]> git.ipfire.org Git - thirdparty/git.git/blob - builtin/fast-import.c
4cd0b079b616879c9c8eaa848fc34260230b15a4
[thirdparty/git.git] / builtin / fast-import.c
1 #define USE_THE_REPOSITORY_VARIABLE
2 #define DISABLE_SIGN_COMPARE_WARNINGS
3
4 #include "builtin.h"
5 #include "abspath.h"
6 #include "environment.h"
7 #include "gettext.h"
8 #include "hex.h"
9 #include "config.h"
10 #include "lockfile.h"
11 #include "object.h"
12 #include "blob.h"
13 #include "tree.h"
14 #include "commit.h"
15 #include "delta.h"
16 #include "pack.h"
17 #include "path.h"
18 #include "read-cache-ll.h"
19 #include "refs.h"
20 #include "csum-file.h"
21 #include "quote.h"
22 #include "dir.h"
23 #include "run-command.h"
24 #include "packfile.h"
25 #include "object-file.h"
26 #include "object-name.h"
27 #include "odb.h"
28 #include "mem-pool.h"
29 #include "commit-reach.h"
30 #include "khash.h"
31 #include "date.h"
32 #include "gpg-interface.h"
33
34 #define PACK_ID_BITS 16
35 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
36 #define DEPTH_BITS 13
37 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
38
39 /*
40 * We abuse the setuid bit on directories to mean "do not delta".
41 */
42 #define NO_DELTA S_ISUID
43
44 /*
45 * The amount of additional space required in order to write an object into the
46 * current pack. This is the hash lengths at the end of the pack, plus the
47 * length of one object ID.
48 */
49 #define PACK_SIZE_THRESHOLD (the_hash_algo->rawsz * 3)
50
51 struct object_entry {
52 struct pack_idx_entry idx;
53 struct hashmap_entry ent;
54 uint32_t type : TYPE_BITS,
55 pack_id : PACK_ID_BITS,
56 depth : DEPTH_BITS;
57 };
58
59 static int object_entry_hashcmp(const void *map_data UNUSED,
60 const struct hashmap_entry *eptr,
61 const struct hashmap_entry *entry_or_key,
62 const void *keydata)
63 {
64 const struct object_id *oid = keydata;
65 const struct object_entry *e1, *e2;
66
67 e1 = container_of(eptr, const struct object_entry, ent);
68 if (oid)
69 return oidcmp(&e1->idx.oid, oid);
70
71 e2 = container_of(entry_or_key, const struct object_entry, ent);
72 return oidcmp(&e1->idx.oid, &e2->idx.oid);
73 }
74
75 struct object_entry_pool {
76 struct object_entry_pool *next_pool;
77 struct object_entry *next_free;
78 struct object_entry *end;
79 struct object_entry entries[FLEX_ARRAY]; /* more */
80 };
81
82 struct mark_set {
83 union {
84 struct object_id *oids[1024];
85 struct object_entry *marked[1024];
86 struct mark_set *sets[1024];
87 } data;
88 unsigned int shift;
89 };
90
91 struct last_object {
92 struct strbuf data;
93 off_t offset;
94 unsigned int depth;
95 unsigned no_swap : 1;
96 };
97
98 struct atom_str {
99 struct atom_str *next_atom;
100 unsigned short str_len;
101 char str_dat[FLEX_ARRAY]; /* more */
102 };
103
104 struct tree_content;
105 struct tree_entry {
106 struct tree_content *tree;
107 struct atom_str *name;
108 struct tree_entry_ms {
109 uint16_t mode;
110 struct object_id oid;
111 } versions[2];
112 };
113
114 struct tree_content {
115 unsigned int entry_capacity; /* must match avail_tree_content */
116 unsigned int entry_count;
117 unsigned int delta_depth;
118 struct tree_entry *entries[FLEX_ARRAY]; /* more */
119 };
120
121 struct avail_tree_content {
122 unsigned int entry_capacity; /* must match tree_content */
123 struct avail_tree_content *next_avail;
124 };
125
126 struct branch {
127 struct branch *table_next_branch;
128 struct branch *active_next_branch;
129 const char *name;
130 struct tree_entry branch_tree;
131 uintmax_t last_commit;
132 uintmax_t num_notes;
133 unsigned active : 1;
134 unsigned delete : 1;
135 unsigned pack_id : PACK_ID_BITS;
136 struct object_id oid;
137 };
138
139 struct tag {
140 struct tag *next_tag;
141 const char *name;
142 unsigned int pack_id;
143 struct object_id oid;
144 };
145
146 struct hash_list {
147 struct hash_list *next;
148 struct object_id oid;
149 };
150
151 typedef enum {
152 WHENSPEC_RAW = 1,
153 WHENSPEC_RAW_PERMISSIVE,
154 WHENSPEC_RFC2822,
155 WHENSPEC_NOW
156 } whenspec_type;
157
158 struct recent_command {
159 struct recent_command *prev;
160 struct recent_command *next;
161 char *buf;
162 };
163
164 typedef void (*mark_set_inserter_t)(struct mark_set **s, struct object_id *oid, uintmax_t mark);
165 typedef void (*each_mark_fn_t)(uintmax_t mark, void *obj, void *cbp);
166
167 /* Configured limits on output */
168 static unsigned long max_depth = 50;
169 static off_t max_packsize;
170 static int unpack_limit = 100;
171 static int force_update;
172
173 /* Stats and misc. counters */
174 static uintmax_t alloc_count;
175 static uintmax_t marks_set_count;
176 static uintmax_t object_count_by_type[1 << TYPE_BITS];
177 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
178 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
179 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
180 static unsigned long object_count;
181 static unsigned long branch_count;
182 static unsigned long branch_load_count;
183 static int failure;
184 static FILE *pack_edges;
185 static unsigned int show_stats = 1;
186 static unsigned int quiet;
187 static int global_argc;
188 static const char **global_argv;
189 static const char *global_prefix;
190
191 static enum sign_mode signed_tag_mode = SIGN_VERBATIM;
192 static enum sign_mode signed_commit_mode = SIGN_VERBATIM;
193
194 /* Memory pools */
195 static struct mem_pool fi_mem_pool = {
196 .block_alloc = 2*1024*1024 - sizeof(struct mp_block),
197 };
198
199 /* Atom management */
200 static unsigned int atom_table_sz = 4451;
201 static unsigned int atom_cnt;
202 static struct atom_str **atom_table;
203
204 /* The .pack file being generated */
205 static struct pack_idx_option pack_idx_opts;
206 static unsigned int pack_id;
207 static struct hashfile *pack_file;
208 static struct packed_git *pack_data;
209 static struct packed_git **all_packs;
210 static off_t pack_size;
211
212 /* Table of objects we've written. */
213 static unsigned int object_entry_alloc = 5000;
214 static struct object_entry_pool *blocks;
215 static struct hashmap object_table;
216 static struct mark_set *marks;
217 static char *export_marks_file;
218 static char *import_marks_file;
219 static int import_marks_file_from_stream;
220 static int import_marks_file_ignore_missing;
221 static int import_marks_file_done;
222 static int relative_marks_paths;
223
224 /* Our last blob */
225 static struct last_object last_blob = {
226 .data = STRBUF_INIT,
227 };
228
229 /* Tree management */
230 static unsigned int tree_entry_alloc = 1000;
231 static void *avail_tree_entry;
232 static unsigned int avail_tree_table_sz = 100;
233 static struct avail_tree_content **avail_tree_table;
234 static size_t tree_entry_allocd;
235 static struct strbuf old_tree = STRBUF_INIT;
236 static struct strbuf new_tree = STRBUF_INIT;
237
238 /* Branch data */
239 static unsigned long max_active_branches = 5;
240 static unsigned long cur_active_branches;
241 static unsigned long branch_table_sz = 1039;
242 static struct branch **branch_table;
243 static struct branch *active_branches;
244
245 /* Tag data */
246 static struct tag *first_tag;
247 static struct tag *last_tag;
248
249 /* Input stream parsing */
250 static whenspec_type whenspec = WHENSPEC_RAW;
251 static struct strbuf command_buf = STRBUF_INIT;
252 static int unread_command_buf;
253 static struct recent_command cmd_hist = {
254 .prev = &cmd_hist,
255 .next = &cmd_hist,
256 };
257 static struct recent_command *cmd_tail = &cmd_hist;
258 static struct recent_command *rc_free;
259 static unsigned int cmd_save = 100;
260 static uintmax_t next_mark;
261 static struct strbuf new_data = STRBUF_INIT;
262 static int seen_data_command;
263 static int require_explicit_termination;
264 static int allow_unsafe_features;
265
266 /* Signal handling */
267 static volatile sig_atomic_t checkpoint_requested;
268
269 /* Submodule marks */
270 static struct string_list sub_marks_from = STRING_LIST_INIT_DUP;
271 static struct string_list sub_marks_to = STRING_LIST_INIT_DUP;
272 static kh_oid_map_t *sub_oid_map;
273
274 /* Where to write output of cat-blob commands */
275 static int cat_blob_fd = STDOUT_FILENO;
276
277 static void parse_argv(void);
278 static void parse_get_mark(const char *p);
279 static void parse_cat_blob(const char *p);
280 static void parse_ls(const char *p, struct branch *b);
281
282 static void for_each_mark(struct mark_set *m, uintmax_t base, each_mark_fn_t callback, void *p)
283 {
284 uintmax_t k;
285 if (m->shift) {
286 for (k = 0; k < 1024; k++) {
287 if (m->data.sets[k])
288 for_each_mark(m->data.sets[k], base + (k << m->shift), callback, p);
289 }
290 } else {
291 for (k = 0; k < 1024; k++) {
292 if (m->data.marked[k])
293 callback(base + k, m->data.marked[k], p);
294 }
295 }
296 }
297
298 static void dump_marks_fn(uintmax_t mark, void *object, void *cbp) {
299 struct object_entry *e = object;
300 FILE *f = cbp;
301
302 fprintf(f, ":%" PRIuMAX " %s\n", mark, oid_to_hex(&e->idx.oid));
303 }
304
305 static void write_branch_report(FILE *rpt, struct branch *b)
306 {
307 fprintf(rpt, "%s:\n", b->name);
308
309 fprintf(rpt, " status :");
310 if (b->active)
311 fputs(" active", rpt);
312 if (b->branch_tree.tree)
313 fputs(" loaded", rpt);
314 if (is_null_oid(&b->branch_tree.versions[1].oid))
315 fputs(" dirty", rpt);
316 fputc('\n', rpt);
317
318 fprintf(rpt, " tip commit : %s\n", oid_to_hex(&b->oid));
319 fprintf(rpt, " old tree : %s\n",
320 oid_to_hex(&b->branch_tree.versions[0].oid));
321 fprintf(rpt, " cur tree : %s\n",
322 oid_to_hex(&b->branch_tree.versions[1].oid));
323 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
324
325 fputs(" last pack : ", rpt);
326 if (b->pack_id < MAX_PACK_ID)
327 fprintf(rpt, "%u", b->pack_id);
328 fputc('\n', rpt);
329
330 fputc('\n', rpt);
331 }
332
333 static void write_crash_report(const char *err)
334 {
335 char *loc = repo_git_path(the_repository, "fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
336 FILE *rpt = fopen(loc, "w");
337 struct branch *b;
338 unsigned long lu;
339 struct recent_command *rc;
340
341 if (!rpt) {
342 error_errno(_("can't write crash report %s"), loc);
343 free(loc);
344 return;
345 }
346
347 fprintf(stderr, _("fast-import: dumping crash report to %s\n"), loc);
348
349 fprintf(rpt, "fast-import crash report:\n");
350 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
351 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
352 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_MODE(ISO8601)));
353 fputc('\n', rpt);
354
355 fputs("fatal: ", rpt);
356 fputs(err, rpt);
357 fputc('\n', rpt);
358
359 fputc('\n', rpt);
360 fputs("Most Recent Commands Before Crash\n", rpt);
361 fputs("---------------------------------\n", rpt);
362 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
363 if (rc->next == &cmd_hist)
364 fputs("* ", rpt);
365 else
366 fputs(" ", rpt);
367 fputs(rc->buf, rpt);
368 fputc('\n', rpt);
369 }
370
371 fputc('\n', rpt);
372 fputs("Active Branch LRU\n", rpt);
373 fputs("-----------------\n", rpt);
374 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
375 cur_active_branches,
376 max_active_branches);
377 fputc('\n', rpt);
378 fputs(" pos clock name\n", rpt);
379 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
380 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
381 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
382 ++lu, b->last_commit, b->name);
383
384 fputc('\n', rpt);
385 fputs("Inactive Branches\n", rpt);
386 fputs("-----------------\n", rpt);
387 for (lu = 0; lu < branch_table_sz; lu++) {
388 for (b = branch_table[lu]; b; b = b->table_next_branch)
389 write_branch_report(rpt, b);
390 }
391
392 if (first_tag) {
393 struct tag *tg;
394 fputc('\n', rpt);
395 fputs("Annotated Tags\n", rpt);
396 fputs("--------------\n", rpt);
397 for (tg = first_tag; tg; tg = tg->next_tag) {
398 fputs(oid_to_hex(&tg->oid), rpt);
399 fputc(' ', rpt);
400 fputs(tg->name, rpt);
401 fputc('\n', rpt);
402 }
403 }
404
405 fputc('\n', rpt);
406 fputs("Marks\n", rpt);
407 fputs("-----\n", rpt);
408 if (export_marks_file)
409 fprintf(rpt, " exported to %s\n", export_marks_file);
410 else
411 for_each_mark(marks, 0, dump_marks_fn, rpt);
412
413 fputc('\n', rpt);
414 fputs("-------------------\n", rpt);
415 fputs("END OF CRASH REPORT\n", rpt);
416 fclose(rpt);
417 free(loc);
418 }
419
420 static void end_packfile(void);
421 static void unkeep_all_packs(void);
422 static void dump_marks(void);
423
424 static NORETURN void die_nicely(const char *err, va_list params)
425 {
426 va_list cp;
427 static int zombie;
428 report_fn die_message_fn = get_die_message_routine();
429
430 va_copy(cp, params);
431 die_message_fn(err, params);
432
433 if (!zombie) {
434 char message[2 * PATH_MAX];
435
436 zombie = 1;
437 vsnprintf(message, sizeof(message), err, cp);
438 write_crash_report(message);
439 end_packfile();
440 unkeep_all_packs();
441 dump_marks();
442 }
443 exit(128);
444 }
445
446 #ifndef SIGUSR1 /* Windows, for example */
447
448 static void set_checkpoint_signal(void)
449 {
450 }
451
452 #else
453
454 static void checkpoint_signal(int signo UNUSED)
455 {
456 checkpoint_requested = 1;
457 }
458
459 static void set_checkpoint_signal(void)
460 {
461 struct sigaction sa;
462
463 memset(&sa, 0, sizeof(sa));
464 sa.sa_handler = checkpoint_signal;
465 sigemptyset(&sa.sa_mask);
466 sa.sa_flags = SA_RESTART;
467 sigaction(SIGUSR1, &sa, NULL);
468 }
469
470 #endif
471
472 static void alloc_objects(unsigned int cnt)
473 {
474 struct object_entry_pool *b;
475
476 b = xmalloc(sizeof(struct object_entry_pool)
477 + cnt * sizeof(struct object_entry));
478 b->next_pool = blocks;
479 b->next_free = b->entries;
480 b->end = b->entries + cnt;
481 blocks = b;
482 alloc_count += cnt;
483 }
484
485 static struct object_entry *new_object(struct object_id *oid)
486 {
487 struct object_entry *e;
488
489 if (blocks->next_free == blocks->end)
490 alloc_objects(object_entry_alloc);
491
492 e = blocks->next_free++;
493 oidcpy(&e->idx.oid, oid);
494 return e;
495 }
496
497 static struct object_entry *find_object(struct object_id *oid)
498 {
499 return hashmap_get_entry_from_hash(&object_table, oidhash(oid), oid,
500 struct object_entry, ent);
501 }
502
503 static struct object_entry *insert_object(struct object_id *oid)
504 {
505 struct object_entry *e;
506 unsigned int hash = oidhash(oid);
507
508 e = hashmap_get_entry_from_hash(&object_table, hash, oid,
509 struct object_entry, ent);
510 if (!e) {
511 e = new_object(oid);
512 e->idx.offset = 0;
513 hashmap_entry_init(&e->ent, hash);
514 hashmap_add(&object_table, &e->ent);
515 }
516
