commit: add repository argument to lookup_commit
[git/git.git] / builtin / fast-export.c
1 /*
2 * "git fast-export" builtin command
3 *
4 * Copyright (C) 2007 Johannes E. Schindelin
5 */
6 #include "builtin.h"
7 #include "cache.h"
8 #include "config.h"
9 #include "refs.h"
10 #include "refspec.h"
11 #include "object-store.h"
12 #include "commit.h"
13 #include "object.h"
14 #include "tag.h"
15 #include "diff.h"
16 #include "diffcore.h"
17 #include "log-tree.h"
18 #include "revision.h"
19 #include "decorate.h"
20 #include "string-list.h"
21 #include "utf8.h"
22 #include "parse-options.h"
23 #include "quote.h"
24 #include "remote.h"
25 #include "blob.h"
26 #include "commit-slab.h"
27
28 static const char *fast_export_usage[] = {
29 N_("git fast-export [rev-list-opts]"),
30 NULL
31 };
32
33 static int progress;
34 static enum { ABORT, VERBATIM, WARN, WARN_STRIP, STRIP } signed_tag_mode = ABORT;
35 static enum { ERROR, DROP, REWRITE } tag_of_filtered_mode = ERROR;
36 static int fake_missing_tagger;
37 static int use_done_feature;
38 static int no_data;
39 static int full_tree;
40 static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
41 static struct refspec refspecs = REFSPEC_INIT_FETCH;
42 static int anonymize;
43 static struct revision_sources revision_sources;
44
45 static int parse_opt_signed_tag_mode(const struct option *opt,
46 const char *arg, int unset)
47 {
48 if (unset || !strcmp(arg, "abort"))
49 signed_tag_mode = ABORT;
50 else if (!strcmp(arg, "verbatim") || !strcmp(arg, "ignore"))
51 signed_tag_mode = VERBATIM;
52 else if (!strcmp(arg, "warn"))
53 signed_tag_mode = WARN;
54 else if (!strcmp(arg, "warn-strip"))
55 signed_tag_mode = WARN_STRIP;
56 else if (!strcmp(arg, "strip"))
57 signed_tag_mode = STRIP;
58 else
59 return error("Unknown signed-tags mode: %s", arg);
60 return 0;
61 }
62
63 static int parse_opt_tag_of_filtered_mode(const struct option *opt,
64 const char *arg, int unset)
65 {
66 if (unset || !strcmp(arg, "abort"))
67 tag_of_filtered_mode = ERROR;
68 else if (!strcmp(arg, "drop"))
69 tag_of_filtered_mode = DROP;
70 else if (!strcmp(arg, "rewrite"))
71 tag_of_filtered_mode = REWRITE;
72 else
73 return error("Unknown tag-of-filtered mode: %s", arg);
74 return 0;
75 }
76
77 static struct decoration idnums;
78 static uint32_t last_idnum;
79
80 static int has_unshown_parent(struct commit *commit)
81 {
82 struct commit_list *parent;
83
84 for (parent = commit->parents; parent; parent = parent->next)
85 if (!(parent->item->object.flags & SHOWN) &&
86 !(parent->item->object.flags & UNINTERESTING))
87 return 1;
88 return 0;
89 }
90
91 struct anonymized_entry {
92 struct hashmap_entry hash;
93 const char *orig;
94 size_t orig_len;
95 const char *anon;
96 size_t anon_len;
97 };
98
99 static int anonymized_entry_cmp(const void *unused_cmp_data,
100 const void *va, const void *vb,
101 const void *unused_keydata)
102 {
103 const struct anonymized_entry *a = va, *b = vb;
104 return a->orig_len != b->orig_len ||
105 memcmp(a->orig, b->orig, a->orig_len);
106 }
107
108 /*
109 * Basically keep a cache of X->Y so that we can repeatedly replace
110 * the same anonymized string with another. The actual generation
111 * is farmed out to the generate function.
