convert object type handling from a string to a number
[git/git.git] / builtin-blame.c
CommitLineData
cee7f245
JH
1/*
2 * Pickaxe
3 *
4 * Copyright (c) 2006, Junio C Hamano
5 */
6
7#include "cache.h"
8#include "builtin.h"
9#include "blob.h"
10#include "commit.h"
11#include "tag.h"
12#include "tree-walk.h"
13#include "diff.h"
14#include "diffcore.h"
15#include "revision.h"
717d1462 16#include "quote.h"
cee7f245 17#include "xdiff-interface.h"
1cfe7733 18#include "cache-tree.h"
cee7f245 19
acca687f 20static char blame_usage[] =
06e75a72 21"git-blame [-c] [-l] [-t] [-f] [-n] [-p] [-L n,m] [-S <revs-file>] [-M] [-C] [-C] [--contents <filename>] [--incremental] [commit] [--] file\n"
cee7f245 22" -c, --compatibility Use the same output mode as git-annotate (Default: off)\n"
4c10a5ca 23" -b Show blank SHA-1 for boundary commits (Default: off)\n"
cee7f245 24" -l, --long Show long commit SHA1 (Default: off)\n"
4c10a5ca 25" --root Do not treat root commits as boundaries (Default: off)\n"
cee7f245
JH
26" -t, --time Show raw timestamp (Default: off)\n"
27" -f, --show-name Show original filename (Default: auto)\n"
28" -n, --show-number Show original linenumber (Default: off)\n"
29" -p, --porcelain Show in a format designed for machine consumption\n"
30" -L n,m Process only line range n,m, counting from 1\n"
18abd745 31" -M, -C Find line movements within and across files\n"
717d1462 32" --incremental Show blame entries as we find them, incrementally\n"
1cfe7733 33" --contents file Use <file>'s contents as the final image\n"
cee7f245
JH
34" -S revs-file Use revisions from revs-file instead of calling git-rev-list\n";
35
36static int longest_file;
37static int longest_author;
38static int max_orig_digits;
39static int max_digits;
5ff62c30 40static int max_score_digits;
4c10a5ca
JH
41static int show_root;
42static int blank_boundary;
717d1462 43static int incremental;
e68989a7 44static int cmd_is_annotate;
cee7f245 45
54a4c617
JH
46#ifndef DEBUG
47#define DEBUG 0
48#endif
49
c2e525d9
JH
50/* stats */
51static int num_read_blob;
52static int num_get_patch;
53static int num_commits;
54
d24bba80 55#define PICKAXE_BLAME_MOVE 01
18abd745
JH
56#define PICKAXE_BLAME_COPY 02
57#define PICKAXE_BLAME_COPY_HARDER 04
d24bba80 58
4a0fc95f
JH
59/*
60 * blame for a blame_entry with score lower than these thresholds
61 * is not passed to the parent using move/copy logic.
62 */
63static unsigned blame_move_score;
64static unsigned blame_copy_score;
65#define BLAME_DEFAULT_MOVE_SCORE 20
66#define BLAME_DEFAULT_COPY_SCORE 40
67
cee7f245
JH
68/* bits #0..7 in revision.h, #8..11 used for merge_bases() in commit.c */
69#define METAINFO_SHOWN (1u<<12)
70#define MORE_THAN_ONE_PATH (1u<<13)
71
72/*
54a4c617 73 * One blob in a commit that is being suspected
cee7f245
JH
74 */
75struct origin {
54a4c617 76 int refcnt;
cee7f245 77 struct commit *commit;
c2e525d9 78 mmfile_t file;
cee7f245
JH
79 unsigned char blob_sha1[20];
80 char path[FLEX_ARRAY];
81};
82
1732a1fd
JH
83/*
84 * Given an origin, prepare mmfile_t structure to be used by the
85 * diff machinery
86 */
c2e525d9
JH
87static char *fill_origin_blob(struct origin *o, mmfile_t *file)
88{
89 if (!o->file.ptr) {
21666f1a 90 enum object_type type;
c2e525d9 91 num_read_blob++;
21666f1a 92 file->ptr = read_sha1_file(o->blob_sha1, &type,
c2e525d9
JH
93 (unsigned long *)(&(file->size)));
94 o->file = *file;
95 }
96 else
97 *file = o->file;
98 return file->ptr;
99}
100
1732a1fd
JH
101/*
102 * Origin is refcounted and usually we keep the blob contents to be
103 * reused.
104 */
54a4c617
JH
105static inline struct origin *origin_incref(struct origin *o)
106{
107 if (o)
108 o->refcnt++;
109 return o;
110}
111
112static void origin_decref(struct origin *o)
113{
114 if (o && --o->refcnt <= 0) {
c2e525d9
JH
115 if (o->file.ptr)
116 free(o->file.ptr);
54a4c617
JH
117 memset(o, 0, sizeof(*o));
118 free(o);
119 }
120}
121
1732a1fd
JH
122/*
123 * Each group of lines is described by a blame_entry; it can be split
124 * as we pass blame to the parents. They form a linked list in the
125 * scoreboard structure, sorted by the target line number.
126 */
cee7f245
JH
127struct blame_entry {
128 struct blame_entry *prev;
129 struct blame_entry *next;
130
131 /* the first line of this group in the final image;
132 * internally all line numbers are 0 based.
133 */
134 int lno;
135
136 /* how many lines this group has */
137 int num_lines;
138
139 /* the commit that introduced this group into the final image */
140 struct origin *suspect;
141
142 /* true if the suspect is truly guilty; false while we have not
143 * checked if the group came from one of its parents.
144 */
145 char guilty;
146
147 /* the line number of the first line of this group in the
148 * suspect's file; internally all line numbers are 0 based.
149 */
150 int s_lno;
5ff62c30
JH
151
152 /* how significant this entry is -- cached to avoid
1732a1fd 153 * scanning the lines over and over.
5ff62c30
JH
154 */
155 unsigned score;
cee7f245
JH
156};
157
1732a1fd
JH
158/*
159 * The current state of the blame assignment.
160 */
cee7f245
JH
161struct scoreboard {
162 /* the final commit (i.e. where we started digging from) */
163 struct commit *final;
164
165 const char *path;
166
1732a1fd
JH
167 /*
168 * The contents in the final image.
169 * Used by many functions to obtain contents of the nth line,
170 * indexed with scoreboard.lineno[blame_entry.lno].
cee7f245
JH
171 */
172 const char *final_buf;
173 unsigned long final_buf_size;
174
175 /* linked list of blames */
176 struct blame_entry *ent;
177
612702e8 178 /* look-up a line in the final buffer */
cee7f245
JH
179 int num_lines;
180 int *lineno;
181};
182
46014766
JH
183static int cmp_suspect(struct origin *a, struct origin *b)
184{
185 int cmp = hashcmp(a->commit->object.sha1, b->commit->object.sha1);
186 if (cmp)
187 return cmp;
188 return strcmp(a->path, b->path);
189}
190
0d981c67
JH
191#define cmp_suspect(a, b) ( ((a)==(b)) ? 0 : cmp_suspect(a,b) )
192
54a4c617
JH
193static void sanity_check_refcnt(struct scoreboard *);
194
1732a1fd
JH
195/*
196 * If two blame entries that are next to each other came from
197 * contiguous lines in the same origin (i.e. <commit, path> pair),
198 * merge them together.
199 */
cee7f245
JH
200static void coalesce(struct scoreboard *sb)
201{
202 struct blame_entry *ent, *next;
203
204 for (ent = sb->ent; ent && (next = ent->next); ent = next) {
46014766 205 if (!cmp_suspect(ent->suspect, next->suspect) &&
cee7f245
JH
206 ent->guilty == next->guilty &&
207 ent->s_lno + ent->num_lines == next->s_lno) {
208 ent->num_lines += next->num_lines;
209 ent->next = next->next;
210 if (ent->next)
211 ent->next->prev = ent;
54a4c617 212 origin_decref(next->suspect);
cee7f245 213 free(next);
46014766 214 ent->score = 0;
cee7f245
JH
215 next = ent; /* again */
216 }
217 }
54a4c617
JH
218
219 if (DEBUG) /* sanity */
220 sanity_check_refcnt(sb);
cee7f245
JH
221}
222
1732a1fd
JH
223/*
224 * Given a commit and a path in it, create a new origin structure.
225 * The callers that add blame to the scoreboard should use
226 * get_origin() to obtain shared, refcounted copy instead of calling
227 * this function directly.
228 */
854b97f6
JH
229static struct origin *make_origin(struct commit *commit, const char *path)
230{
231 struct origin *o;
232 o = xcalloc(1, sizeof(*o) + strlen(path) + 1);
233 o->commit = commit;
234 o->refcnt = 1;
235 strcpy(o->path, path);
236 return o;
237}
238
1732a1fd
JH
239/*
240 * Locate an existing origin or create a new one.
241 */
f6c0e191
JH
242static struct origin *get_origin(struct scoreboard *sb,
243 struct commit *commit,
244 const char *path)
cee7f245 245{
f6c0e191 246 struct blame_entry *e;
cee7f245 247
f6c0e191
JH
248 for (e = sb->ent; e; e = e->next) {
249 if (e->suspect->commit == commit &&
250 !strcmp(e->suspect->path, path))
54a4c617 251 return origin_incref(e->suspect);
cee7f245 252 }
854b97f6 253 return make_origin(commit, path);
cee7f245
JH
254}
255
1732a1fd
JH
256/*
257 * Fill the blob_sha1 field of an origin if it hasn't, so that later
258 * call to fill_origin_blob() can use it to locate the data. blob_sha1
259 * for an origin is also used to pass the blame for the entire file to
260 * the parent to detect the case where a child's blob is identical to
261 * that of its parent's.
262 */
f6c0e191
JH
263static int fill_blob_sha1(struct origin *origin)
264{
265 unsigned mode;
f6c0e191
JH
266
267 if (!is_null_sha1(origin->blob_sha1))
268 return 0;
269 if (get_tree_entry(origin->commit->object.sha1,
270 origin->path,
271 origin->blob_sha1, &mode))
272 goto error_out;
21666f1a 273 if (sha1_object_info(origin->blob_sha1, NULL) != OBJ_BLOB)
f6c0e191
JH
274 goto error_out;
275 return 0;
276 error_out:
277 hashclr(origin->blob_sha1);
278 return -1;
279}
280
1732a1fd
JH
281/*
282 * We have an origin -- check if the same path exists in the
283 * parent and return an origin structure to represent it.
284 */
f6c0e191 285static struct origin *find_origin(struct scoreboard *sb,
cee7f245
JH
286 struct commit *parent,
287 struct origin *origin)
288{
289 struct origin *porigin = NULL;
290 struct diff_options diff_opts;
f6c0e191
JH
291 const char *paths[2];
292
0d981c67 293 if (parent->util) {
1732a1fd
JH
294 /*
295 * Each commit object can cache one origin in that
296 * commit. This is a freestanding copy of origin and
297 * not refcounted.
854b97f6 298 */
0d981c67 299 struct origin *cached = parent->util;
854b97f6 300 if (!strcmp(cached->path, origin->path)) {
1732a1fd
JH
301 /*
302 * The same path between origin and its parent
303 * without renaming -- the most common case.
