]> git.ipfire.org Git - thirdparty/git.git/blob - builtin/fast-export.c
Merge branch 'en/merge-recursive-directory-rename-fixes'
[thirdparty/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 "refspec.h"
11 #include "object-store.h"
12 #include "commit.h"
13 #include "object.h"
14 #include "tag.h"
15 #include "diff.h"
16 #include "diffcore.h"
17 #include "log-tree.h"
18 #include "revision.h"
19 #include "decorate.h"
20 #include "string-list.h"
21 #include "utf8.h"
22 #include "parse-options.h"
23 #include "quote.h"
24 #include "remote.h"
25 #include "blob.h"
26 #include "commit-slab.h"
27
28 static const char *fast_export_usage[] = {
29 N_("git fast-export [rev-list-opts]"),
30 NULL
31 };
32
33 static int progress;
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;
39 static int no_data;
40 static int full_tree;
41 static int reference_excluded_commits;
42 static int show_original_ids;
43 static int mark_tags;
44 static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
45 static struct string_list tag_refs = STRING_LIST_INIT_NODUP;
46 static struct refspec refspecs = REFSPEC_INIT_FETCH;
47 static int anonymize;
48 static struct revision_sources revision_sources;
49
50 static int parse_opt_signed_tag_mode(const struct option *opt,
51 const char *arg, int unset)
52 {
53 if (unset || !strcmp(arg, "abort"))
54 signed_tag_mode = SIGNED_TAG_ABORT;
55 else if (!strcmp(arg, "verbatim") || !strcmp(arg, "ignore"))
56 signed_tag_mode = VERBATIM;
57 else if (!strcmp(arg, "warn"))
58 signed_tag_mode = WARN;
59 else if (!strcmp(arg, "warn-strip"))
60 signed_tag_mode = WARN_STRIP;
61 else if (!strcmp(arg, "strip"))
62 signed_tag_mode = STRIP;
63 else
64 return error("Unknown signed-tags mode: %s", arg);
65 return 0;
66 }
67
68 static int parse_opt_tag_of_filtered_mode(const struct option *opt,
69 const char *arg, int unset)
70 {
71 if (unset || !strcmp(arg, "abort"))
72 tag_of_filtered_mode = TAG_FILTERING_ABORT;
73 else if (!strcmp(arg, "drop"))
74 tag_of_filtered_mode = DROP;
75 else if (!strcmp(arg, "rewrite"))
76 tag_of_filtered_mode = REWRITE;
77 else
78 return error("Unknown tag-of-filtered mode: %s", arg);
79 return 0;
80 }
81
82 static int parse_opt_reencode_mode(const struct option *opt,
83 const char *arg, int unset)
84 {
85 if (unset) {
86 reencode_mode = REENCODE_ABORT;
87 return 0;
88 }
89
90 switch (git_parse_maybe_bool(arg)) {
91 case 0:
92 reencode_mode = REENCODE_NO;
93 break;
94 case 1:
95 reencode_mode = REENCODE_YES;
96 break;
97 default:
98 if (!strcasecmp(arg, "abort"))
99 reencode_mode = REENCODE_ABORT;
100 else
101 return error("Unknown reencoding mode: %s", arg);
102 }
103
104 return 0;
105 }
106
107 static struct decoration idnums;
108 static uint32_t last_idnum;
109
110 static int has_unshown_parent(struct commit *commit)
111 {
112 struct commit_list *parent;
113
114 for (parent = commit->parents; parent; parent = parent->next)
115 if (!(parent->item->object.flags & SHOWN) &&
116 !(parent->item->object.flags & UNINTERESTING))
117 return 1;
118 return 0;
119 }
120
121 struct anonymized_entry {
122 struct hashmap_entry hash;
123 const char *orig;
124 size_t orig_len;
125 const char *anon;
126 size_t anon_len;
127 };
128
129 static int anonymized_entry_cmp(const void *unused_cmp_data,
130 const struct hashmap_entry *eptr,
131 const struct hashmap_entry *entry_or_key,
132 const void *unused_keydata)
133 {
134 const struct anonymized_entry *a, *b;
135
136 a = container_of(eptr, const struct anonymized_entry, hash);
137 b = container_of(entry_or_key, const struct anonymized_entry, hash);
138
139 return a->orig_len != b->orig_len ||
140 memcmp(a->orig, b->orig, a->orig_len);
141 }
142
143 /*
144 * Basically keep a cache of X->Y so that we can repeatedly replace
145 * the same anonymized string with another. The actual generation
146 * is farmed out to the generate function.
147 */
148 static const void *anonymize_mem(struct hashmap *map,
149 void *(*generate)(const void *, size_t *),
150 const void *orig, size_t *len)
151 {
152 struct anonymized_entry key, *ret;
153
154 if (!map->cmpfn)
155 hashmap_init(map, anonymized_entry_cmp, NULL, 0);
156
157 hashmap_entry_init(&key.hash, memhash(orig, *len));
158 key.orig = orig;
159 key.orig_len = *len;
160 ret = hashmap_get_entry(map, &key, hash, NULL);
161
162 if (!ret) {
163 ret = xmalloc(sizeof(*ret));
164 hashmap_entry_init(&ret->hash, key.hash.hash);
165 ret->orig = xstrdup(orig);
166 ret->orig_len = *len;
167 ret->anon = generate(orig, len);
168 ret->anon_len = *len;
169 hashmap_put(map, &ret->hash);
170 }
171
172 *len = ret->anon_len;
173 return ret->anon;
174 }
175
176 /*
177 * We anonymize each component of a path individually,
178 * so that paths a/b and a/c will share a common root.