112 */
113 static const void *anonymize_mem(struct hashmap *map,
114 void *(*generate)(const void *, size_t *),
115 const void *orig, size_t *len)
116 {
117 struct anonymized_entry key, *ret;
118
119 if (!map->cmpfn)
120 hashmap_init(map, anonymized_entry_cmp, NULL, 0);
121
122 hashmap_entry_init(&key, memhash(orig, *len));
123 key.orig = orig;
124 key.orig_len = *len;
125 ret = hashmap_get(map, &key, NULL);
126
127 if (!ret) {
128 ret = xmalloc(sizeof(*ret));
129 hashmap_entry_init(&ret->hash, key.hash.hash);
130 ret->orig = xstrdup(orig);
131 ret->orig_len = *len;
132 ret->anon = generate(orig, len);
133 ret->anon_len = *len;
134 hashmap_put(map, ret);
135 }
136
137 *len = ret->anon_len;
138 return ret->anon;
139 }
140
141 /*
142 * We anonymize each component of a path individually,
143 * so that paths a/b and a/c will share a common root.
144 * The paths are cached via anonymize_mem so that repeated
145 * lookups for "a" will yield the same value.
146 */
147 static void anonymize_path(struct strbuf *out, const char *path,
148 struct hashmap *map,
149 void *(*generate)(const void *, size_t *))
150 {
151 while (*path) {
152 const char *end_of_component = strchrnul(path, '/');
153 size_t len = end_of_component - path;
154 const char *c = anonymize_mem(map, generate, path, &len);
155 strbuf_add(out, c, len);
156 path = end_of_component;
157 if (*path)
158 strbuf_addch(out, *path++);
159 }
160 }
161
162 static inline void *mark_to_ptr(uint32_t mark)
163 {
164 return (void *)(uintptr_t)mark;
165 }
166
167 static inline uint32_t ptr_to_mark(void * mark)
168 {
169 return (uint32_t)(uintptr_t)mark;
170 }
171
172 static inline void mark_object(struct object *object, uint32_t mark)
173 {
174 add_decoration(&idnums, object, mark_to_ptr(mark));
175 }
176
177 static inline void mark_next_object(struct object *object)
178 {
179 mark_object(object, ++last_idnum);
180 }
181
182 static int get_object_mark(struct object *object)
183 {
184 void *decoration = lookup_decoration(&idnums, object);
185 if (!decoration)
186 return 0;
187 return ptr_to_mark(decoration);
188 }
189
190 static void show_progress(void)
191 {
192 static int counter = 0;
193 if (!progress)
194 return;
195 if ((++counter % progress) == 0)
196 printf("progress %d objects\n", counter);
197 }
198
199 /*
200 * Ideally we would want some transformation of the blob data here
201 * that is unreversible, but would still be the same size and have
202 * the same data relationship to other blobs (so that we get the same
203 * delta and packing behavior as the original). But the first and last
204 * requirements there are probably mutually exclusive, so let's take
205 * the easy way out for now, and just generate arbitrary content.
206 *
207 * There's no need to cache this result with anonymize_mem, since
208 * we already handle blob content caching with marks.
209 */
210 static char *anonymize_blob(unsigned long *size)
211 {
212 static int counter;
213 struct strbuf out = STRBUF_INIT;
214 strbuf_addf(&out, "anonymous blob %d", counter++);
215 *size = out.len;
216 return strbuf_detach(&out, NULL);
217 }
218
219 static void export_blob(const struct object_id *oid)
220 {
221 unsigned long size;
222 enum object_type type;
223 char *buf;
224 struct object *object;
225 int eaten;
226
227 if (no_data)
228 return;
229
230 if (is_null_oid(oid))
231 return;
232
233 object = lookup_object(the_repository, oid->hash);
234 if (object && object->flags & SHOWN)
235 return;
236
237 if (anonymize) {
238 buf = anonymize_blob(&size);
239 object = (struct object *)lookup_blob(the_repository, oid);
240 eaten = 0;
241 } else {
242 buf = read_object_file(oid, &type, &size);
243 if (!buf)
244 die ("Could not read blob %s", oid_to_hex(oid));
245 if (check_object_signature(oid, buf, size, type_name(type)) < 0)
246 die("sha1 mismatch in blob %s", oid_to_hex(oid));
247 object = parse_object_buffer(the_repository, oid, type,
248 size, buf, &eaten);
249 }
250
251 if (!object)
252 die("Could not read blob %s", oid_to_hex(oid));
253
254 mark_next_object(object);
255
256 printf("blob\nmark :%"PRIu32"\ndata %lu\n", last_idnum, size);
257 if (size && fwrite(buf, size, 1, stdout) != 1)
258 die_errno ("Could not write blob '%s'", oid_to_hex(oid));
259 printf("\n");
260
261 show_progress();
262
263 object->flags |= SHOWN;
264 if (!eaten)
265 free(buf);
266 }
267
268 static int depth_first(const void *a_, const void *b_)
269 {
270 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
271 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
272 const char *name_a, *name_b;
273 int len_a, len_b, len;
274 int cmp;
275
276 name_a = a->one ? a->one->path : a->two->path;
277 name_b = b->one ? b->one->path : b->two->path;
278
279 len_a = strlen(name_a);
280 len_b = strlen(name_b);
281 len = (len_a < len_b) ? len_a : len_b;
282
283 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
284 cmp = memcmp(name_a, name_b, len);
285 if (cmp)
286 return cmp;
287 cmp = len_b - len_a;
288 if (cmp)
289 return cmp;
290 /*
291 * Move 'R'ename entries last so that all references of the file
292 * appear in the output before it is renamed (e.g., when a file
293 * was copied and renamed in the same commit).
