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"
27 static char blame_usage
[] = N_("git blame [options] [rev-opts] [rev] [--] file");
29 static const char *blame_opt_usage
[] = {
32 N_("[rev-opts] are documented in git-rev-list(1)"),
36 static int longest_file
;
37 static int longest_author
;
38 static int max_orig_digits
;
39 static int max_digits
;
40 static int max_score_digits
;
43 static int blank_boundary
;
44 static int incremental
;
46 static int abbrev
= -1;
47 static int no_whole_file_rename
;
49 static enum date_mode blame_date_mode
= DATE_ISO8601
;
50 static size_t blame_date_width
;
52 static struct string_list mailmap
;
59 static int num_read_blob
;
60 static int num_get_patch
;
61 static int num_commits
;
63 #define PICKAXE_BLAME_MOVE 01
64 #define PICKAXE_BLAME_COPY 02
65 #define PICKAXE_BLAME_COPY_HARDER 04
66 #define PICKAXE_BLAME_COPY_HARDEST 010
69 * blame for a blame_entry with score lower than these thresholds
70 * is not passed to the parent using move/copy logic.
72 static unsigned blame_move_score
;
73 static unsigned blame_copy_score
;
74 #define BLAME_DEFAULT_MOVE_SCORE 20
75 #define BLAME_DEFAULT_COPY_SCORE 40
77 /* bits #0..7 in revision.h, #8..11 used for merge_bases() in commit.c */
78 #define METAINFO_SHOWN (1u<<12)
79 #define MORE_THAN_ONE_PATH (1u<<13)
82 * One blob in a commit that is being suspected
86 struct origin
*previous
;
87 struct commit
*commit
;
89 unsigned char blob_sha1
[20];
91 char path
[FLEX_ARRAY
];
94 static int diff_hunks(mmfile_t
*file_a
, mmfile_t
*file_b
, long ctxlen
,
95 xdl_emit_hunk_consume_func_t hunk_func
, void *cb_data
)
98 xdemitconf_t xecfg
= {0};
99 xdemitcb_t ecb
= {NULL
};
101 xpp
.flags
= xdl_opts
;
102 xecfg
.ctxlen
= ctxlen
;
103 xecfg
.hunk_func
= hunk_func
;
105 return xdi_diff(file_a
, file_b
, &xpp
, &xecfg
, &ecb
);
109 * Prepare diff_filespec and convert it using diff textconv API
110 * if the textconv driver exists.
111 * Return 1 if the conversion succeeds, 0 otherwise.
113 int textconv_object(const char *path
,
115 const unsigned char *sha1
,
118 unsigned long *buf_size
)
120 struct diff_filespec
*df
;
121 struct userdiff_driver
*textconv
;
123 df
= alloc_filespec(path
);
124 fill_filespec(df
, sha1
, sha1_valid
, mode
);
125 textconv
= get_textconv(df
);
131 *buf_size
= fill_textconv(textconv
, df
, buf
);
137 * Given an origin, prepare mmfile_t structure to be used by the
140 static void fill_origin_blob(struct diff_options
*opt
,
141 struct origin
*o
, mmfile_t
*file
)
144 enum object_type type
;
145 unsigned long file_size
;
148 if (DIFF_OPT_TST(opt
, ALLOW_TEXTCONV
) &&
149 textconv_object(o
->path
, o
->mode
, o
->blob_sha1
, 1, &file
->ptr
, &file_size
))
152 file
->ptr
= read_sha1_file(o
->blob_sha1
, &type
, &file_size
);
153 file
->size
= file_size
;
156 die("Cannot read blob %s for path %s",
157 sha1_to_hex(o
->blob_sha1
),
166 * Origin is refcounted and usually we keep the blob contents to be
169 static inline struct origin
*origin_incref(struct origin
*o
)
176 static void origin_decref(struct origin
*o
)
178 if (o
&& --o
->refcnt
<= 0) {
180 origin_decref(o
->previous
);
186 static void drop_origin_blob(struct origin
*o
)
195 * Each group of lines is described by a blame_entry; it can be split
196 * as we pass blame to the parents. They form a linked list in the
197 * scoreboard structure, sorted by the target line number.
200 struct blame_entry
*prev
;
201 struct blame_entry
*next
;
203 /* the first line of this group in the final image;
204 * internally all line numbers are 0 based.
208 /* how many lines this group has */
211 /* the commit that introduced this group into the final image */
212 struct origin
*suspect
;
214 /* true if the suspect is truly guilty; false while we have not
215 * checked if the group came from one of its parents.
219 /* true if the entry has been scanned for copies in the current parent
223 /* the line number of the first line of this group in the
224 * suspect's file; internally all line numbers are 0 based.
228 /* how significant this entry is -- cached to avoid
229 * scanning the lines over and over.
235 * The current state of the blame assignment.
238 /* the final commit (i.e. where we started digging from) */
239 struct commit
*final
;
240 struct rev_info
*revs
;
244 * The contents in the final image.
245 * Used by many functions to obtain contents of the nth line,
246 * indexed with scoreboard.lineno[blame_entry.lno].
248 const char *final_buf
;
249 unsigned long final_buf_size
;
251 /* linked list of blames */
252 struct blame_entry
*ent
;
254 /* look-up a line in the final buffer */
259 static inline int same_suspect(struct origin
*a
, struct origin
*b
)
263 if (a
->commit
!= b
->commit
)
265 return !strcmp(a
->path
, b
->path
);
268 static void sanity_check_refcnt(struct scoreboard
*);
271 * If two blame entries that are next to each other came from
272 * contiguous lines in the same origin (i.e. <commit, path> pair),
273 * merge them together.
275 static void coalesce(struct scoreboard
*sb
)
277 struct blame_entry
*ent
, *next
;
279 for (ent
= sb
->ent
; ent
&& (next
= ent
->next
); ent
= next
) {
280 if (same_suspect(ent
->suspect
, next
->suspect
) &&
281 ent
->guilty
== next
->guilty
&&
282 ent
->s_lno
+ ent
->num_lines
== next
->s_lno
) {
283 ent
->num_lines
+= next
->num_lines
;
284 ent
->next
= next
->next
;
286 ent
->next
->prev
= ent
;
287 origin_decref(next
->suspect
);
290 next
= ent
; /* again */
294 if (DEBUG
) /* sanity */
295 sanity_check_refcnt(sb
);
299 * Given a commit and a path in it, create a new origin structure.
300 * The callers that add blame to the scoreboard should use
301 * get_origin() to obtain shared, refcounted copy instead of calling
302 * this function directly.
304 static struct origin
*make_origin(struct commit
*commit
, const char *path
)
307 o
= xcalloc(1, sizeof(*o
) + strlen(path
) + 1);
310 strcpy(o
->path
, path
);
315 * Locate an existing origin or create a new one.
317 static struct origin
*get_origin(struct scoreboard
*sb
,
318 struct commit
*commit
,
321 struct blame_entry
*e
;
323 for (e
= sb
->ent
; e
; e
= e
->next
) {
324 if (e
->suspect
->commit
== commit
&&
325 !strcmp(e
->suspect
->path
, path
))
326 return origin_incref(e
->suspect
);
328 return make_origin(commit
, path
);
332 * Fill the blob_sha1 field of an origin if it hasn't, so that later
333 * call to fill_origin_blob() can use it to locate the data. blob_sha1
334 * for an origin is also used to pass the blame for the entire file to
335 * the parent to detect the case where a child's blob is identical to
336 * that of its parent's.
338 * This also fills origin->mode for corresponding tree path.
340 static int fill_blob_sha1_and_mode(struct origin
*origin
)
342 if (!is_null_sha1(origin
->blob_sha1
))
344 if (get_tree_entry(origin
->commit
->object
.sha1
,
346 origin
->blob_sha1
, &origin
->mode
))
348 if (sha1_object_info(origin
->blob_sha1
, NULL
) != OBJ_BLOB
)
352 hashclr(origin
->blob_sha1
);
353 origin
->mode
= S_IFINVALID
;
358 * We have an origin -- check if the same path exists in the
359 * parent and return an origin structure to represent it.
361 static struct origin
*find_origin(struct scoreboard
*sb
,
362 struct commit
*parent
,
363 struct origin
*origin
)
365 struct origin
*porigin
= NULL
;
366 struct diff_options diff_opts
;
367 const char *paths
[2];
371 * Each commit object can cache one origin in that
372 * commit. This is a freestanding copy of origin and
375 struct origin
*cached
= parent
->util
;
376 if (!strcmp(cached
->path
, origin
->path
)) {
378 * The same path between origin and its parent
379 * without renaming -- the most common case.
381 porigin
= get_origin(sb
, parent
, cached
->path
);
384 * If the origin was newly created (i.e. get_origin
385 * would call make_origin if none is found in the
386 * scoreboard), it does not know the blob_sha1/mode,
387 * so copy it. Otherwise porigin was in the
388 * scoreboard and already knows blob_sha1/mode.
390 if (porigin
->refcnt
== 1) {
391 hashcpy(porigin
->blob_sha1
, cached
->blob_sha1
);
392 porigin
->mode
= cached
->mode
;
396 /* otherwise it was not very useful; free it */
401 /* See if the origin->path is different between parent
402 * and origin first. Most of the time they are the
403 * same and diff-tree is fairly efficient about this.
