4 * Copyright (c) 2006, Junio C Hamano
12 #include "tree-walk.h"
17 #include "xdiff-interface.h"
18 #include "cache-tree.h"
19 #include "string-list.h"
21 #include "parse-options.h"
24 #include "line-range.h"
26 static char blame_usage
[] = N_("git blame [options] [rev-opts] [rev] [--] file");
28 static const char *blame_opt_usage
[] = {
31 N_("[rev-opts] are documented in git-rev-list(1)"),
35 static int longest_file
;
36 static int longest_author
;
37 static int max_orig_digits
;
38 static int max_digits
;
39 static int max_score_digits
;
42 static int blank_boundary
;
43 static int incremental
;
45 static int abbrev
= -1;
46 static int no_whole_file_rename
;
48 static enum date_mode blame_date_mode
= DATE_ISO8601
;
49 static size_t blame_date_width
;
51 static struct string_list mailmap
;
58 static int num_read_blob
;
59 static int num_get_patch
;
60 static int num_commits
;
62 #define PICKAXE_BLAME_MOVE 01
63 #define PICKAXE_BLAME_COPY 02
64 #define PICKAXE_BLAME_COPY_HARDER 04
65 #define PICKAXE_BLAME_COPY_HARDEST 010
68 * blame for a blame_entry with score lower than these thresholds
69 * is not passed to the parent using move/copy logic.
71 static unsigned blame_move_score
;
72 static unsigned blame_copy_score
;
73 #define BLAME_DEFAULT_MOVE_SCORE 20
74 #define BLAME_DEFAULT_COPY_SCORE 40
76 /* bits #0..7 in revision.h, #8..11 used for merge_bases() in commit.c */
77 #define METAINFO_SHOWN (1u<<12)
78 #define MORE_THAN_ONE_PATH (1u<<13)
81 * One blob in a commit that is being suspected
85 struct origin
*previous
;
86 struct commit
*commit
;
88 unsigned char blob_sha1
[20];
90 char path
[FLEX_ARRAY
];
93 static int diff_hunks(mmfile_t
*file_a
, mmfile_t
*file_b
, long ctxlen
,
94 xdl_emit_hunk_consume_func_t hunk_func
, void *cb_data
)
97 xdemitconf_t xecfg
= {0};
98 xdemitcb_t ecb
= {NULL
};
100 xpp
.flags
= xdl_opts
;
101 xecfg
.ctxlen
= ctxlen
;
102 xecfg
.hunk_func
= hunk_func
;
104 return xdi_diff(file_a
, file_b
, &xpp
, &xecfg
, &ecb
);
108 * Prepare diff_filespec and convert it using diff textconv API
109 * if the textconv driver exists.
110 * Return 1 if the conversion succeeds, 0 otherwise.
112 int textconv_object(const char *path
,
114 const unsigned char *sha1
,
117 unsigned long *buf_size
)
119 struct diff_filespec
*df
;
120 struct userdiff_driver
*textconv
;
122 df
= alloc_filespec(path
);
123 fill_filespec(df
, sha1
, sha1_valid
, mode
);
124 textconv
= get_textconv(df
);
130 *buf_size
= fill_textconv(textconv
, df
, buf
);
136 * Given an origin, prepare mmfile_t structure to be used by the
139 static void fill_origin_blob(struct diff_options
*opt
,
140 struct origin
*o
, mmfile_t
*file
)
143 enum object_type type
;
144 unsigned long file_size
;
147 if (DIFF_OPT_TST(opt
, ALLOW_TEXTCONV
) &&
148 textconv_object(o
->path
, o
->mode
, o
->blob_sha1
, 1, &file
->ptr
, &file_size
))
151 file
->ptr
= read_sha1_file(o
->blob_sha1
, &type
, &file_size
);
152 file
->size
= file_size
;
155 die("Cannot read blob %s for path %s",
156 sha1_to_hex(o
->blob_sha1
),
165 * Origin is refcounted and usually we keep the blob contents to be
168 static inline struct origin
*origin_incref(struct origin
*o
)
175 static void origin_decref(struct origin
*o
)
177 if (o
&& --o
->refcnt
<= 0) {
179 origin_decref(o
->previous
);
185 static void drop_origin_blob(struct origin
*o
)
194 * Each group of lines is described by a blame_entry; it can be split
195 * as we pass blame to the parents. They form a linked list in the
196 * scoreboard structure, sorted by the target line number.
199 struct blame_entry
*prev
;
200 struct blame_entry
*next
;
202 /* the first line of this group in the final image;
203 * internally all line numbers are 0 based.
207 /* how many lines this group has */
210 /* the commit that introduced this group into the final image */
211 struct origin
*suspect
;
213 /* true if the suspect is truly guilty; false while we have not
214 * checked if the group came from one of its parents.
218 /* true if the entry has been scanned for copies in the current parent
222 /* the line number of the first line of this group in the
223 * suspect's file; internally all line numbers are 0 based.
227 /* how significant this entry is -- cached to avoid
228 * scanning the lines over and over.
234 * The current state of the blame assignment.
237 /* the final commit (i.e. where we started digging from) */
238 struct commit
*final
;
239 struct rev_info
*revs
;
243 * The contents in the final image.
244 * Used by many functions to obtain contents of the nth line,
245 * indexed with scoreboard.lineno[blame_entry.lno].
247 const char *final_buf
;
248 unsigned long final_buf_size
;
250 /* linked list of blames */
251 struct blame_entry
*ent
;
253 /* look-up a line in the final buffer */
258 static inline int same_suspect(struct origin
*a
, struct origin
*b
)
262 if (a
->commit
!= b
->commit
)
264 return !strcmp(a
->path
, b
->path
);
267 static void sanity_check_refcnt(struct scoreboard
*);
270 * If two blame entries that are next to each other came from
271 * contiguous lines in the same origin (i.e. <commit, path> pair),
272 * merge them together.
274 static void coalesce(struct scoreboard
*sb
)
276 struct blame_entry
*ent
, *next
;
278 for (ent
= sb
->ent
; ent
&& (next
= ent
->next
); ent
= next
) {
279 if (same_suspect(ent
->suspect
, next
->suspect
) &&
280 ent
->guilty
== next
->guilty
&&
281 ent
->s_lno
+ ent
->num_lines
== next
->s_lno
) {
282 ent
->num_lines
+= next
->num_lines
;
283 ent
->next
= next
->next
;
285 ent
->next
->prev
= ent
;
286 origin_decref(next
->suspect
);
289 next
= ent
; /* again */
293 if (DEBUG
) /* sanity */
294 sanity_check_refcnt(sb
);
298 * Given a commit and a path in it, create a new origin structure.
299 * The callers that add blame to the scoreboard should use
300 * get_origin() to obtain shared, refcounted copy instead of calling
301 * this function directly.
303 static struct origin
*make_origin(struct commit
*commit
, const char *path
)
306 o
= xcalloc(1, sizeof(*o
) + strlen(path
) + 1);
309 strcpy(o
->path
, path
);
314 * Locate an existing origin or create a new one.
316 static struct origin
*get_origin(struct scoreboard
*sb
,
317 struct commit
*commit
,
320 struct blame_entry
*e
;
322 for (e
= sb
->ent
; e
; e
= e
->next
) {
323 if (e
->suspect
->commit
== commit
&&
324 !strcmp(e
->suspect
->path
, path
))
325 return origin_incref(e
->suspect
);
327 return make_origin(commit
, path
);
331 * Fill the blob_sha1 field of an origin if it hasn't, so that later
332 * call to fill_origin_blob() can use it to locate the data. blob_sha1
333 * for an origin is also used to pass the blame for the entire file to
334 * the parent to detect the case where a child's blob is identical to
335 * that of its parent's.
337 * This also fills origin->mode for corresponding tree path.
339 static int fill_blob_sha1_and_mode(struct origin
*origin
)
341 if (!is_null_sha1(origin
->blob_sha1
))
343 if (get_tree_entry(origin
->commit
->object
.sha1
,
345 origin
->blob_sha1
, &origin
->mode
))
347 if (sha1_object_info(origin
->blob_sha1
, NULL
) != OBJ_BLOB
)
351 hashclr(origin
->blob_sha1
);
352 origin
->mode
= S_IFINVALID
;
357 * We have an origin -- check if the same path exists in the
358 * parent and return an origin structure to represent it.
360 static struct origin
*find_origin(struct scoreboard
*sb
,
361 struct commit
*parent
,
362 struct origin
*origin
)
364 struct origin
*porigin
= NULL
;
365 struct diff_options diff_opts
;
366 const char *paths
[2];
370 * Each commit object can cache one origin in that
371 * commit. This is a freestanding copy of origin and
374 struct origin
*cached
= parent
->util
;
375 if (!strcmp(cached
->path
, origin
->path
)) {
377 * The same path between origin and its parent
378 * without renaming -- the most common case.
380 porigin
= get_origin(sb
, parent
, cached
->path
);
383 * If the origin was newly created (i.e. get_origin
384 * would call make_origin if none is found in the
385 * scoreboard), it does not know the blob_sha1/mode,
386 * so copy it. Otherwise porigin was in the
387 * scoreboard and already knows blob_sha1/mode.
389 if (porigin
->refcnt
== 1) {
390 hashcpy(porigin
->blob_sha1
, cached
->blob_sha1
);
391 porigin
->mode
= cached
->mode
;
395 /* otherwise it was not very useful; free it */
400 /* See if the origin->path is different between parent
401 * and origin first. Most of the time they are the
402 * same and diff-tree is fairly efficient about this.
