2 * "git fast-export" builtin command
4 * Copyright (C) 2007 Johannes E. Schindelin
11 #include "object-store.h"
20 #include "string-list.h"
22 #include "parse-options.h"
26 #include "commit-slab.h"
28 static const char *fast_export_usage[] = {
29 N_("git fast-export [rev-list-opts]"),
34 static enum { SIGNED_TAG_ABORT, VERBATIM, WARN, WARN_STRIP, STRIP } signed_tag_mode = SIGNED_TAG_ABORT;
35 static enum { TAG_FILTERING_ABORT, DROP, REWRITE } tag_of_filtered_mode = TAG_FILTERING_ABORT;
36 static enum { REENCODE_ABORT, REENCODE_YES, REENCODE_NO } reencode_mode = REENCODE_ABORT;
37 static int fake_missing_tagger;
38 static int use_done_feature;
41 static int reference_excluded_commits;
42 static int show_original_ids;
44 static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
45 static struct string_list tag_refs = STRING_LIST_INIT_NODUP;
46 static struct refspec refspecs = REFSPEC_INIT_FETCH;
48 static struct revision_sources revision_sources;
50 static int parse_opt_signed_tag_mode(const struct option *opt,
51 const char *arg, int unset)
53 if (unset || !strcmp(arg, "abort"))
54 signed_tag_mode = SIGNED_TAG_ABORT;
55 else if (!strcmp(arg, "verbatim") || !strcmp(arg, "ignore"))
56 signed_tag_mode = VERBATIM;
57 else if (!strcmp(arg, "warn"))
58 signed_tag_mode = WARN;
59 else if (!strcmp(arg, "warn-strip"))
60 signed_tag_mode = WARN_STRIP;
61 else if (!strcmp(arg, "strip"))
62 signed_tag_mode = STRIP;
64 return error("Unknown signed-tags mode: %s", arg);
68 static int parse_opt_tag_of_filtered_mode(const struct option *opt,
69 const char *arg, int unset)
71 if (unset || !strcmp(arg, "abort"))
72 tag_of_filtered_mode = TAG_FILTERING_ABORT;
73 else if (!strcmp(arg, "drop"))
74 tag_of_filtered_mode = DROP;
75 else if (!strcmp(arg, "rewrite"))
76 tag_of_filtered_mode = REWRITE;
78 return error("Unknown tag-of-filtered mode: %s", arg);
82 static int parse_opt_reencode_mode(const struct option *opt,
83 const char *arg, int unset)
86 reencode_mode = REENCODE_ABORT;
90 switch (git_parse_maybe_bool(arg)) {
92 reencode_mode = REENCODE_NO;
95 reencode_mode = REENCODE_YES;
98 if (!strcasecmp(arg, "abort"))
99 reencode_mode = REENCODE_ABORT;
101 return error("Unknown reencoding mode: %s", arg);
107 static struct decoration idnums;
108 static uint32_t last_idnum;
110 static int has_unshown_parent(struct commit *commit)
112 struct commit_list *parent;
114 for (parent = commit->parents; parent; parent = parent->next)
115 if (!(parent->item->object.flags & SHOWN) &&
116 !(parent->item->object.flags & UNINTERESTING))
121 struct anonymized_entry {
122 struct hashmap_entry hash;
129 static int anonymized_entry_cmp(const void *unused_cmp_data,
130 const struct hashmap_entry *eptr,
131 const struct hashmap_entry *entry_or_key,
132 const void *unused_keydata)
134 const struct anonymized_entry *a, *b;
136 a = container_of(eptr, const struct anonymized_entry, hash);
137 b = container_of(entry_or_key, const struct anonymized_entry, hash);
139 return a->orig_len != b->orig_len ||
140 memcmp(a->orig, b->orig, a->orig_len);
144 * Basically keep a cache of X->Y so that we can repeatedly replace
145 * the same anonymized string with another. The actual generation
146 * is farmed out to the generate function.