294 */
295 return (a->status == 'R') - (b->status == 'R');
296 }
297
298 static void print_path_1(const char *path)
299 {
300 int need_quote = quote_c_style(path, NULL, NULL, 0);
301 if (need_quote)
302 quote_c_style(path, NULL, stdout, 0);
303 else if (strchr(path, ' '))
304 printf("\"%s\"", path);
305 else
306 printf("%s", path);
307 }
308
309 static void *anonymize_path_component(const void *path, size_t *len)
310 {
311 static int counter;
312 struct strbuf out = STRBUF_INIT;
313 strbuf_addf(&out, "path%d", counter++);
314 return strbuf_detach(&out, len);
315 }
316
317 static void print_path(const char *path)
318 {
319 if (!anonymize)
320 print_path_1(path);
321 else {
322 static struct hashmap paths;
323 static struct strbuf anon = STRBUF_INIT;
324
325 anonymize_path(&anon, path, &paths, anonymize_path_component);
326 print_path_1(anon.buf);
327 strbuf_reset(&anon);
328 }
329 }
330
331 static void *generate_fake_oid(const void *old, size_t *len)
332 {
333 static uint32_t counter = 1; /* avoid null sha1 */
334 unsigned char *out = xcalloc(GIT_SHA1_RAWSZ, 1);
335 put_be32(out + GIT_SHA1_RAWSZ - 4, counter++);
336 return out;
337 }
338
339 static const unsigned char *anonymize_sha1(const struct object_id *oid)
340 {
341 static struct hashmap sha1s;
342 size_t len = GIT_SHA1_RAWSZ;
343 return anonymize_mem(&sha1s, generate_fake_oid, oid, &len);
344 }
345
346 static void show_filemodify(struct diff_queue_struct *q,
347 struct diff_options *options, void *data)
348 {
349 int i;
350 struct string_list *changed = data;
351
352 /*
353 * Handle files below a directory first, in case they are all deleted
354 * and the directory changes to a file or symlink.
355 */
356 QSORT(q->queue, q->nr, depth_first);
357
358 for (i = 0; i < q->nr; i++) {
359 struct diff_filespec *ospec = q->queue[i]->one;
360 struct diff_filespec *spec = q->queue[i]->two;
361
362 switch (q->queue[i]->status) {
363 case DIFF_STATUS_DELETED:
364 printf("D ");
365 print_path(spec->path);
366 string_list_insert(changed, spec->path);
367 putchar('\n');
368 break;
369
370 case DIFF_STATUS_COPIED:
371 case DIFF_STATUS_RENAMED:
372 /*
373 * If a change in the file corresponding to ospec->path
374 * has been observed, we cannot trust its contents
375 * because the diff is calculated based on the prior
376 * contents, not the current contents. So, declare a
377 * copy or rename only if there was no change observed.
378 */
379 if (!string_list_has_string(changed, ospec->path)) {
380 printf("%c ", q->queue[i]->status);
381 print_path(ospec->path);
382 putchar(' ');
383 print_path(spec->path);
384 string_list_insert(changed, spec->path);
385 putchar('\n');
386
387 if (!oidcmp(&ospec->oid, &spec->oid) &&
388 ospec->mode == spec->mode)
389 break;
390 }
391 /* fallthrough */
392
393 case DIFF_STATUS_TYPE_CHANGED:
394 case DIFF_STATUS_MODIFIED:
395 case DIFF_STATUS_ADDED:
396 /*
397 * Links refer to objects in another repositories;
398 * output the SHA-1 verbatim.
