1/*
2 * "git fast-export" builtin command
3 *
4 * Copyright (C) 2007 Johannes E. Schindelin
5 */
6#include "builtin.h"
7#include "cache.h"
8#include "config.h"
9#include "refs.h"
10#include "commit.h"
11#include "object.h"
12#include "tag.h"
13#include "diff.h"
14#include "diffcore.h"
15#include "log-tree.h"
16#include "revision.h"
17#include "decorate.h"
18#include "string-list.h"
19#include "utf8.h"
20#include "parse-options.h"
21#include "quote.h"
22#include "remote.h"
23#include "blob.h"
24
25static const char *fast_export_usage[] = {
26 N_("git fast-export [rev-list-opts]"),
27 NULL
28};
29
30static int progress;
31static enum { ABORT, VERBATIM, WARN, WARN_STRIP, STRIP } signed_tag_mode = ABORT;
32static enum { ERROR, DROP, REWRITE } tag_of_filtered_mode = ERROR;
33static int fake_missing_tagger;
34static int use_done_feature;
35static int no_data;
36static int full_tree;
37static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
38static struct refspec *refspecs;
39static int refspecs_nr;
40static int anonymize;
41
42static int parse_opt_signed_tag_mode(const struct option *opt,
43 const char *arg, int unset)
44{
45 if (unset || !strcmp(arg, "abort"))
46 signed_tag_mode = ABORT;
47 else if (!strcmp(arg, "verbatim") || !strcmp(arg, "ignore"))
48 signed_tag_mode = VERBATIM;
49 else if (!strcmp(arg, "warn"))
50 signed_tag_mode = WARN;
51 else if (!strcmp(arg, "warn-strip"))
52 signed_tag_mode = WARN_STRIP;
53 else if (!strcmp(arg, "strip"))
54 signed_tag_mode = STRIP;
55 else
56 return error("Unknown signed-tags mode: %s", arg);
57 return 0;
58}
59
60static int parse_opt_tag_of_filtered_mode(const struct option *opt,
61 const char *arg, int unset)
62{
63 if (unset || !strcmp(arg, "abort"))
64 tag_of_filtered_mode = ERROR;
65 else if (!strcmp(arg, "drop"))
66 tag_of_filtered_mode = DROP;
67 else if (!strcmp(arg, "rewrite"))
68 tag_of_filtered_mode = REWRITE;
69 else
70 return error("Unknown tag-of-filtered mode: %s", arg);
71 return 0;
72}
73
74static struct decoration idnums;
75static uint32_t last_idnum;
76
77static int has_unshown_parent(struct commit *commit)
78{
79 struct commit_list *parent;
80
81 for (parent = commit->parents; parent; parent = parent->next)
82 if (!(parent->item->object.flags & SHOWN) &&
83 !(parent->item->object.flags & UNINTERESTING))
84 return 1;
85 return 0;
86}
87
88struct anonymized_entry {
89 struct hashmap_entry hash;
90 const char *orig;
91 size_t orig_len;
92 const char *anon;
93 size_t anon_len;
94};
95
96static int anonymized_entry_cmp(const void *va, const void *vb,
97 const void *data)
98{
99 const struct anonymized_entry *a = va, *b = vb;
100 return a->orig_len != b->orig_len ||
101 memcmp(a->orig, b->orig, a->orig_len);
102}
103
104/*
105 * Basically keep a cache of X->Y so that we can repeatedly replace
106 * the same anonymized string with another. The actual generation
107 * is farmed out to the generate function.
108 */
109static const void *anonymize_mem(struct hashmap *map,
110 void *(*generate)(const void *, size_t *),
111 const void *orig, size_t *len)
112{
113 struct anonymized_entry key, *ret;
114
115 if (!map->cmpfn)
116 hashmap_init(map, anonymized_entry_cmp, 0);
117
118 hashmap_entry_init(&key, memhash(orig, *len));
119 key.orig = orig;
120 key.orig_len = *len;
121 ret = hashmap_get(map, &key, NULL);
122
123 if (!ret) {
124 ret = xmalloc(sizeof(*ret));
125 hashmap_entry_init(&ret->hash, key.hash.hash);
126 ret->orig = xstrdup(orig);
127 ret->orig_len = *len;
128 ret->anon = generate(orig, len);
129 ret->anon_len = *len;
130 hashmap_put(map, ret);
131 }
132
133 *len = ret->anon_len;
134 return ret->anon;
135}
136
137/*
138 * We anonymize each component of a path individually,
139 * so that paths a/b and a/c will share a common root.
