blame: -C -C -C
[git/git.git] / builtin-blame.c
1 /*
2 * Pickaxe
3 *
4 * Copyright (c) 2006, Junio C Hamano
5 */
6
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 "quote.h"
17 #include "xdiff-interface.h"
18 #include "cache-tree.h"
19
20 static char blame_usage[] =
21 "git-blame [-c] [-l] [-t] [-f] [-n] [-p] [-L n,m] [-S <revs-file>] [-M] [-C] [-C] [--contents <filename>] [--incremental] [commit] [--] file\n"
22 " -c Use the same output mode as git-annotate (Default: off)\n"
23 " -b Show blank SHA-1 for boundary commits (Default: off)\n"
24 " -l Show long commit SHA1 (Default: off)\n"
25 " --root Do not treat root commits as boundaries (Default: off)\n"
26 " -t 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 " --incremental Show blame entries as we find them, incrementally\n"
33 " --contents file Use <file>'s contents as the final image\n"
34 " -S revs-file Use revisions from revs-file instead of calling git-rev-list\n";
35
36 static int longest_file;
37 static int longest_author;
38 static int max_orig_digits;
39 static int max_digits;
40 static int max_score_digits;
41 static int show_root;
42 static int blank_boundary;
43 static int incremental;
44 static int cmd_is_annotate;
45
46 #ifndef DEBUG
47 #define DEBUG 0
48 #endif
49
50 /* stats */
51 static int num_read_blob;
52 static int num_get_patch;
53 static int num_commits;
54
55 #define PICKAXE_BLAME_MOVE 01
56 #define PICKAXE_BLAME_COPY 02
57 #define PICKAXE_BLAME_COPY_HARDER 04
58 #define PICKAXE_BLAME_COPY_HARDEST 010
59
60 /*
61 * blame for a blame_entry with score lower than these thresholds
62 * is not passed to the parent using move/copy logic.
63 */
64 static unsigned blame_move_score;
65 static unsigned blame_copy_score;
66 #define BLAME_DEFAULT_MOVE_SCORE 20
67 #define BLAME_DEFAULT_COPY_SCORE 40
68
69 /* bits #0..7 in revision.h, #8..11 used for merge_bases() in commit.c */
70 #define METAINFO_SHOWN (1u<<12)
71 #define MORE_THAN_ONE_PATH (1u<<13)
72
73 /*
74 * One blob in a commit that is being suspected
75 */
76 struct origin {
77 int refcnt;
78 struct commit *commit;
79 mmfile_t file;
80 unsigned char blob_sha1[20];
81 char path[FLEX_ARRAY];
82 };
83
84 /*
85 * Given an origin, prepare mmfile_t structure to be used by the
86 * diff machinery
87 */
88 static char *fill_origin_blob(struct origin *o, mmfile_t *file)
89 {
90 if (!o->file.ptr) {
91 enum object_type type;
92 num_read_blob++;
93 file->ptr = read_sha1_file(o->blob_sha1, &type,
94 (unsigned long *)(&(file->size)));
95 o->file = *file;
96 }
97 else
98 *file = o->file;
99 return file->ptr;
100 }
101
102 /*
103 * Origin is refcounted and usually we keep the blob contents to be
104 * reused.
105 */
106 static inline struct origin *origin_incref(struct origin *o)
107 {
108 if (o)
109 o->refcnt++;
110 return o;
111 }
112
113 static void origin_decref(struct origin *o)
114 {
115 if (o && --o->refcnt <= 0) {
116 if (o->file.ptr)
117 free(o->file.ptr);
118 memset(o, 0, sizeof(*o));
119 free(o);
120 }
121 }
122
123 /*
124 * Each group of lines is described by a blame_entry; it can be split
125 * as we pass blame to the parents. They form a linked list in the
126 * scoreboard structure, sorted by the target line number.
127 */
128 struct blame_entry {
129 struct blame_entry *prev;
130 struct blame_entry *next;
131
132 /* the first line of this group in the final image;
133 * internally all line numbers are 0 based.
134 */
135 int lno;
136
137 /* how many lines this group has */
138 int num_lines;
139
140 /* the commit that introduced this group into the final image */
141 struct origin *suspect;
142
143 /* true if the suspect is truly guilty; false while we have not
144 * checked if the group came from one of its parents.
145 */
146 char guilty;
147
148 /* the line number of the first line of this group in the
149 * suspect's file; internally all line numbers are 0 based.
150 */
151 int s_lno;
152
153 /* how significant this entry is -- cached to avoid
154 * scanning the lines over and over.
155 */
156 unsigned score;
157 };
158
159 /*
160 * The current state of the blame assignment.
161 */
162 struct scoreboard {
163 /* the final commit (i.e. where we started digging from) */
164 struct commit *final;
165
166 const char *path;
167
168 /*
169 * The contents in the final image.
170 * Used by many functions to obtain contents of the nth line,
171 * indexed with scoreboard.lineno[blame_entry.lno].
172 */
173 const char *final_buf;
174 unsigned long final_buf_size;
175
176 /* linked list of blames */
177 struct blame_entry *ent;
178
179 /* look-up a line in the final buffer */
180 int num_lines;
181 int *lineno;
182 };
183
184 static inline int same_suspect(struct origin *a, struct origin *b)
185 {
186 if (a == b)
187 return 1;
188 if (a->commit != b->commit)
189 return 0;
190 return !strcmp(a->path, b->path);
191 }
192
193 static void sanity_check_refcnt(struct scoreboard *);
194
195 /*
196 * If two blame entries that are next to each other came from
197 * contiguous lines in the same origin (i.e. <commit, path> pair),
198 * merge them together.
199 */
200 static void coalesce(struct scoreboard *sb)
201 {
202 struct blame_entry *ent, *next;
203
204 for (ent = sb->ent; ent && (next = ent->next); ent = next) {
205 if (same_suspect(ent->suspect, next->suspect) &&
206 ent->guilty == next->guilty &&
207 ent->s_lno + ent->num_lines == next->s_lno) {
208 ent->num_lines += next->num_lines;
209 ent->next = next->next;
210 if (ent->next)
211 ent->next->prev = ent;
212 origin_decref(next->suspect);
213 free(next);
214 ent->score = 0;
215 next = ent; /* again */
216 }
217 }
218
219 if (DEBUG) /* sanity */
220 sanity_check_refcnt(sb);
221 }
222
223 /*
224 * Given a commit and a path in it, create a new origin structure.
225 * The callers that add blame to the scoreboard should use
226 * get_origin() to obtain shared, refcounted copy instead of calling
227 * this function directly.
228 */
229 static struct origin *make_origin(struct commit *commit, const char *path)
230 {
231 struct origin *o;
232 o = xcalloc(1, sizeof(*o) + strlen(path) + 1);
233 o->commit = commit;
234 o->refcnt = 1;
235 strcpy(o->path, path);
236 return o;
237 }
238
239 /*
240 * Locate an existing origin or create a new one.
241 */
242 static struct origin *get_origin(struct scoreboard *sb,
243 struct commit *commit,
244 const char *path)
245 {
246 struct blame_entry *e;
247
248 for (e = sb->ent; e; e = e->next) {
249 if (e->suspect->commit == commit &&
250 !strcmp(e->suspect->path, path))
251 return origin_incref(e->suspect);
252 }
253 return make_origin(commit, path);
254 }
255
256 /*
257 * Fill the blob_sha1 field of an origin if it hasn't, so that later
258 * call to fill_origin_blob() can use it to locate the data. blob_sha1
259 * for an origin is also used to pass the blame for the entire file to
260 * the parent to detect the case where a child's blob is identical to
261 * that of its parent's.
262 */
263 static int fill_blob_sha1(struct origin *origin)
264 {
265 unsigned mode;
266
267 if (!is_null_sha1(origin->blob_sha1))
268 return 0;
269 if (get_tree_entry(origin->commit->object.sha1,
270 origin->path,
271 origin->blob_sha1, &mode))
272 goto error_out;
273 if (sha1_object_info(origin->blob_sha1, NULL) != OBJ_BLOB)
274 goto error_out;
275 return 0;
276 error_out:
277 hashclr(origin->blob_sha1);
278 return -1;
279 }
280
281 /*
282 * We have an origin -- check if the same path exists in the
283 * parent and return an origin structure to represent it.
284 */
285 static struct origin *find_origin(struct scoreboard *sb,
286 struct commit *parent,
287 struct origin *origin)
288 {
289 struct origin *porigin = NULL;
290 struct diff_options diff_opts;
291 const char *paths[2];
292
293 if (parent->util) {
294 /*
295 * Each commit object can cache one origin in that
296 * commit. This is a freestanding copy of origin and
297 * not refcounted.
298 */
299 struct origin *cached = parent->util;
300 if (!strcmp(cached->path, origin->path)) {
301 /*
302 * The same path between origin and its parent
303 * without renaming -- the most common case.
304 */
305 porigin = get_origin(sb, parent, cached->path);
306
307 /*
308 * If the origin was newly created (i.e. get_origin
309 * would call make_origin if none is found in the
310 * scoreboard), it does not know the blob_sha1,
311 * so copy it. Otherwise porigin was in the
312 * scoreboard and already knows blob_sha1.
313 */
314 if (porigin->refcnt == 1)
315 hashcpy(porigin->blob_sha1, cached->blob_sha1);
316 return porigin;
317 }
318 /* otherwise it was not very useful; free it */
319 free(parent->util);
320 parent->util = NULL;
321 }
322
323 /* See if the origin->path is different between parent
324 * and origin first. Most of the time they are the
325 * same and diff-tree is fairly efficient about this.
326 */
327 diff_setup(&diff_opts);
328 diff_opts.recursive = 1;
329 diff_opts.detect_rename = 0;
330 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
331 paths[0] = origin->path;
332 paths[1] = NULL;
333
334 diff_tree_setup_paths(paths, &diff_opts);
335 if (diff_setup_done(&diff_opts) < 0)
336 die("diff-setup");
337
338 if (is_null_sha1(origin->commit->object.sha1))
339 do_diff_cache(parent->tree->object.sha1, &diff_opts);
340 else
341 diff_tree_sha1(parent->tree->object.sha1,
342 origin->commit->tree->object.sha1,
343 "", &diff_opts);
344 diffcore_std(&diff_opts);
345
346 /* It is either one entry that says "modified", or "created",
347 * or nothing.
