object_array: add and use `object_array_pop()`
[git/git.git] / builtin / fast-export.c
1 /*
2 * "git fast-export" builtin command
3 *
4 * Copyright (C) 2007 Johannes E. Schindelin
5 */
6 #include "builtin.h"
7 #include "cache.h"
8 #include "config.h"
9 #include "refs.h"
10 #include "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"
24
25 static const char *fast_export_usage[] = {
26 N_("git fast-export [rev-list-opts]"),
27 NULL
28 };
29
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;
41
42 static int parse_opt_signed_tag_mode(const struct option *opt,
43 const char *arg, int unset)
44 {
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;
58 }
59
60 static int parse_opt_tag_of_filtered_mode(const struct option *opt,
61 const char *arg, int unset)
62 {
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;
72 }
73
74 static struct decoration idnums;
75 static uint32_t last_idnum;
76
77 static int has_unshown_parent(struct commit *commit)
78 {
79 struct commit_list *parent;
80
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;
86 }
87
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;
94 };
95
96 static int anonymized_entry_cmp(const void *unused_cmp_data,
97 const void *va, const void *vb,
98 const void *unused_keydata)
99 {
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);
103 }
104
105 /*
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.
109 */
110 static const void *anonymize_mem(struct hashmap *map,
111 void *(*generate)(const void *, size_t *),
112 const void *orig, size_t *len)
113 {
114 struct anonymized_entry key, *ret;
115
116 if (!map->cmpfn)
117 hashmap_init(map, anonymized_entry_cmp, NULL, 0);
118
119 hashmap_entry_init(&key, memhash(orig, *len));
120 key.orig = orig;
121 key.orig_len = *len;
122 ret = hashmap_get(map, &key, NULL);
123
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);
132 }
133
134 *len = ret->anon_len;
135 return ret->anon;
136 }
137
138 /*
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.
143 */
144 static void anonymize_path(struct strbuf *out, const char *path,
145 struct hashmap *map,
146 void *(*generate)(const void *, size_t *))
147 {
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++);
156 }
157 }
158
159 /* Since intptr_t is C99, we do not use it here */
160 static inline uint32_t *mark_to_ptr(uint32_t mark)
161 {
162 return ((uint32_t *)NULL) + mark;
163 }
164
165 static inline uint32_t ptr_to_mark(void * mark)
166 {
167 return (uint32_t *)mark - (uint32_t *)NULL;
168 }
169
170 static inline void mark_object(struct object *object, uint32_t mark)
171 {
172 add_decoration(&idnums, object, mark_to_ptr(mark));
173 }
174
175 static inline void mark_next_object(struct object *object)
176 {
177 mark_object(object, ++last_idnum);
178 }
179
180 static int get_object_mark(struct object *object)
181 {
182 void *decoration = lookup_decoration(&idnums, object);
183 if (!decoration)
184 return 0;
185 return ptr_to_mark(decoration);
186 }
187
188 static void show_progress(void)
189 {
190 static int counter = 0;
191 if (!progress)
192 return;
193 if ((++counter % progress) == 0)
194 printf("progress %d objects\n", counter);
195 }
196
197 /*
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.
204 *
205 * There's no need to cache this result with anonymize_mem, since
206 * we already handle blob content caching with marks.
207 */
208 static char *anonymize_blob(unsigned long *size)
209 {
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);
215 }
216
217 static void export_blob(const struct object_id *oid)
218 {
219 unsigned long size;
220 enum object_type type;
221 char *buf;
222 struct object *object;
223 int eaten;
224
225 if (no_data)
226 return;
227
228 if (is_null_oid(oid))
229 return;
230
231 object = lookup_object(oid->hash);
232 if (object && object->flags & SHOWN)
233 return;
234
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);
246 }
247
248 if (!object)
249 die("Could not read blob %s", oid_to_hex(oid));
250
251 mark_next_object(object);
252
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");
257
258 show_progress();
259
260 object->flags |= SHOWN;
261 if (!eaten)
262 free(buf);
263 }
264
265 static int depth_first(const void *a_, const void *b_)
266 {
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;
272
273 name_a = a->one ? a->one->path : a->two->path;
274 name_b = b->one ? b->one->path : b->two->path;
275
276 len_a = strlen(name_a);
277 len_b = strlen(name_b);
278 len = (len_a < len_b) ? len_a : len_b;
279
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;
287 /*
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).
