2 * Copyright (c) 2018, 2019, 2020 Stefan Sperling <stsp@openbsd.org>
3 * Copyright (c) 2020 Neels Hofmeyr <neels@hofmeyr.de>
5 * Permission to use, copy, modify, and distribute this software for any
6 * purpose with or without fee is hereby granted, provided that the above
7 * copyright notice and this permission notice appear in all copies.
9 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
29 #include "got_compat.h"
31 #include "got_error.h"
32 #include "got_object.h"
33 #include "got_cancel.h"
34 #include "got_blame.h"
35 #include "got_commit_graph.h"
36 #include "got_opentemp.h"
38 #include "got_lib_inflate.h"
39 #include "got_lib_delta.h"
40 #include "got_lib_object.h"
41 #include "got_lib_diff.h"
44 #define MAX(_a,_b) ((_a) > (_b) ? (_a) : (_b))
47 struct got_blame_line {
49 struct got_object_id id;
53 struct diff_config *cfg;
54 int nlines; /* number of lines in file being blamed */
55 int nannotated; /* number of lines already annotated */
56 struct got_blame_line *lines; /* one per line */
60 * These change with every traversed commit. After diffing
61 * commits N:N-1, in preparation for diffing commits N-1:N-2,
62 * data for commit N is retained and flipped into data for N-1.
65 FILE *f1; /* older version from commit N-1. */
66 FILE *f2; /* newer version from commit N. */
77 * Map line numbers of an older version of the file to valid line
78 * numbers in the version of the file being blamed. This map is
79 * updated with each commit we traverse throughout the file's history.
80 * Lines mapped to -1 do not correspond to any line in the version
86 struct diff_data *data1;
87 struct diff_data *data2;
90 static const struct got_error *
91 annotate_line(struct got_blame *blame, int lineno, struct got_object_id *id,
92 const struct got_error *(*cb)(void *, int, int, struct got_object_id *),
95 const struct got_error *err = NULL;
96 struct got_blame_line *line;
98 if (lineno < 0 || lineno >= blame->nlines)
101 line = &blame->lines[lineno];
105 memcpy(&line->id, id, sizeof(line->id));
109 err = cb(arg, blame->nlines, lineno + 1, id);
113 static const struct got_error *
114 blame_changes(struct got_blame *blame, struct diff_result *diff_result,
115 struct got_object_id *commit_id,
116 const struct got_error *(*cb)(void *, int, int, struct got_object_id *),
119 const struct got_error *err = NULL;
121 int idx1 = 0, idx2 = 0;
123 for (i = 0; i < diff_result->chunks.len &&
124 blame->nannotated < blame->nlines; i++) {
125 struct diff_chunk *c = diff_chunk_get(diff_result, i);
126 unsigned int left_count, right_count;
130 * We do not need to worry about idx1/idx2 growing out
131 * of bounds because the diff implementation ensures
132 * that chunk ranges never exceed the number of lines
133 * in the left/right input files.
