Update draft release notes to 2.6.5
[git/git.git] / builtin / blame.c
CommitLineData
cee7f245 1/*
31653c1a 2 * Blame
cee7f245 3 *
7e6ac6e4
DK
4 * Copyright (c) 2006, 2014 by its authors
5 * See COPYING for licensing conditions
cee7f245
JH
6 */
7
8#include "cache.h"
fb58c8d5 9#include "refs.h"
cee7f245
JH
10#include "builtin.h"
11#include "blob.h"
12#include "commit.h"
13#include "tag.h"
14#include "tree-walk.h"
15#include "diff.h"
16#include "diffcore.h"
17#include "revision.h"
717d1462 18#include "quote.h"
cee7f245 19#include "xdiff-interface.h"
1cfe7733 20#include "cache-tree.h"
c455c87c 21#include "string-list.h"
f9567384 22#include "mailmap.h"
7e6ac6e4 23#include "mergesort.h"
5817da01 24#include "parse-options.h"
7e6ac6e4 25#include "prio-queue.h"
ffaf9cc0 26#include "utf8.h"
3b8a12e8 27#include "userdiff.h"
25ed3412 28#include "line-range.h"
58dbfa2e 29#include "line-log.h"
dbe44faa 30#include "dir.h"
cee7f245 31
ce41720c 32static char blame_usage[] = N_("git blame [<options>] [<rev-opts>] [<rev>] [--] <file>");
5817da01
PH
33
34static const char *blame_opt_usage[] = {
35 blame_usage,
36 "",
9c9b4f2f 37 N_("<rev-opts> are documented in git-rev-list(1)"),
5817da01
PH
38 NULL
39};
cee7f245
JH
40
41static int longest_file;
42static int longest_author;
43static int max_orig_digits;
44static int max_digits;
5ff62c30 45static int max_score_digits;
4c10a5ca 46static int show_root;
85af7929 47static int reverse;
4c10a5ca 48static int blank_boundary;
717d1462 49static int incremental;
582aa00b 50static int xdl_opts;
84393bfd 51static int abbrev = -1;
3d1aa566 52static int no_whole_file_rename;
31653c1a 53
a5481a6c 54static struct date_mode blame_date_mode = { DATE_ISO8601 };
31653c1a
EL
55static size_t blame_date_width;
56
c455c87c 57static struct string_list mailmap;
cee7f245 58
54a4c617
JH
59#ifndef DEBUG
60#define DEBUG 0
61#endif
62
c2e525d9
JH
63/* stats */
64static int num_read_blob;
65static int num_get_patch;
66static int num_commits;
67
d24bba80 68#define PICKAXE_BLAME_MOVE 01
18abd745
JH
69#define PICKAXE_BLAME_COPY 02
70#define PICKAXE_BLAME_COPY_HARDER 04
c63777c0 71#define PICKAXE_BLAME_COPY_HARDEST 010
d24bba80 72
4a0fc95f
JH
73/*
74 * blame for a blame_entry with score lower than these thresholds
75 * is not passed to the parent using move/copy logic.
76 */
77static unsigned blame_move_score;
78static unsigned blame_copy_score;
79#define BLAME_DEFAULT_MOVE_SCORE 20
80#define BLAME_DEFAULT_COPY_SCORE 40
81
208acbfb 82/* Remember to update object flag allocation in object.h */
cee7f245
JH
83#define METAINFO_SHOWN (1u<<12)
84#define MORE_THAN_ONE_PATH (1u<<13)
85
86/*
54a4c617 87 * One blob in a commit that is being suspected
cee7f245
JH
88 */
89struct origin {
54a4c617 90 int refcnt;
7e6ac6e4 91 /* Record preceding blame record for this blob */
96e11709 92 struct origin *previous;
7e6ac6e4
DK
93 /* origins are put in a list linked via `next' hanging off the
94 * corresponding commit's util field in order to make finding
95 * them fast. The presence in this chain does not count
96 * towards the origin's reference count. It is tempting to
97 * let it count as long as the commit is pending examination,
98 * but even under circumstances where the commit will be
99 * present multiple times in the priority queue of unexamined
100 * commits, processing the first instance will not leave any
101 * work requiring the origin data for the second instance. An
102 * interspersed commit changing that would have to be
103 * preexisting with a different ancestry and with the same
104 * commit date in order to wedge itself between two instances
105 * of the same commit in the priority queue _and_ produce
106 * blame entries relevant for it. While we don't want to let
107 * us get tripped up by this case, it certainly does not seem
108 * worth optimizing for.
109 */
110 struct origin *next;
cee7f245 111 struct commit *commit;
7e6ac6e4
DK
112 /* `suspects' contains blame entries that may be attributed to
113 * this origin's commit or to parent commits. When a commit
114 * is being processed, all suspects will be moved, either by
115 * assigning them to an origin in a different commit, or by
116 * shipping them to the scoreboard's ent list because they
117 * cannot be attributed to a different commit.
118 */
119 struct blame_entry *suspects;
c2e525d9 120 mmfile_t file;
cee7f245 121 unsigned char blob_sha1[20];
90064710 122 unsigned mode;
7e6ac6e4
DK
123 /* guilty gets set when shipping any suspects to the final
124 * blame list instead of other commits
125 */
126 char guilty;
cee7f245
JH
127 char path[FLEX_ARRAY];
128};
129
4b4132fd
RS
130static int diff_hunks(mmfile_t *file_a, mmfile_t *file_b, long ctxlen,
131 xdl_emit_hunk_consume_func_t hunk_func, void *cb_data)
132{
133 xpparam_t xpp = {0};
134 xdemitconf_t xecfg = {0};
85c20c30 135 xdemitcb_t ecb = {NULL};
4b4132fd
RS
136
137 xpp.flags = xdl_opts;
138 xecfg.ctxlen = ctxlen;
139 xecfg.hunk_func = hunk_func;
140 ecb.priv = cb_data;
141 return xdi_diff(file_a, file_b, &xpp, &xecfg, &ecb);
142}
143
3b8a12e8
AB
144/*
145 * Prepare diff_filespec and convert it using diff textconv API
146 * if the textconv driver exists.
147 * Return 1 if the conversion succeeds, 0 otherwise.
148 */
e5fba602 149int textconv_object(const char *path,
90064710 150 unsigned mode,
e5fba602 151 const unsigned char *sha1,
e5450100 152 int sha1_valid,
e5fba602
CP
153 char **buf,
154 unsigned long *buf_size)
3b8a12e8
AB
155{
156 struct diff_filespec *df;
157 struct userdiff_driver *textconv;
158
159 df = alloc_filespec(path);
e5450100 160 fill_filespec(df, sha1, sha1_valid, mode);
3b8a12e8
AB
161 textconv = get_textconv(df);
162 if (!textconv) {
163 free_filespec(df);
164 return 0;
165 }
166
167 *buf_size = fill_textconv(textconv, df, buf);
168 free_filespec(df);
169 return 1;
170}
171
1732a1fd
JH
172/*
173 * Given an origin, prepare mmfile_t structure to be used by the
174 * diff machinery
175 */
3b8a12e8
AB
176static void fill_origin_blob(struct diff_options *opt,
177 struct origin *o, mmfile_t *file)
c2e525d9
JH
178{
179 if (!o->file.ptr) {
21666f1a 180 enum object_type type;
3b8a12e8
AB
181 unsigned long file_size;
182
c2e525d9 183 num_read_blob++;
3b8a12e8 184 if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) &&
e5450100 185 textconv_object(o->path, o->mode, o->blob_sha1, 1, &file->ptr, &file_size))
3b8a12e8
AB
186 ;
187 else
188 file->ptr = read_sha1_file(o->blob_sha1, &type, &file_size);
189 file->size = file_size;
190
ab43e495
JH
191 if (!file->ptr)
192 die("Cannot read blob %s for path %s",
193 sha1_to_hex(o->blob_sha1),
194 o->path);
c2e525d9
JH
195 o->file = *file;
196 }
197 else
198 *file = o->file;
c2e525d9
JH
199}
200
1732a1fd
JH
201/*
202 * Origin is refcounted and usually we keep the blob contents to be
203 * reused.
204 */
54a4c617
JH
205static inline struct origin *origin_incref(struct origin *o)
206{
207 if (o)
208 o->refcnt++;
209 return o;
210}
211
212static void origin_decref(struct origin *o)
213{
214 if (o && --o->refcnt <= 0) {
7e6ac6e4 215 struct origin *p, *l = NULL;
96e11709
JH
216 if (o->previous)
217 origin_decref(o->previous);
8e0f7003 218 free(o->file.ptr);
7e6ac6e4
DK
219 /* Should be present exactly once in commit chain */
220 for (p = o->commit->util; p; l = p, p = p->next) {
221 if (p == o) {
222 if (l)
223 l->next = p->next;
224 else
225 o->commit->util = p->next;
226 free(o);
227 return;
228 }
229 }
230 die("internal error in blame::origin_decref");
54a4c617
JH
231 }
232}
233
7c3c7962
JH
234static void drop_origin_blob(struct origin *o)
235{
236 if (o->file.ptr) {
237 free(o->file.ptr);
238 o->file.ptr = NULL;
239 }
240}
241
1732a1fd
JH
242/*
243 * Each group of lines is described by a blame_entry; it can be split
7e6ac6e4
DK
244 * as we pass blame to the parents. They are arranged in linked lists
245 * kept as `suspects' of some unprocessed origin, or entered (when the
246 * blame origin has been finalized) into the scoreboard structure.
247 * While the scoreboard structure is only sorted at the end of
248 * processing (according to final image line number), the lists
249 * attached to an origin are sorted by the target line number.
1732a1fd 250 */
cee7f245 251struct blame_entry {
cee7f245
JH
252 struct blame_entry *next;
253
254 /* the first line of this group in the final image;
255 * internally all line numbers are 0 based.
256 */
257 int lno;
258
259 /* how many lines this group has */
260 int num_lines;
261
262 /* the commit that introduced this group into the final image */
263 struct origin *suspect;
264
cee7f245
JH
265 /* the line number of the first line of this group in the
266 * suspect's file; internally all line numbers are 0 based.
267 */
268 int s_lno;
5ff62c30
JH
269
270 /* how significant this entry is -- cached to avoid
1732a1fd 271 * scanning the lines over and over.
5ff62c30
JH
272 */
273 unsigned score;
cee7f245
JH
274};
275
7e6ac6e4
DK
276/*
277 * Any merge of blames happens on lists of blames that arrived via
278 * different parents in a single suspect. In this case, we want to
279 * sort according to the suspect line numbers as opposed to the final
280 * image line numbers. The function body is somewhat longish because
281 * it avoids unnecessary writes.
282 */
283
284static struct blame_entry *blame_merge(struct blame_entry *list1,
285 struct blame_entry *list2)
286{
287 struct blame_entry *p1 = list1, *p2 = list2,
288 **tail = &list1;
289
290 if (!p1)
291 return p2;
292 if (!p2)
293 return p1;
294
295 if (p1->s_lno <= p2->s_lno) {
296 do {
297 tail = &p1->next;
298 if ((p1 = *tail) == NULL) {
299 *tail = p2;
300 return list1;
301 }
302 } while (p1->s_lno <= p2->s_lno);
303 }
304 for (;;) {
305 *tail = p2;
306 do {
307 tail = &p2->next;
308 if ((p2 = *tail) == NULL) {
309 *tail = p1;
310 return list1;
311 }
312 } while (p1->s_lno > p2->s_lno);
313 *tail = p1;
314 do {
315 tail = &p1->next;
316 if ((p1 = *tail) == NULL) {
317 *tail = p2;
318 return list1;
319 }
320 } while (p1->s_lno <= p2->s_lno);
321 }
322}
323
324static void *get_next_blame(const void *p)
325{
326 return ((struct blame_entry *)p)->next;
327}
328
329static void set_next_blame(void *p1, void *p2)
330{
331 ((struct blame_entry *)p1)->next = p2;
332}
333
334/*
335 * Final image line numbers are all different, so we don't need a
336 * three-way comparison here.
337 */
338
339static int compare_blame_final(const void *p1, const void *p2)
340{
341 return ((struct blame_entry *)p1)->lno > ((struct blame_entry *)p2)->lno
342 ? 1 : -1;
343}
344
345static int compare_blame_suspect(const void *p1, const void *p2)
346{
347 const struct blame_entry *s1 = p1, *s2 = p2;
348 /*
349 * to allow for collating suspects, we sort according to the
350 * respective pointer value as the primary sorting criterion.
351 * The actual relation is pretty unimportant as long as it
352 * establishes a total order. Comparing as integers gives us
353 * that.
354 */
355 if (s1->suspect != s2->suspect)
356 return (intptr_t)s1->suspect > (intptr_t)s2->suspect ? 1 : -1;
357 if (s1->s_lno == s2->s_lno)
358 return 0;
359 return s1->s_lno > s2->s_lno ? 1 : -1;
360}
361
362static struct blame_entry *blame_sort(struct blame_entry *head,
363 int (*compare_fn)(const void *, const void *))
364{
365 return llist_mergesort (head, get_next_blame, set_next_blame, compare_fn);
366}
367
368static int compare_commits_by_reverse_commit_date(const void *a,
369 const void *b,
370 void *c)
371{
372 return -compare_commits_by_commit_date(a, b, c);
373}
374
1732a1fd
JH
375/*
376 * The current state of the blame assignment.
377 */
cee7f245
JH
378struct scoreboard {
379 /* the final commit (i.e. where we started digging from) */
380 struct commit *final;
7e6ac6e4
DK
381 /* Priority queue for commits with unassigned blame records */
382 struct prio_queue commits;
85af7929 383 struct rev_info *revs;
cee7f245
JH
384 const char *path;
385
1732a1fd
JH
386 /*
387 * The contents in the final image.
388 * Used by many functions to obtain contents of the nth line,
389 * indexed with scoreboard.lineno[blame_entry.lno].
cee7f245
JH
390 */
391 const char *final_buf;
392 unsigned long final_buf_size;
393
394 /* linked list of blames */
395 struct blame_entry *ent;
396
612702e8 397 /* look-up a line in the final buffer */
cee7f245
JH
398 int num_lines;
399 int *lineno;
400};
401
54a4c617
JH
402static void sanity_check_refcnt(struct scoreboard *);
403
1732a1fd
JH
404/*
405 * If two blame entries that are next to each other came from
406 * contiguous lines in the same origin (i.e. <commit, path> pair),
407 * merge them together.