404 diff_setup(&diff_opts
);
405 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
406 diff_opts
.detect_rename
= 0;
407 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
408 paths
[0] = origin
->path
;
411 diff_tree_setup_paths(paths
, &diff_opts
);
412 diff_setup_done(&diff_opts
);
414 if (is_null_sha1(origin
->commit
->object
.sha1
))
415 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
417 diff_tree_sha1(parent
->tree
->object
.sha1
,
418 origin
->commit
->tree
->object
.sha1
,
420 diffcore_std(&diff_opts
);
422 if (!diff_queued_diff
.nr
) {
423 /* The path is the same as parent */
424 porigin
= get_origin(sb
, parent
, origin
->path
);
425 hashcpy(porigin
->blob_sha1
, origin
->blob_sha1
);
426 porigin
->mode
= origin
->mode
;
429 * Since origin->path is a pathspec, if the parent
430 * commit had it as a directory, we will see a whole
431 * bunch of deletion of files in the directory that we
435 struct diff_filepair
*p
= NULL
;
436 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
438 p
= diff_queued_diff
.queue
[i
];
439 name
= p
->one
->path
? p
->one
->path
: p
->two
->path
;
440 if (!strcmp(name
, origin
->path
))
444 die("internal error in blame::find_origin");
447 die("internal error in blame::find_origin (%c)",
450 porigin
= get_origin(sb
, parent
, origin
->path
);
451 hashcpy(porigin
->blob_sha1
, p
->one
->sha1
);
452 porigin
->mode
= p
->one
->mode
;
456 /* Did not exist in parent, or type changed */
460 diff_flush(&diff_opts
);
461 diff_tree_release_paths(&diff_opts
);
464 * Create a freestanding copy that is not part of
465 * the refcounted origin found in the scoreboard, and
466 * cache it in the commit.
468 struct origin
*cached
;
470 cached
= make_origin(porigin
->commit
, porigin
->path
);
471 hashcpy(cached
->blob_sha1
, porigin
->blob_sha1
);
472 cached
->mode
= porigin
->mode
;
473 parent
->util
= cached
;
479 * We have an origin -- find the path that corresponds to it in its
480 * parent and return an origin structure to represent it.
482 static struct origin
*find_rename(struct scoreboard
*sb
,
483 struct commit
*parent
,
484 struct origin
*origin
)
486 struct origin
*porigin
= NULL
;
487 struct diff_options diff_opts
;
489 const char *paths
[2];
491 diff_setup(&diff_opts
);
492 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
493 diff_opts
.detect_rename
= DIFF_DETECT_RENAME
;
494 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
495 diff_opts
.single_follow
= origin
->path
;
497 diff_tree_setup_paths(paths
, &diff_opts
);
498 diff_setup_done(&diff_opts
);
500 if (is_null_sha1(origin
->commit
->object
.sha1
))
501 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
503 diff_tree_sha1(parent
->tree
->object
.sha1
,
504 origin
->commit
->tree
->object
.sha1
,
506 diffcore_std(&diff_opts
);
508 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
509 struct diff_filepair
*p
= diff_queued_diff
.queue
[i
];
510 if ((p
->status
== 'R' || p
->status
== 'C') &&
511 !strcmp(p
->two
->path
, origin
->path
)) {
512 porigin
= get_origin(sb
, parent
, p
->one
->path
);
513 hashcpy(porigin
->blob_sha1
, p
->one
->sha1
);
514 porigin
->mode
= p
->one
->mode
;
518 diff_flush(&diff_opts
);
519 diff_tree_release_paths(&diff_opts
);
524 * Link in a new blame entry to the scoreboard. Entries that cover the
525 * same line range have been removed from the scoreboard previously.
527 static void add_blame_entry(struct scoreboard
*sb
, struct blame_entry
*e
)
529 struct blame_entry
*ent
, *prev
= NULL
;
531 origin_incref(e
->suspect
);
533 for (ent
= sb
->ent
; ent
&& ent
->lno
< e
->lno
; ent
= ent
->next
)
536 /* prev, if not NULL, is the last one that is below e */
539 e
->next
= prev
->next
;
551 * src typically is on-stack; we want to copy the information in it to
552 * a malloced blame_entry that is already on the linked list of the
553 * scoreboard. The origin of dst loses a refcnt while the origin of src
556 static void dup_entry(struct blame_entry
*dst
, struct blame_entry
*src
)
558 struct blame_entry
*p
, *n
;
562 origin_incref(src
->suspect
);
563 origin_decref(dst
->suspect
);
564 memcpy(dst
, src
, sizeof(*src
));
570 static const char *nth_line(struct scoreboard
*sb
, long lno
)
572 return sb
->final_buf
+ sb
->lineno
[lno
];
575 static const char *nth_line_cb(void *data
, long lno
)
577 return nth_line((struct scoreboard
*)data
, lno
);
581 * It is known that lines between tlno to same came from parent, and e
582 * has an overlap with that range. it also is known that parent's
583 * line plno corresponds to e's line tlno.
589 * <------------------>
591 * Split e into potentially three parts; before this chunk, the chunk
592 * to be blamed for the parent, and after that portion.
594 static void split_overlap(struct blame_entry
*split
,
595 struct blame_entry
*e
,
596 int tlno
, int plno
, int same
,
597 struct origin
*parent
)
600 memset(split
, 0, sizeof(struct blame_entry
[3]));
602 if (e
->s_lno
< tlno
) {
603 /* there is a pre-chunk part not blamed on parent */
604 split
[0].suspect
= origin_incref(e
->suspect
);
605 split
[0].lno
= e
->lno
;
606 split
[0].s_lno
= e
->s_lno
;
607 split
[0].num_lines
= tlno
- e
->s_lno
;
608 split
[1].lno
= e
->lno
+ tlno
- e
->s_lno
;
609 split
[1].s_lno
= plno
;
612 split
[1].lno
= e
->lno
;
613 split
[1].s_lno
= plno
+ (e
->s_lno
- tlno
);
616 if (same
< e
->s_lno
+ e
->num_lines
) {
617 /* there is a post-chunk part not blamed on parent */
618 split
[2].suspect
= origin_incref(e
->suspect
);
619 split
[2].lno
= e
->lno
+ (same
- e
->s_lno
);
620 split
[2].s_lno
= e
->s_lno
+ (same
- e
->s_lno
);
621 split
[2].num_lines
= e
->s_lno
+ e
->num_lines
- same
;
622 chunk_end_lno
= split
[2].lno
;
625 chunk_end_lno
= e
->lno
+ e
->num_lines
;
626 split
[1].num_lines
= chunk_end_lno
- split
[1].lno
;
629 * if it turns out there is nothing to blame the parent for,
630 * forget about the splitting. !split[1].suspect signals this.
632 if (split
[1].num_lines
< 1)
634 split
[1].suspect
= origin_incref(parent
);
638 * split_overlap() divided an existing blame e into up to three parts
639 * in split. Adjust the linked list of blames in the scoreboard to
642 static void split_blame(struct scoreboard
*sb
,
643 struct blame_entry
*split
,
644 struct blame_entry
*e
)
646 struct blame_entry
*new_entry
;
648 if (split
[0].suspect
&& split
[2].suspect
) {
649 /* The first part (reuse storage for the existing entry e) */
650 dup_entry(e
, &split
[0]);
652 /* The last part -- me */
653 new_entry
= xmalloc(sizeof(*new_entry
));
654 memcpy(new_entry
, &(split
[2]), sizeof(struct blame_entry
));
655 add_blame_entry(sb
, new_entry
);
657 /* ... and the middle part -- parent */
658 new_entry
= xmalloc(sizeof(*new_entry
));
659 memcpy(new_entry
, &(split
[1]), sizeof(struct blame_entry
));
660 add_blame_entry(sb
, new_entry
);
662 else if (!split
[0].suspect
&& !split
[2].suspect
)
664 * The parent covers the entire area; reuse storage for
665 * e and replace it with the parent.
667 dup_entry(e
, &split
[1]);
668 else if (split
[0].suspect
) {
669 /* me and then parent */
670 dup_entry(e
, &split
[0]);
672 new_entry
= xmalloc(sizeof(*new_entry
));
673 memcpy(new_entry
, &(split
[1]), sizeof(struct blame_entry
));
674 add_blame_entry(sb
, new_entry
);
677 /* parent and then me */
678 dup_entry(e
, &split
[1]);
680 new_entry
= xmalloc(sizeof(*new_entry
));
681 memcpy(new_entry
, &(split
[2]), sizeof(struct blame_entry
));
682 add_blame_entry(sb
, new_entry
);
685 if (DEBUG
) { /* sanity */
686 struct blame_entry
*ent
;
687 int lno
= sb
->ent
->lno
, corrupt
= 0;
689 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
694 lno
+= ent
->num_lines
;
698 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
699 printf("L %8d l %8d n %8d\n",
700 lno
, ent
->lno
, ent
->num_lines
);
701 lno
= ent
->lno
+ ent
->num_lines
;
709 * After splitting the blame, the origins used by the
710 * on-stack blame_entry should lose one refcnt each.
712 static void decref_split(struct blame_entry
*split
)
716 for (i
= 0; i
< 3; i
++)
717 origin_decref(split
[i
].suspect
);
721 * Helper for blame_chunk(). blame_entry e is known to overlap with
722 * the patch hunk; split it and pass blame to the parent.
724 static void blame_overlap(struct scoreboard
*sb
, struct blame_entry
*e
,
725 int tlno
, int plno
, int same
,
726 struct origin
*parent
)
728 struct blame_entry split
[3];
730 split_overlap(split
, e
, tlno
, plno
, same
, parent
);
731 if (split
[1].suspect
)
732 split_blame(sb
, split
, e
);
737 * Find the line number of the last line the target is suspected for.