348 */
349 if (!diff_queued_diff.nr) {
350 /* The path is the same as parent */
351 porigin = get_origin(sb, parent, origin->path);
352 hashcpy(porigin->blob_sha1, origin->blob_sha1);
353 }
354 else if (diff_queued_diff.nr != 1)
355 die("internal error in blame::find_origin");
356 else {
357 struct diff_filepair *p = diff_queued_diff.queue[0];
358 switch (p->status) {
359 default:
360 die("internal error in blame::find_origin (%c)",
361 p->status);
362 case 'M':
363 porigin = get_origin(sb, parent, origin->path);
364 hashcpy(porigin->blob_sha1, p->one->sha1);
365 break;
366 case 'A':
367 case 'T':
368 /* Did not exist in parent, or type changed */
369 break;
370 }
371 }
372 diff_flush(&diff_opts);
373 if (porigin) {
374 /*
375 * Create a freestanding copy that is not part of
376 * the refcounted origin found in the scoreboard, and
377 * cache it in the commit.
378 */
379 struct origin *cached;
380
381 cached = make_origin(porigin->commit, porigin->path);
382 hashcpy(cached->blob_sha1, porigin->blob_sha1);
383 parent->util = cached;
384 }
385 return porigin;
386 }
387
388 /*
389 * We have an origin -- find the path that corresponds to it in its
390 * parent and return an origin structure to represent it.
391 */
392 static struct origin *find_rename(struct scoreboard *sb,
393 struct commit *parent,
394 struct origin *origin)
395 {
396 struct origin *porigin = NULL;
397 struct diff_options diff_opts;
398 int i;
399 const char *paths[2];
400
401 diff_setup(&diff_opts);
402 diff_opts.recursive = 1;
403 diff_opts.detect_rename = DIFF_DETECT_RENAME;
404 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
405 diff_opts.single_follow = origin->path;
406 paths[0] = NULL;
407 diff_tree_setup_paths(paths, &diff_opts);
408 if (diff_setup_done(&diff_opts) < 0)
409 die("diff-setup");
410
411 if (is_null_sha1(origin->commit->object.sha1))
412 do_diff_cache(parent->tree->object.sha1, &diff_opts);
413 else
414 diff_tree_sha1(parent->tree->object.sha1,
415 origin->commit->tree->object.sha1,
416 "", &diff_opts);
417 diffcore_std(&diff_opts);
418
419 for (i = 0; i < diff_queued_diff.nr; i++) {
420 struct diff_filepair *p = diff_queued_diff.queue[i];
421 if ((p->status == 'R' || p->status == 'C') &&
422 !strcmp(p->two->path, origin->path)) {
423 porigin = get_origin(sb, parent, p->one->path);
424 hashcpy(porigin->blob_sha1, p->one->sha1);
425 break;
426 }
427 }
428 diff_flush(&diff_opts);
429 return porigin;
430 }
431
432 /*
433 * Parsing of patch chunks...
434 */
435 struct chunk {
436 /* line number in postimage; up to but not including this
437 * line is the same as preimage
438 */
439 int same;
440
441 /* preimage line number after this chunk */
442 int p_next;
443
444 /* postimage line number after this chunk */
445 int t_next;
446 };
447
448 struct patch {
449 struct chunk *chunks;
450 int num;
451 };
452
453 struct blame_diff_state {
454 struct xdiff_emit_state xm;
455 struct patch *ret;
456 unsigned hunk_post_context;
457 unsigned hunk_in_pre_context : 1;
458 };
459
460 static void process_u_diff(void *state_, char *line, unsigned long len)
461 {
462 struct blame_diff_state *state = state_;
463 struct chunk *chunk;
464 int off1, off2, len1, len2, num;
465
466 num = state->ret->num;
467 if (len < 4 || line[0] != '@' || line[1] != '@') {
468 if (state->hunk_in_pre_context && line[0] == ' ')
469 state->ret->chunks[num - 1].same++;
470 else {
471 state->hunk_in_pre_context = 0;
472 if (line[0] == ' ')
473 state->hunk_post_context++;
474 else
475 state->hunk_post_context = 0;
476 }
477 return;
478 }
479
480 if (num && state->hunk_post_context) {
481 chunk = &state->ret->chunks[num - 1];
482 chunk->p_next -= state->hunk_post_context;
483 chunk->t_next -= state->hunk_post_context;
484 }
485 state->ret->num = ++num;
486 state->ret->chunks = xrealloc(state->ret->chunks,
487 sizeof(struct chunk) * num);
488 chunk = &state->ret->chunks[num - 1];
489 if (parse_hunk_header(line, len, &off1, &len1, &off2, &len2)) {
490 state->ret->num--;
491 return;
492 }
493
494 /* Line numbers in patch output are one based. */
495 off1--;
496 off2--;
497
498 chunk->same = len2 ? off2 : (off2 + 1);
499
500 chunk->p_next = off1 + (len1 ? len1 : 1);
501 chunk->t_next = chunk->same + len2;
502 state->hunk_in_pre_context = 1;
503 state->hunk_post_context = 0;
504 }
505
506 static struct patch *compare_buffer(mmfile_t *file_p, mmfile_t *file_o,
507 int context)
508 {
509 struct blame_diff_state state;
510 xpparam_t xpp;
511 xdemitconf_t xecfg;
512 xdemitcb_t ecb;
513
514 xpp.flags = XDF_NEED_MINIMAL;
515 xecfg.ctxlen = context;
516 xecfg.flags = 0;
517 ecb.outf = xdiff_outf;
518 ecb.priv = &state;
519 memset(&state, 0, sizeof(state));
520 state.xm.consume = process_u_diff;
521 state.ret = xmalloc(sizeof(struct patch));
522 state.ret->chunks = NULL;
523 state.ret->num = 0;
524
525 xdl_diff(file_p, file_o, &xpp, &xecfg, &ecb);
526
527 if (state.ret->num) {
528 struct chunk *chunk;
529 chunk = &state.ret->chunks[state.ret->num - 1];
530 chunk->p_next -= state.hunk_post_context;
531 chunk->t_next -= state.hunk_post_context;
532 }
533 return state.ret;
534 }
535
536 /*
537 * Run diff between two origins and grab the patch output, so that
538 * we can pass blame for lines origin is currently suspected for
539 * to its parent.
540 */
541 static struct patch *get_patch(struct origin *parent, struct origin *origin)
542 {
543 mmfile_t file_p, file_o;
544 struct patch *patch;
545
546 fill_origin_blob(parent, &file_p);
547 fill_origin_blob(origin, &file_o);
548 if (!file_p.ptr || !file_o.ptr)
549 return NULL;
550 patch = compare_buffer(&file_p, &file_o, 0);
551 num_get_patch++;
552 return patch;
553 }
554
555 static void free_patch(struct patch *p)
556 {
557 free(p->chunks);
558 free(p);
559 }
560
561 /*
562 * Link in a new blame entry to the scoreboard. Entries that cover the
563 * same line range have been removed from the scoreboard previously.
564 */
565 static void add_blame_entry(struct scoreboard *sb, struct blame_entry *e)
566 {
567 struct blame_entry *ent, *prev = NULL;
568
569 origin_incref(e->suspect);
570
571 for (ent = sb->ent; ent && ent->lno < e->lno; ent = ent->next)
572 prev = ent;
573
574 /* prev, if not NULL, is the last one that is below e */
575 e->prev = prev;
576 if (prev) {
577 e->next = prev->next;
578 prev->next = e;
579 }
580 else {
581 e->next = sb->ent;
582 sb->ent = e;
583 }
584 if (e->next)
585 e->next->prev = e;
586 }
587
588 /*
589 * src typically is on-stack; we want to copy the information in it to
590 * an malloced blame_entry that is already on the linked list of the
591 * scoreboard. The origin of dst loses a refcnt while the origin of src
592 * gains one.
593 */
594 static void dup_entry(struct blame_entry *dst, struct blame_entry *src)
595 {
596 struct blame_entry *p, *n;
597
598 p = dst->prev;
599 n = dst->next;
600 origin_incref(src->suspect);
601 origin_decref(dst->suspect);
602 memcpy(dst, src, sizeof(*src));
603 dst->prev = p;
604 dst->next = n;
605 dst->score = 0;
606 }
607
608 static const char *nth_line(struct scoreboard *sb, int lno)
609 {
610 return sb->final_buf + sb->lineno[lno];
611 }
612
613 /*
614 * It is known that lines between tlno to same came from parent, and e
615 * has an overlap with that range. it also is known that parent's
616 * line plno corresponds to e's line tlno.
617 *
618 * <---- e ----->
619 * <------>
620 * <------------>
621 * <------------>
622 * <------------------>
623 *
624 * Split e into potentially three parts; before this chunk, the chunk
625 * to be blamed for the parent, and after that portion.
626 */
627 static void split_overlap(struct blame_entry *split,
628 struct blame_entry *e,
629 int tlno, int plno, int same,
630 struct origin *parent)
631 {
632 int chunk_end_lno;
633 memset(split, 0, sizeof(struct blame_entry [3]));
634
635 if (e->s_lno < tlno) {
636 /* there is a pre-chunk part not blamed on parent */
637 split[0].suspect = origin_incref(e->suspect);
638 split[0].lno = e->lno;
639 split[0].s_lno = e->s_lno;
640 split[0].num_lines = tlno - e->s_lno;
641 split[1].lno = e->lno + tlno - e->s_lno;
642 split[1].s_lno = plno;
643 }
644 else {
645 split[1].lno = e->lno;
646 split[1].s_lno = plno + (e->s_lno - tlno);
647 }
648
649 if (same < e->s_lno + e->num_lines) {
650 /* there is a post-chunk part not blamed on parent */
651 split[2].suspect = origin_incref(e->suspect);
652 split[2].lno = e->lno + (same - e->s_lno);
653 split[2].s_lno = e->s_lno + (same - e->s_lno);
654 split[2].num_lines = e->s_lno + e->num_lines - same;
655 chunk_end_lno = split[2].lno;
656 }
657 else
658 chunk_end_lno = e->lno + e->num_lines;
659 split[1].num_lines = chunk_end_lno - split[1].lno;
660
661 /*
662 * if it turns out there is nothing to blame the parent for,
663 * forget about the splitting. !split[1].suspect signals this.
