4 * Copyright (c) 2006, 2014 by its authors
5 * See COPYING for licensing conditions
14 #include "tree-walk.h"
19 #include "xdiff-interface.h"
20 #include "cache-tree.h"
21 #include "string-list.h"
23 #include "mergesort.h"
24 #include "parse-options.h"
25 #include "prio-queue.h"
28 #include "line-range.h"
32 static char blame_usage
[] = N_("git blame [<options>] [<rev-opts>] [<rev>] [--] <file>");
34 static const char *blame_opt_usage
[] = {
37 N_("<rev-opts> are documented in git-rev-list(1)"),
41 static int longest_file
;
42 static int longest_author
;
43 static int max_orig_digits
;
44 static int max_digits
;
45 static int max_score_digits
;
48 static int blank_boundary
;
49 static int incremental
;
51 static int abbrev
= -1;
52 static int no_whole_file_rename
;
54 static struct date_mode blame_date_mode
= { DATE_ISO8601
};
55 static size_t blame_date_width
;
57 static struct string_list mailmap
;
64 static int num_read_blob
;
65 static int num_get_patch
;
66 static int num_commits
;
68 #define PICKAXE_BLAME_MOVE 01
69 #define PICKAXE_BLAME_COPY 02
70 #define PICKAXE_BLAME_COPY_HARDER 04
71 #define PICKAXE_BLAME_COPY_HARDEST 010
74 * blame for a blame_entry with score lower than these thresholds
75 * is not passed to the parent using move/copy logic.
77 static unsigned blame_move_score
;
78 static unsigned blame_copy_score
;
79 #define BLAME_DEFAULT_MOVE_SCORE 20
80 #define BLAME_DEFAULT_COPY_SCORE 40
82 /* Remember to update object flag allocation in object.h */
83 #define METAINFO_SHOWN (1u<<12)
84 #define MORE_THAN_ONE_PATH (1u<<13)
87 * One blob in a commit that is being suspected
91 /* Record preceding blame record for this blob */
92 struct origin
*previous
;
93 /* origins are put in a list linked via `next' hanging off the
94 * corresponding commit's util field in order to make finding
95 * them fast. The presence in this chain does not count
96 * towards the origin's reference count. It is tempting to
97 * let it count as long as the commit is pending examination,
98 * but even under circumstances where the commit will be
99 * present multiple times in the priority queue of unexamined
100 * commits, processing the first instance will not leave any
101 * work requiring the origin data for the second instance. An
102 * interspersed commit changing that would have to be
103 * preexisting with a different ancestry and with the same
104 * commit date in order to wedge itself between two instances
105 * of the same commit in the priority queue _and_ produce
106 * blame entries relevant for it. While we don't want to let
107 * us get tripped up by this case, it certainly does not seem
108 * worth optimizing for.
111 struct commit
*commit
;
112 /* `suspects' contains blame entries that may be attributed to
113 * this origin's commit or to parent commits. When a commit
114 * is being processed, all suspects will be moved, either by
115 * assigning them to an origin in a different commit, or by
116 * shipping them to the scoreboard's ent list because they
117 * cannot be attributed to a different commit.
119 struct blame_entry
*suspects
;
121 unsigned char blob_sha1
[20];
123 /* guilty gets set when shipping any suspects to the final
124 * blame list instead of other commits
127 char path
[FLEX_ARRAY
];
130 static int diff_hunks(mmfile_t
*file_a
, mmfile_t
*file_b
, long ctxlen
,
131 xdl_emit_hunk_consume_func_t hunk_func
, void *cb_data
)
134 xdemitconf_t xecfg
= {0};
135 xdemitcb_t ecb
= {NULL
};
137 xpp
.flags
= xdl_opts
;
138 xecfg
.ctxlen
= ctxlen
;
139 xecfg
.hunk_func
= hunk_func
;
141 return xdi_diff(file_a
, file_b
, &xpp
, &xecfg
, &ecb
);
145 * Prepare diff_filespec and convert it using diff textconv API
146 * if the textconv driver exists.
147 * Return 1 if the conversion succeeds, 0 otherwise.
149 int textconv_object(const char *path
,
151 const unsigned char *sha1
,
154 unsigned long *buf_size
)
156 struct diff_filespec
*df
;
157 struct userdiff_driver
*textconv
;
159 df
= alloc_filespec(path
);
160 fill_filespec(df
, sha1
, sha1_valid
, mode
);
161 textconv
= get_textconv(df
);
167 *buf_size
= fill_textconv(textconv
, df
, buf
);
173 * Given an origin, prepare mmfile_t structure to be used by the
176 static void fill_origin_blob(struct diff_options
*opt
,
177 struct origin
*o
, mmfile_t
*file
)
180 enum object_type type
;
181 unsigned long file_size
;
184 if (DIFF_OPT_TST(opt
, ALLOW_TEXTCONV
) &&
185 textconv_object(o
->path
, o
->mode
, o
->blob_sha1
, 1, &file
->ptr
, &file_size
))
188 file
->ptr
= read_sha1_file(o
->blob_sha1
, &type
, &file_size
);
189 file
->size
= file_size
;
192 die("Cannot read blob %s for path %s",
193 sha1_to_hex(o
->blob_sha1
),
202 * Origin is refcounted and usually we keep the blob contents to be
205 static inline struct origin
*origin_incref(struct origin
*o
)
212 static void origin_decref(struct origin
*o
)
214 if (o
&& --o
->refcnt
<= 0) {
215 struct origin
*p
, *l
= NULL
;
217 origin_decref(o
->previous
);
219 /* Should be present exactly once in commit chain */
220 for (p
= o
->commit
->util
; p
; l
= p
, p
= p
->next
) {
225 o
->commit
->util
= p
->next
;
230 die("internal error in blame::origin_decref");
234 static void drop_origin_blob(struct origin
*o
)
243 * Each group of lines is described by a blame_entry; it can be split
244 * as we pass blame to the parents. They are arranged in linked lists
245 * kept as `suspects' of some unprocessed origin, or entered (when the
246 * blame origin has been finalized) into the scoreboard structure.
247 * While the scoreboard structure is only sorted at the end of
248 * processing (according to final image line number), the lists
249 * attached to an origin are sorted by the target line number.
252 struct blame_entry
*next
;
254 /* the first line of this group in the final image;
255 * internally all line numbers are 0 based.
259 /* how many lines this group has */
262 /* the commit that introduced this group into the final image */
263 struct origin
*suspect
;
265 /* the line number of the first line of this group in the
266 * suspect's file; internally all line numbers are 0 based.
270 /* how significant this entry is -- cached to avoid
271 * scanning the lines over and over.
277 * Any merge of blames happens on lists of blames that arrived via
278 * different parents in a single suspect. In this case, we want to
279 * sort according to the suspect line numbers as opposed to the final
280 * image line numbers. The function body is somewhat longish because
281 * it avoids unnecessary writes.
284 static struct blame_entry
*blame_merge(struct blame_entry
*list1
,
285 struct blame_entry
*list2
)
287 struct blame_entry
*p1
= list1
, *p2
= list2
,
295 if (p1
->s_lno
<= p2
->s_lno
) {
298 if ((p1
= *tail
) == NULL
) {
302 } while (p1
->s_lno
<= p2
->s_lno
);
308 if ((p2
= *tail
) == NULL
) {
312 } while (p1
->s_lno
> p2
->s_lno
);
316 if ((p1
= *tail
) == NULL
) {
320 } while (p1
->s_lno
<= p2
->s_lno
);
324 static void *get_next_blame(const void *p
)
326 return ((struct blame_entry
*)p
)->next
;
329 static void set_next_blame(void *p1
, void *p2
)
331 ((struct blame_entry
*)p1
)->next
= p2
;
335 * Final image line numbers are all different, so we don't need a
336 * three-way comparison here.
339 static int compare_blame_final(const void *p1
, const void *p2
)
341 return ((struct blame_entry
*)p1
)->lno
> ((struct blame_entry
*)p2
)->lno
345 static int compare_blame_suspect(const void *p1
, const void *p2
)
347 const struct blame_entry
*s1
= p1
, *s2
= p2
;
349 * to allow for collating suspects, we sort according to the
350 * respective pointer value as the primary sorting criterion.
351 * The actual relation is pretty unimportant as long as it
352 * establishes a total order. Comparing as integers gives us
355 if (s1
->suspect
!= s2
->suspect
)
356 return (intptr_t)s1
->suspect
> (intptr_t)s2
->suspect
? 1 : -1;
357 if (s1
->s_lno
== s2
->s_lno
)
359 return s1
->s_lno
> s2
->s_lno
? 1 : -1;
362 static struct blame_entry
*blame_sort(struct blame_entry
*head
,
363 int (*compare_fn
)(const void *, const void *))
365 return llist_mergesort (head
, get_next_blame
, set_next_blame
, compare_fn
);
368 static int compare_commits_by_reverse_commit_date(const void *a
,
372 return -compare_commits_by_commit_date(a
, b
, c
);
376 * The current state of the blame assignment.
379 /* the final commit (i.e. where we started digging from) */
380 struct commit
*final
;
381 /* Priority queue for commits with unassigned blame records */
382 struct prio_queue commits
;
383 struct rev_info
*revs
;
387 * The contents in the final image.
388 * Used by many functions to obtain contents of the nth line,
389 * indexed with scoreboard.lineno[blame_entry.lno].
391 const char *final_buf
;
392 unsigned long final_buf_size
;
394 /* linked list of blames */
395 struct blame_entry
*ent
;
397 /* look-up a line in the final buffer */
402 static void sanity_check_refcnt(struct scoreboard
*);
405 * If two blame entries that are next to each other came from
406 * contiguous lines in the same origin (i.e. <commit, path> pair),
407 * merge them together.
409 static void coalesce(struct scoreboard
*sb
)
411 struct blame_entry
*ent
, *next
;
413 for (ent
= sb
->ent
; ent
&& (next
= ent
->next
); ent
= next
) {
414 if (ent
->suspect
== next
->suspect
&&
415 ent
->s_lno
+ ent
->num_lines
== next
->s_lno
) {
416 ent
->num_lines
+= next
->num_lines
;
417 ent
->next
= next
->next
;
418 origin_decref(next
->suspect
);
421 next
= ent
; /* again */
425 if (DEBUG
) /* sanity */
426 sanity_check_refcnt(sb
);
430 * Merge the given sorted list of blames into a preexisting origin.
431 * If there were no previous blames to that commit, it is entered into
432 * the commit priority queue of the score board.
435 static void queue_blames(struct scoreboard
*sb
, struct origin
*porigin
,
436 struct blame_entry
*sorted
)
438 if (porigin
->suspects
)
439 porigin
->suspects
= blame_merge(porigin
->suspects
, sorted
);
442 for (o
= porigin
->commit
->util
; o
; o
= o
->next
) {
444 porigin
->suspects
= sorted
;
448 porigin
->suspects
= sorted
;
449 prio_queue_put(&sb
->commits
, porigin
->commit
);
454 * Given a commit and a path in it, create a new origin structure.