739 static int find_last_in_target(struct scoreboard
*sb
, struct origin
*target
)
741 struct blame_entry
*e
;
742 int last_in_target
= -1;
744 for (e
= sb
->ent
; e
; e
= e
->next
) {
745 if (e
->guilty
|| !same_suspect(e
->suspect
, target
))
747 if (last_in_target
< e
->s_lno
+ e
->num_lines
)
748 last_in_target
= e
->s_lno
+ e
->num_lines
;
750 return last_in_target
;
754 * Process one hunk from the patch between the current suspect for
755 * blame_entry e and its parent. Find and split the overlap, and
756 * pass blame to the overlapping part to the parent.
758 static void blame_chunk(struct scoreboard
*sb
,
759 int tlno
, int plno
, int same
,
760 struct origin
*target
, struct origin
*parent
)
762 struct blame_entry
*e
;
764 for (e
= sb
->ent
; e
; e
= e
->next
) {
765 if (e
->guilty
|| !same_suspect(e
->suspect
, target
))
767 if (same
<= e
->s_lno
)
769 if (tlno
< e
->s_lno
+ e
->num_lines
)
770 blame_overlap(sb
, e
, tlno
, plno
, same
, parent
);
774 struct blame_chunk_cb_data
{
775 struct scoreboard
*sb
;
776 struct origin
*target
;
777 struct origin
*parent
;
782 static int blame_chunk_cb(long start_a
, long count_a
,
783 long start_b
, long count_b
, void *data
)
785 struct blame_chunk_cb_data
*d
= data
;
786 blame_chunk(d
->sb
, d
->tlno
, d
->plno
, start_b
, d
->target
, d
->parent
);
787 d
->plno
= start_a
+ count_a
;
788 d
->tlno
= start_b
+ count_b
;
793 * We are looking at the origin 'target' and aiming to pass blame
794 * for the lines it is suspected to its parent. Run diff to find
795 * which lines came from parent and pass blame for them.
797 static int pass_blame_to_parent(struct scoreboard
*sb
,
798 struct origin
*target
,
799 struct origin
*parent
)
802 mmfile_t file_p
, file_o
;
803 struct blame_chunk_cb_data d
;
805 memset(&d
, 0, sizeof(d
));
806 d
.sb
= sb
; d
.target
= target
; d
.parent
= parent
;
807 last_in_target
= find_last_in_target(sb
, target
);
808 if (last_in_target
< 0)
809 return 1; /* nothing remains for this target */
811 fill_origin_blob(&sb
->revs
->diffopt
, parent
, &file_p
);
812 fill_origin_blob(&sb
->revs
->diffopt
, target
, &file_o
);
815 diff_hunks(&file_p
, &file_o
, 0, blame_chunk_cb
, &d
);
816 /* The rest (i.e. anything after tlno) are the same as the parent */
817 blame_chunk(sb
, d
.tlno
, d
.plno
, last_in_target
, target
, parent
);
823 * The lines in blame_entry after splitting blames many times can become
824 * very small and trivial, and at some point it becomes pointless to
825 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
826 * ordinary C program, and it is not worth to say it was copied from
827 * totally unrelated file in the parent.
829 * Compute how trivial the lines in the blame_entry are.
831 static unsigned ent_score(struct scoreboard
*sb
, struct blame_entry
*e
)
840 cp
= nth_line(sb
, e
->lno
);
841 ep
= nth_line(sb
, e
->lno
+ e
->num_lines
);
843 unsigned ch
= *((unsigned char *)cp
);
853 * best_so_far[] and this[] are both a split of an existing blame_entry
854 * that passes blame to the parent. Maintain best_so_far the best split
855 * so far, by comparing this and best_so_far and copying this into
856 * bst_so_far as needed.
858 static void copy_split_if_better(struct scoreboard
*sb
,
859 struct blame_entry
*best_so_far
,
860 struct blame_entry
*this)
864 if (!this[1].suspect
)
866 if (best_so_far
[1].suspect
) {
867 if (ent_score(sb
, &this[1]) < ent_score(sb
, &best_so_far
[1]))
871 for (i
= 0; i
< 3; i
++)
872 origin_incref(this[i
].suspect
);
873 decref_split(best_so_far
);
874 memcpy(best_so_far
, this, sizeof(struct blame_entry
[3]));
878 * We are looking at a part of the final image represented by
879 * ent (tlno and same are offset by ent->s_lno).
880 * tlno is where we are looking at in the final image.
881 * up to (but not including) same match preimage.
882 * plno is where we are looking at in the preimage.
884 * <-------------- final image ---------------------->
887 * <---------preimage----->
890 * All line numbers are 0-based.
892 static void handle_split(struct scoreboard
*sb
,
893 struct blame_entry
*ent
,
894 int tlno
, int plno
, int same
,
895 struct origin
*parent
,
896 struct blame_entry
*split
)
898 if (ent
->num_lines
<= tlno
)
901 struct blame_entry
this[3];
904 split_overlap(this, ent
, tlno
, plno
, same
, parent
);
905 copy_split_if_better(sb
, split
, this);
910 struct handle_split_cb_data
{
911 struct scoreboard
*sb
;
912 struct blame_entry
*ent
;
913 struct origin
*parent
;
914 struct blame_entry
*split
;
919 static int handle_split_cb(long start_a
, long count_a
,
920 long start_b
, long count_b
, void *data
)
922 struct handle_split_cb_data
*d
= data
;
923 handle_split(d
->sb
, d
->ent
, d
->tlno
, d
->plno
, start_b
, d
->parent
,
925 d
->plno
= start_a
+ count_a
;
926 d
->tlno
= start_b
+ count_b
;
931 * Find the lines from parent that are the same as ent so that
932 * we can pass blames to it. file_p has the blob contents for
935 static void find_copy_in_blob(struct scoreboard
*sb
,
936 struct blame_entry
*ent
,
937 struct origin
*parent
,
938 struct blame_entry
*split
,
944 struct handle_split_cb_data d
;
946 memset(&d
, 0, sizeof(d
));
947 d
.sb
= sb
; d
.ent
= ent
; d
.parent
= parent
; d
.split
= split
;
949 * Prepare mmfile that contains only the lines in ent.
951 cp
= nth_line(sb
, ent
->lno
);
952 file_o
.ptr
= (char *) cp
;
953 cnt
= ent
->num_lines
;
955 while (cnt
&& cp
< sb
->final_buf
+ sb
->final_buf_size
) {
959 file_o
.size
= cp
- file_o
.ptr
;
962 * file_o is a part of final image we are annotating.
963 * file_p partially may match that image.
965 memset(split
, 0, sizeof(struct blame_entry
[3]));
966 diff_hunks(file_p
, &file_o
, 1, handle_split_cb
, &d
);
967 /* remainder, if any, all match the preimage */
968 handle_split(sb
, ent
, d
.tlno
, d
.plno
, ent
->num_lines
, parent
, split
);
972 * See if lines currently target is suspected for can be attributed to
975 static int find_move_in_parent(struct scoreboard
*sb
,
976 struct origin
*target
,
977 struct origin
*parent
)
979 int last_in_target
, made_progress
;
980 struct blame_entry
*e
, split
[3];
983 last_in_target
= find_last_in_target(sb
, target
);
984 if (last_in_target
< 0)
985 return 1; /* nothing remains for this target */
987 fill_origin_blob(&sb
->revs
->diffopt
, parent
, &file_p
);
992 while (made_progress
) {
994 for (e
= sb
->ent
; e
; e
= e
->next
) {
995 if (e
->guilty
|| !same_suspect(e
->suspect
, target
) ||
996 ent_score(sb
, e
) < blame_move_score
)
998 find_copy_in_blob(sb
, e
, parent
, split
, &file_p
);
999 if (split
[1].suspect
&&
1000 blame_move_score
< ent_score(sb
, &split
[1])) {
1001 split_blame(sb
, split
, e
);
1004 decref_split(split
);
1011 struct blame_entry
*ent
;
1012 struct blame_entry split
[3];
1016 * Count the number of entries the target is suspected for,
1017 * and prepare a list of entry and the best split.
1019 static struct blame_list
*setup_blame_list(struct scoreboard
*sb
,
1020 struct origin
*target
,
1024 struct blame_entry
*e
;
1026 struct blame_list
*blame_list
= NULL
;
1028 for (e
= sb
->ent
, num_ents
= 0; e
; e
= e
->next
)
1029 if (!e
->scanned
&& !e
->guilty
&&
1030 same_suspect(e
->suspect
, target
) &&
1031 min_score
< ent_score(sb
, e
))
1034 blame_list
= xcalloc(num_ents
, sizeof(struct blame_list
));
1035 for (e
= sb
->ent
, i
= 0; e
; e
= e
->next
)
1036 if (!e
->scanned
&& !e
->guilty
&&
1037 same_suspect(e
->suspect
, target
) &&
1038 min_score
< ent_score(sb
, e
))
1039 blame_list
[i
++].ent
= e
;
1041 *num_ents_p
= num_ents
;
1046 * Reset the scanned status on all entries.
1048 static void reset_scanned_flag(struct scoreboard
*sb
)
1050 struct blame_entry
*e
;
1051 for (e
= sb
->ent
; e
; e
= e
->next
)
1056 * For lines target is suspected for, see if we can find code movement
1057 * across file boundary from the parent commit. porigin is the path
1058 * in the parent we already tried.
1060 static int find_copy_in_parent(struct scoreboard
*sb
,
1061 struct origin
*target
,
1062 struct commit
*parent
,
1063 struct origin
*porigin
,
1066 struct diff_options diff_opts
;
1067 const char *paths
[1];
1070 struct blame_list
*blame_list
;
1073 blame_list
= setup_blame_list(sb
, target
, blame_copy_score
, &num_ents
);
1075 return 1; /* nothing remains for this target */
1077 diff_setup(&diff_opts
);
1078 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
1079 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
1082 diff_tree_setup_paths(paths
, &diff_opts
);
1083 diff_setup_done(&diff_opts
);
1085 /* Try "find copies harder" on new path if requested;
1086 * we do not want to use diffcore_rename() actually to
1087 * match things up; find_copies_harder is set only to
1088 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1089 * and this code needs to be after diff_setup_done(), which
1090 * usually makes find-copies-harder imply copy detection.