399 */
400 if (no_data || S_ISGITLINK(spec->mode))
401 printf("M %06o %s ", spec->mode,
402 sha1_to_hex(anonymize ?
403 anonymize_sha1(&spec->oid) :
404 spec->oid.hash));
405 else {
406 struct object *object = lookup_object(the_repository,
407 spec->oid.hash);
408 printf("M %06o :%d ", spec->mode,
409 get_object_mark(object));
410 }
411 print_path(spec->path);
412 string_list_insert(changed, spec->path);
413 putchar('\n');
414 break;
415
416 default:
417 die("Unexpected comparison status '%c' for %s, %s",
418 q->queue[i]->status,
419 ospec->path ? ospec->path : "none",
420 spec->path ? spec->path : "none");
421 }
422 }
423 }
424
425 static const char *find_encoding(const char *begin, const char *end)
426 {
427 const char *needle = "\nencoding ";
428 char *bol, *eol;
429
430 bol = memmem(begin, end ? end - begin : strlen(begin),
431 needle, strlen(needle));
432 if (!bol)
433 return git_commit_encoding;
434 bol += strlen(needle);
435 eol = strchrnul(bol, '\n');
436 *eol = '\0';
437 return bol;
438 }
439
440 static void *anonymize_ref_component(const void *old, size_t *len)
441 {
442 static int counter;
443 struct strbuf out = STRBUF_INIT;
444 strbuf_addf(&out, "ref%d", counter++);
445 return strbuf_detach(&out, len);
446 }
447
448 static const char *anonymize_refname(const char *refname)
449 {
450 /*
451 * If any of these prefixes is found, we will leave it intact
452 * so that tags remain tags and so forth.
453 */
454 static const char *prefixes[] = {
455 "refs/heads/",
456 "refs/tags/",
457 "refs/remotes/",
458 "refs/"
459 };
460 static struct hashmap refs;
461 static struct strbuf anon = STRBUF_INIT;
462 int i;
463
464 /*
465 * We also leave "master" as a special case, since it does not reveal
466 * anything interesting.
467 */
468 if (!strcmp(refname, "refs/heads/master"))
469 return refname;
470
471 strbuf_reset(&anon);
472 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
473 if (skip_prefix(refname, prefixes[i], &refname)) {
474 strbuf_addstr(&anon, prefixes[i]);
475 break;
476 }
477 }
478
479 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
480 return anon.buf;
481 }
482
483 /*
484 * We do not even bother to cache commit messages, as they are unlikely
485 * to be repeated verbatim, and it is not that interesting when they are.
486 */
487 static char *anonymize_commit_message(const char *old)
488 {
489 static int counter;
490 return xstrfmt("subject %d\n\nbody\n", counter++);
491 }
492
493 static struct hashmap idents;
494 static void *anonymize_ident(const void *old, size_t *len)
495 {
496 static int counter;
497 struct strbuf out = STRBUF_INIT;
498 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
499 counter++;
500 return strbuf_detach(&out, len);
501 }
502
503 /*
504 * Our strategy here is to anonymize the names and email addresses,
505 * but keep timestamps intact, as they influence things like traversal
506 * order (and by themselves should not be too revealing).