455 * The callers that add blame to the scoreboard should use
456 * get_origin() to obtain shared, refcounted copy instead of calling
457 * this function directly.
459 static struct origin
*make_origin(struct commit
*commit
, const char *path
)
462 o
= xcalloc(1, sizeof(*o
) + strlen(path
) + 1);
465 o
->next
= commit
->util
;
467 strcpy(o
->path
, path
);
472 * Locate an existing origin or create a new one.
473 * This moves the origin to front position in the commit util list.
475 static struct origin
*get_origin(struct scoreboard
*sb
,
476 struct commit
*commit
,
479 struct origin
*o
, *l
;
481 for (o
= commit
->util
, l
= NULL
; o
; l
= o
, o
= o
->next
) {
482 if (!strcmp(o
->path
, path
)) {
486 o
->next
= commit
->util
;
489 return origin_incref(o
);
492 return make_origin(commit
, path
);
496 * Fill the blob_sha1 field of an origin if it hasn't, so that later
497 * call to fill_origin_blob() can use it to locate the data. blob_sha1
498 * for an origin is also used to pass the blame for the entire file to
499 * the parent to detect the case where a child's blob is identical to
500 * that of its parent's.
502 * This also fills origin->mode for corresponding tree path.
504 static int fill_blob_sha1_and_mode(struct origin
*origin
)
506 if (!is_null_sha1(origin
->blob_sha1
))
508 if (get_tree_entry(origin
->commit
->object
.sha1
,
510 origin
->blob_sha1
, &origin
->mode
))
512 if (sha1_object_info(origin
->blob_sha1
, NULL
) != OBJ_BLOB
)
516 hashclr(origin
->blob_sha1
);
517 origin
->mode
= S_IFINVALID
;
522 * We have an origin -- check if the same path exists in the
523 * parent and return an origin structure to represent it.
525 static struct origin
*find_origin(struct scoreboard
*sb
,
526 struct commit
*parent
,
527 struct origin
*origin
)
529 struct origin
*porigin
;
530 struct diff_options diff_opts
;
531 const char *paths
[2];
533 /* First check any existing origins */
534 for (porigin
= parent
->util
; porigin
; porigin
= porigin
->next
)
535 if (!strcmp(porigin
->path
, origin
->path
)) {
537 * The same path between origin and its parent
538 * without renaming -- the most common case.
540 return origin_incref (porigin
);
543 /* See if the origin->path is different between parent
544 * and origin first. Most of the time they are the
545 * same and diff-tree is fairly efficient about this.
547 diff_setup(&diff_opts
);
548 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
549 diff_opts
.detect_rename
= 0;
550 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
551 paths
[0] = origin
->path
;
554 parse_pathspec(&diff_opts
.pathspec
,
555 PATHSPEC_ALL_MAGIC
& ~PATHSPEC_LITERAL
,
556 PATHSPEC_LITERAL_PATH
, "", paths
);
557 diff_setup_done(&diff_opts
);
559 if (is_null_sha1(origin
->commit
->object
.sha1
))
560 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
562 diff_tree_sha1(parent
->tree
->object
.sha1
,
563 origin
->commit
->tree
->object
.sha1
,
565 diffcore_std(&diff_opts
);
567 if (!diff_queued_diff
.nr
) {
568 /* The path is the same as parent */
569 porigin
= get_origin(sb
, parent
, origin
->path
);
570 hashcpy(porigin
->blob_sha1
, origin
->blob_sha1
);
571 porigin
->mode
= origin
->mode
;
574 * Since origin->path is a pathspec, if the parent
575 * commit had it as a directory, we will see a whole
576 * bunch of deletion of files in the directory that we
580 struct diff_filepair
*p
= NULL
;
581 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
583 p
= diff_queued_diff
.queue
[i
];
584 name
= p
->one
->path
? p
->one
->path
: p
->two
->path
;
585 if (!strcmp(name
, origin
->path
))
589 die("internal error in blame::find_origin");
592 die("internal error in blame::find_origin (%c)",
595 porigin
= get_origin(sb
, parent
, origin
->path
);
596 hashcpy(porigin
->blob_sha1
, p
->one
->sha1
);
597 porigin
->mode
= p
->one
->mode
;
601 /* Did not exist in parent, or type changed */
605 diff_flush(&diff_opts
);
606 free_pathspec(&diff_opts
.pathspec
);
611 * We have an origin -- find the path that corresponds to it in its
612 * parent and return an origin structure to represent it.
614 static struct origin
*find_rename(struct scoreboard
*sb
,
615 struct commit
*parent
,
616 struct origin
*origin
)
618 struct origin
*porigin
= NULL
;
619 struct diff_options diff_opts
;
622 diff_setup(&diff_opts
);
623 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
624 diff_opts
.detect_rename
= DIFF_DETECT_RENAME
;
625 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
626 diff_opts
.single_follow
= origin
->path
;
627 diff_setup_done(&diff_opts
);
629 if (is_null_sha1(origin
->commit
->object
.sha1
))
630 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
632 diff_tree_sha1(parent
->tree
->object
.sha1
,
633 origin
->commit
->tree
->object
.sha1
,
635 diffcore_std(&diff_opts
);
637 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
638 struct diff_filepair
*p
= diff_queued_diff
.queue
[i
];
639 if ((p
->status
== 'R' || p
->status
== 'C') &&
640 !strcmp(p
->two
->path
, origin
->path
)) {
641 porigin
= get_origin(sb
, parent
, p
->one
->path
);
642 hashcpy(porigin
->blob_sha1
, p
->one
->sha1
);
643 porigin
->mode
= p
->one
->mode
;
647 diff_flush(&diff_opts
);
648 free_pathspec(&diff_opts
.pathspec
);
653 * Append a new blame entry to a given output queue.
655 static void add_blame_entry(struct blame_entry
***queue
, struct blame_entry
*e
)
657 origin_incref(e
->suspect
);
665 * src typically is on-stack; we want to copy the information in it to
666 * a malloced blame_entry that gets added to the given queue. The
667 * origin of dst loses a refcnt.
669 static void dup_entry(struct blame_entry
***queue
,
670 struct blame_entry
*dst
, struct blame_entry
*src
)
672 origin_incref(src
->suspect
);
673 origin_decref(dst
->suspect
);
674 memcpy(dst
, src
, sizeof(*src
));
680 static const char *nth_line(struct scoreboard
*sb
, long lno
)
682 return sb
->final_buf
+ sb
->lineno
[lno
];
685 static const char *nth_line_cb(void *data
, long lno
)
687 return nth_line((struct scoreboard
*)data
, lno
);
691 * It is known that lines between tlno to same came from parent, and e
692 * has an overlap with that range. it also is known that parent's
693 * line plno corresponds to e's line tlno.
699 * <------------------>
701 * Split e into potentially three parts; before this chunk, the chunk
702 * to be blamed for the parent, and after that portion.
704 static void split_overlap(struct blame_entry
*split
,
705 struct blame_entry
*e
,
706 int tlno
, int plno
, int same
,
707 struct origin
*parent
)
710 memset(split
, 0, sizeof(struct blame_entry
[3]));
712 if (e
->s_lno
< tlno
) {
713 /* there is a pre-chunk part not blamed on parent */
714 split
[0].suspect
= origin_incref(e
->suspect
);
715 split
[0].lno
= e
->lno
;
716 split
[0].s_lno
= e
->s_lno
;
717 split
[0].num_lines
= tlno
- e
->s_lno
;
718 split
[1].lno
= e
->lno
+ tlno
- e
->s_lno
;
719 split
[1].s_lno
= plno
;
722 split
[1].lno
= e
->lno
;
723 split
[1].s_lno
= plno
+ (e
->s_lno
- tlno
);
726 if (same
< e
->s_lno
+ e
->num_lines
) {
727 /* there is a post-chunk part not blamed on parent */
728 split
[2].suspect
= origin_incref(e
->suspect
);
729 split
[2].lno
= e
->lno
+ (same
- e
->s_lno
);
730 split
[2].s_lno
= e
->s_lno
+ (same
- e
->s_lno
);
731 split
[2].num_lines
= e
->s_lno
+ e
->num_lines
- same
;
732 chunk_end_lno
= split
[2].lno
;
735 chunk_end_lno
= e
->lno
+ e
->num_lines
;
736 split
[1].num_lines
= chunk_end_lno
- split
[1].lno
;
739 * if it turns out there is nothing to blame the parent for,
740 * forget about the splitting. !split[1].suspect signals this.
742 if (split
[1].num_lines
< 1)
744 split
[1].suspect
= origin_incref(parent
);
748 * split_overlap() divided an existing blame e into up to three parts
749 * in split. Any assigned blame is moved to queue to
752 static void split_blame(struct blame_entry
***blamed
,
753 struct blame_entry
***unblamed
,
754 struct blame_entry
*split
,
755 struct blame_entry
*e
)
757 struct blame_entry
*new_entry
;
759 if (split
[0].suspect
&& split
[2].suspect
) {
760 /* The first part (reuse storage for the existing entry e) */
761 dup_entry(unblamed
, e
, &split
[0]);
763 /* The last part -- me */
764 new_entry
= xmalloc(sizeof(*new_entry
));
765 memcpy(new_entry
, &(split
[2]), sizeof(struct blame_entry
));
766 add_blame_entry(unblamed
, new_entry
);
768 /* ... and the middle part -- parent */
769 new_entry
= xmalloc(sizeof(*new_entry
));
770 memcpy(new_entry
, &(split
[1]), sizeof(struct blame_entry
));
771 add_blame_entry(blamed
, new_entry
);
773 else if (!split
[0].suspect
&& !split
[2].suspect
)
775 * The parent covers the entire area; reuse storage for
776 * e and replace it with the parent.
778 dup_entry(blamed
, e
, &split
[1]);
779 else if (split
[0].suspect
) {
780 /* me and then parent */
781 dup_entry(unblamed
, e
, &split
[0]);
783 new_entry
= xmalloc(sizeof(*new_entry
));
784 memcpy(new_entry
, &(split
[1]), sizeof(struct blame_entry
));
785 add_blame_entry(blamed
, new_entry
);
788 /* parent and then me */
789 dup_entry(blamed
, e
, &split
[1]);
791 new_entry
= xmalloc(sizeof(*new_entry
));
792 memcpy(new_entry
, &(split
[2]), sizeof(struct blame_entry
));
793 add_blame_entry(unblamed
, new_entry
);
798 * After splitting the blame, the origins used by the
799 * on-stack blame_entry should lose one refcnt each.
801 static void decref_split(struct blame_entry
*split
)
805 for (i
= 0; i
< 3; i
++)
806 origin_decref(split
[i
].suspect
);
810 * reverse_blame reverses the list given in head, appending tail.
