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#include "path-list.h"
20#include "mailmap.h"
21
22static char blame_usage[] =
23"git-blame [-c] [-b] [-l] [--root] [-t] [-f] [-n] [-s] [-p] [-L n,m] [-S <revs-file>] [-M] [-C] [-C] [--contents <filename>] [--incremental] [commit] [--] file\n"
24" -c Use the same output mode as git-annotate (Default: off)\n"
25" -b Show blank SHA-1 for boundary commits (Default: off)\n"
26" -l Show long commit SHA1 (Default: off)\n"
27" --root Do not treat root commits as boundaries (Default: off)\n"
28" -t Show raw timestamp (Default: off)\n"
29" -f, --show-name Show original filename (Default: auto)\n"
30" -n, --show-number Show original linenumber (Default: off)\n"
31" -s Suppress author name and timestamp (Default: off)\n"
32" -p, --porcelain Show in a format designed for machine consumption\n"
33" -L n,m Process only line range n,m, counting from 1\n"
34" -M, -C Find line movements within and across files\n"
35" --incremental Show blame entries as we find them, incrementally\n"
36" --contents file Use <file>'s contents as the final image\n"
37" -S revs-file Use revisions from revs-file instead of calling git-rev-list\n";
38
39static int longest_file;
40static int longest_author;
41static int max_orig_digits;
42static int max_digits;
43static int max_score_digits;
44static int show_root;
45static int blank_boundary;
46static int incremental;
47static int cmd_is_annotate;
48static struct path_list mailmap;
49
50#ifndef DEBUG
51#define DEBUG 0
52#endif
53
54/* stats */
55static int num_read_blob;
56static int num_get_patch;
57static int num_commits;
58
59#define PICKAXE_BLAME_MOVE 01
60#define PICKAXE_BLAME_COPY 02
61#define PICKAXE_BLAME_COPY_HARDER 04
62#define PICKAXE_BLAME_COPY_HARDEST 010
63
64/*
65 * blame for a blame_entry with score lower than these thresholds
66 * is not passed to the parent using move/copy logic.
67 */
68static unsigned blame_move_score;
69static unsigned blame_copy_score;
70#define BLAME_DEFAULT_MOVE_SCORE 20
71#define BLAME_DEFAULT_COPY_SCORE 40
72
73/* bits #0..7 in revision.h, #8..11 used for merge_bases() in commit.c */
74#define METAINFO_SHOWN (1u<<12)
75#define MORE_THAN_ONE_PATH (1u<<13)
76
77/*
78 * One blob in a commit that is being suspected
79 */
80struct origin {
81 int refcnt;
82 struct commit *commit;
83 mmfile_t file;
84 unsigned char blob_sha1[20];
85 char path[FLEX_ARRAY];
86};
87
88/*
89 * Given an origin, prepare mmfile_t structure to be used by the
90 * diff machinery
91 */
92static char *fill_origin_blob(struct origin *o, mmfile_t *file)
93{
94 if (!o->file.ptr) {
95 enum object_type type;
96 num_read_blob++;
97 file->ptr = read_sha1_file(o->blob_sha1, &type,
98 (unsigned long *)(&(file->size)));
99 o->file = *file;
100 }
101 else
102 *file = o->file;
103 return file->ptr;
104}
105
106/*
107 * Origin is refcounted and usually we keep the blob contents to be
108 * reused.
109 */
110static inline struct origin *origin_incref(struct origin *o)
111{
112 if (o)
113 o->refcnt++;
114 return o;
115}
116
117static void origin_decref(struct origin *o)
118{
119 if (o && --o->refcnt <= 0) {
120 if (o->file.ptr)
121 free(o->file.ptr);
122 memset(o, 0, sizeof(*o));
123 free(o);
124 }
125}
126
127/*
128 * Each group of lines is described by a blame_entry; it can be split
129 * as we pass blame to the parents. They form a linked list in the
130 * scoreboard structure, sorted by the target line number.
131 */
132struct blame_entry {
133 struct blame_entry *prev;
134 struct blame_entry *next;
135
136 /* the first line of this group in the final image;
137 * internally all line numbers are 0 based.
138 */
139 int lno;
140
141 /* how many lines this group has */
142 int num_lines;
143
144 /* the commit that introduced this group into the final image */
145 struct origin *suspect;
146
147 /* true if the suspect is truly guilty; false while we have not
148 * checked if the group came from one of its parents.
149 */
150 char guilty;
151
152 /* the line number of the first line of this group in the
153 * suspect's file; internally all line numbers are 0 based.
154 */
155 int s_lno;
156
157 /* how significant this entry is -- cached to avoid
158 * scanning the lines over and over.
159 */
160 unsigned score;
161};
162
163/*
164 * The current state of the blame assignment.
165 */
166struct scoreboard {
167 /* the final commit (i.e. where we started digging from) */
168 struct commit *final;
169
170 const char *path;
171
172 /*
173 * The contents in the final image.
174 * Used by many functions to obtain contents of the nth line,
175 * indexed with scoreboard.lineno[blame_entry.lno].
176 */
177 const char *final_buf;
178 unsigned long final_buf_size;
179
180 /* linked list of blames */
181 struct blame_entry *ent;
182
183 /* look-up a line in the final buffer */
184 int num_lines;
185 int *lineno;
186};
187
188static inline int same_suspect(struct origin *a, struct origin *b)
189{
190 if (a == b)
191 return 1;
192 if (a->commit != b->commit)
193 return 0;
194 return !strcmp(a->path, b->path);
195}
196
197static void sanity_check_refcnt(struct scoreboard *);
198
199/*
200 * If two blame entries that are next to each other came from
201 * contiguous lines in the same origin (i.e. <commit, path> pair),
202 * merge them together.
203 */
204static void coalesce(struct scoreboard *sb)
205{
206 struct blame_entry *ent, *next;
207
208 for (ent = sb->ent; ent && (next = ent->next); ent = next) {
209 if (same_suspect(ent->suspect, next->suspect) &&
210 ent->guilty == next->guilty &&
211 ent->s_lno + ent->num_lines == next->s_lno) {
212 ent->num_lines += next->num_lines;
213 ent->next = next->next;
214 if (ent->next)
215 ent->next->prev = ent;
216 origin_decref(next->suspect);
217 free(next);
218 ent->score = 0;
219 next = ent; /* again */
220 }
221 }
222
223 if (DEBUG) /* sanity */
224 sanity_check_refcnt(sb);
225}
226
227/*
228 * Given a commit and a path in it, create a new origin structure.
229 * The callers that add blame to the scoreboard should use
230 * get_origin() to obtain shared, refcounted copy instead of calling
231 * this function directly.
232 */
233static struct origin *make_origin(struct commit *commit, const char *path)
234{
235 struct origin *o;
236 o = xcalloc(1, sizeof(*o) + strlen(path) + 1);
237 o->commit = commit;
238 o->refcnt = 1;
239 strcpy(o->path, path);
240 return o;
241}
242
243/*
244 * Locate an existing origin or create a new one.
245 */
246static struct origin *get_origin(struct scoreboard *sb,
247 struct commit *commit,
248 const char *path)
249{
250 struct blame_entry *e;
251
252 for (e = sb->ent; e; e = e->next) {
253 if (e->suspect->commit == commit &&
254 !strcmp(e->suspect->path, path))
255 return origin_incref(e->suspect);
256 }
257 return make_origin(commit, path);
258}
259
260/*
261 * Fill the blob_sha1 field of an origin if it hasn't, so that later
262 * call to fill_origin_blob() can use it to locate the data. blob_sha1
263 * for an origin is also used to pass the blame for the entire file to
264 * the parent to detect the case where a child's blob is identical to
265 * that of its parent's.
266 */
267static int fill_blob_sha1(struct origin *origin)
268{
269 unsigned mode;
270
271 if (!is_null_sha1(origin->blob_sha1))
272 return 0;
273 if (get_tree_entry(origin->commit->object.sha1,
274 origin->path,
275 origin->blob_sha1, &mode))
276 goto error_out;
277 if (sha1_object_info(origin->blob_sha1, NULL) != OBJ_BLOB)
278 goto error_out;
279 return 0;
280 error_out:
281 hashclr(origin->blob_sha1);
282 return -1;
283}
284
285/*
286 * We have an origin -- check if the same path exists in the
287 * parent and return an origin structure to represent it.
288 */
289static struct origin *find_origin(struct scoreboard *sb,
290 struct commit *parent,
291 struct origin *origin)
292{
293 struct origin *porigin = NULL;
294 struct diff_options diff_opts;
295 const char *paths[2];
296
297 if (parent->util) {
298 /*
299 * Each commit object can cache one origin in that
300 * commit. This is a freestanding copy of origin and
301 * not refcounted.
302 */
303 struct origin *cached = parent->util;
304 if (!strcmp(cached->path, origin->path)) {
305 /*
306 * The same path between origin and its parent
307 * without renaming -- the most common case.
308 */
309 porigin = get_origin(sb, parent, cached->path);
310
311 /*
312 * If the origin was newly created (i.e. get_origin
313 * would call make_origin if none is found in the
314 * scoreboard), it does not know the blob_sha1,
315 * so copy it. Otherwise porigin was in the
316 * scoreboard and already knows blob_sha1.
317 */
318 if (porigin->refcnt == 1)
319 hashcpy(porigin->blob_sha1, cached->blob_sha1);
320 return porigin;
321 }
322 /* otherwise it was not very useful; free it */
323 free(parent->util);
324 parent->util = NULL;
325 }
326
327 /* See if the origin->path is different between parent
328 * and origin first. Most of the time they are the
329 * same and diff-tree is fairly efficient about this.
330 */
331 diff_setup(&diff_opts);
332 diff_opts.recursive = 1;
333 diff_opts.detect_rename = 0;
334 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
335 paths[0] = origin->path;
336 paths[1] = NULL;
337
338 diff_tree_setup_paths(paths, &diff_opts);
339 if (diff_setup_done(&diff_opts) < 0)
340 die("diff-setup");
341
342 if (is_null_sha1(origin->commit->object.sha1))
343 do_diff_cache(parent->tree->object.sha1, &diff_opts);
344 else
345 diff_tree_sha1(parent->tree->object.sha1,
346 origin->commit->tree->object.sha1,
347 "", &diff_opts);
348 diffcore_std(&diff_opts);
349
350 /* It is either one entry that says "modified", or "created",
351 * or nothing.
