t1300: add a few more hairy examples of sections becoming empty
[git/debian.git] / builtin / fast-export.c
blob796d0cd66c7750a20613dea77cd75b1b8e7665a4
1 /*
2 * "git fast-export" builtin command
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 "commit.h"
11 #include "object.h"
12 #include "tag.h"
13 #include "diff.h"
14 #include "diffcore.h"
15 #include "log-tree.h"
16 #include "revision.h"
17 #include "decorate.h"
18 #include "string-list.h"
19 #include "utf8.h"
20 #include "parse-options.h"
21 #include "quote.h"
22 #include "remote.h"
23 #include "blob.h"
25 static const char *fast_export_usage[] = {
26 N_("git fast-export [rev-list-opts]"),
27 NULL
30 static int progress;
31 static enum { ABORT, VERBATIM, WARN, WARN_STRIP, STRIP } signed_tag_mode = ABORT;
32 static enum { ERROR, DROP, REWRITE } tag_of_filtered_mode = ERROR;
33 static int fake_missing_tagger;
34 static int use_done_feature;
35 static int no_data;
36 static int full_tree;
37 static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
38 static struct refspec *refspecs;
39 static int refspecs_nr;
40 static int anonymize;
42 static int parse_opt_signed_tag_mode(const struct option *opt,
43 const char *arg, int unset)
45 if (unset || !strcmp(arg, "abort"))
46 signed_tag_mode = ABORT;
47 else if (!strcmp(arg, "verbatim") || !strcmp(arg, "ignore"))
48 signed_tag_mode = VERBATIM;
49 else if (!strcmp(arg, "warn"))
50 signed_tag_mode = WARN;
51 else if (!strcmp(arg, "warn-strip"))
52 signed_tag_mode = WARN_STRIP;
53 else if (!strcmp(arg, "strip"))
54 signed_tag_mode = STRIP;
55 else
56 return error("Unknown signed-tags mode: %s", arg);
57 return 0;
60 static int parse_opt_tag_of_filtered_mode(const struct option *opt,
61 const char *arg, int unset)
63 if (unset || !strcmp(arg, "abort"))
64 tag_of_filtered_mode = ERROR;
65 else if (!strcmp(arg, "drop"))
66 tag_of_filtered_mode = DROP;
67 else if (!strcmp(arg, "rewrite"))
68 tag_of_filtered_mode = REWRITE;
69 else
70 return error("Unknown tag-of-filtered mode: %s", arg);
71 return 0;
74 static struct decoration idnums;
75 static uint32_t last_idnum;
77 static int has_unshown_parent(struct commit *commit)
79 struct commit_list *parent;
81 for (parent = commit->parents; parent; parent = parent->next)
82 if (!(parent->item->object.flags & SHOWN) &&
83 !(parent->item->object.flags & UNINTERESTING))
84 return 1;
85 return 0;
88 struct anonymized_entry {
89 struct hashmap_entry hash;
90 const char *orig;
91 size_t orig_len;
92 const char *anon;
93 size_t anon_len;
96 static int anonymized_entry_cmp(const void *unused_cmp_data,
97 const void *va, const void *vb,
98 const void *unused_keydata)
100 const struct anonymized_entry *a = va, *b = vb;
101 return a->orig_len != b->orig_len ||
102 memcmp(a->orig, b->orig, a->orig_len);
106 * Basically keep a cache of X->Y so that we can repeatedly replace
107 * the same anonymized string with another. The actual generation
108 * is farmed out to the generate function.
110 static const void *anonymize_mem(struct hashmap *map,
111 void *(*generate)(const void *, size_t *),
112 const void *orig, size_t *len)
114 struct anonymized_entry key, *ret;
116 if (!map->cmpfn)
117 hashmap_init(map, anonymized_entry_cmp, NULL, 0);
119 hashmap_entry_init(&key, memhash(orig, *len));
120 key.orig = orig;
121 key.orig_len = *len;
122 ret = hashmap_get(map, &key, NULL);
124 if (!ret) {
125 ret = xmalloc(sizeof(*ret));
126 hashmap_entry_init(&ret->hash, key.hash.hash);
127 ret->orig = xstrdup(orig);
128 ret->orig_len = *len;
129 ret->anon = generate(orig, len);
130 ret->anon_len = *len;
131 hashmap_put(map, ret);
134 *len = ret->anon_len;
135 return ret->anon;
139 * We anonymize each component of a path individually,
140 * so that paths a/b and a/c will share a common root.