811 * That allows us to build lists in reverse order, then reverse them
812 * afterwards. This can be faster than building the list in proper
813 * order right away. The reason is that building in proper order
814 * requires writing a link in the _previous_ element, while building
815 * in reverse order just requires placing the list head into the
819 static struct blame_entry
*reverse_blame(struct blame_entry
*head
,
820 struct blame_entry
*tail
)
823 struct blame_entry
*next
= head
->next
;
832 * Process one hunk from the patch between the current suspect for
833 * blame_entry e and its parent. This first blames any unfinished
834 * entries before the chunk (which is where target and parent start
835 * differing) on the parent, and then splits blame entries at the
836 * start and at the end of the difference region. Since use of -M and
837 * -C options may lead to overlapping/duplicate source line number
838 * ranges, all we can rely on from sorting/merging is the order of the
839 * first suspect line number.
841 static void blame_chunk(struct blame_entry
***dstq
, struct blame_entry
***srcq
,
842 int tlno
, int offset
, int same
,
843 struct origin
*parent
)
845 struct blame_entry
*e
= **srcq
;
846 struct blame_entry
*samep
= NULL
, *diffp
= NULL
;
848 while (e
&& e
->s_lno
< tlno
) {
849 struct blame_entry
*next
= e
->next
;
851 * current record starts before differing portion. If
852 * it reaches into it, we need to split it up and
853 * examine the second part separately.
855 if (e
->s_lno
+ e
->num_lines
> tlno
) {
856 /* Move second half to a new record */
857 int len
= tlno
- e
->s_lno
;
858 struct blame_entry
*n
= xcalloc(1, sizeof (struct blame_entry
));
859 n
->suspect
= e
->suspect
;
860 n
->lno
= e
->lno
+ len
;
861 n
->s_lno
= e
->s_lno
+ len
;
862 n
->num_lines
= e
->num_lines
- len
;
865 /* Push new record to diffp */
869 origin_decref(e
->suspect
);
870 /* Pass blame for everything before the differing
871 * chunk to the parent */
872 e
->suspect
= origin_incref(parent
);
879 * As we don't know how much of a common stretch after this
880 * diff will occur, the currently blamed parts are all that we
881 * can assign to the parent for now.
885 **dstq
= reverse_blame(samep
, **dstq
);
886 *dstq
= &samep
->next
;
889 * Prepend the split off portions: everything after e starts
890 * after the blameable portion.
892 e
= reverse_blame(diffp
, e
);
895 * Now retain records on the target while parts are different
900 while (e
&& e
->s_lno
< same
) {
901 struct blame_entry
*next
= e
->next
;
904 * If current record extends into sameness, need to split.
906 if (e
->s_lno
+ e
->num_lines
> same
) {
908 * Move second half to a new record to be
909 * processed by later chunks
911 int len
= same
- e
->s_lno
;
912 struct blame_entry
*n
= xcalloc(1, sizeof (struct blame_entry
));
913 n
->suspect
= origin_incref(e
->suspect
);
914 n
->lno
= e
->lno
+ len
;
915 n
->s_lno
= e
->s_lno
+ len
;
916 n
->num_lines
= e
->num_lines
- len
;
919 /* Push new record to samep */
927 **srcq
= reverse_blame(diffp
, reverse_blame(samep
, e
));
928 /* Move across elements that are in the unblamable portion */
930 *srcq
= &diffp
->next
;
933 struct blame_chunk_cb_data
{
934 struct origin
*parent
;
936 struct blame_entry
**dstq
;
937 struct blame_entry
**srcq
;
940 /* diff chunks are from parent to target */
941 static int blame_chunk_cb(long start_a
, long count_a
,
942 long start_b
, long count_b
, void *data
)
944 struct blame_chunk_cb_data
*d
= data
;
945 if (start_a
- start_b
!= d
->offset
)
946 die("internal error in blame::blame_chunk_cb");
947 blame_chunk(&d
->dstq
, &d
->srcq
, start_b
, start_a
- start_b
,
948 start_b
+ count_b
, d
->parent
);
949 d
->offset
= start_a
+ count_a
- (start_b
+ count_b
);
954 * We are looking at the origin 'target' and aiming to pass blame
955 * for the lines it is suspected to its parent. Run diff to find
956 * which lines came from parent and pass blame for them.
958 static void pass_blame_to_parent(struct scoreboard
*sb
,
959 struct origin
*target
,
960 struct origin
*parent
)
962 mmfile_t file_p
, file_o
;
963 struct blame_chunk_cb_data d
;
964 struct blame_entry
*newdest
= NULL
;
966 if (!target
->suspects
)
967 return; /* nothing remains for this target */
971 d
.dstq
= &newdest
; d
.srcq
= &target
->suspects
;
973 fill_origin_blob(&sb
->revs
->diffopt
, parent
, &file_p
);
974 fill_origin_blob(&sb
->revs
->diffopt
, target
, &file_o
);
977 if (diff_hunks(&file_p
, &file_o
, 0, blame_chunk_cb
, &d
))
978 die("unable to generate diff (%s -> %s)",
979 sha1_to_hex(parent
->commit
->object
.sha1
),
980 sha1_to_hex(target
->commit
->object
.sha1
));
981 /* The rest are the same as the parent */
982 blame_chunk(&d
.dstq
, &d
.srcq
, INT_MAX
, d
.offset
, INT_MAX
, parent
);
984 queue_blames(sb
, parent
, newdest
);
990 * The lines in blame_entry after splitting blames many times can become
991 * very small and trivial, and at some point it becomes pointless to
992 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
993 * ordinary C program, and it is not worth to say it was copied from
994 * totally unrelated file in the parent.
996 * Compute how trivial the lines in the blame_entry are.
998 static unsigned ent_score(struct scoreboard
*sb
, struct blame_entry
*e
)
1001 const char *cp
, *ep
;
1007 cp
= nth_line(sb
, e
->lno
);
1008 ep
= nth_line(sb
, e
->lno
+ e
->num_lines
);
1010 unsigned ch
= *((unsigned char *)cp
);
1020 * best_so_far[] and this[] are both a split of an existing blame_entry
1021 * that passes blame to the parent. Maintain best_so_far the best split
1022 * so far, by comparing this and best_so_far and copying this into
1023 * bst_so_far as needed.
1025 static void copy_split_if_better(struct scoreboard
*sb
,
1026 struct blame_entry
*best_so_far
,
1027 struct blame_entry
*this)
1031 if (!this[1].suspect
)
1033 if (best_so_far
[1].suspect
) {
1034 if (ent_score(sb
, &this[1]) < ent_score(sb
, &best_so_far
[1]))
1038 for (i
= 0; i
< 3; i
++)
1039 origin_incref(this[i
].suspect
);
1040 decref_split(best_so_far
);
1041 memcpy(best_so_far
, this, sizeof(struct blame_entry
[3]));
1045 * We are looking at a part of the final image represented by
1046 * ent (tlno and same are offset by ent->s_lno).
1047 * tlno is where we are looking at in the final image.
1048 * up to (but not including) same match preimage.
1049 * plno is where we are looking at in the preimage.
1051 * <-------------- final image ---------------------->
1054 * <---------preimage----->
1057 * All line numbers are 0-based.
1059 static void handle_split(struct scoreboard
*sb
,
1060 struct blame_entry
*ent
,
1061 int tlno
, int plno
, int same
,
1062 struct origin
*parent
,
1063 struct blame_entry
*split
)
1065 if (ent
->num_lines
<= tlno
)
1068 struct blame_entry
this[3];
1071 split_overlap(this, ent
, tlno
, plno
, same
, parent
);
1072 copy_split_if_better(sb
, split
, this);
1077 struct handle_split_cb_data
{
1078 struct scoreboard
*sb
;
1079 struct blame_entry
*ent
;
1080 struct origin
*parent
;
1081 struct blame_entry
*split
;
1086 static int handle_split_cb(long start_a
, long count_a
,
1087 long start_b
, long count_b
, void *data
)
1089 struct handle_split_cb_data
*d
= data
;
1090 handle_split(d
->sb
, d
->ent
, d
->tlno
, d
->plno
, start_b
, d
->parent
,
1092 d
->plno
= start_a
+ count_a
;
1093 d
->tlno
= start_b
+ count_b
;
1098 * Find the lines from parent that are the same as ent so that
1099 * we can pass blames to it. file_p has the blob contents for
1102 static void find_copy_in_blob(struct scoreboard
*sb
,
1103 struct blame_entry
*ent
,
1104 struct origin
*parent
,
1105 struct blame_entry
*split
,
1110 struct handle_split_cb_data d
;
1112 memset(&d
, 0, sizeof(d
));
1113 d
.sb
= sb
; d
.ent
= ent
; d
.parent
= parent
; d
.split
= split
;
1115 * Prepare mmfile that contains only the lines in ent.
1117 cp
= nth_line(sb
, ent
->lno
);
1118 file_o
.ptr
= (char *) cp
;
1119 file_o
.size
= nth_line(sb
, ent
->lno
+ ent
->num_lines
) - cp
;
1122 * file_o is a part of final image we are annotating.
1123 * file_p partially may match that image.
1125 memset(split
, 0, sizeof(struct blame_entry
[3]));
1126 if (diff_hunks(file_p
, &file_o
, 1, handle_split_cb
, &d
))
1127 die("unable to generate diff (%s)",
1128 sha1_to_hex(parent
->commit
->object
.sha1
));
1129 /* remainder, if any, all match the preimage */
1130 handle_split(sb
, ent
, d
.tlno
, d
.plno
, ent
->num_lines
, parent
, split
);
1133 /* Move all blame entries from list *source that have a score smaller
1134 * than score_min to the front of list *small.
1135 * Returns a pointer to the link pointing to the old head of the small list.
1138 static struct blame_entry
**filter_small(struct scoreboard
*sb
,
1139 struct blame_entry
**small
,
1140 struct blame_entry
**source
,
1143 struct blame_entry
*p
= *source
;
1144 struct blame_entry
*oldsmall
= *small
;
1146 if (ent_score(sb
, p
) <= score_min
) {
1162 * See if lines currently target is suspected for can be attributed to
1165 static void find_move_in_parent(struct scoreboard
*sb
,
1166 struct blame_entry
***blamed
,
1167 struct blame_entry
**toosmall
,
1168 struct origin
*target
,
1169 struct origin
*parent
)
1171 struct blame_entry
*e
, split
[3];
1172 struct blame_entry
*unblamed
= target
->suspects
;
1173 struct blame_entry
*leftover
= NULL
;
1177 return; /* nothing remains for this target */
1179 fill_origin_blob(&sb
->revs
->diffopt
, parent
, &file_p
);
1183 /* At each iteration, unblamed has a NULL-terminated list of
1184 * entries that have not yet been tested for blame. leftover
1185 * contains the reversed list of entries that have been tested
1186 * without being assignable to the parent.
