 0; e; e = e->next)
844 if (!e->guilty && !cmp_suspect(e->suspect, target))
845 num_ents++;
846 if (num_ents) {
847 blame_list = xcalloc(num_ents, sizeof(struct blame_list));
848 for (e = sb->ent, i = 0; e; e = e->next)
849 if (!e->guilty && !cmp_suspect(e->suspect, target))
850 blame_list[i++].ent = e;
852 *num_ents_p = num_ents;
853 return blame_list;
856 static int find_copy_in_parent(struct scoreboard *sb,
857 struct origin *target,
858 struct commit *parent,
859 struct origin *porigin,
860 int opt)
862 struct diff_options diff_opts;
863 const char *paths[1];
864 int i, j;
865 int retval;
866 struct blame_list *blame_list;
867 int num_ents;
869 blame_list = setup_blame_list(sb, target, &num_ents);
870 if (!blame_list)
871 return 1; /* nothing remains for this target */
873 diff_setup(&diff_opts);
874 diff_opts.recursive = 1;
875 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
877 paths[0] = NULL;
878 diff_tree_setup_paths(paths, &diff_opts);
879 if (diff_setup_done(&diff_opts) < 0)
880 die("diff-setup");
882 /* Try "find copies harder" on new path if requested;
883 * we do not want to use diffcore_rename() actually to
884 * match things up; find_copies_harder is set only to
885 * force diff_tree_sha1() to feed all filepairs to diff_queue,
886 * and this code needs to be after diff_setup_done(), which
887 * usually makes find-copies-harder imply copy detection.
889 if ((opt & PICKAXE_BLAME_COPY_HARDER) &&
890 (!porigin || strcmp(target->path, porigin->path)))
891 diff_opts.find_copies_harder = 1;
893 diff_tree_sha1(parent->tree->object.sha1,
894 target->commit->tree->object.sha1,
895 "", &diff_opts);
897 if (!diff_opts.find_copies_harder)
898 diffcore_std(&diff_opts);
900 retval = 0;
901 while (1) {
902 int made_progress = 0;
904 for (i = 0; i < diff_queued_diff.nr; i++) {
905 struct diff_filepair *p = diff_queued_diff.queue[i];
906 struct origin *norigin;
907 mmfile_t file_p;
908 struct blame_entry this[3];
910 if (!DIFF_FILE_VALID(p->one))
911 continue; /* does not exist in parent */
912 if (porigin && !strcmp(p->one->path, porigin->path))
913 /* find_move already dealt with this path */
914 continue;
916 norigin = get_origin(sb, parent, p->one->path);
917 hashcpy(norigin->blob_sha1, p->one->sha1);
918 fill_origin_blob(norigin, &file_p);
919 if (!file_p.ptr)
920 continue;
922 for (j = 0; j < num_ents; j++) {
923 find_copy_in_blob(sb, blame_list[j].ent,
924 norigin, this, &file_p);
925 copy_split_if_better(sb, blame_list[j].split,
926 this);
927 decref_split(this);
929 origin_decref(norigin);
932 for (j = 0; j < num_ents; j++) {
933 struct blame_entry *split = blame_list[j].split;
934 if (split[1].suspect &&
935 blame_copy_score < ent_score(sb, &split[1])) {
936 split_blame(sb, split, blame_list[j].ent);
937 made_progress = 1;
939 decref_split(split);
941 free(blame_list);
943 if (!made_progress)
944 break;
945 blame_list = setup_blame_list(sb, target, &num_ents);
946 if (!blame_list) {
947 retval = 1;
948 break;
951 diff_flush(&diff_opts);
953 return retval;
956 /* The blobs of origin and porigin exactly match, so everything
957 * origin is suspected for can be blamed on the parent.
959 static void pass_whole_blame(struct scoreboard *sb,
960 struct origin *origin, struct origin *porigin)
962 struct blame_entry *e;
964 if (!porigin->file.ptr && origin->file.ptr) {
965 /* Steal its file */
966 porigin->file = origin->file;
967 origin->file.ptr = NULL;
969 for (e = sb->ent; e; e = e->next) {
970 if (cmp_suspect(e->suspect, origin))
971 continue;
972 origin_incref(porigin);
973 origin_decref(e->suspect);
974 e->suspect = porigin;
978 #define MAXPARENT 16
980 static void pass_blame(struct scoreboard *sb, struct origin *origin, int opt)
982 int i, pass;
983 struct commit *commit = origin->commit;
984 struct commit_list *parent;
985 struct origin *parent_origin[MAXPARENT], *porigin;
987 memset(parent_origin, 0, sizeof(parent_origin));
989 /* The first pass looks for unrenamed path to optimize for
990 * common cases, then we look for renames in the second pass.
