Merge branch 'maint' of git://github.com/git-l10n/git-po into maint
[git.git] / fast-import.c
bloba1479e980da154f9179d71aeeda7746770686baa
1 /*
2 (See Documentation/git-fast-import.txt for maintained documentation.)
3 Format of STDIN stream:
5 stream ::= cmd*;
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
20 new_commit ::= 'commit' sp ref_str lf
21 mark?
22 ('author' (sp name)? sp '<' email '>' sp when lf)?
23 'committer' (sp name)? sp '<' email '>' sp when lf
24 commit_msg
25 ('from' sp commit-ish lf)?
26 ('merge' sp commit-ish lf)*
27 (file_change | ls)*
28 lf?;
29 commit_msg ::= data;
31 ls ::= 'ls' sp '"' quoted(path) '"' lf;
33 file_change ::= file_clr
34 | file_del
35 | file_rnm
36 | file_cpy
37 | file_obm
38 | file_inm;
39 file_clr ::= 'deleteall' lf;
40 file_del ::= 'D' sp path_str lf;
41 file_rnm ::= 'R' sp path_str sp path_str lf;
42 file_cpy ::= 'C' sp path_str sp path_str lf;
43 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
44 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
45 data;
46 note_obm ::= 'N' sp (hexsha1 | idnum) sp commit-ish lf;
47 note_inm ::= 'N' sp 'inline' sp commit-ish lf
48 data;
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp commit-ish lf
52 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
53 tag_msg;
54 tag_msg ::= data;
56 reset_branch ::= 'reset' sp ref_str lf
57 ('from' sp commit-ish lf)?
58 lf?;
60 checkpoint ::= 'checkpoint' lf
61 lf?;
63 progress ::= 'progress' sp not_lf* lf
64 lf?;
66 # note: the first idnum in a stream should be 1 and subsequent
67 # idnums should not have gaps between values as this will cause
68 # the stream parser to reserve space for the gapped values. An
69 # idnum can be updated in the future to a new object by issuing
70 # a new mark directive with the old idnum.
72 mark ::= 'mark' sp idnum lf;
73 data ::= (delimited_data | exact_data)
74 lf?;
76 # note: delim may be any string but must not contain lf.
77 # data_line may contain any data but must not be exactly
78 # delim.
79 delimited_data ::= 'data' sp '<<' delim lf
80 (data_line lf)*
81 delim lf;
83 # note: declen indicates the length of binary_data in bytes.
84 # declen does not include the lf preceding the binary data.
86 exact_data ::= 'data' sp declen lf
87 binary_data;
89 # note: quoted strings are C-style quoting supporting \c for
90 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
91 # is the signed byte value in octal. Note that the only
92 # characters which must actually be escaped to protect the
93 # stream formatting is: \, " and LF. Otherwise these values
94 # are UTF8.
96 commit-ish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
97 ref_str ::= ref;
98 sha1exp_str ::= sha1exp;
99 tag_str ::= tag;
100 path_str ::= path | '"' quoted(path) '"' ;
101 mode ::= '100644' | '644'
102 | '100755' | '755'
103 | '120000'
106 declen ::= # unsigned 32 bit value, ascii base10 notation;
107 bigint ::= # unsigned integer value, ascii base10 notation;
108 binary_data ::= # file content, not interpreted;
110 when ::= raw_when | rfc2822_when;
111 raw_when ::= ts sp tz;
112 rfc2822_when ::= # Valid RFC 2822 date and time;
114 sp ::= # ASCII space character;
115 lf ::= # ASCII newline (LF) character;
117 # note: a colon (':') must precede the numerical value assigned to
118 # an idnum. This is to distinguish it from a ref or tag name as
119 # GIT does not permit ':' in ref or tag strings.
121 idnum ::= ':' bigint;
122 path ::= # GIT style file path, e.g. "a/b/c";
123 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
124 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
125 sha1exp ::= # Any valid GIT SHA1 expression;
126 hexsha1 ::= # SHA1 in hexadecimal format;
128 # note: name and email are UTF8 strings, however name must not
129 # contain '<' or lf and email must not contain any of the
130 # following: '<', '>', lf.
132 name ::= # valid GIT author/committer name;
133 email ::= # valid GIT author/committer email;
134 ts ::= # time since the epoch in seconds, ascii base10 notation;
135 tz ::= # GIT style timezone;
137 # note: comments, ls and cat requests may appear anywhere
138 # in the input, except within a data command. Any form
139 # of the data command always escapes the related input
140 # from comment processing.
142 # In case it is not clear, the '#' that starts the comment
143 # must be the first character on that line (an lf
144 # preceded it).
147 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
148 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
150 comment ::= '#' not_lf* lf;
151 not_lf ::= # Any byte that is not ASCII newline (LF);
154 #include "builtin.h"
155 #include "cache.h"
156 #include "object.h"
157 #include "blob.h"
158 #include "tree.h"
159 #include "commit.h"
160 #include "delta.h"
161 #include "pack.h"
162 #include "refs.h"
163 #include "csum-file.h"
164 #include "quote.h"
165 #include "exec_cmd.h"
166 #include "dir.h"
168 #define PACK_ID_BITS 16
169 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
170 #define DEPTH_BITS 13
171 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
174 * We abuse the setuid bit on directories to mean "do not delta".
176 #define NO_DELTA S_ISUID
178 struct object_entry {
179 struct pack_idx_entry idx;
180 struct object_entry *next;
181 uint32_t type : TYPE_BITS,
182 pack_id : PACK_ID_BITS,
183 depth : DEPTH_BITS;
186 struct object_entry_pool {
187 struct object_entry_pool *next_pool;
188 struct object_entry *next_free;
189 struct object_entry *end;
190 struct object_entry entries[FLEX_ARRAY]; /* more */
193 struct mark_set {
194 union {
195 struct object_entry *marked[1024];
196 struct mark_set *sets[1024];
197 } data;
198 unsigned int shift;
201 struct last_object {
202 struct strbuf data;
203 off_t offset;
204 unsigned int depth;
205 unsigned no_swap : 1;
208 struct mem_pool {
209 struct mem_pool *next_pool;
210 char *next_free;
211 char *end;
212 uintmax_t space[FLEX_ARRAY]; /* more */
215 struct atom_str {
216 struct atom_str *next_atom;
217 unsigned short str_len;
218 char str_dat[FLEX_ARRAY]; /* more */
221 struct tree_content;
222 struct tree_entry {
223 struct tree_content *tree;
224 struct atom_str *name;
225 struct tree_entry_ms {
226 uint16_t mode;
227 unsigned char sha1[20];
228 } versions[2];
231 struct tree_content {
232 unsigned int entry_capacity; /* must match avail_tree_content */
233 unsigned int entry_count;
234 unsigned int delta_depth;
235 struct tree_entry *entries[FLEX_ARRAY]; /* more */
238 struct avail_tree_content {
239 unsigned int entry_capacity; /* must match tree_content */
240 struct avail_tree_content *next_avail;
243 struct branch {
244 struct branch *table_next_branch;
245 struct branch *active_next_branch;
246 const char *name;
247 struct tree_entry branch_tree;
248 uintmax_t last_commit;
249 uintmax_t num_notes;
250 unsigned active : 1;
251 unsigned delete : 1;
252 unsigned pack_id : PACK_ID_BITS;
253 unsigned char sha1[20];
256 struct tag {
257 struct tag *next_tag;
258 const char *name;
259 unsigned int pack_id;
260 unsigned char sha1[20];
263 struct hash_list {
264 struct hash_list *next;
265 unsigned char sha1[20];
268 typedef enum {
269 WHENSPEC_RAW = 1,
270 WHENSPEC_RFC2822,
271 WHENSPEC_NOW
272 } whenspec_type;
274 struct recent_command {
275 struct recent_command *prev;
276 struct recent_command *next;
277 char *buf;
280 /* Configured limits on output */
281 static unsigned long max_depth = 10;
282 static off_t max_packsize;
283 static int force_update;
284 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
285 static int pack_compression_seen;
287 /* Stats and misc. counters */
288 static uintmax_t alloc_count;
289 static uintmax_t marks_set_count;
290 static uintmax_t object_count_by_type[1 << TYPE_BITS];
291 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
292 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
293 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
294 static unsigned long object_count;
295 static unsigned long branch_count;
296 static unsigned long branch_load_count;
297 static int failure;
298 static FILE *pack_edges;
299 static unsigned int show_stats = 1;
300 static int global_argc;
301 static char **global_argv;
303 /* Memory pools */
304 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
305 static size_t total_allocd;
306 static struct mem_pool *mem_pool;
308 /* Atom management */
309 static unsigned int atom_table_sz = 4451;
310 static unsigned int atom_cnt;
311 static struct atom_str **atom_table;
313 /* The .pack file being generated */
314 static struct pack_idx_option pack_idx_opts;
315 static unsigned int pack_id;
316 static struct sha1file *pack_file;
317 static struct packed_git *pack_data;
318 static struct packed_git **all_packs;
319 static off_t pack_size;
321 /* Table of objects we've written. */
322 static unsigned int object_entry_alloc = 5000;
323 static struct object_entry_pool *blocks;
324 static struct object_entry *object_table[1 << 16];
325 static struct mark_set *marks;
326 static const char *export_marks_file;
327 static const char *import_marks_file;
328 static int import_marks_file_from_stream;
329 static int import_marks_file_ignore_missing;
330 static int relative_marks_paths;
332 /* Our last blob */
333 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
335 /* Tree management */
336 static unsigned int tree_entry_alloc = 1000;
337 static void *avail_tree_entry;
338 static unsigned int avail_tree_table_sz = 100;
339 static struct avail_tree_content **avail_tree_table;
340 static struct strbuf old_tree = STRBUF_INIT;
341 static struct strbuf new_tree = STRBUF_INIT;
343 /* Branch data */
344 static unsigned long max_active_branches = 5;
345 static unsigned long cur_active_branches;
346 static unsigned long branch_table_sz = 1039;
347 static struct branch **branch_table;
348 static struct branch *active_branches;
350 /* Tag data */
351 static struct tag *first_tag;
352 static struct tag *last_tag;
354 /* Input stream parsing */
355 static whenspec_type whenspec = WHENSPEC_RAW;
356 static struct strbuf command_buf = STRBUF_INIT;
357 static int unread_command_buf;
358 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
359 static struct recent_command *cmd_tail = &cmd_hist;
360 static struct recent_command *rc_free;
361 static unsigned int cmd_save = 100;
362 static uintmax_t next_mark;
363 static struct strbuf new_data = STRBUF_INIT;
364 static int seen_data_command;
365 static int require_explicit_termination;
367 /* Signal handling */
368 static volatile sig_atomic_t checkpoint_requested;
370 /* Where to write output of cat-blob commands */
371 static int cat_blob_fd = STDOUT_FILENO;
373 static void parse_argv(void);
374 static void parse_cat_blob(const char *p);
375 static void parse_ls(const char *p, struct branch *b);
377 static void write_branch_report(FILE *rpt, struct branch *b)
379 fprintf(rpt, "%s:\n", b->name);
381 fprintf(rpt, " status :");
382 if (b->active)
383 fputs(" active", rpt);
384 if (b->branch_tree.tree)
385 fputs(" loaded", rpt);
386 if (is_null_sha1(b->branch_tree.versions[1].sha1))
387 fputs(" dirty", rpt);
388 fputc('\n', rpt);
390 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
