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;
43 static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
44 static struct string_list tag_refs = STRING_LIST_INIT_NODUP;
45 static struct refspec refspecs = REFSPEC_INIT_FETCH;
47 static struct revision_sources revision_sources;
49 static int parse_opt_signed_tag_mode(const struct option *opt,
50 const char *arg, int unset)
52 if (unset || !strcmp(arg, "abort"))
53 signed_tag_mode = SIGNED_TAG_ABORT;
54 else if (!strcmp(arg, "verbatim") || !strcmp(arg, "ignore"))
55 signed_tag_mode = VERBATIM;
56 else if (!strcmp(arg, "warn"))
57 signed_tag_mode = WARN;
58 else if (!strcmp(arg, "warn-strip"))
59 signed_tag_mode = WARN_STRIP;
60 else if (!strcmp(arg, "strip"))
61 signed_tag_mode = STRIP;
63 return error("Unknown signed-tags mode: %s", arg);
67 static int parse_opt_tag_of_filtered_mode(const struct option *opt,
68 const char *arg, int unset)
70 if (unset || !strcmp(arg, "abort"))
71 tag_of_filtered_mode = TAG_FILTERING_ABORT;
72 else if (!strcmp(arg, "drop"))
73 tag_of_filtered_mode = DROP;
74 else if (!strcmp(arg, "rewrite"))
75 tag_of_filtered_mode = REWRITE;
77 return error("Unknown tag-of-filtered mode: %s", arg);
81 static int parse_opt_reencode_mode(const struct option *opt,
82 const char *arg, int unset)
85 reencode_mode = REENCODE_ABORT;
89 switch (git_parse_maybe_bool(arg)) {
91 reencode_mode = REENCODE_NO;
94 reencode_mode = REENCODE_YES;
97 if (!strcasecmp(arg, "abort"))
98 reencode_mode = REENCODE_ABORT;
100 return error("Unknown reencoding mode: %s", arg);
106 static struct decoration idnums;
107 static uint32_t last_idnum;
109 static int has_unshown_parent(struct commit *commit)
111 struct commit_list *parent;
113 for (parent = commit->parents; parent; parent = parent->next)
114 if (!(parent->item->object.flags & SHOWN) &&
115 !(parent->item->object.flags & UNINTERESTING))
120 struct anonymized_entry {
121 struct hashmap_entry hash;
128 static int anonymized_entry_cmp(const void *unused_cmp_data,
129 const void *va, const void *vb,
130 const void *unused_keydata)
132 const struct anonymized_entry *a = va, *b = vb;
133 return a->orig_len != b->orig_len ||
134 memcmp(a->orig, b->orig, a->orig_len);
138 * Basically keep a cache of X->Y so that we can repeatedly replace
139 * the same anonymized string with another. The actual generation
140 * is farmed out to the generate function.
142 static const void *anonymize_mem(struct hashmap *map,
143 void *(*generate)(const void *, size_t *),
144 const void *orig, size_t *len)
146 struct anonymized_entry key, *ret;
149 hashmap_init(map, anonymized_entry_cmp, NULL, 0);
151 hashmap_entry_init(&key, memhash(orig, *len));
154 ret = hashmap_get(map, &key, NULL);
157 ret = xmalloc(sizeof(*ret));
158 hashmap_entry_init(&ret->hash, key.hash.hash);
159 ret->orig = xstrdup(orig);
160 ret->orig_len = *len;
161 ret->anon = generate(orig, len);
162 ret->anon_len = *len;
163 hashmap_put(map, ret);
166 *len = ret->anon_len;
171 * We anonymize each component of a path individually,
172 * so that paths a/b and a/c will share a common root.
173 * The paths are cached via anonymize_mem so that repeated
174 * lookups for "a" will yield the same value.