405 diff_setup(&diff_opts
);
406 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
407 diff_opts
.detect_rename
= 0;
408 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
409 paths
[0] = origin
->path
;
412 diff_tree_setup_paths(paths
, &diff_opts
);
413 diff_setup_done(&diff_opts
);
415 if (is_null_sha1(origin
->commit
->object
.sha1
))
416 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
418 diff_tree_sha1(parent
->tree
->object
.sha1
,
419 origin
->commit
->tree
->object
.sha1
,
421 diffcore_std(&diff_opts
);
423 if (!diff_queued_diff
.nr
) {
424 /* The path is the same as parent */
425 porigin
= get_origin(sb
, parent
, origin
->path
);
426 hashcpy(porigin
->blob_sha1
, origin
->blob_sha1
);
427 porigin
->mode
= origin
->mode
;
430 * Since origin->path is a pathspec, if the parent
431 * commit had it as a directory, we will see a whole
432 * bunch of deletion of files in the directory that we
436 struct diff_filepair
*p
= NULL
;
437 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
439 p
= diff_queued_diff
.queue
[i
];
440 name
= p
->one
->path
? p
->one
->path
: p
->two
->path
;
441 if (!strcmp(name
, origin
->path
))
445 die("internal error in blame::find_origin");
448 die("internal error in blame::find_origin (%c)",
451 porigin
= get_origin(sb
, parent
, origin
->path
);
452 hashcpy(porigin
->blob_sha1
, p
->one
->sha1
);
453 porigin
->mode
= p
->one
->mode
;
457 /* Did not exist in parent, or type changed */
461 diff_flush(&diff_opts
);
462 diff_tree_release_paths(&diff_opts
);
465 * Create a freestanding copy that is not part of
466 * the refcounted origin found in the scoreboard, and
467 * cache it in the commit.
469 struct origin
*cached
;
471 cached
= make_origin(porigin
->commit
, porigin
->path
);
472 hashcpy(cached
->blob_sha1
, porigin
->blob_sha1
);
473 cached
->mode
= porigin
->mode
;
474 parent
->util
= cached
;
480 * We have an origin -- find the path that corresponds to it in its
481 * parent and return an origin structure to represent it.
483 static struct origin
*find_rename(struct scoreboard
*sb
,
484 struct commit
*parent
,
485 struct origin
*origin
)
487 struct origin
*porigin
= NULL
;
488 struct diff_options diff_opts
;
490 const char *paths
[2];
492 diff_setup(&diff_opts
);
493 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
494 diff_opts
.detect_rename
= DIFF_DETECT_RENAME
;
495 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
496 diff_opts
.single_follow
= origin
->path
;
498 diff_tree_setup_paths(paths
, &diff_opts
);
499 diff_setup_done(&diff_opts
);
501 if (is_null_sha1(origin
->commit
->object
.sha1
))
502 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
504 diff_tree_sha1(parent
->tree
->object
.sha1
,
505 origin
->commit
->tree
->object
.sha1
,
507 diffcore_std(&diff_opts
);
509 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
510 struct diff_filepair
*p
= diff_queued_diff
.queue
[i
];
511 if ((p
->status
== 'R' || p
->status
== 'C') &&
512 !strcmp(p
->two
->path
, origin
->path
)) {
513 porigin
= get_origin(sb
, parent
, p
->one
->path
);
514 hashcpy(porigin
->blob_sha1
, p
->one
->sha1
);
515 porigin
->mode
= p
->one
->mode
;
519 diff_flush(&diff_opts
);
520 diff_tree_release_paths(&diff_opts
);
525 * Link in a new blame entry to the scoreboard. Entries that cover the
526 * same line range have been removed from the scoreboard previously.
528 static void add_blame_entry(struct scoreboard
*sb
, struct blame_entry
*e
)
530 struct blame_entry
*ent
, *prev
= NULL
;
532 origin_incref(e
->suspect
);
534 for (ent
= sb
->ent
; ent
&& ent
->lno
< e
->lno
; ent
= ent
->next
)
537 /* prev, if not NULL, is the last one that is below e */
540 e
->next
= prev
->next
;
552 * src typically is on-stack; we want to copy the information in it to
553 * a malloced blame_entry that is already on the linked list of the
554 * scoreboard. The origin of dst loses a refcnt while the origin of src
557 static void dup_entry(struct blame_entry
*dst
, struct blame_entry
*src
)
559 struct blame_entry
*p
, *n
;
563 origin_incref(src
->suspect
);
564 origin_decref(dst
->suspect
);
565 memcpy(dst
, src
, sizeof(*src
));
571 static const char *nth_line(struct scoreboard
*sb
, long lno
)
573 return sb
->final_buf
+ sb
->lineno
[lno
];
576 static const char *nth_line_cb(void *data
, long lno
)
578 return nth_line((struct scoreboard
*)data
, lno
);
582 * It is known that lines between tlno to same came from parent, and e
583 * has an overlap with that range. it also is known that parent's
584 * line plno corresponds to e's line tlno.
590 * <------------------>
592 * Split e into potentially three parts; before this chunk, the chunk
593 * to be blamed for the parent, and after that portion.
595 static void split_overlap(struct blame_entry
*split
,
596 struct blame_entry
*e
,
597 int tlno
, int plno
, int same
,
598 struct origin
*parent
)
601 memset(split
, 0, sizeof(struct blame_entry
[3]));
603 if (e
->s_lno
< tlno
) {
604 /* there is a pre-chunk part not blamed on parent */
605 split
[0].suspect
= origin_incref(e
->suspect
);
606 split
[0].lno
= e
->lno
;
607 split
[0].s_lno
= e
->s_lno
;
608 split
[0].num_lines
= tlno
- e
->s_lno
;
609 split
[1].lno
= e
->lno
+ tlno
- e
->s_lno
;
610 split
[1].s_lno
= plno
;
613 split
[1].lno
= e
->lno
;
614 split
[1].s_lno
= plno
+ (e
->s_lno
- tlno
);
617 if (same
< e
->s_lno
+ e
->num_lines
) {
618 /* there is a post-chunk part not blamed on parent */
619 split
[2].suspect
= origin_incref(e
->suspect
);
620 split
[2].lno
= e
->lno
+ (same
- e
->s_lno
);
621 split
[2].s_lno
= e
->s_lno
+ (same
- e
->s_lno
);
622 split
[2].num_lines
= e
->s_lno
+ e
->num_lines
- same
;
623 chunk_end_lno
= split
[2].lno
;
626 chunk_end_lno
= e
->lno
+ e
->num_lines
;
627 split
[1].num_lines
= chunk_end_lno
- split
[1].lno
;
630 * if it turns out there is nothing to blame the parent for,
631 * forget about the splitting. !split[1].suspect signals this.
633 if (split
[1].num_lines
< 1)
635 split
[1].suspect
= origin_incref(parent
);
639 * split_overlap() divided an existing blame e into up to three parts
640 * in split. Adjust the linked list of blames in the scoreboard to
643 static void split_blame(struct scoreboard
*sb
,
644 struct blame_entry
*split
,
645 struct blame_entry
*e
)
647 struct blame_entry
*new_entry
;
649 if (split
[0].suspect
&& split
[2].suspect
) {
650 /* The first part (reuse storage for the existing entry e) */
651 dup_entry(e
, &split
[0]);
653 /* The last part -- me */
654 new_entry
= xmalloc(sizeof(*new_entry
));
655 memcpy(new_entry
, &(split
[2]), sizeof(struct blame_entry
));
656 add_blame_entry(sb
, new_entry
);
658 /* ... and the middle part -- parent */
659 new_entry
= xmalloc(sizeof(*new_entry
));
660 memcpy(new_entry
, &(split
[1]), sizeof(struct blame_entry
));
661 add_blame_entry(sb
, new_entry
);
663 else if (!split
[0].suspect
&& !split
[2].suspect
)
665 * The parent covers the entire area; reuse storage for
666 * e and replace it with the parent.
668 dup_entry(e
, &split
[1]);
669 else if (split
[0].suspect
) {
670 /* me and then parent */
671 dup_entry(e
, &split
[0]);
673 new_entry
= xmalloc(sizeof(*new_entry
));
674 memcpy(new_entry
, &(split
[1]), sizeof(struct blame_entry
));
675 add_blame_entry(sb
, new_entry
);
678 /* parent and then me */
679 dup_entry(e
, &split
[1]);
681 new_entry
= xmalloc(sizeof(*new_entry
));
682 memcpy(new_entry
, &(split
[2]), sizeof(struct blame_entry
));
683 add_blame_entry(sb
, new_entry
);
686 if (DEBUG
) { /* sanity */
687 struct blame_entry
*ent
;
688 int lno
= sb
->ent
->lno
, corrupt
= 0;
690 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
695 lno
+= ent
->num_lines
;
699 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
700 printf("L %8d l %8d n %8d\n",
701 lno
, ent
->lno
, ent
->num_lines
);
702 lno
= ent
->lno
+ ent
->num_lines
;
710 * After splitting the blame, the origins used by the
711 * on-stack blame_entry should lose one refcnt each.
713 static void decref_split(struct blame_entry
*split
)
717 for (i
= 0; i
< 3; i
++)
718 origin_decref(split
[i
].suspect
);
722 * Helper for blame_chunk(). blame_entry e is known to overlap with
723 * the patch hunk; split it and pass blame to the parent.
725 static void blame_overlap(struct scoreboard
*sb
, struct blame_entry
*e
,
726 int tlno
, int plno
, int same
,
727 struct origin
*parent
)
729 struct blame_entry split
[3];
731 split_overlap(split
, e
, tlno
, plno
, same
, parent
);
732 if (split
[1].suspect
)
733 split_blame(sb
, split
, e
);
738 * Find the line number of the last line the target is suspected for.
740 static int find_last_in_target(struct scoreboard
*sb
, struct origin
*target
)
742 struct blame_entry
*e
;
743 int last_in_target
= -1;
745 for (e
= sb
->ent
; e
; e
= e
->next
) {
746 if (e
->guilty
|| !same_suspect(e
->suspect
, target
))
748 if (last_in_target
< e
->s_lno
+ e
->num_lines
)
749 last_in_target
= e
->s_lno
+ e
->num_lines
;
751 return last_in_target
;
755 * Process one hunk from the patch between the current suspect for
756 * blame_entry e and its parent. Find and split the overlap, and
757 * pass blame to the overlapping part to the parent.
