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 "refs.h"
9#include "commit.h"
10#include "object.h"
11#include "tag.h"
12#include "diff.h"
13#include "diffcore.h"
14#include "log-tree.h"
15#include "revision.h"
16#include "decorate.h"
17#include "string-list.h"
18#include "utf8.h"
19#include "parse-options.h"
20#include "quote.h"
21#include "remote.h"
22#include "blob.h"
23
24static const char *fast_export_usage[] = {
25 N_("git fast-export [rev-list-opts]"),
26 NULL
27};
28
29static int progress;
30static enum { ABORT, VERBATIM, WARN, WARN_STRIP, STRIP } signed_tag_mode = ABORT;
31static enum { ERROR, DROP, REWRITE } tag_of_filtered_mode = ERROR;
32static int fake_missing_tagger;
33static int use_done_feature;
34static int no_data;
35static int full_tree;
36static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
37static struct refspec *refspecs;
38static int refspecs_nr;
39static int anonymize;
40
41static int parse_opt_signed_tag_mode(const struct option *opt,
42 const char *arg, int unset)
43{
44 if (unset || !strcmp(arg, "abort"))
45 signed_tag_mode = ABORT;
46 else if (!strcmp(arg, "verbatim") || !strcmp(arg, "ignore"))
47 signed_tag_mode = VERBATIM;
48 else if (!strcmp(arg, "warn"))
49 signed_tag_mode = WARN;
50 else if (!strcmp(arg, "warn-strip"))
51 signed_tag_mode = WARN_STRIP;
52 else if (!strcmp(arg, "strip"))
53 signed_tag_mode = STRIP;
54 else
55 return error("Unknown signed-tags mode: %s", arg);
56 return 0;
57}
58
59static int parse_opt_tag_of_filtered_mode(const struct option *opt,
60 const char *arg, int unset)
61{
62 if (unset || !strcmp(arg, "abort"))
63 tag_of_filtered_mode = ERROR;
64 else if (!strcmp(arg, "drop"))
65 tag_of_filtered_mode = DROP;
66 else if (!strcmp(arg, "rewrite"))
67 tag_of_filtered_mode = REWRITE;
68 else
69 return error("Unknown tag-of-filtered mode: %s", arg);
70 return 0;
71}
72
73static struct decoration idnums;
74static uint32_t last_idnum;
75
76static int has_unshown_parent(struct commit *commit)
77{
78 struct commit_list *parent;
79
80 for (parent = commit->parents; parent; parent = parent->next)
81 if (!(parent->item->object.flags & SHOWN) &&
82 !(parent->item->object.flags & UNINTERESTING))
83 return 1;
84 return 0;
85}
86
87struct anonymized_entry {
88 struct hashmap_entry hash;
89 const char *orig;
90 size_t orig_len;
91 const char *anon;
92 size_t anon_len;
93};
94
95static int anonymized_entry_cmp(const void *va, const void *vb,
96 const void *data)
97{
98 const struct anonymized_entry *a = va, *b = vb;
99 return a->orig_len != b->orig_len ||
100 memcmp(a->orig, b->orig, a->orig_len);
101}
102
103/*
104 * Basically keep a cache of X->Y so that we can repeatedly replace
105 * the same anonymized string with another. The actual generation
106 * is farmed out to the generate function.
107 */
108static const void *anonymize_mem(struct hashmap *map,
109 void *(*generate)(const void *, size_t *),
110 const void *orig, size_t *len)
111{
112 struct anonymized_entry key, *ret;
113
114 if (!map->cmpfn)
115 hashmap_init(map, anonymized_entry_cmp, 0);
116
117 hashmap_entry_init(&key, memhash(orig, *len));
118 key.orig = orig;
119 key.orig_len = *len;
120 ret = hashmap_get(map, &key, NULL);
121
122 if (!ret) {
123 ret = xmalloc(sizeof(*ret));
124 hashmap_entry_init(&ret->hash, key.hash.hash);
125 ret->orig = xstrdup(orig);
126 ret->orig_len = *len;
127 ret->anon = generate(orig, len);
128 ret->anon_len = *len;
129 hashmap_put(map, ret);
130 }
131
132 *len = ret->anon_len;
133 return ret->anon;
134}
135
136/*
137 * We anonymize each component of a path individually,
138 * so that paths a/b and a/c will share a common root.