148 static const void *anonymize_mem(struct hashmap *map,
149 void *(*generate)(const void *, size_t *),
150 const void *orig, size_t *len)
152 struct anonymized_entry key, *ret;
155 hashmap_init(map, anonymized_entry_cmp, NULL, 0);
157 hashmap_entry_init(&key.hash, memhash(orig, *len));
160 ret = hashmap_get_entry(map, &key, hash, NULL);
163 ret = xmalloc(sizeof(*ret));
164 hashmap_entry_init(&ret->hash, key.hash.hash);
165 ret->orig = xmemdupz(orig, *len);
166 ret->orig_len = *len;
167 ret->anon = generate(orig, len);
168 ret->anon_len = *len;
169 hashmap_put(map, &ret->hash);
172 *len = ret->anon_len;
177 * We anonymize each component of a path individually,
178 * so that paths a/b and a/c will share a common root.
179 * The paths are cached via anonymize_mem so that repeated
180 * lookups for "a" will yield the same value.
182 static void anonymize_path(struct strbuf *out, const char *path,
184 void *(*generate)(const void *, size_t *))
187 const char *end_of_component = strchrnul(path, '/');
188 size_t len = end_of_component - path;
189 const char *c = anonymize_mem(map, generate, path, &len);
190 strbuf_add(out, c, len);
191 path = end_of_component;
193 strbuf_addch(out, *path++);
197 static inline void *mark_to_ptr(uint32_t mark)
199 return (void *)(uintptr_t)mark;
202 static inline uint32_t ptr_to_mark(void * mark)
204 return (uint32_t)(uintptr_t)mark;
207 static inline void mark_object(struct object *object, uint32_t mark)
209 add_decoration(&idnums, object, mark_to_ptr(mark));
212 static inline void mark_next_object(struct object *object)
214 mark_object(object, ++last_idnum);
217 static int get_object_mark(struct object *object)
219 void *decoration = lookup_decoration(&idnums, object);
222 return ptr_to_mark(decoration);
225 static struct commit *rewrite_commit(struct commit *p)
228 if (p->parents && p->parents->next)
230 if (p->object.flags & UNINTERESTING)
232 if (!(p->object.flags & TREESAME))
236 p = p->parents->item;
241 static void show_progress(void)
243 static int counter = 0;
246 if ((++counter % progress) == 0)
247 printf("progress %d objects\n", counter);
251 * Ideally we would want some transformation of the blob data here
252 * that is unreversible, but would still be the same size and have
253 * the same data relationship to other blobs (so that we get the same
254 * delta and packing behavior as the original). But the first and last
255 * requirements there are probably mutually exclusive, so let's take
256 * the easy way out for now, and just generate arbitrary content.
258 * There's no need to cache this result with anonymize_mem, since
259 * we already handle blob content caching with marks.
261 static char *anonymize_blob(unsigned long *size)
264 struct strbuf out = STRBUF_INIT;
265 strbuf_addf(&out, "anonymous blob %d", counter++);
267 return strbuf_detach(&out, NULL);
270 static void export_blob(const struct object_id *oid)
273 enum object_type type;
275 struct object *object;
281 if (is_null_oid(oid))
284 object = lookup_object(the_repository, oid);
285 if (object && object->flags & SHOWN)
289 buf = anonymize_blob(&size);
290 object = (struct object *)lookup_blob(the_repository, oid);
293 buf = read_object_file(oid, &type, &size);
295 die("could not read blob %s", oid_to_hex(oid));
296 if (check_object_signature(the_repository, oid, buf, size,
297 type_name(type)) < 0)
298 die("oid mismatch in blob %s", oid_to_hex(oid));
299 object = parse_object_buffer(the_repository, oid, type,
304 die("Could not read blob %s", oid_to_hex(oid));
306 mark_next_object(object);
308 printf("blob\nmark :%"PRIu32"\n", last_idnum);
309 if (show_original_ids)
310 printf("original-oid %s\n", oid_to_hex(oid));
311 printf("data %"PRIuMAX"\n", (uintmax_t)size);
312 if (size && fwrite(buf, size, 1, stdout) != 1)
313 die_errno("could not write blob '%s'", oid_to_hex(oid));
318 object->flags |= SHOWN;
323 static int depth_first(const void *a_, const void *b_)
325 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
326 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
327 const char *name_a, *name_b;
328 int len_a, len_b, len;
331 name_a = a->one ? a->one->path : a->two->path;
332 name_b = b->one ? b->one->path : b->two->path;
334 len_a = strlen(name_a);
335 len_b = strlen(name_b);
336 len = (len_a < len_b) ? len_a : len_b;
338 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
339 cmp = memcmp(name_a, name_b, len);
346 * Move 'R'ename entries last so that all references of the file
347 * appear in the output before it is renamed (e.g., when a file
348 * was copied and renamed in the same commit).
