blame: honor the diff heuristic options and config
[git/git.git] / builtin / blame.c
CommitLineData
cee7f245 1/*
31653c1a 2 * Blame
cee7f245 3 *
7e6ac6e4
DK
4 * Copyright (c) 2006, 2014 by its authors
5 * See COPYING for licensing conditions
cee7f245
JH
6 */
7
8#include "cache.h"
fb58c8d5 9#include "refs.h"
cee7f245
JH
10#include "builtin.h"
11#include "blob.h"
12#include "commit.h"
13#include "tag.h"
14#include "tree-walk.h"
15#include "diff.h"
16#include "diffcore.h"
17#include "revision.h"
717d1462 18#include "quote.h"
cee7f245 19#include "xdiff-interface.h"
1cfe7733 20#include "cache-tree.h"
c455c87c 21#include "string-list.h"
f9567384 22#include "mailmap.h"
7e6ac6e4 23#include "mergesort.h"
5817da01 24#include "parse-options.h"
7e6ac6e4 25#include "prio-queue.h"
ffaf9cc0 26#include "utf8.h"
3b8a12e8 27#include "userdiff.h"
25ed3412 28#include "line-range.h"
58dbfa2e 29#include "line-log.h"
dbe44faa 30#include "dir.h"
aba37f49 31#include "progress.h"
cee7f245 32
ce41720c 33static char blame_usage[] = N_("git blame [<options>] [<rev-opts>] [<rev>] [--] <file>");
5817da01
PH
34
35static const char *blame_opt_usage[] = {
36 blame_usage,
37 "",
9c9b4f2f 38 N_("<rev-opts> are documented in git-rev-list(1)"),
5817da01
PH
39 NULL
40};
cee7f245
JH
41
42static int longest_file;
43static int longest_author;
44static int max_orig_digits;
45static int max_digits;
5ff62c30 46static int max_score_digits;
4c10a5ca 47static int show_root;
85af7929 48static int reverse;
4c10a5ca 49static int blank_boundary;
717d1462 50static int incremental;
582aa00b 51static int xdl_opts;
84393bfd 52static int abbrev = -1;
3d1aa566 53static int no_whole_file_rename;
aba37f49 54static int show_progress;
31653c1a 55
a5481a6c 56static struct date_mode blame_date_mode = { DATE_ISO8601 };
31653c1a
EL
57static size_t blame_date_width;
58
2721ce21 59static struct string_list mailmap = STRING_LIST_INIT_NODUP;
cee7f245 60
54a4c617
JH
61#ifndef DEBUG
62#define DEBUG 0
63#endif
64
c2e525d9
JH
65/* stats */
66static int num_read_blob;
67static int num_get_patch;
68static int num_commits;
69
d24bba80 70#define PICKAXE_BLAME_MOVE 01
18abd745
JH
71#define PICKAXE_BLAME_COPY 02
72#define PICKAXE_BLAME_COPY_HARDER 04
c63777c0 73#define PICKAXE_BLAME_COPY_HARDEST 010
d24bba80 74
4a0fc95f
JH
75/*
76 * blame for a blame_entry with score lower than these thresholds
77 * is not passed to the parent using move/copy logic.
78 */
79static unsigned blame_move_score;
80static unsigned blame_copy_score;
81#define BLAME_DEFAULT_MOVE_SCORE 20
82#define BLAME_DEFAULT_COPY_SCORE 40
83
208acbfb 84/* Remember to update object flag allocation in object.h */
cee7f245
JH
85#define METAINFO_SHOWN (1u<<12)
86#define MORE_THAN_ONE_PATH (1u<<13)
87
88/*
54a4c617 89 * One blob in a commit that is being suspected
cee7f245
JH
90 */
91struct origin {
54a4c617 92 int refcnt;
7e6ac6e4 93 /* Record preceding blame record for this blob */
96e11709 94 struct origin *previous;
7e6ac6e4
DK
95 /* origins are put in a list linked via `next' hanging off the
96 * corresponding commit's util field in order to make finding
97 * them fast. The presence in this chain does not count
98 * towards the origin's reference count. It is tempting to
99 * let it count as long as the commit is pending examination,
100 * but even under circumstances where the commit will be
101 * present multiple times in the priority queue of unexamined
102 * commits, processing the first instance will not leave any
103 * work requiring the origin data for the second instance. An
104 * interspersed commit changing that would have to be
105 * preexisting with a different ancestry and with the same
106 * commit date in order to wedge itself between two instances
107 * of the same commit in the priority queue _and_ produce
108 * blame entries relevant for it. While we don't want to let
109 * us get tripped up by this case, it certainly does not seem
110 * worth optimizing for.
111 */
112 struct origin *next;
cee7f245 113 struct commit *commit;
7e6ac6e4
DK
114 /* `suspects' contains blame entries that may be attributed to
115 * this origin's commit or to parent commits. When a commit
116 * is being processed, all suspects will be moved, either by
117 * assigning them to an origin in a different commit, or by
118 * shipping them to the scoreboard's ent list because they
119 * cannot be attributed to a different commit.
120 */
121 struct blame_entry *suspects;
c2e525d9 122 mmfile_t file;
cee7f245 123 unsigned char blob_sha1[20];
90064710 124 unsigned mode;
7e6ac6e4
DK
125 /* guilty gets set when shipping any suspects to the final
126 * blame list instead of other commits
127 */
128 char guilty;
cee7f245
JH
129 char path[FLEX_ARRAY];
130};
131
aba37f49
ECA
132struct progress_info {
133 struct progress *progress;
134 int blamed_lines;
135};
136
4b4132fd
RS
137static int diff_hunks(mmfile_t *file_a, mmfile_t *file_b, long ctxlen,
138 xdl_emit_hunk_consume_func_t hunk_func, void *cb_data)
139{
140 xpparam_t xpp = {0};
141 xdemitconf_t xecfg = {0};
85c20c30 142 xdemitcb_t ecb = {NULL};
4b4132fd
RS
143
144 xpp.flags = xdl_opts;
145 xecfg.ctxlen = ctxlen;
146 xecfg.hunk_func = hunk_func;
147 ecb.priv = cb_data;
148 return xdi_diff(file_a, file_b, &xpp, &xecfg, &ecb);
149}
150
3b8a12e8
AB
151/*
152 * Prepare diff_filespec and convert it using diff textconv API
153 * if the textconv driver exists.
154 * Return 1 if the conversion succeeds, 0 otherwise.
155 */
e5fba602 156int textconv_object(const char *path,
90064710 157 unsigned mode,
e5fba602 158 const unsigned char *sha1,
e5450100 159 int sha1_valid,
e5fba602
CP
160 char **buf,
161 unsigned long *buf_size)
3b8a12e8
AB
162{
163 struct diff_filespec *df;
164 struct userdiff_driver *textconv;
165
166 df = alloc_filespec(path);
e5450100 167 fill_filespec(df, sha1, sha1_valid, mode);
3b8a12e8
AB
168 textconv = get_textconv(df);
169 if (!textconv) {
170 free_filespec(df);
171 return 0;
172 }
173
174 *buf_size = fill_textconv(textconv, df, buf);
175 free_filespec(df);
176 return 1;
177}
178
1732a1fd
JH
179/*
180 * Given an origin, prepare mmfile_t structure to be used by the
181 * diff machinery
182 */
3b8a12e8
AB
183static void fill_origin_blob(struct diff_options *opt,
184 struct origin *o, mmfile_t *file)
c2e525d9
JH
185{
186 if (!o->file.ptr) {
21666f1a 187 enum object_type type;
3b8a12e8
AB
188 unsigned long file_size;
189
c2e525d9 190 num_read_blob++;
3b8a12e8 191 if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) &&
e5450100 192 textconv_object(o->path, o->mode, o->blob_sha1, 1, &file->ptr, &file_size))
3b8a12e8
AB
193 ;
194 else
195 file->ptr = read_sha1_file(o->blob_sha1, &type, &file_size);
196 file->size = file_size;
197
ab43e495
JH
198 if (!file->ptr)
199 die("Cannot read blob %s for path %s",
200 sha1_to_hex(o->blob_sha1),
201 o->path);
c2e525d9
JH
202 o->file = *file;
203 }
204 else
205 *file = o->file;
c2e525d9
JH
206}
207
1732a1fd
JH
208/*
209 * Origin is refcounted and usually we keep the blob contents to be
210 * reused.
211 */
54a4c617
JH
212static inline struct origin *origin_incref(struct origin *o)
213{
214 if (o)
215 o->refcnt++;
216 return o;
217}
218
219static void origin_decref(struct origin *o)
220{
221 if (o && --o->refcnt <= 0) {
7e6ac6e4 222 struct origin *p, *l = NULL;
96e11709
JH
223 if (o->previous)
224 origin_decref(o->previous);
8e0f7003 225 free(o->file.ptr);
7e6ac6e4
DK
226 /* Should be present exactly once in commit chain */
227 for (p = o->commit->util; p; l = p, p = p->next) {
228 if (p == o) {
229 if (l)
230 l->next = p->next;
231 else
232 o->commit->util = p->next;
233 free(o);
234 return;
235 }
236 }
237 die("internal error in blame::origin_decref");
54a4c617
JH
238 }
239}
240
7c3c7962
JH
241static void drop_origin_blob(struct origin *o)
242{
243 if (o->file.ptr) {
244 free(o->file.ptr);
245 o->file.ptr = NULL;
246 }
247}
248
1732a1fd
JH
249/*
250 * Each group of lines is described by a blame_entry; it can be split
7e6ac6e4
DK
251 * as we pass blame to the parents. They are arranged in linked lists
252 * kept as `suspects' of some unprocessed origin, or entered (when the
253 * blame origin has been finalized) into the scoreboard structure.
254 * While the scoreboard structure is only sorted at the end of
255 * processing (according to final image line number), the lists
256 * attached to an origin are sorted by the target line number.
1732a1fd 257 */
cee7f245 258struct blame_entry {
cee7f245
JH
259 struct blame_entry *next;
260
261 /* the first line of this group in the final image;
262 * internally all line numbers are 0 based.
263 */
264 int lno;
265
266 /* how many lines this group has */
267 int num_lines;
268
269 /* the commit that introduced this group into the final image */
270 struct origin *suspect;
271
cee7f245
JH
272 /* the line number of the first line of this group in the
273 * suspect's file; internally all line numbers are 0 based.
274 */
275 int s_lno;
5ff62c30
JH
276
277 /* how significant this entry is -- cached to avoid
1732a1fd 278 * scanning the lines over and over.
5ff62c30
JH
279 */
280 unsigned score;
cee7f245
JH
281};
282
7e6ac6e4
DK
283/*
284 * Any merge of blames happens on lists of blames that arrived via
285 * different parents in a single suspect. In this case, we want to
286 * sort according to the suspect line numbers as opposed to the final
287 * image line numbers. The function body is somewhat longish because
288 * it avoids unnecessary writes.
289 */
290
291static struct blame_entry *blame_merge(struct blame_entry *list1,
292 struct blame_entry *list2)
293{
294 struct blame_entry *p1 = list1, *p2 = list2,
295 **tail = &list1;
296
297 if (!p1)
298 return p2;
299 if (!p2)
300 return p1;
301
302 if (p1->s_lno <= p2->s_lno) {
303 do {
304 tail = &p1->next;
305 if ((p1 = *tail) == NULL) {
306 *tail = p2;
307 return list1;
308 }
309 } while (p1->s_lno <= p2->s_lno);
310 }
311 for (;;) {
312 *tail = p2;
313 do {
314 tail = &p2->next;
315 if ((p2 = *tail) == NULL) {
316 *tail = p1;
317 return list1;
318 }
319 } while (p1->s_lno > p2->s_lno);
320 *tail = p1;
321 do {
322 tail = &p1->next;
323 if ((p1 = *tail) == NULL) {
324 *tail = p2;
325 return list1;
326 }
327 } while (p1->s_lno <= p2->s_lno);
328 }
329}
330
331static void *get_next_blame(const void *p)
332{
333 return ((struct blame_entry *)p)->next;
334}
335
336static void set_next_blame(void *p1, void *p2)
337{
338 ((struct blame_entry *)p1)->next = p2;
339}
340
341/*
342 * Final image line numbers are all different, so we don't need a
343 * three-way comparison here.
344 */
345
346static int compare_blame_final(const void *p1, const void *p2)
347{
348 return ((struct blame_entry *)p1)->lno > ((struct blame_entry *)p2)->lno
349 ? 1 : -1;
350}
351
352static int compare_blame_suspect(const void *p1, const void *p2)
353{
354 const struct blame_entry *s1 = p1, *s2 = p2;
355 /*
356 * to allow for collating suspects, we sort according to the
357 * respective pointer value as the primary sorting criterion.
358 * The actual relation is pretty unimportant as long as it
359 * establishes a total order. Comparing as integers gives us
360 * that.
361 */
362 if (s1->suspect != s2->suspect)
363 return (intptr_t)s1->suspect > (intptr_t)s2->suspect ? 1 : -1;
364 if (s1->s_lno == s2->s_lno)
365 return 0;
366 return s1->s_lno > s2->s_lno ? 1 : -1;
367}
368
369static struct blame_entry *blame_sort(struct blame_entry *head,
370 int (*compare_fn)(const void *, const void *))
371{
372 return llist_mergesort (head, get_next_blame, set_next_blame, compare_fn);
373}
374
375static int compare_commits_by_reverse_commit_date(const void *a,
376 const void *b,
377 void *c)
378{
379 return -compare_commits_by_commit_date(a, b, c);
380}
381
1732a1fd
JH
382/*
383 * The current state of the blame assignment.
384 */
cee7f245
JH
385struct scoreboard {
386 /* the final commit (i.e. where we started digging from) */
387 struct commit *final;
7e6ac6e4
DK
388 /* Priority queue for commits with unassigned blame records */
389 struct prio_queue commits;
85af7929 390 struct rev_info *revs;
cee7f245
JH
391 const char *path;
392
1732a1fd
JH
393 /*
394 * The contents in the final image.
395 * Used by many functions to obtain contents of the nth line,
396 * indexed with scoreboard.lineno[blame_entry.lno].
cee7f245
JH
397 */
398 const char *final_buf;
399 unsigned long final_buf_size;
400
401 /* linked list of blames */
402 struct blame_entry *ent;
403
612702e8 404 /* look-up a line in the final buffer */
cee7f245
JH
405 int num_lines;
406 int *lineno;
407};
408
54a4c617
JH
409static void sanity_check_refcnt(struct scoreboard *);
410
1732a1fd
JH
411/*
412 * If two blame entries that are next to each other came from
413 * contiguous lines in the same origin (i.e. <commit, path> pair),
414 * merge them together.
415 */
cee7f245
JH
416static void coalesce(struct scoreboard *sb)
417{
418 struct blame_entry *ent, *next;
419
420 for (ent = sb->ent; ent && (next = ent->next); ent = next) {
0a88f08e 421 if (ent->suspect == next->suspect &&
cee7f245
JH
422 ent->s_lno + ent->num_lines == next->s_lno) {
423 ent->num_lines += next->num_lines;
424 ent->next = next->next;
54a4c617 425 origin_decref(next->suspect);
cee7f245 426 free(next);
46014766 427 ent->score = 0;
cee7f245
JH
428 next = ent; /* again */
429 }
430 }
54a4c617
JH
431
432 if (DEBUG) /* sanity */
433 sanity_check_refcnt(sb);
cee7f245
JH
434}
435
7e6ac6e4
DK
436/*
437 * Merge the given sorted list of blames into a preexisting origin.