304 */
854b97f6 305 porigin = get_origin(sb, parent, cached->path);
1732a1fd
JH
306
307 /*
308 * If the origin was newly created (i.e. get_origin
309 * would call make_origin if none is found in the
310 * scoreboard), it does not know the blob_sha1,
311 * so copy it. Otherwise porigin was in the
312 * scoreboard and already knows blob_sha1.
313 */
854b97f6
JH
314 if (porigin->refcnt == 1)
315 hashcpy(porigin->blob_sha1, cached->blob_sha1);
316 return porigin;
317 }
318 /* otherwise it was not very useful; free it */
319 free(parent->util);
320 parent->util = NULL;
0d981c67
JH
321 }
322
f6c0e191
JH
323 /* See if the origin->path is different between parent
324 * and origin first. Most of the time they are the
325 * same and diff-tree is fairly efficient about this.
326 */
327 diff_setup(&diff_opts);
328 diff_opts.recursive = 1;
329 diff_opts.detect_rename = 0;
330 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
331 paths[0] = origin->path;
332 paths[1] = NULL;
333
334 diff_tree_setup_paths(paths, &diff_opts);
335 if (diff_setup_done(&diff_opts) < 0)
336 die("diff-setup");
1cfe7733
JH
337
338 if (is_null_sha1(origin->commit->object.sha1))
339 do_diff_cache(parent->tree->object.sha1, &diff_opts);
340 else
341 diff_tree_sha1(parent->tree->object.sha1,
342 origin->commit->tree->object.sha1,
343 "", &diff_opts);
f6c0e191
JH
344 diffcore_std(&diff_opts);
345
346 /* It is either one entry that says "modified", or "created",
347 * or nothing.
348 */
349 if (!diff_queued_diff.nr) {
350 /* The path is the same as parent */
351 porigin = get_origin(sb, parent, origin->path);
352 hashcpy(porigin->blob_sha1, origin->blob_sha1);
353 }
354 else if (diff_queued_diff.nr != 1)
acca687f 355 die("internal error in blame::find_origin");
f6c0e191
JH
356 else {
357 struct diff_filepair *p = diff_queued_diff.queue[0];
358 switch (p->status) {
359 default:
acca687f 360 die("internal error in blame::find_origin (%c)",
f6c0e191
JH
361 p->status);
362 case 'M':
363 porigin = get_origin(sb, parent, origin->path);
364 hashcpy(porigin->blob_sha1, p->one->sha1);
365 break;
366 case 'A':
367 case 'T':
368 /* Did not exist in parent, or type changed */
369 break;
370 }
371 }
372 diff_flush(&diff_opts);
0d981c67 373 if (porigin) {
1732a1fd
JH
374 /*
375 * Create a freestanding copy that is not part of
376 * the refcounted origin found in the scoreboard, and
377 * cache it in the commit.
378 */
854b97f6 379 struct origin *cached;
1732a1fd 380
854b97f6
JH
381 cached = make_origin(porigin->commit, porigin->path);
382 hashcpy(cached->blob_sha1, porigin->blob_sha1);
383 parent->util = cached;
0d981c67 384 }
f69e743d
JH
385 return porigin;
386}
f6c0e191 387
1732a1fd
JH
388/*
389 * We have an origin -- find the path that corresponds to it in its
390 * parent and return an origin structure to represent it.
391 */
f69e743d
JH
392static struct origin *find_rename(struct scoreboard *sb,
393 struct commit *parent,
394 struct origin *origin)
395{
396 struct origin *porigin = NULL;
397 struct diff_options diff_opts;
398 int i;
399 const char *paths[2];
cee7f245
JH
400
401 diff_setup(&diff_opts);
402 diff_opts.recursive = 1;
403 diff_opts.detect_rename = DIFF_DETECT_RENAME;
404 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
2f3f8b21 405 diff_opts.single_follow = origin->path;
cee7f245
JH
406 paths[0] = NULL;
407 diff_tree_setup_paths(paths, &diff_opts);
408 if (diff_setup_done(&diff_opts) < 0)
409 die("diff-setup");
1cfe7733
JH
410
411 if (is_null_sha1(origin->commit->object.sha1))
412 do_diff_cache(parent->tree->object.sha1, &diff_opts);
413 else
414 diff_tree_sha1(parent->tree->object.sha1,
415 origin->commit->tree->object.sha1,
416 "", &diff_opts);
cee7f245
JH
417 diffcore_std(&diff_opts);
418
419 for (i = 0; i < diff_queued_diff.nr; i++) {
420 struct diff_filepair *p = diff_queued_diff.queue[i];
612702e8 421 if ((p->status == 'R' || p->status == 'C') &&
f6c0e191
JH
422 !strcmp(p->two->path, origin->path)) {
423 porigin = get_origin(sb, parent, p->one->path);
424 hashcpy(porigin->blob_sha1, p->one->sha1);
cee7f245
JH
425 break;
426 }
427 }
428 diff_flush(&diff_opts);
429 return porigin;
430}
431
1732a1fd
JH
432/*
433 * Parsing of patch chunks...
434 */
cee7f245
JH
435struct chunk {
436 /* line number in postimage; up to but not including this
437 * line is the same as preimage
438 */
439 int same;
440
441 /* preimage line number after this chunk */
442 int p_next;
443
444 /* postimage line number after this chunk */
445 int t_next;
446};
447
448struct patch {
449 struct chunk *chunks;
450 int num;
451};
452
453struct blame_diff_state {
454 struct xdiff_emit_state xm;
455 struct patch *ret;
456 unsigned hunk_post_context;
457 unsigned hunk_in_pre_context : 1;
458};
459
460static void process_u_diff(void *state_, char *line, unsigned long len)
461{
462 struct blame_diff_state *state = state_;
463 struct chunk *chunk;
464 int off1, off2, len1, len2, num;
465
cee7f245
JH
466 num = state->ret->num;
467 if (len < 4 || line[0] != '@' || line[1] != '@') {
468 if (state->hunk_in_pre_context && line[0] == ' ')
469 state->ret->chunks[num - 1].same++;
470 else {
471 state->hunk_in_pre_context = 0;
472 if (line[0] == ' ')
473 state->hunk_post_context++;
474 else
475 state->hunk_post_context = 0;
476 }
477 return;
478 }
479
480 if (num && state->hunk_post_context) {
481 chunk = &state->ret->chunks[num - 1];
482 chunk->p_next -= state->hunk_post_context;
483 chunk->t_next -= state->hunk_post_context;
484 }
485 state->ret->num = ++num;
486 state->ret->chunks = xrealloc(state->ret->chunks,
487 sizeof(struct chunk) * num);
488 chunk = &state->ret->chunks[num - 1];
489 if (parse_hunk_header(line, len, &off1, &len1, &off2, &len2)) {
490 state->ret->num--;
491 return;
492 }
493
494 /* Line numbers in patch output are one based. */
495 off1--;
496 off2--;
497
498 chunk->same = len2 ? off2 : (off2 + 1);
499
500 chunk->p_next = off1 + (len1 ? len1 : 1);
501 chunk->t_next = chunk->same + len2;
502 state->hunk_in_pre_context = 1;
503 state->hunk_post_context = 0;
504}
505
506static struct patch *compare_buffer(mmfile_t *file_p, mmfile_t *file_o,
507 int context)
508{
509 struct blame_diff_state state;
510 xpparam_t xpp;
511 xdemitconf_t xecfg;
512 xdemitcb_t ecb;
513
514 xpp.flags = XDF_NEED_MINIMAL;
515 xecfg.ctxlen = context;
516 xecfg.flags = 0;
517 ecb.outf = xdiff_outf;
518 ecb.priv = &state;
519 memset(&state, 0, sizeof(state));
520 state.xm.consume = process_u_diff;
521 state.ret = xmalloc(sizeof(struct patch));
522 state.ret->chunks = NULL;
523 state.ret->num = 0;
524
525 xdl_diff(file_p, file_o, &xpp, &xecfg, &ecb);
526
527 if (state.ret->num) {
528 struct chunk *chunk;
529 chunk = &state.ret->chunks[state.ret->num - 1];
530 chunk->p_next -= state.hunk_post_context;
531 chunk->t_next -= state.hunk_post_context;
532 }
533 return state.ret;
534}
535
1732a1fd
JH
536/*
537 * Run diff between two origins and grab the patch output, so that
538 * we can pass blame for lines origin is currently suspected for
539 * to its parent.
540 */
cee7f245
JH
541static struct patch *get_patch(struct origin *parent, struct origin *origin)
542{
543 mmfile_t file_p, file_o;
cee7f245
JH
544 struct patch *patch;
545
c2e525d9
JH
546 fill_origin_blob(parent, &file_p);
547 fill_origin_blob(origin, &file_o);
548 if (!file_p.ptr || !file_o.ptr)
cee7f245 549 return NULL;
cee7f245 550 patch = compare_buffer(&file_p, &file_o, 0);
c2e525d9 551 num_get_patch++;
cee7f245
JH
552 return patch;
553}
554
555static void free_patch(struct patch *p)
556{
557 free(p->chunks);
558 free(p);
559}
560
1732a1fd 561/*
3dff5379 562 * Link in a new blame entry to the scoreboard. Entries that cover the
1732a1fd
JH
563 * same line range have been removed from the scoreboard previously.
564 */
cee7f245
JH
565static void add_blame_entry(struct scoreboard *sb, struct blame_entry *e)
566{
567 struct blame_entry *ent, *prev = NULL;
568
54a4c617
JH
569 origin_incref(e->suspect);
570
cee7f245
JH
571 for (ent = sb->ent; ent && ent->lno < e->lno; ent = ent->next)
572 prev = ent;
573
574 /* prev, if not NULL, is the last one that is below e */
575 e->prev = prev;
576 if (prev) {
577 e->next = prev->next;
578 prev->next = e;
579 }
580 else {
581 e->next = sb->ent;
582 sb->ent = e;
583 }
584 if (e->next)
585 e->next->prev = e;
586}
587
1732a1fd
JH
588/*
589 * src typically is on-stack; we want to copy the information in it to
590 * an malloced blame_entry that is already on the linked list of the
591 * scoreboard. The origin of dst loses a refcnt while the origin of src
592 * gains one.
593 */
cee7f245
JH
594static void dup_entry(struct blame_entry *dst, struct blame_entry *src)
595{
596 struct blame_entry *p, *n;
54a4c617 597
cee7f245
JH
598 p = dst->prev;
599 n = dst->next;
54a4c617
JH
600 origin_incref(src->suspect);
601 origin_decref(dst->suspect);
cee7f245
JH
602 memcpy(dst, src, sizeof(*src));
603 dst->prev = p;
604 dst->next = n;
5ff62c30 605 dst->score = 0;
cee7f245
JH
606}
607
608static const char *nth_line(struct scoreboard *sb, int lno)
609{
610 return sb->final_buf + sb->lineno[lno];
611}
612
1732a1fd
JH
613/*
614 * It is known that lines between tlno to same came from parent, and e
615 * has an overlap with that range. it also is known that parent's
616 * line plno corresponds to e's line tlno.
