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 = xstrdup(orig);
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 unsigned char *out = xcalloc(hashsz, 1);
391 put_be32(out + hashsz - 4, counter++);
395 static const struct object_id *anonymize_oid(const struct object_id *oid)
397 static struct hashmap objs;
398 size_t len = the_hash_algo->rawsz;
399 return anonymize_mem(&objs, generate_fake_oid, oid, &len);
402 static void show_filemodify(struct diff_queue_struct *q,
403 struct diff_options *options, void *data)
406 struct string_list *changed = data;
409 * Handle files below a directory first, in case they are all deleted
410 * and the directory changes to a file or symlink.
412 QSORT(q->queue, q->nr, depth_first);
414 for (i = 0; i < q->nr; i++) {
415 struct diff_filespec *ospec = q->queue[i]->one;
416 struct diff_filespec *spec = q->queue[i]->two;
418 switch (q->queue[i]->status) {
419 case DIFF_STATUS_DELETED:
421 print_path(spec->path);
422 string_list_insert(changed, spec->path);
426 case DIFF_STATUS_COPIED:
427 case DIFF_STATUS_RENAMED:
429 * If a change in the file corresponding to ospec->path
430 * has been observed, we cannot trust its contents
431 * because the diff is calculated based on the prior
432 * contents, not the current contents. So, declare a
433 * copy or rename only if there was no change observed.
435 if (!string_list_has_string(changed, ospec->path)) {
436 printf("%c ", q->queue[i]->status);
437 print_path(ospec->path);
439 print_path(spec->path);
440 string_list_insert(changed, spec->path);
443 if (oideq(&ospec->oid, &spec->oid) &&
444 ospec->mode == spec->mode)
449 case DIFF_STATUS_TYPE_CHANGED:
450 case DIFF_STATUS_MODIFIED:
451 case DIFF_STATUS_ADDED:
453 * Links refer to objects in another repositories;
454 * output the SHA-1 verbatim.
456 if (no_data || S_ISGITLINK(spec->mode))
457 printf("M %06o %s ", spec->mode,
458 oid_to_hex(anonymize ?
459 anonymize_oid(&spec->oid) :
462 struct object *object = lookup_object(the_repository,
464 printf("M %06o :%d ", spec->mode,
465 get_object_mark(object));
467 print_path(spec->path);
468 string_list_insert(changed, spec->path);
473 die("Unexpected comparison status '%c' for %s, %s",
475 ospec->path ? ospec->path : "none",
476 spec->path ? spec->path : "none");
481 static const char *find_encoding(const char *begin, const char *end)
483 const char *needle = "\nencoding ";
486 bol = memmem(begin, end ? end - begin : strlen(begin),
487 needle, strlen(needle));
490 bol += strlen(needle);
491 eol = strchrnul(bol, '\n');
496 static void *anonymize_ref_component(const void *old, size_t *len)
499 struct strbuf out = STRBUF_INIT;
500 strbuf_addf(&out, "ref%d", counter++);
501 return strbuf_detach(&out, len);
504 static const char *anonymize_refname(const char *refname)
507 * If any of these prefixes is found, we will leave it intact
508 * so that tags remain tags and so forth.
510 static const char *prefixes[] = {
516 static struct hashmap refs;
517 static struct strbuf anon = STRBUF_INIT;
521 * We also leave "master" as a special case, since it does not reveal
522 * anything interesting.
524 if (!strcmp(refname, "refs/heads/master"))
528 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
529 if (skip_prefix(refname, prefixes[i], &refname)) {
530 strbuf_addstr(&anon, prefixes[i]);
535 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
540 * We do not even bother to cache commit messages, as they are unlikely
541 * to be repeated verbatim, and it is not that interesting when they are.
543 static char *anonymize_commit_message(const char *old)
546 return xstrfmt("subject %d\n\nbody\n", counter++);
549 static struct hashmap idents;
550 static void *anonymize_ident(const void *old, size_t *len)
553 struct strbuf out = STRBUF_INIT;
554 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
556 return strbuf_detach(&out, len);
560 * Our strategy here is to anonymize the names and email addresses,
561 * but keep timestamps intact, as they influence things like traversal
562 * order (and by themselves should not be too revealing).