179 * The paths are cached via anonymize_mem so that repeated
180 * lookups for "a" will yield the same value.
181 */
182 static void anonymize_path(struct strbuf *out, const char *path,
183 struct hashmap *map,
184 void *(*generate)(const void *, size_t *))
185 {
186 while (*path) {
187 const char *end_of_component = strchrnul(path, '/');
188 size_t len = end_of_component - path;
189 const char *c = anonymize_mem(map, generate, path, &len);
190 strbuf_add(out, c, len);
191 path = end_of_component;
192 if (*path)
193 strbuf_addch(out, *path++);
194 }
195 }
196
197 static inline void *mark_to_ptr(uint32_t mark)
198 {
199 return (void *)(uintptr_t)mark;
200 }
201
202 static inline uint32_t ptr_to_mark(void * mark)
203 {
204 return (uint32_t)(uintptr_t)mark;
205 }
206
207 static inline void mark_object(struct object *object, uint32_t mark)
208 {
209 add_decoration(&idnums, object, mark_to_ptr(mark));
210 }
211
212 static inline void mark_next_object(struct object *object)
213 {
214 mark_object(object, ++last_idnum);
215 }
216
217 static int get_object_mark(struct object *object)
218 {
219 void *decoration = lookup_decoration(&idnums, object);
220 if (!decoration)
221 return 0;
222 return ptr_to_mark(decoration);
223 }
224
225 static struct commit *rewrite_commit(struct commit *p)
226 {
227 for (;;) {
228 if (p->parents && p->parents->next)
229 break;
230 if (p->object.flags & UNINTERESTING)
231 break;
232 if (!(p->object.flags & TREESAME))
233 break;
234 if (!p->parents)
235 return NULL;
236 p = p->parents->item;
237 }
238 return p;
239 }
240
241 static void show_progress(void)
242 {
243 static int counter = 0;
244 if (!progress)
245 return;
246 if ((++counter % progress) == 0)
247 printf("progress %d objects\n", counter);
248 }
249
250 /*
251 * Ideally we would want some transformation of the blob data here
252 * that is unreversible, but would still be the same size and have
253 * the same data relationship to other blobs (so that we get the same
254 * delta and packing behavior as the original). But the first and last
255 * requirements there are probably mutually exclusive, so let's take
256 * the easy way out for now, and just generate arbitrary content.
257 *
258 * There's no need to cache this result with anonymize_mem, since
259 * we already handle blob content caching with marks.
260 */
261 static char *anonymize_blob(unsigned long *size)
262 {
263 static int counter;
264 struct strbuf out = STRBUF_INIT;
265 strbuf_addf(&out, "anonymous blob %d", counter++);
266 *size = out.len;
267 return strbuf_detach(&out, NULL);
268 }
269
270 static void export_blob(const struct object_id *oid)
271 {
272 unsigned long size;
273 enum object_type type;
274 char *buf;
275 struct object *object;
276 int eaten;
277
278 if (no_data)
279 return;
280
281 if (is_null_oid(oid))
282 return;
283
284 object = lookup_object(the_repository, oid);
285 if (object && object->flags & SHOWN)
286 return;
287
288 if (anonymize) {
289 buf = anonymize_blob(&size);
290 object = (struct object *)lookup_blob(the_repository, oid);
291 eaten = 0;
292 } else {
293 buf = read_object_file(oid, &type, &size);
294 if (!buf)
295 die("could not read blob %s", oid_to_hex(oid));
296 if (check_object_signature(oid, buf, size, type_name(type)) < 0)
297 die("oid mismatch in blob %s", oid_to_hex(oid));
298 object = parse_object_buffer(the_repository, oid, type,
299 size, buf, &eaten);
300 }
301
302 if (!object)
303 die("Could not read blob %s", oid_to_hex(oid));
304
305 mark_next_object(object);
306
307 printf("blob\nmark :%"PRIu32"\n", last_idnum);
308 if (show_original_ids)
309 printf("original-oid %s\n", oid_to_hex(oid));
310 printf("data %"PRIuMAX"\n", (uintmax_t)size);
311 if (size && fwrite(buf, size, 1, stdout) != 1)
312 die_errno("could not write blob '%s'", oid_to_hex(oid));
313 printf("\n");
314
315 show_progress();
316
317 object->flags |= SHOWN;
318 if (!eaten)
319 free(buf);
320 }
321
322 static int depth_first(const void *a_, const void *b_)
323 {
324 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
325 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
326 const char *name_a, *name_b;
327 int len_a, len_b, len;
328 int cmp;
329
330 name_a = a->one ? a->one->path : a->two->path;
331 name_b = b->one ? b->one->path : b->two->path;
332
333 len_a = strlen(name_a);
334 len_b = strlen(name_b);
335 len = (len_a < len_b) ? len_a : len_b;
336
337 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
338 cmp = memcmp(name_a, name_b, len);
339 if (cmp)
340 return cmp;
341 cmp = len_b - len_a;
342 if (cmp)
343 return cmp;
344 /*
345 * Move 'R'ename entries last so that all references of the file
346 * appear in the output before it is renamed (e.g., when a file
347 * was copied and renamed in the same commit).