141 * The paths are cached via anonymize_mem so that repeated
142 * lookups for "a" will yield the same value.
144 static void anonymize_path(struct strbuf *out, const char *path,
145 struct hashmap *map,
146 void *(*generate)(const void *, size_t *))
148 while (*path) {
149 const char *end_of_component = strchrnul(path, '/');
150 size_t len = end_of_component - path;
151 const char *c = anonymize_mem(map, generate, path, &len);
152 strbuf_add(out, c, len);
153 path = end_of_component;
154 if (*path)
155 strbuf_addch(out, *path++);
159 /* Since intptr_t is C99, we do not use it here */
160 static inline uint32_t *mark_to_ptr(uint32_t mark)
162 return ((uint32_t *)NULL) + mark;
165 static inline uint32_t ptr_to_mark(void * mark)
167 return (uint32_t *)mark - (uint32_t *)NULL;
170 static inline void mark_object(struct object *object, uint32_t mark)
172 add_decoration(&idnums, object, mark_to_ptr(mark));
175 static inline void mark_next_object(struct object *object)
177 mark_object(object, ++last_idnum);
180 static int get_object_mark(struct object *object)
182 void *decoration = lookup_decoration(&idnums, object);
183 if (!decoration)
184 return 0;
185 return ptr_to_mark(decoration);
188 static void show_progress(void)
190 static int counter = 0;
191 if (!progress)
192 return;
193 if ((++counter % progress) == 0)
194 printf("progress %d objects\n", counter);
198 * Ideally we would want some transformation of the blob data here
199 * that is unreversible, but would still be the same size and have
200 * the same data relationship to other blobs (so that we get the same
201 * delta and packing behavior as the original). But the first and last
202 * requirements there are probably mutually exclusive, so let's take
203 * the easy way out for now, and just generate arbitrary content.
205 * There's no need to cache this result with anonymize_mem, since
206 * we already handle blob content caching with marks.
208 static char *anonymize_blob(unsigned long *size)
210 static int counter;
211 struct strbuf out = STRBUF_INIT;
212 strbuf_addf(&out, "anonymous blob %d", counter++);
213 *size = out.len;
214 return strbuf_detach(&out, NULL);
217 static void export_blob(const struct object_id *oid)
219 unsigned long size;
220 enum object_type type;
221 char *buf;
222 struct object *object;
223 int eaten;
225 if (no_data)
226 return;
228 if (is_null_oid(oid))
229 return;
231 object = lookup_object(oid->hash);
232 if (object && object->flags & SHOWN)
233 return;
235 if (anonymize) {
236 buf = anonymize_blob(&size);
237 object = (struct object *)lookup_blob(oid);
238 eaten = 0;
239 } else {
240 buf = read_sha1_file(oid->hash, &type, &size);
241 if (!buf)
242 die ("Could not read blob %s", oid_to_hex(oid));
243 if (check_sha1_signature(oid->hash, buf, size, typename(type)) < 0)
244 die("sha1 mismatch in blob %s", oid_to_hex(oid));
245 object = parse_object_buffer(oid, type, size, buf, &eaten);
248 if (!object)
249 die("Could not read blob %s", oid_to_hex(oid));
251 mark_next_object(object);
253 printf("blob\nmark :%"PRIu32"\ndata %lu\n", last_idnum, size);
254 if (size && fwrite(buf, size, 1, stdout) != 1)
255 die_errno ("Could not write blob '%s'", oid_to_hex(oid));
256 printf("\n");
258 show_progress();
260 object->flags |= SHOWN;
261 if (!eaten)
262 free(buf);
265 static int depth_first(const void *a_, const void *b_)
267 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
268 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
269 const char *name_a, *name_b;
270 int len_a, len_b, len;
271 int cmp;
273 name_a = a->one ? a->one->path : a->two->path;
274 name_b = b->one ? b->one->path : b->two->path;
276 len_a = strlen(name_a);
277 len_b = strlen(name_b);
278 len = (len_a < len_b) ? len_a : len_b;
280 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
281 cmp = memcmp(name_a, name_b, len);
282 if (cmp)
283 return cmp;
284 cmp = len_b - len_a;
285 if (cmp)
286 return cmp;
288 * Move 'R'ename entries last so that all references of the file
289 * appear in the output before it is renamed (e.g., when a file
290 * was copied and renamed in the same commit).