176 static void anonymize_path(struct strbuf *out, const char *path,
178 void *(*generate)(const void *, size_t *))
181 const char *end_of_component = strchrnul(path, '/');
182 size_t len = end_of_component - path;
183 const char *c = anonymize_mem(map, generate, path, &len);
184 strbuf_add(out, c, len);
185 path = end_of_component;
187 strbuf_addch(out, *path++);
191 static inline void *mark_to_ptr(uint32_t mark)
193 return (void *)(uintptr_t)mark;
196 static inline uint32_t ptr_to_mark(void * mark)
198 return (uint32_t)(uintptr_t)mark;
201 static inline void mark_object(struct object *object, uint32_t mark)
203 add_decoration(&idnums, object, mark_to_ptr(mark));
206 static inline void mark_next_object(struct object *object)
208 mark_object(object, ++last_idnum);
211 static int get_object_mark(struct object *object)
213 void *decoration = lookup_decoration(&idnums, object);
216 return ptr_to_mark(decoration);
219 static struct commit *rewrite_commit(struct commit *p)
222 if (p->parents && p->parents->next)
224 if (p->object.flags & UNINTERESTING)
226 if (!(p->object.flags & TREESAME))
230 p = p->parents->item;
235 static void show_progress(void)
237 static int counter = 0;
240 if ((++counter % progress) == 0)
241 printf("progress %d objects\n", counter);
245 * Ideally we would want some transformation of the blob data here
246 * that is unreversible, but would still be the same size and have
247 * the same data relationship to other blobs (so that we get the same
248 * delta and packing behavior as the original). But the first and last
249 * requirements there are probably mutually exclusive, so let's take
250 * the easy way out for now, and just generate arbitrary content.
252 * There's no need to cache this result with anonymize_mem, since
253 * we already handle blob content caching with marks.
255 static char *anonymize_blob(unsigned long *size)
258 struct strbuf out = STRBUF_INIT;
259 strbuf_addf(&out, "anonymous blob %d", counter++);
261 return strbuf_detach(&out, NULL);
264 static void export_blob(const struct object_id *oid)
267 enum object_type type;
269 struct object *object;
275 if (is_null_oid(oid))
278 object = lookup_object(the_repository, oid->hash);
279 if (object && object->flags & SHOWN)
283 buf = anonymize_blob(&size);
284 object = (struct object *)lookup_blob(the_repository, oid);
287 buf = read_object_file(oid, &type, &size);
289 die("could not read blob %s", oid_to_hex(oid));
290 if (check_object_signature(oid, buf, size, type_name(type)) < 0)
291 die("oid mismatch in blob %s", oid_to_hex(oid));
292 object = parse_object_buffer(the_repository, oid, type,
297 die("Could not read blob %s", oid_to_hex(oid));
299 mark_next_object(object);
301 printf("blob\nmark :%"PRIu32"\n", last_idnum);
302 if (show_original_ids)
303 printf("original-oid %s\n", oid_to_hex(oid));
304 printf("data %"PRIuMAX"\n", (uintmax_t)size);
305 if (size && fwrite(buf, size, 1, stdout) != 1)
306 die_errno("could not write blob '%s'", oid_to_hex(oid));
311 object->flags |= SHOWN;
316 static int depth_first(const void *a_, const void *b_)
318 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
319 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
320 const char *name_a, *name_b;
321 int len_a, len_b, len;
324 name_a = a->one ? a->one->path : a->two->path;
325 name_b = b->one ? b->one->path : b->two->path;
327 len_a = strlen(name_a);
328 len_b = strlen(name_b);
329 len = (len_a < len_b) ? len_a : len_b;
331 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
332 cmp = memcmp(name_a, name_b, len);
339 * Move 'R'ename entries last so that all references of the file
340 * appear in the output before it is renamed (e.g., when a file
341 * was copied and renamed in the same commit).
343 return (a->status == 'R') - (b->status == 'R');
346 static void print_path_1(const char *path)
348 int need_quote = quote_c_style(path, NULL, NULL, 0);
350 quote_c_style(path, NULL, stdout, 0);
351 else if (strchr(path, ' '))
352 printf("\"%s\"", path);
357 static void *anonymize_path_component(const void *path, size_t *len)
360 struct strbuf out = STRBUF_INIT;
361 strbuf_addf(&out, "path%d", counter++);
362 return strbuf_detach(&out, len);
365 static void print_path(const char *path)
370 static struct hashmap paths;
371 static struct strbuf anon = STRBUF_INIT;
373 anonymize_path(&anon, path, &paths, anonymize_path_component);
374 print_path_1(anon.buf);
379 static void *generate_fake_oid(const void *old, size_t *len)
381 static uint32_t counter = 1; /* avoid null oid */
382 const unsigned hashsz = the_hash_algo->rawsz;
383 unsigned char *out = xcalloc(hashsz, 1);
384 put_be32(out + hashsz - 4, counter++);
388 static const struct object_id *anonymize_oid(const struct object_id *oid)
390 static struct hashmap objs;
391 size_t len = the_hash_algo->rawsz;
392 return anonymize_mem(&objs, generate_fake_oid, oid, &len);
395 static void show_filemodify(struct diff_queue_struct *q,
396 struct diff_options *options, void *data)
399 struct string_list *changed = data;
402 * Handle files below a directory first, in case they are all deleted
403 * and the directory changes to a file or symlink.