348 */
349 return (a->status == 'R') - (b->status == 'R');
350 }
351
352 static void print_path_1(const char *path)
353 {
354 int need_quote = quote_c_style(path, NULL, NULL, 0);
355 if (need_quote)
356 quote_c_style(path, NULL, stdout, 0);
357 else if (strchr(path, ' '))
358 printf("\"%s\"", path);
359 else
360 printf("%s", path);
361 }
362
363 static void *anonymize_path_component(const void *path, size_t *len)
364 {
365 static int counter;
366 struct strbuf out = STRBUF_INIT;
367 strbuf_addf(&out, "path%d", counter++);
368 return strbuf_detach(&out, len);
369 }
370
371 static void print_path(const char *path)
372 {
373 if (!anonymize)
374 print_path_1(path);
375 else {
376 static struct hashmap paths;
377 static struct strbuf anon = STRBUF_INIT;
378
379 anonymize_path(&anon, path, &paths, anonymize_path_component);
380 print_path_1(anon.buf);
381 strbuf_reset(&anon);
382 }
383 }
384
385 static void *generate_fake_oid(const void *old, size_t *len)
386 {
387 static uint32_t counter = 1; /* avoid null oid */
388 const unsigned hashsz = the_hash_algo->rawsz;
389 unsigned char *out = xcalloc(hashsz, 1);
390 put_be32(out + hashsz - 4, counter++);
391 return out;
392 }
393
394 static const struct object_id *anonymize_oid(const struct object_id *oid)
395 {
396 static struct hashmap objs;
397 size_t len = the_hash_algo->rawsz;
398 return anonymize_mem(&objs, generate_fake_oid, oid, &len);
399 }
400
401 static void show_filemodify(struct diff_queue_struct *q,
402 struct diff_options *options, void *data)
403 {
404 int i;
405 struct string_list *changed = data;
406
407 /*
408 * Handle files below a directory first, in case they are all deleted
409 * and the directory changes to a file or symlink.
410 */
411 QSORT(q->queue, q->nr, depth_first);
412
413 for (i = 0; i < q->nr; i++) {
414 struct diff_filespec *ospec = q->queue[i]->one;
415 struct diff_filespec *spec = q->queue[i]->two;
416
417 switch (q->queue[i]->status) {
418 case DIFF_STATUS_DELETED:
419 printf("D ");
420 print_path(spec->path);
421 string_list_insert(changed, spec->path);
422 putchar('\n');
423 break;
424
425 case DIFF_STATUS_COPIED:
426 case DIFF_STATUS_RENAMED:
427 /*
428 * If a change in the file corresponding to ospec->path
429 * has been observed, we cannot trust its contents
430 * because the diff is calculated based on the prior
431 * contents, not the current contents. So, declare a
432 * copy or rename only if there was no change observed.
433 */
434 if (!string_list_has_string(changed, ospec->path)) {
435 printf("%c ", q->queue[i]->status);
436 print_path(ospec->path);
437 putchar(' ');
438 print_path(spec->path);
439 string_list_insert(changed, spec->path);
440 putchar('\n');
441
442 if (oideq(&ospec->oid, &spec->oid) &&
443 ospec->mode == spec->mode)
444 break;
445 }
446 /* fallthrough */
447
448 case DIFF_STATUS_TYPE_CHANGED:
449 case DIFF_STATUS_MODIFIED:
450 case DIFF_STATUS_ADDED:
451 /*
452 * Links refer to objects in another repositories;
453 * output the SHA-1 verbatim.
454 */
455 if (no_data || S_ISGITLINK(spec->mode))
456 printf("M %06o %s ", spec->mode,
457 oid_to_hex(anonymize ?
458 anonymize_oid(&spec->oid) :
459 &spec->oid));
460 else {
461 struct object *object = lookup_object(the_repository,
462 &spec->oid);
463 printf("M %06o :%d ", spec->mode,
464 get_object_mark(object));
465 }
466 print_path(spec->path);
467 string_list_insert(changed, spec->path);
468 putchar('\n');
469 break;
470
471 default:
472 die("Unexpected comparison status '%c' for %s, %s",
473 q->queue[i]->status,
474 ospec->path ? ospec->path : "none",
475 spec->path ? spec->path : "none");
476 }
477 }
478 }
479
480 static const char *find_encoding(const char *begin, const char *end)
481 {
482 const char *needle = "\nencoding ";
483 char *bol, *eol;
484
485 bol = memmem(begin, end ? end - begin : strlen(begin),
486 needle, strlen(needle));
487 if (!bol)
488 return NULL;
489 bol += strlen(needle);
490 eol = strchrnul(bol, '\n');
491 *eol = '\0';
492 return bol;
493 }
494
495 static void *anonymize_ref_component(const void *old, size_t *len)
496 {
497 static int counter;
498 struct strbuf out = STRBUF_INIT;
499 strbuf_addf(&out, "ref%d", counter++);
500 return strbuf_detach(&out, len);
501 }
502
503 static const char *anonymize_refname(const char *refname)
504 {
505 /*
506 * If any of these prefixes is found, we will leave it intact
507 * so that tags remain tags and so forth.
