a16aeaa826f90a6f7e20f8e7e0e5137d1c3006bc
[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(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(spec->oid.hash);
406 printf("M %06o :%d ", spec->mode,
407 get_object_mark(object));
408 }
409 print_path(spec->path);
410 string_list_insert(changed, spec->path);
411 putchar('\n');
412 break;
413
414 default:
415 die("Unexpected comparison status '%c' for %s, %s",
416 q->queue[i]->status,
417 ospec->path ? ospec->path : "none",
418 spec->path ? spec->path : "none");
419 }
420 }
421 }
422
423 static const char *find_encoding(const char *begin, const char *end)
424 {
425 const char *needle = "\nencoding ";
426 char *bol, *eol;
427
428 bol = memmem(begin, end ? end - begin : strlen(begin),
429 needle, strlen(needle));
430 if (!bol)
431 return git_commit_encoding;
432 bol += strlen(needle);
433 eol = strchrnul(bol, '\n');
434 *eol = '\0';
435 return bol;
436 }
437
438 static void *anonymize_ref_component(const void *old, size_t *len)
439 {
440 static int counter;
441 struct strbuf out = STRBUF_INIT;
442 strbuf_addf(&out, "ref%d", counter++);
443 return strbuf_detach(&out, len);
444 }
445
446 static const char *anonymize_refname(const char *refname)
447 {
448 /*
449 * If any of these prefixes is found, we will leave it intact
450 * so that tags remain tags and so forth.
451 */
452 static const char *prefixes[] = {
453 "refs/heads/",
454 "refs/tags/",
455 "refs/remotes/",
456 "refs/"
457 };
458 static struct hashmap refs;
459 static struct strbuf anon = STRBUF_INIT;
460 int i;
461
462 /*
463 * We also leave "master" as a special case, since it does not reveal
464 * anything interesting.
465 */
466 if (!strcmp(refname, "refs/heads/master"))
467 return refname;
468
469 strbuf_reset(&anon);
470 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
471 if (skip_prefix(refname, prefixes[i], &refname)) {
472 strbuf_addstr(&anon, prefixes[i]);
473 break;
474 }
475 }
476
477 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
478 return anon.buf;
479 }
480
481 /*
482 * We do not even bother to cache commit messages, as they are unlikely
483 * to be repeated verbatim, and it is not that interesting when they are.
484 */
485 static char *anonymize_commit_message(const char *old)
486 {
487 static int counter;
488 return xstrfmt("subject %d\n\nbody\n", counter++);
489 }
490
491 static struct hashmap idents;
492 static void *anonymize_ident(const void *old, size_t *len)
493 {
494 static int counter;
495 struct strbuf out = STRBUF_INIT;
496 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
497 counter++;
498 return strbuf_detach(&out, len);
499 }
500
501 /*
502 * Our strategy here is to anonymize the names and email addresses,
503 * but keep timestamps intact, as they influence things like traversal
504 * order (and by themselves should not be too revealing).
505 */
506 static void anonymize_ident_line(const char **beg, const char **end)
507 {
508 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
509 static unsigned which_buffer;
510
511 struct strbuf *out;
512 struct ident_split split;
513 const char *end_of_header;
514
515 out = &buffers[which_buffer++];
516 which_buffer %= ARRAY_SIZE(buffers);
517 strbuf_reset(out);
518
519 /* skip "committer", "author", "tagger", etc */
520 end_of_header = strchr(*beg, ' ');
521 if (!end_of_header)
522 BUG("malformed line fed to anonymize_ident_line: %.*s",
523 (int)(*end - *beg), *beg);
524 end_of_header++;
525 strbuf_add(out, *beg, end_of_header - *beg);
526
527 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
528 split.date_begin) {
529 const char *ident;
530 size_t len;
531
532 len = split.mail_end - split.name_begin;
533 ident = anonymize_mem(&idents, anonymize_ident,
534 split.name_begin, &len);
535 strbuf_add(out, ident, len);
536 strbuf_addch(out, ' ');
537 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
538 } else {
539 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
540 }
541
542 *beg = out->buf;
543 *end = out->buf + out->len;
544 }
545
546 static void handle_commit(struct commit *commit, struct rev_info *rev,
547 struct string_list *paths_of_changed_objects)