139 * The paths are cached via anonymize_mem so that repeated
140 * lookups for "a" will yield the same value.
141 */
142static void anonymize_path(struct strbuf *out, const char *path,
143 struct hashmap *map,
144 void *(*generate)(const void *, size_t *))
145{
146 while (*path) {
147 const char *end_of_component = strchrnul(path, '/');
148 size_t len = end_of_component - path;
149 const char *c = anonymize_mem(map, generate, path, &len);
150 strbuf_add(out, c, len);
151 path = end_of_component;
152 if (*path)
153 strbuf_addch(out, *path++);
154 }
155}
156
157/* Since intptr_t is C99, we do not use it here */
158static inline uint32_t *mark_to_ptr(uint32_t mark)
159{
160 return ((uint32_t *)NULL) + mark;
161}
162
163static inline uint32_t ptr_to_mark(void * mark)
164{
165 return (uint32_t *)mark - (uint32_t *)NULL;
166}
167
168static inline void mark_object(struct object *object, uint32_t mark)
169{
170 add_decoration(&idnums, object, mark_to_ptr(mark));
171}
172
173static inline void mark_next_object(struct object *object)
174{
175 mark_object(object, ++last_idnum);
176}
177
178static int get_object_mark(struct object *object)
179{
180 void *decoration = lookup_decoration(&idnums, object);
181 if (!decoration)
182 return 0;
183 return ptr_to_mark(decoration);
184}
185
186static void show_progress(void)
187{
188 static int counter = 0;
189 if (!progress)
190 return;
191 if ((++counter % progress) == 0)
192 printf("progress %d objects\n", counter);
193}
194
195/*
196 * Ideally we would want some transformation of the blob data here
197 * that is unreversible, but would still be the same size and have
198 * the same data relationship to other blobs (so that we get the same
199 * delta and packing behavior as the original). But the first and last
200 * requirements there are probably mutually exclusive, so let's take
201 * the easy way out for now, and just generate arbitrary content.
202 *
203 * There's no need to cache this result with anonymize_mem, since
204 * we already handle blob content caching with marks.
205 */
206static char *anonymize_blob(unsigned long *size)
207{
208 static int counter;
209 struct strbuf out = STRBUF_INIT;
210 strbuf_addf(&out, "anonymous blob %d", counter++);
211 *size = out.len;
212 return strbuf_detach(&out, NULL);
213}
214
215static void export_blob(const struct object_id *oid)
216{
217 unsigned long size;
218 enum object_type type;
219 char *buf;
220 struct object *object;
221 int eaten;
222
223 if (no_data)
224 return;
225
226 if (is_null_oid(oid))
227 return;
228
229 object = lookup_object(oid->hash);
230 if (object && object->flags & SHOWN)
231 return;
232
233 if (anonymize) {
234 buf = anonymize_blob(&size);
235 object = (struct object *)lookup_blob(oid->hash);
236 eaten = 0;
237 } else {
238 buf = read_sha1_file(oid->hash, &type, &size);
239 if (!buf)
240 die ("Could not read blob %s", oid_to_hex(oid));
241 if (check_sha1_signature(oid->hash, buf, size, typename(type)) < 0)
242 die("sha1 mismatch in blob %s", oid_to_hex(oid));
243 object = parse_object_buffer(oid->hash, type, size, buf, &eaten);
244 }
245
246 if (!object)
247 die("Could not read blob %s", oid_to_hex(oid));
248
249 mark_next_object(object);
250
251 printf("blob\nmark :%"PRIu32"\ndata %lu\n", last_idnum, size);
252 if (size && fwrite(buf, size, 1, stdout) != 1)
253 die_errno ("Could not write blob '%s'", oid_to_hex(oid));
254 printf("\n");
255
256 show_progress();
257
258 object->flags |= SHOWN;
259 if (!eaten)
260 free(buf);
261}
262
263static int depth_first(const void *a_, const void *b_)
264{
265 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
266 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
267 const char *name_a, *name_b;
268 int len_a, len_b, len;
269 int cmp;
270
271 name_a = a->one ? a->one->path : a->two->path;
272 name_b = b->one ? b->one->path : b->two->path;
273
274 len_a = strlen(name_a);
275 len_b = strlen(name_b);
276 len = (len_a < len_b) ? len_a : len_b;
277
278 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
279 cmp = memcmp(name_a, name_b, len);
280 if (cmp)
281 return cmp;
282 cmp = len_b - len_a;
283 if (cmp)
284 return cmp;
285 /*
286 * Move 'R'ename entries last so that all references of the file
287 * appear in the output before it is renamed (e.g., when a file
288 * was copied and renamed in the same commit).