508 */
509 static const char *prefixes[] = {
510 "refs/heads/",
511 "refs/tags/",
512 "refs/remotes/",
513 "refs/"
514 };
515 static struct hashmap refs;
516 static struct strbuf anon = STRBUF_INIT;
517 int i;
518
519 /*
520 * We also leave "master" as a special case, since it does not reveal
521 * anything interesting.
522 */
523 if (!strcmp(refname, "refs/heads/master"))
524 return refname;
525
526 strbuf_reset(&anon);
527 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
528 if (skip_prefix(refname, prefixes[i], &refname)) {
529 strbuf_addstr(&anon, prefixes[i]);
530 break;
531 }
532 }
533
534 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
535 return anon.buf;
536 }
537
538 /*
539 * We do not even bother to cache commit messages, as they are unlikely
540 * to be repeated verbatim, and it is not that interesting when they are.
541 */
542 static char *anonymize_commit_message(const char *old)
543 {
544 static int counter;
545 return xstrfmt("subject %d\n\nbody\n", counter++);
546 }
547
548 static struct hashmap idents;
549 static void *anonymize_ident(const void *old, size_t *len)
550 {
551 static int counter;
552 struct strbuf out = STRBUF_INIT;
553 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
554 counter++;
555 return strbuf_detach(&out, len);
556 }
557
558 /*
559 * Our strategy here is to anonymize the names and email addresses,
560 * but keep timestamps intact, as they influence things like traversal
561 * order (and by themselves should not be too revealing).
562 */
563 static void anonymize_ident_line(const char **beg, const char **end)
564 {
565 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
566 static unsigned which_buffer;
567
568 struct strbuf *out;
569 struct ident_split split;
570 const char *end_of_header;
571
572 out = &buffers[which_buffer++];
573 which_buffer %= ARRAY_SIZE(buffers);
574 strbuf_reset(out);
575
576 /* skip "committer", "author", "tagger", etc */
577 end_of_header = strchr(*beg, ' ');
578 if (!end_of_header)
579 BUG("malformed line fed to anonymize_ident_line: %.*s",
580 (int)(*end - *beg), *beg);
581 end_of_header++;
582 strbuf_add(out, *beg, end_of_header - *beg);
583
584 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
585 split.date_begin) {
586 const char *ident;
587 size_t len;
588
589 len = split.mail_end - split.name_begin;
590 ident = anonymize_mem(&idents, anonymize_ident,
591 split.name_begin, &len);
592 strbuf_add(out, ident, len);
593 strbuf_addch(out, ' ');
594 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
595 } else {
596 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
597 }
598
599 *beg = out->buf;
600 *end = out->buf + out->len;
601 }
602
603 static void handle_commit(struct commit *commit, struct rev_info *rev,
604 struct string_list *paths_of_changed_objects)
605 {
606 int saved_output_format = rev->diffopt.output_format;
607 const char *commit_buffer;
608 const char *author, *author_end, *committer, *committer_end;
609 const char *encoding, *message;
610 char *reencoded = NULL;
611 struct commit_list *p;
612 const char *refname;
613 int i;
614
615 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
616
617 parse_commit_or_die(commit);
618 commit_buffer = get_commit_buffer(commit, NULL);
619 author = strstr(commit_buffer, "\nauthor ");
620 if (!author)
621 die("could not find author in commit %s",
622 oid_to_hex(&commit->object.oid));
623 author++;
624 author_end = strchrnul(author, '\n');
625 committer = strstr(author_end, "\ncommitter ");
626 if (!committer)
627 die("could not find committer in commit %s",
628 oid_to_hex(&commit->object.oid));
629 committer++;
630 committer_end = strchrnul(committer, '\n');
631 message = strstr(committer_end, "\n\n");
632 encoding = find_encoding(committer_end, message);
633 if (message)
634 message += 2;
635
636 if (commit->parents &&
637 (get_object_mark(&commit->parents->item->object) != 0 ||
638 reference_excluded_commits) &&
639 !full_tree) {
640 parse_commit_or_die(commit->parents->item);
641 diff_tree_oid(get_commit_tree_oid(commit->parents->item),
642 get_commit_tree_oid(commit), "", &rev->diffopt);
643 }
644 else
645 diff_root_tree_oid(get_commit_tree_oid(commit),
646 "", &rev->diffopt);
647
648 /* Export the referenced blobs, and remember the marks. */
649 for (i = 0; i < diff_queued_diff.nr; i++)
650 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
651 export_blob(&diff_queued_diff.queue[i]->two->oid);
652
653 refname = *revision_sources_at(&revision_sources, commit);
654 /*
655 * FIXME: string_list_remove() below for each ref is overall
656 * O(N^2). Compared to a history walk and diffing trees, this is
657 * just lost in the noise in practice. However, theoretically a
658 * repo may have enough refs for this to become slow.