140 * The paths are cached via anonymize_mem so that repeated
141 * lookups for "a" will yield the same value.
142 */
143static void anonymize_path(struct strbuf *out, const char *path,
144 struct hashmap *map,
145 void *(*generate)(const void *, size_t *))
146{
147 while (*path) {
148 const char *end_of_component = strchrnul(path, '/');
149 size_t len = end_of_component - path;
150 const char *c = anonymize_mem(map, generate, path, &len);
151 strbuf_add(out, c, len);
152 path = end_of_component;
153 if (*path)
154 strbuf_addch(out, *path++);
155 }
156}
157
158/* Since intptr_t is C99, we do not use it here */
159static inline uint32_t *mark_to_ptr(uint32_t mark)
160{
161 return ((uint32_t *)NULL) + mark;
162}
163
164static inline uint32_t ptr_to_mark(void * mark)
165{
166 return (uint32_t *)mark - (uint32_t *)NULL;
167}
168
169static inline void mark_object(struct object *object, uint32_t mark)
170{
171 add_decoration(&idnums, object, mark_to_ptr(mark));
172}
173
174static inline void mark_next_object(struct object *object)
175{
176 mark_object(object, ++last_idnum);
177}
178
179static int get_object_mark(struct object *object)
180{
181 void *decoration = lookup_decoration(&idnums, object);
182 if (!decoration)
183 return 0;
184 return ptr_to_mark(decoration);
185}
186
187static void show_progress(void)
188{
189 static int counter = 0;
190 if (!progress)
191 return;
192 if ((++counter % progress) == 0)
193 printf("progress %d objects\n", counter);
194}
195
196/*
197 * Ideally we would want some transformation of the blob data here
198 * that is unreversible, but would still be the same size and have
199 * the same data relationship to other blobs (so that we get the same
200 * delta and packing behavior as the original). But the first and last
201 * requirements there are probably mutually exclusive, so let's take
202 * the easy way out for now, and just generate arbitrary content.
203 *
204 * There's no need to cache this result with anonymize_mem, since
205 * we already handle blob content caching with marks.
206 */
207static char *anonymize_blob(unsigned long *size)
208{
209 static int counter;
210 struct strbuf out = STRBUF_INIT;
211 strbuf_addf(&out, "anonymous blob %d", counter++);
212 *size = out.len;
213 return strbuf_detach(&out, NULL);
214}
215
216static void export_blob(const struct object_id *oid)
217{
218 unsigned long size;
219 enum object_type type;
220 char *buf;
221 struct object *object;
222 int eaten;
223
224 if (no_data)
225 return;
226
227 if (is_null_oid(oid))
228 return;
229
230 object = lookup_object(oid->hash);
231 if (object && object->flags & SHOWN)
232 return;
233
234 if (anonymize) {
235 buf = anonymize_blob(&size);
236 object = (struct object *)lookup_blob(oid);
237 eaten = 0;
238 } else {
239 buf = read_sha1_file(oid->hash, &type, &size);
240 if (!buf)
241 die ("Could not read blob %s", oid_to_hex(oid));
242 if (check_sha1_signature(oid->hash, buf, size, typename(type)) < 0)
243 die("sha1 mismatch in blob %s", oid_to_hex(oid));
244 object = parse_object_buffer(oid, type, size, buf, &eaten);
245 }
246
247 if (!object)
248 die("Could not read blob %s", oid_to_hex(oid));
249
250 mark_next_object(object);
251
252 printf("blob\nmark :%"PRIu32"\ndata %lu\n", last_idnum, size);
253 if (size && fwrite(buf, size, 1, stdout) != 1)
254 die_errno ("Could not write blob '%s'", oid_to_hex(oid));
255 printf("\n");
256
257 show_progress();
258
259 object->flags |= SHOWN;
260 if (!eaten)
261 free(buf);
262}
263
264static int depth_first(const void *a_, const void *b_)
265{
266 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
267 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
268 const char *name_a, *name_b;
269 int len_a, len_b, len;
270 int cmp;
271
272 name_a = a->one ? a->one->path : a->two->path;
273 name_b = b->one ? b->one->path : b->two->path;
274
275 len_a = strlen(name_a);
276 len_b = strlen(name_b);
277 len = (len_a < len_b) ? len_a : len_b;
278
279 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
280 cmp = memcmp(name_a, name_b, len);
281 if (cmp)
282 return cmp;
283 cmp = len_b - len_a;
284 if (cmp)
285 return cmp;
286 /*
287 * Move 'R'ename entries last so that all references of the file
288 * appear in the output before it is renamed (e.g., when a file
289 * was copied and renamed in the same commit).
