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 "object-store.h"
11#include "commit.h"
12#include "object.h"
13#include "tag.h"
14#include "diff.h"
15#include "diffcore.h"
16#include "log-tree.h"
17#include "revision.h"
18#include "decorate.h"
19#include "string-list.h"
20#include "utf8.h"
21#include "parse-options.h"
22#include "quote.h"
23#include "remote.h"
24#include "blob.h"
25
26static const char *fast_export_usage[] = {
27 N_("git fast-export [rev-list-opts]"),
28 NULL
29};
30
31static int progress;
32static enum { ABORT, VERBATIM, WARN, WARN_STRIP, STRIP } signed_tag_mode = ABORT;
33static enum { ERROR, DROP, REWRITE } tag_of_filtered_mode = ERROR;
34static int fake_missing_tagger;
35static int use_done_feature;
36static int no_data;
37static int full_tree;
38static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
39static struct refspec *refspecs;
40static int refspecs_nr;
41static int anonymize;
42
43static int parse_opt_signed_tag_mode(const struct option *opt,
44 const char *arg, int unset)
45{
46 if (unset || !strcmp(arg, "abort"))
47 signed_tag_mode = ABORT;
48 else if (!strcmp(arg, "verbatim") || !strcmp(arg, "ignore"))
49 signed_tag_mode = VERBATIM;
50 else if (!strcmp(arg, "warn"))
51 signed_tag_mode = WARN;
52 else if (!strcmp(arg, "warn-strip"))
53 signed_tag_mode = WARN_STRIP;
54 else if (!strcmp(arg, "strip"))
55 signed_tag_mode = STRIP;
56 else
57 return error("Unknown signed-tags mode: %s", arg);
58 return 0;
59}
60
61static int parse_opt_tag_of_filtered_mode(const struct option *opt,
62 const char *arg, int unset)
63{
64 if (unset || !strcmp(arg, "abort"))
65 tag_of_filtered_mode = ERROR;
66 else if (!strcmp(arg, "drop"))
67 tag_of_filtered_mode = DROP;
68 else if (!strcmp(arg, "rewrite"))
69 tag_of_filtered_mode = REWRITE;
70 else
71 return error("Unknown tag-of-filtered mode: %s", arg);
72 return 0;
73}
74
75static struct decoration idnums;
76static uint32_t last_idnum;
77
78static int has_unshown_parent(struct commit *commit)
79{
80 struct commit_list *parent;
81
82 for (parent = commit->parents; parent; parent = parent->next)
83 if (!(parent->item->object.flags & SHOWN) &&
84 !(parent->item->object.flags & UNINTERESTING))
85 return 1;
86 return 0;
87}
88
89struct anonymized_entry {
90 struct hashmap_entry hash;
91 const char *orig;
92 size_t orig_len;
93 const char *anon;
94 size_t anon_len;
95};
96
97static int anonymized_entry_cmp(const void *unused_cmp_data,
98 const void *va, const void *vb,
99 const void *unused_keydata)
100{
101 const struct anonymized_entry *a = va, *b = vb;
102 return a->orig_len != b->orig_len ||
103 memcmp(a->orig, b->orig, a->orig_len);
104}
105
106/*
107 * Basically keep a cache of X->Y so that we can repeatedly replace
108 * the same anonymized string with another. The actual generation
109 * is farmed out to the generate function.
110 */
111static const void *anonymize_mem(struct hashmap *map,
112 void *(*generate)(const void *, size_t *),
113 const void *orig, size_t *len)
114{
115 struct anonymized_entry key, *ret;
116
117 if (!map->cmpfn)
118 hashmap_init(map, anonymized_entry_cmp, NULL, 0);
119
120 hashmap_entry_init(&key, memhash(orig, *len));
121 key.orig = orig;
122 key.orig_len = *len;
123 ret = hashmap_get(map, &key, NULL);
124
125 if (!ret) {
126 ret = xmalloc(sizeof(*ret));
127 hashmap_entry_init(&ret->hash, key.hash.hash);
128 ret->orig = xstrdup(orig);
129 ret->orig_len = *len;
130 ret->anon = generate(orig, len);
131 ret->anon_len = *len;
132 hashmap_put(map, ret);
133 }
134
135 *len = ret->anon_len;
136 return ret->anon;
137}
138
139/*
140 * We anonymize each component of a path individually,
141 * so that paths a/b and a/c will share a common root.