1092 if ((opt
& PICKAXE_BLAME_COPY_HARDEST
)
1093 || ((opt
& PICKAXE_BLAME_COPY_HARDER
)
1094 && (!porigin
|| strcmp(target
->path
, porigin
->path
))))
1095 DIFF_OPT_SET(&diff_opts
, FIND_COPIES_HARDER
);
1097 if (is_null_sha1(target
->commit
->object
.sha1
))
1098 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
1100 diff_tree_sha1(parent
->tree
->object
.sha1
,
1101 target
->commit
->tree
->object
.sha1
,
1104 if (!DIFF_OPT_TST(&diff_opts
, FIND_COPIES_HARDER
))
1105 diffcore_std(&diff_opts
);
1109 int made_progress
= 0;
1111 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
1112 struct diff_filepair
*p
= diff_queued_diff
.queue
[i
];
1113 struct origin
*norigin
;
1115 struct blame_entry
this[3];
1117 if (!DIFF_FILE_VALID(p
->one
))
1118 continue; /* does not exist in parent */
1119 if (S_ISGITLINK(p
->one
->mode
))
1120 continue; /* ignore git links */
1121 if (porigin
&& !strcmp(p
->one
->path
, porigin
->path
))
1122 /* find_move already dealt with this path */
1125 norigin
= get_origin(sb
, parent
, p
->one
->path
);
1126 hashcpy(norigin
->blob_sha1
, p
->one
->sha1
);
1127 norigin
->mode
= p
->one
->mode
;
1128 fill_origin_blob(&sb
->revs
->diffopt
, norigin
, &file_p
);
1132 for (j
= 0; j
< num_ents
; j
++) {
1133 find_copy_in_blob(sb
, blame_list
[j
].ent
,
1134 norigin
, this, &file_p
);
1135 copy_split_if_better(sb
, blame_list
[j
].split
,
1139 origin_decref(norigin
);
1142 for (j
= 0; j
< num_ents
; j
++) {
1143 struct blame_entry
*split
= blame_list
[j
].split
;
1144 if (split
[1].suspect
&&
1145 blame_copy_score
< ent_score(sb
, &split
[1])) {
1146 split_blame(sb
, split
, blame_list
[j
].ent
);
1150 blame_list
[j
].ent
->scanned
= 1;
1151 decref_split(split
);
1157 blame_list
= setup_blame_list(sb
, target
, blame_copy_score
, &num_ents
);
1163 reset_scanned_flag(sb
);
1164 diff_flush(&diff_opts
);
1165 diff_tree_release_paths(&diff_opts
);
1170 * The blobs of origin and porigin exactly match, so everything
1171 * origin is suspected for can be blamed on the parent.
1173 static void pass_whole_blame(struct scoreboard
*sb
,
1174 struct origin
*origin
, struct origin
*porigin
)
1176 struct blame_entry
*e
;
1178 if (!porigin
->file
.ptr
&& origin
->file
.ptr
) {
1179 /* Steal its file */
1180 porigin
->file
= origin
->file
;
1181 origin
->file
.ptr
= NULL
;
1183 for (e
= sb
->ent
; e
; e
= e
->next
) {
1184 if (!same_suspect(e
->suspect
, origin
))
1186 origin_incref(porigin
);
1187 origin_decref(e
->suspect
);
1188 e
->suspect
= porigin
;
1193 * We pass blame from the current commit to its parents. We keep saying
1194 * "parent" (and "porigin"), but what we mean is to find scapegoat to
1195 * exonerate ourselves.
1197 static struct commit_list
*first_scapegoat(struct rev_info
*revs
, struct commit
*commit
)
1200 return commit
->parents
;
1201 return lookup_decoration(&revs
->children
, &commit
->object
);
1204 static int num_scapegoats(struct rev_info
*revs
, struct commit
*commit
)
1207 struct commit_list
*l
= first_scapegoat(revs
, commit
);
1208 for (cnt
= 0; l
; l
= l
->next
)
1215 static void pass_blame(struct scoreboard
*sb
, struct origin
*origin
, int opt
)
1217 struct rev_info
*revs
= sb
->revs
;
1218 int i
, pass
, num_sg
;
1219 struct commit
*commit
= origin
->commit
;
1220 struct commit_list
*sg
;
1221 struct origin
*sg_buf
[MAXSG
];
1222 struct origin
*porigin
, **sg_origin
= sg_buf
;
1224 num_sg
= num_scapegoats(revs
, commit
);
1227 else if (num_sg
< ARRAY_SIZE(sg_buf
))
1228 memset(sg_buf
, 0, sizeof(sg_buf
));
1230 sg_origin
= xcalloc(num_sg
, sizeof(*sg_origin
));
1233 * The first pass looks for unrenamed path to optimize for
1234 * common cases, then we look for renames in the second pass.
1236 for (pass
= 0; pass
< 2 - no_whole_file_rename
; pass
++) {
1237 struct origin
*(*find
)(struct scoreboard
*,
1238 struct commit
*, struct origin
*);
1239 find
= pass
? find_rename
: find_origin
;
1241 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1243 sg
= sg
->next
, i
++) {
1244 struct commit
*p
= sg
->item
;
1249 if (parse_commit(p
))
1251 porigin
= find(sb
, p
, origin
);
1254 if (!hashcmp(porigin
->blob_sha1
, origin
->blob_sha1
)) {
1255 pass_whole_blame(sb
, origin
, porigin
);
1256 origin_decref(porigin
);
1259 for (j
= same
= 0; j
< i
; j
++)
1261 !hashcmp(sg_origin
[j
]->blob_sha1
,
1262 porigin
->blob_sha1
)) {
1267 sg_origin
[i
] = porigin
;
1269 origin_decref(porigin
);
1274 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1276 sg
= sg
->next
, i
++) {
1277 struct origin
*porigin
= sg_origin
[i
];
1280 if (!origin
->previous
) {
1281 origin_incref(porigin
);
1282 origin
->previous
= porigin
;
1284 if (pass_blame_to_parent(sb
, origin
, porigin
))
1289 * Optionally find moves in parents' files.
1291 if (opt
& PICKAXE_BLAME_MOVE
)
1292 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1294 sg
= sg
->next
, i
++) {
1295 struct origin
*porigin
= sg_origin
[i
];
1298 if (find_move_in_parent(sb
, origin
, porigin
))
1303 * Optionally find copies from parents' files.
1305 if (opt
& PICKAXE_BLAME_COPY
)
1306 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1308 sg
= sg
->next
, i
++) {
1309 struct origin
*porigin
= sg_origin
[i
];
1310 if (find_copy_in_parent(sb
, origin
, sg
->item
,
1316 for (i
= 0; i
< num_sg
; i
++) {
1318 drop_origin_blob(sg_origin
[i
]);
1319 origin_decref(sg_origin
[i
]);
1322 drop_origin_blob(origin
);
1323 if (sg_buf
!= sg_origin
)
1328 * Information on commits, used for output.
1330 struct commit_info
{
1331 struct strbuf author
;
1332 struct strbuf author_mail
;
1333 unsigned long author_time
;
1334 struct strbuf author_tz
;
1336 /* filled only when asked for details */
1337 struct strbuf committer
;
1338 struct strbuf committer_mail
;
1339 unsigned long committer_time
;
1340 struct strbuf committer_tz
;
1342 struct strbuf summary
;
1346 * Parse author/committer line in the commit object buffer
1348 static void get_ac_line(const char *inbuf
, const char *what
,
1349 struct strbuf
*name
, struct strbuf
*mail
,
1350 unsigned long *time
, struct strbuf
*tz
)
1352 struct ident_split ident
;
1353 size_t len
, maillen
, namelen
;
1355 const char *namebuf
, *mailbuf
;
1357 tmp
= strstr(inbuf
, what
);
1360 tmp
+= strlen(what
);
1361 endp
= strchr(tmp
, '\n');
1367 if (split_ident_line(&ident
, tmp
, len
)) {
1371 strbuf_addstr(name
, tmp
);
1372 strbuf_addstr(mail
, tmp
);
1373 strbuf_addstr(tz
, tmp
);
1378 namelen
= ident
.name_end
- ident
.name_begin
;
1379 namebuf
= ident
.name_begin
;
1381 maillen
= ident
.mail_end
- ident
.mail_begin
;
1382 mailbuf
= ident
.mail_begin
;
1384 if (ident
.date_begin
&& ident
.date_end
)
1385 *time
= strtoul(ident
.date_begin
, NULL
, 10);
1389 if (ident
.tz_begin
&& ident
.tz_end
)
1390 strbuf_add(tz
, ident
.tz_begin
, ident
.tz_end
- ident
.tz_begin
);
1392 strbuf_addstr(tz
, "(unknown)");
1395 * Now, convert both name and e-mail using mailmap
1397 map_user(&mailmap
, &mailbuf
, &maillen
,
1398 &namebuf
, &namelen
);
1400 strbuf_addf(mail
, "<%.*s>", (int)maillen
, mailbuf
);
1401 strbuf_add(name
, namebuf
, namelen
);
1404 static void commit_info_init(struct commit_info
*ci
)
1407 strbuf_init(&ci
->author
, 0);
1408 strbuf_init(&ci
->author_mail
, 0);
1409 strbuf_init(&ci
->author_tz
, 0);
1410 strbuf_init(&ci
->committer
, 0);
1411 strbuf_init(&ci
->committer_mail
, 0);
1412 strbuf_init(&ci
->committer_tz
, 0);
1413 strbuf_init(&ci
->summary
, 0);
1416 static void commit_info_destroy(struct commit_info
*ci
)
1419 strbuf_release(&ci
->author
);
1420 strbuf_release(&ci
->author_mail
);
1421 strbuf_release(&ci
->author_tz
);
1422 strbuf_release(&ci
->committer
);
1423 strbuf_release(&ci
->committer_mail
);
1424 strbuf_release(&ci
->committer_tz
);
1425 strbuf_release(&ci
->summary
);
1428 static void get_commit_info(struct commit
*commit
,
1429 struct commit_info
*ret
,
1433 const char *subject
, *encoding
;
1436 commit_info_init(ret
);
1438 encoding
= get_log_output_encoding();
1439 message
= logmsg_reencode(commit
, NULL
, encoding
);
1440 get_ac_line(message
, "\nauthor ",
1441 &ret
->author
, &ret
->author_mail
,
1442 &ret
->author_time
, &ret
->author_tz
);
1445 logmsg_free(message
, commit
);
1449 get_ac_line(message
, "\ncommitter ",
1450 &ret
->committer
, &ret
->committer_mail
,
1451 &ret
->committer_time
, &ret
->committer_tz
);
1453 len
= find_commit_subject(message
, &subject
);
1455 strbuf_add(&ret
->summary
, subject
, len
);
1457 strbuf_addf(&ret
->summary
, "(%s)", sha1_to_hex(commit
->object
.sha1
));
1459 logmsg_free(message
, commit
);
1463 * To allow LF and other nonportable characters in pathnames,
1464 * they are c-style quoted as needed.