289 */
290 return (a->status == 'R') - (b->status == 'R');
291}
292
293static void print_path_1(const char *path)
294{
295 int need_quote = quote_c_style(path, NULL, NULL, 0);
296 if (need_quote)
297 quote_c_style(path, NULL, stdout, 0);
298 else if (strchr(path, ' '))
299 printf("\"%s\"", path);
300 else
301 printf("%s", path);
302}
303
304static void *anonymize_path_component(const void *path, size_t *len)
305{
306 static int counter;
307 struct strbuf out = STRBUF_INIT;
308 strbuf_addf(&out, "path%d", counter++);
309 return strbuf_detach(&out, len);
310}
311
312static void print_path(const char *path)
313{
314 if (!anonymize)
315 print_path_1(path);
316 else {
317 static struct hashmap paths;
318 static struct strbuf anon = STRBUF_INIT;
319
320 anonymize_path(&anon, path, &paths, anonymize_path_component);
321 print_path_1(anon.buf);
322 strbuf_reset(&anon);
323 }
324}
325
326static void *generate_fake_oid(const void *old, size_t *len)
327{
328 static uint32_t counter = 1; /* avoid null sha1 */
329 unsigned char *out = xcalloc(GIT_SHA1_RAWSZ, 1);
330 put_be32(out + GIT_SHA1_RAWSZ - 4, counter++);
331 return out;
332}
333
334static const unsigned char *anonymize_sha1(const struct object_id *oid)
335{
336 static struct hashmap sha1s;
337 size_t len = GIT_SHA1_RAWSZ;
338 return anonymize_mem(&sha1s, generate_fake_oid, oid, &len);
339}
340
341static void show_filemodify(struct diff_queue_struct *q,
342 struct diff_options *options, void *data)
343{
344 int i;
345
346 /*
347 * Handle files below a directory first, in case they are all deleted
348 * and the directory changes to a file or symlink.
349 */
350 QSORT(q->queue, q->nr, depth_first);
351
352 for (i = 0; i < q->nr; i++) {
353 struct diff_filespec *ospec = q->queue[i]->one;
354 struct diff_filespec *spec = q->queue[i]->two;
355
356 switch (q->queue[i]->status) {
357 case DIFF_STATUS_DELETED:
358 printf("D ");
359 print_path(spec->path);
360 putchar('\n');
361 break;
362
363 case DIFF_STATUS_COPIED:
364 case DIFF_STATUS_RENAMED:
365 printf("%c ", q->queue[i]->status);
366 print_path(ospec->path);
367 putchar(' ');
368 print_path(spec->path);
369 putchar('\n');
370
371 if (!oidcmp(&ospec->oid, &spec->oid) &&
372 ospec->mode == spec->mode)
373 break;
374 /* fallthrough */
375
376 case DIFF_STATUS_TYPE_CHANGED:
377 case DIFF_STATUS_MODIFIED:
378 case DIFF_STATUS_ADDED:
379 /*
380 * Links refer to objects in another repositories;
381 * output the SHA-1 verbatim.
382 */
383 if (no_data || S_ISGITLINK(spec->mode))
384 printf("M %06o %s ", spec->mode,
385 sha1_to_hex(anonymize ?