759 static void blame_chunk(struct scoreboard
*sb
,
760 int tlno
, int plno
, int same
,
761 struct origin
*target
, struct origin
*parent
)
763 struct blame_entry
*e
;
765 for (e
= sb
->ent
; e
; e
= e
->next
) {
766 if (e
->guilty
|| !same_suspect(e
->suspect
, target
))
768 if (same
<= e
->s_lno
)
770 if (tlno
< e
->s_lno
+ e
->num_lines
)
771 blame_overlap(sb
, e
, tlno
, plno
, same
, parent
);
775 struct blame_chunk_cb_data
{
776 struct scoreboard
*sb
;
777 struct origin
*target
;
778 struct origin
*parent
;
783 static int blame_chunk_cb(long start_a
, long count_a
,
784 long start_b
, long count_b
, void *data
)
786 struct blame_chunk_cb_data
*d
= data
;
787 blame_chunk(d
->sb
, d
->tlno
, d
->plno
, start_b
, d
->target
, d
->parent
);
788 d
->plno
= start_a
+ count_a
;
789 d
->tlno
= start_b
+ count_b
;
794 * We are looking at the origin 'target' and aiming to pass blame
795 * for the lines it is suspected to its parent. Run diff to find
796 * which lines came from parent and pass blame for them.
798 static int pass_blame_to_parent(struct scoreboard
*sb
,
799 struct origin
*target
,
800 struct origin
*parent
)
803 mmfile_t file_p
, file_o
;
804 struct blame_chunk_cb_data d
;
806 memset(&d
, 0, sizeof(d
));
807 d
.sb
= sb
; d
.target
= target
; d
.parent
= parent
;
808 last_in_target
= find_last_in_target(sb
, target
);
809 if (last_in_target
< 0)
810 return 1; /* nothing remains for this target */
812 fill_origin_blob(&sb
->revs
->diffopt
, parent
, &file_p
);
813 fill_origin_blob(&sb
->revs
->diffopt
, target
, &file_o
);
816 diff_hunks(&file_p
, &file_o
, 0, blame_chunk_cb
, &d
);
817 /* The rest (i.e. anything after tlno) are the same as the parent */
818 blame_chunk(sb
, d
.tlno
, d
.plno
, last_in_target
, target
, parent
);
824 * The lines in blame_entry after splitting blames many times can become
825 * very small and trivial, and at some point it becomes pointless to
826 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
827 * ordinary C program, and it is not worth to say it was copied from
828 * totally unrelated file in the parent.
830 * Compute how trivial the lines in the blame_entry are.
832 static unsigned ent_score(struct scoreboard
*sb
, struct blame_entry
*e
)
841 cp
= nth_line(sb
, e
->lno
);
842 ep
= nth_line(sb
, e
->lno
+ e
->num_lines
);
844 unsigned ch
= *((unsigned char *)cp
);
854 * best_so_far[] and this[] are both a split of an existing blame_entry
855 * that passes blame to the parent. Maintain best_so_far the best split
856 * so far, by comparing this and best_so_far and copying this into
857 * bst_so_far as needed.
859 static void copy_split_if_better(struct scoreboard
*sb
,
860 struct blame_entry
*best_so_far
,
861 struct blame_entry
*this)
865 if (!this[1].suspect
)
867 if (best_so_far
[1].suspect
) {
868 if (ent_score(sb
, &this[1]) < ent_score(sb
, &best_so_far
[1]))
872 for (i
= 0; i
< 3; i
++)
873 origin_incref(this[i
].suspect
);
874 decref_split(best_so_far
);
875 memcpy(best_so_far
, this, sizeof(struct blame_entry
[3]));
879 * We are looking at a part of the final image represented by
880 * ent (tlno and same are offset by ent->s_lno).
881 * tlno is where we are looking at in the final image.
882 * up to (but not including) same match preimage.
883 * plno is where we are looking at in the preimage.
885 * <-------------- final image ---------------------->
888 * <---------preimage----->
891 * All line numbers are 0-based.
893 static void handle_split(struct scoreboard
*sb
,
894 struct blame_entry
*ent
,
895 int tlno
, int plno
, int same
,
896 struct origin
*parent
,
897 struct blame_entry
*split
)
899 if (ent
->num_lines
<= tlno
)
902 struct blame_entry
this[3];
905 split_overlap(this, ent
, tlno
, plno
, same
, parent
);
906 copy_split_if_better(sb
, split
, this);
911 struct handle_split_cb_data
{
912 struct scoreboard
*sb
;
913 struct blame_entry
*ent
;
914 struct origin
*parent
;
915 struct blame_entry
*split
;
920 static int handle_split_cb(long start_a
, long count_a
,
921 long start_b
, long count_b
, void *data
)
923 struct handle_split_cb_data
*d
= data
;
924 handle_split(d
->sb
, d
->ent
, d
->tlno
, d
->plno
, start_b
, d
->parent
,
926 d
->plno
= start_a
+ count_a
;
927 d
->tlno
= start_b
+ count_b
;
932 * Find the lines from parent that are the same as ent so that
933 * we can pass blames to it. file_p has the blob contents for
936 static void find_copy_in_blob(struct scoreboard
*sb
,
937 struct blame_entry
*ent
,
938 struct origin
*parent
,
939 struct blame_entry
*split
,
945 struct handle_split_cb_data d
;
947 memset(&d
, 0, sizeof(d
));
948 d
.sb
= sb
; d
.ent
= ent
; d
.parent
= parent
; d
.split
= split
;
950 * Prepare mmfile that contains only the lines in ent.
952 cp
= nth_line(sb
, ent
->lno
);
953 file_o
.ptr
= (char *) cp
;
954 cnt
= ent
->num_lines
;
956 while (cnt
&& cp
< sb
->final_buf
+ sb
->final_buf_size
) {
960 file_o
.size
= cp
- file_o
.ptr
;
963 * file_o is a part of final image we are annotating.
964 * file_p partially may match that image.
966 memset(split
, 0, sizeof(struct blame_entry
[3]));
967 diff_hunks(file_p
, &file_o
, 1, handle_split_cb
, &d
);
968 /* remainder, if any, all match the preimage */
969 handle_split(sb
, ent
, d
.tlno
, d
.plno
, ent
->num_lines
, parent
, split
);
973 * See if lines currently target is suspected for can be attributed to
976 static int find_move_in_parent(struct scoreboard
*sb
,
977 struct origin
*target
,
978 struct origin
*parent
)
980 int last_in_target
, made_progress
;
981 struct blame_entry
*e
, split
[3];
984 last_in_target
= find_last_in_target(sb
, target
);
985 if (last_in_target
< 0)
986 return 1; /* nothing remains for this target */
988 fill_origin_blob(&sb
->revs
->diffopt
, parent
, &file_p
);
993 while (made_progress
) {
995 for (e
= sb
->ent
; e
; e
= e
->next
) {
996 if (e
->guilty
|| !same_suspect(e
->suspect
, target
) ||
997 ent_score(sb
, e
) < blame_move_score
)
999 find_copy_in_blob(sb
, e
, parent
, split
, &file_p
);
1000 if (split
[1].suspect
&&
1001 blame_move_score
< ent_score(sb
, &split
[1])) {
1002 split_blame(sb
, split
, e
);
1005 decref_split(split
);
1012 struct blame_entry
*ent
;
1013 struct blame_entry split
[3];
1017 * Count the number of entries the target is suspected for,
1018 * and prepare a list of entry and the best split.
1020 static struct blame_list
*setup_blame_list(struct scoreboard
*sb
,
1021 struct origin
*target
,
1025 struct blame_entry
*e
;
1027 struct blame_list
*blame_list
= NULL
;
1029 for (e
= sb
->ent
, num_ents
= 0; e
; e
= e
->next
)
1030 if (!e
->scanned
&& !e
->guilty
&&
1031 same_suspect(e
->suspect
, target
) &&
1032 min_score
< ent_score(sb
, e
))
1035 blame_list
= xcalloc(num_ents
, sizeof(struct blame_list
));
1036 for (e
= sb
->ent
, i
= 0; e
; e
= e
->next
)
1037 if (!e
->scanned
&& !e
->guilty
&&
1038 same_suspect(e
->suspect
, target
) &&
1039 min_score
< ent_score(sb
, e
))
1040 blame_list
[i
++].ent
= e
;
1042 *num_ents_p
= num_ents
;
1047 * Reset the scanned status on all entries.
1049 static void reset_scanned_flag(struct scoreboard
*sb
)
1051 struct blame_entry
*e
;
1052 for (e
= sb
->ent
; e
; e
= e
->next
)
1057 * For lines target is suspected for, see if we can find code movement
1058 * across file boundary from the parent commit. porigin is the path
1059 * in the parent we already tried.
1061 static int find_copy_in_parent(struct scoreboard
*sb
,
1062 struct origin
*target
,
1063 struct commit
*parent
,
1064 struct origin
*porigin
,
1067 struct diff_options diff_opts
;
1068 const char *paths
[1];
1071 struct blame_list
*blame_list
;
1074 blame_list
= setup_blame_list(sb
, target
, blame_copy_score
, &num_ents
);
1076 return 1; /* nothing remains for this target */
1078 diff_setup(&diff_opts
);
1079 DIFF_OPT_SET(&diff_opts
, RECURSIVE
);
1080 diff_opts
.output_format
= DIFF_FORMAT_NO_OUTPUT
;
1083 diff_tree_setup_paths(paths
, &diff_opts
);
1084 diff_setup_done(&diff_opts
);
1086 /* Try "find copies harder" on new path if requested;
1087 * we do not want to use diffcore_rename() actually to
1088 * match things up; find_copies_harder is set only to
1089 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1090 * and this code needs to be after diff_setup_done(), which
1091 * usually makes find-copies-harder imply copy detection.