350 return (a->status == 'R') - (b->status == 'R');
353 static void print_path_1(const char *path)
355 int need_quote = quote_c_style(path, NULL, NULL, 0);
357 quote_c_style(path, NULL, stdout, 0);
358 else if (strchr(path, ' '))
359 printf("\"%s\"", path);
364 static void *anonymize_path_component(const void *path, size_t *len)
367 struct strbuf out = STRBUF_INIT;
368 strbuf_addf(&out, "path%d", counter++);
369 return strbuf_detach(&out, len);
372 static void print_path(const char *path)
377 static struct hashmap paths;
378 static struct strbuf anon = STRBUF_INIT;
380 anonymize_path(&anon, path, &paths, anonymize_path_component);
381 print_path_1(anon.buf);
386 static void *generate_fake_oid(const void *old, size_t *len)
388 static uint32_t counter = 1; /* avoid null oid */
389 const unsigned hashsz = the_hash_algo->rawsz;
390 struct object_id oid;
391 char *hex = xmallocz(GIT_MAX_HEXSZ);
394 put_be32(oid.hash + hashsz - 4, counter++);
395 return oid_to_hex_r(hex, &oid);
398 static const char *anonymize_oid(const char *oid_hex)
400 static struct hashmap objs;
401 size_t len = strlen(oid_hex);
402 return anonymize_mem(&objs, generate_fake_oid, oid_hex, &len);
405 static void show_filemodify(struct diff_queue_struct *q,
406 struct diff_options *options, void *data)
409 struct string_list *changed = data;
412 * Handle files below a directory first, in case they are all deleted
413 * and the directory changes to a file or symlink.
415 QSORT(q->queue, q->nr, depth_first);
417 for (i = 0; i < q->nr; i++) {
418 struct diff_filespec *ospec = q->queue[i]->one;
419 struct diff_filespec *spec = q->queue[i]->two;
421 switch (q->queue[i]->status) {
422 case DIFF_STATUS_DELETED:
424 print_path(spec->path);
425 string_list_insert(changed, spec->path);
429 case DIFF_STATUS_COPIED:
430 case DIFF_STATUS_RENAMED:
432 * If a change in the file corresponding to ospec->path
433 * has been observed, we cannot trust its contents
434 * because the diff is calculated based on the prior
435 * contents, not the current contents. So, declare a
436 * copy or rename only if there was no change observed.
438 if (!string_list_has_string(changed, ospec->path)) {
439 printf("%c ", q->queue[i]->status);
440 print_path(ospec->path);
442 print_path(spec->path);
443 string_list_insert(changed, spec->path);
446 if (oideq(&ospec->oid, &spec->oid) &&
447 ospec->mode == spec->mode)
452 case DIFF_STATUS_TYPE_CHANGED:
453 case DIFF_STATUS_MODIFIED:
454 case DIFF_STATUS_ADDED:
456 * Links refer to objects in another repositories;
457 * output the SHA-1 verbatim.
