Raw
1 /*
2 * "git fast-export" builtin command
3 *
4 * Copyright (C) 2007 Johannes E. Schindelin
5 */
6
7 #define USE_THE_REPOSITORY_VARIABLE
8 #define DISABLE_SIGN_COMPARE_WARNINGS
9
10 #include "builtin.h"
11 #include "config.h"
12 #include "environment.h"
13 #include "gettext.h"
14 #include "hex.h"
15 #include "refs.h"
16 #include "refspec.h"
17 #include "object-file.h"
18 #include "odb.h"
19 #include "commit.h"
20 #include "object.h"
21 #include "tag.h"
22 #include "diff.h"
23 #include "diffcore.h"
24 #include "log-tree.h"
25 #include "revision.h"
26 #include "decorate.h"
27 #include "string-list.h"
28 #include "utf8.h"
29 #include "parse-options.h"
30 #include "quote.h"
31 #include "remote.h"
32 #include "blob.h"
33 #include "gpg-interface.h"
34
35 static const char *const fast_export_usage[] = {
36 N_("git fast-export [<rev-list-opts>]"),
37 NULL
38 };
39
40 static int progress;
41 static enum sign_mode signed_tag_mode = SIGN_ABORT;
42 static enum sign_mode signed_commit_mode = SIGN_STRIP;
43 static enum tag_of_filtered_mode { TAG_FILTERING_ABORT, DROP, REWRITE } tag_of_filtered_mode = TAG_FILTERING_ABORT;
44 static enum reencode_mode { REENCODE_ABORT, REENCODE_YES, REENCODE_NO } reencode_mode = REENCODE_ABORT;
45 static int fake_missing_tagger;
46 static int use_done_feature;
47 static int no_data;
48 static int full_tree;
49 static int reference_excluded_commits;
50 static int show_original_ids;
51 static int mark_tags;
52 static struct string_list extra_refs = STRING_LIST_INIT_DUP;
53 static struct string_list tag_refs = STRING_LIST_INIT_DUP;
54 static struct refspec refspecs = REFSPEC_INIT_FETCH;
55 static int anonymize;
56 static struct hashmap anonymized_seeds;
57 static struct revision_sources revision_sources;
58
59 static int parse_opt_sign_mode(const struct option *opt,
60 const char *arg, int unset)
61 {
62 enum sign_mode *val = opt->value;
63
64 if (unset)
65 return 0;
66
67 if (parse_sign_mode(arg, val, NULL) || (*val == SIGN_STRIP_IF_INVALID) ||
68 (*val == SIGN_SIGN_IF_INVALID) || (*val == SIGN_ABORT_IF_INVALID))
69 return error(_("unknown %s mode: %s"), opt->long_name, arg);
70
71 return 0;
72 }
73
74 static int parse_opt_tag_of_filtered_mode(const struct option *opt,
75 const char *arg, int unset)
76 {
77 enum tag_of_filtered_mode *val = opt->value;
78
79 if (unset || !strcmp(arg, "abort"))
80 *val = TAG_FILTERING_ABORT;
81 else if (!strcmp(arg, "drop"))
82 *val = DROP;
83 else if (!strcmp(arg, "rewrite"))
84 *val = REWRITE;
85 else
86 return error(_("unknown tag-of-filtered mode: %s"), arg);
87 return 0;
88 }
89
90 static int parse_opt_reencode_mode(const struct option *opt,
91 const char *arg, int unset)
92 {
93 enum reencode_mode *val = opt->value;
94
95 if (unset) {
96 *val = REENCODE_ABORT;
97 return 0;
98 }
99
100 switch (git_parse_maybe_bool(arg)) {
101 case 0:
102 *val = REENCODE_NO;
103 break;
104 case 1:
105 *val = REENCODE_YES;
106 break;
107 default:
108 if (!strcasecmp(arg, "abort"))
109 *val = REENCODE_ABORT;
110 else
111 return error(_("unknown reencoding mode: %s"), arg);
112 }
113
114 return 0;
115 }
116
117 static struct decoration idnums;
118 static uint32_t last_idnum;
119 struct anonymized_entry {
120 struct hashmap_entry hash;
121 char *anon;
122 const char orig[FLEX_ARRAY];
123 };
124
125 struct anonymized_entry_key {
126 struct hashmap_entry hash;
127 const char *orig;
128 size_t orig_len;
129 };
130
131 static int anonymized_entry_cmp(const void *cmp_data UNUSED,
132 const struct hashmap_entry *eptr,
133 const struct hashmap_entry *entry_or_key,
134 const void *keydata)
135 {
136 const struct anonymized_entry *a, *b;
137
138 a = container_of(eptr, const struct anonymized_entry, hash);
139 if (keydata) {
140 const struct anonymized_entry_key *key = keydata;
141 int equal = !xstrncmpz(a->orig, key->orig, key->orig_len);
142 return !equal;
143 }
144
145 b = container_of(entry_or_key, const struct anonymized_entry, hash);
146 return strcmp(a->orig, b->orig);
147 }
148
149 static struct anonymized_entry *add_anonymized_entry(struct hashmap *map,
150 unsigned hash,
151 const char *orig, size_t len,
152 char *anon)
153 {
154 struct anonymized_entry *ret, *old;
155
156 if (!map->cmpfn)
157 hashmap_init(map, anonymized_entry_cmp, NULL, 0);
158
159 FLEX_ALLOC_MEM(ret, orig, orig, len);
160 hashmap_entry_init(&ret->hash, hash);
161 ret->anon = anon;
162 old = hashmap_put_entry(map, ret, hash);
163
164 if (old) {
165 free(old->anon);
166 free(old);
167 }
168
169 return ret;
170 }
171
172 /*
173 * Basically keep a cache of X->Y so that we can repeatedly replace
174 * the same anonymized string with another. The actual generation
175 * is farmed out to the generate function.