1093 if ((opt
& PICKAXE_BLAME_COPY_HARDEST
)
1094 || ((opt
& PICKAXE_BLAME_COPY_HARDER
)
1095 && (!porigin
|| strcmp(target
->path
, porigin
->path
))))
1096 DIFF_OPT_SET(&diff_opts
, FIND_COPIES_HARDER
);
1098 if (is_null_sha1(target
->commit
->object
.sha1
))
1099 do_diff_cache(parent
->tree
->object
.sha1
, &diff_opts
);
1101 diff_tree_sha1(parent
->tree
->object
.sha1
,
1102 target
->commit
->tree
->object
.sha1
,
1105 if (!DIFF_OPT_TST(&diff_opts
, FIND_COPIES_HARDER
))
1106 diffcore_std(&diff_opts
);
1110 int made_progress
= 0;
1112 for (i
= 0; i
< diff_queued_diff
.nr
; i
++) {
1113 struct diff_filepair
*p
= diff_queued_diff
.queue
[i
];
1114 struct origin
*norigin
;
1116 struct blame_entry
this[3];
1118 if (!DIFF_FILE_VALID(p
->one
))
1119 continue; /* does not exist in parent */
1120 if (S_ISGITLINK(p
->one
->mode
))
1121 continue; /* ignore git links */
1122 if (porigin
&& !strcmp(p
->one
->path
, porigin
->path
))
1123 /* find_move already dealt with this path */
1126 norigin
= get_origin(sb
, parent
, p
->one
->path
);
1127 hashcpy(norigin
->blob_sha1
, p
->one
->sha1
);
1128 norigin
->mode
= p
->one
->mode
;
1129 fill_origin_blob(&sb
->revs
->diffopt
, norigin
, &file_p
);
1133 for (j
= 0; j
< num_ents
; j
++) {
1134 find_copy_in_blob(sb
, blame_list
[j
].ent
,
1135 norigin
, this, &file_p
);
1136 copy_split_if_better(sb
, blame_list
[j
].split
,
1140 origin_decref(norigin
);
1143 for (j
= 0; j
< num_ents
; j
++) {
1144 struct blame_entry
*split
= blame_list
[j
].split
;
1145 if (split
[1].suspect
&&
1146 blame_copy_score
< ent_score(sb
, &split
[1])) {
1147 split_blame(sb
, split
, blame_list
[j
].ent
);
1151 blame_list
[j
].ent
->scanned
= 1;
1152 decref_split(split
);
1158 blame_list
= setup_blame_list(sb
, target
, blame_copy_score
, &num_ents
);
1164 reset_scanned_flag(sb
);
1165 diff_flush(&diff_opts
);
1166 diff_tree_release_paths(&diff_opts
);
1171 * The blobs of origin and porigin exactly match, so everything
1172 * origin is suspected for can be blamed on the parent.
1174 static void pass_whole_blame(struct scoreboard
*sb
,
1175 struct origin
*origin
, struct origin
*porigin
)
1177 struct blame_entry
*e
;
1179 if (!porigin
->file
.ptr
&& origin
->file
.ptr
) {
1180 /* Steal its file */
1181 porigin
->file
= origin
->file
;
1182 origin
->file
.ptr
= NULL
;
1184 for (e
= sb
->ent
; e
; e
= e
->next
) {
1185 if (!same_suspect(e
->suspect
, origin
))
1187 origin_incref(porigin
);
1188 origin_decref(e
->suspect
);
1189 e
->suspect
= porigin
;
1194 * We pass blame from the current commit to its parents. We keep saying
1195 * "parent" (and "porigin"), but what we mean is to find scapegoat to
1196 * exonerate ourselves.
1198 static struct commit_list
*first_scapegoat(struct rev_info
*revs
, struct commit
*commit
)
1201 return commit
->parents
;
1202 return lookup_decoration(&revs
->children
, &commit
->object
);
1205 static int num_scapegoats(struct rev_info
*revs
, struct commit
*commit
)
1208 struct commit_list
*l
= first_scapegoat(revs
, commit
);
1209 for (cnt
= 0; l
; l
= l
->next
)
1216 static void pass_blame(struct scoreboard
*sb
, struct origin
*origin
, int opt
)
1218 struct rev_info
*revs
= sb
->revs
;
1219 int i
, pass
, num_sg
;
1220 struct commit
*commit
= origin
->commit
;
1221 struct commit_list
*sg
;
1222 struct origin
*sg_buf
[MAXSG
];
1223 struct origin
*porigin
, **sg_origin
= sg_buf
;
1225 num_sg
= num_scapegoats(revs
, commit
);
1228 else if (num_sg
< ARRAY_SIZE(sg_buf
))
1229 memset(sg_buf
, 0, sizeof(sg_buf
));
1231 sg_origin
= xcalloc(num_sg
, sizeof(*sg_origin
));
1234 * The first pass looks for unrenamed path to optimize for
1235 * common cases, then we look for renames in the second pass.
1237 for (pass
= 0; pass
< 2 - no_whole_file_rename
; pass
++) {
1238 struct origin
*(*find
)(struct scoreboard
*,
1239 struct commit
*, struct origin
*);
1240 find
= pass
? find_rename
: find_origin
;
1242 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1244 sg
= sg
->next
, i
++) {
1245 struct commit
*p
= sg
->item
;
1250 if (parse_commit(p
))
1252 porigin
= find(sb
, p
, origin
);
1255 if (!hashcmp(porigin
->blob_sha1
, origin
->blob_sha1
)) {
1256 pass_whole_blame(sb
, origin
, porigin
);
1257 origin_decref(porigin
);
1260 for (j
= same
= 0; j
< i
; j
++)
1262 !hashcmp(sg_origin
[j
]->blob_sha1
,
1263 porigin
->blob_sha1
)) {
1268 sg_origin
[i
] = porigin
;
1270 origin_decref(porigin
);
1275 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1277 sg
= sg
->next
, i
++) {
1278 struct origin
*porigin
= sg_origin
[i
];
1281 if (!origin
->previous
) {
1282 origin_incref(porigin
);
1283 origin
->previous
= porigin
;
1285 if (pass_blame_to_parent(sb
, origin
, porigin
))
1290 * Optionally find moves in parents' files.
1292 if (opt
& PICKAXE_BLAME_MOVE
)
1293 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1295 sg
= sg
->next
, i
++) {
1296 struct origin
*porigin
= sg_origin
[i
];
1299 if (find_move_in_parent(sb
, origin
, porigin
))
1304 * Optionally find copies from parents' files.
1306 if (opt
& PICKAXE_BLAME_COPY
)
1307 for (i
= 0, sg
= first_scapegoat(revs
, commit
);
1309 sg
= sg
->next
, i
++) {
1310 struct origin
*porigin
= sg_origin
[i
];
1311 if (find_copy_in_parent(sb
, origin
, sg
->item
,
1317 for (i
= 0; i
< num_sg
; i
++) {
1319 drop_origin_blob(sg_origin
[i
]);
1320 origin_decref(sg_origin
[i
]);
1323 drop_origin_blob(origin
);
1324 if (sg_buf
!= sg_origin
)
1329 * Information on commits, used for output.
1331 struct commit_info
{
1332 struct strbuf author
;
1333 struct strbuf author_mail
;
1334 unsigned long author_time
;
1335 struct strbuf author_tz
;
1337 /* filled only when asked for details */
1338 struct strbuf committer
;
1339 struct strbuf committer_mail
;
1340 unsigned long committer_time
;
1341 struct strbuf committer_tz
;
1343 struct strbuf summary
;
1347 * Parse author/committer line in the commit object buffer
1349 static void get_ac_line(const char *inbuf
, const char *what
,
1350 struct strbuf
*name
, struct strbuf
*mail
,
1351 unsigned long *time
, struct strbuf
*tz
)
1353 struct ident_split ident
;
1354 size_t len
, maillen
, namelen
;
1356 const char *namebuf
, *mailbuf
;
1358 tmp
= strstr(inbuf
, what
);
1361 tmp
+= strlen(what
);
1362 endp
= strchr(tmp
, '\n');
1368 if (split_ident_line(&ident
, tmp
, len
)) {
1372 strbuf_addstr(name
, tmp
);
1373 strbuf_addstr(mail
, tmp
);
1374 strbuf_addstr(tz
, tmp
);
1379 namelen
= ident
.name_end
- ident
.name_begin
;
1380 namebuf
= ident
.name_begin
;
1382 maillen
= ident
.mail_end
- ident
.mail_begin
;
1383 mailbuf
= ident
.mail_begin
;
1385 if (ident
.date_begin
&& ident
.date_end
)
1386 *time
= strtoul(ident
.date_begin
, NULL
, 10);
1390 if (ident
.tz_begin
&& ident
.tz_end
)
1391 strbuf_add(tz
, ident
.tz_begin
, ident
.tz_end
- ident
.tz_begin
);
1393 strbuf_addstr(tz
, "(unknown)");
1396 * Now, convert both name and e-mail using mailmap
1398 map_user(&mailmap
, &mailbuf
, &maillen
,
1399 &namebuf
, &namelen
);
1401 strbuf_addf(mail
, "<%.*s>", (int)maillen
, mailbuf
);
1402 strbuf_add(name
, namebuf
, namelen
);
1405 static void commit_info_init(struct commit_info
*ci
)
1408 strbuf_init(&ci
->author
, 0);
1409 strbuf_init(&ci
->author_mail
, 0);
1410 strbuf_init(&ci
->author_tz
, 0);
1411 strbuf_init(&ci
->committer
, 0);
1412 strbuf_init(&ci
->committer_mail
, 0);
1413 strbuf_init(&ci
->committer_tz
, 0);
1414 strbuf_init(&ci
->summary
, 0);
1417 static void commit_info_destroy(struct commit_info
*ci
)
1420 strbuf_release(&ci
->author
);
1421 strbuf_release(&ci
->author_mail
);
1422 strbuf_release(&ci
->author_tz
);
1423 strbuf_release(&ci
->committer
);
1424 strbuf_release(&ci
->committer_mail
);
1425 strbuf_release(&ci
->committer_tz
);
1426 strbuf_release(&ci
->summary
);
1429 static void get_commit_info(struct commit
*commit
,
1430 struct commit_info
*ret
,
1434 const char *subject
, *encoding
;
1437 commit_info_init(ret
);
1439 encoding
= get_log_output_encoding();
1440 message
= logmsg_reencode(commit
, NULL
, encoding
);
1441 get_ac_line(message
, "\nauthor ",
1442 &ret
->author
, &ret
->author_mail
,
1443 &ret
->author_time
, &ret
->author_tz
);
1446 logmsg_free(message
, commit
);
1450 get_ac_line(message
, "\ncommitter ",
1451 &ret
->committer
, &ret
->committer_mail
,
1452 &ret
->committer_time
, &ret
->committer_tz
);
1454 len
= find_commit_subject(message
, &subject
);
1456 strbuf_add(&ret
->summary
, subject
, len
);
1458 strbuf_addf(&ret
->summary
, "(%s)", sha1_to_hex(commit
->object
.sha1
));
1460 logmsg_free(message
, commit
);
1464 * To allow LF and other nonportable characters in pathnames,
1465 * they are c-style quoted as needed.