459 if (no_data || S_ISGITLINK(spec->mode))
460 printf("M %06o %s ", spec->mode,
462 anonymize_oid(oid_to_hex(&spec->oid)) :
463 oid_to_hex(&spec->oid));
465 struct object *object = lookup_object(the_repository,
467 printf("M %06o :%d ", spec->mode,
468 get_object_mark(object));
470 print_path(spec->path);
471 string_list_insert(changed, spec->path);
476 die("Unexpected comparison status '%c' for %s, %s",
478 ospec->path ? ospec->path : "none",
479 spec->path ? spec->path : "none");
484 static const char *find_encoding(const char *begin, const char *end)
486 const char *needle = "\nencoding ";
489 bol = memmem(begin, end ? end - begin : strlen(begin),
490 needle, strlen(needle));
493 bol += strlen(needle);
494 eol = strchrnul(bol, '\n');
499 static void *anonymize_ref_component(const void *old, size_t *len)
502 struct strbuf out = STRBUF_INIT;
503 strbuf_addf(&out, "ref%d", counter++);
504 return strbuf_detach(&out, len);
507 static const char *anonymize_refname(const char *refname)
510 * If any of these prefixes is found, we will leave it intact
511 * so that tags remain tags and so forth.
513 static const char *prefixes[] = {
519 static struct hashmap refs;
520 static struct strbuf anon = STRBUF_INIT;
524 * We also leave "master" as a special case, since it does not reveal
525 * anything interesting.
527 if (!strcmp(refname, "refs/heads/master"))
531 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
532 if (skip_prefix(refname, prefixes[i], &refname)) {
533 strbuf_addstr(&anon, prefixes[i]);
538 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
543 * We do not even bother to cache commit messages, as they are unlikely
544 * to be repeated verbatim, and it is not that interesting when they are.
546 static char *anonymize_commit_message(const char *old)
549 return xstrfmt("subject %d\n\nbody\n", counter++);
552 static struct hashmap idents;
553 static void *anonymize_ident(const void *old, size_t *len)
556 struct strbuf out = STRBUF_INIT;
557 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
559 return strbuf_detach(&out, len);
563 * Our strategy here is to anonymize the names and email addresses,
564 * but keep timestamps intact, as they influence things like traversal
565 * order (and by themselves should not be too revealing).
567 static void anonymize_ident_line(const char **beg, const char **end)
569 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
570 static unsigned which_buffer;
573 struct ident_split split;
574 const char *end_of_header;
576 out = &buffers[which_buffer++];
577 which_buffer %= ARRAY_SIZE(buffers);
580 /* skip "committer", "author", "tagger", etc */
581 end_of_header = strchr(*beg, ' ');
583 BUG("malformed line fed to anonymize_ident_line: %.*s",
584 (int)(*end - *beg), *beg);
586 strbuf_add(out, *beg, end_of_header - *beg);
588 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
593 len = split.mail_end - split.name_begin;
594 ident = anonymize_mem(&idents, anonymize_ident,
595 split.name_begin, &len);
596 strbuf_add(out, ident, len);
597 strbuf_addch(out, ' ');
598 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
600 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
604 *end = out->buf + out->len;
607 static void handle_commit(struct commit *commit, struct rev_info *rev,
608 struct string_list *paths_of_changed_objects)
610 int saved_output_format = rev->diffopt.output_format;
611 const char *commit_buffer;
612 const char *author, *author_end, *committer, *committer_end;
613 const char *encoding, *message;
614 char *reencoded = NULL;
615 struct commit_list *p;
619 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
621 parse_commit_or_die(commit);
622 commit_buffer = get_commit_buffer(commit, NULL);
623 author = strstr(commit_buffer, "\nauthor ");
625 die("could not find author in commit %s",
626 oid_to_hex(&commit->object.oid));
628 author_end = strchrnul(author, '\n');
629 committer = strstr(author_end, "\ncommitter ");
631 die("could not find committer in commit %s",
632 oid_to_hex(&commit->object.oid));
634 committer_end = strchrnul(committer, '\n');
635 message = strstr(committer_end, "\n\n");
636 encoding = find_encoding(committer_end, message);
640 if (commit->parents &&
641 (get_object_mark(&commit->parents->item->object) != 0 ||
642 reference_excluded_commits) &&
644 parse_commit_or_die(commit->parents->item);
645 diff_tree_oid(get_commit_tree_oid(commit->parents->item),
646 get_commit_tree_oid(commit), "", &rev->diffopt);
649 diff_root_tree_oid(get_commit_tree_oid(commit),
652 /* Export the referenced blobs, and remember the marks. */
653 for (i = 0; i < diff_queued_diff.nr; i++)
654 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
655 export_blob(&diff_queued_diff.queue[i]->two->oid);
657 refname = *revision_sources_at(&revision_sources, commit);
659 * FIXME: string_list_remove() below for each ref is overall
660 * O(N^2). Compared to a history walk and diffing trees, this is
661 * just lost in the noise in practice. However, theoretically a
662 * repo may have enough refs for this to become slow.