548 {
549 int saved_output_format = rev->diffopt.output_format;
550 const char *commit_buffer;
551 const char *author, *author_end, *committer, *committer_end;
552 const char *encoding, *message;
553 char *reencoded = NULL;
554 struct commit_list *p;
555 const char *refname;
556 int i;
557
558 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
559
560 parse_commit_or_die(commit);
561 commit_buffer = get_commit_buffer(commit, NULL);
562 author = strstr(commit_buffer, "\nauthor ");
563 if (!author)
564 die ("Could not find author in commit %s",
565 oid_to_hex(&commit->object.oid));
566 author++;
567 author_end = strchrnul(author, '\n');
568 committer = strstr(author_end, "\ncommitter ");
569 if (!committer)
570 die ("Could not find committer in commit %s",
571 oid_to_hex(&commit->object.oid));
572 committer++;
573 committer_end = strchrnul(committer, '\n');
574 message = strstr(committer_end, "\n\n");
575 encoding = find_encoding(committer_end, message);
576 if (message)
577 message += 2;
578
579 if (commit->parents &&
580 get_object_mark(&commit->parents->item->object) != 0 &&
581 !full_tree) {
582 parse_commit_or_die(commit->parents->item);
583 diff_tree_oid(get_commit_tree_oid(commit->parents->item),
584 get_commit_tree_oid(commit), "", &rev->diffopt);
585 }
586 else
587 diff_root_tree_oid(get_commit_tree_oid(commit),
588 "", &rev->diffopt);
589
590 /* Export the referenced blobs, and remember the marks. */
591 for (i = 0; i < diff_queued_diff.nr; i++)
592 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
593 export_blob(&diff_queued_diff.queue[i]->two->oid);
594
595 refname = *revision_sources_at(&revision_sources, commit);
596 if (anonymize) {
597 refname = anonymize_refname(refname);
598 anonymize_ident_line(&committer, &committer_end);
599 anonymize_ident_line(&author, &author_end);
600 }
601
602 mark_next_object(&commit->object);
603 if (anonymize)
604 reencoded = anonymize_commit_message(message);
605 else if (!is_encoding_utf8(encoding))
606 reencoded = reencode_string(message, "UTF-8", encoding);
607 if (!commit->parents)
608 printf("reset %s\n", refname);
609 printf("commit %s\nmark :%"PRIu32"\n%.*s\n%.*s\ndata %u\n%s",
610 refname, last_idnum,
611 (int)(author_end - author), author,
612 (int)(committer_end - committer), committer,
613 (unsigned)(reencoded
614 ? strlen(reencoded) : message
615 ? strlen(message) : 0),
616 reencoded ? reencoded : message ? message : "");
617 free(reencoded);
618 unuse_commit_buffer(commit, commit_buffer);
619
620 for (i = 0, p = commit->parents; p; p = p->next) {
621 int mark = get_object_mark(&p->item->object);
622 if (!mark)
623 continue;
624 if (i == 0)
625 printf("from :%d\n", mark);
626 else
627 printf("merge :%d\n", mark);
628 i++;
629 }
630
631 if (full_tree)
632 printf("deleteall\n");
633 log_tree_diff_flush(rev);
634 string_list_clear(paths_of_changed_objects, 0);
635 rev->diffopt.output_format = saved_output_format;
636
637 printf("\n");
638
639 show_progress();
640 }
641
642 static void *anonymize_tag(const void *old, size_t *len)
643 {
644 static int counter;
645 struct strbuf out = STRBUF_INIT;
646 strbuf_addf(&out, "tag message %d", counter++);
647 return strbuf_detach(&out, len);
648 }
649
650 static void handle_tail(struct object_array *commits, struct rev_info *revs,
651 struct string_list *paths_of_changed_objects)
652 {
653 struct commit *commit;
654 while (commits->nr) {
655 commit = (struct commit *)object_array_pop(commits);
656 if (has_unshown_parent(commit)) {
657 /* Queue again, to be handled later */
658 add_object_array(&commit->object, NULL, commits);
659 return;
660 }
661 handle_commit(commit, revs, paths_of_changed_objects);
662 }
663 }
664
665 static void handle_tag(const char *name, struct tag *tag)
666 {
667 unsigned long size;
668 enum object_type type;
669 char *buf;
670 const char *tagger, *tagger_end, *message;
671 size_t message_size = 0;
672 struct object *tagged;
673 int tagged_mark;
674 struct commit *p;
675
676 /* Trees have no identifier in fast-export output, thus we have no way
677 * to output tags of trees, tags of tags of trees, etc. Simply omit
678 * such tags.