408 */
cee7f245
JH
409static void coalesce(struct scoreboard *sb)
410{
411 struct blame_entry *ent, *next;
412
413 for (ent = sb->ent; ent && (next = ent->next); ent = next) {
0a88f08e 414 if (ent->suspect == next->suspect &&
cee7f245
JH
415 ent->s_lno + ent->num_lines == next->s_lno) {
416 ent->num_lines += next->num_lines;
417 ent->next = next->next;
54a4c617 418 origin_decref(next->suspect);
cee7f245 419 free(next);
46014766 420 ent->score = 0;
cee7f245
JH
421 next = ent; /* again */
422 }
423 }
54a4c617
JH
424
425 if (DEBUG) /* sanity */
426 sanity_check_refcnt(sb);
cee7f245
JH
427}
428
7e6ac6e4
DK
429/*
430 * Merge the given sorted list of blames into a preexisting origin.
431 * If there were no previous blames to that commit, it is entered into
432 * the commit priority queue of the score board.
433 */
434
435static void queue_blames(struct scoreboard *sb, struct origin *porigin,
436 struct blame_entry *sorted)
437{
438 if (porigin->suspects)
439 porigin->suspects = blame_merge(porigin->suspects, sorted);
440 else {
441 struct origin *o;
442 for (o = porigin->commit->util; o; o = o->next) {
443 if (o->suspects) {
444 porigin->suspects = sorted;
445 return;
446 }
447 }
448 porigin->suspects = sorted;
449 prio_queue_put(&sb->commits, porigin->commit);
450 }
451}
452
1732a1fd
JH
453/*
454 * Given a commit and a path in it, create a new origin structure.
455 * The callers that add blame to the scoreboard should use
456 * get_origin() to obtain shared, refcounted copy instead of calling
457 * this function directly.
458 */
854b97f6
JH
459static struct origin *make_origin(struct commit *commit, const char *path)
460{
461 struct origin *o;
462 o = xcalloc(1, sizeof(*o) + strlen(path) + 1);
463 o->commit = commit;
464 o->refcnt = 1;
7e6ac6e4
DK
465 o->next = commit->util;
466 commit->util = o;
854b97f6
JH
467 strcpy(o->path, path);
468 return o;
469}
470
1732a1fd
JH
471/*
472 * Locate an existing origin or create a new one.
7e6ac6e4 473 * This moves the origin to front position in the commit util list.
1732a1fd 474 */
f6c0e191
JH
475static struct origin *get_origin(struct scoreboard *sb,
476 struct commit *commit,
477 const char *path)
cee7f245 478{
7e6ac6e4 479 struct origin *o, *l;
cee7f245 480
7e6ac6e4
DK
481 for (o = commit->util, l = NULL; o; l = o, o = o->next) {
482 if (!strcmp(o->path, path)) {
483 /* bump to front */
484 if (l) {
485 l->next = o->next;
486 o->next = commit->util;
487 commit->util = o;
488 }
489 return origin_incref(o);
490 }
cee7f245 491 }
854b97f6 492 return make_origin(commit, path);
cee7f245
JH
493}
494
1732a1fd
JH
495/*
496 * Fill the blob_sha1 field of an origin if it hasn't, so that later
497 * call to fill_origin_blob() can use it to locate the data. blob_sha1
498 * for an origin is also used to pass the blame for the entire file to
499 * the parent to detect the case where a child's blob is identical to
500 * that of its parent's.
90064710
KS
501 *
502 * This also fills origin->mode for corresponding tree path.
1732a1fd 503 */
90064710 504static int fill_blob_sha1_and_mode(struct origin *origin)
f6c0e191 505{
f6c0e191
JH
506 if (!is_null_sha1(origin->blob_sha1))
507 return 0;
508 if (get_tree_entry(origin->commit->object.sha1,
509 origin->path,
90064710 510 origin->blob_sha1, &origin->mode))
f6c0e191 511 goto error_out;
21666f1a 512 if (sha1_object_info(origin->blob_sha1, NULL) != OBJ_BLOB)
f6c0e191
JH
513 goto error_out;
514 return 0;
515 error_out:
516 hashclr(origin->blob_sha1);
90064710 517 origin->mode = S_IFINVALID;
f6c0e191
JH
518 return -1;
519}
520
1732a1fd
JH
521/*
522 * We have an origin -- check if the same path exists in the
523 * parent and return an origin structure to represent it.
524 */
f6c0e191 525static struct origin *find_origin(struct scoreboard *sb,
cee7f245
JH
526 struct commit *parent,
527 struct origin *origin)
528{
7e6ac6e4 529 struct origin *porigin;
cee7f245 530 struct diff_options diff_opts;
f6c0e191
JH
531 const char *paths[2];
532
7e6ac6e4
DK
533 /* First check any existing origins */
534 for (porigin = parent->util; porigin; porigin = porigin->next)
535 if (!strcmp(porigin->path, origin->path)) {
1732a1fd
JH
536 /*
537 * The same path between origin and its parent
538 * without renaming -- the most common case.
539 */
7e6ac6e4 540 return origin_incref (porigin);
854b97f6 541 }
0d981c67 542
f6c0e191
JH
543 /* See if the origin->path is different between parent
544 * and origin first. Most of the time they are the
545 * same and diff-tree is fairly efficient about this.
546 */
547 diff_setup(&diff_opts);
8f67f8ae 548 DIFF_OPT_SET(&diff_opts, RECURSIVE);
f6c0e191
JH
549 diff_opts.detect_rename = 0;
550 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
551 paths[0] = origin->path;
552 paths[1] = NULL;
553
4a2d5ae2
NTND
554 parse_pathspec(&diff_opts.pathspec,
555 PATHSPEC_ALL_MAGIC & ~PATHSPEC_LITERAL,
556 PATHSPEC_LITERAL_PATH, "", paths);
28452655 557 diff_setup_done(&diff_opts);
1cfe7733
JH
558
559 if (is_null_sha1(origin->commit->object.sha1))
560 do_diff_cache(parent->tree->object.sha1, &diff_opts);
561 else
562 diff_tree_sha1(parent->tree->object.sha1,
563 origin->commit->tree->object.sha1,
564 "", &diff_opts);
f6c0e191
JH
565 diffcore_std(&diff_opts);
566
f6c0e191
JH
567 if (!diff_queued_diff.nr) {
568 /* The path is the same as parent */
569 porigin = get_origin(sb, parent, origin->path);
570 hashcpy(porigin->blob_sha1, origin->blob_sha1);
90064710 571 porigin->mode = origin->mode;
a9b2d424
JH
572 } else {
573 /*
574 * Since origin->path is a pathspec, if the parent
575 * commit had it as a directory, we will see a whole
576 * bunch of deletion of files in the directory that we
577 * do not care about.
578 */
579 int i;
580 struct diff_filepair *p = NULL;
581 for (i = 0; i < diff_queued_diff.nr; i++) {
582 const char *name;
583 p = diff_queued_diff.queue[i];
584 name = p->one->path ? p->one->path : p->two->path;
585 if (!strcmp(name, origin->path))
586 break;
587 }
588 if (!p)
589 die("internal error in blame::find_origin");
f6c0e191
JH
590 switch (p->status) {
591 default:
acca687f 592 die("internal error in blame::find_origin (%c)",
f6c0e191
JH
593 p->status);
594 case 'M':
595 porigin = get_origin(sb, parent, origin->path);
596 hashcpy(porigin->blob_sha1, p->one->sha1);
90064710 597 porigin->mode = p->one->mode;
f6c0e191
JH
598 break;
599 case 'A':
600 case 'T':
601 /* Did not exist in parent, or type changed */
602 break;
603 }
604 }
605 diff_flush(&diff_opts);
bd1928df 606 free_pathspec(&diff_opts.pathspec);
f69e743d
JH
607 return porigin;
608}
f6c0e191 609
1732a1fd
JH
610/*
611 * We have an origin -- find the path that corresponds to it in its
612 * parent and return an origin structure to represent it.
613 */
f69e743d
JH
614static struct origin *find_rename(struct scoreboard *sb,
615 struct commit *parent,
616 struct origin *origin)
617{
618 struct origin *porigin = NULL;
619 struct diff_options diff_opts;
620 int i;
cee7f245
JH
621
622 diff_setup(&diff_opts);
8f67f8ae 623 DIFF_OPT_SET(&diff_opts, RECURSIVE);
cee7f245
JH
624 diff_opts.detect_rename = DIFF_DETECT_RENAME;
625 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
2f3f8b21 626 diff_opts.single_follow = origin->path;
28452655 627 diff_setup_done(&diff_opts);
1cfe7733
JH
628
629 if (is_null_sha1(origin->commit->object.sha1))
630 do_diff_cache(parent->tree->object.sha1, &diff_opts);
631 else
632 diff_tree_sha1(parent->tree->object.sha1,
633 origin->commit->tree->object.sha1,
634 "", &diff_opts);
cee7f245
JH
635 diffcore_std(&diff_opts);
636
637 for (i = 0; i < diff_queued_diff.nr; i++) {
638 struct diff_filepair *p = diff_queued_diff.queue[i];
612702e8 639 if ((p->status == 'R' || p->status == 'C') &&
f6c0e191
JH
640 !strcmp(p->two->path, origin->path)) {
641 porigin = get_origin(sb, parent, p->one->path);
642 hashcpy(porigin->blob_sha1, p->one->sha1);
90064710 643 porigin->mode = p->one->mode;
cee7f245
JH
644 break;
645 }
646 }
647 diff_flush(&diff_opts);
bd1928df 648 free_pathspec(&diff_opts.pathspec);
cee7f245
JH
649 return porigin;
650}
651
1732a1fd 652/*
7e6ac6e4 653 * Append a new blame entry to a given output queue.
1732a1fd 654 */
7e6ac6e4 655static void add_blame_entry(struct blame_entry ***queue, struct blame_entry *e)
cee7f245 656{
54a4c617
JH
657 origin_incref(e->suspect);
658
7e6ac6e4
DK
659 e->next = **queue;
660 **queue = e;
661 *queue = &e->next;
cee7f245
JH
662}
663
1732a1fd
JH
664/*
665 * src typically is on-stack; we want to copy the information in it to
7e6ac6e4
DK
666 * a malloced blame_entry that gets added to the given queue. The
667 * origin of dst loses a refcnt.
1732a1fd 668 */
7e6ac6e4
DK
669static void dup_entry(struct blame_entry ***queue,
670 struct blame_entry *dst, struct blame_entry *src)
cee7f245 671{
54a4c617
JH
672 origin_incref(src->suspect);
673 origin_decref(dst->suspect);
cee7f245 674 memcpy(dst, src, sizeof(*src));
7e6ac6e4
DK
675 dst->next = **queue;
676 **queue = dst;
677 *queue = &dst->next;
cee7f245
JH
678}
679
25ed3412 680static const char *nth_line(struct scoreboard *sb, long lno)
cee7f245
JH
681{
682 return sb->final_buf + sb->lineno[lno];
683}
684
25ed3412
BY
685static const char *nth_line_cb(void *data, long lno)
686{
687 return nth_line((struct scoreboard *)data, lno);
688}
689
1732a1fd
JH
690/*
691 * It is known that lines between tlno to same came from parent, and e
692 * has an overlap with that range. it also is known that parent's
693 * line plno corresponds to e's line tlno.
694 *
695 * <---- e ----->
696 * <------>
697 * <------------>
698 * <------------>
699 * <------------------>
700 *
701 * Split e into potentially three parts; before this chunk, the chunk
702 * to be blamed for the parent, and after that portion.
703 */
54a4c617 704static void split_overlap(struct blame_entry *split,
cee7f245
JH
705 struct blame_entry *e,
706 int tlno, int plno, int same,
707 struct origin *parent)
708{
cee7f245
JH
709 int chunk_end_lno;
710 memset(split, 0, sizeof(struct blame_entry [3]));
711
712 if (e->s_lno < tlno) {
713 /* there is a pre-chunk part not blamed on parent */
54a4c617 714 split[0].suspect = origin_incref(e->suspect);
cee7f245
JH
715 split[0].lno = e->lno;
716 split[0].s_lno = e->s_lno;
717 split[0].num_lines = tlno - e->s_lno;
718 split[1].lno = e->lno + tlno - e->s_lno;
719 split[1].s_lno = plno;
720 }
721 else {
722 split[1].lno = e->lno;
723 split[1].s_lno = plno + (e->s_lno - tlno);
724 }
725
726 if (same < e->s_lno + e->num_lines) {
727 /* there is a post-chunk part not blamed on parent */
54a4c617 728 split[2].suspect = origin_incref(e->suspect);
cee7f245
JH
729 split[2].lno = e->lno + (same - e->s_lno);
730 split[2].s_lno = e->s_lno + (same - e->s_lno);
731 split[2].num_lines = e->s_lno + e->num_lines - same;
732 chunk_end_lno = split[2].lno;
733 }
734 else
735 chunk_end_lno = e->lno + e->num_lines;
736 split[1].num_lines = chunk_end_lno - split[1].lno;
737
1732a1fd
JH
738 /*
739 * if it turns out there is nothing to blame the parent for,
740 * forget about the splitting. !split[1].suspect signals this.
741 */
cee7f245
JH
742 if (split[1].num_lines < 1)
743 return;
54a4c617 744 split[1].suspect = origin_incref(parent);
cee7f245
JH
745}
746
1732a1fd
JH
747/*
748 * split_overlap() divided an existing blame e into up to three parts
7e6ac6e4 749 * in split. Any assigned blame is moved to queue to
1732a1fd
JH
750 * reflect the split.
751 */
7e6ac6e4
DK
752static void split_blame(struct blame_entry ***blamed,
753 struct blame_entry ***unblamed,
54a4c617 754 struct blame_entry *split,
cee7f245
JH
755 struct blame_entry *e)
756{
757 struct blame_entry *new_entry;
758
759 if (split[0].suspect && split[2].suspect) {
1732a1fd 760 /* The first part (reuse storage for the existing entry e) */
7e6ac6e4 761 dup_entry(unblamed, e, &split[0]);
cee7f245 762
1732a1fd 763 /* The last part -- me */
cee7f245
JH
764 new_entry = xmalloc(sizeof(*new_entry));
765 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
7e6ac6e4 766 add_blame_entry(unblamed, new_entry);
cee7f245 767
1732a1fd 768 /* ... and the middle part -- parent */
cee7f245
JH
769 new_entry = xmalloc(sizeof(*new_entry));
770 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
7e6ac6e4 771 add_blame_entry(blamed, new_entry);
cee7f245
JH
772 }
773 else if (!split[0].suspect && !split[2].suspect)
1732a1fd
JH
774 /*
775 * The parent covers the entire area; reuse storage for
776 * e and replace it with the parent.
777 */
7e6ac6e4 778 dup_entry(blamed, e, &split[1]);
cee7f245 779 else if (split[0].suspect) {
1732a1fd 780 /* me and then parent */
7e6ac6e4 781 dup_entry(unblamed, e, &split[0]);
cee7f245
JH
782
783 new_entry = xmalloc(sizeof(*new_entry));
784 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
7e6ac6e4 785 add_blame_entry(blamed, new_entry);
cee7f245
JH
786 }
787 else {
1732a1fd 788 /* parent and then me */
7e6ac6e4 789 dup_entry(blamed, e, &split[1]);
cee7f245
JH
790
791 new_entry = xmalloc(sizeof(*new_entry));
792 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
7e6ac6e4 793 add_blame_entry(unblamed, new_entry);
cee7f245
JH
794 }
795}
796
1732a1fd
JH
797/*
798 * After splitting the blame, the origins used by the
799 * on-stack blame_entry should lose one refcnt each.