352 */
353 if (!diff_queued_diff.nr) {
354 /* The path is the same as parent */
355 porigin = get_origin(sb, parent, origin->path);
356 hashcpy(porigin->blob_sha1, origin->blob_sha1);
357 }
358 else if (diff_queued_diff.nr != 1)
359 die("internal error in blame::find_origin");
360 else {
361 struct diff_filepair *p = diff_queued_diff.queue[0];
362 switch (p->status) {
363 default:
364 die("internal error in blame::find_origin (%c)",
365 p->status);
366 case 'M':
367 porigin = get_origin(sb, parent, origin->path);
368 hashcpy(porigin->blob_sha1, p->one->sha1);
369 break;
370 case 'A':
371 case 'T':
372 /* Did not exist in parent, or type changed */
373 break;
374 }
375 }
376 diff_flush(&diff_opts);
377 if (porigin) {
378 /*
379 * Create a freestanding copy that is not part of
380 * the refcounted origin found in the scoreboard, and
381 * cache it in the commit.
382 */
383 struct origin *cached;
384
385 cached = make_origin(porigin->commit, porigin->path);
386 hashcpy(cached->blob_sha1, porigin->blob_sha1);
387 parent->util = cached;
388 }
389 return porigin;
390}
391
392/*
393 * We have an origin -- find the path that corresponds to it in its
394 * parent and return an origin structure to represent it.
395 */
396static struct origin *find_rename(struct scoreboard *sb,
397 struct commit *parent,
398 struct origin *origin)
399{
400 struct origin *porigin = NULL;
401 struct diff_options diff_opts;
402 int i;
403 const char *paths[2];
404
405 diff_setup(&diff_opts);
406 diff_opts.recursive = 1;
407 diff_opts.detect_rename = DIFF_DETECT_RENAME;
408 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
409 diff_opts.single_follow = origin->path;
410 paths[0] = NULL;
411 diff_tree_setup_paths(paths, &diff_opts);
412 if (diff_setup_done(&diff_opts) < 0)
413 die("diff-setup");
414
415 if (is_null_sha1(origin->commit->object.sha1))
416 do_diff_cache(parent->tree->object.sha1, &diff_opts);
417 else
418 diff_tree_sha1(parent->tree->object.sha1,
419 origin->commit->tree->object.sha1,
420 "", &diff_opts);
421 diffcore_std(&diff_opts);
422
423 for (i = 0; i < diff_queued_diff.nr; i++) {
424 struct diff_filepair *p = diff_queued_diff.queue[i];
425 if ((p->status == 'R' || p->status == 'C') &&
426 !strcmp(p->two->path, origin->path)) {
427 porigin = get_origin(sb, parent, p->one->path);
428 hashcpy(porigin->blob_sha1, p->one->sha1);
429 break;
430 }
431 }
432 diff_flush(&diff_opts);
433 return porigin;
434}
435
436/*
437 * Parsing of patch chunks...
438 */
439struct chunk {
440 /* line number in postimage; up to but not including this
441 * line is the same as preimage
442 */
443 int same;
444
445 /* preimage line number after this chunk */
446 int p_next;
447
448 /* postimage line number after this chunk */
449 int t_next;
450};
451
452struct patch {
453 struct chunk *chunks;
454 int num;
455};
456
457struct blame_diff_state {
458 struct xdiff_emit_state xm;
459 struct patch *ret;
460 unsigned hunk_post_context;
461 unsigned hunk_in_pre_context : 1;
462};
463
464static void process_u_diff(void *state_, char *line, unsigned long len)
465{
466 struct blame_diff_state *state = state_;
467 struct chunk *chunk;
468 int off1, off2, len1, len2, num;
469
470 num = state->ret->num;
471 if (len < 4 || line[0] != '@' || line[1] != '@') {
472 if (state->hunk_in_pre_context && line[0] == ' ')
473 state->ret->chunks[num - 1].same++;
474 else {
475 state->hunk_in_pre_context = 0;
476 if (line[0] == ' ')
477 state->hunk_post_context++;
478 else
479 state->hunk_post_context = 0;
480 }
481 return;
482 }
483
484 if (num && state->hunk_post_context) {
485 chunk = &state->ret->chunks[num - 1];
486 chunk->p_next -= state->hunk_post_context;
487 chunk->t_next -= state->hunk_post_context;
488 }
489 state->ret->num = ++num;
490 state->ret->chunks = xrealloc(state->ret->chunks,
491 sizeof(struct chunk) * num);
492 chunk = &state->ret->chunks[num - 1];
493 if (parse_hunk_header(line, len, &off1, &len1, &off2, &len2)) {
494 state->ret->num--;
495 return;
496 }
497
498 /* Line numbers in patch output are one based. */
499 off1--;
500 off2--;
501
502 chunk->same = len2 ? off2 : (off2 + 1);
503
504 chunk->p_next = off1 + (len1 ? len1 : 1);
505 chunk->t_next = chunk->same + len2;
506 state->hunk_in_pre_context = 1;
507 state->hunk_post_context = 0;
508}
509
510static struct patch *compare_buffer(mmfile_t *file_p, mmfile_t *file_o,
511 int context)
512{
513 struct blame_diff_state state;
514 xpparam_t xpp;
515 xdemitconf_t xecfg;
516 xdemitcb_t ecb;
517
518 xpp.flags = XDF_NEED_MINIMAL;
519 xecfg.ctxlen = context;
520 xecfg.flags = 0;
521 ecb.outf = xdiff_outf;
522 ecb.priv = &state;
523 memset(&state, 0, sizeof(state));
524 state.xm.consume = process_u_diff;
525 state.ret = xmalloc(sizeof(struct patch));
526 state.ret->chunks = NULL;
527 state.ret->num = 0;
528
529 xdl_diff(file_p, file_o, &xpp, &xecfg, &ecb);
530
531 if (state.ret->num) {
532 struct chunk *chunk;
533 chunk = &state.ret->chunks[state.ret->num - 1];
534 chunk->p_next -= state.hunk_post_context;
535 chunk->t_next -= state.hunk_post_context;
536 }
537 return state.ret;
538}
539
540/*
541 * Run diff between two origins and grab the patch output, so that
542 * we can pass blame for lines origin is currently suspected for
543 * to its parent.
544 */
545static struct patch *get_patch(struct origin *parent, struct origin *origin)
546{
547 mmfile_t file_p, file_o;
548 struct patch *patch;
549
550 fill_origin_blob(parent, &file_p);
551 fill_origin_blob(origin, &file_o);
552 if (!file_p.ptr || !file_o.ptr)
553 return NULL;
554 patch = compare_buffer(&file_p, &file_o, 0);
555 num_get_patch++;
556 return patch;
557}
558
559static void free_patch(struct patch *p)
560{
561 free(p->chunks);
562 free(p);
563}
564
565/*
566 * Link in a new blame entry to the scoreboard. Entries that cover the
567 * same line range have been removed from the scoreboard previously.
568 */
569static void add_blame_entry(struct scoreboard *sb, struct blame_entry *e)
570{
571 struct blame_entry *ent, *prev = NULL;
572
573 origin_incref(e->suspect);
574
575 for (ent = sb->ent; ent && ent->lno < e->lno; ent = ent->next)
576 prev = ent;
577
578 /* prev, if not NULL, is the last one that is below e */
579 e->prev = prev;
580 if (prev) {
581 e->next = prev->next;
582 prev->next = e;
583 }
584 else {
585 e->next = sb->ent;
586 sb->ent = e;
587 }
588 if (e->next)
589 e->next->prev = e;
590}
591
592/*
593 * src typically is on-stack; we want to copy the information in it to
594 * an malloced blame_entry that is already on the linked list of the
595 * scoreboard. The origin of dst loses a refcnt while the origin of src
596 * gains one.
597 */
598static void dup_entry(struct blame_entry *dst, struct blame_entry *src)
599{
600 struct blame_entry *p, *n;
601
602 p = dst->prev;
603 n = dst->next;
604 origin_incref(src->suspect);
605 origin_decref(dst->suspect);
606 memcpy(dst, src, sizeof(*src));
607 dst->prev = p;
608 dst->next = n;
609 dst->score = 0;
610}
611
612static const char *nth_line(struct scoreboard *sb, int lno)
613{
614 return sb->final_buf + sb->lineno[lno];
615}
616
617/*
618 * It is known that lines between tlno to same came from parent, and e
619 * has an overlap with that range. it also is known that parent's
620 * line plno corresponds to e's line tlno.
621 *
622 * <---- e ----->
623 * <------>
624 * <------------>
625 * <------------>
626 * <------------------>
627 *
628 * Split e into potentially three parts; before this chunk, the chunk
629 * to be blamed for the parent, and after that portion.
630 */
631static void split_overlap(struct blame_entry *split,
632 struct blame_entry *e,
633 int tlno, int plno, int same,
634 struct origin *parent)
635{
636 int chunk_end_lno;
637 memset(split, 0, sizeof(struct blame_entry [3]));
638
639 if (e->s_lno < tlno) {
640 /* there is a pre-chunk part not blamed on parent */
641 split[0].suspect = origin_incref(e->suspect);
642 split[0].lno = e->lno;
643 split[0].s_lno = e->s_lno;
644 split[0].num_lines = tlno - e->s_lno;
645 split[1].lno = e->lno + tlno - e->s_lno;
646 split[1].s_lno = plno;
647 }
648 else {
649 split[1].lno = e->lno;
650 split[1].s_lno = plno + (e->s_lno - tlno);
651 }
652
653 if (same < e->s_lno + e->num_lines) {
654 /* there is a post-chunk part not blamed on parent */
655 split[2].suspect = origin_incref(e->suspect);
656 split[2].lno = e->lno + (same - e->s_lno);
657 split[2].s_lno = e->s_lno + (same - e->s_lno);
658 split[2].num_lines = e->s_lno + e->num_lines - same;
659 chunk_end_lno = split[2].lno;
660 }
661 else
662 chunk_end_lno = e->lno + e->num_lines;
663 split[1].num_lines = chunk_end_lno - split[1].lno;
664
665 /*
666 * if it turns out there is nothing to blame the parent for,
667 * forget about the splitting. !split[1].suspect signals this.