291 */
292 return (a->status == 'R') - (b->status == 'R');
293 }
294
295 static void print_path_1(const char *path)
296 {
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);
304 }
305
306 static void *anonymize_path_component(const void *path, size_t *len)
307 {
308 static int counter;
309 struct strbuf out = STRBUF_INIT;
310 strbuf_addf(&out, "path%d", counter++);
311 return strbuf_detach(&out, len);
312 }
313
314 static void print_path(const char *path)
315 {
316 if (!anonymize)
317 print_path_1(path);
318 else {
319 static struct hashmap paths;
320 static struct strbuf anon = STRBUF_INIT;
321
322 anonymize_path(&anon, path, &paths, anonymize_path_component);
323 print_path_1(anon.buf);
324 strbuf_reset(&anon);
325 }
326 }
327
328 static void *generate_fake_oid(const void *old, size_t *len)
329 {
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;
334 }
335
336 static const unsigned char *anonymize_sha1(const struct object_id *oid)
337 {
338 static struct hashmap sha1s;
339 size_t len = GIT_SHA1_RAWSZ;
340 return anonymize_mem(&sha1s, generate_fake_oid, oid, &len);
341 }
342
343 static void show_filemodify(struct diff_queue_struct *q,
344 struct diff_options *options, void *data)
345 {
346 int i;
347
348 /*
349 * Handle files below a directory first, in case they are all deleted
350 * and the directory changes to a file or symlink.
351 */
352 QSORT(q->queue, q->nr, depth_first);
353
354 for (i = 0; i < q->nr; i++) {
355 struct diff_filespec *ospec = q->queue[i]->one;
356 struct diff_filespec *spec = q->queue[i]->two;
357
358 switch (q->queue[i]->status) {
359 case DIFF_STATUS_DELETED:
360 printf("D ");
361 print_path(spec->path);
362 putchar('\n');
363 break;
364
365 case DIFF_STATUS_COPIED:
366 case DIFF_STATUS_RENAMED:
367 printf("%c ", q->queue[i]->status);
368 print_path(ospec->path);
369 putchar(' ');
370 print_path(spec->path);
371 putchar('\n');
372
373 if (!oidcmp(&ospec->oid, &spec->oid) &&
374 ospec->mode == spec->mode)
375 break;
376 /* fallthrough */
377
378 case DIFF_STATUS_TYPE_CHANGED:
379 case DIFF_STATUS_MODIFIED:
380 case DIFF_STATUS_ADDED:
381 /*
382 * Links refer to objects in another repositories;
383 * output the SHA-1 verbatim.
384 */
385 if (no_data || S_ISGITLINK(spec->mode))
386 printf("M %06o %s ", spec->mode,
387 sha1_to_hex(anonymize ?
388 anonymize_sha1(&spec->oid) :
389 spec->oid.hash));
390 else {
391 struct object *object = lookup_object(spec->oid.hash);
392 printf("M %06o :%d ", spec->mode,
393 get_object_mark(object));
394 }
395 print_path(spec->path);
396 putchar('\n');
397 break;
398
399 default:
400 die("Unexpected comparison status '%c' for %s, %s",
401 q->queue[i]->status,
402 ospec->path ? ospec->path : "none",
403 spec->path ? spec->path : "none");
404 }
405 }
406 }
407
408 static const char *find_encoding(const char *begin, const char *end)
409 {
410 const char *needle = "\nencoding ";
411 char *bol, *eol;
412
413 bol = memmem(begin, end ? end - begin : strlen(begin),
414 needle, strlen(needle));
415 if (!bol)
416 return git_commit_encoding;
417 bol += strlen(needle);
418 eol = strchrnul(bol, '\n');
419 *eol = '\0';
420 return bol;
421 }
422
423 static void *anonymize_ref_component(const void *old, size_t *len)
424 {
425 static int counter;
426 struct strbuf out = STRBUF_INIT;
427 strbuf_addf(&out, "ref%d", counter++);
428 return strbuf_detach(&out, len);
429 }
430
431 static const char *anonymize_refname(const char *refname)
432 {
433 /*
434 * If any of these prefixes is found, we will leave it intact
435 * so that tags remain tags and so forth.