664 */
665 if (split[1].num_lines < 1)
666 return;
667 split[1].suspect = origin_incref(parent);
668 }
669
670 /*
671 * split_overlap() divided an existing blame e into up to three parts
672 * in split. Adjust the linked list of blames in the scoreboard to
673 * reflect the split.
674 */
675 static void split_blame(struct scoreboard *sb,
676 struct blame_entry *split,
677 struct blame_entry *e)
678 {
679 struct blame_entry *new_entry;
680
681 if (split[0].suspect && split[2].suspect) {
682 /* The first part (reuse storage for the existing entry e) */
683 dup_entry(e, &split[0]);
684
685 /* The last part -- me */
686 new_entry = xmalloc(sizeof(*new_entry));
687 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
688 add_blame_entry(sb, new_entry);
689
690 /* ... and the middle part -- parent */
691 new_entry = xmalloc(sizeof(*new_entry));
692 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
693 add_blame_entry(sb, new_entry);
694 }
695 else if (!split[0].suspect && !split[2].suspect)
696 /*
697 * The parent covers the entire area; reuse storage for
698 * e and replace it with the parent.
699 */
700 dup_entry(e, &split[1]);
701 else if (split[0].suspect) {
702 /* me and then parent */
703 dup_entry(e, &split[0]);
704
705 new_entry = xmalloc(sizeof(*new_entry));
706 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
707 add_blame_entry(sb, new_entry);
708 }
709 else {
710 /* parent and then me */
711 dup_entry(e, &split[1]);
712
713 new_entry = xmalloc(sizeof(*new_entry));
714 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
715 add_blame_entry(sb, new_entry);
716 }
717
718 if (DEBUG) { /* sanity */
719 struct blame_entry *ent;
720 int lno = sb->ent->lno, corrupt = 0;
721
722 for (ent = sb->ent; ent; ent = ent->next) {
723 if (lno != ent->lno)
724 corrupt = 1;
725 if (ent->s_lno < 0)
726 corrupt = 1;
727 lno += ent->num_lines;
728 }
729 if (corrupt) {
730 lno = sb->ent->lno;
731 for (ent = sb->ent; ent; ent = ent->next) {
732 printf("L %8d l %8d n %8d\n",
733 lno, ent->lno, ent->num_lines);
734 lno = ent->lno + ent->num_lines;
735 }
736 die("oops");
737 }
738 }
739 }
740
741 /*
742 * After splitting the blame, the origins used by the
743 * on-stack blame_entry should lose one refcnt each.
744 */
745 static void decref_split(struct blame_entry *split)
746 {
747 int i;
748
749 for (i = 0; i < 3; i++)
750 origin_decref(split[i].suspect);
751 }
752
753 /*
754 * Helper for blame_chunk(). blame_entry e is known to overlap with
755 * the patch hunk; split it and pass blame to the parent.
756 */
757 static void blame_overlap(struct scoreboard *sb, struct blame_entry *e,
758 int tlno, int plno, int same,
759 struct origin *parent)
760 {
761 struct blame_entry split[3];
762
763 split_overlap(split, e, tlno, plno, same, parent);
764 if (split[1].suspect)
765 split_blame(sb, split, e);
766 decref_split(split);
767 }
768
769 /*
770 * Find the line number of the last line the target is suspected for.
771 */
772 static int find_last_in_target(struct scoreboard *sb, struct origin *target)
773 {
774 struct blame_entry *e;
775 int last_in_target = -1;
776
777 for (e = sb->ent; e; e = e->next) {
778 if (e->guilty || !same_suspect(e->suspect, target))
779 continue;
780 if (last_in_target < e->s_lno + e->num_lines)
781 last_in_target = e->s_lno + e->num_lines;
782 }
783 return last_in_target;
784 }
785
786 /*
787 * Process one hunk from the patch between the current suspect for
788 * blame_entry e and its parent. Find and split the overlap, and
789 * pass blame to the overlapping part to the parent.
790 */
791 static void blame_chunk(struct scoreboard *sb,
792 int tlno, int plno, int same,
793 struct origin *target, struct origin *parent)
794 {
795 struct blame_entry *e;
796
797 for (e = sb->ent; e; e = e->next) {
798 if (e->guilty || !same_suspect(e->suspect, target))
799 continue;
800 if (same <= e->s_lno)
801 continue;
802 if (tlno < e->s_lno + e->num_lines)
803 blame_overlap(sb, e, tlno, plno, same, parent);
804 }
805 }
806
807 /*
808 * We are looking at the origin 'target' and aiming to pass blame
809 * for the lines it is suspected to its parent. Run diff to find
810 * which lines came from parent and pass blame for them.
811 */
812 static int pass_blame_to_parent(struct scoreboard *sb,
813 struct origin *target,
814 struct origin *parent)
815 {
816 int i, last_in_target, plno, tlno;
817 struct patch *patch;
818
819 last_in_target = find_last_in_target(sb, target);
820 if (last_in_target < 0)
821 return 1; /* nothing remains for this target */
822
823 patch = get_patch(parent, target);
824 plno = tlno = 0;
825 for (i = 0; i < patch->num; i++) {
826 struct chunk *chunk = &patch->chunks[i];
827
828 blame_chunk(sb, tlno, plno, chunk->same, target, parent);
829 plno = chunk->p_next;
830 tlno = chunk->t_next;
831 }
832 /* The rest (i.e. anything after tlno) are the same as the parent */
833 blame_chunk(sb, tlno, plno, last_in_target, target, parent);
834
835 free_patch(patch);
836 return 0;
837 }
838
839 /*
840 * The lines in blame_entry after splitting blames many times can become
841 * very small and trivial, and at some point it becomes pointless to
842 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
843 * ordinary C program, and it is not worth to say it was copied from
844 * totally unrelated file in the parent.
845 *
846 * Compute how trivial the lines in the blame_entry are.
847 */
848 static unsigned ent_score(struct scoreboard *sb, struct blame_entry *e)
849 {
850 unsigned score;
851 const char *cp, *ep;
852
853 if (e->score)
854 return e->score;
855
856 score = 1;
857 cp = nth_line(sb, e->lno);
858 ep = nth_line(sb, e->lno + e->num_lines);
859 while (cp < ep) {
860 unsigned ch = *((unsigned char *)cp);
861 if (isalnum(ch))
862 score++;
863 cp++;
864 }
865 e->score = score;
866 return score;
867 }
868
869 /*
870 * best_so_far[] and this[] are both a split of an existing blame_entry
871 * that passes blame to the parent. Maintain best_so_far the best split
872 * so far, by comparing this and best_so_far and copying this into
873 * bst_so_far as needed.
874 */
875 static void copy_split_if_better(struct scoreboard *sb,
876 struct blame_entry *best_so_far,
877 struct blame_entry *this)
878 {
879 int i;
880
881 if (!this[1].suspect)
882 return;
883 if (best_so_far[1].suspect) {
884 if (ent_score(sb, &this[1]) < ent_score(sb, &best_so_far[1]))
885 return;
886 }
887
888 for (i = 0; i < 3; i++)
889 origin_incref(this[i].suspect);
890 decref_split(best_so_far);
891 memcpy(best_so_far, this, sizeof(struct blame_entry [3]));
892 }
893
894 /*
895 * We are looking at a part of the final image represented by
896 * ent (tlno and same are offset by ent->s_lno).
897 * tlno is where we are looking at in the final image.
898 * up to (but not including) same match preimage.
899 * plno is where we are looking at in the preimage.
900 *
901 * <-------------- final image ---------------------->
902 * <------ent------>
903 * ^tlno ^same
904 * <---------preimage----->
905 * ^plno
906 *
907 * All line numbers are 0-based.
908 */
909 static void handle_split(struct scoreboard *sb,
910 struct blame_entry *ent,
911 int tlno, int plno, int same,
912 struct origin *parent,
913 struct blame_entry *split)
914 {
915 if (ent->num_lines <= tlno)
916 return;
917 if (tlno < same) {
918 struct blame_entry this[3];
919 tlno += ent->s_lno;
920 same += ent->s_lno;
921 split_overlap(this, ent, tlno, plno, same, parent);
922 copy_split_if_better(sb, split, this);
923 decref_split(this);
924 }
925 }
926
927 /*
928 * Find the lines from parent that are the same as ent so that
929 * we can pass blames to it. file_p has the blob contents for
930 * the parent.
931 */
932 static void find_copy_in_blob(struct scoreboard *sb,
933 struct blame_entry *ent,
934 struct origin *parent,
935 struct blame_entry *split,
936 mmfile_t *file_p)
937 {
938 const char *cp;
939 int cnt;
940 mmfile_t file_o;
941 struct patch *patch;
942 int i, plno, tlno;
943
944 /*
945 * Prepare mmfile that contains only the lines in ent.
946 */
947 cp = nth_line(sb, ent->lno);
948 file_o.ptr = (char*) cp;
949 cnt = ent->num_lines;
950
951 while (cnt && cp < sb->final_buf + sb->final_buf_size) {
952 if (*cp++ == '\n')
953 cnt--;
954 }
955 file_o.size = cp - file_o.ptr;
956
957 patch = compare_buffer(file_p, &file_o, 1);
958
959 /*
960 * file_o is a part of final image we are annotating.
961 * file_p partially may match that image.
