4 * Copyright (c) 2006, Junio C Hamano
12 #include "tree-walk.h"
17 #include "xdiff-interface.h"
18 #include "cache-tree.h"
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>] [commit] [--] file\n"
22 " -c, --compatibility 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, --long Show long commit SHA1 (Default: off)\n"
25 " --root Do not treat root commits as boundaries (Default: off)\n"
26 " -t, --time Show raw timestamp (Default: off)\n"
27 " -f, --show-name Show original filename (Default: auto)\n"
28 " -n, --show-number Show original linenumber (Default: off)\n"
29 " -p, --porcelain Show in a format designed for machine consumption\n"
30 " -L n,m Process only line range n,m, counting from 1\n"
31 " -M, -C Find line movements within and across files\n"
32 " --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";
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
;
42 static int blank_boundary
;
43 static int incremental
;
50 static int num_read_blob
;
51 static int num_get_patch
;
52 static int num_commits
;
54 #define PICKAXE_BLAME_MOVE 01
55 #define PICKAXE_BLAME_COPY 02
56 #define PICKAXE_BLAME_COPY_HARDER 04
59 * blame for a blame_entry with score lower than these thresholds
60 * is not passed to the parent using move/copy logic.
62 static unsigned blame_move_score
;
63 static unsigned blame_copy_score
;
64 #define BLAME_DEFAULT_MOVE_SCORE 20
65 #define BLAME_DEFAULT_COPY_SCORE 40
67 /* bits #0..7 in revision.h, #8..11 used for merge_bases() in commit.c */
68 #define METAINFO_SHOWN (1u<<12)
69 #define MORE_THAN_ONE_PATH (1u<<13)
72 * One blob in a commit that is being suspected
76 struct commit
*commit
;
78 unsigned char blob_sha1
[20];
79 char path
[FLEX_ARRAY
];
83 * Given an origin, prepare mmfile_t structure to be used by the
86 static char *fill_origin_blob(struct origin
*o
, mmfile_t
*file
)
91 file
->ptr
= read_sha1_file(o
->blob_sha1
, type
,
92 (unsigned long *)(&(file
->size
)));
101 * Origin is refcounted and usually we keep the blob contents to be
104 static inline struct origin
*origin_incref(struct origin
*o
)
111 static void origin_decref(struct origin
*o
)
113 if (o
&& --o
->refcnt
<= 0) {
116 memset(o
, 0, sizeof(*o
));
122 * Each group of lines is described by a blame_entry; it can be split
123 * as we pass blame to the parents. They form a linked list in the
124 * scoreboard structure, sorted by the target line number.
127 struct blame_entry
*prev
;
128 struct blame_entry
*next
;
130 /* the first line of this group in the final image;
131 * internally all line numbers are 0 based.
135 /* how many lines this group has */
138 /* the commit that introduced this group into the final image */
139 struct origin
*suspect
;
141 /* true if the suspect is truly guilty; false while we have not
142 * checked if the group came from one of its parents.
146 /* the line number of the first line of this group in the
147 * suspect's file; internally all line numbers are 0 based.
151 /* how significant this entry is -- cached to avoid
152 * scanning the lines over and over.
158 * The current state of the blame assignment.
161 /* the final commit (i.e. where we started digging from) */
162 struct commit
*final
;
167 * The contents in the final image.
168 * Used by many functions to obtain contents of the nth line,
169 * indexed with scoreboard.lineno[blame_entry.lno].
171 const char *final_buf
;
172 unsigned long final_buf_size
;
174 /* linked list of blames */
175 struct blame_entry
*ent
;
177 /* look-up a line in the final buffer */
182 static int cmp_suspect(struct origin
*a
, struct origin
*b
)
184 int cmp
= hashcmp(a
->commit
->object
.sha1
, b
->commit
->object
.sha1
);
187 return strcmp(a
->path
, b
->path
);
190 #define cmp_suspect(a, b) ( ((a)==(b)) ? 0 : cmp_suspect(a,b) )
192 static void sanity_check_refcnt(struct scoreboard
*);
195 * If two blame entries that are next to each other came from
196 * contiguous lines in the same origin (i.e. <commit, path> pair),
197 * merge them together.
199 static void coalesce(struct scoreboard
*sb
)
201 struct blame_entry
*ent
, *next
;
203 for (ent
= sb
->ent
; ent
&& (next
= ent
->next
); ent
= next
) {
204 if (!cmp_suspect(ent
->suspect
, next
->suspect
) &&
205 ent
->guilty
== next
->guilty
&&
206 ent
->s_lno
+ ent
->num_lines
== next
->s_lno
) {
207 ent
->num_lines
+= next
->num_lines
;
208 ent
->next
= next
->next
;
210 ent
->next
->prev
= ent
;
211 origin_decref(next
->suspect
);
214 next
= ent
; /* again */
218 if (DEBUG
) /* sanity */
219 sanity_check_refcnt(sb
);
223 * Given a commit and a path in it, create a new origin structure.
224 * The callers that add blame to the scoreboard should use
225 * get_origin() to obtain shared, refcounted copy instead of calling
226 * this function directly.
228 static struct origin
*make_origin(struct commit
*commit
, const char *path
)
231 o
= xcalloc(1, sizeof(*o
) + strlen(path
) + 1);
234 strcpy(o
->path
, path
);
239 * Locate an existing origin or create a new one.
241 static struct origin
*get_origin(struct scoreboard
*sb
,
242 struct commit
*commit
,
245 struct blame_entry
*e
;
247 for (e
= sb
->ent
; e
; e
= e
->next
) {
248 if (e
->suspect
->commit
== commit
&&
249 !strcmp(e
->suspect
->path
, path
))
250 return origin_incref(e
->suspect
);
252 return make_origin(commit
, path
);
256 * Fill the blob_sha1 field of an origin if it hasn't, so that later
257 * call to fill_origin_blob() can use it to locate the data. blob_sha1
258 * for an origin is also used to pass the blame for the entire file to
259 * the parent to detect the case where a child's blob is identical to
260 * that of its parent's.
262 static int fill_blob_sha1(struct origin
*origin
)
267 if (!is_null_sha1(origin
->blob_sha1
))
269 if (get_tree_entry(origin
->commit
->object
.sha1
,
271 origin
->blob_sha1
, &mode
))
273 if (sha1_object_info(origin
->blob_sha1
, type
, NULL
) ||
274 strcmp(type
, blob_type
))
278 hashclr(origin
->blob_sha1
);
283 * We have an origin -- check if the same path exists in the
284 * parent and return an origin structure to represent it.
286 static struct origin
*find_origin(struct scoreboard
*sb
,
287 struct commit
*parent
,
288 struct origin
*origin
)
290 struct origin
*porigin
= NULL
;
291 struct diff_options diff_opts
;
292 const char *paths
[2];
296 * Each commit object can cache one origin in that
297 * commit. This is a freestanding copy of origin and
300 struct origin
*cached
= parent
->util
;
301 if (!strcmp(cached
->path
, origin
->path
)) {
303 * The same path between origin and its parent
304 * without renaming -- the most common case.
306 porigin
= get_origin(sb
, parent
, cached
->path
);
309 * If the origin was newly created (i.e. get_origin
310 * would call make_origin if none is found in the
311 * scoreboard), it does not know the blob_sha1,
312 * so copy it. Otherwise porigin was in the
313 * scoreboard and already knows blob_sha1.
315 if (porigin
->refcnt
== 1)
316 hashcpy(porigin
->blob_sha1
, cached
->blob_sha1
);
319 /* otherwise it was not very useful; free it */
324 /* See if the origin->path is different between parent
325 * and origin first. Most of the time they are the
326 * same and diff-tree is fairly efficient about this.