391 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
392 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
393 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
395 fputs(" last pack : ", rpt);
396 if (b->pack_id < MAX_PACK_ID)
397 fprintf(rpt, "%u", b->pack_id);
398 fputc('\n', rpt);
400 fputc('\n', rpt);
403 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
405 static void write_crash_report(const char *err)
407 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
408 FILE *rpt = fopen(loc, "w");
409 struct branch *b;
410 unsigned long lu;
411 struct recent_command *rc;
413 if (!rpt) {
414 error("can't write crash report %s: %s", loc, strerror(errno));
415 return;
418 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
420 fprintf(rpt, "fast-import crash report:\n");
421 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
422 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
423 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
424 fputc('\n', rpt);
426 fputs("fatal: ", rpt);
427 fputs(err, rpt);
428 fputc('\n', rpt);
430 fputc('\n', rpt);
431 fputs("Most Recent Commands Before Crash\n", rpt);
432 fputs("---------------------------------\n", rpt);
433 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
434 if (rc->next == &cmd_hist)
435 fputs("* ", rpt);
436 else
437 fputs(" ", rpt);
438 fputs(rc->buf, rpt);
439 fputc('\n', rpt);
442 fputc('\n', rpt);
443 fputs("Active Branch LRU\n", rpt);
444 fputs("-----------------\n", rpt);
445 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
446 cur_active_branches,
447 max_active_branches);
448 fputc('\n', rpt);
449 fputs(" pos clock name\n", rpt);
450 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
451 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
452 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
453 ++lu, b->last_commit, b->name);
455 fputc('\n', rpt);
456 fputs("Inactive Branches\n", rpt);
457 fputs("-----------------\n", rpt);
458 for (lu = 0; lu < branch_table_sz; lu++) {
459 for (b = branch_table[lu]; b; b = b->table_next_branch)
460 write_branch_report(rpt, b);
463 if (first_tag) {
464 struct tag *tg;
465 fputc('\n', rpt);
466 fputs("Annotated Tags\n", rpt);
467 fputs("--------------\n", rpt);
468 for (tg = first_tag; tg; tg = tg->next_tag) {
469 fputs(sha1_to_hex(tg->sha1), rpt);
470 fputc(' ', rpt);
471 fputs(tg->name, rpt);
472 fputc('\n', rpt);
476 fputc('\n', rpt);
477 fputs("Marks\n", rpt);
478 fputs("-----\n", rpt);
479 if (export_marks_file)
480 fprintf(rpt, " exported to %s\n", export_marks_file);
481 else
482 dump_marks_helper(rpt, 0, marks);
484 fputc('\n', rpt);
485 fputs("-------------------\n", rpt);
486 fputs("END OF CRASH REPORT\n", rpt);
487 fclose(rpt);
490 static void end_packfile(void);
491 static void unkeep_all_packs(void);
492 static void dump_marks(void);
494 static NORETURN void die_nicely(const char *err, va_list params)
496 static int zombie;
497 char message[2 * PATH_MAX];
499 vsnprintf(message, sizeof(message), err, params);
500 fputs("fatal: ", stderr);
501 fputs(message, stderr);
502 fputc('\n', stderr);
504 if (!zombie) {
505 zombie = 1;
506 write_crash_report(message);
507 end_packfile();
508 unkeep_all_packs();
509 dump_marks();
511 exit(128);
514 #ifndef SIGUSR1 /* Windows, for example */
516 static void set_checkpoint_signal(void)
520 #else
522 static void checkpoint_signal(int signo)
524 checkpoint_requested = 1;
527 static void set_checkpoint_signal(void)
529 struct sigaction sa;
531 memset(&sa, 0, sizeof(sa));
532 sa.sa_handler = checkpoint_signal;
533 sigemptyset(&sa.sa_mask);
534 sa.sa_flags = SA_RESTART;
535 sigaction(SIGUSR1, &sa, NULL);
538 #endif
540 static void alloc_objects(unsigned int cnt)
542 struct object_entry_pool *b;
544 b = xmalloc(sizeof(struct object_entry_pool)
545 + cnt * sizeof(struct object_entry));
546 b->next_pool = blocks;
547 b->next_free = b->entries;
548 b->end = b->entries + cnt;
549 blocks = b;
550 alloc_count += cnt;
553 static struct object_entry *new_object(unsigned char *sha1)
555 struct object_entry *e;
557 if (blocks->next_free == blocks->end)
558 alloc_objects(object_entry_alloc);
560 e = blocks->next_free++;
561 hashcpy(e->idx.sha1, sha1);
562 return e;
565 static struct object_entry *find_object(unsigned char *sha1)
567 unsigned int h = sha1[0] << 8 | sha1[1];
568 struct object_entry *e;
569 for (e = object_table[h]; e; e = e->next)
570 if (!hashcmp(sha1, e->idx.sha1))
571 return e;
572 return NULL;
575 static struct object_entry *insert_object(unsigned char *sha1)
577 unsigned int h = sha1[0] << 8 | sha1[1];
578 struct object_entry *e = object_table[h];
580 while (e) {
581 if (!hashcmp(sha1, e->idx.sha1))
582 return e;
583 e = e->next;
586 e = new_object(sha1);
587 e->next = object_table[h];
588 e->idx.offset = 0;
589 object_table[h] = e;
590 return e;
593 static unsigned int hc_str(const char *s, size_t len)
595 unsigned int r = 0;
596 while (len-- > 0)
597 r = r * 31 + *s++;
598 return r;
601 static void *pool_alloc(size_t len)
603 struct mem_pool *p;
604 void *r;
606 /* round up to a 'uintmax_t' alignment */
607 if (len & (sizeof(uintmax_t) - 1))
608 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
610 for (p = mem_pool; p; p = p->next_pool)
611 if ((p->end - p->next_free >= len))
612 break;
614 if (!p) {
615 if (len >= (mem_pool_alloc/2)) {
616 total_allocd += len;
617 return xmalloc(len);
619 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
620 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
621 p->next_pool = mem_pool;
622 p->next_free = (char *) p->space;
623 p->end = p->next_free + mem_pool_alloc;
624 mem_pool = p;
627 r = p->next_free;
628 p->next_free += len;
629 return r;
632 static void *pool_calloc(size_t count, size_t size)
634 size_t len = count * size;
635 void *r = pool_alloc(len);
636 memset(r, 0, len);
637 return r;
640 static char *pool_strdup(const char *s)
642 char *r = pool_alloc(strlen(s) + 1);
643 strcpy(r, s);
644 return r;
647 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
649 struct mark_set *s = marks;
650 while ((idnum >> s->shift) >= 1024) {
651 s = pool_calloc(1, sizeof(struct mark_set));
652 s->shift = marks->shift + 10;
653 s->data.sets[0] = marks;
654 marks = s;
656 while (s->shift) {
657 uintmax_t i = idnum >> s->shift;
658 idnum -= i << s->shift;
659 if (!s->data.sets[i]) {
660 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
661 s->data.sets[i]->shift = s->shift - 10;
663 s = s->data.sets[i];
665 if (!s->data.marked[idnum])
666 marks_set_count++;
667 s->data.marked[idnum] = oe;
670 static struct object_entry *find_mark(uintmax_t idnum)
672 uintmax_t orig_idnum = idnum;
673 struct mark_set *s = marks;
674 struct object_entry *oe = NULL;
675 if ((idnum >> s->shift) < 1024) {
676 while (s && s->shift) {
677 uintmax_t i = idnum >> s->shift;
678 idnum -= i << s->shift;
679 s = s->data.sets[i];
681 if (s)
682 oe = s->data.marked[idnum];
684 if (!oe)
685 die("mark :%" PRIuMAX " not declared", orig_idnum);
686 return oe;
689 static struct atom_str *to_atom(const char *s, unsigned short len)
691 unsigned int hc = hc_str(s, len) % atom_table_sz;
692 struct atom_str *c;
694 for (c = atom_table[hc]; c; c = c->next_atom)
695 if (c->str_len == len && !strncmp(s, c->str_dat, len))
696 return c;
698 c = pool_alloc(sizeof(struct atom_str) + len + 1);
699 c->str_len = len;
700 strncpy(c->str_dat, s, len);
701 c->str_dat[len] = 0;
702 c->next_atom = atom_table[hc];
703 atom_table[hc] = c;
704 atom_cnt++;
705 return c;
708 static struct branch *lookup_branch(const char *name)
710 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
711 struct branch *b;
713 for (b = branch_table[hc]; b; b = b->table_next_branch)
714 if (!strcmp(name, b->name))
715 return b;
716 return NULL;
719 static struct branch *new_branch(const char *name)
721 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
722 struct branch *b = lookup_branch(name);
724 if (b)
725 die("Invalid attempt to create duplicate branch: %s", name);
726 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
727 die("Branch name doesn't conform to GIT standards: %s", name);
729 b = pool_calloc(1, sizeof(struct branch));
730 b->name = pool_strdup(name);
731 b->table_next_branch = branch_table[hc];
732 b->branch_tree.versions[0].mode = S_IFDIR;
733 b->branch_tree.versions[1].mode = S_IFDIR;
734 b->num_notes = 0;
735 b->active = 0;
736 b->pack_id = MAX_PACK_ID;
737 branch_table[hc] = b;
738 branch_count++;
739 return b;
742 static unsigned int hc_entries(unsigned int cnt)
744 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
745 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
748 static struct tree_content *new_tree_content(unsigned int cnt)
750 struct avail_tree_content *f, *l = NULL;
751 struct tree_content *t;
752 unsigned int hc = hc_entries(cnt);
754 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
755 if (f->entry_capacity >= cnt)
756 break;
758 if (f) {
759 if (l)
760 l->next_avail = f->next_avail;
761 else
762 avail_tree_table[hc] = f->next_avail;
763 } else {
764 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
765 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
766 f->entry_capacity = cnt;
769 t = (struct tree_content*)f;
770 t->entry_count = 0;
771 t->delta_depth = 0;
772 return t;
775 static void release_tree_entry(struct tree_entry *e);
776 static void release_tree_content(struct tree_content *t)
778 struct avail_tree_content *f = (struct avail_tree_content*)t;
779 unsigned int hc = hc_entries(f->entry_capacity);
780 f->next_avail = avail_tree_table[hc];
781 avail_tree_table[hc] = f;
784 static void release_tree_content_recursive(struct tree_content *t)
786 unsigned int i;
787 for (i = 0; i < t->entry_count; i++)
788 release_tree_entry(t->entries[i]);
789 release_tree_content(t);
792 static struct tree_content *grow_tree_content(
793 struct tree_content *t,
794 int amt)
796 struct tree_content *r = new_tree_content(t->entry_count + amt);
797 r->entry_count = t->entry_count;
798 r->delta_depth = t->delta_depth;
799 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
800 release_tree_content(t);
801 return r;
804 static struct tree_entry *new_tree_entry(void)
806 struct tree_entry *e;
808 if (!avail_tree_entry) {
809 unsigned int n = tree_entry_alloc;
810 total_allocd += n * sizeof(struct tree_entry);
811 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
812 while (n-- > 1) {
813 *((void**)e) = e + 1;
814 e++;
816 *((void**)e) = NULL;
819 e = avail_tree_entry;
820 avail_tree_entry = *((void**)e);
821 return e;
824 static void release_tree_entry(struct tree_entry *e)
826 if (e->tree)
827 release_tree_content_recursive(e->tree);
828 *((void**)e) = avail_tree_entry;
829 avail_tree_entry = e;
832 static struct tree_content *dup_tree_content(struct tree_content *s)
834 struct tree_content *d;
835 struct tree_entry *a, *b;
836 unsigned int i;
838 if (!s)
839 return NULL;
840 d = new_tree_content(s->entry_count);
841 for (i = 0; i < s->entry_count; i++) {
842 a = s->entries[i];
843 b = new_tree_entry();
844 memcpy(b, a, sizeof(*a));
845 if (a->tree && is_null_sha1(b->versions[1].sha1))