800 */
54a4c617
JH
801static void decref_split(struct blame_entry *split)
802{
803 int i;
804
805 for (i = 0; i < 3; i++)
806 origin_decref(split[i].suspect);
807}
808
1732a1fd 809/*
7e6ac6e4
DK
810 * reverse_blame reverses the list given in head, appending tail.
811 * That allows us to build lists in reverse order, then reverse them
812 * afterwards. This can be faster than building the list in proper
813 * order right away. The reason is that building in proper order
814 * requires writing a link in the _previous_ element, while building
815 * in reverse order just requires placing the list head into the
816 * _current_ element.
1732a1fd 817 */
cee7f245 818
7e6ac6e4
DK
819static struct blame_entry *reverse_blame(struct blame_entry *head,
820 struct blame_entry *tail)
cee7f245 821{
7e6ac6e4
DK
822 while (head) {
823 struct blame_entry *next = head->next;
824 head->next = tail;
825 tail = head;
826 head = next;
cee7f245 827 }
7e6ac6e4 828 return tail;
cee7f245
JH
829}
830
1732a1fd
JH
831/*
832 * Process one hunk from the patch between the current suspect for
7e6ac6e4
DK
833 * blame_entry e and its parent. This first blames any unfinished
834 * entries before the chunk (which is where target and parent start
835 * differing) on the parent, and then splits blame entries at the
836 * start and at the end of the difference region. Since use of -M and
837 * -C options may lead to overlapping/duplicate source line number
838 * ranges, all we can rely on from sorting/merging is the order of the
839 * first suspect line number.
1732a1fd 840 */
7e6ac6e4
DK
841static void blame_chunk(struct blame_entry ***dstq, struct blame_entry ***srcq,
842 int tlno, int offset, int same,
843 struct origin *parent)
cee7f245 844{
7e6ac6e4
DK
845 struct blame_entry *e = **srcq;
846 struct blame_entry *samep = NULL, *diffp = NULL;
cee7f245 847
7e6ac6e4
DK
848 while (e && e->s_lno < tlno) {
849 struct blame_entry *next = e->next;
850 /*
851 * current record starts before differing portion. If
852 * it reaches into it, we need to split it up and
853 * examine the second part separately.
854 */
855 if (e->s_lno + e->num_lines > tlno) {
856 /* Move second half to a new record */
857 int len = tlno - e->s_lno;
858 struct blame_entry *n = xcalloc(1, sizeof (struct blame_entry));
859 n->suspect = e->suspect;
860 n->lno = e->lno + len;
861 n->s_lno = e->s_lno + len;
862 n->num_lines = e->num_lines - len;
863 e->num_lines = len;
864 e->score = 0;
865 /* Push new record to diffp */
866 n->next = diffp;
867 diffp = n;
868 } else
869 origin_decref(e->suspect);
870 /* Pass blame for everything before the differing
871 * chunk to the parent */
872 e->suspect = origin_incref(parent);
873 e->s_lno += offset;
874 e->next = samep;
875 samep = e;
876 e = next;
877 }
878 /*
879 * As we don't know how much of a common stretch after this
880 * diff will occur, the currently blamed parts are all that we
881 * can assign to the parent for now.
882 */
883
884 if (samep) {
885 **dstq = reverse_blame(samep, **dstq);
886 *dstq = &samep->next;
cee7f245 887 }
7e6ac6e4
DK
888 /*
889 * Prepend the split off portions: everything after e starts
890 * after the blameable portion.
891 */
892 e = reverse_blame(diffp, e);
893
894 /*
895 * Now retain records on the target while parts are different
896 * from the parent.
897 */
898 samep = NULL;
899 diffp = NULL;
900 while (e && e->s_lno < same) {
901 struct blame_entry *next = e->next;
902
903 /*
904 * If current record extends into sameness, need to split.
905 */
906 if (e->s_lno + e->num_lines > same) {
907 /*
908 * Move second half to a new record to be
909 * processed by later chunks
910 */
911 int len = same - e->s_lno;
912 struct blame_entry *n = xcalloc(1, sizeof (struct blame_entry));
913 n->suspect = origin_incref(e->suspect);
914 n->lno = e->lno + len;
915 n->s_lno = e->s_lno + len;
916 n->num_lines = e->num_lines - len;
917 e->num_lines = len;
918 e->score = 0;
919 /* Push new record to samep */
920 n->next = samep;
921 samep = n;
922 }
923 e->next = diffp;
924 diffp = e;
925 e = next;
926 }
927 **srcq = reverse_blame(diffp, reverse_blame(samep, e));
928 /* Move across elements that are in the unblamable portion */
929 if (diffp)
930 *srcq = &diffp->next;
cee7f245
JH
931}
932
cdcab133 933struct blame_chunk_cb_data {
cdcab133 934 struct origin *parent;
7e6ac6e4
DK
935 long offset;
936 struct blame_entry **dstq;
937 struct blame_entry **srcq;
cdcab133
RS
938};
939
7e6ac6e4 940/* diff chunks are from parent to target */
0af596c6
RS
941static int blame_chunk_cb(long start_a, long count_a,
942 long start_b, long count_b, void *data)
cdcab133
RS
943{
944 struct blame_chunk_cb_data *d = data;
7e6ac6e4
DK
945 if (start_a - start_b != d->offset)
946 die("internal error in blame::blame_chunk_cb");
947 blame_chunk(&d->dstq, &d->srcq, start_b, start_a - start_b,
948 start_b + count_b, d->parent);
949 d->offset = start_a + count_a - (start_b + count_b);
0af596c6 950 return 0;
cdcab133
RS
951}
952
1732a1fd
JH
953/*
954 * We are looking at the origin 'target' and aiming to pass blame
955 * for the lines it is suspected to its parent. Run diff to find
956 * which lines came from parent and pass blame for them.
957 */
7e6ac6e4
DK
958static void pass_blame_to_parent(struct scoreboard *sb,
959 struct origin *target,
960 struct origin *parent)
cee7f245 961{
7c4ed8c8 962 mmfile_t file_p, file_o;
66dbfd55 963 struct blame_chunk_cb_data d;
7e6ac6e4 964 struct blame_entry *newdest = NULL;
0af596c6 965
7e6ac6e4
DK
966 if (!target->suspects)
967 return; /* nothing remains for this target */
968
969 d.parent = parent;
970 d.offset = 0;
971 d.dstq = &newdest; d.srcq = &target->suspects;
cee7f245 972
3b8a12e8
AB
973 fill_origin_blob(&sb->revs->diffopt, parent, &file_p);
974 fill_origin_blob(&sb->revs->diffopt, target, &file_o);
7c4ed8c8 975 num_get_patch++;
cee7f245 976
3efb9880
JK
977 if (diff_hunks(&file_p, &file_o, 0, blame_chunk_cb, &d))
978 die("unable to generate diff (%s -> %s)",
979 sha1_to_hex(parent->commit->object.sha1),
980 sha1_to_hex(target->commit->object.sha1));
7e6ac6e4
DK
981 /* The rest are the same as the parent */
982 blame_chunk(&d.dstq, &d.srcq, INT_MAX, d.offset, INT_MAX, parent);
983 *d.dstq = NULL;
984 queue_blames(sb, parent, newdest);
cee7f245 985
7e6ac6e4 986 return;
cee7f245
JH
987}
988
1732a1fd
JH
989/*
990 * The lines in blame_entry after splitting blames many times can become
991 * very small and trivial, and at some point it becomes pointless to
992 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
993 * ordinary C program, and it is not worth to say it was copied from
994 * totally unrelated file in the parent.
995 *
996 * Compute how trivial the lines in the blame_entry are.
997 */
5ff62c30
JH
998static unsigned ent_score(struct scoreboard *sb, struct blame_entry *e)
999{
1000 unsigned score;
1001 const char *cp, *ep;
1002
1003 if (e->score)
1004 return e->score;
1005
612702e8 1006 score = 1;
5ff62c30
JH
1007 cp = nth_line(sb, e->lno);
1008 ep = nth_line(sb, e->lno + e->num_lines);
1009 while (cp < ep) {
1010 unsigned ch = *((unsigned char *)cp);
1011 if (isalnum(ch))
1012 score++;
1013 cp++;
1014 }
1015 e->score = score;
1016 return score;
1017}
1018
1732a1fd
JH
1019/*
1020 * best_so_far[] and this[] are both a split of an existing blame_entry
1021 * that passes blame to the parent. Maintain best_so_far the best split
1022 * so far, by comparing this and best_so_far and copying this into
1023 * bst_so_far as needed.
1024 */
5ff62c30 1025static void copy_split_if_better(struct scoreboard *sb,
54a4c617
JH
1026 struct blame_entry *best_so_far,
1027 struct blame_entry *this)
d24bba80 1028{
54a4c617
JH
1029 int i;
1030
d24bba80
JH
1031 if (!this[1].suspect)
1032 return;
5ff62c30
JH
1033 if (best_so_far[1].suspect) {
1034 if (ent_score(sb, &this[1]) < ent_score(sb, &best_so_far[1]))
1035 return;
1036 }
54a4c617
JH
1037
1038 for (i = 0; i < 3; i++)
1039 origin_incref(this[i].suspect);
1040 decref_split(best_so_far);
d24bba80
JH
1041 memcpy(best_so_far, this, sizeof(struct blame_entry [3]));
1042}
1043
dd166aa8
JH
1044/*
1045 * We are looking at a part of the final image represented by
1046 * ent (tlno and same are offset by ent->s_lno).
1047 * tlno is where we are looking at in the final image.
1048 * up to (but not including) same match preimage.
1049 * plno is where we are looking at in the preimage.
1050 *
1051 * <-------------- final image ---------------------->
1052 * <------ent------>
1053 * ^tlno ^same
1054 * <---------preimage----->
1055 * ^plno
1056 *
1057 * All line numbers are 0-based.
1058 */
1059static void handle_split(struct scoreboard *sb,
1060 struct blame_entry *ent,
1061 int tlno, int plno, int same,
1062 struct origin *parent,
1063 struct blame_entry *split)
1064{
1065 if (ent->num_lines <= tlno)
1066 return;
1067 if (tlno < same) {
1068 struct blame_entry this[3];
1069 tlno += ent->s_lno;
1070 same += ent->s_lno;
1071 split_overlap(this, ent, tlno, plno, same, parent);
1072 copy_split_if_better(sb, split, this);
1073 decref_split(this);
1074 }
1075}
1076
cdcab133
RS
1077struct handle_split_cb_data {
1078 struct scoreboard *sb;
1079 struct blame_entry *ent;
1080 struct origin *parent;
1081 struct blame_entry *split;
1082 long plno;
1083 long tlno;
1084};
1085
5d23ec76
RS
1086static int handle_split_cb(long start_a, long count_a,
1087 long start_b, long count_b, void *data)
cdcab133
RS
1088{
1089 struct handle_split_cb_data *d = data;
5d23ec76
RS
1090 handle_split(d->sb, d->ent, d->tlno, d->plno, start_b, d->parent,
1091 d->split);
1092 d->plno = start_a + count_a;
1093 d->tlno = start_b + count_b;
1094 return 0;
cdcab133
RS
1095}
1096
1732a1fd
JH
1097/*
1098 * Find the lines from parent that are the same as ent so that
1099 * we can pass blames to it. file_p has the blob contents for
1100 * the parent.
1101 */
d24bba80
JH
1102static void find_copy_in_blob(struct scoreboard *sb,
1103 struct blame_entry *ent,
1104 struct origin *parent,
54a4c617 1105 struct blame_entry *split,
d24bba80
JH
1106 mmfile_t *file_p)
1107{
1108 const char *cp;
d24bba80 1109 mmfile_t file_o;
66dbfd55 1110 struct handle_split_cb_data d;
5d23ec76 1111
66dbfd55
GV
1112 memset(&d, 0, sizeof(d));
1113 d.sb = sb; d.ent = ent; d.parent = parent; d.split = split;
1732a1fd
JH
1114 /*
1115 * Prepare mmfile that contains only the lines in ent.
1116 */
d24bba80 1117 cp = nth_line(sb, ent->lno);
4b25d091 1118 file_o.ptr = (char *) cp;
3ee8944f 1119 file_o.size = nth_line(sb, ent->lno + ent->num_lines) - cp;
d24bba80 1120
dd166aa8
JH
1121 /*
1122 * file_o is a part of final image we are annotating.
1123 * file_p partially may match that image.
1124 */
d24bba80 1125 memset(split, 0, sizeof(struct blame_entry [3]));
3efb9880
JK
1126 if (diff_hunks(file_p, &file_o, 1, handle_split_cb, &d))
1127 die("unable to generate diff (%s)",
1128 sha1_to_hex(parent->commit->object.sha1));
dd166aa8 1129 /* remainder, if any, all match the preimage */
cdcab133 1130 handle_split(sb, ent, d.tlno, d.plno, ent->num_lines, parent, split);
d24bba80
JH
1131}
1132
7e6ac6e4
DK
1133/* Move all blame entries from list *source that have a score smaller
1134 * than score_min to the front of list *small.
1135 * Returns a pointer to the link pointing to the old head of the small list.
1136 */
1137
1138static struct blame_entry **filter_small(struct scoreboard *sb,
1139 struct blame_entry **small,
1140 struct blame_entry **source,
1141 unsigned score_min)
1142{
1143 struct blame_entry *p = *source;
1144 struct blame_entry *oldsmall = *small;
1145 while (p) {
1146 if (ent_score(sb, p) <= score_min) {
1147 *small = p;
1148 small = &p->next;
1149 p = *small;
1150 } else {
1151 *source = p;
1152 source = &p->next;
1153 p = *source;
1154 }
1155 }
1156 *small = oldsmall;
1157 *source = NULL;
1158 return small;
1159}
1160
1732a1fd
JH
1161/*
1162 * See if lines currently target is suspected for can be attributed to
1163 * parent.
1164 */
7e6ac6e4
DK
1165static void find_move_in_parent(struct scoreboard *sb,
1166 struct blame_entry ***blamed,
1167 struct blame_entry **toosmall,
1168 struct origin *target,
1169 struct origin *parent)
d24bba80 1170{
46014766 1171 struct blame_entry *e, split[3];
7e6ac6e4
DK
1172 struct blame_entry *unblamed = target->suspects;
1173 struct blame_entry *leftover = NULL;
d24bba80 1174 mmfile_t file_p;
d24bba80 1175
7e6ac6e4
DK
1176 if (!unblamed)
1177 return; /* nothing remains for this target */
d24bba80 1178
3b8a12e8 1179 fill_origin_blob(&sb->revs->diffopt, parent, &file_p);
c2e525d9 1180 if (!file_p.ptr)
7e6ac6e4 1181 return;
d24bba80 1182
7e6ac6e4
DK
1183 /* At each iteration, unblamed has a NULL-terminated list of
1184 * entries that have not yet been tested for blame. leftover
1185 * contains the reversed list of entries that have been tested
1186 * without being assignable to the parent.