405 QSORT(q->queue, q->nr, depth_first);
407 for (i = 0; i < q->nr; i++) {
408 struct diff_filespec *ospec = q->queue[i]->one;
409 struct diff_filespec *spec = q->queue[i]->two;
411 switch (q->queue[i]->status) {
412 case DIFF_STATUS_DELETED:
414 print_path(spec->path);
415 string_list_insert(changed, spec->path);
419 case DIFF_STATUS_COPIED:
420 case DIFF_STATUS_RENAMED:
422 * If a change in the file corresponding to ospec->path
423 * has been observed, we cannot trust its contents
424 * because the diff is calculated based on the prior
425 * contents, not the current contents. So, declare a
426 * copy or rename only if there was no change observed.
428 if (!string_list_has_string(changed, ospec->path)) {
429 printf("%c ", q->queue[i]->status);
430 print_path(ospec->path);
432 print_path(spec->path);
433 string_list_insert(changed, spec->path);
436 if (oideq(&ospec->oid, &spec->oid) &&
437 ospec->mode == spec->mode)
442 case DIFF_STATUS_TYPE_CHANGED:
443 case DIFF_STATUS_MODIFIED:
444 case DIFF_STATUS_ADDED:
446 * Links refer to objects in another repositories;
447 * output the SHA-1 verbatim.
449 if (no_data || S_ISGITLINK(spec->mode))
450 printf("M %06o %s ", spec->mode,
451 oid_to_hex(anonymize ?
452 anonymize_oid(&spec->oid) :
455 struct object *object = lookup_object(the_repository,
457 printf("M %06o :%d ", spec->mode,
458 get_object_mark(object));
460 print_path(spec->path);
461 string_list_insert(changed, spec->path);
466 die("Unexpected comparison status '%c' for %s, %s",
468 ospec->path ? ospec->path : "none",
469 spec->path ? spec->path : "none");
474 static const char *find_encoding(const char *begin, const char *end)
476 const char *needle = "\nencoding ";
479 bol = memmem(begin, end ? end - begin : strlen(begin),
480 needle, strlen(needle));
483 bol += strlen(needle);
484 eol = strchrnul(bol, '\n');
489 static void *anonymize_ref_component(const void *old, size_t *len)
492 struct strbuf out = STRBUF_INIT;
493 strbuf_addf(&out, "ref%d", counter++);
494 return strbuf_detach(&out, len);
497 static const char *anonymize_refname(const char *refname)
500 * If any of these prefixes is found, we will leave it intact
501 * so that tags remain tags and so forth.
503 static const char *prefixes[] = {
509 static struct hashmap refs;
510 static struct strbuf anon = STRBUF_INIT;
514 * We also leave "master" as a special case, since it does not reveal
515 * anything interesting.
517 if (!strcmp(refname, "refs/heads/master"))
521 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
522 if (skip_prefix(refname, prefixes[i], &refname)) {
523 strbuf_addstr(&anon, prefixes[i]);
528 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
533 * We do not even bother to cache commit messages, as they are unlikely
534 * to be repeated verbatim, and it is not that interesting when they are.
536 static char *anonymize_commit_message(const char *old)
539 return xstrfmt("subject %d\n\nbody\n", counter++);
542 static struct hashmap idents;
543 static void *anonymize_ident(const void *old, size_t *len)
546 struct strbuf out = STRBUF_INIT;
547 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
549 return strbuf_detach(&out, len);
553 * Our strategy here is to anonymize the names and email addresses,
554 * but keep timestamps intact, as they influence things like traversal
555 * order (and by themselves should not be too revealing).