438 * If there were no previous blames to that commit, it is entered into
439 * the commit priority queue of the score board.
440 */
441
442static void queue_blames(struct scoreboard *sb, struct origin *porigin,
443 struct blame_entry *sorted)
444{
445 if (porigin->suspects)
446 porigin->suspects = blame_merge(porigin->suspects, sorted);
447 else {
448 struct origin *o;
449 for (o = porigin->commit->util; o; o = o->next) {
450 if (o->suspects) {
451 porigin->suspects = sorted;
452 return;
453 }
454 }
455 porigin->suspects = sorted;
456 prio_queue_put(&sb->commits, porigin->commit);
457 }
458}
459
1732a1fd
JH
460/*
461 * Given a commit and a path in it, create a new origin structure.
462 * The callers that add blame to the scoreboard should use
463 * get_origin() to obtain shared, refcounted copy instead of calling
464 * this function directly.
465 */
854b97f6
JH
466static struct origin *make_origin(struct commit *commit, const char *path)
467{
468 struct origin *o;
96ffc06f 469 FLEX_ALLOC_STR(o, path, path);
854b97f6
JH
470 o->commit = commit;
471 o->refcnt = 1;
7e6ac6e4
DK
472 o->next = commit->util;
473 commit->util = o;
854b97f6
JH
474 return o;
475}
476
1732a1fd
JH
477/*
478 * Locate an existing origin or create a new one.
7e6ac6e4 479 * This moves the origin to front position in the commit util list.
1732a1fd 480 */
f6c0e191
JH
481static struct origin *get_origin(struct scoreboard *sb,
482 struct commit *commit,
483 const char *path)
cee7f245 484{
7e6ac6e4 485 struct origin *o, *l;
cee7f245 486
7e6ac6e4
DK
487 for (o = commit->util, l = NULL; o; l = o, o = o->next) {
488 if (!strcmp(o->path, path)) {
489 /* bump to front */
490 if (l) {
491 l->next = o->next;
492 o->next = commit->util;
493 commit->util = o;
494 }
495 return origin_incref(o);
496 }
cee7f245 497 }
854b97f6 498 return make_origin(commit, path);
cee7f245
JH
499}
500
1732a1fd
JH
501/*
502 * Fill the blob_sha1 field of an origin if it hasn't, so that later
503 * call to fill_origin_blob() can use it to locate the data. blob_sha1
504 * for an origin is also used to pass the blame for the entire file to
505 * the parent to detect the case where a child's blob is identical to
506 * that of its parent's.
90064710
KS
507 *
508 * This also fills origin->mode for corresponding tree path.
1732a1fd 509 */
90064710 510static int fill_blob_sha1_and_mode(struct origin *origin)
f6c0e191 511{
f6c0e191
JH
512 if (!is_null_sha1(origin->blob_sha1))
513 return 0;
ed1c9977 514 if (get_tree_entry(origin->commit->object.oid.hash,
f6c0e191 515 origin->path,
90064710 516 origin->blob_sha1, &origin->mode))
f6c0e191 517 goto error_out;
21666f1a 518 if (sha1_object_info(origin->blob_sha1, NULL) != OBJ_BLOB)
f6c0e191
JH
519 goto error_out;
520 return 0;
521 error_out:
522 hashclr(origin->blob_sha1);
90064710 523 origin->mode = S_IFINVALID;
f6c0e191
JH
524 return -1;
525}
526
1732a1fd
JH
527/*
528 * We have an origin -- check if the same path exists in the
529 * parent and return an origin structure to represent it.
530 */
f6c0e191 531static struct origin *find_origin(struct scoreboard *sb,
cee7f245
JH
532 struct commit *parent,
533 struct origin *origin)
534{
7e6ac6e4 535 struct origin *porigin;
cee7f245 536 struct diff_options diff_opts;
f6c0e191
JH
537 const char *paths[2];
538
7e6ac6e4
DK
539 /* First check any existing origins */
540 for (porigin = parent->util; porigin; porigin = porigin->next)
541 if (!strcmp(porigin->path, origin->path)) {
1732a1fd
JH
542 /*
543 * The same path between origin and its parent
544 * without renaming -- the most common case.
545 */
7e6ac6e4 546 return origin_incref (porigin);
854b97f6 547 }
0d981c67 548
f6c0e191
JH
549 /* See if the origin->path is different between parent
550 * and origin first. Most of the time they are the
551 * same and diff-tree is fairly efficient about this.
552 */
553 diff_setup(&diff_opts);
8f67f8ae 554 DIFF_OPT_SET(&diff_opts, RECURSIVE);
f6c0e191
JH
555 diff_opts.detect_rename = 0;
556 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
557 paths[0] = origin->path;
558 paths[1] = NULL;
559
4a2d5ae2
NTND
560 parse_pathspec(&diff_opts.pathspec,
561 PATHSPEC_ALL_MAGIC & ~PATHSPEC_LITERAL,
562 PATHSPEC_LITERAL_PATH, "", paths);
28452655 563 diff_setup_done(&diff_opts);
1cfe7733 564
f2fd0760 565 if (is_null_oid(&origin->commit->object.oid))
ed1c9977 566 do_diff_cache(parent->tree->object.oid.hash, &diff_opts);
1cfe7733 567 else
ed1c9977 568 diff_tree_sha1(parent->tree->object.oid.hash,
569 origin->commit->tree->object.oid.hash,
1cfe7733 570 "", &diff_opts);
f6c0e191
JH
571 diffcore_std(&diff_opts);
572
f6c0e191
JH
573 if (!diff_queued_diff.nr) {
574 /* The path is the same as parent */
575 porigin = get_origin(sb, parent, origin->path);
576 hashcpy(porigin->blob_sha1, origin->blob_sha1);
90064710 577 porigin->mode = origin->mode;
a9b2d424
JH
578 } else {
579 /*
580 * Since origin->path is a pathspec, if the parent
581 * commit had it as a directory, we will see a whole
582 * bunch of deletion of files in the directory that we
583 * do not care about.
584 */
585 int i;
586 struct diff_filepair *p = NULL;
587 for (i = 0; i < diff_queued_diff.nr; i++) {
588 const char *name;
589 p = diff_queued_diff.queue[i];
590 name = p->one->path ? p->one->path : p->two->path;
591 if (!strcmp(name, origin->path))
592 break;
593 }
594 if (!p)
595 die("internal error in blame::find_origin");
f6c0e191
JH
596 switch (p->status) {
597 default:
acca687f 598 die("internal error in blame::find_origin (%c)",
f6c0e191
JH
599 p->status);
600 case 'M':
601 porigin = get_origin(sb, parent, origin->path);
602 hashcpy(porigin->blob_sha1, p->one->sha1);
90064710 603 porigin->mode = p->one->mode;
f6c0e191
JH
604 break;
605 case 'A':
606 case 'T':
607 /* Did not exist in parent, or type changed */
608 break;
609 }
610 }
611 diff_flush(&diff_opts);
ed6e8038 612 clear_pathspec(&diff_opts.pathspec);
f69e743d
JH
613 return porigin;
614}
f6c0e191 615
1732a1fd
JH
616/*
617 * We have an origin -- find the path that corresponds to it in its
618 * parent and return an origin structure to represent it.
619 */
f69e743d
JH
620static struct origin *find_rename(struct scoreboard *sb,
621 struct commit *parent,
622 struct origin *origin)
623{
624 struct origin *porigin = NULL;
625 struct diff_options diff_opts;
626 int i;
cee7f245
JH
627
628 diff_setup(&diff_opts);
8f67f8ae 629 DIFF_OPT_SET(&diff_opts, RECURSIVE);
cee7f245
JH
630 diff_opts.detect_rename = DIFF_DETECT_RENAME;
631 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
2f3f8b21 632 diff_opts.single_follow = origin->path;
28452655 633 diff_setup_done(&diff_opts);
1cfe7733 634
f2fd0760 635 if (is_null_oid(&origin->commit->object.oid))
ed1c9977 636 do_diff_cache(parent->tree->object.oid.hash, &diff_opts);
1cfe7733 637 else
ed1c9977 638 diff_tree_sha1(parent->tree->object.oid.hash,
639 origin->commit->tree->object.oid.hash,
1cfe7733 640 "", &diff_opts);
cee7f245
JH
641 diffcore_std(&diff_opts);
642
643 for (i = 0; i < diff_queued_diff.nr; i++) {
644 struct diff_filepair *p = diff_queued_diff.queue[i];
612702e8 645 if ((p->status == 'R' || p->status == 'C') &&
f6c0e191
JH
646 !strcmp(p->two->path, origin->path)) {
647 porigin = get_origin(sb, parent, p->one->path);
648 hashcpy(porigin->blob_sha1, p->one->sha1);
90064710 649 porigin->mode = p->one->mode;
cee7f245
JH
650 break;
651 }
652 }
653 diff_flush(&diff_opts);
ed6e8038 654 clear_pathspec(&diff_opts.pathspec);
cee7f245
JH
655 return porigin;
656}
657
1732a1fd 658/*
7e6ac6e4 659 * Append a new blame entry to a given output queue.
1732a1fd 660 */
7e6ac6e4 661static void add_blame_entry(struct blame_entry ***queue, struct blame_entry *e)
cee7f245 662{
54a4c617
JH
663 origin_incref(e->suspect);
664
7e6ac6e4
DK
665 e->next = **queue;
666 **queue = e;
667 *queue = &e->next;
cee7f245
JH
668}
669
1732a1fd
JH
670/*
671 * src typically is on-stack; we want to copy the information in it to
7e6ac6e4
DK
672 * a malloced blame_entry that gets added to the given queue. The
673 * origin of dst loses a refcnt.
1732a1fd 674 */
7e6ac6e4
DK
675static void dup_entry(struct blame_entry ***queue,
676 struct blame_entry *dst, struct blame_entry *src)
cee7f245 677{
54a4c617
JH
678 origin_incref(src->suspect);
679 origin_decref(dst->suspect);
cee7f245 680 memcpy(dst, src, sizeof(*src));
7e6ac6e4
DK
681 dst->next = **queue;
682 **queue = dst;
683 *queue = &dst->next;
cee7f245
JH
684}
685
25ed3412 686static const char *nth_line(struct scoreboard *sb, long lno)
cee7f245
JH
687{
688 return sb->final_buf + sb->lineno[lno];
689}
690
25ed3412
BY
691static const char *nth_line_cb(void *data, long lno)
692{
693 return nth_line((struct scoreboard *)data, lno);
694}
695
1732a1fd
JH
696/*
697 * It is known that lines between tlno to same came from parent, and e
698 * has an overlap with that range. it also is known that parent's
699 * line plno corresponds to e's line tlno.
700 *
701 * <---- e ----->
702 * <------>
703 * <------------>
704 * <------------>
705 * <------------------>
706 *
707 * Split e into potentially three parts; before this chunk, the chunk
708 * to be blamed for the parent, and after that portion.
709 */
54a4c617 710static void split_overlap(struct blame_entry *split,
cee7f245
JH
711 struct blame_entry *e,
712 int tlno, int plno, int same,
713 struct origin *parent)
714{
cee7f245
JH
715 int chunk_end_lno;
716 memset(split, 0, sizeof(struct blame_entry [3]));
717
718 if (e->s_lno < tlno) {
719 /* there is a pre-chunk part not blamed on parent */
54a4c617 720 split[0].suspect = origin_incref(e->suspect);
cee7f245
JH
721 split[0].lno = e->lno;
722 split[0].s_lno = e->s_lno;
723 split[0].num_lines = tlno - e->s_lno;
724 split[1].lno = e->lno + tlno - e->s_lno;
725 split[1].s_lno = plno;
726 }
727 else {
728 split[1].lno = e->lno;
729 split[1].s_lno = plno + (e->s_lno - tlno);
730 }
731
732 if (same < e->s_lno + e->num_lines) {
733 /* there is a post-chunk part not blamed on parent */
54a4c617 734 split[2].suspect = origin_incref(e->suspect);
cee7f245
JH
735 split[2].lno = e->lno + (same - e->s_lno);
736 split[2].s_lno = e->s_lno + (same - e->s_lno);
737 split[2].num_lines = e->s_lno + e->num_lines - same;
738 chunk_end_lno = split[2].lno;
739 }
740 else
741 chunk_end_lno = e->lno + e->num_lines;
742 split[1].num_lines = chunk_end_lno - split[1].lno;
743
1732a1fd
JH
744 /*
745 * if it turns out there is nothing to blame the parent for,
746 * forget about the splitting. !split[1].suspect signals this.
747 */
cee7f245
JH
748 if (split[1].num_lines < 1)
749 return;
54a4c617 750 split[1].suspect = origin_incref(parent);
cee7f245
JH
751}
752
1732a1fd
JH
753/*
754 * split_overlap() divided an existing blame e into up to three parts
7e6ac6e4 755 * in split. Any assigned blame is moved to queue to
1732a1fd
JH
756 * reflect the split.
757 */
7e6ac6e4
DK
758static void split_blame(struct blame_entry ***blamed,
759 struct blame_entry ***unblamed,
54a4c617 760 struct blame_entry *split,
cee7f245
JH
761 struct blame_entry *e)
762{
763 struct blame_entry *new_entry;
764
765 if (split[0].suspect && split[2].suspect) {
1732a1fd 766 /* The first part (reuse storage for the existing entry e) */
7e6ac6e4 767 dup_entry(unblamed, e, &split[0]);
cee7f245 768
1732a1fd 769 /* The last part -- me */
cee7f245
JH
770 new_entry = xmalloc(sizeof(*new_entry));
771 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
7e6ac6e4 772 add_blame_entry(unblamed, new_entry);
cee7f245 773
1732a1fd 774 /* ... and the middle part -- parent */
cee7f245
JH
775 new_entry = xmalloc(sizeof(*new_entry));
776 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
7e6ac6e4 777 add_blame_entry(blamed, new_entry);
cee7f245
JH
778 }
779 else if (!split[0].suspect && !split[2].suspect)
1732a1fd
JH
780 /*
781 * The parent covers the entire area; reuse storage for
782 * e and replace it with the parent.
783 */
7e6ac6e4 784 dup_entry(blamed, e, &split[1]);
cee7f245 785 else if (split[0].suspect) {
1732a1fd 786 /* me and then parent */
7e6ac6e4 787 dup_entry(unblamed, e, &split[0]);
cee7f245
JH
788
789 new_entry = xmalloc(sizeof(*new_entry));
790 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
7e6ac6e4 791 add_blame_entry(blamed, new_entry);
cee7f245
JH
792 }
793 else {
1732a1fd 794 /* parent and then me */
7e6ac6e4 795 dup_entry(blamed, e, &split[1]);
cee7f245
JH
796
797 new_entry = xmalloc(sizeof(*new_entry));
798 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
7e6ac6e4 799 add_blame_entry(unblamed, new_entry);
cee7f245
JH
800 }
801}
802
1732a1fd
JH
803/*
804 * After splitting the blame, the origins used by the
805 * on-stack blame_entry should lose one refcnt each.
806 */
54a4c617
JH
807static void decref_split(struct blame_entry *split)
808{
809 int i;
810
811 for (i = 0; i < 3; i++)
812 origin_decref(split[i].suspect);
813}
814
1732a1fd 815/*
7e6ac6e4
DK
816 * reverse_blame reverses the list given in head, appending tail.