290 */
291 return (a->status == 'R') - (b->status == 'R');
292}
293
294static void print_path_1(const char *path)
295{
296 int need_quote = quote_c_style(path, NULL, NULL, 0);
297 if (need_quote)
298 quote_c_style(path, NULL, stdout, 0);
299 else if (strchr(path, ' '))
300 printf("\"%s\"", path);
301 else
302 printf("%s", path);
303}
304
305static void *anonymize_path_component(const void *path, size_t *len)
306{
307 static int counter;
308 struct strbuf out = STRBUF_INIT;
309 strbuf_addf(&out, "path%d", counter++);
310 return strbuf_detach(&out, len);
311}
312
313static void print_path(const char *path)
314{
315 if (!anonymize)
316 print_path_1(path);
317 else {
318 static struct hashmap paths;
319 static struct strbuf anon = STRBUF_INIT;
320
321 anonymize_path(&anon, path, &paths, anonymize_path_component);
322 print_path_1(anon.buf);
323 strbuf_reset(&anon);
324 }
325}
326
327static void *generate_fake_oid(const void *old, size_t *len)
328{
329 static uint32_t counter = 1; /* avoid null sha1 */
330 unsigned char *out = xcalloc(GIT_SHA1_RAWSZ, 1);
331 put_be32(out + GIT_SHA1_RAWSZ - 4, counter++);
332 return out;
333}
334
335static const unsigned char *anonymize_sha1(const struct object_id *oid)
336{
337 static struct hashmap sha1s;
338 size_t len = GIT_SHA1_RAWSZ;
339 return anonymize_mem(&sha1s, generate_fake_oid, oid, &len);
340}
341
342static void show_filemodify(struct diff_queue_struct *q,
343 struct diff_options *options, void *data)
344{
345 int i;
346
347 /*
348 * Handle files below a directory first, in case they are all deleted
349 * and the directory changes to a file or symlink.
350 */
351 QSORT(q->queue, q->nr, depth_first);
352
353 for (i = 0; i < q->nr; i++) {
354 struct diff_filespec *ospec = q->queue[i]->one;
355 struct diff_filespec *spec = q->queue[i]->two;
356
357 switch (q->queue[i]->status) {
358 case DIFF_STATUS_DELETED:
359 printf("D ");
360 print_path(spec->path);
361 putchar('\n');
362 break;
363
364 case DIFF_STATUS_COPIED:
365 case DIFF_STATUS_RENAMED:
366 printf("%c ", q->queue[i]->status);
367 print_path(ospec->path);
368 putchar(' ');
369 print_path(spec->path);
370 putchar('\n');
371
372 if (!oidcmp(&ospec->oid, &spec->oid) &&
373 ospec->mode == spec->mode)
374 break;
375 /* fallthrough */
376
377 case DIFF_STATUS_TYPE_CHANGED:
378 case DIFF_STATUS_MODIFIED:
379 case DIFF_STATUS_ADDED:
380 /*
381 * Links refer to objects in another repositories;
382 * output the SHA-1 verbatim.
383 */
384 if (no_data || S_ISGITLINK(spec->mode))
385 printf("M %06o %s ", spec->mode,
386 sha1_to_hex(anonymize ?