992 for (pass = 0; pass < 2; pass++) {
993 struct origin *(*find)(struct scoreboard *,
994 struct commit *, struct origin *);
995 find = pass ? find_rename : find_origin;
997 for (i = 0, parent = commit->parents;
998 i < MAXPARENT && parent;
999 parent = parent->next, i++) {
1000 struct commit *p = parent->item;
1001 int j, same;
1003 if (parent_origin[i])
1004 continue;
1005 if (parse_commit(p))
1006 continue;
1007 porigin = find(sb, p, origin);
1008 if (!porigin)
1009 continue;
1010 if (!hashcmp(porigin->blob_sha1, origin->blob_sha1)) {
1011 pass_whole_blame(sb, origin, porigin);
1012 origin_decref(porigin);
1013 goto finish;
1015 for (j = same = 0; j < i; j++)
1016 if (parent_origin[j] &&
1017 !hashcmp(parent_origin[j]->blob_sha1,
1018 porigin->blob_sha1)) {
1019 same = 1;
1020 break;
1022 if (!same)
1023 parent_origin[i] = porigin;
1024 else
1025 origin_decref(porigin);
1029 num_commits++;
1030 for (i = 0, parent = commit->parents;
1031 i < MAXPARENT && parent;
1032 parent = parent->next, i++) {
1033 struct origin *porigin = parent_origin[i];
1034 if (!porigin)
1035 continue;
1036 if (pass_blame_to_parent(sb, origin, porigin))
1037 goto finish;
1041 * Optionally run "miff" to find moves in parents' files here.
1043 if (opt & PICKAXE_BLAME_MOVE)
1044 for (i = 0, parent = commit->parents;
1045 i < MAXPARENT && parent;
1046 parent = parent->next, i++) {
1047 struct origin *porigin = parent_origin[i];
1048 if (!porigin)
1049 continue;
1050 if (find_move_in_parent(sb, origin, porigin))
1051 goto finish;
1055 * Optionally run "ciff" to find copies from parents' files here.
1057 if (opt & PICKAXE_BLAME_COPY)
1058 for (i = 0, parent = commit->parents;
1059 i < MAXPARENT && parent;
1060 parent = parent->next, i++) {
1061 struct origin *porigin = parent_origin[i];
1062 if (find_copy_in_parent(sb, origin, parent->item,
1063 porigin, opt))
1064 goto finish;
1067 finish:
1068 for (i = 0; i < MAXPARENT; i++)
1069 origin_decref(parent_origin[i]);
1072 static void assign_blame(struct scoreboard *sb, struct rev_info *revs, int opt)
1074 while (1) {
1075 struct blame_entry *ent;
1076 struct commit *commit;
1077 struct origin *suspect = NULL;
1079 /* find one suspect to break down */
1080 for (ent = sb->ent; !suspect && ent; ent = ent->next)
1081 if (!ent->guilty)
1082 suspect = ent->suspect;
1083 if (!suspect)
1084 return; /* all done */
1086 origin_incref(suspect);
1087 commit = suspect->commit;
1088 if (!commit->object.parsed)
1089 parse_commit(commit);
1090 if (!(commit->object.flags & UNINTERESTING) &&
1091 !(revs->max_age != -1 && commit->date < revs->max_age))
1092 pass_blame(sb, suspect, opt);
1094 /* Take responsibility for the remaining entries */
1095 for (ent = sb->ent; ent; ent = ent->next)
1096 if (!cmp_suspect(ent->suspect, suspect))
1097 ent->guilty = 1;
1098 origin_decref(suspect);
1100 if (DEBUG) /* sanity */
1101 sanity_check_refcnt(sb);
1105 static const char *format_time(unsigned long time, const char *tz_str,
1106 int show_raw_time)
1108 static char time_buf[128];
1109 time_t t = time;
1110 int minutes, tz;
1111 struct tm *tm;
1113 if (show_raw_time) {
1114 sprintf(time_buf, "%lu %s", time, tz_str);
1115 return time_buf;
1118 tz = atoi(tz_str);
1119 minutes = tz < 0 ? -tz : tz;