617 *
618 * <---- e ----->
619 * <------>
620 * <------------>
621 * <------------>
622 * <------------------>
623 *
624 * Split e into potentially three parts; before this chunk, the chunk
625 * to be blamed for the parent, and after that portion.
626 */
54a4c617 627static void split_overlap(struct blame_entry *split,
cee7f245
JH
628 struct blame_entry *e,
629 int tlno, int plno, int same,
630 struct origin *parent)
631{
cee7f245
JH
632 int chunk_end_lno;
633 memset(split, 0, sizeof(struct blame_entry [3]));
634
635 if (e->s_lno < tlno) {
636 /* there is a pre-chunk part not blamed on parent */
54a4c617 637 split[0].suspect = origin_incref(e->suspect);
cee7f245
JH
638 split[0].lno = e->lno;
639 split[0].s_lno = e->s_lno;
640 split[0].num_lines = tlno - e->s_lno;
641 split[1].lno = e->lno + tlno - e->s_lno;
642 split[1].s_lno = plno;
643 }
644 else {
645 split[1].lno = e->lno;
646 split[1].s_lno = plno + (e->s_lno - tlno);
647 }
648
649 if (same < e->s_lno + e->num_lines) {
650 /* there is a post-chunk part not blamed on parent */
54a4c617 651 split[2].suspect = origin_incref(e->suspect);
cee7f245
JH
652 split[2].lno = e->lno + (same - e->s_lno);
653 split[2].s_lno = e->s_lno + (same - e->s_lno);
654 split[2].num_lines = e->s_lno + e->num_lines - same;
655 chunk_end_lno = split[2].lno;
656 }
657 else
658 chunk_end_lno = e->lno + e->num_lines;
659 split[1].num_lines = chunk_end_lno - split[1].lno;
660
1732a1fd
JH
661 /*
662 * if it turns out there is nothing to blame the parent for,
663 * forget about the splitting. !split[1].suspect signals this.
664 */
cee7f245
JH
665 if (split[1].num_lines < 1)
666 return;
54a4c617 667 split[1].suspect = origin_incref(parent);
cee7f245
JH
668}
669
1732a1fd
JH
670/*
671 * split_overlap() divided an existing blame e into up to three parts
672 * in split. Adjust the linked list of blames in the scoreboard to
673 * reflect the split.
674 */
cee7f245 675static void split_blame(struct scoreboard *sb,
54a4c617 676 struct blame_entry *split,
cee7f245
JH
677 struct blame_entry *e)
678{
679 struct blame_entry *new_entry;
680
681 if (split[0].suspect && split[2].suspect) {
1732a1fd 682 /* The first part (reuse storage for the existing entry e) */
cee7f245
JH
683 dup_entry(e, &split[0]);
684
1732a1fd 685 /* The last part -- me */
cee7f245
JH
686 new_entry = xmalloc(sizeof(*new_entry));
687 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
688 add_blame_entry(sb, new_entry);
689
1732a1fd 690 /* ... and the middle part -- parent */
cee7f245
JH
691 new_entry = xmalloc(sizeof(*new_entry));
692 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
693 add_blame_entry(sb, new_entry);
694 }
695 else if (!split[0].suspect && !split[2].suspect)
1732a1fd
JH
696 /*
697 * The parent covers the entire area; reuse storage for
698 * e and replace it with the parent.
699 */
cee7f245
JH
700 dup_entry(e, &split[1]);
701 else if (split[0].suspect) {
1732a1fd 702 /* me and then parent */
cee7f245
JH
703 dup_entry(e, &split[0]);
704
705 new_entry = xmalloc(sizeof(*new_entry));
706 memcpy(new_entry, &(split[1]), sizeof(struct blame_entry));
707 add_blame_entry(sb, new_entry);
708 }
709 else {
1732a1fd 710 /* parent and then me */
cee7f245
JH
711 dup_entry(e, &split[1]);
712
713 new_entry = xmalloc(sizeof(*new_entry));
714 memcpy(new_entry, &(split[2]), sizeof(struct blame_entry));
715 add_blame_entry(sb, new_entry);
716 }
717
54a4c617 718 if (DEBUG) { /* sanity */
cee7f245 719 struct blame_entry *ent;
612702e8 720 int lno = sb->ent->lno, corrupt = 0;
cee7f245
JH
721
722 for (ent = sb->ent; ent; ent = ent->next) {
723 if (lno != ent->lno)
724 corrupt = 1;
725 if (ent->s_lno < 0)
726 corrupt = 1;
727 lno += ent->num_lines;
728 }
729 if (corrupt) {
612702e8 730 lno = sb->ent->lno;
cee7f245
JH
731 for (ent = sb->ent; ent; ent = ent->next) {
732 printf("L %8d l %8d n %8d\n",
733 lno, ent->lno, ent->num_lines);
734 lno = ent->lno + ent->num_lines;
735 }
736 die("oops");
737 }
738 }
739}
740
1732a1fd
JH
741/*
742 * After splitting the blame, the origins used by the
743 * on-stack blame_entry should lose one refcnt each.
744 */
54a4c617
JH
745static void decref_split(struct blame_entry *split)
746{
747 int i;
748
749 for (i = 0; i < 3; i++)
750 origin_decref(split[i].suspect);
751}
752
1732a1fd
JH
753/*
754 * Helper for blame_chunk(). blame_entry e is known to overlap with
755 * the patch hunk; split it and pass blame to the parent.
756 */
cee7f245
JH
757static void blame_overlap(struct scoreboard *sb, struct blame_entry *e,
758 int tlno, int plno, int same,
759 struct origin *parent)
760{
761 struct blame_entry split[3];
762
763 split_overlap(split, e, tlno, plno, same, parent);
54a4c617
JH
764 if (split[1].suspect)
765 split_blame(sb, split, e);
766 decref_split(split);
cee7f245
JH
767}
768
1732a1fd
JH
769/*
770 * Find the line number of the last line the target is suspected for.
771 */
cee7f245
JH
772static int find_last_in_target(struct scoreboard *sb, struct origin *target)
773{
774 struct blame_entry *e;
775 int last_in_target = -1;
776
777 for (e = sb->ent; e; e = e->next) {
46014766 778 if (e->guilty || cmp_suspect(e->suspect, target))
cee7f245
JH
779 continue;
780 if (last_in_target < e->s_lno + e->num_lines)
781 last_in_target = e->s_lno + e->num_lines;
782 }
783 return last_in_target;
784}
785
1732a1fd
JH
786/*
787 * Process one hunk from the patch between the current suspect for
788 * blame_entry e and its parent. Find and split the overlap, and
789 * pass blame to the overlapping part to the parent.
790 */
cee7f245
JH
791static void blame_chunk(struct scoreboard *sb,
792 int tlno, int plno, int same,
793 struct origin *target, struct origin *parent)
794{
612702e8 795 struct blame_entry *e;
cee7f245 796
612702e8 797 for (e = sb->ent; e; e = e->next) {
46014766 798 if (e->guilty || cmp_suspect(e->suspect, target))
cee7f245
JH
799 continue;
800 if (same <= e->s_lno)
801 continue;
802 if (tlno < e->s_lno + e->num_lines)
803 blame_overlap(sb, e, tlno, plno, same, parent);
804 }
805}
806
1732a1fd
JH
807/*
808 * We are looking at the origin 'target' and aiming to pass blame
809 * for the lines it is suspected to its parent. Run diff to find
810 * which lines came from parent and pass blame for them.
811 */
cee7f245
JH
812static int pass_blame_to_parent(struct scoreboard *sb,
813 struct origin *target,
814 struct origin *parent)
815{
816 int i, last_in_target, plno, tlno;
817 struct patch *patch;
818
819 last_in_target = find_last_in_target(sb, target);
820 if (last_in_target < 0)
821 return 1; /* nothing remains for this target */
822
823 patch = get_patch(parent, target);
824 plno = tlno = 0;
825 for (i = 0; i < patch->num; i++) {
826 struct chunk *chunk = &patch->chunks[i];
827
cee7f245
JH
828 blame_chunk(sb, tlno, plno, chunk->same, target, parent);
829 plno = chunk->p_next;
830 tlno = chunk->t_next;
831 }
1732a1fd 832 /* The rest (i.e. anything after tlno) are the same as the parent */
cee7f245
JH
833 blame_chunk(sb, tlno, plno, last_in_target, target, parent);
834
835 free_patch(patch);
836 return 0;
837}
838
1732a1fd
JH
839/*
840 * The lines in blame_entry after splitting blames many times can become
841 * very small and trivial, and at some point it becomes pointless to
842 * blame the parents. E.g. "\t\t}\n\t}\n\n" appears everywhere in any
843 * ordinary C program, and it is not worth to say it was copied from
844 * totally unrelated file in the parent.
845 *
846 * Compute how trivial the lines in the blame_entry are.
847 */
5ff62c30
JH
848static unsigned ent_score(struct scoreboard *sb, struct blame_entry *e)
849{
850 unsigned score;
851 const char *cp, *ep;
852
853 if (e->score)
854 return e->score;
855
612702e8 856 score = 1;
5ff62c30
JH
857 cp = nth_line(sb, e->lno);
858 ep = nth_line(sb, e->lno + e->num_lines);
859 while (cp < ep) {
860 unsigned ch = *((unsigned char *)cp);
861 if (isalnum(ch))
862 score++;
863 cp++;
864 }
865 e->score = score;
866 return score;
867}
868
1732a1fd
JH
869/*
870 * best_so_far[] and this[] are both a split of an existing blame_entry
871 * that passes blame to the parent. Maintain best_so_far the best split
872 * so far, by comparing this and best_so_far and copying this into
873 * bst_so_far as needed.
874 */
5ff62c30 875static void copy_split_if_better(struct scoreboard *sb,
54a4c617
JH
876 struct blame_entry *best_so_far,
877 struct blame_entry *this)
d24bba80 878{
54a4c617
JH
879 int i;
880
d24bba80
JH
881 if (!this[1].suspect)
882 return;
5ff62c30
JH
883 if (best_so_far[1].suspect) {
884 if (ent_score(sb, &this[1]) < ent_score(sb, &best_so_far[1]))
885 return;
886 }
54a4c617
JH
887
888 for (i = 0; i < 3; i++)
889 origin_incref(this[i].suspect);
890 decref_split(best_so_far);
d24bba80
JH
891 memcpy(best_so_far, this, sizeof(struct blame_entry [3]));
892}
893
1732a1fd
JH
894/*
895 * Find the lines from parent that are the same as ent so that
896 * we can pass blames to it. file_p has the blob contents for
897 * the parent.
898 */
d24bba80
JH
899static void find_copy_in_blob(struct scoreboard *sb,
900 struct blame_entry *ent,
901 struct origin *parent,
54a4c617 902 struct blame_entry *split,
d24bba80
JH
903 mmfile_t *file_p)
904{
905 const char *cp;
906 int cnt;
907 mmfile_t file_o;
908 struct patch *patch;
909 int i, plno, tlno;
910
1732a1fd
JH
911 /*
912 * Prepare mmfile that contains only the lines in ent.