328 diff_setup(&diff_opts
);
329 diff_opts
.recursive
= 1;
330 diff_opts
.detect_rename
= 0;
331 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
332 paths
[0] = origin
->path
;
335 diff_tree_setup_paths(paths
, &diff_opts
);
336 if (diff_setup_done(&diff_opts
) < 0)
339 if (is_null_sha1(origin
->commit
->object
.sha1
))
340 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
342 diff_tree_sha1(parent
->tree
->object
.sha1
,
343 origin
->commit
->tree
->object
.sha1
,
345 diffcore_std(&diff_opts
);
347 /* It is either one entry that says "modified", or "created",
350 if (!diff_queued_diff
.nr
) {
351 /* The path is the same as parent */
352 porigin
= get_origin(sb
, parent
, origin
->path
);
353 hashcpy(porigin
->blob_sha1
, origin
->blob_sha1
);
355 else if (diff_queued_diff
.nr
!= 1)
356 die("internal error in blame::find_origin");
358 struct diff_filepair
*p
= diff_queued_diff
.queue
[0];
361 die("internal error in blame::find_origin (%c)",
364 porigin
= get_origin(sb
, parent
, origin
->path
);
365 hashcpy(porigin
->blob_sha1
, p
->one
->sha1
);
369 /* Did not exist in parent, or type changed */
373 diff_flush(&diff_opts
);
376 * Create a freestanding copy that is not part of
377 * the refcounted origin found in the scoreboard, and
378 * cache it in the commit.
380 struct origin
*cached
;
382 cached
= make_origin(porigin
->commit
, porigin
->path
);
383 hashcpy(cached
->blob_sha1
, porigin
->blob_sha1
);
384 parent
->util
= cached
;
390 * We have an origin -- find the path that corresponds to it in its
391 * parent and return an origin structure to represent it.
393 static struct origin
*find_rename(struct scoreboard
*sb
,
394 struct commit
*parent
,
395 struct origin
*origin
)
397 struct origin
*porigin
= NULL
;
398 struct diff_options diff_opts
;
400 const char *paths
[2];
402 diff_setup(&diff_opts
);
403 diff_opts
.recursive
= 1;
404 diff_opts
.detect_rename
= DIFF_DETECT_RENAME
;
405 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
406 diff_opts
.single_follow
= origin
->path
;
408 diff_tree_setup_paths(paths
, &diff_opts
);
409 if (diff_setup_done(&diff_opts
) < 0)
412 if (is_null_sha1(origin
->commit
->object
.sha1
))
413 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
415 diff_tree_sha1(parent
->tree
->object
.sha1
,
416 origin
->commit
->tree
->object
.sha1
,
418 diffcore_std(&diff_opts
);
420 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
421 struct diff_filepair
*p
= diff_queued_diff
.queue
[i
];
422 if ((p
->status
== 'R' || p
->status
== 'C') &&
423 !strcmp(p
->two
->path
, origin
->path
)) {
424 porigin
= get_origin(sb
, parent
, p
->one
->path
);
425 hashcpy(porigin
->blob_sha1
, p
->one
->sha1
);
429 diff_flush(&diff_opts
);
434 * Parsing of patch chunks...
437 /* line number in postimage; up to but not including this
438 * line is the same as preimage
442 /* preimage line number after this chunk */
445 /* postimage line number after this chunk */
450 struct chunk
*chunks
;
454 struct blame_diff_state
{
455 struct xdiff_emit_state xm
;
457 unsigned hunk_post_context
;
458 unsigned hunk_in_pre_context
: 1;
461 static void process_u_diff(void *state_
, char *line
, unsigned long len
)
463 struct blame_diff_state
*state
= state_
;
465 int off1
, off2
, len1
, len2
, num
;
467 num
= state
->ret
->num
;
468 if (len
< 4 || line
[0] != '@' || line
[1] != '@') {
469 if (state
->hunk_in_pre_context
&& line
[0] == ' ')
470 state
->ret
->chunks
[num
- 1].same
++;
472 state
->hunk_in_pre_context
= 0;
474 state
->hunk_post_context
++;
476 state
->hunk_post_context
= 0;
481 if (num
&& state
->hunk_post_context
) {
482 chunk
= &state
->ret
->chunks
[num
- 1];
483 chunk
->p_next
-= state
->hunk_post_context
;
484 chunk
->t_next
-= state
->hunk_post_context
;
486 state
->ret
->num
= ++num
;
487 state
->ret
->chunks
= xrealloc(state
->ret
->chunks
,
488 sizeof(struct chunk
) * num
);
489 chunk
= &state
->ret
->chunks
[num
- 1];
490 if (parse_hunk_header(line
, len
, &off1
, &len1
, &off2
, &len2
)) {
495 /* Line numbers in patch output are one based. */
499 chunk
->same
= len2
? off2
: (off2
+ 1);
501 chunk
->p_next
= off1
+ (len1
? len1
: 1);
502 chunk
->t_next
= chunk
->same
+ len2
;
503 state
->hunk_in_pre_context
= 1;
504 state
->hunk_post_context
= 0;
507 static struct patch
*compare_buffer(mmfile_t
*file_p
, mmfile_t
*file_o
,
510 struct blame_diff_state state
;
515 xpp
.flags
= XDF_NEED_MINIMAL
;
516 xecfg
.ctxlen
= context
;
518 ecb
.outf
= xdiff_outf
;
520 memset(&state
, 0, sizeof(state
));
521 state
.xm
.consume
= process_u_diff
;
522 state
.ret
= xmalloc(sizeof(struct patch
));
523 state
.ret
->chunks
= NULL
;
526 xdl_diff(file_p
, file_o
, &xpp
, &xecfg
, &ecb
);
528 if (state
.ret
->num
) {
530 chunk
= &state
.ret
->chunks
[state
.ret
->num
- 1];
531 chunk
->p_next
-= state
.hunk_post_context
;
532 chunk
->t_next
-= state
.hunk_post_context
;
538 * Run diff between two origins and grab the patch output, so that
539 * we can pass blame for lines origin is currently suspected for
542 static struct patch
*get_patch(struct origin
*parent
, struct origin
*origin
)
544 mmfile_t file_p
, file_o
;
547 fill_origin_blob(parent
, &file_p
);
548 fill_origin_blob(origin
, &file_o
);
549 if (!file_p
.ptr
|| !file_o
.ptr
)
551 patch
= compare_buffer(&file_p
, &file_o
, 0);
556 static void free_patch(struct patch
*p
)
563 * Link in a new blame entry to the scoreboard. Entries that cover the
564 * same line range have been removed from the scoreboard previously.
566 static void add_blame_entry(struct scoreboard
*sb
, struct blame_entry
*e
)
568 struct blame_entry
*ent
, *prev
= NULL
;
570 origin_incref(e
->suspect
);
572 for (ent
= sb
->ent
; ent
&& ent
->lno
< e
->lno
; ent
= ent
->next
)
575 /* prev, if not NULL, is the last one that is below e */
578 e
->next
= prev
->next
;
590 * src typically is on-stack; we want to copy the information in it to
591 * an malloced blame_entry that is already on the linked list of the
592 * scoreboard. The origin of dst loses a refcnt while the origin of src
595 static void dup_entry(struct blame_entry
*dst
, struct blame_entry
*src
)
597 struct blame_entry
*p
, *n
;
601 origin_incref(src
->suspect
);
602 origin_decref(dst
->suspect
);
603 memcpy(dst
, src
, sizeof(*src
));
609 static const char *nth_line(struct scoreboard
*sb
, int lno
)
611 return sb
->final_buf
+ sb
->lineno
[lno
];
615 * It is known that lines between tlno to same came from parent, and e
616 * has an overlap with that range. it also is known that parent's
617 * line plno corresponds to e's line tlno.
623 * <------------------>
625 * Split e into potentially three parts; before this chunk, the chunk
626 * to be blamed for the parent, and after that portion.
628 static void split_overlap(struct blame_entry
*split
,
629 struct blame_entry
*e
,
630 int tlno
, int plno
, int same
,
631 struct origin
*parent
)
634 memset(split
, 0, sizeof(struct blame_entry
[3]));
636 if (e
->s_lno
< tlno
) {
637 /* there is a pre-chunk part not blamed on parent */
638 split
[0].suspect
= origin_incref(e
->suspect
);
639 split
[0].lno
= e
->lno
;
640 split
[0].s_lno
= e
->s_lno
;
641 split
[0].num_lines
= tlno
- e
->s_lno
;
642 split
[1].lno
= e
->lno
+ tlno
- e
->s_lno
;
643 split
[1].s_lno
= plno
;
646 split
[1].lno
= e
->lno
;
647 split
[1].s_lno
= plno
+ (e
->s_lno
- tlno
);
650 if (same
< e
->s_lno
+ e
->num_lines
) {
651 /* there is a post-chunk part not blamed on parent */
652 split
[2].suspect
= origin_incref(e
->suspect
);
653 split
[2].lno
= e
->lno
+ (same
- e
->s_lno
);
654 split
[2].s_lno
= e
->s_lno
+ (same
- e
->s_lno
);
655 split
[2].num_lines
= e
->s_lno
+ e
->num_lines
- same
;
656 chunk_end_lno
= split
[2].lno
;
659 chunk_end_lno
= e
->lno
+ e
->num_lines
;
660 split
[1].num_lines
= chunk_end_lno
- split
[1].lno
;
663 * if it turns out there is nothing to blame the parent for,
664 * forget about the splitting. !split[1].suspect signals this.
