blame: extract find_single_final
[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"
9#include "builtin.h"
10#include "blob.h"
11#include "commit.h"
12#include "tag.h"
13#include "tree-walk.h"
14#include "diff.h"
15#include "diffcore.h"
16#include "revision.h"
717d1462 17#include "quote.h"
cee7f245 18#include "xdiff-interface.h"
1cfe7733 19#include "cache-tree.h"
c455c87c 20#include "string-list.h"
f9567384 21#include "mailmap.h"
7e6ac6e4 22#include "mergesort.h"
5817da01 23#include "parse-options.h"
7e6ac6e4 24#include "prio-queue.h"
ffaf9cc0 25#include "utf8.h"
3b8a12e8 26#include "userdiff.h"
25ed3412 27#include "line-range.h"
58dbfa2e 28#include "line-log.h"
dbe44faa 29#include "dir.h"
cee7f245 30
ce41720c 31static char blame_usage[] = N_("git blame [<options>] [<rev-opts>] [<rev>] [--] <file>");
5817da01
PH
32
33static const char *blame_opt_usage[] = {
34 blame_usage,
35 "",
9c9b4f2f 36 N_("<rev-opts> are documented in git-rev-list(1)"),
5817da01
PH
37 NULL
38};
cee7f245
JH
39
40static int longest_file;
41static int longest_author;
42static int max_orig_digits;
43static int max_digits;
5ff62c30 44static int max_score_digits;
4c10a5ca 45static int show_root;
85af7929 46static int reverse;
4c10a5ca 47static int blank_boundary;
717d1462 48static int incremental;
582aa00b 49static int xdl_opts;
84393bfd 50static int abbrev = -1;
3d1aa566 51static int no_whole_file_rename;
31653c1a
EL
52
53static enum date_mode blame_date_mode = DATE_ISO8601;
54static size_t blame_date_width;
55
c455c87c 56static struct string_list mailmap;
cee7f245 57
54a4c617
JH
58#ifndef DEBUG
59#define DEBUG 0
60#endif
61
c2e525d9
JH
62/* stats */
63static int num_read_blob;
64static int num_get_patch;
65static int num_commits;
66
d24bba80 67#define PICKAXE_BLAME_MOVE 01
18abd745
JH
68#define PICKAXE_BLAME_COPY 02
69#define PICKAXE_BLAME_COPY_HARDER 04
c63777c0 70#define PICKAXE_BLAME_COPY_HARDEST 010
d24bba80 71
4a0fc95f
JH
72/*
73 * blame for a blame_entry with score lower than these thresholds
74 * is not passed to the parent using move/copy logic.
75 */
76static unsigned blame_move_score;
77static unsigned blame_copy_score;
78#define BLAME_DEFAULT_MOVE_SCORE 20
79#define BLAME_DEFAULT_COPY_SCORE 40
80
208acbfb 81/* Remember to update object flag allocation in object.h */
cee7f245
JH
82#define METAINFO_SHOWN (1u<<12)
83#define MORE_THAN_ONE_PATH (1u<<13)
84
85/*
54a4c617 86 * One blob in a commit that is being suspected
cee7f245
JH
87 */
88struct origin {
54a4c617 89 int refcnt;
7e6ac6e4 90 /* Record preceding blame record for this blob */
96e11709 91 struct origin *previous;
7e6ac6e4
DK
92 /* origins are put in a list linked via `next' hanging off the
93 * corresponding commit's util field in order to make finding
94 * them fast. The presence in this chain does not count
95 * towards the origin's reference count. It is tempting to
96 * let it count as long as the commit is pending examination,
97 * but even under circumstances where the commit will be
98 * present multiple times in the priority queue of unexamined
99 * commits, processing the first instance will not leave any
100 * work requiring the origin data for the second instance. An
101 * interspersed commit changing that would have to be
102 * preexisting with a different ancestry and with the same
103 * commit date in order to wedge itself between two instances
104 * of the same commit in the priority queue _and_ produce
105 * blame entries relevant for it. While we don't want to let
106 * us get tripped up by this case, it certainly does not seem
107 * worth optimizing for.
108 */
109 struct origin *next;
cee7f245 110 struct commit *commit;
7e6ac6e4
DK
111 /* `suspects' contains blame entries that may be attributed to
112 * this origin's commit or to parent commits. When a commit
113 * is being processed, all suspects will be moved, either by
114 * assigning them to an origin in a different commit, or by
115 * shipping them to the scoreboard's ent list because they
116 * cannot be attributed to a different commit.
117 */
118 struct blame_entry *suspects;
c2e525d9 119 mmfile_t file;
cee7f245 120 unsigned char blob_sha1[20];
90064710 121 unsigned mode;
7e6ac6e4
DK
122 /* guilty gets set when shipping any suspects to the final
123 * blame list instead of other commits
124 */
125 char guilty;
cee7f245
JH
126 char path[FLEX_ARRAY];
127};
128
4b4132fd
RS
129static int diff_hunks(mmfile_t *file_a, mmfile_t *file_b, long ctxlen,
130 xdl_emit_hunk_consume_func_t hunk_func, void *cb_data)
131{
132 xpparam_t xpp = {0};
133 xdemitconf_t xecfg = {0};
85c20c30 134 xdemitcb_t ecb = {NULL};
4b4132fd
RS
135
136 xpp.flags = xdl_opts;
137 xecfg.ctxlen = ctxlen;
138 xecfg.hunk_func = hunk_func;
139 ecb.priv = cb_data;
140 return xdi_diff(file_a, file_b, &xpp, &xecfg, &ecb);
141}
142
3b8a12e8
AB
143/*
144 * Prepare diff_filespec and convert it using diff textconv API
145 * if the textconv driver exists.
146 * Return 1 if the conversion succeeds, 0 otherwise.
147 */
e5fba602 148int textconv_object(const char *path,
90064710 149 unsigned mode,
e5fba602 150 const unsigned char *sha1,
e5450100 151 int sha1_valid,
e5fba602
CP
152 char **buf,
153 unsigned long *buf_size)
3b8a12e8
AB
154{
155 struct diff_filespec *df;
156 struct userdiff_driver *textconv;
157
158 df = alloc_filespec(path);
e5450100 159 fill_filespec(df, sha1, sha1_valid, mode);
3b8a12e8
AB
160 textconv = get_textconv(df);
161 if (!textconv) {
162 free_filespec(df);
163 return 0;
164 }
165
166 *buf_size = fill_textconv(textconv, df, buf);
167 free_filespec(df);
168 return 1;
169}
170
1732a1fd
JH
171/*
172 * Given an origin, prepare mmfile_t structure to be used by the
173 * diff machinery
174 */
3b8a12e8
AB
175static void fill_origin_blob(struct diff_options *opt,
176 struct origin *o, mmfile_t *file)
c2e525d9
JH
177{
178 if (!o->file.ptr) {
21666f1a 179 enum object_type type;
3b8a12e8
AB
180 unsigned long file_size;
181
c2e525d9 182 num_read_blob++;
3b8a12e8 183 if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) &&
e5450100 184 textconv_object(o->path, o->mode, o->blob_sha1, 1, &file->ptr, &file_size))
3b8a12e8
AB
185 ;
186 else
187 file->ptr = read_sha1_file(o->blob_sha1, &type, &file_size);
188 file->size = file_size;
189
ab43e495
JH
190 if (!file->ptr)
191 die("Cannot read blob %s for path %s",
192 sha1_to_hex(o->blob_sha1),
193 o->path);
c2e525d9
JH
194 o->file = *file;
195 }
196 else
197 *file = o->file;
c2e525d9
JH
198}
199
1732a1fd
JH
200/*
201 * Origin is refcounted and usually we keep the blob contents to be
202 * reused.
203 */
54a4c617
JH
204static inline struct origin *origin_incref(struct origin *o)
205{
206 if (o)
207 o->refcnt++;
208 return o;
209}
210
211static void origin_decref(struct origin *o)
212{
213 if (o && --o->refcnt <= 0) {
7e6ac6e4 214 struct origin *p, *l = NULL;
96e11709
JH
215 if (o->previous)
216 origin_decref(o->previous);
8e0f7003 217 free(o->file.ptr);
7e6ac6e4
DK
218 /* Should be present exactly once in commit chain */
219 for (p = o->commit->util; p; l = p, p = p->next) {
220 if (p == o) {
221 if (l)
222 l->next = p->next;
223 else
224 o->commit->util = p->next;
225 free(o);
226 return;
227 }
228 }
229 die("internal error in blame::origin_decref");
54a4c617
JH
230 }
231}
232
7c3c7962
JH
233static void drop_origin_blob(struct origin *o)
234{
235 if (o->file.ptr) {
236 free(o->file.ptr);
237 o->file.ptr = NULL;
238 }
239}
240
1732a1fd
JH
241/*
242 * Each group of lines is described by a blame_entry; it can be split
7e6ac6e4
DK
243 * as we pass blame to the parents. They are arranged in linked lists
244 * kept as `suspects' of some unprocessed origin, or entered (when the
245 * blame origin has been finalized) into the scoreboard structure.
246 * While the scoreboard structure is only sorted at the end of
247 * processing (according to final image line number), the lists
248 * attached to an origin are sorted by the target line number.
1732a1fd 249 */
cee7f245 250struct blame_entry {
cee7f245
JH
251 struct blame_entry *next;
252
253 /* the first line of this group in the final image;
254 * internally all line numbers are 0 based.
255 */
256 int lno;
257
258 /* how many lines this group has */
259 int num_lines;
260
261 /* the commit that introduced this group into the final image */
262 struct origin *suspect;
263
cee7f245
JH
264 /* the line number of the first line of this group in the
265 * suspect's file; internally all line numbers are 0 based.
266 */
267 int s_lno;
5ff62c30
JH
268
269 /* how significant this entry is -- cached to avoid
1732a1fd 270 * scanning the lines over and over.
5ff62c30
JH
271 */
272 unsigned score;
cee7f245
JH
273};
274
7e6ac6e4
DK
275/*
276 * Any merge of blames happens on lists of blames that arrived via
277 * different parents in a single suspect. In this case, we want to
278 * sort according to the suspect line numbers as opposed to the final
279 * image line numbers. The function body is somewhat longish because
280 * it avoids unnecessary writes.
281 */
282
283static struct blame_entry *blame_merge(struct blame_entry *list1,
284 struct blame_entry *list2)
285{
286 struct blame_entry *p1 = list1, *p2 = list2,
287 **tail = &list1;
288
289 if (!p1)
290 return p2;
291 if (!p2)
292 return p1;
293
294 if (p1->s_lno <= p2->s_lno) {
295 do {
296 tail = &p1->next;
297 if ((p1 = *tail) == NULL) {
298 *tail = p2;
299 return list1;
300 }
301 } while (p1->s_lno <= p2->s_lno);
302 }
303 for (;;) {
304 *tail = p2;
305 do {
306 tail = &p2->next;
307 if ((p2 = *tail) == NULL) {
308 *tail = p1;
309 return list1;
310 }
311 } while (p1->s_lno > p2->s_lno);
312 *tail = p1;
313 do {
314 tail = &p1->next;
315 if ((p1 = *tail) == NULL) {
316 *tail = p2;
317 return list1;
318 }
319 } while (p1->s_lno <= p2->s_lno);
320 }
321}
322
323static void *get_next_blame(const void *p)
324{
325 return ((struct blame_entry *)p)->next;
326}
327
328static void set_next_blame(void *p1, void *p2)
329{
330 ((struct blame_entry *)p1)->next = p2;
331}
332
333/*
334 * Final image line numbers are all different, so we don't need a
335 * three-way comparison here.
336 */
337
338static int compare_blame_final(const void *p1, const void *p2)
339{
340 return ((struct blame_entry *)p1)->lno > ((struct blame_entry *)p2)->lno
341 ? 1 : -1;
342}
343
344static int compare_blame_suspect(const void *p1, const void *p2)
345{
346 const struct blame_entry *s1 = p1, *s2 = p2;
347 /*
348 * to allow for collating suspects, we sort according to the
349 * respective pointer value as the primary sorting criterion.
350 * The actual relation is pretty unimportant as long as it
351 * establishes a total order. Comparing as integers gives us
352 * that.
353 */
354 if (s1->suspect != s2->suspect)
355 return (intptr_t)s1->suspect > (intptr_t)s2->suspect ? 1 : -1;
356 if (s1->s_lno == s2->s_lno)
357 return 0;
358 return s1->s_lno > s2->s_lno ? 1 : -1;
359}
360
361static struct blame_entry *blame_sort(struct blame_entry *head,
362 int (*compare_fn)(const void *, const void *))
363{
364 return llist_mergesort (head, get_next_blame, set_next_blame, compare_fn);
365}
366
367static int compare_commits_by_reverse_commit_date(const void *a,
368 const void *b,
369 void *c)
370{
371 return -compare_commits_by_commit_date(a, b, c);
372}
373
1732a1fd
JH
374/*
375 * The current state of the blame assignment.
376 */
cee7f245
JH
377struct scoreboard {
378 /* the final commit (i.e. where we started digging from) */
379 struct commit *final;
7e6ac6e4
DK
380 /* Priority queue for commits with unassigned blame records */
381 struct prio_queue commits;
85af7929 382 struct rev_info *revs;
cee7f245
JH
383 const char *path;
384
1732a1fd
JH
385 /*
386 * The contents in the final image.