913 */
d24bba80
JH
914 cp = nth_line(sb, ent->lno);
915 file_o.ptr = (char*) cp;
916 cnt = ent->num_lines;
917
918 while (cnt && cp < sb->final_buf + sb->final_buf_size) {
919 if (*cp++ == '\n')
920 cnt--;
921 }
922 file_o.size = cp - file_o.ptr;
923
924 patch = compare_buffer(file_p, &file_o, 1);
925
926 memset(split, 0, sizeof(struct blame_entry [3]));
927 plno = tlno = 0;
928 for (i = 0; i < patch->num; i++) {
929 struct chunk *chunk = &patch->chunks[i];
930
931 /* tlno to chunk->same are the same as ent */
932 if (ent->num_lines <= tlno)
933 break;
934 if (tlno < chunk->same) {
935 struct blame_entry this[3];
936 split_overlap(this, ent,
937 tlno + ent->s_lno, plno,
938 chunk->same + ent->s_lno,
939 parent);
5ff62c30 940 copy_split_if_better(sb, split, this);
54a4c617 941 decref_split(this);
d24bba80
JH
942 }
943 plno = chunk->p_next;
944 tlno = chunk->t_next;
945 }
946 free_patch(patch);
947}
948
1732a1fd
JH
949/*
950 * See if lines currently target is suspected for can be attributed to
951 * parent.
952 */
d24bba80
JH
953static int find_move_in_parent(struct scoreboard *sb,
954 struct origin *target,
955 struct origin *parent)
956{
650e2f67 957 int last_in_target, made_progress;
46014766 958 struct blame_entry *e, split[3];
d24bba80 959 mmfile_t file_p;
d24bba80
JH
960
961 last_in_target = find_last_in_target(sb, target);
962 if (last_in_target < 0)
963 return 1; /* nothing remains for this target */
964
c2e525d9
JH
965 fill_origin_blob(parent, &file_p);
966 if (!file_p.ptr)
d24bba80 967 return 0;
d24bba80 968
650e2f67
JH
969 made_progress = 1;
970 while (made_progress) {
971 made_progress = 0;
972 for (e = sb->ent; e; e = e->next) {
973 if (e->guilty || cmp_suspect(e->suspect, target))
974 continue;
975 find_copy_in_blob(sb, e, parent, split, &file_p);
976 if (split[1].suspect &&
977 blame_move_score < ent_score(sb, &split[1])) {
978 split_blame(sb, split, e);
979 made_progress = 1;
980 }
981 decref_split(split);
982 }
d24bba80 983 }
d24bba80
JH
984 return 0;
985}
986
33494784
JH
987struct blame_list {
988 struct blame_entry *ent;
989 struct blame_entry split[3];
990};
991
1732a1fd
JH
992/*
993 * Count the number of entries the target is suspected for,
994 * and prepare a list of entry and the best split.
995 */
33494784
JH
996static struct blame_list *setup_blame_list(struct scoreboard *sb,
997 struct origin *target,
998 int *num_ents_p)
999{
1000 struct blame_entry *e;
1001 int num_ents, i;
1002 struct blame_list *blame_list = NULL;
1003
33494784
JH
1004 for (e = sb->ent, num_ents = 0; e; e = e->next)
1005 if (!e->guilty && !cmp_suspect(e->suspect, target))
1006 num_ents++;
1007 if (num_ents) {
1008 blame_list = xcalloc(num_ents, sizeof(struct blame_list));
1009 for (e = sb->ent, i = 0; e; e = e->next)
1010 if (!e->guilty && !cmp_suspect(e->suspect, target))
1011 blame_list[i++].ent = e;
1012 }
1013 *num_ents_p = num_ents;
1014 return blame_list;
1015}
1016
1732a1fd
JH
1017/*
1018 * For lines target is suspected for, see if we can find code movement
1019 * across file boundary from the parent commit. porigin is the path
1020 * in the parent we already tried.
1021 */
18abd745
JH
1022static int find_copy_in_parent(struct scoreboard *sb,
1023 struct origin *target,
1024 struct commit *parent,
1025 struct origin *porigin,
1026 int opt)
1027{
1028 struct diff_options diff_opts;
1029 const char *paths[1];
aec8fa1f 1030 int i, j;
33494784
JH
1031 int retval;
1032 struct blame_list *blame_list;
aec8fa1f 1033 int num_ents;
18abd745 1034
33494784
JH
1035 blame_list = setup_blame_list(sb, target, &num_ents);
1036 if (!blame_list)
18abd745
JH
1037 return 1; /* nothing remains for this target */
1038
1039 diff_setup(&diff_opts);
1040 diff_opts.recursive = 1;
1041 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
1042
18abd745
JH
1043 paths[0] = NULL;
1044 diff_tree_setup_paths(paths, &diff_opts);
1045 if (diff_setup_done(&diff_opts) < 0)
1046 die("diff-setup");
33494784
JH
1047
1048 /* Try "find copies harder" on new path if requested;
1049 * we do not want to use diffcore_rename() actually to
1050 * match things up; find_copies_harder is set only to
1051 * force diff_tree_sha1() to feed all filepairs to diff_queue,
1052 * and this code needs to be after diff_setup_done(), which
1053 * usually makes find-copies-harder imply copy detection.
1054 */
1055 if ((opt & PICKAXE_BLAME_COPY_HARDER) &&
1056 (!porigin || strcmp(target->path, porigin->path)))
1057 diff_opts.find_copies_harder = 1;
1058
1cfe7733
JH
1059 if (is_null_sha1(target->commit->object.sha1))
1060 do_diff_cache(parent->tree->object.sha1, &diff_opts);
1061 else
1062 diff_tree_sha1(parent->tree->object.sha1,
1063 target->commit->tree->object.sha1,
1064 "", &diff_opts);
18abd745 1065
33494784
JH
1066 if (!diff_opts.find_copies_harder)
1067 diffcore_std(&diff_opts);
18abd745 1068
33494784
JH
1069 retval = 0;
1070 while (1) {
1071 int made_progress = 0;
1072
1073 for (i = 0; i < diff_queued_diff.nr; i++) {
1074 struct diff_filepair *p = diff_queued_diff.queue[i];
1075 struct origin *norigin;
1076 mmfile_t file_p;
33494784
JH
1077 struct blame_entry this[3];
1078
1079 if (!DIFF_FILE_VALID(p->one))
1080 continue; /* does not exist in parent */
1081 if (porigin && !strcmp(p->one->path, porigin->path))
1082 /* find_move already dealt with this path */
1083 continue;
1084
1085 norigin = get_origin(sb, parent, p->one->path);
1086 hashcpy(norigin->blob_sha1, p->one->sha1);
c2e525d9 1087 fill_origin_blob(norigin, &file_p);
33494784
JH
1088 if (!file_p.ptr)
1089 continue;
1090
1091 for (j = 0; j < num_ents; j++) {
1092 find_copy_in_blob(sb, blame_list[j].ent,
1093 norigin, this, &file_p);
1094 copy_split_if_better(sb, blame_list[j].split,
1095 this);
1096 decref_split(this);
1097 }
33494784
JH
1098 origin_decref(norigin);
1099 }
18abd745 1100
aec8fa1f 1101 for (j = 0; j < num_ents; j++) {
33494784
JH
1102 struct blame_entry *split = blame_list[j].split;
1103 if (split[1].suspect &&
1104 blame_copy_score < ent_score(sb, &split[1])) {
1105 split_blame(sb, split, blame_list[j].ent);
1106 made_progress = 1;
1107 }
1108 decref_split(split);
18abd745 1109 }
33494784 1110 free(blame_list);
aec8fa1f 1111
33494784
JH
1112 if (!made_progress)
1113 break;
1114 blame_list = setup_blame_list(sb, target, &num_ents);
1115 if (!blame_list) {
1116 retval = 1;
1117 break;
1118 }
18abd745 1119 }
33494784 1120 diff_flush(&diff_opts);
18abd745 1121
33494784 1122 return retval;
18abd745
JH
1123}
1124
1732a1fd
JH
1125/*
1126 * The blobs of origin and porigin exactly match, so everything
c2e525d9
JH
1127 * origin is suspected for can be blamed on the parent.
1128 */
1129static void pass_whole_blame(struct scoreboard *sb,
1130 struct origin *origin, struct origin *porigin)
1131{
1132 struct blame_entry *e;
1133
1134 if (!porigin->file.ptr && origin->file.ptr) {
1135 /* Steal its file */
1136 porigin->file = origin->file;
1137 origin->file.ptr = NULL;
1138 }
1139 for (e = sb->ent; e; e = e->next) {
1140 if (cmp_suspect(e->suspect, origin))
1141 continue;
1142 origin_incref(porigin);
1143 origin_decref(e->suspect);
1144 e->suspect = porigin;
1145 }
1146}
1147
cee7f245
JH
1148#define MAXPARENT 16
1149
d24bba80 1150static void pass_blame(struct scoreboard *sb, struct origin *origin, int opt)
cee7f245 1151{
f69e743d 1152 int i, pass;
cee7f245
JH
1153 struct commit *commit = origin->commit;
1154 struct commit_list *parent;
1155 struct origin *parent_origin[MAXPARENT], *porigin;
1156
1157 memset(parent_origin, 0, sizeof(parent_origin));
cee7f245 1158
f69e743d
JH
1159 /* The first pass looks for unrenamed path to optimize for
1160 * common cases, then we look for renames in the second pass.
1161 */
1162 for (pass = 0; pass < 2; pass++) {
1163 struct origin *(*find)(struct scoreboard *,
1164 struct commit *, struct origin *);
1165 find = pass ? find_rename : find_origin;
1166
1167 for (i = 0, parent = commit->parents;
1168 i < MAXPARENT && parent;
1169 parent = parent->next, i++) {
1170 struct commit *p = parent->item;
0421d9f8 1171 int j, same;
f69e743d
JH
1172
1173 if (parent_origin[i])
1174 continue;
1175 if (parse_commit(p))
1176 continue;
0d981c67 1177 porigin = find(sb, p, origin);
f69e743d
JH
1178 if (!porigin)
1179 continue;
1180 if (!hashcmp(porigin->blob_sha1, origin->blob_sha1)) {
c2e525d9 1181 pass_whole_blame(sb, origin, porigin);
f69e743d
JH
1182 origin_decref(porigin);
1183 goto finish;
1184 }
0421d9f8 1185 for (j = same = 0; j < i; j++)
33494784
JH
1186 if (parent_origin[j] &&
1187 !hashcmp(parent_origin[j]->blob_sha1,
0421d9f8
JH
1188 porigin->blob_sha1)) {
1189 same = 1;
1190 break;
1191 }
1192 if (!same)
1193 parent_origin[i] = porigin;
1194 else
1195 origin_decref(porigin);
cee7f245 1196 }
cee7f245
JH
1197 }
1198
c2e525d9 1199 num_commits++;
cee7f245
JH
1200 for (i = 0, parent = commit->parents;
1201 i < MAXPARENT && parent;
1202 parent = parent->next, i++) {
1203 struct origin *porigin = parent_origin[i];
1204 if (!porigin)
1205 continue;
1206 if (pass_blame_to_parent(sb, origin, porigin))
54a4c617 1207 goto finish;
cee7f245 1208 }
d24bba80
JH
1209
1210 /*
1732a1fd 1211 * Optionally find moves in parents' files.
d24bba80
JH
1212 */
1213 if (opt & PICKAXE_BLAME_MOVE)
1214 for (i = 0, parent = commit->parents;
1215 i < MAXPARENT && parent;
1216 parent = parent->next, i++) {
1217 struct origin *porigin = parent_origin[i];
1218 if (!porigin)
1219 continue;
1220 if (find_move_in_parent(sb, origin, porigin))
54a4c617 1221 goto finish;
d24bba80
JH
1222 }
1223
18abd745 1224 /*
1732a1fd 1225 * Optionally find copies from parents' files.