962 */
963 memset(split, 0, sizeof(struct blame_entry [3]));
964 plno = tlno = 0;
965 for (i = 0; i < patch->num; i++) {
966 struct chunk *chunk = &patch->chunks[i];
967
968 handle_split(sb, ent, tlno, plno, chunk->same, parent, split);
969 plno = chunk->p_next;
970 tlno = chunk->t_next;
971 }
972 /* remainder, if any, all match the preimage */
973 handle_split(sb, ent, tlno, plno, ent->num_lines, parent, split);
974 free_patch(patch);
975 }
976
977 /*
978 * See if lines currently target is suspected for can be attributed to
979 * parent.
980 */
981 static int find_move_in_parent(struct scoreboard *sb,
982 struct origin *target,
983 struct origin *parent)
984 {
985 int last_in_target, made_progress;
986 struct blame_entry *e, split[3];
987 mmfile_t file_p;
988
989 last_in_target = find_last_in_target(sb, target);
990 if (last_in_target < 0)
991 return 1; /* nothing remains for this target */
992
993 fill_origin_blob(parent, &file_p);
994 if (!file_p.ptr)
995 return 0;
996
997 made_progress = 1;
998 while (made_progress) {
999 made_progress = 0;
1000 for (e = sb->ent; e; e = e->next) {
1001 if (e->guilty || !same_suspect(e->suspect, target))
1002 continue;
1003 find_copy_in_blob(sb, e, parent, split, &file_p);
1004 if (split[1].suspect &&
1005 blame_move_score < ent_score(sb, &split[1])) {
1006 split_blame(sb, split, e);
1007 made_progress = 1;
1008 }
1009 decref_split(split);
1010 }
1011 }
1012 return 0;
1013 }
1014
1015 struct blame_list {
1016 struct blame_entry *ent;
1017 struct blame_entry split[3];
1018 };
1019
1020 /*
1021 * Count the number of entries the target is suspected for,
1022 * and prepare a list of entry and the best split.
1023 */
1024 static struct blame_list *setup_blame_list(struct scoreboard *sb,
1025 struct origin *target,
1026 int *num_ents_p)
1027 {
1028 struct blame_entry *e;
1029 int num_ents, i;
1030 struct blame_list *blame_list = NULL;
1031
1032 for (e = sb->ent, num_ents = 0; e; e = e->next)
1033 if (!e->guilty && same_suspect(e->suspect, target))
1034 num_ents++;
1035 if (num_ents) {
1036 blame_list = xcalloc(num_ents, sizeof(struct blame_list));
1037 for (e = sb->ent, i = 0; e; e = e->next)
1038 if (!e->guilty && same_suspect(e->suspect, target))
1039 blame_list[i++].ent = e;
1040 }
1041 *num_ents_p = num_ents;
1042 return blame_list;
1043 }
1044
1045 /*
1046 * For lines target is suspected for, see if we can find code movement
1047 * across file boundary from the parent commit. porigin is the path
1048 * in the parent we already tried.
1049 */
1050 static int find_copy_in_parent(struct scoreboard *sb,
1051 struct origin *target,
1052 struct commit *parent,
1053 struct origin *porigin,
1054 int opt)
1055 {
1056 struct diff_options diff_opts;
1057 const char *paths[1];
1058 int i, j;
1059 int retval;
1060 struct blame_list *blame_list;
1061 int num_ents;
1062
1063 blame_list = setup_blame_list(sb, target, &num_ents);
1064 if (!blame_list)
1065 return 1; /* nothing remains for this target */
1066
1067 diff_setup(&diff_opts);
1068 diff_opts.recursive = 1;
1069 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
1070
1071 paths[0] = NULL;
1072 diff_tree_setup_paths(paths, &diff_opts);
1073 if (diff_setup_done(&diff_opts) < 0)
1074 die("diff-setup");
1075
1076 /* Try "find copies harder" on new path if requested;
1077 * we do not want to use diffcore_rename() actually to
1078 * match things up; find_copies_harder is set only to
1079 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1080 * and this code needs to be after diff_setup_done(), which
1081 * usually makes find-copies-harder imply copy detection.
1082 */
1083 if ((opt & PICKAXE_BLAME_COPY_HARDEST)
1084 || ((opt & PICKAXE_BLAME_COPY_HARDER)
1085 && (!porigin || strcmp(target->path, porigin->path))))
1086 diff_opts.find_copies_harder = 1;
1087
1088 if (is_null_sha1(target->commit->object.sha1))
1089 do_diff_cache(parent->tree->object.sha1, &diff_opts);
1090 else
1091 diff_tree_sha1(parent->tree->object.sha1,
1092 target->commit->tree->object.sha1,
1093 "", &diff_opts);
1094
1095 if (!diff_opts.find_copies_harder)
1096 diffcore_std(&diff_opts);
1097
1098 retval = 0;
1099 while (1) {
1100 int made_progress = 0;
1101
1102 for (i = 0; i < diff_queued_diff.nr; i++) {
1103 struct diff_filepair *p = diff_queued_diff.queue[i];
1104 struct origin *norigin;
1105 mmfile_t file_p;
1106 struct blame_entry this[3];
1107
1108 if (!DIFF_FILE_VALID(p->one))
1109 continue; /* does not exist in parent */
1110 if (porigin && !strcmp(p->one->path, porigin->path))
1111 /* find_move already dealt with this path */
1112 continue;
1113
1114 norigin = get_origin(sb, parent, p->one->path);
1115 hashcpy(norigin->blob_sha1, p->one->sha1);
1116 fill_origin_blob(norigin, &file_p);
1117 if (!file_p.ptr)
1118 continue;
1119
1120 for (j = 0; j < num_ents; j++) {
1121 find_copy_in_blob(sb, blame_list[j].ent,
1122 norigin, this, &file_p);
1123 copy_split_if_better(sb, blame_list[j].split,
1124 this);
1125 decref_split(this);
1126 }
1127 origin_decref(norigin);
1128 }
1129
1130 for (j = 0; j < num_ents; j++) {
1131 struct blame_entry *split = blame_list[j].split;
1132 if (split[1].suspect &&
1133 blame_copy_score < ent_score(sb, &split[1])) {
1134 split_blame(sb, split, blame_list[j].ent);
1135 made_progress = 1;
1136 }
1137 decref_split(split);
1138 }
1139 free(blame_list);
1140
1141 if (!made_progress)
1142 break;
1143 blame_list = setup_blame_list(sb, target, &num_ents);
1144 if (!blame_list) {
1145 retval = 1;
1146 break;
1147 }
1148 }
1149 diff_flush(&diff_opts);
1150
1151 return retval;
1152 }
1153
1154 /*
1155 * The blobs of origin and porigin exactly match, so everything
1156 * origin is suspected for can be blamed on the parent.
1157 */
1158 static void pass_whole_blame(struct scoreboard *sb,
1159 struct origin *origin, struct origin *porigin)
1160 {
1161 struct blame_entry *e;
1162
1163 if (!porigin->file.ptr && origin->file.ptr) {
1164 /* Steal its file */
1165 porigin->file = origin->file;
1166 origin->file.ptr = NULL;
1167 }
1168 for (e = sb->ent; e; e = e->next) {
1169 if (!same_suspect(e->suspect, origin))
1170 continue;
1171 origin_incref(porigin);
1172 origin_decref(e->suspect);
1173 e->suspect = porigin;
1174 }
1175 }
1176
1177 #define MAXPARENT 16
1178
1179 static void pass_blame(struct scoreboard *sb, struct origin *origin, int opt)
1180 {
1181 int i, pass;
1182 struct commit *commit = origin->commit;
1183 struct commit_list *parent;
1184 struct origin *parent_origin[MAXPARENT], *porigin;
1185
1186 memset(parent_origin, 0, sizeof(parent_origin));
1187
1188 /* The first pass looks for unrenamed path to optimize for
1189 * common cases, then we look for renames in the second pass.
1190 */
1191 for (pass = 0; pass < 2; pass++) {
1192 struct origin *(*find)(struct scoreboard *,
1193 struct commit *, struct origin *);
1194 find = pass ? find_rename : find_origin;
1195
1196 for (i = 0, parent = commit->parents;
1197 i < MAXPARENT && parent;
1198 parent = parent->next, i++) {
1199 struct commit *p = parent->item;
1200 int j, same;
1201
1202 if (parent_origin[i])
1203 continue;
1204 if (parse_commit(p))
1205 continue;
1206 porigin = find(sb, p, origin);
1207 if (!porigin)
1208 continue;
1209 if (!hashcmp(porigin->blob_sha1, origin->blob_sha1)) {
1210 pass_whole_blame(sb, origin, porigin);
1211 origin_decref(porigin);
1212 goto finish;
1213 }
1214 for (j = same = 0; j < i; j++)
1215 if (parent_origin[j] &&
1216 !hashcmp(parent_origin[j]->blob_sha1,
1217 porigin->blob_sha1)) {
1218 same = 1;
1219 break;
1220 }
1221 if (!same)
1222 parent_origin[i] = porigin;
1223 else
1224 origin_decref(porigin);
1225 }
1226 }
1227
1228 num_commits++;
1229 for (i = 0, parent = commit->parents;
1230 i < MAXPARENT && parent;
1231 parent = parent->next, i++) {
1232 struct origin *porigin = parent_origin[i];
1233 if (!porigin)
1234 continue;
1235 if (pass_blame_to_parent(sb, origin, porigin))
1236 goto finish;
1237 }
1238
1239 /*
1240 * Optionally find moves in parents' files.
1241 */
1242 if (opt & PICKAXE_BLAME_MOVE)
1243 for (i = 0, parent = commit->parents;
1244 i < MAXPARENT && parent;
1245 parent = parent->next, i++) {
1246 struct origin *porigin = parent_origin[i];
1247 if (!porigin)
1248 continue;
1249 if (find_move_in_parent(sb, origin, porigin))
1250 goto finish;
1251 }
1252
1253 /*
1254 * Optionally find copies from parents' files.