386 anonymize_sha1(&spec->oid) :
387 spec->oid.hash));
388 else {
389 struct object *object = lookup_object(spec->oid.hash);
390 printf("M %06o :%d ", spec->mode,
391 get_object_mark(object));
392 }
393 print_path(spec->path);
394 putchar('\n');
395 break;
396
397 default:
398 die("Unexpected comparison status '%c' for %s, %s",
399 q->queue[i]->status,
400 ospec->path ? ospec->path : "none",
401 spec->path ? spec->path : "none");
402 }
403 }
404}
405
406static const char *find_encoding(const char *begin, const char *end)
407{
408 const char *needle = "\nencoding ";
409 char *bol, *eol;
410
411 bol = memmem(begin, end ? end - begin : strlen(begin),
412 needle, strlen(needle));
413 if (!bol)
414 return git_commit_encoding;
415 bol += strlen(needle);
416 eol = strchrnul(bol, '\n');
417 *eol = '\0';
418 return bol;
419}
420
421static void *anonymize_ref_component(const void *old, size_t *len)
422{
423 static int counter;
424 struct strbuf out = STRBUF_INIT;
425 strbuf_addf(&out, "ref%d", counter++);
426 return strbuf_detach(&out, len);
427}
428
429static const char *anonymize_refname(const char *refname)
430{
431 /*
432 * If any of these prefixes is found, we will leave it intact
433 * so that tags remain tags and so forth.
434 */
435 static const char *prefixes[] = {
436 "refs/heads/",
437 "refs/tags/",
438 "refs/remotes/",
439 "refs/"
440 };
441 static struct hashmap refs;
442 static struct strbuf anon = STRBUF_INIT;
443 int i;
444
445 /*
446 * We also leave "master" as a special case, since it does not reveal
447 * anything interesting.
448 */
449 if (!strcmp(refname, "refs/heads/master"))
450 return refname;
451
452 strbuf_reset(&anon);
453 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
454 if (skip_prefix(refname, prefixes[i], &refname)) {
455 strbuf_addstr(&anon, prefixes[i]);
456 break;
457 }
458 }
459
460 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
461 return anon.buf;
462}
463
464/*
465 * We do not even bother to cache commit messages, as they are unlikely
466 * to be repeated verbatim, and it is not that interesting when they are.
467 */
468static char *anonymize_commit_message(const char *old)
469{
470 static int counter;
471 return xstrfmt("subject %d\n\nbody\n", counter++);
472}
473
474static struct hashmap idents;
475static void *anonymize_ident(const void *old, size_t *len)
476{
477 static int counter;
478 struct strbuf out = STRBUF_INIT;
479 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
480 counter++;
481 return strbuf_detach(&out, len);
482}
483
484/*
485 * Our strategy here is to anonymize the names and email addresses,
486 * but keep timestamps intact, as they influence things like traversal
487 * order (and by themselves should not be too revealing).
488 */
489static void anonymize_ident_line(const char **beg, const char **end)
490{
491 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
492 static unsigned which_buffer;
493
494 struct strbuf *out;
495 struct ident_split split;
496 const char *end_of_header;
497
498 out = &buffers[which_buffer++];
499 which_buffer %= ARRAY_SIZE(buffers);
500 strbuf_reset(out);
501
502 /* skip "committer", "author", "tagger", etc */
503 end_of_header = strchr(*beg, ' ');
504 if (!end_of_header)
505 die("BUG: malformed line fed to anonymize_ident_line: %.*s",
506 (int)(*end - *beg), *beg);
507 end_of_header++;
508 strbuf_add(out, *beg, end_of_header - *beg);
509
510 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
511 split.date_begin) {
512 const char *ident;
513 size_t len;
514
515 len = split.mail_end - split.name_begin;
516 ident = anonymize_mem(&idents, anonymize_ident,
517 split.name_begin, &len);
518 strbuf_add(out, ident, len);
519 strbuf_addch(out, ' ');
520 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
521 } else {
522 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
523 }
524
525 *beg = out->buf;
526 *end = out->buf + out->len;
527}
528