176 */
177 static const char *anonymize_str(struct hashmap *map,
178 char *(*generate)(void),
179 const char *orig, size_t len)
180 {
181 struct anonymized_entry_key key;
182 struct anonymized_entry *ret;
183
184 hashmap_entry_init(&key.hash, memhash(orig, len));
185 key.orig = orig;
186 key.orig_len = len;
187
188 /* First check if it's a token the user configured manually... */
189 ret = hashmap_get_entry(&anonymized_seeds, &key, hash, &key);
190
191 /* ...otherwise check if we've already seen it in this context... */
192 if (!ret)
193 ret = hashmap_get_entry(map, &key, hash, &key);
194
195 /* ...and finally generate a new mapping if necessary */
196 if (!ret)
197 ret = add_anonymized_entry(map, key.hash.hash,
198 orig, len, generate());
199
200 return ret->anon;
201 }
202
203 /*
204 * We anonymize each component of a path individually,
205 * so that paths a/b and a/c will share a common root.
206 * The paths are cached via anonymize_mem so that repeated
207 * lookups for "a" will yield the same value.
208 */
209 static void anonymize_path(struct strbuf *out, const char *path,
210 struct hashmap *map,
211 char *(*generate)(void))
212 {
213 while (*path) {
214 const char *end_of_component = strchrnul(path, '/');
215 size_t len = end_of_component - path;
216 const char *c = anonymize_str(map, generate, path, len);
217 strbuf_addstr(out, c);
218 path = end_of_component;
219 if (*path)
220 strbuf_addch(out, *path++);
221 }
222 }
223
224 static inline void *mark_to_ptr(uint32_t mark)
225 {
226 return (void *)(uintptr_t)mark;
227 }
228
229 static inline uint32_t ptr_to_mark(void * mark)
230 {
231 return (uint32_t)(uintptr_t)mark;
232 }
233
234 static inline void mark_object(struct object *object, uint32_t mark)
235 {
236 add_decoration(&idnums, object, mark_to_ptr(mark));
237 }
238
239 static inline void mark_next_object(struct object *object)
240 {
241 mark_object(object, ++last_idnum);
242 }
243
244 static int get_object_mark(struct object *object)
245 {
246 void *decoration = lookup_decoration(&idnums, object);
247 if (!decoration)
248 return 0;
249 return ptr_to_mark(decoration);
250 }
251
252 static struct commit *rewrite_commit(struct commit *p)
253 {
254 for (;;) {
255 if (p->parents && p->parents->next)
256 break;
257 if (p->object.flags & UNINTERESTING)
258 break;
259 if (!(p->object.flags & TREESAME))
260 break;
261 if (!p->parents)
262 return NULL;
263 p = p->parents->item;
264 }
265 return p;
266 }
267
268 static void show_progress(void)
269 {
270 static int counter = 0;
271 if (!progress)
272 return;
273 if ((++counter % progress) == 0)
274 printf("progress %d objects\n", counter);
275 }
276
277 /*
278 * Ideally we would want some transformation of the blob data here
279 * that is unreversible, but would still be the same size and have
280 * the same data relationship to other blobs (so that we get the same
281 * delta and packing behavior as the original). But the first and last
282 * requirements there are probably mutually exclusive, so let's take
283 * the easy way out for now, and just generate arbitrary content.
284 *
285 * There's no need to cache this result with anonymize_mem, since
286 * we already handle blob content caching with marks.
287 */
288 static char *anonymize_blob(unsigned long *size)
289 {
290 static int counter;
291 struct strbuf out = STRBUF_INIT;
292 strbuf_addf(&out, "anonymous blob %d", counter++);
293 *size = out.len;
294 return strbuf_detach(&out, NULL);
295 }
296
297 static void export_blob(const struct object_id *oid)
298 {
299 unsigned long size;
300 enum object_type type;
301 char *buf;
302 struct object *object;
303 int eaten;
304
305 if (no_data)
306 return;
307
308 if (is_null_oid(oid))
309 return;
310
311 object = lookup_object(the_repository, oid);
312 if (object && object->flags & SHOWN)
313 return;
314
315 if (anonymize) {
316 buf = anonymize_blob(&size);
317 object = (struct object *)lookup_blob(the_repository, oid);
318 eaten = 0;
319 } else {
320 buf = odb_read_object(the_repository->objects, oid, &type, &size);
321 if (!buf)
322 die(_("could not read blob %s"), oid_to_hex(oid));
323 if (check_object_signature(the_repository, oid, buf, size,
324 type) < 0)
325 die(_("oid mismatch in blob %s"), oid_to_hex(oid));
326 object = parse_object_buffer(the_repository, oid, type,
327 size, buf, &eaten);
328 }
329
330 if (!object)
331 die(_("could not read blob %s"), oid_to_hex(oid));
332
333 mark_next_object(object);
334
335 printf("blob\nmark :%"PRIu32"\n", last_idnum);
336 if (show_original_ids)
337 printf("original-oid %s\n", oid_to_hex(oid));
338 printf("data %"PRIuMAX"\n", (uintmax_t)size);
339 if (size && fwrite(buf, size, 1, stdout) != 1)
340 die_errno(_("could not write blob '%s'"), oid_to_hex(oid));
341 printf("\n");
342
343 show_progress();
344
345 object->flags |= SHOWN;
346 if (!eaten)
347 free(buf);
348 }
349
350 static int depth_first(const void *a_, const void *b_)
351 {
352 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
353 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
354 const char *name_a, *name_b;
355 int len_a, len_b, len;
356 int cmp;
357
358 name_a = a->one ? a->one->path : a->two->path;
359 name_b = b->one ? b->one->path : b->two->path;
360
361 len_a = strlen(name_a);
362 len_b = strlen(name_b);
363 len = (len_a < len_b) ? len_a : len_b;
364
365 /* strcmp will sort 'd' before 'd/e', we want 'd/e' before 'd' */
366 cmp = memcmp(name_a, name_b, len);
367 if (cmp)
368 return cmp;
369 cmp = len_b - len_a;
370 if (cmp)
371 return cmp;
372 /*
373 * Move 'R'ename entries last so that all references of the file
374 * appear in the output before it is renamed (e.g., when a file
375 * was copied and renamed in the same commit).