1466 static void write_filename_info(const char *path
)
1468 printf("filename ");
1469 write_name_quoted(path
, stdout
, '\n');
1473 * Porcelain/Incremental format wants to show a lot of details per
1474 * commit. Instead of repeating this every line, emit it only once,
1475 * the first time each commit appears in the output (unless the
1476 * user has specifically asked for us to repeat).
1478 static int emit_one_suspect_detail(struct origin
*suspect
, int repeat
)
1480 struct commit_info ci
;
1482 if (!repeat
&& (suspect
->commit
->object
.flags
& METAINFO_SHOWN
))
1485 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
1486 get_commit_info(suspect
->commit
, &ci
, 1);
1487 printf("author %s\n", ci
.author
.buf
);
1488 printf("author-mail %s\n", ci
.author_mail
.buf
);
1489 printf("author-time %lu\n", ci
.author_time
);
1490 printf("author-tz %s\n", ci
.author_tz
.buf
);
1491 printf("committer %s\n", ci
.committer
.buf
);
1492 printf("committer-mail %s\n", ci
.committer_mail
.buf
);
1493 printf("committer-time %lu\n", ci
.committer_time
);
1494 printf("committer-tz %s\n", ci
.committer_tz
.buf
);
1495 printf("summary %s\n", ci
.summary
.buf
);
1496 if (suspect
->commit
->object
.flags
& UNINTERESTING
)
1497 printf("boundary\n");
1498 if (suspect
->previous
) {
1499 struct origin
*prev
= suspect
->previous
;
1500 printf("previous %s ", sha1_to_hex(prev
->commit
->object
.sha1
));
1501 write_name_quoted(prev
->path
, stdout
, '\n');
1504 commit_info_destroy(&ci
);
1510 * The blame_entry is found to be guilty for the range. Mark it
1511 * as such, and show it in incremental output.
1513 static void found_guilty_entry(struct blame_entry
*ent
)
1519 struct origin
*suspect
= ent
->suspect
;
1521 printf("%s %d %d %d\n",
1522 sha1_to_hex(suspect
->commit
->object
.sha1
),
1523 ent
->s_lno
+ 1, ent
->lno
+ 1, ent
->num_lines
);
1524 emit_one_suspect_detail(suspect
, 0);
1525 write_filename_info(suspect
->path
);
1526 maybe_flush_or_die(stdout
, "stdout");
1531 * The main loop -- while the scoreboard has lines whose true origin
1532 * is still unknown, pick one blame_entry, and allow its current
1533 * suspect to pass blames to its parents.
1535 static void assign_blame(struct scoreboard
*sb
, int opt
)
1537 struct rev_info
*revs
= sb
->revs
;
1540 struct blame_entry
*ent
;
1541 struct commit
*commit
;
1542 struct origin
*suspect
= NULL
;
1544 /* find one suspect to break down */
1545 for (ent
= sb
->ent
; !suspect
&& ent
; ent
= ent
->next
)
1547 suspect
= ent
->suspect
;
1549 return; /* all done */
1552 * We will use this suspect later in the loop,
1553 * so hold onto it in the meantime.
1555 origin_incref(suspect
);
1556 commit
= suspect
->commit
;
1557 if (!commit
->object
.parsed
)
1558 parse_commit(commit
);
1560 (!(commit
->object
.flags
& UNINTERESTING
) &&
1561 !(revs
->max_age
!= -1 && commit
->date
< revs
->max_age
)))
1562 pass_blame(sb
, suspect
, opt
);
1564 commit
->object
.flags
|= UNINTERESTING
;
1565 if (commit
->object
.parsed
)
1566 mark_parents_uninteresting(commit
);
1568 /* treat root commit as boundary */
1569 if (!commit
->parents
&& !show_root
)
1570 commit
->object
.flags
|= UNINTERESTING
;
1572 /* Take responsibility for the remaining entries */
1573 for (ent
= sb
->ent
; ent
; ent
= ent
->next
)
1574 if (same_suspect(ent
->suspect
, suspect
))
1575 found_guilty_entry(ent
);
1576 origin_decref(suspect
);
1578 if (DEBUG
) /* sanity */
1579 sanity_check_refcnt(sb
);
1583 static const char *format_time(unsigned long time
, const char *tz_str
,
1586 static char time_buf
[128];
1587 const char *time_str
;
1591 if (show_raw_time
) {
1592 snprintf(time_buf
, sizeof(time_buf
), "%lu %s", time
, tz_str
);
1596 time_str
= show_date(time
, tz
, blame_date_mode
);
1597 time_len
= strlen(time_str
);
1598 memcpy(time_buf
, time_str
, time_len
);
1599 memset(time_buf
+ time_len
, ' ', blame_date_width
- time_len
);
1604 #define OUTPUT_ANNOTATE_COMPAT 001
1605 #define OUTPUT_LONG_OBJECT_NAME 002
1606 #define OUTPUT_RAW_TIMESTAMP 004
1607 #define OUTPUT_PORCELAIN 010
1608 #define OUTPUT_SHOW_NAME 020
1609 #define OUTPUT_SHOW_NUMBER 040
1610 #define OUTPUT_SHOW_SCORE 0100
1611 #define OUTPUT_NO_AUTHOR 0200
1612 #define OUTPUT_SHOW_EMAIL 0400
1613 #define OUTPUT_LINE_PORCELAIN 01000
1615 static void emit_porcelain_details(struct origin
*suspect
, int repeat
)
1617 if (emit_one_suspect_detail(suspect
, repeat
) ||
1618 (suspect
->commit
->object
.flags
& MORE_THAN_ONE_PATH
))
1619 write_filename_info(suspect
->path
);
1622 static void emit_porcelain(struct scoreboard
*sb
, struct blame_entry
*ent
,
1625 int repeat
= opt
& OUTPUT_LINE_PORCELAIN
;
1628 struct origin
*suspect
= ent
->suspect
;
1631 strcpy(hex
, sha1_to_hex(suspect
->commit
->object
.sha1
));
1632 printf("%s%c%d %d %d\n",
1634 ent
->guilty
? ' ' : '*', /* purely for debugging */
1638 emit_porcelain_details(suspect
, repeat
);
1640 cp
= nth_line(sb
, ent
->lno
);
1641 for (cnt
= 0; cnt
< ent
->num_lines
; cnt
++) {
1644 printf("%s %d %d\n", hex
,
1645 ent
->s_lno
+ 1 + cnt
,
1646 ent
->lno
+ 1 + cnt
);
1648 emit_porcelain_details(suspect
, 1);
1654 } while (ch
!= '\n' &&
1655 cp
< sb
->final_buf
+ sb
->final_buf_size
);
1658 if (sb
->final_buf_size
&& cp
[-1] != '\n')
1662 static void emit_other(struct scoreboard
*sb
, struct blame_entry
*ent
, int opt
)
1666 struct origin
*suspect
= ent
->suspect
;
1667 struct commit_info ci
;
1669 int show_raw_time
= !!(opt
& OUTPUT_RAW_TIMESTAMP
);
1671 get_commit_info(suspect
->commit
, &ci
, 1);
1672 strcpy(hex
, sha1_to_hex(suspect
->commit
->object
.sha1
));
1674 cp
= nth_line(sb
, ent
->lno
);
1675 for (cnt
= 0; cnt
< ent
->num_lines
; cnt
++) {
1677 int length
= (opt
& OUTPUT_LONG_OBJECT_NAME
) ? 40 : abbrev
;
1679 if (suspect
->commit
->object
.flags
& UNINTERESTING
) {
1681 memset(hex
, ' ', length
);
1682 else if (!(opt
& OUTPUT_ANNOTATE_COMPAT
)) {
1688 printf("%.*s", length
, hex
);
1689 if (opt
& OUTPUT_ANNOTATE_COMPAT
) {
1691 if (opt
& OUTPUT_SHOW_EMAIL
)
1692 name
= ci
.author_mail
.buf
;
1694 name
= ci
.author
.buf
;
1695 printf("\t(%10s\t%10s\t%d)", name
,
1696 format_time(ci
.author_time
, ci
.author_tz
.buf
,
1698 ent
->lno
+ 1 + cnt
);
1700 if (opt
& OUTPUT_SHOW_SCORE
)
1702 max_score_digits
, ent
->score
,
1703 ent
->suspect
->refcnt
);
1704 if (opt
& OUTPUT_SHOW_NAME
)
1705 printf(" %-*.*s", longest_file
, longest_file
,
1707 if (opt
& OUTPUT_SHOW_NUMBER
)
1708 printf(" %*d", max_orig_digits
,
1709 ent
->s_lno
+ 1 + cnt
);
1711 if (!(opt
& OUTPUT_NO_AUTHOR
)) {
1714 if (opt
& OUTPUT_SHOW_EMAIL
)
1715 name
= ci
.author_mail
.buf
;
1717 name
= ci
.author
.buf
;
1718 pad
= longest_author
- utf8_strwidth(name
);
1719 printf(" (%s%*s %10s",
1721 format_time(ci
.author_time
,
1726 max_digits
, ent
->lno
+ 1 + cnt
);
1731 } while (ch
!= '\n' &&
1732 cp
< sb
->final_buf
+ sb
->final_buf_size
);
1735 if (sb
->final_buf_size
&& cp
[-1] != '\n')
1738 commit_info_destroy(&ci
);
1741 static void output(struct scoreboard
*sb
, int option
)
1743 struct blame_entry
*ent
;
1745 if (option
& OUTPUT_PORCELAIN
) {
1746 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1747 struct blame_entry
*oth
;
1748 struct origin
*suspect
= ent
->suspect
;
1749 struct commit
*commit
= suspect
->commit
;
1750 if (commit
->object
.flags
& MORE_THAN_ONE_PATH
)
1752 for (oth
= ent
->next
; oth
; oth
= oth
->next
) {
1753 if ((oth
->suspect
->commit
!= commit
) ||
1754 !strcmp(oth
->suspect
->path
, suspect
->path
))
1756 commit
->object
.flags
|= MORE_THAN_ONE_PATH
;
1762 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1763 if (option
& OUTPUT_PORCELAIN
)
1764 emit_porcelain(sb
, ent
, option
);
1766 emit_other(sb
, ent
, option
);
1772 * To allow quick access to the contents of nth line in the
1773 * final image, prepare an index in the scoreboard.