387 * Used by many functions to obtain contents of the nth line,
388 * indexed with scoreboard.lineno[blame_entry.lno].
cee7f245
JH
389 */
390 const char *final_buf;
391 unsigned long final_buf_size;
392
393 /* linked list of blames */
394 struct blame_entry *ent;
395
612702e8 396 /* look-up a line in the final buffer */
cee7f245
JH
397 int num_lines;
398 int *lineno;
399};
400
54a4c617
JH
401static void sanity_check_refcnt(struct scoreboard *);
402
1732a1fd
JH
403/*
404 * If two blame entries that are next to each other came from
405 * contiguous lines in the same origin (i.e. <commit, path> pair),
406 * merge them together.
407 */
cee7f245
JH
408static void coalesce(struct scoreboard *sb)
409{
410 struct blame_entry *ent, *next;
411
412 for (ent = sb->ent; ent && (next = ent->next); ent = next) {
0a88f08e 413 if (ent->suspect == next->suspect &&
cee7f245
JH
414 ent->s_lno + ent->num_lines == next->s_lno) {
415 ent->num_lines += next->num_lines;
416 ent->next = next->next;
54a4c617 417 origin_decref(next->suspect);
cee7f245 418 free(next);
46014766 419 ent->score = 0;
cee7f245
JH
420 next = ent; /* again */
421 }
422 }
54a4c617
JH
423
424 if (DEBUG) /* sanity */
425 sanity_check_refcnt(sb);
cee7f245
JH
426}
427
7e6ac6e4
DK
428/*
429 * Merge the given sorted list of blames into a preexisting origin.
430 * If there were no previous blames to that commit, it is entered into
431 * the commit priority queue of the score board.
432 */
433
434static void queue_blames(struct scoreboard *sb, struct origin *porigin,
435 struct blame_entry *sorted)
436{
437 if (porigin->suspects)
438 porigin->suspects = blame_merge(porigin->suspects, sorted);
439 else {
440 struct origin *o;
441 for (o = porigin->commit->util; o; o = o->next) {
442 if (o->suspects) {
443 porigin->suspects = sorted;
444 return;
445 }
446 }
447 porigin->suspects = sorted;
448 prio_queue_put(&sb->commits, porigin->commit);
449 }
450}
451
1732a1fd
JH
452/*
453 * Given a commit and a path in it, create a new origin structure.
454 * The callers that add blame to the scoreboard should use
455 * get_origin() to obtain shared, refcounted copy instead of calling
456 * this function directly.
457 */
854b97f6
JH
458static struct origin *make_origin(struct commit *commit, const char *path)
459{
460 struct origin *o;
461 o = xcalloc(1, sizeof(*o) + strlen(path) + 1);
462 o->commit = commit;
463 o->refcnt = 1;
7e6ac6e4
DK
464 o->next = commit->util;
465 commit->util = o;
854b97f6
JH
466 strcpy(o->path, path);
467 return o;
468}
469
1732a1fd
JH
470/*
471 * Locate an existing origin or create a new one.
7e6ac6e4 472 * This moves the origin to front position in the commit util list.
1732a1fd 473 */
f6c0e191
JH
474static struct origin *get_origin(struct scoreboard *sb,
475 struct commit *commit,
476 const char *path)
cee7f245 477{
7e6ac6e4 478 struct origin *o, *l;
cee7f245 479
7e6ac6e4
DK
480 for (o = commit->util, l = NULL; o; l = o, o = o->next) {
481 if (!strcmp(o->path, path)) {
482 /* bump to front */
483 if (l) {
484 l->next = o->next;
485 o->next = commit->util;
486 commit->util = o;
487 }
488 return origin_incref(o);
489 }
cee7f245 490 }
854b97f6 491 return make_origin(commit, path);
cee7f245
JH
492}
493
1732a1fd
JH
494/*
495 * Fill the blob_sha1 field of an origin if it hasn't, so that later
496 * call to fill_origin_blob() can use it to locate the data. blob_sha1
497 * for an origin is also used to pass the blame for the entire file to
498 * the parent to detect the case where a child's blob is identical to
499 * that of its parent's.
90064710
KS
500 *
501 * This also fills origin->mode for corresponding tree path.
1732a1fd 502 */
90064710 503static int fill_blob_sha1_and_mode(struct origin *origin)
f6c0e191 504{
f6c0e191
JH
505 if (!is_null_sha1(origin->blob_sha1))
506 return 0;
507 if (get_tree_entry(origin->commit->object.sha1,
508 origin->path,
90064710 509 origin->blob_sha1, &origin->mode))
f6c0e191 510 goto error_out;
21666f1a 511 if (sha1_object_info(origin->blob_sha1, NULL) != OBJ_BLOB)
f6c0e191
JH
512 goto error_out;
513 return 0;
514 error_out:
515 hashclr(origin->blob_sha1);
90064710 516 origin->mode = S_IFINVALID;
f6c0e191
JH
517 return -1;
518}
519
1732a1fd
JH
520/*
521 * We have an origin -- check if the same path exists in the
522 * parent and return an origin structure to represent it.
523 */
f6c0e191 524static struct origin *find_origin(struct scoreboard *sb,
cee7f245
JH
525 struct commit *parent,
526 struct origin *origin)
527{
7e6ac6e4 528 struct origin *porigin;
cee7f245 529 struct diff_options diff_opts;
f6c0e191
JH
530 const char *paths[2];
531
7e6ac6e4
DK
532 /* First check any existing origins */
533 for (porigin = parent->util; porigin; porigin = porigin->next)
534 if (!strcmp(porigin->path, origin->path)) {
1732a1fd
JH
535 /*
536 * The same path between origin and its parent
537 * without renaming -- the most common case.
538 */
7e6ac6e4 539 return origin_incref (porigin);
854b97f6 540 }
0d981c67 541
f6c0e191
JH
542 /* See if the origin->path is different between parent
543 * and origin first. Most of the time they are the
544 * same and diff-tree is fairly efficient about this.
545 */
546 diff_setup(&diff_opts);
8f67f8ae 547 DIFF_OPT_SET(&diff_opts, RECURSIVE);
f6c0e191
JH
548 diff_opts.detect_rename = 0;
549 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
550 paths[0] = origin->path;
551 paths[1] = NULL;
552
4a2d5ae2
NTND
553 parse_pathspec(&diff_opts.pathspec,
554 PATHSPEC_ALL_MAGIC & ~PATHSPEC_LITERAL,
555 PATHSPEC_LITERAL_PATH, "", paths);
28452655 556 diff_setup_done(&diff_opts);
1cfe7733
JH
557
558 if (is_null_sha1(origin->commit->object.sha1))
559 do_diff_cache(parent->tree->object.sha1, &diff_opts);
560 else
561 diff_tree_sha1(parent->tree->object.sha1,
562 origin->commit->tree->object.sha1,
563 "", &diff_opts);
f6c0e191
JH
564 diffcore_std(&diff_opts);
565
f6c0e191
JH
566 if (!diff_queued_diff.nr) {
567 /* The path is the same as parent */
568 porigin = get_origin(sb, parent, origin->path);
569 hashcpy(porigin->blob_sha1, origin->blob_sha1);
90064710 570 porigin->mode = origin->mode;
a9b2d424
JH
571 } else {
572 /*
573 * Since origin->path is a pathspec, if the parent
574 * commit had it as a directory, we will see a whole
575 * bunch of deletion of files in the directory that we
576 * do not care about.
577 */
578 int i;
579 struct diff_filepair *p = NULL;
580 for (i = 0; i < diff_queued_diff.nr; i++) {
581 const char *name;
582 p = diff_queued_diff.queue[i];
583 name = p->one->path ? p->one->path : p->two->path;
584 if (!strcmp(name, origin->path))
585 break;
586 }
587 if (!p)
588 die("internal error in blame::find_origin");
f6c0e191
JH
589 switch (p->status) {
590 default:
acca687f 591 die("internal error in blame::find_origin (%c)",
f6c0e191
JH
592 p->status);
593 case 'M':
594 porigin = get_origin(sb, parent, origin->path);
595 hashcpy(porigin->blob_sha1, p->one->sha1);
90064710 596 porigin->mode = p->one->mode;
f6c0e191
JH
597 break;
598 case 'A':
599 case 'T':
600 /* Did not exist in parent, or type changed */
601 break;
602 }
603 }
604 diff_flush(&diff_opts);
bd1928df 605 free_pathspec(&diff_opts.pathspec);
f69e743d
JH
606 return porigin;
607}
f6c0e191 608
1732a1fd
JH
609/*
610 * We have an origin -- find the path that corresponds to it in its
611 * parent and return an origin structure to represent it.
612 */
f69e743d
JH
613static struct origin *find_rename(struct scoreboard *sb,
614 struct commit *parent,
615 struct origin *origin)
616{
617 struct origin *porigin = NULL;
618 struct diff_options diff_opts;
619 int i;
cee7f245
JH
620
621 diff_setup(&diff_opts);
8f67f8ae 622 DIFF_OPT_SET(&diff_opts, RECURSIVE);
cee7f245
JH
623 diff_opts.detect_rename = DIFF_DETECT_RENAME;
624 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
2f3f8b21 625 diff_opts.single_follow = origin->path;
28452655 626 diff_setup_done(&diff_opts);
1cfe7733
JH
627
628 if (is_null_sha1(origin->commit->object.sha1))
629 do_diff_cache(parent->tree->object.sha1, &diff_opts);
630 else
631 diff_tree_sha1(parent->tree->object.sha1,
632 origin->commit->tree->object.sha1,
633 "", &diff_opts);
cee7f245
JH
634 diffcore_std(&diff_opts);
635
636 for (i = 0; i < diff_queued_diff.nr; i++) {
637 struct diff_filepair *p = diff_queued_diff.queue[i];
612702e8 638 if ((p->status == 'R' || p->status == 'C') &&
f6c0e191
JH
639 !strcmp(p->two->path, origin->path)) {
640 porigin = get_origin(sb, parent, p->one->path);
641 hashcpy(porigin->blob_sha1, p->one->sha1);
90064710 642 porigin->mode = p->one->mode;
cee7f245
JH
643 break;
644 }
645 }
646 diff_flush(&diff_opts);
bd1928df 647 free_pathspec(&diff_opts.pathspec);
cee7f245
JH
648 return porigin;
649}
650
1732a1fd 651/*
7e6ac6e4 652 * Append a new blame entry to a given output queue.
1732a1fd 653 */
7e6ac6e4 654static void add_blame_entry(struct blame_entry ***queue, struct blame_entry *e)
cee7f245 655{
54a4c617
JH
656 origin_incref(e->suspect);
657
7e6ac6e4
DK
658 e->next = **queue;
659 **queue = e;
660 *queue = &e->next;
cee7f245
JH
661}
662
1732a1fd
JH
663/*
664 * src typically is on-stack; we want to copy the information in it to
7e6ac6e4
DK
665 * a malloced blame_entry that gets added to the given queue. The
666 * origin of dst loses a refcnt.
1732a1fd 667 */
7e6ac6e4
DK
668static void dup_entry(struct blame_entry ***queue,
669 struct blame_entry *dst, struct blame_entry *src)
cee7f245 670{
54a4c617
JH
671 origin_incref(src->suspect);
672 origin_decref(dst->suspect);
cee7f245 673 memcpy(dst, src, sizeof(*src));
7e6ac6e4
DK
674 dst->next = **queue;
675 **queue = dst;
676 *queue = &dst->next;
cee7f245
JH
677}
678
25ed3412 679static const char *nth_line(struct scoreboard *sb, long lno)
cee7f245
JH
680{
681 return sb->final_buf + sb->lineno[lno];
682}
683
25ed3412
BY
684static const char *nth_line_cb(void *data, long lno)
685{
686 return nth_line((struct scoreboard *)data, lno);
687}
688
1732a1fd
JH
689/*
690 * It is known that lines between tlno to same came from parent, and e
691 * has an overlap with that range. it also is known that parent's
692 * line plno corresponds to e's line tlno.
693 *
694 * <---- e ----->
695 * <------>
696 * <------------>
697 * <------------>
698 * <------------------>
699 *
700 * Split e into potentially three parts; before this chunk, the chunk
701 * to be blamed for the parent, and after that portion.
702 */
54a4c617 703static void split_overlap(struct blame_entry *split,
cee7f245
JH
704 struct blame_entry *e,
705 int tlno, int plno, int same,
706 struct origin *parent)
707{
cee7f245
JH
708 int chunk_end_lno;
709 memset(split, 0, sizeof(struct blame_entry [3]));
710
711 if (e->s_lno < tlno) {
712 /* there is a pre-chunk part not blamed on parent */
54a4c617 713 split[0].suspect = origin_incref(e->suspect);
cee7f245
JH
714 split[0].lno = e->lno;
715 split[0].s_lno = e->s_lno;
716 split[0].num_lines = tlno - e->s_lno;
717 split[1].lno = e->lno + tlno - e->s_lno;
718 split[1].s_lno = plno;
719 }
720 else {
721 split[1].lno = e->lno;
722 split[1].s_lno = plno + (e->s_lno - tlno);
723 }
724
725 if (same < e->s_lno + e->num_lines) {
726 /* there is a post-chunk part not blamed on parent */
54a4c617 727 split[2].suspect = origin_incref(e->suspect);
cee7f245
JH
728 split[2].lno = e->lno + (same - e->s_lno);
729 split[2].s_lno = e->s_lno + (same - e->s_lno);
730 split[2].num_lines = e->s_lno + e->num_lines - same;
731 chunk_end_lno = split[2].lno;
732 }
733 else
734 chunk_end_lno = e->lno + e->num_lines;
735 split[1].num_lines = chunk_end_lno - split[1].lno;
736
1732a1fd
JH
737 /*
738 * if it turns out there is nothing to blame the parent for,
739 * forget about the splitting. !split[1].suspect signals this.