668 */
669 if (split[1].num_lines < 1)
670 return;
671 split[1].suspect = origin_incref(parent);
672}
673
674/*
675 * split_overlap() divided an existing blame e into up to three parts
676 * in split. Adjust the linked list of blames in the scoreboard to
677 * reflect the split.
678 */
679static void split_blame(struct scoreboard *sb,
680 struct blame_entry *split,
681 struct blame_entry *e)
682{
683 struct blame_entry *new_entry;
684
685 if (split[0].suspect && split[2].suspect) {
686 /* The first part (reuse storage for the existing entry e) */
687 dup_entry(e, &split[0]);
688
689 /* The last part -- me */
690 new_entry = xmalloc(sizeof(*new_entry));
691 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
692 add_blame_entry(sb, new_entry);
693
694 /* ... and the middle part -- parent */
695 new_entry = xmalloc(sizeof(*new_entry));
696 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
697 add_blame_entry(sb, new_entry);
698 }
699 else if (!split[0].suspect && !split[2].suspect)
700 /*
701 * The parent covers the entire area; reuse storage for
702 * e and replace it with the parent.
703 */
704 dup_entry(e, &split[1]);
705 else if (split[0].suspect) {
706 /* me and then parent */
707 dup_entry(e, &split[0]);
708
709 new_entry = xmalloc(sizeof(*new_entry));
710 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
711 add_blame_entry(sb, new_entry);
712 }
713 else {
714 /* parent and then me */
715 dup_entry(e, &split[1]);
716
717 new_entry = xmalloc(sizeof(*new_entry));
718 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
719 add_blame_entry(sb, new_entry);
720 }
721
722 if (DEBUG) { /* sanity */
723 struct blame_entry *ent;
724 int lno = sb->ent->lno, corrupt = 0;
725
726 for (ent = sb->ent; ent; ent = ent->next) {
727 if (lno != ent->lno)
728 corrupt = 1;
729 if (ent->s_lno < 0)
730 corrupt = 1;
731 lno += ent->num_lines;
732 }
733 if (corrupt) {
734 lno = sb->ent->lno;
735 for (ent = sb->ent; ent; ent = ent->next) {
736 printf("L %8d l %8d n %8d\n",
737 lno, ent->lno, ent->num_lines);
738 lno = ent->lno + ent->num_lines;
739 }
740 die("oops");
741 }
742 }
743}
744
745/*
746 * After splitting the blame, the origins used by the
747 * on-stack blame_entry should lose one refcnt each.
748 */
749static void decref_split(struct blame_entry *split)
750{
751 int i;
752
753 for (i = 0; i < 3; i++)
754 origin_decref(split[i].suspect);
755}
756
757/*
758 * Helper for blame_chunk(). blame_entry e is known to overlap with
759 * the patch hunk; split it and pass blame to the parent.
760 */
761static void blame_overlap(struct scoreboard *sb, struct blame_entry *e,
762 int tlno, int plno, int same,
763 struct origin *parent)
764{
765 struct blame_entry split[3];
766
767 split_overlap(split, e, tlno, plno, same, parent);
768 if (split[1].suspect)
769 split_blame(sb, split, e);
770 decref_split(split);
771}
772
773/*
774 * Find the line number of the last line the target is suspected for.
775 */
776static int find_last_in_target(struct scoreboard *sb, struct origin *target)
777{
778 struct blame_entry *e;
779 int last_in_target = -1;
780
781 for (e = sb->ent; e; e = e->next) {
782 if (e->guilty || !same_suspect(e->suspect, target))
783 continue;
784 if (last_in_target < e->s_lno + e->num_lines)
785 last_in_target = e->s_lno + e->num_lines;
786 }
787 return last_in_target;
788}
789
790/*
791 * Process one hunk from the patch between the current suspect for
792 * blame_entry e and its parent. Find and split the overlap, and
793 * pass blame to the overlapping part to the parent.
794 */
795static void blame_chunk(struct scoreboard *sb,
796 int tlno, int plno, int same,
797 struct origin *target, struct origin *parent)
798{
799 struct blame_entry *e;
800
801 for (e = sb->ent; e; e = e->next) {
802 if (e->guilty || !same_suspect(e->suspect, target))
803 continue;
804 if (same <= e->s_lno)
805 continue;
806 if (tlno < e->s_lno + e->num_lines)
807 blame_overlap(sb, e, tlno, plno, same, parent);
808 }
809}
810
811/*
812 * We are looking at the origin 'target' and aiming to pass blame
813 * for the lines it is suspected to its parent. Run diff to find
814 * which lines came from parent and pass blame for them.
815 */
816static int pass_blame_to_parent(struct scoreboard *sb,
817 struct origin *target,
818 struct origin *parent)
819{
820 int i, last_in_target, plno, tlno;
821 struct patch *patch;
822
823 last_in_target = find_last_in_target(sb, target);
824 if (last_in_target < 0)
825 return 1; /* nothing remains for this target */
826
827 patch = get_patch(parent, target);
828 plno = tlno = 0;
829 for (i = 0; i < patch->num; i++) {
830 struct chunk *chunk = &patch->chunks[i];
831
832 blame_chunk(sb, tlno, plno, chunk->same, target, parent);
833 plno = chunk->p_next;
834 tlno = chunk->t_next;
835 }
836 /* The rest (i.e. anything after tlno) are the same as the parent */
837 blame_chunk(sb, tlno, plno, last_in_target, target, parent);
838
839 free_patch(patch);
840 return 0;
841}
842
843/*
844 * The lines in blame_entry after splitting blames many times can become
845 * very small and trivial, and at some point it becomes pointless to
846 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
847 * ordinary C program, and it is not worth to say it was copied from
848 * totally unrelated file in the parent.
849 *
850 * Compute how trivial the lines in the blame_entry are.
851 */
852static unsigned ent_score(struct scoreboard *sb, struct blame_entry *e)
853{
854 unsigned score;
855 const char *cp, *ep;
856
857 if (e->score)
858 return e->score;
859
860 score = 1;
861 cp = nth_line(sb, e->lno);
862 ep = nth_line(sb, e->lno + e->num_lines);
863 while (cp < ep) {
864 unsigned ch = *((unsigned char *)cp);
865 if (isalnum(ch))
866 score++;
867 cp++;
868 }
869 e->score = score;
870 return score;
871}
872
873/*
874 * best_so_far[] and this[] are both a split of an existing blame_entry
875 * that passes blame to the parent. Maintain best_so_far the best split
876 * so far, by comparing this and best_so_far and copying this into
877 * bst_so_far as needed.
878 */
879static void copy_split_if_better(struct scoreboard *sb,
880 struct blame_entry *best_so_far,
881 struct blame_entry *this)
882{
883 int i;
884
885 if (!this[1].suspect)
886 return;
887 if (best_so_far[1].suspect) {
888 if (ent_score(sb, &this[1]) < ent_score(sb, &best_so_far[1]))
889 return;
890 }
891
892 for (i = 0; i < 3; i++)
893 origin_incref(this[i].suspect);
894 decref_split(best_so_far);
895 memcpy(best_so_far, this, sizeof(struct blame_entry [3]));
896}
897
898/*
899 * We are looking at a part of the final image represented by
900 * ent (tlno and same are offset by ent->s_lno).
901 * tlno is where we are looking at in the final image.
902 * up to (but not including) same match preimage.
903 * plno is where we are looking at in the preimage.
904 *
905 * <-------------- final image ---------------------->
906 * <------ent------>
907 * ^tlno ^same
908 * <---------preimage----->
909 * ^plno
910 *
911 * All line numbers are 0-based.
912 */
913static void handle_split(struct scoreboard *sb,
914 struct blame_entry *ent,
915 int tlno, int plno, int same,
916 struct origin *parent,
917 struct blame_entry *split)
918{
919 if (ent->num_lines <= tlno)
920 return;
921 if (tlno < same) {
922 struct blame_entry this[3];
923 tlno += ent->s_lno;
924 same += ent->s_lno;
925 split_overlap(this, ent, tlno, plno, same, parent);
926 copy_split_if_better(sb, split, this);
927 decref_split(this);
928 }
929}
930
931/*
932 * Find the lines from parent that are the same as ent so that
933 * we can pass blames to it. file_p has the blob contents for
934 * the parent.
935 */
936static void find_copy_in_blob(struct scoreboard *sb,
937 struct blame_entry *ent,
938 struct origin *parent,
939 struct blame_entry *split,
940 mmfile_t *file_p)
941{
942 const char *cp;
943 int cnt;
944 mmfile_t file_o;
945 struct patch *patch;
946 int i, plno, tlno;
947
948 /*
949 * Prepare mmfile that contains only the lines in ent.
950 */
951 cp = nth_line(sb, ent->lno);
952 file_o.ptr = (char*) cp;
953 cnt = ent->num_lines;
954
955 while (cnt && cp < sb->final_buf + sb->final_buf_size) {
956 if (*cp++ == '\n')
957 cnt--;
958 }
959 file_o.size = cp - file_o.ptr;
960
961 patch = compare_buffer(file_p, &file_o, 1);
962
963 /*
964 * file_o is a part of final image we are annotating.
965 * file_p partially may match that image.
966 */
967 memset(split, 0, sizeof(struct blame_entry [3]));
968 plno = tlno = 0;
969 for (i = 0; i < patch->num; i++) {
970 struct chunk *chunk = &patch->chunks[i];
971
972 handle_split(sb, ent, tlno, plno, chunk->same, parent, split);
973 plno = chunk->p_next;
974 tlno = chunk->t_next;
975 }
976 /* remainder, if any, all match the preimage */
977 handle_split(sb, ent, tlno, plno, ent->num_lines, parent, split);
978 free_patch(patch);
979}
980
981/*
982 * See if lines currently target is suspected for can be attributed to
983 * parent.