18abd745
JH
1226 */
1227 if (opt & PICKAXE_BLAME_COPY)
1228 for (i = 0, parent = commit->parents;
1229 i < MAXPARENT && parent;
1230 parent = parent->next, i++) {
1231 struct origin *porigin = parent_origin[i];
1232 if (find_copy_in_parent(sb, origin, parent->item,
1233 porigin, opt))
54a4c617 1234 goto finish;
18abd745 1235 }
54a4c617
JH
1236
1237 finish:
1238 for (i = 0; i < MAXPARENT; i++)
1239 origin_decref(parent_origin[i]);
cee7f245
JH
1240}
1241
1732a1fd
JH
1242/*
1243 * Information on commits, used for output.
1244 */
cee7f245
JH
1245struct commit_info
1246{
1247 char *author;
1248 char *author_mail;
1249 unsigned long author_time;
1250 char *author_tz;
1251
1252 /* filled only when asked for details */
1253 char *committer;
1254 char *committer_mail;
1255 unsigned long committer_time;
1256 char *committer_tz;
1257
1258 char *summary;
1259};
1260
1732a1fd
JH
1261/*
1262 * Parse author/committer line in the commit object buffer
1263 */
cee7f245
JH
1264static void get_ac_line(const char *inbuf, const char *what,
1265 int bufsz, char *person, char **mail,
1266 unsigned long *time, char **tz)
1267{
1268 int len;
1269 char *tmp, *endp;
1270
1271 tmp = strstr(inbuf, what);
1272 if (!tmp)
1273 goto error_out;
1274 tmp += strlen(what);
1275 endp = strchr(tmp, '\n');
1276 if (!endp)
1277 len = strlen(tmp);
1278 else
1279 len = endp - tmp;
1280 if (bufsz <= len) {
1281 error_out:
1282 /* Ugh */
1283 person = *mail = *tz = "(unknown)";
1284 *time = 0;
1285 return;
1286 }
1287 memcpy(person, tmp, len);
1288
1289 tmp = person;
1290 tmp += len;
1291 *tmp = 0;
1292 while (*tmp != ' ')
1293 tmp--;
1294 *tz = tmp+1;
1295
1296 *tmp = 0;
1297 while (*tmp != ' ')
1298 tmp--;
1299 *time = strtoul(tmp, NULL, 10);
1300
1301 *tmp = 0;
1302 while (*tmp != ' ')
1303 tmp--;
1304 *mail = tmp + 1;
1305 *tmp = 0;
1306}
1307
1308static void get_commit_info(struct commit *commit,
1309 struct commit_info *ret,
1310 int detailed)
1311{
1312 int len;
1313 char *tmp, *endp;
1314 static char author_buf[1024];
1315 static char committer_buf[1024];
1316 static char summary_buf[1024];
1317
1732a1fd
JH
1318 /*
1319 * We've operated without save_commit_buffer, so
612702e8
JH
1320 * we now need to populate them for output.
1321 */
1322 if (!commit->buffer) {
21666f1a 1323 enum object_type type;
612702e8
JH
1324 unsigned long size;
1325 commit->buffer =
21666f1a 1326 read_sha1_file(commit->object.sha1, &type, &size);
612702e8 1327 }
cee7f245
JH
1328 ret->author = author_buf;
1329 get_ac_line(commit->buffer, "\nauthor ",
1330 sizeof(author_buf), author_buf, &ret->author_mail,
1331 &ret->author_time, &ret->author_tz);
1332
1333 if (!detailed)
1334 return;
1335
1336 ret->committer = committer_buf;
1337 get_ac_line(commit->buffer, "\ncommitter ",
1338 sizeof(committer_buf), committer_buf, &ret->committer_mail,
1339 &ret->committer_time, &ret->committer_tz);
1340
1341 ret->summary = summary_buf;
1342 tmp = strstr(commit->buffer, "\n\n");
1343 if (!tmp) {
1344 error_out:
1345 sprintf(summary_buf, "(%s)", sha1_to_hex(commit->object.sha1));
1346 return;
1347 }
1348 tmp += 2;
1349 endp = strchr(tmp, '\n');
1350 if (!endp)
1cfe7733 1351 endp = tmp + strlen(tmp);
cee7f245 1352 len = endp - tmp;
1cfe7733 1353 if (len >= sizeof(summary_buf) || len == 0)
cee7f245
JH
1354 goto error_out;
1355 memcpy(summary_buf, tmp, len);
1356 summary_buf[len] = 0;
1357}
1358
1732a1fd
JH
1359/*
1360 * To allow LF and other nonportable characters in pathnames,
1361 * they are c-style quoted as needed.
1362 */
46e5e69d
JH
1363static void write_filename_info(const char *path)
1364{
1365 printf("filename ");
1366 write_name_quoted(NULL, 0, path, 1, stdout);
1367 putchar('\n');
1368}
1369
1732a1fd
JH
1370/*
1371 * The blame_entry is found to be guilty for the range. Mark it
1372 * as such, and show it in incremental output.
1373 */
717d1462
LT
1374static void found_guilty_entry(struct blame_entry *ent)
1375{
1376 if (ent->guilty)
1377 return;
1378 ent->guilty = 1;
1379 if (incremental) {
1380 struct origin *suspect = ent->suspect;
1381
1382 printf("%s %d %d %d\n",
1383 sha1_to_hex(suspect->commit->object.sha1),
1384 ent->s_lno + 1, ent->lno + 1, ent->num_lines);
1385 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1386 struct commit_info ci;
1387 suspect->commit->object.flags |= METAINFO_SHOWN;
1388 get_commit_info(suspect->commit, &ci, 1);
1389 printf("author %s\n", ci.author);
1390 printf("author-mail %s\n", ci.author_mail);
1391 printf("author-time %lu\n", ci.author_time);
1392 printf("author-tz %s\n", ci.author_tz);
1393 printf("committer %s\n", ci.committer);
1394 printf("committer-mail %s\n", ci.committer_mail);
1395 printf("committer-time %lu\n", ci.committer_time);
1396 printf("committer-tz %s\n", ci.committer_tz);
1397 printf("summary %s\n", ci.summary);
1398 if (suspect->commit->object.flags & UNINTERESTING)
1399 printf("boundary\n");
1400 }
46e5e69d 1401 write_filename_info(suspect->path);
717d1462
LT
1402 }
1403}
1404
1732a1fd
JH
1405/*
1406 * The main loop -- while the scoreboard has lines whose true origin
3dff5379 1407 * is still unknown, pick one blame_entry, and allow its current
1732a1fd
JH
1408 * suspect to pass blames to its parents.
1409 */
717d1462
LT
1410static void assign_blame(struct scoreboard *sb, struct rev_info *revs, int opt)
1411{
1412 while (1) {
1413 struct blame_entry *ent;
1414 struct commit *commit;
1415 struct origin *suspect = NULL;
1416
1417 /* find one suspect to break down */
1418 for (ent = sb->ent; !suspect && ent; ent = ent->next)
1419 if (!ent->guilty)
1420 suspect = ent->suspect;
1421 if (!suspect)
1422 return; /* all done */
1423
1732a1fd
JH
1424 /*
1425 * We will use this suspect later in the loop,
1426 * so hold onto it in the meantime.