846 b->tree = dup_tree_content(a->tree);
847 else
848 b->tree = NULL;
849 d->entries[i] = b;
851 d->entry_count = s->entry_count;
852 d->delta_depth = s->delta_depth;
854 return d;
857 static void start_packfile(void)
859 static char tmp_file[PATH_MAX];
860 struct packed_git *p;
861 struct pack_header hdr;
862 int pack_fd;
864 pack_fd = odb_mkstemp(tmp_file, sizeof(tmp_file),
865 "pack/tmp_pack_XXXXXX");
866 p = xcalloc(1, sizeof(*p) + strlen(tmp_file) + 2);
867 strcpy(p->pack_name, tmp_file);
868 p->pack_fd = pack_fd;
869 p->do_not_close = 1;
870 pack_file = sha1fd(pack_fd, p->pack_name);
872 hdr.hdr_signature = htonl(PACK_SIGNATURE);
873 hdr.hdr_version = htonl(2);
874 hdr.hdr_entries = 0;
875 sha1write(pack_file, &hdr, sizeof(hdr));
877 pack_data = p;
878 pack_size = sizeof(hdr);
879 object_count = 0;
881 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
882 all_packs[pack_id] = p;
885 static const char *create_index(void)
887 const char *tmpfile;
888 struct pack_idx_entry **idx, **c, **last;
889 struct object_entry *e;
890 struct object_entry_pool *o;
892 /* Build the table of object IDs. */
893 idx = xmalloc(object_count * sizeof(*idx));
894 c = idx;
895 for (o = blocks; o; o = o->next_pool)
896 for (e = o->next_free; e-- != o->entries;)
897 if (pack_id == e->pack_id)
898 *c++ = &e->idx;
899 last = idx + object_count;
900 if (c != last)
901 die("internal consistency error creating the index");
903 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
904 free(idx);
905 return tmpfile;
908 static char *keep_pack(const char *curr_index_name)
910 static char name[PATH_MAX];
911 static const char *keep_msg = "fast-import";
912 int keep_fd;
914 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
915 if (keep_fd < 0)
916 die_errno("cannot create keep file");
917 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
918 if (close(keep_fd))
919 die_errno("failed to write keep file");
921 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
922 get_object_directory(), sha1_to_hex(pack_data->sha1));
923 if (move_temp_to_file(pack_data->pack_name, name))
924 die("cannot store pack file");
926 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
927 get_object_directory(), sha1_to_hex(pack_data->sha1));
928 if (move_temp_to_file(curr_index_name, name))
929 die("cannot store index file");
930 free((void *)curr_index_name);
931 return name;
934 static void unkeep_all_packs(void)
936 static char name[PATH_MAX];
937 int k;
939 for (k = 0; k < pack_id; k++) {
940 struct packed_git *p = all_packs[k];
941 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
942 get_object_directory(), sha1_to_hex(p->sha1));
943 unlink_or_warn(name);
947 static void end_packfile(void)
949 if (!pack_data)
950 return;
952 clear_delta_base_cache();
953 if (object_count) {
954 struct packed_git *new_p;
955 unsigned char cur_pack_sha1[20];
956 char *idx_name;
957 int i;
958 struct branch *b;
959 struct tag *t;
961 close_pack_windows(pack_data);
962 sha1close(pack_file, cur_pack_sha1, 0);
963 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
964 pack_data->pack_name, object_count,
965 cur_pack_sha1, pack_size);
966 close(pack_data->pack_fd);
967 idx_name = keep_pack(create_index());
969 /* Register the packfile with core git's machinery. */
970 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
971 if (!new_p)
972 die("core git rejected index %s", idx_name);
973 all_packs[pack_id] = new_p;
974 install_packed_git(new_p);
976 /* Print the boundary */
977 if (pack_edges) {
978 fprintf(pack_edges, "%s:", new_p->pack_name);
979 for (i = 0; i < branch_table_sz; i++) {
980 for (b = branch_table[i]; b; b = b->table_next_branch) {
981 if (b->pack_id == pack_id)
982 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
985 for (t = first_tag; t; t = t->next_tag) {
986 if (t->pack_id == pack_id)
987 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
989 fputc('\n', pack_edges);
990 fflush(pack_edges);
993 pack_id++;
995 else {
996 close(pack_data->pack_fd);
997 unlink_or_warn(pack_data->pack_name);
999 free(pack_data);
1000 pack_data = NULL;
1002 /* We can't carry a delta across packfiles. */
1003 strbuf_release(&last_blob.data);
1004 last_blob.offset = 0;
1005 last_blob.depth = 0;
1008 static void cycle_packfile(void)
1010 end_packfile();
1011 start_packfile();
1014 static int store_object(
1015 enum object_type type,
1016 struct strbuf *dat,
1017 struct last_object *last,
1018 unsigned char *sha1out,
1019 uintmax_t mark)
1021 void *out, *delta;
1022 struct object_entry *e;
1023 unsigned char hdr[96];
1024 unsigned char sha1[20];
1025 unsigned long hdrlen, deltalen;
1026 git_SHA_CTX c;
1027 git_zstream s;
1029 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1030 (unsigned long)dat->len) + 1;
1031 git_SHA1_Init(&c);
1032 git_SHA1_Update(&c, hdr, hdrlen);
1033 git_SHA1_Update(&c, dat->buf, dat->len);
1034 git_SHA1_Final(sha1, &c);
1035 if (sha1out)
1036 hashcpy(sha1out, sha1);
1038 e = insert_object(sha1);
1039 if (mark)
1040 insert_mark(mark, e);
1041 if (e->idx.offset) {
1042 duplicate_count_by_type[type]++;
1043 return 1;
1044 } else if (find_sha1_pack(sha1, packed_git)) {
1045 e->type = type;
1046 e->pack_id = MAX_PACK_ID;
1047 e->idx.offset = 1; /* just not zero! */
1048 duplicate_count_by_type[type]++;
1049 return 1;
1052 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1053 delta_count_attempts_by_type[type]++;
1054 delta = diff_delta(last->data.buf, last->data.len,
1055 dat->buf, dat->len,
1056 &deltalen, dat->len - 20);
1057 } else
1058 delta = NULL;
1060 memset(&s, 0, sizeof(s));
1061 git_deflate_init(&s, pack_compression_level);
1062 if (delta) {
1063 s.next_in = delta;
1064 s.avail_in = deltalen;
1065 } else {
1066 s.next_in = (void *)dat->buf;
1067 s.avail_in = dat->len;
1069 s.avail_out = git_deflate_bound(&s, s.avail_in);
1070 s.next_out = out = xmalloc(s.avail_out);
1071 while (git_deflate(&s, Z_FINISH) == Z_OK)
1072 ; /* nothing */
1073 git_deflate_end(&s);
1075 /* Determine if we should auto-checkpoint. */
1076 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1077 || (pack_size + 60 + s.total_out) < pack_size) {
1079 /* This new object needs to *not* have the current pack_id. */
1080 e->pack_id = pack_id + 1;
1081 cycle_packfile();
1083 /* We cannot carry a delta into the new pack. */
1084 if (delta) {
1085 free(delta);
1086 delta = NULL;
1088 memset(&s, 0, sizeof(s));
1089 git_deflate_init(&s, pack_compression_level);
1090 s.next_in = (void *)dat->buf;
1091 s.avail_in = dat->len;
1092 s.avail_out = git_deflate_bound(&s, s.avail_in);
1093 s.next_out = out = xrealloc(out, s.avail_out);
1094 while (git_deflate(&s, Z_FINISH) == Z_OK)
1095 ; /* nothing */
1096 git_deflate_end(&s);
1100 e->type = type;
1101 e->pack_id = pack_id;
1102 e->idx.offset = pack_size;
1103 object_count++;
1104 object_count_by_type[type]++;
1106 crc32_begin(pack_file);
1108 if (delta) {
1109 off_t ofs = e->idx.offset - last->offset;
1110 unsigned pos = sizeof(hdr) - 1;
1112 delta_count_by_type[type]++;
1113 e->depth = last->depth + 1;
1115 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1116 sha1write(pack_file, hdr, hdrlen);
1117 pack_size += hdrlen;
1119 hdr[pos] = ofs & 127;
1120 while (ofs >>= 7)
1121 hdr[--pos] = 128 | (--ofs & 127);
1122 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1123 pack_size += sizeof(hdr) - pos;
1124 } else {
1125 e->depth = 0;
1126 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1127 sha1write(pack_file, hdr, hdrlen);
1128 pack_size += hdrlen;
1131 sha1write(pack_file, out, s.total_out);
1132 pack_size += s.total_out;
1134 e->idx.crc32 = crc32_end(pack_file);
1136 free(out);
1137 free(delta);
1138 if (last) {
1139 if (last->no_swap) {
1140 last->data = *dat;
1141 } else {
1142 strbuf_swap(&last->data, dat);
1144 last->offset = e->idx.offset;
1145 last->depth = e->depth;
1147 return 0;
1150 static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1152 if (sha1file_truncate(pack_file, checkpoint))
1153 die_errno("cannot truncate pack to skip duplicate");
1154 pack_size = checkpoint->offset;
1157 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1159 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1160 unsigned char *in_buf = xmalloc(in_sz);
1161 unsigned char *out_buf = xmalloc(out_sz);
1162 struct object_entry *e;
1163 unsigned char sha1[20];
1164 unsigned long hdrlen;
1165 off_t offset;
1166 git_SHA_CTX c;
1167 git_zstream s;
1168 struct sha1file_checkpoint checkpoint;
1169 int status = Z_OK;
1171 /* Determine if we should auto-checkpoint. */
1172 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1173 || (pack_size + 60 + len) < pack_size)
1174 cycle_packfile();
1176 sha1file_checkpoint(pack_file, &checkpoint);
1177 offset = checkpoint.offset;
1179 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1180 if (out_sz <= hdrlen)
1181 die("impossibly large object header");
1183 git_SHA1_Init(&c);
1184 git_SHA1_Update(&c, out_buf, hdrlen);
1186 crc32_begin(pack_file);
1188 memset(&s, 0, sizeof(s));
1189 git_deflate_init(&s, pack_compression_level);
1191 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1192 if (out_sz <= hdrlen)
1193 die("impossibly large object header");
1195 s.next_out = out_buf + hdrlen;
1196 s.avail_out = out_sz - hdrlen;
1198 while (status != Z_STREAM_END) {
1199 if (0 < len && !s.avail_in) {
1200 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1201 size_t n = fread(in_buf, 1, cnt, stdin);
1202 if (!n && feof(stdin))
1203 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1205 git_SHA1_Update(&c, in_buf, n);
1206 s.next_in = in_buf;
1207 s.avail_in = n;
1208 len -= n;
1211 status = git_deflate(&s, len ? 0 : Z_FINISH);
1213 if (!s.avail_out || status == Z_STREAM_END) {
1214 size_t n = s.next_out - out_buf;
1215 sha1write(pack_file, out_buf, n);
1216 pack_size += n;
1217 s.next_out = out_buf;
1218 s.avail_out = out_sz;
1221 switch (status) {
1222 case Z_OK:
1223 case Z_BUF_ERROR:
1224 case Z_STREAM_END:
1225 continue;
1226 default:
1227 die("unexpected deflate failure: %d", status);
1230 git_deflate_end(&s);
1231 git_SHA1_Final(sha1, &c);
1233 if (sha1out)
1234 hashcpy(sha1out, sha1);
1236 e = insert_object(sha1);
1238 if (mark)
1239 insert_mark(mark, e);
1241 if (e->idx.offset) {
1242 duplicate_count_by_type[OBJ_BLOB]++;
1243 truncate_pack(&checkpoint);
1245 } else if (find_sha1_pack(sha1, packed_git)) {
1246 e->type = OBJ_BLOB;
1247 e->pack_id = MAX_PACK_ID;
1248 e->idx.offset = 1; /* just not zero! */
1249 duplicate_count_by_type[OBJ_BLOB]++;
1250 truncate_pack(&checkpoint);
1252 } else {
1253 e->depth = 0;
1254 e->type = OBJ_BLOB;
1255 e->pack_id = pack_id;
1256 e->idx.offset = offset;
1257 e->idx.crc32 = crc32_end(pack_file);
1258 object_count++;
1259 object_count_by_type[OBJ_BLOB]++;
1262 free(in_buf);
1263 free(out_buf);
1266 /* All calls must be guarded by find_object() or find_mark() to
1267 * ensure the 'struct object_entry' passed was written by this
1268 * process instance. We unpack the entry by the offset, avoiding
1269 * the need for the corresponding .idx file. This unpacking rule
1270 * works because we only use OBJ_REF_DELTA within the packfiles
1271 * created by fast-import.