142 * The paths are cached via anonymize_mem so that repeated
143 * lookups for "a" will yield the same value.
144 */
145static void anonymize_path(struct strbuf *out, const char *path,
146 struct hashmap *map,
147 void *(*generate)(const void *, size_t *))
148{
149 while (*path) {
150 const char *end_of_component = strchrnul(path, '/');
151 size_t len = end_of_component - path;
152 const char *c = anonymize_mem(map, generate, path, &len);
153 strbuf_add(out, c, len);
154 path = end_of_component;
155 if (*path)
156 strbuf_addch(out, *path++);
157 }
158}
159
160/* Since intptr_t is C99, we do not use it here */
161static inline uint32_t *mark_to_ptr(uint32_t mark)
162{
163 return ((uint32_t *)NULL) + mark;
164}
165
166static inline uint32_t ptr_to_mark(void * mark)
167{
168 return (uint32_t *)mark - (uint32_t *)NULL;
169}
170
171static inline void mark_object(struct object *object, uint32_t mark)
172{
173 add_decoration(&idnums, object, mark_to_ptr(mark));
174}
175
176static inline void mark_next_object(struct object *object)
177{
178 mark_object(object, ++last_idnum);
179}
180
181static int get_object_mark(struct object *object)
182{
183 void *decoration = lookup_decoration(&idnums, object);
184 if (!decoration)
185 return 0;
186 return ptr_to_mark(decoration);
187}
188
189static void show_progress(void)
190{
191 static int counter = 0;
192 if (!progress)
193 return;
194 if ((++counter % progress) == 0)
195 printf("progress %d objects\n", counter);
196}
197
198/*
199 * Ideally we would want some transformation of the blob data here
200 * that is unreversible, but would still be the same size and have
201 * the same data relationship to other blobs (so that we get the same
202 * delta and packing behavior as the original). But the first and last
203 * requirements there are probably mutually exclusive, so let's take
204 * the easy way out for now, and just generate arbitrary content.
205 *
206 * There's no need to cache this result with anonymize_mem, since
207 * we already handle blob content caching with marks.
208 */
209static char *anonymize_blob(unsigned long *size)
210{
211 static int counter;
212 struct strbuf out = STRBUF_INIT;
213 strbuf_addf(&out, "anonymous blob %d", counter++);
214 *size = out.len;
215 return strbuf_detach(&out, NULL);
216}
217
218static void export_blob(const struct object_id *oid)
219{
220 unsigned long size;
221 enum object_type type;
222 char *buf;
223 struct object *object;
224 int eaten;
225
226 if (no_data)
227 return;
228
229 if (is_null_oid(oid))
230 return;
231
232 object = lookup_object(oid->hash);
233 if (object && object->flags & SHOWN)
234 return;
235
236 if (anonymize) {
237 buf = anonymize_blob(&size);
238 object = (struct object *)lookup_blob(oid);
239 eaten = 0;
240 } else {
241 buf = read_object_file(oid, &type, &size);
242 if (!buf)
243 die ("Could not read blob %s", oid_to_hex(oid));
244 if (check_object_signature(oid, buf, size, type_name(type)) < 0)
245 die("sha1 mismatch in blob %s", oid_to_hex(oid));
246 object = parse_object_buffer(oid, type, size, buf, &eaten);
247 }
248
249 if (!object)
250 die("Could not read blob %s", oid_to_hex(oid));
251
252 mark_next_object(object);
253
254 printf("blob\nmark :%"PRIu32"\ndata %lu\n", last_idnum, size);
255 if (size && fwrite(buf, size, 1, stdout) != 1)
256 die_errno ("Could not write blob '%s'", oid_to_hex(oid));
257 printf("\n");
258
259 show_progress();
260
261 object->flags |= SHOWN;
262 if (!eaten)
263 free(buf);
264}
265
266static int depth_first(const void *a_, const void *b_)
267{
268 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
269 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
270 const char *name_a, *name_b;
271 int len_a, len_b, len;
272 int cmp;
273
274 name_a = a->one ? a->one->path : a->two->path;
275 name_b = b->one ? b->one->path : b->two->path;
276
277 len_a = strlen(name_a);
278 len_b = strlen(name_b);
279 len = (len_a < len_b) ? len_a : len_b;
280
281 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
282 cmp = memcmp(name_a, name_b, len);
283 if (cmp)
284 return cmp;
285 cmp = len_b - len_a;
286 if (cmp)
287 return cmp;
288 /*
289 * Move 'R'ename entries last so that all references of the file
290 * appear in the output before it is renamed (e.g., when a file
291 * was copied and renamed in the same commit).