564 static void anonymize_ident_line(const char **beg, const char **end)
566 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
567 static unsigned which_buffer;
570 struct ident_split split;
571 const char *end_of_header;
573 out = &buffers[which_buffer++];
574 which_buffer %= ARRAY_SIZE(buffers);
577 /* skip "committer", "author", "tagger", etc */
578 end_of_header = strchr(*beg, ' ');
580 BUG("malformed line fed to anonymize_ident_line: %.*s",
581 (int)(*end - *beg), *beg);
583 strbuf_add(out, *beg, end_of_header - *beg);
585 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
590 len = split.mail_end - split.name_begin;
591 ident = anonymize_mem(&idents, anonymize_ident,
592 split.name_begin, &len);
593 strbuf_add(out, ident, len);
594 strbuf_addch(out, ' ');
595 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
597 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
601 *end = out->buf + out->len;
604 static void handle_commit(struct commit *commit, struct rev_info *rev,
605 struct string_list *paths_of_changed_objects)
607 int saved_output_format = rev->diffopt.output_format;
608 const char *commit_buffer;
609 const char *author, *author_end, *committer, *committer_end;
610 const char *encoding, *message;
611 char *reencoded = NULL;
612 struct commit_list *p;
616 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
618 parse_commit_or_die(commit);
619 commit_buffer = get_commit_buffer(commit, NULL);
620 author = strstr(commit_buffer, "\nauthor ");
622 die("could not find author in commit %s",
623 oid_to_hex(&commit->object.oid));
625 author_end = strchrnul(author, '\n');
626 committer = strstr(author_end, "\ncommitter ");
628 die("could not find committer in commit %s",
629 oid_to_hex(&commit->object.oid));
631 committer_end = strchrnul(committer, '\n');
632 message = strstr(committer_end, "\n\n");
633 encoding = find_encoding(committer_end, message);
637 if (commit->parents &&
638 (get_object_mark(&commit->parents->item->object) != 0 ||
639 reference_excluded_commits) &&
641 parse_commit_or_die(commit->parents->item);
642 diff_tree_oid(get_commit_tree_oid(commit->parents->item),
643 get_commit_tree_oid(commit), "", &rev->diffopt);
646 diff_root_tree_oid(get_commit_tree_oid(commit),
649 /* Export the referenced blobs, and remember the marks. */
650 for (i = 0; i < diff_queued_diff.nr; i++)
651 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
652 export_blob(&diff_queued_diff.queue[i]->two->oid);
654 refname = *revision_sources_at(&revision_sources, commit);
656 * FIXME: string_list_remove() below for each ref is overall
657 * O(N^2). Compared to a history walk and diffing trees, this is
658 * just lost in the noise in practice. However, theoretically a
659 * repo may have enough refs for this to become slow.
661 string_list_remove(&extra_refs, refname, 0);
663 refname = anonymize_refname(refname);
664 anonymize_ident_line(&committer, &committer_end);
665 anonymize_ident_line(&author, &author_end);
668 mark_next_object(&commit->object);
670 reencoded = anonymize_commit_message(message);
671 } else if (encoding) {
672 switch(reencode_mode) {
674 reencoded = reencode_string(message, "UTF-8", encoding);
679 die("Encountered commit-specific encoding %s in commit "
680 "%s; use --reencode=[yes|no] to handle it",
681 encoding, oid_to_hex(&commit->object.oid));
684 if (!commit->parents)
685 printf("reset %s\n", refname);
686 printf("commit %s\nmark :%"PRIu32"\n", refname, last_idnum);
687 if (show_original_ids)
688 printf("original-oid %s\n", oid_to_hex(&commit->object.oid));
689 printf("%.*s\n%.*s\n",
690 (int)(author_end - author), author,
691 (int)(committer_end - committer), committer);
692 if (!reencoded && encoding)
693 printf("encoding %s\n", encoding);
694 printf("data %u\n%s",
696 ? strlen(reencoded) : message
697 ? strlen(message) : 0),
698 reencoded ? reencoded : message ? message : "");
700 unuse_commit_buffer(commit, commit_buffer);
702 for (i = 0, p = commit->parents; p; p = p->next) {
703 struct object *obj = &p->item->object;
704 int mark = get_object_mark(obj);
706 if (!mark && !reference_excluded_commits)
713 printf(":%d\n", mark);
715 printf("%s\n", oid_to_hex(anonymize ?