679 */
680 tagged = tag->tagged;
681 while (tagged->type == OBJ_TAG) {
682 tagged = ((struct tag *)tagged)->tagged;
683 }
684 if (tagged->type == OBJ_TREE) {
685 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
686 oid_to_hex(&tag->object.oid));
687 return;
688 }
689
690 buf = read_object_file(&tag->object.oid, &type, &size);
691 if (!buf)
692 die ("Could not read tag %s", oid_to_hex(&tag->object.oid));
693 message = memmem(buf, size, "\n\n", 2);
694 if (message) {
695 message += 2;
696 message_size = strlen(message);
697 }
698 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
699 if (!tagger) {
700 if (fake_missing_tagger)
701 tagger = "tagger Unspecified Tagger "
702 "<unspecified-tagger> 0 +0000";
703 else
704 tagger = "";
705 tagger_end = tagger + strlen(tagger);
706 } else {
707 tagger++;
708 tagger_end = strchrnul(tagger, '\n');
709 if (anonymize)
710 anonymize_ident_line(&tagger, &tagger_end);
711 }
712
713 if (anonymize) {
714 name = anonymize_refname(name);
715 if (message) {
716 static struct hashmap tags;
717 message = anonymize_mem(&tags, anonymize_tag,
718 message, &message_size);
719 }
720 }
721
722 /* handle signed tags */
723 if (message) {
724 const char *signature = strstr(message,
725 "\n-----BEGIN PGP SIGNATURE-----\n");
726 if (signature)
727 switch(signed_tag_mode) {
728 case ABORT:
729 die ("Encountered signed tag %s; use "
730 "--signed-tags=<mode> to handle it.",
731 oid_to_hex(&tag->object.oid));
732 case WARN:
733 warning ("Exporting signed tag %s",
734 oid_to_hex(&tag->object.oid));
735 /* fallthru */
736 case VERBATIM:
737 break;
738 case WARN_STRIP:
739 warning ("Stripping signature from tag %s",
740 oid_to_hex(&tag->object.oid));
741 /* fallthru */
742 case STRIP:
743 message_size = signature + 1 - message;
744 break;
745 }
746 }
747
748 /* handle tag->tagged having been filtered out due to paths specified */
749 tagged = tag->tagged;
750 tagged_mark = get_object_mark(tagged);
751 if (!tagged_mark) {
752 switch(tag_of_filtered_mode) {
753 case ABORT:
754 die ("Tag %s tags unexported object; use "
755 "--tag-of-filtered-object=<mode> to handle it.",
756 oid_to_hex(&tag->object.oid));
757 case DROP:
758 /* Ignore this tag altogether */
759 free(buf);
760 return;
761 case REWRITE:
762 if (tagged->type != OBJ_COMMIT) {
763 die ("Tag %s tags unexported %s!",
764 oid_to_hex(&tag->object.oid),
765 type_name(tagged->type));
766 }
767 p = (struct commit *)tagged;
768 for (;;) {
769 if (p->parents && p->parents->next)
770 break;
771 if (p->object.flags & UNINTERESTING)
772 break;
773 if (!(p->object.flags & TREESAME))
774 break;
775 if (!p->parents)
776 die ("Can't find replacement commit for tag %s\n",
777 oid_to_hex(&tag->object.oid));
778 p = p->parents->item;
779 }
780 tagged_mark = get_object_mark(&p->object);
781 }
782 }
783
784 if (starts_with(name, "refs/tags/"))
785 name += 10;
786 printf("tag %s\nfrom :%d\n%.*s%sdata %d\n%.*s\n",
787 name, tagged_mark,
788 (int)(tagger_end - tagger), tagger,
789 tagger == tagger_end ? "" : "\n",
790 (int)message_size, (int)message_size, message ? message : "");
791 free(buf);
792 }
793
794 static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
795 {
796 switch (e->item->type) {
797 case OBJ_COMMIT:
798 return (struct commit *)e->item;
799 case OBJ_TAG: {
800 struct tag *tag = (struct tag *)e->item;
801
802 /* handle nested tags */
803 while (tag && tag->object.type == OBJ_TAG) {
804 parse_object(the_repository, &tag->object.oid);
805 string_list_append(&extra_refs, full_name)->util = tag;
806 tag = (struct tag *)tag->tagged;
807 }
808 if (!tag)
809 die("Tag %s points nowhere?", e->name);
810 return (struct commit *)tag;
811 break;
812 }
813 default:
814 return NULL;
815 }
816 }
817
818 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
819 {
820 int i;
821
822 for (i = 0; i < info->nr; i++) {