557 static void anonymize_ident_line(const char **beg, const char **end)
559 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
560 static unsigned which_buffer;
563 struct ident_split split;
564 const char *end_of_header;
566 out = &buffers[which_buffer++];
567 which_buffer %= ARRAY_SIZE(buffers);
570 /* skip "committer", "author", "tagger", etc */
571 end_of_header = strchr(*beg, ' ');
573 BUG("malformed line fed to anonymize_ident_line: %.*s",
574 (int)(*end - *beg), *beg);
576 strbuf_add(out, *beg, end_of_header - *beg);
578 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
583 len = split.mail_end - split.name_begin;
584 ident = anonymize_mem(&idents, anonymize_ident,
585 split.name_begin, &len);
586 strbuf_add(out, ident, len);
587 strbuf_addch(out, ' ');
588 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
590 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
594 *end = out->buf + out->len;
597 static void handle_commit(struct commit *commit, struct rev_info *rev,
598 struct string_list *paths_of_changed_objects)
600 int saved_output_format = rev->diffopt.output_format;
601 const char *commit_buffer;
602 const char *author, *author_end, *committer, *committer_end;
603 const char *encoding, *message;
604 char *reencoded = NULL;
605 struct commit_list *p;
609 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
611 parse_commit_or_die(commit);
612 commit_buffer = get_commit_buffer(commit, NULL);
613 author = strstr(commit_buffer, "\nauthor ");
615 die("could not find author in commit %s",
616 oid_to_hex(&commit->object.oid));
618 author_end = strchrnul(author, '\n');
619 committer = strstr(author_end, "\ncommitter ");
621 die("could not find committer in commit %s",
622 oid_to_hex(&commit->object.oid));
624 committer_end = strchrnul(committer, '\n');
625 message = strstr(committer_end, "\n\n");
626 encoding = find_encoding(committer_end, message);
630 if (commit->parents &&
631 (get_object_mark(&commit->parents->item->object) != 0 ||
632 reference_excluded_commits) &&
634 parse_commit_or_die(commit->parents->item);
635 diff_tree_oid(get_commit_tree_oid(commit->parents->item),
636 get_commit_tree_oid(commit), "", &rev->diffopt);
639 diff_root_tree_oid(get_commit_tree_oid(commit),
642 /* Export the referenced blobs, and remember the marks. */
643 for (i = 0; i < diff_queued_diff.nr; i++)
644 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
645 export_blob(&diff_queued_diff.queue[i]->two->oid);
647 refname = *revision_sources_at(&revision_sources, commit);
649 * FIXME: string_list_remove() below for each ref is overall
650 * O(N^2). Compared to a history walk and diffing trees, this is
651 * just lost in the noise in practice. However, theoretically a
652 * repo may have enough refs for this to become slow.
654 string_list_remove(&extra_refs, refname, 0);
656 refname = anonymize_refname(refname);
657 anonymize_ident_line(&committer, &committer_end);
658 anonymize_ident_line(&author, &author_end);
661 mark_next_object(&commit->object);
663 reencoded = anonymize_commit_message(message);
664 } else if (encoding) {
665 switch(reencode_mode) {
667 reencoded = reencode_string(message, "UTF-8", encoding);
672 die("Encountered commit-specific encoding %s in commit "
673 "%s; use --reencode=[yes|no] to handle it",
674 encoding, oid_to_hex(&commit->object.oid));
677 if (!commit->parents)
678 printf("reset %s\n", refname);
679 printf("commit %s\nmark :%"PRIu32"\n", refname, last_idnum);
680 if (show_original_ids)
681 printf("original-oid %s\n", oid_to_hex(&commit->object.oid));
682 printf("%.*s\n%.*s\n",
683 (int)(author_end - author), author,
684 (int)(committer_end - committer), committer);
685 if (!reencoded && encoding)
686 printf("encoding %s\n", encoding);
687 printf("data %u\n%s",
689 ? strlen(reencoded) : message
690 ? strlen(message) : 0),
691 reencoded ? reencoded : message ? message : "");
693 unuse_commit_buffer(commit, commit_buffer);
695 for (i = 0, p = commit->parents; p; p = p->next) {
696 struct object *obj = &p->item->object;
697 int mark = get_object_mark(obj);
699 if (!mark && !reference_excluded_commits)
706 printf(":%d\n", mark);
708 printf("%s\n", oid_to_hex(anonymize ?