1467 static void write_filename_info(const char *path
)
1469 printf("filename ");
1470 write_name_quoted(path
, stdout
, '\n');
1474 * Porcelain/Incremental format wants to show a lot of details per
1475 * commit. Instead of repeating this every line, emit it only once,
1476 * the first time each commit appears in the output (unless the
1477 * user has specifically asked for us to repeat).
1479 static int emit_one_suspect_detail(struct origin
*suspect
, int repeat
)
1481 struct commit_info ci
;
1483 if (!repeat
&& (suspect
->commit
->object
.flags
& METAINFO_SHOWN
))
1486 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
1487 get_commit_info(suspect
->commit
, &ci
, 1);
1488 printf("author %s\n", ci
.author
.buf
);
1489 printf("author-mail %s\n", ci
.author_mail
.buf
);
1490 printf("author-time %lu\n", ci
.author_time
);
1491 printf("author-tz %s\n", ci
.author_tz
.buf
);
1492 printf("committer %s\n", ci
.committer
.buf
);
1493 printf("committer-mail %s\n", ci
.committer_mail
.buf
);
1494 printf("committer-time %lu\n", ci
.committer_time
);
1495 printf("committer-tz %s\n", ci
.committer_tz
.buf
);
1496 printf("summary %s\n", ci
.summary
.buf
);
1497 if (suspect
->commit
->object
.flags
& UNINTERESTING
)
1498 printf("boundary\n");
1499 if (suspect
->previous
) {
1500 struct origin
*prev
= suspect
->previous
;
1501 printf("previous %s ", sha1_to_hex(prev
->commit
->object
.sha1
));
1502 write_name_quoted(prev
->path
, stdout
, '\n');
1505 commit_info_destroy(&ci
);
1511 * The blame_entry is found to be guilty for the range. Mark it
1512 * as such, and show it in incremental output.
1514 static void found_guilty_entry(struct blame_entry
*ent
)
1520 struct origin
*suspect
= ent
->suspect
;
1522 printf("%s %d %d %d\n",
1523 sha1_to_hex(suspect
->commit
->object
.sha1
),
1524 ent
->s_lno
+ 1, ent
->lno
+ 1, ent
->num_lines
);
1525 emit_one_suspect_detail(suspect
, 0);
1526 write_filename_info(suspect
->path
);
1527 maybe_flush_or_die(stdout
, "stdout");
1532 * The main loop -- while the scoreboard has lines whose true origin
1533 * is still unknown, pick one blame_entry, and allow its current
1534 * suspect to pass blames to its parents.
1536 static void assign_blame(struct scoreboard
*sb
, int opt
)
1538 struct rev_info
*revs
= sb
->revs
;
1541 struct blame_entry
*ent
;
1542 struct commit
*commit
;
1543 struct origin
*suspect
= NULL
;
1545 /* find one suspect to break down */
1546 for (ent
= sb
->ent
; !suspect
&& ent
; ent
= ent
->next
)
1548 suspect
= ent
->suspect
;
1550 return; /* all done */
1553 * We will use this suspect later in the loop,
1554 * so hold onto it in the meantime.
1556 origin_incref(suspect
);
1557 commit
= suspect
->commit
;
1558 if (!commit
->object
.parsed
)
1559 parse_commit(commit
);
1561 (!(commit
->object
.flags
& UNINTERESTING
) &&
1562 !(revs
->max_age
!= -1 && commit
->date
< revs
->max_age
)))
1563 pass_blame(sb
, suspect
, opt
);
1565 commit
->object
.flags
|= UNINTERESTING
;
1566 if (commit
->object
.parsed
)
1567 mark_parents_uninteresting(commit
);
1569 /* treat root commit as boundary */
1570 if (!commit
->parents
&& !show_root
)
1571 commit
->object
.flags
|= UNINTERESTING
;
1573 /* Take responsibility for the remaining entries */
1574 for (ent
= sb
->ent
; ent
; ent
= ent
->next
)
1575 if (same_suspect(ent
->suspect
, suspect
))
1576 found_guilty_entry(ent
);
1577 origin_decref(suspect
);
1579 if (DEBUG
) /* sanity */
1580 sanity_check_refcnt(sb
);
1584 static const char *format_time(unsigned long time
, const char *tz_str
,
1587 static char time_buf
[128];
1588 const char *time_str
;
1592 if (show_raw_time
) {
1593 snprintf(time_buf
, sizeof(time_buf
), "%lu %s", time
, tz_str
);
1597 time_str
= show_date(time
, tz
, blame_date_mode
);
1598 time_len
= strlen(time_str
);
1599 memcpy(time_buf
, time_str
, time_len
);
1600 memset(time_buf
+ time_len
, ' ', blame_date_width
- time_len
);
1605 #define OUTPUT_ANNOTATE_COMPAT 001
1606 #define OUTPUT_LONG_OBJECT_NAME 002
1607 #define OUTPUT_RAW_TIMESTAMP 004
1608 #define OUTPUT_PORCELAIN 010
1609 #define OUTPUT_SHOW_NAME 020
1610 #define OUTPUT_SHOW_NUMBER 040
1611 #define OUTPUT_SHOW_SCORE 0100
1612 #define OUTPUT_NO_AUTHOR 0200
1613 #define OUTPUT_SHOW_EMAIL 0400
1614 #define OUTPUT_LINE_PORCELAIN 01000
1616 static void emit_porcelain_details(struct origin
*suspect
, int repeat
)
1618 if (emit_one_suspect_detail(suspect
, repeat
) ||
1619 (suspect
->commit
->object
.flags
& MORE_THAN_ONE_PATH
))
1620 write_filename_info(suspect
->path
);
1623 static void emit_porcelain(struct scoreboard
*sb
, struct blame_entry
*ent
,
1626 int repeat
= opt
& OUTPUT_LINE_PORCELAIN
;
1629 struct origin
*suspect
= ent
->suspect
;
1632 strcpy(hex
, sha1_to_hex(suspect
->commit
->object
.sha1
));
1633 printf("%s%c%d %d %d\n",
1635 ent
->guilty
? ' ' : '*', /* purely for debugging */
1639 emit_porcelain_details(suspect
, repeat
);
1641 cp
= nth_line(sb
, ent
->lno
);
1642 for (cnt
= 0; cnt
< ent
->num_lines
; cnt
++) {
1645 printf("%s %d %d\n", hex
,
1646 ent
->s_lno
+ 1 + cnt
,
1647 ent
->lno
+ 1 + cnt
);
1649 emit_porcelain_details(suspect
, 1);
1655 } while (ch
!= '\n' &&
1656 cp
< sb
->final_buf
+ sb
->final_buf_size
);
1659 if (sb
->final_buf_size
&& cp
[-1] != '\n')
1663 static void emit_other(struct scoreboard
*sb
, struct blame_entry
*ent
, int opt
)
1667 struct origin
*suspect
= ent
->suspect
;
1668 struct commit_info ci
;
1670 int show_raw_time
= !!(opt
& OUTPUT_RAW_TIMESTAMP
);
1672 get_commit_info(suspect
->commit
, &ci
, 1);
1673 strcpy(hex
, sha1_to_hex(suspect
->commit
->object
.sha1
));
1675 cp
= nth_line(sb
, ent
->lno
);
1676 for (cnt
= 0; cnt
< ent
->num_lines
; cnt
++) {
1678 int length
= (opt
& OUTPUT_LONG_OBJECT_NAME
) ? 40 : abbrev
;
1680 if (suspect
->commit
->object
.flags
& UNINTERESTING
) {
1682 memset(hex
, ' ', length
);
1683 else if (!(opt
& OUTPUT_ANNOTATE_COMPAT
)) {
1689 printf("%.*s", length
, hex
);
1690 if (opt
& OUTPUT_ANNOTATE_COMPAT
) {
1692 if (opt
& OUTPUT_SHOW_EMAIL
)
1693 name
= ci
.author_mail
.buf
;
1695 name
= ci
.author
.buf
;
1696 printf("\t(%10s\t%10s\t%d)", name
,
1697 format_time(ci
.author_time
, ci
.author_tz
.buf
,
1699 ent
->lno
+ 1 + cnt
);
1701 if (opt
& OUTPUT_SHOW_SCORE
)
1703 max_score_digits
, ent
->score
,
1704 ent
->suspect
->refcnt
);
1705 if (opt
& OUTPUT_SHOW_NAME
)
1706 printf(" %-*.*s", longest_file
, longest_file
,
1708 if (opt
& OUTPUT_SHOW_NUMBER
)
1709 printf(" %*d", max_orig_digits
,
1710 ent
->s_lno
+ 1 + cnt
);
1712 if (!(opt
& OUTPUT_NO_AUTHOR
)) {
1715 if (opt
& OUTPUT_SHOW_EMAIL
)
1716 name
= ci
.author_mail
.buf
;
1718 name
= ci
.author
.buf
;
1719 pad
= longest_author
- utf8_strwidth(name
);
1720 printf(" (%s%*s %10s",
1722 format_time(ci
.author_time
,
1727 max_digits
, ent
->lno
+ 1 + cnt
);
1732 } while (ch
!= '\n' &&
1733 cp
< sb
->final_buf
+ sb
->final_buf_size
);
1736 if (sb
->final_buf_size
&& cp
[-1] != '\n')
1739 commit_info_destroy(&ci
);
1742 static void output(struct scoreboard
*sb
, int option
)
1744 struct blame_entry
*ent
;
1746 if (option
& OUTPUT_PORCELAIN
) {
1747 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1748 struct blame_entry
*oth
;
1749 struct origin
*suspect
= ent
->suspect
;
1750 struct commit
*commit
= suspect
->commit
;
1751 if (commit
->object
.flags
& MORE_THAN_ONE_PATH
)
1753 for (oth
= ent
->next
; oth
; oth
= oth
->next
) {
1754 if ((oth
->suspect
->commit
!= commit
) ||
1755 !strcmp(oth
->suspect
->path
, suspect
->path
))
1757 commit
->object
.flags
|= MORE_THAN_ONE_PATH
;
1763 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1764 if (option
& OUTPUT_PORCELAIN
)
1765 emit_porcelain(sb
, ent
, option
);
1767 emit_other(sb
, ent
, option
);
1773 * To allow quick access to the contents of nth line in the
1774 * final image, prepare an index in the scoreboard.