135 left_count = diff_chunk_get_left_count(c);
136 right_count = diff_chunk_get_right_count(c);
138 if (left_count == right_count) {
139 for (j = 0; j < left_count; j++) {
140 blame->linemap1[idx1++] =
141 blame->linemap2[idx2++];
146 if (right_count == 0) {
147 for (j = 0; j < left_count; j++) {
148 blame->linemap1[idx1++] = -1;
153 for (j = 0; j < right_count; j++) {
154 int ln = blame->linemap2[idx2++];
155 err = annotate_line(blame, ln, commit_id, cb, arg);
158 if (blame->nlines == blame->nannotated)
166 static const struct got_error *
167 blame_prepare_file(FILE *f, unsigned char **p, off_t *size,
168 int *nlines, off_t **line_offsets, struct diff_data *diff_data,
169 const struct diff_config *cfg, struct got_blob_object *blob)
171 const struct got_error *err = NULL;
172 int diff_flags = 0, rc;
174 err = got_object_blob_dump_to_file(size, nlines, line_offsets,
179 #ifndef GOT_DIFF_NO_MMAP
180 *p = mmap(NULL, *size, PROT_READ, MAP_PRIVATE, fileno(f), 0);
181 if (*p == MAP_FAILED)
183 *p = NULL; /* fall back on file I/O */
185 /* Allow blaming lines in binary files even though it's useless. */
186 diff_flags |= DIFF_FLAG_FORCE_TEXT_DATA;
188 rc = diff_atomize_file(diff_data, cfg, f, *p, *size, diff_flags);
190 return got_error_set_errno(rc, "diff_atomize_file");
195 static const struct got_error *
196 blame_commit(struct got_blame *blame, struct got_object_id *id,
197 const char *path, struct got_repository *repo,
198 const struct got_error *(*cb)(void *, int, int, struct got_object_id *),
201 const struct got_error *err = NULL;
202 struct got_commit_object *commit = NULL;
203 struct got_object_qid *pid = NULL;
204 struct got_object_id *pblob_id = NULL;
205 struct got_blob_object *pblob = NULL;
206 struct diff_result *diff_result = NULL;
208 err = got_object_open_as_commit(&commit, repo, id);
212 pid = STAILQ_FIRST(got_object_commit_get_parent_ids(commit));
214 got_object_commit_close(commit);
218 err = got_object_id_by_path(&pblob_id, repo, pid->id, path);
220 if (err->code == GOT_ERR_NO_TREE_ENTRY)
225 err = got_object_open_as_blob(&pblob, repo, pblob_id, 8192);
229 blame->f1 = got_opentemp();
230 if (blame->f1 == NULL) {
231 err = got_error_from_errno("got_opentemp");
235 err = blame_prepare_file(blame->f1, &blame->map1, &blame->size1,
236 &blame->nlines1, &blame->line_offsets1, blame->data1,
241 diff_result = diff_main(blame->cfg, blame->data1, blame->data2);
242 if (diff_result == NULL) {
243 err = got_error_set_errno(ENOMEM, "malloc");
246 if (diff_result->rc != DIFF_RC_OK) {
247 err = got_error_set_errno(diff_result->rc, "diff");
250 if (diff_result->chunks.len > 0) {
251 if (blame->nlines1 > 0) {
252 blame->linemap1 = calloc(blame->nlines1,
253 sizeof(*blame->linemap1));
254 if (blame->linemap1 == NULL) {
255 err = got_error_from_errno("malloc");
259 err = blame_changes(blame, diff_result, id, cb, arg);
263 err = cb(arg, blame->nlines, -1, id);
266 diff_result_free(diff_result);
268 got_object_commit_close(commit);
271 got_object_blob_close(pblob);
275 static const struct got_error *
276 blame_close(struct got_blame *blame)
278 const struct got_error *err = NULL;
280 diff_data_free(blame->data1);
282 diff_data_free(blame->data2);
285 if (munmap(blame->map1, blame->size1) == -1 && err == NULL)
286 err = got_error_from_errno("munmap");
289 if (munmap(blame->map2, blame->size2) == -1 && err == NULL)