529static void handle_commit(struct commit *commit, struct rev_info *rev)
530{
531 int saved_output_format = rev->diffopt.output_format;
532 const char *commit_buffer;
533 const char *author, *author_end, *committer, *committer_end;
534 const char *encoding, *message;
535 char *reencoded = NULL;
536 struct commit_list *p;
537 const char *refname;
538 int i;
539
540 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
541
542 parse_commit_or_die(commit);
543 commit_buffer = get_commit_buffer(commit, NULL);
544 author = strstr(commit_buffer, "\nauthor ");
545 if (!author)
546 die ("Could not find author in commit %s",
547 oid_to_hex(&commit->object.oid));
548 author++;
549 author_end = strchrnul(author, '\n');
550 committer = strstr(author_end, "\ncommitter ");
551 if (!committer)
552 die ("Could not find committer in commit %s",
553 oid_to_hex(&commit->object.oid));
554 committer++;
555 committer_end = strchrnul(committer, '\n');
556 message = strstr(committer_end, "\n\n");
557 encoding = find_encoding(committer_end, message);
558 if (message)
559 message += 2;
560
561 if (commit->parents &&
562 get_object_mark(&commit->parents->item->object) != 0 &&
563 !full_tree) {
564 parse_commit_or_die(commit->parents->item);
565 diff_tree_sha1(commit->parents->item->tree->object.oid.hash,
566 commit->tree->object.oid.hash, "", &rev->diffopt);
567 }
568 else
569 diff_root_tree_sha1(commit->tree->object.oid.hash,
570 "", &rev->diffopt);
571
572 /* Export the referenced blobs, and remember the marks. */
573 for (i = 0; i < diff_queued_diff.nr; i++)
574 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
575 export_blob(&diff_queued_diff.queue[i]->two->oid);
576
577 refname = commit->util;
578 if (anonymize) {
579 refname = anonymize_refname(refname);
580 anonymize_ident_line(&committer, &committer_end);
581 anonymize_ident_line(&author, &author_end);
582 }
583
584 mark_next_object(&commit->object);
585 if (anonymize)
586 reencoded = anonymize_commit_message(message);
587 else if (!is_encoding_utf8(encoding))
588 reencoded = reencode_string(message, "UTF-8", encoding);
589 if (!commit->parents)
590 printf("reset %s\n", refname);
591 printf("commit %s\nmark :%"PRIu32"\n%.*s\n%.*s\ndata %u\n%s",
592 refname, last_idnum,
593 (int)(author_end - author), author,
594 (int)(committer_end - committer), committer,
595 (unsigned)(reencoded
596 ? strlen(reencoded) : message
597 ? strlen(message) : 0),
598 reencoded ? reencoded : message ? message : "");
599 free(reencoded);
600 unuse_commit_buffer(commit, commit_buffer);
601
602 for (i = 0, p = commit->parents; p; p = p->next) {
603 int mark = get_object_mark(&p->item->object);
604 if (!mark)
605 continue;
606 if (i == 0)
607 printf("from :%d\n", mark);
608 else
609 printf("merge :%d\n", mark);
610 i++;
611 }
612
613 if (full_tree)
614 printf("deleteall\n");
615 log_tree_diff_flush(rev);
616 rev->diffopt.output_format = saved_output_format;
617
618 printf("\n");
619
620 show_progress();
621}
622
623static void *anonymize_tag(const void *old, size_t *len)
624{
625 static int counter;
626 struct strbuf out = STRBUF_INIT;
627 strbuf_addf(&out, "tag message %d", counter++);
628 return strbuf_detach(&out, len);
629}
630
631static void handle_tail(struct object_array *commits, struct rev_info *revs)
632{
633 struct commit *commit;
634 while (commits->nr) {
635 commit = (struct commit *)commits->objects[commits->nr - 1].item;
636 if (has_unshown_parent(commit))
637 return;
638 handle_commit(commit, revs);
639 commits->nr--;
640 }
641}
642
643static void handle_tag(const char *name, struct tag *tag)
644{
645 unsigned long size;
646 enum object_type type;
647 char *buf;
648 const char *tagger, *tagger_end, *message;
649 size_t message_size = 0;
650 struct object *tagged;
651 int tagged_mark;
652 struct commit *p;
653
654 /* Trees have no identifier in fast-export output, thus we have no way
655 * to output tags of trees, tags of tags of trees, etc. Simply omit
656 * such tags.