1120 minutes = (minutes / 100)*60 + (minutes % 100);
1121 minutes = tz < 0 ? -minutes : minutes;
1122 t = time + minutes * 60;
1123 tm = gmtime(&t);
1125 strftime(time_buf, sizeof(time_buf), "%Y-%m-%d %H:%M:%S ", tm);
1126 strcat(time_buf, tz_str);
1127 return time_buf;
1130 struct commit_info
1132 char *author;
1133 char *author_mail;
1134 unsigned long author_time;
1135 char *author_tz;
1137 /* filled only when asked for details */
1138 char *committer;
1139 char *committer_mail;
1140 unsigned long committer_time;
1141 char *committer_tz;
1143 char *summary;
1146 static void get_ac_line(const char *inbuf, const char *what,
1147 int bufsz, char *person, char **mail,
1148 unsigned long *time, char **tz)
1150 int len;
1151 char *tmp, *endp;
1153 tmp = strstr(inbuf, what);
1154 if (!tmp)
1155 goto error_out;
1156 tmp += strlen(what);
1157 endp = strchr(tmp, '\n');
1158 if (!endp)
1159 len = strlen(tmp);
1160 else
1161 len = endp - tmp;
1162 if (bufsz <= len) {
1163 error_out:
1164 /* Ugh */
1165 person = *mail = *tz = "(unknown)";
1166 *time = 0;
1167 return;
1169 memcpy(person, tmp, len);
1171 tmp = person;
1172 tmp += len;
1173 *tmp = 0;
1174 while (*tmp != ' ')
1175 tmp--;
1176 *tz = tmp+1;
1178 *tmp = 0;
1179 while (*tmp != ' ')
1180 tmp--;
1181 *time = strtoul(tmp, NULL, 10);
1183 *tmp = 0;
1184 while (*tmp != ' ')
1185 tmp--;
1186 *mail = tmp + 1;
1187 *tmp = 0;
1190 static void get_commit_info(struct commit *commit,
1191 struct commit_info *ret,
1192 int detailed)
1194 int len;
1195 char *tmp, *endp;
1196 static char author_buf[1024];
1197 static char committer_buf[1024];
1198 static char summary_buf[1024];
1200 /* We've operated without save_commit_buffer, so
1201 * we now need to populate them for output.
1203 if (!commit->buffer) {
1204 char type[20];
1205 unsigned long size;
1206 commit->buffer =
1207 read_sha1_file(commit->object.sha1, type, &size);
1209 ret->author = author_buf;
1210 get_ac_line(commit->buffer, "\nauthor ",
1211 sizeof(author_buf), author_buf, &ret->author_mail,
1212 &ret->author_time, &ret->author_tz);
1214 if (!detailed)
1215 return;
1217 ret->committer = committer_buf;
1218 get_ac_line(commit->buffer, "\ncommitter ",
1219 sizeof(committer_buf), committer_buf, &ret->committer_mail,
1220 &ret->committer_time, &ret->committer_tz);
1222 ret->summary = summary_buf;
1223 tmp = strstr(commit->buffer, "\n\n");
1224 if (!tmp) {
1225 error_out:
1226 sprintf(summary_buf, "(%s)", sha1_to_hex(commit->object.sha1));
1227 return;
1229 tmp += 2;
1230 endp = strchr(tmp, '\n');
1231 if (!endp)
1232 goto error_out;
1233 len = endp - tmp;
1234 if (len >= sizeof(summary_buf))
1235 goto error_out;
1236 memcpy(summary_buf, tmp, len);
1237 summary_buf[len] = 0;
1240 #define OUTPUT_ANNOTATE_COMPAT 001
1241 #define OUTPUT_LONG_OBJECT_NAME 002
1242 #define OUTPUT_RAW_TIMESTAMP 004
1243 #define OUTPUT_PORCELAIN 010
1244 #define OUTPUT_SHOW_NAME 020
1245 #define OUTPUT_SHOW_NUMBER 040
1246 #define OUTPUT_SHOW_SCORE 0100
1248 static void emit_porcelain(struct scoreboard *sb, struct blame_entry *ent)
1250 int cnt;
1251 const char *cp;
1252 struct origin *suspect = ent->suspect;
1253 char hex[41];
1255 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1256 printf("%s%c%d %d %d\n",
1257 hex,