517 return e;
518 }
519
520 static void invalidate_pack_id(unsigned int id)
521 {
522 unsigned long lu;
523 struct tag *t;
524 struct hashmap_iter iter;
525 struct object_entry *e;
526
527 hashmap_for_each_entry(&object_table, &iter, e, ent) {
528 if (e->pack_id == id)
529 e->pack_id = MAX_PACK_ID;
530 }
531
532 for (lu = 0; lu < branch_table_sz; lu++) {
533 struct branch *b;
534
535 for (b = branch_table[lu]; b; b = b->table_next_branch)
536 if (b->pack_id == id)
537 b->pack_id = MAX_PACK_ID;
538 }
539
540 for (t = first_tag; t; t = t->next_tag)
541 if (t->pack_id == id)
542 t->pack_id = MAX_PACK_ID;
543 }
544
545 static unsigned int hc_str(const char *s, size_t len)
546 {
547 unsigned int r = 0;
548 while (len-- > 0)
549 r = r * 31 + *s++;
550 return r;
551 }
552
553 static void insert_mark(struct mark_set **top, uintmax_t idnum, struct object_entry *oe)
554 {
555 struct mark_set *s = *top;
556
557 while ((idnum >> s->shift) >= 1024) {
558 s = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
559 s->shift = (*top)->shift + 10;
560 s->data.sets[0] = *top;
561 *top = s;
562 }
563 while (s->shift) {
564 uintmax_t i = idnum >> s->shift;
565 idnum -= i << s->shift;
566 if (!s->data.sets[i]) {
567 s->data.sets[i] = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
568 s->data.sets[i]->shift = s->shift - 10;
569 }
570 s = s->data.sets[i];
571 }
572 if (!s->data.marked[idnum])
573 marks_set_count++;
574 s->data.marked[idnum] = oe;
575 }
576
577 static void *find_mark(struct mark_set *s, uintmax_t idnum)
578 {
579 uintmax_t orig_idnum = idnum;
580 struct object_entry *oe = NULL;
581 if ((idnum >> s->shift) < 1024) {
582 while (s && s->shift) {
583 uintmax_t i = idnum >> s->shift;
584 idnum -= i << s->shift;
585 s = s->data.sets[i];
586 }
587 if (s)
588 oe = s->data.marked[idnum];
589 }
590 if (!oe)
591 die(_("mark :%" PRIuMAX " not declared"), orig_idnum);
592 return oe;
593 }
594
595 static struct atom_str *to_atom(const char *s, unsigned short len)
596 {
597 unsigned int hc = hc_str(s, len) % atom_table_sz;
598 struct atom_str *c;
599
600 for (c = atom_table[hc]; c; c = c->next_atom)
601 if (c->str_len == len && !strncmp(s, c->str_dat, len))
602 return c;
603
604 c = mem_pool_alloc(&fi_mem_pool, sizeof(struct atom_str) + len + 1);
605 c->str_len = len;
606 memcpy(c->str_dat, s, len);
607 c->str_dat[len] = 0;
608 c->next_atom = atom_table[hc];
609 atom_table[hc] = c;
610 atom_cnt++;
611 return c;
612 }
613
614 static struct branch *lookup_branch(const char *name)
615 {
616 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
617 struct branch *b;
618
619 for (b = branch_table[hc]; b; b = b->table_next_branch)
620 if (!strcmp(name, b->name))
621 return b;
622 return NULL;
623 }
624
625 static struct branch *new_branch(const char *name)
626 {
627 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
628 struct branch *b = lookup_branch(name);
629
630 if (b)
631 die(_("invalid attempt to create duplicate branch: %s"), name);
632 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
633 die(_("branch name doesn't conform to Git standards: %s"), name);
634
635 b = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct branch));
636 b->name = mem_pool_strdup(&fi_mem_pool, name);
637 b->table_next_branch = branch_table[hc];
638 b->branch_tree.versions[0].mode = S_IFDIR;
639 b->branch_tree.versions[1].mode = S_IFDIR;
640 b->num_notes = 0;
641 b->active = 0;
642 b->pack_id = MAX_PACK_ID;
643 branch_table[hc] = b;
644 branch_count++;
645 return b;
646 }
647
648 static unsigned int hc_entries(unsigned int cnt)
649 {
650 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
651 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
652 }
653
654 static struct tree_content *new_tree_content(unsigned int cnt)
655 {
656 struct avail_tree_content *f, *l = NULL;
657 struct tree_content *t;
658 unsigned int hc = hc_entries(cnt);
659
660 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
661 if (f->entry_capacity >= cnt)
662 break;
663
664 if (f) {
665 if (l)
666 l->next_avail = f->next_avail;
667 else
668 avail_tree_table[hc] = f->next_avail;
669 } else {
670 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
671 f = mem_pool_alloc(&fi_mem_pool, sizeof(*t) + sizeof(t->entries[0]) * cnt);
672 f->entry_capacity = cnt;
673 }
674
675 t = (struct tree_content*)f;
676 t->entry_count = 0;
677 t->delta_depth = 0;
678 return t;
679 }
680
681 static void release_tree_entry(struct tree_entry *e);
682 static void release_tree_content(struct tree_content *t)
683 {
684 struct avail_tree_content *f = (struct avail_tree_content*)t;
685 unsigned int hc = hc_entries(f->entry_capacity);
686 f->next_avail = avail_tree_table[hc];
687 avail_tree_table[hc] = f;
688 }
689
690 static void release_tree_content_recursive(struct tree_content *t)
691 {
692 unsigned int i;
693 for (i = 0; i < t->entry_count; i++)
694 release_tree_entry(t->entries[i]);
695 release_tree_content(t);
696 }
697
698 static struct tree_content *grow_tree_content(
699 struct tree_content *t,
700 int amt)
701 {
702 struct tree_content *r = new_tree_content(t->entry_count + amt);
703 r->entry_count = t->entry_count;
704 r->delta_depth = t->delta_depth;
705 COPY_ARRAY(r->entries, t->entries, t->entry_count);
706 release_tree_content(t);
707 return r;
708 }
709
710 static struct tree_entry *new_tree_entry(void)
711 {
712 struct tree_entry *e;
713
714 if (!avail_tree_entry) {
715 unsigned int n = tree_entry_alloc;
716 tree_entry_allocd += n * sizeof(struct tree_entry);
717 ALLOC_ARRAY(e, n);
718 avail_tree_entry = e;
719 while (n-- > 1) {
720 *((void**)e) = e + 1;
721 e++;
722 }
723 *((void**)e) = NULL;
724 }
725
726 e = avail_tree_entry;
727 avail_tree_entry = *((void**)e);
728 return e;
729 }
730
731 static void release_tree_entry(struct tree_entry *e)
732 {
733 if (e->tree)
734 release_tree_content_recursive(e->tree);
735 *((void**)e) = avail_tree_entry;
736 avail_tree_entry = e;
737 }
738
739 static struct tree_content *dup_tree_content(struct tree_content *s)
740 {
741 struct tree_content *d;
742 struct tree_entry *a, *b;
743 unsigned int i;
744
745 if (!s)
746 return NULL;
747 d = new_tree_content(s->entry_count);
748 for (i = 0; i < s->entry_count; i++) {
749 a = s->entries[i];
750 b = new_tree_entry();
751 memcpy(b, a, sizeof(*a));
752 if (a->tree && is_null_oid(&b->versions[1].oid))
753 b->tree = dup_tree_content(a->tree);
754 else
755 b->tree = NULL;
756 d->entries[i] = b;
757 }
758 d->entry_count = s->entry_count;
759 d->delta_depth = s->delta_depth;
760
761 return d;
762 }
763
764 static void start_packfile(void)
765 {
766 struct strbuf tmp_file = STRBUF_INIT;
767 struct packed_git *p;
768 int pack_fd;
769
770 pack_fd = odb_mkstemp(the_repository->objects, &tmp_file,
771 "pack/tmp_pack_XXXXXX");
772 FLEX_ALLOC_STR(p, pack_name, tmp_file.buf);
773 strbuf_release(&tmp_file);
774
775 p->pack_fd = pack_fd;
776 p->do_not_close = 1;
777 p->repo = the_repository;
778 pack_file = hashfd(the_repository->hash_algo, pack_fd, p->pack_name);
779
780 pack_data = p;
781 pack_size = write_pack_header(pack_file, 0);
782 object_count = 0;
783
784 REALLOC_ARRAY(all_packs, pack_id + 1);
785 all_packs[pack_id] = p;
786 }
787
788 static const char *create_index(void)
789 {
790 const char *tmpfile;
791 struct pack_idx_entry **idx, **c, **last;
792 struct object_entry *e;
793 struct object_entry_pool *o;
794
795 /* Build the table of object IDs. */
796 ALLOC_ARRAY(idx, object_count);
797 c = idx;
798 for (o = blocks; o; o = o->next_pool)
799 for (e = o->next_free; e-- != o->entries;)
800 if (pack_id == e->pack_id)
801 *c++ = &e->idx;
802 last = idx + object_count;
803 if (c != last)
804 die(_("internal consistency error creating the index"));
805
806 tmpfile = write_idx_file(the_repository, NULL, idx, object_count,
807 &pack_idx_opts, pack_data->hash);
808 free(idx);
809 return tmpfile;
810 }
811
812 static char *keep_pack(const char *curr_index_name)
813 {
814 static const char *keep_msg = "fast-import";
815 struct strbuf name = STRBUF_INIT;
816 int keep_fd;
817
818 odb_pack_name(pack_data->repo, &name, pack_data->hash, "keep");
819 keep_fd = safe_create_file_with_leading_directories(pack_data->repo,
820 name.buf);
821 if (keep_fd < 0)
822 die_errno(_("cannot create keep file"));
823 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
824 if (close(keep_fd))
825 die_errno(_("failed to write keep file"));
826
827 odb_pack_name(pack_data->repo, &name, pack_data->hash, "pack");
828 if (finalize_object_file(pack_data->repo, pack_data->pack_name, name.buf))
829 die(_("cannot store pack file"));
830
831 odb_pack_name(pack_data->repo, &name, pack_data->hash, "idx");
832 if (finalize_object_file(pack_data->repo, curr_index_name, name.buf))
833 die(_("cannot store index file"));
834 free((void *)curr_index_name);
835 return strbuf_detach(&name, NULL);
836 }
837
838 static void unkeep_all_packs(void)
839 {
840 struct strbuf name = STRBUF_INIT;
841 int k;
842
843 for (k = 0; k < pack_id; k++) {
844 struct packed_git *p = all_packs[k];
845 odb_pack_name(p->repo, &name, p->hash, "keep");
846 unlink_or_warn(name.buf);
847 }
848 strbuf_release(&name);
849 }
850
851 static int loosen_small_pack(const struct packed_git *p)
852 {
853 struct child_process unpack = CHILD_PROCESS_INIT;
854
855 if (lseek(p->pack_fd, 0, SEEK_SET) < 0)
856 die_errno(_("failed seeking to start of '%s'"), p->pack_name);
857
858 unpack.in = p->pack_fd;
859 unpack.git_cmd = 1;
860 unpack.stdout_to_stderr = 1;
861 strvec_push(&unpack.args, "unpack-objects");
862 if (!show_stats)
863 strvec_push(&unpack.args, "-q");
864
865 return run_command(&unpack);
866 }
867
868 static void end_packfile(void)
869 {
870 static int running;
871
872 if (running || !pack_data)
873 return;
874
875 running = 1;
876 clear_delta_base_cache();
877 if (object_count) {
878 struct packed_git *new_p;
879 struct object_id cur_pack_oid;
880 char *idx_name;
881 int i;
882 struct branch *b;
883 struct tag *t;
884
885 close_pack_windows(pack_data);
886 finalize_hashfile(pack_file, cur_pack_oid.hash, FSYNC_COMPONENT_PACK, 0);
887 fixup_pack_header_footer(the_hash_algo, pack_data->pack_fd,
888 pack_data->hash, pack_data->pack_name,
889 object_count, cur_pack_oid.hash,
890 pack_size);
891
892 if (object_count <= unpack_limit) {
893 if (!loosen_small_pack(pack_data)) {
894 invalidate_pack_id(pack_id);
895 goto discard_pack;
896 }
897 }
898
899 close(pack_data->pack_fd);
900 idx_name = keep_pack(create_index());
901
902 /* Register the packfile with core git's machinery. */
903 new_p = packfile_store_load_pack(pack_data->repo->objects->packfiles,
904 idx_name, 1);
905 if (!new_p)
906 die(_("core Git rejected index %s"), idx_name);
907 all_packs[pack_id] = new_p;
908 free(idx_name);
909
910 /* Print the boundary */
911 if (pack_edges) {
912 fprintf(pack_edges, "%s:", new_p->pack_name);
913 for (i = 0; i < branch_table_sz; i++) {
914 for (b = branch_table[i]; b; b = b->table_next_branch) {
915 if (b->pack_id == pack_id)
916 fprintf(pack_edges, " %s",
917 oid_to_hex(&b->oid));
918 }
919 }
920 for (t = first_tag; t; t = t->next_tag) {
921 if (t->pack_id == pack_id)
922 fprintf(pack_edges, " %s",
923 oid_to_hex(&t->oid));
924 }
925 fputc('\n', pack_edges);
926 fflush(pack_edges);
927 }
928
929 pack_id++;
930 }
931 else {
932 discard_pack:
933 close(pack_data->pack_fd);
934 unlink_or_warn(pack_data->pack_name);
935 }
936 FREE_AND_NULL(pack_data);
937 running = 0;
938
939 /* We can't carry a delta across packfiles. */
940 strbuf_release(&last_blob.data);
941 last_blob.offset = 0;
942 last_blob.depth = 0;
943 }
944
945 static void cycle_packfile(void)
946 {
947 end_packfile();
948 start_packfile();
949 }
950
951 static int store_object(
952 enum object_type type,
953 struct strbuf *dat,
954 struct last_object *last,
955 struct object_id *oidout,
956 uintmax_t mark)
957 {
958 struct packfile_store *packs = the_repository->objects->packfiles;
959 void *out, *delta;
960 struct object_entry *e;
961 unsigned char hdr[96];
962 struct object_id oid;
963 unsigned long hdrlen, deltalen;
964 struct git_hash_ctx c;
965 git_zstream s;
966
967 hdrlen = format_object_header((char *)hdr, sizeof(hdr), type,
968 dat->len);
969 the_hash_algo->init_fn(&c);
970 git_hash_update(&c, hdr, hdrlen);
971 git_hash_update(&c, dat->buf, dat->len);
972 git_hash_final_oid(&oid, &c);
973 if (oidout)
974 oidcpy(oidout, &oid);
975
976 e = insert_object(&oid);
977 if (mark)
978 insert_mark(&marks, mark, e);
979 if (e->idx.offset) {
980 duplicate_count_by_type[type]++;
981 return 1;
982 } else if (packfile_list_find_oid(packfile_store_get_packs(packs), &oid)) {
983 e->type = type;
984 e->pack_id = MAX_PACK_ID;
985 e->idx.offset = 1; /* just not zero! */
986 duplicate_count_by_type[type]++;
987 return 1;
988 }
989
990 if (last && last->data.len && last->data.buf && last->depth < max_depth
991 && dat->len > the_hash_algo->rawsz) {
992
993 delta_count_attempts_by_type[type]++;
994 delta = diff_delta(last->data.buf, last->data.len,
995 dat->buf, dat->len,
996 &deltalen, dat->len - the_hash_algo->rawsz);
997 } else
998 delta = NULL;
999
1000 git_deflate_init(&s, pack_compression_level);
1001 if (delta) {
1002 s.next_in = delta;
1003 s.avail_in = deltalen;
1004 } else {
1005 s.next_in = (void *)dat->buf;
1006 s.avail_in = dat->len;
1007 }
1008 s.avail_out = git_deflate_bound(&s, s.avail_in);
1009 s.next_out = out = xmalloc(s.avail_out);
1010 while (git_deflate(&s, Z_FINISH) == Z_OK)
1011 ; /* nothing */
1012 git_deflate_end(&s);
1013
1014 /* Determine if we should auto-checkpoint. */
1015 if ((max_packsize
1016 && (pack_size + PACK_SIZE_THRESHOLD + s.total_out) > max_packsize)
1017 || (pack_size + PACK_SIZE_THRESHOLD + s.total_out) < pack_size) {
1018
1019 /* This new object needs to *not* have the current pack_id. */
1020 e->pack_id = pack_id + 1;
1021 cycle_packfile();
1022
1023 /* We cannot carry a delta into the new pack. */
1024 if (delta) {
1025 FREE_AND_NULL(delta);
1026
1027 git_deflate_init(&s, pack_compression_level);
1028 s.next_in = (void *)dat->buf;
1029 s.avail_in = dat->len;
1030 s.avail_out = git_deflate_bound(&s, s.avail_in);
1031 s.next_out = out = xrealloc(out, s.avail_out);
1032 while (git_deflate(&s, Z_FINISH) == Z_OK)
1033 ; /* nothing */
1034 git_deflate_end(&s);
1035 }
1036 }
1037
1038 e->type = type;
1039 e->pack_id = pack_id;
1040 e->idx.offset = pack_size;
1041 object_count++;
1042 object_count_by_type[type]++;
1043
1044 crc32_begin(pack_file);
1045
1046 if (delta) {
1047 off_t ofs = e->idx.offset - last->offset;
1048 unsigned pos = sizeof(hdr) - 1;
1049
1050 delta_count_by_type[type]++;