436 */
437 static const char *prefixes[] = {
438 "refs/heads/",
439 "refs/tags/",
440 "refs/remotes/",
441 "refs/"
442 };
443 static struct hashmap refs;
444 static struct strbuf anon = STRBUF_INIT;
445 int i;
446
447 /*
448 * We also leave "master" as a special case, since it does not reveal
449 * anything interesting.
450 */
451 if (!strcmp(refname, "refs/heads/master"))
452 return refname;
453
454 strbuf_reset(&anon);
455 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
456 if (skip_prefix(refname, prefixes[i], &refname)) {
457 strbuf_addstr(&anon, prefixes[i]);
458 break;
459 }
460 }
461
462 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
463 return anon.buf;
464 }
465
466 /*
467 * We do not even bother to cache commit messages, as they are unlikely
468 * to be repeated verbatim, and it is not that interesting when they are.
469 */
470 static char *anonymize_commit_message(const char *old)
471 {
472 static int counter;
473 return xstrfmt("subject %d\n\nbody\n", counter++);
474 }
475
476 static struct hashmap idents;
477 static void *anonymize_ident(const void *old, size_t *len)
478 {
479 static int counter;
480 struct strbuf out = STRBUF_INIT;
481 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
482 counter++;
483 return strbuf_detach(&out, len);
484 }
485
486 /*
487 * Our strategy here is to anonymize the names and email addresses,
488 * but keep timestamps intact, as they influence things like traversal
489 * order (and by themselves should not be too revealing).
490 */
491 static void anonymize_ident_line(const char **beg, const char **end)
492 {
493 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
494 static unsigned which_buffer;
495
496 struct strbuf *out;
497 struct ident_split split;
498 const char *end_of_header;
499
500 out = &buffers[which_buffer++];
501 which_buffer %= ARRAY_SIZE(buffers);
502 strbuf_reset(out);
503
504 /* skip "committer", "author", "tagger", etc */
505 end_of_header = strchr(*beg, ' ');
506 if (!end_of_header)
507 die("BUG: malformed line fed to anonymize_ident_line: %.*s",
508 (int)(*end - *beg), *beg);
509 end_of_header++;
510 strbuf_add(out, *beg, end_of_header - *beg);
511
512 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
513 split.date_begin) {
514 const char *ident;
515 size_t len;
516
517 len = split.mail_end - split.name_begin;
518 ident = anonymize_mem(&idents, anonymize_ident,
519 split.name_begin, &len);
520 strbuf_add(out, ident, len);
521 strbuf_addch(out, ' ');
522 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
523 } else {
524 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
525 }
526
527 *beg = out->buf;
528 *end = out->buf + out->len;
529 }
530
531 static void handle_commit(struct commit *commit, struct rev_info *rev)
532 {
533 int saved_output_format = rev->diffopt.output_format;
534 const char *commit_buffer;
535 const char *author, *author_end, *committer, *committer_end;
536 const char *encoding, *message;
537 char *reencoded = NULL;
538 struct commit_list *p;
539 const char *refname;
540 int i;
541
542 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
543
544 parse_commit_or_die(commit);
545 commit_buffer = get_commit_buffer(commit, NULL);
546 author = strstr(commit_buffer, "\nauthor ");
547 if (!author)
548 die ("Could not find author in commit %s",