1258 ent->guilty ? ' ' : '*', // purely for debugging
1259 ent->s_lno + 1,
1260 ent->lno + 1,
1261 ent->num_lines);
1262 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1263 struct commit_info ci;
1264 suspect->commit->object.flags |= METAINFO_SHOWN;
1265 get_commit_info(suspect->commit, &ci, 1);
1266 printf("author %s\n", ci.author);
1267 printf("author-mail %s\n", ci.author_mail);
1268 printf("author-time %lu\n", ci.author_time);
1269 printf("author-tz %s\n", ci.author_tz);
1270 printf("committer %s\n", ci.committer);
1271 printf("committer-mail %s\n", ci.committer_mail);
1272 printf("committer-time %lu\n", ci.committer_time);
1273 printf("committer-tz %s\n", ci.committer_tz);
1274 printf("filename %s\n", suspect->path);
1275 printf("summary %s\n", ci.summary);
1277 else if (suspect->commit->object.flags & MORE_THAN_ONE_PATH)
1278 printf("filename %s\n", suspect->path);
1280 cp = nth_line(sb, ent->lno);
1281 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1282 char ch;
1283 if (cnt)
1284 printf("%s %d %d\n", hex,
1285 ent->s_lno + 1 + cnt,
1286 ent->lno + 1 + cnt);
1287 putchar('\t');
1288 do {
1289 ch = *cp++;
1290 putchar(ch);
1291 } while (ch != '\n' &&
1292 cp < sb->final_buf + sb->final_buf_size);
1296 static void emit_other(struct scoreboard *sb, struct blame_entry *ent, int opt)
1298 int cnt;
1299 const char *cp;
1300 struct origin *suspect = ent->suspect;
1301 struct commit_info ci;
1302 char hex[41];
1303 int show_raw_time = !!(opt & OUTPUT_RAW_TIMESTAMP);
1305 get_commit_info(suspect->commit, &ci, 1);
1306 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1308 cp = nth_line(sb, ent->lno);
1309 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1310 char ch;
1312 printf("%.*s", (opt & OUTPUT_LONG_OBJECT_NAME) ? 40 : 8, hex);
1313 if (opt & OUTPUT_ANNOTATE_COMPAT)
1314 printf("\t(%10s\t%10s\t%d)", ci.author,
1315 format_time(ci.author_time, ci.author_tz,
1316 show_raw_time),
1317 ent->lno + 1 + cnt);
1318 else {
1319 if (opt & OUTPUT_SHOW_SCORE)
1320 printf(" %*d %02d",
1321 max_score_digits, ent->score,
1322 ent->suspect->refcnt);
1323 if (opt & OUTPUT_SHOW_NAME)
1324 printf(" %-*.*s", longest_file, longest_file,
1325 suspect->path);
1326 if (opt & OUTPUT_SHOW_NUMBER)
1327 printf(" %*d", max_orig_digits,
1328 ent->s_lno + 1 + cnt);
1329 printf(" (%-*.*s %10s %*d) ",
1330 longest_author, longest_author, ci.author,
1331 format_time(ci.author_time, ci.author_tz,
1332 show_raw_time),
1333 max_digits, ent->lno + 1 + cnt);
1335 do {
1336 ch = *cp++;
1337 putchar(ch);
1338 } while (ch != '\n' &&
1339 cp < sb->final_buf + sb->final_buf_size);
1343 static void output(struct scoreboard *sb, int option)
1345 struct blame_entry *ent;
1347 if (option & OUTPUT_PORCELAIN) {
1348 for (ent = sb->ent; ent; ent = ent->next) {
1349 struct blame_entry *oth;
1350 struct origin *suspect = ent->suspect;
1351 struct commit *commit = suspect->commit;
1352 if (commit->object.flags & MORE_THAN_ONE_PATH)
1353 continue;
1354 for (oth = ent->next; oth; oth = oth->next) {
1355 if ((oth->suspect->commit != commit) ||
1356 !strcmp(oth->suspect->path, suspect->path))
1357 continue;
1358 commit->object.flags |= MORE_THAN_ONE_PATH;
1359 break;
1364 for (ent = sb->ent; ent; ent = ent->next) {