664 string_list_remove(&extra_refs, refname, 0);
666 refname = anonymize_refname(refname);
667 anonymize_ident_line(&committer, &committer_end);
668 anonymize_ident_line(&author, &author_end);
671 mark_next_object(&commit->object);
673 reencoded = anonymize_commit_message(message);
674 } else if (encoding) {
675 switch(reencode_mode) {
677 reencoded = reencode_string(message, "UTF-8", encoding);
682 die("Encountered commit-specific encoding %s in commit "
683 "%s; use --reencode=[yes|no] to handle it",
684 encoding, oid_to_hex(&commit->object.oid));
687 if (!commit->parents)
688 printf("reset %s\n", refname);
689 printf("commit %s\nmark :%"PRIu32"\n", refname, last_idnum);
690 if (show_original_ids)
691 printf("original-oid %s\n", oid_to_hex(&commit->object.oid));
692 printf("%.*s\n%.*s\n",
693 (int)(author_end - author), author,
694 (int)(committer_end - committer), committer);
695 if (!reencoded && encoding)
696 printf("encoding %s\n", encoding);
697 printf("data %u\n%s",
699 ? strlen(reencoded) : message
700 ? strlen(message) : 0),
701 reencoded ? reencoded : message ? message : "");
703 unuse_commit_buffer(commit, commit_buffer);
705 for (i = 0, p = commit->parents; p; p = p->next) {
706 struct object *obj = &p->item->object;
707 int mark = get_object_mark(obj);
709 if (!mark && !reference_excluded_commits)
716 printf(":%d\n", mark);
720 anonymize_oid(oid_to_hex(&obj->oid)) :
721 oid_to_hex(&obj->oid));
726 printf("deleteall\n");
727 log_tree_diff_flush(rev);
728 string_list_clear(paths_of_changed_objects, 0);
729 rev->diffopt.output_format = saved_output_format;
736 static void *anonymize_tag(const void *old, size_t *len)
739 struct strbuf out = STRBUF_INIT;
740 strbuf_addf(&out, "tag message %d", counter++);
741 return strbuf_detach(&out, len);
744 static void handle_tail(struct object_array *commits, struct rev_info *revs,
745 struct string_list *paths_of_changed_objects)
747 struct commit *commit;
748 while (commits->nr) {
749 commit = (struct commit *)object_array_pop(commits);
750 if (has_unshown_parent(commit)) {
751 /* Queue again, to be handled later */
752 add_object_array(&commit->object, NULL, commits);
755 handle_commit(commit, revs, paths_of_changed_objects);
759 static void handle_tag(const char *name, struct tag *tag)
762 enum object_type type;
764 const char *tagger, *tagger_end, *message;
765 size_t message_size = 0;
766 struct object *tagged;
770 /* Trees have no identifier in fast-export output, thus we have no way
771 * to output tags of trees, tags of tags of trees, etc. Simply omit
774 tagged = tag->tagged;
775 while (tagged->type == OBJ_TAG) {
776 tagged = ((struct tag *)tagged)->tagged;
778 if (tagged->type == OBJ_TREE) {
779 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
780 oid_to_hex(&tag->object.oid));
784 buf = read_object_file(&tag->object.oid, &type, &size);
786 die("could not read tag %s", oid_to_hex(&tag->object.oid));
787 message = memmem(buf, size, "\n\n", 2);
790 message_size = strlen(message);
792 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
794 if (fake_missing_tagger)
795 tagger = "tagger Unspecified Tagger "
796 "<unspecified-tagger> 0 +0000";
799 tagger_end = tagger + strlen(tagger);
802 tagger_end = strchrnul(tagger, '\n');
804 anonymize_ident_line(&tagger, &tagger_end);
808 name = anonymize_refname(name);