507 */
508 static void anonymize_ident_line(const char **beg, const char **end)
509 {
510 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
511 static unsigned which_buffer;
512
513 struct strbuf *out;
514 struct ident_split split;
515 const char *end_of_header;
516
517 out = &buffers[which_buffer++];
518 which_buffer %= ARRAY_SIZE(buffers);
519 strbuf_reset(out);
520
521 /* skip "committer", "author", "tagger", etc */
522 end_of_header = strchr(*beg, ' ');
523 if (!end_of_header)
524 BUG("malformed line fed to anonymize_ident_line: %.*s",
525 (int)(*end - *beg), *beg);
526 end_of_header++;
527 strbuf_add(out, *beg, end_of_header - *beg);
528
529 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
530 split.date_begin) {
531 const char *ident;
532 size_t len;
533
534 len = split.mail_end - split.name_begin;
535 ident = anonymize_mem(&idents, anonymize_ident,
536 split.name_begin, &len);
537 strbuf_add(out, ident, len);
538 strbuf_addch(out, ' ');
539 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
540 } else {
541 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
542 }
543
544 *beg = out->buf;
545 *end = out->buf + out->len;
546 }
547
548 static void handle_commit(struct commit *commit, struct rev_info *rev,
549 struct string_list *paths_of_changed_objects)
550 {
551 int saved_output_format = rev->diffopt.output_format;
552 const char *commit_buffer;
553 const char *author, *author_end, *committer, *committer_end;
554 const char *encoding, *message;
555 char *reencoded = NULL;
556 struct commit_list *p;
557 const char *refname;
558 int i;
559
560 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
561
562 parse_commit_or_die(commit);
563 commit_buffer = get_commit_buffer(commit, NULL);
564 author = strstr(commit_buffer, "\nauthor ");
565 if (!author)
566 die ("Could not find author in commit %s",
567 oid_to_hex(&commit->object.oid));
568 author++;
569 author_end = strchrnul(author, '\n');
570 committer = strstr(author_end, "\ncommitter ");
571 if (!committer)
572 die ("Could not find committer in commit %s",
573 oid_to_hex(&commit->object.oid));
574 committer++;
575 committer_end = strchrnul(committer, '\n');
576 message = strstr(committer_end, "\n\n");
577 encoding = find_encoding(committer_end, message);
578 if (message)
579 message += 2;
580
581 if (commit->parents &&
582 get_object_mark(&commit->parents->item->object) != 0 &&
583 !full_tree) {
584 parse_commit_or_die(commit->parents->item);
585 diff_tree_oid(get_commit_tree_oid(commit->parents->item),
586 get_commit_tree_oid(commit), "", &rev->diffopt);
587 }
588 else
589 diff_root_tree_oid(get_commit_tree_oid(commit),
590 "", &rev->diffopt);
591
592 /* Export the referenced blobs, and remember the marks. */
593 for (i = 0; i < diff_queued_diff.nr; i++)
594 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
595 export_blob(&diff_queued_diff.queue[i]->two->oid);
596
597 refname = *revision_sources_at(&revision_sources, commit);
598 if (anonymize) {
599 refname = anonymize_refname(refname);
600 anonymize_ident_line(&committer, &committer_end);
601 anonymize_ident_line(&author, &author_end);
602 }
603
604 mark_next_object(&commit->object);
605 if (anonymize)
606 reencoded = anonymize_commit_message(message);
607 else if (!is_encoding_utf8(encoding))
608 reencoded = reencode_string(message, "UTF-8", encoding);
609 if (!commit->parents)
610 printf("reset %s\n", refname);
611 printf("commit %s\nmark :%"PRIu32"\n%.*s\n%.*s\ndata %u\n%s",
612 refname, last_idnum,
613 (int)(author_end - author), author,
614 (int)(committer_end - committer), committer,
615 (unsigned)(reencoded
616 ? strlen(reencoded) : message
617 ? strlen(message) : 0),
618 reencoded ? reencoded : message ? message : "");
619 free(reencoded);
620 unuse_commit_buffer(commit, commit_buffer);
621
622 for (i = 0, p = commit->parents; p; p = p->next) {
623 int mark = get_object_mark(&p->item->object);
624 if (!mark)
625 continue;
626 if (i == 0)
627 printf("from :%d\n", mark);
628 else
629 printf("merge :%d\n", mark);
630 i++;
631 }
632
633 if (full_tree)
634 printf("deleteall\n");
635 log_tree_diff_flush(rev);
636 string_list_clear(paths_of_changed_objects, 0);
637 rev->diffopt.output_format = saved_output_format;
638
639 printf("\n");
640
641 show_progress();
642 }
643
644 static void *anonymize_tag(const void *old, size_t *len)
645 {
646 static int counter;
647 struct strbuf out = STRBUF_INIT;
648 strbuf_addf(&out, "tag message %d", counter++);
649 return strbuf_detach(&out, len);
650 }
651
652 static void handle_tail(struct object_array *commits, struct rev_info *revs,
653 struct string_list *paths_of_changed_objects)
654 {
655 struct commit *commit;
656 while (commits->nr) {
657 commit = (struct commit *)object_array_pop(commits);
658 if (has_unshown_parent(commit)) {
659 /* Queue again, to be handled later */
660 add_object_array(&commit->object, NULL, commits);
661 return;
662 }
663 handle_commit(commit, revs, paths_of_changed_objects);
664 }
665 }
666
667 static void handle_tag(const char *name, struct tag *tag)
668 {
669 unsigned long size;
670 enum object_type type;
671 char *buf;
672 const char *tagger, *tagger_end, *message;
673 size_t message_size = 0;
674 struct object *tagged;
675 int tagged_mark;
676 struct commit *p;
677
678 /* Trees have no identifier in fast-export output, thus we have no way
679 * to output tags of trees, tags of tags of trees, etc. Simply omit
680 * such tags.