1255 */
1256 if (opt & PICKAXE_BLAME_COPY)
1257 for (i = 0, parent = commit->parents;
1258 i < MAXPARENT && parent;
1259 parent = parent->next, i++) {
1260 struct origin *porigin = parent_origin[i];
1261 if (find_copy_in_parent(sb, origin, parent->item,
1262 porigin, opt))
1263 goto finish;
1264 }
1265
1266 finish:
1267 for (i = 0; i < MAXPARENT; i++)
1268 origin_decref(parent_origin[i]);
1269 }
1270
1271 /*
1272 * Information on commits, used for output.
1273 */
1274 struct commit_info
1275 {
1276 const char *author;
1277 const char *author_mail;
1278 unsigned long author_time;
1279 const char *author_tz;
1280
1281 /* filled only when asked for details */
1282 const char *committer;
1283 const char *committer_mail;
1284 unsigned long committer_time;
1285 const char *committer_tz;
1286
1287 const char *summary;
1288 };
1289
1290 /*
1291 * Parse author/committer line in the commit object buffer
1292 */
1293 static void get_ac_line(const char *inbuf, const char *what,
1294 int bufsz, char *person, const char **mail,
1295 unsigned long *time, const char **tz)
1296 {
1297 int len;
1298 char *tmp, *endp;
1299
1300 tmp = strstr(inbuf, what);
1301 if (!tmp)
1302 goto error_out;
1303 tmp += strlen(what);
1304 endp = strchr(tmp, '\n');
1305 if (!endp)
1306 len = strlen(tmp);
1307 else
1308 len = endp - tmp;
1309 if (bufsz <= len) {
1310 error_out:
1311 /* Ugh */
1312 *mail = *tz = "(unknown)";
1313 *time = 0;
1314 return;
1315 }
1316 memcpy(person, tmp, len);
1317
1318 tmp = person;
1319 tmp += len;
1320 *tmp = 0;
1321 while (*tmp != ' ')
1322 tmp--;
1323 *tz = tmp+1;
1324
1325 *tmp = 0;
1326 while (*tmp != ' ')
1327 tmp--;
1328 *time = strtoul(tmp, NULL, 10);
1329
1330 *tmp = 0;
1331 while (*tmp != ' ')
1332 tmp--;
1333 *mail = tmp + 1;
1334 *tmp = 0;
1335 }
1336
1337 static void get_commit_info(struct commit *commit,
1338 struct commit_info *ret,
1339 int detailed)
1340 {
1341 int len;
1342 char *tmp, *endp;
1343 static char author_buf[1024];
1344 static char committer_buf[1024];
1345 static char summary_buf[1024];
1346
1347 /*
1348 * We've operated without save_commit_buffer, so
1349 * we now need to populate them for output.
1350 */
1351 if (!commit->buffer) {
1352 enum object_type type;
1353 unsigned long size;
1354 commit->buffer =
1355 read_sha1_file(commit->object.sha1, &type, &size);
1356 }
1357 ret->author = author_buf;
1358 get_ac_line(commit->buffer, "\nauthor ",
1359 sizeof(author_buf), author_buf, &ret->author_mail,
1360 &ret->author_time, &ret->author_tz);
1361
1362 if (!detailed)
1363 return;
1364
1365 ret->committer = committer_buf;
1366 get_ac_line(commit->buffer, "\ncommitter ",
1367 sizeof(committer_buf), committer_buf, &ret->committer_mail,
1368 &ret->committer_time, &ret->committer_tz);
1369
1370 ret->summary = summary_buf;
1371 tmp = strstr(commit->buffer, "\n\n");
1372 if (!tmp) {
1373 error_out:
1374 sprintf(summary_buf, "(%s)", sha1_to_hex(commit->object.sha1));
1375 return;
1376 }
1377 tmp += 2;
1378 endp = strchr(tmp, '\n');
1379 if (!endp)
1380 endp = tmp + strlen(tmp);
1381 len = endp - tmp;
1382 if (len >= sizeof(summary_buf) || len == 0)
1383 goto error_out;
1384 memcpy(summary_buf, tmp, len);
1385 summary_buf[len] = 0;
1386 }
1387
1388 /*
1389 * To allow LF and other nonportable characters in pathnames,
1390 * they are c-style quoted as needed.
1391 */
1392 static void write_filename_info(const char *path)
1393 {
1394 printf("filename ");
1395 write_name_quoted(NULL, 0, path, 1, stdout);
1396 putchar('\n');
1397 }
1398
1399 /*
1400 * The blame_entry is found to be guilty for the range. Mark it
1401 * as such, and show it in incremental output.
1402 */
1403 static void found_guilty_entry(struct blame_entry *ent)
1404 {
1405 if (ent->guilty)
1406 return;
1407 ent->guilty = 1;
1408 if (incremental) {
1409 struct origin *suspect = ent->suspect;
1410
1411 printf("%s %d %d %d\n",
1412 sha1_to_hex(suspect->commit->object.sha1),
1413 ent->s_lno + 1, ent->lno + 1, ent->num_lines);
1414 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1415 struct commit_info ci;
1416 suspect->commit->object.flags |= METAINFO_SHOWN;
1417 get_commit_info(suspect->commit, &ci, 1);
1418 printf("author %s\n", ci.author);
1419 printf("author-mail %s\n", ci.author_mail);
1420 printf("author-time %lu\n", ci.author_time);
1421 printf("author-tz %s\n", ci.author_tz);
1422 printf("committer %s\n", ci.committer);
1423 printf("committer-mail %s\n", ci.committer_mail);
1424 printf("committer-time %lu\n", ci.committer_time);
1425 printf("committer-tz %s\n", ci.committer_tz);
1426 printf("summary %s\n", ci.summary);
1427 if (suspect->commit->object.flags & UNINTERESTING)
1428 printf("boundary\n");
1429 }
1430 write_filename_info(suspect->path);
1431 }
1432 }
1433
1434 /*
1435 * The main loop -- while the scoreboard has lines whose true origin
1436 * is still unknown, pick one blame_entry, and allow its current
1437 * suspect to pass blames to its parents.
1438 */
1439 static void assign_blame(struct scoreboard *sb, struct rev_info *revs, int opt)
1440 {
1441 while (1) {
1442 struct blame_entry *ent;
1443 struct commit *commit;
1444 struct origin *suspect = NULL;
1445
1446 /* find one suspect to break down */
1447 for (ent = sb->ent; !suspect && ent; ent = ent->next)
1448 if (!ent->guilty)
1449 suspect = ent->suspect;
1450 if (!suspect)
1451 return; /* all done */
1452
1453 /*
1454 * We will use this suspect later in the loop,
1455 * so hold onto it in the meantime.
1456 */
1457 origin_incref(suspect);
1458 commit = suspect->commit;
1459 if (!commit->object.parsed)
1460 parse_commit(commit);
1461 if (!(commit->object.flags & UNINTERESTING) &&
1462 !(revs->max_age != -1 && commit->date < revs->max_age))
1463 pass_blame(sb, suspect, opt);
1464 else {
1465 commit->object.flags |= UNINTERESTING;
1466 if (commit->object.parsed)
1467 mark_parents_uninteresting(commit);
1468 }
1469 /* treat root commit as boundary */
1470 if (!commit->parents && !show_root)
1471 commit->object.flags |= UNINTERESTING;
1472
1473 /* Take responsibility for the remaining entries */
1474 for (ent = sb->ent; ent; ent = ent->next)
1475 if (same_suspect(ent->suspect, suspect))
1476 found_guilty_entry(ent);
1477 origin_decref(suspect);
1478
1479 if (DEBUG) /* sanity */
1480 sanity_check_refcnt(sb);
1481 }
1482 }
1483
1484 static const char *format_time(unsigned long time, const char *tz_str,
1485 int show_raw_time)
1486 {
1487 static char time_buf[128];
1488 time_t t = time;
1489 int minutes, tz;
1490 struct tm *tm;
1491
1492 if (show_raw_time) {
1493 sprintf(time_buf, "%lu %s", time, tz_str);
1494 return time_buf;
1495 }
1496
1497 tz = atoi(tz_str);
1498 minutes = tz < 0 ? -tz : tz;
1499 minutes = (minutes / 100)*60 + (minutes % 100);
1500 minutes = tz < 0 ? -minutes : minutes;
1501 t = time + minutes * 60;
1502 tm = gmtime(&t);
1503
1504 strftime(time_buf, sizeof(time_buf), "%Y-%m-%d %H:%M:%S ", tm);
1505 strcat(time_buf, tz_str);
1506 return time_buf;
1507 }
1508
1509 #define OUTPUT_ANNOTATE_COMPAT 001
1510 #define OUTPUT_LONG_OBJECT_NAME 002
1511 #define OUTPUT_RAW_TIMESTAMP 004
1512 #define OUTPUT_PORCELAIN 010
1513 #define OUTPUT_SHOW_NAME 020
1514 #define OUTPUT_SHOW_NUMBER 040
1515 #define OUTPUT_SHOW_SCORE 0100
1516
1517 static void emit_porcelain(struct scoreboard *sb, struct blame_entry *ent)
1518 {
1519 int cnt;
1520 const char *cp;
1521 struct origin *suspect = ent->suspect;
1522 char hex[41];
1523
1524 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1525 printf("%s%c%d %d %d\n",
1526 hex,
1527 ent->guilty ? ' ' : '*', // purely for debugging
1528 ent->s_lno + 1,
1529 ent->lno + 1,
1530 ent->num_lines);
1531 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1532 struct commit_info ci;
1533 suspect->commit->object.flags |= METAINFO_SHOWN;
1534 get_commit_info(suspect->commit, &ci, 1);
1535 printf("author %s\n", ci.author);
1536 printf("author-mail %s\n", ci.author_mail);
1537 printf("author-time %lu\n", ci.author_time);
1538 printf("author-tz %s\n", ci.author_tz);
1539 printf("committer %s\n", ci.committer);