549 oid_to_hex(&commit->object.oid));
550 author++;
551 author_end = strchrnul(author, '\n');
552 committer = strstr(author_end, "\ncommitter ");
553 if (!committer)
554 die ("Could not find committer in commit %s",
555 oid_to_hex(&commit->object.oid));
556 committer++;
557 committer_end = strchrnul(committer, '\n');
558 message = strstr(committer_end, "\n\n");
559 encoding = find_encoding(committer_end, message);
560 if (message)
561 message += 2;
562
563 if (commit->parents &&
564 get_object_mark(&commit->parents->item->object) != 0 &&
565 !full_tree) {
566 parse_commit_or_die(commit->parents->item);
567 diff_tree_oid(&commit->parents->item->tree->object.oid,
568 &commit->tree->object.oid, "", &rev->diffopt);
569 }
570 else
571 diff_root_tree_oid(&commit->tree->object.oid,
572 "", &rev->diffopt);
573
574 /* Export the referenced blobs, and remember the marks. */
575 for (i = 0; i < diff_queued_diff.nr; i++)
576 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
577 export_blob(&diff_queued_diff.queue[i]->two->oid);
578
579 refname = commit->util;
580 if (anonymize) {
581 refname = anonymize_refname(refname);
582 anonymize_ident_line(&committer, &committer_end);
583 anonymize_ident_line(&author, &author_end);
584 }
585
586 mark_next_object(&commit->object);
587 if (anonymize)
588 reencoded = anonymize_commit_message(message);
589 else if (!is_encoding_utf8(encoding))
590 reencoded = reencode_string(message, "UTF-8", encoding);
591 if (!commit->parents)
592 printf("reset %s\n", refname);
593 printf("commit %s\nmark :%"PRIu32"\n%.*s\n%.*s\ndata %u\n%s",
594 refname, last_idnum,
595 (int)(author_end - author), author,
596 (int)(committer_end - committer), committer,
597 (unsigned)(reencoded
598 ? strlen(reencoded) : message
599 ? strlen(message) : 0),
600 reencoded ? reencoded : message ? message : "");
601 free(reencoded);
602 unuse_commit_buffer(commit, commit_buffer);
603
604 for (i = 0, p = commit->parents; p; p = p->next) {
605 int mark = get_object_mark(&p->item->object);
606 if (!mark)
607 continue;
608 if (i == 0)
609 printf("from :%d\n", mark);
610 else
611 printf("merge :%d\n", mark);
612 i++;
613 }
614
615 if (full_tree)
616 printf("deleteall\n");
617 log_tree_diff_flush(rev);
618 rev->diffopt.output_format = saved_output_format;
619
620 printf("\n");
621
622 show_progress();
623 }
624
625 static void *anonymize_tag(const void *old, size_t *len)
626 {
627 static int counter;
628 struct strbuf out = STRBUF_INIT;
629 strbuf_addf(&out, "tag message %d", counter++);
630 return strbuf_detach(&out, len);
631 }
632
633 static void handle_tail(struct object_array *commits, struct rev_info *revs)
634 {
635 struct commit *commit;
636 while (commits->nr) {
637 commit = (struct commit *)object_array_pop(commits);
638 if (has_unshown_parent(commit))
639 return;
640 handle_commit(commit, revs);
641 }
642 }
643
644 static void handle_tag(const char *name, struct tag *tag)
645 {
646 unsigned long size;
647 enum object_type type;
648 char *buf;
649 const char *tagger, *tagger_end, *message;
650 size_t message_size = 0;
651 struct object *tagged;
652 int tagged_mark;
653 struct commit *p;
654
655 /* Trees have no identifier in fast-export output, thus we have no way
656 * to output tags of trees, tags of tags of trees, etc. Simply omit
657 * such tags.