810 static struct hashmap tags;
811 message = anonymize_mem(&tags, anonymize_tag,
812 message, &message_size);
816 /* handle signed tags */
818 const char *signature = strstr(message,
819 "\n-----BEGIN PGP SIGNATURE-----\n");
821 switch(signed_tag_mode) {
822 case SIGNED_TAG_ABORT:
823 die("encountered signed tag %s; use "
824 "--signed-tags=<mode> to handle it",
825 oid_to_hex(&tag->object.oid));
827 warning("exporting signed tag %s",
828 oid_to_hex(&tag->object.oid));
833 warning("stripping signature from tag %s",
834 oid_to_hex(&tag->object.oid));
837 message_size = signature + 1 - message;
842 /* handle tag->tagged having been filtered out due to paths specified */
843 tagged = tag->tagged;
844 tagged_mark = get_object_mark(tagged);
846 switch(tag_of_filtered_mode) {
847 case TAG_FILTERING_ABORT:
848 die("tag %s tags unexported object; use "
849 "--tag-of-filtered-object=<mode> to handle it",
850 oid_to_hex(&tag->object.oid));
852 /* Ignore this tag altogether */
856 if (tagged->type == OBJ_TAG && !mark_tags) {
857 die(_("Error: Cannot export nested tags unless --mark-tags is specified."));
858 } else if (tagged->type == OBJ_COMMIT) {
859 p = rewrite_commit((struct commit *)tagged);
861 printf("reset %s\nfrom %s\n\n",
862 name, oid_to_hex(&null_oid));
866 tagged_mark = get_object_mark(&p->object);
868 /* tagged->type is either OBJ_BLOB or OBJ_TAG */
869 tagged_mark = get_object_mark(tagged);
874 if (tagged->type == OBJ_TAG) {
875 printf("reset %s\nfrom %s\n\n",
876 name, oid_to_hex(&null_oid));
878 skip_prefix(name, "refs/tags/", &name);
879 printf("tag %s\n", name);
881 mark_next_object(&tag->object);
882 printf("mark :%"PRIu32"\n", last_idnum);
885 printf("from :%d\n", tagged_mark);
887 printf("from %s\n", oid_to_hex(&tagged->oid));
889 if (show_original_ids)
890 printf("original-oid %s\n", oid_to_hex(&tag->object.oid));
891 printf("%.*s%sdata %d\n%.*s\n",
892 (int)(tagger_end - tagger), tagger,
893 tagger == tagger_end ? "" : "\n",
894 (int)message_size, (int)message_size, message ? message : "");
898 static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
900 switch (e->item->type) {
902 return (struct commit *)e->item;
904 struct tag *tag = (struct tag *)e->item;
906 /* handle nested tags */
907 while (tag && tag->object.type == OBJ_TAG) {
908 parse_object(the_repository, &tag->object.oid);
909 string_list_append(&tag_refs, full_name)->util = tag;
910 tag = (struct tag *)tag->tagged;
913 die("Tag %s points nowhere?", e->name);
914 return (struct commit *)tag;
922 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
926 for (i = 0; i < info->nr; i++) {
927 struct rev_cmdline_entry *e = info->rev + i;
928 struct object_id oid;
929 struct commit *commit;
932 if (e->flags & UNINTERESTING)
935 if (dwim_ref(e->name, strlen(e->name), &oid, &full_name) != 1)
940 private = apply_refspecs(&refspecs, full_name);
947 commit = get_commit(e, full_name);
949 warning("%s: Unexpected object of type %s, skipping.",
951 type_name(e->item->type));
955 switch(commit->object.type) {
959 export_blob(&commit->object.oid);
961 default: /* OBJ_TAG (nested tags) is already handled */
962 warning("Tag points to object of unexpected type %s, skipping.",
963 type_name(commit->object.type));
968 * Make sure this ref gets properly updated eventually, whether
969 * through a commit or manually at the end.