659 */
660 string_list_remove(&extra_refs, refname, 0);
661 if (anonymize) {
662 refname = anonymize_refname(refname);
663 anonymize_ident_line(&committer, &committer_end);
664 anonymize_ident_line(&author, &author_end);
665 }
666
667 mark_next_object(&commit->object);
668 if (anonymize) {
669 reencoded = anonymize_commit_message(message);
670 } else if (encoding) {
671 switch(reencode_mode) {
672 case REENCODE_YES:
673 reencoded = reencode_string(message, "UTF-8", encoding);
674 break;
675 case REENCODE_NO:
676 break;
677 case REENCODE_ABORT:
678 die("Encountered commit-specific encoding %s in commit "
679 "%s; use --reencode=[yes|no] to handle it",
680 encoding, oid_to_hex(&commit->object.oid));
681 }
682 }
683 if (!commit->parents)
684 printf("reset %s\n", refname);
685 printf("commit %s\nmark :%"PRIu32"\n", refname, last_idnum);
686 if (show_original_ids)
687 printf("original-oid %s\n", oid_to_hex(&commit->object.oid));
688 printf("%.*s\n%.*s\n",
689 (int)(author_end - author), author,
690 (int)(committer_end - committer), committer);
691 if (!reencoded && encoding)
692 printf("encoding %s\n", encoding);
693 printf("data %u\n%s",
694 (unsigned)(reencoded
695 ? strlen(reencoded) : message
696 ? strlen(message) : 0),
697 reencoded ? reencoded : message ? message : "");
698 free(reencoded);
699 unuse_commit_buffer(commit, commit_buffer);
700
701 for (i = 0, p = commit->parents; p; p = p->next) {
702 struct object *obj = &p->item->object;
703 int mark = get_object_mark(obj);
704
705 if (!mark && !reference_excluded_commits)
706 continue;
707 if (i == 0)
708 printf("from ");
709 else
710 printf("merge ");
711 if (mark)
712 printf(":%d\n", mark);
713 else
714 printf("%s\n", oid_to_hex(anonymize ?
715 anonymize_oid(&obj->oid) :
716 &obj->oid));
717 i++;
718 }
719
720 if (full_tree)
721 printf("deleteall\n");
722 log_tree_diff_flush(rev);
723 string_list_clear(paths_of_changed_objects, 0);
724 rev->diffopt.output_format = saved_output_format;
725
726 printf("\n");
727
728 show_progress();
729 }
730
731 static void *anonymize_tag(const void *old, size_t *len)
732 {
733 static int counter;
734 struct strbuf out = STRBUF_INIT;
735 strbuf_addf(&out, "tag message %d", counter++);
736 return strbuf_detach(&out, len);
737 }
738
739 static void handle_tail(struct object_array *commits, struct rev_info *revs,
740 struct string_list *paths_of_changed_objects)
741 {
742 struct commit *commit;
743 while (commits->nr) {
744 commit = (struct commit *)object_array_pop(commits);
745 if (has_unshown_parent(commit)) {
746 /* Queue again, to be handled later */
747 add_object_array(&commit->object, NULL, commits);
748 return;
749 }
750 handle_commit(commit, revs, paths_of_changed_objects);
751 }
752 }
753
754 static void handle_tag(const char *name, struct tag *tag)
755 {
756 unsigned long size;
757 enum object_type type;
758 char *buf;
759 const char *tagger, *tagger_end, *message;
760 size_t message_size = 0;
761 struct object *tagged;
762 int tagged_mark;
763 struct commit *p;
764
765 /* Trees have no identifier in fast-export output, thus we have no way
766 * to output tags of trees, tags of tags of trees, etc. Simply omit
767 * such tags.