984 */
985static int find_move_in_parent(struct scoreboard *sb,
986 struct origin *target,
987 struct origin *parent)
988{
989 int last_in_target, made_progress;
990 struct blame_entry *e, split[3];
991 mmfile_t file_p;
992
993 last_in_target = find_last_in_target(sb, target);
994 if (last_in_target < 0)
995 return 1; /* nothing remains for this target */
996
997 fill_origin_blob(parent, &file_p);
998 if (!file_p.ptr)
999 return 0;
1000
1001 made_progress = 1;
1002 while (made_progress) {
1003 made_progress = 0;
1004 for (e = sb->ent; e; e = e->next) {
1005 if (e->guilty || !same_suspect(e->suspect, target))
1006 continue;
1007 find_copy_in_blob(sb, e, parent, split, &file_p);
1008 if (split[1].suspect &&
1009 blame_move_score < ent_score(sb, &split[1])) {
1010 split_blame(sb, split, e);
1011 made_progress = 1;
1012 }
1013 decref_split(split);
1014 }
1015 }
1016 return 0;
1017}
1018
1019struct blame_list {
1020 struct blame_entry *ent;
1021 struct blame_entry split[3];
1022};
1023
1024/*
1025 * Count the number of entries the target is suspected for,
1026 * and prepare a list of entry and the best split.
1027 */
1028static struct blame_list *setup_blame_list(struct scoreboard *sb,
1029 struct origin *target,
1030 int *num_ents_p)
1031{
1032 struct blame_entry *e;
1033 int num_ents, i;
1034 struct blame_list *blame_list = NULL;
1035
1036 for (e = sb->ent, num_ents = 0; e; e = e->next)
1037 if (!e->guilty && same_suspect(e->suspect, target))
1038 num_ents++;
1039 if (num_ents) {
1040 blame_list = xcalloc(num_ents, sizeof(struct blame_list));
1041 for (e = sb->ent, i = 0; e; e = e->next)
1042 if (!e->guilty && same_suspect(e->suspect, target))
1043 blame_list[i++].ent = e;
1044 }
1045 *num_ents_p = num_ents;
1046 return blame_list;
1047}
1048
1049/*
1050 * For lines target is suspected for, see if we can find code movement
1051 * across file boundary from the parent commit. porigin is the path
1052 * in the parent we already tried.
1053 */
1054static int find_copy_in_parent(struct scoreboard *sb,
1055 struct origin *target,
1056 struct commit *parent,
1057 struct origin *porigin,
1058 int opt)
1059{
1060 struct diff_options diff_opts;
1061 const char *paths[1];
1062 int i, j;
1063 int retval;
1064 struct blame_list *blame_list;
1065 int num_ents;
1066
1067 blame_list = setup_blame_list(sb, target, &num_ents);
1068 if (!blame_list)
1069 return 1; /* nothing remains for this target */
1070
1071 diff_setup(&diff_opts);
1072 diff_opts.recursive = 1;
1073 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
1074
1075 paths[0] = NULL;
1076 diff_tree_setup_paths(paths, &diff_opts);
1077 if (diff_setup_done(&diff_opts) < 0)
1078 die("diff-setup");
1079
1080 /* Try "find copies harder" on new path if requested;
1081 * we do not want to use diffcore_rename() actually to
1082 * match things up; find_copies_harder is set only to
1083 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1084 * and this code needs to be after diff_setup_done(), which
1085 * usually makes find-copies-harder imply copy detection.
1086 */
1087 if ((opt & PICKAXE_BLAME_COPY_HARDEST)
1088 || ((opt & PICKAXE_BLAME_COPY_HARDER)
1089 && (!porigin || strcmp(target->path, porigin->path))))
1090 diff_opts.find_copies_harder = 1;
1091
1092 if (is_null_sha1(target->commit->object.sha1))
1093 do_diff_cache(parent->tree->object.sha1, &diff_opts);
1094 else
1095 diff_tree_sha1(parent->tree->object.sha1,
1096 target->commit->tree->object.sha1,
1097 "", &diff_opts);
1098
1099 if (!diff_opts.find_copies_harder)
1100 diffcore_std(&diff_opts);
1101
1102 retval = 0;
1103 while (1) {
1104 int made_progress = 0;
1105
1106 for (i = 0; i < diff_queued_diff.nr; i++) {
1107 struct diff_filepair *p = diff_queued_diff.queue[i];
1108 struct origin *norigin;
1109 mmfile_t file_p;
1110 struct blame_entry this[3];
1111
1112 if (!DIFF_FILE_VALID(p->one))
1113 continue; /* does not exist in parent */
1114 if (porigin && !strcmp(p->one->path, porigin->path))
1115 /* find_move already dealt with this path */
1116 continue;
1117
1118 norigin = get_origin(sb, parent, p->one->path);
1119 hashcpy(norigin->blob_sha1, p->one->sha1);
1120 fill_origin_blob(norigin, &file_p);
1121 if (!file_p.ptr)
1122 continue;
1123
1124 for (j = 0; j < num_ents; j++) {
1125 find_copy_in_blob(sb, blame_list[j].ent,
1126 norigin, this, &file_p);
1127 copy_split_if_better(sb, blame_list[j].split,
1128 this);
1129 decref_split(this);
1130 }
1131 origin_decref(norigin);
1132 }
1133
1134 for (j = 0; j < num_ents; j++) {
1135 struct blame_entry *split = blame_list[j].split;
1136 if (split[1].suspect &&
1137 blame_copy_score < ent_score(sb, &split[1])) {
1138 split_blame(sb, split, blame_list[j].ent);
1139 made_progress = 1;
1140 }
1141 decref_split(split);
1142 }
1143 free(blame_list);
1144
1145 if (!made_progress)
1146 break;
1147 blame_list = setup_blame_list(sb, target, &num_ents);
1148 if (!blame_list) {
1149 retval = 1;
1150 break;
1151 }
1152 }
1153 diff_flush(&diff_opts);
1154
1155 return retval;
1156}
1157
1158/*
1159 * The blobs of origin and porigin exactly match, so everything
1160 * origin is suspected for can be blamed on the parent.
1161 */
1162static void pass_whole_blame(struct scoreboard *sb,
1163 struct origin *origin, struct origin *porigin)
1164{
1165 struct blame_entry *e;
1166
1167 if (!porigin->file.ptr && origin->file.ptr) {
1168 /* Steal its file */
1169 porigin->file = origin->file;
1170 origin->file.ptr = NULL;
1171 }
1172 for (e = sb->ent; e; e = e->next) {
1173 if (!same_suspect(e->suspect, origin))
1174 continue;
1175 origin_incref(porigin);
1176 origin_decref(e->suspect);
1177 e->suspect = porigin;
1178 }
1179}
1180
1181#define MAXPARENT 16
1182
1183static void pass_blame(struct scoreboard *sb, struct origin *origin, int opt)
1184{
1185 int i, pass;
1186 struct commit *commit = origin->commit;
1187 struct commit_list *parent;
1188 struct origin *parent_origin[MAXPARENT], *porigin;
1189
1190 memset(parent_origin, 0, sizeof(parent_origin));
1191
1192 /* The first pass looks for unrenamed path to optimize for
1193 * common cases, then we look for renames in the second pass.
1194 */
1195 for (pass = 0; pass < 2; pass++) {
1196 struct origin *(*find)(struct scoreboard *,
1197 struct commit *, struct origin *);
1198 find = pass ? find_rename : find_origin;
1199
1200 for (i = 0, parent = commit->parents;
1201 i < MAXPARENT && parent;
1202 parent = parent->next, i++) {
1203 struct commit *p = parent->item;
1204 int j, same;
1205
1206 if (parent_origin[i])
1207 continue;
1208 if (parse_commit(p))
1209 continue;
1210 porigin = find(sb, p, origin);
1211 if (!porigin)
1212 continue;
1213 if (!hashcmp(porigin->blob_sha1, origin->blob_sha1)) {
1214 pass_whole_blame(sb, origin, porigin);
1215 origin_decref(porigin);
1216 goto finish;
1217 }
1218 for (j = same = 0; j < i; j++)
1219 if (parent_origin[j] &&
1220 !hashcmp(parent_origin[j]->blob_sha1,
1221 porigin->blob_sha1)) {
1222 same = 1;
1223 break;
1224 }
1225 if (!same)
1226 parent_origin[i] = porigin;
1227 else
1228 origin_decref(porigin);
1229 }
1230 }
1231
1232 num_commits++;
1233 for (i = 0, parent = commit->parents;
1234 i < MAXPARENT && parent;
1235 parent = parent->next, i++) {
1236 struct origin *porigin = parent_origin[i];
1237 if (!porigin)
1238 continue;
1239 if (pass_blame_to_parent(sb, origin, porigin))
1240 goto finish;
1241 }
1242
1243 /*
1244 * Optionally find moves in parents' files.
1245 */
1246 if (opt & PICKAXE_BLAME_MOVE)
1247 for (i = 0, parent = commit->parents;
1248 i < MAXPARENT && parent;
1249 parent = parent->next, i++) {
1250 struct origin *porigin = parent_origin[i];
1251 if (!porigin)
1252 continue;
1253 if (find_move_in_parent(sb, origin, porigin))
1254 goto finish;
1255 }
1256
1257 /*
1258 * Optionally find copies from parents' files.
1259 */
1260 if (opt & PICKAXE_BLAME_COPY)
1261 for (i = 0, parent = commit->parents;
1262 i < MAXPARENT && parent;
1263 parent = parent->next, i++) {
1264 struct origin *porigin = parent_origin[i];
1265 if (find_copy_in_parent(sb, origin, parent->item,
1266 porigin, opt))
1267 goto finish;
1268 }
1269
1270 finish:
1271 for (i = 0; i < MAXPARENT; i++)
1272 origin_decref(parent_origin[i]);
1273}
1274
1275/*
1276 * Information on commits, used for output.