666 if (split
[1].num_lines
< 1)
668 split
[1].suspect
= origin_incref(parent
);
672 * split_overlap() divided an existing blame e into up to three parts
673 * in split. Adjust the linked list of blames in the scoreboard to
676 static void split_blame(struct scoreboard
*sb
,
677 struct blame_entry
*split
,
678 struct blame_entry
*e
)
680 struct blame_entry
*new_entry
;
682 if (split
[0].suspect
&& split
[2].suspect
) {
683 /* The first part (reuse storage for the existing entry e) */
684 dup_entry(e
, &split
[0]);
686 /* The last part -- me */
687 new_entry
= xmalloc(sizeof(*new_entry
));
688 memcpy(new_entry
, &(split
[2]), sizeof(struct blame_entry
));
689 add_blame_entry(sb
, new_entry
);
691 /* ... and the middle part -- parent */
692 new_entry
= xmalloc(sizeof(*new_entry
));
693 memcpy(new_entry
, &(split
[1]), sizeof(struct blame_entry
));
694 add_blame_entry(sb
, new_entry
);
696 else if (!split
[0].suspect
&& !split
[2].suspect
)
698 * The parent covers the entire area; reuse storage for
699 * e and replace it with the parent.
701 dup_entry(e
, &split
[1]);
702 else if (split
[0].suspect
) {
703 /* me and then parent */
704 dup_entry(e
, &split
[0]);
706 new_entry
= xmalloc(sizeof(*new_entry
));
707 memcpy(new_entry
, &(split
[1]), sizeof(struct blame_entry
));
708 add_blame_entry(sb
, new_entry
);
711 /* parent and then me */
712 dup_entry(e
, &split
[1]);
714 new_entry
= xmalloc(sizeof(*new_entry
));
715 memcpy(new_entry
, &(split
[2]), sizeof(struct blame_entry
));
716 add_blame_entry(sb
, new_entry
);
719 if (DEBUG
) { /* sanity */
720 struct blame_entry
*ent
;
721 int lno
= sb
->ent
->lno
, corrupt
= 0;
723 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
728 lno
+= ent
->num_lines
;
732 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
733 printf("L %8d l %8d n %8d\n",
734 lno
, ent
->lno
, ent
->num_lines
);
735 lno
= ent
->lno
+ ent
->num_lines
;
743 * After splitting the blame, the origins used by the
744 * on-stack blame_entry should lose one refcnt each.
746 static void decref_split(struct blame_entry
*split
)
750 for (i
= 0; i
< 3; i
++)
751 origin_decref(split
[i
].suspect
);
755 * Helper for blame_chunk(). blame_entry e is known to overlap with
756 * the patch hunk; split it and pass blame to the parent.
758 static void blame_overlap(struct scoreboard
*sb
, struct blame_entry
*e
,
759 int tlno
, int plno
, int same
,
760 struct origin
*parent
)
762 struct blame_entry split
[3];
764 split_overlap(split
, e
, tlno
, plno
, same
, parent
);
765 if (split
[1].suspect
)
766 split_blame(sb
, split
, e
);
771 * Find the line number of the last line the target is suspected for.
773 static int find_last_in_target(struct scoreboard
*sb
, struct origin
*target
)
775 struct blame_entry
*e
;
776 int last_in_target
= -1;
778 for (e
= sb
->ent
; e
; e
= e
->next
) {
779 if (e
->guilty
|| cmp_suspect(e
->suspect
, target
))
781 if (last_in_target
< e
->s_lno
+ e
->num_lines
)
782 last_in_target
= e
->s_lno
+ e
->num_lines
;
784 return last_in_target
;
788 * Process one hunk from the patch between the current suspect for
789 * blame_entry e and its parent. Find and split the overlap, and
790 * pass blame to the overlapping part to the parent.
792 static void blame_chunk(struct scoreboard
*sb
,
793 int tlno
, int plno
, int same
,
794 struct origin
*target
, struct origin
*parent
)
796 struct blame_entry
*e
;
798 for (e
= sb
->ent
; e
; e
= e
->next
) {
799 if (e
->guilty
|| cmp_suspect(e
->suspect
, target
))
801 if (same
<= e
->s_lno
)
803 if (tlno
< e
->s_lno
+ e
->num_lines
)
804 blame_overlap(sb
, e
, tlno
, plno
, same
, parent
);
809 * We are looking at the origin 'target' and aiming to pass blame
810 * for the lines it is suspected to its parent. Run diff to find
811 * which lines came from parent and pass blame for them.
813 static int pass_blame_to_parent(struct scoreboard
*sb
,
814 struct origin
*target
,
815 struct origin
*parent
)
817 int i
, last_in_target
, plno
, tlno
;
820 last_in_target
= find_last_in_target(sb
, target
);
821 if (last_in_target
< 0)
822 return 1; /* nothing remains for this target */
824 patch
= get_patch(parent
, target
);
826 for (i
= 0; i
< patch
->num
; i
++) {
827 struct chunk
*chunk
= &patch
->chunks
[i
];
829 blame_chunk(sb
, tlno
, plno
, chunk
->same
, target
, parent
);
830 plno
= chunk
->p_next
;
831 tlno
= chunk
->t_next
;
833 /* The rest (i.e. anything after tlno) are the same as the parent */
834 blame_chunk(sb
, tlno
, plno
, last_in_target
, target
, parent
);
841 * The lines in blame_entry after splitting blames many times can become
842 * very small and trivial, and at some point it becomes pointless to
843 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
844 * ordinary C program, and it is not worth to say it was copied from
845 * totally unrelated file in the parent.
847 * Compute how trivial the lines in the blame_entry are.
849 static unsigned ent_score(struct scoreboard
*sb
, struct blame_entry
*e
)
858 cp
= nth_line(sb
, e
->lno
);
859 ep
= nth_line(sb
, e
->lno
+ e
->num_lines
);
861 unsigned ch
= *((unsigned char *)cp
);
871 * best_so_far[] and this[] are both a split of an existing blame_entry
872 * that passes blame to the parent. Maintain best_so_far the best split
873 * so far, by comparing this and best_so_far and copying this into
874 * bst_so_far as needed.
876 static void copy_split_if_better(struct scoreboard
*sb
,
877 struct blame_entry
*best_so_far
,
878 struct blame_entry
*this)
882 if (!this[1].suspect
)
884 if (best_so_far
[1].suspect
) {
885 if (ent_score(sb
, &this[1]) < ent_score(sb
, &best_so_far
[1]))
889 for (i
= 0; i
< 3; i
++)
890 origin_incref(this[i
].suspect
);
891 decref_split(best_so_far
);
892 memcpy(best_so_far
, this, sizeof(struct blame_entry
[3]));
896 * Find the lines from parent that are the same as ent so that
897 * we can pass blames to it. file_p has the blob contents for
900 static void find_copy_in_blob(struct scoreboard
*sb
,
901 struct blame_entry
*ent
,
902 struct origin
*parent
,
903 struct blame_entry
*split
,
913 * Prepare mmfile that contains only the lines in ent.