768 */
769 tagged = tag->tagged;
770 while (tagged->type == OBJ_TAG) {
771 tagged = ((struct tag *)tagged)->tagged;
772 }
773 if (tagged->type == OBJ_TREE) {
774 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
775 oid_to_hex(&tag->object.oid));
776 return;
777 }
778
779 buf = read_object_file(&tag->object.oid, &type, &size);
780 if (!buf)
781 die("could not read tag %s", oid_to_hex(&tag->object.oid));
782 message = memmem(buf, size, "\n\n", 2);
783 if (message) {
784 message += 2;
785 message_size = strlen(message);
786 }
787 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
788 if (!tagger) {
789 if (fake_missing_tagger)
790 tagger = "tagger Unspecified Tagger "
791 "<unspecified-tagger> 0 +0000";
792 else
793 tagger = "";
794 tagger_end = tagger + strlen(tagger);
795 } else {
796 tagger++;
797 tagger_end = strchrnul(tagger, '\n');
798 if (anonymize)
799 anonymize_ident_line(&tagger, &tagger_end);
800 }
801
802 if (anonymize) {
803 name = anonymize_refname(name);
804 if (message) {
805 static struct hashmap tags;
806 message = anonymize_mem(&tags, anonymize_tag,
807 message, &message_size);
808 }
809 }
810
811 /* handle signed tags */
812 if (message) {
813 const char *signature = strstr(message,
814 "\n-----BEGIN PGP SIGNATURE-----\n");
815 if (signature)
816 switch(signed_tag_mode) {
817 case SIGNED_TAG_ABORT:
818 die("encountered signed tag %s; use "
819 "--signed-tags=<mode> to handle it",
820 oid_to_hex(&tag->object.oid));
821 case WARN:
822 warning("exporting signed tag %s",
823 oid_to_hex(&tag->object.oid));
824 /* fallthru */
825 case VERBATIM:
826 break;
827 case WARN_STRIP:
828 warning("stripping signature from tag %s",
829 oid_to_hex(&tag->object.oid));
830 /* fallthru */
831 case STRIP:
832 message_size = signature + 1 - message;
833 break;
834 }
835 }
836
837 /* handle tag->tagged having been filtered out due to paths specified */
838 tagged = tag->tagged;
839 tagged_mark = get_object_mark(tagged);
840 if (!tagged_mark) {
841 switch(tag_of_filtered_mode) {
842 case TAG_FILTERING_ABORT:
843 die("tag %s tags unexported object; use "
844 "--tag-of-filtered-object=<mode> to handle it",
845 oid_to_hex(&tag->object.oid));
846 case DROP:
847 /* Ignore this tag altogether */
848 free(buf);
849 return;
850 case REWRITE:
851 if (tagged->type == OBJ_TAG && !mark_tags) {
852 die(_("Error: Cannot export nested tags unless --mark-tags is specified."));
853 } else if (tagged->type == OBJ_COMMIT) {
854 p = rewrite_commit((struct commit *)tagged);
855 if (!p) {
856 printf("reset %s\nfrom %s\n\n",
857 name, oid_to_hex(&null_oid));
858 free(buf);
859 return;
860 }
861 tagged_mark = get_object_mark(&p->object);
862 } else {
863 /* tagged->type is either OBJ_BLOB or OBJ_TAG */
864 tagged_mark = get_object_mark(tagged);
865 }
866 }
867 }
868
869 if (tagged->type == OBJ_TAG) {
870 printf("reset %s\nfrom %s\n\n",
871 name, oid_to_hex(&null_oid));
872 }
873 if (starts_with(name, "refs/tags/"))
874 name += 10;
875 printf("tag %s\n", name);
876 if (mark_tags) {
877 mark_next_object(&tag->object);
878 printf("mark :%"PRIu32"\n", last_idnum);
879 }
880 if (tagged_mark)
881 printf("from :%d\n", tagged_mark);
882 else
883 printf("from %s\n", oid_to_hex(&tagged->oid));
884
885 if (show_original_ids)
886 printf("original-oid %s\n", oid_to_hex(&tag->object.oid));
887 printf("%.*s%sdata %d\n%.*s\n",
888 (int)(tagger_end - tagger), tagger,
889 tagger == tagger_end ? "" : "\n",
890 (int)message_size, (int)message_size, message ? message : "");
891 free(buf);
892 }
893
894 static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
895 {
896 switch (e->item->type) {
897 case OBJ_COMMIT:
898 return (struct commit *)e->item;
899 case OBJ_TAG: {
900 struct tag *tag = (struct tag *)e->item;
901
902 /* handle nested tags */
903 while (tag && tag->object.type == OBJ_TAG) {
904 parse_object(the_repository, &tag->object.oid);
905 string_list_append(&tag_refs, full_name)->util = tag;
906 tag = (struct tag *)tag->tagged;
907 }
908 if (!tag)
909 die("Tag %s points nowhere?", e->name);
910 return (struct commit *)tag;
911 break;
912 }
913 default:
914 return NULL;
915 }
916 }
917
918 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
919 {
920 int i;
921
922 for (i = 0; i < info->nr; i++) {
923 struct rev_cmdline_entry *e = info->rev + i;
924 struct object_id oid;
925 struct commit *commit;
926 char *full_name;
927
928 if (e->flags & UNINTERESTING)
929 continue;
930
931 if (dwim_ref(e->name, strlen(e->name), &oid, &full_name) != 1)
932 continue;
933
934 if (refspecs.nr) {
935 char *private;
936 private = apply_refspecs(&refspecs, full_name);
937 if (private) {
938 free(full_name);
939 full_name = private;
940 }
941 }
942
943 commit = get_commit(e, full_name);
944 if (!commit) {
945 warning("%s: Unexpected object of type %s, skipping.",
946 e->name,
947 type_name(e->item->type));
948 continue;
949 }
950
951 switch(commit->object.type) {
952 case OBJ_COMMIT:
953 break;
954 case OBJ_BLOB:
955 export_blob(&commit->object.oid);
956 continue;
957 default: /* OBJ_TAG (nested tags) is already handled */
958 warning("Tag points to object of unexpected type %s, skipping.",
959 type_name(commit->object.type));
960 continue;
961 }
962
963 /*
964 * Make sure this ref gets properly updated eventually, whether
965 * through a commit or manually at the end.