290 err = got_error_from_errno("munmap");
292 if (blame->f1 && fclose(blame->f1) == EOF && err == NULL)
293 err = got_error_from_errno("fclose");
294 if (blame->f2 && fclose(blame->f2) == EOF && err == NULL)
295 err = got_error_from_errno("fclose");
297 free(blame->line_offsets1);
298 free(blame->line_offsets2);
299 free(blame->linemap1);
300 free(blame->linemap2);
307 atomize_file(struct diff_data *d, FILE *f, off_t filesize, int nlines,
310 int i, rc = DIFF_RC_OK;
311 int embedded_nul = 0;
313 ARRAYLIST_INIT(d->atoms, nlines);
315 for (i = 0; i < nlines; i++) {
316 struct diff_atom *atom;
317 off_t len, pos = line_offsets[i];
318 unsigned int hash = 0;
321 ARRAYLIST_ADD(atom, d->atoms);
328 len = line_offsets[i + 1] - pos;
330 len = filesize - pos;
332 if (fseeko(f, pos, SEEK_SET) == -1) {
336 for (j = 0; j < len; j++) {
340 rc = EIO; /* unexpected EOF */
346 hash = diff_atom_hash_update(hash, (unsigned char)c);
352 *atom = (struct diff_atom){
355 .at = NULL, /* atom data is not memory-mapped */
361 /* File are considered binary if they contain embedded '\0' bytes. */
363 d->atomizer_flags |= DIFF_ATOMIZER_FOUND_BINARY_DATA;
366 ARRAYLIST_FREE(d->atoms);
372 atomize_file_mmap(struct diff_data *d, unsigned char *p,
373 off_t filesize, int nlines, off_t *line_offsets)
375 int i, rc = DIFF_RC_OK;
376 int embedded_nul = 0;
378 ARRAYLIST_INIT(d->atoms, nlines);
380 for (i = 0; i < nlines; i++) {
381 struct diff_atom *atom;
382 off_t len, pos = line_offsets[i];
383 unsigned int hash = 0;
386 ARRAYLIST_ADD(atom, d->atoms);
393 len = line_offsets[i + 1] - pos;
395 len = filesize - pos;
397 for (j = 0; j < len; j++)
398 hash = diff_atom_hash_update(hash, p[pos + j]);
400 if (!embedded_nul && memchr(&p[pos], '\0', len) != NULL)
403 *atom = (struct diff_atom){
412 /* File are considered binary if they contain embedded '\0' bytes. */
414 d->atomizer_flags |= DIFF_ATOMIZER_FOUND_BINARY_DATA;
417 ARRAYLIST_FREE(d->atoms);
422 /* Implements diff_atomize_func_t */
424 blame_atomize_file(void *arg, struct diff_data *d)
426 struct got_blame *blame = arg;
428 if (d->f == blame->f1) {
430 return atomize_file_mmap(d, blame->map1,
431 blame->size1, blame->nlines1,
432 blame->line_offsets1);
434 return atomize_file(d, blame->f1, blame->size1,
435 blame->nlines1, blame->line_offsets1);
436 } else if (d->f == blame->f2) {
437 if (d->atoms.len > 0) {
438 /* Re-use data from previous commit. */
442 return atomize_file_mmap(d, blame->map2,
443 blame->size2, blame->nlines2,
444 blame->line_offsets2);
446 return atomize_file(d, blame->f2, blame->size2,
447 blame->nlines2, blame->line_offsets2);
453 static const struct got_error *
454 close_file2_and_reuse_file1(struct got_blame *blame)
458 free(blame->line_offsets2);
459 blame->line_offsets2 = blame->line_offsets1;
460 blame->line_offsets1 = NULL;
462 free(blame->linemap2);
463 blame->linemap2 = blame->linemap1;
464 blame->linemap1 = NULL;
467 if (munmap(blame->map2, blame->size2) == -1)
468 return got_error_from_errno("munmap");
469 blame->map2 = blame->map1;
473 blame->size2 = blame->size1;
476 if (fclose(blame->f2) == EOF)
477 return got_error_from_errno("fclose");
478 blame->f2 = blame->f1;
481 blame->nlines2 = blame->nlines1;
484 diff_data_free(blame->data2); /* does not free pointer itself */
485 memset(blame->data2, 0, sizeof(*blame->data2));
487 blame->data2 = blame->data1;
493 static const struct got_error *