292 */
293 return (a->status == 'R') - (b->status == 'R');
294}
295
296static void print_path_1(const char *path)
297{
298 int need_quote = quote_c_style(path, NULL, NULL, 0);
299 if (need_quote)
300 quote_c_style(path, NULL, stdout, 0);
301 else if (strchr(path, ' '))
302 printf("\"%s\"", path);
303 else
304 printf("%s", path);
305}
306
307static void *anonymize_path_component(const void *path, size_t *len)
308{
309 static int counter;
310 struct strbuf out = STRBUF_INIT;
311 strbuf_addf(&out, "path%d", counter++);
312 return strbuf_detach(&out, len);
313}
314
315static void print_path(const char *path)
316{
317 if (!anonymize)
318 print_path_1(path);
319 else {
320 static struct hashmap paths;
321 static struct strbuf anon = STRBUF_INIT;
322
323 anonymize_path(&anon, path, &paths, anonymize_path_component);
324 print_path_1(anon.buf);
325 strbuf_reset(&anon);
326 }
327}
328
329static void *generate_fake_oid(const void *old, size_t *len)
330{
331 static uint32_t counter = 1; /* avoid null sha1 */
332 unsigned char *out = xcalloc(GIT_SHA1_RAWSZ, 1);
333 put_be32(out + GIT_SHA1_RAWSZ - 4, counter++);
334 return out;
335}
336
337static const unsigned char *anonymize_sha1(const struct object_id *oid)
338{
339 static struct hashmap sha1s;
340 size_t len = GIT_SHA1_RAWSZ;
341 return anonymize_mem(&sha1s, generate_fake_oid, oid, &len);
342}
343
344static void show_filemodify(struct diff_queue_struct *q,
345 struct diff_options *options, void *data)
346{
347 int i;
348 struct string_list *changed = data;
349
350 /*
351 * Handle files below a directory first, in case they are all deleted
352 * and the directory changes to a file or symlink.
353 */
354 QSORT(q->queue, q->nr, depth_first);
355
356 for (i = 0; i < q->nr; i++) {
357 struct diff_filespec *ospec = q->queue[i]->one;
358 struct diff_filespec *spec = q->queue[i]->two;
359
360 switch (q->queue[i]->status) {
361 case DIFF_STATUS_DELETED:
362 printf("D ");
363 print_path(spec->path);
364 string_list_insert(changed, spec->path);
365 putchar('\n');
366 break;
367
368 case DIFF_STATUS_COPIED:
369 case DIFF_STATUS_RENAMED:
370 /*
371 * If a change in the file corresponding to ospec->path
372 * has been observed, we cannot trust its contents
373 * because the diff is calculated based on the prior
374 * contents, not the current contents. So, declare a
375 * copy or rename only if there was no change observed.
376 */
377 if (!string_list_has_string(changed, ospec->path)) {
378 printf("%c ", q->queue[i]->status);
379 print_path(ospec->path);
380 putchar(' ');
381 print_path(spec->path);
382 string_list_insert(changed, spec->path);
383 putchar('\n');
384
385 if (!oidcmp(&ospec->oid, &spec->oid) &&
386 ospec->mode == spec->mode)
387 break;
388 }
389 /* fallthrough */
390
391 case DIFF_STATUS_TYPE_CHANGED:
392 case DIFF_STATUS_MODIFIED:
393 case DIFF_STATUS_ADDED:
394 /*
395 * Links refer to objects in another repositories;
396 * output the SHA-1 verbatim.
397 */
398 if (no_data || S_ISGITLINK(spec->mode))
399 printf("M %06o %s ", spec->mode,
400 sha1_to_hex(anonymize ?