817 * That allows us to build lists in reverse order, then reverse them
818 * afterwards. This can be faster than building the list in proper
819 * order right away. The reason is that building in proper order
820 * requires writing a link in the _previous_ element, while building
821 * in reverse order just requires placing the list head into the
822 * _current_ element.
1732a1fd 823 */
cee7f245 824
7e6ac6e4
DK
825static struct blame_entry *reverse_blame(struct blame_entry *head,
826 struct blame_entry *tail)
cee7f245 827{
7e6ac6e4
DK
828 while (head) {
829 struct blame_entry *next = head->next;
830 head->next = tail;
831 tail = head;
832 head = next;
cee7f245 833 }
7e6ac6e4 834 return tail;
cee7f245
JH
835}
836
1732a1fd
JH
837/*
838 * Process one hunk from the patch between the current suspect for
7e6ac6e4
DK
839 * blame_entry e and its parent. This first blames any unfinished
840 * entries before the chunk (which is where target and parent start
841 * differing) on the parent, and then splits blame entries at the
842 * start and at the end of the difference region. Since use of -M and
843 * -C options may lead to overlapping/duplicate source line number
844 * ranges, all we can rely on from sorting/merging is the order of the
845 * first suspect line number.
1732a1fd 846 */
7e6ac6e4
DK
847static void blame_chunk(struct blame_entry ***dstq, struct blame_entry ***srcq,
848 int tlno, int offset, int same,
849 struct origin *parent)
cee7f245 850{
7e6ac6e4
DK
851 struct blame_entry *e = **srcq;
852 struct blame_entry *samep = NULL, *diffp = NULL;
cee7f245 853
7e6ac6e4
DK
854 while (e && e->s_lno < tlno) {
855 struct blame_entry *next = e->next;
856 /*
857 * current record starts before differing portion. If
858 * it reaches into it, we need to split it up and
859 * examine the second part separately.
860 */
861 if (e->s_lno + e->num_lines > tlno) {
862 /* Move second half to a new record */
863 int len = tlno - e->s_lno;
864 struct blame_entry *n = xcalloc(1, sizeof (struct blame_entry));
865 n->suspect = e->suspect;
866 n->lno = e->lno + len;
867 n->s_lno = e->s_lno + len;
868 n->num_lines = e->num_lines - len;
869 e->num_lines = len;
870 e->score = 0;
871 /* Push new record to diffp */
872 n->next = diffp;
873 diffp = n;
874 } else
875 origin_decref(e->suspect);
876 /* Pass blame for everything before the differing
877 * chunk to the parent */
878 e->suspect = origin_incref(parent);
879 e->s_lno += offset;
880 e->next = samep;
881 samep = e;
882 e = next;
883 }
884 /*
885 * As we don't know how much of a common stretch after this
886 * diff will occur, the currently blamed parts are all that we
887 * can assign to the parent for now.
888 */
889
890 if (samep) {
891 **dstq = reverse_blame(samep, **dstq);
892 *dstq = &samep->next;
cee7f245 893 }
7e6ac6e4
DK
894 /*
895 * Prepend the split off portions: everything after e starts
896 * after the blameable portion.
897 */
898 e = reverse_blame(diffp, e);
899
900 /*
901 * Now retain records on the target while parts are different
902 * from the parent.
903 */
904 samep = NULL;
905 diffp = NULL;
906 while (e && e->s_lno < same) {
907 struct blame_entry *next = e->next;
908
909 /*
910 * If current record extends into sameness, need to split.
911 */
912 if (e->s_lno + e->num_lines > same) {
913 /*
914 * Move second half to a new record to be
915 * processed by later chunks
916 */
917 int len = same - e->s_lno;
918 struct blame_entry *n = xcalloc(1, sizeof (struct blame_entry));
919 n->suspect = origin_incref(e->suspect);
920 n->lno = e->lno + len;
921 n->s_lno = e->s_lno + len;
922 n->num_lines = e->num_lines - len;
923 e->num_lines = len;
924 e->score = 0;
925 /* Push new record to samep */
926 n->next = samep;
927 samep = n;
928 }
929 e->next = diffp;
930 diffp = e;
931 e = next;
932 }
933 **srcq = reverse_blame(diffp, reverse_blame(samep, e));
934 /* Move across elements that are in the unblamable portion */
935 if (diffp)
936 *srcq = &diffp->next;
cee7f245
JH
937}
938
cdcab133 939struct blame_chunk_cb_data {
cdcab133 940 struct origin *parent;
7e6ac6e4
DK
941 long offset;
942 struct blame_entry **dstq;
943 struct blame_entry **srcq;
cdcab133
RS
944};
945
7e6ac6e4 946/* diff chunks are from parent to target */
0af596c6
RS
947static int blame_chunk_cb(long start_a, long count_a,
948 long start_b, long count_b, void *data)
cdcab133
RS
949{
950 struct blame_chunk_cb_data *d = data;
7e6ac6e4
DK
951 if (start_a - start_b != d->offset)
952 die("internal error in blame::blame_chunk_cb");
953 blame_chunk(&d->dstq, &d->srcq, start_b, start_a - start_b,
954 start_b + count_b, d->parent);
955 d->offset = start_a + count_a - (start_b + count_b);
0af596c6 956 return 0;
cdcab133
RS
957}
958
1732a1fd
JH
959/*
960 * We are looking at the origin 'target' and aiming to pass blame
961 * for the lines it is suspected to its parent. Run diff to find
962 * which lines came from parent and pass blame for them.
963 */
7e6ac6e4
DK
964static void pass_blame_to_parent(struct scoreboard *sb,
965 struct origin *target,
966 struct origin *parent)
cee7f245 967{
7c4ed8c8 968 mmfile_t file_p, file_o;
66dbfd55 969 struct blame_chunk_cb_data d;
7e6ac6e4 970 struct blame_entry *newdest = NULL;
0af596c6 971
7e6ac6e4
DK
972 if (!target->suspects)
973 return; /* nothing remains for this target */
974
975 d.parent = parent;
976 d.offset = 0;
977 d.dstq = &newdest; d.srcq = &target->suspects;
cee7f245 978
3b8a12e8
AB
979 fill_origin_blob(&sb->revs->diffopt, parent, &file_p);
980 fill_origin_blob(&sb->revs->diffopt, target, &file_o);
7c4ed8c8 981 num_get_patch++;
cee7f245 982
3efb9880
JK
983 if (diff_hunks(&file_p, &file_o, 0, blame_chunk_cb, &d))
984 die("unable to generate diff (%s -> %s)",
f2fd0760 985 oid_to_hex(&parent->commit->object.oid),
986 oid_to_hex(&target->commit->object.oid));
7e6ac6e4
DK
987 /* The rest are the same as the parent */
988 blame_chunk(&d.dstq, &d.srcq, INT_MAX, d.offset, INT_MAX, parent);
989 *d.dstq = NULL;
990 queue_blames(sb, parent, newdest);
cee7f245 991
7e6ac6e4 992 return;
cee7f245
JH
993}
994
1732a1fd
JH
995/*
996 * The lines in blame_entry after splitting blames many times can become
997 * very small and trivial, and at some point it becomes pointless to
998 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
999 * ordinary C program, and it is not worth to say it was copied from
1000 * totally unrelated file in the parent.
1001 *
1002 * Compute how trivial the lines in the blame_entry are.
1003 */
5ff62c30
JH
1004static unsigned ent_score(struct scoreboard *sb, struct blame_entry *e)
1005{
1006 unsigned score;
1007 const char *cp, *ep;
1008
1009 if (e->score)
1010 return e->score;
1011
612702e8 1012 score = 1;
5ff62c30
JH
1013 cp = nth_line(sb, e->lno);
1014 ep = nth_line(sb, e->lno + e->num_lines);
1015 while (cp < ep) {
1016 unsigned ch = *((unsigned char *)cp);
1017 if (isalnum(ch))
1018 score++;
1019 cp++;
1020 }
1021 e->score = score;
1022 return score;
1023}
1024
1732a1fd
JH
1025/*
1026 * best_so_far[] and this[] are both a split of an existing blame_entry
1027 * that passes blame to the parent. Maintain best_so_far the best split
1028 * so far, by comparing this and best_so_far and copying this into
1029 * bst_so_far as needed.
1030 */
5ff62c30 1031static void copy_split_if_better(struct scoreboard *sb,
54a4c617
JH
1032 struct blame_entry *best_so_far,
1033 struct blame_entry *this)
d24bba80 1034{
54a4c617
JH
1035 int i;
1036
d24bba80
JH
1037 if (!this[1].suspect)
1038 return;
5ff62c30
JH
1039 if (best_so_far[1].suspect) {
1040 if (ent_score(sb, &this[1]) < ent_score(sb, &best_so_far[1]))
1041 return;
1042 }
54a4c617
JH
1043
1044 for (i = 0; i < 3; i++)
1045 origin_incref(this[i].suspect);
1046 decref_split(best_so_far);
d24bba80
JH
1047 memcpy(best_so_far, this, sizeof(struct blame_entry [3]));
1048}
1049
dd166aa8
JH
1050/*
1051 * We are looking at a part of the final image represented by
1052 * ent (tlno and same are offset by ent->s_lno).
1053 * tlno is where we are looking at in the final image.
1054 * up to (but not including) same match preimage.
1055 * plno is where we are looking at in the preimage.
1056 *
1057 * <-------------- final image ---------------------->
1058 * <------ent------>
1059 * ^tlno ^same
1060 * <---------preimage----->
1061 * ^plno
1062 *
1063 * All line numbers are 0-based.
1064 */
1065static void handle_split(struct scoreboard *sb,
1066 struct blame_entry *ent,
1067 int tlno, int plno, int same,
1068 struct origin *parent,
1069 struct blame_entry *split)
1070{
1071 if (ent->num_lines <= tlno)
1072 return;
1073 if (tlno < same) {
1074 struct blame_entry this[3];
1075 tlno += ent->s_lno;
1076 same += ent->s_lno;
1077 split_overlap(this, ent, tlno, plno, same, parent);
1078 copy_split_if_better(sb, split, this);
1079 decref_split(this);
1080 }
1081}
1082
cdcab133
RS
1083struct handle_split_cb_data {
1084 struct scoreboard *sb;
1085 struct blame_entry *ent;
1086 struct origin *parent;
1087 struct blame_entry *split;
1088 long plno;
1089 long tlno;
1090};
1091
5d23ec76
RS
1092static int handle_split_cb(long start_a, long count_a,
1093 long start_b, long count_b, void *data)
cdcab133
RS
1094{
1095 struct handle_split_cb_data *d = data;
5d23ec76
RS
1096 handle_split(d->sb, d->ent, d->tlno, d->plno, start_b, d->parent,
1097 d->split);
1098 d->plno = start_a + count_a;
1099 d->tlno = start_b + count_b;
1100 return 0;
cdcab133
RS
1101}
1102
1732a1fd
JH
1103/*
1104 * Find the lines from parent that are the same as ent so that
1105 * we can pass blames to it. file_p has the blob contents for
1106 * the parent.
1107 */
d24bba80
JH
1108static void find_copy_in_blob(struct scoreboard *sb,
1109 struct blame_entry *ent,
1110 struct origin *parent,
54a4c617 1111 struct blame_entry *split,
d24bba80
JH
1112 mmfile_t *file_p)
1113{
1114 const char *cp;
d24bba80 1115 mmfile_t file_o;
66dbfd55 1116 struct handle_split_cb_data d;
5d23ec76 1117
66dbfd55
GV
1118 memset(&d, 0, sizeof(d));
1119 d.sb = sb; d.ent = ent; d.parent = parent; d.split = split;
1732a1fd
JH
1120 /*
1121 * Prepare mmfile that contains only the lines in ent.
1122 */
d24bba80 1123 cp = nth_line(sb, ent->lno);
4b25d091 1124 file_o.ptr = (char *) cp;
3ee8944f 1125 file_o.size = nth_line(sb, ent->lno + ent->num_lines) - cp;
d24bba80 1126
dd166aa8
JH
1127 /*
1128 * file_o is a part of final image we are annotating.
1129 * file_p partially may match that image.
1130 */
d24bba80 1131 memset(split, 0, sizeof(struct blame_entry [3]));
3efb9880
JK
1132 if (diff_hunks(file_p, &file_o, 1, handle_split_cb, &d))
1133 die("unable to generate diff (%s)",
f2fd0760 1134 oid_to_hex(&parent->commit->object.oid));
dd166aa8 1135 /* remainder, if any, all match the preimage */
cdcab133 1136 handle_split(sb, ent, d.tlno, d.plno, ent->num_lines, parent, split);
d24bba80
JH
1137}
1138
7e6ac6e4
DK
1139/* Move all blame entries from list *source that have a score smaller
1140 * than score_min to the front of list *small.
1141 * Returns a pointer to the link pointing to the old head of the small list.
1142 */
1143
1144static struct blame_entry **filter_small(struct scoreboard *sb,
1145 struct blame_entry **small,
1146 struct blame_entry **source,
1147 unsigned score_min)
1148{
1149 struct blame_entry *p = *source;
1150 struct blame_entry *oldsmall = *small;
1151 while (p) {
1152 if (ent_score(sb, p) <= score_min) {
1153 *small = p;
1154 small = &p->next;
1155 p = *small;
1156 } else {
1157 *source = p;
1158 source = &p->next;
1159 p = *source;
1160 }
1161 }
1162 *small = oldsmall;
1163 *source = NULL;
1164 return small;
1165}
1166
1732a1fd
JH
1167/*
1168 * See if lines currently target is suspected for can be attributed to
1169 * parent.
1170 */
7e6ac6e4
DK
1171static void find_move_in_parent(struct scoreboard *sb,
1172 struct blame_entry ***blamed,
1173 struct blame_entry **toosmall,
1174 struct origin *target,
1175 struct origin *parent)
d24bba80 1176{
46014766 1177 struct blame_entry *e, split[3];
7e6ac6e4
DK
1178 struct blame_entry *unblamed = target->suspects;
1179 struct blame_entry *leftover = NULL;
d24bba80 1180 mmfile_t file_p;
d24bba80 1181
7e6ac6e4
DK
1182 if (!unblamed)
1183 return; /* nothing remains for this target */
d24bba80 1184
3b8a12e8 1185 fill_origin_blob(&sb->revs->diffopt, parent, &file_p);
c2e525d9 1186 if (!file_p.ptr)
7e6ac6e4 1187 return;
d24bba80 1188
7e6ac6e4
DK
1189 /* At each iteration, unblamed has a NULL-terminated list of
1190 * entries that have not yet been tested for blame. leftover
1191 * contains the reversed list of entries that have been tested
1192 * without being assignable to the parent.
1193 */
1194 do {
1195 struct blame_entry **unblamedtail = &unblamed;
1196 struct blame_entry *next;
1197 for (e = unblamed; e; e = next) {
1198 next = e->next;
650e2f67
JH
1199 find_copy_in_blob(sb, e, parent, split, &file_p);
1200 if (split[1].suspect &&
1201 blame_move_score < ent_score(sb, &split[1])) {
7e6ac6e4
DK
1202 split_blame(blamed, &unblamedtail, split, e);
1203 } else {
1204 e->next = leftover;
1205 leftover = e;
650e2f67
JH
1206 }
1207 decref_split(split);
1208 }
7e6ac6e4
DK
1209 *unblamedtail = NULL;
1210 toosmall = filter_small(sb, toosmall, &unblamed, blame_move_score);
1211 } while (unblamed);
1212 target->suspects = reverse_blame(leftover, NULL);
d24bba80
JH
1213}
1214
33494784
JH
1215struct blame_list {
1216 struct blame_entry *ent;
1217 struct blame_entry split[3];
1218};
1219
1732a1fd
JH
1220/*
1221 * Count the number of entries the target is suspected for,
1222 * and prepare a list of entry and the best split.