387 anonymize_sha1(&spec->oid) :
388 spec->oid.hash));
389 else {
390 struct object *object = lookup_object(spec->oid.hash);
391 printf("M %06o :%d ", spec->mode,
392 get_object_mark(object));
393 }
394 print_path(spec->path);
395 putchar('\n');
396 break;
397
398 default:
399 die("Unexpected comparison status '%c' for %s, %s",
400 q->queue[i]->status,
401 ospec->path ? ospec->path : "none",
402 spec->path ? spec->path : "none");
403 }
404 }
405}
406
407static const char *find_encoding(const char *begin, const char *end)
408{
409 const char *needle = "\nencoding ";
410 char *bol, *eol;
411
412 bol = memmem(begin, end ? end - begin : strlen(begin),
413 needle, strlen(needle));
414 if (!bol)
415 return git_commit_encoding;
416 bol += strlen(needle);
417 eol = strchrnul(bol, '\n');
418 *eol = '\0';
419 return bol;
420}
421
422static void *anonymize_ref_component(const void *old, size_t *len)
423{
424 static int counter;
425 struct strbuf out = STRBUF_INIT;
426 strbuf_addf(&out, "ref%d", counter++);
427 return strbuf_detach(&out, len);
428}
429
430static const char *anonymize_refname(const char *refname)
431{
432 /*
433 * If any of these prefixes is found, we will leave it intact
434 * so that tags remain tags and so forth.
435 */
436 static const char *prefixes[] = {
437 "refs/heads/",
438 "refs/tags/",
439 "refs/remotes/",
440 "refs/"
441 };
442 static struct hashmap refs;
443 static struct strbuf anon = STRBUF_INIT;
444 int i;
445
446 /*
447 * We also leave "master" as a special case, since it does not reveal
448 * anything interesting.
449 */
450 if (!strcmp(refname, "refs/heads/master"))
451 return refname;
452
453 strbuf_reset(&anon);
454 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
455 if (skip_prefix(refname, prefixes[i], &refname)) {
456 strbuf_addstr(&anon, prefixes[i]);
457 break;
458 }
459 }
460
461 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
462 return anon.buf;
463}
464
465/*
466 * We do not even bother to cache commit messages, as they are unlikely
467 * to be repeated verbatim, and it is not that interesting when they are.
468 */
469static char *anonymize_commit_message(const char *old)
470{
471 static int counter;
472 return xstrfmt("subject %d\n\nbody\n", counter++);
473}
474
475static struct hashmap idents;
476static void *anonymize_ident(const void *old, size_t *len)
477{
478 static int counter;
479 struct strbuf out = STRBUF_INIT;
480 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
481 counter++;
482 return strbuf_detach(&out, len);
483}
484
485/*
486 * Our strategy here is to anonymize the names and email addresses,
487 * but keep timestamps intact, as they influence things like traversal
488 * order (and by themselves should not be too revealing).
489 */
490static void anonymize_ident_line(const char **beg, const char **end)
491{
492 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
493 static unsigned which_buffer;
494
495 struct strbuf *out;
496 struct ident_split split;
497 const char *end_of_header;
498
499 out = &buffers[which_buffer++];
500 which_buffer %= ARRAY_SIZE(buffers);
501 strbuf_reset(out);
502
503 /* skip "committer", "author", "tagger", etc */
504 end_of_header = strchr(*beg, ' ');
505 if (!end_of_header)
506 die("BUG: malformed line fed to anonymize_ident_line: %.*s",
507 (int)(*end - *beg), *beg);
508 end_of_header++;
509 strbuf_add(out, *beg, end_of_header - *beg);
510
511 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
512 split.date_begin) {
513 const char *ident;
514 size_t len;
515
516 len = split.mail_end - split.name_begin;
517 ident = anonymize_mem(&idents, anonymize_ident,
518 split.name_begin, &len);
519 strbuf_add(out, ident, len);
520 strbuf_addch(out, ' ');
521 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
522 } else {
523 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
524 }
525
526 *beg = out->buf;
527 *end = out->buf + out->len;
528}
529