823 struct rev_cmdline_entry *e = info->rev + i;
824 struct object_id oid;
825 struct commit *commit;
826 char *full_name;
827
828 if (e->flags & UNINTERESTING)
829 continue;
830
831 if (dwim_ref(e->name, strlen(e->name), &oid, &full_name) != 1)
832 continue;
833
834 if (refspecs.nr) {
835 char *private;
836 private = apply_refspecs(&refspecs, full_name);
837 if (private) {
838 free(full_name);
839 full_name = private;
840 }
841 }
842
843 commit = get_commit(e, full_name);
844 if (!commit) {
845 warning("%s: Unexpected object of type %s, skipping.",
846 e->name,
847 type_name(e->item->type));
848 continue;
849 }
850
851 switch(commit->object.type) {
852 case OBJ_COMMIT:
853 break;
854 case OBJ_BLOB:
855 export_blob(&commit->object.oid);
856 continue;
857 default: /* OBJ_TAG (nested tags) is already handled */
858 warning("Tag points to object of unexpected type %s, skipping.",
859 type_name(commit->object.type));
860 continue;
861 }
862
863 /*
864 * This ref will not be updated through a commit, lets make
865 * sure it gets properly updated eventually.
866 */
867 if (*revision_sources_at(&revision_sources, commit) ||
868 commit->object.flags & SHOWN)
869 string_list_append(&extra_refs, full_name)->util = commit;
870 if (!*revision_sources_at(&revision_sources, commit))
871 *revision_sources_at(&revision_sources, commit) = full_name;
872 }
873 }
874
875 static void handle_tags_and_duplicates(void)
876 {
877 struct commit *commit;
878 int i;
879
880 for (i = extra_refs.nr - 1; i >= 0; i--) {
881 const char *name = extra_refs.items[i].string;
882 struct object *object = extra_refs.items[i].util;
883 switch (object->type) {
884 case OBJ_TAG:
885 handle_tag(name, (struct tag *)object);
886 break;
887 case OBJ_COMMIT:
888 if (anonymize)
889 name = anonymize_refname(name);
890 /* create refs pointing to already seen commits */
891 commit = (struct commit *)object;
892 printf("reset %s\nfrom :%d\n\n", name,
893 get_object_mark(&commit->object));
894 show_progress();
895 break;
896 }
897 }
898 }
899
900 static void export_marks(char *file)
901 {
902 unsigned int i;
903 uint32_t mark;
904 struct decoration_entry *deco = idnums.entries;
905 FILE *f;
906 int e = 0;
907
908 f = fopen_for_writing(file);
909 if (!f)
910 die_errno("Unable to open marks file %s for writing.", file);
911
912 for (i = 0; i < idnums.size; i++) {
913 if (deco->base && deco->base->type == 1) {
914 mark = ptr_to_mark(deco->decoration);
915 if (fprintf(f, ":%"PRIu32" %s\n", mark,
916 oid_to_hex(&deco->base->oid)) < 0) {
917 e = 1;
918 break;
919 }
920 }
921 deco++;
922 }
923
924 e |= ferror(f);
925 e |= fclose(f);
926 if (e)
927 error("Unable to write marks file %s.", file);
928 }
929
930 static void import_marks(char *input_file)
931 {
932 char line[512];
933 FILE *f = xfopen(input_file, "r");
934
935 while (fgets(line, sizeof(line), f)) {
936 uint32_t mark;
937 char *line_end, *mark_end;
938 struct object_id oid;
939 struct object *object;
940 struct commit *commit;
941 enum object_type type;
942
943 line_end = strchr(line, '\n');
944 if (line[0] != ':' || !line_end)
945 die("corrupt mark line: %s", line);
946 *line_end = '\0';
947
948 mark = strtoumax(line + 1, &mark_end, 10);
949 if (!mark || mark_end == line + 1
950 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
951 die("corrupt mark line: %s", line);
952
953 if (last_idnum < mark)
954 last_idnum = mark;
955
956 type = oid_object_info(the_repository, &oid, NULL);
957 if (type < 0)
958 die("object not found: %s", oid_to_hex(&oid));
959
960 if (type != OBJ_COMMIT)
961 /* only commits */
962 continue;
963
964 commit = lookup_commit(&oid);
965 if (!commit)
966 die("not a commit? can't happen: %s", oid_to_hex(&oid));
967
968 object = &commit->object;
969
970 if (object->flags & SHOWN)
971 error("Object %s already has a mark", oid_to_hex(&oid));