292 return (a->status == 'R') - (b->status == 'R');
295 static void print_path_1(const char *path)
297 int need_quote = quote_c_style(path, NULL, NULL, 0);
298 if (need_quote)
299 quote_c_style(path, NULL, stdout, 0);
300 else if (strchr(path, ' '))
301 printf("\"%s\"", path);
302 else
303 printf("%s", path);
306 static void *anonymize_path_component(const void *path, size_t *len)
308 static int counter;
309 struct strbuf out = STRBUF_INIT;
310 strbuf_addf(&out, "path%d", counter++);
311 return strbuf_detach(&out, len);
314 static void print_path(const char *path)
316 if (!anonymize)
317 print_path_1(path);
318 else {
319 static struct hashmap paths;
320 static struct strbuf anon = STRBUF_INIT;
322 anonymize_path(&anon, path, &paths, anonymize_path_component);
323 print_path_1(anon.buf);
324 strbuf_reset(&anon);
328 static void *generate_fake_oid(const void *old, size_t *len)
330 static uint32_t counter = 1; /* avoid null sha1 */
331 unsigned char *out = xcalloc(GIT_SHA1_RAWSZ, 1);
332 put_be32(out + GIT_SHA1_RAWSZ - 4, counter++);
333 return out;
336 static const unsigned char *anonymize_sha1(const struct object_id *oid)
338 static struct hashmap sha1s;
339 size_t len = GIT_SHA1_RAWSZ;
340 return anonymize_mem(&sha1s, generate_fake_oid, oid, &len);
343 static void show_filemodify(struct diff_queue_struct *q,
344 struct diff_options *options, void *data)
346 int i;
347 struct string_list *changed = data;
350 * Handle files below a directory first, in case they are all deleted
351 * and the directory changes to a file or symlink.
353 QSORT(q->queue, q->nr, depth_first);
355 for (i = 0; i < q->nr; i++) {
356 struct diff_filespec *ospec = q->queue[i]->one;
357 struct diff_filespec *spec = q->queue[i]->two;
359 switch (q->queue[i]->status) {
360 case DIFF_STATUS_DELETED:
361 printf("D ");
362 print_path(spec->path);
363 string_list_insert(changed, spec->path);
364 putchar('\n');
365 break;
367 case DIFF_STATUS_COPIED:
368 case DIFF_STATUS_RENAMED:
370 * If a change in the file corresponding to ospec->path
371 * has been observed, we cannot trust its contents
372 * because the diff is calculated based on the prior
373 * contents, not the current contents. So, declare a
374 * copy or rename only if there was no change observed.
376 if (!string_list_has_string(changed, ospec->path)) {
377 printf("%c ", q->queue[i]->status);
378 print_path(ospec->path);
379 putchar(' ');
380 print_path(spec->path);
381 string_list_insert(changed, spec->path);
382 putchar('\n');
384 if (!oidcmp(&ospec->oid, &spec->oid) &&
385 ospec->mode == spec->mode)
386 break;
388 /* fallthrough */
390 case DIFF_STATUS_TYPE_CHANGED:
391 case DIFF_STATUS_MODIFIED:
392 case DIFF_STATUS_ADDED:
394 * Links refer to objects in another repositories;
395 * output the SHA-1 verbatim.
397 if (no_data || S_ISGITLINK(spec->mode))
398 printf("M %06o %s ", spec->mode,
399 sha1_to_hex(anonymize ?