709 anonymize_oid(&obj->oid) :
715 printf("deleteall\n");
716 log_tree_diff_flush(rev);
717 string_list_clear(paths_of_changed_objects, 0);
718 rev->diffopt.output_format = saved_output_format;
725 static void *anonymize_tag(const void *old, size_t *len)
728 struct strbuf out = STRBUF_INIT;
729 strbuf_addf(&out, "tag message %d", counter++);
730 return strbuf_detach(&out, len);
733 static void handle_tail(struct object_array *commits, struct rev_info *revs,
734 struct string_list *paths_of_changed_objects)
736 struct commit *commit;
737 while (commits->nr) {
738 commit = (struct commit *)object_array_pop(commits);
739 if (has_unshown_parent(commit)) {
740 /* Queue again, to be handled later */
741 add_object_array(&commit->object, NULL, commits);
744 handle_commit(commit, revs, paths_of_changed_objects);
748 static void handle_tag(const char *name, struct tag *tag)
751 enum object_type type;
753 const char *tagger, *tagger_end, *message;
754 size_t message_size = 0;
755 struct object *tagged;
759 /* Trees have no identifier in fast-export output, thus we have no way
760 * to output tags of trees, tags of tags of trees, etc. Simply omit
763 tagged = tag->tagged;
764 while (tagged->type == OBJ_TAG) {
765 tagged = ((struct tag *)tagged)->tagged;
767 if (tagged->type == OBJ_TREE) {
768 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
769 oid_to_hex(&tag->object.oid));
773 buf = read_object_file(&tag->object.oid, &type, &size);
775 die("could not read tag %s", oid_to_hex(&tag->object.oid));
776 message = memmem(buf, size, "\n\n", 2);
779 message_size = strlen(message);
781 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
783 if (fake_missing_tagger)
784 tagger = "tagger Unspecified Tagger "
785 "<unspecified-tagger> 0 +0000";
788 tagger_end = tagger + strlen(tagger);
791 tagger_end = strchrnul(tagger, '\n');
793 anonymize_ident_line(&tagger, &tagger_end);
797 name = anonymize_refname(name);
799 static struct hashmap tags;
800 message = anonymize_mem(&tags, anonymize_tag,
801 message, &message_size);
805 /* handle signed tags */
807 const char *signature = strstr(message,
808 "\n-----BEGIN PGP SIGNATURE-----\n");
810 switch(signed_tag_mode) {
811 case SIGNED_TAG_ABORT:
812 die("encountered signed tag %s; use "
813 "--signed-tags=<mode> to handle it",
814 oid_to_hex(&tag->object.oid));
816 warning("exporting signed tag %s",
817 oid_to_hex(&tag->object.oid));
822 warning("stripping signature from tag %s",
823 oid_to_hex(&tag->object.oid));
826 message_size = signature + 1 - message;
831 /* handle tag->tagged having been filtered out due to paths specified */
832 tagged = tag->tagged;
833 tagged_mark = get_object_mark(tagged);
835 switch(tag_of_filtered_mode) {
836 case TAG_FILTERING_ABORT:
837 die("tag %s tags unexported object; use "
838 "--tag-of-filtered-object=<mode> to handle it",
839 oid_to_hex(&tag->object.oid));
841 /* Ignore this tag altogether */
845 if (tagged->type != OBJ_COMMIT) {
846 die("tag %s tags unexported %s!",
847 oid_to_hex(&tag->object.oid),
848 type_name(tagged->type));
850 p = rewrite_commit((struct commit *)tagged);
852 printf("reset %s\nfrom %s\n\n",
853 name, oid_to_hex(&null_oid));
857 tagged_mark = get_object_mark(&p->object);
861 if (starts_with(name, "refs/tags/"))
863 printf("tag %s\nfrom :%d\n", name, tagged_mark);
864 if (show_original_ids)
865 printf("original-oid %s\n", oid_to_hex(&tag->object.oid));
866 printf("%.*s%sdata %d\n%.*s\n",
867 (int)(tagger_end - tagger), tagger,
868 tagger == tagger_end ? "" : "\n",
869 (int)message_size, (int)message_size, message ? message : "");
873 static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
875 switch (e->item->type) {
877 return (struct commit *)e->item;
879 struct tag *tag = (struct tag *)e->item;
881 /* handle nested tags */
882 while (tag && tag->object.type == OBJ_TAG) {
883 parse_object(the_repository, &tag->object.oid);
884 string_list_append(&tag_refs, full_name)->util = tag;
885 tag = (struct tag *)tag->tagged;
888 die("Tag %s points nowhere?", e->name);
889 return (struct commit *)tag;
897 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
901 for (i = 0; i < info->nr; i++) {
902 struct rev_cmdline_entry *e = info->rev + i;
903 struct object_id oid;
904 struct commit *commit;
907 if (e->flags & UNINTERESTING)
910 if (dwim_ref(e->name, strlen(e->name), &oid, &full_name) != 1)
915 private = apply_refspecs(&refspecs, full_name);
922 commit = get_commit(e, full_name);
924 warning("%s: Unexpected object of type %s, skipping.",
926 type_name(e->item->type));
930 switch(commit->object.type) {
934 export_blob(&commit->object.oid);
936 default: /* OBJ_TAG (nested tags) is already handled */
937 warning("Tag points to object of unexpected type %s, skipping.",
938 type_name(commit->object.type));
943 * Make sure this ref gets properly updated eventually, whether
944 * through a commit or manually at the end.