966 */
967 if (e->item->type != OBJ_TAG)
968 string_list_append(&extra_refs, full_name)->util = commit;
969
970 if (!*revision_sources_at(&revision_sources, commit))
971 *revision_sources_at(&revision_sources, commit) = full_name;
972 }
973
974 string_list_sort(&extra_refs);
975 string_list_remove_duplicates(&extra_refs, 0);
976 }
977
978 static void handle_tags_and_duplicates(struct string_list *extras)
979 {
980 struct commit *commit;
981 int i;
982
983 for (i = extras->nr - 1; i >= 0; i--) {
984 const char *name = extras->items[i].string;
985 struct object *object = extras->items[i].util;
986 int mark;
987
988 switch (object->type) {
989 case OBJ_TAG:
990 handle_tag(name, (struct tag *)object);
991 break;
992 case OBJ_COMMIT:
993 if (anonymize)
994 name = anonymize_refname(name);
995 /* create refs pointing to already seen commits */
996 commit = rewrite_commit((struct commit *)object);
997 if (!commit) {
998 /*
999 * Neither this object nor any of its
1000 * ancestors touch any relevant paths, so
1001 * it has been filtered to nothing. Delete
1002 * it.
1003 */
1004 printf("reset %s\nfrom %s\n\n",
1005 name, oid_to_hex(&null_oid));
1006 continue;
1007 }
1008
1009 mark = get_object_mark(&commit->object);
1010 if (!mark) {
1011 /*
1012 * Getting here means we have a commit which
1013 * was excluded by a negative refspec (e.g.
1014 * fast-export ^master master). If we are
1015 * referencing excluded commits, set the ref
1016 * to the exact commit. Otherwise, the user
1017 * wants the branch exported but every commit
1018 * in its history to be deleted, which basically
1019 * just means deletion of the ref.
1020 */
1021 if (!reference_excluded_commits) {
1022 /* delete the ref */
1023 printf("reset %s\nfrom %s\n\n",
1024 name, oid_to_hex(&null_oid));
1025 continue;
1026 }
1027 /* set ref to commit using oid, not mark */
1028 printf("reset %s\nfrom %s\n\n", name,
1029 oid_to_hex(&commit->object.oid));
1030 continue;
1031 }
1032
1033 printf("reset %s\nfrom :%d\n\n", name, mark
1034 );
1035 show_progress();
1036 break;
1037 }
1038 }
1039 }
1040
1041 static void export_marks(char *file)
1042 {
1043 unsigned int i;
1044 uint32_t mark;
1045 struct decoration_entry *deco = idnums.entries;
1046 FILE *f;
1047 int e = 0;
1048
1049 f = fopen_for_writing(file);
1050 if (!f)
1051 die_errno("Unable to open marks file %s for writing.", file);
1052
1053 for (i = 0; i < idnums.size; i++) {
1054 if (deco->base && deco->base->type == 1) {
1055 mark = ptr_to_mark(deco->decoration);
1056 if (fprintf(f, ":%"PRIu32" %s\n", mark,
1057 oid_to_hex(&deco->base->oid)) < 0) {
1058 e = 1;
1059 break;
1060 }
1061 }
1062 deco++;
1063 }
1064
1065 e |= ferror(f);
1066 e |= fclose(f);
1067 if (e)
1068 error("Unable to write marks file %s.", file);
1069 }
1070
1071 static void import_marks(char *input_file, int check_exists)
1072 {
1073 char line[512];
1074 FILE *f;
1075 struct stat sb;
1076
1077 if (check_exists && stat(input_file, &sb))
1078 return;
1079
1080 f = xfopen(input_file, "r");
1081 while (fgets(line, sizeof(line), f)) {
1082 uint32_t mark;
1083 char *line_end, *mark_end;
1084 struct object_id oid;
1085 struct object *object;
1086 struct commit *commit;
1087 enum object_type type;
1088
1089 line_end = strchr(line, '\n');
1090 if (line[0] != ':' || !line_end)
1091 die("corrupt mark line: %s", line);
1092 *line_end = '\0';
1093
1094 mark = strtoumax(line + 1, &mark_end, 10);
1095 if (!mark || mark_end == line + 1
1096 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
1097 die("corrupt mark line: %s", line);
1098
1099 if (last_idnum < mark)
1100 last_idnum = mark;
1101
1102 type = oid_object_info(the_repository, &oid, NULL);
1103 if (type < 0)
1104 die("object not found: %s", oid_to_hex(&oid));
1105
1106 if (type != OBJ_COMMIT)
1107 /* only commits */
1108 continue;
1109
1110 commit = lookup_commit(the_repository, &oid);
1111 if (!commit)
1112 die("not a commit? can't happen: %s", oid_to_hex(&oid));
1113
1114 object = &commit->object;
1115
1116 if (object->flags & SHOWN)
1117 error("Object %s already has a mark", oid_to_hex(&oid));
1118
1119 mark_object(object, mark);
1120
1121 object->flags |= SHOWN;
1122 }
1123 fclose(f);
1124 }
1125
1126 static void handle_deletes(void)
1127 {
1128 int i;
1129 for (i = 0; i < refspecs.nr; i++) {
1130 struct refspec_item *refspec = &refspecs.items[i];
1131 if (*refspec->src)
1132 continue;
1133
1134 printf("reset %s\nfrom %s\n\n",
1135 refspec->dst, oid_to_hex(&null_oid));
1136 }