400 anonymize_sha1(&spec->oid) :
401 spec->oid.hash));
402 else {
403 struct object *object = lookup_object(spec->oid.hash);
404 printf("M %06o :%d ", spec->mode,
405 get_object_mark(object));
407 print_path(spec->path);
408 string_list_insert(changed, spec->path);
409 putchar('\n');
410 break;
412 default:
413 die("Unexpected comparison status '%c' for %s, %s",
414 q->queue[i]->status,
415 ospec->path ? ospec->path : "none",
416 spec->path ? spec->path : "none");
421 static const char *find_encoding(const char *begin, const char *end)
423 const char *needle = "\nencoding ";
424 char *bol, *eol;
426 bol = memmem(begin, end ? end - begin : strlen(begin),
427 needle, strlen(needle));
428 if (!bol)
429 return git_commit_encoding;
430 bol += strlen(needle);
431 eol = strchrnul(bol, '\n');
432 *eol = '\0';
433 return bol;
436 static void *anonymize_ref_component(const void *old, size_t *len)
438 static int counter;
439 struct strbuf out = STRBUF_INIT;
440 strbuf_addf(&out, "ref%d", counter++);
441 return strbuf_detach(&out, len);
444 static const char *anonymize_refname(const char *refname)
447 * If any of these prefixes is found, we will leave it intact
448 * so that tags remain tags and so forth.
450 static const char *prefixes[] = {
451 "refs/heads/",
452 "refs/tags/",
453 "refs/remotes/",
454 "refs/"
456 static struct hashmap refs;
457 static struct strbuf anon = STRBUF_INIT;
458 int i;
461 * We also leave "master" as a special case, since it does not reveal
462 * anything interesting.
464 if (!strcmp(refname, "refs/heads/master"))
465 return refname;
467 strbuf_reset(&anon);
468 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
469 if (skip_prefix(refname, prefixes[i], &refname)) {
470 strbuf_addstr(&anon, prefixes[i]);
471 break;
475 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
476 return anon.buf;
480 * We do not even bother to cache commit messages, as they are unlikely
481 * to be repeated verbatim, and it is not that interesting when they are.
483 static char *anonymize_commit_message(const char *old)
485 static int counter;
486 return xstrfmt("subject %d\n\nbody\n", counter++);
489 static struct hashmap idents;
490 static void *anonymize_ident(const void *old, size_t *len)
492 static int counter;
493 struct strbuf out = STRBUF_INIT;
494 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
495 counter++;
496 return strbuf_detach(&out, len);
500 * Our strategy here is to anonymize the names and email addresses,
501 * but keep timestamps intact, as they influence things like traversal
502 * order (and by themselves should not be too revealing).
504 static void anonymize_ident_line(const char **beg, const char **end)
506 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
507 static unsigned which_buffer;
509 struct strbuf *out;
510 struct ident_split split;
511 const char *end_of_header;
513 out = &buffers[which_buffer++];
514 which_buffer %= ARRAY_SIZE(buffers);
515 strbuf_reset(out);
517 /* skip "committer", "author", "tagger", etc */
518 end_of_header = strchr(*beg, ' ');
519 if (!end_of_header)
520 die("BUG: malformed line fed to anonymize_ident_line: %.*s",
521 (int)(*end - *beg), *beg);
522 end_of_header++;
523 strbuf_add(out, *beg, end_of_header - *beg);
525 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
526 split.date_begin) {
527 const char *ident;
528 size_t len;
530 len = split.mail_end - split.name_begin;
531 ident = anonymize_mem(&idents, anonymize_ident,
532 split.name_begin, &len);
533 strbuf_add(out, ident, len);
534 strbuf_addch(out, ' ');
535 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
536 } else {
537 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
540 *beg = out->buf;
541 *end = out->buf + out->len;
544 static void handle_commit(struct commit *commit, struct rev_info *rev,
545 struct string_list *paths_of_changed_objects)