494 blame_open(struct got_blame **blamep, const char *path,
495 struct got_object_id *start_commit_id, struct got_repository *repo,
496 const struct got_error *(*cb)(void *, int, int, struct got_object_id *),
497 void *arg, got_cancel_cb cancel_cb, void *cancel_arg)
499 const struct got_error *err = NULL;
500 struct got_object_id *obj_id = NULL;
501 struct got_blob_object *blob = NULL;
502 struct got_blame *blame = NULL;
503 struct got_object_id *id = NULL;
505 struct got_commit_graph *graph = NULL;
509 err = got_object_id_by_path(&obj_id, repo, start_commit_id, path);
513 err = got_object_open_as_blob(&blob, repo, obj_id, 8192);
517 blame = calloc(1, sizeof(*blame));
519 err = got_error_from_errno("calloc");
523 blame->data1 = calloc(1, sizeof(*blame->data1));
524 if (blame->data1 == NULL) {
525 err = got_error_from_errno("calloc");
528 blame->data2 = calloc(1, sizeof(*blame->data2));
529 if (blame->data2 == NULL) {
530 err = got_error_from_errno("calloc");
534 blame->f2 = got_opentemp();
535 if (blame->f2 == NULL) {
536 err = got_error_from_errno("got_opentemp");
539 err = got_diff_get_config(&blame->cfg, GOT_DIFF_ALGORITHM_PATIENCE,
540 blame_atomize_file, blame);
544 err = blame_prepare_file(blame->f2, &blame->map2, &blame->size2,
545 &blame->nlines2, &blame->line_offsets2, blame->data2,
547 blame->nlines = blame->nlines2;
548 if (err || blame->nlines == 0)
551 got_object_blob_close(blob);
554 /* Don't include \n at EOF in the blame line count. */
555 if (blame->line_offsets2[blame->nlines - 1] == blame->size2)
558 blame->lines = calloc(blame->nlines, sizeof(*blame->lines));
559 if (blame->lines == NULL) {
560 err = got_error_from_errno("calloc");
564 blame->linemap2 = calloc(blame->nlines2, sizeof(*blame->linemap2));
565 if (blame->linemap2 == NULL) {
566 err = got_error_from_errno("calloc");
569 for (lineno = 0; lineno < blame->nlines2; lineno++)
570 blame->linemap2[lineno] = lineno;
572 err = got_commit_graph_open(&graph, path, 1);
576 err = got_commit_graph_iter_start(graph, start_commit_id, repo,
577 cancel_cb, cancel_arg);
581 struct got_object_id *next_id;
582 err = got_commit_graph_iter_next(&next_id, graph, repo,
583 cancel_cb, cancel_arg);
585 if (err->code == GOT_ERR_ITER_COMPLETED) {
593 err = blame_commit(blame, id, path, repo, cb, arg);
595 if (err->code == GOT_ERR_ITER_COMPLETED)
599 if (blame->nannotated == blame->nlines)
602 err = close_file2_and_reuse_file1(blame);
608 if (id && blame->nannotated < blame->nlines) {
609 /* Annotate remaining non-annotated lines with last commit. */
610 for (lineno = 0; lineno < blame->nlines; lineno++) {
611 err = annotate_line(blame, lineno, id, cb, arg);
619 got_commit_graph_close(graph);
622 got_object_blob_close(blob);
632 const struct got_error *
633 got_blame(const char *path, struct got_object_id *commit_id,
634 struct got_repository *repo,
635 const struct got_error *(*cb)(void *, int, int, struct got_object_id *),
636 void *arg, got_cancel_cb cancel_cb, void* cancel_arg)
638 const struct got_error *err = NULL, *close_err = NULL;
639 struct got_blame *blame;
642 if (asprintf(&abspath, "%s%s", path[0] == '/' ? "" : "/", path) == -1)
643 return got_error_from_errno2("asprintf", path);
645 err = blame_open(&blame, abspath, commit_id, repo, cb, arg,
646 cancel_cb, cancel_arg);
649 close_err = blame_close(blame);
650 return err ? err : close_err;