657 */
658 tagged = tag->tagged;
659 while (tagged->type == OBJ_TAG) {
660 tagged = ((struct tag *)tagged)->tagged;
661 }
662 if (tagged->type == OBJ_TREE) {
663 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
664 oid_to_hex(&tag->object.oid));
665 return;
666 }
667
668 buf = read_sha1_file(tag->object.oid.hash, &type, &size);
669 if (!buf)
670 die ("Could not read tag %s", oid_to_hex(&tag->object.oid));
671 message = memmem(buf, size, "\n\n", 2);
672 if (message) {
673 message += 2;
674 message_size = strlen(message);
675 }
676 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
677 if (!tagger) {
678 if (fake_missing_tagger)
679 tagger = "tagger Unspecified Tagger "
680 "<unspecified-tagger> 0 +0000";
681 else
682 tagger = "";
683 tagger_end = tagger + strlen(tagger);
684 } else {
685 tagger++;
686 tagger_end = strchrnul(tagger, '\n');
687 if (anonymize)
688 anonymize_ident_line(&tagger, &tagger_end);
689 }
690
691 if (anonymize) {
692 name = anonymize_refname(name);
693 if (message) {
694 static struct hashmap tags;
695 message = anonymize_mem(&tags, anonymize_tag,
696 message, &message_size);
697 }
698 }
699
700 /* handle signed tags */
701 if (message) {
702 const char *signature = strstr(message,
703 "\n-----BEGIN PGP SIGNATURE-----\n");
704 if (signature)
705 switch(signed_tag_mode) {
706 case ABORT:
707 die ("Encountered signed tag %s; use "
708 "--signed-tags=<mode> to handle it.",
709 oid_to_hex(&tag->object.oid));
710 case WARN:
711 warning ("Exporting signed tag %s",
712 oid_to_hex(&tag->object.oid));
713 /* fallthru */
714 case VERBATIM:
715 break;
716 case WARN_STRIP:
717 warning ("Stripping signature from tag %s",
718 oid_to_hex(&tag->object.oid));
719 /* fallthru */
720 case STRIP:
721 message_size = signature + 1 - message;
722 break;
723 }
724 }
725
726 /* handle tag->tagged having been filtered out due to paths specified */
727 tagged = tag->tagged;
728 tagged_mark = get_object_mark(tagged);
729 if (!tagged_mark) {
730 switch(tag_of_filtered_mode) {
731 case ABORT:
732 die ("Tag %s tags unexported object; use "
733 "--tag-of-filtered-object=<mode> to handle it.",
734 oid_to_hex(&tag->object.oid));
735 case DROP:
736 /* Ignore this tag altogether */
737 return;
738 case REWRITE:
739 if (tagged->type != OBJ_COMMIT) {
740 die ("Tag %s tags unexported %s!",
741 oid_to_hex(&tag->object.oid),
742 typename(tagged->type));
743 }
744 p = (struct commit *)tagged;
745 for (;;) {
746 if (p->parents && p->parents->next)
747 break;
748 if (p->object.flags & UNINTERESTING)
749 break;
750 if (!(p->object.flags & TREESAME))
751 break;
752 if (!p->parents)
753 die ("Can't find replacement commit for tag %s\n",
754 oid_to_hex(&tag->object.oid));
755 p = p->parents->item;
756 }
757 tagged_mark = get_object_mark(&p->object);
758 }
759 }
760
761 if (starts_with(name, "refs/tags/"))
762 name += 10;
763 printf("tag %s\nfrom :%d\n%.*s%sdata %d\n%.*s\n",
764 name, tagged_mark,
765 (int)(tagger_end - tagger), tagger,
766 tagger == tagger_end ? "" : "\n",
767 (int)message_size, (int)message_size, message ? message : "");
768}
769
770static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
771{
772 switch (e->item->type) {
773 case OBJ_COMMIT:
774 return (struct commit *)e->item;
775 case OBJ_TAG: {
776 struct tag *tag = (struct tag *)e->item;
777
778 /* handle nested tags */
779 while (tag && tag->object.type == OBJ_TAG) {
780 parse_object(tag->object.oid.hash);
781 string_list_append(&extra_refs, full_name)->util = tag;
782 tag = (struct tag *)tag->tagged;
783 }
784 if (!tag)
785 die("Tag %s points nowhere?", e->name);
786 return (struct commit *)tag;
787 break;
788 }
789 default:
790 return NULL;
791 }
792}
793
794static void get_tags_and_duplicates(struct rev_cmdline_info *info)
795{
796 int i;