530static void handle_commit(struct commit *commit, struct rev_info *rev)
531{
532 int saved_output_format = rev->diffopt.output_format;
533 const char *commit_buffer;
534 const char *author, *author_end, *committer, *committer_end;
535 const char *encoding, *message;
536 char *reencoded = NULL;
537 struct commit_list *p;
538 const char *refname;
539 int i;
540
541 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
542
543 parse_commit_or_die(commit);
544 commit_buffer = get_commit_buffer(commit, NULL);
545 author = strstr(commit_buffer, "\nauthor ");
546 if (!author)
547 die ("Could not find author in commit %s",
548 oid_to_hex(&commit->object.oid));
549 author++;
550 author_end = strchrnul(author, '\n');
551 committer = strstr(author_end, "\ncommitter ");
552 if (!committer)
553 die ("Could not find committer in commit %s",
554 oid_to_hex(&commit->object.oid));
555 committer++;
556 committer_end = strchrnul(committer, '\n');
557 message = strstr(committer_end, "\n\n");
558 encoding = find_encoding(committer_end, message);
559 if (message)
560 message += 2;
561
562 if (commit->parents &&
563 get_object_mark(&commit->parents->item->object) != 0 &&
564 !full_tree) {
565 parse_commit_or_die(commit->parents->item);
566 diff_tree_oid(&commit->parents->item->tree->object.oid,
567 &commit->tree->object.oid, "", &rev->diffopt);
568 }
569 else
570 diff_root_tree_oid(&commit->tree->object.oid,
571 "", &rev->diffopt);
572
573 /* Export the referenced blobs, and remember the marks. */
574 for (i = 0; i < diff_queued_diff.nr; i++)
575 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
576 export_blob(&diff_queued_diff.queue[i]->two->oid);
577
578 refname = commit->util;
579 if (anonymize) {
580 refname = anonymize_refname(refname);
581 anonymize_ident_line(&committer, &committer_end);
582 anonymize_ident_line(&author, &author_end);
583 }
584
585 mark_next_object(&commit->object);
586 if (anonymize)
587 reencoded = anonymize_commit_message(message);
588 else if (!is_encoding_utf8(encoding))
589 reencoded = reencode_string(message, "UTF-8", encoding);
590 if (!commit->parents)
591 printf("reset %s\n", refname);
592 printf("commit %s\nmark :%"PRIu32"\n%.*s\n%.*s\ndata %u\n%s",
593 refname, last_idnum,
594 (int)(author_end - author), author,
595 (int)(committer_end - committer), committer,
596 (unsigned)(reencoded
597 ? strlen(reencoded) : message
598 ? strlen(message) : 0),
599 reencoded ? reencoded : message ? message : "");
600 free(reencoded);
601 unuse_commit_buffer(commit, commit_buffer);
602
603 for (i = 0, p = commit->parents; p; p = p->next) {
604 int mark = get_object_mark(&p->item->object);
605 if (!mark)
606 continue;
607 if (i == 0)
608 printf("from :%d\n", mark);
609 else
610 printf("merge :%d\n", mark);
611 i++;
612 }
613
614 if (full_tree)
615 printf("deleteall\n");
616 log_tree_diff_flush(rev);
617 rev->diffopt.output_format = saved_output_format;
618
619 printf("\n");
620
621 show_progress();
622}
623
624static void *anonymize_tag(const void *old, size_t *len)
625{
626 static int counter;
627 struct strbuf out = STRBUF_INIT;
628 strbuf_addf(&out, "tag message %d", counter++);
629 return strbuf_detach(&out, len);
630}
631
632static void handle_tail(struct object_array *commits, struct rev_info *revs)
633{
634 struct commit *commit;
635 while (commits->nr) {
636 commit = (struct commit *)commits->objects[commits->nr - 1].item;
637 if (has_unshown_parent(commit))
638 return;
639 handle_commit(commit, revs);
640 commits->nr--;
641 }
642}
643
644static void handle_tag(const char *name, struct tag *tag)
645{
646 unsigned long size;
647 enum object_type type;
648 char *buf;
649 const char *tagger, *tagger_end, *message;
650 size_t message_size = 0;
651 struct object *tagged;
652 int tagged_mark;
653 struct commit *p;
654
655 /* Trees have no identifier in fast-export output, thus we have no way
656 * to output tags of trees, tags of tags of trees, etc. Simply omit
657 * such tags.