1277 */
1278struct commit_info
1279{
1280 const char *author;
1281 const char *author_mail;
1282 unsigned long author_time;
1283 const char *author_tz;
1284
1285 /* filled only when asked for details */
1286 const char *committer;
1287 const char *committer_mail;
1288 unsigned long committer_time;
1289 const char *committer_tz;
1290
1291 const char *summary;
1292};
1293
1294/*
1295 * Parse author/committer line in the commit object buffer
1296 */
1297static void get_ac_line(const char *inbuf, const char *what,
1298 int bufsz, char *person, const char **mail,
1299 unsigned long *time, const char **tz)
1300{
1301 int len, tzlen, maillen;
1302 char *tmp, *endp, *timepos;
1303
1304 tmp = strstr(inbuf, what);
1305 if (!tmp)
1306 goto error_out;
1307 tmp += strlen(what);
1308 endp = strchr(tmp, '\n');
1309 if (!endp)
1310 len = strlen(tmp);
1311 else
1312 len = endp - tmp;
1313 if (bufsz <= len) {
1314 error_out:
1315 /* Ugh */
1316 *mail = *tz = "(unknown)";
1317 *time = 0;
1318 return;
1319 }
1320 memcpy(person, tmp, len);
1321
1322 tmp = person;
1323 tmp += len;
1324 *tmp = 0;
1325 while (*tmp != ' ')
1326 tmp--;
1327 *tz = tmp+1;
1328 tzlen = (person+len)-(tmp+1);
1329
1330 *tmp = 0;
1331 while (*tmp != ' ')
1332 tmp--;
1333 *time = strtoul(tmp, NULL, 10);
1334 timepos = tmp;
1335
1336 *tmp = 0;
1337 while (*tmp != ' ')
1338 tmp--;
1339 *mail = tmp + 1;
1340 *tmp = 0;
1341 maillen = timepos - tmp;
1342
1343 if (!mailmap.nr)
1344 return;
1345
1346 /*
1347 * mailmap expansion may make the name longer.
1348 * make room by pushing stuff down.
1349 */
1350 tmp = person + bufsz - (tzlen + 1);
1351 memmove(tmp, *tz, tzlen);
1352 tmp[tzlen] = 0;
1353 *tz = tmp;
1354
1355 tmp = tmp - (maillen + 1);
1356 memmove(tmp, *mail, maillen);
1357 tmp[maillen] = 0;
1358 *mail = tmp;
1359
1360 /*
1361 * Now, convert e-mail using mailmap
1362 */
1363 map_email(&mailmap, tmp + 1, person, tmp-person-1);
1364}
1365
1366static void get_commit_info(struct commit *commit,
1367 struct commit_info *ret,
1368 int detailed)
1369{
1370 int len;
1371 char *tmp, *endp;
1372 static char author_buf[1024];
1373 static char committer_buf[1024];
1374 static char summary_buf[1024];
1375
1376 /*
1377 * We've operated without save_commit_buffer, so
1378 * we now need to populate them for output.
1379 */
1380 if (!commit->buffer) {
1381 enum object_type type;
1382 unsigned long size;
1383 commit->buffer =
1384 read_sha1_file(commit->object.sha1, &type, &size);
1385 }
1386 ret->author = author_buf;
1387 get_ac_line(commit->buffer, "\nauthor ",
1388 sizeof(author_buf), author_buf, &ret->author_mail,
1389 &ret->author_time, &ret->author_tz);
1390
1391 if (!detailed)
1392 return;
1393
1394 ret->committer = committer_buf;
1395 get_ac_line(commit->buffer, "\ncommitter ",
1396 sizeof(committer_buf), committer_buf, &ret->committer_mail,
1397 &ret->committer_time, &ret->committer_tz);
1398
1399 ret->summary = summary_buf;
1400 tmp = strstr(commit->buffer, "\n\n");
1401 if (!tmp) {
1402 error_out:
1403 sprintf(summary_buf, "(%s)", sha1_to_hex(commit->object.sha1));
1404 return;
1405 }
1406 tmp += 2;
1407 endp = strchr(tmp, '\n');
1408 if (!endp)
1409 endp = tmp + strlen(tmp);
1410 len = endp - tmp;
1411 if (len >= sizeof(summary_buf) || len == 0)
1412 goto error_out;
1413 memcpy(summary_buf, tmp, len);
1414 summary_buf[len] = 0;
1415}
1416
1417/*
1418 * To allow LF and other nonportable characters in pathnames,
1419 * they are c-style quoted as needed.
1420 */
1421static void write_filename_info(const char *path)
1422{
1423 printf("filename ");
1424 write_name_quoted(NULL, 0, path, 1, stdout);
1425 putchar('\n');
1426}
1427
1428/*
1429 * The blame_entry is found to be guilty for the range. Mark it
1430 * as such, and show it in incremental output.
1431 */
1432static void found_guilty_entry(struct blame_entry *ent)
1433{
1434 if (ent->guilty)
1435 return;
1436 ent->guilty = 1;
1437 if (incremental) {
1438 struct origin *suspect = ent->suspect;
1439
1440 printf("%s %d %d %d\n",
1441 sha1_to_hex(suspect->commit->object.sha1),
1442 ent->s_lno + 1, ent->lno + 1, ent->num_lines);
1443 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1444 struct commit_info ci;
1445 suspect->commit->object.flags |= METAINFO_SHOWN;
1446 get_commit_info(suspect->commit, &ci, 1);
1447 printf("author %s\n", ci.author);
1448 printf("author-mail %s\n", ci.author_mail);
1449 printf("author-time %lu\n", ci.author_time);
1450 printf("author-tz %s\n", ci.author_tz);
1451 printf("committer %s\n", ci.committer);
1452 printf("committer-mail %s\n", ci.committer_mail);
1453 printf("committer-time %lu\n", ci.committer_time);
1454 printf("committer-tz %s\n", ci.committer_tz);
1455 printf("summary %s\n", ci.summary);
1456 if (suspect->commit->object.flags & UNINTERESTING)
1457 printf("boundary\n");
1458 }
1459 write_filename_info(suspect->path);
1460 }
1461}
1462
1463/*
1464 * The main loop -- while the scoreboard has lines whose true origin
1465 * is still unknown, pick one blame_entry, and allow its current
1466 * suspect to pass blames to its parents.
1467 */
1468static void assign_blame(struct scoreboard *sb, struct rev_info *revs, int opt)
1469{
1470 while (1) {
1471 struct blame_entry *ent;
1472 struct commit *commit;
1473 struct origin *suspect = NULL;
1474
1475 /* find one suspect to break down */
1476 for (ent = sb->ent; !suspect && ent; ent = ent->next)
1477 if (!ent->guilty)
1478 suspect = ent->suspect;
1479 if (!suspect)
1480 return; /* all done */
1481
1482 /*
1483 * We will use this suspect later in the loop,
1484 * so hold onto it in the meantime.
1485 */
1486 origin_incref(suspect);
1487 commit = suspect->commit;
1488 if (!commit->object.parsed)
1489 parse_commit(commit);
1490 if (!(commit->object.flags & UNINTERESTING) &&
1491 !(revs->max_age != -1 && commit->date < revs->max_age))
1492 pass_blame(sb, suspect, opt);
1493 else {
1494 commit->object.flags |= UNINTERESTING;
1495 if (commit->object.parsed)
1496 mark_parents_uninteresting(commit);
1497 }
1498 /* treat root commit as boundary */
1499 if (!commit->parents && !show_root)
1500 commit->object.flags |= UNINTERESTING;
1501
1502 /* Take responsibility for the remaining entries */
1503 for (ent = sb->ent; ent; ent = ent->next)
1504 if (same_suspect(ent->suspect, suspect))
1505 found_guilty_entry(ent);
1506 origin_decref(suspect);
1507
1508 if (DEBUG) /* sanity */
1509 sanity_check_refcnt(sb);
1510 }
1511}
1512
1513static const char *format_time(unsigned long time, const char *tz_str,
1514 int show_raw_time)
1515{
1516 static char time_buf[128];
1517 time_t t = time;
1518 int minutes, tz;
1519 struct tm *tm;
1520
1521 if (show_raw_time) {
1522 sprintf(time_buf, "%lu %s", time, tz_str);
1523 return time_buf;
1524 }
1525
1526 tz = atoi(tz_str);
1527 minutes = tz < 0 ? -tz : tz;
1528 minutes = (minutes / 100)*60 + (minutes % 100);
1529 minutes = tz < 0 ? -minutes : minutes;
1530 t = time + minutes * 60;
1531 tm = gmtime(&t);
1532
1533 strftime(time_buf, sizeof(time_buf), "%Y-%m-%d %H:%M:%S ", tm);
1534 strcat(time_buf, tz_str);
1535 return time_buf;
1536}
1537
1538#define OUTPUT_ANNOTATE_COMPAT 001
1539#define OUTPUT_LONG_OBJECT_NAME 002
1540#define OUTPUT_RAW_TIMESTAMP 004
1541#define OUTPUT_PORCELAIN 010
1542#define OUTPUT_SHOW_NAME 020
1543#define OUTPUT_SHOW_NUMBER 040
1544#define OUTPUT_SHOW_SCORE 0100
1545#define OUTPUT_NO_AUTHOR 0200
1546
1547static void emit_porcelain(struct scoreboard *sb, struct blame_entry *ent)
1548{
1549 int cnt;
1550 const char *cp;
1551 struct origin *suspect = ent->suspect;
1552 char hex[41];
1553
1554 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1555 printf("%s%c%d %d %d\n",
1556 hex,
1557 ent->guilty ? ' ' : '*', // purely for debugging
1558 ent->s_lno + 1,
1559 ent->lno + 1,
1560 ent->num_lines);
1561 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1562 struct commit_info ci;
1563 suspect->commit->object.flags |= METAINFO_SHOWN;
1564 get_commit_info(suspect->commit, &ci, 1);
1565 printf("author %s\n", ci.author);
1566 printf("author-mail %s\n", ci.author_mail);
1567 printf("author-time %lu\n", ci.author_time);
1568 printf("author-tz %s\n", ci.author_tz);