547 int saved_output_format = rev->diffopt.output_format;
548 const char *commit_buffer;
549 const char *author, *author_end, *committer, *committer_end;
550 const char *encoding, *message;
551 char *reencoded = NULL;
552 struct commit_list *p;
553 const char *refname;
554 int i;
556 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
558 parse_commit_or_die(commit);
559 commit_buffer = get_commit_buffer(commit, NULL);
560 author = strstr(commit_buffer, "\nauthor ");
561 if (!author)
562 die ("Could not find author in commit %s",
563 oid_to_hex(&commit->object.oid));
564 author++;
565 author_end = strchrnul(author, '\n');
566 committer = strstr(author_end, "\ncommitter ");
567 if (!committer)
568 die ("Could not find committer in commit %s",
569 oid_to_hex(&commit->object.oid));
570 committer++;
571 committer_end = strchrnul(committer, '\n');
572 message = strstr(committer_end, "\n\n");
573 encoding = find_encoding(committer_end, message);
574 if (message)
575 message += 2;
577 if (commit->parents &&
578 get_object_mark(&commit->parents->item->object) != 0 &&
579 !full_tree) {
580 parse_commit_or_die(commit->parents->item);
581 diff_tree_oid(&commit->parents->item->tree->object.oid,
582 &commit->tree->object.oid, "", &rev->diffopt);
584 else
585 diff_root_tree_oid(&commit->tree->object.oid,
586 "", &rev->diffopt);
588 /* Export the referenced blobs, and remember the marks. */
589 for (i = 0; i < diff_queued_diff.nr; i++)
590 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
591 export_blob(&diff_queued_diff.queue[i]->two->oid);
593 refname = commit->util;
594 if (anonymize) {
595 refname = anonymize_refname(refname);
596 anonymize_ident_line(&committer, &committer_end);
597 anonymize_ident_line(&author, &author_end);
600 mark_next_object(&commit->object);
601 if (anonymize)
602 reencoded = anonymize_commit_message(message);
603 else if (!is_encoding_utf8(encoding))
604 reencoded = reencode_string(message, "UTF-8", encoding);
605 if (!commit->parents)
606 printf("reset %s\n", refname);
607 printf("commit %s\nmark :%"PRIu32"\n%.*s\n%.*s\ndata %u\n%s",
608 refname, last_idnum,
609 (int)(author_end - author), author,
610 (int)(committer_end - committer), committer,
611 (unsigned)(reencoded
612 ? strlen(reencoded) : message
613 ? strlen(message) : 0),
614 reencoded ? reencoded : message ? message : "");
615 free(reencoded);
616 unuse_commit_buffer(commit, commit_buffer);
618 for (i = 0, p = commit->parents; p; p = p->next) {
619 int mark = get_object_mark(&p->item->object);
620 if (!mark)
621 continue;
622 if (i == 0)
623 printf("from :%d\n", mark);
624 else
625 printf("merge :%d\n", mark);
626 i++;
629 if (full_tree)
630 printf("deleteall\n");
631 log_tree_diff_flush(rev);
632 string_list_clear(paths_of_changed_objects, 0);
633 rev->diffopt.output_format = saved_output_format;
635 printf("\n");
637 show_progress();
640 static void *anonymize_tag(const void *old, size_t *len)
642 static int counter;
643 struct strbuf out = STRBUF_INIT;
644 strbuf_addf(&out, "tag message %d", counter++);
645 return strbuf_detach(&out, len);
648 static void handle_tail(struct object_array *commits, struct rev_info *revs,
649 struct string_list *paths_of_changed_objects)
651 struct commit *commit;
652 while (commits->nr) {
653 commit = (struct commit *)object_array_pop(commits);
654 if (has_unshown_parent(commit))
655 return;
656 handle_commit(commit, revs, paths_of_changed_objects);
660 static void handle_tag(const char *name, struct tag *tag)
662 unsigned long size;
663 enum object_type type;
664 char *buf;
665 const char *tagger, *tagger_end, *message;
666 size_t message_size = 0;
667 struct object *tagged;
668 int tagged_mark;
669 struct commit *p;
671 /* Trees have no identifier in fast-export output, thus we have no way
672 * to output tags of trees, tags of tags of trees, etc. Simply omit
673 * such tags.