1051 e->depth = last->depth + 1;
1052
1053 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1054 OBJ_OFS_DELTA, deltalen);
1055 hashwrite(pack_file, hdr, hdrlen);
1056 pack_size += hdrlen;
1057
1058 hdr[pos] = ofs & 127;
1059 while (ofs >>= 7)
1060 hdr[--pos] = 128 | (--ofs & 127);
1061 hashwrite(pack_file, hdr + pos, sizeof(hdr) - pos);
1062 pack_size += sizeof(hdr) - pos;
1063 } else {
1064 e->depth = 0;
1065 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1066 type, dat->len);
1067 hashwrite(pack_file, hdr, hdrlen);
1068 pack_size += hdrlen;
1069 }
1070
1071 hashwrite(pack_file, out, s.total_out);
1072 pack_size += s.total_out;
1073
1074 e->idx.crc32 = crc32_end(pack_file);
1075
1076 free(out);
1077 free(delta);
1078 if (last) {
1079 if (last->no_swap) {
1080 last->data = *dat;
1081 } else {
1082 strbuf_swap(&last->data, dat);
1083 }
1084 last->offset = e->idx.offset;
1085 last->depth = e->depth;
1086 }
1087 return 0;
1088 }
1089
1090 static void truncate_pack(struct hashfile_checkpoint *checkpoint)
1091 {
1092 if (hashfile_truncate(pack_file, checkpoint))
1093 die_errno(_("cannot truncate pack to skip duplicate"));
1094 pack_size = checkpoint->offset;
1095 }
1096
1097 static void stream_blob(uintmax_t len, struct object_id *oidout, uintmax_t mark)
1098 {
1099 struct packfile_store *packs = the_repository->objects->packfiles;
1100 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1101 unsigned char *in_buf = xmalloc(in_sz);
1102 unsigned char *out_buf = xmalloc(out_sz);
1103 struct object_entry *e;
1104 struct object_id oid;
1105 unsigned long hdrlen;
1106 off_t offset;
1107 struct git_hash_ctx c;
1108 git_zstream s;
1109 struct hashfile_checkpoint checkpoint;
1110 int status = Z_OK;
1111
1112 /* Determine if we should auto-checkpoint. */
1113 if ((max_packsize
1114 && (pack_size + PACK_SIZE_THRESHOLD + len) > max_packsize)
1115 || (pack_size + PACK_SIZE_THRESHOLD + len) < pack_size)
1116 cycle_packfile();
1117
1118 hashfile_checkpoint_init(pack_file, &checkpoint);
1119 hashfile_checkpoint(pack_file, &checkpoint);
1120 offset = checkpoint.offset;
1121
1122 hdrlen = format_object_header((char *)out_buf, out_sz, OBJ_BLOB, len);
1123
1124 the_hash_algo->init_fn(&c);
1125 git_hash_update(&c, out_buf, hdrlen);
1126
1127 crc32_begin(pack_file);
1128
1129 git_deflate_init(&s, pack_compression_level);
1130
1131 hdrlen = encode_in_pack_object_header(out_buf, out_sz, OBJ_BLOB, len);
1132
1133 s.next_out = out_buf + hdrlen;
1134 s.avail_out = out_sz - hdrlen;
1135
1136 while (status != Z_STREAM_END) {
1137 if (0 < len && !s.avail_in) {
1138 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1139 size_t n = fread(in_buf, 1, cnt, stdin);
1140 if (!n && feof(stdin))
1141 die(_("EOF in data (%" PRIuMAX " bytes remaining)"), len);
1142
1143 git_hash_update(&c, in_buf, n);
1144 s.next_in = in_buf;
1145 s.avail_in = n;
1146 len -= n;
1147 }
1148
1149 status = git_deflate(&s, len ? 0 : Z_FINISH);
1150
1151 if (!s.avail_out || status == Z_STREAM_END) {
1152 size_t n = s.next_out - out_buf;
1153 hashwrite(pack_file, out_buf, n);
1154 pack_size += n;
1155 s.next_out = out_buf;
1156 s.avail_out = out_sz;
1157 }
1158
1159 switch (status) {
1160 case Z_OK:
1161 case Z_BUF_ERROR:
1162 case Z_STREAM_END:
1163 continue;
1164 default:
1165 die(_("unexpected deflate failure: %d"), status);
1166 }
1167 }
1168 git_deflate_end(&s);
1169 git_hash_final_oid(&oid, &c);
1170
1171 if (oidout)
1172 oidcpy(oidout, &oid);
1173
1174 e = insert_object(&oid);
1175
1176 if (mark)
1177 insert_mark(&marks, mark, e);
1178
1179 if (e->idx.offset) {
1180 duplicate_count_by_type[OBJ_BLOB]++;
1181 truncate_pack(&checkpoint);
1182
1183 } else if (packfile_list_find_oid(packfile_store_get_packs(packs), &oid)) {
1184 e->type = OBJ_BLOB;
1185 e->pack_id = MAX_PACK_ID;
1186 e->idx.offset = 1; /* just not zero! */
1187 duplicate_count_by_type[OBJ_BLOB]++;
1188 truncate_pack(&checkpoint);
1189
1190 } else {
1191 e->depth = 0;
1192 e->type = OBJ_BLOB;
1193 e->pack_id = pack_id;
1194 e->idx.offset = offset;
1195 e->idx.crc32 = crc32_end(pack_file);
1196 object_count++;
1197 object_count_by_type[OBJ_BLOB]++;
1198 }
1199
1200 free(in_buf);
1201 free(out_buf);
1202 }
1203
1204 /* All calls must be guarded by find_object() or find_mark() to
1205 * ensure the 'struct object_entry' passed was written by this
1206 * process instance. We unpack the entry by the offset, avoiding
1207 * the need for the corresponding .idx file. This unpacking rule
1208 * works because we only use OBJ_REF_DELTA within the packfiles
1209 * created by fast-import.
1210 *
1211 * oe must not be NULL. Such an oe usually comes from giving
1212 * an unknown SHA-1 to find_object() or an undefined mark to
1213 * find_mark(). Callers must test for this condition and use
1214 * the standard read_sha1_file() when it happens.
1215 *
1216 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1217 * find_mark(), where the mark was reloaded from an existing marks
1218 * file and is referencing an object that this fast-import process
1219 * instance did not write out to a packfile. Callers must test for
1220 * this condition and use read_sha1_file() instead.
1221 */
1222 static void *gfi_unpack_entry(
1223 struct object_entry *oe,
1224 unsigned long *sizep)
1225 {
1226 enum object_type type;
1227 struct packed_git *p = all_packs[oe->pack_id];
1228 if (p == pack_data && p->pack_size < (pack_size + the_hash_algo->rawsz)) {
1229 /* The object is stored in the packfile we are writing to
1230 * and we have modified it since the last time we scanned
1231 * back to read a previously written object. If an old
1232 * window covered [p->pack_size, p->pack_size + rawsz) its
1233 * data is stale and is not valid. Closing all windows
1234 * and updating the packfile length ensures we can read
1235 * the newly written data.
1236 */
1237 close_pack_windows(p);
1238 hashflush(pack_file);
1239
1240 /* We have to offer rawsz bytes additional on the end of
1241 * the packfile as the core unpacker code assumes the
1242 * footer is present at the file end and must promise
1243 * at least rawsz bytes within any window it maps. But
1244 * we don't actually create the footer here.
1245 */
1246 p->pack_size = pack_size + the_hash_algo->rawsz;
1247 }
1248 return unpack_entry(the_repository, p, oe->idx.offset, &type, sizep);
1249 }
1250
1251 static void load_tree(struct tree_entry *root)
1252 {
1253 struct object_id *oid = &root->versions[1].oid;
1254 struct object_entry *myoe;
1255 struct tree_content *t;
1256 unsigned long size;
1257 char *buf;
1258 const char *c;
1259
1260 root->tree = t = new_tree_content(8);
1261 if (is_null_oid(oid))
1262 return;
1263
1264 myoe = find_object(oid);
1265 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1266 if (myoe->type != OBJ_TREE)
1267 die(_("not a tree: %s"), oid_to_hex(oid));
1268 t->delta_depth = myoe->depth;
1269 buf = gfi_unpack_entry(myoe, &size);
1270 if (!buf)
1271 die(_("can't load tree %s"), oid_to_hex(oid));
1272 } else {
1273 enum object_type type;
1274 buf = odb_read_object(the_repository->objects, oid, &type, &size);
1275 if (!buf || type != OBJ_TREE)
1276 die(_("can't load tree %s"), oid_to_hex(oid));
1277 }
1278
1279 c = buf;
1280 while (c != (buf + size)) {
1281 struct tree_entry *e = new_tree_entry();
1282
1283 if (t->entry_count == t->entry_capacity)
1284 root->tree = t = grow_tree_content(t, t->entry_count);
1285 t->entries[t->entry_count++] = e;
1286
1287 e->tree = NULL;
1288 c = parse_mode(c, &e->versions[1].mode);
1289 if (!c)
1290 die(_("corrupt mode in %s"), oid_to_hex(oid));
1291 e->versions[0].mode = e->versions[1].mode;
1292 e->name = to_atom(c, strlen(c));
1293 c += e->name->str_len + 1;
1294 oidread(&e->versions[0].oid, (unsigned char *)c,
1295 the_repository->hash_algo);
1296 oidread(&e->versions[1].oid, (unsigned char *)c,
1297 the_repository->hash_algo);
1298 c += the_hash_algo->rawsz;
1299 }
1300 free(buf);
1301 }
1302
1303 static int tecmp0 (const void *_a, const void *_b)
1304 {
1305 struct tree_entry *a = *((struct tree_entry**)_a);
1306 struct tree_entry *b = *((struct tree_entry**)_b);
1307 return base_name_compare(
1308 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1309 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1310 }
1311
1312 static int tecmp1 (const void *_a, const void *_b)
1313 {
1314 struct tree_entry *a = *((struct tree_entry**)_a);
1315 struct tree_entry *b = *((struct tree_entry**)_b);
1316 return base_name_compare(
1317 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1318 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1319 }
1320
1321 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1322 {
1323 size_t maxlen = 0;
1324 unsigned int i;
1325
1326 if (!v)
1327 QSORT(t->entries, t->entry_count, tecmp0);
1328 else
1329 QSORT(t->entries, t->entry_count, tecmp1);
1330
1331 for (i = 0; i < t->entry_count; i++) {
1332 if (t->entries[i]->versions[v].mode)
1333 maxlen += t->entries[i]->name->str_len + 34;
1334 }
1335
1336 strbuf_reset(b);
1337 strbuf_grow(b, maxlen);
1338 for (i = 0; i < t->entry_count; i++) {
1339 struct tree_entry *e = t->entries[i];
1340 if (!e->versions[v].mode)
1341 continue;
1342 strbuf_addf(b, "%o %s%c",
1343 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1344 e->name->str_dat, '\0');
1345 strbuf_add(b, e->versions[v].oid.hash, the_hash_algo->rawsz);
1346 }
1347 }
1348
1349 static void store_tree(struct tree_entry *root)
1350 {
1351 struct tree_content *t;
1352 unsigned int i, j, del;
1353 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1354 struct object_entry *le = NULL;
1355
1356 if (!is_null_oid(&root->versions[1].oid))
1357 return;
1358
1359 if (!root->tree)
1360 load_tree(root);
1361 t = root->tree;
1362
1363 for (i = 0; i < t->entry_count; i++) {
1364 if (t->entries[i]->tree)
1365 store_tree(t->entries[i]);
1366 }
1367
1368 if (!(root->versions[0].mode & NO_DELTA))
1369 le = find_object(&root->versions[0].oid);
1370 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1371 mktree(t, 0, &old_tree);
1372 lo.data = old_tree;
1373 lo.offset = le->idx.offset;
1374 lo.depth = t->delta_depth;
1375 }
1376
1377 mktree(t, 1, &new_tree);
1378 store_object(OBJ_TREE, &new_tree, &lo, &root->versions[1].oid, 0);
1379
1380 t->delta_depth = lo.depth;
1381 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1382 struct tree_entry *e = t->entries[i];
1383 if (e->versions[1].mode) {
1384 e->versions[0].mode = e->versions[1].mode;
1385 oidcpy(&e->versions[0].oid, &e->versions[1].oid);
1386 t->entries[j++] = e;
1387 } else {
1388 release_tree_entry(e);
1389 del++;
1390 }
1391 }
1392 t->entry_count -= del;
1393 }
1394
1395 static void tree_content_replace(
1396 struct tree_entry *root,
1397 const struct object_id *oid,
1398 const uint16_t mode,
1399 struct tree_content *newtree)
1400 {
1401 if (!S_ISDIR(mode))
1402 die(_("root cannot be a non-directory"));
1403 oidclr(&root->versions[0].oid, the_repository->hash_algo);
1404 oidcpy(&root->versions[1].oid, oid);
1405 if (root->tree)
1406 release_tree_content_recursive(root->tree);
1407 root->tree = newtree;
1408 }
1409
1410 static int tree_content_set(
1411 struct tree_entry *root,
1412 const char *p,
1413 const struct object_id *oid,
1414 const uint16_t mode,
1415 struct tree_content *subtree)
1416 {
1417 struct tree_content *t;
1418 const char *slash1;
1419 unsigned int i, n;
1420 struct tree_entry *e;
1421
1422 slash1 = strchrnul(p, '/');
1423 n = slash1 - p;
1424 if (!n)
1425 die(_("empty path component found in input"));
1426 if (!*slash1 && !S_ISDIR(mode) && subtree)
1427 die(_("non-directories cannot have subtrees"));
1428
1429 if (!root->tree)
1430 load_tree(root);
1431 t = root->tree;
1432 for (i = 0; i < t->entry_count; i++) {
1433 e = t->entries[i];
1434 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1435 if (!*slash1) {
1436 if (!S_ISDIR(mode)
1437 && e->versions[1].mode == mode
1438 && oideq(&e->versions[1].oid, oid))
1439 return 0;
1440 e->versions[1].mode = mode;
1441 oidcpy(&e->versions[1].oid, oid);
1442 if (e->tree)
1443 release_tree_content_recursive(e->tree);
1444 e->tree = subtree;
1445
1446 /*
1447 * We need to leave e->versions[0].sha1 alone
1448 * to avoid modifying the preimage tree used
1449 * when writing out the parent directory.
1450 * But after replacing the subdir with a
1451 * completely different one, it's not a good
1452 * delta base any more, and besides, we've
1453 * thrown away the tree entries needed to
1454 * make a delta against it.
1455 *
1456 * So let's just explicitly disable deltas
1457 * for the subtree.
1458 */
1459 if (S_ISDIR(e->versions[0].mode))
1460 e->versions[0].mode |= NO_DELTA;
1461
1462 oidclr(&root->versions[1].oid, the_repository->hash_algo);
1463 return 1;
1464 }
1465 if (!S_ISDIR(e->versions[1].mode)) {
1466 e->tree = new_tree_content(8);
1467 e->versions[1].mode = S_IFDIR;
1468 }
1469 if (!e->tree)
1470 load_tree(e);
1471 if (tree_content_set(e, slash1 + 1, oid, mode, subtree)) {
1472 oidclr(&root->versions[1].oid, the_repository->hash_algo);
1473 return 1;
1474 }
1475 return 0;
1476 }
1477 }
1478
1479 if (t->entry_count == t->entry_capacity)
1480 root->tree = t = grow_tree_content(t, t->entry_count);
1481 e = new_tree_entry();
1482 e->name = to_atom(p, n);
1483 e->versions[0].mode = 0;
1484 oidclr(&e->versions[0].oid, the_repository->hash_algo);
1485 t->entries[t->entry_count++] = e;
1486 if (*slash1) {
1487 e->tree = new_tree_content(8);
1488 e->versions[1].mode = S_IFDIR;
1489 tree_content_set(e, slash1 + 1, oid, mode, subtree);
1490 } else {
1491 e->tree = subtree;
1492 e->versions[1].mode = mode;
1493 oidcpy(&e->versions[1].oid, oid);
1494 }
1495 oidclr(&root->versions[1].oid, the_repository->hash_algo);
1496 return 1;
1497 }
1498
1499 static int tree_content_remove(
1500 struct tree_entry *root,
1501 const char *p,
1502 struct tree_entry *backup_leaf,
1503 int allow_root)
1504 {
1505 struct tree_content *t;
1506 const char *slash1;
1507 unsigned int i, n;
1508 struct tree_entry *e;
1509
1510 slash1 = strchrnul(p, '/');
1511 n = slash1 - p;
1512
1513 if (!root->tree)
1514 load_tree(root);
1515
1516 if (!*p && allow_root) {
1517 e = root;
1518 goto del_entry;
1519 }
1520
1521 t = root->tree;
1522 for (i = 0; i < t->entry_count; i++) {
1523 e = t->entries[i];
1524 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1525 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1526 /*
1527 * If p names a file in some subdirectory, and a
1528 * file or symlink matching the name of the
1529 * parent directory of p exists, then p cannot
1530 * exist and need not be deleted.