1569 printf("committer %s\n", ci.committer);
1570 printf("committer-mail %s\n", ci.committer_mail);
1571 printf("committer-time %lu\n", ci.committer_time);
1572 printf("committer-tz %s\n", ci.committer_tz);
1573 write_filename_info(suspect->path);
1574 printf("summary %s\n", ci.summary);
1575 if (suspect->commit->object.flags & UNINTERESTING)
1576 printf("boundary\n");
1577 }
1578 else if (suspect->commit->object.flags & MORE_THAN_ONE_PATH)
1579 write_filename_info(suspect->path);
1580
1581 cp = nth_line(sb, ent->lno);
1582 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1583 char ch;
1584 if (cnt)
1585 printf("%s %d %d\n", hex,
1586 ent->s_lno + 1 + cnt,
1587 ent->lno + 1 + cnt);
1588 putchar('\t');
1589 do {
1590 ch = *cp++;
1591 putchar(ch);
1592 } while (ch != '\n' &&
1593 cp < sb->final_buf + sb->final_buf_size);
1594 }
1595}
1596
1597static void emit_other(struct scoreboard *sb, struct blame_entry *ent, int opt)
1598{
1599 int cnt;
1600 const char *cp;
1601 struct origin *suspect = ent->suspect;
1602 struct commit_info ci;
1603 char hex[41];
1604 int show_raw_time = !!(opt & OUTPUT_RAW_TIMESTAMP);
1605
1606 get_commit_info(suspect->commit, &ci, 1);
1607 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1608
1609 cp = nth_line(sb, ent->lno);
1610 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1611 char ch;
1612 int length = (opt & OUTPUT_LONG_OBJECT_NAME) ? 40 : 8;
1613
1614 if (suspect->commit->object.flags & UNINTERESTING) {
1615 if (blank_boundary)
1616 memset(hex, ' ', length);
1617 else if (!cmd_is_annotate) {
1618 length--;
1619 putchar('^');
1620 }
1621 }
1622
1623 printf("%.*s", length, hex);
1624 if (opt & OUTPUT_ANNOTATE_COMPAT)
1625 printf("\t(%10s\t%10s\t%d)", ci.author,
1626 format_time(ci.author_time, ci.author_tz,
1627 show_raw_time),
1628 ent->lno + 1 + cnt);
1629 else {
1630 if (opt & OUTPUT_SHOW_SCORE)
1631 printf(" %*d %02d",
1632 max_score_digits, ent->score,
1633 ent->suspect->refcnt);
1634 if (opt & OUTPUT_SHOW_NAME)
1635 printf(" %-*.*s", longest_file, longest_file,
1636 suspect->path);
1637 if (opt & OUTPUT_SHOW_NUMBER)
1638 printf(" %*d", max_orig_digits,
1639 ent->s_lno + 1 + cnt);
1640
1641 if (!(opt & OUTPUT_NO_AUTHOR))
1642 printf(" (%-*.*s %10s",
1643 longest_author, longest_author,
1644 ci.author,
1645 format_time(ci.author_time,
1646 ci.author_tz,
1647 show_raw_time));
1648 printf(" %*d) ",
1649 max_digits, ent->lno + 1 + cnt);
1650 }
1651 do {
1652 ch = *cp++;
1653 putchar(ch);
1654 } while (ch != '\n' &&
1655 cp < sb->final_buf + sb->final_buf_size);
1656 }
1657}
1658
1659static void output(struct scoreboard *sb, int option)
1660{
1661 struct blame_entry *ent;
1662
1663 if (option & OUTPUT_PORCELAIN) {
1664 for (ent = sb->ent; ent; ent = ent->next) {
1665 struct blame_entry *oth;
1666 struct origin *suspect = ent->suspect;
1667 struct commit *commit = suspect->commit;
1668 if (commit->object.flags & MORE_THAN_ONE_PATH)
1669 continue;
1670 for (oth = ent->next; oth; oth = oth->next) {
1671 if ((oth->suspect->commit != commit) ||
1672 !strcmp(oth->suspect->path, suspect->path))
1673 continue;
1674 commit->object.flags |= MORE_THAN_ONE_PATH;
1675 break;
1676 }
1677 }
1678 }
1679
1680 for (ent = sb->ent; ent; ent = ent->next) {
1681 if (option & OUTPUT_PORCELAIN)
1682 emit_porcelain(sb, ent);
1683 else {
1684 emit_other(sb, ent, option);
1685 }
1686 }
1687}
1688
1689/*
1690 * To allow quick access to the contents of nth line in the
1691 * final image, prepare an index in the scoreboard.
1692 */
1693static int prepare_lines(struct scoreboard *sb)
1694{
1695 const char *buf = sb->final_buf;
1696 unsigned long len = sb->final_buf_size;
1697 int num = 0, incomplete = 0, bol = 1;
1698
1699 if (len && buf[len-1] != '\n')
1700 incomplete++; /* incomplete line at the end */
1701 while (len--) {
1702 if (bol) {
1703 sb->lineno = xrealloc(sb->lineno,
1704 sizeof(int* ) * (num + 1));
1705 sb->lineno[num] = buf - sb->final_buf;
1706 bol = 0;
1707 }
1708 if (*buf++ == '\n') {
1709 num++;
1710 bol = 1;
1711 }
1712 }
1713 sb->lineno = xrealloc(sb->lineno,
1714 sizeof(int* ) * (num + incomplete + 1));
1715 sb->lineno[num + incomplete] = buf - sb->final_buf;
1716 sb->num_lines = num + incomplete;
1717 return sb->num_lines;
1718}
1719
1720/*
1721 * Add phony grafts for use with -S; this is primarily to
1722 * support git-cvsserver that wants to give a linear history
1723 * to its clients.
1724 */
1725static int read_ancestry(const char *graft_file)
1726{
1727 FILE *fp = fopen(graft_file, "r");
1728 char buf[1024];
1729 if (!fp)
1730 return -1;
1731 while (fgets(buf, sizeof(buf), fp)) {
1732 /* The format is just "Commit Parent1 Parent2 ...\n" */
1733 int len = strlen(buf);
1734 struct commit_graft *graft = read_graft_line(buf, len);
1735 if (graft)
1736 register_commit_graft(graft, 0);
1737 }
1738 fclose(fp);
1739 return 0;
1740}
1741
1742/*
1743 * How many columns do we need to show line numbers in decimal?
1744 */
1745static int lineno_width(int lines)
1746{
1747 int i, width;
1748
1749 for (width = 1, i = 10; i <= lines + 1; width++)
1750 i *= 10;
1751 return width;
1752}
1753
1754/*
1755 * How many columns do we need to show line numbers, authors,
1756 * and filenames?
1757 */
1758static void find_alignment(struct scoreboard *sb, int *option)
1759{
1760 int longest_src_lines = 0;
1761 int longest_dst_lines = 0;
1762 unsigned largest_score = 0;
1763 struct blame_entry *e;
1764
1765 for (e = sb->ent; e; e = e->next) {
1766 struct origin *suspect = e->suspect;
1767 struct commit_info ci;
1768 int num;
1769
1770 if (strcmp(suspect->path, sb->path))
1771 *option |= OUTPUT_SHOW_NAME;
1772 num = strlen(suspect->path);
1773 if (longest_file < num)
1774 longest_file = num;
1775 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1776 suspect->commit->object.flags |= METAINFO_SHOWN;
1777 get_commit_info(suspect->commit, &ci, 1);
1778 num = strlen(ci.author);
1779 if (longest_author < num)
1780 longest_author = num;
1781 }
1782 num = e->s_lno + e->num_lines;
1783 if (longest_src_lines < num)
1784 longest_src_lines = num;
1785 num = e->lno + e->num_lines;
1786 if (longest_dst_lines < num)
1787 longest_dst_lines = num;
1788 if (largest_score < ent_score(sb, e))
1789 largest_score = ent_score(sb, e);
1790 }
1791 max_orig_digits = lineno_width(longest_src_lines);
1792 max_digits = lineno_width(longest_dst_lines);
1793 max_score_digits = lineno_width(largest_score);
1794}
1795
1796/*
1797 * For debugging -- origin is refcounted, and this asserts that
1798 * we do not underflow.
1799 */
1800static void sanity_check_refcnt(struct scoreboard *sb)
1801{
1802 int baa = 0;
1803 struct blame_entry *ent;
1804
1805 for (ent = sb->ent; ent; ent = ent->next) {
1806 /* Nobody should have zero or negative refcnt */
1807 if (ent->suspect->refcnt <= 0) {
1808 fprintf(stderr, "%s in %s has negative refcnt %d\n",
1809 ent->suspect->path,
1810 sha1_to_hex(ent->suspect->commit->object.sha1),
1811 ent->suspect->refcnt);
1812 baa = 1;
1813 }
1814 }
1815 for (ent = sb->ent; ent; ent = ent->next) {
1816 /* Mark the ones that haven't been checked */
1817 if (0 < ent->suspect->refcnt)
1818 ent->suspect->refcnt = -ent->suspect->refcnt;
1819 }
1820 for (ent = sb->ent; ent; ent = ent->next) {
1821 /*
1822 * ... then pick each and see if they have the the
1823 * correct refcnt.
1824 */
1825 int found;
1826 struct blame_entry *e;
1827 struct origin *suspect = ent->suspect;
1828
1829 if (0 < suspect->refcnt)
1830 continue;
1831 suspect->refcnt = -suspect->refcnt; /* Unmark */
1832 for (found = 0, e = sb->ent; e; e = e->next) {
1833 if (e->suspect != suspect)
1834 continue;
1835 found++;
1836 }
1837 if (suspect->refcnt != found) {
1838 fprintf(stderr, "%s in %s has refcnt %d, not %d\n",
1839 ent->suspect->path,
1840 sha1_to_hex(ent->suspect->commit->object.sha1),
1841 ent->suspect->refcnt, found);
1842 baa = 2;
1843 }
1844 }
1845 if (baa) {
1846 int opt = 0160;
1847 find_alignment(sb, &opt);
1848 output(sb, opt);
1849 die("Baa %d!", baa);
1850 }
1851}
1852
1853/*
1854 * Used for the command line parsing; check if the path exists
1855 * in the working tree.