401 anonymize_sha1(&spec->oid) :
402 spec->oid.hash));
403 else {
404 struct object *object = lookup_object(spec->oid.hash);
405 printf("M %06o :%d ", spec->mode,
406 get_object_mark(object));
407 }
408 print_path(spec->path);
409 string_list_insert(changed, spec->path);
410 putchar('\n');
411 break;
412
413 default:
414 die("Unexpected comparison status '%c' for %s, %s",
415 q->queue[i]->status,
416 ospec->path ? ospec->path : "none",
417 spec->path ? spec->path : "none");
418 }
419 }
420}
421
422static const char *find_encoding(const char *begin, const char *end)
423{
424 const char *needle = "\nencoding ";
425 char *bol, *eol;
426
427 bol = memmem(begin, end ? end - begin : strlen(begin),
428 needle, strlen(needle));
429 if (!bol)
430 return git_commit_encoding;
431 bol += strlen(needle);
432 eol = strchrnul(bol, '\n');
433 *eol = '\0';
434 return bol;
435}
436
437static void *anonymize_ref_component(const void *old, size_t *len)
438{
439 static int counter;
440 struct strbuf out = STRBUF_INIT;
441 strbuf_addf(&out, "ref%d", counter++);
442 return strbuf_detach(&out, len);
443}
444
445static const char *anonymize_refname(const char *refname)
446{
447 /*
448 * If any of these prefixes is found, we will leave it intact
449 * so that tags remain tags and so forth.
450 */
451 static const char *prefixes[] = {
452 "refs/heads/",
453 "refs/tags/",
454 "refs/remotes/",
455 "refs/"
456 };
457 static struct hashmap refs;
458 static struct strbuf anon = STRBUF_INIT;
459 int i;
460
461 /*
462 * We also leave "master" as a special case, since it does not reveal
463 * anything interesting.
464 */
465 if (!strcmp(refname, "refs/heads/master"))
466 return refname;
467
468 strbuf_reset(&anon);
469 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
470 if (skip_prefix(refname, prefixes[i], &refname)) {
471 strbuf_addstr(&anon, prefixes[i]);
472 break;
473 }
474 }
475
476 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
477 return anon.buf;
478}
479
480/*
481 * We do not even bother to cache commit messages, as they are unlikely
482 * to be repeated verbatim, and it is not that interesting when they are.
483 */
484static char *anonymize_commit_message(const char *old)
485{
486 static int counter;
487 return xstrfmt("subject %d\n\nbody\n", counter++);
488}
489
490static struct hashmap idents;
491static void *anonymize_ident(const void *old, size_t *len)
492{
493 static int counter;
494 struct strbuf out = STRBUF_INIT;
495 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
496 counter++;
497 return strbuf_detach(&out, len);
498}
499
500/*
501 * Our strategy here is to anonymize the names and email addresses,
502 * but keep timestamps intact, as they influence things like traversal
503 * order (and by themselves should not be too revealing).
504 */
505static void anonymize_ident_line(const char **beg, const char **end)
506{
507 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
508 static unsigned which_buffer;
509
510 struct strbuf *out;
511 struct ident_split split;
512 const char *end_of_header;
513
514 out = &buffers[which_buffer++];
515 which_buffer %= ARRAY_SIZE(buffers);
516 strbuf_reset(out);
517
518 /* skip "committer", "author", "tagger", etc */
519 end_of_header = strchr(*beg, ' ');
520 if (!end_of_header)
521 die("BUG: malformed line fed to anonymize_ident_line: %.*s",
522 (int)(*end - *beg), *beg);
523 end_of_header++;
524 strbuf_add(out, *beg, end_of_header - *beg);
525
526 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
527 split.date_begin) {
528 const char *ident;
529 size_t len;
530
531 len = split.mail_end - split.name_begin;
532 ident = anonymize_mem(&idents, anonymize_ident,
533 split.name_begin, &len);
534 strbuf_add(out, ident, len);
535 strbuf_addch(out, ' ');
536 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
537 } else {
538 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
539 }
540
541 *beg = out->buf;
542 *end = out->buf + out->len;
543}
544
545static void handle_commit(struct commit *commit, struct rev_info *rev,