1365 if (option & OUTPUT_PORCELAIN)
1366 emit_porcelain(sb, ent);
1367 else {
1368 emit_other(sb, ent, option);
1373 static int prepare_lines(struct scoreboard *sb)
1375 const char *buf = sb->final_buf;
1376 unsigned long len = sb->final_buf_size;
1377 int num = 0, incomplete = 0, bol = 1;
1379 if (len && buf[len-1] != '\n')
1380 incomplete++; /* incomplete line at the end */
1381 while (len--) {
1382 if (bol) {
1383 sb->lineno = xrealloc(sb->lineno,
1384 sizeof(int* ) * (num + 1));
1385 sb->lineno[num] = buf - sb->final_buf;
1386 bol = 0;
1388 if (*buf++ == '\n') {
1389 num++;
1390 bol = 1;
1393 sb->lineno = xrealloc(sb->lineno,
1394 sizeof(int* ) * (num + incomplete + 1));
1395 sb->lineno[num + incomplete] = buf - sb->final_buf;
1396 sb->num_lines = num + incomplete;
1397 return sb->num_lines;
1400 static int read_ancestry(const char *graft_file)
1402 FILE *fp = fopen(graft_file, "r");
1403 char buf[1024];
1404 if (!fp)
1405 return -1;
1406 while (fgets(buf, sizeof(buf), fp)) {
1407 /* The format is just "Commit Parent1 Parent2 ...\n" */
1408 int len = strlen(buf);
1409 struct commit_graft *graft = read_graft_line(buf, len);
1410 register_commit_graft(graft, 0);
1412 fclose(fp);
1413 return 0;
1416 static int lineno_width(int lines)
1418 int i, width;
1420 for (width = 1, i = 10; i <= lines + 1; width++)
1421 i *= 10;
1422 return width;
1425 static void find_alignment(struct scoreboard *sb, int *option)
1427 int longest_src_lines = 0;
1428 int longest_dst_lines = 0;
1429 unsigned largest_score = 0;
1430 struct blame_entry *e;
1432 for (e = sb->ent; e; e = e->next) {
1433 struct origin *suspect = e->suspect;
1434 struct commit_info ci;
1435 int num;
1437 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1438 suspect->commit->object.flags |= METAINFO_SHOWN;
1439 get_commit_info(suspect->commit, &ci, 1);
1440 if (strcmp(suspect->path, sb->path))
1441 *option |= OUTPUT_SHOW_NAME;
1442 num = strlen(suspect->path);
1443 if (longest_file < num)
1444 longest_file = num;
1445 num = strlen(ci.author);
1446 if (longest_author < num)
1447 longest_author = num;
1449 num = e->s_lno + e->num_lines;
1450 if (longest_src_lines < num)
1451 longest_src_lines = num;
1452 num = e->lno + e->num_lines;
1453 if (longest_dst_lines < num)
1454 longest_dst_lines = num;
1455 if (largest_score < ent_score(sb, e))
1456 largest_score = ent_score(sb, e);
1458 max_orig_digits = lineno_width(longest_src_lines);
1459 max_digits = lineno_width(longest_dst_lines);
1460 max_score_digits = lineno_width(largest_score);
1463 static void sanity_check_refcnt(struct scoreboard *sb)
1465 int baa = 0;
1466 struct blame_entry *ent;
1468 for (ent = sb->ent; ent; ent = ent->next) {
1469 /* Nobody should have zero or negative refcnt */
1470 if (ent->suspect->refcnt <= 0) {
1471 fprintf(stderr, "%s in %s has negative refcnt %d\n",
1472 ent->suspect->path,
1473 sha1_to_hex(ent->suspect->commit->object.sha1),
1474 ent->suspect->refcnt);
1475 baa = 1;
1478 for (ent = sb->ent; ent; ent = ent->next) {
1479 /* Mark the ones that haven't been checked */
1480 if (0 < ent->suspect->refcnt)
1481 ent->suspect->refcnt = -ent->suspect->refcnt;
1483 for (ent = sb->ent; ent; ent = ent->next) {
1484 /* then pick each and see if they have the the correct
1485 * refcnt.