915 cp
= nth_line(sb
, ent
->lno
);
916 file_o
.ptr
= (char*) cp
;
917 cnt
= ent
->num_lines
;
919 while (cnt
&& cp
< sb
->final_buf
+ sb
->final_buf_size
) {
923 file_o
.size
= cp
- file_o
.ptr
;
925 patch
= compare_buffer(file_p
, &file_o
, 1);
927 memset(split
, 0, sizeof(struct blame_entry
[3]));
929 for (i
= 0; i
< patch
->num
; i
++) {
930 struct chunk
*chunk
= &patch
->chunks
[i
];
932 /* tlno to chunk->same are the same as ent */
933 if (ent
->num_lines
<= tlno
)
935 if (tlno
< chunk
->same
) {
936 struct blame_entry
this[3];
937 split_overlap(this, ent
,
938 tlno
+ ent
->s_lno
, plno
,
939 chunk
->same
+ ent
->s_lno
,
941 copy_split_if_better(sb
, split
, this);
944 plno
= chunk
->p_next
;
945 tlno
= chunk
->t_next
;
951 * See if lines currently target is suspected for can be attributed to
954 static int find_move_in_parent(struct scoreboard
*sb
,
955 struct origin
*target
,
956 struct origin
*parent
)
958 int last_in_target
, made_progress
;
959 struct blame_entry
*e
, split
[3];
962 last_in_target
= find_last_in_target(sb
, target
);
963 if (last_in_target
< 0)
964 return 1; /* nothing remains for this target */
966 fill_origin_blob(parent
, &file_p
);
971 while (made_progress
) {
973 for (e
= sb
->ent
; e
; e
= e
->next
) {
974 if (e
->guilty
|| cmp_suspect(e
->suspect
, target
))
976 find_copy_in_blob(sb
, e
, parent
, split
, &file_p
);
977 if (split
[1].suspect
&&
978 blame_move_score
< ent_score(sb
, &split
[1])) {
979 split_blame(sb
, split
, e
);
989 struct blame_entry
*ent
;
990 struct blame_entry split
[3];
994 * Count the number of entries the target is suspected for,
995 * and prepare a list of entry and the best split.
997 static struct blame_list
*setup_blame_list(struct scoreboard
*sb
,
998 struct origin
*target
,
1001 struct blame_entry
*e
;
1003 struct blame_list
*blame_list
= NULL
;
1005 for (e
= sb
->ent
, num_ents
= 0; e
; e
= e
->next
)
1006 if (!e
->guilty
&& !cmp_suspect(e
->suspect
, target
))
1009 blame_list
= xcalloc(num_ents
, sizeof(struct blame_list
));
1010 for (e
= sb
->ent
, i
= 0; e
; e
= e
->next
)
1011 if (!e
->guilty
&& !cmp_suspect(e
->suspect
, target
))
1012 blame_list
[i
++].ent
= e
;
1014 *num_ents_p
= num_ents
;
1019 * For lines target is suspected for, see if we can find code movement
1020 * across file boundary from the parent commit. porigin is the path
1021 * in the parent we already tried.
1023 static int find_copy_in_parent(struct scoreboard
*sb
,
1024 struct origin
*target
,
1025 struct commit
*parent
,
1026 struct origin
*porigin
,
1029 struct diff_options diff_opts
;
1030 const char *paths
[1];
1033 struct blame_list
*blame_list
;
1036 blame_list
= setup_blame_list(sb
, target
, &num_ents
);
1038 return 1; /* nothing remains for this target */
1040 diff_setup(&diff_opts
);
1041 diff_opts
.recursive
= 1;
1042 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
1045 diff_tree_setup_paths(paths
, &diff_opts
);
1046 if (diff_setup_done(&diff_opts
) < 0)
1049 /* Try "find copies harder" on new path if requested;
1050 * we do not want to use diffcore_rename() actually to
1051 * match things up; find_copies_harder is set only to
1052 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1053 * and this code needs to be after diff_setup_done(), which
1054 * usually makes find-copies-harder imply copy detection.
1056 if ((opt
& PICKAXE_BLAME_COPY_HARDER
) &&
1057 (!porigin
|| strcmp(target
->path
, porigin
->path
)))
1058 diff_opts
.find_copies_harder
= 1;
1060 if (is_null_sha1(target
->commit
->object
.sha1
))
1061 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
1063 diff_tree_sha1(parent
->tree
->object
.sha1
,
1064 target
->commit
->tree
->object
.sha1
,
1067 if (!diff_opts
.find_copies_harder
)
1068 diffcore_std(&diff_opts
);
1072 int made_progress
= 0;
1074 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
1075 struct diff_filepair
*p
= diff_queued_diff
.queue
[i
];
1076 struct origin
*norigin
;
1078 struct blame_entry
this[3];
1080 if (!DIFF_FILE_VALID(p
->one
))
1081 continue; /* does not exist in parent */
1082 if (porigin
&& !strcmp(p
->one
->path
, porigin
->path
))
1083 /* find_move already dealt with this path */
1086 norigin
= get_origin(sb
, parent
, p
->one
->path
);
1087 hashcpy(norigin
->blob_sha1
, p
->one
->sha1
);
1088 fill_origin_blob(norigin
, &file_p
);
1092 for (j
= 0; j
< num_ents
; j
++) {
1093 find_copy_in_blob(sb
, blame_list
[j
].ent
,
1094 norigin
, this, &file_p
);
1095 copy_split_if_better(sb
, blame_list
[j
].split
,
1099 origin_decref(norigin
);
1102 for (j
= 0; j
< num_ents
; j
++) {
1103 struct blame_entry
*split
= blame_list
[j
].split
;
1104 if (split
[1].suspect
&&
1105 blame_copy_score
< ent_score(sb
, &split
[1])) {
1106 split_blame(sb
, split
, blame_list
[j
].ent
);
1109 decref_split(split
);
1115 blame_list
= setup_blame_list(sb
, target
, &num_ents
);
1121 diff_flush(&diff_opts
);
1127 * The blobs of origin and porigin exactly match, so everything
1128 * origin is suspected for can be blamed on the parent.
1130 static void pass_whole_blame(struct scoreboard
*sb
,
1131 struct origin
*origin
, struct origin
*porigin
)
1133 struct blame_entry
*e
;
1135 if (!porigin
->file
.ptr
&& origin
->file
.ptr
) {
1136 /* Steal its file */
1137 porigin
->file
= origin
->file
;
1138 origin
->file
.ptr
= NULL
;
1140 for (e
= sb
->ent
; e
; e
= e
->next
) {
1141 if (cmp_suspect(e
->suspect
, origin
))
1143 origin_incref(porigin
);
1144 origin_decref(e
->suspect
);
1145 e
->suspect
= porigin
;
1149 #define MAXPARENT 16
1151 static void pass_blame(struct scoreboard
*sb
, struct origin
*origin
, int opt
)
1154 struct commit
*commit
= origin
->commit
;
1155 struct commit_list
*parent
;
1156 struct origin
*parent_origin
[MAXPARENT
], *porigin
;
1158 memset(parent_origin
, 0, sizeof(parent_origin
));
1160 /* The first pass looks for unrenamed path to optimize for
1161 * common cases, then we look for renames in the second pass.
1163 for (pass
= 0; pass
< 2; pass
++) {
1164 struct origin
*(*find
)(struct scoreboard
*,
1165 struct commit
*, struct origin
*);
1166 find
= pass
? find_rename
: find_origin
;
1168 for (i
= 0, parent
= commit
->parents
;
1169 i
< MAXPARENT
&& parent
;
1170 parent
= parent
->next
, i
++) {
1171 struct commit
*p
= parent
->item
;
1174 if (parent_origin
[i
])
1176 if (parse_commit(p
))
1178 porigin
= find(sb
, p
, origin
);
1181 if (!hashcmp(porigin
->blob_sha1
, origin
->blob_sha1
)) {
1182 pass_whole_blame(sb
, origin
, porigin
);
1183 origin_decref(porigin
);
1186 for (j
= same
= 0; j
< i
; j
++)
1187 if (parent_origin
[j
] &&
1188 !hashcmp(parent_origin
[j
]->blob_sha1
,
1189 porigin
->blob_sha1
)) {
1194 parent_origin
[i
] = porigin
;
1196 origin_decref(porigin
);
1201 for (i
= 0, parent
= commit
->parents
;
1202 i
< MAXPARENT
&& parent
;
1203 parent
= parent
->next
, i
++) {
1204 struct origin
*porigin
= parent_origin
[i
];
1207 if (pass_blame_to_parent(sb
, origin
, porigin
))
1212 * Optionally find moves in parents' files.