1427 */
717d1462
LT
1428 origin_incref(suspect);
1429 commit = suspect->commit;
1430 if (!commit->object.parsed)
1431 parse_commit(commit);
1432 if (!(commit->object.flags & UNINTERESTING) &&
1732a1fd 1433 !(revs->max_age != -1 && commit->date < revs->max_age))
717d1462
LT
1434 pass_blame(sb, suspect, opt);
1435 else {
1436 commit->object.flags |= UNINTERESTING;
1437 if (commit->object.parsed)
1438 mark_parents_uninteresting(commit);
1439 }
1440 /* treat root commit as boundary */
1441 if (!commit->parents && !show_root)
1442 commit->object.flags |= UNINTERESTING;
1443
1444 /* Take responsibility for the remaining entries */
1445 for (ent = sb->ent; ent; ent = ent->next)
1446 if (!cmp_suspect(ent->suspect, suspect))
1447 found_guilty_entry(ent);
1448 origin_decref(suspect);
1449
1450 if (DEBUG) /* sanity */
1451 sanity_check_refcnt(sb);
1452 }
1453}
1454
1455static const char *format_time(unsigned long time, const char *tz_str,
1456 int show_raw_time)
1457{
1458 static char time_buf[128];
1459 time_t t = time;
1460 int minutes, tz;
1461 struct tm *tm;
1462
1463 if (show_raw_time) {
1464 sprintf(time_buf, "%lu %s", time, tz_str);
1465 return time_buf;
1466 }
1467
1468 tz = atoi(tz_str);
1469 minutes = tz < 0 ? -tz : tz;
1470 minutes = (minutes / 100)*60 + (minutes % 100);
1471 minutes = tz < 0 ? -minutes : minutes;
1472 t = time + minutes * 60;
1473 tm = gmtime(&t);
1474
1475 strftime(time_buf, sizeof(time_buf), "%Y-%m-%d %H:%M:%S ", tm);
1476 strcat(time_buf, tz_str);
1477 return time_buf;
1478}
1479
cee7f245
JH
1480#define OUTPUT_ANNOTATE_COMPAT 001
1481#define OUTPUT_LONG_OBJECT_NAME 002
1482#define OUTPUT_RAW_TIMESTAMP 004
1483#define OUTPUT_PORCELAIN 010
1484#define OUTPUT_SHOW_NAME 020
1485#define OUTPUT_SHOW_NUMBER 040
5ff62c30 1486#define OUTPUT_SHOW_SCORE 0100
cee7f245
JH
1487
1488static void emit_porcelain(struct scoreboard *sb, struct blame_entry *ent)
1489{
1490 int cnt;
1491 const char *cp;
1492 struct origin *suspect = ent->suspect;
1493 char hex[41];
1494
1495 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1496 printf("%s%c%d %d %d\n",
1497 hex,
1498 ent->guilty ? ' ' : '*', // purely for debugging
1499 ent->s_lno + 1,
1500 ent->lno + 1,
1501 ent->num_lines);
1502 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1503 struct commit_info ci;
1504 suspect->commit->object.flags |= METAINFO_SHOWN;
1505 get_commit_info(suspect->commit, &ci, 1);
1506 printf("author %s\n", ci.author);
1507 printf("author-mail %s\n", ci.author_mail);
1508 printf("author-time %lu\n", ci.author_time);
1509 printf("author-tz %s\n", ci.author_tz);
1510 printf("committer %s\n", ci.committer);
1511 printf("committer-mail %s\n", ci.committer_mail);
1512 printf("committer-time %lu\n", ci.committer_time);
1513 printf("committer-tz %s\n", ci.committer_tz);
46e5e69d 1514 write_filename_info(suspect->path);
cee7f245 1515 printf("summary %s\n", ci.summary);
b11121d9
JH
1516 if (suspect->commit->object.flags & UNINTERESTING)
1517 printf("boundary\n");
cee7f245
JH
1518 }
1519 else if (suspect->commit->object.flags & MORE_THAN_ONE_PATH)
46e5e69d 1520 write_filename_info(suspect->path);
cee7f245
JH
1521
1522 cp = nth_line(sb, ent->lno);
1523 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1524 char ch;
1525 if (cnt)
1526 printf("%s %d %d\n", hex,
1527 ent->s_lno + 1 + cnt,
1528 ent->lno + 1 + cnt);
1529 putchar('\t');
1530 do {
1531 ch = *cp++;
1532 putchar(ch);
1533 } while (ch != '\n' &&
1534 cp < sb->final_buf + sb->final_buf_size);
1535 }
1536}
1537
1538static void emit_other(struct scoreboard *sb, struct blame_entry *ent, int opt)
1539{
1540 int cnt;
1541 const char *cp;
1542 struct origin *suspect = ent->suspect;
1543 struct commit_info ci;
1544 char hex[41];
1545 int show_raw_time = !!(opt & OUTPUT_RAW_TIMESTAMP);
1546
1547 get_commit_info(suspect->commit, &ci, 1);
1548 strcpy(hex, sha1_to_hex(suspect->commit->object.sha1));
1549
1550 cp = nth_line(sb, ent->lno);
1551 for (cnt = 0; cnt < ent->num_lines; cnt++) {
1552 char ch;
b11121d9
JH
1553 int length = (opt & OUTPUT_LONG_OBJECT_NAME) ? 40 : 8;
1554
1555 if (suspect->commit->object.flags & UNINTERESTING) {
e68989a7
JH
1556 if (blank_boundary)
1557 memset(hex, ' ', length);
1558 else if (!cmd_is_annotate) {
4c10a5ca
JH
1559 length--;
1560 putchar('^');
1561 }
b11121d9 1562 }
cee7f245 1563
b11121d9 1564 printf("%.*s", length, hex);
cee7f245
JH
1565 if (opt & OUTPUT_ANNOTATE_COMPAT)
1566 printf("\t(%10s\t%10s\t%d)", ci.author,
1567 format_time(ci.author_time, ci.author_tz,
1568 show_raw_time),
1569 ent->lno + 1 + cnt);
1570 else {
5ff62c30 1571 if (opt & OUTPUT_SHOW_SCORE)
54a4c617
JH
1572 printf(" %*d %02d",
1573 max_score_digits, ent->score,
1574 ent->suspect->refcnt);
cee7f245
JH
1575 if (opt & OUTPUT_SHOW_NAME)
1576 printf(" %-*.*s", longest_file, longest_file,
1577 suspect->path);
1578 if (opt & OUTPUT_SHOW_NUMBER)
1579 printf(" %*d", max_orig_digits,
1580 ent->s_lno + 1 + cnt);
1581 printf(" (%-*.*s %10s %*d) ",
1582 longest_author, longest_author, ci.author,
1583 format_time(ci.author_time, ci.author_tz,
1584 show_raw_time),
1585 max_digits, ent->lno + 1 + cnt);
1586 }
1587 do {
1588 ch = *cp++;
1589 putchar(ch);
1590 } while (ch != '\n' &&
1591 cp < sb->final_buf + sb->final_buf_size);
1592 }
1593}
1594
1595static void output(struct scoreboard *sb, int option)
1596{
1597 struct blame_entry *ent;
1598
1599 if (option & OUTPUT_PORCELAIN) {
1600 for (ent = sb->ent; ent; ent = ent->next) {
1601 struct blame_entry *oth;
1602 struct origin *suspect = ent->suspect;
1603 struct commit *commit = suspect->commit;
1604 if (commit->object.flags & MORE_THAN_ONE_PATH)
1605 continue;
1606 for (oth = ent->next; oth; oth = oth->next) {
1607 if ((oth->suspect->commit != commit) ||
1608 !strcmp(oth->suspect->path, suspect->path))
1609 continue;
1610 commit->object.flags |= MORE_THAN_ONE_PATH;
1611 break;
1612 }
1613 }
1614 }
1615
1616 for (ent = sb->ent; ent; ent = ent->next) {
1617 if (option & OUTPUT_PORCELAIN)
1618 emit_porcelain(sb, ent);
5ff62c30 1619 else {
cee7f245 1620 emit_other(sb, ent, option);
5ff62c30 1621 }
cee7f245
JH
1622 }
1623}
1624
1732a1fd
JH
1625/*
1626 * To allow quick access to the contents of nth line in the
1627 * final image, prepare an index in the scoreboard.
1628 */
cee7f245
JH
1629static int prepare_lines(struct scoreboard *sb)
1630{
1631 const char *buf = sb->final_buf;
1632 unsigned long len = sb->final_buf_size;
1633 int num = 0, incomplete = 0, bol = 1;
1634
1635 if (len && buf[len-1] != '\n')
1636 incomplete++; /* incomplete line at the end */
1637 while (len--) {
1638 if (bol) {
1639 sb->lineno = xrealloc(sb->lineno,
1640 sizeof(int* ) * (num + 1));
1641 sb->lineno[num] = buf - sb->final_buf;
1642 bol = 0;
1643 }
1644 if (*buf++ == '\n') {
1645 num++;
1646 bol = 1;
1647 }
1648 }
1ca6ca87
JH
1649 sb->lineno = xrealloc(sb->lineno,
1650 sizeof(int* ) * (num + incomplete + 1));
1651 sb->lineno[num + incomplete] = buf - sb->final_buf;
cee7f245
JH
1652 sb->num_lines = num + incomplete;
1653 return sb->num_lines;
1654}
1655
1732a1fd
JH
1656/*
1657 * Add phony grafts for use with -S; this is primarily to
1658 * support git-cvsserver that wants to give a linear history
1659 * to its clients.
1660 */
cee7f245
JH
1661static int read_ancestry(const char *graft_file)
1662{
1663 FILE *fp = fopen(graft_file, "r");
1664 char buf[1024];
1665 if (!fp)
1666 return -1;
1667 while (fgets(buf, sizeof(buf), fp)) {
1668 /* The format is just "Commit Parent1 Parent2 ...\n" */
1669 int len = strlen(buf);
1670 struct commit_graft *graft = read_graft_line(buf, len);
8eaf7986
JH
1671 if (graft)
1672 register_commit_graft(graft, 0);
cee7f245
JH
1673 }
1674 fclose(fp);
1675 return 0;
1676}
1677
1732a1fd
JH
1678/*
1679 * How many columns do we need to show line numbers in decimal?
1680 */
cee7f245
JH
1681static int lineno_width(int lines)
1682{
1683 int i, width;
1684
1685 for (width = 1, i = 10; i <= lines + 1; width++)
1686 i *= 10;
1687 return width;
1688}
1689
1732a1fd
JH
1690/*
1691 * How many columns do we need to show line numbers, authors,
1692 * and filenames?
1693 */
cee7f245
JH
1694static void find_alignment(struct scoreboard *sb, int *option)
1695{
1696 int longest_src_lines = 0;
1697 int longest_dst_lines = 0;
5ff62c30 1698 unsigned largest_score = 0;
cee7f245
JH
1699 struct blame_entry *e;
1700
1701 for (e = sb->ent; e; e = e->next) {
1702 struct origin *suspect = e->suspect;
1703 struct commit_info ci;
1704 int num;
1705
ab3bb800
JH
1706 if (strcmp(suspect->path, sb->path))
1707 *option |= OUTPUT_SHOW_NAME;
1708 num = strlen(suspect->path);
1709 if (longest_file < num)
1710 longest_file = num;
cee7f245
JH
1711 if (!(suspect->commit->object.flags & METAINFO_SHOWN)) {
1712 suspect->commit->object.flags |= METAINFO_SHOWN;
1713 get_commit_info(suspect->commit, &ci, 1);
cee7f245
JH
1714 num = strlen(ci.author);
1715 if (longest_author < num)
1716 longest_author = num;
1717 }
1718 num = e->s_lno + e->num_lines;
1719 if (longest_src_lines < num)
1720 longest_src_lines = num;
1721 num = e->lno + e->num_lines;
1722 if (longest_dst_lines < num)
1723 longest_dst_lines = num;
5ff62c30
JH
1724 if (largest_score < ent_score(sb, e))
1725 largest_score = ent_score(sb, e);
cee7f245
JH
1726 }
1727 max_orig_digits = lineno_width(longest_src_lines);
1728 max_digits = lineno_width(longest_dst_lines);
5ff62c30 1729 max_score_digits = lineno_width(largest_score);
cee7f245
JH
1730}
1731
1732a1fd
JH
1732/*
1733 * For debugging -- origin is refcounted, and this asserts that
1734 * we do not underflow.
1735 */
54a4c617
JH
1736static void sanity_check_refcnt(struct scoreboard *sb)
1737{
1738 int baa = 0;
1739 struct blame_entry *ent;
1740
1741 for (ent = sb->ent; ent; ent = ent->next) {
ae86ad65 1742 /* Nobody should have zero or negative refcnt */
854b97f6
JH
1743 if (ent->suspect->refcnt <= 0) {
1744 fprintf(stderr, "%s in %s has negative refcnt %d\n",
1745 ent->suspect->path,
1746 sha1_to_hex(ent->suspect->commit->object.sha1),
1747 ent->suspect->refcnt);
ae86ad65 1748 baa = 1;
854b97f6 1749 }
ae86ad65
JH
1750 }
1751 for (ent = sb->ent; ent; ent = ent->next) {
1752 /* Mark the ones that haven't been checked */
54a4c617
JH
1753 if (0 < ent->suspect->refcnt)
1754 ent->suspect->refcnt = -ent->suspect->refcnt;
54a4c617
JH
1755 }
1756 for (ent = sb->ent; ent; ent = ent->next) {
1732a1fd
JH
1757 /*
1758 * ... then pick each and see if they have the the
1759 * correct refcnt.
54a4c617
JH
1760 */
1761 int found;
1762 struct blame_entry *e;
1763 struct origin *suspect = ent->suspect;
1764
1765 if (0 < suspect->refcnt)
1766 continue;
ae86ad65 1767 suspect->refcnt = -suspect->refcnt; /* Unmark */
54a4c617
JH
1768 for (found = 0, e = sb->ent; e; e = e->next) {
1769 if (e->suspect != suspect)
1770 continue;
1771 found++;
1772 }
854b97f6
JH
1773 if (suspect->refcnt != found) {
1774 fprintf(stderr, "%s in %s has refcnt %d, not %d\n",
1775 ent->suspect->path,
1776 sha1_to_hex(ent->suspect->commit->object.sha1),
1777 ent->suspect->refcnt, found);
1778 baa = 2;
1779 }
54a4c617
JH
1780 }
1781 if (baa) {
1782 int opt = 0160;
1783 find_alignment(sb, &opt);
1784 output(sb, opt);
854b97f6 1785 die("Baa %d!", baa);
54a4c617
JH
1786 }
1787}
1788
1732a1fd
JH
1789/*
1790 * Used for the command line parsing; check if the path exists
1791 * in the working tree.