1223 */
7e6ac6e4 1224static struct blame_list *setup_blame_list(struct blame_entry *unblamed,
33494784
JH
1225 int *num_ents_p)
1226{
1227 struct blame_entry *e;
1228 int num_ents, i;
1229 struct blame_list *blame_list = NULL;
1230
7e6ac6e4
DK
1231 for (e = unblamed, num_ents = 0; e; e = e->next)
1232 num_ents++;
33494784
JH
1233 if (num_ents) {
1234 blame_list = xcalloc(num_ents, sizeof(struct blame_list));
7e6ac6e4
DK
1235 for (e = unblamed, i = 0; e; e = e->next)
1236 blame_list[i++].ent = e;
33494784
JH
1237 }
1238 *num_ents_p = num_ents;
1239 return blame_list;
1240}
1241
1732a1fd
JH
1242/*
1243 * For lines target is suspected for, see if we can find code movement
1244 * across file boundary from the parent commit. porigin is the path
1245 * in the parent we already tried.
1246 */
7e6ac6e4
DK
1247static void find_copy_in_parent(struct scoreboard *sb,
1248 struct blame_entry ***blamed,
1249 struct blame_entry **toosmall,
1250 struct origin *target,
1251 struct commit *parent,
1252 struct origin *porigin,
1253 int opt)
18abd745
JH
1254{
1255 struct diff_options diff_opts;
aec8fa1f 1256 int i, j;
33494784 1257 struct blame_list *blame_list;
aec8fa1f 1258 int num_ents;
7e6ac6e4
DK
1259 struct blame_entry *unblamed = target->suspects;
1260 struct blame_entry *leftover = NULL;
18abd745 1261
7e6ac6e4
DK
1262 if (!unblamed)
1263 return; /* nothing remains for this target */
18abd745
JH
1264
1265 diff_setup(&diff_opts);
8f67f8ae 1266 DIFF_OPT_SET(&diff_opts, RECURSIVE);
18abd745
JH
1267 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
1268
28452655 1269 diff_setup_done(&diff_opts);
33494784
JH
1270
1271 /* Try "find copies harder" on new path if requested;
1272 * we do not want to use diffcore_rename() actually to
1273 * match things up; find_copies_harder is set only to
1274 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1275 * and this code needs to be after diff_setup_done(), which
1276 * usually makes find-copies-harder imply copy detection.
1277 */
c63777c0
JH
1278 if ((opt & PICKAXE_BLAME_COPY_HARDEST)
1279 || ((opt & PICKAXE_BLAME_COPY_HARDER)
1280 && (!porigin || strcmp(target->path, porigin->path))))
8f67f8ae 1281 DIFF_OPT_SET(&diff_opts, FIND_COPIES_HARDER);
33494784 1282
f2fd0760 1283 if (is_null_oid(&target->commit->object.oid))
ed1c9977 1284 do_diff_cache(parent->tree->object.oid.hash, &diff_opts);
1cfe7733 1285 else
ed1c9977 1286 diff_tree_sha1(parent->tree->object.oid.hash,
1287 target->commit->tree->object.oid.hash,
1cfe7733 1288 "", &diff_opts);
18abd745 1289
8f67f8ae 1290 if (!DIFF_OPT_TST(&diff_opts, FIND_COPIES_HARDER))
33494784 1291 diffcore_std(&diff_opts);
18abd745 1292
7e6ac6e4
DK
1293 do {
1294 struct blame_entry **unblamedtail = &unblamed;
1295 blame_list = setup_blame_list(unblamed, &num_ents);
33494784
JH
1296
1297 for (i = 0; i < diff_queued_diff.nr; i++) {
1298 struct diff_filepair *p = diff_queued_diff.queue[i];
1299 struct origin *norigin;
1300 mmfile_t file_p;
33494784
JH
1301 struct blame_entry this[3];
1302
1303 if (!DIFF_FILE_VALID(p->one))
1304 continue; /* does not exist in parent */
5209ac4d
AG
1305 if (S_ISGITLINK(p->one->mode))
1306 continue; /* ignore git links */
33494784
JH
1307 if (porigin && !strcmp(p->one->path, porigin->path))
1308 /* find_move already dealt with this path */
1309 continue;
1310
1311 norigin = get_origin(sb, parent, p->one->path);
1312 hashcpy(norigin->blob_sha1, p->one->sha1);
90064710 1313 norigin->mode = p->one->mode;
3b8a12e8 1314 fill_origin_blob(&sb->revs->diffopt, norigin, &file_p);
33494784
JH
1315 if (!file_p.ptr)
1316 continue;
1317
1318 for (j = 0; j < num_ents; j++) {
1319 find_copy_in_blob(sb, blame_list[j].ent,
1320 norigin, this, &file_p);
1321 copy_split_if_better(sb, blame_list[j].split,
1322 this);
1323 decref_split(this);
1324 }
33494784
JH
1325 origin_decref(norigin);
1326 }
18abd745 1327
aec8fa1f 1328 for (j = 0; j < num_ents; j++) {
33494784
JH
1329 struct blame_entry *split = blame_list[j].split;
1330 if (split[1].suspect &&
1331 blame_copy_score < ent_score(sb, &split[1])) {
7e6ac6e4
DK
1332 split_blame(blamed, &unblamedtail, split,
1333 blame_list[j].ent);
1334 } else {
1335 blame_list[j].ent->next = leftover;
1336 leftover = blame_list[j].ent;
33494784
JH
1337 }
1338 decref_split(split);
18abd745 1339 }
33494784 1340 free(blame_list);
7e6ac6e4
DK
1341 *unblamedtail = NULL;
1342 toosmall = filter_small(sb, toosmall, &unblamed, blame_copy_score);
1343 } while (unblamed);
1344 target->suspects = reverse_blame(leftover, NULL);
33494784 1345 diff_flush(&diff_opts);
ed6e8038 1346 clear_pathspec(&diff_opts.pathspec);
18abd745
JH
1347}
1348
1732a1fd
JH
1349/*
1350 * The blobs of origin and porigin exactly match, so everything
c2e525d9
JH
1351 * origin is suspected for can be blamed on the parent.
1352 */
1353static void pass_whole_blame(struct scoreboard *sb,
1354 struct origin *origin, struct origin *porigin)
1355{
7e6ac6e4 1356 struct blame_entry *e, *suspects;
c2e525d9
JH
1357
1358 if (!porigin->file.ptr && origin->file.ptr) {
1359 /* Steal its file */
1360 porigin->file = origin->file;
1361 origin->file.ptr = NULL;
1362 }
7e6ac6e4
DK
1363 suspects = origin->suspects;
1364 origin->suspects = NULL;
1365 for (e = suspects; e; e = e->next) {
c2e525d9
JH
1366 origin_incref(porigin);
1367 origin_decref(e->suspect);
1368 e->suspect = porigin;
1369 }
7e6ac6e4 1370 queue_blames(sb, porigin, suspects);
c2e525d9
JH
1371}
1372
69264f46
JH
1373/*
1374 * We pass blame from the current commit to its parents. We keep saying
1375 * "parent" (and "porigin"), but what we mean is to find scapegoat to
1376 * exonerate ourselves.
1377 */
85af7929 1378static struct commit_list *first_scapegoat(struct rev_info *revs, struct commit *commit)
69264f46 1379{
95a4fb0e
JK
1380 if (!reverse) {
1381 if (revs->first_parent_only &&
1382 commit->parents &&
1383 commit->parents->next) {
1384 free_commit_list(commit->parents->next);
1385 commit->parents->next = NULL;
1386 }
85af7929 1387 return commit->parents;
95a4fb0e 1388 }
85af7929 1389 return lookup_decoration(&revs->children, &commit->object);
69264f46
JH
1390}
1391
85af7929 1392static int num_scapegoats(struct rev_info *revs, struct commit *commit)
69264f46 1393{
85af7929 1394 struct commit_list *l = first_scapegoat(revs, commit);
4bbaa1eb 1395 return commit_list_count(l);
69264f46
JH
1396}
1397
7e6ac6e4
DK
1398/* Distribute collected unsorted blames to the respected sorted lists
1399 * in the various origins.
1400 */
1401static void distribute_blame(struct scoreboard *sb, struct blame_entry *blamed)
1402{
1403 blamed = blame_sort(blamed, compare_blame_suspect);
1404 while (blamed)
1405 {
1406 struct origin *porigin = blamed->suspect;
1407 struct blame_entry *suspects = NULL;
1408 do {
1409 struct blame_entry *next = blamed->next;
1410 blamed->next = suspects;
1411 suspects = blamed;
1412 blamed = next;
1413 } while (blamed && blamed->suspect == porigin);
1414 suspects = reverse_blame(suspects, NULL);
1415 queue_blames(sb, porigin, suspects);
1416 }
1417}
1418
69264f46 1419#define MAXSG 16
cee7f245 1420
d24bba80 1421static void pass_blame(struct scoreboard *sb, struct origin *origin, int opt)
cee7f245 1422{
85af7929 1423 struct rev_info *revs = sb->revs;
69264f46 1424 int i, pass, num_sg;
cee7f245 1425 struct commit *commit = origin->commit;
69264f46
JH
1426 struct commit_list *sg;
1427 struct origin *sg_buf[MAXSG];
1428 struct origin *porigin, **sg_origin = sg_buf;
7e6ac6e4
DK
1429 struct blame_entry *toosmall = NULL;
1430 struct blame_entry *blames, **blametail = &blames;
69264f46 1431
85af7929 1432 num_sg = num_scapegoats(revs, commit);
69264f46
JH
1433 if (!num_sg)
1434 goto finish;
1435 else if (num_sg < ARRAY_SIZE(sg_buf))
1436 memset(sg_buf, 0, sizeof(sg_buf));
1437 else
1438 sg_origin = xcalloc(num_sg, sizeof(*sg_origin));
cee7f245 1439
69264f46
JH
1440 /*
1441 * The first pass looks for unrenamed path to optimize for
f69e743d
JH
1442 * common cases, then we look for renames in the second pass.
1443 */
3d1aa566 1444 for (pass = 0; pass < 2 - no_whole_file_rename; pass++) {
f69e743d
JH
1445 struct origin *(*find)(struct scoreboard *,
1446 struct commit *, struct origin *);
1447 find = pass ? find_rename : find_origin;
1448
85af7929 1449 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1450 i < num_sg && sg;
1451 sg = sg->next, i++) {
1452 struct commit *p = sg->item;
0421d9f8 1453 int j, same;
f69e743d 1454
69264f46 1455 if (sg_origin[i])
f69e743d
JH
1456 continue;
1457 if (parse_commit(p))
1458 continue;
0d981c67 1459 porigin = find(sb, p, origin);
f69e743d
JH
1460 if (!porigin)
1461 continue;
1462 if (!hashcmp(porigin->blob_sha1, origin->blob_sha1)) {
c2e525d9 1463 pass_whole_blame(sb, origin, porigin);
f69e743d
JH
1464 origin_decref(porigin);
1465 goto finish;
1466 }
0421d9f8 1467 for (j = same = 0; j < i; j++)
69264f46
JH
1468 if (sg_origin[j] &&
1469 !hashcmp(sg_origin[j]->blob_sha1,
0421d9f8
JH
1470 porigin->blob_sha1)) {
1471 same = 1;
1472 break;
1473 }
1474 if (!same)
69264f46 1475 sg_origin[i] = porigin;
0421d9f8
JH
1476 else
1477 origin_decref(porigin);
cee7f245 1478 }
cee7f245
JH
1479 }
1480
c2e525d9 1481 num_commits++;
85af7929 1482 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1483 i < num_sg && sg;
1484 sg = sg->next, i++) {
1485 struct origin *porigin = sg_origin[i];
cee7f245
JH
1486 if (!porigin)
1487 continue;
96e11709
JH
1488 if (!origin->previous) {
1489 origin_incref(porigin);
1490 origin->previous = porigin;
1491 }
7e6ac6e4
DK
1492 pass_blame_to_parent(sb, origin, porigin);
1493 if (!origin->suspects)
54a4c617 1494 goto finish;
cee7f245 1495 }
d24bba80
JH
1496
1497 /*
1732a1fd 1498 * Optionally find moves in parents' files.
d24bba80 1499 */
7e6ac6e4
DK
1500 if (opt & PICKAXE_BLAME_MOVE) {
1501 filter_small(sb, &toosmall, &origin->suspects, blame_move_score);
1502 if (origin->suspects) {
1503 for (i = 0, sg = first_scapegoat(revs, commit);
1504 i < num_sg && sg;
1505 sg = sg->next, i++) {
1506 struct origin *porigin = sg_origin[i];
1507 if (!porigin)
1508 continue;
1509 find_move_in_parent(sb, &blametail, &toosmall, origin, porigin);
1510 if (!origin->suspects)
1511 break;
1512 }
d24bba80 1513 }
7e6ac6e4 1514 }
d24bba80 1515
18abd745 1516 /*
1732a1fd 1517 * Optionally find copies from parents' files.
18abd745 1518 */
7e6ac6e4
DK
1519 if (opt & PICKAXE_BLAME_COPY) {
1520 if (blame_copy_score > blame_move_score)
1521 filter_small(sb, &toosmall, &origin->suspects, blame_copy_score);
1522 else if (blame_copy_score < blame_move_score) {
1523 origin->suspects = blame_merge(origin->suspects, toosmall);
1524 toosmall = NULL;
1525 filter_small(sb, &toosmall, &origin->suspects, blame_copy_score);
1526 }
1527 if (!origin->suspects)
1528 goto finish;
1529
85af7929 1530 for (i = 0, sg = first_scapegoat(revs, commit);
69264f46
JH
1531 i < num_sg && sg;
1532 sg = sg->next, i++) {
1533 struct origin *porigin = sg_origin[i];
7e6ac6e4
DK
1534 find_copy_in_parent(sb, &blametail, &toosmall,
1535 origin, sg->item, porigin, opt);
1536 if (!origin->suspects)
54a4c617 1537 goto finish;
18abd745 1538 }
7e6ac6e4 1539 }
54a4c617 1540
7e6ac6e4
DK
1541finish:
1542 *blametail = NULL;
1543 distribute_blame(sb, blames);
1544 /*
1545 * prepend toosmall to origin->suspects
1546 *
1547 * There is no point in sorting: this ends up on a big
1548 * unsorted list in the caller anyway.
1549 */
1550 if (toosmall) {
1551 struct blame_entry **tail = &toosmall;
1552 while (*tail)
1553 tail = &(*tail)->next;
1554 *tail = origin->suspects;
1555 origin->suspects = toosmall;
1556 }
69264f46
JH
1557 for (i = 0; i < num_sg; i++) {
1558 if (sg_origin[i]) {
1559 drop_origin_blob(sg_origin[i]);
1560 origin_decref(sg_origin[i]);
7c3c7962
JH
1561 }
1562 }
1563 drop_origin_blob(origin);
69264f46
JH
1564 if (sg_buf != sg_origin)
1565 free(sg_origin);
cee7f245
JH
1566}
1567
1732a1fd
JH
1568/*
1569 * Information on commits, used for output.