1775 static int prepare_lines(struct scoreboard
*sb
)
1777 const char *buf
= sb
->final_buf
;
1778 unsigned long len
= sb
->final_buf_size
;
1779 int num
= 0, incomplete
= 0, bol
= 1;
1781 if (len
&& buf
[len
-1] != '\n')
1782 incomplete
++; /* incomplete line at the end */
1785 sb
->lineno
= xrealloc(sb
->lineno
,
1786 sizeof(int *) * (num
+ 1));
1787 sb
->lineno
[num
] = buf
- sb
->final_buf
;
1790 if (*buf
++ == '\n') {
1795 sb
->lineno
= xrealloc(sb
->lineno
,
1796 sizeof(int *) * (num
+ incomplete
+ 1));
1797 sb
->lineno
[num
+ incomplete
] = buf
- sb
->final_buf
;
1798 sb
->num_lines
= num
+ incomplete
;
1799 return sb
->num_lines
;
1803 * Add phony grafts for use with -S; this is primarily to
1804 * support git's cvsserver that wants to give a linear history
1807 static int read_ancestry(const char *graft_file
)
1809 FILE *fp
= fopen(graft_file
, "r");
1813 while (fgets(buf
, sizeof(buf
), fp
)) {
1814 /* The format is just "Commit Parent1 Parent2 ...\n" */
1815 int len
= strlen(buf
);
1816 struct commit_graft
*graft
= read_graft_line(buf
, len
);
1818 register_commit_graft(graft
, 0);
1824 static int update_auto_abbrev(int auto_abbrev
, struct origin
*suspect
)
1826 const char *uniq
= find_unique_abbrev(suspect
->commit
->object
.sha1
,
1828 int len
= strlen(uniq
);
1829 if (auto_abbrev
< len
)
1835 * How many columns do we need to show line numbers, authors,
1838 static void find_alignment(struct scoreboard
*sb
, int *option
)
1840 int longest_src_lines
= 0;
1841 int longest_dst_lines
= 0;
1842 unsigned largest_score
= 0;
1843 struct blame_entry
*e
;
1844 int compute_auto_abbrev
= (abbrev
< 0);
1845 int auto_abbrev
= default_abbrev
;
1847 for (e
= sb
->ent
; e
; e
= e
->next
) {
1848 struct origin
*suspect
= e
->suspect
;
1849 struct commit_info ci
;
1852 if (compute_auto_abbrev
)
1853 auto_abbrev
= update_auto_abbrev(auto_abbrev
, suspect
);
1854 if (strcmp(suspect
->path
, sb
->path
))
1855 *option
|= OUTPUT_SHOW_NAME
;
1856 num
= strlen(suspect
->path
);
1857 if (longest_file
< num
)
1859 if (!(suspect
->commit
->object
.flags
& METAINFO_SHOWN
)) {
1860 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
1861 get_commit_info(suspect
->commit
, &ci
, 1);
1862 if (*option
& OUTPUT_SHOW_EMAIL
)
1863 num
= utf8_strwidth(ci
.author_mail
.buf
);
1865 num
= utf8_strwidth(ci
.author
.buf
);
1866 if (longest_author
< num
)
1867 longest_author
= num
;
1869 num
= e
->s_lno
+ e
->num_lines
;
1870 if (longest_src_lines
< num
)
1871 longest_src_lines
= num
;
1872 num
= e
->lno
+ e
->num_lines
;
1873 if (longest_dst_lines
< num
)
1874 longest_dst_lines
= num
;
1875 if (largest_score
< ent_score(sb
, e
))
1876 largest_score
= ent_score(sb
, e
);
1878 commit_info_destroy(&ci
);
1880 max_orig_digits
= decimal_width(longest_src_lines
);
1881 max_digits
= decimal_width(longest_dst_lines
);
1882 max_score_digits
= decimal_width(largest_score
);
1884 if (compute_auto_abbrev
)
1885 /* one more abbrev length is needed for the boundary commit */
1886 abbrev
= auto_abbrev
+ 1;
1890 * For debugging -- origin is refcounted, and this asserts that
1891 * we do not underflow.
1893 static void sanity_check_refcnt(struct scoreboard
*sb
)
1896 struct blame_entry
*ent
;
1898 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1899 /* Nobody should have zero or negative refcnt */
1900 if (ent
->suspect
->refcnt
<= 0) {
1901 fprintf(stderr
, "%s in %s has negative refcnt %d\n",
1903 sha1_to_hex(ent
->suspect
->commit
->object
.sha1
),
1904 ent
->suspect
->refcnt
);
1910 find_alignment(sb
, &opt
);
1912 die("Baa %d!", baa
);
1917 * Used for the command line parsing; check if the path exists
1918 * in the working tree.
1920 static int has_string_in_work_tree(const char *path
)
1923 return !lstat(path
, &st
);
1926 static unsigned parse_score(const char *arg
)
1929 unsigned long score
= strtoul(arg
, &end
, 10);
1935 static const char *add_prefix(const char *prefix
, const char *path
)
1937 return prefix_path(prefix
, prefix
? strlen(prefix
) : 0, path
);
1941 * Parsing of -L option
1943 static void prepare_blame_range(struct scoreboard
*sb
,
1944 const char *bottomtop
,
1946 long *bottom
, long *top
)
1948 if (parse_range_arg(bottomtop
, nth_line_cb
, sb
, lno
, bottom
, top
, sb
->path
))
1952 static int git_blame_config(const char *var
, const char *value
, void *cb
)
1954 if (!strcmp(var
, "blame.showroot")) {
1955 show_root
= git_config_bool(var
, value
);
1958 if (!strcmp(var
, "blame.blankboundary")) {
1959 blank_boundary
= git_config_bool(var
, value
);
1962 if (!strcmp(var
, "blame.date")) {
1964 return config_error_nonbool(var
);
1965 blame_date_mode
= parse_date_format(value
);
1969 if (userdiff_config(var
, value
) < 0)
1972 return git_default_config(var
, value
, cb
);
1975 static void verify_working_tree_path(struct commit
*work_tree
, const char *path
)
1977 struct commit_list
*parents
;
1979 for (parents
= work_tree
->parents
; parents
; parents
= parents
->next
) {
1980 const unsigned char *commit_sha1
= parents
->item
->object
.sha1
;
1981 unsigned char blob_sha1
[20];
1984 if (!get_tree_entry(commit_sha1
, path
, blob_sha1
, &mode
) &&
1985 sha1_object_info(blob_sha1
, NULL
) == OBJ_BLOB
)
1988 die("no such path '%s' in HEAD", path
);
1991 static struct commit_list
**append_parent(struct commit_list
**tail
, const unsigned char *sha1
)
1993 struct commit
*parent
;
1995 parent
= lookup_commit_reference(sha1
);
1997 die("no such commit %s", sha1_to_hex(sha1
));
1998 return &commit_list_insert(parent
, tail
)->next
;
2001 static void append_merge_parents(struct commit_list
**tail
)
2004 const char *merge_head_file
= git_path("MERGE_HEAD");
2005 struct strbuf line
= STRBUF_INIT
;
2007 merge_head
= open(merge_head_file
, O_RDONLY
);
2008 if (merge_head
< 0) {
2009 if (errno
== ENOENT
)
2011 die("cannot open '%s' for reading", merge_head_file
);
2014 while (!strbuf_getwholeline_fd(&line
, merge_head
, '\n')) {
2015 unsigned char sha1
[20];
2016 if (line
.len
< 40 || get_sha1_hex(line
.buf
, sha1
))
2017 die("unknown line in '%s': %s", merge_head_file
, line
.buf
);
2018 tail
= append_parent(tail
, sha1
);
2021 strbuf_release(&line
);
2025 * Prepare a dummy commit that represents the work tree (or staged) item.