658 */
659 tagged = tag->tagged;
660 while (tagged->type == OBJ_TAG) {
661 tagged = ((struct tag *)tagged)->tagged;
662 }
663 if (tagged->type == OBJ_TREE) {
664 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
665 oid_to_hex(&tag->object.oid));
666 return;
667 }
668
669 buf = read_sha1_file(tag->object.oid.hash, &type, &size);
670 if (!buf)
671 die ("Could not read tag %s", oid_to_hex(&tag->object.oid));
672 message = memmem(buf, size, "\n\n", 2);
673 if (message) {
674 message += 2;
675 message_size = strlen(message);
676 }
677 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
678 if (!tagger) {
679 if (fake_missing_tagger)
680 tagger = "tagger Unspecified Tagger "
681 "<unspecified-tagger> 0 +0000";
682 else
683 tagger = "";
684 tagger_end = tagger + strlen(tagger);
685 } else {
686 tagger++;
687 tagger_end = strchrnul(tagger, '\n');
688 if (anonymize)
689 anonymize_ident_line(&tagger, &tagger_end);
690 }
691
692 if (anonymize) {
693 name = anonymize_refname(name);
694 if (message) {
695 static struct hashmap tags;
696 message = anonymize_mem(&tags, anonymize_tag,
697 message, &message_size);
698 }
699 }
700
701 /* handle signed tags */
702 if (message) {
703 const char *signature = strstr(message,
704 "\n-----BEGIN PGP SIGNATURE-----\n");
705 if (signature)
706 switch(signed_tag_mode) {
707 case ABORT:
708 die ("Encountered signed tag %s; use "
709 "--signed-tags=<mode> to handle it.",
710 oid_to_hex(&tag->object.oid));
711 case WARN:
712 warning ("Exporting signed tag %s",
713 oid_to_hex(&tag->object.oid));
714 /* fallthru */
715 case VERBATIM:
716 break;
717 case WARN_STRIP:
718 warning ("Stripping signature from tag %s",
719 oid_to_hex(&tag->object.oid));
720 /* fallthru */
721 case STRIP:
722 message_size = signature + 1 - message;
723 break;
724 }
725 }
726
727 /* handle tag->tagged having been filtered out due to paths specified */
728 tagged = tag->tagged;
729 tagged_mark = get_object_mark(tagged);
730 if (!tagged_mark) {
731 switch(tag_of_filtered_mode) {
732 case ABORT:
733 die ("Tag %s tags unexported object; use "
734 "--tag-of-filtered-object=<mode> to handle it.",
735 oid_to_hex(&tag->object.oid));
736 case DROP:
737 /* Ignore this tag altogether */
738 free(buf);
739 return;
740 case REWRITE:
741 if (tagged->type != OBJ_COMMIT) {
742 die ("Tag %s tags unexported %s!",
743 oid_to_hex(&tag->object.oid),
744 typename(tagged->type));
745 }
746 p = (struct commit *)tagged;
747 for (;;) {
748 if (p->parents && p->parents->next)
749 break;
750 if (p->object.flags & UNINTERESTING)
751 break;
752 if (!(p->object.flags & TREESAME))
753 break;
754 if (!p->parents)
755 die ("Can't find replacement commit for tag %s\n",
756 oid_to_hex(&tag->object.oid));
757 p = p->parents->item;
758 }
759 tagged_mark = get_object_mark(&p->object);
760 }
761 }
762
763 if (starts_with(name, "refs/tags/"))
764 name += 10;
765 printf("tag %s\nfrom :%d\n%.*s%sdata %d\n%.*s\n",
766 name, tagged_mark,
767 (int)(tagger_end - tagger), tagger,
768 tagger == tagger_end ? "" : "\n",
769 (int)message_size, (int)message_size, message ? message : "");
770 free(buf);
771 }
772
773 static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
774 {
775 switch (e->item->type) {
776 case OBJ_COMMIT:
777 return (struct commit *)e->item;
778 case OBJ_TAG: {
779 struct tag *tag = (struct tag *)e->item;
780
781 /* handle nested tags */
782 while (tag && tag->object.type == OBJ_TAG) {
783 parse_object(&tag->object.oid);
784 string_list_append(&extra_refs, full_name)->util = tag;
785 tag = (struct tag *)tag->tagged;
786 }
787 if (!tag)
788 die("Tag %s points nowhere?", e->name);
789 return (struct commit *)tag;
790 break;
791 }
792 default:
793 return NULL;
794 }
795 }
796
797 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
798 {
799 int i;
800
801 for (i = 0; i < info->nr; i++) {
802 struct rev_cmdline_entry *e = info->rev + i;
803 struct object_id oid;
804 struct commit *commit;
805 char *full_name;
806
807 if (e->flags & UNINTERESTING)
808 continue;
809
810 if (dwim_ref(e->name, strlen(e->name), oid.hash, &full_name) != 1)
811 continue;
812
813 if (refspecs) {
814 char *private;
815 private = apply_refspecs(refspecs, refspecs_nr, full_name);
816 if (private) {
817 free(full_name);
818 full_name = private;
819 }
820 }
821
822 commit = get_commit(e, full_name);
823 if (!commit) {
824 warning("%s: Unexpected object of type %s, skipping.",
825 e->name,
826 typename(e->item->type));
827 continue;
828 }
829
830 switch(commit->object.type) {
831 case OBJ_COMMIT:
832 break;
833 case OBJ_BLOB:
834 export_blob(&commit->object.oid);
835 continue;
836 default: /* OBJ_TAG (nested tags) is already handled */
837 warning("Tag points to object of unexpected type %s, skipping.",
838 typename(commit->object.type));
839 continue;
840 }
841
842 /*
843 * This ref will not be updated through a commit, lets make
844 * sure it gets properly updated eventually.