740 */
cee7f245
JH
741 if (split[1].num_lines < 1)
742 return;
54a4c617 743 split[1].suspect = origin_incref(parent);
cee7f245
JH
744}
745
1732a1fd
JH
746/*
747 * split_overlap() divided an existing blame e into up to three parts
7e6ac6e4 748 * in split. Any assigned blame is moved to queue to
1732a1fd
JH
749 * reflect the split.
750 */
7e6ac6e4
DK
751static void split_blame(struct blame_entry ***blamed,
752 struct blame_entry ***unblamed,
54a4c617 753 struct blame_entry *split,
cee7f245
JH
754 struct blame_entry *e)
755{
756 struct blame_entry *new_entry;
757
758 if (split[0].suspect && split[2].suspect) {
1732a1fd 759 /* The first part (reuse storage for the existing entry e) */
7e6ac6e4 760 dup_entry(unblamed, e, &split[0]);
cee7f245 761
1732a1fd 762 /* The last part -- me */
cee7f245
JH
763 new_entry = xmalloc(sizeof(*new_entry));
764 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
7e6ac6e4 765 add_blame_entry(unblamed, new_entry);
cee7f245 766
1732a1fd 767 /* ... and the middle part -- parent */
cee7f245
JH
768 new_entry = xmalloc(sizeof(*new_entry));
769 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
7e6ac6e4 770 add_blame_entry(blamed, new_entry);
cee7f245
JH
771 }
772 else if (!split[0].suspect && !split[2].suspect)
1732a1fd
JH
773 /*
774 * The parent covers the entire area; reuse storage for
775 * e and replace it with the parent.
776 */
7e6ac6e4 777 dup_entry(blamed, e, &split[1]);
cee7f245 778 else if (split[0].suspect) {
1732a1fd 779 /* me and then parent */
7e6ac6e4 780 dup_entry(unblamed, e, &split[0]);
cee7f245
JH
781
782 new_entry = xmalloc(sizeof(*new_entry));
783 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
7e6ac6e4 784 add_blame_entry(blamed, new_entry);
cee7f245
JH
785 }
786 else {
1732a1fd 787 /* parent and then me */
7e6ac6e4 788 dup_entry(blamed, e, &split[1]);
cee7f245
JH
789
790 new_entry = xmalloc(sizeof(*new_entry));
791 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
7e6ac6e4 792 add_blame_entry(unblamed, new_entry);
cee7f245
JH
793 }
794}
795
1732a1fd
JH
796/*
797 * After splitting the blame, the origins used by the
798 * on-stack blame_entry should lose one refcnt each.
799 */
54a4c617
JH
800static void decref_split(struct blame_entry *split)
801{
802 int i;
803
804 for (i = 0; i < 3; i++)
805 origin_decref(split[i].suspect);
806}
807
1732a1fd 808/*
7e6ac6e4
DK
809 * reverse_blame reverses the list given in head, appending tail.
810 * That allows us to build lists in reverse order, then reverse them
811 * afterwards. This can be faster than building the list in proper
812 * order right away. The reason is that building in proper order
813 * requires writing a link in the _previous_ element, while building
814 * in reverse order just requires placing the list head into the
815 * _current_ element.
1732a1fd 816 */
cee7f245 817
7e6ac6e4
DK
818static struct blame_entry *reverse_blame(struct blame_entry *head,
819 struct blame_entry *tail)
cee7f245 820{
7e6ac6e4
DK
821 while (head) {
822 struct blame_entry *next = head->next;
823 head->next = tail;
824 tail = head;
825 head = next;
cee7f245 826 }
7e6ac6e4 827 return tail;
cee7f245
JH
828}
829
1732a1fd
JH
830/*
831 * Process one hunk from the patch between the current suspect for
7e6ac6e4
DK
832 * blame_entry e and its parent. This first blames any unfinished
833 * entries before the chunk (which is where target and parent start
834 * differing) on the parent, and then splits blame entries at the
835 * start and at the end of the difference region. Since use of -M and
836 * -C options may lead to overlapping/duplicate source line number
837 * ranges, all we can rely on from sorting/merging is the order of the
838 * first suspect line number.
1732a1fd 839 */
7e6ac6e4
DK
840static void blame_chunk(struct blame_entry ***dstq, struct blame_entry ***srcq,
841 int tlno, int offset, int same,
842 struct origin *parent)
cee7f245 843{
7e6ac6e4
DK
844 struct blame_entry *e = **srcq;
845 struct blame_entry *samep = NULL, *diffp = NULL;
cee7f245 846
7e6ac6e4
DK
847 while (e && e->s_lno < tlno) {
848 struct blame_entry *next = e->next;
849 /*
850 * current record starts before differing portion. If
851 * it reaches into it, we need to split it up and
852 * examine the second part separately.
853 */
854 if (e->s_lno + e->num_lines > tlno) {
855 /* Move second half to a new record */
856 int len = tlno - e->s_lno;
857 struct blame_entry *n = xcalloc(1, sizeof (struct blame_entry));
858 n->suspect = e->suspect;
859 n->lno = e->lno + len;
860 n->s_lno = e->s_lno + len;
861 n->num_lines = e->num_lines - len;
862 e->num_lines = len;
863 e->score = 0;
864 /* Push new record to diffp */
865 n->next = diffp;
866 diffp = n;
867 } else
868 origin_decref(e->suspect);
869 /* Pass blame for everything before the differing
870 * chunk to the parent */
871 e->suspect = origin_incref(parent);
872 e->s_lno += offset;
873 e->next = samep;
874 samep = e;
875 e = next;
876 }
877 /*
878 * As we don't know how much of a common stretch after this
879 * diff will occur, the currently blamed parts are all that we
880 * can assign to the parent for now.
881 */
882
883 if (samep) {
884 **dstq = reverse_blame(samep, **dstq);
885 *dstq = &samep->next;
cee7f245 886 }
7e6ac6e4
DK
887 /*
888 * Prepend the split off portions: everything after e starts
889 * after the blameable portion.
890 */
891 e = reverse_blame(diffp, e);
892
893 /*
894 * Now retain records on the target while parts are different
895 * from the parent.
896 */
897 samep = NULL;
898 diffp = NULL;
899 while (e && e->s_lno < same) {
900 struct blame_entry *next = e->next;
901
902 /*
903 * If current record extends into sameness, need to split.
904 */
905 if (e->s_lno + e->num_lines > same) {
906 /*
907 * Move second half to a new record to be
908 * processed by later chunks
909 */
910 int len = same - e->s_lno;
911 struct blame_entry *n = xcalloc(1, sizeof (struct blame_entry));
912 n->suspect = origin_incref(e->suspect);
913 n->lno = e->lno + len;
914 n->s_lno = e->s_lno + len;
915 n->num_lines = e->num_lines - len;
916 e->num_lines = len;
917 e->score = 0;
918 /* Push new record to samep */
919 n->next = samep;
920 samep = n;
921 }
922 e->next = diffp;
923 diffp = e;
924 e = next;
925 }
926 **srcq = reverse_blame(diffp, reverse_blame(samep, e));
927 /* Move across elements that are in the unblamable portion */
928 if (diffp)
929 *srcq = &diffp->next;
cee7f245
JH
930}
931
cdcab133 932struct blame_chunk_cb_data {
cdcab133 933 struct origin *parent;
7e6ac6e4
DK
934 long offset;
935 struct blame_entry **dstq;
936 struct blame_entry **srcq;
cdcab133
RS
937};
938
7e6ac6e4 939/* diff chunks are from parent to target */
0af596c6
RS
940static int blame_chunk_cb(long start_a, long count_a,
941 long start_b, long count_b, void *data)
cdcab133
RS
942{
943 struct blame_chunk_cb_data *d = data;
7e6ac6e4
DK
944 if (start_a - start_b != d->offset)
945 die("internal error in blame::blame_chunk_cb");
946 blame_chunk(&d->dstq, &d->srcq, start_b, start_a - start_b,
947 start_b + count_b, d->parent);
948 d->offset = start_a + count_a - (start_b + count_b);
0af596c6 949 return 0;
cdcab133
RS
950}
951
1732a1fd
JH
952/*
953 * We are looking at the origin 'target' and aiming to pass blame
954 * for the lines it is suspected to its parent. Run diff to find
955 * which lines came from parent and pass blame for them.
956 */
7e6ac6e4
DK
957static void pass_blame_to_parent(struct scoreboard *sb,
958 struct origin *target,
959 struct origin *parent)
cee7f245 960{
7c4ed8c8 961 mmfile_t file_p, file_o;
66dbfd55 962 struct blame_chunk_cb_data d;
7e6ac6e4 963 struct blame_entry *newdest = NULL;
0af596c6 964
7e6ac6e4
DK
965 if (!target->suspects)
966 return; /* nothing remains for this target */
967
968 d.parent = parent;
969 d.offset = 0;
970 d.dstq = &newdest; d.srcq = &target->suspects;
cee7f245 971
3b8a12e8
AB
972 fill_origin_blob(&sb->revs->diffopt, parent, &file_p);
973 fill_origin_blob(&sb->revs->diffopt, target, &file_o);
7c4ed8c8 974 num_get_patch++;
cee7f245 975
4b4132fd 976 diff_hunks(&file_p, &file_o, 0, blame_chunk_cb, &d);
7e6ac6e4
DK
977 /* The rest are the same as the parent */
978 blame_chunk(&d.dstq, &d.srcq, INT_MAX, d.offset, INT_MAX, parent);
979 *d.dstq = NULL;
980 queue_blames(sb, parent, newdest);
cee7f245 981
7e6ac6e4 982 return;
cee7f245
JH
983}
984
1732a1fd
JH
985/*
986 * The lines in blame_entry after splitting blames many times can become
987 * very small and trivial, and at some point it becomes pointless to
988 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
989 * ordinary C program, and it is not worth to say it was copied from
990 * totally unrelated file in the parent.
991 *
992 * Compute how trivial the lines in the blame_entry are.
993 */
5ff62c30
JH
994static unsigned ent_score(struct scoreboard *sb, struct blame_entry *e)
995{
996 unsigned score;
997 const char *cp, *ep;
998
999 if (e->score)
1000 return e->score;
1001
612702e8 1002 score = 1;
5ff62c30
JH
1003 cp = nth_line(sb, e->lno);
1004 ep = nth_line(sb, e->lno + e->num_lines);
1005 while (cp < ep) {
1006 unsigned ch = *((unsigned char *)cp);
1007 if (isalnum(ch))
1008 score++;
1009 cp++;
1010 }
1011 e->score = score;
1012 return score;
1013}
1014
1732a1fd
JH
1015/*
1016 * best_so_far[] and this[] are both a split of an existing blame_entry
1017 * that passes blame to the parent. Maintain best_so_far the best split
1018 * so far, by comparing this and best_so_far and copying this into
1019 * bst_so_far as needed.
1020 */
5ff62c30 1021static void copy_split_if_better(struct scoreboard *sb,
54a4c617
JH
1022 struct blame_entry *best_so_far,
1023 struct blame_entry *this)
d24bba80 1024{
54a4c617
JH
1025 int i;
1026
d24bba80
JH
1027 if (!this[1].suspect)
1028 return;
5ff62c30
JH
1029 if (best_so_far[1].suspect) {
1030 if (ent_score(sb, &this[1]) < ent_score(sb, &best_so_far[1]))
1031 return;
1032 }
54a4c617
JH
1033
1034 for (i = 0; i < 3; i++)
1035 origin_incref(this[i].suspect);
1036 decref_split(best_so_far);
d24bba80
JH
1037 memcpy(best_so_far, this, sizeof(struct blame_entry [3]));
1038}
1039
dd166aa8
JH
1040/*
1041 * We are looking at a part of the final image represented by
1042 * ent (tlno and same are offset by ent->s_lno).
1043 * tlno is where we are looking at in the final image.
1044 * up to (but not including) same match preimage.
1045 * plno is where we are looking at in the preimage.
1046 *
1047 * <-------------- final image ---------------------->
1048 * <------ent------>
1049 * ^tlno ^same
1050 * <---------preimage----->
1051 * ^plno
1052 *
1053 * All line numbers are 0-based.
1054 */
1055static void handle_split(struct scoreboard *sb,
1056 struct blame_entry *ent,
1057 int tlno, int plno, int same,
1058 struct origin *parent,
1059 struct blame_entry *split)
1060{
1061 if (ent->num_lines <= tlno)
1062 return;
1063 if (tlno < same) {
1064 struct blame_entry this[3];
1065 tlno += ent->s_lno;
1066 same += ent->s_lno;
1067 split_overlap(this, ent, tlno, plno, same, parent);
1068 copy_split_if_better(sb, split, this);
1069 decref_split(this);
1070 }
1071}
1072
cdcab133
RS
1073struct handle_split_cb_data {
1074 struct scoreboard *sb;
1075 struct blame_entry *ent;
1076 struct origin *parent;
1077 struct blame_entry *split;
1078 long plno;
1079 long tlno;
1080};
1081
5d23ec76
RS
1082static int handle_split_cb(long start_a, long count_a,
1083 long start_b, long count_b, void *data)
cdcab133
RS
1084{
1085 struct handle_split_cb_data *d = data;
5d23ec76
RS
1086 handle_split(d->sb, d->ent, d->tlno, d->plno, start_b, d->parent,
1087 d->split);
1088 d->plno = start_a + count_a;
1089 d->tlno = start_b + count_b;
1090 return 0;
cdcab133
RS
1091}
1092
1732a1fd
JH
1093/*
1094 * Find the lines from parent that are the same as ent so that
1095 * we can pass blames to it. file_p has the blob contents for
1096 * the parent.