1540 printf("committer-mail %s\n", ci.committer_mail);
1541 printf("committer-time %lu\n", ci.committer_time);
1542 printf("committer-tz %s\n", ci.committer_tz);
1543 write_filename_info(suspect->path);
1544 printf("summary %s\n", ci.summary);
1545 if (suspect->commit->object.flags & UNINTERESTING)
1546 printf("boundary\n");
1547 }
1548 else if (suspect->commit->object.flags & MORE_THAN_ONE_PATH)
1549 write_filename_info(suspect->path);
1550
1551 cp = nth_line(sb, ent->lno);
1552 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1553 char ch;
1554 if (cnt)
1555 printf("%s %d %d\n", hex,
1556 ent->s_lno + 1 + cnt,
1557 ent->lno + 1 + cnt);
1558 putchar('\t');
1559 do {
1560 ch = *cp++;
1561 putchar(ch);
1562 } while (ch != '\n' &&
1563 cp < sb->final_buf + sb->final_buf_size);
1564 }
1565 }
1566
1567 static void emit_other(struct scoreboard *sb, struct blame_entry *ent, int opt)
1568 {
1569 int cnt;
1570 const char *cp;
1571 struct origin *suspect = ent->suspect;
1572 struct commit_info ci;
1573 char hex[41];
1574 int show_raw_time = !!(opt & OUTPUT_RAW_TIMESTAMP);
1575
1576 get_commit_info(suspect->commit, &ci, 1);
1577 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1578
1579 cp = nth_line(sb, ent->lno);
1580 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1581 char ch;
1582 int length = (opt & OUTPUT_LONG_OBJECT_NAME) ? 40 : 8;
1583
1584 if (suspect->commit->object.flags & UNINTERESTING) {
1585 if (blank_boundary)
1586 memset(hex, ' ', length);
1587 else if (!cmd_is_annotate) {
1588 length--;
1589 putchar('^');
1590 }
1591 }
1592
1593 printf("%.*s", length, hex);
1594 if (opt & OUTPUT_ANNOTATE_COMPAT)
1595 printf("\t(%10s\t%10s\t%d)", ci.author,
1596 format_time(ci.author_time, ci.author_tz,
1597 show_raw_time),
1598 ent->lno + 1 + cnt);
1599 else {
1600 if (opt & OUTPUT_SHOW_SCORE)
1601 printf(" %*d %02d",
1602 max_score_digits, ent->score,
1603 ent->suspect->refcnt);
1604 if (opt & OUTPUT_SHOW_NAME)
1605 printf(" %-*.*s", longest_file, longest_file,
1606 suspect->path);
1607 if (opt & OUTPUT_SHOW_NUMBER)
1608 printf(" %*d", max_orig_digits,
1609 ent->s_lno + 1 + cnt);
1610 printf(" (%-*.*s %10s %*d) ",
1611 longest_author, longest_author, ci.author,
1612 format_time(ci.author_time, ci.author_tz,
1613 show_raw_time),
1614 max_digits, ent->lno + 1 + cnt);
1615 }
1616 do {
1617 ch = *cp++;
1618 putchar(ch);
1619 } while (ch != '\n' &&
1620 cp < sb->final_buf + sb->final_buf_size);
1621 }
1622 }
1623
1624 static void output(struct scoreboard *sb, int option)
1625 {
1626 struct blame_entry *ent;
1627
1628 if (option & OUTPUT_PORCELAIN) {
1629 for (ent = sb->ent; ent; ent = ent->next) {
1630 struct blame_entry *oth;
1631 struct origin *suspect = ent->suspect;
1632 struct commit *commit = suspect->commit;
1633 if (commit->object.flags & MORE_THAN_ONE_PATH)
1634 continue;
1635 for (oth = ent->next; oth; oth = oth->next) {
1636 if ((oth->suspect->commit != commit) ||
1637 !strcmp(oth->suspect->path, suspect->path))
1638 continue;
1639 commit->object.flags |= MORE_THAN_ONE_PATH;
1640 break;
1641 }
1642 }
1643 }
1644
1645 for (ent = sb->ent; ent; ent = ent->next) {
1646 if (option & OUTPUT_PORCELAIN)
1647 emit_porcelain(sb, ent);
1648 else {
1649 emit_other(sb, ent, option);
1650 }
1651 }
1652 }
1653
1654 /*
1655 * To allow quick access to the contents of nth line in the
1656 * final image, prepare an index in the scoreboard.
1657 */
1658 static int prepare_lines(struct scoreboard *sb)
1659 {
1660 const char *buf = sb->final_buf;
1661 unsigned long len = sb->final_buf_size;
1662 int num = 0, incomplete = 0, bol = 1;
1663
1664 if (len && buf[len-1] != '\n')
1665 incomplete++; /* incomplete line at the end */
1666 while (len--) {
1667 if (bol) {
1668 sb->lineno = xrealloc(sb->lineno,
1669 sizeof(int* ) * (num + 1));
1670 sb->lineno[num] = buf - sb->final_buf;
1671 bol = 0;
1672 }
1673 if (*buf++ == '\n') {
1674 num++;
1675 bol = 1;
1676 }
1677 }
1678 sb->lineno = xrealloc(sb->lineno,
1679 sizeof(int* ) * (num + incomplete + 1));
1680 sb->lineno[num + incomplete] = buf - sb->final_buf;
1681 sb->num_lines = num + incomplete;
1682 return sb->num_lines;
1683 }
1684
1685 /*
1686 * Add phony grafts for use with -S; this is primarily to
1687 * support git-cvsserver that wants to give a linear history
1688 * to its clients.
1689 */
1690 static int read_ancestry(const char *graft_file)
1691 {
1692 FILE *fp = fopen(graft_file, "r");
1693 char buf[1024];
1694 if (!fp)
1695 return -1;
1696 while (fgets(buf, sizeof(buf), fp)) {
1697 /* The format is just "Commit Parent1 Parent2 ...\n" */
1698 int len = strlen(buf);
1699 struct commit_graft *graft = read_graft_line(buf, len);
1700 if (graft)
1701 register_commit_graft(graft, 0);
1702 }
1703 fclose(fp);
1704 return 0;
1705 }
1706
1707 /*
1708 * How many columns do we need to show line numbers in decimal?
1709 */
1710 static int lineno_width(int lines)
1711 {
1712 int i, width;
1713
1714 for (width = 1, i = 10; i <= lines + 1; width++)
1715 i *= 10;
1716 return width;
1717 }
1718
1719 /*
1720 * How many columns do we need to show line numbers, authors,
1721 * and filenames?
1722 */
1723 static void find_alignment(struct scoreboard *sb, int *option)
1724 {
1725 int longest_src_lines = 0;
1726 int longest_dst_lines = 0;
1727 unsigned largest_score = 0;
1728 struct blame_entry *e;
1729
1730 for (e = sb->ent; e; e = e->next) {
1731 struct origin *suspect = e->suspect;
1732 struct commit_info ci;
1733 int num;
1734
1735 if (strcmp(suspect->path, sb->path))
1736 *option |= OUTPUT_SHOW_NAME;
1737 num = strlen(suspect->path);
1738 if (longest_file < num)
1739 longest_file = num;
1740 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1741 suspect->commit->object.flags |= METAINFO_SHOWN;
1742 get_commit_info(suspect->commit, &ci, 1);
1743 num = strlen(ci.author);
1744 if (longest_author < num)
1745 longest_author = num;
1746 }
1747 num = e->s_lno + e->num_lines;
1748 if (longest_src_lines < num)
1749 longest_src_lines = num;
1750 num = e->lno + e->num_lines;
1751 if (longest_dst_lines < num)
1752 longest_dst_lines = num;
1753 if (largest_score < ent_score(sb, e))
1754 largest_score = ent_score(sb, e);
1755 }
1756 max_orig_digits = lineno_width(longest_src_lines);
1757 max_digits = lineno_width(longest_dst_lines);
1758 max_score_digits = lineno_width(largest_score);
1759 }
1760
1761 /*
1762 * For debugging -- origin is refcounted, and this asserts that
1763 * we do not underflow.
1764 */
1765 static void sanity_check_refcnt(struct scoreboard *sb)
1766 {
1767 int baa = 0;
1768 struct blame_entry *ent;
1769
1770 for (ent = sb->ent; ent; ent = ent->next) {
1771 /* Nobody should have zero or negative refcnt */
1772 if (ent->suspect->refcnt <= 0) {
1773 fprintf(stderr, "%s in %s has negative refcnt %d\n",
1774 ent->suspect->path,
1775 sha1_to_hex(ent->suspect->commit->object.sha1),
1776 ent->suspect->refcnt);
1777 baa = 1;
1778 }
1779 }
1780 for (ent = sb->ent; ent; ent = ent->next) {
1781 /* Mark the ones that haven't been checked */
1782 if (0 < ent->suspect->refcnt)
1783 ent->suspect->refcnt = -ent->suspect->refcnt;
1784 }
1785 for (ent = sb->ent; ent; ent = ent->next) {
1786 /*
1787 * ... then pick each and see if they have the the
1788 * correct refcnt.
1789 */
1790 int found;
1791 struct blame_entry *e;
1792 struct origin *suspect = ent->suspect;
1793
1794 if (0 < suspect->refcnt)
1795 continue;
1796 suspect->refcnt = -suspect->refcnt; /* Unmark */
1797 for (found = 0, e = sb->ent; e; e = e->next) {
1798 if (e->suspect != suspect)
1799 continue;
1800 found++;
1801 }
1802 if (suspect->refcnt != found) {
1803 fprintf(stderr, "%s in %s has refcnt %d, not %d\n",
1804 ent->suspect->path,
1805 sha1_to_hex(ent->suspect->commit->object.sha1),
1806 ent->suspect->refcnt, found);
1807 baa = 2;
1808 }
1809 }
1810 if (baa) {
1811 int opt = 0160;
1812 find_alignment(sb, &opt);
1813 output(sb, opt);
1814 die("Baa %d!", baa);
1815 }
1816 }
1817
1818 /*
1819 * Used for the command line parsing; check if the path exists
1820 * in the working tree.