972
973 mark_object(object, mark);
974
975 object->flags |= SHOWN;
976 }
977 fclose(f);
978 }
979
980 static void handle_deletes(void)
981 {
982 int i;
983 for (i = 0; i < refspecs.nr; i++) {
984 struct refspec_item *refspec = &refspecs.items[i];
985 if (*refspec->src)
986 continue;
987
988 printf("reset %s\nfrom %s\n\n",
989 refspec->dst, sha1_to_hex(null_sha1));
990 }
991 }
992
993 int cmd_fast_export(int argc, const char **argv, const char *prefix)
994 {
995 struct rev_info revs;
996 struct object_array commits = OBJECT_ARRAY_INIT;
997 struct commit *commit;
998 char *export_filename = NULL, *import_filename = NULL;
999 uint32_t lastimportid;
1000 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
1001 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
1002 struct option options[] = {
1003 OPT_INTEGER(0, "progress", &progress,
1004 N_("show progress after <n> objects")),
1005 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1006 N_("select handling of signed tags"),
1007 parse_opt_signed_tag_mode),
1008 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1009 N_("select handling of tags that tag filtered objects"),
1010 parse_opt_tag_of_filtered_mode),
1011 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1012 N_("Dump marks to this file")),
1013 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1014 N_("Import marks from this file")),
1015 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1016 N_("Fake a tagger when tags lack one")),
1017 OPT_BOOL(0, "full-tree", &full_tree,
1018 N_("Output full tree for each commit")),
1019 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1020 N_("Use the done feature to terminate the stream")),
1021 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
1022 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1023 N_("Apply refspec to exported refs")),
1024 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1025 OPT_END()
1026 };
1027
1028 if (argc == 1)
1029 usage_with_options (fast_export_usage, options);
1030
1031 /* we handle encodings */
1032 git_config(git_default_config, NULL);
1033
1034 init_revisions(&revs, prefix);
1035 init_revision_sources(&revision_sources);
1036 revs.topo_order = 1;
1037 revs.sources = &revision_sources;
1038 revs.rewrite_parents = 1;
1039 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1040 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1041 argc = setup_revisions(argc, argv, &revs, NULL);
1042 if (argc > 1)
1043 usage_with_options (fast_export_usage, options);
1044
1045 if (refspecs_list.nr) {
1046 int i;
1047
1048 for (i = 0; i < refspecs_list.nr; i++)
1049 refspec_append(&refspecs, refspecs_list.items[i].string);
1050
1051 string_list_clear(&refspecs_list, 1);
1052 }
1053
1054 if (use_done_feature)
1055 printf("feature done\n");
1056
1057 if (import_filename)
1058 import_marks(import_filename);
1059 lastimportid = last_idnum;
1060
1061 if (import_filename && revs.prune_data.nr)
1062 full_tree = 1;
1063
1064 get_tags_and_duplicates(&revs.cmdline);
1065
1066 if (prepare_revision_walk(&revs))
1067 die("revision walk setup failed");
1068 revs.diffopt.format_callback = show_filemodify;
1069 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1070 revs.diffopt.flags.recursive = 1;
1071 while ((commit = get_revision(&revs))) {
1072 if (has_unshown_parent(commit)) {
1073 add_object_array(&commit->object, NULL, &commits);
1074 }
1075 else {
1076 handle_commit(commit, &revs, &paths_of_changed_objects);
1077 handle_tail(&commits, &revs, &paths_of_changed_objects);
1078 }
1079 }
1080
1081 handle_tags_and_duplicates();
1082 handle_deletes();
1083
1084 if (export_filename && lastimportid != last_idnum)
1085 export_marks(export_filename);
1086
1087 if (use_done_feature)
1088 printf("done\n");
1089
1090 refspec_clear(&refspecs);
1091
1092 return 0;
1093 }