675 tagged = tag->tagged;
676 while (tagged->type == OBJ_TAG) {
677 tagged = ((struct tag *)tagged)->tagged;
679 if (tagged->type == OBJ_TREE) {
680 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
681 oid_to_hex(&tag->object.oid));
682 return;
685 buf = read_sha1_file(tag->object.oid.hash, &type, &size);
686 if (!buf)
687 die ("Could not read tag %s", oid_to_hex(&tag->object.oid));
688 message = memmem(buf, size, "\n\n", 2);
689 if (message) {
690 message += 2;
691 message_size = strlen(message);
693 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
694 if (!tagger) {
695 if (fake_missing_tagger)
696 tagger = "tagger Unspecified Tagger "
697 "<unspecified-tagger> 0 +0000";
698 else
699 tagger = "";
700 tagger_end = tagger + strlen(tagger);
701 } else {
702 tagger++;
703 tagger_end = strchrnul(tagger, '\n');
704 if (anonymize)
705 anonymize_ident_line(&tagger, &tagger_end);
708 if (anonymize) {
709 name = anonymize_refname(name);
710 if (message) {
711 static struct hashmap tags;
712 message = anonymize_mem(&tags, anonymize_tag,
713 message, &message_size);
717 /* handle signed tags */
718 if (message) {
719 const char *signature = strstr(message,
720 "\n-----BEGIN PGP SIGNATURE-----\n");
721 if (signature)
722 switch(signed_tag_mode) {
723 case ABORT:
724 die ("Encountered signed tag %s; use "
725 "--signed-tags=<mode> to handle it.",
726 oid_to_hex(&tag->object.oid));
727 case WARN:
728 warning ("Exporting signed tag %s",
729 oid_to_hex(&tag->object.oid));
730 /* fallthru */
731 case VERBATIM:
732 break;
733 case WARN_STRIP:
734 warning ("Stripping signature from tag %s",
735 oid_to_hex(&tag->object.oid));
736 /* fallthru */
737 case STRIP:
738 message_size = signature + 1 - message;
739 break;
743 /* handle tag->tagged having been filtered out due to paths specified */
744 tagged = tag->tagged;
745 tagged_mark = get_object_mark(tagged);
746 if (!tagged_mark) {
747 switch(tag_of_filtered_mode) {
748 case ABORT:
749 die ("Tag %s tags unexported object; use "
750 "--tag-of-filtered-object=<mode> to handle it.",
751 oid_to_hex(&tag->object.oid));
752 case DROP:
753 /* Ignore this tag altogether */
754 free(buf);
755 return;
756 case REWRITE:
757 if (tagged->type != OBJ_COMMIT) {
758 die ("Tag %s tags unexported %s!",
759 oid_to_hex(&tag->object.oid),
760 typename(tagged->type));
762 p = (struct commit *)tagged;
763 for (;;) {
764 if (p->parents && p->parents->next)
765 break;
766 if (p->object.flags & UNINTERESTING)
767 break;
768 if (!(p->object.flags & TREESAME))
769 break;
770 if (!p->parents)
771 die ("Can't find replacement commit for tag %s\n",
772 oid_to_hex(&tag->object.oid));
773 p = p->parents->item;
775 tagged_mark = get_object_mark(&p->object);
779 if (starts_with(name, "refs/tags/"))
780 name += 10;
781 printf("tag %s\nfrom :%d\n%.*s%sdata %d\n%.*s\n",
782 name, tagged_mark,
783 (int)(tagger_end - tagger), tagger,
784 tagger == tagger_end ? "" : "\n",
785 (int)message_size, (int)message_size, message ? message : "");
786 free(buf);
789 static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
791 switch (e->item->type) {
792 case OBJ_COMMIT:
793 return (struct commit *)e->item;
794 case OBJ_TAG: {
795 struct tag *tag = (struct tag *)e->item;
797 /* handle nested tags */
798 while (tag && tag->object.type == OBJ_TAG) {
799 parse_object(&tag->object.oid);
800 string_list_append(&extra_refs, full_name)->util = tag;
801 tag = (struct tag *)tag->tagged;
803 if (!tag)
804 die("Tag %s points nowhere?", e->name);
805 return (struct commit *)tag;
806 break;
808 default:
809 return NULL;
813 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
815 int i;
817 for (i = 0; i < info->nr; i++) {