1792 */
cee7f245
JH
1793static int has_path_in_work_tree(const char *path)
1794{
1795 struct stat st;
1796 return !lstat(path, &st);
1797}
1798
4a0fc95f
JH
1799static unsigned parse_score(const char *arg)
1800{
1801 char *end;
1802 unsigned long score = strtoul(arg, &end, 10);
1803 if (*end)
1804 return 0;
1805 return score;
1806}
1807
20239bae
JK
1808static const char *add_prefix(const char *prefix, const char *path)
1809{
1810 if (!prefix || !prefix[0])
1811 return path;
1812 return prefix_path(prefix, strlen(prefix), path);
1813}
1814
1732a1fd
JH
1815/*
1816 * Parsing of (comma separated) one item in the -L option
1817 */
931233bc
JH
1818static const char *parse_loc(const char *spec,
1819 struct scoreboard *sb, long lno,
1820 long begin, long *ret)
1821{
1822 char *term;
1823 const char *line;
1824 long num;
1825 int reg_error;
1826 regex_t regexp;
1827 regmatch_t match[1];
1828
7bd9641d
JH
1829 /* Allow "-L <something>,+20" to mean starting at <something>
1830 * for 20 lines, or "-L <something>,-5" for 5 lines ending at
1831 * <something>.
1832 */
1833 if (1 < begin && (spec[0] == '+' || spec[0] == '-')) {
1834 num = strtol(spec + 1, &term, 10);
1835 if (term != spec + 1) {
1836 if (spec[0] == '-')
1837 num = 0 - num;
1838 if (0 < num)
1839 *ret = begin + num - 2;
1840 else if (!num)
1841 *ret = begin;
1842 else
1843 *ret = begin + num;
1844 return term;
1845 }
1846 return spec;
1847 }
931233bc
JH
1848 num = strtol(spec, &term, 10);
1849 if (term != spec) {
1850 *ret = num;
1851 return term;
1852 }
1853 if (spec[0] != '/')
1854 return spec;
1855
1856 /* it could be a regexp of form /.../ */
1857 for (term = (char*) spec + 1; *term && *term != '/'; term++) {
1858 if (*term == '\\')
1859 term++;
1860 }
1861 if (*term != '/')
1862 return spec;
1863
1864 /* try [spec+1 .. term-1] as regexp */
1865 *term = 0;
1866 begin--; /* input is in human terms */
1867 line = nth_line(sb, begin);
1868
1869 if (!(reg_error = regcomp(&regexp, spec + 1, REG_NEWLINE)) &&
1870 !(reg_error = regexec(&regexp, line, 1, match, 0))) {
1871 const char *cp = line + match[0].rm_so;
1872 const char *nline;
1873
1874 while (begin++ < lno) {
1875 nline = nth_line(sb, begin);
1876 if (line <= cp && cp < nline)
1877 break;
1878 line = nline;
1879 }
1880 *ret = begin;
1881 regfree(&regexp);
1882 *term++ = '/';
1883 return term;
1884 }
1885 else {
1886 char errbuf[1024];
1887 regerror(reg_error, &regexp, errbuf, 1024);
1888 die("-L parameter '%s': %s", spec + 1, errbuf);
1889 }
1890}
1891
1732a1fd
JH
1892/*
1893 * Parsing of -L option
1894 */
931233bc
JH
1895static void prepare_blame_range(struct scoreboard *sb,
1896 const char *bottomtop,
1897 long lno,
1898 long *bottom, long *top)
1899{
1900 const char *term;
1901
1902 term = parse_loc(bottomtop, sb, lno, 1, bottom);
1903 if (*term == ',') {
1904 term = parse_loc(term + 1, sb, lno, *bottom + 1, top);
1905 if (*term)
acca687f 1906 usage(blame_usage);
931233bc
JH
1907 }
1908 if (*term)
acca687f 1909 usage(blame_usage);
931233bc
JH
1910}
1911
4c10a5ca
JH
1912static int git_blame_config(const char *var, const char *value)
1913{
1914 if (!strcmp(var, "blame.showroot")) {
1915 show_root = git_config_bool(var, value);
1916 return 0;
1917 }
1918 if (!strcmp(var, "blame.blankboundary")) {
1919 blank_boundary = git_config_bool(var, value);
1920 return 0;
1921 }
1922 return git_default_config(var, value);
1923}
1924
1cfe7733
JH
1925static struct commit *fake_working_tree_commit(const char *path, const char *contents_from)
1926{
1927 struct commit *commit;
1928 struct origin *origin;
1929 unsigned char head_sha1[20];
1930 char *buf;
1931 const char *ident;
1932 int fd;
1933 time_t now;
1934 unsigned long fin_size;
1935 int size, len;
1936 struct cache_entry *ce;
1937 unsigned mode;
1938
1939 if (get_sha1("HEAD", head_sha1))
1940 die("No such ref: HEAD");
1941
1942 time(&now);
1943 commit = xcalloc(1, sizeof(*commit));
1944 commit->parents = xcalloc(1, sizeof(*commit->parents));
1945 commit->parents->item = lookup_commit_reference(head_sha1);
1946 commit->object.parsed = 1;
1947 commit->date = now;
1948 commit->object.type = OBJ_COMMIT;
1949
1950 origin = make_origin(commit, path);
1951
1952 if (!contents_from || strcmp("-", contents_from)) {
1953 struct stat st;
1954 const char *read_from;
1955
1956 if (contents_from) {
1957 if (stat(contents_from, &st) < 0)
1958 die("Cannot stat %s", contents_from);
1959 read_from = contents_from;
1960 }
1961 else {
1962 if (lstat(path, &st) < 0)
1963 die("Cannot lstat %s", path);
1964 read_from = path;
1965 }
1966 fin_size = st.st_size;
1967 buf = xmalloc(fin_size+1);
1968 mode = canon_mode(st.st_mode);
1969 switch (st.st_mode & S_IFMT) {
1970 case S_IFREG:
1971 fd = open(read_from, O_RDONLY);
1972 if (fd < 0)
1973 die("cannot open %s", read_from);
1974 if (read_in_full(fd, buf, fin_size) != fin_size)
1975 die("cannot read %s", read_from);
1976 break;
1977 case S_IFLNK:
1978 if (readlink(read_from, buf, fin_size+1) != fin_size)
1979 die("cannot readlink %s", read_from);
1980 break;
1981 default:
1982 die("unsupported file type %s", read_from);
1983 }
1984 }
1985 else {
1986 /* Reading from stdin */
1987 contents_from = "standard input";
1988 buf = NULL;
1989 fin_size = 0;
1990 mode = 0;
1991 while (1) {
1992 ssize_t cnt = 8192;
1993 buf = xrealloc(buf, fin_size + cnt);
1994 cnt = xread(0, buf + fin_size, cnt);
1995 if (cnt < 0)
1996 die("read error %s from stdin",
1997 strerror(errno));
1998 if (!cnt)
1999 break;
2000 fin_size += cnt;
2001 }
2002 buf = xrealloc(buf, fin_size + 1);
2003 }
2004 buf[fin_size] = 0;
2005 origin->file.ptr = buf;
2006 origin->file.size = fin_size;
21666f1a 2007 pretend_sha1_file(buf, fin_size, OBJ_BLOB, origin->blob_sha1);
1cfe7733
JH
2008 commit->util = origin;
2009
2010 /*
2011 * Read the current index, replace the path entry with
2012 * origin->blob_sha1 without mucking with its mode or type
2013 * bits; we are not going to write this index out -- we just
2014 * want to run "diff-index --cached".
2015 */
2016 discard_cache();
2017 read_cache();
2018
2019 len = strlen(path);
2020 if (!mode) {
2021 int pos = cache_name_pos(path, len);
2022 if (0 <= pos)
2023 mode = ntohl(active_cache[pos]->ce_mode);
2024 else
2025 /* Let's not bother reading from HEAD tree */
2026 mode = S_IFREG | 0644;
2027 }
2028 size = cache_entry_size(len);
2029 ce = xcalloc(1, size);
2030 hashcpy(ce->sha1, origin->blob_sha1);
2031 memcpy(ce->name, path, len);
2032 ce->ce_flags = create_ce_flags(len, 0);
2033 ce->ce_mode = create_ce_mode(mode);
2034 add_cache_entry(ce, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
2035
2036 /*
2037 * We are not going to write this out, so this does not matter
2038 * right now, but someday we might optimize diff-index --cached
2039 * with cache-tree information.