1273 * oe must not be NULL. Such an oe usually comes from giving
1274 * an unknown SHA-1 to find_object() or an undefined mark to
1275 * find_mark(). Callers must test for this condition and use
1276 * the standard read_sha1_file() when it happens.
1278 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1279 * find_mark(), where the mark was reloaded from an existing marks
1280 * file and is referencing an object that this fast-import process
1281 * instance did not write out to a packfile. Callers must test for
1282 * this condition and use read_sha1_file() instead.
1284 static void *gfi_unpack_entry(
1285 struct object_entry *oe,
1286 unsigned long *sizep)
1288 enum object_type type;
1289 struct packed_git *p = all_packs[oe->pack_id];
1290 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1291 /* The object is stored in the packfile we are writing to
1292 * and we have modified it since the last time we scanned
1293 * back to read a previously written object. If an old
1294 * window covered [p->pack_size, p->pack_size + 20) its
1295 * data is stale and is not valid. Closing all windows
1296 * and updating the packfile length ensures we can read
1297 * the newly written data.
1299 close_pack_windows(p);
1300 sha1flush(pack_file);
1302 /* We have to offer 20 bytes additional on the end of
1303 * the packfile as the core unpacker code assumes the
1304 * footer is present at the file end and must promise
1305 * at least 20 bytes within any window it maps. But
1306 * we don't actually create the footer here.
1308 p->pack_size = pack_size + 20;
1310 return unpack_entry(p, oe->idx.offset, &type, sizep);
1313 static const char *get_mode(const char *str, uint16_t *modep)
1315 unsigned char c;
1316 uint16_t mode = 0;
1318 while ((c = *str++) != ' ') {
1319 if (c < '0' || c > '7')
1320 return NULL;
1321 mode = (mode << 3) + (c - '0');
1323 *modep = mode;
1324 return str;
1327 static void load_tree(struct tree_entry *root)
1329 unsigned char *sha1 = root->versions[1].sha1;
1330 struct object_entry *myoe;
1331 struct tree_content *t;
1332 unsigned long size;
1333 char *buf;
1334 const char *c;
1336 root->tree = t = new_tree_content(8);
1337 if (is_null_sha1(sha1))
1338 return;
1340 myoe = find_object(sha1);
1341 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1342 if (myoe->type != OBJ_TREE)
1343 die("Not a tree: %s", sha1_to_hex(sha1));
1344 t->delta_depth = myoe->depth;
1345 buf = gfi_unpack_entry(myoe, &size);
1346 if (!buf)
1347 die("Can't load tree %s", sha1_to_hex(sha1));
1348 } else {
1349 enum object_type type;
1350 buf = read_sha1_file(sha1, &type, &size);
1351 if (!buf || type != OBJ_TREE)
1352 die("Can't load tree %s", sha1_to_hex(sha1));
1355 c = buf;
1356 while (c != (buf + size)) {
1357 struct tree_entry *e = new_tree_entry();
1359 if (t->entry_count == t->entry_capacity)
1360 root->tree = t = grow_tree_content(t, t->entry_count);
1361 t->entries[t->entry_count++] = e;
1363 e->tree = NULL;
1364 c = get_mode(c, &e->versions[1].mode);
1365 if (!c)
1366 die("Corrupt mode in %s", sha1_to_hex(sha1));
1367 e->versions[0].mode = e->versions[1].mode;
1368 e->name = to_atom(c, strlen(c));
1369 c += e->name->str_len + 1;
1370 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1371 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1372 c += 20;
1374 free(buf);
1377 static int tecmp0 (const void *_a, const void *_b)
1379 struct tree_entry *a = *((struct tree_entry**)_a);
1380 struct tree_entry *b = *((struct tree_entry**)_b);
1381 return base_name_compare(
1382 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1383 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1386 static int tecmp1 (const void *_a, const void *_b)
1388 struct tree_entry *a = *((struct tree_entry**)_a);
1389 struct tree_entry *b = *((struct tree_entry**)_b);
1390 return base_name_compare(
1391 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1392 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1395 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1397 size_t maxlen = 0;
1398 unsigned int i;
1400 if (!v)
1401 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1402 else
1403 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1405 for (i = 0; i < t->entry_count; i++) {
1406 if (t->entries[i]->versions[v].mode)
1407 maxlen += t->entries[i]->name->str_len + 34;
1410 strbuf_reset(b);
1411 strbuf_grow(b, maxlen);
1412 for (i = 0; i < t->entry_count; i++) {
1413 struct tree_entry *e = t->entries[i];
1414 if (!e->versions[v].mode)
1415 continue;
1416 strbuf_addf(b, "%o %s%c",
1417 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1418 e->name->str_dat, '\0');
1419 strbuf_add(b, e->versions[v].sha1, 20);
1423 static void store_tree(struct tree_entry *root)
1425 struct tree_content *t = root->tree;
1426 unsigned int i, j, del;
1427 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1428 struct object_entry *le = NULL;
1430 if (!is_null_sha1(root->versions[1].sha1))
1431 return;
1433 for (i = 0; i < t->entry_count; i++) {
1434 if (t->entries[i]->tree)
1435 store_tree(t->entries[i]);
1438 if (!(root->versions[0].mode & NO_DELTA))
1439 le = find_object(root->versions[0].sha1);
1440 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1441 mktree(t, 0, &old_tree);
1442 lo.data = old_tree;
1443 lo.offset = le->idx.offset;
1444 lo.depth = t->delta_depth;
1447 mktree(t, 1, &new_tree);
1448 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1450 t->delta_depth = lo.depth;
1451 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1452 struct tree_entry *e = t->entries[i];
1453 if (e->versions[1].mode) {
1454 e->versions[0].mode = e->versions[1].mode;
1455 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1456 t->entries[j++] = e;
1457 } else {
1458 release_tree_entry(e);
1459 del++;
1462 t->entry_count -= del;
1465 static void tree_content_replace(
1466 struct tree_entry *root,
1467 const unsigned char *sha1,
1468 const uint16_t mode,
1469 struct tree_content *newtree)
1471 if (!S_ISDIR(mode))
1472 die("Root cannot be a non-directory");
1473 hashclr(root->versions[0].sha1);
1474 hashcpy(root->versions[1].sha1, sha1);
1475 if (root->tree)
1476 release_tree_content_recursive(root->tree);
1477 root->tree = newtree;
1480 static int tree_content_set(
1481 struct tree_entry *root,
1482 const char *p,
1483 const unsigned char *sha1,
1484 const uint16_t mode,
1485 struct tree_content *subtree)
1487 struct tree_content *t;
1488 const char *slash1;
1489 unsigned int i, n;
1490 struct tree_entry *e;
1492 slash1 = strchrnul(p, '/');
1493 n = slash1 - p;
1494 if (!n)
1495 die("Empty path component found in input");
1496 if (!*slash1 && !S_ISDIR(mode) && subtree)
1497 die("Non-directories cannot have subtrees");
1499 if (!root->tree)
1500 load_tree(root);
1501 t = root->tree;
1502 for (i = 0; i < t->entry_count; i++) {
1503 e = t->entries[i];
1504 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1505 if (!*slash1) {
1506 if (!S_ISDIR(mode)
1507 && e->versions[1].mode == mode
1508 && !hashcmp(e->versions[1].sha1, sha1))
1509 return 0;
1510 e->versions[1].mode = mode;
1511 hashcpy(e->versions[1].sha1, sha1);
1512 if (e->tree)
1513 release_tree_content_recursive(e->tree);
1514 e->tree = subtree;
1517 * We need to leave e->versions[0].sha1 alone
1518 * to avoid modifying the preimage tree used
1519 * when writing out the parent directory.
1520 * But after replacing the subdir with a
1521 * completely different one, it's not a good
1522 * delta base any more, and besides, we've
1523 * thrown away the tree entries needed to
1524 * make a delta against it.
1526 * So let's just explicitly disable deltas
1527 * for the subtree.
1529 if (S_ISDIR(e->versions[0].mode))
1530 e->versions[0].mode |= NO_DELTA;
1532 hashclr(root->versions[1].sha1);
1533 return 1;
1535 if (!S_ISDIR(e->versions[1].mode)) {
1536 e->tree = new_tree_content(8);
1537 e->versions[1].mode = S_IFDIR;
1539 if (!e->tree)
1540 load_tree(e);
1541 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1542 hashclr(root->versions[1].sha1);
1543 return 1;
1545 return 0;
1549 if (t->entry_count == t->entry_capacity)
1550 root->tree = t = grow_tree_content(t, t->entry_count);
1551 e = new_tree_entry();
1552 e->name = to_atom(p, n);
1553 e->versions[0].mode = 0;
1554 hashclr(e->versions[0].sha1);
1555 t->entries[t->entry_count++] = e;
1556 if (*slash1) {
1557 e->tree = new_tree_content(8);
1558 e->versions[1].mode = S_IFDIR;
1559 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1560 } else {
1561 e->tree = subtree;
1562 e->versions[1].mode = mode;
1563 hashcpy(e->versions[1].sha1, sha1);
1565 hashclr(root->versions[1].sha1);
1566 return 1;
1569 static int tree_content_remove(
1570 struct tree_entry *root,
1571 const char *p,
1572 struct tree_entry *backup_leaf,
1573 int allow_root)
1575 struct tree_content *t;
1576 const char *slash1;
1577 unsigned int i, n;
1578 struct tree_entry *e;
1580 slash1 = strchrnul(p, '/');
1581 n = slash1 - p;
1583 if (!root->tree)
1584 load_tree(root);
1586 if (!*p && allow_root) {
1587 e = root;
1588 goto del_entry;
1591 t = root->tree;
1592 for (i = 0; i < t->entry_count; i++) {
1593 e = t->entries[i];
1594 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1595 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1597 * If p names a file in some subdirectory, and a
1598 * file or symlink matching the name of the
1599 * parent directory of p exists, then p cannot
1600 * exist and need not be deleted.