845 */
846 if (commit->util || commit->object.flags & SHOWN)
847 string_list_append(&extra_refs, full_name)->util = commit;
848 if (!commit->util)
849 commit->util = full_name;
850 }
851 }
852
853 static void handle_tags_and_duplicates(void)
854 {
855 struct commit *commit;
856 int i;
857
858 for (i = extra_refs.nr - 1; i >= 0; i--) {
859 const char *name = extra_refs.items[i].string;
860 struct object *object = extra_refs.items[i].util;
861 switch (object->type) {
862 case OBJ_TAG:
863 handle_tag(name, (struct tag *)object);
864 break;
865 case OBJ_COMMIT:
866 if (anonymize)
867 name = anonymize_refname(name);
868 /* create refs pointing to already seen commits */
869 commit = (struct commit *)object;
870 printf("reset %s\nfrom :%d\n\n", name,
871 get_object_mark(&commit->object));
872 show_progress();
873 break;
874 }
875 }
876 }
877
878 static void export_marks(char *file)
879 {
880 unsigned int i;
881 uint32_t mark;
882 struct object_decoration *deco = idnums.hash;
883 FILE *f;
884 int e = 0;
885
886 f = fopen_for_writing(file);
887 if (!f)
888 die_errno("Unable to open marks file %s for writing.", file);
889
890 for (i = 0; i < idnums.size; i++) {
891 if (deco->base && deco->base->type == 1) {
892 mark = ptr_to_mark(deco->decoration);
893 if (fprintf(f, ":%"PRIu32" %s\n", mark,
894 oid_to_hex(&deco->base->oid)) < 0) {
895 e = 1;
896 break;
897 }
898 }
899 deco++;
900 }
901
902 e |= ferror(f);
903 e |= fclose(f);
904 if (e)
905 error("Unable to write marks file %s.", file);
906 }
907
908 static void import_marks(char *input_file)
909 {
910 char line[512];
911 FILE *f = xfopen(input_file, "r");
912
913 while (fgets(line, sizeof(line), f)) {
914 uint32_t mark;
915 char *line_end, *mark_end;
916 struct object_id oid;
917 struct object *object;
918 struct commit *commit;
919 enum object_type type;
920
921 line_end = strchr(line, '\n');
922 if (line[0] != ':' || !line_end)
923 die("corrupt mark line: %s", line);
924 *line_end = '\0';
925
926 mark = strtoumax(line + 1, &mark_end, 10);
927 if (!mark || mark_end == line + 1
928 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
929 die("corrupt mark line: %s", line);
930
931 if (last_idnum < mark)
932 last_idnum = mark;
933
934 type = sha1_object_info(oid.hash, NULL);
935 if (type < 0)
936 die("object not found: %s", oid_to_hex(&oid));
937
938 if (type != OBJ_COMMIT)
939 /* only commits */
940 continue;
941
942 commit = lookup_commit(&oid);
943 if (!commit)
944 die("not a commit? can't happen: %s", oid_to_hex(&oid));
945
946 object = &commit->object;
947
948 if (object->flags & SHOWN)
949 error("Object %s already has a mark", oid_to_hex(&oid));
950
951 mark_object(object, mark);
952
953 object->flags |= SHOWN;
954 }
955 fclose(f);
956 }
957
958 static void handle_deletes(void)
959 {
960 int i;
961 for (i = 0; i < refspecs_nr; i++) {
962 struct refspec *refspec = &refspecs[i];
963 if (*refspec->src)
964 continue;
965
966 printf("reset %s\nfrom %s\n\n",
967 refspec->dst, sha1_to_hex(null_sha1));
968 }
969 }
970
971 int cmd_fast_export(int argc, const char **argv, const char *prefix)
972 {
973 struct rev_info revs;
974 struct object_array commits = OBJECT_ARRAY_INIT;
975 struct commit *commit;