946 if (e->item->type != OBJ_TAG)
947 string_list_append(&extra_refs, full_name)->util = commit;
949 if (!*revision_sources_at(&revision_sources, commit))
950 *revision_sources_at(&revision_sources, commit) = full_name;
953 string_list_sort(&extra_refs);
954 string_list_remove_duplicates(&extra_refs, 0);
957 static void handle_tags_and_duplicates(struct string_list *extras)
959 struct commit *commit;
962 for (i = extras->nr - 1; i >= 0; i--) {
963 const char *name = extras->items[i].string;
964 struct object *object = extras->items[i].util;
967 switch (object->type) {
969 handle_tag(name, (struct tag *)object);
973 name = anonymize_refname(name);
974 /* create refs pointing to already seen commits */
975 commit = rewrite_commit((struct commit *)object);
978 * Neither this object nor any of its
979 * ancestors touch any relevant paths, so
980 * it has been filtered to nothing. Delete
983 printf("reset %s\nfrom %s\n\n",
984 name, oid_to_hex(&null_oid));
988 mark = get_object_mark(&commit->object);
991 * Getting here means we have a commit which
992 * was excluded by a negative refspec (e.g.
993 * fast-export ^master master). If we are
994 * referencing excluded commits, set the ref
995 * to the exact commit. Otherwise, the user
996 * wants the branch exported but every commit
997 * in its history to be deleted, which basically
998 * just means deletion of the ref.
1000 if (!reference_excluded_commits) {
1001 /* delete the ref */
1002 printf("reset %s\nfrom %s\n\n",
1003 name, oid_to_hex(&null_oid));
1006 /* set ref to commit using oid, not mark */
1007 printf("reset %s\nfrom %s\n\n", name,
1008 oid_to_hex(&commit->object.oid));
1012 printf("reset %s\nfrom :%d\n\n", name, mark
1020 static void export_marks(char *file)
1024 struct decoration_entry *deco = idnums.entries;
1028 f = fopen_for_writing(file);
1030 die_errno("Unable to open marks file %s for writing.", file);
1032 for (i = 0; i < idnums.size; i++) {
1033 if (deco->base && deco->base->type == 1) {
1034 mark = ptr_to_mark(deco->decoration);
1035 if (fprintf(f, ":%"PRIu32" %s\n", mark,
1036 oid_to_hex(&deco->base->oid)) < 0) {
1047 error("Unable to write marks file %s.", file);
1050 static void import_marks(char *input_file)
1053 FILE *f = xfopen(input_file, "r");
1055 while (fgets(line, sizeof(line), f)) {
1057 char *line_end, *mark_end;
1058 struct object_id oid;
1059 struct object *object;
1060 struct commit *commit;
1061 enum object_type type;
1063 line_end = strchr(line, '\n');
1064 if (line[0] != ':' || !line_end)
1065 die("corrupt mark line: %s", line);
1068 mark = strtoumax(line + 1, &mark_end, 10);
1069 if (!mark || mark_end == line + 1
1070 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
1071 die("corrupt mark line: %s", line);
1073 if (last_idnum < mark)
1076 type = oid_object_info(the_repository, &oid, NULL);
1078 die("object not found: %s", oid_to_hex(&oid));
1080 if (type != OBJ_COMMIT)
1084 commit = lookup_commit(the_repository, &oid);
1086 die("not a commit? can't happen: %s", oid_to_hex(&oid));
1088 object = &commit->object;
1090 if (object->flags & SHOWN)
1091 error("Object %s already has a mark", oid_to_hex(&oid));
1093 mark_object(object, mark);
1095 object->flags |= SHOWN;
1100 static void handle_deletes(void)
1103 for (i = 0; i < refspecs.nr; i++) {
1104 struct refspec_item *refspec = &refspecs.items[i];
1108 printf("reset %s\nfrom %s\n\n",
1109 refspec->dst, oid_to_hex(&null_oid));
1113 int cmd_fast_export(int argc, const char **argv, const char *prefix)
1115 struct rev_info revs;