818 struct rev_cmdline_entry *e = info->rev + i;
819 struct object_id oid;
820 struct commit *commit;
821 char *full_name;
823 if (e->flags & UNINTERESTING)
824 continue;
826 if (dwim_ref(e->name, strlen(e->name), &oid, &full_name) != 1)
827 continue;
829 if (refspecs) {
830 char *private;
831 private = apply_refspecs(refspecs, refspecs_nr, full_name);
832 if (private) {
833 free(full_name);
834 full_name = private;
838 commit = get_commit(e, full_name);
839 if (!commit) {
840 warning("%s: Unexpected object of type %s, skipping.",
841 e->name,
842 typename(e->item->type));
843 continue;
846 switch(commit->object.type) {
847 case OBJ_COMMIT:
848 break;
849 case OBJ_BLOB:
850 export_blob(&commit->object.oid);
851 continue;
852 default: /* OBJ_TAG (nested tags) is already handled */
853 warning("Tag points to object of unexpected type %s, skipping.",
854 typename(commit->object.type));
855 continue;
859 * This ref will not be updated through a commit, lets make
860 * sure it gets properly updated eventually.
862 if (commit->util || commit->object.flags & SHOWN)
863 string_list_append(&extra_refs, full_name)->util = commit;
864 if (!commit->util)
865 commit->util = full_name;
869 static void handle_tags_and_duplicates(void)
871 struct commit *commit;
872 int i;
874 for (i = extra_refs.nr - 1; i >= 0; i--) {
875 const char *name = extra_refs.items[i].string;
876 struct object *object = extra_refs.items[i].util;
877 switch (object->type) {
878 case OBJ_TAG:
879 handle_tag(name, (struct tag *)object);
880 break;
881 case OBJ_COMMIT:
882 if (anonymize)
883 name = anonymize_refname(name);
884 /* create refs pointing to already seen commits */
885 commit = (struct commit *)object;
886 printf("reset %s\nfrom :%d\n\n", name,
887 get_object_mark(&commit->object));
888 show_progress();
889 break;
894 static void export_marks(char *file)
896 unsigned int i;
897 uint32_t mark;
898 struct decoration_entry *deco = idnums.entries;
899 FILE *f;
900 int e = 0;
902 f = fopen_for_writing(file);
903 if (!f)
904 die_errno("Unable to open marks file %s for writing.", file);
906 for (i = 0; i < idnums.size; i++) {
907 if (deco->base && deco->base->type == 1) {
908 mark = ptr_to_mark(deco->decoration);
909 if (fprintf(f, ":%"PRIu32" %s\n", mark,
910 oid_to_hex(&deco->base->oid)) < 0) {
911 e = 1;
912 break;
915 deco++;
918 e |= ferror(f);
919 e |= fclose(f);
920 if (e)
921 error("Unable to write marks file %s.", file);
924 static void import_marks(char *input_file)
926 char line[512];
927 FILE *f = xfopen(input_file, "r");
929 while (fgets(line, sizeof(line), f)) {
930 uint32_t mark;
931 char *line_end, *mark_end;
932 struct object_id oid;
933 struct object *object;
934 struct commit *commit;
935 enum object_type type;
937 line_end = strchr(line, '\n');
938 if (line[0] != ':' || !line_end)
939 die("corrupt mark line: %s", line);
940 *line_end = '\0';
942 mark = strtoumax(line + 1, &mark_end, 10);
943 if (!mark || mark_end == line + 1
944 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
945 die("corrupt mark line: %s", line);
947 if (last_idnum < mark)
948 last_idnum = mark;
950 type = sha1_object_info(oid.hash, NULL);
951 if (type < 0)
952 die("object not found: %s", oid_to_hex(&oid));
954 if (type != OBJ_COMMIT)
955 /* only commits */
956 continue;
958 commit = lookup_commit(&oid);
959 if (!commit)
960 die("not a commit? can't happen: %s", oid_to_hex(&oid));
962 object = &commit->object;
964 if (object->flags & SHOWN)
965 error("Object %s already has a mark", oid_to_hex(&oid));
967 mark_object(object, mark);
969 object->flags |= SHOWN;
971 fclose(f);
974 static void handle_deletes(void)
976 int i;