1531 */
1532 return 1;
1533 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1534 goto del_entry;
1535 if (!e->tree)
1536 load_tree(e);
1537 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1538 for (n = 0; n < e->tree->entry_count; n++) {
1539 if (e->tree->entries[n]->versions[1].mode) {
1540 oidclr(&root->versions[1].oid,
1541 the_repository->hash_algo);
1542 return 1;
1543 }
1544 }
1545 backup_leaf = NULL;
1546 goto del_entry;
1547 }
1548 return 0;
1549 }
1550 }
1551 return 0;
1552
1553 del_entry:
1554 if (backup_leaf)
1555 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1556 else if (e->tree)
1557 release_tree_content_recursive(e->tree);
1558 e->tree = NULL;
1559 e->versions[1].mode = 0;
1560 oidclr(&e->versions[1].oid, the_repository->hash_algo);
1561 oidclr(&root->versions[1].oid, the_repository->hash_algo);
1562 return 1;
1563 }
1564
1565 static int tree_content_get(
1566 struct tree_entry *root,
1567 const char *p,
1568 struct tree_entry *leaf,
1569 int allow_root)
1570 {
1571 struct tree_content *t;
1572 const char *slash1;
1573 unsigned int i, n;
1574 struct tree_entry *e;
1575
1576 slash1 = strchrnul(p, '/');
1577 n = slash1 - p;
1578 if (!n && !allow_root)
1579 die(_("empty path component found in input"));
1580
1581 if (!root->tree)
1582 load_tree(root);
1583
1584 if (!n) {
1585 e = root;
1586 goto found_entry;
1587 }
1588
1589 t = root->tree;
1590 for (i = 0; i < t->entry_count; i++) {
1591 e = t->entries[i];
1592 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1593 if (!*slash1)
1594 goto found_entry;
1595 if (!S_ISDIR(e->versions[1].mode))
1596 return 0;
1597 if (!e->tree)
1598 load_tree(e);
1599 return tree_content_get(e, slash1 + 1, leaf, 0);
1600 }
1601 }
1602 return 0;
1603
1604 found_entry:
1605 memcpy(leaf, e, sizeof(*leaf));
1606 if (e->tree && is_null_oid(&e->versions[1].oid))
1607 leaf->tree = dup_tree_content(e->tree);
1608 else
1609 leaf->tree = NULL;
1610 return 1;
1611 }
1612
1613 static int update_branch(struct branch *b)
1614 {
1615 static const char *msg = "fast-import";
1616 struct ref_transaction *transaction;
1617 struct object_id old_oid;
1618 struct strbuf err = STRBUF_INIT;
1619 static const char *replace_prefix = "refs/replace/";
1620
1621 if (starts_with(b->name, replace_prefix) &&
1622 !strcmp(b->name + strlen(replace_prefix),
1623 oid_to_hex(&b->oid))) {
1624 if (!quiet)
1625 warning(_("dropping %s since it would point to "
1626 "itself (i.e. to %s)"),
1627 b->name, oid_to_hex(&b->oid));
1628 refs_delete_ref(get_main_ref_store(the_repository),
1629 NULL, b->name, NULL, 0);
1630 return 0;
1631 }
1632 if (is_null_oid(&b->oid)) {
1633 if (b->delete)
1634 refs_delete_ref(get_main_ref_store(the_repository),
1635 NULL, b->name, NULL, 0);
1636 return 0;
1637 }
1638 if (refs_read_ref(get_main_ref_store(the_repository), b->name, &old_oid))
1639 oidclr(&old_oid, the_repository->hash_algo);
1640 if (!force_update && !is_null_oid(&old_oid)) {
1641 struct commit *old_cmit, *new_cmit;
1642 int ret;
1643
1644 old_cmit = lookup_commit_reference_gently(the_repository,
1645 &old_oid, 0);
1646 new_cmit = lookup_commit_reference_gently(the_repository,
1647 &b->oid, 0);
1648 if (!old_cmit || !new_cmit)
1649 return error(_("branch %s is missing commits."), b->name);
1650
1651 ret = repo_in_merge_bases(the_repository, old_cmit, new_cmit);
1652 if (ret < 0)
1653 exit(128);
1654 if (!ret) {
1655 warning(_("not updating %s"
1656 " (new tip %s does not contain %s)"),
1657 b->name, oid_to_hex(&b->oid),
1658 oid_to_hex(&old_oid));
1659 return -1;
1660 }
1661 }
1662 transaction = ref_store_transaction_begin(get_main_ref_store(the_repository),
1663 0, &err);
1664 if (!transaction ||
1665 ref_transaction_update(transaction, b->name, &b->oid, &old_oid,
1666 NULL, NULL, 0, msg, &err) ||
1667 ref_transaction_commit(transaction, &err)) {
1668 ref_transaction_free(transaction);
1669 error("%s", err.buf);
1670 strbuf_release(&err);
1671 return -1;
1672 }
1673 ref_transaction_free(transaction);
1674 strbuf_release(&err);
1675 return 0;
1676 }
1677
1678 static void dump_branches(void)
1679 {
1680 unsigned int i;
1681 struct branch *b;
1682
1683 for (i = 0; i < branch_table_sz; i++) {
1684 for (b = branch_table[i]; b; b = b->table_next_branch)
1685 failure |= update_branch(b);
1686 }
1687 }
1688
1689 static void dump_tags(void)
1690 {
1691 static const char *msg = "fast-import";
1692 struct tag *t;
1693 struct strbuf ref_name = STRBUF_INIT;
1694 struct strbuf err = STRBUF_INIT;
1695 struct ref_transaction *transaction;
1696
1697 transaction = ref_store_transaction_begin(get_main_ref_store(the_repository),
1698 0, &err);
1699 if (!transaction) {
1700 failure |= error("%s", err.buf);
1701 goto cleanup;
1702 }
1703 for (t = first_tag; t; t = t->next_tag) {
1704 strbuf_reset(&ref_name);
1705 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1706
1707 if (ref_transaction_update(transaction, ref_name.buf,
1708 &t->oid, NULL, NULL, NULL,
1709 0, msg, &err)) {
1710 failure |= error("%s", err.buf);
1711 goto cleanup;
1712 }
1713 }
1714 if (ref_transaction_commit(transaction, &err))
1715 failure |= error("%s", err.buf);
1716
1717 cleanup:
1718 ref_transaction_free(transaction);
1719 strbuf_release(&ref_name);
1720 strbuf_release(&err);
1721 }
1722
1723 static void dump_marks(void)
1724 {
1725 struct lock_file mark_lock = LOCK_INIT;
1726 FILE *f;
1727
1728 if (!export_marks_file || (import_marks_file && !import_marks_file_done))
1729 return;
1730
1731 if (safe_create_leading_directories_const(the_repository, export_marks_file)) {
1732 failure |= error_errno(_("unable to create leading directories of %s"),
1733 export_marks_file);
1734 return;
1735 }
1736
1737 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1738 failure |= error_errno(_("unable to write marks file %s"),
1739 export_marks_file);
1740 return;
1741 }
1742
1743 f = fdopen_lock_file(&mark_lock, "w");
1744 if (!f) {
1745 int saved_errno = errno;
1746 rollback_lock_file(&mark_lock);
1747 failure |= error(_("unable to write marks file %s: %s"),
1748 export_marks_file, strerror(saved_errno));
1749 return;
1750 }
1751
1752 for_each_mark(marks, 0, dump_marks_fn, f);
1753 if (commit_lock_file(&mark_lock)) {
1754 failure |= error_errno(_("unable to write file %s"),
1755 export_marks_file);
1756 return;
1757 }
1758 }
1759
1760 static void insert_object_entry(struct mark_set **s, struct object_id *oid, uintmax_t mark)
1761 {
1762 struct object_entry *e;
1763 e = find_object(oid);
1764 if (!e) {
1765 enum object_type type = odb_read_object_info(the_repository->objects,
1766 oid, NULL);
1767 if (type < 0)
1768 die(_("object not found: %s"), oid_to_hex(oid));
1769 e = insert_object(oid);
1770 e->type = type;
1771 e->pack_id = MAX_PACK_ID;
1772 e->idx.offset = 1; /* just not zero! */
1773 }
1774 insert_mark(s, mark, e);
1775 }
1776
1777 static void insert_oid_entry(struct mark_set **s, struct object_id *oid, uintmax_t mark)
1778 {
1779 insert_mark(s, mark, xmemdupz(oid, sizeof(*oid)));
1780 }
1781
1782 static void read_mark_file(struct mark_set **s, FILE *f, mark_set_inserter_t inserter)
1783 {
1784 char line[512];
1785 while (fgets(line, sizeof(line), f)) {
1786 uintmax_t mark;
1787 char *end;
1788 struct object_id oid;
1789
1790 /* Ensure SHA-1 objects are padded with zeros. */
1791 memset(oid.hash, 0, sizeof(oid.hash));
1792
1793 end = strchr(line, '\n');
1794 if (line[0] != ':' || !end)
1795 die(_("corrupt mark line: %s"), line);
1796 *end = 0;
1797 mark = strtoumax(line + 1, &end, 10);
1798 if (!mark || end == line + 1
1799 || *end != ' '
1800 || get_oid_hex_any(end + 1, &oid) == GIT_HASH_UNKNOWN)
1801 die(_("corrupt mark line: %s"), line);
1802 inserter(s, &oid, mark);
1803 }
1804 }
1805
1806 static void read_marks(void)
1807 {
1808 FILE *f = fopen(import_marks_file, "r");
1809 if (f)
1810 ;
1811 else if (import_marks_file_ignore_missing && errno == ENOENT)
1812 goto done; /* Marks file does not exist */
1813 else
1814 die_errno(_("cannot read '%s'"), import_marks_file);
1815 read_mark_file(&marks, f, insert_object_entry);
1816 fclose(f);
1817 done:
1818 import_marks_file_done = 1;
1819 }
1820
1821
1822 static int read_next_command(void)
1823 {
1824 static int stdin_eof = 0;
1825
1826 if (stdin_eof) {
1827 unread_command_buf = 0;
1828 return EOF;
1829 }
1830
1831 for (;;) {
1832 if (unread_command_buf) {
1833 unread_command_buf = 0;
1834 } else {
1835 struct recent_command *rc;
1836
1837 stdin_eof = strbuf_getline_lf(&command_buf, stdin);
1838 if (stdin_eof)
1839 return EOF;
1840
1841 if (!seen_data_command
1842 && !starts_with(command_buf.buf, "feature ")
1843 && !starts_with(command_buf.buf, "option ")) {
1844 parse_argv();
1845 }
1846
1847 rc = rc_free;
1848 if (rc)
1849 rc_free = rc->next;
1850 else {
1851 rc = cmd_hist.next;
1852 cmd_hist.next = rc->next;
1853 cmd_hist.next->prev = &cmd_hist;
1854 free(rc->buf);
1855 }
1856
1857 rc->buf = xstrdup(command_buf.buf);
1858 rc->prev = cmd_tail;
1859 rc->next = cmd_hist.prev;
1860 rc->prev->next = rc;
1861 cmd_tail = rc;
1862 }
1863 if (command_buf.buf[0] == '#')
1864 continue;
1865 return 0;
1866 }
1867 }
1868
1869 static void skip_optional_lf(void)
1870 {
1871 int term_char = fgetc(stdin);
1872 if (term_char != '\n' && term_char != EOF)
1873 ungetc(term_char, stdin);
1874 }
1875
1876 static void parse_mark(void)
1877 {
1878 const char *v;
1879 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1880 next_mark = strtoumax(v, NULL, 10);
1881 read_next_command();
1882 }
1883 else
1884 next_mark = 0;
1885 }
1886
1887 static void parse_original_identifier(void)
1888 {
1889 const char *v;
1890 if (skip_prefix(command_buf.buf, "original-oid ", &v))
1891 read_next_command();
1892 }
1893
1894 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1895 {
1896 const char *data;
1897 strbuf_reset(sb);
1898
1899 if (!skip_prefix(command_buf.buf, "data ", &data))
1900 die(_("expected 'data n' command, found: %s"), command_buf.buf);
1901
1902 if (skip_prefix(data, "<<", &data)) {
1903 char *term = xstrdup(data);
1904 size_t term_len = command_buf.len - (data - command_buf.buf);
1905
1906 for (;;) {
1907 if (strbuf_getline_lf(&command_buf, stdin) == EOF)
1908 die(_("EOF in data (terminator '%s' not found)"), term);
1909 if (term_len == command_buf.len
1910 && !strcmp(term, command_buf.buf))
1911 break;
1912 strbuf_addbuf(sb, &command_buf);
1913 strbuf_addch(sb, '\n');
1914 }
1915 free(term);
1916 }
1917 else {
1918 uintmax_t len = strtoumax(data, NULL, 10);
1919 size_t n = 0, length = (size_t)len;
1920
1921 if (limit && limit < len) {
1922 *len_res = len;
1923 return 0;
1924 }
1925 if (length < len)
1926 die(_("data is too large to use in this context"));
1927
1928 while (n < length) {
1929 size_t s = strbuf_fread(sb, length - n, stdin);
1930 if (!s && feof(stdin))
1931 die(_("EOF in data (%lu bytes remaining)"),
1932 (unsigned long)(length - n));
1933 n += s;
1934 }
1935 }
1936
1937 skip_optional_lf();
1938 return 1;
1939 }
1940
1941 static int validate_raw_date(const char *src, struct strbuf *result, int strict)
1942 {
1943 const char *orig_src = src;
1944 char *endp;
1945 unsigned long num;
1946
1947 errno = 0;
1948
1949 num = strtoul(src, &endp, 10);
1950 /*
1951 * NEEDSWORK: perhaps check for reasonable values? For example, we
1952 * could error on values representing times more than a
1953 * day in the future.
1954 */
1955 if (errno || endp == src || *endp != ' ')
1956 return -1;
1957
1958 src = endp + 1;
1959 if (*src != '-' && *src != '+')
1960 return -1;
1961
1962 num = strtoul(src + 1, &endp, 10);
1963 /*
1964 * NEEDSWORK: check for brokenness other than num > 1400, such as
1965 * (num % 100) >= 60, or ((num % 100) % 15) != 0 ?