1116 struct object_array commits = OBJECT_ARRAY_INIT;
1117 struct commit *commit;
1118 char *export_filename = NULL, *import_filename = NULL;
1119 uint32_t lastimportid;
1120 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
1121 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
1122 struct option options[] = {
1123 OPT_INTEGER(0, "progress", &progress,
1124 N_("show progress after <n> objects")),
1125 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1126 N_("select handling of signed tags"),
1127 parse_opt_signed_tag_mode),
1128 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1129 N_("select handling of tags that tag filtered objects"),
1130 parse_opt_tag_of_filtered_mode),
1131 OPT_CALLBACK(0, "reencode", &reencode_mode, N_("mode"),
1132 N_("select handling of commit messages in an alternate encoding"),
1133 parse_opt_reencode_mode),
1134 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1135 N_("Dump marks to this file")),
1136 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1137 N_("Import marks from this file")),
1138 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1139 N_("Fake a tagger when tags lack one")),
1140 OPT_BOOL(0, "full-tree", &full_tree,
1141 N_("Output full tree for each commit")),
1142 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1143 N_("Use the done feature to terminate the stream")),
1144 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
1145 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1146 N_("Apply refspec to exported refs")),
1147 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1148 OPT_BOOL(0, "reference-excluded-parents",
1149 &reference_excluded_commits, N_("Reference parents which are not in fast-export stream by object id")),
1150 OPT_BOOL(0, "show-original-ids", &show_original_ids,
1151 N_("Show original object ids of blobs/commits")),
1157 usage_with_options (fast_export_usage, options);
1159 /* we handle encodings */
1160 git_config(git_default_config, NULL);
1162 repo_init_revisions(the_repository, &revs, prefix);
1163 init_revision_sources(&revision_sources);
1164 revs.topo_order = 1;
1165 revs.sources = &revision_sources;
1166 revs.rewrite_parents = 1;
1167 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1168 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1169 argc = setup_revisions(argc, argv, &revs, NULL);
1171 usage_with_options (fast_export_usage, options);
1173 if (refspecs_list.nr) {
1176 for (i = 0; i < refspecs_list.nr; i++)
1177 refspec_append(&refspecs, refspecs_list.items[i].string);
1179 string_list_clear(&refspecs_list, 1);
1182 if (use_done_feature)
1183 printf("feature done\n");
1185 if (import_filename)
1186 import_marks(import_filename);
1187 lastimportid = last_idnum;
1189 if (import_filename && revs.prune_data.nr)
1192 get_tags_and_duplicates(&revs.cmdline);
1194 if (prepare_revision_walk(&revs))
1195 die("revision walk setup failed");
1196 revs.diffopt.format_callback = show_filemodify;
1197 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1198 revs.diffopt.flags.recursive = 1;
1199 while ((commit = get_revision(&revs))) {
1200 if (has_unshown_parent(commit)) {
1201 add_object_array(&commit->object, NULL, &commits);
1204 handle_commit(commit, &revs, &paths_of_changed_objects);
1205 handle_tail(&commits, &revs, &paths_of_changed_objects);
1209 handle_tags_and_duplicates(&extra_refs);
1210 handle_tags_and_duplicates(&tag_refs);
1213 if (export_filename && lastimportid != last_idnum)
1214 export_marks(export_filename);
1216 if (use_done_feature)
1219 refspec_clear(&refspecs);