1187 */
1188 do {
1189 struct blame_entry **unblamedtail = &unblamed;
1190 struct blame_entry *next;
1191 for (e = unblamed; e; e = next) {
1192 next = e->next;
650e2f67
JH
1193 find_copy_in_blob(sb, e, parent, split, &file_p);
1194 if (split[1].suspect &&
1195 blame_move_score < ent_score(sb, &split[1])) {
7e6ac6e4
DK
1196 split_blame(blamed, &unblamedtail, split, e);
1197 } else {
1198 e->next = leftover;
1199 leftover = e;
650e2f67
JH
1200 }
1201 decref_split(split);
1202 }
7e6ac6e4
DK
1203 *unblamedtail = NULL;
1204 toosmall = filter_small(sb, toosmall, &unblamed, blame_move_score);
1205 } while (unblamed);
1206 target->suspects = reverse_blame(leftover, NULL);
d24bba80
JH
1207}
1208
33494784
JH
1209struct blame_list {
1210 struct blame_entry *ent;
1211 struct blame_entry split[3];
1212};
1213
1732a1fd
JH
1214/*
1215 * Count the number of entries the target is suspected for,
1216 * and prepare a list of entry and the best split.
1217 */
7e6ac6e4 1218static struct blame_list *setup_blame_list(struct blame_entry *unblamed,
33494784
JH
1219 int *num_ents_p)
1220{
1221 struct blame_entry *e;
1222 int num_ents, i;
1223 struct blame_list *blame_list = NULL;
1224
7e6ac6e4
DK
1225 for (e = unblamed, num_ents = 0; e; e = e->next)
1226 num_ents++;
33494784
JH
1227 if (num_ents) {
1228 blame_list = xcalloc(num_ents, sizeof(struct blame_list));
7e6ac6e4
DK
1229 for (e = unblamed, i = 0; e; e = e->next)
1230 blame_list[i++].ent = e;
33494784
JH
1231 }
1232 *num_ents_p = num_ents;
1233 return blame_list;
1234}
1235
1732a1fd
JH
1236/*
1237 * For lines target is suspected for, see if we can find code movement
1238 * across file boundary from the parent commit. porigin is the path
1239 * in the parent we already tried.
1240 */
7e6ac6e4
DK
1241static void find_copy_in_parent(struct scoreboard *sb,
1242 struct blame_entry ***blamed,
1243 struct blame_entry **toosmall,
1244 struct origin *target,
1245 struct commit *parent,
1246 struct origin *porigin,
1247 int opt)
18abd745
JH
1248{
1249 struct diff_options diff_opts;
aec8fa1f 1250 int i, j;
33494784 1251 struct blame_list *blame_list;
aec8fa1f 1252 int num_ents;
7e6ac6e4
DK
1253 struct blame_entry *unblamed = target->suspects;
1254 struct blame_entry *leftover = NULL;
18abd745 1255
7e6ac6e4
DK
1256 if (!unblamed)
1257 return; /* nothing remains for this target */
18abd745
JH
1258
1259 diff_setup(&diff_opts);
8f67f8ae 1260 DIFF_OPT_SET(&diff_opts, RECURSIVE);
18abd745
JH
1261 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
1262
28452655 1263 diff_setup_done(&diff_opts);
33494784
JH
1264
1265 /* Try "find copies harder" on new path if requested;
1266 * we do not want to use diffcore_rename() actually to
1267 * match things up; find_copies_harder is set only to
1268 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1269 * and this code needs to be after diff_setup_done(), which
1270 * usually makes find-copies-harder imply copy detection.
1271 */
c63777c0
JH
1272 if ((opt & PICKAXE_BLAME_COPY_HARDEST)
1273 || ((opt & PICKAXE_BLAME_COPY_HARDER)
1274 && (!porigin || strcmp(target->path, porigin->path))))
8f67f8ae 1275 DIFF_OPT_SET(&diff_opts, FIND_COPIES_HARDER);
33494784 1276
1cfe7733
JH
1277 if (is_null_sha1(target->commit->object.sha1))
1278 do_diff_cache(parent->tree->object.sha1, &diff_opts);
1279 else
1280 diff_tree_sha1(parent->tree->object.sha1,
1281 target->commit->tree->object.sha1,
1282 "", &diff_opts);
18abd745 1283
8f67f8ae 1284 if (!DIFF_OPT_TST(&diff_opts, FIND_COPIES_HARDER))
33494784 1285 diffcore_std(&diff_opts);
18abd745 1286
7e6ac6e4
DK
1287 do {
1288 struct blame_entry **unblamedtail = &unblamed;
1289 blame_list = setup_blame_list(unblamed, &num_ents);
33494784
JH
1290
1291 for (i = 0; i < diff_queued_diff.nr; i++) {
1292 struct diff_filepair *p = diff_queued_diff.queue[i];
1293 struct origin *norigin;
1294 mmfile_t file_p;
33494784
JH
1295 struct blame_entry this[3];
1296
1297 if (!DIFF_FILE_VALID(p->one))
1298 continue; /* does not exist in parent */
5209ac4d
AG
1299 if (S_ISGITLINK(p->one->mode))
1300 continue; /* ignore git links */
33494784
JH
1301 if (porigin && !strcmp(p->one->path, porigin->path))
1302 /* find_move already dealt with this path */
1303 continue;
1304
1305 norigin = get_origin(sb, parent, p->one->path);
1306 hashcpy(norigin->blob_sha1, p->one->sha1);
90064710 1307 norigin->mode = p->one->mode;
3b8a12e8 1308 fill_origin_blob(&sb->revs->diffopt, norigin, &file_p);
33494784
JH
1309 if (!file_p.ptr)
1310 continue;
1311
1312 for (j = 0; j < num_ents; j++) {
1313 find_copy_in_blob(sb, blame_list[j].ent,
1314 norigin, this, &file_p);
1315 copy_split_if_better(sb, blame_list[j].split,
1316 this);
1317 decref_split(this);
1318 }
33494784
JH
1319 origin_decref(norigin);
1320 }
18abd745 1321
aec8fa1f 1322 for (j = 0; j < num_ents; j++) {
33494784
JH
1323 struct blame_entry *split = blame_list[j].split;
1324 if (split[1].suspect &&
1325 blame_copy_score < ent_score(sb, &split[1])) {
7e6ac6e4
DK
1326 split_blame(blamed, &unblamedtail, split,
1327 blame_list[j].ent);
1328 } else {
1329 blame_list[j].ent->next = leftover;
1330 leftover = blame_list[j].ent;
33494784
JH
1331 }
1332 decref_split(split);
18abd745 1333 }
33494784 1334 free(blame_list);
7e6ac6e4
DK
1335 *unblamedtail = NULL;
1336 toosmall = filter_small(sb, toosmall, &unblamed, blame_copy_score);
1337 } while (unblamed);
1338 target->suspects = reverse_blame(leftover, NULL);
33494784 1339 diff_flush(&diff_opts);
bd1928df 1340 free_pathspec(&diff_opts.pathspec);
18abd745
JH
1341}
1342
1732a1fd
JH
1343/*
1344 * The blobs of origin and porigin exactly match, so everything
c2e525d9
JH
1345 * origin is suspected for can be blamed on the parent.
1346 */
1347static void pass_whole_blame(struct scoreboard *sb,
1348 struct origin *origin, struct origin *porigin)
1349{
7e6ac6e4 1350 struct blame_entry *e, *suspects;
c2e525d9
JH
1351
1352 if (!porigin->file.ptr && origin->file.ptr) {
1353 /* Steal its file */
1354 porigin->file = origin->file;
1355 origin->file.ptr = NULL;
1356 }
7e6ac6e4
DK
1357 suspects = origin->suspects;
1358 origin->suspects = NULL;
1359 for (e = suspects; e; e = e->next) {
c2e525d9
JH
1360 origin_incref(porigin);
1361 origin_decref(e->suspect);
1362 e->suspect = porigin;
1363 }
7e6ac6e4 1364 queue_blames(sb, porigin, suspects);
c2e525d9
JH
1365}
1366
69264f46
JH
1367/*
1368 * We pass blame from the current commit to its parents. We keep saying
1369 * "parent" (and "porigin"), but what we mean is to find scapegoat to
1370 * exonerate ourselves.
1371 */
85af7929 1372static struct commit_list *first_scapegoat(struct rev_info *revs, struct commit *commit)
69264f46 1373{
95a4fb0e
JK
1374 if (!reverse) {
1375 if (revs->first_parent_only &&
1376 commit->parents &&
1377 commit->parents->next) {
1378 free_commit_list(commit->parents->next);
1379 commit->parents->next = NULL;
1380 }
85af7929 1381 return commit->parents;
95a4fb0e 1382 }
85af7929 1383 return lookup_decoration(&revs->children, &commit->object);
69264f46
JH
1384}
1385
85af7929 1386static int num_scapegoats(struct rev_info *revs, struct commit *commit)
69264f46 1387{
85af7929 1388 struct commit_list *l = first_scapegoat(revs, commit);
4bbaa1eb 1389 return commit_list_count(l);
69264f46
JH
1390}
1391
7e6ac6e4
DK
1392/* Distribute collected unsorted blames to the respected sorted lists
1393 * in the various origins.
1394 */
1395static void distribute_blame(struct scoreboard *sb, struct blame_entry *blamed)
1396{
1397 blamed = blame_sort(blamed, compare_blame_suspect);
1398 while (blamed)
1399 {
1400 struct origin *porigin = blamed->suspect;
1401 struct blame_entry *suspects = NULL;
1402 do {
1403 struct blame_entry *next = blamed->next;
1404 blamed->next = suspects;
1405 suspects = blamed;
1406 blamed = next;
1407 } while (blamed && blamed->suspect == porigin);
1408 suspects = reverse_blame(suspects, NULL);
1409 queue_blames(sb, porigin, suspects);
1410 }
1411}
1412
69264f46 1413#define MAXSG 16
cee7f245 1414
d24bba80 1415static void pass_blame(struct scoreboard *sb, struct origin *origin, int opt)
cee7f245 1416{
85af7929 1417 struct rev_info *revs = sb->revs;
69264f46 1418 int i, pass, num_sg;
cee7f245 1419 struct commit *commit = origin->commit;
69264f46
JH
1420 struct commit_list *sg;
1421 struct origin *sg_buf[MAXSG];
1422 struct origin *porigin, **sg_origin = sg_buf;
7e6ac6e4
DK
1423 struct blame_entry *toosmall = NULL;
1424 struct blame_entry *blames, **blametail = &blames;
69264f46 1425
85af7929 1426 num_sg = num_scapegoats(revs, commit);
69264f46
JH
1427 if (!num_sg)
1428 goto finish;
1429 else if (num_sg < ARRAY_SIZE(sg_buf))
1430 memset(sg_buf, 0, sizeof(sg_buf));
1431 else
1432 sg_origin = xcalloc(num_sg, sizeof(*sg_origin));
cee7f245 1433
69264f46
JH
1434 /*
1435 * The first pass looks for unrenamed path to optimize for
f69e743d
JH
1436 * common cases, then we look for renames in the second pass.
1437 */
3d1aa566 1438 for (pass = 0; pass < 2 - no_whole_file_rename; pass++) {
f69e743d
JH
1439 struct origin *(*find)(struct scoreboard *,
1440 struct commit *, struct origin *);
1441 find = pass ? find_rename : find_origin;
1442
85af7929 1443 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1444 i < num_sg && sg;
1445 sg = sg->next, i++) {
1446 struct commit *p = sg->item;
0421d9f8 1447 int j, same;
f69e743d 1448
69264f46 1449 if (sg_origin[i])
f69e743d
JH
1450 continue;
1451 if (parse_commit(p))
1452 continue;
0d981c67 1453 porigin = find(sb, p, origin);
f69e743d
JH
1454 if (!porigin)
1455 continue;
1456 if (!hashcmp(porigin->blob_sha1, origin->blob_sha1)) {
c2e525d9 1457 pass_whole_blame(sb, origin, porigin);
f69e743d
JH
1458 origin_decref(porigin);
1459 goto finish;
1460 }
0421d9f8 1461 for (j = same = 0; j < i; j++)
69264f46
JH
1462 if (sg_origin[j] &&
1463 !hashcmp(sg_origin[j]->blob_sha1,
0421d9f8
JH
1464 porigin->blob_sha1)) {
1465 same = 1;
1466 break;
1467 }
1468 if (!same)
69264f46 1469 sg_origin[i] = porigin;
0421d9f8
JH
1470 else
1471 origin_decref(porigin);
cee7f245 1472 }
cee7f245
JH
1473 }
1474
c2e525d9 1475 num_commits++;
85af7929 1476 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1477 i < num_sg && sg;
1478 sg = sg->next, i++) {
1479 struct origin *porigin = sg_origin[i];
cee7f245
JH
1480 if (!porigin)
1481 continue;
96e11709
JH
1482 if (!origin->previous) {
1483 origin_incref(porigin);
1484 origin->previous = porigin;
1485 }
7e6ac6e4
DK
1486 pass_blame_to_parent(sb, origin, porigin);
1487 if (!origin->suspects)
54a4c617 1488 goto finish;
cee7f245 1489 }
d24bba80
JH
1490
1491 /*
1732a1fd 1492 * Optionally find moves in parents' files.
d24bba80 1493 */
7e6ac6e4
DK
1494 if (opt & PICKAXE_BLAME_MOVE) {
1495 filter_small(sb, &toosmall, &origin->suspects, blame_move_score);
1496 if (origin->suspects) {
1497 for (i = 0, sg = first_scapegoat(revs, commit);
1498 i < num_sg && sg;
1499 sg = sg->next, i++) {
1500 struct origin *porigin = sg_origin[i];
1501 if (!porigin)
1502 continue;
1503 find_move_in_parent(sb, &blametail, &toosmall, origin, porigin);
1504 if (!origin->suspects)
1505 break;
1506 }
d24bba80 1507 }
7e6ac6e4 1508 }
d24bba80 1509
18abd745 1510 /*
1732a1fd 1511 * Optionally find copies from parents' files.