1776 static int prepare_lines(struct scoreboard
*sb
)
1778 const char *buf
= sb
->final_buf
;
1779 unsigned long len
= sb
->final_buf_size
;
1780 int num
= 0, incomplete
= 0, bol
= 1;
1782 if (len
&& buf
[len
-1] != '\n')
1783 incomplete
++; /* incomplete line at the end */
1786 sb
->lineno
= xrealloc(sb
->lineno
,
1787 sizeof(int *) * (num
+ 1));
1788 sb
->lineno
[num
] = buf
- sb
->final_buf
;
1791 if (*buf
++ == '\n') {
1796 sb
->lineno
= xrealloc(sb
->lineno
,
1797 sizeof(int *) * (num
+ incomplete
+ 1));
1798 sb
->lineno
[num
+ incomplete
] = buf
- sb
->final_buf
;
1799 sb
->num_lines
= num
+ incomplete
;
1800 return sb
->num_lines
;
1804 * Add phony grafts for use with -S; this is primarily to
1805 * support git's cvsserver that wants to give a linear history
1808 static int read_ancestry(const char *graft_file
)
1810 FILE *fp
= fopen(graft_file
, "r");
1814 while (fgets(buf
, sizeof(buf
), fp
)) {
1815 /* The format is just "Commit Parent1 Parent2 ...\n" */
1816 int len
= strlen(buf
);
1817 struct commit_graft
*graft
= read_graft_line(buf
, len
);
1819 register_commit_graft(graft
, 0);
1825 static int update_auto_abbrev(int auto_abbrev
, struct origin
*suspect
)
1827 const char *uniq
= find_unique_abbrev(suspect
->commit
->object
.sha1
,
1829 int len
= strlen(uniq
);
1830 if (auto_abbrev
< len
)
1836 * How many columns do we need to show line numbers, authors,
1839 static void find_alignment(struct scoreboard
*sb
, int *option
)
1841 int longest_src_lines
= 0;
1842 int longest_dst_lines
= 0;
1843 unsigned largest_score
= 0;
1844 struct blame_entry
*e
;
1845 int compute_auto_abbrev
= (abbrev
< 0);
1846 int auto_abbrev
= default_abbrev
;
1848 for (e
= sb
->ent
; e
; e
= e
->next
) {
1849 struct origin
*suspect
= e
->suspect
;
1850 struct commit_info ci
;
1853 if (compute_auto_abbrev
)
1854 auto_abbrev
= update_auto_abbrev(auto_abbrev
, suspect
);
1855 if (strcmp(suspect
->path
, sb
->path
))
1856 *option
|= OUTPUT_SHOW_NAME
;
1857 num
= strlen(suspect
->path
);
1858 if (longest_file
< num
)
1860 if (!(suspect
->commit
->object
.flags
& METAINFO_SHOWN
)) {
1861 suspect
->commit
->object
.flags
|= METAINFO_SHOWN
;
1862 get_commit_info(suspect
->commit
, &ci
, 1);
1863 if (*option
& OUTPUT_SHOW_EMAIL
)
1864 num
= utf8_strwidth(ci
.author_mail
.buf
);
1866 num
= utf8_strwidth(ci
.author
.buf
);
1867 if (longest_author
< num
)
1868 longest_author
= num
;
1870 num
= e
->s_lno
+ e
->num_lines
;
1871 if (longest_src_lines
< num
)
1872 longest_src_lines
= num
;
1873 num
= e
->lno
+ e
->num_lines
;
1874 if (longest_dst_lines
< num
)
1875 longest_dst_lines
= num
;
1876 if (largest_score
< ent_score(sb
, e
))
1877 largest_score
= ent_score(sb
, e
);
1879 commit_info_destroy(&ci
);
1881 max_orig_digits
= decimal_width(longest_src_lines
);
1882 max_digits
= decimal_width(longest_dst_lines
);
1883 max_score_digits
= decimal_width(largest_score
);
1885 if (compute_auto_abbrev
)
1886 /* one more abbrev length is needed for the boundary commit */
1887 abbrev
= auto_abbrev
+ 1;
1891 * For debugging -- origin is refcounted, and this asserts that
1892 * we do not underflow.
1894 static void sanity_check_refcnt(struct scoreboard
*sb
)
1897 struct blame_entry
*ent
;
1899 for (ent
= sb
->ent
; ent
; ent
= ent
->next
) {
1900 /* Nobody should have zero or negative refcnt */
1901 if (ent
->suspect
->refcnt
<= 0) {
1902 fprintf(stderr
, "%s in %s has negative refcnt %d\n",
1904 sha1_to_hex(ent
->suspect
->commit
->object
.sha1
),
1905 ent
->suspect
->refcnt
);
1911 find_alignment(sb
, &opt
);
1913 die("Baa %d!", baa
);
1918 * Used for the command line parsing; check if the path exists
1919 * in the working tree.
1921 static int has_string_in_work_tree(const char *path
)
1924 return !lstat(path
, &st
);
1927 static unsigned parse_score(const char *arg
)
1930 unsigned long score
= strtoul(arg
, &end
, 10);
1936 static const char *add_prefix(const char *prefix
, const char *path
)
1938 return prefix_path(prefix
, prefix
? strlen(prefix
) : 0, path
);
1941 static int git_blame_config(const char *var
, const char *value
, void *cb
)
1943 if (!strcmp(var
, "blame.showroot")) {
1944 show_root
= git_config_bool(var
, value
);
1947 if (!strcmp(var
, "blame.blankboundary")) {
1948 blank_boundary
= git_config_bool(var
, value
);
1951 if (!strcmp(var
, "blame.date")) {
1953 return config_error_nonbool(var
);
1954 blame_date_mode
= parse_date_format(value
);
1958 if (userdiff_config(var
, value
) < 0)
1961 return git_default_config(var
, value
, cb
);
1964 static void verify_working_tree_path(struct commit
*work_tree
, const char *path
)
1966 struct commit_list
*parents
;
1968 for (parents
= work_tree
->parents
; parents
; parents
= parents
->next
) {
1969 const unsigned char *commit_sha1
= parents
->item
->object
.sha1
;
1970 unsigned char blob_sha1
[20];
1973 if (!get_tree_entry(commit_sha1
, path
, blob_sha1
, &mode
) &&
1974 sha1_object_info(blob_sha1
, NULL
) == OBJ_BLOB
)
1977 die("no such path '%s' in HEAD", path
);
1980 static struct commit_list
**append_parent(struct commit_list
**tail
, const unsigned char *sha1
)
1982 struct commit
*parent
;
1984 parent
= lookup_commit_reference(sha1
);
1986 die("no such commit %s", sha1_to_hex(sha1
));
1987 return &commit_list_insert(parent
, tail
)->next
;
1990 static void append_merge_parents(struct commit_list
**tail
)
1993 const char *merge_head_file
= git_path("MERGE_HEAD");
1994 struct strbuf line
= STRBUF_INIT
;
1996 merge_head
= open(merge_head_file
, O_RDONLY
);
1997 if (merge_head
< 0) {
1998 if (errno
== ENOENT
)
2000 die("cannot open '%s' for reading", merge_head_file
);
2003 while (!strbuf_getwholeline_fd(&line
, merge_head
, '\n')) {
2004 unsigned char sha1
[20];
2005 if (line
.len
< 40 || get_sha1_hex(line
.buf
, sha1
))
2006 die("unknown line in '%s': %s", merge_head_file
, line
.buf
);
2007 tail
= append_parent(tail
, sha1
);
2010 strbuf_release(&line
);
2014 * Prepare a dummy commit that represents the work tree (or staged) item.
2015 * Note that annotating work tree item never works in the reverse.