1097 */
d24bba80
JH
1098static void find_copy_in_blob(struct scoreboard *sb,
1099 struct blame_entry *ent,
1100 struct origin *parent,
54a4c617 1101 struct blame_entry *split,
d24bba80
JH
1102 mmfile_t *file_p)
1103{
1104 const char *cp;
d24bba80 1105 mmfile_t file_o;
66dbfd55 1106 struct handle_split_cb_data d;
5d23ec76 1107
66dbfd55
GV
1108 memset(&d, 0, sizeof(d));
1109 d.sb = sb; d.ent = ent; d.parent = parent; d.split = split;
1732a1fd
JH
1110 /*
1111 * Prepare mmfile that contains only the lines in ent.
1112 */
d24bba80 1113 cp = nth_line(sb, ent->lno);
4b25d091 1114 file_o.ptr = (char *) cp;
3ee8944f 1115 file_o.size = nth_line(sb, ent->lno + ent->num_lines) - cp;
d24bba80 1116
dd166aa8
JH
1117 /*
1118 * file_o is a part of final image we are annotating.
1119 * file_p partially may match that image.
1120 */
d24bba80 1121 memset(split, 0, sizeof(struct blame_entry [3]));
4b4132fd 1122 diff_hunks(file_p, &file_o, 1, handle_split_cb, &d);
dd166aa8 1123 /* remainder, if any, all match the preimage */
cdcab133 1124 handle_split(sb, ent, d.tlno, d.plno, ent->num_lines, parent, split);
d24bba80
JH
1125}
1126
7e6ac6e4
DK
1127/* Move all blame entries from list *source that have a score smaller
1128 * than score_min to the front of list *small.
1129 * Returns a pointer to the link pointing to the old head of the small list.
1130 */
1131
1132static struct blame_entry **filter_small(struct scoreboard *sb,
1133 struct blame_entry **small,
1134 struct blame_entry **source,
1135 unsigned score_min)
1136{
1137 struct blame_entry *p = *source;
1138 struct blame_entry *oldsmall = *small;
1139 while (p) {
1140 if (ent_score(sb, p) <= score_min) {
1141 *small = p;
1142 small = &p->next;
1143 p = *small;
1144 } else {
1145 *source = p;
1146 source = &p->next;
1147 p = *source;
1148 }
1149 }
1150 *small = oldsmall;
1151 *source = NULL;
1152 return small;
1153}
1154
1732a1fd
JH
1155/*
1156 * See if lines currently target is suspected for can be attributed to
1157 * parent.
1158 */
7e6ac6e4
DK
1159static void find_move_in_parent(struct scoreboard *sb,
1160 struct blame_entry ***blamed,
1161 struct blame_entry **toosmall,
1162 struct origin *target,
1163 struct origin *parent)
d24bba80 1164{
46014766 1165 struct blame_entry *e, split[3];
7e6ac6e4
DK
1166 struct blame_entry *unblamed = target->suspects;
1167 struct blame_entry *leftover = NULL;
d24bba80 1168 mmfile_t file_p;
d24bba80 1169
7e6ac6e4
DK
1170 if (!unblamed)
1171 return; /* nothing remains for this target */
d24bba80 1172
3b8a12e8 1173 fill_origin_blob(&sb->revs->diffopt, parent, &file_p);
c2e525d9 1174 if (!file_p.ptr)
7e6ac6e4 1175 return;
d24bba80 1176
7e6ac6e4
DK
1177 /* At each iteration, unblamed has a NULL-terminated list of
1178 * entries that have not yet been tested for blame. leftover
1179 * contains the reversed list of entries that have been tested
1180 * without being assignable to the parent.
1181 */
1182 do {
1183 struct blame_entry **unblamedtail = &unblamed;
1184 struct blame_entry *next;
1185 for (e = unblamed; e; e = next) {
1186 next = e->next;
650e2f67
JH
1187 find_copy_in_blob(sb, e, parent, split, &file_p);
1188 if (split[1].suspect &&
1189 blame_move_score < ent_score(sb, &split[1])) {
7e6ac6e4
DK
1190 split_blame(blamed, &unblamedtail, split, e);
1191 } else {
1192 e->next = leftover;
1193 leftover = e;
650e2f67
JH
1194 }
1195 decref_split(split);
1196 }
7e6ac6e4
DK
1197 *unblamedtail = NULL;
1198 toosmall = filter_small(sb, toosmall, &unblamed, blame_move_score);
1199 } while (unblamed);
1200 target->suspects = reverse_blame(leftover, NULL);
d24bba80
JH
1201}
1202
33494784
JH
1203struct blame_list {
1204 struct blame_entry *ent;
1205 struct blame_entry split[3];
1206};
1207
1732a1fd
JH
1208/*
1209 * Count the number of entries the target is suspected for,
1210 * and prepare a list of entry and the best split.
1211 */
7e6ac6e4 1212static struct blame_list *setup_blame_list(struct blame_entry *unblamed,
33494784
JH
1213 int *num_ents_p)
1214{
1215 struct blame_entry *e;
1216 int num_ents, i;
1217 struct blame_list *blame_list = NULL;
1218
7e6ac6e4
DK
1219 for (e = unblamed, num_ents = 0; e; e = e->next)
1220 num_ents++;
33494784
JH
1221 if (num_ents) {
1222 blame_list = xcalloc(num_ents, sizeof(struct blame_list));
7e6ac6e4
DK
1223 for (e = unblamed, i = 0; e; e = e->next)
1224 blame_list[i++].ent = e;
33494784
JH
1225 }
1226 *num_ents_p = num_ents;
1227 return blame_list;
1228}
1229
1732a1fd
JH
1230/*
1231 * For lines target is suspected for, see if we can find code movement
1232 * across file boundary from the parent commit. porigin is the path
1233 * in the parent we already tried.
1234 */
7e6ac6e4
DK
1235static void find_copy_in_parent(struct scoreboard *sb,
1236 struct blame_entry ***blamed,
1237 struct blame_entry **toosmall,
1238 struct origin *target,
1239 struct commit *parent,
1240 struct origin *porigin,
1241 int opt)
18abd745
JH
1242{
1243 struct diff_options diff_opts;
aec8fa1f 1244 int i, j;
33494784 1245 struct blame_list *blame_list;
aec8fa1f 1246 int num_ents;
7e6ac6e4
DK
1247 struct blame_entry *unblamed = target->suspects;
1248 struct blame_entry *leftover = NULL;
18abd745 1249
7e6ac6e4
DK
1250 if (!unblamed)
1251 return; /* nothing remains for this target */
18abd745
JH
1252
1253 diff_setup(&diff_opts);
8f67f8ae 1254 DIFF_OPT_SET(&diff_opts, RECURSIVE);
18abd745
JH
1255 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
1256
28452655 1257 diff_setup_done(&diff_opts);
33494784
JH
1258
1259 /* Try "find copies harder" on new path if requested;
1260 * we do not want to use diffcore_rename() actually to
1261 * match things up; find_copies_harder is set only to
1262 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1263 * and this code needs to be after diff_setup_done(), which
1264 * usually makes find-copies-harder imply copy detection.
1265 */
c63777c0
JH
1266 if ((opt & PICKAXE_BLAME_COPY_HARDEST)
1267 || ((opt & PICKAXE_BLAME_COPY_HARDER)
1268 && (!porigin || strcmp(target->path, porigin->path))))
8f67f8ae 1269 DIFF_OPT_SET(&diff_opts, FIND_COPIES_HARDER);
33494784 1270
1cfe7733
JH
1271 if (is_null_sha1(target->commit->object.sha1))
1272 do_diff_cache(parent->tree->object.sha1, &diff_opts);
1273 else
1274 diff_tree_sha1(parent->tree->object.sha1,
1275 target->commit->tree->object.sha1,
1276 "", &diff_opts);
18abd745 1277
8f67f8ae 1278 if (!DIFF_OPT_TST(&diff_opts, FIND_COPIES_HARDER))
33494784 1279 diffcore_std(&diff_opts);
18abd745 1280
7e6ac6e4
DK
1281 do {
1282 struct blame_entry **unblamedtail = &unblamed;
1283 blame_list = setup_blame_list(unblamed, &num_ents);
33494784
JH
1284
1285 for (i = 0; i < diff_queued_diff.nr; i++) {
1286 struct diff_filepair *p = diff_queued_diff.queue[i];
1287 struct origin *norigin;
1288 mmfile_t file_p;
33494784
JH
1289 struct blame_entry this[3];
1290
1291 if (!DIFF_FILE_VALID(p->one))
1292 continue; /* does not exist in parent */
5209ac4d
AG
1293 if (S_ISGITLINK(p->one->mode))
1294 continue; /* ignore git links */
33494784
JH
1295 if (porigin && !strcmp(p->one->path, porigin->path))
1296 /* find_move already dealt with this path */
1297 continue;
1298
1299 norigin = get_origin(sb, parent, p->one->path);
1300 hashcpy(norigin->blob_sha1, p->one->sha1);
90064710 1301 norigin->mode = p->one->mode;
3b8a12e8 1302 fill_origin_blob(&sb->revs->diffopt, norigin, &file_p);
33494784
JH
1303 if (!file_p.ptr)
1304 continue;
1305
1306 for (j = 0; j < num_ents; j++) {
1307 find_copy_in_blob(sb, blame_list[j].ent,
1308 norigin, this, &file_p);
1309 copy_split_if_better(sb, blame_list[j].split,
1310 this);
1311 decref_split(this);
1312 }
33494784
JH
1313 origin_decref(norigin);
1314 }
18abd745 1315
aec8fa1f 1316 for (j = 0; j < num_ents; j++) {
33494784
JH
1317 struct blame_entry *split = blame_list[j].split;
1318 if (split[1].suspect &&
1319 blame_copy_score < ent_score(sb, &split[1])) {
7e6ac6e4
DK
1320 split_blame(blamed, &unblamedtail, split,
1321 blame_list[j].ent);
1322 } else {
1323 blame_list[j].ent->next = leftover;
1324 leftover = blame_list[j].ent;
33494784
JH
1325 }
1326 decref_split(split);
18abd745 1327 }
33494784 1328 free(blame_list);
7e6ac6e4
DK
1329 *unblamedtail = NULL;
1330 toosmall = filter_small(sb, toosmall, &unblamed, blame_copy_score);
1331 } while (unblamed);
1332 target->suspects = reverse_blame(leftover, NULL);
33494784 1333 diff_flush(&diff_opts);
bd1928df 1334 free_pathspec(&diff_opts.pathspec);
18abd745
JH
1335}
1336
1732a1fd
JH
1337/*
1338 * The blobs of origin and porigin exactly match, so everything
c2e525d9
JH
1339 * origin is suspected for can be blamed on the parent.
1340 */
1341static void pass_whole_blame(struct scoreboard *sb,
1342 struct origin *origin, struct origin *porigin)
1343{
7e6ac6e4 1344 struct blame_entry *e, *suspects;
c2e525d9
JH
1345
1346 if (!porigin->file.ptr && origin->file.ptr) {
1347 /* Steal its file */
1348 porigin->file = origin->file;
1349 origin->file.ptr = NULL;
1350 }
7e6ac6e4
DK
1351 suspects = origin->suspects;
1352 origin->suspects = NULL;
1353 for (e = suspects; e; e = e->next) {
c2e525d9
JH
1354 origin_incref(porigin);
1355 origin_decref(e->suspect);
1356 e->suspect = porigin;
1357 }
7e6ac6e4 1358 queue_blames(sb, porigin, suspects);
c2e525d9
JH
1359}
1360
69264f46
JH
1361/*
1362 * We pass blame from the current commit to its parents. We keep saying
1363 * "parent" (and "porigin"), but what we mean is to find scapegoat to
1364 * exonerate ourselves.
1365 */
85af7929 1366static struct commit_list *first_scapegoat(struct rev_info *revs, struct commit *commit)
69264f46 1367{
95a4fb0e
JK
1368 if (!reverse) {
1369 if (revs->first_parent_only &&
1370 commit->parents &&
1371 commit->parents->next) {
1372 free_commit_list(commit->parents->next);
1373 commit->parents->next = NULL;
1374 }
85af7929 1375 return commit->parents;
95a4fb0e 1376 }
85af7929 1377 return lookup_decoration(&revs->children, &commit->object);
69264f46
JH
1378}
1379
85af7929 1380static int num_scapegoats(struct rev_info *revs, struct commit *commit)
69264f46 1381{
85af7929 1382 struct commit_list *l = first_scapegoat(revs, commit);
4bbaa1eb 1383 return commit_list_count(l);
69264f46
JH
1384}
1385
7e6ac6e4
DK
1386/* Distribute collected unsorted blames to the respected sorted lists
1387 * in the various origins.