18abd745 1512 */
7e6ac6e4
DK
1513 if (opt & PICKAXE_BLAME_COPY) {
1514 if (blame_copy_score > blame_move_score)
1515 filter_small(sb, &toosmall, &origin->suspects, blame_copy_score);
1516 else if (blame_copy_score < blame_move_score) {
1517 origin->suspects = blame_merge(origin->suspects, toosmall);
1518 toosmall = NULL;
1519 filter_small(sb, &toosmall, &origin->suspects, blame_copy_score);
1520 }
1521 if (!origin->suspects)
1522 goto finish;
1523
85af7929 1524 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1525 i < num_sg && sg;
1526 sg = sg->next, i++) {
1527 struct origin *porigin = sg_origin[i];
7e6ac6e4
DK
1528 find_copy_in_parent(sb, &blametail, &toosmall,
1529 origin, sg->item, porigin, opt);
1530 if (!origin->suspects)
54a4c617 1531 goto finish;
18abd745 1532 }
7e6ac6e4 1533 }
54a4c617 1534
7e6ac6e4
DK
1535finish:
1536 *blametail = NULL;
1537 distribute_blame(sb, blames);
1538 /*
1539 * prepend toosmall to origin->suspects
1540 *
1541 * There is no point in sorting: this ends up on a big
1542 * unsorted list in the caller anyway.
1543 */
1544 if (toosmall) {
1545 struct blame_entry **tail = &toosmall;
1546 while (*tail)
1547 tail = &(*tail)->next;
1548 *tail = origin->suspects;
1549 origin->suspects = toosmall;
1550 }
69264f46
JH
1551 for (i = 0; i < num_sg; i++) {
1552 if (sg_origin[i]) {
1553 drop_origin_blob(sg_origin[i]);
1554 origin_decref(sg_origin[i]);
7c3c7962
JH
1555 }
1556 }
1557 drop_origin_blob(origin);
69264f46
JH
1558 if (sg_buf != sg_origin)
1559 free(sg_origin);
cee7f245
JH
1560}
1561
1732a1fd
JH
1562/*
1563 * Information on commits, used for output.
1564 */
9cba13ca 1565struct commit_info {
ea02ffa3
AP
1566 struct strbuf author;
1567 struct strbuf author_mail;
cee7f245 1568 unsigned long author_time;
ea02ffa3 1569 struct strbuf author_tz;
cee7f245
JH
1570
1571 /* filled only when asked for details */
ea02ffa3
AP
1572 struct strbuf committer;
1573 struct strbuf committer_mail;
cee7f245 1574 unsigned long committer_time;
ea02ffa3 1575 struct strbuf committer_tz;
cee7f245 1576
ea02ffa3 1577 struct strbuf summary;
cee7f245
JH
1578};
1579
1732a1fd
JH
1580/*
1581 * Parse author/committer line in the commit object buffer
1582 */
cee7f245 1583static void get_ac_line(const char *inbuf, const char *what,
ea02ffa3
AP
1584 struct strbuf *name, struct strbuf *mail,
1585 unsigned long *time, struct strbuf *tz)
cee7f245 1586{
3c020bd5 1587 struct ident_split ident;
ea02ffa3
AP
1588 size_t len, maillen, namelen;
1589 char *tmp, *endp;
1590 const char *namebuf, *mailbuf;
cee7f245
JH
1591
1592 tmp = strstr(inbuf, what);
1593 if (!tmp)
1594 goto error_out;
1595 tmp += strlen(what);
1596 endp = strchr(tmp, '\n');
1597 if (!endp)
1598 len = strlen(tmp);
1599 else
1600 len = endp - tmp;
3c020bd5
AP
1601
1602 if (split_ident_line(&ident, tmp, len)) {
cee7f245
JH
1603 error_out:
1604 /* Ugh */
ea02ffa3
AP
1605 tmp = "(unknown)";
1606 strbuf_addstr(name, tmp);
1607 strbuf_addstr(mail, tmp);
1608 strbuf_addstr(tz, tmp);
cee7f245
JH
1609 *time = 0;
1610 return;
1611 }
cee7f245 1612
3c020bd5 1613 namelen = ident.name_end - ident.name_begin;
ea02ffa3 1614 namebuf = ident.name_begin;
cee7f245 1615
ea02ffa3
AP
1616 maillen = ident.mail_end - ident.mail_begin;
1617 mailbuf = ident.mail_begin;
f9567384 1618
de5abe9f
RS
1619 if (ident.date_begin && ident.date_end)
1620 *time = strtoul(ident.date_begin, NULL, 10);
1621 else
1622 *time = 0;
f9567384 1623
de5abe9f
RS
1624 if (ident.tz_begin && ident.tz_end)
1625 strbuf_add(tz, ident.tz_begin, ident.tz_end - ident.tz_begin);
1626 else
1627 strbuf_addstr(tz, "(unknown)");
f9567384 1628
f9567384 1629 /*
d20d654f 1630 * Now, convert both name and e-mail using mailmap
f9567384 1631 */
ea02ffa3
AP
1632 map_user(&mailmap, &mailbuf, &maillen,
1633 &namebuf, &namelen);
1634
1635 strbuf_addf(mail, "<%.*s>", (int)maillen, mailbuf);
1636 strbuf_add(name, namebuf, namelen);
1637}
1638
1639static void commit_info_init(struct commit_info *ci)
1640{
1641
1642 strbuf_init(&ci->author, 0);
1643 strbuf_init(&ci->author_mail, 0);
1644 strbuf_init(&ci->author_tz, 0);
1645 strbuf_init(&ci->committer, 0);
1646 strbuf_init(&ci->committer_mail, 0);
1647 strbuf_init(&ci->committer_tz, 0);
1648 strbuf_init(&ci->summary, 0);
1649}
1650
1651static void commit_info_destroy(struct commit_info *ci)
1652{
1653
1654 strbuf_release(&ci->author);
1655 strbuf_release(&ci->author_mail);
1656 strbuf_release(&ci->author_tz);
1657 strbuf_release(&ci->committer);
1658 strbuf_release(&ci->committer_mail);
1659 strbuf_release(&ci->committer_tz);
1660 strbuf_release(&ci->summary);
cee7f245
JH
1661}
1662
1663static void get_commit_info(struct commit *commit,
1664 struct commit_info *ret,
1665 int detailed)
1666{
1667 int len;
e297cf5a 1668 const char *subject, *encoding;
b000c59b 1669 const char *message;
ea02ffa3
AP
1670
1671 commit_info_init(ret);
cee7f245 1672
e297cf5a 1673 encoding = get_log_output_encoding();
5a10d236 1674 message = logmsg_reencode(commit, NULL, encoding);
69cd8f63 1675 get_ac_line(message, "\nauthor ",
ea02ffa3 1676 &ret->author, &ret->author_mail,
cee7f245
JH
1677 &ret->author_time, &ret->author_tz);
1678
69cd8f63 1679 if (!detailed) {
b66103c3 1680 unuse_commit_buffer(commit, message);
cee7f245 1681 return;
69cd8f63 1682 }
cee7f245 1683
69cd8f63 1684 get_ac_line(message, "\ncommitter ",
ea02ffa3 1685 &ret->committer, &ret->committer_mail,
cee7f245
JH
1686 &ret->committer_time, &ret->committer_tz);
1687
ad98a58b 1688 len = find_commit_subject(message, &subject);
ea02ffa3
AP
1689 if (len)
1690 strbuf_add(&ret->summary, subject, len);
1691 else
1692 strbuf_addf(&ret->summary, "(%s)", sha1_to_hex(commit->object.sha1));
1693
b66103c3 1694 unuse_commit_buffer(commit, message);
cee7f245
JH
1695}
1696
1732a1fd
JH
1697/*
1698 * To allow LF and other nonportable characters in pathnames,
1699 * they are c-style quoted as needed.
1700 */
46e5e69d
JH
1701static void write_filename_info(const char *path)
1702{
1703 printf("filename ");
663af342 1704 write_name_quoted(path, stdout, '\n');
46e5e69d
JH
1705}
1706
9991030c
JH
1707/*
1708 * Porcelain/Incremental format wants to show a lot of details per
1709 * commit. Instead of repeating this every line, emit it only once,
e86226e3
JK
1710 * the first time each commit appears in the output (unless the
1711 * user has specifically asked for us to repeat).
9991030c 1712 */
e86226e3 1713static int emit_one_suspect_detail(struct origin *suspect, int repeat)
9991030c
JH
1714{
1715 struct commit_info ci;
1716
e86226e3 1717 if (!repeat && (suspect->commit->object.flags & METAINFO_SHOWN))
9991030c
JH
1718 return 0;
1719
1720 suspect->commit->object.flags |= METAINFO_SHOWN;
1721 get_commit_info(suspect->commit, &ci, 1);
ea02ffa3
AP
1722 printf("author %s\n", ci.author.buf);
1723 printf("author-mail %s\n", ci.author_mail.buf);
9991030c 1724 printf("author-time %lu\n", ci.author_time);
ea02ffa3
AP
1725 printf("author-tz %s\n", ci.author_tz.buf);
1726 printf("committer %s\n", ci.committer.buf);
1727 printf("committer-mail %s\n", ci.committer_mail.buf);
9991030c 1728 printf("committer-time %lu\n", ci.committer_time);
ea02ffa3
AP
1729 printf("committer-tz %s\n", ci.committer_tz.buf);
1730 printf("summary %s\n", ci.summary.buf);
9991030c
JH
1731 if (suspect->commit->object.flags & UNINTERESTING)
1732 printf("boundary\n");
96e11709
JH
1733 if (suspect->previous) {
1734 struct origin *prev = suspect->previous;
1735 printf("previous %s ", sha1_to_hex(prev->commit->object.sha1));
1736 write_name_quoted(prev->path, stdout, '\n');
1737 }
ea02ffa3
AP
1738
1739 commit_info_destroy(&ci);
1740
9991030c
JH
1741 return 1;
1742}
1743
1732a1fd 1744/*
7e6ac6e4
DK
1745 * The blame_entry is found to be guilty for the range.
1746 * Show it in incremental output.
1732a1fd 1747 */
717d1462
LT
1748static void found_guilty_entry(struct blame_entry *ent)
1749{
717d1462
LT
1750 if (incremental) {
1751 struct origin *suspect = ent->suspect;
1752
1753 printf("%s %d %d %d\n",
1754 sha1_to_hex(suspect->commit->object.sha1),
1755 ent->s_lno + 1, ent->lno + 1, ent->num_lines);
e86226e3 1756 emit_one_suspect_detail(suspect, 0);
46e5e69d 1757 write_filename_info(suspect->path);
06f59e9f 1758 maybe_flush_or_die(stdout, "stdout");
717d1462
LT
1759 }
1760}
1761
1732a1fd 1762/*
7e6ac6e4
DK
1763 * The main loop -- while we have blobs with lines whose true origin
1764 * is still unknown, pick one blob, and allow its lines to pass blames
1765 * to its parents. */
85af7929 1766static void assign_blame(struct scoreboard *sb, int opt)
717d1462 1767{
85af7929 1768 struct rev_info *revs = sb->revs;
7e6ac6e4 1769 struct commit *commit = prio_queue_get(&sb->commits);
85af7929 1770
7e6ac6e4 1771 while (commit) {
717d1462 1772 struct blame_entry *ent;
7e6ac6e4 1773 struct origin *suspect = commit->util;
717d1462
LT
1774
1775 /* find one suspect to break down */
7e6ac6e4
DK
1776 while (suspect && !suspect->suspects)
1777 suspect = suspect->next;
1778
1779 if (!suspect) {
1780 commit = prio_queue_get(&sb->commits);
1781 continue;
1782 }
1783
1784 assert(commit == suspect->commit);
717d1462 1785
1732a1fd
JH
1786 /*
1787 * We will use this suspect later in the loop,
1788 * so hold onto it in the meantime.
1789 */
717d1462 1790 origin_incref(suspect);
0064053b 1791 parse_commit(commit);
85af7929
JH
1792 if (reverse ||
1793 (!(commit->object.flags & UNINTERESTING) &&
1794 !(revs->max_age != -1 && commit->date < revs->max_age)))
717d1462
LT
1795 pass_blame(sb, suspect, opt);
1796 else {
1797 commit->object.flags |= UNINTERESTING;
1798 if (commit->object.parsed)
1799 mark_parents_uninteresting(commit);
1800 }
1801 /* treat root commit as boundary */
1802 if (!commit->parents && !show_root)
1803 commit->object.flags |= UNINTERESTING;
1804
1805 /* Take responsibility for the remaining entries */
7e6ac6e4
DK
1806 ent = suspect->suspects;
1807 if (ent) {
1808 suspect->guilty = 1;
1809 for (;;) {
1810 struct blame_entry *next = ent->next;
717d1462 1811 found_guilty_entry(ent);
7e6ac6e4
DK
1812 if (next) {
1813 ent = next;
1814 continue;
1815 }
1816 ent->next = sb->ent;
1817 sb->ent = suspect->suspects;
1818 suspect->suspects = NULL;
1819 break;
1820 }
1821 }
717d1462
LT
1822 origin_decref(suspect);
1823
1824 if (DEBUG) /* sanity */
1825 sanity_check_refcnt(sb);
1826 }
1827}
1828
1829static const char *format_time(unsigned long time, const char *tz_str,
1830 int show_raw_time)
1831{
bccce0f8 1832 static struct strbuf time_buf = STRBUF_INIT;
717d1462 1833
bccce0f8 1834 strbuf_reset(&time_buf);
717d1462 1835 if (show_raw_time) {
bccce0f8 1836 strbuf_addf(&time_buf, "%lu %s", time, tz_str);
717d1462 1837 }
31653c1a 1838 else {
ac39b277 1839 const char *time_str;
bccce0f8 1840 size_t time_width;
ac39b277 1841 int tz;
31653c1a 1842 tz = atoi(tz_str);
a5481a6c 1843 time_str = show_date(time, tz, &blame_date_mode);
bccce0f8
JX
1844 strbuf_addstr(&time_buf, time_str);
1845 /*
1846 * Add space paddings to time_buf to display a fixed width
1847 * string, and use time_width for display width calibration.