2017 static struct commit
*fake_working_tree_commit(struct diff_options
*opt
,
2019 const char *contents_from
)
2021 struct commit
*commit
;
2022 struct origin
*origin
;
2023 struct commit_list
**parent_tail
, *parent
;
2024 unsigned char head_sha1
[20];
2025 struct strbuf buf
= STRBUF_INIT
;
2029 struct cache_entry
*ce
;
2031 struct strbuf msg
= STRBUF_INIT
;
2034 commit
= xcalloc(1, sizeof(*commit
));
2035 commit
->object
.parsed
= 1;
2037 commit
->object
.type
= OBJ_COMMIT
;
2038 parent_tail
= &commit
->parents
;
2040 if (!resolve_ref_unsafe("HEAD", head_sha1
, 1, NULL
))
2041 die("no such ref: HEAD");
2043 parent_tail
= append_parent(parent_tail
, head_sha1
);
2044 append_merge_parents(parent_tail
);
2045 verify_working_tree_path(commit
, path
);
2047 origin
= make_origin(commit
, path
);
2049 ident
= fmt_ident("Not Committed Yet", "not.committed.yet", NULL
, 0);
2050 strbuf_addstr(&msg
, "tree 0000000000000000000000000000000000000000\n");
2051 for (parent
= commit
->parents
; parent
; parent
= parent
->next
)
2052 strbuf_addf(&msg
, "parent %s\n",
2053 sha1_to_hex(parent
->item
->object
.sha1
));
2057 "Version of %s from %s\n",
2059 (!contents_from
? path
:
2060 (!strcmp(contents_from
, "-") ? "standard input" : contents_from
)));
2061 commit
->buffer
= strbuf_detach(&msg
, NULL
);
2063 if (!contents_from
|| strcmp("-", contents_from
)) {
2065 const char *read_from
;
2067 unsigned long buf_len
;
2069 if (contents_from
) {
2070 if (stat(contents_from
, &st
) < 0)
2071 die_errno("Cannot stat '%s'", contents_from
);
2072 read_from
= contents_from
;
2075 if (lstat(path
, &st
) < 0)
2076 die_errno("Cannot lstat '%s'", path
);
2079 mode
= canon_mode(st
.st_mode
);
2081 switch (st
.st_mode
& S_IFMT
) {
2083 if (DIFF_OPT_TST(opt
, ALLOW_TEXTCONV
) &&
2084 textconv_object(read_from
, mode
, null_sha1
, 0, &buf_ptr
, &buf_len
))
2085 strbuf_attach(&buf
, buf_ptr
, buf_len
, buf_len
+ 1);
2086 else if (strbuf_read_file(&buf
, read_from
, st
.st_size
) != st
.st_size
)
2087 die_errno("cannot open or read '%s'", read_from
);
2090 if (strbuf_readlink(&buf
, read_from
, st
.st_size
) < 0)
2091 die_errno("cannot readlink '%s'", read_from
);
2094 die("unsupported file type %s", read_from
);
2098 /* Reading from stdin */
2100 if (strbuf_read(&buf
, 0, 0) < 0)
2101 die_errno("failed to read from stdin");
2103 convert_to_git(path
, buf
.buf
, buf
.len
, &buf
, 0);
2104 origin
->file
.ptr
= buf
.buf
;
2105 origin
->file
.size
= buf
.len
;
2106 pretend_sha1_file(buf
.buf
, buf
.len
, OBJ_BLOB
, origin
->blob_sha1
);
2107 commit
->util
= origin
;
2110 * Read the current index, replace the path entry with
2111 * origin->blob_sha1 without mucking with its mode or type
2112 * bits; we are not going to write this index out -- we just
2113 * want to run "diff-index --cached".
2120 int pos
= cache_name_pos(path
, len
);
2122 mode
= active_cache
[pos
]->ce_mode
;
2124 /* Let's not bother reading from HEAD tree */
2125 mode
= S_IFREG
| 0644;
2127 size
= cache_entry_size(len
);
2128 ce
= xcalloc(1, size
);
2129 hashcpy(ce
->sha1
, origin
->blob_sha1
);
2130 memcpy(ce
->name
, path
, len
);
2131 ce
->ce_flags
= create_ce_flags(0);
2132 ce
->ce_namelen
= len
;
2133 ce
->ce_mode
= create_ce_mode(mode
);
2134 add_cache_entry(ce
, ADD_CACHE_OK_TO_ADD
|ADD_CACHE_OK_TO_REPLACE
);
2137 * We are not going to write this out, so this does not matter
2138 * right now, but someday we might optimize diff-index --cached
2139 * with cache-tree information.
2141 cache_tree_invalidate_path(active_cache_tree
, path
);
2146 static const char *prepare_final(struct scoreboard
*sb
)
2149 const char *final_commit_name
= NULL
;
2150 struct rev_info
*revs
= sb
->revs
;
2153 * There must be one and only one positive commit in the
2154 * revs->pending array.
2156 for (i
= 0; i
< revs
->pending
.nr
; i
++) {
2157 struct object
*obj
= revs
->pending
.objects
[i
].item
;
2158 if (obj
->flags
& UNINTERESTING
)
2160 while (obj
->type
== OBJ_TAG
)
2161 obj
= deref_tag(obj
, NULL
, 0);
2162 if (obj
->type
!= OBJ_COMMIT
)
2163 die("Non commit %s?", revs
->pending
.objects
[i
].name
);
2165 die("More than one commit to dig from %s and %s?",
2166 revs
->pending
.objects
[i
].name
,
2168 sb
->final
= (struct commit
*) obj
;
2169 final_commit_name
= revs
->pending
.objects
[i
].name
;
2171 return final_commit_name
;
2174 static const char *prepare_initial(struct scoreboard
*sb
)
2177 const char *final_commit_name
= NULL
;
2178 struct rev_info
*revs
= sb
->revs
;
2181 * There must be one and only one negative commit, and it must be
2184 for (i
= 0; i
< revs
->pending
.nr
; i
++) {
2185 struct object
*obj
= revs
->pending
.objects
[i
].item
;
2186 if (!(obj
->flags
& UNINTERESTING
))
2188 while (obj
->type
== OBJ_TAG
)
2189 obj
= deref_tag(obj
, NULL
, 0);
2190 if (obj
->type
!= OBJ_COMMIT
)
2191 die("Non commit %s?", revs
->pending
.objects
[i
].name
);
2193 die("More than one commit to dig down to %s and %s?",
2194 revs
->pending
.objects
[i
].name
,
2196 sb
->final
= (struct commit
*) obj
;
2197 final_commit_name
= revs
->pending
.objects
[i
].name
;
2199 if (!final_commit_name
)
2200 die("No commit to dig down to?");
2201 return final_commit_name
;
2204 static int blame_copy_callback(const struct option
*option
, const char *arg
, int unset
)
2206 int *opt
= option
->value
;
2209 * -C enables copy from removed files;
2210 * -C -C enables copy from existing files, but only
2211 * when blaming a new file;
2212 * -C -C -C enables copy from existing files for
2215 if (*opt
& PICKAXE_BLAME_COPY_HARDER
)
2216 *opt
|= PICKAXE_BLAME_COPY_HARDEST
;
2217 if (*opt
& PICKAXE_BLAME_COPY
)
2218 *opt
|= PICKAXE_BLAME_COPY_HARDER
;
2219 *opt
|= PICKAXE_BLAME_COPY
| PICKAXE_BLAME_MOVE
;
2222 blame_copy_score
= parse_score(arg
);
2226 static int blame_move_callback(const struct option
*option
, const char *arg
, int unset
)
2228 int *opt
= option
->value
;
2230 *opt
|= PICKAXE_BLAME_MOVE
;
2233 blame_move_score
= parse_score(arg
);
2237 int cmd_blame(int argc
, const char **argv
, const char *prefix
)
2239 struct rev_info revs
;
2241 struct scoreboard sb
;
2243 struct blame_entry
*ent
= NULL
;
2244 long dashdash_pos
, lno
;
2245 const char *final_commit_name
= NULL
;
2246 enum object_type type
;
2248 static struct string_list range_list
;
2249 static int output_option
= 0, opt
= 0;
2250 static int show_stats
= 0;
2251 static const char *revs_file
= NULL
;
2252 static const char *contents_from
= NULL
;
2253 static const struct option options
[] = {
2254 OPT_BOOLEAN(0, "incremental", &incremental
, N_("Show blame entries as we find them, incrementally")),
2255 OPT_BOOLEAN('b', NULL
, &blank_boundary
, N_("Show blank SHA-1 for boundary commits (Default: off)")),
2256 OPT_BOOLEAN(0, "root", &show_root
, N_("Do not treat root commits as boundaries (Default: off)")),
2257 OPT_BOOLEAN(0, "show-stats", &show_stats
, N_("Show work cost statistics")),
2258 OPT_BIT(0, "score-debug", &output_option
, N_("Show output score for blame entries"), OUTPUT_SHOW_SCORE
),
2259 OPT_BIT('f', "show-name", &output_option
, N_("Show original filename (Default: auto)"), OUTPUT_SHOW_NAME
),
2260 OPT_BIT('n', "show-number", &output_option
, N_("Show original linenumber (Default: off)"), OUTPUT_SHOW_NUMBER
),
2261 OPT_BIT('p', "porcelain", &output_option
, N_("Show in a format designed for machine consumption"), OUTPUT_PORCELAIN
),
2262 OPT_BIT(0, "line-porcelain", &output_option
, N_("Show porcelain format with per-line commit information"), OUTPUT_PORCELAIN
|OUTPUT_LINE_PORCELAIN
),
2263 OPT_BIT('c', NULL
, &output_option