1388 */
1389static void distribute_blame(struct scoreboard *sb, struct blame_entry *blamed)
1390{
1391 blamed = blame_sort(blamed, compare_blame_suspect);
1392 while (blamed)
1393 {
1394 struct origin *porigin = blamed->suspect;
1395 struct blame_entry *suspects = NULL;
1396 do {
1397 struct blame_entry *next = blamed->next;
1398 blamed->next = suspects;
1399 suspects = blamed;
1400 blamed = next;
1401 } while (blamed && blamed->suspect == porigin);
1402 suspects = reverse_blame(suspects, NULL);
1403 queue_blames(sb, porigin, suspects);
1404 }
1405}
1406
69264f46 1407#define MAXSG 16
cee7f245 1408
d24bba80 1409static void pass_blame(struct scoreboard *sb, struct origin *origin, int opt)
cee7f245 1410{
85af7929 1411 struct rev_info *revs = sb->revs;
69264f46 1412 int i, pass, num_sg;
cee7f245 1413 struct commit *commit = origin->commit;
69264f46
JH
1414 struct commit_list *sg;
1415 struct origin *sg_buf[MAXSG];
1416 struct origin *porigin, **sg_origin = sg_buf;
7e6ac6e4
DK
1417 struct blame_entry *toosmall = NULL;
1418 struct blame_entry *blames, **blametail = &blames;
69264f46 1419
85af7929 1420 num_sg = num_scapegoats(revs, commit);
69264f46
JH
1421 if (!num_sg)
1422 goto finish;
1423 else if (num_sg < ARRAY_SIZE(sg_buf))
1424 memset(sg_buf, 0, sizeof(sg_buf));
1425 else
1426 sg_origin = xcalloc(num_sg, sizeof(*sg_origin));
cee7f245 1427
69264f46
JH
1428 /*
1429 * The first pass looks for unrenamed path to optimize for
f69e743d
JH
1430 * common cases, then we look for renames in the second pass.
1431 */
3d1aa566 1432 for (pass = 0; pass < 2 - no_whole_file_rename; pass++) {
f69e743d
JH
1433 struct origin *(*find)(struct scoreboard *,
1434 struct commit *, struct origin *);
1435 find = pass ? find_rename : find_origin;
1436
85af7929 1437 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1438 i < num_sg && sg;
1439 sg = sg->next, i++) {
1440 struct commit *p = sg->item;
0421d9f8 1441 int j, same;
f69e743d 1442
69264f46 1443 if (sg_origin[i])
f69e743d
JH
1444 continue;
1445 if (parse_commit(p))
1446 continue;
0d981c67 1447 porigin = find(sb, p, origin);
f69e743d
JH
1448 if (!porigin)
1449 continue;
1450 if (!hashcmp(porigin->blob_sha1, origin->blob_sha1)) {
c2e525d9 1451 pass_whole_blame(sb, origin, porigin);
f69e743d
JH
1452 origin_decref(porigin);
1453 goto finish;
1454 }
0421d9f8 1455 for (j = same = 0; j < i; j++)
69264f46
JH
1456 if (sg_origin[j] &&
1457 !hashcmp(sg_origin[j]->blob_sha1,
0421d9f8
JH
1458 porigin->blob_sha1)) {
1459 same = 1;
1460 break;
1461 }
1462 if (!same)
69264f46 1463 sg_origin[i] = porigin;
0421d9f8
JH
1464 else
1465 origin_decref(porigin);
cee7f245 1466 }
cee7f245
JH
1467 }
1468
c2e525d9 1469 num_commits++;
85af7929 1470 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1471 i < num_sg && sg;
1472 sg = sg->next, i++) {
1473 struct origin *porigin = sg_origin[i];
cee7f245
JH
1474 if (!porigin)
1475 continue;
96e11709
JH
1476 if (!origin->previous) {
1477 origin_incref(porigin);
1478 origin->previous = porigin;
1479 }
7e6ac6e4
DK
1480 pass_blame_to_parent(sb, origin, porigin);
1481 if (!origin->suspects)
54a4c617 1482 goto finish;
cee7f245 1483 }
d24bba80
JH
1484
1485 /*
1732a1fd 1486 * Optionally find moves in parents' files.
d24bba80 1487 */
7e6ac6e4
DK
1488 if (opt & PICKAXE_BLAME_MOVE) {
1489 filter_small(sb, &toosmall, &origin->suspects, blame_move_score);
1490 if (origin->suspects) {
1491 for (i = 0, sg = first_scapegoat(revs, commit);
1492 i < num_sg && sg;
1493 sg = sg->next, i++) {
1494 struct origin *porigin = sg_origin[i];
1495 if (!porigin)
1496 continue;
1497 find_move_in_parent(sb, &blametail, &toosmall, origin, porigin);
1498 if (!origin->suspects)
1499 break;
1500 }
d24bba80 1501 }
7e6ac6e4 1502 }
d24bba80 1503
18abd745 1504 /*
1732a1fd 1505 * Optionally find copies from parents' files.
18abd745 1506 */
7e6ac6e4
DK
1507 if (opt & PICKAXE_BLAME_COPY) {
1508 if (blame_copy_score > blame_move_score)
1509 filter_small(sb, &toosmall, &origin->suspects, blame_copy_score);
1510 else if (blame_copy_score < blame_move_score) {
1511 origin->suspects = blame_merge(origin->suspects, toosmall);
1512 toosmall = NULL;
1513 filter_small(sb, &toosmall, &origin->suspects, blame_copy_score);
1514 }
1515 if (!origin->suspects)
1516 goto finish;
1517
85af7929 1518 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1519 i < num_sg && sg;
1520 sg = sg->next, i++) {
1521 struct origin *porigin = sg_origin[i];
7e6ac6e4
DK
1522 find_copy_in_parent(sb, &blametail, &toosmall,
1523 origin, sg->item, porigin, opt);
1524 if (!origin->suspects)
54a4c617 1525 goto finish;
18abd745 1526 }
7e6ac6e4 1527 }
54a4c617 1528
7e6ac6e4
DK
1529finish:
1530 *blametail = NULL;
1531 distribute_blame(sb, blames);
1532 /*
1533 * prepend toosmall to origin->suspects
1534 *
1535 * There is no point in sorting: this ends up on a big
1536 * unsorted list in the caller anyway.
1537 */
1538 if (toosmall) {
1539 struct blame_entry **tail = &toosmall;
1540 while (*tail)
1541 tail = &(*tail)->next;
1542 *tail = origin->suspects;
1543 origin->suspects = toosmall;
1544 }
69264f46
JH
1545 for (i = 0; i < num_sg; i++) {
1546 if (sg_origin[i]) {
1547 drop_origin_blob(sg_origin[i]);
1548 origin_decref(sg_origin[i]);
7c3c7962
JH
1549 }
1550 }
1551 drop_origin_blob(origin);
69264f46
JH
1552 if (sg_buf != sg_origin)
1553 free(sg_origin);
cee7f245
JH
1554}
1555
1732a1fd
JH
1556/*
1557 * Information on commits, used for output.
1558 */
9cba13ca 1559struct commit_info {
ea02ffa3
AP
1560 struct strbuf author;
1561 struct strbuf author_mail;
cee7f245 1562 unsigned long author_time;
ea02ffa3 1563 struct strbuf author_tz;
cee7f245
JH
1564
1565 /* filled only when asked for details */
ea02ffa3
AP
1566 struct strbuf committer;
1567 struct strbuf committer_mail;
cee7f245 1568 unsigned long committer_time;
ea02ffa3 1569 struct strbuf committer_tz;
cee7f245 1570
ea02ffa3 1571 struct strbuf summary;
cee7f245
JH
1572};
1573
1732a1fd
JH
1574/*
1575 * Parse author/committer line in the commit object buffer
1576 */
cee7f245 1577static void get_ac_line(const char *inbuf, const char *what,
ea02ffa3
AP
1578 struct strbuf *name, struct strbuf *mail,
1579 unsigned long *time, struct strbuf *tz)
cee7f245 1580{
3c020bd5 1581 struct ident_split ident;
ea02ffa3
AP
1582 size_t len, maillen, namelen;
1583 char *tmp, *endp;
1584 const char *namebuf, *mailbuf;
cee7f245
JH
1585
1586 tmp = strstr(inbuf, what);
1587 if (!tmp)
1588 goto error_out;
1589 tmp += strlen(what);
1590 endp = strchr(tmp, '\n');
1591 if (!endp)
1592 len = strlen(tmp);
1593 else
1594 len = endp - tmp;
3c020bd5
AP
1595
1596 if (split_ident_line(&ident, tmp, len)) {
cee7f245
JH
1597 error_out:
1598 /* Ugh */
ea02ffa3
AP
1599 tmp = "(unknown)";
1600 strbuf_addstr(name, tmp);
1601 strbuf_addstr(mail, tmp);
1602 strbuf_addstr(tz, tmp);
cee7f245
JH
1603 *time = 0;
1604 return;
1605 }
cee7f245 1606
3c020bd5 1607 namelen = ident.name_end - ident.name_begin;
ea02ffa3 1608 namebuf = ident.name_begin;
cee7f245 1609
ea02ffa3
AP
1610 maillen = ident.mail_end - ident.mail_begin;
1611 mailbuf = ident.mail_begin;
f9567384 1612
de5abe9f
RS
1613 if (ident.date_begin && ident.date_end)
1614 *time = strtoul(ident.date_begin, NULL, 10);
1615 else
1616 *time = 0;
f9567384 1617
de5abe9f
RS
1618 if (ident.tz_begin && ident.tz_end)
1619 strbuf_add(tz, ident.tz_begin, ident.tz_end - ident.tz_begin);
1620 else
1621 strbuf_addstr(tz, "(unknown)");
f9567384 1622
f9567384 1623 /*
d20d654f 1624 * Now, convert both name and e-mail using mailmap
f9567384 1625 */
ea02ffa3
AP
1626 map_user(&mailmap, &mailbuf, &maillen,
1627 &namebuf, &namelen);
1628
1629 strbuf_addf(mail, "<%.*s>", (int)maillen, mailbuf);
1630 strbuf_add(name, namebuf, namelen);
1631}
1632
1633static void commit_info_init(struct commit_info *ci)
1634{
1635
1636 strbuf_init(&ci->author, 0);
1637 strbuf_init(&ci->author_mail, 0);
1638 strbuf_init(&ci->author_tz, 0);
1639 strbuf_init(&ci->committer, 0);
1640 strbuf_init(&ci->committer_mail, 0);
1641 strbuf_init(&ci->committer_tz, 0);
1642 strbuf_init(&ci->summary, 0);
1643}
1644
1645static void commit_info_destroy(struct commit_info *ci)
1646{
1647
1648 strbuf_release(&ci->author);
1649 strbuf_release(&ci->author_mail);
1650 strbuf_release(&ci->author_tz);
1651 strbuf_release(&ci->committer);
1652 strbuf_release(&ci->committer_mail);
1653 strbuf_release(&ci->committer_tz);
1654 strbuf_release(&ci->summary);
cee7f245
JH
1655}
1656
1657static void get_commit_info(struct commit *commit,
1658 struct commit_info *ret,
1659 int detailed)
1660{
1661 int len;
e297cf5a 1662 const char *subject, *encoding;
b000c59b 1663 const char *message;
ea02ffa3
AP
1664
1665 commit_info_init(ret);
cee7f245 1666
e297cf5a 1667 encoding = get_log_output_encoding();
5a10d236 1668 message = logmsg_reencode(commit, NULL, encoding);
69cd8f63 1669 get_ac_line(message, "\nauthor ",
ea02ffa3 1670 &ret->author, &ret->author_mail,
cee7f245
JH
1671 &ret->author_time, &ret->author_tz);
1672
69cd8f63 1673 if (!detailed) {
b66103c3 1674 unuse_commit_buffer(commit, message);
cee7f245 1675 return;
69cd8f63 1676 }
cee7f245 1677
69cd8f63 1678 get_ac_line(message, "\ncommitter ",
ea02ffa3 1679 &ret->committer, &ret->committer_mail,
cee7f245
JH
1680 &ret->committer_time, &ret->committer_tz);
1681
ad98a58b 1682 len = find_commit_subject(message, &subject);
ea02ffa3
AP
1683 if (len)
1684 strbuf_add(&ret->summary, subject, len);
1685 else
1686 strbuf_addf(&ret->summary, "(%s)", sha1_to_hex(commit->object.sha1));
1687
b66103c3 1688 unuse_commit_buffer(commit, message);
cee7f245
JH
1689}
1690
1732a1fd
JH
1691/*
1692 * To allow LF and other nonportable characters in pathnames,
1693 * they are c-style quoted as needed.
1694 */
46e5e69d
JH
1695static void write_filename_info(const char *path)
1696{
1697 printf("filename ");
663af342 1698 write_name_quoted(path, stdout, '\n');
46e5e69d
JH
1699}
1700
9991030c
JH
1701/*
1702 * Porcelain/Incremental format wants to show a lot of details per
1703 * commit. Instead of repeating this every line, emit it only once,
e86226e3
JK
1704 * the first time each commit appears in the output (unless the
1705 * user has specifically asked for us to repeat).