1570 */
9cba13ca 1571struct commit_info {
ea02ffa3
AP
1572 struct strbuf author;
1573 struct strbuf author_mail;
cee7f245 1574 unsigned long author_time;
ea02ffa3 1575 struct strbuf author_tz;
cee7f245
JH
1576
1577 /* filled only when asked for details */
ea02ffa3
AP
1578 struct strbuf committer;
1579 struct strbuf committer_mail;
cee7f245 1580 unsigned long committer_time;
ea02ffa3 1581 struct strbuf committer_tz;
cee7f245 1582
ea02ffa3 1583 struct strbuf summary;
cee7f245
JH
1584};
1585
1732a1fd
JH
1586/*
1587 * Parse author/committer line in the commit object buffer
1588 */
cee7f245 1589static void get_ac_line(const char *inbuf, const char *what,
ea02ffa3
AP
1590 struct strbuf *name, struct strbuf *mail,
1591 unsigned long *time, struct strbuf *tz)
cee7f245 1592{
3c020bd5 1593 struct ident_split ident;
ea02ffa3
AP
1594 size_t len, maillen, namelen;
1595 char *tmp, *endp;
1596 const char *namebuf, *mailbuf;
cee7f245
JH
1597
1598 tmp = strstr(inbuf, what);
1599 if (!tmp)
1600 goto error_out;
1601 tmp += strlen(what);
1602 endp = strchr(tmp, '\n');
1603 if (!endp)
1604 len = strlen(tmp);
1605 else
1606 len = endp - tmp;
3c020bd5
AP
1607
1608 if (split_ident_line(&ident, tmp, len)) {
cee7f245
JH
1609 error_out:
1610 /* Ugh */
ea02ffa3
AP
1611 tmp = "(unknown)";
1612 strbuf_addstr(name, tmp);
1613 strbuf_addstr(mail, tmp);
1614 strbuf_addstr(tz, tmp);
cee7f245
JH
1615 *time = 0;
1616 return;
1617 }
cee7f245 1618
3c020bd5 1619 namelen = ident.name_end - ident.name_begin;
ea02ffa3 1620 namebuf = ident.name_begin;
cee7f245 1621
ea02ffa3
AP
1622 maillen = ident.mail_end - ident.mail_begin;
1623 mailbuf = ident.mail_begin;
f9567384 1624
de5abe9f
RS
1625 if (ident.date_begin && ident.date_end)
1626 *time = strtoul(ident.date_begin, NULL, 10);
1627 else
1628 *time = 0;
f9567384 1629
de5abe9f
RS
1630 if (ident.tz_begin && ident.tz_end)
1631 strbuf_add(tz, ident.tz_begin, ident.tz_end - ident.tz_begin);
1632 else
1633 strbuf_addstr(tz, "(unknown)");
f9567384 1634
f9567384 1635 /*
d20d654f 1636 * Now, convert both name and e-mail using mailmap
f9567384 1637 */
ea02ffa3
AP
1638 map_user(&mailmap, &mailbuf, &maillen,
1639 &namebuf, &namelen);
1640
1641 strbuf_addf(mail, "<%.*s>", (int)maillen, mailbuf);
1642 strbuf_add(name, namebuf, namelen);
1643}
1644
1645static void commit_info_init(struct commit_info *ci)
1646{
1647
1648 strbuf_init(&ci->author, 0);
1649 strbuf_init(&ci->author_mail, 0);
1650 strbuf_init(&ci->author_tz, 0);
1651 strbuf_init(&ci->committer, 0);
1652 strbuf_init(&ci->committer_mail, 0);
1653 strbuf_init(&ci->committer_tz, 0);
1654 strbuf_init(&ci->summary, 0);
1655}
1656
1657static void commit_info_destroy(struct commit_info *ci)
1658{
1659
1660 strbuf_release(&ci->author);
1661 strbuf_release(&ci->author_mail);
1662 strbuf_release(&ci->author_tz);
1663 strbuf_release(&ci->committer);
1664 strbuf_release(&ci->committer_mail);
1665 strbuf_release(&ci->committer_tz);
1666 strbuf_release(&ci->summary);
cee7f245
JH
1667}
1668
1669static void get_commit_info(struct commit *commit,
1670 struct commit_info *ret,
1671 int detailed)
1672{
1673 int len;
e297cf5a 1674 const char *subject, *encoding;
b000c59b 1675 const char *message;
ea02ffa3
AP
1676
1677 commit_info_init(ret);
cee7f245 1678
e297cf5a 1679 encoding = get_log_output_encoding();
5a10d236 1680 message = logmsg_reencode(commit, NULL, encoding);
69cd8f63 1681 get_ac_line(message, "\nauthor ",
ea02ffa3 1682 &ret->author, &ret->author_mail,
cee7f245
JH
1683 &ret->author_time, &ret->author_tz);
1684
69cd8f63 1685 if (!detailed) {
b66103c3 1686 unuse_commit_buffer(commit, message);
cee7f245 1687 return;
69cd8f63 1688 }
cee7f245 1689
69cd8f63 1690 get_ac_line(message, "\ncommitter ",
ea02ffa3 1691 &ret->committer, &ret->committer_mail,
cee7f245
JH
1692 &ret->committer_time, &ret->committer_tz);
1693
ad98a58b 1694 len = find_commit_subject(message, &subject);
ea02ffa3
AP
1695 if (len)
1696 strbuf_add(&ret->summary, subject, len);
1697 else
f2fd0760 1698 strbuf_addf(&ret->summary, "(%s)", oid_to_hex(&commit->object.oid));
ea02ffa3 1699
b66103c3 1700 unuse_commit_buffer(commit, message);
cee7f245
JH
1701}
1702
1732a1fd
JH
1703/*
1704 * To allow LF and other nonportable characters in pathnames,
1705 * they are c-style quoted as needed.
1706 */
46e5e69d
JH
1707static void write_filename_info(const char *path)
1708{
1709 printf("filename ");
663af342 1710 write_name_quoted(path, stdout, '\n');
46e5e69d
JH
1711}
1712
9991030c
JH
1713/*
1714 * Porcelain/Incremental format wants to show a lot of details per
1715 * commit. Instead of repeating this every line, emit it only once,
e86226e3
JK
1716 * the first time each commit appears in the output (unless the
1717 * user has specifically asked for us to repeat).
9991030c 1718 */
e86226e3 1719static int emit_one_suspect_detail(struct origin *suspect, int repeat)
9991030c
JH
1720{
1721 struct commit_info ci;
1722
e86226e3 1723 if (!repeat && (suspect->commit->object.flags & METAINFO_SHOWN))
9991030c
JH
1724 return 0;
1725
1726 suspect->commit->object.flags |= METAINFO_SHOWN;
1727 get_commit_info(suspect->commit, &ci, 1);
ea02ffa3
AP
1728 printf("author %s\n", ci.author.buf);
1729 printf("author-mail %s\n", ci.author_mail.buf);
9991030c 1730 printf("author-time %lu\n", ci.author_time);
ea02ffa3
AP
1731 printf("author-tz %s\n", ci.author_tz.buf);
1732 printf("committer %s\n", ci.committer.buf);
1733 printf("committer-mail %s\n", ci.committer_mail.buf);
9991030c 1734 printf("committer-time %lu\n", ci.committer_time);
ea02ffa3
AP
1735 printf("committer-tz %s\n", ci.committer_tz.buf);
1736 printf("summary %s\n", ci.summary.buf);
9991030c
JH
1737 if (suspect->commit->object.flags & UNINTERESTING)
1738 printf("boundary\n");
96e11709
JH
1739 if (suspect->previous) {
1740 struct origin *prev = suspect->previous;
f2fd0760 1741 printf("previous %s ", oid_to_hex(&prev->commit->object.oid));
96e11709
JH
1742 write_name_quoted(prev->path, stdout, '\n');
1743 }
ea02ffa3
AP
1744
1745 commit_info_destroy(&ci);
1746
9991030c
JH
1747 return 1;
1748}
1749
1732a1fd 1750/*
7e6ac6e4
DK
1751 * The blame_entry is found to be guilty for the range.
1752 * Show it in incremental output.
1732a1fd 1753 */
aba37f49
ECA
1754static void found_guilty_entry(struct blame_entry *ent,
1755 struct progress_info *pi)
717d1462 1756{
717d1462
LT
1757 if (incremental) {
1758 struct origin *suspect = ent->suspect;
1759
1760 printf("%s %d %d %d\n",
f2fd0760 1761 oid_to_hex(&suspect->commit->object.oid),
717d1462 1762 ent->s_lno + 1, ent->lno + 1, ent->num_lines);
e86226e3 1763 emit_one_suspect_detail(suspect, 0);
46e5e69d 1764 write_filename_info(suspect->path);
06f59e9f 1765 maybe_flush_or_die(stdout, "stdout");
717d1462 1766 }
aba37f49
ECA
1767 pi->blamed_lines += ent->num_lines;
1768 display_progress(pi->progress, pi->blamed_lines);
717d1462
LT
1769}
1770
1732a1fd 1771/*
7e6ac6e4
DK
1772 * The main loop -- while we have blobs with lines whose true origin
1773 * is still unknown, pick one blob, and allow its lines to pass blames
1774 * to its parents. */
85af7929 1775static void assign_blame(struct scoreboard *sb, int opt)
717d1462 1776{
85af7929 1777 struct rev_info *revs = sb->revs;
7e6ac6e4 1778 struct commit *commit = prio_queue_get(&sb->commits);
aba37f49
ECA
1779 struct progress_info pi = { NULL, 0 };
1780
1781 if (show_progress)
1782 pi.progress = start_progress_delay(_("Blaming lines"),
1783 sb->num_lines, 50, 1);
85af7929 1784
7e6ac6e4 1785 while (commit) {
717d1462 1786 struct blame_entry *ent;
7e6ac6e4 1787 struct origin *suspect = commit->util;
717d1462
LT
1788
1789 /* find one suspect to break down */
7e6ac6e4
DK
1790 while (suspect && !suspect->suspects)
1791 suspect = suspect->next;
1792
1793 if (!suspect) {
1794 commit = prio_queue_get(&sb->commits);
1795 continue;
1796 }
1797
1798 assert(commit == suspect->commit);
717d1462 1799
1732a1fd
JH
1800 /*
1801 * We will use this suspect later in the loop,
1802 * so hold onto it in the meantime.
1803 */
717d1462 1804 origin_incref(suspect);
0064053b 1805 parse_commit(commit);
85af7929
JH
1806 if (reverse ||
1807 (!(commit->object.flags & UNINTERESTING) &&
1808 !(revs->max_age != -1 && commit->date < revs->max_age)))
717d1462
LT
1809 pass_blame(sb, suspect, opt);
1810 else {
1811 commit->object.flags |= UNINTERESTING;
1812 if (commit->object.parsed)
1813 mark_parents_uninteresting(commit);
1814 }
1815 /* treat root commit as boundary */
1816 if (!commit->parents && !show_root)
1817 commit->object.flags |= UNINTERESTING;
1818
1819 /* Take responsibility for the remaining entries */
7e6ac6e4
DK
1820 ent = suspect->suspects;
1821 if (ent) {
1822 suspect->guilty = 1;
1823 for (;;) {
1824 struct blame_entry *next = ent->next;
aba37f49 1825 found_guilty_entry(ent, &pi);
7e6ac6e4
DK
1826 if (next) {
1827 ent = next;
1828 continue;
1829 }
1830 ent->next = sb->ent;
1831 sb->ent = suspect->suspects;
1832 suspect->suspects = NULL;
1833 break;
1834 }
1835 }
717d1462
LT
1836 origin_decref(suspect);
1837
1838 if (DEBUG) /* sanity */
1839 sanity_check_refcnt(sb);
1840 }
aba37f49
ECA
1841
1842 stop_progress(&pi.progress);
717d1462
LT
1843}
1844
1845static const char *format_time(unsigned long time, const char *tz_str,
1846 int show_raw_time)
1847{
bccce0f8 1848 static struct strbuf time_buf = STRBUF_INIT;
717d1462 1849
bccce0f8 1850 strbuf_reset(&time_buf);
717d1462 1851 if (show_raw_time) {
bccce0f8 1852 strbuf_addf(&time_buf, "%lu %s", time, tz_str);
717d1462 1853 }
31653c1a 1854 else {
ac39b277 1855 const char *time_str;
bccce0f8 1856 size_t time_width;
ac39b277 1857 int tz;
31653c1a 1858 tz = atoi(tz_str);
a5481a6c 1859 time_str = show_date(time, tz, &blame_date_mode);
bccce0f8
JX
1860 strbuf_addstr(&time_buf, time_str);
1861 /*
1862 * Add space paddings to time_buf to display a fixed width
1863 * string, and use time_width for display width calibration.