1137 }
1138
1139 int cmd_fast_export(int argc, const char **argv, const char *prefix)
1140 {
1141 struct rev_info revs;
1142 struct object_array commits = OBJECT_ARRAY_INIT;
1143 struct commit *commit;
1144 char *export_filename = NULL,
1145 *import_filename = NULL,
1146 *import_filename_if_exists = NULL;
1147 uint32_t lastimportid;
1148 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
1149 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
1150 struct option options[] = {
1151 OPT_INTEGER(0, "progress", &progress,
1152 N_("show progress after <n> objects")),
1153 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1154 N_("select handling of signed tags"),
1155 parse_opt_signed_tag_mode),
1156 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1157 N_("select handling of tags that tag filtered objects"),
1158 parse_opt_tag_of_filtered_mode),
1159 OPT_CALLBACK(0, "reencode", &reencode_mode, N_("mode"),
1160 N_("select handling of commit messages in an alternate encoding"),
1161 parse_opt_reencode_mode),
1162 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1163 N_("Dump marks to this file")),
1164 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1165 N_("Import marks from this file")),
1166 OPT_STRING(0, "import-marks-if-exists",
1167 &import_filename_if_exists,
1168 N_("file"),
1169 N_("Import marks from this file if it exists")),
1170 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1171 N_("Fake a tagger when tags lack one")),
1172 OPT_BOOL(0, "full-tree", &full_tree,
1173 N_("Output full tree for each commit")),
1174 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1175 N_("Use the done feature to terminate the stream")),
1176 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
1177 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1178 N_("Apply refspec to exported refs")),
1179 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1180 OPT_BOOL(0, "reference-excluded-parents",
1181 &reference_excluded_commits, N_("Reference parents which are not in fast-export stream by object id")),
1182 OPT_BOOL(0, "show-original-ids", &show_original_ids,
1183 N_("Show original object ids of blobs/commits")),
1184 OPT_BOOL(0, "mark-tags", &mark_tags,
1185 N_("Label tags with mark ids")),
1186
1187 OPT_END()
1188 };
1189
1190 if (argc == 1)
1191 usage_with_options (fast_export_usage, options);
1192
1193 /* we handle encodings */
1194 git_config(git_default_config, NULL);
1195
1196 repo_init_revisions(the_repository, &revs, prefix);
1197 init_revision_sources(&revision_sources);
1198 revs.topo_order = 1;
1199 revs.sources = &revision_sources;
1200 revs.rewrite_parents = 1;
1201 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1202 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1203 argc = setup_revisions(argc, argv, &revs, NULL);
1204 if (argc > 1)
1205 usage_with_options (fast_export_usage, options);
1206
1207 if (refspecs_list.nr) {
1208 int i;
1209
1210 for (i = 0; i < refspecs_list.nr; i++)
1211 refspec_append(&refspecs, refspecs_list.items[i].string);
1212
1213 string_list_clear(&refspecs_list, 1);
1214 }
1215
1216 if (use_done_feature)
1217 printf("feature done\n");
1218
1219 if (import_filename && import_filename_if_exists)
1220 die(_("Cannot pass both --import-marks and --import-marks-if-exists"));
1221 if (import_filename)
1222 import_marks(import_filename, 0);
1223 else if (import_filename_if_exists)
1224 import_marks(import_filename_if_exists, 1);
1225 lastimportid = last_idnum;
1226
1227 if (import_filename && revs.prune_data.nr)
1228 full_tree = 1;
1229
1230 get_tags_and_duplicates(&revs.cmdline);
1231
1232 if (prepare_revision_walk(&revs))
1233 die("revision walk setup failed");
1234 revs.diffopt.format_callback = show_filemodify;
1235 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1236 revs.diffopt.flags.recursive = 1;
1237 while ((commit = get_revision(&revs))) {
1238 if (has_unshown_parent(commit)) {
1239 add_object_array(&commit->object, NULL, &commits);
1240 }
1241 else {
1242 handle_commit(commit, &revs, &paths_of_changed_objects);
1243 handle_tail(&commits, &revs, &paths_of_changed_objects);
1244 }
1245 }
1246
1247 handle_tags_and_duplicates(&extra_refs);
1248 handle_tags_and_duplicates(&tag_refs);
1249 handle_deletes();
1250
1251 if (export_filename && lastimportid != last_idnum)
1252 export_marks(export_filename);
1253
1254 if (use_done_feature)
1255 printf("done\n");
1256
1257 refspec_clear(&refspecs);
1258
1259 return 0;
1260 }