716 anonymize_oid(&obj->oid) :
722 printf("deleteall\n");
723 log_tree_diff_flush(rev);
724 string_list_clear(paths_of_changed_objects, 0);
725 rev->diffopt.output_format = saved_output_format;
732 static void *anonymize_tag(const void *old, size_t *len)
735 struct strbuf out = STRBUF_INIT;
736 strbuf_addf(&out, "tag message %d", counter++);
737 return strbuf_detach(&out, len);
740 static void handle_tail(struct object_array *commits, struct rev_info *revs,
741 struct string_list *paths_of_changed_objects)
743 struct commit *commit;
744 while (commits->nr) {
745 commit = (struct commit *)object_array_pop(commits);
746 if (has_unshown_parent(commit)) {
747 /* Queue again, to be handled later */
748 add_object_array(&commit->object, NULL, commits);
751 handle_commit(commit, revs, paths_of_changed_objects);
755 static void handle_tag(const char *name, struct tag *tag)
758 enum object_type type;
760 const char *tagger, *tagger_end, *message;
761 size_t message_size = 0;
762 struct object *tagged;
766 /* Trees have no identifier in fast-export output, thus we have no way
767 * to output tags of trees, tags of tags of trees, etc. Simply omit
770 tagged = tag->tagged;
771 while (tagged->type == OBJ_TAG) {
772 tagged = ((struct tag *)tagged)->tagged;
774 if (tagged->type == OBJ_TREE) {
775 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
776 oid_to_hex(&tag->object.oid));
780 buf = read_object_file(&tag->object.oid, &type, &size);
782 die("could not read tag %s", oid_to_hex(&tag->object.oid));
783 message = memmem(buf, size, "\n\n", 2);
786 message_size = strlen(message);
788 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
790 if (fake_missing_tagger)
791 tagger = "tagger Unspecified Tagger "
792 "<unspecified-tagger> 0 +0000";
795 tagger_end = tagger + strlen(tagger);
798 tagger_end = strchrnul(tagger, '\n');
800 anonymize_ident_line(&tagger, &tagger_end);
804 name = anonymize_refname(name);
806 static struct hashmap tags;
807 message = anonymize_mem(&tags, anonymize_tag,
808 message, &message_size);
812 /* handle signed tags */
814 const char *signature = strstr(message,
815 "\n-----BEGIN PGP SIGNATURE-----\n");
817 switch(signed_tag_mode) {
818 case SIGNED_TAG_ABORT:
819 die("encountered signed tag %s; use "
820 "--signed-tags=<mode> to handle it",
821 oid_to_hex(&tag->object.oid));
823 warning("exporting signed tag %s",
824 oid_to_hex(&tag->object.oid));
829 warning("stripping signature from tag %s",
830 oid_to_hex(&tag->object.oid));
833 message_size = signature + 1 - message;
838 /* handle tag->tagged having been filtered out due to paths specified */
839 tagged = tag->tagged;
840 tagged_mark = get_object_mark(tagged);
842 switch(tag_of_filtered_mode) {
843 case TAG_FILTERING_ABORT:
844 die("tag %s tags unexported object; use "
845 "--tag-of-filtered-object=<mode> to handle it",
846 oid_to_hex(&tag->object.oid));
848 /* Ignore this tag altogether */
852 if (tagged->type == OBJ_TAG && !mark_tags) {
853 die(_("Error: Cannot export nested tags unless --mark-tags is specified."));
854 } else if (tagged->type == OBJ_COMMIT) {
855 p = rewrite_commit((struct commit *)tagged);
857 printf("reset %s\nfrom %s\n\n",
858 name, oid_to_hex(&null_oid));
862 tagged_mark = get_object_mark(&p->object);
864 /* tagged->type is either OBJ_BLOB or OBJ_TAG */
865 tagged_mark = get_object_mark(tagged);
870 if (tagged->type == OBJ_TAG) {
871 printf("reset %s\nfrom %s\n\n",
872 name, oid_to_hex(&null_oid));
874 skip_prefix(name, "refs/tags/", &name);
875 printf("tag %s\n", name);
877 mark_next_object(&tag->object);
878 printf("mark :%"PRIu32"\n", last_idnum);
881 printf("from :%d\n", tagged_mark);
883 printf("from %s\n", oid_to_hex(&tagged->oid));
885 if (show_original_ids)
886 printf("original-oid %s\n", oid_to_hex(&tag->object.oid));
887 printf("%.*s%sdata %d\n%.*s\n",
888 (int)(tagger_end - tagger), tagger,
889 tagger == tagger_end ? "" : "\n",
890 (int)message_size, (int)message_size, message ? message : "");
894 static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
896 switch (e->item->type) {
898 return (struct commit *)e->item;
900 struct tag *tag = (struct tag *)e->item;
902 /* handle nested tags */
903 while (tag && tag->object.type == OBJ_TAG) {
904 parse_object(the_repository, &tag->object.oid);
905 string_list_append(&tag_refs, full_name)->util = tag;
906 tag = (struct tag *)tag->tagged;
909 die("Tag %s points nowhere?", e->name);
910 return (struct commit *)tag;
918 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
922 for (i = 0; i < info->nr; i++) {
923 struct rev_cmdline_entry *e = info->rev + i;
924 struct object_id oid;
925 struct commit *commit;
928 if (e->flags & UNINTERESTING)
931 if (dwim_ref(e->name, strlen(e->name), &oid, &full_name) != 1)
936 private = apply_refspecs(&refspecs, full_name);
943 commit = get_commit(e, full_name);
945 warning("%s: Unexpected object of type %s, skipping.",
947 type_name(e->item->type));
951 switch(commit->object.type) {
955 export_blob(&commit->object.oid);
957 default: /* OBJ_TAG (nested tags) is already handled */
958 warning("Tag points to object of unexpected type %s, skipping.",
959 type_name(commit->object.type));
964 * Make sure this ref gets properly updated eventually, whether
965 * through a commit or manually at the end.