1864 */
1865 for (time_width = utf8_strwidth(time_str);
1866 time_width < blame_date_width;
1867 time_width++)
1868 strbuf_addch(&time_buf, ' ');
31653c1a 1869 }
bccce0f8 1870 return time_buf.buf;
717d1462
LT
1871}
1872
cee7f245
JH
1873#define OUTPUT_ANNOTATE_COMPAT 001
1874#define OUTPUT_LONG_OBJECT_NAME 002
1875#define OUTPUT_RAW_TIMESTAMP 004
1876#define OUTPUT_PORCELAIN 010
1877#define OUTPUT_SHOW_NAME 020
1878#define OUTPUT_SHOW_NUMBER 040
5ff62c30 1879#define OUTPUT_SHOW_SCORE 0100
093dc5be 1880#define OUTPUT_NO_AUTHOR 0200
1b8cdce9 1881#define OUTPUT_SHOW_EMAIL 0400
ed747dd5 1882#define OUTPUT_LINE_PORCELAIN 01000
cee7f245 1883
e86226e3
JK
1884static void emit_porcelain_details(struct origin *suspect, int repeat)
1885{
1886 if (emit_one_suspect_detail(suspect, repeat) ||
1887 (suspect->commit->object.flags & MORE_THAN_ONE_PATH))
1888 write_filename_info(suspect->path);
1889}
1890
1891static void emit_porcelain(struct scoreboard *sb, struct blame_entry *ent,
1892 int opt)
cee7f245 1893{
ed747dd5 1894 int repeat = opt & OUTPUT_LINE_PORCELAIN;
cee7f245
JH
1895 int cnt;
1896 const char *cp;
1897 struct origin *suspect = ent->suspect;
d59f765a 1898 char hex[GIT_SHA1_HEXSZ + 1];
cee7f245 1899
f2fd0760 1900 sha1_to_hex_r(hex, suspect->commit->object.oid.hash);
7e6ac6e4 1901 printf("%s %d %d %d\n",
cee7f245 1902 hex,
cee7f245
JH
1903 ent->s_lno + 1,
1904 ent->lno + 1,
1905 ent->num_lines);
ed747dd5 1906 emit_porcelain_details(suspect, repeat);
cee7f245
JH
1907
1908 cp = nth_line(sb, ent->lno);
1909 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1910 char ch;
ed747dd5 1911 if (cnt) {
cee7f245
JH
1912 printf("%s %d %d\n", hex,
1913 ent->s_lno + 1 + cnt,
1914 ent->lno + 1 + cnt);
ed747dd5
JK
1915 if (repeat)
1916 emit_porcelain_details(suspect, 1);
1917 }
cee7f245
JH
1918 putchar('\t');
1919 do {
1920 ch = *cp++;
1921 putchar(ch);
1922 } while (ch != '\n' &&
1923 cp < sb->final_buf + sb->final_buf_size);
1924 }
a5ca8367
JS
1925
1926 if (sb->final_buf_size && cp[-1] != '\n')
1927 putchar('\n');
cee7f245
JH
1928}
1929
1930static void emit_other(struct scoreboard *sb, struct blame_entry *ent, int opt)
1931{
1932 int cnt;
1933 const char *cp;
1934 struct origin *suspect = ent->suspect;
1935 struct commit_info ci;
d59f765a 1936 char hex[GIT_SHA1_HEXSZ + 1];
cee7f245
JH
1937 int show_raw_time = !!(opt & OUTPUT_RAW_TIMESTAMP);
1938
1939 get_commit_info(suspect->commit, &ci, 1);
f2fd0760 1940 sha1_to_hex_r(hex, suspect->commit->object.oid.hash);
cee7f245
JH
1941
1942 cp = nth_line(sb, ent->lno);
1943 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1944 char ch;
84393bfd 1945 int length = (opt & OUTPUT_LONG_OBJECT_NAME) ? 40 : abbrev;
b11121d9
JH
1946
1947 if (suspect->commit->object.flags & UNINTERESTING) {
e68989a7
JH
1948 if (blank_boundary)
1949 memset(hex, ' ', length);
7ceacdff 1950 else if (!(opt & OUTPUT_ANNOTATE_COMPAT)) {
4c10a5ca
JH
1951 length--;
1952 putchar('^');
1953 }
b11121d9 1954 }
cee7f245 1955
b11121d9 1956 printf("%.*s", length, hex);
1b8cdce9
KB
1957 if (opt & OUTPUT_ANNOTATE_COMPAT) {
1958 const char *name;
1959 if (opt & OUTPUT_SHOW_EMAIL)
ea02ffa3 1960 name = ci.author_mail.buf;
1b8cdce9 1961 else
ea02ffa3 1962 name = ci.author.buf;
1b8cdce9 1963 printf("\t(%10s\t%10s\t%d)", name,
ea02ffa3 1964 format_time(ci.author_time, ci.author_tz.buf,
cee7f245
JH
1965 show_raw_time),
1966 ent->lno + 1 + cnt);
1b8cdce9 1967 } else {
5ff62c30 1968 if (opt & OUTPUT_SHOW_SCORE)
54a4c617
JH
1969 printf(" %*d %02d",
1970 max_score_digits, ent->score,
1971 ent->suspect->refcnt);
cee7f245
JH
1972 if (opt & OUTPUT_SHOW_NAME)
1973 printf(" %-*.*s", longest_file, longest_file,
1974 suspect->path);
1975 if (opt & OUTPUT_SHOW_NUMBER)
1976 printf(" %*d", max_orig_digits,
1977 ent->s_lno + 1 + cnt);
093dc5be 1978
ffaf9cc0 1979 if (!(opt & OUTPUT_NO_AUTHOR)) {
1b8cdce9
KB
1980 const char *name;
1981 int pad;
1982 if (opt & OUTPUT_SHOW_EMAIL)
ea02ffa3 1983 name = ci.author_mail.buf;
1b8cdce9 1984 else
ea02ffa3 1985 name = ci.author.buf;
1b8cdce9 1986 pad = longest_author - utf8_strwidth(name);
ffaf9cc0 1987 printf(" (%s%*s %10s",
1b8cdce9 1988 name, pad, "",
093dc5be 1989 format_time(ci.author_time,
ea02ffa3 1990 ci.author_tz.buf,
093dc5be 1991 show_raw_time));
ffaf9cc0 1992 }
093dc5be 1993 printf(" %*d) ",
cee7f245
JH
1994 max_digits, ent->lno + 1 + cnt);
1995 }
1996 do {
1997 ch = *cp++;
1998 putchar(ch);
1999 } while (ch != '\n' &&
2000 cp < sb->final_buf + sb->final_buf_size);
2001 }
a5ca8367
JS
2002
2003 if (sb->final_buf_size && cp[-1] != '\n')
2004 putchar('\n');
ea02ffa3
AP
2005
2006 commit_info_destroy(&ci);
cee7f245
JH
2007}
2008
2009static void output(struct scoreboard *sb, int option)
2010{
2011 struct blame_entry *ent;
2012
2013 if (option & OUTPUT_PORCELAIN) {
2014 for (ent = sb->ent; ent; ent = ent->next) {
7e6ac6e4
DK
2015 int count = 0;
2016 struct origin *suspect;
2017 struct commit *commit = ent->suspect->commit;
cee7f245
JH
2018 if (commit->object.flags & MORE_THAN_ONE_PATH)
2019 continue;
7e6ac6e4
DK
2020 for (suspect = commit->util; suspect; suspect = suspect->next) {
2021 if (suspect->guilty && count++) {
2022 commit->object.flags |= MORE_THAN_ONE_PATH;
2023 break;
2024 }
cee7f245
JH
2025 }
2026 }
2027 }
2028
2029 for (ent = sb->ent; ent; ent = ent->next) {
2030 if (option & OUTPUT_PORCELAIN)
e86226e3 2031 emit_porcelain(sb, ent, option);
5ff62c30 2032 else {
cee7f245 2033 emit_other(sb, ent, option);
5ff62c30 2034 }
cee7f245
JH
2035 }
2036}
2037
29aa0b20
RS
2038static const char *get_next_line(const char *start, const char *end)
2039{
2040 const char *nl = memchr(start, '\n', end - start);
60d85e11 2041 return nl ? nl + 1 : end;
29aa0b20
RS
2042}
2043
1732a1fd
JH
2044/*
2045 * To allow quick access to the contents of nth line in the
2046 * final image, prepare an index in the scoreboard.
2047 */
cee7f245
JH
2048static int prepare_lines(struct scoreboard *sb)
2049{
2050 const char *buf = sb->final_buf;
2051 unsigned long len = sb->final_buf_size;
352bbbd9
DK
2052 const char *end = buf + len;
2053 const char *p;
2054 int *lineno;
60d85e11 2055 int num = 0;
352bbbd9 2056
60d85e11 2057 for (p = buf; p < end; p = get_next_line(p, end))
29aa0b20 2058 num++;
352bbbd9 2059
b32fa95f
JK
2060 ALLOC_ARRAY(sb->lineno, num + 1);
2061 lineno = sb->lineno;
352bbbd9 2062
60d85e11 2063 for (p = buf; p < end; p = get_next_line(p, end))
29aa0b20 2064 *lineno++ = p - buf;
352bbbd9 2065
60d85e11 2066 *lineno = len;
352bbbd9 2067
60d85e11 2068 sb->num_lines = num;
cee7f245
JH
2069 return sb->num_lines;
2070}
2071
1732a1fd
JH
2072/*
2073 * Add phony grafts for use with -S; this is primarily to
34baebce 2074 * support git's cvsserver that wants to give a linear history
1732a1fd
JH
2075 * to its clients.
2076 */
cee7f245
JH
2077static int read_ancestry(const char *graft_file)
2078{
2079 FILE *fp = fopen(graft_file, "r");
e228c173 2080 struct strbuf buf = STRBUF_INIT;
cee7f245
JH
2081 if (!fp)
2082 return -1;
e228c173 2083 while (!strbuf_getwholeline(&buf, fp, '\n')) {
cee7f245 2084 /* The format is just "Commit Parent1 Parent2 ...\n" */
e228c173 2085 struct commit_graft *graft = read_graft_line(buf.buf, buf.len);
8eaf7986
JH
2086 if (graft)
2087 register_commit_graft(graft, 0);
cee7f245
JH
2088 }
2089 fclose(fp);
e228c173 2090 strbuf_release(&buf);
cee7f245
JH
2091 return 0;
2092}
2093
b31272f7
JH
2094static int update_auto_abbrev(int auto_abbrev, struct origin *suspect)
2095{
ed1c9977 2096 const char *uniq = find_unique_abbrev(suspect->commit->object.oid.hash,
b31272f7
JH
2097 auto_abbrev);
2098 int len = strlen(uniq);
2099 if (auto_abbrev < len)
2100 return len;
2101 return auto_abbrev;
2102}
2103
1732a1fd
JH
2104/*
2105 * How many columns do we need to show line numbers, authors,
2106 * and filenames?
2107 */
cee7f245
JH
2108static void find_alignment(struct scoreboard *sb, int *option)
2109{
2110 int longest_src_lines = 0;
2111 int longest_dst_lines = 0;
5ff62c30 2112 unsigned largest_score = 0;
cee7f245 2113 struct blame_entry *e;
b31272f7
JH
2114 int compute_auto_abbrev = (abbrev < 0);
2115 int auto_abbrev = default_abbrev;
cee7f245
JH
2116
2117 for (e = sb->ent; e; e = e->next) {
2118 struct origin *suspect = e->suspect;
cee7f245
JH
2119 int num;
2120
b31272f7
JH
2121 if (compute_auto_abbrev)
2122 auto_abbrev = update_auto_abbrev(auto_abbrev, suspect);
ab3bb800
JH
2123 if (strcmp(suspect->path, sb->path))
2124 *option |= OUTPUT_SHOW_NAME;
2125 num = strlen(suspect->path);
2126 if (longest_file < num)
2127 longest_file = num;
cee7f245 2128 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
e6005927 2129 struct commit_info ci;
cee7f245
JH
2130 suspect->commit->object.flags |= METAINFO_SHOWN;
2131 get_commit_info(suspect->commit, &ci, 1);
1b8cdce9 2132 if (*option & OUTPUT_SHOW_EMAIL)
ea02ffa3 2133 num = utf8_strwidth(ci.author_mail.buf);
1b8cdce9 2134 else
ea02ffa3 2135 num = utf8_strwidth(ci.author.buf);
cee7f245
JH
2136 if (longest_author < num)
2137 longest_author = num;
e6005927 2138 commit_info_destroy(&ci);
cee7f245
JH
2139 }
2140 num = e->s_lno + e->num_lines;
2141 if (longest_src_lines < num)
2142 longest_src_lines = num;
2143 num = e->lno + e->num_lines;
2144 if (longest_dst_lines < num)
2145 longest_dst_lines = num;
5ff62c30
JH
2146 if (largest_score < ent_score(sb, e))
2147 largest_score = ent_score(sb, e);
cee7f245 2148 }
ec7ff5ba
ZJS
2149 max_orig_digits = decimal_width(longest_src_lines);
2150 max_digits = decimal_width(longest_dst_lines);
2151 max_score_digits = decimal_width(largest_score);
b31272f7
JH
2152
2153 if (compute_auto_abbrev)
2154 /* one more abbrev length is needed for the boundary commit */
2155 abbrev = auto_abbrev + 1;
cee7f245
JH
2156}
2157
1732a1fd
JH
2158/*
2159 * For debugging -- origin is refcounted, and this asserts that
2160 * we do not underflow.
2161 */
54a4c617
JH
2162static void sanity_check_refcnt(struct scoreboard *sb)
2163{
2164 int baa = 0;
2165 struct blame_entry *ent;
2166
2167 for (ent = sb->ent; ent; ent = ent->next) {
ae86ad65 2168 /* Nobody should have zero or negative refcnt */
854b97f6
JH
2169 if (ent->suspect->refcnt <= 0) {
2170 fprintf(stderr, "%s in %s has negative refcnt %d\n",
2171 ent->suspect->path,
f2fd0760 2172 oid_to_hex(&ent->suspect->commit->object.oid),
854b97f6 2173 ent->suspect->refcnt);
ae86ad65 2174 baa = 1;
854b97f6 2175 }
ae86ad65 2176 }
54a4c617
JH
2177 if (baa) {
2178 int opt = 0160;
2179 find_alignment(sb, &opt);
2180 output(sb, opt);
854b97f6 2181 die("Baa %d!", baa);
54a4c617
JH
2182 }
2183}
2184
4a0fc95f
JH
2185static unsigned parse_score(const char *arg)
2186{
2187 char *end;
2188 unsigned long score = strtoul(arg, &end, 10);
2189 if (*end)
2190 return 0;
2191 return score;
2192}
2193
20239bae
JK
2194static const char *add_prefix(const char *prefix, const char *path)
2195{
097971f5 2196 return prefix_path(prefix, prefix ? strlen(prefix) : 0, path);
20239bae
JK
2197}
2198
ef90d6d4 2199static int git_blame_config(const char *var, const char *value, void *cb)
4c10a5ca
JH
2200{
2201 if (!strcmp(var, "blame.showroot")) {
2202 show_root = git_config_bool(var, value);
2203 return 0;
2204 }
2205 if (!strcmp(var, "blame.blankboundary")) {
2206 blank_boundary = git_config_bool(var, value);
2207 return 0;
2208 }
8b504db3
QN
2209 if (!strcmp(var, "blame.showemail")) {
2210 int *output_option = cb;
2211 if (git_config_bool(var, value))
2212 *output_option |= OUTPUT_SHOW_EMAIL;
2213 else
2214 *output_option &= ~OUTPUT_SHOW_EMAIL;
2215 return 0;
2216 }
31653c1a
EL
2217 if (!strcmp(var, "blame.date")) {
2218 if (!value)
2219 return config_error_nonbool(var);
a5481a6c 2220 parse_date_format(value, &blame_date_mode);
31653c1a
EL
2221 return 0;
2222 }
3b8a12e8 2223
5b162879
MH
2224 if (git_diff_heuristic_config(var, value, cb) < 0)
2225 return -1;
6680a087 2226 if (userdiff_config(var, value) < 0)
3b8a12e8 2227 return -1;
3b8a12e8 2228
ef90d6d4 2229 return git_default_config(var, value, cb);
4c10a5ca
JH
2230}
2231
9aeaab68 2232static void verify_working_tree_path(struct commit *work_tree, const char *path)
ffcabccf 2233{
9aeaab68 2234 struct commit_list *parents;
ffcabccf 2235
9aeaab68 2236 for (parents = work_tree->parents; parents; parents = parents->next) {
ed1c9977 2237 const unsigned char *commit_sha1 = parents->item->object.oid.hash;
9aeaab68
JH
2238 unsigned char blob_sha1[20];
2239 unsigned mode;
2240
2241 if (!get_tree_entry(commit_sha1, path, blob_sha1, &mode) &&
2242 sha1_object_info(blob_sha1, NULL) == OBJ_BLOB)
2243 return;
2244 }
2245 die("no such path '%s' in HEAD", path);
2246}
2247
2248static struct commit_list **append_parent(struct commit_list **tail, const unsigned char *sha1)
2249{
2250 struct commit *parent;
2251
2252 parent = lookup_commit_reference(sha1);
2253 if (!parent)
2254 die("no such commit %s", sha1_to_hex(sha1));
2255 return &commit_list_insert(parent, tail)->next;
2256}
2257
2258static void append_merge_parents(struct commit_list **tail)
2259{
2260 int merge_head;
9aeaab68
JH
2261 struct strbuf line = STRBUF_INIT;
2262
f932729c 2263 merge_head = open(git_path_merge_head(), O_RDONLY);
9aeaab68
JH
2264 if (merge_head < 0) {
2265 if (errno == ENOENT)
2266 return;
f932729c 2267 die("cannot open '%s' for reading", git_path_merge_head());
9aeaab68
JH
2268 }
2269
2270 while (!strbuf_getwholeline_fd(&line, merge_head, '\n')) {
2271 unsigned char sha1[20];
2272 if (line.len < 40 || get_sha1_hex(line.buf, sha1))
f932729c 2273 die("unknown line in '%s': %s", git_path_merge_head(), line.buf);
9aeaab68
JH
2274 tail = append_parent(tail, sha1);
2275 }
2276 close(merge_head);
2277 strbuf_release(&line);
ffcabccf
JH
2278}
2279
8597ea3a
JK
2280/*
2281 * This isn't as simple as passing sb->buf and sb->len, because we
2282 * want to transfer ownership of the buffer to the commit (so we
2283 * must use detach).