1848 */
1849 for (time_width = utf8_strwidth(time_str);
1850 time_width < blame_date_width;
1851 time_width++)
1852 strbuf_addch(&time_buf, ' ');
31653c1a 1853 }
bccce0f8 1854 return time_buf.buf;
717d1462
LT
1855}
1856
cee7f245
JH
1857#define OUTPUT_ANNOTATE_COMPAT 001
1858#define OUTPUT_LONG_OBJECT_NAME 002
1859#define OUTPUT_RAW_TIMESTAMP 004
1860#define OUTPUT_PORCELAIN 010
1861#define OUTPUT_SHOW_NAME 020
1862#define OUTPUT_SHOW_NUMBER 040
5ff62c30 1863#define OUTPUT_SHOW_SCORE 0100
093dc5be 1864#define OUTPUT_NO_AUTHOR 0200
1b8cdce9 1865#define OUTPUT_SHOW_EMAIL 0400
ed747dd5 1866#define OUTPUT_LINE_PORCELAIN 01000
cee7f245 1867
e86226e3
JK
1868static void emit_porcelain_details(struct origin *suspect, int repeat)
1869{
1870 if (emit_one_suspect_detail(suspect, repeat) ||
1871 (suspect->commit->object.flags & MORE_THAN_ONE_PATH))
1872 write_filename_info(suspect->path);
1873}
1874
1875static void emit_porcelain(struct scoreboard *sb, struct blame_entry *ent,
1876 int opt)
cee7f245 1877{
ed747dd5 1878 int repeat = opt & OUTPUT_LINE_PORCELAIN;
cee7f245
JH
1879 int cnt;
1880 const char *cp;
1881 struct origin *suspect = ent->suspect;
1882 char hex[41];
1883
1884 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
7e6ac6e4 1885 printf("%s %d %d %d\n",
cee7f245 1886 hex,
cee7f245
JH
1887 ent->s_lno + 1,
1888 ent->lno + 1,
1889 ent->num_lines);
ed747dd5 1890 emit_porcelain_details(suspect, repeat);
cee7f245
JH
1891
1892 cp = nth_line(sb, ent->lno);
1893 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1894 char ch;
ed747dd5 1895 if (cnt) {
cee7f245
JH
1896 printf("%s %d %d\n", hex,
1897 ent->s_lno + 1 + cnt,
1898 ent->lno + 1 + cnt);
ed747dd5
JK
1899 if (repeat)
1900 emit_porcelain_details(suspect, 1);
1901 }
cee7f245
JH
1902 putchar('\t');
1903 do {
1904 ch = *cp++;
1905 putchar(ch);
1906 } while (ch != '\n' &&
1907 cp < sb->final_buf + sb->final_buf_size);
1908 }
a5ca8367
JS
1909
1910 if (sb->final_buf_size && cp[-1] != '\n')
1911 putchar('\n');
cee7f245
JH
1912}
1913
1914static void emit_other(struct scoreboard *sb, struct blame_entry *ent, int opt)
1915{
1916 int cnt;
1917 const char *cp;
1918 struct origin *suspect = ent->suspect;
1919 struct commit_info ci;
1920 char hex[41];
1921 int show_raw_time = !!(opt & OUTPUT_RAW_TIMESTAMP);
1922
1923 get_commit_info(suspect->commit, &ci, 1);
1924 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1925
1926 cp = nth_line(sb, ent->lno);
1927 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1928 char ch;
84393bfd 1929 int length = (opt & OUTPUT_LONG_OBJECT_NAME) ? 40 : abbrev;
b11121d9
JH
1930
1931 if (suspect->commit->object.flags & UNINTERESTING) {
e68989a7
JH
1932 if (blank_boundary)
1933 memset(hex, ' ', length);
7ceacdff 1934 else if (!(opt & OUTPUT_ANNOTATE_COMPAT)) {
4c10a5ca
JH
1935 length--;
1936 putchar('^');
1937 }
b11121d9 1938 }
cee7f245 1939
b11121d9 1940 printf("%.*s", length, hex);
1b8cdce9
KB
1941 if (opt & OUTPUT_ANNOTATE_COMPAT) {
1942 const char *name;
1943 if (opt & OUTPUT_SHOW_EMAIL)
ea02ffa3 1944 name = ci.author_mail.buf;
1b8cdce9 1945 else
ea02ffa3 1946 name = ci.author.buf;
1b8cdce9 1947 printf("\t(%10s\t%10s\t%d)", name,
ea02ffa3 1948 format_time(ci.author_time, ci.author_tz.buf,
cee7f245
JH
1949 show_raw_time),
1950 ent->lno + 1 + cnt);
1b8cdce9 1951 } else {
5ff62c30 1952 if (opt & OUTPUT_SHOW_SCORE)
54a4c617
JH
1953 printf(" %*d %02d",
1954 max_score_digits, ent->score,
1955 ent->suspect->refcnt);
cee7f245
JH
1956 if (opt & OUTPUT_SHOW_NAME)
1957 printf(" %-*.*s", longest_file, longest_file,
1958 suspect->path);
1959 if (opt & OUTPUT_SHOW_NUMBER)
1960 printf(" %*d", max_orig_digits,
1961 ent->s_lno + 1 + cnt);
093dc5be 1962
ffaf9cc0 1963 if (!(opt & OUTPUT_NO_AUTHOR)) {
1b8cdce9
KB
1964 const char *name;
1965 int pad;
1966 if (opt & OUTPUT_SHOW_EMAIL)
ea02ffa3 1967 name = ci.author_mail.buf;
1b8cdce9 1968 else
ea02ffa3 1969 name = ci.author.buf;
1b8cdce9 1970 pad = longest_author - utf8_strwidth(name);
ffaf9cc0 1971 printf(" (%s%*s %10s",
1b8cdce9 1972 name, pad, "",
093dc5be 1973 format_time(ci.author_time,
ea02ffa3 1974 ci.author_tz.buf,
093dc5be 1975 show_raw_time));
ffaf9cc0 1976 }
093dc5be 1977 printf(" %*d) ",
cee7f245
JH
1978 max_digits, ent->lno + 1 + cnt);
1979 }
1980 do {
1981 ch = *cp++;
1982 putchar(ch);
1983 } while (ch != '\n' &&
1984 cp < sb->final_buf + sb->final_buf_size);
1985 }
a5ca8367
JS
1986
1987 if (sb->final_buf_size && cp[-1] != '\n')
1988 putchar('\n');
ea02ffa3
AP
1989
1990 commit_info_destroy(&ci);
cee7f245
JH
1991}
1992
1993static void output(struct scoreboard *sb, int option)
1994{
1995 struct blame_entry *ent;
1996
1997 if (option & OUTPUT_PORCELAIN) {
1998 for (ent = sb->ent; ent; ent = ent->next) {
7e6ac6e4
DK
1999 int count = 0;
2000 struct origin *suspect;
2001 struct commit *commit = ent->suspect->commit;
cee7f245
JH
2002 if (commit->object.flags & MORE_THAN_ONE_PATH)
2003 continue;
7e6ac6e4
DK
2004 for (suspect = commit->util; suspect; suspect = suspect->next) {
2005 if (suspect->guilty && count++) {
2006 commit->object.flags |= MORE_THAN_ONE_PATH;
2007 break;
2008 }
cee7f245
JH
2009 }
2010 }
2011 }
2012
2013 for (ent = sb->ent; ent; ent = ent->next) {
2014 if (option & OUTPUT_PORCELAIN)
e86226e3 2015 emit_porcelain(sb, ent, option);
5ff62c30 2016 else {
cee7f245 2017 emit_other(sb, ent, option);
5ff62c30 2018 }
cee7f245
JH
2019 }
2020}
2021
29aa0b20
RS
2022static const char *get_next_line(const char *start, const char *end)
2023{
2024 const char *nl = memchr(start, '\n', end - start);
60d85e11 2025 return nl ? nl + 1 : end;
29aa0b20
RS
2026}
2027
1732a1fd
JH
2028/*
2029 * To allow quick access to the contents of nth line in the
2030 * final image, prepare an index in the scoreboard.
2031 */
cee7f245
JH
2032static int prepare_lines(struct scoreboard *sb)
2033{
2034 const char *buf = sb->final_buf;
2035 unsigned long len = sb->final_buf_size;
352bbbd9
DK
2036 const char *end = buf + len;
2037 const char *p;
2038 int *lineno;
60d85e11 2039 int num = 0;
352bbbd9 2040
60d85e11 2041 for (p = buf; p < end; p = get_next_line(p, end))
29aa0b20 2042 num++;
352bbbd9 2043
60d85e11 2044 sb->lineno = lineno = xmalloc(sizeof(*sb->lineno) * (num + 1));
352bbbd9 2045
60d85e11 2046 for (p = buf; p < end; p = get_next_line(p, end))
29aa0b20 2047 *lineno++ = p - buf;
352bbbd9 2048
60d85e11 2049 *lineno = len;
352bbbd9 2050
60d85e11 2051 sb->num_lines = num;
cee7f245
JH
2052 return sb->num_lines;
2053}
2054
1732a1fd
JH
2055/*
2056 * Add phony grafts for use with -S; this is primarily to
34baebce 2057 * support git's cvsserver that wants to give a linear history
1732a1fd
JH
2058 * to its clients.
2059 */
cee7f245
JH
2060static int read_ancestry(const char *graft_file)
2061{
2062 FILE *fp = fopen(graft_file, "r");
e228c173 2063 struct strbuf buf = STRBUF_INIT;
cee7f245
JH
2064 if (!fp)
2065 return -1;
e228c173 2066 while (!strbuf_getwholeline(&buf, fp, '\n')) {
cee7f245 2067 /* The format is just "Commit Parent1 Parent2 ...\n" */
e228c173 2068 struct commit_graft *graft = read_graft_line(buf.buf, buf.len);
8eaf7986
JH
2069 if (graft)
2070 register_commit_graft(graft, 0);
cee7f245
JH
2071 }
2072 fclose(fp);
e228c173 2073 strbuf_release(&buf);
cee7f245
JH
2074 return 0;
2075}
2076
b31272f7
JH
2077static int update_auto_abbrev(int auto_abbrev, struct origin *suspect)
2078{
2079 const char *uniq = find_unique_abbrev(suspect->commit->object.sha1,
2080 auto_abbrev);
2081 int len = strlen(uniq);
2082 if (auto_abbrev < len)
2083 return len;
2084 return auto_abbrev;
2085}
2086
1732a1fd
JH
2087/*
2088 * How many columns do we need to show line numbers, authors,
2089 * and filenames?
2090 */
cee7f245
JH
2091static void find_alignment(struct scoreboard *sb, int *option)
2092{
2093 int longest_src_lines = 0;
2094 int longest_dst_lines = 0;
5ff62c30 2095 unsigned largest_score = 0;
cee7f245 2096 struct blame_entry *e;
b31272f7
JH
2097 int compute_auto_abbrev = (abbrev < 0);
2098 int auto_abbrev = default_abbrev;
cee7f245
JH
2099
2100 for (e = sb->ent; e; e = e->next) {
2101 struct origin *suspect = e->suspect;
cee7f245
JH
2102 int num;
2103
b31272f7
JH
2104 if (compute_auto_abbrev)
2105 auto_abbrev = update_auto_abbrev(auto_abbrev, suspect);
ab3bb800
JH
2106 if (strcmp(suspect->path, sb->path))
2107 *option |= OUTPUT_SHOW_NAME;
2108 num = strlen(suspect->path);
2109 if (longest_file < num)
2110 longest_file = num;
cee7f245 2111 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
e6005927 2112 struct commit_info ci;
cee7f245
JH
2113 suspect->commit->object.flags |= METAINFO_SHOWN;
2114 get_commit_info(suspect->commit, &ci, 1);
1b8cdce9 2115 if (*option & OUTPUT_SHOW_EMAIL)
ea02ffa3 2116 num = utf8_strwidth(ci.author_mail.buf);
1b8cdce9 2117 else
ea02ffa3 2118 num = utf8_strwidth(ci.author.buf);
cee7f245
JH
2119 if (longest_author < num)
2120 longest_author = num;
e6005927 2121 commit_info_destroy(&ci);
cee7f245
JH
2122 }
2123 num = e->s_lno + e->num_lines;
2124 if (longest_src_lines < num)
2125 longest_src_lines = num;
2126 num = e->lno + e->num_lines;
2127 if (longest_dst_lines < num)
2128 longest_dst_lines = num;
5ff62c30
JH
2129 if (largest_score < ent_score(sb, e))
2130 largest_score = ent_score(sb, e);
cee7f245 2131 }
ec7ff5ba
ZJS
2132 max_orig_digits = decimal_width(longest_src_lines);
2133 max_digits = decimal_width(longest_dst_lines);
2134 max_score_digits = decimal_width(largest_score);
b31272f7
JH
2135
2136 if (compute_auto_abbrev)
2137 /* one more abbrev length is needed for the boundary commit */
2138 abbrev = auto_abbrev + 1;
cee7f245
JH
2139}
2140
1732a1fd
JH
2141/*
2142 * For debugging -- origin is refcounted, and this asserts that
2143 * we do not underflow.
2144 */
54a4c617
JH
2145static void sanity_check_refcnt(struct scoreboard *sb)
2146{
2147 int baa = 0;
2148 struct blame_entry *ent;
2149
2150 for (ent = sb->ent; ent; ent = ent->next) {
ae86ad65 2151 /* Nobody should have zero or negative refcnt */
854b97f6
JH
2152 if (ent->suspect->refcnt <= 0) {
2153 fprintf(stderr, "%s in %s has negative refcnt %d\n",
2154 ent->suspect->path,
2155 sha1_to_hex(ent->suspect->commit->object.sha1),
2156 ent->suspect->refcnt);
ae86ad65 2157 baa = 1;
854b97f6 2158 }
ae86ad65 2159 }
54a4c617
JH
2160 if (baa) {
2161 int opt = 0160;
2162 find_alignment(sb, &opt);
2163 output(sb, opt);
854b97f6 2164 die("Baa %d!", baa);
54a4c617
JH
2165 }
2166}
2167
4a0fc95f
JH
2168static unsigned parse_score(const char *arg)
2169{
2170 char *end;
2171 unsigned long score = strtoul(arg, &end, 10);
2172 if (*end)
2173 return 0;
2174 return score;
2175}
2176
20239bae
JK
2177static const char *add_prefix(const char *prefix, const char *path)
2178{
097971f5 2179 return prefix_path(prefix, prefix ? strlen(prefix) : 0, path);
20239bae
JK
2180}
2181
ef90d6d4 2182static int git_blame_config(const char *var, const char *value, void *cb)
4c10a5ca
JH
2183{
2184 if (!strcmp(var, "blame.showroot")) {
2185 show_root = git_config_bool(var, value);
2186 return 0;
2187 }
2188 if (!strcmp(var, "blame.blankboundary")) {
2189 blank_boundary = git_config_bool(var, value);
2190 return 0;
2191 }
8b504db3
QN
2192 if (!strcmp(var, "blame.showemail")) {
2193 int *output_option = cb;
2194 if (git_config_bool(var, value))
2195 *output_option |= OUTPUT_SHOW_EMAIL;
2196 else
2197 *output_option &= ~OUTPUT_SHOW_EMAIL;
2198 return 0;
2199 }
31653c1a
EL
2200 if (!strcmp(var, "blame.date")) {
2201 if (!value)
2202 return config_error_nonbool(var);
a5481a6c 2203 parse_date_format(value, &blame_date_mode);
31653c1a
EL
2204 return 0;
2205 }
3b8a12e8 2206
6680a087 2207 if (userdiff_config(var, value) < 0)
3b8a12e8 2208 return -1;
3b8a12e8 2209
ef90d6d4 2210 return git_default_config(var, value, cb);
4c10a5ca
JH
2211}
2212
9aeaab68 2213static void verify_working_tree_path(struct commit *work_tree, const char *path)
ffcabccf 2214{
9aeaab68 2215 struct commit_list *parents;
ffcabccf 2216
9aeaab68
JH
2217 for (parents = work_tree->parents; parents; parents = parents->next) {
2218 const unsigned char *commit_sha1 = parents->item->object.sha1;
2219 unsigned char blob_sha1[20];
2220 unsigned mode;
2221
2222 if (!get_tree_entry(commit_sha1, path, blob_sha1, &mode) &&
2223 sha1_object_info(blob_sha1, NULL) == OBJ_BLOB)
2224 return;
2225 }
2226 die("no such path '%s' in HEAD", path);
2227}
2228
2229static struct commit_list **append_parent(struct commit_list **tail, const unsigned char *sha1)
2230{
2231 struct commit *parent;
2232
2233 parent = lookup_commit_reference(sha1);
2234 if (!parent)
2235 die("no such commit %s", sha1_to_hex(sha1));
2236 return &commit_list_insert(parent, tail)->next;
2237}
2238
2239static void append_merge_parents(struct commit_list **tail)
2240{
2241 int merge_head;
9aeaab68
JH
2242 struct strbuf line = STRBUF_INIT;
2243
f932729c 2244 merge_head = open(git_path_merge_head(), O_RDONLY);
9aeaab68
JH
2245 if (merge_head < 0) {
2246 if (errno == ENOENT)
2247 return;
f932729c 2248 die("cannot open '%s' for reading", git_path_merge_head());
9aeaab68
JH
2249 }
2250
2251 while (!strbuf_getwholeline_fd(&line, merge_head, '\n')) {
2252 unsigned char sha1[20];
2253 if (line.len < 40 || get_sha1_hex(line.buf, sha1))
f932729c 2254 die("unknown line in '%s': %s", git_path_merge_head(), line.buf);
9aeaab68
JH
2255 tail = append_parent(tail, sha1);
2256 }
2257 close(merge_head);
2258 strbuf_release(&line);
ffcabccf
JH
2259}
2260
8597ea3a
JK
2261/*
2262 * This isn't as simple as passing sb->buf and sb->len, because we
2263 * want to transfer ownership of the buffer to the commit (so we
2264 * must use detach).