376 */
377 return (a->status == 'R') - (b->status == 'R');
378 }
379
380 static void print_path_1(const char *path)
381 {
382 int need_quote = quote_c_style(path, NULL, NULL, 0);
383 if (need_quote)
384 quote_c_style(path, NULL, stdout, 0);
385 else if (strchr(path, ' '))
386 printf("\"%s\"", path);
387 else
388 printf("%s", path);
389 }
390
391 static char *anonymize_path_component(void)
392 {
393 static int counter;
394 struct strbuf out = STRBUF_INIT;
395 strbuf_addf(&out, "path%d", counter++);
396 return strbuf_detach(&out, NULL);
397 }
398
399 static void print_path(const char *path)
400 {
401 if (!anonymize)
402 print_path_1(path);
403 else {
404 static struct hashmap paths;
405 static struct strbuf anon = STRBUF_INIT;
406
407 anonymize_path(&anon, path, &paths, anonymize_path_component);
408 print_path_1(anon.buf);
409 strbuf_reset(&anon);
410 }
411 }
412
413 static char *generate_fake_oid(void)
414 {
415 static uint32_t counter = 1; /* avoid null oid */
416 const unsigned hashsz = the_hash_algo->rawsz;
417 struct object_id oid;
418 char *hex = xmallocz(GIT_MAX_HEXSZ);
419
420 oidclr(&oid, the_repository->hash_algo);
421 put_be32(oid.hash + hashsz - 4, counter++);
422 return oid_to_hex_r(hex, &oid);
423 }
424
425 static const char *anonymize_oid(const char *oid_hex)
426 {
427 static struct hashmap objs;
428 size_t len = strlen(oid_hex);
429 return anonymize_str(&objs, generate_fake_oid, oid_hex, len);
430 }
431
432 static void show_filemodify(struct diff_queue_struct *q,
433 struct diff_options *options UNUSED, void *data)
434 {
435 int i;
436 struct string_list *changed = data;
437
438 /*
439 * Handle files below a directory first, in case they are all deleted
440 * and the directory changes to a file or symlink.
441 */
442 QSORT(q->queue, q->nr, depth_first);
443
444 for (i = 0; i < q->nr; i++) {
445 struct diff_filespec *ospec = q->queue[i]->one;
446 struct diff_filespec *spec = q->queue[i]->two;
447
448 switch (q->queue[i]->status) {
449 case DIFF_STATUS_DELETED:
450 printf("D ");
451 print_path(spec->path);
452 string_list_insert(changed, spec->path);
453 putchar('\n');
454 break;
455
456 case DIFF_STATUS_COPIED:
457 case DIFF_STATUS_RENAMED:
458 /*
459 * If a change in the file corresponding to ospec->path
460 * has been observed, we cannot trust its contents
461 * because the diff is calculated based on the prior
462 * contents, not the current contents. So, declare a
463 * copy or rename only if there was no change observed.
464 */
465 if (!string_list_has_string(changed, ospec->path)) {
466 printf("%c ", q->queue[i]->status);
467 print_path(ospec->path);
468 putchar(' ');
469 print_path(spec->path);
470 string_list_insert(changed, spec->path);
471 putchar('\n');
472
473 if (oideq(&ospec->oid, &spec->oid) &&
474 ospec->mode == spec->mode)
475 break;
476 }
477 /* fallthrough */
478
479 case DIFF_STATUS_TYPE_CHANGED:
480 case DIFF_STATUS_MODIFIED:
481 case DIFF_STATUS_ADDED:
482 /*
483 * Links refer to objects in another repositories;
484 * output the SHA-1 verbatim.
485 */
486 if (no_data || S_ISGITLINK(spec->mode))
487 printf("M %06o %s ", spec->mode,
488 anonymize ?
489 anonymize_oid(oid_to_hex(&spec->oid)) :
490 oid_to_hex(&spec->oid));
491 else {
492 struct object *object = lookup_object(the_repository,
493 &spec->oid);
494 printf("M %06o :%d ", spec->mode,
495 get_object_mark(object));
496 }
497 print_path(spec->path);
498 string_list_insert(changed, spec->path);
499 putchar('\n');
500 break;
501
502 default:
503 die(_("unexpected comparison status '%c' for %s, %s"),
504 q->queue[i]->status,
505 ospec->path ? ospec->path : _("none"),
506 spec->path ? spec->path : _("none"));
507 }
508 }
509 }
510
511 static char *anonymize_ref_component(void)
512 {
513 static int counter;
514 struct strbuf out = STRBUF_INIT;
515 strbuf_addf(&out, "ref%d", counter++);
516 return strbuf_detach(&out, NULL);
517 }
518
519 static const char *anonymize_refname(const char *refname)
520 {
521 /*
522 * If any of these prefixes is found, we will leave it intact
523 * so that tags remain tags and so forth.
524 */
525 static const char *prefixes[] = {
526 "refs/heads/",
527 "refs/tags/",
528 "refs/remotes/",
529 "refs/"
530 };
531 static struct hashmap refs;
532 static struct strbuf anon = STRBUF_INIT;
533 int i;
534
535 strbuf_reset(&anon);
536 for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
537 if (skip_prefix(refname, prefixes[i], &refname)) {
538 strbuf_addstr(&anon, prefixes[i]);
539 break;
540 }
541 }
542
543 anonymize_path(&anon, refname, &refs, anonymize_ref_component);
544 return anon.buf;
545 }
546
547 /*
548 * We do not even bother to cache commit messages, as they are unlikely
549 * to be repeated verbatim, and it is not that interesting when they are.
550 */
551 static char *anonymize_commit_message(void)
552 {
553 static int counter;
554 return xstrfmt("subject %d\n\nbody\n", counter++);
555 }
556
557 static char *anonymize_ident(void)
558 {
559 static int counter;
560 struct strbuf out = STRBUF_INIT;
561 strbuf_addf(&out, "User %d <user%d@example.com>", counter, counter);
562 counter++;
563 return strbuf_detach(&out, NULL);
564 }
565
566 /*
567 * Our strategy here is to anonymize the names and email addresses,
568 * but keep timestamps intact, as they influence things like traversal
569 * order (and by themselves should not be too revealing).