658 */
659 tagged = tag->tagged;
660 while (tagged->type == OBJ_TAG) {
661 tagged = ((struct tag *)tagged)->tagged;
662 }
663 if (tagged->type == OBJ_TREE) {
664 warning("Omitting tag %s,\nsince tags of trees (or tags of tags of trees, etc.) are not supported.",
665 oid_to_hex(&tag->object.oid));
666 return;
667 }
668
669 buf = read_sha1_file(tag->object.oid.hash, &type, &size);
670 if (!buf)
671 die ("Could not read tag %s", oid_to_hex(&tag->object.oid));
672 message = memmem(buf, size, "\n\n", 2);
673 if (message) {
674 message += 2;
675 message_size = strlen(message);
676 }
677 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
678 if (!tagger) {
679 if (fake_missing_tagger)
680 tagger = "tagger Unspecified Tagger "
681 "<unspecified-tagger> 0 +0000";
682 else
683 tagger = "";
684 tagger_end = tagger + strlen(tagger);
685 } else {
686 tagger++;
687 tagger_end = strchrnul(tagger, '\n');
688 if (anonymize)
689 anonymize_ident_line(&tagger, &tagger_end);
690 }
691
692 if (anonymize) {
693 name = anonymize_refname(name);
694 if (message) {
695 static struct hashmap tags;
696 message = anonymize_mem(&tags, anonymize_tag,
697 message, &message_size);
698 }
699 }
700
701 /* handle signed tags */
702 if (message) {
703 const char *signature = strstr(message,
704 "\n-----BEGIN PGP SIGNATURE-----\n");
705 if (signature)
706 switch(signed_tag_mode) {
707 case ABORT:
708 die ("Encountered signed tag %s; use "
709 "--signed-tags=<mode> to handle it.",
710 oid_to_hex(&tag->object.oid));
711 case WARN:
712 warning ("Exporting signed tag %s",
713 oid_to_hex(&tag->object.oid));
714 /* fallthru */
715 case VERBATIM:
716 break;
717 case WARN_STRIP:
718 warning ("Stripping signature from tag %s",
719 oid_to_hex(&tag->object.oid));
720 /* fallthru */
721 case STRIP:
722 message_size = signature + 1 - message;
723 break;
724 }
725 }
726
727 /* handle tag->tagged having been filtered out due to paths specified */
728 tagged = tag->tagged;
729 tagged_mark = get_object_mark(tagged);
730 if (!tagged_mark) {
731 switch(tag_of_filtered_mode) {
732 case ABORT:
733 die ("Tag %s tags unexported object; use "
734 "--tag-of-filtered-object=<mode> to handle it.",
735 oid_to_hex(&tag->object.oid));
736 case DROP:
737 /* Ignore this tag altogether */
738 free(buf);
739 return;
740 case REWRITE:
741 if (tagged->type != OBJ_COMMIT) {
742 die ("Tag %s tags unexported %s!",
743 oid_to_hex(&tag->object.oid),
744 typename(tagged->type));
745 }
746 p = (struct commit *)tagged;
747 for (;;) {
748 if (p->parents && p->parents->next)
749 break;
750 if (p->object.flags & UNINTERESTING)
751 break;
752 if (!(p->object.flags & TREESAME))
753 break;
754 if (!p->parents)
755 die ("Can't find replacement commit for tag %s\n",
756 oid_to_hex(&tag->object.oid));
757 p = p->parents->item;
758 }
759 tagged_mark = get_object_mark(&p->object);
760 }
761 }
762
763 if (starts_with(name, "refs/tags/"))
764 name += 10;
765 printf("tag %s\nfrom :%d\n%.*s%sdata %d\n%.*s\n",
766 name, tagged_mark,
767 (int)(tagger_end - tagger), tagger,
768 tagger == tagger_end ? "" : "\n",
769 (int)message_size, (int)message_size, message ? message : "");
770 free(buf);
771}
772
773static struct commit *get_commit(struct rev_cmdline_entry *e, char *full_name)
774{
775 switch (e->item->type) {
776 case OBJ_COMMIT:
777 return (struct commit *)e->item;
778 case OBJ_TAG: {
779 struct tag *tag = (struct tag *)e->item;
780
781 /* handle nested tags */
782 while (tag && tag->object.type == OBJ_TAG) {
783 parse_object(&tag->object.oid);
784 string_list_append(&extra_refs, full_name)->util = tag;
785 tag = (struct tag *)tag->tagged;
786 }
787 if (!tag)
788 die("Tag %s points nowhere?", e->name);
789 return (struct commit *)tag;
790 break;
791 }
792 default:
793 return NULL;
794 }
795}
796