2265 */
2266static void set_commit_buffer_from_strbuf(struct commit *c, struct strbuf *sb)
2267{
2268 size_t len;
2269 void *buf = strbuf_detach(sb, &len);
2270 set_commit_buffer(c, buf, len);
2271}
2272
f6c07d7d
JH
2273/*
2274 * Prepare a dummy commit that represents the work tree (or staged) item.
2275 * Note that annotating work tree item never works in the reverse.
2276 */
3b8a12e8
AB
2277static struct commit *fake_working_tree_commit(struct diff_options *opt,
2278 const char *path,
2279 const char *contents_from)
1cfe7733
JH
2280{
2281 struct commit *commit;
2282 struct origin *origin;
9aeaab68 2283 struct commit_list **parent_tail, *parent;
1cfe7733 2284 unsigned char head_sha1[20];
f285a2d7 2285 struct strbuf buf = STRBUF_INIT;
1cfe7733 2286 const char *ident;
1cfe7733 2287 time_t now;
1cfe7733
JH
2288 int size, len;
2289 struct cache_entry *ce;
2290 unsigned mode;
9aeaab68 2291 struct strbuf msg = STRBUF_INIT;
1cfe7733
JH
2292
2293 time(&now);
10322a0a 2294 commit = alloc_commit_node();
1cfe7733
JH
2295 commit->object.parsed = 1;
2296 commit->date = now;
9aeaab68
JH
2297 parent_tail = &commit->parents;
2298
7695d118 2299 if (!resolve_ref_unsafe("HEAD", RESOLVE_REF_READING, head_sha1, NULL))
9aeaab68
JH
2300 die("no such ref: HEAD");
2301
2302 parent_tail = append_parent(parent_tail, head_sha1);
2303 append_merge_parents(parent_tail);
2304 verify_working_tree_path(commit, path);
1cfe7733
JH
2305
2306 origin = make_origin(commit, path);
2307
9aeaab68
JH
2308 ident = fmt_ident("Not Committed Yet", "not.committed.yet", NULL, 0);
2309 strbuf_addstr(&msg, "tree 0000000000000000000000000000000000000000\n");
2310 for (parent = commit->parents; parent; parent = parent->next)
2311 strbuf_addf(&msg, "parent %s\n",
2312 sha1_to_hex(parent->item->object.sha1));
2313 strbuf_addf(&msg,
2314 "author %s\n"
2315 "committer %s\n\n"
2316 "Version of %s from %s\n",
2317 ident, ident, path,
2318 (!contents_from ? path :
2319 (!strcmp(contents_from, "-") ? "standard input" : contents_from)));
8597ea3a 2320 set_commit_buffer_from_strbuf(commit, &msg);
9aeaab68 2321
1cfe7733
JH
2322 if (!contents_from || strcmp("-", contents_from)) {
2323 struct stat st;
2324 const char *read_from;
8518088f 2325 char *buf_ptr;
3b8a12e8 2326 unsigned long buf_len;
1cfe7733
JH
2327
2328 if (contents_from) {
2329 if (stat(contents_from, &st) < 0)
0721c314 2330 die_errno("Cannot stat '%s'", contents_from);
1cfe7733
JH
2331 read_from = contents_from;
2332 }
2333 else {
2334 if (lstat(path, &st) < 0)
0721c314 2335 die_errno("Cannot lstat '%s'", path);
1cfe7733
JH
2336 read_from = path;
2337 }
1cfe7733 2338 mode = canon_mode(st.st_mode);
3b8a12e8 2339
1cfe7733
JH
2340 switch (st.st_mode & S_IFMT) {
2341 case S_IFREG:
3b8a12e8 2342 if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) &&
e5450100 2343 textconv_object(read_from, mode, null_sha1, 0, &buf_ptr, &buf_len))
8518088f 2344 strbuf_attach(&buf, buf_ptr, buf_len, buf_len + 1);
3b8a12e8 2345 else if (strbuf_read_file(&buf, read_from, st.st_size) != st.st_size)
0721c314 2346 die_errno("cannot open or read '%s'", read_from);
1cfe7733
JH
2347 break;
2348 case S_IFLNK:
edfd45d2 2349 if (strbuf_readlink(&buf, read_from, st.st_size) < 0)
0721c314 2350 die_errno("cannot readlink '%s'", read_from);
1cfe7733
JH
2351 break;
2352 default:
2353 die("unsupported file type %s", read_from);
2354 }
2355 }
2356 else {
2357 /* Reading from stdin */
1cfe7733 2358 mode = 0;
f1696ee3 2359 if (strbuf_read(&buf, 0, 0) < 0)
d824cbba 2360 die_errno("failed to read from stdin");
1cfe7733 2361 }
4bf256d6 2362 convert_to_git(path, buf.buf, buf.len, &buf, 0);
af6eb822
PH
2363 origin->file.ptr = buf.buf;
2364 origin->file.size = buf.len;
2365 pretend_sha1_file(buf.buf, buf.len, OBJ_BLOB, origin->blob_sha1);
1cfe7733
JH
2366
2367 /*
2368 * Read the current index, replace the path entry with
2369 * origin->blob_sha1 without mucking with its mode or type
2370 * bits; we are not going to write this index out -- we just
2371 * want to run "diff-index --cached".
2372 */
2373 discard_cache();
2374 read_cache();
2375
2376 len = strlen(path);
2377 if (!mode) {
2378 int pos = cache_name_pos(path, len);
2379 if (0 <= pos)
7a51ed66 2380 mode = active_cache[pos]->ce_mode;
1cfe7733
JH
2381 else
2382 /* Let's not bother reading from HEAD tree */
2383 mode = S_IFREG | 0644;
2384 }
2385 size = cache_entry_size(len);
2386 ce = xcalloc(1, size);
2387 hashcpy(ce->sha1, origin->blob_sha1);
2388 memcpy(ce->name, path, len);
b60e188c
TG
2389 ce->ce_flags = create_ce_flags(0);
2390 ce->ce_namelen = len;
1cfe7733
JH
2391 ce->ce_mode = create_ce_mode(mode);
2392 add_cache_entry(ce, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
2393
2394 /*
2395 * We are not going to write this out, so this does not matter
2396 * right now, but someday we might optimize diff-index --cached
2397 * with cache-tree information.
2398 */
a5400efe 2399 cache_tree_invalidate_path(&the_index, path);
1cfe7733 2400
1cfe7733
JH
2401 return commit;
2402}
2403
a46442f1 2404static char *prepare_final(struct scoreboard *sb)
f6c07d7d
JH
2405{
2406 int i;
2407 const char *final_commit_name = NULL;
85af7929 2408 struct rev_info *revs = sb->revs;
f6c07d7d
JH
2409
2410 /*
2411 * There must be one and only one positive commit in the
2412 * revs->pending array.
2413 */
2414 for (i = 0; i < revs->pending.nr; i++) {
2415 struct object *obj = revs->pending.objects[i].item;
2416 if (obj->flags & UNINTERESTING)
2417 continue;
2418 while (obj->type == OBJ_TAG)
2419 obj = deref_tag(obj, NULL, 0);
2420 if (obj->type != OBJ_COMMIT)
2421 die("Non commit %s?", revs->pending.objects[i].name);
2422 if (sb->final)
2423 die("More than one commit to dig from %s and %s?",
2424 revs->pending.objects[i].name,
2425 final_commit_name);
2426 sb->final = (struct commit *) obj;
2427 final_commit_name = revs->pending.objects[i].name;
2428 }
a46442f1 2429 return xstrdup_or_null(final_commit_name);
f6c07d7d
JH
2430}
2431
a46442f1 2432static char *prepare_initial(struct scoreboard *sb)
85af7929
JH
2433{
2434 int i;
2435 const char *final_commit_name = NULL;
2436 struct rev_info *revs = sb->revs;
2437
2438 /*
2439 * There must be one and only one negative commit, and it must be
2440 * the boundary.
2441 */
2442 for (i = 0; i < revs->pending.nr; i++) {
2443 struct object *obj = revs->pending.objects[i].item;
2444 if (!(obj->flags & UNINTERESTING))
2445 continue;
2446 while (obj->type == OBJ_TAG)
2447 obj = deref_tag(obj, NULL, 0);
2448 if (obj->type != OBJ_COMMIT)
2449 die("Non commit %s?", revs->pending.objects[i].name);
2450 if (sb->final)
2451 die("More than one commit to dig down to %s and %s?",
2452 revs->pending.objects[i].name,
2453 final_commit_name);
2454 sb->final = (struct commit *) obj;
2455 final_commit_name = revs->pending.objects[i].name;
2456 }
2457 if (!final_commit_name)
2458 die("No commit to dig down to?");
a46442f1 2459 return xstrdup(final_commit_name);
85af7929
JH
2460}
2461
5817da01
PH
2462static int blame_copy_callback(const struct option *option, const char *arg, int unset)
2463{
2464 int *opt = option->value;
2465
2466 /*
2467 * -C enables copy from removed files;
2468 * -C -C enables copy from existing files, but only
2469 * when blaming a new file;
2470 * -C -C -C enables copy from existing files for
2471 * everybody
2472 */
2473 if (*opt & PICKAXE_BLAME_COPY_HARDER)
2474 *opt |= PICKAXE_BLAME_COPY_HARDEST;
2475 if (*opt & PICKAXE_BLAME_COPY)
2476 *opt |= PICKAXE_BLAME_COPY_HARDER;
2477 *opt |= PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE;
2478
2479 if (arg)
2480 blame_copy_score = parse_score(arg);
2481 return 0;
2482}
2483
2484static int blame_move_callback(const struct option *option, const char *arg, int unset)
2485{
2486 int *opt = option->value;
2487
2488 *opt |= PICKAXE_BLAME_MOVE;
2489
2490 if (arg)
2491 blame_move_score = parse_score(arg);
2492 return 0;
2493}
2494
acca687f 2495int cmd_blame(int argc, const char **argv, const char *prefix)
cee7f245
JH
2496{
2497 struct rev_info revs;
2498 const char *path;
2499 struct scoreboard sb;
2500 struct origin *o;
58dbfa2e
ES
2501 struct blame_entry *ent = NULL;
2502 long dashdash_pos, lno;
a46442f1 2503 char *final_commit_name = NULL;
21666f1a 2504 enum object_type type;
5817da01 2505
58dbfa2e 2506 static struct string_list range_list;
5817da01
PH
2507 static int output_option = 0, opt = 0;
2508 static int show_stats = 0;
2509 static const char *revs_file = NULL;
2510 static const char *contents_from = NULL;
2511 static const struct option options[] = {
d5d09d47
SB
2512 OPT_BOOL(0, "incremental", &incremental, N_("Show blame entries as we find them, incrementally")),
2513 OPT_BOOL('b', NULL, &blank_boundary, N_("Show blank SHA-1 for boundary commits (Default: off)")),
2514 OPT_BOOL(0, "root", &show_root, N_("Do not treat root commits as boundaries (Default: off)")),
2515 OPT_BOOL(0, "show-stats", &show_stats, N_("Show work cost statistics")),
efd2a8bd
NTND
2516 OPT_BIT(0, "score-debug", &output_option, N_("Show output score for blame entries"), OUTPUT_SHOW_SCORE),
2517 OPT_BIT('f', "show-name", &output_option, N_("Show original filename (Default: auto)"), OUTPUT_SHOW_NAME),
2518 OPT_BIT('n', "show-number", &output_option, N_("Show original linenumber (Default: off)"), OUTPUT_SHOW_NUMBER),
2519 OPT_BIT('p', "porcelain", &output_option, N_("Show in a format designed for machine consumption"), OUTPUT_PORCELAIN),
2520 OPT_BIT(0, "line-porcelain", &output_option, N_("Show porcelain format with per-line commit information"), OUTPUT_PORCELAIN|OUTPUT_LINE_PORCELAIN),
2521 OPT_BIT('c', NULL, &output_option, N_("Use the same output mode as git-annotate (Default: off)"), OUTPUT_ANNOTATE_COMPAT),
2522 OPT_BIT('t', NULL, &output_option, N_("Show raw timestamp (Default: off)"), OUTPUT_RAW_TIMESTAMP),
2523 OPT_BIT('l', NULL, &output_option, N_("Show long commit SHA1 (Default: off)"), OUTPUT_LONG_OBJECT_NAME),
2524 OPT_BIT('s', NULL, &output_option, N_("Suppress author name and timestamp (Default: off)"), OUTPUT_NO_AUTHOR),
2525 OPT_BIT('e', "show-email", &output_option, N_("Show author email instead of name (Default: off)"), OUTPUT_SHOW_EMAIL),
2526 OPT_BIT('w', NULL, &xdl_opts, N_("Ignore whitespace differences"), XDF_IGNORE_WHITESPACE),
2527 OPT_BIT(0, "minimal", &xdl_opts, N_("Spend extra cycles to find better match"), XDF_NEED_MINIMAL),
2528 OPT_STRING('S', NULL, &revs_file, N_("file"), N_("Use revisions from <file> instead of calling git-rev-list")),
2529 OPT_STRING(0, "contents", &contents_from, N_("file"), N_("Use <file>'s contents as the final image")),
2530 { OPTION_CALLBACK, 'C', NULL, &opt, N_("score"), N_("Find line copies within and across files"), PARSE_OPT_OPTARG, blame_copy_callback },