2026 * Note that annotating work tree item never works in the reverse.
2028 static struct commit
*fake_working_tree_commit(struct diff_options
*opt
,
2030 const char *contents_from
)
2032 struct commit
*commit
;
2033 struct origin
*origin
;
2034 struct commit_list
**parent_tail
, *parent
;
2035 unsigned char head_sha1
[20];
2036 struct strbuf buf
= STRBUF_INIT
;
2040 struct cache_entry
*ce
;
2042 struct strbuf msg
= STRBUF_INIT
;
2045 commit
= xcalloc(1, sizeof(*commit
));
2046 commit
->object
.parsed
= 1;
2048 commit
->object
.type
= OBJ_COMMIT
;
2049 parent_tail
= &commit
->parents
;
2051 if (!resolve_ref_unsafe("HEAD", head_sha1
, 1, NULL
))
2052 die("no such ref: HEAD");
2054 parent_tail
= append_parent(parent_tail
, head_sha1
);
2055 append_merge_parents(parent_tail
);
2056 verify_working_tree_path(commit
, path
);
2058 origin
= make_origin(commit
, path
);
2060 ident
= fmt_ident("Not Committed Yet", "not.committed.yet", NULL
, 0);
2061 strbuf_addstr(&msg
, "tree 0000000000000000000000000000000000000000\n");
2062 for (parent
= commit
->parents
; parent
; parent
= parent
->next
)
2063 strbuf_addf(&msg
, "parent %s\n",
2064 sha1_to_hex(parent
->item
->object
.sha1
));
2068 "Version of %s from %s\n",
2070 (!contents_from
? path
:
2071 (!strcmp(contents_from
, "-") ? "standard input" : contents_from
)));
2072 commit
->buffer
= strbuf_detach(&msg
, NULL
);
2074 if (!contents_from
|| strcmp("-", contents_from
)) {
2076 const char *read_from
;
2078 unsigned long buf_len
;
2080 if (contents_from
) {
2081 if (stat(contents_from
, &st
) < 0)
2082 die_errno("Cannot stat '%s'", contents_from
);
2083 read_from
= contents_from
;
2086 if (lstat(path
, &st
) < 0)
2087 die_errno("Cannot lstat '%s'", path
);
2090 mode
= canon_mode(st
.st_mode
);
2092 switch (st
.st_mode
& S_IFMT
) {
2094 if (DIFF_OPT_TST(opt
, ALLOW_TEXTCONV
) &&
2095 textconv_object(read_from
, mode
, null_sha1
, 0, &buf_ptr
, &buf_len
))
2096 strbuf_attach(&buf
, buf_ptr
, buf_len
, buf_len
+ 1);
2097 else if (strbuf_read_file(&buf
, read_from
, st
.st_size
) != st
.st_size
)
2098 die_errno("cannot open or read '%s'", read_from
);
2101 if (strbuf_readlink(&buf
, read_from
, st
.st_size
) < 0)
2102 die_errno("cannot readlink '%s'", read_from
);
2105 die("unsupported file type %s", read_from
);
2109 /* Reading from stdin */
2111 if (strbuf_read(&buf
, 0, 0) < 0)
2112 die_errno("failed to read from stdin");
2114 convert_to_git(path
, buf
.buf
, buf
.len
, &buf
, 0);
2115 origin
->file
.ptr
= buf
.buf
;
2116 origin
->file
.size
= buf
.len
;
2117 pretend_sha1_file(buf
.buf
, buf
.len
, OBJ_BLOB
, origin
->blob_sha1
);
2118 commit
->util
= origin
;
2121 * Read the current index, replace the path entry with
2122 * origin->blob_sha1 without mucking with its mode or type
2123 * bits; we are not going to write this index out -- we just
2124 * want to run "diff-index --cached".
2131 int pos
= cache_name_pos(path
, len
);
2133 mode
= active_cache
[pos
]->ce_mode
;
2135 /* Let's not bother reading from HEAD tree */
2136 mode
= S_IFREG
| 0644;
2138 size
= cache_entry_size(len
);
2139 ce
= xcalloc(1, size
);
2140 hashcpy(ce
->sha1
, origin
->blob_sha1
);
2141 memcpy(ce
->name
, path
, len
);
2142 ce
->ce_flags
= create_ce_flags(0);
2143 ce
->ce_namelen
= len
;
2144 ce
->ce_mode
= create_ce_mode(mode
);
2145 add_cache_entry(ce
, ADD_CACHE_OK_TO_ADD
|ADD_CACHE_OK_TO_REPLACE
);
2148 * We are not going to write this out, so this does not matter
2149 * right now, but someday we might optimize diff-index --cached
2150 * with cache-tree information.
2152 cache_tree_invalidate_path(active_cache_tree
, path
);
2157 static const char *prepare_final(struct scoreboard
*sb
)
2160 const char *final_commit_name
= NULL
;
2161 struct rev_info
*revs
= sb
->revs
;
2164 * There must be one and only one positive commit in the
2165 * revs->pending array.
2167 for (i
= 0; i
< revs
->pending
.nr
; i
++) {
2168 struct object
*obj
= revs
->pending
.objects
[i
].item
;
2169 if (obj
->flags
& UNINTERESTING
)
2171 while (obj
->type
== OBJ_TAG
)
2172 obj
= deref_tag(obj
, NULL
, 0);
2173 if (obj
->type
!= OBJ_COMMIT
)
2174 die("Non commit %s?", revs
->pending
.objects
[i
].name
);
2176 die("More than one commit to dig from %s and %s?",
2177 revs
->pending
.objects
[i
].name
,
2179 sb
->final
= (struct commit
*) obj
;
2180 final_commit_name
= revs
->pending
.objects
[i
].name
;
2182 return final_commit_name
;
2185 static const char *prepare_initial(struct scoreboard
*sb
)
2188 const char *final_commit_name
= NULL
;
2189 struct rev_info
*revs
= sb
->revs
;
2192 * There must be one and only one negative commit, and it must be
2195 for (i
= 0; i
< revs
->pending
.nr
; i
++) {
2196 struct object
*obj
= revs
->pending
.objects
[i
].item
;
2197 if (!(obj
->flags
& UNINTERESTING
))
2199 while (obj
->type
== OBJ_TAG
)
2200 obj
= deref_tag(obj
, NULL
, 0);
2201 if (obj
->type
!= OBJ_COMMIT
)
2202 die("Non commit %s?", revs
->pending
.objects
[i
].name
);
2204 die("More than one commit to dig down to %s and %s?",
2205 revs
->pending
.objects
[i
].name
,
2207 sb
->final
= (struct commit
*) obj
;
2208 final_commit_name
= revs
->pending
.objects
[i
].name
;
2210 if (!final_commit_name
)
2211 die("No commit to dig down to?");
2212 return final_commit_name
;
2215 static int blame_copy_callback(const struct option
*option
, const char *arg
, int unset
)
2217 int *opt
= option
->value
;
2220 * -C enables copy from removed files;
2221 * -C -C enables copy from existing files, but only
2222 * when blaming a new file;
2223 * -C -C -C enables copy from existing files for
2226 if (*opt
& PICKAXE_BLAME_COPY_HARDER
)
2227 *opt
|= PICKAXE_BLAME_COPY_HARDEST
;
2228 if (*opt
& PICKAXE_BLAME_COPY
)
2229 *opt
|= PICKAXE_BLAME_COPY_HARDER
;
2230 *opt
|= PICKAXE_BLAME_COPY
| PICKAXE_BLAME_MOVE
;
2233 blame_copy_score
= parse_score(arg
);
2237 static int blame_move_callback(const struct option
*option
, const char *arg
, int unset
)
2239 int *opt
= option
->value
;
2241 *opt
|= PICKAXE_BLAME_MOVE
;
2244 blame_move_score
= parse_score(arg
);
2248 static int blame_bottomtop_callback(const struct option
*option
, const char *arg
, int unset
)
2250 const char **bottomtop
= option
->value
;
2254 die("More than one '-L n,m' option given");
2259 int cmd_blame(int argc
, const char **argv
, const char *prefix
)
2261 struct rev_info revs
;
2263 struct scoreboard sb
;
2265 struct blame_entry
*ent
;
2266 long dashdash_pos
, bottom
, top
, lno
;
2267 const char *final_commit_name
= NULL
;
2268 enum object_type type
;
2270 static const char *bottomtop
= NULL
;
2271 static int output_option
= 0, opt
= 0;
2272 static int show_stats
= 0;
2273 static const char *revs_file
= NULL
;
2274 static const char *contents_from
= NULL
;
2275 static const struct option options
[] = {
2276 OPT_BOOLEAN(0, "incremental", &incremental
, N_("Show blame entries as we find them, incrementally")),
2277 OPT_BOOLEAN('b', NULL
, &blank_boundary
, N_("Show blank SHA-1 for boundary commits (Default: off)")),
2278 OPT_BOOLEAN(0, "root", &show_root
, N_("Do not treat root commits as boundaries (Default: off)")),
2279 OPT_BOOLEAN(0, "show-stats", &show_stats
, N_("Show work cost statistics")),
2280 OPT_BIT(0, "score-debug", &output_option
, N_("Show output score for blame entries"), OUTPUT_SHOW_SCORE
),
2281 OPT_BIT('f', "show-name", &output_option
, N_("Show original filename (Default: auto)"), OUTPUT_SHOW_NAME
),