1487 int found;
1488 struct blame_entry *e;
1489 struct origin *suspect = ent->suspect;
1491 if (0 < suspect->refcnt)
1492 continue;
1493 suspect->refcnt = -suspect->refcnt; /* Unmark */
1494 for (found = 0, e = sb->ent; e; e = e->next) {
1495 if (e->suspect != suspect)
1496 continue;
1497 found++;
1499 if (suspect->refcnt != found) {
1500 fprintf(stderr, "%s in %s has refcnt %d, not %d\n",
1501 ent->suspect->path,
1502 sha1_to_hex(ent->suspect->commit->object.sha1),
1503 ent->suspect->refcnt, found);
1504 baa = 2;
1507 if (baa) {
1508 int opt = 0160;
1509 find_alignment(sb, &opt);
1510 output(sb, opt);
1511 die("Baa %d!", baa);
1515 static int has_path_in_work_tree(const char *path)
1517 struct stat st;
1518 return !lstat(path, &st);
1521 static unsigned parse_score(const char *arg)
1523 char *end;
1524 unsigned long score = strtoul(arg, &end, 10);
1525 if (*end)
1526 return 0;
1527 return score;
1530 static const char *add_prefix(const char *prefix, const char *path)
1532 if (!prefix || !prefix[0])
1533 return path;
1534 return prefix_path(prefix, strlen(prefix), path);
1537 static const char *parse_loc(const char *spec,
1538 struct scoreboard *sb, long lno,
1539 long begin, long *ret)
1541 char *term;
1542 const char *line;
1543 long num;
1544 int reg_error;
1545 regex_t regexp;
1546 regmatch_t match[1];
1548 /* Allow "-L <something>,+20" to mean starting at <something>
1549 * for 20 lines, or "-L <something>,-5" for 5 lines ending at
1550 * <something>.
1552 if (1 < begin && (spec[0] == '+' || spec[0] == '-')) {
1553 num = strtol(spec + 1, &term, 10);
1554 if (term != spec + 1) {
1555 if (spec[0] == '-')
1556 num = 0 - num;
1557 if (0 < num)
1558 *ret = begin + num - 2;
1559 else if (!num)
1560 *ret = begin;
1561 else
1562 *ret = begin + num;
1563 return term;
1565 return spec;
1567 num = strtol(spec, &term, 10);
1568 if (term != spec) {
1569 *ret = num;
1570 return term;
1572 if (spec[0] != '/')
1573 return spec;
1575 /* it could be a regexp of form /.../ */
1576 for (term = (char*) spec + 1; *term && *term != '/'; term++) {
1577 if (*term == '\\')
1578 term++;
1580 if (*term != '/')
1581 return spec;
1583 /* try [spec+1 .. term-1] as regexp */
1584 *term = 0;
1585 begin--; /* input is in human terms */
1586 line = nth_line(sb, begin);
1588 if (!(reg_error = regcomp(&regexp, spec + 1, REG_NEWLINE)) &&
1589 !(reg_error = regexec(&regexp, line, 1, match, 0))) {
1590 const char *cp = line + match[0].rm_so;
1591 const char *nline;
1593 while (begin++ < lno) {
1594 nline = nth_line(sb, begin);
1595 if (line <= cp && cp < nline)
1596 break;
1597 line = nline;
1599 *ret = begin;
1600 regfree(&regexp);
1601 *term++ = '/';
1602 return term;
1604 else {
1605 char errbuf[1024];
1606 regerror(reg_error, &regexp, errbuf, 1024);
1607 die("-L parameter '%s': %s", spec + 1, errbuf);
1611 static void prepare_blame_range(struct scoreboard *sb,
1612 const char *bottomtop,
1613 long lno,
1614 long *bottom, long *top)
1616 const char *term;
1618 term = parse_loc(bottomtop, sb, lno, 1, bottom);
1619 if (*term == ',') {
1620 term = parse_loc(term + 1, sb, lno, *bottom + 1, top);
1621 if (*term)
1622 usage(blame_usage);
1624 if (*term)
1625 usage(blame_usage);
1628 int cmd_blame(int argc, const char **argv, const char *prefix)
1630 struct rev_info revs;
1631 const char *path;
1632 struct scoreboard sb;
1633 struct origin *o;
1634 struct blame_entry *ent;
1635 int i, seen_dashdash, unk, opt;
1636 long bottom, top, lno;
1637 int output_option = 0;
1638 const char *revs_file = NULL;
1639 const char *final_commit_name = NULL;
1640 char type[10];
1641 const char *bottomtop = NULL;
1643 save_commit_buffer = 0;
1645 opt = 0;
1646 seen_dashdash = 0;
1647 for (unk = i = 1; i < argc; i++) {
1648 const char *arg = argv[i];
1649 if (*arg != '-')
1650 break;
1651 else if (!strcmp("-c", arg))
1652 output_option |= OUTPUT_ANNOTATE_COMPAT;