1189 struct blame_entry
**unblamedtail
= &unblamed
;
1190 struct blame_entry
*next
;
1191 for (e
= unblamed
; e
; e
= next
) {
1193 find_copy_in_blob(sb
, e
, parent
, split
, &file_p
);
1194 if (split
[1].suspect
&&
1195 blame_move_score
< ent_score(sb
, &split
[1])) {
1196 split_blame(blamed
, &unblamedtail
, split
, e
);
1201 decref_split(split
);
1203 *unblamedtail
= NULL
;
1204 toosmall
= filter_small(sb
, toosmall
, &unblamed
, blame_move_score
);
1206 target
->suspects
= reverse_blame(leftover
, NULL
);
1210 struct blame_entry
*ent
;
1211 struct blame_entry split
[3];
1215 * Count the number of entries the target is suspected for,
1216 * and prepare a list of entry and the best split.
1218 static struct blame_list
*setup_blame_list(struct blame_entry
*unblamed
,
1221 struct blame_entry
*e
;
1223 struct blame_list
*blame_list
= NULL
;
1225 for (e
= unblamed
, num_ents
= 0; e
; e
= e
->next
)
1228 blame_list
= xcalloc(num_ents
, sizeof(struct blame_list
));
1229 for (e
= unblamed
, i
= 0; e
; e
= e
->next
)
1230 blame_list
[i
++].ent
= e
;
1232 *num_ents_p
= num_ents
;
1237 * For lines target is suspected for, see if we can find code movement
1238 * across file boundary from the parent commit. porigin is the path
1239 * in the parent we already tried.
1241 static void find_copy_in_parent(struct scoreboard
*sb
,
1242 struct blame_entry
***blamed
,
1243 struct blame_entry
**toosmall
,
1244 struct origin
*target
,
1245 struct commit
*parent
,
1246 struct origin
*porigin
,
1249 struct diff_options diff_opts
;
1251 struct blame_list
*blame_list
;
1253 struct blame_entry
*unblamed
= target
->suspects
;
1254 struct blame_entry
*leftover
= NULL
;
1257 return; /* nothing remains for this target */
1259 diff_setup(&diff_opts
);
1260 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
1261 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
1263 diff_setup_done(&diff_opts
);
1265 /* Try "find copies harder" on new path if requested;
1266 * we do not want to use diffcore_rename() actually to
1267 * match things up; find_copies_harder is set only to
1268 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1269 * and this code needs to be after diff_setup_done(), which
1270 * usually makes find-copies-harder imply copy detection.
1272 if ((opt
& PICKAXE_BLAME_COPY_HARDEST
)
1273 || ((opt
& PICKAXE_BLAME_COPY_HARDER
)
1274 && (!porigin
|| strcmp(target
->path
, porigin
->path
))))
1275 DIFF_OPT_SET(&diff_opts
, FIND_COPIES_HARDER
);
1277 if (is_null_sha1(target
->commit
->object
.sha1
))
1278 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
1280 diff_tree_sha1(parent
->tree
->object
.sha1
,
1281 target
->commit
->tree
->object
.sha1
,
1284 if (!DIFF_OPT_TST(&diff_opts
, FIND_COPIES_HARDER
))
1285 diffcore_std(&diff_opts
);
1288 struct blame_entry
**unblamedtail
= &unblamed
;
1289 blame_list
= setup_blame_list(unblamed
, &num_ents
);
1291 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
1292 struct diff_filepair
*p
= diff_queued_diff
.queue
[i
];
1293 struct origin
*norigin
;
1295 struct blame_entry
this[3];
1297 if (!DIFF_FILE_VALID(p
->one
))
1298 continue; /* does not exist in parent */
1299 if (S_ISGITLINK(p
->one
->mode
))
1300 continue; /* ignore git links */
1301 if (porigin
&& !strcmp(p
->one
->path
, porigin
->path
))
1302 /* find_move already dealt with this path */
1305 norigin
= get_origin(sb
, parent
, p
->one
->path
);
1306 hashcpy(norigin
->blob_sha1
, p
->one
->sha1
);
1307 norigin
->mode
= p
->one
->mode
;
1308 fill_origin_blob(&sb
->revs
->diffopt
, norigin
, &file_p
);
1312 for (j
= 0; j
< num_ents
; j
++) {
1313 find_copy_in_blob(sb
, blame_list
[j
].ent
,
1314 norigin
, this, &file_p
);
1315 copy_split_if_better(sb
, blame_list
[j
].split
,
1319 origin_decref(norigin
);
1322 for (j
= 0; j
< num_ents
; j
++) {
1323 struct blame_entry
*split
= blame_list
[j
].split
;
1324 if (split
[1].suspect
&&
1325 blame_copy_score
< ent_score(sb
, &split
[1])) {
1326 split_blame(blamed
, &unblamedtail
, split
,
1329 blame_list
[j
].ent
->next
= leftover
;
1330 leftover
= blame_list
[j
].ent
;
1332 decref_split(split
);
1335 *unblamedtail
= NULL
;
1336 toosmall
= filter_small(sb
, toosmall
, &unblamed
, blame_copy_score
);
1338 target
->suspects
= reverse_blame(leftover
, NULL
);
1339 diff_flush(&diff_opts
);
1340 free_pathspec(&diff_opts
.pathspec
);
1344 * The blobs of origin and porigin exactly match, so everything
1345 * origin is suspected for can be blamed on the parent.
1347 static void pass_whole_blame(struct scoreboard
*sb
,
1348 struct origin
*origin
, struct origin
*porigin
)
1350 struct blame_entry
*e
, *suspects
;
1352 if (!porigin
->file
.ptr
&& origin
->file
.ptr
) {
1353 /* Steal its file */
1354 porigin
->file
= origin
->file
;
1355 origin
->file
.ptr
= NULL
;
1357 suspects
= origin
->suspects
;
1358 origin
->suspects
= NULL
;
1359 for (e
= suspects
; e
; e
= e
->next
) {
1360 origin_incref(porigin
);
1361 origin_decref(e
->suspect
);
1362 e
->suspect
= porigin
;
1364 queue_blames(sb
, porigin
, suspects
);
1368 * We pass blame from the current commit to its parents. We keep saying
1369 * "parent" (and "porigin"), but what we mean is to find scapegoat to
1370 * exonerate ourselves.
1372 static struct commit_list
*first_scapegoat(struct rev_info
*revs
, struct commit
*commit
)
1375 return commit
->parents
;
1376 return lookup_decoration(&revs
->children
, &commit
->object
);
1379 static int num_scapegoats(struct rev_info
*revs
, struct commit
*commit
)
1381 struct commit_list
*l
= first_scapegoat(revs
, commit
);
1382 return commit_list_count(l
);
1385 /* Distribute collected unsorted blames to the respected sorted lists
1386 * in the various origins.
1388 static void distribute_blame(struct scoreboard
*sb
, struct blame_entry
*blamed
)
1390 blamed
= blame_sort(blamed
, compare_blame_suspect
);
1393 struct origin
*porigin
= blamed
->suspect
;
1394 struct blame_entry
*suspects
= NULL
;
1396 struct blame_entry
*next
= blamed
->next
;
1397 blamed
->next
= suspects
;
1400 } while (blamed
&& blamed
->suspect
== porigin
);
1401 suspects
= reverse_blame(suspects
, NULL
);
1402 queue_blames(sb
, porigin
, suspects
);
1408 static void pass_blame(struct scoreboard
*sb
, struct origin
*origin
, int opt
)
1410 struct rev_info
*revs
= sb
->revs
;
1411 int i
, pass
, num_sg
;
1412 struct commit
*commit
= origin
->commit
;
1413 struct commit_list
*sg
;
1414 struct origin
*sg_buf
[MAXSG
];
1415 struct origin
*porigin
, **sg_origin
= sg_buf
;
1416 struct blame_entry
*toosmall
= NULL
;
1417 struct blame_entry
*blames
, **blametail
= &blames
;
1419 num_sg
= num_scapegoats(revs
, commit
);
1422 else if (num_sg
< ARRAY_SIZE(sg_buf
))
1423 memset(sg_buf
, 0, sizeof(sg_buf
));
1425 sg_origin
= xcalloc(num_sg
, sizeof(*sg_origin
));
1428 * The first pass looks for unrenamed path to optimize for
1429 * common cases, then we look for renames in the second pass.
1431 for (pass
= 0; pass
< 2 - no_whole_file_rename
; pass
++) {
1432 struct origin
*(*find
)(struct scoreboard
*,
1433 struct commit
*, struct origin
*);
1434 find
= pass
? find_rename
: find_origin
;
1436 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1438 sg
= sg
->next
, i
++) {
1439 struct commit
*p
= sg
->item
;
1444 if (parse_commit(p
))
1446 porigin
= find(sb
, p
, origin
);
1449 if (!hashcmp(porigin
->blob_sha1
, origin
->blob_sha1
)) {
1450 pass_whole_blame(sb
, origin
, porigin
);
1451 origin_decref(porigin
);
1454 for (j
= same
= 0; j
< i
; j
++)
1456 !hashcmp(sg_origin
[j
]->blob_sha1
,
1457 porigin
->blob_sha1
)) {
1462 sg_origin
[i
] = porigin
;
1464 origin_decref(porigin
);
1469 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1471 sg
= sg
->next
, i
++) {
1472 struct origin
*porigin
= sg_origin
[i
];
1475 if (!origin
->previous
) {
1476 origin_incref(porigin
);
1477 origin
->previous
= porigin
;
1479 pass_blame_to_parent(sb
, origin
, porigin
);
1480 if (!origin
->suspects
)
1485 * Optionally find moves in parents' files.
1487 if (opt
& PICKAXE_BLAME_MOVE
) {
1488 filter_small(sb
, &toosmall
, &origin
->suspects
, blame_move_score
);
1489 if (origin
->suspects
) {
1490 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1492 sg
= sg
->next
, i
++) {
1493 struct origin
*porigin
= sg_origin
[i
];
1496 find_move_in_parent(sb
, &blametail
, &toosmall
, origin
, porigin
);
1497 if (!origin
->suspects
)
1504 * Optionally find copies from parents' files.
1506 if (opt
& PICKAXE_BLAME_COPY
) {
1507 if (blame_copy_score
> blame_move_score
)
1508 filter_small(sb
, &toosmall
, &origin
->suspects
, blame_copy_score
);
1509 else if (blame_copy_score
< blame_move_score
) {
1510 origin
->suspects
= blame_merge(origin
->suspects
, toosmall
);
1512 filter_small(sb
, &toosmall
, &origin
->suspects
, blame_copy_score
);
1514 if (!origin
->suspects
)
1517 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1519 sg
= sg
->next
, i
++) {
1520 struct origin
*porigin
= sg_origin
[i
];
1521 find_copy_in_parent(sb
, &blametail
, &toosmall
,
1522 origin
, sg
->item
, porigin
, opt
);
1523 if (!origin
->suspects
)
1530 distribute_blame(sb
, blames
);
1532 * prepend toosmall to origin->suspects
1534 * There is no point in sorting: this ends up on a big
1535 * unsorted list in the caller anyway.