546 struct string_list *paths_of_changed_objects)
547{
548 int saved_output_format = rev->diffopt.output_format;
549 const char *commit_buffer;
550 const char *author, *author_end, *committer, *committer_end;
551 const char *encoding, *message;
552 char *reencoded = NULL;
553 struct commit_list *p;
554 const char *refname;
555 int i;
556
557 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
558
559 parse_commit_or_die(commit);
560 commit_buffer = get_commit_buffer(commit, NULL);
561 author = strstr(commit_buffer, "\nauthor ");
562 if (!author)
563 die ("Could not find author in commit %s",
564 oid_to_hex(&commit->object.oid));
565 author++;
566 author_end = strchrnul(author, '\n');
567 committer = strstr(author_end, "\ncommitter ");
568 if (!committer)
569 die ("Could not find committer in commit %s",
570 oid_to_hex(&commit->object.oid));
571 committer++;
572 committer_end = strchrnul(committer, '\n');
573 message = strstr(committer_end, "\n\n");
574 encoding = find_encoding(committer_end, message);
575 if (message)
576 message += 2;
577
578 if (commit->parents &&
579 get_object_mark(&commit->parents->item->object) != 0 &&
580 !full_tree) {
581 parse_commit_or_die(commit->parents->item);
582 diff_tree_oid(&commit->parents->item->tree->object.oid,
583 &commit->tree->object.oid, "", &rev->diffopt);
584 }
585 else
586 diff_root_tree_oid(&commit->tree->object.oid,
587 "", &rev->diffopt);
588
589 /* Export the referenced blobs, and remember the marks. */
590 for (i = 0; i < diff_queued_diff.nr; i++)
591 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
592 export_blob(&diff_queued_diff.queue[i]->two->oid);
593
594 refname = commit->util;
595 if (anonymize) {
596 refname = anonymize_refname(refname);
597 anonymize_ident_line(&committer, &committer_end);
598 anonymize_ident_line(&author, &author_end);
599 }
600
601 mark_next_object(&commit->object);
602 if (anonymize)
603 reencoded = anonymize_commit_message(message);
604 else if (!is_encoding_utf8(encoding))
605 reencoded = reencode_string(message, "UTF-8", encoding);
606 if (!commit->parents)
607 printf("reset %s\n", refname);
608 printf("commit %s\nmark :%"PRIu32"\n%.*s\n%.*s\ndata %u\n%s",
609 refname, last_idnum,
610 (int)(author_end - author), author,
611 (int)(committer_end - committer), committer,
612 (unsigned)(reencoded
613 ? strlen(reencoded) : message
614 ? strlen(message) : 0),
615 reencoded ? reencoded : message ? message : "");
616 free(reencoded);
617 unuse_commit_buffer(commit, commit_buffer);
618
619 for (i = 0, p = commit->parents; p; p = p->next) {
620 int mark = get_object_mark(&p->item->object);
621 if (!mark)
622 continue;
623 if (i == 0)
624 printf("from :%d\n", mark);
625 else
626 printf("merge :%d\n", mark);
627 i++;
628 }
629
630 if (full_tree)
631 printf("deleteall\n");
632 log_tree_diff_flush(rev);
633 string_list_clear(paths_of_changed_objects, 0);
634 rev->diffopt.output_format = saved_output_format;
635
636 printf("\n");
637
638 show_progress();
639}
640
641static void *anonymize_tag(const void *old, size_t *len)
642{
643 static int counter;
644 struct strbuf out = STRBUF_INIT;
645 strbuf_addf(&out, "tag message %d", counter++);
646 return strbuf_detach(&out, len);
647}
648
649static void handle_tail(struct object_array *commits, struct rev_info *revs,
650 struct string_list *paths_of_changed_objects)
651{
652 struct commit *commit;
653 while (commits->nr) {
654 commit = (struct commit *)object_array_pop(commits);
655 if (has_unshown_parent(commit))
656 return;
657 handle_commit(commit, revs, paths_of_changed_objects);
658 }
659}
660
661static void handle_tag(const char *name, struct tag *tag)
662{
663 unsigned long size;
664 enum object_type type;
665 char *buf;
666 const char *tagger, *tagger_end, *message;
667 size_t message_size = 0;
668 struct object *tagged;
669 int tagged_mark;
670 struct commit *p;
671
672 /* Trees have no identifier in fast-export output, thus we have no way
673 * to output tags of trees, tags of tags of trees, etc. Simply omit
674 * such tags.