570 */
571 static void anonymize_ident_line(const char **beg, const char **end)
572 {
573 static struct hashmap idents;
574 static struct strbuf buffers[] = { STRBUF_INIT, STRBUF_INIT };
575 static unsigned which_buffer;
576
577 struct strbuf *out;
578 struct ident_split split;
579 const char *end_of_header;
580
581 out = &buffers[which_buffer++];
582 which_buffer %= ARRAY_SIZE(buffers);
583 strbuf_reset(out);
584
585 /* skip "committer", "author", "tagger", etc */
586 end_of_header = strchr(*beg, ' ');
587 if (!end_of_header)
588 BUG("malformed line fed to anonymize_ident_line: %.*s",
589 (int)(*end - *beg), *beg);
590 end_of_header++;
591 strbuf_add(out, *beg, end_of_header - *beg);
592
593 if (!split_ident_line(&split, end_of_header, *end - end_of_header) &&
594 split.date_begin) {
595 const char *ident;
596 size_t len;
597
598 len = split.mail_end - split.name_begin;
599 ident = anonymize_str(&idents, anonymize_ident,
600 split.name_begin, len);
601 strbuf_addstr(out, ident);
602 strbuf_addch(out, ' ');
603 strbuf_add(out, split.date_begin, split.tz_end - split.date_begin);
604 } else {
605 strbuf_addstr(out, "Malformed Ident <malformed@example.com> 0 -0000");
606 }
607
608 *beg = out->buf;
609 *end = out->buf + out->len;
610 }
611
612 /*
613 * find_commit_multiline_header is similar to find_commit_header,
614 * except that it handles multi-line headers, rather than simply
615 * returning the first line of the header.
616 *
617 * The returned string has had the ' ' line continuation markers
618 * removed, and points to allocated memory that must be free()d (not
619 * to memory within 'msg').
620 *
621 * If the header is found, then *end is set to point at the '\n' in
622 * msg that immediately follows the header value.
623 */
624 static const char *find_commit_multiline_header(const char *msg,
625 const char *key,
626 const char **end)
627 {
628 struct strbuf val = STRBUF_INIT;
629 const char *bol, *eol;
630 size_t len;
631
632 bol = find_commit_header(msg, key, &len);
633 if (!bol)
634 return NULL;
635 eol = bol + len;
636 strbuf_add(&val, bol, len);
637
638 while (eol[0] == '\n' && eol[1] == ' ') {
639 bol = eol + 2;
640 eol = strchrnul(bol, '\n');
641 strbuf_addch(&val, '\n');
642 strbuf_add(&val, bol, eol - bol);
643 }
644
645 *end = eol;
646 return strbuf_detach(&val, NULL);
647 }
648
649 static void print_signature(const char *signature, const char *object_hash)
650 {
651 if (!signature)
652 return;
653
654 printf("gpgsig %s %s\ndata %u\n%s\n",
655 object_hash,
656 get_signature_format(signature),
657 (unsigned)strlen(signature),
658 signature);
659 }
660
661 static const char *append_signatures_for_header(struct string_list *signatures,
662 const char *pos,
663 const char *header,
664 const char *object_hash)
665 {
666 const char *signature;
667 const char *start = pos;
668 const char *end = pos;
669
670 while ((signature = find_commit_multiline_header(start + 1,
671 header,
672 &end))) {
673 string_list_append(signatures, signature)->util = (void *)object_hash;
674 free((char *)signature);
675 start = end;
676 }
677
678 return end;
679 }
680
681 static void handle_commit(struct commit *commit, struct rev_info *rev,
682 struct string_list *paths_of_changed_objects)
683 {
684 int saved_output_format = rev->diffopt.output_format;
685 const char *commit_buffer, *commit_buffer_cursor;
686 const char *author, *author_end, *committer, *committer_end;
687 const char *encoding = NULL;
688 size_t encoding_len;
689 struct string_list signatures = STRING_LIST_INIT_DUP;
690 const char *message;
691 char *reencoded = NULL;
692 struct commit_list *p;
693 const char *refname;
694 int i;
695
696 rev->diffopt.output_format = DIFF_FORMAT_CALLBACK;
697
698 parse_commit_or_die(commit);
699 commit_buffer_cursor = commit_buffer = repo_get_commit_buffer(the_repository, commit, NULL);
700
701 author = strstr(commit_buffer_cursor, "\nauthor ");
702 if (!author)
703 die(_("could not find author in commit %s"),
704 oid_to_hex(&commit->object.oid));
705 author++;
706 commit_buffer_cursor = author_end = strchrnul(author, '\n');
707
708 committer = strstr(commit_buffer_cursor, "\ncommitter ");
709 if (!committer)
710 die(_("could not find committer in commit %s"),
711 oid_to_hex(&commit->object.oid));
712 committer++;
713 commit_buffer_cursor = committer_end = strchrnul(committer, '\n');
714
715 /*
716 * find_commit_header() and find_commit_multiline_header() get
717 * a `+ 1` because commit_buffer_cursor points at the trailing
718 * "\n" at the end of the previous line, but they want a
719 * pointer to the beginning of the next line.
720 */
721
722 if (*commit_buffer_cursor == '\n') {
723 encoding = find_commit_header(commit_buffer_cursor + 1, "encoding", &encoding_len);
724 if (encoding)
725 commit_buffer_cursor = encoding + encoding_len;
726 }
727
728 if (*commit_buffer_cursor == '\n') {
729 const char *after_sha1 = append_signatures_for_header(&signatures, commit_buffer_cursor,
730 "gpgsig", "sha1");
731 const char *after_sha256 = append_signatures_for_header(&signatures, commit_buffer_cursor,
732 "gpgsig-sha256", "sha256");
733 commit_buffer_cursor = (after_sha1 > after_sha256) ? after_sha1 : after_sha256;
734 }
735
736 message = strstr(commit_buffer_cursor, "\n\n");
737 if (message)
738 message += 2;
739
740 if (commit->parents &&
741 (get_object_mark(&commit->parents->item->object) != 0 ||
742 reference_excluded_commits) &&
743 !full_tree) {
744 parse_commit_or_die(commit->parents->item);
745 diff_tree_oid(get_commit_tree_oid(commit->parents->item),
746 get_commit_tree_oid(commit), "", &rev->diffopt);
747 }
748 else
749 diff_root_tree_oid(get_commit_tree_oid(commit),
750 "", &rev->diffopt);
751
752 /* Export the referenced blobs, and remember the marks. */
753 for (i = 0; i < diff_queued_diff.nr; i++)
754 if (!S_ISGITLINK(diff_queued_diff.queue[i]->two->mode))
755 export_blob(&diff_queued_diff.queue[i]->two->oid);
756
757 refname = *revision_sources_at(&revision_sources, commit);
758 /*
759 * FIXME: string_list_remove() below for each ref is overall
760 * O(N^2). Compared to a history walk and diffing trees, this is
761 * just lost in the noise in practice. However, theoretically a
762 * repo may have enough refs for this to become slow.