1966 */
1967 if (errno || endp == src + 1 || *endp || /* did not parse */
1968 (strict && (1400 < num)) /* parsed a broken timezone */
1969 )
1970 return -1;
1971
1972 strbuf_addstr(result, orig_src);
1973 return 0;
1974 }
1975
1976 static char *parse_ident(const char *buf)
1977 {
1978 const char *ltgt;
1979 size_t name_len;
1980 struct strbuf ident = STRBUF_INIT;
1981
1982 /* ensure there is a space delimiter even if there is no name */
1983 if (*buf == '<')
1984 --buf;
1985
1986 ltgt = buf + strcspn(buf, "<>");
1987 if (*ltgt != '<')
1988 die(_("missing < in ident string: %s"), buf);
1989 if (ltgt != buf && ltgt[-1] != ' ')
1990 die(_("missing space before < in ident string: %s"), buf);
1991 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
1992 if (*ltgt != '>')
1993 die(_("missing > in ident string: %s"), buf);
1994 ltgt++;
1995 if (*ltgt != ' ')
1996 die(_("missing space after > in ident string: %s"), buf);
1997 ltgt++;
1998 name_len = ltgt - buf;
1999 strbuf_add(&ident, buf, name_len);
2000
2001 switch (whenspec) {
2002 case WHENSPEC_RAW:
2003 if (validate_raw_date(ltgt, &ident, 1) < 0)
2004 die(_("invalid raw date \"%s\" in ident: %s"), ltgt, buf);
2005 break;
2006 case WHENSPEC_RAW_PERMISSIVE:
2007 if (validate_raw_date(ltgt, &ident, 0) < 0)
2008 die(_("invalid raw date \"%s\" in ident: %s"), ltgt, buf);
2009 break;
2010 case WHENSPEC_RFC2822:
2011 if (parse_date(ltgt, &ident) < 0)
2012 die(_("invalid rfc2822 date \"%s\" in ident: %s"), ltgt, buf);
2013 break;
2014 case WHENSPEC_NOW:
2015 if (strcmp("now", ltgt))
2016 die(_("date in ident must be 'now': %s"), buf);
2017 datestamp(&ident);
2018 break;
2019 }
2020
2021 return strbuf_detach(&ident, NULL);
2022 }
2023
2024 static void parse_and_store_blob(
2025 struct last_object *last,
2026 struct object_id *oidout,
2027 uintmax_t mark)
2028 {
2029 static struct strbuf buf = STRBUF_INIT;
2030 uintmax_t len;
2031
2032 if (parse_data(&buf, repo_settings_get_big_file_threshold(the_repository), &len))
2033 store_object(OBJ_BLOB, &buf, last, oidout, mark);
2034 else {
2035 if (last) {
2036 strbuf_release(&last->data);
2037 last->offset = 0;
2038 last->depth = 0;
2039 }
2040 stream_blob(len, oidout, mark);
2041 skip_optional_lf();
2042 }
2043 }
2044
2045 static void parse_new_blob(void)
2046 {
2047 read_next_command();
2048 parse_mark();
2049 parse_original_identifier();
2050 parse_and_store_blob(&last_blob, NULL, next_mark);
2051 }
2052
2053 static void unload_one_branch(void)
2054 {
2055 while (cur_active_branches
2056 && cur_active_branches >= max_active_branches) {
2057 uintmax_t min_commit = ULONG_MAX;
2058 struct branch *e, *l = NULL, *p = NULL;
2059
2060 for (e = active_branches; e; e = e->active_next_branch) {
2061 if (e->last_commit < min_commit) {
2062 p = l;
2063 min_commit = e->last_commit;
2064 }
2065 l = e;
2066 }
2067
2068 if (p) {
2069 e = p->active_next_branch;
2070 p->active_next_branch = e->active_next_branch;
2071 } else {
2072 e = active_branches;
2073 active_branches = e->active_next_branch;
2074 }
2075 e->active = 0;
2076 e->active_next_branch = NULL;
2077 if (e->branch_tree.tree) {
2078 release_tree_content_recursive(e->branch_tree.tree);
2079 e->branch_tree.tree = NULL;
2080 }
2081 cur_active_branches--;
2082 }
2083 }
2084
2085 static void load_branch(struct branch *b)
2086 {
2087 load_tree(&b->branch_tree);
2088 if (!b->active) {
2089 b->active = 1;
2090 b->active_next_branch = active_branches;
2091 active_branches = b;
2092 cur_active_branches++;
2093 branch_load_count++;
2094 }
2095 }
2096
2097 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2098 {
2099 unsigned char fanout = 0;
2100 while ((num_notes >>= 8))
2101 fanout++;
2102 return fanout;
2103 }
2104
2105 static void construct_path_with_fanout(const char *hex_sha1,
2106 unsigned char fanout, char *path)
2107 {
2108 unsigned int i = 0, j = 0;
2109 if (fanout >= the_hash_algo->rawsz)
2110 die(_("too large fanout (%u)"), fanout);
2111 while (fanout) {
2112 path[i++] = hex_sha1[j++];
2113 path[i++] = hex_sha1[j++];
2114 path[i++] = '/';
2115 fanout--;
2116 }
2117 memcpy(path + i, hex_sha1 + j, the_hash_algo->hexsz - j);
2118 path[i + the_hash_algo->hexsz - j] = '\0';
2119 }
2120
2121 static uintmax_t do_change_note_fanout(
2122 struct tree_entry *orig_root, struct tree_entry *root,
2123 char *hex_oid, unsigned int hex_oid_len,
2124 char *fullpath, unsigned int fullpath_len,
2125 unsigned char fanout)
2126 {
2127 struct tree_content *t;
2128 struct tree_entry *e, leaf;
2129 unsigned int i, tmp_hex_oid_len, tmp_fullpath_len;
2130 uintmax_t num_notes = 0;
2131 struct object_id oid;
2132 /* hex oid + '/' between each pair of hex digits + NUL */
2133 char realpath[GIT_MAX_HEXSZ + ((GIT_MAX_HEXSZ / 2) - 1) + 1];
2134 const unsigned hexsz = the_hash_algo->hexsz;
2135
2136 if (!root->tree)
2137 load_tree(root);
2138 t = root->tree;
2139
2140 for (i = 0; t && i < t->entry_count; i++) {
2141 e = t->entries[i];
2142 tmp_hex_oid_len = hex_oid_len + e->name->str_len;
2143 tmp_fullpath_len = fullpath_len;
2144
2145 /*
2146 * We're interested in EITHER existing note entries (entries
2147 * with exactly 40 hex chars in path, not including directory
2148 * separators), OR directory entries that may contain note
2149 * entries (with < 40 hex chars in path).
2150 * Also, each path component in a note entry must be a multiple
2151 * of 2 chars.
2152 */
2153 if (!e->versions[1].mode ||
2154 tmp_hex_oid_len > hexsz ||
2155 e->name->str_len % 2)
2156 continue;
2157
2158 /* This _may_ be a note entry, or a subdir containing notes */
2159 memcpy(hex_oid + hex_oid_len, e->name->str_dat,
2160 e->name->str_len);
2161 if (tmp_fullpath_len)
2162 fullpath[tmp_fullpath_len++] = '/';
2163 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2164 e->name->str_len);
2165 tmp_fullpath_len += e->name->str_len;
2166 fullpath[tmp_fullpath_len] = '\0';
2167
2168 if (tmp_hex_oid_len == hexsz && !get_oid_hex(hex_oid, &oid)) {
2169 /* This is a note entry */
2170 if (fanout == 0xff) {
2171 /* Counting mode, no rename */
2172 num_notes++;
2173 continue;
2174 }
2175 construct_path_with_fanout(hex_oid, fanout, realpath);
2176 if (!strcmp(fullpath, realpath)) {
2177 /* Note entry is in correct location */
2178 num_notes++;
2179 continue;
2180 }
2181
2182 /* Rename fullpath to realpath */
2183 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2184 die(_("failed to remove path %s"), fullpath);
2185 tree_content_set(orig_root, realpath,
2186 &leaf.versions[1].oid,
2187 leaf.versions[1].mode,
2188 leaf.tree);
2189 } else if (S_ISDIR(e->versions[1].mode)) {
2190 /* This is a subdir that may contain note entries */
2191 num_notes += do_change_note_fanout(orig_root, e,
2192 hex_oid, tmp_hex_oid_len,
2193 fullpath, tmp_fullpath_len, fanout);
2194 }
2195
2196 /* The above may have reallocated the current tree_content */
2197 t = root->tree;
2198 }
2199 return num_notes;
2200 }
2201
2202 static uintmax_t change_note_fanout(struct tree_entry *root,
2203 unsigned char fanout)
2204 {
2205 /*
2206 * The size of path is due to one slash between every two hex digits,
2207 * plus the terminating NUL. Note that there is no slash at the end, so
2208 * the number of slashes is one less than half the number of hex
2209 * characters.
2210 */
2211 char hex_oid[GIT_MAX_HEXSZ], path[GIT_MAX_HEXSZ + (GIT_MAX_HEXSZ / 2) - 1 + 1];
2212 return do_change_note_fanout(root, root, hex_oid, 0, path, 0, fanout);
2213 }
2214
2215 static int parse_mapped_oid_hex(const char *hex, struct object_id *oid, const char **end)
2216 {
2217 int algo;
2218 khiter_t it;
2219
2220 /* Make SHA-1 object IDs have all-zero padding. */
2221 memset(oid->hash, 0, sizeof(oid->hash));
2222
2223 algo = parse_oid_hex_any(hex, oid, end);
2224 if (algo == GIT_HASH_UNKNOWN)
2225 return -1;
2226
2227 it = kh_get_oid_map(sub_oid_map, *oid);
2228 /* No such object? */
2229 if (it == kh_end(sub_oid_map)) {
2230 /* If we're using the same algorithm, pass it through. */
2231 if (hash_algos[algo].format_id == the_hash_algo->format_id)
2232 return 0;
2233 return -1;
2234 }
2235 oidcpy(oid, kh_value(sub_oid_map, it));
2236 return 0;
2237 }
2238
2239 /*
2240 * Given a pointer into a string, parse a mark reference:
2241 *
2242 * idnum ::= ':' bigint;
2243 *
2244 * Update *endptr to point to the first character after the value.
2245 *
2246 * Complain if the following character is not what is expected,
2247 * either a space or end of the string.
2248 */
2249 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2250 {
2251 uintmax_t mark;
2252
2253 assert(*p == ':');
2254 p++;
2255 mark = strtoumax(p, endptr, 10);
2256 if (*endptr == p)
2257 die(_("no value after ':' in mark: %s"), command_buf.buf);
2258 return mark;
2259 }
2260
2261 /*
2262 * Parse the mark reference, and complain if this is not the end of
2263 * the string.
2264 */
2265 static uintmax_t parse_mark_ref_eol(const char *p)
2266 {
2267 char *end;
2268 uintmax_t mark;
2269
2270 mark = parse_mark_ref(p, &end);
2271 if (*end != '\0')
2272 die(_("garbage after mark: %s"), command_buf.buf);
2273 return mark;
2274 }
2275
2276 /*
2277 * Parse the mark reference, demanding a trailing space. Update *p to
2278 * point to the first character after the space.
2279 */
2280 static uintmax_t parse_mark_ref_space(const char **p)
2281 {
2282 uintmax_t mark;
2283 char *end;
2284
2285 mark = parse_mark_ref(*p, &end);
2286 if (*end++ != ' ')
2287 die(_("missing space after mark: %s"), command_buf.buf);
2288 *p = end;
2289 return mark;
2290 }
2291
2292 /*
2293 * Parse the path string into the strbuf. The path can either be quoted with
2294 * escape sequences or unquoted without escape sequences. Unquoted strings may
2295 * contain spaces only if `is_last_field` is nonzero; otherwise, it stops
2296 * parsing at the first space.
2297 */
2298 static void parse_path(struct strbuf *sb, const char *p, const char **endp,
2299 int is_last_field, const char *field)
2300 {
2301 if (*p == '"') {
2302 if (unquote_c_style(sb, p, endp))
2303 die(_("invalid %s: %s"), field, command_buf.buf);
2304 if (strlen(sb->buf) != sb->len)
2305 die(_("NUL in %s: %s"), field, command_buf.buf);
2306 } else {
2307 /*
2308 * Unless we are parsing the last field of a line,
2309 * SP is the end of this field.
2310 */
2311 *endp = is_last_field
2312 ? p + strlen(p)
2313 : strchrnul(p, ' ');
2314 strbuf_add(sb, p, *endp - p);
2315 }
2316 }
2317
2318 /*
2319 * Parse the path string into the strbuf, and complain if this is not the end of
2320 * the string. Unquoted strings may contain spaces.
2321 */
2322 static void parse_path_eol(struct strbuf *sb, const char *p, const char *field)
2323 {
2324 const char *end;
2325
2326 parse_path(sb, p, &end, 1, field);
2327 if (*end)
2328 die(_("garbage after %s: %s"), field, command_buf.buf);
2329 }
2330
2331 /*
2332 * Parse the path string into the strbuf, and ensure it is followed by a space.
2333 * Unquoted strings may not contain spaces. Update *endp to point to the first
2334 * character after the space.
2335 */
2336 static void parse_path_space(struct strbuf *sb, const char *p,
2337 const char **endp, const char *field)
2338 {
2339 parse_path(sb, p, endp, 0, field);
2340 if (**endp != ' ')
2341 die(_("missing space after %s: %s"), field, command_buf.buf);
2342 (*endp)++;
2343 }
2344
2345 static void file_change_m(const char *p, struct branch *b)
2346 {
2347 static struct strbuf path = STRBUF_INIT;
2348 struct object_entry *oe;
2349 struct object_id oid;
2350 uint16_t mode, inline_data = 0;
2351
2352 p = parse_mode(p, &mode);
2353 if (!p)
2354 die(_("corrupt mode: %s"), command_buf.buf);
2355 switch (mode) {
2356 case 0644:
2357 case 0755:
2358 mode |= S_IFREG;
2359 case S_IFREG | 0644:
2360 case S_IFREG | 0755:
2361 case S_IFLNK:
2362 case S_IFDIR:
2363 case S_IFGITLINK:
2364 /* ok */
2365 break;
2366 default:
2367 die(_("corrupt mode: %s"), command_buf.buf);
2368 }
2369
2370 if (*p == ':') {
2371 oe = find_mark(marks, parse_mark_ref_space(&p));
2372 oidcpy(&oid, &oe->idx.oid);
2373 } else if (skip_prefix(p, "inline ", &p)) {
2374 inline_data = 1;
2375 oe = NULL; /* not used with inline_data, but makes gcc happy */
2376 } else {
2377 if (parse_mapped_oid_hex(p, &oid, &p))
2378 die(_("invalid dataref: %s"), command_buf.buf);
2379 oe = find_object(&oid);
2380 if (*p++ != ' ')
2381 die(_("missing space after SHA1: %s"), command_buf.buf);
2382 }
2383
2384 strbuf_reset(&path);
2385 parse_path_eol(&path, p, "path");
2386
2387 /* Git does not track empty, non-toplevel directories. */
2388 if (S_ISDIR(mode) &&
2389 is_empty_tree_oid(&oid, the_repository->hash_algo) &&
2390 *path.buf) {
2391 tree_content_remove(&b->branch_tree, path.buf, NULL, 0);
2392 return;
2393 }
2394
2395 if (S_ISGITLINK(mode)) {
2396 if (inline_data)
2397 die(_("Git links cannot be specified 'inline': %s"),
2398 command_buf.buf);
2399 else if (oe) {
2400 if (oe->type != OBJ_COMMIT)
2401 die(_("not a commit (actually a %s): %s"),
2402 type_name(oe->type), command_buf.buf);
2403 }
2404 /*
2405 * Accept the sha1 without checking; it expected to be in
2406 * another repository.
2407 */
2408 } else if (inline_data) {
2409 if (S_ISDIR(mode))
2410 die(_("directories cannot be specified 'inline': %s"),
2411 command_buf.buf);
2412 while (read_next_command() != EOF) {
2413 const char *v;
2414 if (skip_prefix(command_buf.buf, "cat-blob ", &v))
2415 parse_cat_blob(v);
2416 else {
2417 parse_and_store_blob(&last_blob, &oid, 0);
2418 break;
2419 }
2420 }
2421 } else {
2422 enum object_type expected = S_ISDIR(mode) ?
2423 OBJ_TREE: OBJ_BLOB;
2424 enum object_type type = oe ? oe->type :
2425 odb_read_object_info(the_repository->objects,
2426 &oid, NULL);
2427 if (type < 0)
2428 die(_("%s not found: %s"),
2429 S_ISDIR(mode) ? _("tree") : _("blob"),
2430 command_buf.buf);
2431 if (type != expected)
2432 die(_("not a %s (actually a %s): %s"),
2433 type_name(expected), type_name(type),
2434 command_buf.buf);
2435 }
2436
2437 if (!*path.buf) {
2438 tree_content_replace(&b->branch_tree, &oid, mode, NULL);
2439 return;
2440 }
2441
2442 if (!verify_path(path.buf, mode))
2443 die(_("invalid path '%s'"), path.buf);
2444 tree_content_set(&b->branch_tree, path.buf, &oid, mode, NULL);
2445 }
2446
2447 static void file_change_d(const char *p, struct branch *b)
2448 {
2449 static struct strbuf path = STRBUF_INIT;
2450
2451 strbuf_reset(&path);
2452 parse_path_eol(&path, p, "path");
2453 tree_content_remove(&b->branch_tree, path.buf, NULL, 1);
2454 }
2455
2456 static void file_change_cr(const char *p, struct branch *b, int rename)
2457 {
2458 static struct strbuf source = STRBUF_INIT;
2459 static struct strbuf dest = STRBUF_INIT;
2460 struct tree_entry leaf;
2461
2462 strbuf_reset(&source);
2463 parse_path_space(&source, p, &p, "source");
2464 strbuf_reset(&dest);
2465 parse_path_eol(&dest, p, "dest");
2466
2467 memset(&leaf, 0, sizeof(leaf));
2468 if (rename)
2469 tree_content_remove(&b->branch_tree, source.buf, &leaf, 1);
2470 else
2471 tree_content_get(&b->branch_tree, source.buf, &leaf, 1);
2472 if (!leaf.versions[1].mode)
2473 die(_("path %s not in branch"), source.buf);
2474 if (!*dest.buf) { /* C "path/to/subdir" "" */
2475 tree_content_replace(&b->branch_tree,
2476 &leaf.versions[1].oid,
2477 leaf.versions[1].mode,
2478 leaf.tree);
2479 return;
2480 }
2481 if (!verify_path(dest.buf, leaf.versions[1].mode))
2482 die(_("invalid path '%s'"), dest.buf);
2483 tree_content_set(&b->branch_tree, dest.buf,
2484 &leaf.versions[1].oid,
2485 leaf.versions[1].mode,
2486 leaf.tree);
2487 }
2488
2489 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2490 {
2491 struct object_entry *oe;
2492 struct branch *s;
2493 struct object_id oid, commit_oid;
2494 char path[GIT_MAX_RAWSZ * 3];
2495 uint16_t inline_data = 0;
2496 unsigned char new_fanout;
2497
2498 /*
2499 * When loading a branch, we don't traverse its tree to count the real
2500 * number of notes (too expensive to do this for all non-note refs).