681 */
682 tagged = tag->tagged;
683 while (tagged->type == OBJ_TAG) {
684 tagged = ((struct tag *)tagged)->tagged;
685 }
686 if (tagged->type == OBJ_TREE) {
687 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
688 oid_to_hex(&tag->object.oid));
689 return;
690 }
691
692 buf = read_object_file(&tag->object.oid, &type, &size);
693 if (!buf)
694 die ("Could not read tag %s", oid_to_hex(&tag->object.oid));
695 message = memmem(buf, size, "\n\n", 2);
696 if (message) {
697 message += 2;
698 message_size = strlen(message);
699 }
700 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
701 if (!tagger) {
702 if (fake_missing_tagger)
703 tagger = "tagger Unspecified Tagger "
704 "<unspecified-tagger> 0 +0000";
705 else
706 tagger = "";
707 tagger_end = tagger + strlen(tagger);
708 } else {
709 tagger++;
710 tagger_end = strchrnul(tagger, '\n');
711 if (anonymize)
712 anonymize_ident_line(&tagger, &tagger_end);
713 }
714
715 if (anonymize) {
716 name = anonymize_refname(name);
717 if (message) {
718 static struct hashmap tags;
719 message = anonymize_mem(&tags, anonymize_tag,
720 message, &message_size);
721 }
722 }
723
724 /* handle signed tags */
725 if (message) {
726 const char *signature = strstr(message,
727 "\n-----BEGIN PGP SIGNATURE-----\n");
728 if (signature)
729 switch(signed_tag_mode) {
730 case ABORT:
731 die ("Encountered signed tag %s; use "
732 "--signed-tags=<mode> to handle it.",
733 oid_to_hex(&tag->object.oid));
734 case WARN:
735 warning ("Exporting signed tag %s",
736 oid_to_hex(&tag->object.oid));
737 /* fallthru */
738 case VERBATIM:
739 break;
740 case WARN_STRIP:
741 warning ("Stripping signature from tag %s",
742 oid_to_hex(&tag->object.oid));
743 /* fallthru */
744 case STRIP:
745 message_size = signature + 1 - message;
746 break;
747 }
748 }
749
750 /* handle tag->tagged having been filtered out due to paths specified */
751 tagged = tag->tagged;
752 tagged_mark = get_object_mark(tagged);
753 if (!tagged_mark) {
754 switch(tag_of_filtered_mode) {
755 case ABORT:
756 die ("Tag %s tags unexported object; use "
757 "--tag-of-filtered-object=<mode> to handle it.",
758 oid_to_hex(&tag->object.oid));
759 case DROP:
760 /* Ignore this tag altogether */
761 free(buf);
762 return;
763 case REWRITE:
764 if (tagged->type != OBJ_COMMIT) {
765 die ("Tag %s tags unexported %s!",
766 oid_to_hex(&tag->object.oid),
767 type_name(tagged->type));
768 }
769 p = (struct commit *)tagged;
770 for (;;) {
771 if (p->parents && p->parents->next)
772 break;
773 if (p->object.flags & UNINTERESTING)
774 break;
775 if (!(p->object.flags & TREESAME))
776 break;
777 if (!p->parents)
778 die ("Can't find replacement commit for tag %s\n",
779 oid_to_hex(&tag->object.oid));
780 p = p->parents->item;
781 }
782 tagged_mark = get_object_mark(&p->object);
783 }
784 }
785
786 if (starts_with(name, "refs/tags/"))
787 name += 10;
788 printf("tag %s\nfrom :%d\n%.*s%sdata %d\n%.*s\n",
789 name, tagged_mark,
790 (int)(tagger_end - tagger), tagger,
791 tagger == tagger_end ? "" : "\n",
792 (int)message_size, (int)message_size, message ? message : "");
793 free(buf);
794 }
795
796 static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
797 {
798 switch (e->item->type) {
799 case OBJ_COMMIT:
800 return (struct commit *)e->item;
801 case OBJ_TAG: {
802 struct tag *tag = (struct tag *)e->item;
803
804 /* handle nested tags */
805 while (tag && tag->object.type == OBJ_TAG) {
806 parse_object(the_repository, &tag->object.oid);
807 string_list_append(&extra_refs, full_name)->util = tag;
808 tag = (struct tag *)tag->tagged;
809 }
810 if (!tag)
811 die("Tag %s points nowhere?", e->name);
812 return (struct commit *)tag;
813 break;
814 }
815 default:
816 return NULL;
817 }
818 }
819
820 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
821 {
822 int i;
823
824 for (i = 0; i < info->nr; i++) {