, N_("Use the same output mode as git-annotate (Default: off)"), OUTPUT_ANNOTATE_COMPAT
),
2264 OPT_BIT('t', NULL
, &output_option
, N_("Show raw timestamp (Default: off)"), OUTPUT_RAW_TIMESTAMP
),
2265 OPT_BIT('l', NULL
, &output_option
, N_("Show long commit SHA1 (Default: off)"), OUTPUT_LONG_OBJECT_NAME
),
2266 OPT_BIT('s', NULL
, &output_option
, N_("Suppress author name and timestamp (Default: off)"), OUTPUT_NO_AUTHOR
),
2267 OPT_BIT('e', "show-email", &output_option
, N_("Show author email instead of name (Default: off)"), OUTPUT_SHOW_EMAIL
),
2268 OPT_BIT('w', NULL
, &xdl_opts
, N_("Ignore whitespace differences"), XDF_IGNORE_WHITESPACE
),
2269 OPT_BIT(0, "minimal", &xdl_opts
, N_("Spend extra cycles to find better match"), XDF_NEED_MINIMAL
),
2270 OPT_STRING('S', NULL
, &revs_file
, N_("file"), N_("Use revisions from <file> instead of calling git-rev-list")),
2271 OPT_STRING(0, "contents", &contents_from
, N_("file"), N_("Use <file>'s contents as the final image")),
2272 { OPTION_CALLBACK
, 'C', NULL
, &opt
, N_("score"), N_("Find line copies within and across files"), PARSE_OPT_OPTARG
, blame_copy_callback
},
2273 { OPTION_CALLBACK
, 'M', NULL
, &opt
, N_("score"), N_("Find line movements within and across files"), PARSE_OPT_OPTARG
, blame_move_callback
},
2274 OPT_STRING_LIST('L', NULL
, &range_list
, N_("n,m"), N_("Process only line range n,m, counting from 1")),
2275 OPT__ABBREV(&abbrev
),
2279 struct parse_opt_ctx_t ctx
;
2280 int cmd_is_annotate
= !strcmp(argv
[0], "annotate");
2281 struct range_set ranges
;
2282 unsigned int range_i
;
2284 git_config(git_blame_config
, NULL
);
2285 init_revisions(&revs
, NULL
);
2286 revs
.date_mode
= blame_date_mode
;
2287 DIFF_OPT_SET(&revs
.diffopt
, ALLOW_TEXTCONV
);
2288 DIFF_OPT_SET(&revs
.diffopt
, FOLLOW_RENAMES
);
2290 save_commit_buffer
= 0;
2293 parse_options_start(&ctx
, argc
, argv
, prefix
, options
,
2294 PARSE_OPT_KEEP_DASHDASH
| PARSE_OPT_KEEP_ARGV0
);
2296 switch (parse_options_step(&ctx
, options
, blame_opt_usage
)) {
2297 case PARSE_OPT_HELP
:
2299 case PARSE_OPT_DONE
:
2301 dashdash_pos
= ctx
.cpidx
;
2305 if (!strcmp(ctx
.argv
[0], "--reverse")) {
2306 ctx
.argv
[0] = "--children";
2309 parse_revision_opt(&revs
, &ctx
, options
, blame_opt_usage
);
2312 no_whole_file_rename
= !DIFF_OPT_TST(&revs
.diffopt
, FOLLOW_RENAMES
);
2313 DIFF_OPT_CLR(&revs
.diffopt
, FOLLOW_RENAMES
);
2314 argc
= parse_options_end(&ctx
);
2317 /* one more abbrev length is needed for the boundary commit */
2320 if (revs_file
&& read_ancestry(revs_file
))
2321 die_errno("reading graft file '%s' failed", revs_file
);
2323 if (cmd_is_annotate
) {
2324 output_option
|= OUTPUT_ANNOTATE_COMPAT
;
2325 blame_date_mode
= DATE_ISO8601
;
2327 blame_date_mode
= revs
.date_mode
;
2330 /* The maximum width used to show the dates */
2331 switch (blame_date_mode
) {
2333 blame_date_width
= sizeof("Thu, 19 Oct 2006 16:00:04 -0700");
2336 blame_date_width
= sizeof("2006-10-19 16:00:04 -0700");
2339 blame_date_width
= sizeof("1161298804 -0700");
2342 blame_date_width
= sizeof("2006-10-19");
2345 /* "normal" is used as the fallback for "relative" */
2348 blame_date_width
= sizeof("Thu Oct 19 16:00:04 2006 -0700");
2351 blame_date_width
-= 1; /* strip the null */
2353 if (DIFF_OPT_TST(&revs
.diffopt
, FIND_COPIES_HARDER
))
2354 opt
|= (PICKAXE_BLAME_COPY
| PICKAXE_BLAME_MOVE
|
2355 PICKAXE_BLAME_COPY_HARDER
);
2357 if (!blame_move_score
)
2358 blame_move_score
= BLAME_DEFAULT_MOVE_SCORE
;
2359 if (!blame_copy_score
)
2360 blame_copy_score
= BLAME_DEFAULT_COPY_SCORE
;
2363 * We have collected options unknown to us in argv[1..unk]
2364 * which are to be passed to revision machinery if we are
2365 * going to do the "bottom" processing.
2367 * The remaining are:
2369 * (1) if dashdash_pos != 0, it is either
2370 * "blame [revisions] -- <path>" or
2371 * "blame -- <path> <rev>"
2373 * (2) otherwise, it is one of the two:
2374 * "blame [revisions] <path>"
2375 * "blame <path> <rev>"
2377 * Note that we must strip out <path> from the arguments: we do not
2378 * want the path pruning but we may want "bottom" processing.
2381 switch (argc
- dashdash_pos
- 1) {
2384 usage_with_options(blame_opt_usage
, options
);
2385 /* reorder for the new way: <rev> -- <path> */
2391 path
= add_prefix(prefix
, argv
[--argc
]);
2395 usage_with_options(blame_opt_usage
, options
);
2399 usage_with_options(blame_opt_usage
, options
);
2400 path
= add_prefix(prefix
, argv
[argc
- 1]);
2401 if (argc
== 3 && !has_string_in_work_tree(path
)) { /* (2b) */
2402 path
= add_prefix(prefix
, argv
[1]);
2405 argv
[argc
- 1] = "--";
2408 if (!has_string_in_work_tree(path
))
2409 die_errno("cannot stat path '%s'", path
);
2412 revs
.disable_stdin
= 1;
2413 setup_revisions(argc
, argv
, &revs
, NULL
);
2414 memset(&sb
, 0, sizeof(sb
));
2418 final_commit_name
= prepare_final(&sb
);
2419 else if (contents_from
)
2420 die("--contents and --children do not blend well.");
2422 final_commit_name
= prepare_initial(&sb
);
2426 * "--not A B -- path" without anything positive;
2427 * do not default to HEAD, but use the working tree
2431 sb
.final
= fake_working_tree_commit(&sb
.revs
->diffopt
,
2432 path
, contents_from
);
2433 add_pending_object(&revs
, &(sb
.final
->object
), ":");
2435 else if (contents_from
)
2436 die("Cannot use --contents with final commit object name");
2439 * If we have bottom, this will mark the ancestors of the
2440 * bottom commits we would reach while traversing as
2443 if (prepare_revision_walk(&revs
))
2444 die("revision walk setup failed");
2446 if (is_null_sha1(sb
.final
->object
.sha1
)) {
2449 buf
= xmalloc(o
->file
.size
+ 1);
2450 memcpy(buf
, o
->file
.ptr
, o
->file
.size
+ 1);
2452 sb
.final_buf_size
= o
->file
.size
;
2455 o
= get_origin(&sb
, sb
.final
, path
);
2456 if (fill_blob_sha1_and_mode(o
))
2457 die("no such path %s in %s", path
, final_commit_name
);
2459 if (DIFF_OPT_TST(&sb
.revs
->diffopt
, ALLOW_TEXTCONV
) &&
2460 textconv_object(path
, o
->mode
, o
->blob_sha1
, 1, (char **) &sb
.final_buf
,
2461 &sb
.final_buf_size
))
2464 sb
.final_buf
= read_sha1_file(o
->blob_sha1
, &type
,
2465 &sb
.final_buf_size
);
2468 die("Cannot read blob %s for path %s",
2469 sha1_to_hex(o
->blob_sha1
),
2473 lno
= prepare_lines(&sb
);
2475 if (lno
&& !range_list
.nr
)
2476 string_list_append(&range_list
, xstrdup("1"));
2478 range_set_init(&ranges
, range_list
.nr
);
2479 for (range_i
= 0; range_i
< range_list
.nr
; ++range_i
) {
2481 if (parse_range_arg(range_list
.items
[range_i
].string
,
2482 nth_line_cb
, &sb
, lno
,
2483 &bottom
, &top
, sb
.path
))
2485 if (lno
< top
|| ((lno
|| bottom
) && lno
< bottom
))
2486 die("file %s has only %lu lines", path
, lno
);
2492 range_set_append_unsafe(&ranges
, bottom
, top
);
2494 sort_and_merge_range_set(&ranges
);
2496 for (range_i
= ranges
.nr
; range_i
> 0; --range_i
) {
2497 const struct range
*r
= &ranges
.ranges
[range_i
- 1];
2498 long bottom
= r
->start
;
2500 struct blame_entry
*next
= ent
;
2501 ent
= xcalloc(1, sizeof(*ent
));
2503 ent
->num_lines
= top
- bottom
;
2505 ent
->s_lno
= bottom
;
2513 range_set_release(&ranges
);
2514 string_list_clear(&range_list
, 0);
2519 read_mailmap(&mailmap
, NULL
);
2524 assign_blame(&sb
, opt
);
2531 if (!(output_option
& OUTPUT_PORCELAIN
))
2532 find_alignment(&sb
, &output_option
);
2534 output(&sb
, output_option
);
2535 free((void *)sb
.final_buf
);
2536 for (ent
= sb
.ent
; ent
; ) {
2537 struct blame_entry
*e
= ent
->next
;
2543 printf("num read blob: %d\n", num_read_blob
);
2544 printf("num get patch: %d\n", num_get_patch
);
2545 printf("num commits: %d\n", num_commits
);