1856 */
1857static int has_path_in_work_tree(const char *path)
1858{
1859 struct stat st;
1860 return !lstat(path, &st);
1861}
1862
1863static unsigned parse_score(const char *arg)
1864{
1865 char *end;
1866 unsigned long score = strtoul(arg, &end, 10);
1867 if (*end)
1868 return 0;
1869 return score;
1870}
1871
1872static const char *add_prefix(const char *prefix, const char *path)
1873{
1874 if (!prefix || !prefix[0])
1875 return path;
1876 return prefix_path(prefix, strlen(prefix), path);
1877}
1878
1879/*
1880 * Parsing of (comma separated) one item in the -L option
1881 */
1882static const char *parse_loc(const char *spec,
1883 struct scoreboard *sb, long lno,
1884 long begin, long *ret)
1885{
1886 char *term;
1887 const char *line;
1888 long num;
1889 int reg_error;
1890 regex_t regexp;
1891 regmatch_t match[1];
1892
1893 /* Allow "-L <something>,+20" to mean starting at <something>
1894 * for 20 lines, or "-L <something>,-5" for 5 lines ending at
1895 * <something>.
1896 */
1897 if (1 < begin && (spec[0] == '+' || spec[0] == '-')) {
1898 num = strtol(spec + 1, &term, 10);
1899 if (term != spec + 1) {
1900 if (spec[0] == '-')
1901 num = 0 - num;
1902 if (0 < num)
1903 *ret = begin + num - 2;
1904 else if (!num)
1905 *ret = begin;
1906 else
1907 *ret = begin + num;
1908 return term;
1909 }
1910 return spec;
1911 }
1912 num = strtol(spec, &term, 10);
1913 if (term != spec) {
1914 *ret = num;
1915 return term;
1916 }
1917 if (spec[0] != '/')
1918 return spec;
1919
1920 /* it could be a regexp of form /.../ */
1921 for (term = (char*) spec + 1; *term && *term != '/'; term++) {
1922 if (*term == '\\')
1923 term++;
1924 }
1925 if (*term != '/')
1926 return spec;
1927
1928 /* try [spec+1 .. term-1] as regexp */
1929 *term = 0;
1930 begin--; /* input is in human terms */
1931 line = nth_line(sb, begin);
1932
1933 if (!(reg_error = regcomp(®exp, spec + 1, REG_NEWLINE)) &&
1934 !(reg_error = regexec(®exp, line, 1, match, 0))) {
1935 const char *cp = line + match[0].rm_so;
1936 const char *nline;
1937
1938 while (begin++ < lno) {
1939 nline = nth_line(sb, begin);
1940 if (line <= cp && cp < nline)
1941 break;
1942 line = nline;
1943 }
1944 *ret = begin;
1945 regfree(®exp);
1946 *term++ = '/';
1947 return term;
1948 }
1949 else {
1950 char errbuf[1024];
1951 regerror(reg_error, ®exp, errbuf, 1024);
1952 die("-L parameter '%s': %s", spec + 1, errbuf);
1953 }
1954}
1955
1956/*
1957 * Parsing of -L option
1958 */
1959static void prepare_blame_range(struct scoreboard *sb,
1960 const char *bottomtop,
1961 long lno,
1962 long *bottom, long *top)
1963{
1964 const char *term;
1965
1966 term = parse_loc(bottomtop, sb, lno, 1, bottom);
1967 if (*term == ',') {
1968 term = parse_loc(term + 1, sb, lno, *bottom + 1, top);
1969 if (*term)
1970 usage(blame_usage);
1971 }
1972 if (*term)
1973 usage(blame_usage);
1974}
1975
1976static int git_blame_config(const char *var, const char *value)
1977{
1978 if (!strcmp(var, "blame.showroot")) {
1979 show_root = git_config_bool(var, value);
1980 return 0;
1981 }
1982 if (!strcmp(var, "blame.blankboundary")) {
1983 blank_boundary = git_config_bool(var, value);
1984 return 0;
1985 }
1986 return git_default_config(var, value);
1987}
1988
1989static struct commit *fake_working_tree_commit(const char *path, const char *contents_from)
1990{
1991 struct commit *commit;
1992 struct origin *origin;
1993 unsigned char head_sha1[20];
1994 char *buf;
1995 const char *ident;
1996 int fd;
1997 time_t now;
1998 unsigned long fin_size;
1999 int size, len;
2000 struct cache_entry *ce;
2001 unsigned mode;
2002
2003 if (get_sha1("HEAD", head_sha1))
2004 die("No such ref: HEAD");
2005
2006 time(&now);
2007 commit = xcalloc(1, sizeof(*commit));
2008 commit->parents = xcalloc(1, sizeof(*commit->parents));
2009 commit->parents->item = lookup_commit_reference(head_sha1);
2010 commit->object.parsed = 1;
2011 commit->date = now;
2012 commit->object.type = OBJ_COMMIT;
2013
2014 origin = make_origin(commit, path);
2015
2016 if (!contents_from || strcmp("-", contents_from)) {
2017 struct stat st;
2018 const char *read_from;
2019
2020 if (contents_from) {
2021 if (stat(contents_from, &st) < 0)
2022 die("Cannot stat %s", contents_from);
2023 read_from = contents_from;
2024 }
2025 else {
2026 if (lstat(path, &st) < 0)
2027 die("Cannot lstat %s", path);
2028 read_from = path;
2029 }
2030 fin_size = xsize_t(st.st_size);
2031 buf = xmalloc(fin_size+1);
2032 mode = canon_mode(st.st_mode);
2033 switch (st.st_mode & S_IFMT) {
2034 case S_IFREG:
2035 fd = open(read_from, O_RDONLY);
2036 if (fd < 0)
2037 die("cannot open %s", read_from);
2038 if (read_in_full(fd, buf, fin_size) != fin_size)
2039 die("cannot read %s", read_from);
2040 break;
2041 case S_IFLNK:
2042 if (readlink(read_from, buf, fin_size+1) != fin_size)
2043 die("cannot readlink %s", read_from);
2044 break;
2045 default:
2046 die("unsupported file type %s", read_from);
2047 }
2048 }
2049 else {
2050 /* Reading from stdin */
2051 contents_from = "standard input";
2052 buf = NULL;
2053 fin_size = 0;
2054 mode = 0;
2055 while (1) {
2056 ssize_t cnt = 8192;
2057 buf = xrealloc(buf, fin_size + cnt);
2058 cnt = xread(0, buf + fin_size, cnt);
2059 if (cnt < 0)
2060 die("read error %s from stdin",
2061 strerror(errno));
2062 if (!cnt)
2063 break;
2064 fin_size += cnt;
2065 }
2066 buf = xrealloc(buf, fin_size + 1);
2067 }
2068 buf[fin_size] = 0;
2069 origin->file.ptr = buf;
2070 origin->file.size = fin_size;
2071 pretend_sha1_file(buf, fin_size, OBJ_BLOB, origin->blob_sha1);
2072 commit->util = origin;
2073
2074 /*
2075 * Read the current index, replace the path entry with
2076 * origin->blob_sha1 without mucking with its mode or type
2077 * bits; we are not going to write this index out -- we just
2078 * want to run "diff-index --cached".
2079 */
2080 discard_cache();
2081 read_cache();
2082
2083 len = strlen(path);
2084 if (!mode) {
2085 int pos = cache_name_pos(path, len);
2086 if (0 <= pos)
2087 mode = ntohl(active_cache[pos]->ce_mode);
2088 else
2089 /* Let's not bother reading from HEAD tree */
2090 mode = S_IFREG | 0644;
2091 }
2092 size = cache_entry_size(len);
2093 ce = xcalloc(1, size);
2094 hashcpy(ce->sha1, origin->blob_sha1);
2095 memcpy(ce->name, path, len);
2096 ce->ce_flags = create_ce_flags(len, 0);
2097 ce->ce_mode = create_ce_mode(mode);
2098 add_cache_entry(ce, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
2099
2100 /*
2101 * We are not going to write this out, so this does not matter
2102 * right now, but someday we might optimize diff-index --cached
2103 * with cache-tree information.