1602 return 1;
1603 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1604 goto del_entry;
1605 if (!e->tree)
1606 load_tree(e);
1607 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1608 for (n = 0; n < e->tree->entry_count; n++) {
1609 if (e->tree->entries[n]->versions[1].mode) {
1610 hashclr(root->versions[1].sha1);
1611 return 1;
1614 backup_leaf = NULL;
1615 goto del_entry;
1617 return 0;
1620 return 0;
1622 del_entry:
1623 if (backup_leaf)
1624 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1625 else if (e->tree)
1626 release_tree_content_recursive(e->tree);
1627 e->tree = NULL;
1628 e->versions[1].mode = 0;
1629 hashclr(e->versions[1].sha1);
1630 hashclr(root->versions[1].sha1);
1631 return 1;
1634 static int tree_content_get(
1635 struct tree_entry *root,
1636 const char *p,
1637 struct tree_entry *leaf,
1638 int allow_root)
1640 struct tree_content *t;
1641 const char *slash1;
1642 unsigned int i, n;
1643 struct tree_entry *e;
1645 slash1 = strchrnul(p, '/');
1646 n = slash1 - p;
1647 if (!n && !allow_root)
1648 die("Empty path component found in input");
1650 if (!root->tree)
1651 load_tree(root);
1653 if (!n) {
1654 e = root;
1655 goto found_entry;
1658 t = root->tree;
1659 for (i = 0; i < t->entry_count; i++) {
1660 e = t->entries[i];
1661 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1662 if (!*slash1)
1663 goto found_entry;
1664 if (!S_ISDIR(e->versions[1].mode))
1665 return 0;
1666 if (!e->tree)
1667 load_tree(e);
1668 return tree_content_get(e, slash1 + 1, leaf, 0);
1671 return 0;
1673 found_entry:
1674 memcpy(leaf, e, sizeof(*leaf));
1675 if (e->tree && is_null_sha1(e->versions[1].sha1))
1676 leaf->tree = dup_tree_content(e->tree);
1677 else
1678 leaf->tree = NULL;
1679 return 1;
1682 static int update_branch(struct branch *b)
1684 static const char *msg = "fast-import";
1685 struct ref_lock *lock;
1686 unsigned char old_sha1[20];
1688 if (read_ref(b->name, old_sha1))
1689 hashclr(old_sha1);
1690 if (is_null_sha1(b->sha1)) {
1691 if (b->delete)
1692 delete_ref(b->name, old_sha1, 0);
1693 return 0;
1695 lock = lock_any_ref_for_update(b->name, old_sha1, 0, NULL);
1696 if (!lock)
1697 return error("Unable to lock %s", b->name);
1698 if (!force_update && !is_null_sha1(old_sha1)) {
1699 struct commit *old_cmit, *new_cmit;
1701 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1702 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1703 if (!old_cmit || !new_cmit) {
1704 unlock_ref(lock);
1705 return error("Branch %s is missing commits.", b->name);
1708 if (!in_merge_bases(old_cmit, new_cmit)) {
1709 unlock_ref(lock);
1710 warning("Not updating %s"
1711 " (new tip %s does not contain %s)",
1712 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1713 return -1;
1716 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1717 return error("Unable to update %s", b->name);
1718 return 0;
1721 static void dump_branches(void)
1723 unsigned int i;
1724 struct branch *b;
1726 for (i = 0; i < branch_table_sz; i++) {
1727 for (b = branch_table[i]; b; b = b->table_next_branch)
1728 failure |= update_branch(b);
1732 static void dump_tags(void)
1734 static const char *msg = "fast-import";
1735 struct tag *t;
1736 struct ref_lock *lock;
1737 struct strbuf ref_name = STRBUF_INIT;
1739 for (t = first_tag; t; t = t->next_tag) {
1740 strbuf_reset(&ref_name);
1741 strbuf_addf(&ref_name, "tags/%s", t->name);
1742 lock = lock_ref_sha1(ref_name.buf, NULL);
1743 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1744 failure |= error("Unable to update %s", ref_name.buf);
1746 strbuf_release(&ref_name);
1749 static void dump_marks_helper(FILE *f,
1750 uintmax_t base,
1751 struct mark_set *m)
1753 uintmax_t k;
1754 if (m->shift) {
1755 for (k = 0; k < 1024; k++) {
1756 if (m->data.sets[k])
1757 dump_marks_helper(f, base + (k << m->shift),
1758 m->data.sets[k]);
1760 } else {
1761 for (k = 0; k < 1024; k++) {
1762 if (m->data.marked[k])
1763 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1764 sha1_to_hex(m->data.marked[k]->idx.sha1));
1769 static void dump_marks(void)
1771 static struct lock_file mark_lock;
1772 int mark_fd;
1773 FILE *f;
1775 if (!export_marks_file)
1776 return;
1778 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1779 if (mark_fd < 0) {
1780 failure |= error("Unable to write marks file %s: %s",
1781 export_marks_file, strerror(errno));
1782 return;
1785 f = fdopen(mark_fd, "w");
1786 if (!f) {
1787 int saved_errno = errno;
1788 rollback_lock_file(&mark_lock);
1789 failure |= error("Unable to write marks file %s: %s",
1790 export_marks_file, strerror(saved_errno));
1791 return;
1795 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1796 * Assign -1 to the lock file descriptor so that commit_lock_file()
1797 * won't try to close() it.
1799 mark_lock.fd = -1;
1801 dump_marks_helper(f, 0, marks);
1802 if (ferror(f) || fclose(f)) {
1803 int saved_errno = errno;
1804 rollback_lock_file(&mark_lock);
1805 failure |= error("Unable to write marks file %s: %s",
1806 export_marks_file, strerror(saved_errno));
1807 return;
1810 if (commit_lock_file(&mark_lock)) {
1811 int saved_errno = errno;
1812 rollback_lock_file(&mark_lock);
1813 failure |= error("Unable to commit marks file %s: %s",
1814 export_marks_file, strerror(saved_errno));
1815 return;
1819 static void read_marks(void)
1821 char line[512];
1822 FILE *f = fopen(import_marks_file, "r");
1823 if (f)
1825 else if (import_marks_file_ignore_missing && errno == ENOENT)
1826 return; /* Marks file does not exist */
1827 else
1828 die_errno("cannot read '%s'", import_marks_file);
1829 while (fgets(line, sizeof(line), f)) {
1830 uintmax_t mark;
1831 char *end;
1832 unsigned char sha1[20];
1833 struct object_entry *e;
1835 end = strchr(line, '\n');
1836 if (line[0] != ':' || !end)
1837 die("corrupt mark line: %s", line);
1838 *end = 0;
1839 mark = strtoumax(line + 1, &end, 10);
1840 if (!mark || end == line + 1
1841 || *end != ' ' || get_sha1_hex(end + 1, sha1))
1842 die("corrupt mark line: %s", line);
1843 e = find_object(sha1);
1844 if (!e) {
1845 enum object_type type = sha1_object_info(sha1, NULL);
1846 if (type < 0)
1847 die("object not found: %s", sha1_to_hex(sha1));
1848 e = insert_object(sha1);
1849 e->type = type;
1850 e->pack_id = MAX_PACK_ID;
1851 e->idx.offset = 1; /* just not zero! */
1853 insert_mark(mark, e);
1855 fclose(f);
1859 static int read_next_command(void)
1861 static int stdin_eof = 0;
1863 if (stdin_eof) {
1864 unread_command_buf = 0;
1865 return EOF;
1868 for (;;) {
1869 const char *p;
1871 if (unread_command_buf) {
1872 unread_command_buf = 0;
1873 } else {
1874 struct recent_command *rc;
1876 strbuf_detach(&command_buf, NULL);
1877 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1878 if (stdin_eof)
1879 return EOF;
1881 if (!seen_data_command
1882 && !starts_with(command_buf.buf, "feature ")
1883 && !starts_with(command_buf.buf, "option ")) {
1884 parse_argv();
1887 rc = rc_free;
1888 if (rc)
1889 rc_free = rc->next;
1890 else {
1891 rc = cmd_hist.next;
1892 cmd_hist.next = rc->next;
1893 cmd_hist.next->prev = &cmd_hist;
1894 free(rc->buf);
1897 rc->buf = command_buf.buf;
1898 rc->prev = cmd_tail;
1899 rc->next = cmd_hist.prev;
1900 rc->prev->next = rc;
1901 cmd_tail = rc;
1903 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1904 parse_cat_blob(p);
1905 continue;
1907 if (command_buf.buf[0] == '#')
1908 continue;
1909 return 0;
1913 static void skip_optional_lf(void)
1915 int term_char = fgetc(stdin);
1916 if (term_char != '\n' && term_char != EOF)
1917 ungetc(term_char, stdin);
1920 static void parse_mark(void)
1922 const char *v;
1923 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1924 next_mark = strtoumax(v, NULL, 10);
1925 read_next_command();
1927 else
1928 next_mark = 0;
1931 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1933 const char *data;
1934 strbuf_reset(sb);
1936 if (!skip_prefix(command_buf.buf, "data ", &data))
1937 die("Expected 'data n' command, found: %s", command_buf.buf);
1939 if (skip_prefix(data, "<<", &data)) {
1940 char *term = xstrdup(data);
1941 size_t term_len = command_buf.len - (data - command_buf.buf);
1943 strbuf_detach(&command_buf, NULL);
1944 for (;;) {
1945 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1946 die("EOF in data (terminator '%s' not found)", term);
1947 if (term_len == command_buf.len
1948 && !strcmp(term, command_buf.buf))
1949 break;
1950 strbuf_addbuf(sb, &command_buf);
1951 strbuf_addch(sb, '\n');
1953 free(term);
1955 else {
1956 uintmax_t len = strtoumax(data, NULL, 10);
1957 size_t n = 0, length = (size_t)len;
1959 if (limit && limit < len) {
1960 *len_res = len;
1961 return 0;
1963 if (length < len)
1964 die("data is too large to use in this context");
1966 while (n < length) {
1967 size_t s = strbuf_fread(sb, length - n, stdin);
1968 if (!s && feof(stdin))
1969 die("EOF in data (%lu bytes remaining)",
1970 (unsigned long)(length - n));
1971 n += s;
1975 skip_optional_lf();
1976 return 1;
1979 static int validate_raw_date(const char *src, char *result, int maxlen)
1981 const char *orig_src = src;
1982 char *endp;
1983 unsigned long num;
1985 errno = 0;
1987 num = strtoul(src, &endp, 10);
1988 /* NEEDSWORK: perhaps check for reasonable values? */
1989 if (errno || endp == src || *endp != ' ')
1990 return -1;
1992 src = endp + 1;
1993 if (*src != '-' && *src != '+')
1994 return -1;
1996 num = strtoul(src + 1, &endp, 10);
1997 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1998 1400 < num)
1999 return -1;
2001 strcpy(result, orig_src);
2002 return 0;
2005 static char *parse_ident(const char *buf)
2007 const char *ltgt;
2008 size_t name_len;
2009 char *ident;
2011 /* ensure there is a space delimiter even if there is no name */
2012 if (*buf == '<')
2013 --buf;
2015 ltgt = buf + strcspn(buf, "<>");
2016 if (*ltgt != '<')
2017 die("Missing < in ident string: %s", buf);
2018 if (ltgt != buf && ltgt[-1] != ' ')
2019 die("Missing space before < in ident string: %s", buf);
2020 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2021 if (*ltgt != '>')
2022 die("Missing > in ident string: %s", buf);
2023 ltgt++;
2024 if (*ltgt != ' ')
2025 die("Missing space after > in ident string: %s", buf);
2026 ltgt++;
2027 name_len = ltgt - buf;
2028 ident = xmalloc(name_len + 24);
2029 strncpy(ident, buf, name_len);
2031 switch (whenspec) {
2032 case WHENSPEC_RAW:
2033 if (validate_raw_date(ltgt, ident + name_len, 24) < 0)
2034 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2035 break;
2036 case WHENSPEC_RFC2822:
2037 if (parse_date(ltgt, ident + name_len, 24) < 0)
2038 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2039 break;
2040 case WHENSPEC_NOW:
2041 if (strcmp("now", ltgt))
2042 die("Date in ident must be 'now': %s", buf);
2043 datestamp(ident + name_len, 24);
2044 break;
2047 return ident;
2050 static void parse_and_store_blob(
2051 struct last_object *last,
2052 unsigned char *sha1out,
2053 uintmax_t mark)
2055 static struct strbuf buf = STRBUF_INIT;
2056 uintmax_t len;
2058 if (parse_data(&buf, big_file_threshold, &len))
2059 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2060 else {
2061 if (last) {
2062 strbuf_release(&last->data);
2063 last->offset = 0;
2064 last->depth = 0;
2066 stream_blob(len, sha1out, mark);
2067 skip_optional_lf();
2071 static void parse_new_blob(void)
2073 read_next_command();
2074 parse_mark();
2075 parse_and_store_blob(&last_blob, NULL, next_mark);
2078 static void unload_one_branch(void)
2080 while (cur_active_branches
2081 && cur_active_branches >= max_active_branches) {
2082 uintmax_t min_commit = ULONG_MAX;
2083 struct branch *e, *l = NULL, *p = NULL;
2085 for (e = active_branches; e; e = e->active_next_branch) {
2086 if (e->last_commit < min_commit) {
2087 p = l;
2088 min_commit = e->last_commit;
2090 l = e;
2093 if (p) {
2094 e = p->active_next_branch;
2095 p->active_next_branch = e->active_next_branch;
2096 } else {
2097 e = active_branches;
2098 active_branches = e->active_next_branch;
2100 e->active = 0;
2101 e->active_next_branch = NULL;
2102 if (e->branch_tree.tree) {
2103 release_tree_content_recursive(e->branch_tree.tree);
2104 e->branch_tree.tree = NULL;
2106 cur_active_branches--;
2110 static void load_branch(struct branch *b)
2112 load_tree(&b->branch_tree);
2113 if (!b->active) {
2114 b->active = 1;
2115 b->active_next_branch = active_branches;
2116 active_branches = b;
2117 cur_active_branches++;
2118 branch_load_count++;
2122 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2124 unsigned char fanout = 0;
2125 while ((num_notes >>= 8))
2126 fanout++;
2127 return fanout;
2130 static void construct_path_with_fanout(const char *hex_sha1,
2131 unsigned char fanout, char *path)
2133 unsigned int i = 0, j = 0;
2134 if (fanout >= 20)
2135 die("Too large fanout (%u)", fanout);
2136 while (fanout) {
2137 path[i++] = hex_sha1[j++];
2138 path[i++] = hex_sha1[j++];
2139 path[i++] = '/';
2140 fanout--;
2142 memcpy(path + i, hex_sha1 + j, 40 - j);
2143 path[i + 40 - j] = '\0';
2146 static uintmax_t do_change_note_fanout(
2147 struct tree_entry *orig_root, struct tree_entry *root,
2148 char *hex_sha1, unsigned int hex_sha1_len,
2149 char *fullpath, unsigned int fullpath_len,
2150 unsigned char fanout)
2152 struct tree_content *t = root->tree;
2153 struct tree_entry *e, leaf;
2154 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2155 uintmax_t num_notes = 0;
2156 unsigned char sha1[20];
2157 char realpath[60];
2159 for (i = 0; t && i < t->entry_count; i++) {
2160 e = t->entries[i];
2161 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2162 tmp_fullpath_len = fullpath_len;
2165 * We're interested in EITHER existing note entries (entries
2166 * with exactly 40 hex chars in path, not including directory
2167 * separators), OR directory entries that may contain note
2168 * entries (with < 40 hex chars in path).