763 */
764 string_list_remove(&extra_refs, refname, 0);
765 if (anonymize) {
766 refname = anonymize_refname(refname);
767 anonymize_ident_line(&committer, &committer_end);
768 anonymize_ident_line(&author, &author_end);
769 }
770
771 mark_next_object(&commit->object);
772 if (anonymize) {
773 reencoded = anonymize_commit_message();
774 } else if (encoding) {
775 char *buf;
776 switch (reencode_mode) {
777 case REENCODE_YES:
778 buf = xstrfmt("%.*s", (int)encoding_len, encoding);
779 reencoded = reencode_string(message, "UTF-8", buf);
780 free(buf);
781 break;
782 case REENCODE_NO:
783 break;
784 case REENCODE_ABORT:
785 die(_("encountered commit-specific encoding %.*s in commit "
786 "%s; use --reencode=[yes|no] to handle it"),
787 (int)encoding_len, encoding,
788 oid_to_hex(&commit->object.oid));
789 }
790 }
791 if (!commit->parents)
792 printf("reset %s\n", refname);
793 printf("commit %s\nmark :%"PRIu32"\n", refname, last_idnum);
794 if (show_original_ids)
795 printf("original-oid %s\n", oid_to_hex(&commit->object.oid));
796 printf("%.*s\n%.*s\n",
797 (int)(author_end - author), author,
798 (int)(committer_end - committer), committer);
799 if (signatures.nr) {
800 switch (signed_commit_mode) {
801 /* Exporting modes */
802 case SIGN_WARN_VERBATIM:
803 warning(_("exporting %"PRIuMAX" signature(s) for commit %s"),
804 (uintmax_t)signatures.nr, oid_to_hex(&commit->object.oid));
805 /* fallthru */
806 case SIGN_VERBATIM:
807 for (size_t i = 0; i < signatures.nr; i++) {
808 struct string_list_item *item = &signatures.items[i];
809 print_signature(item->string, item->util);
810 }
811 break;
812
813 /* Stripping modes */
814 case SIGN_WARN_STRIP:
815 warning(_("stripping signature(s) from commit %s"),
816 oid_to_hex(&commit->object.oid));
817 /* fallthru */
818 case SIGN_STRIP:
819 break;
820
821 /* Aborting modes */
822 case SIGN_ABORT:
823 die(_("encountered signed commit %s; use "
824 "--signed-commits=<mode> to handle it"),
825 oid_to_hex(&commit->object.oid));
826 default:
827 BUG("invalid signed_commit_mode value %d", signed_commit_mode);
828 }
829 string_list_clear(&signatures, 0);
830 }
831 if (!reencoded && encoding)
832 printf("encoding %.*s\n", (int)encoding_len, encoding);
833 printf("data %u\n%s",
834 (unsigned)(reencoded
835 ? strlen(reencoded) : message
836 ? strlen(message) : 0),
837 reencoded ? reencoded : message ? message : "");
838 free(reencoded);
839 repo_unuse_commit_buffer(the_repository, commit, commit_buffer);
840
841 for (i = 0, p = commit->parents; p; p = p->next) {
842 struct object *obj = &p->item->object;
843 int mark = get_object_mark(obj);
844
845 if (!mark && !reference_excluded_commits)
846 continue;
847 if (i == 0)
848 printf("from ");
849 else
850 printf("merge ");
851 if (mark)
852 printf(":%d\n", mark);
853 else
854 printf("%s\n",
855 anonymize ?
856 anonymize_oid(oid_to_hex(&obj->oid)) :
857 oid_to_hex(&obj->oid));
858 i++;
859 }
860
861 if (full_tree)
862 printf("deleteall\n");
863 log_tree_diff_flush(rev);
864 string_list_clear(paths_of_changed_objects, 0);
865 rev->diffopt.output_format = saved_output_format;
866
867 printf("\n");
868
869 show_progress();
870 }
871
872 static char *anonymize_tag(void)
873 {
874 static int counter;
875 struct strbuf out = STRBUF_INIT;
876 strbuf_addf(&out, "tag message %d", counter++);
877 return strbuf_detach(&out, NULL);
878 }
879
880
881 static void handle_tag(const char *name, struct tag *tag)
882 {
883 unsigned long size;
884 enum object_type type;
885 char *buf;
886 const char *tagger, *tagger_end, *message;
887 size_t message_size = 0;
888 struct object *tagged;
889 int tagged_mark;
890 struct commit *p;
891
892 /* Trees have no identifier in fast-export output, thus we have no way
893 * to output tags of trees, tags of tags of trees, etc. Simply omit
894 * such tags.