2104 */
2105 cache_tree_invalidate_path(active_cache_tree, path);
2106
2107 commit->buffer = xmalloc(400);
2108 ident = fmt_ident("Not Committed Yet", "not.committed.yet", NULL, 0);
2109 snprintf(commit->buffer, 400,
2110 "tree 0000000000000000000000000000000000000000\n"
2111 "parent %s\n"
2112 "author %s\n"
2113 "committer %s\n\n"
2114 "Version of %s from %s\n",
2115 sha1_to_hex(head_sha1),
2116 ident, ident, path, contents_from ? contents_from : path);
2117 return commit;
2118}
2119
2120int cmd_blame(int argc, const char **argv, const char *prefix)
2121{
2122 struct rev_info revs;
2123 const char *path;
2124 struct scoreboard sb;
2125 struct origin *o;
2126 struct blame_entry *ent;
2127 int i, seen_dashdash, unk, opt;
2128 long bottom, top, lno;
2129 int output_option = 0;
2130 int show_stats = 0;
2131 const char *revs_file = NULL;
2132 const char *final_commit_name = NULL;
2133 enum object_type type;
2134 const char *bottomtop = NULL;
2135 const char *contents_from = NULL;
2136
2137 cmd_is_annotate = !strcmp(argv[0], "annotate");
2138
2139 git_config(git_blame_config);
2140 save_commit_buffer = 0;
2141
2142 opt = 0;
2143 seen_dashdash = 0;
2144 for (unk = i = 1; i < argc; i++) {
2145 const char *arg = argv[i];
2146 if (*arg != '-')
2147 break;
2148 else if (!strcmp("-b", arg))
2149 blank_boundary = 1;
2150 else if (!strcmp("--root", arg))
2151 show_root = 1;
2152 else if (!strcmp(arg, "--show-stats"))
2153 show_stats = 1;
2154 else if (!strcmp("-c", arg))
2155 output_option |= OUTPUT_ANNOTATE_COMPAT;
2156 else if (!strcmp("-t", arg))
2157 output_option |= OUTPUT_RAW_TIMESTAMP;
2158 else if (!strcmp("-l", arg))
2159 output_option |= OUTPUT_LONG_OBJECT_NAME;
2160 else if (!strcmp("-s", arg))
2161 output_option |= OUTPUT_NO_AUTHOR;
2162 else if (!strcmp("-S", arg) && ++i < argc)
2163 revs_file = argv[i];
2164 else if (!prefixcmp(arg, "-M")) {
2165 opt |= PICKAXE_BLAME_MOVE;
2166 blame_move_score = parse_score(arg+2);
2167 }
2168 else if (!prefixcmp(arg, "-C")) {
2169 /*
2170 * -C enables copy from removed files;
2171 * -C -C enables copy from existing files, but only
2172 * when blaming a new file;
2173 * -C -C -C enables copy from existing files for
2174 * everybody
2175 */
2176 if (opt & PICKAXE_BLAME_COPY_HARDER)
2177 opt |= PICKAXE_BLAME_COPY_HARDEST;
2178 if (opt & PICKAXE_BLAME_COPY)
2179 opt |= PICKAXE_BLAME_COPY_HARDER;
2180 opt |= PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE;
2181 blame_copy_score = parse_score(arg+2);
2182 }
2183 else if (!prefixcmp(arg, "-L")) {
2184 if (!arg[2]) {
2185 if (++i >= argc)
2186 usage(blame_usage);
2187 arg = argv[i];
2188 }
2189 else
2190 arg += 2;
2191 if (bottomtop)
2192 die("More than one '-L n,m' option given");
2193 bottomtop = arg;
2194 }
2195 else if (!strcmp("--contents", arg)) {
2196 if (++i >= argc)
2197 usage(blame_usage);
2198 contents_from = argv[i];
2199 }
2200 else if (!strcmp("--incremental", arg))
2201 incremental = 1;
2202 else if (!strcmp("--score-debug", arg))
2203 output_option |= OUTPUT_SHOW_SCORE;
2204 else if (!strcmp("-f", arg) ||
2205 !strcmp("--show-name", arg))
2206 output_option |= OUTPUT_SHOW_NAME;
2207 else if (!strcmp("-n", arg) ||
2208 !strcmp("--show-number", arg))
2209 output_option |= OUTPUT_SHOW_NUMBER;
2210 else if (!strcmp("-p", arg) ||
2211 !strcmp("--porcelain", arg))
2212 output_option |= OUTPUT_PORCELAIN;
2213 else if (!strcmp("--", arg)) {
2214 seen_dashdash = 1;
2215 i++;
2216 break;
2217 }
2218 else
2219 argv[unk++] = arg;
2220 }
2221
2222 if (!incremental)
2223 setup_pager();
2224
2225 if (!blame_move_score)
2226 blame_move_score = BLAME_DEFAULT_MOVE_SCORE;
2227 if (!blame_copy_score)
2228 blame_copy_score = BLAME_DEFAULT_COPY_SCORE;
2229
2230 /*
2231 * We have collected options unknown to us in argv[1..unk]
2232 * which are to be passed to revision machinery if we are
2233 * going to do the "bottom" processing.
2234 *
2235 * The remaining are:
2236 *
2237 * (1) if seen_dashdash, its either
2238 * "-options -- <path>" or
2239 * "-options -- <path> <rev>".
2240 * but the latter is allowed only if there is no
2241 * options that we passed to revision machinery.
2242 *
2243 * (2) otherwise, we may have "--" somewhere later and
2244 * might be looking at the first one of multiple 'rev'
2245 * parameters (e.g. " master ^next ^maint -- path").
2246 * See if there is a dashdash first, and give the
2247 * arguments before that to revision machinery.
2248 * After that there must be one 'path'.
2249 *
2250 * (3) otherwise, its one of the three:
2251 * "-options <path> <rev>"
2252 * "-options <rev> <path>"
2253 * "-options <path>"
2254 * but again the first one is allowed only if
2255 * there is no options that we passed to revision
2256 * machinery.
2257 */
2258
2259 if (seen_dashdash) {
2260 /* (1) */
2261 if (argc <= i)
2262 usage(blame_usage);
2263 path = add_prefix(prefix, argv[i]);
2264 if (i + 1 == argc - 1) {
2265 if (unk != 1)
2266 usage(blame_usage);
2267 argv[unk++] = argv[i + 1];
2268 }
2269 else if (i + 1 != argc)
2270 /* garbage at end */
2271 usage(blame_usage);
2272 }
2273 else {
2274 int j;
2275 for (j = i; !seen_dashdash && j < argc; j++)
2276 if (!strcmp(argv[j], "--"))
2277 seen_dashdash = j;
2278 if (seen_dashdash) {
2279 /* (2) */
2280 if (seen_dashdash + 1 != argc - 1)
2281 usage(blame_usage);
2282 path = add_prefix(prefix, argv[seen_dashdash + 1]);
2283 for (j = i; j < seen_dashdash; j++)
2284 argv[unk++] = argv[j];
2285 }
2286 else {
2287 /* (3) */
2288 if (argc <= i)
2289 usage(blame_usage);
2290 path = add_prefix(prefix, argv[i]);
2291 if (i + 1 == argc - 1) {
2292 final_commit_name = argv[i + 1];
2293
2294 /* if (unk == 1) we could be getting
2295 * old-style
2296 */
2297 if (unk == 1 && !has_path_in_work_tree(path)) {
2298 path = add_prefix(prefix, argv[i + 1]);
2299 final_commit_name = argv[i];
2300 }
2301 }
2302 else if (i != argc - 1)
2303 usage(blame_usage); /* garbage at end */
2304
2305 if (!has_path_in_work_tree(path))
2306 die("cannot stat path %s: %s",
2307 path, strerror(errno));
2308 }
2309 }
2310
2311 if (final_commit_name)
2312 argv[unk++] = final_commit_name;
2313
2314 /*
2315 * Now we got rev and path. We do not want the path pruning
2316 * but we may want "bottom" processing.
2317 */
2318 argv[unk++] = "--"; /* terminate the rev name */
2319 argv[unk] = NULL;
2320
2321 init_revisions(&revs, NULL);
2322 setup_revisions(unk, argv, &revs, NULL);
2323 memset(&sb, 0, sizeof(sb));
2324
2325 /*
2326 * There must be one and only one positive commit in the
2327 * revs->pending array.
2328 */
2329 for (i = 0; i < revs.pending.nr; i++) {
2330 struct object *obj = revs.pending.objects[i].item;
2331 if (obj->flags & UNINTERESTING)
2332 continue;
2333 while (obj->type == OBJ_TAG)
2334 obj = deref_tag(obj, NULL, 0);
2335 if (obj->type != OBJ_COMMIT)
2336 die("Non commit %s?",
2337 revs.pending.objects[i].name);
2338 if (sb.final)
2339 die("More than one commit to dig from %s and %s?",
2340 revs.pending.objects[i].name,
2341 final_commit_name);
2342 sb.final = (struct commit *) obj;
2343 final_commit_name = revs.pending.objects[i].name;
2344 }
2345
2346 if (!sb.final) {
2347 /*
2348 * "--not A B -- path" without anything positive;
2349 * do not default to HEAD, but use the working tree
2350 * or "--contents".
2351 */
2352 sb.final = fake_working_tree_commit(path, contents_from);
2353 add_pending_object(&revs, &(sb.final->object), ":");
2354 }
2355 else if (contents_from)
2356 die("Cannot use --contents with final commit object name");
2357
2358 /*
2359 * If we have bottom, this will mark the ancestors of the
2360 * bottom commits we would reach while traversing as
2361 * uninteresting.
2362 */
2363 prepare_revision_walk(&revs);
2364
2365 if (is_null_sha1(sb.final->object.sha1)) {
2366 char *buf;
2367 o = sb.final->util;
2368 buf = xmalloc(o->file.size + 1);
2369 memcpy(buf, o->file.ptr, o->file.size + 1);
2370 sb.final_buf = buf;
2371 sb.final_buf_size = o->file.size;
2372 }
2373 else {
2374 o = get_origin(&sb, sb.final, path);
2375 if (fill_blob_sha1(o))
2376 die("no such path %s in %s", path, final_commit_name);
2377
2378 sb.final_buf = read_sha1_file(o->blob_sha1, &type,
2379 &sb.final_buf_size);
2380 }
2381 num_read_blob++;
2382 lno = prepare_lines(&sb);
2383
2384 bottom = top = 0;
2385 if (bottomtop)
2386 prepare_blame_range(&sb, bottomtop, lno, &bottom, &top);
2387 if (bottom && top && top < bottom) {
2388 long tmp;
2389 tmp = top; top = bottom; bottom = tmp;
2390 }
2391 if (bottom < 1)
2392 bottom = 1;
2393 if (top < 1)
2394 top = lno;
2395 bottom--;
2396 if (lno < top)
2397 die("file %s has only %lu lines", path, lno);
2398
2399 ent = xcalloc(1, sizeof(*ent));
2400 ent->lno = bottom;
2401 ent->num_lines = top - bottom;
2402 ent->suspect = o;
2403 ent->s_lno = bottom;
2404
2405 sb.ent = ent;
2406 sb.path = path;
2407
2408 if (revs_file && read_ancestry(revs_file))
2409 die("reading graft file %s failed: %s",
2410 revs_file, strerror(errno));
2411
2412 read_mailmap(&mailmap, ".mailmap", NULL);
2413
2414 assign_blame(&sb, &revs, opt);
2415
2416 if (incremental)
2417 return 0;
2418
2419 coalesce(&sb);
2420
2421 if (!(output_option & OUTPUT_PORCELAIN))
2422 find_alignment(&sb, &output_option);
2423
2424 output(&sb, output_option);
2425 free((void *)sb.final_buf);
2426 for (ent = sb.ent; ent; ) {
2427 struct blame_entry *e = ent->next;
2428 free(ent);
2429 ent = e;
2430 }
2431
2432 if (show_stats) {
2433 printf("num read blob: %d\n", num_read_blob);
2434 printf("num get patch: %d\n", num_get_patch);
2435 printf("num commits: %d\n", num_commits);
2436 }
2437 return 0;
2438}