2282 OPT_BIT('n', "show-number", &output_option
, N_("Show original linenumber (Default: off)"), OUTPUT_SHOW_NUMBER
),
2283 OPT_BIT('p', "porcelain", &output_option
, N_("Show in a format designed for machine consumption"), OUTPUT_PORCELAIN
),
2284 OPT_BIT(0, "line-porcelain", &output_option
, N_("Show porcelain format with per-line commit information"), OUTPUT_PORCELAIN
|OUTPUT_LINE_PORCELAIN
),
2285 OPT_BIT('c', NULL
, &output_option
, N_("Use the same output mode as git-annotate (Default: off)"), OUTPUT_ANNOTATE_COMPAT
),
2286 OPT_BIT('t', NULL
, &output_option
, N_("Show raw timestamp (Default: off)"), OUTPUT_RAW_TIMESTAMP
),
2287 OPT_BIT('l', NULL
, &output_option
, N_("Show long commit SHA1 (Default: off)"), OUTPUT_LONG_OBJECT_NAME
),
2288 OPT_BIT('s', NULL
, &output_option
, N_("Suppress author name and timestamp (Default: off)"), OUTPUT_NO_AUTHOR
),
2289 OPT_BIT('e', "show-email", &output_option
, N_("Show author email instead of name (Default: off)"), OUTPUT_SHOW_EMAIL
),
2290 OPT_BIT('w', NULL
, &xdl_opts
, N_("Ignore whitespace differences"), XDF_IGNORE_WHITESPACE
),
2291 OPT_BIT(0, "minimal", &xdl_opts
, N_("Spend extra cycles to find better match"), XDF_NEED_MINIMAL
),
2292 OPT_STRING('S', NULL
, &revs_file
, N_("file"), N_("Use revisions from <file> instead of calling git-rev-list")),
2293 OPT_STRING(0, "contents", &contents_from
, N_("file"), N_("Use <file>'s contents as the final image")),
2294 { OPTION_CALLBACK
, 'C', NULL
, &opt
, N_("score"), N_("Find line copies within and across files"), PARSE_OPT_OPTARG
, blame_copy_callback
},
2295 { OPTION_CALLBACK
, 'M', NULL
, &opt
, N_("score"), N_("Find line movements within and across files"), PARSE_OPT_OPTARG
, blame_move_callback
},
2296 OPT_CALLBACK('L', NULL
, &bottomtop
, N_("n,m"), N_("Process only line range n,m, counting from 1"), blame_bottomtop_callback
),
2297 OPT__ABBREV(&abbrev
),
2301 struct parse_opt_ctx_t ctx
;
2302 int cmd_is_annotate
= !strcmp(argv
[0], "annotate");
2304 git_config(git_blame_config
, NULL
);
2305 init_revisions(&revs
, NULL
);
2306 revs
.date_mode
= blame_date_mode
;
2307 DIFF_OPT_SET(&revs
.diffopt
, ALLOW_TEXTCONV
);
2308 DIFF_OPT_SET(&revs
.diffopt
, FOLLOW_RENAMES
);
2310 save_commit_buffer
= 0;
2313 parse_options_start(&ctx
, argc
, argv
, prefix
, options
,
2314 PARSE_OPT_KEEP_DASHDASH
| PARSE_OPT_KEEP_ARGV0
);
2316 switch (parse_options_step(&ctx
, options
, blame_opt_usage
)) {
2317 case PARSE_OPT_HELP
:
2319 case PARSE_OPT_DONE
:
2321 dashdash_pos
= ctx
.cpidx
;
2325 if (!strcmp(ctx
.argv
[0], "--reverse")) {
2326 ctx
.argv
[0] = "--children";
2329 parse_revision_opt(&revs
, &ctx
, options
, blame_opt_usage
);
2332 no_whole_file_rename
= !DIFF_OPT_TST(&revs
.diffopt
, FOLLOW_RENAMES
);
2333 DIFF_OPT_CLR(&revs
.diffopt
, FOLLOW_RENAMES
);
2334 argc
= parse_options_end(&ctx
);
2337 /* one more abbrev length is needed for the boundary commit */
2340 if (revs_file
&& read_ancestry(revs_file
))
2341 die_errno("reading graft file '%s' failed", revs_file
);
2343 if (cmd_is_annotate
) {
2344 output_option
|= OUTPUT_ANNOTATE_COMPAT
;
2345 blame_date_mode
= DATE_ISO8601
;
2347 blame_date_mode
= revs
.date_mode
;
2350 /* The maximum width used to show the dates */
2351 switch (blame_date_mode
) {
2353 blame_date_width
= sizeof("Thu, 19 Oct 2006 16:00:04 -0700");
2356 blame_date_width
= sizeof("2006-10-19 16:00:04 -0700");
2359 blame_date_width
= sizeof("1161298804 -0700");
2362 blame_date_width
= sizeof("2006-10-19");
2365 /* "normal" is used as the fallback for "relative" */
2368 blame_date_width
= sizeof("Thu Oct 19 16:00:04 2006 -0700");
2371 blame_date_width
-= 1; /* strip the null */
2373 if (DIFF_OPT_TST(&revs
.diffopt
, FIND_COPIES_HARDER
))
2374 opt
|= (PICKAXE_BLAME_COPY
| PICKAXE_BLAME_MOVE
|
2375 PICKAXE_BLAME_COPY_HARDER
);
2377 if (!blame_move_score
)
2378 blame_move_score
= BLAME_DEFAULT_MOVE_SCORE
;
2379 if (!blame_copy_score
)
2380 blame_copy_score
= BLAME_DEFAULT_COPY_SCORE
;
2383 * We have collected options unknown to us in argv[1..unk]
2384 * which are to be passed to revision machinery if we are
2385 * going to do the "bottom" processing.
2387 * The remaining are:
2389 * (1) if dashdash_pos != 0, it is either
2390 * "blame [revisions] -- <path>" or
2391 * "blame -- <path> <rev>"
2393 * (2) otherwise, it is one of the two:
2394 * "blame [revisions] <path>"
2395 * "blame <path> <rev>"
2397 * Note that we must strip out <path> from the arguments: we do not
2398 * want the path pruning but we may want "bottom" processing.
2401 switch (argc
- dashdash_pos
- 1) {
2404 usage_with_options(blame_opt_usage
, options
);
2405 /* reorder for the new way: <rev> -- <path> */
2411 path
= add_prefix(prefix
, argv
[--argc
]);
2415 usage_with_options(blame_opt_usage
, options
);
2419 usage_with_options(blame_opt_usage
, options
);
2420 path
= add_prefix(prefix
, argv
[argc
- 1]);
2421 if (argc
== 3 && !has_string_in_work_tree(path
)) { /* (2b) */
2422 path
= add_prefix(prefix
, argv
[1]);
2425 argv
[argc
- 1] = "--";
2428 if (!has_string_in_work_tree(path
))
2429 die_errno("cannot stat path '%s'", path
);
2432 revs
.disable_stdin
= 1;
2433 setup_revisions(argc
, argv
, &revs
, NULL
);
2434 memset(&sb
, 0, sizeof(sb
));
2438 final_commit_name
= prepare_final(&sb
);
2439 else if (contents_from
)
2440 die("--contents and --children do not blend well.");
2442 final_commit_name
= prepare_initial(&sb
);
2446 * "--not A B -- path" without anything positive;
2447 * do not default to HEAD, but use the working tree
2451 sb
.final
= fake_working_tree_commit(&sb
.revs
->diffopt
,
2452 path
, contents_from
);
2453 add_pending_object(&revs
, &(sb
.final
->object
), ":");
2455 else if (contents_from
)
2456 die("Cannot use --contents with final commit object name");
2459 * If we have bottom, this will mark the ancestors of the
2460 * bottom commits we would reach while traversing as
2463 if (prepare_revision_walk(&revs
))
2464 die("revision walk setup failed");
2466 if (is_null_sha1(sb
.final
->object
.sha1
)) {
2469 buf
= xmalloc(o
->file
.size
+ 1);
2470 memcpy(buf
, o
->file
.ptr
, o
->file
.size
+ 1);
2472 sb
.final_buf_size
= o
->file
.size
;
2475 o
= get_origin(&sb
, sb
.final
, path
);
2476 if (fill_blob_sha1_and_mode(o
))
2477 die("no such path %s in %s", path
, final_commit_name
);
2479 if (DIFF_OPT_TST(&sb
.revs
->diffopt
, ALLOW_TEXTCONV
) &&
2480 textconv_object(path
, o
->mode
, o
->blob_sha1
, 1, (char **) &sb
.final_buf
,
2481 &sb
.final_buf_size
))
2484 sb
.final_buf
= read_sha1_file(o
->blob_sha1
, &type
,
2485 &sb
.final_buf_size
);
2488 die("Cannot read blob %s for path %s",
2489 sha1_to_hex(o
->blob_sha1
),
2493 lno
= prepare_lines(&sb
);
2497 prepare_blame_range(&sb
, bottomtop
, lno
, &bottom
, &top
);
2498 if (lno
< top
|| ((lno
|| bottom
) && lno
< bottom
))
2499 die("file %s has only %lu lines", path
, lno
);
2506 ent
= xcalloc(1, sizeof(*ent
));
2508 ent
->num_lines
= top
- bottom
;
2510 ent
->s_lno
= bottom
;
2515 read_mailmap(&mailmap
, NULL
);
2520 assign_blame(&sb
, opt
);
2527 if (!(output_option
& OUTPUT_PORCELAIN
))
2528 find_alignment(&sb
, &output_option
);
2530 output(&sb
, output_option
);
2531 free((void *)sb
.final_buf
);
2532 for (ent
= sb
.ent
; ent
; ) {
2533 struct blame_entry
*e
= ent
->next
;
2539 printf("num read blob: %d\n", num_read_blob
);
2540 printf("num get patch: %d\n", num_get_patch
);
2541 printf("num commits: %d\n", num_commits
);