895 */
896 tagged = tag->tagged;
897 while (tagged->type == OBJ_TAG) {
898 tagged = ((struct tag *)tagged)->tagged;
899 }
900 if (tagged->type == OBJ_TREE) {
901 warning(_("omitting tag %s,\nsince tags of trees (or tags "
902 "of tags of trees, etc.) are not supported."),
903 oid_to_hex(&tag->object.oid));
904 return;
905 }
906
907 buf = odb_read_object(the_repository->objects, &tag->object.oid,
908 &type, &size);
909 if (!buf)
910 die(_("could not read tag %s"), oid_to_hex(&tag->object.oid));
911 message = memmem(buf, size, "\n\n", 2);
912 if (message) {
913 message += 2;
914 message_size = strlen(message);
915 }
916 tagger = memmem(buf, message ? message - buf : size, "\ntagger ", 8);
917 if (!tagger) {
918 if (fake_missing_tagger)
919 tagger = "tagger Unspecified Tagger "
920 "<unspecified-tagger> 0 +0000";
921 else
922 tagger = "";
923 tagger_end = tagger + strlen(tagger);
924 } else {
925 tagger++;
926 tagger_end = strchrnul(tagger, '\n');
927 if (anonymize)
928 anonymize_ident_line(&tagger, &tagger_end);
929 }
930
931 if (anonymize) {
932 name = anonymize_refname(name);
933 if (message) {
934 static struct hashmap tags;
935 message = anonymize_str(&tags, anonymize_tag,
936 message, message_size);
937 message_size = strlen(message);
938 }
939 }
940
941 /* handle signed tags */
942 if (message) {
943 size_t sig_offset = parse_signed_buffer(message, message_size);
944 if (sig_offset < message_size)
945 switch (signed_tag_mode) {
946 /* Exporting modes */
947 case SIGN_WARN_VERBATIM:
948 warning(_("exporting signed tag %s"),
949 oid_to_hex(&tag->object.oid));
950 /* fallthru */
951 case SIGN_VERBATIM:
952 break;
953
954 /* Stripping modes */
955 case SIGN_WARN_STRIP:
956 warning(_("stripping signature from tag %s"),
957 oid_to_hex(&tag->object.oid));
958 /* fallthru */
959 case SIGN_STRIP:
960 message_size = sig_offset;
961 break;
962
963 /* Aborting modes */
964 case SIGN_ABORT:
965 die(_("encountered signed tag %s; use "
966 "--signed-tags=<mode> to handle it"),
967 oid_to_hex(&tag->object.oid));
968 default:
969 BUG("invalid signed_commit_mode value %d", signed_commit_mode);
970 }
971 }
972
973 /* handle tag->tagged having been filtered out due to paths specified */
974 tagged = tag->tagged;
975 tagged_mark = get_object_mark(tagged);
976 if (!tagged_mark) {
977 switch (tag_of_filtered_mode) {
978 case TAG_FILTERING_ABORT:
979 die(_("tag %s tags unexported object; use "
980 "--tag-of-filtered-object=<mode> to handle it"),
981 oid_to_hex(&tag->object.oid));
982 case DROP:
983 /* Ignore this tag altogether */
984 free(buf);
985 return;
986 case REWRITE:
987 if (tagged->type == OBJ_TAG && !mark_tags) {
988 die(_("cannot export nested tags unless --mark-tags is specified."));
989 } else if (tagged->type == OBJ_COMMIT) {
990 p = rewrite_commit((struct commit *)tagged);
991 if (!p) {
992 printf("reset %s\nfrom %s\n\n",
993 name, oid_to_hex(null_oid(the_hash_algo)));
994 free(buf);
995 return;
996 }
997 tagged_mark = get_object_mark(&p->object);
998 } else {
999 /* tagged->type is either OBJ_BLOB or OBJ_TAG */
1000 tagged_mark = get_object_mark(tagged);
1001 }
1002 }
1003 }
1004
1005 if (tagged->type == OBJ_TAG) {
1006 printf("reset %s\nfrom %s\n\n",
1007 name, oid_to_hex(null_oid(the_hash_algo)));
1008 }
1009 skip_prefix(name, "refs/tags/", &name);
1010 printf("tag %s\n", name);
1011 if (mark_tags) {
1012 mark_next_object(&tag->object);
1013 printf("mark :%"PRIu32"\n", last_idnum);
1014 }
1015 if (tagged_mark)
1016 printf("from :%d\n", tagged_mark);
1017 else
1018 printf("from %s\n", oid_to_hex(&tagged->oid));
1019
1020 if (show_original_ids)
1021 printf("original-oid %s\n", oid_to_hex(&tag->object.oid));
1022 printf("%.*s%sdata %d\n%.*s\n",
1023 (int)(tagger_end - tagger), tagger,
1024 tagger == tagger_end ? "" : "\n",
1025 (int)message_size, (int)message_size, message ? message : "");
1026 free(buf);
1027 }
1028
1029 static struct commit *get_commit(struct rev_cmdline_entry *e, const char *full_name)
1030 {
1031 switch (e->item->type) {
1032 case OBJ_COMMIT:
1033 return (struct commit *)e->item;
1034 case OBJ_TAG: {
1035 struct tag *tag = (struct tag *)e->item;
1036
1037 /* handle nested tags */
1038 while (tag && tag->object.type == OBJ_TAG) {
1039 parse_object(the_repository, &tag->object.oid);
1040 string_list_append(&tag_refs, full_name)->util = tag;
1041 tag = (struct tag *)tag->tagged;
1042 }
1043 if (!tag)
1044 die(_("tag %s points nowhere?"), e->name);
1045 return (struct commit *)tag;
1046 }
1047 default:
1048 return NULL;
1049 }
1050 }
1051
1052 static void get_tags_and_duplicates(struct rev_cmdline_info *info)
1053 {
1054 int i;
1055
1056 for (i = 0; i < info->nr; i++) {
1057 struct rev_cmdline_entry *e = info->rev + i;
1058 struct object_id oid;
1059 struct commit *commit;
1060 char *full_name = NULL;
1061
1062 if (e->flags & UNINTERESTING)
1063 continue;
1064
1065 if (repo_dwim_ref(the_repository, e->name, strlen(e->name),
1066 &oid, &full_name, 0) != 1) {
1067 free(full_name);
1068 continue;
1069 }
1070
1071 if (refspecs.nr) {
1072 char *private;
1073 private = apply_refspecs(&refspecs, full_name);
1074 if (private) {
1075 free(full_name);
1076 full_name = private;
1077 }
1078 }
1079
1080 commit = get_commit(e, full_name);
1081 if (!commit) {
1082 warning(_("%s: unexpected object of type %s, skipping."),
1083 e->name,
1084 type_name(e->item->type));
1085 free(full_name);
1086 continue;
1087 }
1088
1089 switch (commit->object.type) {
1090 case OBJ_COMMIT:
1091 break;
1092 case OBJ_BLOB:
1093 export_blob(&commit->object.oid);
1094 free(full_name);
1095 continue;
1096 default: /* OBJ_TAG (nested tags) is already handled */
1097 warning(_("tag points to object of unexpected type %s, skipping."),
1098 type_name(commit->object.type));
1099 free(full_name);
1100 continue;
1101 }
1102
1103 /*
1104 * Make sure this ref gets properly updated eventually, whether
1105 * through a commit or manually at the end.