1214 if (opt
& PICKAXE_BLAME_MOVE
)
1215 for (i
= 0, parent
= commit
->parents
;
1216 i
< MAXPARENT
&& parent
;
1217 parent
= parent
->next
, i
++) {
1218 struct origin
*porigin
= parent_origin
[i
];
1221 if (find_move_in_parent(sb
, origin
, porigin
))
1226 * Optionally find copies from parents' files.
1228 if (opt
& PICKAXE_BLAME_COPY
)
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 (find_copy_in_parent(sb
, origin
, parent
->item
,
1239 for (i
= 0; i
< MAXPARENT
; i
++)
1240 origin_decref(parent_origin
[i
]);
1244 * Information on commits, used for output.
1250 unsigned long author_time
;
1253 /* filled only when asked for details */
1255 char *committer_mail
;
1256 unsigned long committer_time
;
1263 * Parse author/committer line in the commit object buffer
1265 static void get_ac_line(const char *inbuf
, const char *what
,
1266 int bufsz
, char *person
, char **mail
,
1267 unsigned long *time
, char **tz
)
1272 tmp
= strstr(inbuf
, what
);
1275 tmp
+= strlen(what
);
1276 endp
= strchr(tmp
, '\n');
1284 person
= *mail
= *tz
= "(unknown)";
1288 memcpy(person
, tmp
, len
);
1300 *time
= strtoul(tmp
, NULL
, 10);
1309 static void get_commit_info(struct commit
*commit
,
1310 struct commit_info
*ret
,
1315 static char author_buf
[1024];
1316 static char committer_buf
[1024];
1317 static char summary_buf
[1024];
1320 * We've operated without save_commit_buffer, so
1321 * we now need to populate them for output.
1323 if (!commit
->buffer
) {
1327 read_sha1_file(commit
->object
.sha1
, type
, &size
);
1329 ret
->author
= author_buf
;
1330 get_ac_line(commit
->buffer
, "\nauthor ",
1331 sizeof(author_buf
), author_buf
, &ret
->author_mail
,
1332 &ret
->author_time
, &ret
->author_tz
);
1337 ret
->committer
= committer_buf
;
1338 get_ac_line(commit
->buffer
, "\ncommitter ",
1339 sizeof(committer_buf
), committer_buf
, &ret
->committer_mail
,
1340 &ret
->committer_time
, &ret
->committer_tz
);
1342 ret
->summary
= summary_buf
;
1343 tmp
= strstr(commit
->buffer
, "\n\n");
1346 sprintf(summary_buf
, "(%s)", sha1_to_hex(commit
->object
.sha1
));
1350 endp
= strchr(tmp
, '\n');
1352 endp
= tmp
+ strlen(tmp
);
1354 if (len
>= sizeof(summary_buf
) || len
== 0)
1356 memcpy(summary_buf
, tmp
, len
);
1357 summary_buf
[len
] = 0;
1361 * To allow LF and other nonportable characters in pathnames,
1362 * they are c-style quoted as needed.
1364 static void write_filename_info(const char *path
)
1366 printf("filename ");
1367 write_name_quoted(NULL
, 0, path
, 1, stdout
);
1372 * The blame_entry is found to be guilty for the range. Mark it
1373 * as such, and show it in incremental output.
1375 static void found_guilty_entry(struct blame_entry
*ent
)
1381 struct origin
*suspect
= ent
->suspect
;
1383 printf("%s %d %d %d\n",
1384 sha1_to_hex(suspect
->commit
->object
.sha1
),
1385 ent
->s_lno
+ 1, ent
->lno
+ 1, ent
->num_lines
);
1386 if (!(suspect
->commit
->object
.flags
& METAINFO_SHOWN
)) {
1387 struct commit_info ci
;
1388 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
1389 get_commit_info(suspect
->commit
, &ci
, 1);
1390 printf("author %s\n", ci
.author
);
1391 printf("author-mail %s\n", ci
.author_mail
);
1392 printf("author-time %lu\n", ci
.author_time
);
1393 printf("author-tz %s\n", ci
.author_tz
);
1394 printf("committer %s\n", ci
.committer
);
1395 printf("committer-mail %s\n", ci
.committer_mail
);
1396 printf("committer-time %lu\n", ci
.committer_time
);
1397 printf("committer-tz %s\n", ci
.committer_tz
);
1398 printf("summary %s\n", ci
.summary
);
1399 if (suspect
->commit
->object
.flags
& UNINTERESTING
)
1400 printf("boundary\n");
1402 write_filename_info(suspect
->path
);
1407 * The main loop -- while the scoreboard has lines whose true origin
1408 * is still unknown, pick one blame_entry, and allow its current
1409 * suspect to pass blames to its parents.
1411 static void assign_blame(struct scoreboard
*sb
, struct rev_info
*revs
, int opt
)
1414 struct blame_entry
*ent
;
1415 struct commit
*commit
;
1416 struct origin
*suspect
= NULL
;
1418 /* find one suspect to break down */
1419 for (ent
= sb
->ent
; !suspect
&& ent
; ent
= ent
->next
)
1421 suspect
= ent
->suspect
;
1423 return; /* all done */
1426 * We will use this suspect later in the loop,
1427 * so hold onto it in the meantime.
1429 origin_incref(suspect
);
1430 commit
= suspect
->commit
;
1431 if (!commit
->object
.parsed
)
1432 parse_commit(commit
);
1433 if (!(commit
->object
.flags
& UNINTERESTING
) &&
1434 !(revs
->max_age
!= -1 && commit
->date
< revs
->max_age
))
1435 pass_blame(sb
, suspect
, opt
);
1437 commit
->object
.flags
|= UNINTERESTING
;
1438 if (commit
->object
.parsed
)
1439 mark_parents_uninteresting(commit
);
1441 /* treat root commit as boundary */
1442 if (!commit
->parents
&& !show_root
)
1443 commit
->object
.flags
|= UNINTERESTING
;
1445 /* Take responsibility for the remaining entries */
1446 for (ent
= sb
->ent
; ent
; ent
= ent
->next
)
1447 if (!cmp_suspect(ent
->suspect
, suspect
))
1448 found_guilty_entry(ent
);
1449 origin_decref(suspect
);
1451 if (DEBUG
) /* sanity */
1452 sanity_check_refcnt(sb
);
1456 static const char *format_time(unsigned long time
, const char *tz_str
,
1459 static char time_buf
[128];
1464 if (show_raw_time
) {
1465 sprintf(time_buf
, "%lu %s", time
, tz_str
);
1470 minutes
= tz
< 0 ? -tz
: tz
;
1471 minutes
= (minutes
/ 100)*60 + (minutes
% 100);
1472 minutes
= tz
< 0 ? -minutes
: minutes
;
1473 t
= time
+ minutes
* 60;
1476 strftime(time_buf
, sizeof(time_buf
), "%Y-%m-%d %H:%M:%S ", tm
);
1477 strcat(time_buf
, tz_str
);
1481 #define OUTPUT_ANNOTATE_COMPAT 001
1482 #define OUTPUT_LONG_OBJECT_NAME 002
1483 #define OUTPUT_RAW_TIMESTAMP 004
1484 #define OUTPUT_PORCELAIN 010
1485 #define OUTPUT_SHOW_NAME 020
1486 #define OUTPUT_SHOW_NUMBER 040
1487 #define OUTPUT_SHOW_SCORE 0100
1489 static void emit_porcelain(struct scoreboard
*sb
, struct blame_entry
*ent
)
1493 struct origin
*suspect
= ent
->suspect
;
1496 strcpy(hex
, sha1_to_hex(suspect
->commit
->object
.sha1
));
1497 printf("%s%c%d %d %d\n",
1499 ent