2531 { OPTION_CALLBACK, 'M', NULL, &opt, N_("score"), N_("Find line movements within and across files"), PARSE_OPT_OPTARG, blame_move_callback },
58dbfa2e 2532 OPT_STRING_LIST('L', NULL, &range_list, N_("n,m"), N_("Process only line range n,m, counting from 1")),
84393bfd 2533 OPT__ABBREV(&abbrev),
5817da01
PH
2534 OPT_END()
2535 };
2536
2537 struct parse_opt_ctx_t ctx;
7ceacdff 2538 int cmd_is_annotate = !strcmp(argv[0], "annotate");
58dbfa2e
ES
2539 struct range_set ranges;
2540 unsigned int range_i;
52f4d126 2541 long anchor;
e68989a7 2542
8b504db3 2543 git_config(git_blame_config, &output_option);
5817da01 2544 init_revisions(&revs, NULL);
31653c1a 2545 revs.date_mode = blame_date_mode;
3b8a12e8 2546 DIFF_OPT_SET(&revs.diffopt, ALLOW_TEXTCONV);
3d1aa566 2547 DIFF_OPT_SET(&revs.diffopt, FOLLOW_RENAMES);
31653c1a 2548
612702e8 2549 save_commit_buffer = 0;
3f8d5204 2550 dashdash_pos = 0;
612702e8 2551
9ca1169f
SB
2552 parse_options_start(&ctx, argc, argv, prefix, options,
2553 PARSE_OPT_KEEP_DASHDASH | PARSE_OPT_KEEP_ARGV0);
5817da01 2554 for (;;) {
5817da01
PH
2555 switch (parse_options_step(&ctx, options, blame_opt_usage)) {
2556 case PARSE_OPT_HELP:
2557 exit(129);
2558 case PARSE_OPT_DONE:
3f8d5204
PH
2559 if (ctx.argv[0])
2560 dashdash_pos = ctx.cpidx;
5817da01
PH
2561 goto parse_done;
2562 }
2563
2564 if (!strcmp(ctx.argv[0], "--reverse")) {
2565 ctx.argv[0] = "--children";
2566 reverse = 1;
2567 }
6b61ec05 2568 parse_revision_opt(&revs, &ctx, options, blame_opt_usage);
5817da01
PH
2569 }
2570parse_done:
3d1aa566
JH
2571 no_whole_file_rename = !DIFF_OPT_TST(&revs.diffopt, FOLLOW_RENAMES);
2572 DIFF_OPT_CLR(&revs.diffopt, FOLLOW_RENAMES);
5817da01
PH
2573 argc = parse_options_end(&ctx);
2574
b31272f7
JH
2575 if (0 < abbrev)
2576 /* one more abbrev length is needed for the boundary commit */
2577 abbrev++;
84393bfd 2578
aa9ea77d 2579 if (revs_file && read_ancestry(revs_file))
d824cbba 2580 die_errno("reading graft file '%s' failed", revs_file);
aa9ea77d 2581
31653c1a 2582 if (cmd_is_annotate) {
7ceacdff 2583 output_option |= OUTPUT_ANNOTATE_COMPAT;
a5481a6c 2584 blame_date_mode.type = DATE_ISO8601;
31653c1a
EL
2585 } else {
2586 blame_date_mode = revs.date_mode;
2587 }
2588
2589 /* The maximum width used to show the dates */
a5481a6c 2590 switch (blame_date_mode.type) {
31653c1a
EL
2591 case DATE_RFC2822:
2592 blame_date_width = sizeof("Thu, 19 Oct 2006 16:00:04 -0700");
2593 break;
466fb674
BB
2594 case DATE_ISO8601_STRICT:
2595 blame_date_width = sizeof("2006-10-19T16:00:04-07:00");
2596 break;
31653c1a
EL
2597 case DATE_ISO8601:
2598 blame_date_width = sizeof("2006-10-19 16:00:04 -0700");
2599 break;
2600 case DATE_RAW:
2601 blame_date_width = sizeof("1161298804 -0700");
2602 break;
2603 case DATE_SHORT:
2604 blame_date_width = sizeof("2006-10-19");
2605 break;
2606 case DATE_RELATIVE:
dd75553b
JX
2607 /* TRANSLATORS: This string is used to tell us the maximum
2608 display width for a relative timestamp in "git blame"
2609 output. For C locale, "4 years, 11 months ago", which
2610 takes 22 places, is the longest among various forms of
2611 relative timestamps, but your language may need more or
2612 fewer display columns. */
2613 blame_date_width = utf8_strwidth(_("4 years, 11 months ago")) + 1; /* add the null */
2614 break;
31653c1a
EL
2615 case DATE_LOCAL:
2616 case DATE_NORMAL:
2617 blame_date_width = sizeof("Thu Oct 19 16:00:04 2006 -0700");
2618 break;
aa1462cc
JK
2619 case DATE_STRFTIME:
2620 blame_date_width = strlen(show_date(0, 0, &blame_date_mode)) + 1; /* add the null */
2621 break;
31653c1a
EL
2622 }
2623 blame_date_width -= 1; /* strip the null */
7ceacdff 2624
b3123f98
JH
2625 if (DIFF_OPT_TST(&revs.diffopt, FIND_COPIES_HARDER))
2626 opt |= (PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE |
2627 PICKAXE_BLAME_COPY_HARDER);
2628
4a0fc95f
JH
2629 if (!blame_move_score)
2630 blame_move_score = BLAME_DEFAULT_MOVE_SCORE;
2631 if (!blame_copy_score)
2632 blame_copy_score = BLAME_DEFAULT_COPY_SCORE;
2633
1732a1fd
JH
2634 /*
2635 * We have collected options unknown to us in argv[1..unk]
cee7f245 2636 * which are to be passed to revision machinery if we are
3dff5379 2637 * going to do the "bottom" processing.
cee7f245
JH
2638 *
2639 * The remaining are:
2640 *
22e5e58a 2641 * (1) if dashdash_pos != 0, it is either
3f8d5204
PH
2642 * "blame [revisions] -- <path>" or
2643 * "blame -- <path> <rev>"
cee7f245 2644 *
22e5e58a 2645 * (2) otherwise, it is one of the two:
3f8d5204
PH
2646 * "blame [revisions] <path>"
2647 * "blame <path> <rev>"
cee7f245 2648 *
3f8d5204
PH
2649 * Note that we must strip out <path> from the arguments: we do not
2650 * want the path pruning but we may want "bottom" processing.
cee7f245 2651 */
3f8d5204
PH
2652 if (dashdash_pos) {
2653 switch (argc - dashdash_pos - 1) {
2654 case 2: /* (1b) */
2655 if (argc != 4)
5817da01 2656 usage_with_options(blame_opt_usage, options);
3f8d5204
PH
2657 /* reorder for the new way: <rev> -- <path> */
2658 argv[1] = argv[3];
2659 argv[3] = argv[2];
2660 argv[2] = "--";
2661 /* FALLTHROUGH */
2662 case 1: /* (1a) */
2663 path = add_prefix(prefix, argv[--argc]);
2664 argv[argc] = NULL;
2665 break;
2666 default:
2667 usage_with_options(blame_opt_usage, options);
cee7f245 2668 }
3f8d5204
PH
2669 } else {
2670 if (argc < 2)
5817da01 2671 usage_with_options(blame_opt_usage, options);
3f8d5204 2672 path = add_prefix(prefix, argv[argc - 1]);
dbe44faa 2673 if (argc == 3 && !file_exists(path)) { /* (2b) */
3f8d5204
PH
2674 path = add_prefix(prefix, argv[1]);
2675 argv[1] = argv[2];
cee7f245 2676 }
3f8d5204 2677 argv[argc - 1] = "--";
cee7f245 2678
3f8d5204 2679 setup_work_tree();
dbe44faa 2680 if (!file_exists(path))
d824cbba 2681 die_errno("cannot stat path '%s'", path);
cee7f245
JH
2682 }
2683
8b3dce56 2684 revs.disable_stdin = 1;
3f8d5204 2685 setup_revisions(argc, argv, &revs, NULL);
cee7f245
JH
2686 memset(&sb, 0, sizeof(sb));
2687
85af7929 2688 sb.revs = &revs;
7e6ac6e4 2689 if (!reverse) {
85af7929 2690 final_commit_name = prepare_final(&sb);
7e6ac6e4
DK
2691 sb.commits.compare = compare_commits_by_commit_date;
2692 }
85af7929 2693 else if (contents_from)
95261974 2694 die("--contents and --reverse do not blend well.");
95a4fb0e
JK
2695 else if (revs.first_parent_only)
2696 die("combining --first-parent and --reverse is not supported");
7e6ac6e4 2697 else {
85af7929 2698 final_commit_name = prepare_initial(&sb);
7e6ac6e4
DK
2699 sb.commits.compare = compare_commits_by_reverse_commit_date;
2700 }
cee7f245
JH
2701
2702 if (!sb.final) {
1732a1fd
JH
2703 /*
2704 * "--not A B -- path" without anything positive;
1cfe7733
JH
2705 * do not default to HEAD, but use the working tree
2706 * or "--contents".
1732a1fd 2707 */
1981820b 2708 setup_work_tree();
3b8a12e8
AB
2709 sb.final = fake_working_tree_commit(&sb.revs->diffopt,
2710 path, contents_from);
1cfe7733 2711 add_pending_object(&revs, &(sb.final->object), ":");
cee7f245 2712 }
1cfe7733
JH
2713 else if (contents_from)
2714 die("Cannot use --contents with final commit object name");
cee7f245 2715
1732a1fd
JH
2716 /*
2717 * If we have bottom, this will mark the ancestors of the
cee7f245
JH
2718 * bottom commits we would reach while traversing as
2719 * uninteresting.
2720 */
3d51e1b5 2721 if (prepare_revision_walk(&revs))
20108742 2722 die(_("revision walk setup failed"));
cee7f245 2723
1cfe7733 2724 if (is_null_sha1(sb.final->object.sha1)) {
1cfe7733 2725 o = sb.final->util;
5c0b13f8 2726 sb.final_buf = xmemdupz(o->file.ptr, o->file.size);
1cfe7733
JH
2727 sb.final_buf_size = o->file.size;
2728 }
2729 else {
2730 o = get_origin(&sb, sb.final, path);
90064710 2731 if (fill_blob_sha1_and_mode(o))
1cfe7733 2732 die("no such path %s in %s", path, final_commit_name);
cee7f245 2733
3b8a12e8 2734 if (DIFF_OPT_TST(&sb.revs->diffopt, ALLOW_TEXTCONV) &&
e5450100 2735 textconv_object(path, o->mode, o->blob_sha1, 1, (char **) &sb.final_buf,
3b8a12e8
AB
2736 &sb.final_buf_size))
2737 ;
2738 else
2739 sb.final_buf = read_sha1_file(o->blob_sha1, &type,
2740 &sb.final_buf_size);
2741
ab43e495
JH
2742 if (!sb.final_buf)
2743 die("Cannot read blob %s for path %s",
2744 sha1_to_hex(o->blob_sha1),
2745 path);
1cfe7733 2746 }
c2e525d9 2747 num_read_blob++;
cee7f245
JH
2748 lno = prepare_lines(&sb);
2749
58dbfa2e
ES
2750 if (lno && !range_list.nr)
2751 string_list_append(&range_list, xstrdup("1"));
2752
52f4d126 2753 anchor = 1;
58dbfa2e
ES
2754 range_set_init(&ranges, range_list.nr);
2755 for (range_i = 0; range_i < range_list.nr; ++range_i) {
2756 long bottom, top;
2757 if (parse_range_arg(range_list.items[range_i].string,
52f4d126 2758 nth_line_cb, &sb, lno, anchor,
58dbfa2e
ES
2759 &bottom, &top, sb.path))
2760 usage(blame_usage);
2761 if (lno < top || ((lno || bottom) && lno < bottom))
2762 die("file %s has only %lu lines", path, lno);
2763 if (bottom < 1)
2764 bottom = 1;
2765 if (top < 1)
2766 top = lno;
2767 bottom--;
2768 range_set_append_unsafe(&ranges, bottom, top);
52f4d126 2769 anchor = top + 1;
58dbfa2e
ES
2770 }
2771 sort_and_merge_range_set(&ranges);
2772
2773 for (range_i = ranges.nr; range_i > 0; --range_i) {
2774 const struct range *r = &ranges.ranges[range_i - 1];
2775 long bottom = r->start;
2776 long top = r->end;
2777 struct blame_entry *next = ent;
2778 ent = xcalloc(1, sizeof(*ent));
2779 ent->lno = bottom;
2780 ent->num_lines = top - bottom;
2781 ent->suspect = o;
2782 ent->s_lno = bottom;
2783 ent->next = next;
58dbfa2e
ES
2784 origin_incref(o);
2785 }
7e6ac6e4
DK
2786
2787 o->suspects = ent;
2788 prio_queue_put(&sb.commits, o->commit);
2789
58dbfa2e
ES
2790 origin_decref(o);
2791
2792 range_set_release(&ranges);
2793 string_list_clear(&range_list, 0);
cee7f245 2794
7e6ac6e4 2795 sb.ent = NULL;
cee7f245
JH
2796 sb.path = path;
2797
d551a488 2798 read_mailmap(&mailmap, NULL);
f9567384 2799
b92565dc
MH
2800 if (!incremental)
2801 setup_pager();
2802
85af7929 2803 assign_blame(&sb, opt);
cee7f245 2804
a46442f1
LF
2805 free(final_commit_name);
2806
717d1462
LT
2807 if (incremental)
2808 return 0;
2809
7e6ac6e4
DK
2810 sb.ent = blame_sort(sb.ent, compare_blame_final);
2811
cee7f245
JH
2812 coalesce(&sb);
2813
2814 if (!(output_option & OUTPUT_PORCELAIN))
2815 find_alignment(&sb, &output_option);
2816
2817 output(&sb, output_option);
2818 free((void *)sb.final_buf);
2819 for (ent = sb.ent; ent; ) {
2820 struct blame_entry *e = ent->next;
2821 free(ent);
2822 ent = e;
2823 }
c2e525d9 2824
870b39c1 2825 if (show_stats) {
c2e525d9
JH
2826 printf("num read blob: %d\n", num_read_blob);
2827 printf("num get patch: %d\n", num_get_patch);
2828 printf("num commits: %d\n", num_commits);
2829 }
cee7f245
JH
2830 return 0;
2831}