2501 * This means that recently loaded notes refs might incorrectly have
2502 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2503 *
2504 * Fix this by traversing the tree and counting the number of notes
2505 * when b->num_notes == 0. If the notes tree is truly empty, the
2506 * calculation should not take long.
2507 */
2508 if (b->num_notes == 0 && *old_fanout == 0) {
2509 /* Invoke change_note_fanout() in "counting mode". */
2510 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2511 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2512 }
2513
2514 /* Now parse the notemodify command. */
2515 /* <dataref> or 'inline' */
2516 if (*p == ':') {
2517 oe = find_mark(marks, parse_mark_ref_space(&p));
2518 oidcpy(&oid, &oe->idx.oid);
2519 } else if (skip_prefix(p, "inline ", &p)) {
2520 inline_data = 1;
2521 oe = NULL; /* not used with inline_data, but makes gcc happy */
2522 } else {
2523 if (parse_mapped_oid_hex(p, &oid, &p))
2524 die(_("invalid dataref: %s"), command_buf.buf);
2525 oe = find_object(&oid);
2526 if (*p++ != ' ')
2527 die(_("missing space after SHA1: %s"), command_buf.buf);
2528 }
2529
2530 /* <commit-ish> */
2531 s = lookup_branch(p);
2532 if (s) {
2533 if (is_null_oid(&s->oid))
2534 die(_("can't add a note on empty branch."));
2535 oidcpy(&commit_oid, &s->oid);
2536 } else if (*p == ':') {
2537 uintmax_t commit_mark = parse_mark_ref_eol(p);
2538 struct object_entry *commit_oe = find_mark(marks, commit_mark);
2539 if (commit_oe->type != OBJ_COMMIT)
2540 die(_("mark :%" PRIuMAX " not a commit"), commit_mark);
2541 oidcpy(&commit_oid, &commit_oe->idx.oid);
2542 } else if (!repo_get_oid(the_repository, p, &commit_oid)) {
2543 unsigned long size;
2544 char *buf = odb_read_object_peeled(the_repository->objects,
2545 &commit_oid, OBJ_COMMIT, &size,
2546 &commit_oid);
2547 if (!buf || size < the_hash_algo->hexsz + 6)
2548 die(_("not a valid commit: %s"), p);
2549 free(buf);
2550 } else
2551 die(_("invalid ref name or SHA1 expression: %s"), p);
2552
2553 if (inline_data) {
2554 read_next_command();
2555 parse_and_store_blob(&last_blob, &oid, 0);
2556 } else if (oe) {
2557 if (oe->type != OBJ_BLOB)
2558 die(_("not a blob (actually a %s): %s"),
2559 type_name(oe->type), command_buf.buf);
2560 } else if (!is_null_oid(&oid)) {
2561 enum object_type type = odb_read_object_info(the_repository->objects, &oid,
2562 NULL);
2563 if (type < 0)
2564 die(_("blob not found: %s"), command_buf.buf);
2565 if (type != OBJ_BLOB)
2566 die(_("not a blob (actually a %s): %s"),
2567 type_name(type), command_buf.buf);
2568 }
2569
2570 construct_path_with_fanout(oid_to_hex(&commit_oid), *old_fanout, path);
2571 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2572 b->num_notes--;
2573
2574 if (is_null_oid(&oid))
2575 return; /* nothing to insert */
2576
2577 b->num_notes++;
2578 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2579 construct_path_with_fanout(oid_to_hex(&commit_oid), new_fanout, path);
2580 tree_content_set(&b->branch_tree, path, &oid, S_IFREG | 0644, NULL);
2581 }
2582
2583 static void file_change_deleteall(struct branch *b)
2584 {
2585 release_tree_content_recursive(b->branch_tree.tree);
2586 oidclr(&b->branch_tree.versions[0].oid, the_repository->hash_algo);
2587 oidclr(&b->branch_tree.versions[1].oid, the_repository->hash_algo);
2588 load_tree(&b->branch_tree);
2589 b->num_notes = 0;
2590 }
2591
2592 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2593 {
2594 if (!buf || size < the_hash_algo->hexsz + 6)
2595 die(_("not a valid commit: %s"), oid_to_hex(&b->oid));
2596 if (memcmp("tree ", buf, 5)
2597 || get_oid_hex(buf + 5, &b->branch_tree.versions[1].oid))
2598 die(_("the commit %s is corrupt"), oid_to_hex(&b->oid));
2599 oidcpy(&b->branch_tree.versions[0].oid,
2600 &b->branch_tree.versions[1].oid);
2601 }
2602
2603 static void parse_from_existing(struct branch *b)
2604 {
2605 if (is_null_oid(&b->oid)) {
2606 oidclr(&b->branch_tree.versions[0].oid, the_repository->hash_algo);
2607 oidclr(&b->branch_tree.versions[1].oid, the_repository->hash_algo);
2608 } else {
2609 unsigned long size;
2610 char *buf;
2611
2612 buf = odb_read_object_peeled(the_repository->objects, &b->oid,
2613 OBJ_COMMIT, &size, &b->oid);
2614 parse_from_commit(b, buf, size);
2615 free(buf);
2616 }
2617 }
2618
2619 static int parse_objectish(struct branch *b, const char *objectish)
2620 {
2621 struct branch *s;
2622 struct object_id oid;
2623
2624 oidcpy(&oid, &b->branch_tree.versions[1].oid);
2625
2626 s = lookup_branch(objectish);
2627 if (b == s)
2628 die(_("can't create a branch from itself: %s"), b->name);
2629 else if (s) {
2630 struct object_id *t = &s->branch_tree.versions[1].oid;
2631 oidcpy(&b->oid, &s->oid);
2632 oidcpy(&b->branch_tree.versions[0].oid, t);
2633 oidcpy(&b->branch_tree.versions[1].oid, t);
2634 } else if (*objectish == ':') {
2635 uintmax_t idnum = parse_mark_ref_eol(objectish);
2636 struct object_entry *oe = find_mark(marks, idnum);
2637 if (oe->type != OBJ_COMMIT)
2638 die(_("mark :%" PRIuMAX " not a commit"), idnum);
2639 if (!oideq(&b->oid, &oe->idx.oid)) {
2640 oidcpy(&b->oid, &oe->idx.oid);
2641 if (oe->pack_id != MAX_PACK_ID) {
2642 unsigned long size;
2643 char *buf = gfi_unpack_entry(oe, &size);
2644 parse_from_commit(b, buf, size);
2645 free(buf);
2646 } else
2647 parse_from_existing(b);
2648 }
2649 } else if (!repo_get_oid(the_repository, objectish, &b->oid)) {
2650 parse_from_existing(b);
2651 if (is_null_oid(&b->oid))
2652 b->delete = 1;
2653 }
2654 else
2655 die(_("invalid ref name or SHA1 expression: %s"), objectish);
2656
2657 if (b->branch_tree.tree && !oideq(&oid, &b->branch_tree.versions[1].oid)) {
2658 release_tree_content_recursive(b->branch_tree.tree);
2659 b->branch_tree.tree = NULL;
2660 }
2661
2662 read_next_command();
2663 return 1;
2664 }
2665
2666 static int parse_from(struct branch *b)
2667 {
2668 const char *from;
2669
2670 if (!skip_prefix(command_buf.buf, "from ", &from))
2671 return 0;
2672
2673 return parse_objectish(b, from);
2674 }
2675
2676 static int parse_objectish_with_prefix(struct branch *b, const char *prefix)
2677 {
2678 const char *base;
2679
2680 if (!skip_prefix(command_buf.buf, prefix, &base))
2681 return 0;
2682
2683 return parse_objectish(b, base);
2684 }
2685
2686 static struct hash_list *parse_merge(unsigned int *count)
2687 {
2688 struct hash_list *list = NULL, **tail = &list, *n;
2689 const char *from;
2690 struct branch *s;
2691
2692 *count = 0;
2693 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2694 n = xmalloc(sizeof(*n));
2695 s = lookup_branch(from);
2696 if (s)
2697 oidcpy(&n->oid, &s->oid);
2698 else if (*from == ':') {
2699 uintmax_t idnum = parse_mark_ref_eol(from);
2700 struct object_entry *oe = find_mark(marks, idnum);
2701 if (oe->type != OBJ_COMMIT)
2702 die(_("mark :%" PRIuMAX " not a commit"), idnum);
2703 oidcpy(&n->oid, &oe->idx.oid);
2704 } else if (!repo_get_oid(the_repository, from, &n->oid)) {
2705 unsigned long size;
2706 char *buf = odb_read_object_peeled(the_repository->objects,
2707 &n->oid, OBJ_COMMIT,
2708 &size, &n->oid);
2709 if (!buf || size < the_hash_algo->hexsz + 6)
2710 die(_("not a valid commit: %s"), from);
2711 free(buf);
2712 } else
2713 die(_("invalid ref name or SHA1 expression: %s"), from);
2714
2715 n->next = NULL;
2716 *tail = n;
2717 tail = &n->next;
2718
2719 (*count)++;
2720 read_next_command();
2721 }
2722 return list;
2723 }
2724
2725 struct signature_data {
2726 char *hash_algo; /* "sha1" or "sha256" */
2727 char *sig_format; /* "openpgp", "x509", "ssh", or "unknown" */
2728 struct strbuf data; /* The actual signature data */
2729 };
2730
2731 static void parse_one_signature(struct signature_data *sig, const char *v)
2732 {
2733 char *args = xstrdup(v); /* Will be freed when sig->hash_algo is freed */
2734 char *space = strchr(args, ' ');
2735
2736 if (!space)
2737 die(_("expected gpgsig format: 'gpgsig <hash-algo> <signature-format>', "
2738 "got 'gpgsig %s'"), args);
2739 *space = '\0';
2740
2741 sig->hash_algo = args;
2742 sig->sig_format = space + 1;
2743
2744 /* Validate hash algorithm */
2745 if (strcmp(sig->hash_algo, "sha1") &&
2746 strcmp(sig->hash_algo, "sha256"))
2747 die(_("unknown git hash algorithm in gpgsig: '%s'"), sig->hash_algo);
2748
2749 /* Validate signature format */
2750 if (!valid_signature_format(sig->sig_format))
2751 die(_("invalid signature format in gpgsig: '%s'"), sig->sig_format);
2752 if (!strcmp(sig->sig_format, "unknown"))
2753 warning(_("'unknown' signature format in gpgsig"));
2754
2755 /* Read signature data */
2756 read_next_command();
2757 parse_data(&sig->data, 0, NULL);
2758 }
2759
2760 static void discard_one_signature(void)
2761 {
2762 struct strbuf data = STRBUF_INIT;
2763
2764 read_next_command();
2765 parse_data(&data, 0, NULL);
2766 strbuf_release(&data);
2767 }
2768
2769 static void add_gpgsig_to_commit(struct strbuf *commit_data,
2770 const char *header,
2771 struct signature_data *sig)
2772 {
2773 struct string_list siglines = STRING_LIST_INIT_NODUP;
2774
2775 if (!sig->hash_algo)
2776 return;
2777
2778 strbuf_addstr(commit_data, header);
2779 string_list_split_in_place(&siglines, sig->data.buf, "\n", -1);
2780 strbuf_add_separated_string_list(commit_data, "\n ", &siglines);
2781 strbuf_addch(commit_data, '\n');
2782 string_list_clear(&siglines, 1);
2783 strbuf_release(&sig->data);
2784 free(sig->hash_algo);
2785 }
2786
2787 static void store_signature(struct signature_data *stored_sig,
2788 struct signature_data *new_sig,
2789 const char *hash_type)
2790 {
2791 if (stored_sig->hash_algo) {
2792 warning(_("multiple %s signatures found, "
2793 "ignoring additional signature"),
2794 hash_type);
2795 strbuf_release(&new_sig->data);
2796 free(new_sig->hash_algo);
2797 } else {
2798 *stored_sig = *new_sig;
2799 }
2800 }
2801
2802 static void import_one_signature(struct signature_data *sig_sha1,
2803 struct signature_data *sig_sha256,
2804 const char *v)
2805 {
2806 struct signature_data sig = { NULL, NULL, STRBUF_INIT };
2807
2808 parse_one_signature(&sig, v);
2809
2810 if (!strcmp(sig.hash_algo, "sha1"))
2811 store_signature(sig_sha1, &sig, "SHA-1");
2812 else if (!strcmp(sig.hash_algo, "sha256"))
2813 store_signature(sig_sha256, &sig, "SHA-256");
2814 else
2815 die(_("parse_one_signature() returned unknown hash algo"));
2816 }
2817
2818 static void parse_new_commit(const char *arg)
2819 {
2820 static struct strbuf msg = STRBUF_INIT;
2821 struct signature_data sig_sha1 = { NULL, NULL, STRBUF_INIT };
2822 struct signature_data sig_sha256 = { NULL, NULL, STRBUF_INIT };
2823 struct branch *b;
2824 char *author = NULL;
2825 char *committer = NULL;
2826 char *encoding = NULL;
2827 struct hash_list *merge_list = NULL;
2828 unsigned int merge_count;
2829 unsigned char prev_fanout, new_fanout;
2830 const char *v;
2831
2832 b = lookup_branch(arg);
2833 if (!b)
2834 b = new_branch(arg);
2835
2836 read_next_command();
2837 parse_mark();
2838 parse_original_identifier();
2839 if (skip_prefix(command_buf.buf, "author ", &v)) {
2840 author = parse_ident(v);
2841 read_next_command();
2842 }
2843 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2844 committer = parse_ident(v);
2845 read_next_command();
2846 }
2847 if (!committer)
2848 die(_("expected committer but didn't get one"));
2849
2850 while (skip_prefix(command_buf.buf, "gpgsig ", &v)) {
2851 switch (signed_commit_mode) {
2852
2853 /* First, modes that don't need the signature to be parsed */
2854 case SIGN_ABORT:
2855 die(_("encountered signed commit; use "
2856 "--signed-commits=<mode> to handle it"));
2857 case SIGN_WARN_STRIP:
2858 warning(_("stripping a commit signature"));
2859 /* fallthru */
2860 case SIGN_STRIP:
2861 discard_one_signature();
2862 break;
2863
2864 /* Second, modes that parse the signature */
2865 case SIGN_WARN_VERBATIM:
2866 warning(_("importing a commit signature verbatim"));
2867 /* fallthru */
2868 case SIGN_VERBATIM:
2869 import_one_signature(&sig_sha1, &sig_sha256, v);
2870 break;
2871
2872 /* Third, BUG */
2873 default:
2874 BUG("invalid signed_commit_mode value %d", signed_commit_mode);
2875 }
2876 read_next_command();
2877 }
2878
2879 if (skip_prefix(command_buf.buf, "encoding ", &v)) {
2880 encoding = xstrdup(v);
2881 read_next_command();
2882 }
2883 parse_data(&msg, 0, NULL);
2884 read_next_command();
2885 parse_from(b);
2886 merge_list = parse_merge(&merge_count);
2887
2888 /* ensure the branch is active/loaded */
2889 if (!b->branch_tree.tree || !max_active_branches) {
2890 unload_one_branch();
2891 load_branch(b);
2892 }
2893
2894 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2895
2896 /* file_change* */
2897 while (command_buf.len > 0) {
2898 if (skip_prefix(command_buf.buf, "M ", &v))
2899 file_change_m(v, b);
2900 else if (skip_prefix(command_buf.buf, "D ", &v))
2901 file_change_d(v, b);
2902 else if (skip_prefix(command_buf.buf, "R ", &v))
2903 file_change_cr(v, b, 1);
2904 else if (skip_prefix(command_buf.buf, "C ", &v))
2905 file_change_cr(v, b, 0);
2906 else if (skip_prefix(command_buf.buf, "N ", &v))
2907 note_change_n(v, b, &prev_fanout);
2908 else if (!strcmp("deleteall", command_buf.buf))
2909 file_change_deleteall(b);
2910 else if (skip_prefix(command_buf.buf, "ls ", &v))
2911 parse_ls(v, b);
2912 else if (skip_prefix(command_buf.buf, "cat-blob ", &v))
2913 parse_cat_blob(v);
2914 else {
2915 unread_command_buf = 1;
2916 break;
2917 }
2918 if (read_next_command() == EOF)
2919 break;
2920 }
2921
2922 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2923 if (new_fanout != prev_fanout)
2924 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2925
2926 /* build the tree and the commit */
2927 store_tree(&b->branch_tree);