797
798 for (i = 0; i < info->nr; i++) {
799 struct rev_cmdline_entry *e = info->rev + i;
800 struct object_id oid;
801 struct commit *commit;
802 char *full_name;
803
804 if (e->flags & UNINTERESTING)
805 continue;
806
807 if (dwim_ref(e->name, strlen(e->name), oid.hash, &full_name) != 1)
808 continue;
809
810 if (refspecs) {
811 char *private;
812 private = apply_refspecs(refspecs, refspecs_nr, full_name);
813 if (private) {
814 free(full_name);
815 full_name = private;
816 }
817 }
818
819 commit = get_commit(e, full_name);
820 if (!commit) {
821 warning("%s: Unexpected object of type %s, skipping.",
822 e->name,
823 typename(e->item->type));
824 continue;
825 }
826
827 switch(commit->object.type) {
828 case OBJ_COMMIT:
829 break;
830 case OBJ_BLOB:
831 export_blob(&commit->object.oid);
832 continue;
833 default: /* OBJ_TAG (nested tags) is already handled */
834 warning("Tag points to object of unexpected type %s, skipping.",
835 typename(commit->object.type));
836 continue;
837 }
838
839 /*
840 * This ref will not be updated through a commit, lets make
841 * sure it gets properly updated eventually.
842 */
843 if (commit->util || commit->object.flags & SHOWN)
844 string_list_append(&extra_refs, full_name)->util = commit;
845 if (!commit->util)
846 commit->util = full_name;
847 }
848}
849
850static void handle_tags_and_duplicates(void)
851{
852 struct commit *commit;
853 int i;
854
855 for (i = extra_refs.nr - 1; i >= 0; i--) {
856 const char *name = extra_refs.items[i].string;
857 struct object *object = extra_refs.items[i].util;
858 switch (object->type) {
859 case OBJ_TAG:
860 handle_tag(name, (struct tag *)object);
861 break;
862 case OBJ_COMMIT:
863 if (anonymize)
864 name = anonymize_refname(name);
865 /* create refs pointing to already seen commits */
866 commit = (struct commit *)object;
867 printf("reset %s\nfrom :%d\n\n", name,
868 get_object_mark(&commit->object));
869 show_progress();
870 break;
871 }
872 }
873}
874
875static void export_marks(char *file)
876{
877 unsigned int i;
878 uint32_t mark;
879 struct object_decoration *deco = idnums.hash;
880 FILE *f;
881 int e = 0;
882
883 f = fopen_for_writing(file);
884 if (!f)
885 die_errno("Unable to open marks file %s for writing.", file);
886
887 for (i = 0; i < idnums.size; i++) {
888 if (deco->base && deco->base->type == 1) {
889 mark = ptr_to_mark(deco->decoration);
890 if (fprintf(f, ":%"PRIu32" %s\n", mark,
891 oid_to_hex(&deco->base->oid)) < 0) {
892 e = 1;
893 break;
894 }
895 }
896 deco++;
897 }
898
899 e |= ferror(f);
900 e |= fclose(f);
901 if (e)
902 error("Unable to write marks file %s.", file);
903}
904
905static void import_marks(char *input_file)
906{
907 char line[512];
908 FILE *f = xfopen(input_file, "r");
909
910 while (fgets(line, sizeof(line), f)) {
911 uint32_t mark;
912 char *line_end, *mark_end;
913 struct object_id oid;
914 struct object *object;
915 struct commit *commit;
916 enum object_type type;
917
918 line_end = strchr(line, '\n');
919 if (line[0] != ':' || !line_end)
920 die("corrupt mark line: %s", line);
921 *line_end = '\0';
922
923 mark = strtoumax(line + 1, &mark_end, 10);
924 if (!mark || mark_end == line + 1
925 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
926 die("corrupt mark line: %s", line);
927
928 if (last_idnum < mark)
929 last_idnum = mark;
930
931 type = sha1_object_info(oid.hash, NULL);
932 if (type < 0)
933 die("object not found: %s", oid_to_hex(&oid));
934
935 if (type != OBJ_COMMIT)
936 /* only commits */
937 continue;
938
939 commit = lookup_commit(oid.hash);
940 if (!commit)
941 die("not a commit? can't happen: %s", oid_to_hex(&oid));
942
943 object = &commit->object;
944
945 if (object->flags & SHOWN)
946 error("Object %s already has a mark", oid_to_hex(&oid));
947