1821 */
1822 static int has_path_in_work_tree(const char *path)
1823 {
1824 struct stat st;
1825 return !lstat(path, &st);
1826 }
1827
1828 static unsigned parse_score(const char *arg)
1829 {
1830 char *end;
1831 unsigned long score = strtoul(arg, &end, 10);
1832 if (*end)
1833 return 0;
1834 return score;
1835 }
1836
1837 static const char *add_prefix(const char *prefix, const char *path)
1838 {
1839 if (!prefix || !prefix[0])
1840 return path;
1841 return prefix_path(prefix, strlen(prefix), path);
1842 }
1843
1844 /*
1845 * Parsing of (comma separated) one item in the -L option
1846 */
1847 static const char *parse_loc(const char *spec,
1848 struct scoreboard *sb, long lno,
1849 long begin, long *ret)
1850 {
1851 char *term;
1852 const char *line;
1853 long num;
1854 int reg_error;
1855 regex_t regexp;
1856 regmatch_t match[1];
1857
1858 /* Allow "-L <something>,+20" to mean starting at <something>
1859 * for 20 lines, or "-L <something>,-5" for 5 lines ending at
1860 * <something>.
1861 */
1862 if (1 < begin && (spec[0] == '+' || spec[0] == '-')) {
1863 num = strtol(spec + 1, &term, 10);
1864 if (term != spec + 1) {
1865 if (spec[0] == '-')
1866 num = 0 - num;
1867 if (0 < num)
1868 *ret = begin + num - 2;
1869 else if (!num)
1870 *ret = begin;
1871 else
1872 *ret = begin + num;
1873 return term;
1874 }
1875 return spec;
1876 }
1877 num = strtol(spec, &term, 10);
1878 if (term != spec) {
1879 *ret = num;
1880 return term;
1881 }
1882 if (spec[0] != '/')
1883 return spec;
1884
1885 /* it could be a regexp of form /.../ */
1886 for (term = (char*) spec + 1; *term && *term != '/'; term++) {
1887 if (*term == '\\')
1888 term++;
1889 }
1890 if (*term != '/')
1891 return spec;
1892
1893 /* try [spec+1 .. term-1] as regexp */
1894 *term = 0;
1895 begin--; /* input is in human terms */
1896 line = nth_line(sb, begin);
1897
1898 if (!(reg_error = regcomp(&regexp, spec + 1, REG_NEWLINE)) &&
1899 !(reg_error = regexec(&regexp, line, 1, match, 0))) {
1900 const char *cp = line + match[0].rm_so;
1901 const char *nline;
1902
1903 while (begin++ < lno) {
1904 nline = nth_line(sb, begin);
1905 if (line <= cp && cp < nline)
1906 break;
1907 line = nline;
1908 }
1909 *ret = begin;
1910 regfree(&regexp);
1911 *term++ = '/';
1912 return term;
1913 }
1914 else {
1915 char errbuf[1024];
1916 regerror(reg_error, &regexp, errbuf, 1024);
1917 die("-L parameter '%s': %s", spec + 1, errbuf);
1918 }
1919 }
1920
1921 /*
1922 * Parsing of -L option
1923 */
1924 static void prepare_blame_range(struct scoreboard *sb,
1925 const char *bottomtop,
1926 long lno,
1927 long *bottom, long *top)
1928 {
1929 const char *term;
1930
1931 term = parse_loc(bottomtop, sb, lno, 1, bottom);
1932 if (*term == ',') {
1933 term = parse_loc(term + 1, sb, lno, *bottom + 1, top);
1934 if (*term)
1935 usage(blame_usage);
1936 }
1937 if (*term)
1938 usage(blame_usage);
1939 }
1940
1941 static int git_blame_config(const char *var, const char *value)
1942 {
1943 if (!strcmp(var, "blame.showroot")) {
1944 show_root = git_config_bool(var, value);
1945 return 0;
1946 }
1947 if (!strcmp(var, "blame.blankboundary")) {
1948 blank_boundary = git_config_bool(var, value);
1949 return 0;
1950 }
1951 return git_default_config(var, value);
1952 }
1953
1954 static struct commit *fake_working_tree_commit(const char *path, const char *contents_from)
1955 {
1956 struct commit *commit;
1957 struct origin *origin;
1958 unsigned char head_sha1[20];
1959 char *buf;
1960 const char *ident;
1961 int fd;
1962 time_t now;
1963 unsigned long fin_size;
1964 int size, len;
1965 struct cache_entry *ce;
1966 unsigned mode;
1967
1968 if (get_sha1("HEAD", head_sha1))
1969 die("No such ref: HEAD");
1970
1971 time(&now);
1972 commit = xcalloc(1, sizeof(*commit));
1973 commit->parents = xcalloc(1, sizeof(*commit->parents));
1974 commit->parents->item = lookup_commit_reference(head_sha1);
1975 commit->object.parsed = 1;
1976 commit->date = now;
1977 commit->object.type = OBJ_COMMIT;
1978
1979 origin = make_origin(commit, path);
1980
1981 if (!contents_from || strcmp("-", contents_from)) {
1982 struct stat st;
1983 const char *read_from;
1984
1985 if (contents_from) {
1986 if (stat(contents_from, &st) < 0)
1987 die("Cannot stat %s", contents_from);
1988 read_from = contents_from;
1989 }
1990 else {
1991 if (lstat(path, &st) < 0)
1992 die("Cannot lstat %s", path);
1993 read_from = path;
1994 }
1995 fin_size = xsize_t(st.st_size);
1996 buf = xmalloc(fin_size+1);
1997 mode = canon_mode(st.st_mode);
1998 switch (st.st_mode & S_IFMT) {
1999 case S_IFREG:
2000 fd = open(read_from, O_RDONLY);
2001 if (fd < 0)
2002 die("cannot open %s", read_from);
2003 if (read_in_full(fd, buf, fin_size) != fin_size)
2004 die("cannot read %s", read_from);
2005 break;
2006 case S_IFLNK:
2007 if (readlink(read_from, buf, fin_size+1) != fin_size)
2008 die("cannot readlink %s", read_from);
2009 break;
2010 default:
2011 die("unsupported file type %s", read_from);
2012 }
2013 }
2014 else {
2015 /* Reading from stdin */
2016 contents_from = "standard input";
2017 buf = NULL;
2018 fin_size = 0;
2019 mode = 0;
2020 while (1) {
2021 ssize_t cnt = 8192;
2022 buf = xrealloc(buf, fin_size + cnt);
2023 cnt = xread(0, buf + fin_size, cnt);
2024 if (cnt < 0)
2025 die("read error %s from stdin",
2026 strerror(errno));
2027 if (!cnt)
2028 break;
2029 fin_size += cnt;
2030 }
2031 buf = xrealloc(buf, fin_size + 1);
2032 }
2033 buf[fin_size] = 0;
2034 origin->file.ptr = buf;
2035 origin->file.size = fin_size;
2036 pretend_sha1_file(buf, fin_size, OBJ_BLOB, origin->blob_sha1);
2037 commit->util = origin;
2038
2039 /*
2040 * Read the current index, replace the path entry with
2041 * origin->blob_sha1 without mucking with its mode or type
2042 * bits; we are not going to write this index out -- we just
2043 * want to run "diff-index --cached".
2044 */
2045 discard_cache();
2046 read_cache();
2047
2048 len = strlen(path);
2049 if (!mode) {
2050 int pos = cache_name_pos(path, len);
2051 if (0 <= pos)
2052 mode = ntohl(active_cache[pos]->ce_mode);
2053 else
2054 /* Let's not bother reading from HEAD tree */
2055 mode = S_IFREG | 0644;
2056 }
2057 size = cache_entry_size(len);
2058 ce = xcalloc(1, size);
2059 hashcpy(ce->sha1, origin->blob_sha1);
2060 memcpy(ce->name, path, len);
2061 ce->ce_flags = create_ce_flags(len, 0);
2062 ce->ce_mode = create_ce_mode(mode);
2063 add_cache_entry(ce, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
2064
2065 /*
2066 * We are not going to write this out, so this does not matter
2067 * right now, but someday we might optimize diff-index --cached
2068 * with cache-tree information.