2928 oidcpy(&b->branch_tree.versions[0].oid,
2929 &b->branch_tree.versions[1].oid);
2930
2931 strbuf_reset(&new_data);
2932 strbuf_addf(&new_data, "tree %s\n",
2933 oid_to_hex(&b->branch_tree.versions[1].oid));
2934 if (!is_null_oid(&b->oid))
2935 strbuf_addf(&new_data, "parent %s\n",
2936 oid_to_hex(&b->oid));
2937 while (merge_list) {
2938 struct hash_list *next = merge_list->next;
2939 strbuf_addf(&new_data, "parent %s\n",
2940 oid_to_hex(&merge_list->oid));
2941 free(merge_list);
2942 merge_list = next;
2943 }
2944 strbuf_addf(&new_data,
2945 "author %s\n"
2946 "committer %s\n",
2947 author ? author : committer, committer);
2948 if (encoding)
2949 strbuf_addf(&new_data,
2950 "encoding %s\n",
2951 encoding);
2952
2953 add_gpgsig_to_commit(&new_data, "gpgsig ", &sig_sha1);
2954 add_gpgsig_to_commit(&new_data, "gpgsig-sha256 ", &sig_sha256);
2955
2956 strbuf_addch(&new_data, '\n');
2957 strbuf_addbuf(&new_data, &msg);
2958 free(author);
2959 free(committer);
2960 free(encoding);
2961
2962 if (!store_object(OBJ_COMMIT, &new_data, NULL, &b->oid, next_mark))
2963 b->pack_id = pack_id;
2964 b->last_commit = object_count_by_type[OBJ_COMMIT];
2965 }
2966
2967 static void handle_tag_signature(struct strbuf *msg, const char *name)
2968 {
2969 size_t sig_offset = parse_signed_buffer(msg->buf, msg->len);
2970
2971 /* If there is no signature, there is nothing to do. */
2972 if (sig_offset >= msg->len)
2973 return;
2974
2975 switch (signed_tag_mode) {
2976
2977 /* First, modes that don't change anything */
2978 case SIGN_ABORT:
2979 die(_("encountered signed tag; use "
2980 "--signed-tags=<mode> to handle it"));
2981 case SIGN_WARN_VERBATIM:
2982 warning(_("importing a tag signature verbatim for tag '%s'"), name);
2983 /* fallthru */
2984 case SIGN_VERBATIM:
2985 /* Nothing to do, the signature will be put into the imported tag. */
2986 break;
2987
2988 /* Second, modes that remove the signature */
2989 case SIGN_WARN_STRIP:
2990 warning(_("stripping a tag signature for tag '%s'"), name);
2991 /* fallthru */
2992 case SIGN_STRIP:
2993 /* Truncate the buffer to remove the signature */
2994 strbuf_setlen(msg, sig_offset);
2995 break;
2996
2997 /* Third, BUG */
2998 default:
2999 BUG("invalid signed_tag_mode value %d from tag '%s'",
3000 signed_tag_mode, name);
3001 }
3002 }
3003
3004 static void parse_new_tag(const char *arg)
3005 {
3006 static struct strbuf msg = STRBUF_INIT;
3007 const char *from;
3008 char *tagger;
3009 struct branch *s;
3010 struct tag *t;
3011 uintmax_t from_mark = 0;
3012 struct object_id oid;
3013 enum object_type type;
3014 const char *v;
3015
3016 t = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct tag));
3017 t->name = mem_pool_strdup(&fi_mem_pool, arg);
3018 if (last_tag)
3019 last_tag->next_tag = t;
3020 else
3021 first_tag = t;
3022 last_tag = t;
3023 read_next_command();
3024 parse_mark();
3025
3026 /* from ... */
3027 if (!skip_prefix(command_buf.buf, "from ", &from))
3028 die(_("expected 'from' command, got '%s'"), command_buf.buf);
3029 s = lookup_branch(from);
3030 if (s) {
3031 if (is_null_oid(&s->oid))
3032 die(_("can't tag an empty branch."));
3033 oidcpy(&oid, &s->oid);
3034 type = OBJ_COMMIT;
3035 } else if (*from == ':') {
3036 struct object_entry *oe;
3037 from_mark = parse_mark_ref_eol(from);
3038 oe = find_mark(marks, from_mark);
3039 type = oe->type;
3040 oidcpy(&oid, &oe->idx.oid);
3041 } else if (!repo_get_oid(the_repository, from, &oid)) {
3042 struct object_entry *oe = find_object(&oid);
3043 if (!oe) {
3044 type = odb_read_object_info(the_repository->objects,
3045 &oid, NULL);
3046 if (type < 0)
3047 die(_("not a valid object: %s"), from);
3048 } else
3049 type = oe->type;
3050 } else
3051 die(_("invalid ref name or SHA1 expression: %s"), from);
3052 read_next_command();
3053
3054 /* original-oid ... */
3055 parse_original_identifier();
3056
3057 /* tagger ... */
3058 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
3059 tagger = parse_ident(v);
3060 read_next_command();
3061 } else
3062 tagger = NULL;
3063
3064 /* tag payload/message */
3065 parse_data(&msg, 0, NULL);
3066
3067 handle_tag_signature(&msg, t->name);
3068
3069 /* build the tag object */
3070 strbuf_reset(&new_data);
3071
3072 strbuf_addf(&new_data,
3073 "object %s\n"
3074 "type %s\n"
3075 "tag %s\n",
3076 oid_to_hex(&oid), type_name(type), t->name);
3077 if (tagger)
3078 strbuf_addf(&new_data,
3079 "tagger %s\n", tagger);
3080 strbuf_addch(&new_data, '\n');
3081 strbuf_addbuf(&new_data, &msg);
3082 free(tagger);
3083
3084 if (store_object(OBJ_TAG, &new_data, NULL, &t->oid, next_mark))
3085 t->pack_id = MAX_PACK_ID;
3086 else
3087 t->pack_id = pack_id;
3088 }
3089
3090 static void parse_reset_branch(const char *arg)
3091 {
3092 struct branch *b;
3093 const char *tag_name;
3094
3095 b = lookup_branch(arg);
3096 if (b) {
3097 oidclr(&b->oid, the_repository->hash_algo);
3098 oidclr(&b->branch_tree.versions[0].oid, the_repository->hash_algo);
3099 oidclr(&b->branch_tree.versions[1].oid, the_repository->hash_algo);
3100 if (b->branch_tree.tree) {
3101 release_tree_content_recursive(b->branch_tree.tree);
3102 b->branch_tree.tree = NULL;
3103 }
3104 }
3105 else
3106 b = new_branch(arg);
3107 read_next_command();
3108 parse_from(b);
3109 if (b->delete && skip_prefix(b->name, "refs/tags/", &tag_name)) {
3110 /*
3111 * Elsewhere, we call dump_branches() before dump_tags(),
3112 * and dump_branches() will handle ref deletions first, so
3113 * in order to make sure the deletion actually takes effect,
3114 * we need to remove the tag from our list of tags to update.
3115 *
3116 * NEEDSWORK: replace list of tags with hashmap for faster
3117 * deletion?
3118 */
3119 struct tag *t, *prev = NULL;
3120 for (t = first_tag; t; t = t->next_tag) {
3121 if (!strcmp(t->name, tag_name))
3122 break;
3123 prev = t;
3124 }
3125 if (t) {
3126 if (prev)
3127 prev->next_tag = t->next_tag;
3128 else
3129 first_tag = t->next_tag;
3130 if (!t->next_tag)
3131 last_tag = prev;
3132 /* There is no mem_pool_free(t) function to call. */
3133 }
3134 }
3135 if (command_buf.len > 0)
3136 unread_command_buf = 1;
3137 }
3138
3139 static void cat_blob_write(const char *buf, unsigned long size)
3140 {
3141 if (write_in_full(cat_blob_fd, buf, size) < 0)
3142 die_errno(_("write to frontend failed"));
3143 }
3144
3145 static void cat_blob(struct object_entry *oe, struct object_id *oid)
3146 {
3147 struct strbuf line = STRBUF_INIT;
3148 unsigned long size;
3149 enum object_type type = 0;
3150 char *buf;
3151
3152 if (!oe || oe->pack_id == MAX_PACK_ID) {
3153 buf = odb_read_object(the_repository->objects, oid, &type, &size);
3154 } else {
3155 type = oe->type;
3156 buf = gfi_unpack_entry(oe, &size);
3157 }
3158
3159 /*
3160 * Output based on batch_one_object() from cat-file.c.
3161 */
3162 if (type <= 0) {
3163 strbuf_reset(&line);
3164 strbuf_addf(&line, "%s missing\n", oid_to_hex(oid));
3165 cat_blob_write(line.buf, line.len);
3166 strbuf_release(&line);
3167 free(buf);
3168 return;
3169 }
3170 if (!buf)
3171 die(_("can't read object %s"), oid_to_hex(oid));
3172 if (type != OBJ_BLOB)
3173 die(_("object %s is a %s but a blob was expected."),
3174 oid_to_hex(oid), type_name(type));
3175 strbuf_reset(&line);
3176 strbuf_addf(&line, "%s %s %"PRIuMAX"\n", oid_to_hex(oid),
3177 type_name(type), (uintmax_t)size);
3178 cat_blob_write(line.buf, line.len);
3179 strbuf_release(&line);
3180 cat_blob_write(buf, size);
3181 cat_blob_write("\n", 1);
3182 if (oe && oe->pack_id == pack_id) {
3183 last_blob.offset = oe->idx.offset;
3184 strbuf_attach(&last_blob.data, buf, size, size);
3185 last_blob.depth = oe->depth;
3186 } else
3187 free(buf);
3188 }
3189
3190 static void parse_get_mark(const char *p)
3191 {
3192 struct object_entry *oe;
3193 char output[GIT_MAX_HEXSZ + 2];
3194
3195 /* get-mark SP <object> LF */
3196 if (*p != ':')
3197 die(_("not a mark: %s"), p);
3198
3199 oe = find_mark(marks, parse_mark_ref_eol(p));
3200 if (!oe)
3201 die(_("unknown mark: %s"), command_buf.buf);
3202
3203 xsnprintf(output, sizeof(output), "%s\n", oid_to_hex(&oe->idx.oid));
3204 cat_blob_write(output, the_hash_algo->hexsz + 1);
3205 }
3206
3207 static void parse_cat_blob(const char *p)
3208 {
3209 struct object_entry *oe;
3210 struct object_id oid;
3211
3212 /* cat-blob SP <object> LF */
3213 if (*p == ':') {
3214 oe = find_mark(marks, parse_mark_ref_eol(p));
3215 if (!oe)
3216 die(_("unknown mark: %s"), command_buf.buf);
3217 oidcpy(&oid, &oe->idx.oid);
3218 } else {
3219 if (parse_mapped_oid_hex(p, &oid, &p))
3220 die(_("invalid dataref: %s"), command_buf.buf);
3221 if (*p)
3222 die(_("garbage after SHA1: %s"), command_buf.buf);
3223 oe = find_object(&oid);
3224 }
3225
3226 cat_blob(oe, &oid);
3227 }
3228
3229 static struct object_entry *dereference(struct object_entry *oe,
3230 struct object_id *oid)
3231 {
3232 unsigned long size;
3233 char *buf = NULL;
3234 const unsigned hexsz = the_hash_algo->hexsz;
3235
3236 if (!oe) {
3237 enum object_type type = odb_read_object_info(the_repository->objects,
3238 oid, NULL);
3239 if (type < 0)
3240 die(_("object not found: %s"), oid_to_hex(oid));
3241 /* cache it! */
3242 oe = insert_object(oid);
3243 oe->type = type;
3244 oe->pack_id = MAX_PACK_ID;
3245 oe->idx.offset = 1;
3246 }
3247 switch (oe->type) {
3248 case OBJ_TREE: /* easy case. */
3249 return oe;
3250 case OBJ_COMMIT:
3251 case OBJ_TAG:
3252 break;
3253 default:
3254 die(_("not a tree-ish: %s"), command_buf.buf);
3255 }
3256
3257 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
3258 buf = gfi_unpack_entry(oe, &size);
3259 } else {
3260 enum object_type unused;
3261 buf = odb_read_object(the_repository->objects, oid,
3262 &unused, &size);
3263 }
3264 if (!buf)
3265 die(_("can't load object %s"), oid_to_hex(oid));
3266
3267 /* Peel one layer. */
3268 switch (oe->type) {
3269 case OBJ_TAG:
3270 if (size < hexsz + strlen("object ") ||
3271 get_oid_hex(buf + strlen("object "), oid))
3272 die(_("invalid SHA1 in tag: %s"), command_buf.buf);
3273 break;
3274 case OBJ_COMMIT:
3275 if (size < hexsz + strlen("tree ") ||
3276 get_oid_hex(buf + strlen("tree "), oid))
3277 die(_("invalid SHA1 in commit: %s"), command_buf.buf);
3278 }
3279
3280 free(buf);
3281 return find_object(oid);
3282 }
3283
3284 static void insert_mapped_mark(uintmax_t mark, void *object, void *cbp)
3285 {
3286 struct object_id *fromoid = object;
3287 struct object_id *tooid = find_mark(cbp, mark);
3288 int ret;
3289 khiter_t it;
3290
3291 it = kh_put_oid_map(sub_oid_map, *fromoid, &ret);
3292 /* We've already seen this object. */
3293 if (ret == 0)
3294 return;
3295 kh_value(sub_oid_map, it) = tooid;
3296 }
3297
3298 static void build_mark_map_one(struct mark_set *from, struct mark_set *to)
3299 {
3300 for_each_mark(from, 0, insert_mapped_mark, to);
3301 }
3302
3303 static void build_mark_map(struct string_list *from, struct string_list *to)
3304 {
3305 struct string_list_item *fromp, *top;
3306
3307 sub_oid_map = kh_init_oid_map();
3308
3309 for_each_string_list_item(fromp, from) {
3310 top = string_list_lookup(to, fromp->string);
3311 if (!fromp->util) {
3312 die(_("missing from marks for submodule '%s'"), fromp->string);
3313 } else if (!top || !top->util) {
3314 die(_("missing to marks for submodule '%s'"), fromp->string);
3315 }
3316 build_mark_map_one(fromp->util, top->util);
3317 }
3318 }
3319
3320 static struct object_entry *parse_treeish_dataref(const char **p)
3321 {
3322 struct object_id oid;
3323 struct object_entry *e;
3324
3325 if (**p == ':') { /* <mark> */
3326 e = find_mark(marks, parse_mark_ref_space(p));
3327 if (!e)
3328 die(_("unknown mark: %s"), command_buf.buf);
3329 oidcpy(&oid, &e->idx.oid);
3330 } else { /* <sha1> */
3331 if (parse_mapped_oid_hex(*p, &oid, p))
3332 die(_("invalid dataref: %s"), command_buf.buf);
3333 e = find_object(&oid);
3334 if (*(*p)++ != ' ')
3335 die(_("missing space after tree-ish: %s"), command_buf.buf);
3336 }
3337
3338 while (!e || e->type != OBJ_TREE)
3339 e = dereference(e, &oid);
3340 return e;
3341 }
3342
3343 static void print_ls(int mode, const unsigned char *hash, const char *path)
3344 {
3345 static struct strbuf line = STRBUF_INIT;
3346
3347 /* See show_tree(). */
3348 const char *type =
3349 S_ISGITLINK(mode) ? commit_type :
3350 S_ISDIR(mode) ? tree_type :
3351 blob_type;
3352
3353 if (!mode) {
3354 /* missing SP path LF */
3355 strbuf_reset(&line);
3356 strbuf_addstr(&line, "missing ");
3357 quote_c_style(path, &line, NULL, 0);
3358 strbuf_addch(&line, '\n');
3359 } else {
3360 /* mode SP type SP object_name TAB path LF */
3361 strbuf_reset(&line);
3362 strbuf_addf(&line, "%06o %s %s\t",
3363 mode & ~NO_DELTA, type, hash_to_hex(hash));
3364 quote_c_style(path, &line, NULL, 0);
3365 strbuf_addch(&line, '\n');
3366 }
3367 cat_blob_write(line.buf, line.len);
3368 }
3369
3370 static void parse_ls(const char *p, struct branch *b)
3371 {
3372 static struct strbuf path = STRBUF_INIT;
3373 struct tree_entry *root = NULL;
3374 struct tree_entry leaf = {NULL};
3375
3376 /* ls SP (<tree-ish> SP)? <path> */
3377 if (*p == '"') {
3378 if (!b)
3379 die(_("not in a commit: %s"), command_buf.buf);
3380 root = &b->branch_tree;
3381 } else {
3382 struct object_entry *e = parse_treeish_dataref(&p);
3383 root = new_tree_entry();
3384 oidcpy(&root->versions[1].oid, &e->idx.oid);
3385 if (!is_null_oid(&root->versions[1].oid))
3386 root->versions[1].mode = S_IFDIR;
3387 load_tree(root);
3388 }
3389 strbuf_reset(&path);
3390 parse_path_eol(&path, p, "path");
3391 tree_content_get(root, path.buf, &leaf, 1);
3392 /*
3393 * A directory in preparation would have a sha1 of zero
3394 * until it is saved. Save, for simplicity.
