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