->guilty
? ' ' : '*', // purely for debugging
1503 if (!(suspect
->commit
->object
.flags
& METAINFO_SHOWN
)) {
1504 struct commit_info ci
;
1505 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
1506 get_commit_info(suspect
->commit
, &ci
, 1);
1507 printf("author %s\n", ci
.author
);
1508 printf("author-mail %s\n", ci
.author_mail
);
1509 printf("author-time %lu\n", ci
.author_time
);
1510 printf("author-tz %s\n", ci
.author_tz
);
1511 printf("committer %s\n", ci
.committer
);
1512 printf("committer-mail %s\n", ci
.committer_mail
);
1513 printf("committer-time %lu\n", ci
.committer_time
);
1514 printf("committer-tz %s\n", ci
.committer_tz
);
1515 write_filename_info(suspect
->path
);
1516 printf("summary %s\n", ci
.summary
);
1517 if (suspect
->commit
->object
.flags
& UNINTERESTING
)
1518 printf("boundary\n");
1520 else if (suspect
->commit
->object
.flags
& MORE_THAN_ONE_PATH
)
1521 write_filename_info(suspect
->path
);
1523 cp
= nth_line(sb
, ent
->lno
);
1524 for (cnt
= 0; cnt
< ent
->num_lines
; cnt
++) {
1527 printf("%s %d %d\n", hex
,
1528 ent
->s_lno
+ 1 + cnt
,
1529 ent
->lno
+ 1 + cnt
);
1534 } while (ch
!= '\n' &&
1535 cp
< sb
->final_buf
+ sb
->final_buf_size
);
1539 static void emit_other(struct scoreboard
*sb
, struct blame_entry
*ent
, int opt
)
1543 struct origin
*suspect
= ent
->suspect
;
1544 struct commit_info ci
;
1546 int show_raw_time
= !!(opt
& OUTPUT_RAW_TIMESTAMP
);
1548 get_commit_info(suspect
->commit
, &ci
, 1);
1549 strcpy(hex
, sha1_to_hex(suspect
->commit
->object
.sha1
));
1551 cp
= nth_line(sb
, ent
->lno
);
1552 for (cnt
= 0; cnt
< ent
->num_lines
; cnt
++) {
1554 int length
= (opt
& OUTPUT_LONG_OBJECT_NAME
) ? 40 : 8;
1556 if (suspect
->commit
->object
.flags
& UNINTERESTING
) {
1557 if (!blank_boundary
) {
1562 memset(hex
, ' ', length
);
1565 printf("%.*s", length
, hex
);
1566 if (opt
& OUTPUT_ANNOTATE_COMPAT
)
1567 printf("\t(%10s\t%10s\t%d)", ci
.author
,
1568 format_time(ci
.author_time
, ci
.author_tz
,
1570 ent
->lno
+ 1 + cnt
);
1572 if (opt
& OUTPUT_SHOW_SCORE
)
1574 max_score_digits
, ent
->score
,
1575 ent
->suspect
->refcnt
);
1576 if (opt
& OUTPUT_SHOW_NAME
)
1577 printf(" %-*.*s", longest_file
, longest_file
,
1579 if (opt
& OUTPUT_SHOW_NUMBER
)
1580 printf(" %*d", max_orig_digits
,
1581 ent
->s_lno
+ 1 + cnt
);
1582 printf(" (%-*.*s %10s %*d) ",
1583 longest_author
, longest_author
, ci
.author
,
1584 format_time(ci
.author_time
, ci
.author_tz
,
1586 max_digits
, ent
->lno
+ 1 + cnt
);
1591 } while (ch
!= '\n' &&
1592 cp
< sb
->final_buf
+ sb
->final_buf_size
);
1596 static void output(struct scoreboard
*sb
, int option
)
1598 struct blame_entry
*ent
;
1600 if (option
& OUTPUT_PORCELAIN
) {
1601 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1602 struct blame_entry
*oth
;
1603 struct origin
*suspect
= ent
->suspect
;
1604 struct commit
*commit
= suspect
->commit
;
1605 if (commit
->object
.flags
& MORE_THAN_ONE_PATH
)
1607 for (oth
= ent
->next
; oth
; oth
= oth
->next
) {
1608 if ((oth
->suspect
->commit
!= commit
) ||
1609 !strcmp(oth
->suspect
->path
, suspect
->path
))
1611 commit
->object
.flags
|= MORE_THAN_ONE_PATH
;
1617 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1618 if (option
& OUTPUT_PORCELAIN
)
1619 emit_porcelain(sb
, ent
);
1621 emit_other(sb
, ent
, option
);
1627 * To allow quick access to the contents of nth line in the
1628 * final image, prepare an index in the scoreboard.
1630 static int prepare_lines(struct scoreboard
*sb
)
1632 const char *buf
= sb
->final_buf
;
1633 unsigned long len
= sb
->final_buf_size
;
1634 int num
= 0, incomplete
= 0, bol
= 1;
1636 if (len
&& buf
[len
-1] != '\n')
1637 incomplete
++; /* incomplete line at the end */
1640 sb
->lineno
= xrealloc(sb
->lineno
,
1641 sizeof(int* ) * (num
+ 1));
1642 sb
->lineno
[num
] = buf
- sb
->final_buf
;
1645 if (*buf
++ == '\n') {
1650 sb
->lineno
= xrealloc(sb
->lineno
,
1651 sizeof(int* ) * (num
+ incomplete
+ 1));
1652 sb
->lineno
[num
+ incomplete
] = buf
- sb
->final_buf
;
1653 sb
->num_lines
= num
+ incomplete
;
1654 return sb
->num_lines
;
1658 * Add phony grafts for use with -S; this is primarily to
1659 * support git-cvsserver that wants to give a linear history
1662 static int read_ancestry(const char *graft_file
)
1664 FILE *fp
= fopen(graft_file
, "r");
1668 while (fgets(buf
, sizeof(buf
), fp
)) {
1669 /* The format is just "Commit Parent1 Parent2 ...\n" */
1670 int len
= strlen(buf
);
1671 struct commit_graft
*graft
= read_graft_line(buf
, len
);
1673 register_commit_graft(graft
, 0);
1680 * How many columns do we need to show line numbers in decimal?
1682 static int lineno_width(int lines
)
1686 for (width
= 1, i
= 10; i
<= lines
+ 1; width
++)
1692 * How many columns do we need to show line numbers, authors,
1695 static void find_alignment(struct scoreboard
*sb
, int *option
)
1697 int longest_src_lines
= 0;
1698 int longest_dst_lines
= 0;
1699 unsigned largest_score
= 0;
1700 struct blame_entry
*e
;
1702 for (e
= sb
->ent
; e
; e
= e
->next
) {
1703 struct origin
*suspect
= e
->suspect
;
1704 struct commit_info ci
;
1707 if (strcmp(suspect
->path
, sb
->path
))
1708 *option
|= OUTPUT_SHOW_NAME
;
1709 num
= strlen(suspect
->path
);
1710 if (longest_file
< num
)
1712 if (!(suspect
->commit
->object
.flags
& METAINFO_SHOWN
)) {
1713 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
1714 get_commit_info(suspect
->commit
, &ci
, 1);
1715 num
= strlen(ci
.author
);
1716 if (longest_author
< num
)
1717 longest_author
= num
;
1719 num
= e
->s_lno
+ e
->num_lines
;
1720 if (longest_src_lines
< num
)
1721 longest_src_lines
= num
;
1722 num
= e
->lno
+ e
->num_lines
;
1723 if (longest_dst_lines
< num
)
1724 longest_dst_lines
= num
;
1725 if (largest_score
< ent_score(sb
, e
))
1726 largest_score
= ent_score(sb
, e
);
1728 max_orig_digits
= lineno_width(longest_src_lines
);
1729 max_digits
= lineno_width(longest_dst_lines
);
1730 max_score_digits
= lineno_width(largest_score
);
1734 * For debugging -- origin is refcounted, and this asserts that
1735 * we do not underflow.