797static void get_tags_and_duplicates(struct rev_cmdline_info *info)
798{
799 int i;
800
801 for (i = 0; i < info->nr; i++) {
802 struct rev_cmdline_entry *e = info->rev + i;
803 struct object_id oid;
804 struct commit *commit;
805 char *full_name;
806
807 if (e->flags & UNINTERESTING)
808 continue;
809
810 if (dwim_ref(e->name, strlen(e->name), oid.hash, &full_name) != 1)
811 continue;
812
813 if (refspecs) {
814 char *private;
815 private = apply_refspecs(refspecs, refspecs_nr, full_name);
816 if (private) {
817 free(full_name);
818 full_name = private;
819 }
820 }
821
822 commit = get_commit(e, full_name);
823 if (!commit) {
824 warning("%s: Unexpected object of type %s, skipping.",
825 e->name,
826 typename(e->item->type));
827 continue;
828 }
829
830 switch(commit->object.type) {
831 case OBJ_COMMIT:
832 break;
833 case OBJ_BLOB:
834 export_blob(&commit->object.oid);
835 continue;
836 default: /* OBJ_TAG (nested tags) is already handled */
837 warning("Tag points to object of unexpected type %s, skipping.",
838 typename(commit->object.type));
839 continue;
840 }
841
842 /*
843 * This ref will not be updated through a commit, lets make
844 * sure it gets properly updated eventually.
845 */
846 if (commit->util || commit->object.flags & SHOWN)
847 string_list_append(&extra_refs, full_name)->util = commit;
848 if (!commit->util)
849 commit->util = full_name;
850 }
851}
852
853static void handle_tags_and_duplicates(void)
854{
855 struct commit *commit;
856 int i;
857
858 for (i = extra_refs.nr - 1; i >= 0; i--) {
859 const char *name = extra_refs.items[i].string;
860 struct object *object = extra_refs.items[i].util;
861 switch (object->type) {
862 case OBJ_TAG:
863 handle_tag(name, (struct tag *)object);
864 break;
865 case OBJ_COMMIT:
866 if (anonymize)
867 name = anonymize_refname(name);
868 /* create refs pointing to already seen commits */
869 commit = (struct commit *)object;
870 printf("reset %s\nfrom :%d\n\n", name,
871 get_object_mark(&commit->object));
872 show_progress();
873 break;
874 }
875 }
876}
877
878static void export_marks(char *file)
879{
880 unsigned int i;
881 uint32_t mark;
882 struct object_decoration *deco = idnums.hash;
883 FILE *f;
884 int e = 0;
885
886 f = fopen_for_writing(file);
887 if (!f)
888 die_errno("Unable to open marks file %s for writing.", file);
889
890 for (i = 0; i < idnums.size; i++) {
891 if (deco->base && deco->base->type == 1) {
892 mark = ptr_to_mark(deco->decoration);
893 if (fprintf(f, ":%"PRIu32" %s\n", mark,
894 oid_to_hex(&deco->base->oid)) < 0) {
895 e = 1;
896 break;
897 }
898 }
899 deco++;
900 }
901
902 e |= ferror(f);
903 e |= fclose(f);
904 if (e)
905 error("Unable to write marks file %s.", file);
906}
907
908static void import_marks(char *input_file)
909{
910 char line[512];
911 FILE *f = xfopen(input_file, "r");
912
913 while (fgets(line, sizeof(line), f)) {
914 uint32_t mark;
915 char *line_end, *mark_end;
916 struct object_id oid;
917 struct object *object;
918 struct commit *commit;
919 enum object_type type;
920
921 line_end = strchr(line, '\n');
922 if (line[0] != ':' || !line_end)
923 die("corrupt mark line: %s", line);
924 *line_end = '\0';
925
926 mark = strtoumax(line + 1, &mark_end, 10);
927 if (!mark || mark_end == line + 1
928 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
929 die("corrupt mark line: %s", line);
930
931 if (last_idnum < mark)
932 last_idnum = mark;
933
934 type = sha1_object_info(oid.hash, NULL);
935 if (type < 0)
936 die("object not found: %s", oid_to_hex(&oid));
937
938 if (type != OBJ_COMMIT)
939 /* only commits */
940 continue;
941
942 commit = lookup_commit(&oid);
943 if (!commit)
944 die("not a commit? can't happen: %s", oid_to_hex(&oid));
945
946 object = &commit->object;
947
948 if (object->flags & SHOWN)
949 error("Object %s already has a mark", oid_to_hex(&oid));