9991030c 1706 */
e86226e3 1707static int emit_one_suspect_detail(struct origin *suspect, int repeat)
9991030c
JH
1708{
1709 struct commit_info ci;
1710
e86226e3 1711 if (!repeat && (suspect->commit->object.flags & METAINFO_SHOWN))
9991030c
JH
1712 return 0;
1713
1714 suspect->commit->object.flags |= METAINFO_SHOWN;
1715 get_commit_info(suspect->commit, &ci, 1);
ea02ffa3
AP
1716 printf("author %s\n", ci.author.buf);
1717 printf("author-mail %s\n", ci.author_mail.buf);
9991030c 1718 printf("author-time %lu\n", ci.author_time);
ea02ffa3
AP
1719 printf("author-tz %s\n", ci.author_tz.buf);
1720 printf("committer %s\n", ci.committer.buf);
1721 printf("committer-mail %s\n", ci.committer_mail.buf);
9991030c 1722 printf("committer-time %lu\n", ci.committer_time);
ea02ffa3
AP
1723 printf("committer-tz %s\n", ci.committer_tz.buf);
1724 printf("summary %s\n", ci.summary.buf);
9991030c
JH
1725 if (suspect->commit->object.flags & UNINTERESTING)
1726 printf("boundary\n");
96e11709
JH
1727 if (suspect->previous) {
1728 struct origin *prev = suspect->previous;
1729 printf("previous %s ", sha1_to_hex(prev->commit->object.sha1));
1730 write_name_quoted(prev->path, stdout, '\n');
1731 }
ea02ffa3
AP
1732
1733 commit_info_destroy(&ci);
1734
9991030c
JH
1735 return 1;
1736}
1737
1732a1fd 1738/*
7e6ac6e4
DK
1739 * The blame_entry is found to be guilty for the range.
1740 * Show it in incremental output.
1732a1fd 1741 */
717d1462
LT
1742static void found_guilty_entry(struct blame_entry *ent)
1743{
717d1462
LT
1744 if (incremental) {
1745 struct origin *suspect = ent->suspect;
1746
1747 printf("%s %d %d %d\n",
1748 sha1_to_hex(suspect->commit->object.sha1),
1749 ent->s_lno + 1, ent->lno + 1, ent->num_lines);
e86226e3 1750 emit_one_suspect_detail(suspect, 0);
46e5e69d 1751 write_filename_info(suspect->path);
06f59e9f 1752 maybe_flush_or_die(stdout, "stdout");
717d1462
LT
1753 }
1754}
1755
1732a1fd 1756/*
7e6ac6e4
DK
1757 * The main loop -- while we have blobs with lines whose true origin
1758 * is still unknown, pick one blob, and allow its lines to pass blames
1759 * to its parents. */
85af7929 1760static void assign_blame(struct scoreboard *sb, int opt)
717d1462 1761{
85af7929 1762 struct rev_info *revs = sb->revs;
7e6ac6e4 1763 struct commit *commit = prio_queue_get(&sb->commits);
85af7929 1764
7e6ac6e4 1765 while (commit) {
717d1462 1766 struct blame_entry *ent;
7e6ac6e4 1767 struct origin *suspect = commit->util;
717d1462
LT
1768
1769 /* find one suspect to break down */
7e6ac6e4
DK
1770 while (suspect && !suspect->suspects)
1771 suspect = suspect->next;
1772
1773 if (!suspect) {
1774 commit = prio_queue_get(&sb->commits);
1775 continue;
1776 }
1777
1778 assert(commit == suspect->commit);
717d1462 1779
1732a1fd
JH
1780 /*
1781 * We will use this suspect later in the loop,
1782 * so hold onto it in the meantime.
1783 */
717d1462 1784 origin_incref(suspect);
0064053b 1785 parse_commit(commit);
85af7929
JH
1786 if (reverse ||
1787 (!(commit->object.flags & UNINTERESTING) &&
1788 !(revs->max_age != -1 && commit->date < revs->max_age)))
717d1462
LT
1789 pass_blame(sb, suspect, opt);
1790 else {
1791 commit->object.flags |= UNINTERESTING;
1792 if (commit->object.parsed)
1793 mark_parents_uninteresting(commit);
1794 }
1795 /* treat root commit as boundary */
1796 if (!commit->parents && !show_root)
1797 commit->object.flags |= UNINTERESTING;
1798
1799 /* Take responsibility for the remaining entries */
7e6ac6e4
DK
1800 ent = suspect->suspects;
1801 if (ent) {
1802 suspect->guilty = 1;
1803 for (;;) {
1804 struct blame_entry *next = ent->next;
717d1462 1805 found_guilty_entry(ent);
7e6ac6e4
DK
1806 if (next) {
1807 ent = next;
1808 continue;
1809 }
1810 ent->next = sb->ent;
1811 sb->ent = suspect->suspects;
1812 suspect->suspects = NULL;
1813 break;
1814 }
1815 }
717d1462
LT
1816 origin_decref(suspect);
1817
1818 if (DEBUG) /* sanity */
1819 sanity_check_refcnt(sb);
1820 }
1821}
1822
1823static const char *format_time(unsigned long time, const char *tz_str,
1824 int show_raw_time)
1825{
bccce0f8 1826 static struct strbuf time_buf = STRBUF_INIT;
717d1462 1827
bccce0f8 1828 strbuf_reset(&time_buf);
717d1462 1829 if (show_raw_time) {
bccce0f8 1830 strbuf_addf(&time_buf, "%lu %s", time, tz_str);
717d1462 1831 }
31653c1a 1832 else {
ac39b277 1833 const char *time_str;
bccce0f8 1834 size_t time_width;
ac39b277 1835 int tz;
31653c1a
EL
1836 tz = atoi(tz_str);
1837 time_str = show_date(time, tz, blame_date_mode);
bccce0f8
JX
1838 strbuf_addstr(&time_buf, time_str);
1839 /*
1840 * Add space paddings to time_buf to display a fixed width
1841 * string, and use time_width for display width calibration.
1842 */
1843 for (time_width = utf8_strwidth(time_str);
1844 time_width < blame_date_width;
1845 time_width++)
1846 strbuf_addch(&time_buf, ' ');
31653c1a 1847 }
bccce0f8 1848 return time_buf.buf;
717d1462
LT
1849}
1850
cee7f245
JH
1851#define OUTPUT_ANNOTATE_COMPAT 001
1852#define OUTPUT_LONG_OBJECT_NAME 002
1853#define OUTPUT_RAW_TIMESTAMP 004
1854#define OUTPUT_PORCELAIN 010
1855#define OUTPUT_SHOW_NAME 020
1856#define OUTPUT_SHOW_NUMBER 040
5ff62c30 1857#define OUTPUT_SHOW_SCORE 0100
093dc5be 1858#define OUTPUT_NO_AUTHOR 0200
1b8cdce9 1859#define OUTPUT_SHOW_EMAIL 0400
ed747dd5 1860#define OUTPUT_LINE_PORCELAIN 01000
cee7f245 1861
e86226e3
JK
1862static void emit_porcelain_details(struct origin *suspect, int repeat)
1863{
1864 if (emit_one_suspect_detail(suspect, repeat) ||
1865 (suspect->commit->object.flags & MORE_THAN_ONE_PATH))
1866 write_filename_info(suspect->path);
1867}
1868
1869static void emit_porcelain(struct scoreboard *sb, struct blame_entry *ent,
1870 int opt)
cee7f245 1871{
ed747dd5 1872 int repeat = opt & OUTPUT_LINE_PORCELAIN;
cee7f245
JH
1873 int cnt;
1874 const char *cp;
1875 struct origin *suspect = ent->suspect;
1876 char hex[41];
1877
1878 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
7e6ac6e4 1879 printf("%s %d %d %d\n",
cee7f245 1880 hex,
cee7f245
JH
1881 ent->s_lno + 1,
1882 ent->lno + 1,
1883 ent->num_lines);
ed747dd5 1884 emit_porcelain_details(suspect, repeat);
cee7f245
JH
1885
1886 cp = nth_line(sb, ent->lno);
1887 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1888 char ch;
ed747dd5 1889 if (cnt) {
cee7f245
JH
1890 printf("%s %d %d\n", hex,
1891 ent->s_lno + 1 + cnt,
1892 ent->lno + 1 + cnt);
ed747dd5
JK
1893 if (repeat)
1894 emit_porcelain_details(suspect, 1);
1895 }
cee7f245
JH
1896 putchar('\t');
1897 do {
1898 ch = *cp++;
1899 putchar(ch);
1900 } while (ch != '\n' &&
1901 cp < sb->final_buf + sb->final_buf_size);
1902 }
a5ca8367
JS
1903
1904 if (sb->final_buf_size && cp[-1] != '\n')
1905 putchar('\n');
cee7f245
JH
1906}
1907
1908static void emit_other(struct scoreboard *sb, struct blame_entry *ent, int opt)
1909{
1910 int cnt;
1911 const char *cp;
1912 struct origin *suspect = ent->suspect;
1913 struct commit_info ci;
1914 char hex[41];
1915 int show_raw_time = !!(opt & OUTPUT_RAW_TIMESTAMP);
1916
1917 get_commit_info(suspect->commit, &ci, 1);
1918 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1919
1920 cp = nth_line(sb, ent->lno);
1921 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1922 char ch;
84393bfd 1923 int length = (opt & OUTPUT_LONG_OBJECT_NAME) ? 40 : abbrev;
b11121d9
JH
1924
1925 if (suspect->commit->object.flags & UNINTERESTING) {
e68989a7
JH
1926 if (blank_boundary)
1927 memset(hex, ' ', length);
7ceacdff 1928 else if (!(opt & OUTPUT_ANNOTATE_COMPAT)) {
4c10a5ca
JH
1929 length--;
1930 putchar('^');
1931 }
b11121d9 1932 }
cee7f245 1933
b11121d9 1934 printf("%.*s", length, hex);
1b8cdce9
KB
1935 if (opt & OUTPUT_ANNOTATE_COMPAT) {
1936 const char *name;
1937 if (opt & OUTPUT_SHOW_EMAIL)
ea02ffa3 1938 name = ci.author_mail.buf;
1b8cdce9 1939 else
ea02ffa3 1940 name = ci.author.buf;
1b8cdce9 1941 printf("\t(%10s\t%10s\t%d)", name,
ea02ffa3 1942 format_time(ci.author_time, ci.author_tz.buf,
cee7f245
JH
1943 show_raw_time),
1944 ent->lno + 1 + cnt);
1b8cdce9 1945 } else {
5ff62c30 1946 if (opt & OUTPUT_SHOW_SCORE)
54a4c617
JH
1947 printf(" %*d %02d",
1948 max_score_digits, ent->score,
1949 ent->suspect->refcnt);
cee7f245
JH
1950 if (opt & OUTPUT_SHOW_NAME)
1951 printf(" %-*.*s", longest_file, longest_file,
1952 suspect->path);
1953 if (opt & OUTPUT_SHOW_NUMBER)
1954 printf(" %*d", max_orig_digits,
1955 ent->s_lno + 1 + cnt);
093dc5be 1956
ffaf9cc0 1957 if (!(opt & OUTPUT_NO_AUTHOR)) {
1b8cdce9
KB
1958 const char *name;
1959 int pad;
1960 if (opt & OUTPUT_SHOW_EMAIL)
ea02ffa3 1961 name = ci.author_mail.buf;
1b8cdce9 1962 else
ea02ffa3 1963 name = ci.author.buf;
1b8cdce9 1964 pad = longest_author - utf8_strwidth(name);
ffaf9cc0 1965 printf(" (%s%*s %10s",
1b8cdce9 1966 name, pad, "",
093dc5be 1967 format_time(ci.author_time,
ea02ffa3 1968 ci.author_tz.buf,
093dc5be 1969 show_raw_time));
ffaf9cc0 1970 }
093dc5be 1971 printf(" %*d) ",
cee7f245
JH
1972 max_digits, ent->lno + 1 + cnt);
1973 }
1974 do {
1975 ch = *cp++;
1976 putchar(ch);
1977 } while (ch != '\n' &&
1978 cp < sb->final_buf + sb->final_buf_size);
1979 }
a5ca8367
JS
1980
1981 if (sb->final_buf_size && cp[-1] != '\n')
1982 putchar('\n');
ea02ffa3
AP
1983
1984 commit_info_destroy(&ci);
cee7f245
JH
1985}
1986
1987static void output(struct scoreboard *sb, int option)
1988{
1989 struct blame_entry *ent;
1990
1991 if (option & OUTPUT_PORCELAIN) {
1992 for (ent = sb->ent; ent; ent = ent->next) {
7e6ac6e4
DK
1993 int count = 0;
1994 struct origin *suspect;
1995 struct commit *commit = ent->suspect->commit;
cee7f245
JH
1996 if (commit->object.flags & MORE_THAN_ONE_PATH)
1997 continue;
7e6ac6e4
DK
1998 for (suspect = commit->util; suspect; suspect = suspect->next) {
1999 if (suspect->guilty && count++) {
2000 commit->object.flags |= MORE_THAN_ONE_PATH;
2001 break;
2002 }
cee7f245
JH
2003 }
2004 }
2005 }
2006
2007 for (ent = sb->ent; ent; ent = ent->next) {
2008 if (option & OUTPUT_PORCELAIN)
e86226e3 2009 emit_porcelain(sb, ent, option);
5ff62c30 2010 else {
cee7f245 2011 emit_other(sb, ent, option);
5ff62c30 2012 }
cee7f245
JH
2013 }
2014}
2015
29aa0b20
RS
2016static const char *get_next_line(const char *start, const char *end)
2017{
2018 const char *nl = memchr(start, '\n', end - start);
60d85e11 2019 return nl ? nl + 1 : end;
29aa0b20
RS
2020}
2021
1732a1fd
JH
2022/*
2023 * To allow quick access to the contents of nth line in the
2024 * final image, prepare an index in the scoreboard.