967 if (e->item->type != OBJ_TAG)
968 string_list_append(&extra_refs, full_name)->util = commit;
970 if (!*revision_sources_at(&revision_sources, commit))
971 *revision_sources_at(&revision_sources, commit) = full_name;
974 string_list_sort(&extra_refs);
975 string_list_remove_duplicates(&extra_refs, 0);
978 static void handle_tags_and_duplicates(struct string_list *extras)
980 struct commit *commit;
983 for (i = extras->nr - 1; i >= 0; i--) {
984 const char *name = extras->items[i].string;
985 struct object *object = extras->items[i].util;
988 switch (object->type) {
990 handle_tag(name, (struct tag *)object);
994 name = anonymize_refname(name);
995 /* create refs pointing to already seen commits */
996 commit = rewrite_commit((struct commit *)object);
999 * Neither this object nor any of its
1000 * ancestors touch any relevant paths, so
1001 * it has been filtered to nothing. Delete
1004 printf("reset %s\nfrom %s\n\n",
1005 name, oid_to_hex(&null_oid));
1009 mark = get_object_mark(&commit->object);
1012 * Getting here means we have a commit which
1013 * was excluded by a negative refspec (e.g.
1014 * fast-export ^master master). If we are
1015 * referencing excluded commits, set the ref
1016 * to the exact commit. Otherwise, the user
1017 * wants the branch exported but every commit
1018 * in its history to be deleted, which basically
1019 * just means deletion of the ref.
1021 if (!reference_excluded_commits) {
1022 /* delete the ref */
1023 printf("reset %s\nfrom %s\n\n",
1024 name, oid_to_hex(&null_oid));
1027 /* set ref to commit using oid, not mark */
1028 printf("reset %s\nfrom %s\n\n", name,
1029 oid_to_hex(&commit->object.oid));
1033 printf("reset %s\nfrom :%d\n\n", name, mark
1041 static void export_marks(char *file)
1045 struct decoration_entry *deco = idnums.entries;
1049 f = fopen_for_writing(file);
1051 die_errno("Unable to open marks file %s for writing.", file);
1053 for (i = 0; i < idnums.size; i++) {
1054 if (deco->base && deco->base->type == 1) {
1055 mark = ptr_to_mark(deco->decoration);
1056 if (fprintf(f, ":%"PRIu32" %s\n", mark,
1057 oid_to_hex(&deco->base->oid)) < 0) {
1068 error("Unable to write marks file %s.", file);
1071 static void import_marks(char *input_file, int check_exists)
1077 if (check_exists && stat(input_file, &sb))
1080 f = xfopen(input_file, "r");
1081 while (fgets(line, sizeof(line), f)) {
1083 char *line_end, *mark_end;
1084 struct object_id oid;
1085 struct object *object;
1086 struct commit *commit;
1087 enum object_type type;
1089 line_end = strchr(line, '\n');
1090 if (line[0] != ':' || !line_end)
1091 die("corrupt mark line: %s", line);
1094 mark = strtoumax(line + 1, &mark_end, 10);
1095 if (!mark || mark_end == line + 1
1096 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
1097 die("corrupt mark line: %s", line);
1099 if (last_idnum < mark)
1102 type = oid_object_info(the_repository, &oid, NULL);
1104 die("object not found: %s", oid_to_hex(&oid));
1106 if (type != OBJ_COMMIT)
1110 commit = lookup_commit(the_repository, &oid);
1112 die("not a commit? can't happen: %s", oid_to_hex(&oid));
1114 object = &commit->object;
1116 if (object->flags & SHOWN)
1117 error("Object %s already has a mark", oid_to_hex(&oid));
1119 mark_object(object, mark);
1121 object->flags |= SHOWN;
1126 static void handle_deletes(void)
1129 for (i = 0; i < refspecs.nr; i++) {
1130 struct refspec_item *refspec = &refspecs.items[i];
1134 printf("reset %s\nfrom %s\n\n",
1135 refspec->dst, oid_to_hex(&null_oid));
1139 int cmd_fast_export(int argc, const char **argv, const char *prefix)
1141 struct rev_info revs;
1142 struct object_array commits = OBJECT_ARRAY_INIT;
1143 struct commit *commit;
1144 char *export_filename = NULL,
1145 *import_filename = NULL,
1146 *import_filename_if_exists = NULL;
1147 uint32_t lastimportid;