1538 struct blame_entry
**tail
= &toosmall
;
1540 tail
= &(*tail
)->next
;
1541 *tail
= origin
->suspects
;
1542 origin
->suspects
= toosmall
;
1544 for (i
= 0; i
< num_sg
; i
++) {
1546 drop_origin_blob(sg_origin
[i
]);
1547 origin_decref(sg_origin
[i
]);
1550 drop_origin_blob(origin
);
1551 if (sg_buf
!= sg_origin
)
1556 * Information on commits, used for output.
1558 struct commit_info
{
1559 struct strbuf author
;
1560 struct strbuf author_mail
;
1561 unsigned long author_time
;
1562 struct strbuf author_tz
;
1564 /* filled only when asked for details */
1565 struct strbuf committer
;
1566 struct strbuf committer_mail
;
1567 unsigned long committer_time
;
1568 struct strbuf committer_tz
;
1570 struct strbuf summary
;
1574 * Parse author/committer line in the commit object buffer
1576 static void get_ac_line(const char *inbuf
, const char *what
,
1577 struct strbuf
*name
, struct strbuf
*mail
,
1578 unsigned long *time
, struct strbuf
*tz
)
1580 struct ident_split ident
;
1581 size_t len
, maillen
, namelen
;
1583 const char *namebuf
, *mailbuf
;
1585 tmp
= strstr(inbuf
, what
);
1588 tmp
+= strlen(what
);
1589 endp
= strchr(tmp
, '\n');
1595 if (split_ident_line(&ident
, tmp
, len
)) {
1599 strbuf_addstr(name
, tmp
);
1600 strbuf_addstr(mail
, tmp
);
1601 strbuf_addstr(tz
, tmp
);
1606 namelen
= ident
.name_end
- ident
.name_begin
;
1607 namebuf
= ident
.name_begin
;
1609 maillen
= ident
.mail_end
- ident
.mail_begin
;
1610 mailbuf
= ident
.mail_begin
;
1612 if (ident
.date_begin
&& ident
.date_end
)
1613 *time
= strtoul(ident
.date_begin
, NULL
, 10);
1617 if (ident
.tz_begin
&& ident
.tz_end
)
1618 strbuf_add(tz
, ident
.tz_begin
, ident
.tz_end
- ident
.tz_begin
);
1620 strbuf_addstr(tz
, "(unknown)");
1623 * Now, convert both name and e-mail using mailmap
1625 map_user(&mailmap
, &mailbuf
, &maillen
,
1626 &namebuf
, &namelen
);
1628 strbuf_addf(mail
, "<%.*s>", (int)maillen
, mailbuf
);
1629 strbuf_add(name
, namebuf
, namelen
);
1632 static void commit_info_init(struct commit_info
*ci
)
1635 strbuf_init(&ci
->author
, 0);
1636 strbuf_init(&ci
->author_mail
, 0);
1637 strbuf_init(&ci
->author_tz
, 0);
1638 strbuf_init(&ci
->committer
, 0);
1639 strbuf_init(&ci
->committer_mail
, 0);
1640 strbuf_init(&ci
->committer_tz
, 0);
1641 strbuf_init(&ci
->summary
, 0);
1644 static void commit_info_destroy(struct commit_info
*ci
)
1647 strbuf_release(&ci
->author
);
1648 strbuf_release(&ci
->author_mail
);
1649 strbuf_release(&ci
->author_tz
);
1650 strbuf_release(&ci
->committer
);
1651 strbuf_release(&ci
->committer_mail
);
1652 strbuf_release(&ci
->committer_tz
);
1653 strbuf_release(&ci
->summary
);
1656 static void get_commit_info(struct commit
*commit
,
1657 struct commit_info
*ret
,
1661 const char *subject
, *encoding
;
1662 const char *message
;
1664 commit_info_init(ret
);
1666 encoding
= get_log_output_encoding();
1667 message
= logmsg_reencode(commit
, NULL
, encoding
);
1668 get_ac_line(message
, "\nauthor ",
1669 &ret
->author
, &ret
->author_mail
,
1670 &ret
->author_time
, &ret
->author_tz
);
1673 unuse_commit_buffer(commit
, message
);
1677 get_ac_line(message
, "\ncommitter ",
1678 &ret
->committer
, &ret
->committer_mail
,
1679 &ret
->committer_time
, &ret
->committer_tz
);
1681 len
= find_commit_subject(message
, &subject
);
1683 strbuf_add(&ret
->summary
, subject
, len
);
1685 strbuf_addf(&ret
->summary
, "(%s)", sha1_to_hex(commit
->object
.sha1
));
1687 unuse_commit_buffer(commit
, message
);
1691 * To allow LF and other nonportable characters in pathnames,
1692 * they are c-style quoted as needed.
1694 static void write_filename_info(const char *path
)
1696 printf("filename ");
1697 write_name_quoted(path
, stdout
, '\n');
1701 * Porcelain/Incremental format wants to show a lot of details per
1702 * commit. Instead of repeating this every line, emit it only once,
1703 * the first time each commit appears in the output (unless the
1704 * user has specifically asked for us to repeat).
1706 static int emit_one_suspect_detail(struct origin
*suspect
, int repeat
)
1708 struct commit_info ci
;
1710 if (!repeat
&& (suspect
->commit
->object
.flags
& METAINFO_SHOWN
))
1713 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
1714 get_commit_info(suspect
->commit
, &ci
, 1);
1715 printf("author %s\n", ci
.author
.buf
);
1716 printf("author-mail %s\n", ci
.author_mail
.buf
);
1717 printf("author-time %lu\n", ci
.author_time
);
1718 printf("author-tz %s\n", ci
.author_tz
.buf
);
1719 printf("committer %s\n", ci
.committer
.buf
);
1720 printf("committer-mail %s\n", ci
.committer_mail
.buf
);
1721 printf("committer-time %lu\n", ci
.committer_time
);
1722 printf("committer-tz %s\n", ci
.committer_tz
.buf
);
1723 printf("summary %s\n", ci
.summary
.buf
);
1724 if (suspect
->commit
->object
.flags
& UNINTERESTING
)
1725 printf("boundary\n");
1726 if (suspect
->previous
) {
1727 struct origin
*prev
= suspect
->previous
;
1728 printf("previous %s ", sha1_to_hex(prev
->commit
->object
.sha1
));
1729 write_name_quoted(prev
->path
, stdout
, '\n');
1732 commit_info_destroy(&ci
);
1738 * The blame_entry is found to be guilty for the range.
1739 * Show it in incremental output.
1741 static void found_guilty_entry(struct blame_entry
*ent
)
1744 struct origin
*suspect
= ent
->suspect
;
1746 printf("%s %d %d %d\n",
1747 sha1_to_hex(suspect
->commit
->object
.sha1
),
1748 ent
->s_lno
+ 1, ent
->lno
+ 1, ent
->num_lines
);
1749 emit_one_suspect_detail(suspect
, 0);
1750 write_filename_info(suspect
->path
);
1751 maybe_flush_or_die(stdout
, "stdout");
1756 * The main loop -- while we have blobs with lines whose true origin
1757 * is still unknown, pick one blob, and allow its lines to pass blames
1758 * to its parents. */
1759 static void assign_blame(struct scoreboard
*sb
, int opt
)
1761 struct rev_info
*revs
= sb
->revs
;
1762 struct commit
*commit
= prio_queue_get(&sb
->commits
);
1765 struct blame_entry
*ent
;
1766 struct origin
*suspect
= commit
->util
;
1768 /* find one suspect to break down */
1769 while (suspect
&& !suspect
->suspects
)
1770 suspect
= suspect
->next
;
1773 commit
= prio_queue_get(&sb
->commits
);
1777 assert(commit
== suspect
->commit
);
1780 * We will use this suspect later in the loop,
1781 * so hold onto it in the meantime.
1783 origin_incref(suspect
);
1784 parse_commit(commit
);
1786 (!(commit
->object
.flags
& UNINTERESTING
) &&
1787 !(revs
->max_age
!= -1 && commit
->date
< revs
->max_age
)))
1788 pass_blame(sb
, suspect
, opt
);
1790 commit
->object
.flags
|= UNINTERESTING
;
1791 if (commit
->object
.parsed
)
1792 mark_parents_uninteresting(commit
);
1794 /* treat root commit as boundary */
1795 if (!commit
->parents
&& !show_root
)
1796 commit
->object
.flags
|= UNINTERESTING
;
1798 /* Take responsibility for the remaining entries */
1799 ent
= suspect
->suspects
;
1801 suspect
->guilty
= 1;
1803 struct blame_entry
*next
= ent
->next
;
1804 found_guilty_entry(ent
);
1809 ent
->next
= sb
->ent
;
1810 sb
->ent
= suspect
->suspects
;
1811 suspect
->suspects
= NULL
;
1815 origin_decref(suspect
);
1817 if (DEBUG
) /* sanity */
1818 sanity_check_refcnt(sb
);
1822 static const char *format_time(unsigned long time
, const char *tz_str
,
1825 static struct strbuf time_buf
= STRBUF_INIT
;
1827 strbuf_reset(&time_buf
);
1828 if (show_raw_time
) {
1829 strbuf_addf(&time_buf
, "%lu %s", time
, tz_str
);
1832 const char *time_str
;
1836 time_str
= show_date(time
, tz
, &blame_date_mode
);
1837 strbuf_addstr(&time_buf
, time_str
);
1839 * Add space paddings to time_buf to display a fixed width
1840 * string, and use time_width for display width calibration.