2025 */
cee7f245
JH
2026static int prepare_lines(struct scoreboard *sb)
2027{
2028 const char *buf = sb->final_buf;
2029 unsigned long len = sb->final_buf_size;
352bbbd9
DK
2030 const char *end = buf + len;
2031 const char *p;
2032 int *lineno;
60d85e11 2033 int num = 0;
352bbbd9 2034
60d85e11 2035 for (p = buf; p < end; p = get_next_line(p, end))
29aa0b20 2036 num++;
352bbbd9 2037
60d85e11 2038 sb->lineno = lineno = xmalloc(sizeof(*sb->lineno) * (num + 1));
352bbbd9 2039
60d85e11 2040 for (p = buf; p < end; p = get_next_line(p, end))
29aa0b20 2041 *lineno++ = p - buf;
352bbbd9 2042
60d85e11 2043 *lineno = len;
352bbbd9 2044
60d85e11 2045 sb->num_lines = num;
cee7f245
JH
2046 return sb->num_lines;
2047}
2048
1732a1fd
JH
2049/*
2050 * Add phony grafts for use with -S; this is primarily to
34baebce 2051 * support git's cvsserver that wants to give a linear history
1732a1fd
JH
2052 * to its clients.
2053 */
cee7f245
JH
2054static int read_ancestry(const char *graft_file)
2055{
2056 FILE *fp = fopen(graft_file, "r");
e228c173 2057 struct strbuf buf = STRBUF_INIT;
cee7f245
JH
2058 if (!fp)
2059 return -1;
e228c173 2060 while (!strbuf_getwholeline(&buf, fp, '\n')) {
cee7f245 2061 /* The format is just "Commit Parent1 Parent2 ...\n" */
e228c173 2062 struct commit_graft *graft = read_graft_line(buf.buf, buf.len);
8eaf7986
JH
2063 if (graft)
2064 register_commit_graft(graft, 0);
cee7f245
JH
2065 }
2066 fclose(fp);
e228c173 2067 strbuf_release(&buf);
cee7f245
JH
2068 return 0;
2069}
2070
b31272f7
JH
2071static int update_auto_abbrev(int auto_abbrev, struct origin *suspect)
2072{
2073 const char *uniq = find_unique_abbrev(suspect->commit->object.sha1,
2074 auto_abbrev);
2075 int len = strlen(uniq);
2076 if (auto_abbrev < len)
2077 return len;
2078 return auto_abbrev;
2079}
2080
1732a1fd
JH
2081/*
2082 * How many columns do we need to show line numbers, authors,
2083 * and filenames?
2084 */
cee7f245
JH
2085static void find_alignment(struct scoreboard *sb, int *option)
2086{
2087 int longest_src_lines = 0;
2088 int longest_dst_lines = 0;
5ff62c30 2089 unsigned largest_score = 0;
cee7f245 2090 struct blame_entry *e;
b31272f7
JH
2091 int compute_auto_abbrev = (abbrev < 0);
2092 int auto_abbrev = default_abbrev;
cee7f245
JH
2093
2094 for (e = sb->ent; e; e = e->next) {
2095 struct origin *suspect = e->suspect;
cee7f245
JH
2096 int num;
2097
b31272f7
JH
2098 if (compute_auto_abbrev)
2099 auto_abbrev = update_auto_abbrev(auto_abbrev, suspect);
ab3bb800
JH
2100 if (strcmp(suspect->path, sb->path))
2101 *option |= OUTPUT_SHOW_NAME;
2102 num = strlen(suspect->path);
2103 if (longest_file < num)
2104 longest_file = num;
cee7f245 2105 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
e6005927 2106 struct commit_info ci;
cee7f245
JH
2107 suspect->commit->object.flags |= METAINFO_SHOWN;
2108 get_commit_info(suspect->commit, &ci, 1);
1b8cdce9 2109 if (*option & OUTPUT_SHOW_EMAIL)
ea02ffa3 2110 num = utf8_strwidth(ci.author_mail.buf);
1b8cdce9 2111 else
ea02ffa3 2112 num = utf8_strwidth(ci.author.buf);
cee7f245
JH
2113 if (longest_author < num)
2114 longest_author = num;
e6005927 2115 commit_info_destroy(&ci);
cee7f245
JH
2116 }
2117 num = e->s_lno + e->num_lines;
2118 if (longest_src_lines < num)
2119 longest_src_lines = num;
2120 num = e->lno + e->num_lines;
2121 if (longest_dst_lines < num)
2122 longest_dst_lines = num;
5ff62c30
JH
2123 if (largest_score < ent_score(sb, e))
2124 largest_score = ent_score(sb, e);
cee7f245 2125 }
ec7ff5ba
ZJS
2126 max_orig_digits = decimal_width(longest_src_lines);
2127 max_digits = decimal_width(longest_dst_lines);
2128 max_score_digits = decimal_width(largest_score);
b31272f7
JH
2129
2130 if (compute_auto_abbrev)
2131 /* one more abbrev length is needed for the boundary commit */
2132 abbrev = auto_abbrev + 1;
cee7f245
JH
2133}
2134
1732a1fd
JH
2135/*
2136 * For debugging -- origin is refcounted, and this asserts that
2137 * we do not underflow.
2138 */
54a4c617
JH
2139static void sanity_check_refcnt(struct scoreboard *sb)
2140{
2141 int baa = 0;
2142 struct blame_entry *ent;
2143
2144 for (ent = sb->ent; ent; ent = ent->next) {
ae86ad65 2145 /* Nobody should have zero or negative refcnt */
854b97f6
JH
2146 if (ent->suspect->refcnt <= 0) {
2147 fprintf(stderr, "%s in %s has negative refcnt %d\n",
2148 ent->suspect->path,
2149 sha1_to_hex(ent->suspect->commit->object.sha1),
2150 ent->suspect->refcnt);
ae86ad65 2151 baa = 1;
854b97f6 2152 }
ae86ad65 2153 }
54a4c617
JH
2154 if (baa) {
2155 int opt = 0160;
2156 find_alignment(sb, &opt);
2157 output(sb, opt);
854b97f6 2158 die("Baa %d!", baa);
54a4c617
JH
2159 }
2160}
2161
4a0fc95f
JH
2162static unsigned parse_score(const char *arg)
2163{
2164 char *end;
2165 unsigned long score = strtoul(arg, &end, 10);
2166 if (*end)
2167 return 0;
2168 return score;
2169}
2170
20239bae
JK
2171static const char *add_prefix(const char *prefix, const char *path)
2172{
097971f5 2173 return prefix_path(prefix, prefix ? strlen(prefix) : 0, path);
20239bae
JK
2174}
2175
ef90d6d4 2176static int git_blame_config(const char *var, const char *value, void *cb)
4c10a5ca
JH
2177{
2178 if (!strcmp(var, "blame.showroot")) {
2179 show_root = git_config_bool(var, value);
2180 return 0;
2181 }
2182 if (!strcmp(var, "blame.blankboundary")) {
2183 blank_boundary = git_config_bool(var, value);
2184 return 0;
2185 }
8b504db3
QN
2186 if (!strcmp(var, "blame.showemail")) {
2187 int *output_option = cb;
2188 if (git_config_bool(var, value))
2189 *output_option |= OUTPUT_SHOW_EMAIL;
2190 else
2191 *output_option &= ~OUTPUT_SHOW_EMAIL;
2192 return 0;
2193 }
31653c1a
EL
2194 if (!strcmp(var, "blame.date")) {
2195 if (!value)
2196 return config_error_nonbool(var);
2197 blame_date_mode = parse_date_format(value);
2198 return 0;
2199 }
3b8a12e8 2200
6680a087 2201 if (userdiff_config(var, value) < 0)
3b8a12e8 2202 return -1;
3b8a12e8 2203
ef90d6d4 2204 return git_default_config(var, value, cb);
4c10a5ca
JH
2205}
2206
9aeaab68 2207static void verify_working_tree_path(struct commit *work_tree, const char *path)
ffcabccf 2208{
9aeaab68 2209 struct commit_list *parents;
ffcabccf 2210
9aeaab68
JH
2211 for (parents = work_tree->parents; parents; parents = parents->next) {
2212 const unsigned char *commit_sha1 = parents->item->object.sha1;
2213 unsigned char blob_sha1[20];
2214 unsigned mode;
2215
2216 if (!get_tree_entry(commit_sha1, path, blob_sha1, &mode) &&
2217 sha1_object_info(blob_sha1, NULL) == OBJ_BLOB)
2218 return;
2219 }
2220 die("no such path '%s' in HEAD", path);
2221}
2222
2223static struct commit_list **append_parent(struct commit_list **tail, const unsigned char *sha1)
2224{
2225 struct commit *parent;
2226
2227 parent = lookup_commit_reference(sha1);
2228 if (!parent)
2229 die("no such commit %s", sha1_to_hex(sha1));
2230 return &commit_list_insert(parent, tail)->next;
2231}
2232
2233static void append_merge_parents(struct commit_list **tail)
2234{
2235 int merge_head;
2236 const char *merge_head_file = git_path("MERGE_HEAD");
2237 struct strbuf line = STRBUF_INIT;
2238
2239 merge_head = open(merge_head_file, O_RDONLY);
2240 if (merge_head < 0) {
2241 if (errno == ENOENT)
2242 return;
2243 die("cannot open '%s' for reading", merge_head_file);
2244 }
2245
2246 while (!strbuf_getwholeline_fd(&line, merge_head, '\n')) {
2247 unsigned char sha1[20];
2248 if (line.len < 40 || get_sha1_hex(line.buf, sha1))
2249 die("unknown line in '%s': %s", merge_head_file, line.buf);
2250 tail = append_parent(tail, sha1);
2251 }
2252 close(merge_head);
2253 strbuf_release(&line);
ffcabccf
JH
2254}
2255
8597ea3a
JK
2256/*
2257 * This isn't as simple as passing sb->buf and sb->len, because we
2258 * want to transfer ownership of the buffer to the commit (so we
2259 * must use detach).
2260 */
2261static void set_commit_buffer_from_strbuf(struct commit *c, struct strbuf *sb)
2262{
2263 size_t len;
2264 void *buf = strbuf_detach(sb, &len);
2265 set_commit_buffer(c, buf, len);
2266}
2267
f6c07d7d
JH
2268/*
2269 * Prepare a dummy commit that represents the work tree (or staged) item.
2270 * Note that annotating work tree item never works in the reverse.
2271 */
3b8a12e8
AB
2272static struct commit *fake_working_tree_commit(struct diff_options *opt,
2273 const char *path,
2274 const char *contents_from)
1cfe7733
JH
2275{
2276 struct commit *commit;
2277 struct origin *origin;
9aeaab68 2278 struct commit_list **parent_tail, *parent;
1cfe7733 2279 unsigned char head_sha1[20];
f285a2d7 2280 struct strbuf buf = STRBUF_INIT;
1cfe7733 2281 const char *ident;
1cfe7733 2282 time_t now;
1cfe7733
JH
2283 int size, len;
2284 struct cache_entry *ce;
2285 unsigned mode;
9aeaab68 2286 struct strbuf msg = STRBUF_INIT;
1cfe7733
JH
2287
2288 time(&now);
10322a0a 2289 commit = alloc_commit_node();
1cfe7733
JH
2290 commit->object.parsed = 1;
2291 commit->date = now;
9aeaab68
JH
2292 parent_tail = &commit->parents;
2293
7695d118 2294 if (!resolve_ref_unsafe("HEAD", RESOLVE_REF_READING, head_sha1, NULL))
9aeaab68
JH
2295 die("no such ref: HEAD");
2296
2297 parent_tail = append_parent(parent_tail, head_sha1);
2298 append_merge_parents(parent_tail);
2299 verify_working_tree_path(commit, path);
1cfe7733
JH
2300
2301 origin = make_origin(commit, path);
2302
9aeaab68
JH
2303 ident = fmt_ident("Not Committed Yet", "not.committed.yet", NULL, 0);
2304 strbuf_addstr(&msg, "tree 0000000000000000000000000000000000000000\n");
2305 for (parent = commit->parents; parent; parent = parent->next)
2306 strbuf_addf(&msg, "parent %s\n",
2307 sha1_to_hex(parent->item->object.sha1));
2308 strbuf_addf(&msg,
2309 "author %s\n"
2310 "committer %s\n\n"
2311 "Version of %s from %s\n",
2312 ident, ident, path,
2313 (!contents_from ? path :
2314 (!strcmp(contents_from, "-") ? "standard input" : contents_from)));
8597ea3a 2315 set_commit_buffer_from_strbuf(commit, &msg);
9aeaab68 2316
1cfe7733
JH
2317 if (!contents_from || strcmp("-", contents_from)) {
2318 struct stat st;
2319 const char *read_from;
8518088f 2320 char *buf_ptr;
3b8a12e8 2321 unsigned long buf_len;
1cfe7733
JH
2322
2323 if (contents_from) {
2324 if (stat(contents_from, &st) < 0)
0721c314 2325 die_errno("Cannot stat '%s'", contents_from);
1cfe7733
JH
2326 read_from = contents_from;
2327 }
2328 else {
2329 if (lstat(path, &st) < 0)
0721c314 2330 die_errno("Cannot lstat '%s'", path);
1cfe7733
JH
2331 read_from = path;
2332 }
1cfe7733 2333 mode = canon_mode(st.st_mode);
3b8a12e8 2334
1cfe7733
JH
2335 switch (st.st_mode & S_IFMT) {
2336 case S_IFREG:
3b8a12e8 2337 if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) &&
e5450100 2338 textconv_object(read_from, mode, null_sha1, 0, &buf_ptr, &buf_len))
8518088f 2339 strbuf_attach(&buf, buf_ptr, buf_len, buf_len + 1);
3b8a12e8 2340 else if (strbuf_read_file(&buf, read_from, st.st_size) != st.st_size)
0721c314 2341 die_errno("cannot open or read '%s'", read_from);
1cfe7733
JH
2342 break;
2343 case S_IFLNK:
edfd45d2 2344 if (strbuf_readlink(&buf, read_from, st.st_size) < 0)
0721c314 2345 die_errno("cannot readlink '%s'", read_from);
1cfe7733
JH
2346 break;
2347 default:
2348 die("unsupported file type %s", read_from);
2349 }
2350 }
2351 else {
2352 /* Reading from stdin */
1cfe7733 2353 mode = 0;
f1696ee3 2354 if (strbuf_read(&buf, 0, 0) < 0)
d824cbba 2355 die_errno("failed to read from stdin");
1cfe7733 2356 }
4bf256d6 2357 convert_to_git(path, buf.buf, buf.len, &buf, 0);
af6eb822
PH
2358 origin->file.ptr = buf.buf;
2359 origin->file.size = buf.len;
2360 pretend_sha1_file(buf.buf, buf.len, OBJ_BLOB, origin->blob_sha1);
1cfe7733
JH
2361
2362 /*
2363 * Read the current index, replace the path entry with
2364 * origin->blob_sha1 without mucking with its mode or type
2365 * bits; we are not going to write this index out -- we just
2366 * want to run "diff-index --cached".