2069 */
2070 cache_tree_invalidate_path(active_cache_tree, path);
2071
2072 commit->buffer = xmalloc(400);
2073 ident = fmt_ident("Not Committed Yet", "not.committed.yet", NULL, 0);
2074 snprintf(commit->buffer, 400,
2075 "tree 0000000000000000000000000000000000000000\n"
2076 "parent %s\n"
2077 "author %s\n"
2078 "committer %s\n\n"
2079 "Version of %s from %s\n",
2080 sha1_to_hex(head_sha1),
2081 ident, ident, path, contents_from ? contents_from : path);
2082 return commit;
2083 }
2084
2085 int cmd_blame(int argc, const char **argv, const char *prefix)
2086 {
2087 struct rev_info revs;
2088 const char *path;
2089 struct scoreboard sb;
2090 struct origin *o;
2091 struct blame_entry *ent;
2092 int i, seen_dashdash, unk, opt;
2093 long bottom, top, lno;
2094 int output_option = 0;
2095 int show_stats = 0;
2096 const char *revs_file = NULL;
2097 const char *final_commit_name = NULL;
2098 enum object_type type;
2099 const char *bottomtop = NULL;
2100 const char *contents_from = NULL;
2101
2102 cmd_is_annotate = !strcmp(argv[0], "annotate");
2103
2104 git_config(git_blame_config);
2105 save_commit_buffer = 0;
2106
2107 opt = 0;
2108 seen_dashdash = 0;
2109 for (unk = i = 1; i < argc; i++) {
2110 const char *arg = argv[i];
2111 if (*arg != '-')
2112 break;
2113 else if (!strcmp("-b", arg))
2114 blank_boundary = 1;
2115 else if (!strcmp("--root", arg))
2116 show_root = 1;
2117 else if (!strcmp(arg, "--show-stats"))
2118 show_stats = 1;
2119 else if (!strcmp("-c", arg))
2120 output_option |= OUTPUT_ANNOTATE_COMPAT;
2121 else if (!strcmp("-t", arg))
2122 output_option |= OUTPUT_RAW_TIMESTAMP;
2123 else if (!strcmp("-l", arg))
2124 output_option |= OUTPUT_LONG_OBJECT_NAME;
2125 else if (!strcmp("-S", arg) && ++i < argc)
2126 revs_file = argv[i];
2127 else if (!prefixcmp(arg, "-M")) {
2128 opt |= PICKAXE_BLAME_MOVE;
2129 blame_move_score = parse_score(arg+2);
2130 }
2131 else if (!prefixcmp(arg, "-C")) {
2132 /*
2133 * -C enables copy from removed files;
2134 * -C -C enables copy from existing files, but only
2135 * when blaming a new file;
2136 * -C -C -C enables copy from existing files for
2137 * everybody
2138 */
2139 if (opt & PICKAXE_BLAME_COPY_HARDER)
2140 opt |= PICKAXE_BLAME_COPY_HARDEST;
2141 if (opt & PICKAXE_BLAME_COPY)
2142 opt |= PICKAXE_BLAME_COPY_HARDER;
2143 opt |= PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE;
2144 blame_copy_score = parse_score(arg+2);
2145 }
2146 else if (!prefixcmp(arg, "-L")) {
2147 if (!arg[2]) {
2148 if (++i >= argc)
2149 usage(blame_usage);
2150 arg = argv[i];
2151 }
2152 else
2153 arg += 2;
2154 if (bottomtop)
2155 die("More than one '-L n,m' option given");
2156 bottomtop = arg;
2157 }
2158 else if (!strcmp("--contents", arg)) {
2159 if (++i >= argc)
2160 usage(blame_usage);
2161 contents_from = argv[i];
2162 }
2163 else if (!strcmp("--incremental", arg))
2164 incremental = 1;
2165 else if (!strcmp("--score-debug", arg))
2166 output_option |= OUTPUT_SHOW_SCORE;
2167 else if (!strcmp("-f", arg) ||
2168 !strcmp("--show-name", arg))
2169 output_option |= OUTPUT_SHOW_NAME;
2170 else if (!strcmp("-n", arg) ||
2171 !strcmp("--show-number", arg))
2172 output_option |= OUTPUT_SHOW_NUMBER;
2173 else if (!strcmp("-p", arg) ||
2174 !strcmp("--porcelain", arg))
2175 output_option |= OUTPUT_PORCELAIN;
2176 else if (!strcmp("--", arg)) {
2177 seen_dashdash = 1;
2178 i++;
2179 break;
2180 }
2181 else
2182 argv[unk++] = arg;
2183 }
2184
2185 if (!incremental)
2186 setup_pager();
2187
2188 if (!blame_move_score)
2189 blame_move_score = BLAME_DEFAULT_MOVE_SCORE;
2190 if (!blame_copy_score)
2191 blame_copy_score = BLAME_DEFAULT_COPY_SCORE;
2192
2193 /*
2194 * We have collected options unknown to us in argv[1..unk]
2195 * which are to be passed to revision machinery if we are
2196 * going to do the "bottom" processing.
2197 *
2198 * The remaining are:
2199 *
2200 * (1) if seen_dashdash, its either
2201 * "-options -- <path>" or
2202 * "-options -- <path> <rev>".
2203 * but the latter is allowed only if there is no
2204 * options that we passed to revision machinery.
2205 *
2206 * (2) otherwise, we may have "--" somewhere later and
2207 * might be looking at the first one of multiple 'rev'
2208 * parameters (e.g. " master ^next ^maint -- path").
2209 * See if there is a dashdash first, and give the
2210 * arguments before that to revision machinery.
2211 * After that there must be one 'path'.
2212 *
2213 * (3) otherwise, its one of the three:
2214 * "-options <path> <rev>"
2215 * "-options <rev> <path>"
2216 * "-options <path>"
2217 * but again the first one is allowed only if
2218 * there is no options that we passed to revision
2219 * machinery.
2220 */
2221
2222 if (seen_dashdash) {
2223 /* (1) */
2224 if (argc <= i)
2225 usage(blame_usage);
2226 path = add_prefix(prefix, argv[i]);
2227 if (i + 1 == argc - 1) {
2228 if (unk != 1)
2229 usage(blame_usage);
2230 argv[unk++] = argv[i + 1];
2231 }
2232 else if (i + 1 != argc)
2233 /* garbage at end */
2234 usage(blame_usage);
2235 }
2236 else {
2237 int j;
2238 for (j = i; !seen_dashdash && j < argc; j++)
2239 if (!strcmp(argv[j], "--"))
2240 seen_dashdash = j;
2241 if (seen_dashdash) {
2242 /* (2) */
2243 if (seen_dashdash + 1 != argc - 1)
2244 usage(blame_usage);
2245 path = add_prefix(prefix, argv[seen_dashdash + 1]);
2246 for (j = i; j < seen_dashdash; j++)
2247 argv[unk++] = argv[j];
2248 }
2249 else {
2250 /* (3) */
2251 if (argc <= i)
2252 usage(blame_usage);
2253 path = add_prefix(prefix, argv[i]);
2254 if (i + 1 == argc - 1) {
2255 final_commit_name = argv[i + 1];
2256
2257 /* if (unk == 1) we could be getting
2258 * old-style
2259 */
2260 if (unk == 1 && !has_path_in_work_tree(path)) {
2261 path = add_prefix(prefix, argv[i + 1]);
2262 final_commit_name = argv[i];
2263 }
2264 }
2265 else if (i != argc - 1)
2266 usage(blame_usage); /* garbage at end */
2267
2268 if (!has_path_in_work_tree(path))
2269 die("cannot stat path %s: %s",
2270 path, strerror(errno));
2271 }
2272 }
2273
2274 if (final_commit_name)
2275 argv[unk++] = final_commit_name;
2276
2277 /*
2278 * Now we got rev and path. We do not want the path pruning
2279 * but we may want "bottom" processing.
2280 */
2281 argv[unk++] = "--"; /* terminate the rev name */
2282 argv[unk] = NULL;
2283
2284 init_revisions(&revs, NULL);
2285 setup_revisions(unk, argv, &revs, NULL);
2286 memset(&sb, 0, sizeof(sb));
2287
2288 /*
2289 * There must be one and only one positive commit in the
2290 * revs->pending array.
2291 */
2292 for (i = 0; i < revs.pending.nr; i++) {
2293 struct object *obj = revs.pending.objects[i].item;
2294 if (obj->flags & UNINTERESTING)
2295 continue;
2296 while (obj->type == OBJ_TAG)
2297 obj = deref_tag(obj, NULL, 0);
2298 if (obj->type != OBJ_COMMIT)
2299 die("Non commit %s?",
2300 revs.pending.objects[i].name);
2301 if (sb.final)
2302 die("More than one commit to dig from %s and %s?",
2303 revs.pending.objects[i].name,
2304 final_commit_name);
2305 sb.final = (struct commit *) obj;
2306 final_commit_name = revs.pending.objects[i].name;
2307 }
2308
2309 if (!sb.final) {
2310 /*
2311 * "--not A B -- path" without anything positive;
2312 * do not default to HEAD, but use the working tree
2313 * or "--contents".
2314 */
2315 sb.final = fake_working_tree_commit(path, contents_from);
2316 add_pending_object(&revs, &(sb.final->object), ":");
2317 }
2318 else if (contents_from)
2319 die("Cannot use --contents with final commit object name");
2320
2321 /*
2322 * If we have bottom, this will mark the ancestors of the
2323 * bottom commits we would reach while traversing as
2324 * uninteresting.
2325 */
2326 prepare_revision_walk(&revs);
2327
2328 if (is_null_sha1(sb.final->object.sha1)) {
2329 char *buf;
2330 o = sb.final->util;
2331 buf = xmalloc(o->file.size + 1);
2332 memcpy(buf, o->file.ptr, o->file.size + 1);
2333 sb.final_buf = buf;
2334 sb.final_buf_size = o->file.size;
2335 }
2336 else {
2337 o = get_origin(&sb, sb.final, path);
2338 if (fill_blob_sha1(o))
2339 die("no such path %s in %s", path, final_commit_name);
2340
2341 sb.final_buf = read_sha1_file(o->blob_sha1, &type,
2342 &sb.final_buf_size);
2343 }
2344 num_read_blob++;
2345 lno = prepare_lines(&sb);
2346
2347 bottom = top = 0;
2348 if (bottomtop)
2349 prepare_blame_range(&sb, bottomtop, lno, &bottom, &top);
2350 if (bottom && top && top < bottom) {
2351 long tmp;
2352 tmp = top; top = bottom; bottom = tmp;
2353 }
2354 if (bottom < 1)
2355 bottom = 1;
2356 if (top < 1)
2357 top = lno;
2358 bottom--;
2359 if (lno < top)
2360 die("file %s has only %lu lines", path, lno);
2361
2362 ent = xcalloc(1, sizeof(*ent));
2363 ent->lno = bottom;
2364 ent->num_lines = top - bottom;
2365 ent->suspect = o;
2366 ent->s_lno = bottom;
2367
2368 sb.ent = ent;
2369 sb.path = path;
2370
2371 if (revs_file && read_ancestry(revs_file))
2372 die("reading graft file %s failed: %s",
2373 revs_file, strerror(errno));
2374
2375 assign_blame(&sb, &revs, opt);
2376
2377 if (incremental)
2378 return 0;
2379
2380 coalesce(&sb);
2381
2382 if (!(output_option & OUTPUT_PORCELAIN))
2383 find_alignment(&sb, &output_option);
2384
2385 output(&sb, output_option);
2386 free((void *)sb.final_buf);
2387 for (ent = sb.ent; ent; ) {
2388 struct blame_entry *e = ent->next;
2389 free(ent);
2390 ent = e;
2391 }
2392
2393 if (show_stats) {
2394 printf("num read blob: %d\n", num_read_blob);
2395 printf("num get patch: %d\n", num_get_patch);
2396 printf("num commits: %d\n", num_commits);
2397 }
2398 return 0;
2399 }