1842 for (time_width
= utf8_strwidth(time_str
);
1843 time_width
< blame_date_width
;
1845 strbuf_addch(&time_buf
, ' ');
1847 return time_buf
.buf
;
1850 #define OUTPUT_ANNOTATE_COMPAT 001
1851 #define OUTPUT_LONG_OBJECT_NAME 002
1852 #define OUTPUT_RAW_TIMESTAMP 004
1853 #define OUTPUT_PORCELAIN 010
1854 #define OUTPUT_SHOW_NAME 020
1855 #define OUTPUT_SHOW_NUMBER 040
1856 #define OUTPUT_SHOW_SCORE 0100
1857 #define OUTPUT_NO_AUTHOR 0200
1858 #define OUTPUT_SHOW_EMAIL 0400
1859 #define OUTPUT_LINE_PORCELAIN 01000
1861 static void emit_porcelain_details(struct origin
*suspect
, int repeat
)
1863 if (emit_one_suspect_detail(suspect
, repeat
) ||
1864 (suspect
->commit
->object
.flags
& MORE_THAN_ONE_PATH
))
1865 write_filename_info(suspect
->path
);
1868 static void emit_porcelain(struct scoreboard
*sb
, struct blame_entry
*ent
,
1871 int repeat
= opt
& OUTPUT_LINE_PORCELAIN
;
1874 struct origin
*suspect
= ent
->suspect
;
1877 strcpy(hex
, sha1_to_hex(suspect
->commit
->object
.sha1
));
1878 printf("%s %d %d %d\n",
1883 emit_porcelain_details(suspect
, repeat
);
1885 cp
= nth_line(sb
, ent
->lno
);
1886 for (cnt
= 0; cnt
< ent
->num_lines
; cnt
++) {
1889 printf("%s %d %d\n", hex
,
1890 ent
->s_lno
+ 1 + cnt
,
1891 ent
->lno
+ 1 + cnt
);
1893 emit_porcelain_details(suspect
, 1);
1899 } while (ch
!= '\n' &&
1900 cp
< sb
->final_buf
+ sb
->final_buf_size
);
1903 if (sb
->final_buf_size
&& cp
[-1] != '\n')
1907 static void emit_other(struct scoreboard
*sb
, struct blame_entry
*ent
, int opt
)
1911 struct origin
*suspect
= ent
->suspect
;
1912 struct commit_info ci
;
1914 int show_raw_time
= !!(opt
& OUTPUT_RAW_TIMESTAMP
);
1916 get_commit_info(suspect
->commit
, &ci
, 1);
1917 strcpy(hex
, sha1_to_hex(suspect
->commit
->object
.sha1
));
1919 cp
= nth_line(sb
, ent
->lno
);
1920 for (cnt
= 0; cnt
< ent
->num_lines
; cnt
++) {
1922 int length
= (opt
& OUTPUT_LONG_OBJECT_NAME
) ? 40 : abbrev
;
1924 if (suspect
->commit
->object
.flags
& UNINTERESTING
) {
1926 memset(hex
, ' ', length
);
1927 else if (!(opt
& OUTPUT_ANNOTATE_COMPAT
)) {
1933 printf("%.*s", length
, hex
);
1934 if (opt
& OUTPUT_ANNOTATE_COMPAT
) {
1936 if (opt
& OUTPUT_SHOW_EMAIL
)
1937 name
= ci
.author_mail
.buf
;
1939 name
= ci
.author
.buf
;
1940 printf("\t(%10s\t%10s\t%d)", name
,
1941 format_time(ci
.author_time
, ci
.author_tz
.buf
,
1943 ent
->lno
+ 1 + cnt
);
1945 if (opt
& OUTPUT_SHOW_SCORE
)
1947 max_score_digits
, ent
->score
,
1948 ent
->suspect
->refcnt
);
1949 if (opt
& OUTPUT_SHOW_NAME
)
1950 printf(" %-*.*s", longest_file
, longest_file
,
1952 if (opt
& OUTPUT_SHOW_NUMBER
)
1953 printf(" %*d", max_orig_digits
,
1954 ent
->s_lno
+ 1 + cnt
);
1956 if (!(opt
& OUTPUT_NO_AUTHOR
)) {
1959 if (opt
& OUTPUT_SHOW_EMAIL
)
1960 name
= ci
.author_mail
.buf
;
1962 name
= ci
.author
.buf
;
1963 pad
= longest_author
- utf8_strwidth(name
);
1964 printf(" (%s%*s %10s",
1966 format_time(ci
.author_time
,
1971 max_digits
, ent
->lno
+ 1 + cnt
);
1976 } while (ch
!= '\n' &&
1977 cp
< sb
->final_buf
+ sb
->final_buf_size
);
1980 if (sb
->final_buf_size
&& cp
[-1] != '\n')
1983 commit_info_destroy(&ci
);
1986 static void output(struct scoreboard
*sb
, int option
)
1988 struct blame_entry
*ent
;
1990 if (option
& OUTPUT_PORCELAIN
) {
1991 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1993 struct origin
*suspect
;
1994 struct commit
*commit
= ent
->suspect
->commit
;
1995 if (commit
->object
.flags
& MORE_THAN_ONE_PATH
)
1997 for (suspect
= commit
->util
; suspect
; suspect
= suspect
->next
) {
1998 if (suspect
->guilty
&& count
++) {
1999 commit
->object
.flags
|= MORE_THAN_ONE_PATH
;
2006 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
2007 if (option
& OUTPUT_PORCELAIN
)
2008 emit_porcelain(sb
, ent
, option
);
2010 emit_other(sb
, ent
, option
);
2015 static const char *get_next_line(const char *start
, const char *end
)
2017 const char *nl
= memchr(start
, '\n', end
- start
);
2018 return nl
? nl
+ 1 : end
;
2022 * To allow quick access to the contents of nth line in the
2023 * final image, prepare an index in the scoreboard.
2025 static int prepare_lines(struct scoreboard
*sb
)
2027 const char *buf
= sb
->final_buf
;
2028 unsigned long len
= sb
->final_buf_size
;
2029 const char *end
= buf
+ len
;
2034 for (p
= buf
; p
< end
; p
= get_next_line(p
, end
))
2037 sb
->lineno
= lineno
= xmalloc(sizeof(*sb
->lineno
) * (num
+ 1));
2039 for (p
= buf
; p
< end
; p
= get_next_line(p
, end
))
2040 *lineno
++ = p
- buf
;
2044 sb
->num_lines
= num
;
2045 return sb
->num_lines
;
2049 * Add phony grafts for use with -S; this is primarily to
2050 * support git's cvsserver that wants to give a linear history
2053 static int read_ancestry(const char *graft_file
)
2055 FILE *fp
= fopen(graft_file
, "r");
2056 struct strbuf buf
= STRBUF_INIT
;
2059 while (!strbuf_getwholeline(&buf
, fp
, '\n')) {
2060 /* The format is just "Commit Parent1 Parent2 ...\n" */
2061 struct commit_graft
*graft
= read_graft_line(buf
.buf
, buf
.len
);
2063 register_commit_graft(graft
, 0);
2066 strbuf_release(&buf
);
2070 static int update_auto_abbrev(int auto_abbrev
, struct origin
*suspect
)
2072 const char *uniq
= find_unique_abbrev(suspect
->commit
->object
.sha1
,
2074 int len
= strlen(uniq
);
2075 if (auto_abbrev
< len
)
2081 * How many columns do we need to show line numbers, authors,
2084 static void find_alignment(struct scoreboard
*sb
, int *option
)
2086 int longest_src_lines
= 0;
2087 int longest_dst_lines
= 0;
2088 unsigned largest_score
= 0;
2089 struct blame_entry
*e
;
2090 int compute_auto_abbrev
= (abbrev
< 0);
2091 int auto_abbrev
= default_abbrev
;
2093 for (e
= sb
->ent
; e
; e
= e
->next
) {
2094 struct origin
*suspect
= e
->suspect
;
2097 if (compute_auto_abbrev
)
2098 auto_abbrev
= update_auto_abbrev(auto_abbrev
, suspect
);
2099 if (strcmp(suspect
->path
, sb
->path
))
2100 *option
|= OUTPUT_SHOW_NAME
;
2101 num
= strlen(suspect
->path
);
2102 if (longest_file
< num
)
2104 if (!(suspect
->commit
->object
.flags
& METAINFO_SHOWN
)) {
2105 struct commit_info ci
;
2106 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
2107 get_commit_info(suspect
->commit
, &ci
, 1);
2108 if (*option
& OUTPUT_SHOW_EMAIL
)
2109 num
= utf8_strwidth(ci
.author_mail
.buf
);
2111 num
= utf8_strwidth(ci
.author
.buf
);
2112 if (longest_author
< num
)
2113 longest_author
= num
;
2114 commit_info_destroy(&ci
);
2116 num
= e
->s_lno
+ e
->num_lines
;
2117 if (longest_src_lines
< num
)
2118 longest_src_lines
= num
;
2119 num
= e
->lno
+ e
->num_lines
;
2120 if (longest_dst_lines
< num
)
2121 longest_dst_lines
= num
;
2122 if (largest_score
< ent_score(sb
, e
))
2123 largest_score
= ent_score(sb
, e
);
2125 max_orig_digits
= decimal_width(longest_src_lines
);
2126 max_digits
= decimal_width(longest_dst_lines
);
2127 max_score_digits
= decimal_width(largest_score
);
2129 if (compute_auto_abbrev
)
2130 /* one more abbrev length is needed for the boundary commit */
2131 abbrev
= auto_abbrev
+ 1;
2135 * For debugging -- origin is refcounted, and this asserts that
2136 * we do not underflow.
2138 static void sanity_check_refcnt(struct scoreboard
*sb
)
2141 struct blame_entry
*ent
;
2143 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
2144 /* Nobody should have zero or negative refcnt */
2145 if (ent
->suspect
->refcnt
<= 0) {
2146 fprintf(stderr
, "%s in %s has negative refcnt %d\n",
2148 sha1_to_hex(ent
->suspect
->commit
->object
.sha1
),
2149 ent
->suspect
->refcnt
);
2155 find_alignment(sb
, &opt
);
2157 die("Baa %d!", baa
);
2161 static unsigned parse_score(const char *arg
)
2164 unsigned long score
= strtoul(arg
, &end
, 10);
2170 static const char *add_prefix(const char *prefix
, const char *path
)
2172 return prefix_path(prefix
, prefix
? strlen(prefix
) : 0, path
);
2175 static int git_blame_config(const char *var
, const char *value
, void *cb
)
2177 if (!strcmp(var
, "blame.showroot")) {
2178 show_root
= git_config_bool(var
, value
);
2181 if (!strcmp(var
, "blame.blankboundary")) {
2182 blank_boundary
= git_config_bool(var
, value
);
2185 if (!strcmp(var
, "blame.showemail")) {
2186 int *output_option
= cb
;
2187 if (git_config_bool(var
, value
))
2188 *output_option
|= OUTPUT_SHOW_EMAIL
;
2190 *output_option
&= ~OUTPUT_SHOW_EMAIL
;
2193 if (!strcmp(var
, "blame.date")) {
2195 return config_error_nonbool(var
);
2196 parse_date_format(value
, &blame_date_mode
);
2200 if (userdiff_config(var
, value
) < 0)
2203 return git_default_config(var
, value
, cb
);
2206 static void verify_working_tree_path(struct commit
*work_tree
, const char *path
)
2208 struct commit_list
*parents
;
2210 for (parents
= work_tree
->parents
; parents
; parents
= parents
->next
) {
2211 const unsigned char *commit_sha1
= parents
->item
->object
.sha1
;
2212 unsigned char blob_sha1
[20];
2215 if (!get_tree_entry(commit_sha1
, path
, blob_sha1
, &mode
) &&
2216 sha1_object_info(blob_sha1
, NULL
) == OBJ_BLOB
)
2219 die("no such path '%s' in HEAD", path
);
2222 static struct commit_list
**append_parent(struct commit_list
**tail
, const unsigned char *sha1
)
2224 struct commit
*parent
;
2226 parent
= lookup_commit_reference(sha1
);
2228 die("no such commit %s", sha1_to_hex(sha1
));
2229 return &commit_list_insert(parent
, tail
)->next
;
2232 static void append_merge_parents(struct commit_list
**tail
)
2235 struct strbuf line
= STRBUF_INIT
;
2237 merge_head
= open(git_path_merge_head(), O_RDONLY
);
2238 if (merge_head
< 0) {
2239 if (errno
== ENOENT
)
2241 die("cannot open '%s' for reading", git_path_merge_head());
2244 while (!strbuf_getwholeline_fd(&line
, merge_head
, '\n')) {
2245 unsigned char sha1
[20];
2246 if (line
.len
< 40 || get_sha1_hex(line
.buf
, sha1
))
2247 die("unknown line in '%s': %s", git_path_merge_head(), line
.buf
);
2248 tail
= append_parent(tail
, sha1
);
2251 strbuf_release(&line
);
2255 * This isn't as simple as passing sb->buf and sb->len, because we
2256 * want to transfer ownership of the buffer to the commit (so we
2259 static void set_commit_buffer_from_strbuf(struct commit
*c
, struct strbuf
*sb
)
2262 void *buf
= strbuf_detach(sb
, &len
);
2263 set_commit_buffer(c
, buf
, len
);
2267 * Prepare a dummy commit that represents the work tree (or staged) item.