1106 */
1107 if (e->item->type != OBJ_TAG)
1108 string_list_append(&extra_refs, full_name)->util = commit;
1109
1110 if (!*revision_sources_at(&revision_sources, commit))
1111 *revision_sources_at(&revision_sources, commit) = full_name;
1112 else
1113 free(full_name);
1114 }
1115
1116 string_list_sort_u(&extra_refs, 0);
1117 }
1118
1119 static void handle_tags_and_duplicates(struct string_list *extras)
1120 {
1121 struct commit *commit;
1122 int i;
1123
1124 for (i = extras->nr - 1; i >= 0; i--) {
1125 const char *name = extras->items[i].string;
1126 struct object *object = extras->items[i].util;
1127 int mark;
1128
1129 switch (object->type) {
1130 case OBJ_TAG:
1131 handle_tag(name, (struct tag *)object);
1132 break;
1133 case OBJ_COMMIT:
1134 if (anonymize)
1135 name = anonymize_refname(name);
1136 /* create refs pointing to already seen commits */
1137 commit = rewrite_commit((struct commit *)object);
1138 if (!commit) {
1139 /*
1140 * Neither this object nor any of its
1141 * ancestors touch any relevant paths, so
1142 * it has been filtered to nothing. Delete
1143 * it.
1144 */
1145 printf("reset %s\nfrom %s\n\n",
1146 name, oid_to_hex(null_oid(the_hash_algo)));
1147 continue;
1148 }
1149
1150 mark = get_object_mark(&commit->object);
1151 if (!mark) {
1152 /*
1153 * Getting here means we have a commit which
1154 * was excluded by a negative refspec (e.g.
1155 * fast-export ^HEAD HEAD). If we are
1156 * referencing excluded commits, set the ref
1157 * to the exact commit. Otherwise, the user
1158 * wants the branch exported but every commit
1159 * in its history to be deleted, which basically
1160 * just means deletion of the ref.
1161 */
1162 if (!reference_excluded_commits) {
1163 /* delete the ref */
1164 printf("reset %s\nfrom %s\n\n",
1165 name, oid_to_hex(null_oid(the_hash_algo)));
1166 continue;
1167 }
1168 /* set ref to commit using oid, not mark */
1169 printf("reset %s\nfrom %s\n\n", name,
1170 oid_to_hex(&commit->object.oid));
1171 continue;
1172 }
1173
1174 printf("reset %s\nfrom :%d\n\n", name, mark
1175 );
1176 show_progress();
1177 break;
1178 }
1179 }
1180 }
1181
1182 static void export_marks(char *file)
1183 {
1184 unsigned int i;
1185 uint32_t mark;
1186 struct decoration_entry *deco = idnums.entries;
1187 FILE *f;
1188 int e = 0;
1189
1190 f = fopen_for_writing(file);
1191 if (!f)
1192 die_errno(_("unable to open marks file %s for writing."), file);
1193
1194 for (i = 0; i < idnums.size; i++) {
1195 if (deco->base && deco->base->type == 1) {
1196 mark = ptr_to_mark(deco->decoration);
1197 if (fprintf(f, ":%"PRIu32" %s\n", mark,
1198 oid_to_hex(&deco->base->oid)) < 0) {
1199 e = 1;
1200 break;
1201 }
1202 }
1203 deco++;
1204 }
1205
1206 e |= ferror(f);
1207 e |= fclose(f);
1208 if (e)
1209 error(_("unable to write marks file %s."), file);
1210 }
1211
1212 static void import_marks(char *input_file, int check_exists)
1213 {
1214 char line[512];
1215 FILE *f;
1216 struct stat sb;
1217
1218 if (check_exists && stat(input_file, &sb))
1219 return;
1220
1221 f = xfopen(input_file, "r");
1222 while (fgets(line, sizeof(line), f)) {
1223 uint32_t mark;
1224 char *line_end, *mark_end;
1225 struct object_id oid;
1226 struct object *object;
1227 struct commit *commit;
1228 enum object_type type;
1229
1230 line_end = strchr(line, '\n');
1231 if (line[0] != ':' || !line_end)
1232 die(_("corrupt mark line: %s"), line);
1233 *line_end = '\0';
1234
1235 mark = strtoumax(line + 1, &mark_end, 10);
1236 if (!mark || mark_end == line + 1
1237 || *mark_end != ' ' || get_oid_hex(mark_end + 1, &oid))
1238 die(_("corrupt mark line: %s"), line);
1239
1240 if (last_idnum < mark)
1241 last_idnum = mark;
1242
1243 type = odb_read_object_info(the_repository->objects, &oid, NULL);
1244 if (type < 0)
1245 die(_("object not found: %s"), oid_to_hex(&oid));
1246
1247 if (type != OBJ_COMMIT)
1248 /* only commits */
1249 continue;
1250
1251 commit = lookup_commit(the_repository, &oid);
1252 if (!commit)
1253 die(_("not a commit? can't happen: %s"), oid_to_hex(&oid));
1254
1255 object = &commit->object;
1256
1257 if (object->flags & SHOWN)
1258 error(_("object %s already has a mark"), oid_to_hex(&oid));
1259
1260 mark_object(object, mark);
1261
1262 object->flags |= SHOWN;
1263 }
1264 fclose(f);
1265 }
1266
1267 static void handle_deletes(void)
1268 {
1269 int i;
1270 for (i = 0; i < refspecs.nr; i++) {
1271 struct refspec_item *refspec = &refspecs.items[i];
1272 if (*refspec->src)
1273 continue;
1274
1275 printf("reset %s\nfrom %s\n\n",
1276 refspec->dst, oid_to_hex(null_oid(the_hash_algo)));
1277 }
1278 }
1279
1280 static int parse_opt_anonymize_map(const struct option *opt,
1281 const char *arg, int unset)
1282 {
1283 struct hashmap *map = opt->value;
1284 const char *delim, *value;
1285 size_t keylen;
1286
1287 BUG_ON_OPT_NEG(unset);
1288
1289 delim = strchr(arg, ':');
1290 if (delim) {
1291 keylen = delim - arg;
1292 value = delim + 1;
1293 } else {
1294 keylen = strlen(arg);
1295 value = arg;
1296 }
1297
1298 if (!keylen || !*value)
1299 return error(_("--anonymize-map token cannot be empty"));
1300