2367 */
2368 discard_cache();
2369 read_cache();
2370
2371 len = strlen(path);
2372 if (!mode) {
2373 int pos = cache_name_pos(path, len);
2374 if (0 <= pos)
7a51ed66 2375 mode = active_cache[pos]->ce_mode;
1cfe7733
JH
2376 else
2377 /* Let's not bother reading from HEAD tree */
2378 mode = S_IFREG | 0644;
2379 }
2380 size = cache_entry_size(len);
2381 ce = xcalloc(1, size);
2382 hashcpy(ce->sha1, origin->blob_sha1);
2383 memcpy(ce->name, path, len);
b60e188c
TG
2384 ce->ce_flags = create_ce_flags(0);
2385 ce->ce_namelen = len;
1cfe7733
JH
2386 ce->ce_mode = create_ce_mode(mode);
2387 add_cache_entry(ce, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
2388
2389 /*
2390 * We are not going to write this out, so this does not matter
2391 * right now, but someday we might optimize diff-index --cached
2392 * with cache-tree information.
2393 */
a5400efe 2394 cache_tree_invalidate_path(&the_index, path);
1cfe7733 2395
1cfe7733
JH
2396 return commit;
2397}
2398
1b0d4000 2399static struct object_array_entry *find_single_final(struct rev_info *revs)
f6c07d7d
JH
2400{
2401 int i;
1b0d4000 2402 struct object_array_entry *found = NULL;
f6c07d7d 2403
f6c07d7d
JH
2404 for (i = 0; i < revs->pending.nr; i++) {
2405 struct object *obj = revs->pending.objects[i].item;
2406 if (obj->flags & UNINTERESTING)
2407 continue;
2408 while (obj->type == OBJ_TAG)
2409 obj = deref_tag(obj, NULL, 0);
2410 if (obj->type != OBJ_COMMIT)
2411 die("Non commit %s?", revs->pending.objects[i].name);
1b0d4000 2412 if (found)
f6c07d7d
JH
2413 die("More than one commit to dig from %s and %s?",
2414 revs->pending.objects[i].name,
1b0d4000
MK
2415 found->name);
2416 found = &(revs->pending.objects[i]);
2417 }
2418 return found;
2419}
2420
2421static char *prepare_final(struct scoreboard *sb)
2422{
2423 struct object_array_entry *found = find_single_final(sb->revs);
2424 if (found) {
2425 sb->final = (struct commit *) found->item;
2426 return xstrdup(found->name);
2427 } else {
2428 return NULL;
f6c07d7d 2429 }
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;
31653c1a
EL
2584 blame_date_mode = DATE_ISO8601;
2585 } else {
2586 blame_date_mode = revs.date_mode;
2587 }
2588
2589 /* The maximum width used to show the dates */
2590 switch (blame_date_mode) {
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;
2619 }
2620 blame_date_width -= 1; /* strip the null */
7ceacdff 2621
b3123f98
JH
2622 if (DIFF_OPT_TST(&revs.diffopt, FIND_COPIES_HARDER))
2623 opt |= (PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE |
2624 PICKAXE_BLAME_COPY_HARDER);
2625
4a0fc95f
JH
2626 if (!blame_move_score)
2627 blame_move_score = BLAME_DEFAULT_MOVE_SCORE;
2628 if (!blame_copy_score)
2629 blame_copy_score = BLAME_DEFAULT_COPY_SCORE;
2630
1732a1fd
JH
2631 /*
2632 * We have collected options unknown to us in argv[1..unk]
cee7f245 2633 * which are to be passed to revision machinery if we are
3dff5379 2634 * going to do the "bottom" processing.
cee7f245
JH
2635 *
2636 * The remaining are:
2637 *
22e5e58a 2638 * (1) if dashdash_pos != 0, it is either
3f8d5204
PH
2639 * "blame [revisions] -- <path>" or
2640 * "blame -- <path> <rev>"
cee7f245 2641 *
22e5e58a 2642 * (2) otherwise, it is one of the two:
3f8d5204
PH
2643 * "blame [revisions] <path>"
2644 * "blame <path> <rev>"
cee7f245 2645 *
3f8d5204
PH
2646 * Note that we must strip out <path> from the arguments: we do not
2647 * want the path pruning but we may want "bottom" processing.
cee7f245 2648 */
3f8d5204
PH
2649 if (dashdash_pos) {
2650 switch (argc - dashdash_pos - 1) {
2651 case 2: /* (1b) */
2652 if (argc != 4)
5817da01 2653 usage_with_options(blame_opt_usage, options);
3f8d5204
PH
2654 /* reorder for the new way: <rev> -- <path> */
2655 argv[1] = argv[3];
2656 argv[3] = argv[2];
2657 argv[2] = "--";
2658 /* FALLTHROUGH */
2659 case 1: /* (1a) */
2660 path = add_prefix(prefix, argv[--argc]);
2661 argv[argc] = NULL;
2662 break;
2663 default:
2664 usage_with_options(blame_opt_usage, options);
cee7f245 2665 }
3f8d5204
PH
2666 } else {
2667 if (argc < 2)
5817da01 2668 usage_with_options(blame_opt_usage, options);
3f8d5204 2669 path = add_prefix(prefix, argv[argc - 1]);
dbe44faa 2670 if (argc == 3 && !file_exists(path)) { /* (2b) */
3f8d5204
PH
2671 path = add_prefix(prefix, argv[1]);
2672 argv[1] = argv[2];
cee7f245 2673 }
3f8d5204 2674 argv[argc - 1] = "--";
cee7f245 2675
3f8d5204 2676 setup_work_tree();
dbe44faa 2677 if (!file_exists(path))
d824cbba 2678 die_errno("cannot stat path '%s'", path);
cee7f245
JH
2679 }
2680
8b3dce56 2681 revs.disable_stdin = 1;
3f8d5204 2682 setup_revisions(argc, argv, &revs, NULL);
cee7f245
JH
2683 memset(&sb, 0, sizeof(sb));
2684
85af7929 2685 sb.revs = &revs;
7e6ac6e4 2686 if (!reverse) {
85af7929 2687 final_commit_name = prepare_final(&sb);
7e6ac6e4
DK
2688 sb.commits.compare = compare_commits_by_commit_date;
2689 }
85af7929
JH
2690 else if (contents_from)
2691 die("--contents and --children do not blend well.");
95a4fb0e
JK
2692 else if (revs.first_parent_only)
2693 die("combining --first-parent and --reverse is not supported");
7e6ac6e4 2694 else {
85af7929 2695 final_commit_name = prepare_initial(&sb);
7e6ac6e4
DK
2696 sb.commits.compare = compare_commits_by_reverse_commit_date;
2697 }
cee7f245
JH
2698
2699 if (!sb.final) {
1732a1fd
JH
2700 /*
2701 * "--not A B -- path" without anything positive;
1cfe7733
JH
2702 * do not default to HEAD, but use the working tree
2703 * or "--contents".
1732a1fd 2704 */
1981820b 2705 setup_work_tree();
3b8a12e8
AB
2706 sb.final = fake_working_tree_commit(&sb.revs->diffopt,
2707 path, contents_from);
1cfe7733 2708 add_pending_object(&revs, &(sb.final->object), ":");
cee7f245 2709 }
1cfe7733
JH
2710 else if (contents_from)
2711 die("Cannot use --contents with final commit object name");
cee7f245 2712
1732a1fd
JH
2713 /*
2714 * If we have bottom, this will mark the ancestors of the
cee7f245
JH
2715 * bottom commits we would reach while traversing as
2716 * uninteresting.
2717 */
3d51e1b5 2718 if (prepare_revision_walk(&revs))
20108742 2719 die(_("revision walk setup failed"));
cee7f245 2720
1cfe7733 2721 if (is_null_sha1(sb.final->object.sha1)) {
1cfe7733 2722 o = sb.final->util;
5c0b13f8 2723 sb.final_buf = xmemdupz(o->file.ptr, o->file.size);
1cfe7733
JH
2724 sb.final_buf_size = o->file.size;
2725 }
2726 else {
2727 o = get_origin(&sb, sb.final, path);
90064710 2728 if (fill_blob_sha1_and_mode(o))
1cfe7733 2729 die("no such path %s in %s", path, final_commit_name);
cee7f245 2730
3b8a12e8 2731 if (DIFF_OPT_TST(&sb.revs->diffopt, ALLOW_TEXTCONV) &&
e5450100 2732 textconv_object(path, o->mode, o->blob_sha1, 1, (char **) &sb.final_buf,
3b8a12e8
AB
2733 &sb.final_buf_size))
2734 ;
2735 else
2736 sb.final_buf = read_sha1_file(o->blob_sha1, &type,
2737 &sb.final_buf_size);
2738
ab43e495
JH
2739 if (!sb.final_buf)
2740 die("Cannot read blob %s for path %s",
2741 sha1_to_hex(o->blob_sha1),
2742 path);
1cfe7733 2743 }
c2e525d9 2744 num_read_blob++;
cee7f245
JH
2745 lno = prepare_lines(&sb);
2746
58dbfa2e
ES
2747 if (lno && !range_list.nr)
2748 string_list_append(&range_list, xstrdup("1"));
2749
52f4d126 2750 anchor = 1;
58dbfa2e
ES
2751 range_set_init(&ranges, range_list.nr);
2752 for (range_i = 0; range_i < range_list.nr; ++range_i) {
2753 long bottom, top;
2754 if (parse_range_arg(range_list.items[range_i].string,
52f4d126 2755 nth_line_cb, &sb, lno, anchor,
58dbfa2e
ES
2756 &bottom, &top, sb.path))
2757 usage(blame_usage);
2758 if (lno < top || ((lno || bottom) && lno < bottom))
2759 die("file %s has only %lu lines", path, lno);
2760 if (bottom < 1)
2761 bottom = 1;
2762 if (top < 1)
2763 top = lno;
2764 bottom--;
2765 range_set_append_unsafe(&ranges, bottom, top);
52f4d126 2766 anchor = top + 1;
58dbfa2e
ES
2767 }
2768 sort_and_merge_range_set(&ranges);
2769
2770 for (range_i = ranges.nr; range_i > 0; --range_i) {
2771 const struct range *r = &ranges.ranges[range_i - 1];
2772 long bottom = r->start;
2773 long top = r->end;
2774 struct blame_entry *next = ent;
2775 ent = xcalloc(1, sizeof(*ent));
2776 ent->lno = bottom;
2777 ent->num_lines = top - bottom;
2778 ent->suspect = o;
2779 ent->s_lno = bottom;
2780 ent->next = next;
58dbfa2e
ES
2781 origin_incref(o);
2782 }
7e6ac6e4
DK
2783
2784 o->suspects = ent;
2785 prio_queue_put(&sb.commits, o->commit);
2786
58dbfa2e
ES
2787 origin_decref(o);
2788
2789 range_set_release(&ranges);
2790 string_list_clear(&range_list, 0);
cee7f245 2791
7e6ac6e4 2792 sb.ent = NULL;
cee7f245
JH
2793 sb.path = path;
2794
d551a488 2795 read_mailmap(&mailmap, NULL);
f9567384 2796
b92565dc
MH
2797 if (!incremental)
2798 setup_pager();
2799
85af7929 2800 assign_blame(&sb, opt);
cee7f245 2801
a46442f1
LF
2802 free(final_commit_name);
2803
717d1462
LT
2804 if (incremental)
2805 return 0;
2806
7e6ac6e4
DK
2807 sb.ent = blame_sort(sb.ent, compare_blame_final);
2808
cee7f245
JH
2809 coalesce(&sb);
2810
2811 if (!(output_option & OUTPUT_PORCELAIN))
2812 find_alignment(&sb, &output_option);
2813
2814 output(&sb, output_option);
2815 free((void *)sb.final_buf);
2816 for (ent = sb.ent; ent; ) {
2817 struct blame_entry *e = ent->next;
2818 free(ent);
2819 ent = e;
2820 }
c2e525d9 2821
870b39c1 2822 if (show_stats) {
c2e525d9
JH
2823 printf("num read blob: %d\n", num_read_blob);
2824 printf("num get patch: %d\n", num_get_patch);
2825 printf("num commits: %d\n", num_commits);
2826 }
cee7f245
JH
2827 return 0;
2828}