1653 else if (!strcmp("-t", arg))
1654 output_option |= OUTPUT_RAW_TIMESTAMP;
1655 else if (!strcmp("-l", arg))
1656 output_option |= OUTPUT_LONG_OBJECT_NAME;
1657 else if (!strcmp("-S", arg) && ++i < argc)
1658 revs_file = argv[i];
1659 else if (!strncmp("-M", arg, 2)) {
1660 opt |= PICKAXE_BLAME_MOVE;
1661 blame_move_score = parse_score(arg+2);
1663 else if (!strncmp("-C", arg, 2)) {
1664 if (opt & PICKAXE_BLAME_COPY)
1665 opt |= PICKAXE_BLAME_COPY_HARDER;
1666 opt |= PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE;
1667 blame_copy_score = parse_score(arg+2);
1669 else if (!strncmp("-L", arg, 2)) {
1670 if (!arg[2]) {
1671 if (++i >= argc)
1672 usage(blame_usage);
1673 arg = argv[i];
1675 else
1676 arg += 2;
1677 if (bottomtop)
1678 die("More than one '-L n,m' option given");
1679 bottomtop = arg;
1681 else if (!strcmp("--score-debug", arg))
1682 output_option |= OUTPUT_SHOW_SCORE;
1683 else if (!strcmp("-f", arg) ||
1684 !strcmp("--show-name", arg))
1685 output_option |= OUTPUT_SHOW_NAME;
1686 else if (!strcmp("-n", arg) ||
1687 !strcmp("--show-number", arg))
1688 output_option |= OUTPUT_SHOW_NUMBER;
1689 else if (!strcmp("-p", arg) ||
1690 !strcmp("--porcelain", arg))
1691 output_option |= OUTPUT_PORCELAIN;
1692 else if (!strcmp("--", arg)) {
1693 seen_dashdash = 1;
1694 i++;
1695 break;
1697 else
1698 argv[unk++] = arg;
1701 if (!blame_move_score)
1702 blame_move_score = BLAME_DEFAULT_MOVE_SCORE;
1703 if (!blame_copy_score)
1704 blame_copy_score = BLAME_DEFAULT_COPY_SCORE;
1706 /* We have collected options unknown to us in argv[1..unk]
1707 * which are to be passed to revision machinery if we are
1708 * going to do the "bottom" procesing.
1710 * The remaining are:
1712 * (1) if seen_dashdash, its either
1713 * "-options -- <path>" or
1714 * "-options -- <path> <rev>".
1715 * but the latter is allowed only if there is no
1716 * options that we passed to revision machinery.
1718 * (2) otherwise, we may have "--" somewhere later and
1719 * might be looking at the first one of multiple 'rev'
1720 * parameters (e.g. " master ^next ^maint -- path").
1721 * See if there is a dashdash first, and give the
1722 * arguments before that to revision machinery.
1723 * After that there must be one 'path'.
1725 * (3) otherwise, its one of the three:
1726 * "-options <path> <rev>"
1727 * "-options <rev> <path>"
1728 * "-options <path>"
1729 * but again the first one is allowed only if
1730 * there is no options that we passed to revision
1731 * machinery.
1734 if (seen_dashdash) {
1735 /* (1) */
1736 if (argc <= i)
1737 usage(blame_usage);
1738 path = add_prefix(prefix, argv[i]);
1739 if (i + 1 == argc - 1) {
1740 if (unk != 1)
1741 usage(blame_usage);
1742 argv[unk++] = argv[i + 1];
1744 else if (i + 1 != argc)
1745 /* garbage at end */
1746 usage(blame_usage);
1748 else {
1749 int j;
1750 for (j = i; !seen_dashdash && j < argc; j++)
1751 if (!strcmp(argv[j], "--"))
1752 seen_dashdash = j;
1753 if (seen_dashdash) {
1754 if (seen_dashdash + 1 != argc - 1)
1755 usage(blame_usage);
1756 path = add_prefix(prefix, argv[seen_dashdash + 1]);
1757 for (j = i; j < seen_dashdash; j++)
1758 argv[unk++] = argv[j];
1760 else {
1761 /* (3) */
1762 path = add_prefix(prefix, argv[i]);
1763 if (i + 1 == argc - 1) {
1764 final_commit_name = argv[i + 1];
1766 /* if (unk == 1) we could be getting
1767 * old-style
1769 if (unk == 1 && !has_path_in_work_tree(path)) {
1770 path = add_prefix(prefix, argv[i + 1]);
1771 final_commit_name = argv[i];
1774 else if (i != argc - 1)
1775 usage(blame_usage); /* garbage at end */
1777 if (!has_path_in_work_tree(path))
1778 die("cannot stat path %s: %s",
1779 path, strerror(errno));
1783 if (final_commit_name)
1784 argv[unk++] = final_commit_name;
1786 /* Now we got rev and path. We do not want the path pruning
1787 * but we may want "bottom" processing.