1148 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
1149 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
1150 struct option options[] = {
1151 OPT_INTEGER(0, "progress", &progress,
1152 N_("show progress after <n> objects")),
1153 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1154 N_("select handling of signed tags"),
1155 parse_opt_signed_tag_mode),
1156 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1157 N_("select handling of tags that tag filtered objects"),
1158 parse_opt_tag_of_filtered_mode),
1159 OPT_CALLBACK(0, "reencode", &reencode_mode, N_("mode"),
1160 N_("select handling of commit messages in an alternate encoding"),
1161 parse_opt_reencode_mode),
1162 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1163 N_("Dump marks to this file")),
1164 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1165 N_("Import marks from this file")),
1166 OPT_STRING(0, "import-marks-if-exists",
1167 &import_filename_if_exists,
1169 N_("Import marks from this file if it exists")),
1170 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1171 N_("Fake a tagger when tags lack one")),
1172 OPT_BOOL(0, "full-tree", &full_tree,
1173 N_("Output full tree for each commit")),
1174 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1175 N_("Use the done feature to terminate the stream")),
1176 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
1177 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1178 N_("Apply refspec to exported refs")),
1179 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1180 OPT_BOOL(0, "reference-excluded-parents",
1181 &reference_excluded_commits, N_("Reference parents which are not in fast-export stream by object id")),
1182 OPT_BOOL(0, "show-original-ids", &show_original_ids,
1183 N_("Show original object ids of blobs/commits")),
1184 OPT_BOOL(0, "mark-tags", &mark_tags,
1185 N_("Label tags with mark ids")),
1191 usage_with_options (fast_export_usage, options);
1193 /* we handle encodings */
1194 git_config(git_default_config, NULL);
1196 repo_init_revisions(the_repository, &revs, prefix);
1197 init_revision_sources(&revision_sources);
1198 revs.topo_order = 1;
1199 revs.sources = &revision_sources;
1200 revs.rewrite_parents = 1;
1201 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1202 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1203 argc = setup_revisions(argc, argv, &revs, NULL);
1205 usage_with_options (fast_export_usage, options);
1207 if (refspecs_list.nr) {
1210 for (i = 0; i < refspecs_list.nr; i++)
1211 refspec_append(&refspecs, refspecs_list.items[i].string);
1213 string_list_clear(&refspecs_list, 1);
1216 if (use_done_feature)
1217 printf("feature done\n");
1219 if (import_filename && import_filename_if_exists)
1220 die(_("Cannot pass both --import-marks and --import-marks-if-exists"));
1221 if (import_filename)
1222 import_marks(import_filename, 0);
1223 else if (import_filename_if_exists)
1224 import_marks(import_filename_if_exists, 1);
1225 lastimportid = last_idnum;
1227 if (import_filename && revs.prune_data.nr)
1230 get_tags_and_duplicates(&revs.cmdline);
1232 if (prepare_revision_walk(&revs))
1233 die("revision walk setup failed");
1234 revs.diffopt.format_callback = show_filemodify;
1235 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1236 revs.diffopt.flags.recursive = 1;
1237 while ((commit = get_revision(&revs))) {
1238 if (has_unshown_parent(commit)) {
1239 add_object_array(&commit->object, NULL, &commits);
1242 handle_commit(commit, &revs, &paths_of_changed_objects);
1243 handle_tail(&commits, &revs, &paths_of_changed_objects);
1247 handle_tags_and_duplicates(&extra_refs);
1248 handle_tags_and_duplicates(&tag_refs);
1251 if (export_filename && lastimportid != last_idnum)
1252 export_marks(export_filename);
1254 if (use_done_feature)
1257 refspec_clear(&refspecs);