976 char *export_filename = NULL, *import_filename = NULL;
977 uint32_t lastimportid;
978 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
979 struct option options[] = {
980 OPT_INTEGER(0, "progress", &progress,
981 N_("show progress after <n> objects")),
982 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
983 N_("select handling of signed tags"),
984 parse_opt_signed_tag_mode),
985 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
986 N_("select handling of tags that tag filtered objects"),
987 parse_opt_tag_of_filtered_mode),
988 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
989 N_("Dump marks to this file")),
990 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
991 N_("Import marks from this file")),
992 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
993 N_("Fake a tagger when tags lack one")),
994 OPT_BOOL(0, "full-tree", &full_tree,
995 N_("Output full tree for each commit")),
996 OPT_BOOL(0, "use-done-feature", &use_done_feature,
997 N_("Use the done feature to terminate the stream")),
998 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
999 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1000 N_("Apply refspec to exported refs")),
1001 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1002 OPT_END()
1003 };
1004
1005 if (argc == 1)
1006 usage_with_options (fast_export_usage, options);
1007
1008 /* we handle encodings */
1009 git_config(git_default_config, NULL);
1010
1011 init_revisions(&revs, prefix);
1012 revs.topo_order = 1;
1013 revs.show_source = 1;
1014 revs.rewrite_parents = 1;
1015 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1016 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1017 argc = setup_revisions(argc, argv, &revs, NULL);
1018 if (argc > 1)
1019 usage_with_options (fast_export_usage, options);
1020
1021 if (refspecs_list.nr) {
1022 const char **refspecs_str;
1023 int i;
1024
1025 ALLOC_ARRAY(refspecs_str, refspecs_list.nr);
1026 for (i = 0; i < refspecs_list.nr; i++)
1027 refspecs_str[i] = refspecs_list.items[i].string;
1028
1029 refspecs_nr = refspecs_list.nr;
1030 refspecs = parse_fetch_refspec(refspecs_nr, refspecs_str);
1031
1032 string_list_clear(&refspecs_list, 1);
1033 free(refspecs_str);
1034 }
1035
1036 if (use_done_feature)
1037 printf("feature done\n");
1038
1039 if (import_filename)
1040 import_marks(import_filename);
1041 lastimportid = last_idnum;
1042
1043 if (import_filename && revs.prune_data.nr)
1044 full_tree = 1;
1045
1046 get_tags_and_duplicates(&revs.cmdline);
1047
1048 if (prepare_revision_walk(&revs))
1049 die("revision walk setup failed");
1050 revs.diffopt.format_callback = show_filemodify;
1051 DIFF_OPT_SET(&revs.diffopt, RECURSIVE);
1052 while ((commit = get_revision(&revs))) {
1053 if (has_unshown_parent(commit)) {
1054 add_object_array(&commit->object, NULL, &commits);
1055 }
1056 else {
1057 handle_commit(commit, &revs);
1058 handle_tail(&commits, &revs);
1059 }
1060 }
1061
1062 handle_tags_and_duplicates();
1063 handle_deletes();
1064
1065 if (export_filename && lastimportid != last_idnum)
1066 export_marks(export_filename);
1067
1068 if (use_done_feature)
1069 printf("done\n");
1070
1071 free_refspec(refspecs_nr, refspecs);
1072
1073 return 0;
1074 }