3395 */
3396 if (S_ISDIR(leaf.versions[1].mode))
3397 store_tree(&leaf);
3398
3399 print_ls(leaf.versions[1].mode, leaf.versions[1].oid.hash, path.buf);
3400 if (leaf.tree)
3401 release_tree_content_recursive(leaf.tree);
3402 if (!b || root != &b->branch_tree)
3403 release_tree_entry(root);
3404 }
3405
3406 static void checkpoint(void)
3407 {
3408 checkpoint_requested = 0;
3409 if (object_count) {
3410 cycle_packfile();
3411 }
3412 dump_branches();
3413 dump_tags();
3414 dump_marks();
3415 }
3416
3417 static void parse_checkpoint(void)
3418 {
3419 checkpoint_requested = 1;
3420 skip_optional_lf();
3421 }
3422
3423 static void parse_progress(void)
3424 {
3425 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3426 fputc('\n', stdout);
3427 fflush(stdout);
3428 skip_optional_lf();
3429 }
3430
3431 static void parse_alias(void)
3432 {
3433 struct object_entry *e;
3434 struct branch b;
3435
3436 skip_optional_lf();
3437 read_next_command();
3438
3439 /* mark ... */
3440 parse_mark();
3441 if (!next_mark)
3442 die(_("expected 'mark' command, got %s"), command_buf.buf);
3443
3444 /* to ... */
3445 memset(&b, 0, sizeof(b));
3446 if (!parse_objectish_with_prefix(&b, "to "))
3447 die(_("expected 'to' command, got %s"), command_buf.buf);
3448 e = find_object(&b.oid);
3449 assert(e);
3450 insert_mark(&marks, next_mark, e);
3451 }
3452
3453 static char* make_fast_import_path(const char *path)
3454 {
3455 if (!relative_marks_paths || is_absolute_path(path))
3456 return prefix_filename(global_prefix, path);
3457 return repo_git_path(the_repository, "info/fast-import/%s", path);
3458 }
3459
3460 static void option_import_marks(const char *marks,
3461 int from_stream, int ignore_missing)
3462 {
3463 if (import_marks_file) {
3464 if (from_stream)
3465 die(_("only one import-marks command allowed per stream"));
3466
3467 /* read previous mark file */
3468 if(!import_marks_file_from_stream)
3469 read_marks();
3470 }
3471
3472 free(import_marks_file);
3473 import_marks_file = make_fast_import_path(marks);
3474 import_marks_file_from_stream = from_stream;
3475 import_marks_file_ignore_missing = ignore_missing;
3476 }
3477
3478 static void option_date_format(const char *fmt)
3479 {
3480 if (!strcmp(fmt, "raw"))
3481 whenspec = WHENSPEC_RAW;
3482 else if (!strcmp(fmt, "raw-permissive"))
3483 whenspec = WHENSPEC_RAW_PERMISSIVE;
3484 else if (!strcmp(fmt, "rfc2822"))
3485 whenspec = WHENSPEC_RFC2822;
3486 else if (!strcmp(fmt, "now"))
3487 whenspec = WHENSPEC_NOW;
3488 else
3489 die(_("unknown --date-format argument %s"), fmt);
3490 }
3491
3492 static unsigned long ulong_arg(const char *option, const char *arg)
3493 {
3494 char *endptr;
3495 unsigned long rv = strtoul(arg, &endptr, 0);
3496 if (strchr(arg, '-') || endptr == arg || *endptr)
3497 die(_("%s: argument must be a non-negative integer"), option);
3498 return rv;
3499 }
3500
3501 static void option_depth(const char *depth)
3502 {
3503 max_depth = ulong_arg("--depth", depth);
3504 if (max_depth > MAX_DEPTH)
3505 die(_("--depth cannot exceed %u"), MAX_DEPTH);
3506 }
3507
3508 static void option_active_branches(const char *branches)
3509 {
3510 max_active_branches = ulong_arg("--active-branches", branches);
3511 }
3512
3513 static void option_export_marks(const char *marks)
3514 {
3515 free(export_marks_file);
3516 export_marks_file = make_fast_import_path(marks);
3517 }
3518
3519 static void option_cat_blob_fd(const char *fd)
3520 {
3521 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3522 if (n > (unsigned long) INT_MAX)
3523 die(_("--cat-blob-fd cannot exceed %d"), INT_MAX);
3524 cat_blob_fd = (int) n;
3525 }
3526
3527 static void option_export_pack_edges(const char *edges)
3528 {
3529 char *fn = prefix_filename(global_prefix, edges);
3530 if (pack_edges)
3531 fclose(pack_edges);
3532 pack_edges = xfopen(fn, "a");
3533 free(fn);
3534 }
3535
3536 static void option_rewrite_submodules(const char *arg, struct string_list *list)
3537 {
3538 struct mark_set *ms;
3539 FILE *fp;
3540 char *s = xstrdup(arg);
3541 char *f = strchr(s, ':');
3542 if (!f)
3543 die(_("expected format name:filename for submodule rewrite option"));
3544 *f = '\0';
3545 f++;
3546 CALLOC_ARRAY(ms, 1);
3547
3548 f = prefix_filename(global_prefix, f);
3549 fp = fopen(f, "r");
3550 if (!fp)
3551 die_errno(_("cannot read '%s'"), f);
3552 read_mark_file(&ms, fp, insert_oid_entry);
3553 fclose(fp);
3554 free(f);
3555
3556 string_list_insert(list, s)->util = ms;
3557
3558 free(s);
3559 }
3560
3561 static int parse_one_option(const char *option)
3562 {
3563 if (skip_prefix(option, "max-pack-size=", &option)) {
3564 unsigned long v;
3565 if (!git_parse_ulong(option, &v))
3566 return 0;
3567 if (v < 8192) {
3568 warning(_("max-pack-size is now in bytes, assuming --max-pack-size=%lum"), v);
3569 v *= 1024 * 1024;
3570 } else if (v < 1024 * 1024) {
3571 warning(_("minimum max-pack-size is 1 MiB"));
3572 v = 1024 * 1024;
3573 }
3574 max_packsize = v;
3575 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3576 unsigned long v;
3577 if (!git_parse_ulong(option, &v))
3578 return 0;
3579 repo_settings_set_big_file_threshold(the_repository, v);
3580 } else if (skip_prefix(option, "depth=", &option)) {
3581 option_depth(option);
3582 } else if (skip_prefix(option, "active-branches=", &option)) {
3583 option_active_branches(option);
3584 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3585 option_export_pack_edges(option);
3586 } else if (skip_prefix(option, "signed-commits=", &option)) {
3587 if (parse_sign_mode(option, &signed_commit_mode))
3588 usagef(_("unknown --signed-commits mode '%s'"), option);
3589 } else if (skip_prefix(option, "signed-tags=", &option)) {
3590 if (parse_sign_mode(option, &signed_tag_mode))
3591 usagef(_("unknown --signed-tags mode '%s'"), option);
3592 } else if (!strcmp(option, "quiet")) {
3593 show_stats = 0;
3594 quiet = 1;
3595 } else if (!strcmp(option, "stats")) {
3596 show_stats = 1;
3597 } else if (!strcmp(option, "allow-unsafe-features")) {
3598 ; /* already handled during early option parsing */
3599 } else {
3600 return 0;
3601 }
3602
3603 return 1;
3604 }
3605
3606 static void check_unsafe_feature(const char *feature, int from_stream)
3607 {
3608 if (from_stream && !allow_unsafe_features)
3609 die(_("feature '%s' forbidden in input without --allow-unsafe-features"),
3610 feature);
3611 }
3612
3613 static int parse_one_feature(const char *feature, int from_stream)
3614 {
3615 const char *arg;
3616
3617 if (skip_prefix(feature, "date-format=", &arg)) {
3618 option_date_format(arg);
3619 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3620 check_unsafe_feature("import-marks", from_stream);
3621 option_import_marks(arg, from_stream, 0);
3622 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3623 check_unsafe_feature("import-marks-if-exists", from_stream);
3624 option_import_marks(arg, from_stream, 1);
3625 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3626 check_unsafe_feature(feature, from_stream);
3627 option_export_marks(arg);
3628 } else if (!strcmp(feature, "alias")) {
3629 ; /* Don't die - this feature is supported */
3630 } else if (skip_prefix(feature, "rewrite-submodules-to=", &arg)) {
3631 option_rewrite_submodules(arg, &sub_marks_to);
3632 } else if (skip_prefix(feature, "rewrite-submodules-from=", &arg)) {
3633 option_rewrite_submodules(arg, &sub_marks_from);
3634 } else if (!strcmp(feature, "get-mark")) {
3635 ; /* Don't die - this feature is supported */
3636 } else if (!strcmp(feature, "cat-blob")) {
3637 ; /* Don't die - this feature is supported */
3638 } else if (!strcmp(feature, "relative-marks")) {
3639 relative_marks_paths = 1;
3640 } else if (!strcmp(feature, "no-relative-marks")) {
3641 relative_marks_paths = 0;
3642 } else if (!strcmp(feature, "done")) {
3643 require_explicit_termination = 1;
3644 } else if (!strcmp(feature, "force")) {
3645 force_update = 1;
3646 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3647 ; /* do nothing; we have the feature */
3648 } else {
3649 return 0;
3650 }
3651
3652 return 1;
3653 }
3654
3655 static void parse_feature(const char *feature)
3656 {
3657 if (seen_data_command)
3658 die(_("got feature command '%s' after data command"), feature);
3659
3660 if (parse_one_feature(feature, 1))
3661 return;
3662
3663 die(_("this version of fast-import does not support feature %s."), feature);
3664 }
3665
3666 static void parse_option(const char *option)
3667 {
3668 if (seen_data_command)
3669 die(_("got option command '%s' after data command"), option);
3670
3671 if (parse_one_option(option))
3672 return;
3673
3674 die(_("this version of fast-import does not support option: %s"), option);
3675 }
3676
3677 static void git_pack_config(void)
3678 {
3679 int indexversion_value;
3680 int limit;
3681 unsigned long packsizelimit_value;
3682
3683 if (!repo_config_get_ulong(the_repository, "pack.depth", &max_depth)) {
3684 if (max_depth > MAX_DEPTH)
3685 max_depth = MAX_DEPTH;
3686 }
3687 if (!repo_config_get_int(the_repository, "pack.indexversion", &indexversion_value)) {
3688 pack_idx_opts.version = indexversion_value;
3689 if (pack_idx_opts.version > 2)
3690 git_die_config(the_repository, "pack.indexversion",
3691 "bad pack.indexVersion=%"PRIu32, pack_idx_opts.version);
3692 }
3693 if (!repo_config_get_ulong(the_repository, "pack.packsizelimit", &packsizelimit_value))
3694 max_packsize = packsizelimit_value;
3695
3696 if (!repo_config_get_int(the_repository, "fastimport.unpacklimit", &limit))
3697 unpack_limit = limit;
3698 else if (!repo_config_get_int(the_repository, "transfer.unpacklimit", &limit))
3699 unpack_limit = limit;
3700
3701 repo_config(the_repository, git_default_config, NULL);
3702 }
3703
3704 static const char fast_import_usage[] =
3705 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3706
3707 static void parse_argv(void)
3708 {
3709 unsigned int i;
3710
3711 for (i = 1; i < global_argc; i++) {
3712 const char *a = global_argv[i];
3713
3714 if (*a != '-' || !strcmp(a, "--"))
3715 break;
3716
3717 if (!skip_prefix(a, "--", &a))
3718 die(_("unknown option %s"), a);
3719
3720 if (parse_one_option(a))
3721 continue;
3722
3723 if (parse_one_feature(a, 0))
3724 continue;
3725
3726 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3727 option_cat_blob_fd(a);
3728 continue;
3729 }
3730
3731 die(_("unknown option --%s"), a);
3732 }
3733 if (i != global_argc)
3734 usage(fast_import_usage);
3735
3736 seen_data_command = 1;
3737 if (import_marks_file)
3738 read_marks();
3739 build_mark_map(&sub_marks_from, &sub_marks_to);
3740 }
3741
3742 int cmd_fast_import(int argc,
3743 const char **argv,
3744 const char *prefix,
3745 struct repository *repo)
3746 {
3747 unsigned int i;
3748
3749 show_usage_if_asked(argc, argv, fast_import_usage);
3750
3751 reset_pack_idx_option(&pack_idx_opts);
3752 git_pack_config();
3753
3754 alloc_objects(object_entry_alloc);
3755 strbuf_init(&command_buf, 0);
3756 CALLOC_ARRAY(atom_table, atom_table_sz);
3757 CALLOC_ARRAY(branch_table, branch_table_sz);
3758 CALLOC_ARRAY(avail_tree_table, avail_tree_table_sz);
3759 marks = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
3760
3761 hashmap_init(&object_table, object_entry_hashcmp, NULL, 0);
3762
3763 /*
3764 * We don't parse most options until after we've seen the set of
3765 * "feature" lines at the start of the stream (which allows the command
3766 * line to override stream data). But we must do an early parse of any
3767 * command-line options that impact how we interpret the feature lines.
3768 */
3769 for (i = 1; i < argc; i++) {
3770 const char *arg = argv[i];
3771 if (*arg != '-' || !strcmp(arg, "--"))
3772 break;
3773 if (!strcmp(arg, "--allow-unsafe-features"))
3774 allow_unsafe_features = 1;
3775 }
3776
3777 global_argc = argc;
3778 global_argv = argv;
3779 global_prefix = prefix;
3780
3781 rc_free = mem_pool_alloc(&fi_mem_pool, cmd_save * sizeof(*rc_free));
3782 for (i = 0; i < (cmd_save - 1); i++)
3783 rc_free[i].next = &rc_free[i + 1];
3784 rc_free[cmd_save - 1].next = NULL;
3785
3786 start_packfile();
3787 set_die_routine(die_nicely);
3788 set_checkpoint_signal();
3789 while (read_next_command() != EOF) {
3790 const char *v;
3791 if (!strcmp("blob", command_buf.buf))
3792 parse_new_blob();
3793 else if (skip_prefix(command_buf.buf, "commit ", &v))
3794 parse_new_commit(v);
3795 else if (skip_prefix(command_buf.buf, "tag ", &v))
3796 parse_new_tag(v);
3797 else if (skip_prefix(command_buf.buf, "reset ", &v))
3798 parse_reset_branch(v);
3799 else if (skip_prefix(command_buf.buf, "ls ", &v))
3800 parse_ls(v, NULL);
3801 else if (skip_prefix(command_buf.buf, "cat-blob ", &v))
3802 parse_cat_blob(v);
3803 else if (skip_prefix(command_buf.buf, "get-mark ", &v))
3804 parse_get_mark(v);
3805 else if (!strcmp("checkpoint", command_buf.buf))
3806 parse_checkpoint();
3807 else if (!strcmp("done", command_buf.buf))
3808 break;
3809 else if (!strcmp("alias", command_buf.buf))
3810 parse_alias();
3811 else if (starts_with(command_buf.buf, "progress "))
3812 parse_progress();
3813 else if (skip_prefix(command_buf.buf, "feature ", &v))
3814 parse_feature(v);
3815 else if (skip_prefix(command_buf.buf, "option git ", &v))
3816 parse_option(v);
3817 else if (starts_with(command_buf.buf, "option "))
3818 /* ignore non-git options*/;
3819 else
3820 die(_("unsupported command: %s"), command_buf.buf);
3821
3822 if (checkpoint_requested)
3823 checkpoint();
3824 }
3825
3826 /* argv hasn't been parsed yet, do so */
3827 if (!seen_data_command)
3828 parse_argv();
3829
3830 if (require_explicit_termination && feof(stdin))
3831 die(_("stream ends early"));
3832
3833 end_packfile();
3834
3835 dump_branches();
3836 dump_tags();
3837 unkeep_all_packs();
3838 dump_marks();
3839
3840 if (pack_edges)
3841 fclose(pack_edges);
3842
3843 if (show_stats) {
3844 uintmax_t total_count = 0, duplicate_count = 0;
3845 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3846 total_count += object_count_by_type[i];
3847 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3848 duplicate_count += duplicate_count_by_type[i];
3849
3850 fprintf(stderr, "%s statistics:\n", argv[0]);
3851 fprintf(stderr, "---------------------------------------------------------------------\n");
3852 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3853 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3854 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB], delta_count_attempts_by_type[OBJ_BLOB]);
3855 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE], delta_count_attempts_by_type[OBJ_TREE]);
3856 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT], delta_count_attempts_by_type[OBJ_COMMIT]);
3857 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG], delta_count_attempts_by_type[OBJ_TAG]);
3858 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3859 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3860 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3861 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (tree_entry_allocd + fi_mem_pool.pool_alloc + alloc_count*sizeof(struct object_entry))/1024);
3862 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)((tree_entry_allocd + fi_mem_pool.pool_alloc) /1024));
3863 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3864 fprintf(stderr, "---------------------------------------------------------------------\n");
3865 pack_report(repo);
3866 fprintf(stderr, "---------------------------------------------------------------------\n");
3867 fprintf(stderr, "\n");
3868 }
3869
3870 return failure ? 1 : 0;
3871 }