675 */
676 tagged = tag->tagged;
677 while (tagged->type == OBJ_TAG) {
678 tagged = ((struct tag *)tagged)->tagged;
679 }
680 if (tagged->type == OBJ_TREE) {
681 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
682 oid_to_hex(&tag->object.oid));
683 return;
684 }
685
686 buf = read_object_file(&tag->object.oid, &type, &size);
687 if (!buf)
688 die ("Could not read tag %s", oid_to_hex(&tag->object.oid));
689 message = memmem(buf, size, "\n\n", 2);
690 if (message) {
691 message += 2;
692 message_size = strlen(message);
693 }
694 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
695 if (!tagger) {
696 if (fake_missing_tagger)
697 tagger = "tagger Unspecified Tagger "
698 "<unspecified-tagger> 0 +0000";
699 else
700 tagger = "";
701 tagger_end = tagger + strlen(tagger);
702 } else {
703 tagger++;
704 tagger_end = strchrnul(tagger, '\n');
705 if (anonymize)
706 anonymize_ident_line(&tagger, &tagger_end);
707 }
708
709 if (anonymize) {
710 name = anonymize_refname(name);
711 if (message) {
712 static struct hashmap tags;
713 message = anonymize_mem(&tags, anonymize_tag,
714 message, &message_size);
715 }
716 }
717
718 /* handle signed tags */
719 if (message) {
720 const char *signature = strstr(message,
721 "\n-----BEGIN PGP SIGNATURE-----\n");
722 if (signature)
723 switch(signed_tag_mode) {
724 case ABORT:
725 die ("Encountered signed tag %s; use "
726 "--signed-tags=<mode> to handle it.",
727 oid_to_hex(&tag->object.oid));
728 case WARN:
729 warning ("Exporting signed tag %s",
730 oid_to_hex(&tag->object.oid));
731 /* fallthru */
732 case VERBATIM:
733 break;
734 case WARN_STRIP:
735 warning ("Stripping signature from tag %s",
736 oid_to_hex(&tag->object.oid));
737 /* fallthru */
738 case STRIP:
739 message_size = signature + 1 - message;
740 break;
741 }
742 }
743
744 /* handle tag->tagged having been filtered out due to paths specified */
745 tagged = tag->tagged;
746 tagged_mark = get_object_mark(tagged);
747 if (!tagged_mark) {
748 switch(tag_of_filtered_mode) {
749 case ABORT:
750 die ("Tag %s tags unexported object; use "
751 "--tag-of-filtered-object=<mode> to handle it.",
752 oid_to_hex(&tag->object.oid));
753 case DROP:
754 /* Ignore this tag altogether */
755 free(buf);
756 return;
757 case REWRITE:
758 if (tagged->type != OBJ_COMMIT) {
759 die ("Tag %s tags unexported %s!",
760 oid_to_hex(&tag->object.oid),
761 type_name(tagged->type));
762 }
763 p = (struct commit *)tagged;
764 for (;;) {
765 if (p->parents && p->parents->next)
766 break;
767 if (p->object.flags & UNINTERESTING)
768 break;
769 if (!(p->object.flags & TREESAME))
770 break;
771 if (!p->parents)
772 die ("Can't find replacement commit for tag %s\n",
773 oid_to_hex(&tag->object.oid));
774 p = p->parents->item;
775 }
776 tagged_mark = get_object_mark(&p->object);
777 }
778 }
779
780 if (starts_with(name, "refs/tags/"))
781 name += 10;
782 printf("tag %s\nfrom :%d\n%.*s%sdata %d\n%.*s\n",
783 name, tagged_mark,
784 (int)(tagger_end - tagger), tagger,
785 tagger == tagger_end ? "" : "\n",
786 (int)message_size, (int)message_size, message ? message : "");
787 free(buf);
788}
789
790static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
791{
792 switch (e->item->type) {
793 case OBJ_COMMIT:
794 return (struct commit *)e->item;
795 case OBJ_TAG: {
796 struct tag *tag = (struct tag *)e->item;
797
798 /* handle nested tags */
799 while (tag && tag->object.type == OBJ_TAG) {
800 parse_object(&tag->object.oid);
801 string_list_append(&extra_refs, full_name)->util = tag;
802 tag = (struct tag *)tag->tagged;
803 }
804 if (!tag)
805 die("Tag %s points nowhere?", e->name);
806 return (struct commit *)tag;
807 break;
808 }
809 default:
810 return NULL;
811 }
812}
813
814static void get_tags_and_duplicates(struct rev_cmdline_info *info)
815{
816 int i;
817
818 for (i = 0; i < info->nr; i++) {