2169 * Also, each path component in a note entry must be a multiple
2170 * of 2 chars.
2172 if (!e->versions[1].mode ||
2173 tmp_hex_sha1_len > 40 ||
2174 e->name->str_len % 2)
2175 continue;
2177 /* This _may_ be a note entry, or a subdir containing notes */
2178 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2179 e->name->str_len);
2180 if (tmp_fullpath_len)
2181 fullpath[tmp_fullpath_len++] = '/';
2182 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2183 e->name->str_len);
2184 tmp_fullpath_len += e->name->str_len;
2185 fullpath[tmp_fullpath_len] = '\0';
2187 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2188 /* This is a note entry */
2189 if (fanout == 0xff) {
2190 /* Counting mode, no rename */
2191 num_notes++;
2192 continue;
2194 construct_path_with_fanout(hex_sha1, fanout, realpath);
2195 if (!strcmp(fullpath, realpath)) {
2196 /* Note entry is in correct location */
2197 num_notes++;
2198 continue;
2201 /* Rename fullpath to realpath */
2202 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2203 die("Failed to remove path %s", fullpath);
2204 tree_content_set(orig_root, realpath,
2205 leaf.versions[1].sha1,
2206 leaf.versions[1].mode,
2207 leaf.tree);
2208 } else if (S_ISDIR(e->versions[1].mode)) {
2209 /* This is a subdir that may contain note entries */
2210 if (!e->tree)
2211 load_tree(e);
2212 num_notes += do_change_note_fanout(orig_root, e,
2213 hex_sha1, tmp_hex_sha1_len,
2214 fullpath, tmp_fullpath_len, fanout);
2217 /* The above may have reallocated the current tree_content */
2218 t = root->tree;
2220 return num_notes;
2223 static uintmax_t change_note_fanout(struct tree_entry *root,
2224 unsigned char fanout)
2226 char hex_sha1[40], path[60];
2227 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2231 * Given a pointer into a string, parse a mark reference:
2233 * idnum ::= ':' bigint;
2235 * Return the first character after the value in *endptr.
2237 * Complain if the following character is not what is expected,
2238 * either a space or end of the string.
2240 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2242 uintmax_t mark;
2244 assert(*p == ':');
2245 p++;
2246 mark = strtoumax(p, endptr, 10);
2247 if (*endptr == p)
2248 die("No value after ':' in mark: %s", command_buf.buf);
2249 return mark;
2253 * Parse the mark reference, and complain if this is not the end of
2254 * the string.
2256 static uintmax_t parse_mark_ref_eol(const char *p)
2258 char *end;
2259 uintmax_t mark;
2261 mark = parse_mark_ref(p, &end);
2262 if (*end != '\0')
2263 die("Garbage after mark: %s", command_buf.buf);
2264 return mark;
2268 * Parse the mark reference, demanding a trailing space. Return a
2269 * pointer to the space.
2271 static uintmax_t parse_mark_ref_space(const char **p)
2273 uintmax_t mark;
2274 char *end;
2276 mark = parse_mark_ref(*p, &end);
2277 if (*end++ != ' ')
2278 die("Missing space after mark: %s", command_buf.buf);
2279 *p = end;
2280 return mark;
2283 static void file_change_m(const char *p, struct branch *b)
2285 static struct strbuf uq = STRBUF_INIT;
2286 const char *endp;
2287 struct object_entry *oe;
2288 unsigned char sha1[20];
2289 uint16_t mode, inline_data = 0;
2291 p = get_mode(p, &mode);
2292 if (!p)
2293 die("Corrupt mode: %s", command_buf.buf);
2294 switch (mode) {
2295 case 0644:
2296 case 0755:
2297 mode |= S_IFREG;
2298 case S_IFREG | 0644:
2299 case S_IFREG | 0755:
2300 case S_IFLNK:
2301 case S_IFDIR:
2302 case S_IFGITLINK:
2303 /* ok */
2304 break;
2305 default:
2306 die("Corrupt mode: %s", command_buf.buf);
2309 if (*p == ':') {
2310 oe = find_mark(parse_mark_ref_space(&p));
2311 hashcpy(sha1, oe->idx.sha1);
2312 } else if (skip_prefix(p, "inline ", &p)) {
2313 inline_data = 1;
2314 oe = NULL; /* not used with inline_data, but makes gcc happy */
2315 } else {
2316 if (get_sha1_hex(p, sha1))
2317 die("Invalid dataref: %s", command_buf.buf);
2318 oe = find_object(sha1);
2319 p += 40;
2320 if (*p++ != ' ')
2321 die("Missing space after SHA1: %s", command_buf.buf);
2324 strbuf_reset(&uq);
2325 if (!unquote_c_style(&uq, p, &endp)) {
2326 if (*endp)
2327 die("Garbage after path in: %s", command_buf.buf);
2328 p = uq.buf;
2331 /* Git does not track empty, non-toplevel directories. */
2332 if (S_ISDIR(mode) && !hashcmp(sha1, EMPTY_TREE_SHA1_BIN) && *p) {
2333 tree_content_remove(&b->branch_tree, p, NULL, 0);
2334 return;
2337 if (S_ISGITLINK(mode)) {
2338 if (inline_data)
2339 die("Git links cannot be specified 'inline': %s",
2340 command_buf.buf);
2341 else if (oe) {
2342 if (oe->type != OBJ_COMMIT)
2343 die("Not a commit (actually a %s): %s",
2344 typename(oe->type), command_buf.buf);
2347 * Accept the sha1 without checking; it expected to be in
2348 * another repository.
2350 } else if (inline_data) {
2351 if (S_ISDIR(mode))
2352 die("Directories cannot be specified 'inline': %s",
2353 command_buf.buf);
2354 if (p != uq.buf) {
2355 strbuf_addstr(&uq, p);
2356 p = uq.buf;
2358 read_next_command();
2359 parse_and_store_blob(&last_blob, sha1, 0);
2360 } else {
2361 enum object_type expected = S_ISDIR(mode) ?
2362 OBJ_TREE: OBJ_BLOB;
2363 enum object_type type = oe ? oe->type :
2364 sha1_object_info(sha1, NULL);
2365 if (type < 0)
2366 die("%s not found: %s",
2367 S_ISDIR(mode) ? "Tree" : "Blob",
2368 command_buf.buf);
2369 if (type != expected)
2370 die("Not a %s (actually a %s): %s",
2371 typename(expected), typename(type),
2372 command_buf.buf);
2375 if (!*p) {
2376 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2377 return;
2379 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2382 static void file_change_d(const char *p, struct branch *b)
2384 static struct strbuf uq = STRBUF_INIT;
2385 const char *endp;
2387 strbuf_reset(&uq);
2388 if (!unquote_c_style(&uq, p, &endp)) {
2389 if (*endp)
2390 die("Garbage after path in: %s", command_buf.buf);
2391 p = uq.buf;
2393 tree_content_remove(&b->branch_tree, p, NULL, 1);
2396 static void file_change_cr(const char *s, struct branch *b, int rename)
2398 const char *d;
2399 static struct strbuf s_uq = STRBUF_INIT;
2400 static struct strbuf d_uq = STRBUF_INIT;
2401 const char *endp;
2402 struct tree_entry leaf;
2404 strbuf_reset(&s_uq);
2405 if (!unquote_c_style(&s_uq, s, &endp)) {
2406 if (*endp != ' ')
2407 die("Missing space after source: %s", command_buf.buf);
2408 } else {
2409 endp = strchr(s, ' ');
2410 if (!endp)
2411 die("Missing space after source: %s", command_buf.buf);
2412 strbuf_add(&s_uq, s, endp - s);
2414 s = s_uq.buf;
2416 endp++;
2417 if (!*endp)
2418 die("Missing dest: %s", command_buf.buf);
2420 d = endp;
2421 strbuf_reset(&d_uq);
2422 if (!unquote_c_style(&d_uq, d, &endp)) {
2423 if (*endp)
2424 die("Garbage after dest in: %s", command_buf.buf);
2425 d = d_uq.buf;
2428 memset(&leaf, 0, sizeof(leaf));
2429 if (rename)
2430 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2431 else
2432 tree_content_get(&b->branch_tree, s, &leaf, 1);
2433 if (!leaf.versions[1].mode)
2434 die("Path %s not in branch", s);
2435 if (!*d) { /* C "path/to/subdir" "" */
2436 tree_content_replace(&b->branch_tree,
2437 leaf.versions[1].sha1,
2438 leaf.versions[1].mode,
2439 leaf.tree);
2440 return;
2442 tree_content_set(&b->branch_tree, d,
2443 leaf.versions[1].sha1,
2444 leaf.versions[1].mode,
2445 leaf.tree);
2448 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2450 static struct strbuf uq = STRBUF_INIT;
2451 struct object_entry *oe;
2452 struct branch *s;
2453 unsigned char sha1[20], commit_sha1[20];
2454 char path[60];
2455 uint16_t inline_data = 0;
2456 unsigned char new_fanout;
2459 * When loading a branch, we don't traverse its tree to count the real
2460 * number of notes (too expensive to do this for all non-note refs).
2461 * This means that recently loaded notes refs might incorrectly have
2462 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2464 * Fix this by traversing the tree and counting the number of notes
2465 * when b->num_notes == 0. If the notes tree is truly empty, the
2466 * calculation should not take long.