1737 static void sanity_check_refcnt(struct scoreboard
*sb
)
1740 struct blame_entry
*ent
;
1742 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1743 /* Nobody should have zero or negative refcnt */
1744 if (ent
->suspect
->refcnt
<= 0) {
1745 fprintf(stderr
, "%s in %s has negative refcnt %d\n",
1747 sha1_to_hex(ent
->suspect
->commit
->object
.sha1
),
1748 ent
->suspect
->refcnt
);
1752 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1753 /* Mark the ones that haven't been checked */
1754 if (0 < ent
->suspect
->refcnt
)
1755 ent
->suspect
->refcnt
= -ent
->suspect
->refcnt
;
1757 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1759 * ... then pick each and see if they have the the
1763 struct blame_entry
*e
;
1764 struct origin
*suspect
= ent
->suspect
;
1766 if (0 < suspect
->refcnt
)
1768 suspect
->refcnt
= -suspect
->refcnt
; /* Unmark */
1769 for (found
= 0, e
= sb
->ent
; e
; e
= e
->next
) {
1770 if (e
->suspect
!= suspect
)
1774 if (suspect
->refcnt
!= found
) {
1775 fprintf(stderr
, "%s in %s has refcnt %d, not %d\n",
1777 sha1_to_hex(ent
->suspect
->commit
->object
.sha1
),
1778 ent
->suspect
->refcnt
, found
);
1784 find_alignment(sb
, &opt
);
1786 die("Baa %d!", baa
);
1791 * Used for the command line parsing; check if the path exists
1792 * in the working tree.
1794 static int has_path_in_work_tree(const char *path
)
1797 return !lstat(path
, &st
);
1800 static unsigned parse_score(const char *arg
)
1803 unsigned long score
= strtoul(arg
, &end
, 10);
1809 static const char *add_prefix(const char *prefix
, const char *path
)
1811 if (!prefix
|| !prefix
[0])
1813 return prefix_path(prefix
, strlen(prefix
), path
);
1817 * Parsing of (comma separated) one item in the -L option
1819 static const char *parse_loc(const char *spec
,
1820 struct scoreboard
*sb
, long lno
,
1821 long begin
, long *ret
)
1828 regmatch_t match
[1];
1830 /* Allow "-L <something>,+20" to mean starting at <something>
1831 * for 20 lines, or "-L <something>,-5" for 5 lines ending at
1834 if (1 < begin
&& (spec
[0] == '+' || spec
[0] == '-')) {
1835 num
= strtol(spec
+ 1, &term
, 10);
1836 if (term
!= spec
+ 1) {
1840 *ret
= begin
+ num
- 2;
1849 num
= strtol(spec
, &term
, 10);
1857 /* it could be a regexp of form /.../ */
1858 for (term
= (char*) spec
+ 1; *term
&& *term
!= '/'; term
++) {
1865 /* try [spec+1 .. term-1] as regexp */
1867 begin
--; /* input is in human terms */
1868 line
= nth_line(sb
, begin
);
1870 if (!(reg_error
= regcomp(®exp
, spec
+ 1, REG_NEWLINE
)) &&
1871 !(reg_error
= regexec(®exp
, line
, 1, match
, 0))) {
1872 const char *cp
= line
+ match
[0].rm_so
;
1875 while (begin
++ < lno
) {
1876 nline
= nth_line(sb
, begin
);
1877 if (line
<= cp
&& cp
< nline
)
1888 regerror(reg_error
, ®exp
, errbuf
, 1024);
1889 die("-L parameter '%s': %s", spec
+ 1, errbuf
);
1894 * Parsing of -L option
1896 static void prepare_blame_range(struct scoreboard
*sb
,
1897 const char *bottomtop
,
1899 long *bottom
, long *top
)
1903 term
= parse_loc(bottomtop
, sb
, lno
, 1, bottom
);
1905 term
= parse_loc(term
+ 1, sb
, lno
, *bottom
+ 1, top
);
1913 static int git_blame_config(const char *var
, const char *value
)
1915 if (!strcmp(var
, "blame.showroot")) {
1916 show_root
= git_config_bool(var
, value
);
1919 if (!strcmp(var
, "blame.blankboundary")) {
1920 blank_boundary
= git_config_bool(var
, value
);
1923 return git_default_config(var
, value
);
1926 static struct commit
*fake_working_tree_commit(const char *path
, const char *contents_from
)
1928 struct commit
*commit
;
1929 struct origin
*origin
;
1930 unsigned char head_sha1
[20];
1935 unsigned long fin_size
;
1937 struct cache_entry
*ce
;
1940 if (get_sha1("HEAD", head_sha1
))
1941 die("No such ref: HEAD");
1944 commit
= xcalloc(1, sizeof(*commit
));
1945 commit
->parents
= xcalloc(1, sizeof(*commit
->parents
));
1946 commit
->parents
->item
= lookup_commit_reference(head_sha1
);
1947 commit
->object
.parsed
= 1;
1949 commit
->object
.type
= OBJ_COMMIT
;
1951 origin
= make_origin(commit
, path
);
1953 if (!contents_from
|| strcmp("-", contents_from
)) {
1955 const char *read_from
;
1957 if (contents_from
) {
1958 if (stat(contents_from
, &st
) < 0)
1959 die("Cannot stat %s", contents_from
);
1960 read_from
= contents_from
;
1963 if (lstat(path
, &st
) < 0)
1964 die("Cannot lstat %s", path
);
1967 fin_size
= st
.st_size
;
1968 buf
= xmalloc(fin_size
+1);
1969 mode
= canon_mode(st
.st_mode
);
1970 switch (st
.st_mode
& S_IFMT
) {
1972 fd
= open(read_from
, O_RDONLY
);
1974 die("cannot open %s", read_from
);
1975 if (read_in_full(fd
, buf
, fin_size
) != fin_size
)
1976 die("cannot read %s", read_from
);
1979 if (readlink(read_from
, buf
, fin_size
+1) != fin_size
)
1980 die("cannot readlink %s", read_from
);
1983 die("unsupported file type %s", read_from
);
1987 /* Reading from stdin */
1988 contents_from
= "standard input";
1994 buf
= xrealloc(buf
, fin_size
+ cnt
);
1995 cnt
= xread(0, buf
+ fin_size
, cnt
);
1997 die("read error %s from stdin",
2003 buf
= xrealloc(buf
, fin_size
+ 1);
2006 origin
->file
.ptr
= buf
;
2007 origin
->file
.size
= fin_size
;
2008 pretend_sha1_file(buf
, fin_size
, blob_type
, origin
->blob_sha1
);
2009 commit
->util
= origin
;
2012 * Read the current index, replace the path entry with
2013 * origin->blob_sha1 without mucking with its mode or type
2014 * bits; we are not going to write this index out -- we just
2015 * want to run "diff-index --cached".
2022 int pos
= cache_name_pos(path
, len
);
2024 mode
= ntohl(active_cache
[pos
]->ce_mode
);
2026 /* Let's not bother reading from HEAD tree */
2027 mode
= S_IFREG
| 0644;
2029 size
= cache_entry_size(len
);
2030 ce
= xcalloc(1, size
);
2031 hashcpy(ce
->sha1
, origin
->blob_sha1
);
2032 memcpy(ce
->name
, path
, len
);
2033 ce
->ce_flags
= create_ce_flags(len
, 0);
2034 ce
->ce_mode
= create_ce_mode(mode
);
2035 add_cache_entry(ce
, ADD_CACHE_OK_TO_ADD
|ADD_CACHE_OK_TO_REPLACE
);
2038 * We are not going to write this out, so this does not matter
2039 * right now, but someday we might optimize diff-index --cached
2040 * with cache-tree information.