948 mark_object(object, mark);
949
950 object->flags |= SHOWN;
951 }
952 fclose(f);
953}
954
955static void handle_deletes(void)
956{
957 int i;
958 for (i = 0; i < refspecs_nr; i++) {
959 struct refspec *refspec = &refspecs[i];
960 if (*refspec->src)
961 continue;
962
963 printf("reset %s\nfrom %s\n\n",
964 refspec->dst, sha1_to_hex(null_sha1));
965 }
966}
967
968int cmd_fast_export(int argc, const char **argv, const char *prefix)
969{
970 struct rev_info revs;
971 struct object_array commits = OBJECT_ARRAY_INIT;
972 struct commit *commit;
973 char *export_filename = NULL, *import_filename = NULL;
974 uint32_t lastimportid;
975 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
976 struct option options[] = {
977 OPT_INTEGER(0, "progress", &progress,
978 N_("show progress after <n> objects")),
979 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
980 N_("select handling of signed tags"),
981 parse_opt_signed_tag_mode),
982 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
983 N_("select handling of tags that tag filtered objects"),
984 parse_opt_tag_of_filtered_mode),
985 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
986 N_("Dump marks to this file")),
987 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
988 N_("Import marks from this file")),
989 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
990 N_("Fake a tagger when tags lack one")),
991 OPT_BOOL(0, "full-tree", &full_tree,
992 N_("Output full tree for each commit")),
993 OPT_BOOL(0, "use-done-feature", &use_done_feature,
994 N_("Use the done feature to terminate the stream")),
995 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
996 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
997 N_("Apply refspec to exported refs")),
998 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
999 OPT_END()
1000 };
1001
1002 if (argc == 1)
1003 usage_with_options (fast_export_usage, options);
1004
1005 /* we handle encodings */
1006 git_config(git_default_config, NULL);
1007
1008 init_revisions(&revs, prefix);
1009 revs.topo_order = 1;
1010 revs.show_source = 1;
1011 revs.rewrite_parents = 1;
1012 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1013 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1014 argc = setup_revisions(argc, argv, &revs, NULL);
1015 if (argc > 1)
1016 usage_with_options (fast_export_usage, options);
1017
1018 if (refspecs_list.nr) {
1019 const char **refspecs_str;
1020 int i;
1021
1022 ALLOC_ARRAY(refspecs_str, refspecs_list.nr);
1023 for (i = 0; i < refspecs_list.nr; i++)
1024 refspecs_str[i] = refspecs_list.items[i].string;
1025
1026 refspecs_nr = refspecs_list.nr;
1027 refspecs = parse_fetch_refspec(refspecs_nr, refspecs_str);
1028
1029 string_list_clear(&refspecs_list, 1);
1030 free(refspecs_str);
1031 }
1032
1033 if (use_done_feature)
1034 printf("feature done\n");
1035
1036 if (import_filename)
1037 import_marks(import_filename);
1038 lastimportid = last_idnum;
1039
1040 if (import_filename && revs.prune_data.nr)
1041 full_tree = 1;
1042
1043 get_tags_and_duplicates(&revs.cmdline);
1044
1045 if (prepare_revision_walk(&revs))
1046 die("revision walk setup failed");
1047 revs.diffopt.format_callback = show_filemodify;
1048 DIFF_OPT_SET(&revs.diffopt, RECURSIVE);
1049 while ((commit = get_revision(&revs))) {
1050 if (has_unshown_parent(commit)) {
1051 add_object_array(&commit->object, NULL, &commits);
1052 }
1053 else {
1054 handle_commit(commit, &revs);
1055 handle_tail(&commits, &revs);
1056 }
1057 }
1058
1059 handle_tags_and_duplicates();
1060 handle_deletes();
1061
1062 if (export_filename && lastimportid != last_idnum)
1063 export_marks(export_filename);
1064
1065 if (use_done_feature)
1066 printf("done\n");
1067
1068 free_refspec(refspecs_nr, refspecs);
1069
1070 return 0;
1071}