1789 argv[unk] = NULL;
1791 init_revisions(&revs, NULL);
1792 setup_revisions(unk, argv, &revs, "HEAD");
1793 memset(&sb, 0, sizeof(sb));
1795 /* There must be one and only one positive commit in the
1796 * revs->pending array.
1798 for (i = 0; i < revs.pending.nr; i++) {
1799 struct object *obj = revs.pending.objects[i].item;
1800 if (obj->flags & UNINTERESTING)
1801 continue;
1802 while (obj->type == OBJ_TAG)
1803 obj = deref_tag(obj, NULL, 0);
1804 if (obj->type != OBJ_COMMIT)
1805 die("Non commit %s?",
1806 revs.pending.objects[i].name);
1807 if (sb.final)
1808 die("More than one commit to dig from %s and %s?",
1809 revs.pending.objects[i].name,
1810 final_commit_name);
1811 sb.final = (struct commit *) obj;
1812 final_commit_name = revs.pending.objects[i].name;
1815 if (!sb.final) {
1816 /* "--not A B -- path" without anything positive */
1817 unsigned char head_sha1[20];
1819 final_commit_name = "HEAD";
1820 if (get_sha1(final_commit_name, head_sha1))
1821 die("No such ref: HEAD");
1822 sb.final = lookup_commit_reference(head_sha1);
1823 add_pending_object(&revs, &(sb.final->object), "HEAD");
1826 /* If we have bottom, this will mark the ancestors of the
1827 * bottom commits we would reach while traversing as
1828 * uninteresting.
1830 prepare_revision_walk(&revs);
1832 o = get_origin(&sb, sb.final, path);
1833 if (fill_blob_sha1(o))
1834 die("no such path %s in %s", path, final_commit_name);
1836 sb.final_buf = read_sha1_file(o->blob_sha1, type, &sb.final_buf_size);
1837 num_read_blob++;
1838 lno = prepare_lines(&sb);
1840 bottom = top = 0;
1841 if (bottomtop)
1842 prepare_blame_range(&sb, bottomtop, lno, &bottom, &top);
1843 if (bottom && top && top < bottom) {
1844 long tmp;
1845 tmp = top; top = bottom; bottom = tmp;
1847 if (bottom < 1)
1848 bottom = 1;
1849 if (top < 1)
1850 top = lno;
1851 bottom--;
1852 if (lno < top)
1853 die("file %s has only %lu lines", path, lno);
1855 ent = xcalloc(1, sizeof(*ent));
1856 ent->lno = bottom;
1857 ent->num_lines = top - bottom;
1858 ent->suspect = o;
1859 ent->s_lno = bottom;
1861 sb.ent = ent;
1862 sb.path = path;
1864 if (revs_file && read_ancestry(revs_file))
1865 die("reading graft file %s failed: %s",
1866 revs_file, strerror(errno));
1868 assign_blame(&sb, &revs, opt);
1870 coalesce(&sb);
1872 if (!(output_option & OUTPUT_PORCELAIN))
1873 find_alignment(&sb, &output_option);
1875 output(&sb, output_option);
1876 free((void *)sb.final_buf);
1877 for (ent = sb.ent; ent; ) {
1878 struct blame_entry *e = ent->next;
1879 free(ent);
1880 ent = e;
1883 if (DEBUG) {
1884 printf("num read blob: %d\n", num_read_blob);
1885 printf("num get patch: %d\n", num_get_patch);
1886 printf("num commits: %d\n", num_commits);
1888 return 0;