2042 cache_tree_invalidate_path(active_cache_tree
, path
);
2044 commit
->buffer
= xmalloc(400);
2045 ident
= fmt_ident("Not Committed Yet", "not.committed.yet", NULL
, 0);
2046 sprintf(commit
->buffer
,
2047 "tree 0000000000000000000000000000000000000000\n"
2051 "Version of %s from %s\n",
2052 sha1_to_hex(head_sha1
),
2053 ident
, ident
, path
, contents_from
? contents_from
: path
);
2057 int cmd_blame(int argc
, const char **argv
, const char *prefix
)
2059 struct rev_info revs
;
2061 struct scoreboard sb
;
2063 struct blame_entry
*ent
;
2064 int i
, seen_dashdash
, unk
, opt
;
2065 long bottom
, top
, lno
;
2066 int output_option
= 0;
2067 const char *revs_file
= NULL
;
2068 const char *final_commit_name
= NULL
;
2070 const char *bottomtop
= NULL
;
2071 const char *contents_from
= NULL
;
2073 git_config(git_blame_config
);
2074 save_commit_buffer
= 0;
2078 for (unk
= i
= 1; i
< argc
; i
++) {
2079 const char *arg
= argv
[i
];
2082 else if (!strcmp("-b", arg
))
2084 else if (!strcmp("--root", arg
))
2086 else if (!strcmp("-c", arg
))
2087 output_option
|= OUTPUT_ANNOTATE_COMPAT
;
2088 else if (!strcmp("-t", arg
))
2089 output_option
|= OUTPUT_RAW_TIMESTAMP
;
2090 else if (!strcmp("-l", arg
))
2091 output_option
|= OUTPUT_LONG_OBJECT_NAME
;
2092 else if (!strcmp("-S", arg
) && ++i
< argc
)
2093 revs_file
= argv
[i
];
2094 else if (!strncmp("-M", arg
, 2)) {
2095 opt
|= PICKAXE_BLAME_MOVE
;
2096 blame_move_score
= parse_score(arg
+2);
2098 else if (!strncmp("-C", arg
, 2)) {
2099 if (opt
& PICKAXE_BLAME_COPY
)
2100 opt
|= PICKAXE_BLAME_COPY_HARDER
;
2101 opt
|= PICKAXE_BLAME_COPY
| PICKAXE_BLAME_MOVE
;
2102 blame_copy_score
= parse_score(arg
+2);
2104 else if (!strncmp("-L", arg
, 2)) {
2113 die("More than one '-L n,m' option given");
2116 else if (!strcmp("--contents", arg
)) {
2119 contents_from
= argv
[i
];
2121 else if (!strcmp("--incremental", arg
))
2123 else if (!strcmp("--score-debug", arg
))
2124 output_option
|= OUTPUT_SHOW_SCORE
;
2125 else if (!strcmp("-f", arg
) ||
2126 !strcmp("--show-name", arg
))
2127 output_option
|= OUTPUT_SHOW_NAME
;
2128 else if (!strcmp("-n", arg
) ||
2129 !strcmp("--show-number", arg
))
2130 output_option
|= OUTPUT_SHOW_NUMBER
;
2131 else if (!strcmp("-p", arg
) ||
2132 !strcmp("--porcelain", arg
))
2133 output_option
|= OUTPUT_PORCELAIN
;
2134 else if (!strcmp("--", arg
)) {
2146 if (!blame_move_score
)
2147 blame_move_score
= BLAME_DEFAULT_MOVE_SCORE
;
2148 if (!blame_copy_score
)
2149 blame_copy_score
= BLAME_DEFAULT_COPY_SCORE
;
2152 * We have collected options unknown to us in argv[1..unk]
2153 * which are to be passed to revision machinery if we are
2154 * going to do the "bottom" processing.
2156 * The remaining are:
2158 * (1) if seen_dashdash, its either
2159 * "-options -- <path>" or
2160 * "-options -- <path> <rev>".
2161 * but the latter is allowed only if there is no
2162 * options that we passed to revision machinery.
2164 * (2) otherwise, we may have "--" somewhere later and
2165 * might be looking at the first one of multiple 'rev'
2166 * parameters (e.g. " master ^next ^maint -- path").
2167 * See if there is a dashdash first, and give the
2168 * arguments before that to revision machinery.
2169 * After that there must be one 'path'.
2171 * (3) otherwise, its one of the three:
2172 * "-options <path> <rev>"
2173 * "-options <rev> <path>"
2175 * but again the first one is allowed only if
2176 * there is no options that we passed to revision
2180 if (seen_dashdash
) {
2184 path
= add_prefix(prefix
, argv
[i
]);
2185 if (i
+ 1 == argc
- 1) {
2188 argv
[unk
++] = argv
[i
+ 1];
2190 else if (i
+ 1 != argc
)
2191 /* garbage at end */
2196 for (j
= i
; !seen_dashdash
&& j
< argc
; j
++)
2197 if (!strcmp(argv
[j
], "--"))
2199 if (seen_dashdash
) {
2200 if (seen_dashdash
+ 1 != argc
- 1)
2202 path
= add_prefix(prefix
, argv
[seen_dashdash
+ 1]);
2203 for (j
= i
; j
< seen_dashdash
; j
++)
2204 argv
[unk
++] = argv
[j
];
2208 path
= add_prefix(prefix
, argv
[i
]);
2209 if (i
+ 1 == argc
- 1) {
2210 final_commit_name
= argv
[i
+ 1];
2212 /* if (unk == 1) we could be getting
2215 if (unk
== 1 && !has_path_in_work_tree(path
)) {
2216 path
= add_prefix(prefix
, argv
[i
+ 1]);
2217 final_commit_name
= argv
[i
];
2220 else if (i
!= argc
- 1)
2221 usage(blame_usage
); /* garbage at end */
2223 if (!has_path_in_work_tree(path
))
2224 die("cannot stat path %s: %s",
2225 path
, strerror(errno
));
2229 if (final_commit_name
)
2230 argv
[unk
++] = final_commit_name
;
2233 * Now we got rev and path. We do not want the path pruning
2234 * but we may want "bottom" processing.
2236 argv
[unk
++] = "--"; /* terminate the rev name */
2239 init_revisions(&revs
, NULL
);
2240 setup_revisions(unk
, argv
, &revs
, NULL
);
2241 memset(&sb
, 0, sizeof(sb
));
2244 * There must be one and only one positive commit in the
2245 * revs->pending array.
2247 for (i
= 0; i
< revs
.pending
.nr
; i
++) {
2248 struct object
*obj
= revs
.pending
.objects
[i
].item
;
2249 if (obj
->flags
& UNINTERESTING
)
2251 while (obj
->type
== OBJ_TAG
)
2252 obj
= deref_tag(obj
, NULL
, 0);
2253 if (obj
->type
!= OBJ_COMMIT
)
2254 die("Non commit %s?",
2255 revs
.pending
.objects
[i
].name
);
2257 die("More than one commit to dig from %s and %s?",
2258 revs
.pending
.objects
[i
].name
,
2260 sb
.final
= (struct commit
*) obj
;
2261 final_commit_name
= revs
.pending
.objects
[i
].name
;
2266 * "--not A B -- path" without anything positive;
2267 * do not default to HEAD, but use the working tree
2270 sb
.final
= fake_working_tree_commit(path
, contents_from
);
2271 add_pending_object(&revs
, &(sb
.final
->object
), ":");
2273 else if (contents_from
)
2274 die("Cannot use --contents with final commit object name");
2277 * If we have bottom, this will mark the ancestors of the
2278 * bottom commits we would reach while traversing as
2281 prepare_revision_walk(&revs
);
2283 if (is_null_sha1(sb
.final
->object
.sha1
)) {
2286 buf
= xmalloc(o
->file
.size
+ 1);
2287 memcpy(buf
, o
->file
.ptr
, o
->file
.size
+ 1);
2289 sb
.final_buf_size
= o
->file
.size
;
2292 o
= get_origin(&sb
, sb
.final
, path
);
2293 if (fill_blob_sha1(o
))
2294 die("no such path %s in %s", path
, final_commit_name
);
2296 sb
.final_buf
= read_sha1_file(o
->blob_sha1
, type
,
2297 &sb
.final_buf_size
);
2300 lno
= prepare_lines(&sb
);
2304 prepare_blame_range(&sb
, bottomtop
, lno
, &bottom
, &top
);
2305 if (bottom
&& top
&& top
< bottom
) {
2307 tmp
= top
; top
= bottom
; bottom
= tmp
;
2315 die("file %s has only %lu lines", path
, lno
);
2317 ent
= xcalloc(1, sizeof(*ent
));
2319 ent
->num_lines
= top
- bottom
;
2321 ent
->s_lno
= bottom
;
2326 if (revs_file
&& read_ancestry(revs_file
))
2327 die("reading graft file %s failed: %s",
2328 revs_file
, strerror(errno
));
2330 assign_blame(&sb
, &revs
, opt
);
2337 if (!(output_option
& OUTPUT_PORCELAIN
))
2338 find_alignment(&sb
, &output_option
);
2340 output(&sb
, output_option
);
2341 free((void *)sb
.final_buf
);
2342 for (ent
= sb
.ent
; ent
; ) {
2343 struct blame_entry
*e
= ent
->next
;
2349 printf("num read blob: %d\n", num_read_blob
);
2350 printf("num get patch: %d\n", num_get_patch
);
2351 printf("num commits: %d\n", num_commits
);