950
951 mark_object(object, mark);
952
953 object->flags |= SHOWN;
954 }
955 fclose(f);
956}
957
958static void handle_deletes(void)
959{
960 int i;
961 for (i = 0; i < refspecs_nr; i++) {
962 struct refspec *refspec = &refspecs[i];
963 if (*refspec->src)
964 continue;
965
966 printf("reset %s\nfrom %s\n\n",
967 refspec->dst, sha1_to_hex(null_sha1));
968 }
969}
970
971int cmd_fast_export(int argc, const char **argv, const char *prefix)
972{
973 struct rev_info revs;
974 struct object_array commits = OBJECT_ARRAY_INIT;
975 struct commit *commit;
976 char *export_filename = NULL, *import_filename = NULL;
977 uint32_t lastimportid;
978 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
979 struct option options[] = {
980 OPT_INTEGER(0, "progress", &progress,
981 N_("show progress after <n> objects")),
982 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
983 N_("select handling of signed tags"),
984 parse_opt_signed_tag_mode),
985 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
986 N_("select handling of tags that tag filtered objects"),
987 parse_opt_tag_of_filtered_mode),
988 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
989 N_("Dump marks to this file")),
990 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
991 N_("Import marks from this file")),
992 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
993 N_("Fake a tagger when tags lack one")),
994 OPT_BOOL(0, "full-tree", &full_tree,
995 N_("Output full tree for each commit")),
996 OPT_BOOL(0, "use-done-feature", &use_done_feature,
997 N_("Use the done feature to terminate the stream")),
998 OPT_BOOL(0, "no-data", &no_data, N_("Skip output of blob data")),
999 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1000 N_("Apply refspec to exported refs")),
1001 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1002 OPT_END()
1003 };
1004
1005 if (argc == 1)
1006 usage_with_options (fast_export_usage, options);
1007
1008 /* we handle encodings */
1009 git_config(git_default_config, NULL);
1010
1011 init_revisions(&revs, prefix);
1012 revs.topo_order = 1;
1013 revs.show_source = 1;
1014 revs.rewrite_parents = 1;
1015 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1016 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN);
1017 argc = setup_revisions(argc, argv, &revs, NULL);
1018 if (argc > 1)
1019 usage_with_options (fast_export_usage, options);
1020
1021 if (refspecs_list.nr) {
1022 const char **refspecs_str;
1023 int i;
1024
1025 ALLOC_ARRAY(refspecs_str, refspecs_list.nr);
1026 for (i = 0; i < refspecs_list.nr; i++)
1027 refspecs_str[i] = refspecs_list.items[i].string;
1028
1029 refspecs_nr = refspecs_list.nr;
1030 refspecs = parse_fetch_refspec(refspecs_nr, refspecs_str);
1031
1032 string_list_clear(&refspecs_list, 1);
1033 free(refspecs_str);
1034 }
1035
1036 if (use_done_feature)
1037 printf("feature done\n");
1038
1039 if (import_filename)
1040 import_marks(import_filename);
1041 lastimportid = last_idnum;
1042
1043 if (import_filename && revs.prune_data.nr)
1044 full_tree = 1;
1045
1046 get_tags_and_duplicates(&revs.cmdline);
1047
1048 if (prepare_revision_walk(&revs))
1049 die("revision walk setup failed");
1050 revs.diffopt.format_callback = show_filemodify;
1051 DIFF_OPT_SET(&revs.diffopt, RECURSIVE);
1052 while ((commit = get_revision(&revs))) {
1053 if (has_unshown_parent(commit)) {
1054 add_object_array(&commit->object, NULL, &commits);
1055 }
1056 else {
1057 handle_commit(commit, &revs);
1058 handle_tail(&commits, &revs);
1059 }
1060 }
1061
1062 handle_tags_and_duplicates();
1063 handle_deletes();
1064
1065 if (export_filename && lastimportid != last_idnum)
1066 export_marks(export_filename);
1067
1068 if (use_done_feature)
1069 printf("done\n");
1070
1071 free_refspec(refspecs_nr, refspecs);
1072
1073 return 0;
1074}