2040 */
2041 cache_tree_invalidate_path(active_cache_tree, path);
2042
2043 commit->buffer = xmalloc(400);
2044 ident = fmt_ident("Not Committed Yet", "not.committed.yet", NULL, 0);
2045 sprintf(commit->buffer,
2046 "tree 0000000000000000000000000000000000000000\n"
2047 "parent %s\n"
2048 "author %s\n"
2049 "committer %s\n\n"
2050 "Version of %s from %s\n",
2051 sha1_to_hex(head_sha1),
2052 ident, ident, path, contents_from ? contents_from : path);
2053 return commit;
2054}
2055
acca687f 2056int cmd_blame(int argc, const char **argv, const char *prefix)
cee7f245
JH
2057{
2058 struct rev_info revs;
2059 const char *path;
2060 struct scoreboard sb;
2061 struct origin *o;
2062 struct blame_entry *ent;
d24bba80 2063 int i, seen_dashdash, unk, opt;
cee7f245
JH
2064 long bottom, top, lno;
2065 int output_option = 0;
870b39c1 2066 int show_stats = 0;
cee7f245
JH
2067 const char *revs_file = NULL;
2068 const char *final_commit_name = NULL;
21666f1a 2069 enum object_type type;
931233bc 2070 const char *bottomtop = NULL;
1cfe7733 2071 const char *contents_from = NULL;
cee7f245 2072
e68989a7
JH
2073 cmd_is_annotate = !strcmp(argv[0], "annotate");
2074
4c10a5ca 2075 git_config(git_blame_config);
612702e8
JH
2076 save_commit_buffer = 0;
2077
d24bba80 2078 opt = 0;
cee7f245
JH
2079 seen_dashdash = 0;
2080 for (unk = i = 1; i < argc; i++) {
2081 const char *arg = argv[i];
2082 if (*arg != '-')
2083 break;
4c10a5ca
JH
2084 else if (!strcmp("-b", arg))
2085 blank_boundary = 1;
2086 else if (!strcmp("--root", arg))
2087 show_root = 1;
870b39c1
JH
2088 else if (!strcmp(arg, "--show-stats"))
2089 show_stats = 1;
cee7f245
JH
2090 else if (!strcmp("-c", arg))
2091 output_option |= OUTPUT_ANNOTATE_COMPAT;
2092 else if (!strcmp("-t", arg))
2093 output_option |= OUTPUT_RAW_TIMESTAMP;
2094 else if (!strcmp("-l", arg))
2095 output_option |= OUTPUT_LONG_OBJECT_NAME;
2096 else if (!strcmp("-S", arg) && ++i < argc)
2097 revs_file = argv[i];
599065a3 2098 else if (!prefixcmp(arg, "-M")) {
d24bba80 2099 opt |= PICKAXE_BLAME_MOVE;
4a0fc95f
JH
2100 blame_move_score = parse_score(arg+2);
2101 }
599065a3 2102 else if (!prefixcmp(arg, "-C")) {
18abd745
JH
2103 if (opt & PICKAXE_BLAME_COPY)
2104 opt |= PICKAXE_BLAME_COPY_HARDER;
2105 opt |= PICKAXE_BLAME_COPY | PICKAXE_BLAME_MOVE;
4a0fc95f 2106 blame_copy_score = parse_score(arg+2);
18abd745 2107 }
599065a3 2108 else if (!prefixcmp(arg, "-L")) {
2c40f984
JH
2109 if (!arg[2]) {
2110 if (++i >= argc)
acca687f 2111 usage(blame_usage);
2c40f984
JH
2112 arg = argv[i];
2113 }
2114 else
2115 arg += 2;
931233bc 2116 if (bottomtop)
cee7f245 2117 die("More than one '-L n,m' option given");
931233bc 2118 bottomtop = arg;
cee7f245 2119 }
1cfe7733
JH
2120 else if (!strcmp("--contents", arg)) {
2121 if (++i >= argc)
2122 usage(blame_usage);
2123 contents_from = argv[i];
2124 }
717d1462
LT
2125 else if (!strcmp("--incremental", arg))
2126 incremental = 1;
5ff62c30
JH
2127 else if (!strcmp("--score-debug", arg))
2128 output_option |= OUTPUT_SHOW_SCORE;
cee7f245
JH
2129 else if (!strcmp("-f", arg) ||
2130 !strcmp("--show-name", arg))
2131 output_option |= OUTPUT_SHOW_NAME;
2132 else if (!strcmp("-n", arg) ||
2133 !strcmp("--show-number", arg))
2134 output_option |= OUTPUT_SHOW_NUMBER;
2135 else if (!strcmp("-p", arg) ||
2136 !strcmp("--porcelain", arg))
2137 output_option |= OUTPUT_PORCELAIN;
2138 else if (!strcmp("--", arg)) {
2139 seen_dashdash = 1;
2140 i++;
2141 break;
2142 }
2143 else
2144 argv[unk++] = arg;
2145 }
2146
4f0219a4
RS
2147 if (!incremental)
2148 setup_pager();
2149
4a0fc95f
JH
2150 if (!blame_move_score)
2151 blame_move_score = BLAME_DEFAULT_MOVE_SCORE;
2152 if (!blame_copy_score)
2153 blame_copy_score = BLAME_DEFAULT_COPY_SCORE;
2154
1732a1fd
JH
2155 /*
2156 * We have collected options unknown to us in argv[1..unk]
cee7f245 2157 * which are to be passed to revision machinery if we are
3dff5379 2158 * going to do the "bottom" processing.
cee7f245
JH
2159 *
2160 * The remaining are:
2161 *
2162 * (1) if seen_dashdash, its either
2163 * "-options -- <path>" or
2164 * "-options -- <path> <rev>".
2165 * but the latter is allowed only if there is no
2166 * options that we passed to revision machinery.
2167 *
2168 * (2) otherwise, we may have "--" somewhere later and
2169 * might be looking at the first one of multiple 'rev'
2170 * parameters (e.g. " master ^next ^maint -- path").
2171 * See if there is a dashdash first, and give the
2172 * arguments before that to revision machinery.
2173 * After that there must be one 'path'.
2174 *
2175 * (3) otherwise, its one of the three:
2176 * "-options <path> <rev>"
2177 * "-options <rev> <path>"
2178 * "-options <path>"
2179 * but again the first one is allowed only if
2180 * there is no options that we passed to revision
2181 * machinery.
2182 */
2183
2184 if (seen_dashdash) {
2185 /* (1) */
2186 if (argc <= i)
acca687f 2187 usage(blame_usage);
20239bae 2188 path = add_prefix(prefix, argv[i]);
cee7f245
JH
2189 if (i + 1 == argc - 1) {
2190 if (unk != 1)
acca687f 2191 usage(blame_usage);
cee7f245
JH
2192 argv[unk++] = argv[i + 1];
2193 }
2194 else if (i + 1 != argc)
2195 /* garbage at end */
acca687f 2196 usage(blame_usage);
cee7f245
JH
2197 }
2198 else {
2199 int j;
2200 for (j = i; !seen_dashdash && j < argc; j++)
2201 if (!strcmp(argv[j], "--"))
2202 seen_dashdash = j;
2203 if (seen_dashdash) {
f4421325 2204 /* (2) */
cee7f245 2205 if (seen_dashdash + 1 != argc - 1)
acca687f 2206 usage(blame_usage);
20239bae 2207 path = add_prefix(prefix, argv[seen_dashdash + 1]);
cee7f245
JH
2208 for (j = i; j < seen_dashdash; j++)
2209 argv[unk++] = argv[j];
2210 }
2211 else {
2212 /* (3) */
f4421325
TK
2213 if (argc <= i)
2214 usage(blame_usage);
20239bae 2215 path = add_prefix(prefix, argv[i]);
cee7f245
JH
2216 if (i + 1 == argc - 1) {
2217 final_commit_name = argv[i + 1];
2218
2219 /* if (unk == 1) we could be getting
2220 * old-style
2221 */
2222 if (unk == 1 && !has_path_in_work_tree(path)) {
20239bae 2223 path = add_prefix(prefix, argv[i + 1]);
cee7f245
JH
2224 final_commit_name = argv[i];
2225 }
2226 }
2227 else if (i != argc - 1)
acca687f 2228 usage(blame_usage); /* garbage at end */
cee7f245
JH
2229
2230 if (!has_path_in_work_tree(path))
2231 die("cannot stat path %s: %s",
2232 path, strerror(errno));
2233 }
2234 }
2235
2236 if (final_commit_name)
2237 argv[unk++] = final_commit_name;
2238
1732a1fd
JH
2239 /*
2240 * Now we got rev and path. We do not want the path pruning
cee7f245
JH
2241 * but we may want "bottom" processing.
2242 */
6bee4e40 2243 argv[unk++] = "--"; /* terminate the rev name */
cee7f245
JH
2244 argv[unk] = NULL;
2245
2246 init_revisions(&revs, NULL);
1cfe7733 2247 setup_revisions(unk, argv, &revs, NULL);
cee7f245
JH
2248 memset(&sb, 0, sizeof(sb));
2249
1732a1fd
JH
2250 /*
2251 * There must be one and only one positive commit in the
cee7f245
JH
2252 * revs->pending array.
2253 */
2254 for (i = 0; i < revs.pending.nr; i++) {
2255 struct object *obj = revs.pending.objects[i].item;
2256 if (obj->flags & UNINTERESTING)
2257 continue;
2258 while (obj->type == OBJ_TAG)
2259 obj = deref_tag(obj, NULL, 0);
2260 if (obj->type != OBJ_COMMIT)
2261 die("Non commit %s?",
2262 revs.pending.objects[i].name);
2263 if (sb.final)
2264 die("More than one commit to dig from %s and %s?",
2265 revs.pending.objects[i].name,
2266 final_commit_name);
2267 sb.final = (struct commit *) obj;
2268 final_commit_name = revs.pending.objects[i].name;
2269 }
2270
2271 if (!sb.final) {
1732a1fd
JH
2272 /*
2273 * "--not A B -- path" without anything positive;
1cfe7733
JH
2274 * do not default to HEAD, but use the working tree
2275 * or "--contents".
1732a1fd 2276 */
1cfe7733
JH
2277 sb.final = fake_working_tree_commit(path, contents_from);
2278 add_pending_object(&revs, &(sb.final->object), ":");
cee7f245 2279 }
1cfe7733
JH
2280 else if (contents_from)
2281 die("Cannot use --contents with final commit object name");
cee7f245 2282
1732a1fd
JH
2283 /*
2284 * If we have bottom, this will mark the ancestors of the
cee7f245
JH
2285 * bottom commits we would reach while traversing as
2286 * uninteresting.
2287 */
2288 prepare_revision_walk(&revs);
2289
1cfe7733
JH
2290 if (is_null_sha1(sb.final->object.sha1)) {
2291 char *buf;
2292 o = sb.final->util;
2293 buf = xmalloc(o->file.size + 1);
2294 memcpy(buf, o->file.ptr, o->file.size + 1);
2295 sb.final_buf = buf;
2296 sb.final_buf_size = o->file.size;
2297 }
2298 else {
2299 o = get_origin(&sb, sb.final, path);
2300 if (fill_blob_sha1(o))
2301 die("no such path %s in %s", path, final_commit_name);
cee7f245 2302
21666f1a 2303 sb.final_buf = read_sha1_file(o->blob_sha1, &type,
1cfe7733
JH
2304 &sb.final_buf_size);
2305 }
c2e525d9 2306 num_read_blob++;
cee7f245
JH
2307 lno = prepare_lines(&sb);
2308
931233bc
JH
2309 bottom = top = 0;
2310 if (bottomtop)
2311 prepare_blame_range(&sb, bottomtop, lno, &bottom, &top);
2312 if (bottom && top && top < bottom) {
2313 long tmp;
2314 tmp = top; top = bottom; bottom = tmp;
2315 }
cee7f245
JH
2316 if (bottom < 1)
2317 bottom = 1;
2318 if (top < 1)
2319 top = lno;
2320 bottom--;
2321 if (lno < top)
2322 die("file %s has only %lu lines", path, lno);
2323
2324 ent = xcalloc(1, sizeof(*ent));
2325 ent->lno = bottom;
2326 ent->num_lines = top - bottom;
2327 ent->suspect = o;
2328 ent->s_lno = bottom;
2329
2330 sb.ent = ent;
2331 sb.path = path;
2332
2333 if (revs_file && read_ancestry(revs_file))
2334 die("reading graft file %s failed: %s",
2335 revs_file, strerror(errno));
2336
d24bba80 2337 assign_blame(&sb, &revs, opt);
cee7f245 2338
717d1462
LT
2339 if (incremental)
2340 return 0;
2341
cee7f245
JH
2342 coalesce(&sb);
2343
2344 if (!(output_option & OUTPUT_PORCELAIN))
2345 find_alignment(&sb, &output_option);
2346
2347 output(&sb, output_option);
2348 free((void *)sb.final_buf);
2349 for (ent = sb.ent; ent; ) {
2350 struct blame_entry *e = ent->next;
2351 free(ent);
2352 ent = e;
2353 }
c2e525d9 2354
870b39c1 2355 if (show_stats) {
c2e525d9
JH
2356 printf("num read blob: %d\n", num_read_blob);
2357 printf("num get patch: %d\n", num_get_patch);
2358 printf("num commits: %d\n", num_commits);
2359 }
cee7f245
JH
2360 return 0;
2361}