2284 */
2285static void set_commit_buffer_from_strbuf(struct commit *c, struct strbuf *sb)
2286{
2287 size_t len;
2288 void *buf = strbuf_detach(sb, &len);
2289 set_commit_buffer(c, buf, len);
2290}
2291
f6c07d7d
JH
2292/*
2293 * Prepare a dummy commit that represents the work tree (or staged) item.
2294 * Note that annotating work tree item never works in the reverse.
2295 */
3b8a12e8
AB
2296static struct commit *fake_working_tree_commit(struct diff_options *opt,
2297 const char *path,
2298 const char *contents_from)
1cfe7733
JH
2299{
2300 struct commit *commit;
2301 struct origin *origin;
9aeaab68 2302 struct commit_list **parent_tail, *parent;
1cfe7733 2303 unsigned char head_sha1[20];
f285a2d7 2304 struct strbuf buf = STRBUF_INIT;
1cfe7733 2305 const char *ident;
1cfe7733 2306 time_t now;
1cfe7733
JH
2307 int size, len;
2308 struct cache_entry *ce;
2309 unsigned mode;
9aeaab68 2310 struct strbuf msg = STRBUF_INIT;
1cfe7733 2311
a08feb8e 2312 read_cache();
1cfe7733 2313 time(&now);
10322a0a 2314 commit = alloc_commit_node();
1cfe7733
JH
2315 commit->object.parsed = 1;
2316 commit->date = now;
9aeaab68
JH
2317 parent_tail = &commit->parents;
2318
7695d118 2319 if (!resolve_ref_unsafe("HEAD", RESOLVE_REF_READING, head_sha1, NULL))
9aeaab68
JH
2320 die("no such ref: HEAD");
2321
2322 parent_tail = append_parent(parent_tail, head_sha1);
2323 append_merge_parents(parent_tail);
2324 verify_working_tree_path(commit, path);
1cfe7733
JH
2325
2326 origin = make_origin(commit, path);
2327
9aeaab68
JH
2328 ident = fmt_ident("Not Committed Yet", "not.committed.yet", NULL, 0);
2329 strbuf_addstr(&msg, "tree 0000000000000000000000000000000000000000\n");
2330 for (parent = commit->parents; parent; parent = parent->next)
2331 strbuf_addf(&msg, "parent %s\n",
f2fd0760 2332 oid_to_hex(&parent->item->object.oid));
9aeaab68
JH
2333 strbuf_addf(&msg,
2334 "author %s\n"
2335 "committer %s\n\n"
2336 "Version of %s from %s\n",
2337 ident, ident, path,
2338 (!contents_from ? path :
2339 (!strcmp(contents_from, "-") ? "standard input" : contents_from)));
8597ea3a 2340 set_commit_buffer_from_strbuf(commit, &msg);
9aeaab68 2341
1cfe7733
JH
2342 if (!contents_from || strcmp("-", contents_from)) {
2343 struct stat st;
2344 const char *read_from;
8518088f 2345 char *buf_ptr;
3b8a12e8 2346 unsigned long buf_len;
1cfe7733
JH
2347
2348 if (contents_from) {
2349 if (stat(contents_from, &st) < 0)
0721c314 2350 die_errno("Cannot stat '%s'", contents_from);
1cfe7733
JH
2351 read_from = contents_from;
2352 }
2353 else {
2354 if (lstat(path, &st) < 0)
0721c314 2355 die_errno("Cannot lstat '%s'", path);
1cfe7733
JH
2356 read_from = path;
2357 }
1cfe7733 2358 mode = canon_mode(st.st_mode);
3b8a12e8 2359
1cfe7733
JH
2360 switch (st.st_mode & S_IFMT) {
2361 case S_IFREG:
3b8a12e8 2362 if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) &&
e5450100 2363 textconv_object(read_from, mode, null_sha1, 0, &buf_ptr, &buf_len))
8518088f 2364 strbuf_attach(&buf, buf_ptr, buf_len, buf_len + 1);
3b8a12e8 2365 else if (strbuf_read_file(&buf, read_from, st.st_size) != st.st_size)
0721c314 2366 die_errno("cannot open or read '%s'", read_from);
1cfe7733
JH
2367 break;
2368 case S_IFLNK:
edfd45d2 2369 if (strbuf_readlink(&buf, read_from, st.st_size) < 0)
0721c314 2370 die_errno("cannot readlink '%s'", read_from);
1cfe7733
JH
2371 break;
2372 default:
2373 die("unsupported file type %s", read_from);
2374 }
2375 }
2376 else {
2377 /* Reading from stdin */
1cfe7733 2378 mode = 0;
f1696ee3 2379 if (strbuf_read(&buf, 0, 0) < 0)
d824cbba 2380 die_errno("failed to read from stdin");
1cfe7733 2381 }
4bf256d6 2382 convert_to_git(path, buf.buf, buf.len, &buf, 0);
af6eb822
PH
2383 origin->file.ptr = buf.buf;
2384 origin->file.size = buf.len;
2385 pretend_sha1_file(buf.buf, buf.len, OBJ_BLOB, origin->blob_sha1);
1cfe7733
JH
2386
2387 /*
2388 * Read the current index, replace the path entry with
2389 * origin->blob_sha1 without mucking with its mode or type
2390 * bits; we are not going to write this index out -- we just
2391 * want to run "diff-index --cached".
2392 */
2393 discard_cache();
2394 read_cache();
2395
2396 len = strlen(path);
2397 if (!mode) {
2398 int pos = cache_name_pos(path, len);
2399 if (0 <= pos)
7a51ed66 2400 mode = active_cache[pos]->ce_mode;
1cfe7733
JH
2401 else
2402 /* Let's not bother reading from HEAD tree */
2403 mode = S_IFREG | 0644;
2404 }
2405 size = cache_entry_size(len);
2406 ce = xcalloc(1, size);
2407 hashcpy(ce->sha1, origin->blob_sha1);
2408 memcpy(ce->name, path, len);
b60e188c
TG
2409 ce->ce_flags = create_ce_flags(0);
2410 ce->ce_namelen = len;
1cfe7733
JH
2411 ce->ce_mode = create_ce_mode(mode);
2412 add_cache_entry(ce, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
2413
a5400efe 2414 cache_tree_invalidate_path(&the_index, path);
1cfe7733 2415
1cfe7733
JH
2416 return commit;
2417}
2418
7cb5f7c4
JK
2419static struct commit *find_single_final(struct rev_info *revs,
2420 const char **name_p)
f6c07d7d
JH
2421{
2422 int i;
7cb5f7c4
JK
2423 struct commit *found = NULL;
2424 const char *name = NULL;
f6c07d7d 2425
f6c07d7d
JH
2426 for (i = 0; i < revs->pending.nr; i++) {
2427 struct object *obj = revs->pending.objects[i].item;
2428 if (obj->flags & UNINTERESTING)
2429 continue;
31da121f 2430 obj = deref_tag(obj, NULL, 0);
f6c07d7d
JH
2431 if (obj->type != OBJ_COMMIT)
2432 die("Non commit %s?", revs->pending.objects[i].name);
1b0d4000 2433 if (found)
f6c07d7d 2434 die("More than one commit to dig from %s and %s?",
7cb5f7c4
JK
2435 revs->pending.objects[i].name, name);
2436 found = (struct commit *)obj;
2437 name = revs->pending.objects[i].name;
1b0d4000 2438 }
7cb5f7c4
JK
2439 if (name_p)
2440 *name_p = name;
1b0d4000
MK
2441 return found;
2442}
2443
2444static char *prepare_final(struct scoreboard *sb)
2445{
7cb5f7c4
JK
2446 const char *name;
2447 sb->final = find_single_final(sb->revs, &name);
2448 return xstrdup_or_null(name);
f6c07d7d
JH
2449}
2450
a46442f1 2451static char *prepare_initial(struct scoreboard *sb)
85af7929
JH
2452{
2453 int i;
2454 const char *final_commit_name = NULL;
2455 struct rev_info *revs = sb->revs;
2456
2457 /*
2458 * There must be one and only one negative commit, and it must be
2459 * the boundary.
2460 */
2461 for (i = 0; i < revs->pending.nr; i++) {
2462 struct object *obj = revs->pending.objects[i].item;
2463 if (!(obj->flags & UNINTERESTING))
2464 continue;
31da121f 2465 obj = deref_tag(obj, NULL, 0);
85af7929
JH
2466 if (obj->type != OBJ_COMMIT)
2467 die("Non commit %s?", revs->pending.objects[i].name);
2468 if (sb->final)
2469 die("More than one commit to dig down to %s and %s?",
2470 revs->pending.objects[i].name,
2471 final_commit_name);
2472 sb->final = (struct commit *) obj;
2473 final_commit_name = revs->pending.objects[i].name;
2474 }
2475 if (!final_commit_name)
2476 die("No commit to dig down to?");
a46442f1 2477 return xstrdup(final_commit_name);
85af7929
JH
2478}
2479
5817da01
PH
2480static int blame_copy_callback(const struct option *option, const char *arg, int unset)
2481{
2482 int *opt = option->value;
2483
2484 /*
2485 * -C enables copy from removed files;
2486 * -C -C enables copy from existing files, but only
2487 * when blaming a new file;
2488 * -C -C -C enables copy from existing files for
2489 * everybody
2490 */
2491 if (*opt & PICKAXE_BLAME_COPY_HARDER)
2492 *opt |= PICKAXE_BLAME_COPY_HARDEST;
2493 if (*opt & PICKAXE_BLAME_COPY)
2494 *opt |= PICKAXE_BLAME_COPY_HARDER;
2495 *opt |= PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE;
2496
2497 if (arg)
2498 blame_copy_score = parse_score(arg);
2499 return 0;
2500}
2501
2502static int blame_move_callback(const struct option *option, const char *arg, int unset)
2503{
2504 int *opt = option->value;
2505
2506 *opt |= PICKAXE_BLAME_MOVE;
2507
2508 if (arg)
2509 blame_move_score = parse_score(arg);
2510 return 0;
2511}
2512
acca687f 2513int cmd_blame(int argc, const char **argv, const char *prefix)
cee7f245
JH
2514{
2515 struct rev_info revs;
2516 const char *path;
2517 struct scoreboard sb;
2518 struct origin *o;
58dbfa2e
ES
2519 struct blame_entry *ent = NULL;
2520 long dashdash_pos, lno;
a46442f1 2521 char *final_commit_name = NULL;
21666f1a 2522 enum object_type type;
700fd28e 2523 struct commit *final_commit = NULL;
5817da01 2524
64093fc0
JK
2525 struct string_list range_list = STRING_LIST_INIT_NODUP;
2526 int output_option = 0, opt = 0;
2527 int show_stats = 0;
2528 const char *revs_file = NULL;
2529 const char *contents_from = NULL;
2530 const struct option options[] = {
d5d09d47
SB
2531 OPT_BOOL(0, "incremental", &incremental, N_("Show blame entries as we find them, incrementally")),
2532 OPT_BOOL('b', NULL, &blank_boundary, N_("Show blank SHA-1 for boundary commits (Default: off)")),
2533 OPT_BOOL(0, "root", &show_root, N_("Do not treat root commits as boundaries (Default: off)")),
2534 OPT_BOOL(0, "show-stats", &show_stats, N_("Show work cost statistics")),
aba37f49 2535 OPT_BOOL(0, "progress", &show_progress, N_("Force progress reporting")),
efd2a8bd
NTND
2536 OPT_BIT(0, "score-debug", &output_option, N_("Show output score for blame entries"), OUTPUT_SHOW_SCORE),
2537 OPT_BIT('f', "show-name", &output_option, N_("Show original filename (Default: auto)"), OUTPUT_SHOW_NAME),
2538 OPT_BIT('n', "show-number", &output_option, N_("Show original linenumber (Default: off)"), OUTPUT_SHOW_NUMBER),
2539 OPT_BIT('p', "porcelain", &output_option, N_("Show in a format designed for machine consumption"), OUTPUT_PORCELAIN),
2540 OPT_BIT(0, "line-porcelain", &output_option, N_("Show porcelain format with per-line commit information"), OUTPUT_PORCELAIN|OUTPUT_LINE_PORCELAIN),
2541 OPT_BIT('c', NULL, &output_option, N_("Use the same output mode as git-annotate (Default: off)"), OUTPUT_ANNOTATE_COMPAT),
2542 OPT_BIT('t', NULL, &output_option, N_("Show raw timestamp (Default: off)"), OUTPUT_RAW_TIMESTAMP),
2543 OPT_BIT('l', NULL, &output_option, N_("Show long commit SHA1 (Default: off)"), OUTPUT_LONG_OBJECT_NAME),
2544 OPT_BIT('s', NULL, &output_option, N_("Suppress author name and timestamp (Default: off)"), OUTPUT_NO_AUTHOR),
2545 OPT_BIT('e', "show-email", &output_option, N_("Show author email instead of name (Default: off)"), OUTPUT_SHOW_EMAIL),
2546 OPT_BIT('w', NULL, &xdl_opts, N_("Ignore whitespace differences"), XDF_IGNORE_WHITESPACE),
5b162879
MH
2547
2548 /*
2549 * The following two options are parsed by parse_revision_opt()
2550 * and are only included here to get included in the "-h"
2551 * output:
2552 */
2553 { OPTION_LOWLEVEL_CALLBACK, 0, "indent-heuristic", NULL, NULL, N_("Use an experimental indent-based heuristic to improve diffs"), PARSE_OPT_NOARG, parse_opt_unknown_cb },
2554 { OPTION_LOWLEVEL_CALLBACK, 0, "compaction-heuristic", NULL, NULL, N_("Use an experimental blank-line-based heuristic to improve diffs"), PARSE_OPT_NOARG, parse_opt_unknown_cb },
2555
efd2a8bd
NTND
2556 OPT_BIT(0, "minimal", &xdl_opts, N_("Spend extra cycles to find better match"), XDF_NEED_MINIMAL),
2557 OPT_STRING('S', NULL, &revs_file, N_("file"), N_("Use revisions from <file> instead of calling git-rev-list")),
2558 OPT_STRING(0, "contents", &contents_from, N_("file"), N_("Use <file>'s contents as the final image")),
2559 { OPTION_CALLBACK, 'C', NULL, &opt, N_("score"), N_("Find line copies within and across files"), PARSE_OPT_OPTARG, blame_copy_callback },
2560 { OPTION_CALLBACK, 'M', NULL, &opt, N_("score"), N_("Find line movements within and across files"), PARSE_OPT_OPTARG, blame_move_callback },
58dbfa2e 2561 OPT_STRING_LIST('L', NULL, &range_list, N_("n,m"), N_("Process only line range n,m, counting from 1")),
84393bfd 2562 OPT__ABBREV(&abbrev),