2268 * Note that annotating work tree item never works in the reverse.
2270 static struct commit
*fake_working_tree_commit(struct diff_options
*opt
,
2272 const char *contents_from
)
2274 struct commit
*commit
;
2275 struct origin
*origin
;
2276 struct commit_list
**parent_tail
, *parent
;
2277 unsigned char head_sha1
[20];
2278 struct strbuf buf
= STRBUF_INIT
;
2282 struct cache_entry
*ce
;
2284 struct strbuf msg
= STRBUF_INIT
;
2287 commit
= alloc_commit_node();
2288 commit
->object
.parsed
= 1;
2290 parent_tail
= &commit
->parents
;
2292 if (!resolve_ref_unsafe("HEAD", RESOLVE_REF_READING
, head_sha1
, NULL
))
2293 die("no such ref: HEAD");
2295 parent_tail
= append_parent(parent_tail
, head_sha1
);
2296 append_merge_parents(parent_tail
);
2297 verify_working_tree_path(commit
, path
);
2299 origin
= make_origin(commit
, path
);
2301 ident
= fmt_ident("Not Committed Yet", "not.committed.yet", NULL
, 0);
2302 strbuf_addstr(&msg
, "tree 0000000000000000000000000000000000000000\n");
2303 for (parent
= commit
->parents
; parent
; parent
= parent
->next
)
2304 strbuf_addf(&msg
, "parent %s\n",
2305 sha1_to_hex(parent
->item
->object
.sha1
));
2309 "Version of %s from %s\n",
2311 (!contents_from
? path
:
2312 (!strcmp(contents_from
, "-") ? "standard input" : contents_from
)));
2313 set_commit_buffer_from_strbuf(commit
, &msg
);
2315 if (!contents_from
|| strcmp("-", contents_from
)) {
2317 const char *read_from
;
2319 unsigned long buf_len
;
2321 if (contents_from
) {
2322 if (stat(contents_from
, &st
) < 0)
2323 die_errno("Cannot stat '%s'", contents_from
);
2324 read_from
= contents_from
;
2327 if (lstat(path
, &st
) < 0)
2328 die_errno("Cannot lstat '%s'", path
);
2331 mode
= canon_mode(st
.st_mode
);
2333 switch (st
.st_mode
& S_IFMT
) {
2335 if (DIFF_OPT_TST(opt
, ALLOW_TEXTCONV
) &&
2336 textconv_object(read_from
, mode
, null_sha1
, 0, &buf_ptr
, &buf_len
))
2337 strbuf_attach(&buf
, buf_ptr
, buf_len
, buf_len
+ 1);
2338 else if (strbuf_read_file(&buf
, read_from
, st
.st_size
) != st
.st_size
)
2339 die_errno("cannot open or read '%s'", read_from
);
2342 if (strbuf_readlink(&buf
, read_from
, st
.st_size
) < 0)
2343 die_errno("cannot readlink '%s'", read_from
);
2346 die("unsupported file type %s", read_from
);
2350 /* Reading from stdin */
2352 if (strbuf_read(&buf
, 0, 0) < 0)
2353 die_errno("failed to read from stdin");
2355 convert_to_git(path
, buf
.buf
, buf
.len
, &buf
, 0);
2356 origin
->file
.ptr
= buf
.buf
;
2357 origin
->file
.size
= buf
.len
;
2358 pretend_sha1_file(buf
.buf
, buf
.len
, OBJ_BLOB
, origin
->blob_sha1
);
2361 * Read the current index, replace the path entry with
2362 * origin->blob_sha1 without mucking with its mode or type
2363 * bits; we are not going to write this index out -- we just
2364 * want to run "diff-index --cached".
2371 int pos
= cache_name_pos(path
, len
);
2373 mode
= active_cache
[pos
]->ce_mode
;
2375 /* Let's not bother reading from HEAD tree */
2376 mode
= S_IFREG
| 0644;
2378 size
= cache_entry_size(len
);
2379 ce
= xcalloc(1, size
);
2380 hashcpy(ce
->sha1
, origin
->blob_sha1
);
2381 memcpy(ce
->name
, path
, len
);
2382 ce
->ce_flags
= create_ce_flags(0);
2383 ce
->ce_namelen
= len
;
2384 ce
->ce_mode
= create_ce_mode(mode
);
2385 add_cache_entry(ce
, ADD_CACHE_OK_TO_ADD
|ADD_CACHE_OK_TO_REPLACE
);
2388 * We are not going to write this out, so this does not matter
2389 * right now, but someday we might optimize diff-index --cached
2390 * with cache-tree information.
2392 cache_tree_invalidate_path(&the_index
, path
);
2397 static char *prepare_final(struct scoreboard
*sb
)
2400 const char *final_commit_name
= NULL
;
2401 struct rev_info
*revs
= sb
->revs
;
2404 * There must be one and only one positive commit in the
2405 * revs->pending array.
2407 for (i
= 0; i
< revs
->pending
.nr
; i
++) {
2408 struct object
*obj
= revs
->pending
.objects
[i
].item
;
2409 if (obj
->flags
& UNINTERESTING
)
2411 while (obj
->type
== OBJ_TAG
)
2412 obj
= deref_tag(obj
, NULL
, 0);
2413 if (obj
->type
!= OBJ_COMMIT
)
2414 die("Non commit %s?", revs
->pending
.objects
[i
].name
);
2416 die("More than one commit to dig from %s and %s?",
2417 revs
->pending
.objects
[i
].name
,
2419 sb
->final
= (struct commit
*) obj
;
2420 final_commit_name
= revs
->pending
.objects
[i
].name
;
2422 return xstrdup_or_null(final_commit_name
);
2425 static char *prepare_initial(struct scoreboard
*sb
)
2428 const char *final_commit_name
= NULL
;
2429 struct rev_info
*revs
= sb
->revs
;
2432 * There must be one and only one negative commit, and it must be
2435 for (i
= 0; i
< revs
->pending
.nr
; i
++) {
2436 struct object
*obj
= revs
->pending
.objects
[i
].item
;
2437 if (!(obj
->flags
& UNINTERESTING
))
2439 while (obj
->type
== OBJ_TAG
)
2440 obj
= deref_tag(obj
, NULL
, 0);
2441 if (obj
->type
!= OBJ_COMMIT
)
2442 die("Non commit %s?", revs
->pending
.objects
[i
].name
);
2444 die("More than one commit to dig down to %s and %s?",
2445 revs
->pending
.objects
[i
].name
,
2447 sb
->final
= (struct commit
*) obj
;
2448 final_commit_name
= revs
->pending
.objects
[i
].name
;
2450 if (!final_commit_name
)
2451 die("No commit to dig down to?");
2452 return xstrdup(final_commit_name
);
2455 static int blame_copy_callback(const struct option
*option
, const char *arg
, int unset
)
2457 int *opt
= option
->value
;
2460 * -C enables copy from removed files;
2461 * -C -C enables copy from existing files, but only
2462 * when blaming a new file;
2463 * -C -C -C enables copy from existing files for
2466 if (*opt
& PICKAXE_BLAME_COPY_HARDER
)
2467 *opt
|= PICKAXE_BLAME_COPY_HARDEST
;
2468 if (*opt
& PICKAXE_BLAME_COPY
)
2469 *opt
|= PICKAXE_BLAME_COPY_HARDER
;
2470 *opt
|= PICKAXE_BLAME_COPY
| PICKAXE_BLAME_MOVE
;
2473 blame_copy_score
= parse_score(arg
);
2477 static int blame_move_callback(const struct option
*option
, const char *arg
, int unset
)
2479 int *opt
= option
->value
;
2481 *opt
|= PICKAXE_BLAME_MOVE
;
2484 blame_move_score
= parse_score(arg
);
2488 int cmd_blame(int argc
, const char **argv
, const char *prefix
)
2490 struct rev_info revs
;
2492 struct scoreboard sb
;
2494 struct blame_entry
*ent
= NULL
;
2495 long dashdash_pos
, lno
;
2496 char *final_commit_name
= NULL
;
2497 enum object_type type
;
2499 static struct string_list range_list
;
2500 static int output_option
= 0, opt
= 0;
2501 static int show_stats
= 0;
2502 static const char *revs_file
= NULL
;
2503 static const char *contents_from
= NULL
;
2504 static const struct option options
[] = {
2505 OPT_BOOL(0, "incremental", &incremental
, N_("Show blame entries as we find them, incrementally")),
2506 OPT_BOOL('b', NULL
, &blank_boundary
, N_("Show blank SHA-1 for boundary commits (Default: off)")),
2507 OPT_BOOL(0, "root", &show_root
, N_("Do not treat root commits as boundaries (Default: off)")),
2508 OPT_BOOL(0, "show-stats", &show_stats
, N_("Show work cost statistics")),
2509 OPT_BIT(0, "score-debug", &output_option
, N_("Show output score for blame entries"), OUTPUT_SHOW_SCORE
),
2510 OPT_BIT('f', "show-name", &output_option
, N_("Show original filename (Default: auto)"), OUTPUT_SHOW_NAME
),
2511 OPT_BIT('n', "show-number", &output_option
, N_("Show original linenumber (Default: off)"), OUTPUT_SHOW_NUMBER
),
2512 OPT_BIT('p', "porcelain", &output_option
, N_("Show in a format designed for machine consumption"), OUTPUT_PORCELAIN
),
2513 OPT_BIT(0, "line-porcelain", &output_option
, N_("Show porcelain format with per-line commit information"), OUTPUT_PORCELAIN
|OUTPUT_LINE_PORCELAIN
),
2514 OPT_BIT('c', NULL
, &output_option
, N_("Use the same output mode as git-annotate (Default: off)"), OUTPUT_ANNOTATE_COMPAT
),
2515 OPT_BIT('t', NULL
, &output_option
, N_("Show raw timestamp (Default: off)"), OUTPUT_RAW_TIMESTAMP
),
2516 OPT_BIT('l', NULL
, &output_option
, N_("Show long commit SHA1 (Default: off)"), OUTPUT_LONG_OBJECT_NAME
),
2517 OPT_BIT('s', NULL
, &output_option
, N_("Suppress author name and timestamp (Default: off)"), OUTPUT_NO_AUTHOR
),