2468 if (b->num_notes == 0 && *old_fanout == 0) {
2469 /* Invoke change_note_fanout() in "counting mode". */
2470 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2471 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2474 /* Now parse the notemodify command. */
2475 /* <dataref> or 'inline' */
2476 if (*p == ':') {
2477 oe = find_mark(parse_mark_ref_space(&p));
2478 hashcpy(sha1, oe->idx.sha1);
2479 } else if (skip_prefix(p, "inline ", &p)) {
2480 inline_data = 1;
2481 oe = NULL; /* not used with inline_data, but makes gcc happy */
2482 } else {
2483 if (get_sha1_hex(p, sha1))
2484 die("Invalid dataref: %s", command_buf.buf);
2485 oe = find_object(sha1);
2486 p += 40;
2487 if (*p++ != ' ')
2488 die("Missing space after SHA1: %s", command_buf.buf);
2491 /* <commit-ish> */
2492 s = lookup_branch(p);
2493 if (s) {
2494 if (is_null_sha1(s->sha1))
2495 die("Can't add a note on empty branch.");
2496 hashcpy(commit_sha1, s->sha1);
2497 } else if (*p == ':') {
2498 uintmax_t commit_mark = parse_mark_ref_eol(p);
2499 struct object_entry *commit_oe = find_mark(commit_mark);
2500 if (commit_oe->type != OBJ_COMMIT)
2501 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2502 hashcpy(commit_sha1, commit_oe->idx.sha1);
2503 } else if (!get_sha1(p, commit_sha1)) {
2504 unsigned long size;
2505 char *buf = read_object_with_reference(commit_sha1,
2506 commit_type, &size, commit_sha1);
2507 if (!buf || size < 46)
2508 die("Not a valid commit: %s", p);
2509 free(buf);
2510 } else
2511 die("Invalid ref name or SHA1 expression: %s", p);
2513 if (inline_data) {
2514 if (p != uq.buf) {
2515 strbuf_addstr(&uq, p);
2516 p = uq.buf;
2518 read_next_command();
2519 parse_and_store_blob(&last_blob, sha1, 0);
2520 } else if (oe) {
2521 if (oe->type != OBJ_BLOB)
2522 die("Not a blob (actually a %s): %s",
2523 typename(oe->type), command_buf.buf);
2524 } else if (!is_null_sha1(sha1)) {
2525 enum object_type type = sha1_object_info(sha1, NULL);
2526 if (type < 0)
2527 die("Blob not found: %s", command_buf.buf);
2528 if (type != OBJ_BLOB)
2529 die("Not a blob (actually a %s): %s",
2530 typename(type), command_buf.buf);
2533 construct_path_with_fanout(sha1_to_hex(commit_sha1), *old_fanout, path);
2534 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2535 b->num_notes--;
2537 if (is_null_sha1(sha1))
2538 return; /* nothing to insert */
2540 b->num_notes++;
2541 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2542 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2543 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2546 static void file_change_deleteall(struct branch *b)
2548 release_tree_content_recursive(b->branch_tree.tree);
2549 hashclr(b->branch_tree.versions[0].sha1);
2550 hashclr(b->branch_tree.versions[1].sha1);
2551 load_tree(&b->branch_tree);
2552 b->num_notes = 0;
2555 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2557 if (!buf || size < 46)
2558 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2559 if (memcmp("tree ", buf, 5)
2560 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2561 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2562 hashcpy(b->branch_tree.versions[0].sha1,
2563 b->branch_tree.versions[1].sha1);
2566 static void parse_from_existing(struct branch *b)
2568 if (is_null_sha1(b->sha1)) {
2569 hashclr(b->branch_tree.versions[0].sha1);
2570 hashclr(b->branch_tree.versions[1].sha1);
2571 } else {
2572 unsigned long size;
2573 char *buf;
2575 buf = read_object_with_reference(b->sha1,
2576 commit_type, &size, b->sha1);
2577 parse_from_commit(b, buf, size);
2578 free(buf);
2582 static int parse_from(struct branch *b)
2584 const char *from;
2585 struct branch *s;
2587 if (!skip_prefix(command_buf.buf, "from ", &from))
2588 return 0;
2590 if (b->branch_tree.tree) {
2591 release_tree_content_recursive(b->branch_tree.tree);
2592 b->branch_tree.tree = NULL;
2595 s = lookup_branch(from);
2596 if (b == s)
2597 die("Can't create a branch from itself: %s", b->name);
2598 else if (s) {
2599 unsigned char *t = s->branch_tree.versions[1].sha1;
2600 hashcpy(b->sha1, s->sha1);
2601 hashcpy(b->branch_tree.versions[0].sha1, t);
2602 hashcpy(b->branch_tree.versions[1].sha1, t);
2603 } else if (*from == ':') {
2604 uintmax_t idnum = parse_mark_ref_eol(from);
2605 struct object_entry *oe = find_mark(idnum);
2606 if (oe->type != OBJ_COMMIT)
2607 die("Mark :%" PRIuMAX " not a commit", idnum);
2608 hashcpy(b->sha1, oe->idx.sha1);
2609 if (oe->pack_id != MAX_PACK_ID) {
2610 unsigned long size;
2611 char *buf = gfi_unpack_entry(oe, &size);
2612 parse_from_commit(b, buf, size);
2613 free(buf);
2614 } else
2615 parse_from_existing(b);
2616 } else if (!get_sha1(from, b->sha1)) {
2617 parse_from_existing(b);
2618 if (is_null_sha1(b->sha1))
2619 b->delete = 1;
2621 else
2622 die("Invalid ref name or SHA1 expression: %s", from);
2624 read_next_command();
2625 return 1;
2628 static struct hash_list *parse_merge(unsigned int *count)
2630 struct hash_list *list = NULL, **tail = &list, *n;
2631 const char *from;
2632 struct branch *s;
2634 *count = 0;
2635 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2636 n = xmalloc(sizeof(*n));
2637 s = lookup_branch(from);
2638 if (s)
2639 hashcpy(n->sha1, s->sha1);
2640 else if (*from == ':') {
2641 uintmax_t idnum = parse_mark_ref_eol(from);
2642 struct object_entry *oe = find_mark(idnum);
2643 if (oe->type != OBJ_COMMIT)
2644 die("Mark :%" PRIuMAX " not a commit", idnum);
2645 hashcpy(n->sha1, oe->idx.sha1);
2646 } else if (!get_sha1(from, n->sha1)) {
2647 unsigned long size;
2648 char *buf = read_object_with_reference(n->sha1,
2649 commit_type, &size, n->sha1);
2650 if (!buf || size < 46)
2651 die("Not a valid commit: %s", from);
2652 free(buf);
2653 } else
2654 die("Invalid ref name or SHA1 expression: %s", from);
2656 n->next = NULL;
2657 *tail = n;
2658 tail = &n->next;
2660 (*count)++;
2661 read_next_command();
2663 return list;
2666 static void parse_new_commit(const char *arg)
2668 static struct strbuf msg = STRBUF_INIT;
2669 struct branch *b;
2670 char *author = NULL;
2671 char *committer = NULL;
2672 struct hash_list *merge_list = NULL;
2673 unsigned int merge_count;
2674 unsigned char prev_fanout, new_fanout;
2675 const char *v;
2677 b = lookup_branch(arg);
2678 if (!b)
2679 b = new_branch(arg);
2681 read_next_command();
2682 parse_mark();
2683 if (skip_prefix(command_buf.buf, "author ", &v)) {
2684 author = parse_ident(v);
2685 read_next_command();
2687 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2688 committer = parse_ident(v);
2689 read_next_command();
2691 if (!committer)
2692 die("Expected committer but didn't get one");
2693 parse_data(&msg, 0, NULL);
2694 read_next_command();
2695 parse_from(b);
2696 merge_list = parse_merge(&merge_count);
2698 /* ensure the branch is active/loaded */
2699 if (!b->branch_tree.tree || !max_active_branches) {
2700 unload_one_branch();
2701 load_branch(b);
2704 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2706 /* file_change* */
2707 while (command_buf.len > 0) {
2708 if (skip_prefix(command_buf.buf, "M ", &v))
2709 file_change_m(v, b);
2710 else if (skip_prefix(command_buf.buf, "D ", &v))
2711 file_change_d(v, b);
2712 else if (skip_prefix(command_buf.buf, "R ", &v))
2713 file_change_cr(v, b, 1);
2714 else if (skip_prefix(command_buf.buf, "C ", &v))
2715 file_change_cr(v, b, 0);
2716 else if (skip_prefix(command_buf.buf, "N ", &v))
2717 note_change_n(v, b, &prev_fanout);
2718 else if (!strcmp("deleteall", command_buf.buf))
2719 file_change_deleteall(b);
2720 else if (skip_prefix(command_buf.buf, "ls ", &v))
2721 parse_ls(v, b);
2722 else {
2723 unread_command_buf = 1;
2724 break;
2726 if (read_next_command() == EOF)
2727 break;
2730 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2731 if (new_fanout != prev_fanout)
2732 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2734 /* build the tree and the commit */
2735 store_tree(&b->branch_tree);
2736 hashcpy(b->branch_tree.versions[0].sha1,
2737 b->branch_tree.versions[1].sha1);
2739 strbuf_reset(&new_data);
2740 strbuf_addf(&new_data, "tree %s\n",
2741 sha1_to_hex(b->branch_tree.versions[1].sha1));
2742 if (!is_null_sha1(b->sha1))
2743 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2744 while (merge_list) {
2745 struct hash_list *next = merge_list->next;
2746 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2747 free(merge_list);
2748 merge_list = next;
2750 strbuf_addf(&new_data,
2751 "author %s\n"
2752 "committer %s\n"
2753 "\n",
2754 author ? author : committer, committer);
2755 strbuf_addbuf(&new_data, &msg);
2756 free(author);
2757 free(committer);
2759 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2760 b->pack_id = pack_id;
2761 b->last_commit = object_count_by_type[OBJ_COMMIT];
2764 static void parse_new_tag(const char *arg)
2766 static struct strbuf msg = STRBUF_INIT;
2767 const char *from;
2768 char *tagger;
2769 struct branch *s;
2770 struct tag *t;
2771 uintmax_t from_mark = 0;
2772 unsigned char sha1[20];
2773 enum object_type type;
2774 const char *v;
2776 t = pool_alloc(sizeof(struct tag));
2777 memset(t, 0, sizeof(struct tag));
2778 t->name = pool_strdup(arg);
2779 if (last_tag)
2780 last_tag->next_tag = t;
2781 else
2782 first_tag = t;
2783 last_tag = t;
2784 read_next_command();
2786 /* from ... */
2787 if (!skip_prefix(command_buf.buf, "from ", &from))
2788 die("Expected from command, got %s", command_buf.buf);
2789 s = lookup_branch(from);
2790 if (s) {
2791 if (is_null_sha1(s->sha1))
2792 die("Can't tag an empty branch.");
2793 hashcpy(sha1, s->sha1);
2794 type = OBJ_COMMIT;
2795 } else if (*from == ':') {
2796 struct object_entry *oe;
2797 from_mark = parse_mark_ref_eol(from);
2798 oe = find_mark(from_mark);
2799 type = oe->type;
2800 hashcpy(sha1, oe->idx.sha1);
2801 } else if (!get_sha1(from, sha1)) {
2802 struct object_entry *oe = find_object(sha1);
2803 if (!oe) {
2804 type = sha1_object_info(sha1, NULL);
2805 if (type < 0)
2806 die("Not a valid object: %s", from);
2807 } else
2808 type = oe->type;
2809 } else
2810 die("Invalid ref name or SHA1 expression: %s", from);
2811 read_next_command();
2813 /* tagger ... */
2814 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2815 tagger = parse_ident(v);
2816 read_next_command();
2817 } else
2818 tagger = NULL;
2820 /* tag payload/message */
2821 parse_data(&msg, 0, NULL);
2823 /* build the tag object */
2824 strbuf_reset(&new_data);
2826 strbuf_addf(&new_data,
2827 "object %s\n"
2828 "type %s\n"
2829 "tag %s\n",
2830 sha1_to_hex(sha1), typename(type), t->name);
2831 if (tagger)
2832 strbuf_addf(&new_data,
2833 "tagger %s\n", tagger);
2834 strbuf_addch(&new_data, '\n');
2835 strbuf_addbuf(&new_data, &msg);
2836 free(tagger);
2838 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2839 t->pack_id = MAX_PACK_ID;
2840 else
2841 t->pack_id = pack_id;
2844 static void parse_reset_branch(const char *arg)
2846 struct branch *b;
2848 b = lookup_branch(arg);
2849 if (b) {
2850 hashclr(b->sha1);
2851 hashclr(b->branch_tree.versions[0].sha1);
2852 hashclr(b->branch_tree.versions[1].sha1);
2853 if (b->branch_tree.tree) {
2854 release_tree_content_recursive(b->branch_tree.tree);
2855 b->branch_tree.tree = NULL;
2858 else
2859 b = new_branch(arg);
2860 read_next_command();
2861 parse_from(b);
2862 if (command_buf.len > 0)
2863 unread_command_buf = 1;
2866 static void cat_blob_write(const char *buf, unsigned long size)
2868 if (write_in_full(cat_blob_fd, buf, size) != size)
2869 die_errno("Write to frontend failed");
2872 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2874 struct strbuf line = STRBUF_INIT;
2875 unsigned long size;
2876 enum object_type type = 0;
2877 char *buf;
2879 if (!oe || oe->pack_id == MAX_PACK_ID) {
2880 buf = read_sha1_file(sha1, &type, &size);
2881 } else {
2882 type = oe->type;
2883 buf = gfi_unpack_entry(oe, &size);
2887 * Output based on batch_one_object() from cat-file.c.