5817da01
PH
2563 OPT_END()
2564 };
2565
2566 struct parse_opt_ctx_t ctx;
7ceacdff 2567 int cmd_is_annotate = !strcmp(argv[0], "annotate");
58dbfa2e
ES
2568 struct range_set ranges;
2569 unsigned int range_i;
52f4d126 2570 long anchor;
e68989a7 2571
8b504db3 2572 git_config(git_blame_config, &output_option);
5817da01 2573 init_revisions(&revs, NULL);
31653c1a 2574 revs.date_mode = blame_date_mode;
3b8a12e8 2575 DIFF_OPT_SET(&revs.diffopt, ALLOW_TEXTCONV);
3d1aa566 2576 DIFF_OPT_SET(&revs.diffopt, FOLLOW_RENAMES);
31653c1a 2577
612702e8 2578 save_commit_buffer = 0;
3f8d5204 2579 dashdash_pos = 0;
aba37f49 2580 show_progress = -1;
612702e8 2581
9ca1169f
SB
2582 parse_options_start(&ctx, argc, argv, prefix, options,
2583 PARSE_OPT_KEEP_DASHDASH | PARSE_OPT_KEEP_ARGV0);
5817da01 2584 for (;;) {
5817da01
PH
2585 switch (parse_options_step(&ctx, options, blame_opt_usage)) {
2586 case PARSE_OPT_HELP:
2587 exit(129);
2588 case PARSE_OPT_DONE:
3f8d5204
PH
2589 if (ctx.argv[0])
2590 dashdash_pos = ctx.cpidx;
5817da01
PH
2591 goto parse_done;
2592 }
2593
2594 if (!strcmp(ctx.argv[0], "--reverse")) {
2595 ctx.argv[0] = "--children";
2596 reverse = 1;
2597 }
6b61ec05 2598 parse_revision_opt(&revs, &ctx, options, blame_opt_usage);
5817da01
PH
2599 }
2600parse_done:
3d1aa566 2601 no_whole_file_rename = !DIFF_OPT_TST(&revs.diffopt, FOLLOW_RENAMES);
5b162879 2602 xdl_opts |= revs.diffopt.xdl_opts & (XDF_COMPACTION_HEURISTIC | XDF_INDENT_HEURISTIC);
3d1aa566 2603 DIFF_OPT_CLR(&revs.diffopt, FOLLOW_RENAMES);
5817da01
PH
2604 argc = parse_options_end(&ctx);
2605
aba37f49
ECA
2606 if (incremental || (output_option & OUTPUT_PORCELAIN)) {
2607 if (show_progress > 0)
2608 die("--progress can't be used with --incremental or porcelain formats");
2609 show_progress = 0;
2610 } else if (show_progress < 0)
2611 show_progress = isatty(2);
2612
b31272f7
JH
2613 if (0 < abbrev)
2614 /* one more abbrev length is needed for the boundary commit */
2615 abbrev++;
84393bfd 2616
aa9ea77d 2617 if (revs_file && read_ancestry(revs_file))
d824cbba 2618 die_errno("reading graft file '%s' failed", revs_file);
aa9ea77d 2619
31653c1a 2620 if (cmd_is_annotate) {
7ceacdff 2621 output_option |= OUTPUT_ANNOTATE_COMPAT;
a5481a6c 2622 blame_date_mode.type = DATE_ISO8601;
31653c1a
EL
2623 } else {
2624 blame_date_mode = revs.date_mode;
2625 }
2626
2627 /* The maximum width used to show the dates */
a5481a6c 2628 switch (blame_date_mode.type) {
31653c1a
EL
2629 case DATE_RFC2822:
2630 blame_date_width = sizeof("Thu, 19 Oct 2006 16:00:04 -0700");
2631 break;
466fb674
BB
2632 case DATE_ISO8601_STRICT:
2633 blame_date_width = sizeof("2006-10-19T16:00:04-07:00");
2634 break;
31653c1a
EL
2635 case DATE_ISO8601:
2636 blame_date_width = sizeof("2006-10-19 16:00:04 -0700");
2637 break;
2638 case DATE_RAW:
2639 blame_date_width = sizeof("1161298804 -0700");
2640 break;
2641 case DATE_SHORT:
2642 blame_date_width = sizeof("2006-10-19");
2643 break;
2644 case DATE_RELATIVE:
dd75553b
JX
2645 /* TRANSLATORS: This string is used to tell us the maximum
2646 display width for a relative timestamp in "git blame"
2647 output. For C locale, "4 years, 11 months ago", which
2648 takes 22 places, is the longest among various forms of
2649 relative timestamps, but your language may need more or
2650 fewer display columns. */
2651 blame_date_width = utf8_strwidth(_("4 years, 11 months ago")) + 1; /* add the null */
2652 break;
31653c1a
EL
2653 case DATE_NORMAL:
2654 blame_date_width = sizeof("Thu Oct 19 16:00:04 2006 -0700");
2655 break;
aa1462cc
JK
2656 case DATE_STRFTIME:
2657 blame_date_width = strlen(show_date(0, 0, &blame_date_mode)) + 1; /* add the null */
2658 break;
31653c1a
EL
2659 }
2660 blame_date_width -= 1; /* strip the null */
7ceacdff 2661
b3123f98
JH
2662 if (DIFF_OPT_TST(&revs.diffopt, FIND_COPIES_HARDER))
2663 opt |= (PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE |
2664 PICKAXE_BLAME_COPY_HARDER);
2665
4a0fc95f
JH
2666 if (!blame_move_score)
2667 blame_move_score = BLAME_DEFAULT_MOVE_SCORE;
2668 if (!blame_copy_score)
2669 blame_copy_score = BLAME_DEFAULT_COPY_SCORE;
2670
1732a1fd
JH
2671 /*
2672 * We have collected options unknown to us in argv[1..unk]
cee7f245 2673 * which are to be passed to revision machinery if we are
3dff5379 2674 * going to do the "bottom" processing.
cee7f245
JH
2675 *
2676 * The remaining are:
2677 *
22e5e58a 2678 * (1) if dashdash_pos != 0, it is either
3f8d5204
PH
2679 * "blame [revisions] -- <path>" or
2680 * "blame -- <path> <rev>"
cee7f245 2681 *
22e5e58a 2682 * (2) otherwise, it is one of the two:
3f8d5204
PH
2683 * "blame [revisions] <path>"
2684 * "blame <path> <rev>"
cee7f245 2685 *
3f8d5204
PH
2686 * Note that we must strip out <path> from the arguments: we do not
2687 * want the path pruning but we may want "bottom" processing.
cee7f245 2688 */
3f8d5204
PH
2689 if (dashdash_pos) {
2690 switch (argc - dashdash_pos - 1) {
2691 case 2: /* (1b) */
2692 if (argc != 4)
5817da01 2693 usage_with_options(blame_opt_usage, options);
3f8d5204
PH
2694 /* reorder for the new way: <rev> -- <path> */
2695 argv[1] = argv[3];
2696 argv[3] = argv[2];
2697 argv[2] = "--";
2698 /* FALLTHROUGH */
2699 case 1: /* (1a) */
2700 path = add_prefix(prefix, argv[--argc]);
2701 argv[argc] = NULL;
2702 break;
2703 default:
2704 usage_with_options(blame_opt_usage, options);
cee7f245 2705 }
3f8d5204
PH
2706 } else {
2707 if (argc < 2)
5817da01 2708 usage_with_options(blame_opt_usage, options);
3f8d5204 2709 path = add_prefix(prefix, argv[argc - 1]);
dbe44faa 2710 if (argc == 3 && !file_exists(path)) { /* (2b) */
3f8d5204
PH
2711 path = add_prefix(prefix, argv[1]);
2712 argv[1] = argv[2];
cee7f245 2713 }
3f8d5204 2714 argv[argc - 1] = "--";
cee7f245 2715
3f8d5204 2716 setup_work_tree();
dbe44faa 2717 if (!file_exists(path))
d824cbba 2718 die_errno("cannot stat path '%s'", path);
cee7f245
JH
2719 }
2720
8b3dce56 2721 revs.disable_stdin = 1;
3f8d5204 2722 setup_revisions(argc, argv, &revs, NULL);
cee7f245
JH
2723 memset(&sb, 0, sizeof(sb));
2724
85af7929 2725 sb.revs = &revs;
7e6ac6e4 2726 if (!reverse) {
85af7929 2727 final_commit_name = prepare_final(&sb);
7e6ac6e4
DK
2728 sb.commits.compare = compare_commits_by_commit_date;
2729 }
85af7929 2730 else if (contents_from)
95261974 2731 die("--contents and --reverse do not blend well.");
7e6ac6e4 2732 else {
85af7929 2733 final_commit_name = prepare_initial(&sb);
7e6ac6e4 2734 sb.commits.compare = compare_commits_by_reverse_commit_date;
700fd28e
MK
2735 if (revs.first_parent_only)
2736 revs.children.name = NULL;
7e6ac6e4 2737 }
cee7f245
JH
2738
2739 if (!sb.final) {
1732a1fd
JH
2740 /*
2741 * "--not A B -- path" without anything positive;
1cfe7733
JH
2742 * do not default to HEAD, but use the working tree
2743 * or "--contents".
1732a1fd 2744 */
1981820b 2745 setup_work_tree();
3b8a12e8
AB
2746 sb.final = fake_working_tree_commit(&sb.revs->diffopt,
2747 path, contents_from);
1cfe7733 2748 add_pending_object(&revs, &(sb.final->object), ":");
cee7f245 2749 }
1cfe7733
JH
2750 else if (contents_from)
2751 die("Cannot use --contents with final commit object name");
cee7f245 2752
700fd28e 2753 if (reverse && revs.first_parent_only) {
7cb5f7c4
JK
2754 final_commit = find_single_final(sb.revs, NULL);
2755 if (!final_commit)
700fd28e 2756 die("--reverse and --first-parent together require specified latest commit");
700fd28e
MK
2757 }
2758
1732a1fd
JH
2759 /*
2760 * If we have bottom, this will mark the ancestors of the
cee7f245
JH
2761 * bottom commits we would reach while traversing as
2762 * uninteresting.
2763 */
3d51e1b5 2764 if (prepare_revision_walk(&revs))
20108742 2765 die(_("revision walk setup failed"));
cee7f245 2766
700fd28e
MK
2767 if (reverse && revs.first_parent_only) {
2768 struct commit *c = final_commit;
2769
2770 sb.revs->children.name = "children";
2771 while (c->parents &&
f2fd0760 2772 oidcmp(&c->object.oid, &sb.final->object.oid)) {
700fd28e
MK
2773 struct commit_list *l = xcalloc(1, sizeof(*l));
2774
2775 l->item = c;
2776 if (add_decoration(&sb.revs->children,
2777 &c->parents->item->object, l))
2778 die("BUG: not unique item in first-parent chain");
2779 c = c->parents->item;
2780 }
2781
f2fd0760 2782 if (oidcmp(&c->object.oid, &sb.final->object.oid))
700fd28e
MK
2783 die("--reverse --first-parent together require range along first-parent chain");
2784 }
2785
f2fd0760 2786 if (is_null_oid(&sb.final->object.oid)) {
1cfe7733 2787 o = sb.final->util;
5c0b13f8 2788 sb.final_buf = xmemdupz(o->file.ptr, o->file.size);
1cfe7733
JH
2789 sb.final_buf_size = o->file.size;
2790 }
2791 else {
2792 o = get_origin(&sb, sb.final, path);
90064710 2793 if (fill_blob_sha1_and_mode(o))
1cfe7733 2794 die("no such path %s in %s", path, final_commit_name);
cee7f245 2795
3b8a12e8 2796 if (DIFF_OPT_TST(&sb.revs->diffopt, ALLOW_TEXTCONV) &&
e5450100 2797 textconv_object(path, o->mode, o->blob_sha1, 1, (char **) &sb.final_buf,
3b8a12e8
AB
2798 &sb.final_buf_size))
2799 ;
2800 else
2801 sb.final_buf = read_sha1_file(o->blob_sha1, &type,
2802 &sb.final_buf_size);
2803
ab43e495
JH
2804 if (!sb.final_buf)
2805 die("Cannot read blob %s for path %s",
2806 sha1_to_hex(o->blob_sha1),
2807 path);
1cfe7733 2808 }
c2e525d9 2809 num_read_blob++;
cee7f245
JH
2810 lno = prepare_lines(&sb);
2811
58dbfa2e
ES
2812 if (lno && !range_list.nr)
2813 string_list_append(&range_list, xstrdup("1"));
2814
52f4d126 2815 anchor = 1;
58dbfa2e
ES
2816 range_set_init(&ranges, range_list.nr);
2817 for (range_i = 0; range_i < range_list.nr; ++range_i) {
2818 long bottom, top;
2819 if (parse_range_arg(range_list.items[range_i].string,
52f4d126 2820 nth_line_cb, &sb, lno, anchor,
58dbfa2e
ES
2821 &bottom, &top, sb.path))
2822 usage(blame_usage);
2823 if (lno < top || ((lno || bottom) && lno < bottom))
2824 die("file %s has only %lu lines", path, lno);
2825 if (bottom < 1)
2826 bottom = 1;
2827 if (top < 1)
2828 top = lno;
2829 bottom--;
2830 range_set_append_unsafe(&ranges, bottom, top);
52f4d126 2831 anchor = top + 1;
58dbfa2e
ES
2832 }
2833 sort_and_merge_range_set(&ranges);
2834
2835 for (range_i = ranges.nr; range_i > 0; --range_i) {
2836 const struct range *r = &ranges.ranges[range_i - 1];
2837 long bottom = r->start;
2838 long top = r->end;
2839 struct blame_entry *next = ent;
2840 ent = xcalloc(1, sizeof(*ent));
2841 ent->lno = bottom;
2842 ent->num_lines = top - bottom;
2843 ent->suspect = o;
2844 ent->s_lno = bottom;
2845 ent->next = next;
58dbfa2e
ES
2846 origin_incref(o);
2847 }
7e6ac6e4
DK
2848
2849 o->suspects = ent;
2850 prio_queue_put(&sb.commits, o->commit);
2851
58dbfa2e
ES
2852 origin_decref(o);
2853
2854 range_set_release(&ranges);
2855 string_list_clear(&range_list, 0);
cee7f245 2856
7e6ac6e4 2857 sb.ent = NULL;
cee7f245
JH
2858 sb.path = path;
2859
d551a488 2860 read_mailmap(&mailmap, NULL);
f9567384 2861
aba37f49
ECA
2862 assign_blame(&sb, opt);
2863
b92565dc
MH
2864 if (!incremental)
2865 setup_pager();
2866
a46442f1
LF
2867 free(final_commit_name);
2868
717d1462
LT
2869 if (incremental)
2870 return 0;
2871
7e6ac6e4
DK
2872 sb.ent = blame_sort(sb.ent, compare_blame_final);
2873
cee7f245
JH
2874 coalesce(&sb);
2875
2876 if (!(output_option & OUTPUT_PORCELAIN))
2877 find_alignment(&sb, &output_option);
2878
2879 output(&sb, output_option);
2880 free((void *)sb.final_buf);
2881 for (ent = sb.ent; ent; ) {
2882 struct blame_entry *e = ent->next;
2883 free(ent);
2884 ent = e;
2885 }
c2e525d9 2886
870b39c1 2887 if (show_stats) {
c2e525d9
JH
2888 printf("num read blob: %d\n", num_read_blob);
2889 printf("num get patch: %d\n", num_get_patch);
2890 printf("num commits: %d\n", num_commits);
2891 }
cee7f245
JH
2892 return 0;
2893}