2518 OPT_BIT('e', "show-email", &output_option
, N_("Show author email instead of name (Default: off)"), OUTPUT_SHOW_EMAIL
),
2519 OPT_BIT('w', NULL
, &xdl_opts
, N_("Ignore whitespace differences"), XDF_IGNORE_WHITESPACE
),
2520 OPT_BIT(0, "minimal", &xdl_opts
, N_("Spend extra cycles to find better match"), XDF_NEED_MINIMAL
),
2521 OPT_STRING('S', NULL
, &revs_file
, N_("file"), N_("Use revisions from <file> instead of calling git-rev-list")),
2522 OPT_STRING(0, "contents", &contents_from
, N_("file"), N_("Use <file>'s contents as the final image")),
2523 { OPTION_CALLBACK
, 'C', NULL
, &opt
, N_("score"), N_("Find line copies within and across files"), PARSE_OPT_OPTARG
, blame_copy_callback
},
2524 { OPTION_CALLBACK
, 'M', NULL
, &opt
, N_("score"), N_("Find line movements within and across files"), PARSE_OPT_OPTARG
, blame_move_callback
},
2525 OPT_STRING_LIST('L', NULL
, &range_list
, N_("n,m"), N_("Process only line range n,m, counting from 1")),
2526 OPT__ABBREV(&abbrev
),
2530 struct parse_opt_ctx_t ctx
;
2531 int cmd_is_annotate
= !strcmp(argv
[0], "annotate");
2532 struct range_set ranges
;
2533 unsigned int range_i
;
2536 git_config(git_blame_config
, &output_option
);
2537 init_revisions(&revs
, NULL
);
2538 revs
.date_mode
= blame_date_mode
;
2539 DIFF_OPT_SET(&revs
.diffopt
, ALLOW_TEXTCONV
);
2540 DIFF_OPT_SET(&revs
.diffopt
, FOLLOW_RENAMES
);
2542 save_commit_buffer
= 0;
2545 parse_options_start(&ctx
, argc
, argv
, prefix
, options
,
2546 PARSE_OPT_KEEP_DASHDASH
| PARSE_OPT_KEEP_ARGV0
);
2548 switch (parse_options_step(&ctx
, options
, blame_opt_usage
)) {
2549 case PARSE_OPT_HELP
:
2551 case PARSE_OPT_DONE
:
2553 dashdash_pos
= ctx
.cpidx
;
2557 if (!strcmp(ctx
.argv
[0], "--reverse")) {
2558 ctx
.argv
[0] = "--children";
2561 parse_revision_opt(&revs
, &ctx
, options
, blame_opt_usage
);
2564 no_whole_file_rename
= !DIFF_OPT_TST(&revs
.diffopt
, FOLLOW_RENAMES
);
2565 DIFF_OPT_CLR(&revs
.diffopt
, FOLLOW_RENAMES
);
2566 argc
= parse_options_end(&ctx
);
2569 /* one more abbrev length is needed for the boundary commit */
2572 if (revs_file
&& read_ancestry(revs_file
))
2573 die_errno("reading graft file '%s' failed", revs_file
);
2575 if (cmd_is_annotate
) {
2576 output_option
|= OUTPUT_ANNOTATE_COMPAT
;
2577 blame_date_mode
.type
= DATE_ISO8601
;
2579 blame_date_mode
= revs
.date_mode
;
2582 /* The maximum width used to show the dates */
2583 switch (blame_date_mode
.type
) {
2585 blame_date_width
= sizeof("Thu, 19 Oct 2006 16:00:04 -0700");
2587 case DATE_ISO8601_STRICT
:
2588 blame_date_width
= sizeof("2006-10-19T16:00:04-07:00");
2591 blame_date_width
= sizeof("2006-10-19 16:00:04 -0700");
2594 blame_date_width
= sizeof("1161298804 -0700");
2597 blame_date_width
= sizeof("2006-10-19");
2600 /* TRANSLATORS: This string is used to tell us the maximum
2601 display width for a relative timestamp in "git blame"
2602 output. For C locale, "4 years, 11 months ago", which
2603 takes 22 places, is the longest among various forms of
2604 relative timestamps, but your language may need more or
2605 fewer display columns. */
2606 blame_date_width
= utf8_strwidth(_("4 years, 11 months ago")) + 1; /* add the null */
2610 blame_date_width
= sizeof("Thu Oct 19 16:00:04 2006 -0700");
2613 blame_date_width
= strlen(show_date(0, 0, &blame_date_mode
)) + 1; /* add the null */
2616 blame_date_width
-= 1; /* strip the null */
2618 if (DIFF_OPT_TST(&revs
.diffopt
, FIND_COPIES_HARDER
))
2619 opt
|= (PICKAXE_BLAME_COPY
| PICKAXE_BLAME_MOVE
|
2620 PICKAXE_BLAME_COPY_HARDER
);
2622 if (!blame_move_score
)
2623 blame_move_score
= BLAME_DEFAULT_MOVE_SCORE
;
2624 if (!blame_copy_score
)
2625 blame_copy_score
= BLAME_DEFAULT_COPY_SCORE
;
2628 * We have collected options unknown to us in argv[1..unk]
2629 * which are to be passed to revision machinery if we are
2630 * going to do the "bottom" processing.
2632 * The remaining are:
2634 * (1) if dashdash_pos != 0, it is either
2635 * "blame [revisions] -- <path>" or
2636 * "blame -- <path> <rev>"
2638 * (2) otherwise, it is one of the two:
2639 * "blame [revisions] <path>"
2640 * "blame <path> <rev>"
2642 * Note that we must strip out <path> from the arguments: we do not
2643 * want the path pruning but we may want "bottom" processing.
2646 switch (argc
- dashdash_pos
- 1) {
2649 usage_with_options(blame_opt_usage
, options
);
2650 /* reorder for the new way: <rev> -- <path> */
2656 path
= add_prefix(prefix
, argv
[--argc
]);
2660 usage_with_options(blame_opt_usage
, options
);
2664 usage_with_options(blame_opt_usage
, options
);
2665 path
= add_prefix(prefix
, argv
[argc
- 1]);
2666 if (argc
== 3 && !file_exists(path
)) { /* (2b) */
2667 path
= add_prefix(prefix
, argv
[1]);
2670 argv
[argc
- 1] = "--";
2673 if (!file_exists(path
))
2674 die_errno("cannot stat path '%s'", path
);
2677 revs
.disable_stdin
= 1;
2678 setup_revisions(argc
, argv
, &revs
, NULL
);
2679 memset(&sb
, 0, sizeof(sb
));
2683 final_commit_name
= prepare_final(&sb
);
2684 sb
.commits
.compare
= compare_commits_by_commit_date
;
2686 else if (contents_from
)
2687 die("--contents and --children do not blend well.");
2689 final_commit_name
= prepare_initial(&sb
);
2690 sb
.commits
.compare
= compare_commits_by_reverse_commit_date
;
2695 * "--not A B -- path" without anything positive;
2696 * do not default to HEAD, but use the working tree
2700 sb
.final
= fake_working_tree_commit(&sb
.revs
->diffopt
,
2701 path
, contents_from
);
2702 add_pending_object(&revs
, &(sb
.final
->object
), ":");
2704 else if (contents_from
)
2705 die("Cannot use --contents with final commit object name");
2708 * If we have bottom, this will mark the ancestors of the
2709 * bottom commits we would reach while traversing as
2712 if (prepare_revision_walk(&revs
))
2713 die(_("revision walk setup failed"));
2715 if (is_null_sha1(sb
.final
->object
.sha1
)) {
2717 sb
.final_buf
= xmemdupz(o
->file
.ptr
, o
->file
.size
);
2718 sb
.final_buf_size
= o
->file
.size
;
2721 o
= get_origin(&sb
, sb
.final
, path
);
2722 if (fill_blob_sha1_and_mode(o
))
2723 die("no such path %s in %s", path
, final_commit_name
);
2725 if (DIFF_OPT_TST(&sb
.revs
->diffopt
, ALLOW_TEXTCONV
) &&
2726 textconv_object(path
, o
->mode
, o
->blob_sha1
, 1, (char **) &sb
.final_buf
,
2727 &sb
.final_buf_size
))
2730 sb
.final_buf
= read_sha1_file(o
->blob_sha1
, &type
,
2731 &sb
.final_buf_size
);
2734 die("Cannot read blob %s for path %s",
2735 sha1_to_hex(o
->blob_sha1
),
2739 lno
= prepare_lines(&sb
);
2741 if (lno
&& !range_list
.nr
)
2742 string_list_append(&range_list
, xstrdup("1"));
2745 range_set_init(&ranges
, range_list
.nr
);
2746 for (range_i
= 0; range_i
< range_list
.nr
; ++range_i
) {
2748 if (parse_range_arg(range_list
.items
[range_i
].string
,
2749 nth_line_cb
, &sb
, lno
, anchor
,
2750 &bottom
, &top
, sb
.path
))
2752 if (lno
< top
|| ((lno
|| bottom
) && lno
< bottom
))
2753 die("file %s has only %lu lines", path
, lno
);
2759 range_set_append_unsafe(&ranges
, bottom
, top
);
2762 sort_and_merge_range_set(&ranges
);
2764 for (range_i
= ranges
.nr
; range_i
> 0; --range_i
) {
2765 const struct range
*r
= &ranges
.ranges
[range_i
- 1];
2766 long bottom
= r
->start
;
2768 struct blame_entry
*next
= ent
;
2769 ent
= xcalloc(1, sizeof(*ent
));
2771 ent
->num_lines
= top
- bottom
;
2773 ent
->s_lno
= bottom
;
2779 prio_queue_put(&sb
.commits
, o
->commit
);
2783 range_set_release(&ranges
);
2784 string_list_clear(&range_list
, 0);
2789 read_mailmap(&mailmap
, NULL
);
2794 assign_blame(&sb
, opt
);
2796 free(final_commit_name
);
2801 sb
.ent
= blame_sort(sb
.ent
, compare_blame_final
);
2805 if (!(output_option
& OUTPUT_PORCELAIN
))
2806 find_alignment(&sb
, &output_option
);
2808 output(&sb
, output_option
);
2809 free((void *)sb
.final_buf
);
2810 for (ent
= sb
.ent
; ent
; ) {
2811 struct blame_entry
*e
= ent
->next
;
2817 printf("num read blob: %d\n", num_read_blob
);
2818 printf("num get patch: %d\n", num_get_patch
);
2819 printf("num commits: %d\n", num_commits
);