819 struct rev_cmdline_entry *e = info->rev + i;
820 struct object_id oid;
821 struct commit *commit;
822 char *full_name;
823
824 if (e->flags & UNINTERESTING)
825 continue;
826
827 if (dwim_ref(e->name, strlen(e->name), &oid, &full_name) != 1)
828 continue;
829
830 if (refspecs) {
831 char *private;
832 private = apply_refspecs(refspecs, refspecs_nr, full_name);
833 if (private) {
834 free(full_name);
835 full_name = private;
836 }
837 }
838
839 commit = get_commit(e, full_name);
840 if (!commit) {
841 warning("%s: Unexpected object of type %s, skipping.",
842 e->name,
843 type_name(e->item->type));
844 continue;
845 }
846
847 switch(commit->object.type) {
848 case OBJ_COMMIT:
849 break;
850 case OBJ_BLOB:
851 export_blob(&commit->object.oid);
852 continue;
853 default: /* OBJ_TAG (nested tags) is already handled */
854 warning("Tag points to object of unexpected type %s, skipping.",
855 type_name(commit->object.type));
856 continue;
857 }
858
859 /*
860 * This ref will not be updated through a commit, lets make
861 * sure it gets properly updated eventually.
862 */
863 if (commit->util || commit->object.flags & SHOWN)
864 string_list_append(&extra_refs, full_name)->util = commit;
865 if (!commit->util)
866 commit->util = full_name;
867 }
868}
869
870static void handle_tags_and_duplicates(void)
871{
872 struct commit *commit;
873 int i;
874
875 for (i = extra_refs.nr - 1; i >= 0; i--) {
876 const char *name = extra_refs.items[i].string;
877 struct object *object = extra_refs.items[i].util;
878 switch (object->type) {
879 case OBJ_TAG:
880 handle_tag(name, (struct tag *)object);
881 break;
882 case OBJ_COMMIT:
883 if (anonymize)
884 name = anonymize_refname(name);
885 /* create refs pointing to already seen commits */
886 commit = (struct commit *)object;
887 printf("reset %s\nfrom :%d\n\n", name,
888 get_object_mark(&commit->object));
889 show_progress();
890 break;
891 }
892 }
893}
894
895static void export_marks(char *file)
896{
897 unsigned int i;
898 uint32_t mark;
899 struct decoration_entry *deco = idnums.entries;
900 FILE *f;
901 int e = 0;
902
903 f = fopen_for_writing(file);
904 if (!f)
905 die_errno("Unable to open marks file %s for writing.", file);
906
907 for (i = 0; i < idnums.size; i++) {
908 if (deco->base && deco->base->type == 1) {
909 mark = ptr_to_mark(deco->decoration);
910 if (fprintf(f, ":%"PRIu32" %s\n", mark,
911 oid_to_hex(&deco->base->oid)) < 0) {
912 e = 1;
913 break;
914 }
915 }
916 deco++;
917 }
918
919 e |= ferror(f);
920 e |= fclose(f);
921 if (e)
922 error("Unable to write marks file %s.", file);
923}
924
925static void import_marks(char *input_file)
926{
927 char line[512];
928 FILE *f = xfopen(input_file, "r");
929
930 while (fgets(line, sizeof(line), f)) {
931 uint32_t mark;
932 char *line_end, *mark_end;
933 struct object_id oid;
934 struct object *object;
935 struct commit *commit;
936 enum object_type type;
937
938 line_end = strchr(line, '\n');
939 if (line[0] != ':' || !line_end)
940 die("corrupt mark line: %s", line);
941 *line_end = '\0';
942
943 mark = strtoumax(line + 1, &mark_end, 10);
944 if (!mark || mark_end == line + 1
945 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
946 die("corrupt mark line: %s", line);
947
948 if (last_idnum < mark)
949 last_idnum = mark;
950
951 type = oid_object_info(the_repository, &oid, NULL);
952 if (type < 0)
953 die("object not found: %s", oid_to_hex(&oid));
954
955 if (type != OBJ_COMMIT)
956 /* only commits */
957 continue;
958
959 commit = lookup_commit(&oid);
960 if (!commit)
961 die("not a commit? can't happen: %s", oid_to_hex(&oid));
962
963 object = &commit->object;
964
965 if (object->flags & SHOWN)
966 error("Object %s already has a mark", oid_to_hex(&oid));
967
968 mark_object(object, mark);
969
970 object->flags |= SHOWN;
971 }
972 fclose(f);
973}
974
975static void handle_deletes(void)