1301 add_anonymized_entry(map, memhash(arg, keylen), arg, keylen,
1302 xstrdup(value));
1303
1304 return 0;
1305 }
1306
1307 int cmd_fast_export(int argc,
1308 const char **argv,
1309 const char *prefix,
1310 struct repository *repo UNUSED)
1311 {
1312 struct rev_info revs;
1313 struct commit *commit;
1314 char *export_filename = NULL,
1315 *import_filename = NULL,
1316 *import_filename_if_exists = NULL;
1317 uint32_t lastimportid;
1318 struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
1319 struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
1320 struct option options[] = {
1321 OPT_INTEGER(0, "progress", &progress,
1322 N_("show progress after <n> objects")),
1323 OPT_CALLBACK(0, "signed-tags", &signed_tag_mode, N_("mode"),
1324 N_("select handling of signed tags"),
1325 parse_opt_sign_mode),
1326 OPT_CALLBACK(0, "signed-commits", &signed_commit_mode, N_("mode"),
1327 N_("select handling of signed commits"),
1328 parse_opt_sign_mode),
1329 OPT_CALLBACK(0, "tag-of-filtered-object", &tag_of_filtered_mode, N_("mode"),
1330 N_("select handling of tags that tag filtered objects"),
1331 parse_opt_tag_of_filtered_mode),
1332 OPT_CALLBACK(0, "reencode", &reencode_mode, N_("mode"),
1333 N_("select handling of commit messages in an alternate encoding"),
1334 parse_opt_reencode_mode),
1335 OPT_STRING(0, "export-marks", &export_filename, N_("file"),
1336 N_("dump marks to this file")),
1337 OPT_STRING(0, "import-marks", &import_filename, N_("file"),
1338 N_("import marks from this file")),
1339 OPT_STRING(0, "import-marks-if-exists",
1340 &import_filename_if_exists,
1341 N_("file"),
1342 N_("import marks from this file if it exists")),
1343 OPT_BOOL(0, "fake-missing-tagger", &fake_missing_tagger,
1344 N_("fake a tagger when tags lack one")),
1345 OPT_BOOL(0, "full-tree", &full_tree,
1346 N_("output full tree for each commit")),
1347 OPT_BOOL(0, "use-done-feature", &use_done_feature,
1348 N_("use the done feature to terminate the stream")),
1349 OPT_BOOL(0, "no-data", &no_data, N_("skip output of blob data")),
1350 OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
1351 N_("apply refspec to exported refs")),
1352 OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
1353 OPT_CALLBACK_F(0, "anonymize-map", &anonymized_seeds, N_("from:to"),
1354 N_("convert <from> to <to> in anonymized output"),
1355 PARSE_OPT_NONEG, parse_opt_anonymize_map),
1356 OPT_BOOL(0, "reference-excluded-parents",
1357 &reference_excluded_commits, N_("reference parents which are not in fast-export stream by object id")),
1358 OPT_BOOL(0, "show-original-ids", &show_original_ids,
1359 N_("show original object ids of blobs/commits")),
1360 OPT_BOOL(0, "mark-tags", &mark_tags,
1361 N_("label tags with mark ids")),
1362
1363 OPT_END()
1364 };
1365
1366 if (argc == 1)
1367 usage_with_options (fast_export_usage, options);
1368
1369 /* we handle encodings */
1370 repo_config(the_repository, git_default_config, NULL);
1371
1372 repo_init_revisions(the_repository, &revs, prefix);
1373 init_revision_sources(&revision_sources);
1374 revs.topo_order = 1;
1375 revs.sources = &revision_sources;
1376 revs.rewrite_parents = 1;
1377 argc = parse_options(argc, argv, prefix, options, fast_export_usage,
1378 PARSE_OPT_KEEP_ARGV0 | PARSE_OPT_KEEP_UNKNOWN_OPT);
1379 argc = setup_revisions(argc, argv, &revs, NULL);
1380 if (argc > 1)
1381 usage_with_options (fast_export_usage, options);
1382
1383 if (anonymized_seeds.cmpfn && !anonymize)
1384 die(_("the option '%s' requires '%s'"), "--anonymize-map", "--anonymize");
1385
1386 if (refspecs_list.nr) {
1387 int i;
1388
1389 for (i = 0; i < refspecs_list.nr; i++)
1390 refspec_append(&refspecs, refspecs_list.items[i].string);
1391
1392 string_list_clear(&refspecs_list, 1);
1393 }
1394
1395 if (use_done_feature)
1396 printf("feature done\n");
1397
1398 if (import_filename && import_filename_if_exists)
1399 die(_("options '%s' and '%s' cannot be used together"), "--import-marks", "--import-marks-if-exists");
1400 if (import_filename)
1401 import_marks(import_filename, 0);
1402 else if (import_filename_if_exists)
1403 import_marks(import_filename_if_exists, 1);
1404 lastimportid = last_idnum;
1405
1406 if (import_filename && revs.prune_data.nr)
1407 full_tree = 1;
1408
1409 get_tags_and_duplicates(&revs.cmdline);
1410
1411 if (prepare_revision_walk(&revs))
1412 die(_("revision walk setup failed"));
1413
1414 revs.reverse = 1;
1415 revs.diffopt.format_callback = show_filemodify;
1416 revs.diffopt.format_callback_data = &paths_of_changed_objects;
1417 revs.diffopt.flags.recursive = 1;
1418
1419 revs.diffopt.no_free = 1;
1420 while ((commit = get_revision(&revs)))
1421 handle_commit(commit, &revs, &paths_of_changed_objects);
1422 revs.diffopt.no_free = 0;
1423
1424 handle_tags_and_duplicates(&extra_refs);
1425 handle_tags_and_duplicates(&tag_refs);
1426 handle_deletes();
1427
1428 if (export_filename && lastimportid != last_idnum)
1429 export_marks(export_filename);
1430
1431 if (use_done_feature)
1432 printf("done\n");
1433
1434 refspec_clear(&refspecs);
1435 release_revisions(&revs);
1436
1437 return 0;
1438 }