825 struct rev_cmdline_entry *e = info->rev + i;
826 struct object_id oid;
827 struct commit *commit;
828 char *full_name;
829
830 if (e->flags & UNINTERESTING)
831 continue;
832
833 if (dwim_ref(e->name, strlen(e->name), &oid, &full_name) != 1)
834 continue;
835
836 if (refspecs.nr) {
837 char *private;
838 private = apply_refspecs(&refspecs, full_name);
839 if (private) {
840 free(full_name);
841 full_name = private;
842 }
843 }
844
845 commit = get_commit(e, full_name);
846 if (!commit) {
847 warning("%s: Unexpected object of type %s, skipping.",
848 e->name,
849 type_name(e->item->type));
850 continue;
851 }
852
853 switch(commit->object.type) {
854 case OBJ_COMMIT:
855 break;
856 case OBJ_BLOB:
857 export_blob(&commit->object.oid);
858 continue;
859 default: /* OBJ_TAG (nested tags) is already handled */
860 warning("Tag points to object of unexpected type %s, skipping.",
861 type_name(commit->object.type));
862 continue;
863 }
864
865 /*
866 * This ref will not be updated through a commit, lets make
867 * sure it gets properly updated eventually.
868 */
869 if (*revision_sources_at(&revision_sources, commit) ||
870 commit->object.flags & SHOWN)
871 string_list_append(&extra_refs, full_name)->util = commit;
872 if (!*revision_sources_at(&revision_sources, commit))
873 *revision_sources_at(&revision_sources, commit) = full_name;
874 }
875 }
876
877 static void handle_tags_and_duplicates(void)
878 {
879 struct commit *commit;
880 int i;
881
882 for (i = extra_refs.nr - 1; i >= 0; i--) {
883 const char *name = extra_refs.items[i].string;
884 struct object *object = extra_refs.items[i].util;
885 switch (object->type) {
886 case OBJ_TAG:
887 handle_tag(name, (struct tag *)object);
888 break;
889 case OBJ_COMMIT:
890 if (anonymize)
891 name = anonymize_refname(name);
892 /* create refs pointing to already seen commits */
893 commit = (struct commit *)object;
894 printf("reset %s\nfrom :%d\n\n", name,
895 get_object_mark(&commit->object));
896 show_progress();
897 break;
898 }
899 }
900 }
901
902 static void export_marks(char *file)
903 {
904 unsigned int i;
905 uint32_t mark;
906 struct decoration_entry *deco = idnums.entries;
907 FILE *f;
908 int e = 0;
909
910 f = fopen_for_writing(file);
911 if (!f)
912 die_errno("Unable to open marks file %s for writing.", file);
913
914 for (i = 0; i < idnums.size; i++) {
915 if (deco->base && deco->base->type == 1) {
916 mark = ptr_to_mark(deco->decoration);
917 if (fprintf(f, ":%"PRIu32" %s\n", mark,
918 oid_to_hex(&deco->base->oid)) < 0) {
919 e = 1;
920 break;
921 }
922 }
923 deco++;
924 }
925
926 e |= ferror(f);
927 e |= fclose(f);
928 if (e)
929 error("Unable to write marks file %s.", file);
930 }
931
932 static void import_marks(char *input_file)
933 {
934 char line[512];
935 FILE *f = xfopen(input_file, "r");
936
937 while (fgets(line, sizeof(line), f)) {
938 uint32_t mark;
939 char *line_end, *mark_end;
940 struct object_id oid;
941 struct object *object;
942 struct commit *commit;
943 enum object_type type;
944
945 line_end = strchr(line, '\n');
946 if (line[0] != ':' || !line_end)
947 die("corrupt mark line: %s", line);
948 *line_end = '\0';
949
950 mark = strtoumax(line + 1, &mark_end, 10);
951 if (!mark || mark_end == line + 1
952 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
953 die("corrupt mark line: %s", line);
954
955 if (last_idnum < mark)
956 last_idnum = mark;
957
958 type = oid_object_info(the_repository, &oid, NULL);
959 if (type < 0)
960 die("object not found: %s", oid_to_hex(&oid));
961
962 if (type != OBJ_COMMIT)
963 /* only commits */
964 continue;
965
966 commit = lookup_commit(the_repository, &oid);
967 if (!commit)
968 die("not a commit? can't happen: %s", oid_to_hex(&oid));
969
970 object = &commit->object;
971
972 if (object->flags & SHOWN)
973 error("Object %s already has a mark", oid_to_hex(&oid));