976{
977 int i;
978 for (i = 0; i < refspecs_nr; i++) {
979 struct refspec *refspec = &refspecs[i];
980 if (*refspec->src)
981 continue;
982
983 printf("reset %s\nfrom %s\n\n",
984 refspec->dst, sha1_to_hex(null_sha1));
985 }
986}
987
988int cmd_fast_export(int argc, const char **argv, const char *prefix)
989{
990 struct rev_info revs;
991 struct object_array commits = OBJECT_ARRAY_INIT;
992 struct commit *commit;
993 char *export_filename = NULL, *import_filename = NULL;
994 uint32_t lastimportid;
995 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
996 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
997 struct option options[] = {
998 OPT_INTEGER(0, "progress", &progress,
999 N_("show progress after <n> objects")),
1000 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1001 N_("select handling of signed tags"),
1002 parse_opt_signed_tag_mode),
1003 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1004 N_("select handling of tags that tag filtered objects"),
1005 parse_opt_tag_of_filtered_mode),
1006 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1007 N_("Dump marks to this file")),
1008 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1009 N_("Import marks from this file")),
1010 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1011 N_("Fake a tagger when tags lack one")),
1012 OPT_BOOL(0, "full-tree", &full_tree,
1013 N_("Output full tree for each commit")),
1014 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1015 N_("Use the done feature to terminate the stream")),
1016 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
1017 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1018 N_("Apply refspec to exported refs")),
1019 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1020 OPT_END()
1021 };
1022
1023 if (argc == 1)
1024 usage_with_options (fast_export_usage, options);
1025
1026 /* we handle encodings */
1027 git_config(git_default_config, NULL);
1028
1029 init_revisions(&revs, prefix);
1030 revs.topo_order = 1;
1031 revs.show_source = 1;
1032 revs.rewrite_parents = 1;
1033 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1034 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1035 argc = setup_revisions(argc, argv, &revs, NULL);
1036 if (argc > 1)
1037 usage_with_options (fast_export_usage, options);
1038
1039 if (refspecs_list.nr) {
1040 const char **refspecs_str;
1041 int i;
1042
1043 ALLOC_ARRAY(refspecs_str, refspecs_list.nr);
1044 for (i = 0; i < refspecs_list.nr; i++)
1045 refspecs_str[i] = refspecs_list.items[i].string;
1046
1047 refspecs_nr = refspecs_list.nr;
1048 refspecs = parse_fetch_refspec(refspecs_nr, refspecs_str);
1049
1050 string_list_clear(&refspecs_list, 1);
1051 free(refspecs_str);
1052 }
1053
1054 if (use_done_feature)
1055 printf("feature done\n");
1056
1057 if (import_filename)
1058 import_marks(import_filename);
1059 lastimportid = last_idnum;
1060
1061 if (import_filename && revs.prune_data.nr)
1062 full_tree = 1;
1063
1064 get_tags_and_duplicates(&revs.cmdline);
1065
1066 if (prepare_revision_walk(&revs))
1067 die("revision walk setup failed");
1068 revs.diffopt.format_callback = show_filemodify;
1069 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1070 revs.diffopt.flags.recursive = 1;
1071 while ((commit = get_revision(&revs))) {
1072 if (has_unshown_parent(commit)) {
1073 add_object_array(&commit->object, NULL, &commits);
1074 }
1075 else {
1076 handle_commit(commit, &revs, &paths_of_changed_objects);
1077 handle_tail(&commits, &revs, &paths_of_changed_objects);
1078 }
1079 }
1080
1081 handle_tags_and_duplicates();
1082 handle_deletes();
1083
1084 if (export_filename && lastimportid != last_idnum)
1085 export_marks(export_filename);
1086
1087 if (use_done_feature)
1088 printf("done\n");
1089
1090 free_refspec(refspecs_nr, refspecs);
1091
1092 return 0;
1093}