971 if (e->item->type != OBJ_TAG)
972 string_list_append(&extra_refs, full_name)->util = commit;
974 if (!*revision_sources_at(&revision_sources, commit))
975 *revision_sources_at(&revision_sources, commit) = full_name;
978 string_list_sort(&extra_refs);
979 string_list_remove_duplicates(&extra_refs, 0);
982 static void handle_tags_and_duplicates(struct string_list *extras)
984 struct commit *commit;
987 for (i = extras->nr - 1; i >= 0; i--) {
988 const char *name = extras->items[i].string;
989 struct object *object = extras->items[i].util;
992 switch (object->type) {
994 handle_tag(name, (struct tag *)object);
998 name = anonymize_refname(name);
999 /* create refs pointing to already seen commits */
1000 commit = rewrite_commit((struct commit *)object);
1003 * Neither this object nor any of its
1004 * ancestors touch any relevant paths, so
1005 * it has been filtered to nothing. Delete
1008 printf("reset %s\nfrom %s\n\n",
1009 name, oid_to_hex(&null_oid));
1013 mark = get_object_mark(&commit->object);
1016 * Getting here means we have a commit which
1017 * was excluded by a negative refspec (e.g.
1018 * fast-export ^master master). If we are
1019 * referencing excluded commits, set the ref
1020 * to the exact commit. Otherwise, the user
1021 * wants the branch exported but every commit
1022 * in its history to be deleted, which basically
1023 * just means deletion of the ref.
1025 if (!reference_excluded_commits) {
1026 /* delete the ref */
1027 printf("reset %s\nfrom %s\n\n",
1028 name, oid_to_hex(&null_oid));
1031 /* set ref to commit using oid, not mark */
1032 printf("reset %s\nfrom %s\n\n", name,
1033 oid_to_hex(&commit->object.oid));
1037 printf("reset %s\nfrom :%d\n\n", name, mark
1045 static void export_marks(char *file)
1049 struct decoration_entry *deco = idnums.entries;
1053 f = fopen_for_writing(file);
1055 die_errno("Unable to open marks file %s for writing.", file);
1057 for (i = 0; i < idnums.size; i++) {
1058 if (deco->base && deco->base->type == 1) {
1059 mark = ptr_to_mark(deco->decoration);
1060 if (fprintf(f, ":%"PRIu32" %s\n", mark,
1061 oid_to_hex(&deco->base->oid)) < 0) {
1072 error("Unable to write marks file %s.", file);
1075 static void import_marks(char *input_file, int check_exists)
1081 if (check_exists && stat(input_file, &sb))
1084 f = xfopen(input_file, "r");
1085 while (fgets(line, sizeof(line), f)) {
1087 char *line_end, *mark_end;
1088 struct object_id oid;
1089 struct object *object;
1090 struct commit *commit;
1091 enum object_type type;
1093 line_end = strchr(line, '\n');
1094 if (line[0] != ':' || !line_end)
1095 die("corrupt mark line: %s", line);
1098 mark = strtoumax(line + 1, &mark_end, 10);
1099 if (!mark || mark_end == line + 1
1100 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
1101 die("corrupt mark line: %s", line);
1103 if (last_idnum < mark)
1106 type = oid_object_info(the_repository, &oid, NULL);
1108 die("object not found: %s", oid_to_hex(&oid));
1110 if (type != OBJ_COMMIT)
1114 commit = lookup_commit(the_repository, &oid);
1116 die("not a commit? can't happen: %s", oid_to_hex(&oid));
1118 object = &commit->object;
1120 if (object->flags & SHOWN)
1121 error("Object %s already has a mark", oid_to_hex(&oid));
1123 mark_object(object, mark);
1125 object->flags |= SHOWN;
1130 static void handle_deletes(void)
1133 for (i = 0; i < refspecs.nr; i++) {
1134 struct refspec_item *refspec = &refspecs.items[i];
1138 printf("reset %s\nfrom %s\n\n",
1139 refspec->dst, oid_to_hex(&null_oid));
1143 int cmd_fast_export(int argc, const char **argv, const char *prefix)
1145 struct rev_info revs;
1146 struct object_array commits = OBJECT_ARRAY_INIT;
1147 struct commit *commit;
1148 char *export_filename = NULL,
1149 *import_filename = NULL,
1150 *import_filename_if_exists = NULL;
1151 uint32_t lastimportid;