974
975 mark_object(object, mark);
976
977 object->flags |= SHOWN;
978 }
979 fclose(f);
980 }
981
982 static void handle_deletes(void)
983 {
984 int i;
985 for (i = 0; i < refspecs.nr; i++) {
986 struct refspec_item *refspec = &refspecs.items[i];
987 if (*refspec->src)
988 continue;
989
990 printf("reset %s\nfrom %s\n\n",
991 refspec->dst, sha1_to_hex(null_sha1));
992 }
993 }
994
995 int cmd_fast_export(int argc, const char **argv, const char *prefix)
996 {
997 struct rev_info revs;
998 struct object_array commits = OBJECT_ARRAY_INIT;
999 struct commit *commit;
1000 char *export_filename = NULL, *import_filename = NULL;
1001 uint32_t lastimportid;
1002 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
1003 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
1004 struct option options[] = {
1005 OPT_INTEGER(0, "progress", &progress,
1006 N_("show progress after <n> objects")),
1007 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1008 N_("select handling of signed tags"),
1009 parse_opt_signed_tag_mode),
1010 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1011 N_("select handling of tags that tag filtered objects"),
1012 parse_opt_tag_of_filtered_mode),
1013 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1014 N_("Dump marks to this file")),
1015 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1016 N_("Import marks from this file")),
1017 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1018 N_("Fake a tagger when tags lack one")),
1019 OPT_BOOL(0, "full-tree", &full_tree,
1020 N_("Output full tree for each commit")),
1021 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1022 N_("Use the done feature to terminate the stream")),
1023 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
1024 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1025 N_("Apply refspec to exported refs")),
1026 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1027 OPT_END()
1028 };
1029
1030 if (argc == 1)
1031 usage_with_options (fast_export_usage, options);
1032
1033 /* we handle encodings */
1034 git_config(git_default_config, NULL);
1035
1036 init_revisions(&revs, prefix);
1037 init_revision_sources(&revision_sources);
1038 revs.topo_order = 1;
1039 revs.sources = &revision_sources;
1040 revs.rewrite_parents = 1;
1041 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1042 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1043 argc = setup_revisions(argc, argv, &revs, NULL);
1044 if (argc > 1)
1045 usage_with_options (fast_export_usage, options);
1046
1047 if (refspecs_list.nr) {
1048 int i;
1049
1050 for (i = 0; i < refspecs_list.nr; i++)
1051 refspec_append(&refspecs, refspecs_list.items[i].string);
1052
1053 string_list_clear(&refspecs_list, 1);
1054 }
1055
1056 if (use_done_feature)
1057 printf("feature done\n");
1058
1059 if (import_filename)
1060 import_marks(import_filename);
1061 lastimportid = last_idnum;
1062
1063 if (import_filename && revs.prune_data.nr)
1064 full_tree = 1;
1065
1066 get_tags_and_duplicates(&revs.cmdline);
1067
1068 if (prepare_revision_walk(&revs))
1069 die("revision walk setup failed");
1070 revs.diffopt.format_callback = show_filemodify;
1071 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1072 revs.diffopt.flags.recursive = 1;
1073 while ((commit = get_revision(&revs))) {
1074 if (has_unshown_parent(commit)) {
1075 add_object_array(&commit->object, NULL, &commits);
1076 }
1077 else {
1078 handle_commit(commit, &revs, &paths_of_changed_objects);
1079 handle_tail(&commits, &revs, &paths_of_changed_objects);
1080 }
1081 }
1082
1083 handle_tags_and_duplicates();
1084 handle_deletes();
1085
1086 if (export_filename && lastimportid != last_idnum)
1087 export_marks(export_filename);
1088
1089 if (use_done_feature)
1090 printf("done\n");
1091
1092 refspec_clear(&refspecs);
1093
1094 return 0;
1095 }