977 for (i = 0; i < refspecs_nr; i++) {
978 struct refspec *refspec = &refspecs[i];
979 if (*refspec->src)
980 continue;
982 printf("reset %s\nfrom %s\n\n",
983 refspec->dst, sha1_to_hex(null_sha1));
987 int cmd_fast_export(int argc, const char **argv, const char *prefix)
989 struct rev_info revs;
990 struct object_array commits = OBJECT_ARRAY_INIT;
991 struct commit *commit;
992 char *export_filename = NULL, *import_filename = NULL;
993 uint32_t lastimportid;
994 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
995 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
996 struct option options[] = {
997 OPT_INTEGER(0, "progress", &progress,
998 N_("show progress after <n> objects")),
999 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1000 N_("select handling of signed tags"),
1001 parse_opt_signed_tag_mode),
1002 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1003 N_("select handling of tags that tag filtered objects"),
1004 parse_opt_tag_of_filtered_mode),
1005 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1006 N_("Dump marks to this file")),
1007 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1008 N_("Import marks from this file")),
1009 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1010 N_("Fake a tagger when tags lack one")),
1011 OPT_BOOL(0, "full-tree", &full_tree,
1012 N_("Output full tree for each commit")),
1013 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1014 N_("Use the done feature to terminate the stream")),
1015 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
1016 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1017 N_("Apply refspec to exported refs")),
1018 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1019 OPT_END()
1022 if (argc == 1)
1023 usage_with_options (fast_export_usage, options);
1025 /* we handle encodings */
1026 git_config(git_default_config, NULL);
1028 init_revisions(&revs, prefix);
1029 revs.topo_order = 1;
1030 revs.show_source = 1;
1031 revs.rewrite_parents = 1;
1032 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1033 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1034 argc = setup_revisions(argc, argv, &revs, NULL);
1035 if (argc > 1)
1036 usage_with_options (fast_export_usage, options);
1038 if (refspecs_list.nr) {
1039 const char **refspecs_str;
1040 int i;
1042 ALLOC_ARRAY(refspecs_str, refspecs_list.nr);
1043 for (i = 0; i < refspecs_list.nr; i++)
1044 refspecs_str[i] = refspecs_list.items[i].string;
1046 refspecs_nr = refspecs_list.nr;
1047 refspecs = parse_fetch_refspec(refspecs_nr, refspecs_str);
1049 string_list_clear(&refspecs_list, 1);
1050 free(refspecs_str);
1053 if (use_done_feature)
1054 printf("feature done\n");
1056 if (import_filename)
1057 import_marks(import_filename);
1058 lastimportid = last_idnum;
1060 if (import_filename && revs.prune_data.nr)
1061 full_tree = 1;
1063 get_tags_and_duplicates(&revs.cmdline);
1065 if (prepare_revision_walk(&revs))
1066 die("revision walk setup failed");
1067 revs.diffopt.format_callback = show_filemodify;
1068 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1069 revs.diffopt.flags.recursive = 1;
1070 while ((commit = get_revision(&revs))) {
1071 if (has_unshown_parent(commit)) {
1072 add_object_array(&commit->object, NULL, &commits);
1074 else {
1075 handle_commit(commit, &revs, &paths_of_changed_objects);
1076 handle_tail(&commits, &revs, &paths_of_changed_objects);
1080 handle_tags_and_duplicates();
1081 handle_deletes();
1083 if (export_filename && lastimportid != last_idnum)
1084 export_marks(export_filename);
1086 if (use_done_feature)
1087 printf("done\n");
1089 free_refspec(refspecs_nr, refspecs);
1091 return 0;