1152 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
1153 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
1154 struct option options[] = {
1155 OPT_INTEGER(0, "progress", &progress,
1156 N_("show progress after <n> objects")),
1157 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1158 N_("select handling of signed tags"),
1159 parse_opt_signed_tag_mode),
1160 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1161 N_("select handling of tags that tag filtered objects"),
1162 parse_opt_tag_of_filtered_mode),
1163 OPT_CALLBACK(0, "reencode", &reencode_mode, N_("mode"),
1164 N_("select handling of commit messages in an alternate encoding"),
1165 parse_opt_reencode_mode),
1166 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1167 N_("Dump marks to this file")),
1168 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1169 N_("Import marks from this file")),
1170 OPT_STRING(0, "import-marks-if-exists",
1171 &import_filename_if_exists,
1173 N_("Import marks from this file if it exists")),
1174 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1175 N_("Fake a tagger when tags lack one")),
1176 OPT_BOOL(0, "full-tree", &full_tree,
1177 N_("Output full tree for each commit")),
1178 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1179 N_("Use the done feature to terminate the stream")),
1180 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
1181 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1182 N_("Apply refspec to exported refs")),
1183 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1184 OPT_BOOL(0, "reference-excluded-parents",
1185 &reference_excluded_commits, N_("Reference parents which are not in fast-export stream by object id")),
1186 OPT_BOOL(0, "show-original-ids", &show_original_ids,
1187 N_("Show original object ids of blobs/commits")),
1188 OPT_BOOL(0, "mark-tags", &mark_tags,
1189 N_("Label tags with mark ids")),
1195 usage_with_options (fast_export_usage, options);
1197 /* we handle encodings */
1198 git_config(git_default_config, NULL);
1200 repo_init_revisions(the_repository, &revs, prefix);
1201 init_revision_sources(&revision_sources);
1202 revs.topo_order = 1;
1203 revs.sources = &revision_sources;
1204 revs.rewrite_parents = 1;
1205 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1206 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1207 argc = setup_revisions(argc, argv, &revs, NULL);
1209 usage_with_options (fast_export_usage, options);
1211 if (refspecs_list.nr) {
1214 for (i = 0; i < refspecs_list.nr; i++)
1215 refspec_append(&refspecs, refspecs_list.items[i].string);
1217 string_list_clear(&refspecs_list, 1);
1220 if (use_done_feature)
1221 printf("feature done\n");
1223 if (import_filename && import_filename_if_exists)
1224 die(_("Cannot pass both --import-marks and --import-marks-if-exists"));
1225 if (import_filename)
1226 import_marks(import_filename, 0);
1227 else if (import_filename_if_exists)
1228 import_marks(import_filename_if_exists, 1);
1229 lastimportid = last_idnum;
1231 if (import_filename && revs.prune_data.nr)
1234 get_tags_and_duplicates(&revs.cmdline);
1236 if (prepare_revision_walk(&revs))
1237 die("revision walk setup failed");
1238 revs.diffopt.format_callback = show_filemodify;
1239 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1240 revs.diffopt.flags.recursive = 1;
1241 while ((commit = get_revision(&revs))) {
1242 if (has_unshown_parent(commit)) {
1243 add_object_array(&commit->object, NULL, &commits);
1246 handle_commit(commit, &revs, &paths_of_changed_objects);
1247 handle_tail(&commits, &revs, &paths_of_changed_objects);
1251 handle_tags_and_duplicates(&extra_refs);
1252 handle_tags_and_duplicates(&tag_refs);
1255 if (export_filename && lastimportid != last_idnum)
1256 export_marks(export_filename);
1258 if (use_done_feature)
1261 refspec_clear(&refspecs);