2889 if (type <= 0) {
2890 strbuf_reset(&line);
2891 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2892 cat_blob_write(line.buf, line.len);
2893 strbuf_release(&line);
2894 free(buf);
2895 return;
2897 if (!buf)
2898 die("Can't read object %s", sha1_to_hex(sha1));
2899 if (type != OBJ_BLOB)
2900 die("Object %s is a %s but a blob was expected.",
2901 sha1_to_hex(sha1), typename(type));
2902 strbuf_reset(&line);
2903 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2904 typename(type), size);
2905 cat_blob_write(line.buf, line.len);
2906 strbuf_release(&line);
2907 cat_blob_write(buf, size);
2908 cat_blob_write("\n", 1);
2909 if (oe && oe->pack_id == pack_id) {
2910 last_blob.offset = oe->idx.offset;
2911 strbuf_attach(&last_blob.data, buf, size, size);
2912 last_blob.depth = oe->depth;
2913 } else
2914 free(buf);
2917 static void parse_cat_blob(const char *p)
2919 struct object_entry *oe = oe;
2920 unsigned char sha1[20];
2922 /* cat-blob SP <object> LF */
2923 if (*p == ':') {
2924 oe = find_mark(parse_mark_ref_eol(p));
2925 if (!oe)
2926 die("Unknown mark: %s", command_buf.buf);
2927 hashcpy(sha1, oe->idx.sha1);
2928 } else {
2929 if (get_sha1_hex(p, sha1))
2930 die("Invalid dataref: %s", command_buf.buf);
2931 if (p[40])
2932 die("Garbage after SHA1: %s", command_buf.buf);
2933 oe = find_object(sha1);
2936 cat_blob(oe, sha1);
2939 static struct object_entry *dereference(struct object_entry *oe,
2940 unsigned char sha1[20])
2942 unsigned long size;
2943 char *buf = NULL;
2944 if (!oe) {
2945 enum object_type type = sha1_object_info(sha1, NULL);
2946 if (type < 0)
2947 die("object not found: %s", sha1_to_hex(sha1));
2948 /* cache it! */
2949 oe = insert_object(sha1);
2950 oe->type = type;
2951 oe->pack_id = MAX_PACK_ID;
2952 oe->idx.offset = 1;
2954 switch (oe->type) {
2955 case OBJ_TREE: /* easy case. */
2956 return oe;
2957 case OBJ_COMMIT:
2958 case OBJ_TAG:
2959 break;
2960 default:
2961 die("Not a tree-ish: %s", command_buf.buf);
2964 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2965 buf = gfi_unpack_entry(oe, &size);
2966 } else {
2967 enum object_type unused;
2968 buf = read_sha1_file(sha1, &unused, &size);
2970 if (!buf)
2971 die("Can't load object %s", sha1_to_hex(sha1));
2973 /* Peel one layer. */
2974 switch (oe->type) {
2975 case OBJ_TAG:
2976 if (size < 40 + strlen("object ") ||
2977 get_sha1_hex(buf + strlen("object "), sha1))
2978 die("Invalid SHA1 in tag: %s", command_buf.buf);
2979 break;
2980 case OBJ_COMMIT:
2981 if (size < 40 + strlen("tree ") ||
2982 get_sha1_hex(buf + strlen("tree "), sha1))
2983 die("Invalid SHA1 in commit: %s", command_buf.buf);
2986 free(buf);
2987 return find_object(sha1);
2990 static struct object_entry *parse_treeish_dataref(const char **p)
2992 unsigned char sha1[20];
2993 struct object_entry *e;
2995 if (**p == ':') { /* <mark> */
2996 e = find_mark(parse_mark_ref_space(p));
2997 if (!e)
2998 die("Unknown mark: %s", command_buf.buf);
2999 hashcpy(sha1, e->idx.sha1);
3000 } else { /* <sha1> */
3001 if (get_sha1_hex(*p, sha1))
3002 die("Invalid dataref: %s", command_buf.buf);
3003 e = find_object(sha1);
3004 *p += 40;
3005 if (*(*p)++ != ' ')
3006 die("Missing space after tree-ish: %s", command_buf.buf);
3009 while (!e || e->type != OBJ_TREE)
3010 e = dereference(e, sha1);
3011 return e;
3014 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3016 static struct strbuf line = STRBUF_INIT;
3018 /* See show_tree(). */
3019 const char *type =
3020 S_ISGITLINK(mode) ? commit_type :
3021 S_ISDIR(mode) ? tree_type :
3022 blob_type;
3024 if (!mode) {
3025 /* missing SP path LF */
3026 strbuf_reset(&line);
3027 strbuf_addstr(&line, "missing ");
3028 quote_c_style(path, &line, NULL, 0);
3029 strbuf_addch(&line, '\n');
3030 } else {
3031 /* mode SP type SP object_name TAB path LF */
3032 strbuf_reset(&line);
3033 strbuf_addf(&line, "%06o %s %s\t",
3034 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3035 quote_c_style(path, &line, NULL, 0);
3036 strbuf_addch(&line, '\n');
3038 cat_blob_write(line.buf, line.len);
3041 static void parse_ls(const char *p, struct branch *b)
3043 struct tree_entry *root = NULL;
3044 struct tree_entry leaf = {NULL};
3046 /* ls SP (<tree-ish> SP)? <path> */
3047 if (*p == '"') {
3048 if (!b)
3049 die("Not in a commit: %s", command_buf.buf);
3050 root = &b->branch_tree;
3051 } else {
3052 struct object_entry *e = parse_treeish_dataref(&p);
3053 root = new_tree_entry();
3054 hashcpy(root->versions[1].sha1, e->idx.sha1);
3055 if (!is_null_sha1(root->versions[1].sha1))
3056 root->versions[1].mode = S_IFDIR;
3057 load_tree(root);
3059 if (*p == '"') {
3060 static struct strbuf uq = STRBUF_INIT;
3061 const char *endp;
3062 strbuf_reset(&uq);
3063 if (unquote_c_style(&uq, p, &endp))
3064 die("Invalid path: %s", command_buf.buf);
3065 if (*endp)
3066 die("Garbage after path in: %s", command_buf.buf);
3067 p = uq.buf;
3069 tree_content_get(root, p, &leaf, 1);
3071 * A directory in preparation would have a sha1 of zero
3072 * until it is saved. Save, for simplicity.
3074 if (S_ISDIR(leaf.versions[1].mode))
3075 store_tree(&leaf);
3077 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3078 if (leaf.tree)
3079 release_tree_content_recursive(leaf.tree);
3080 if (!b || root != &b->branch_tree)
3081 release_tree_entry(root);
3084 static void checkpoint(void)
3086 checkpoint_requested = 0;
3087 if (object_count) {
3088 cycle_packfile();
3089 dump_branches();
3090 dump_tags();
3091 dump_marks();
3095 static void parse_checkpoint(void)
3097 checkpoint_requested = 1;
3098 skip_optional_lf();
3101 static void parse_progress(void)
3103 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3104 fputc('\n', stdout);
3105 fflush(stdout);
3106 skip_optional_lf();
3109 static char* make_fast_import_path(const char *path)
3111 struct strbuf abs_path = STRBUF_INIT;
3113 if (!relative_marks_paths || is_absolute_path(path))
3114 return xstrdup(path);
3115 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
3116 return strbuf_detach(&abs_path, NULL);
3119 static void option_import_marks(const char *marks,
3120 int from_stream, int ignore_missing)
3122 if (import_marks_file) {
3123 if (from_stream)
3124 die("Only one import-marks command allowed per stream");
3126 /* read previous mark file */
3127 if(!import_marks_file_from_stream)
3128 read_marks();
3131 import_marks_file = make_fast_import_path(marks);
3132 safe_create_leading_directories_const(import_marks_file);
3133 import_marks_file_from_stream = from_stream;
3134 import_marks_file_ignore_missing = ignore_missing;
3137 static void option_date_format(const char *fmt)
3139 if (!strcmp(fmt, "raw"))
3140 whenspec = WHENSPEC_RAW;
3141 else if (!strcmp(fmt, "rfc2822"))
3142 whenspec = WHENSPEC_RFC2822;
3143 else if (!strcmp(fmt, "now"))
3144 whenspec = WHENSPEC_NOW;
3145 else
3146 die("unknown --date-format argument %s", fmt);
3149 static unsigned long ulong_arg(const char *option, const char *arg)
3151 char *endptr;
3152 unsigned long rv = strtoul(arg, &endptr, 0);
3153 if (strchr(arg, '-') || endptr == arg || *endptr)
3154 die("%s: argument must be a non-negative integer", option);
3155 return rv;
3158 static void option_depth(const char *depth)
3160 max_depth = ulong_arg("--depth", depth);
3161 if (max_depth > MAX_DEPTH)
3162 die("--depth cannot exceed %u", MAX_DEPTH);
3165 static void option_active_branches(const char *branches)
3167 max_active_branches = ulong_arg("--active-branches", branches);
3170 static void option_export_marks(const char *marks)
3172 export_marks_file = make_fast_import_path(marks);
3173 safe_create_leading_directories_const(export_marks_file);
3176 static void option_cat_blob_fd(const char *fd)
3178 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3179 if (n > (unsigned long) INT_MAX)
3180 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3181 cat_blob_fd = (int) n;
3184 static void option_export_pack_edges(const char *edges)
3186 if (pack_edges)
3187 fclose(pack_edges);
3188 pack_edges = fopen(edges, "a");
3189 if (!pack_edges)
3190 die_errno("Cannot open '%s'", edges);
3193 static int parse_one_option(const char *option)
3195 if (skip_prefix(option, "max-pack-size=", &option)) {
3196 unsigned long v;
3197 if (!git_parse_ulong(option, &v))
3198 return 0;
3199 if (v < 8192) {
3200 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3201 v *= 1024 * 1024;
3202 } else if (v < 1024 * 1024) {
3203 warning("minimum max-pack-size is 1 MiB");
3204 v = 1024 * 1024;
3206 max_packsize = v;
3207 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3208 unsigned long v;
3209 if (!git_parse_ulong(option, &v))
3210 return 0;
3211 big_file_threshold = v;
3212 } else if (skip_prefix(option, "depth=", &option)) {
3213 option_depth(option);
3214 } else if (skip_prefix(option, "active-branches=", &option)) {
3215 option_active_branches(option);
3216 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3217 option_export_pack_edges(option);
3218 } else if (starts_with(option, "quiet")) {
3219 show_stats = 0;
3220 } else if (starts_with(option, "stats")) {
3221 show_stats = 1;
3222 } else {
3223 return 0;
3226 return 1;
3229 static int parse_one_feature(const char *feature, int from_stream)
3231 const char *arg;
3233 if (skip_prefix(feature, "date-format=", &arg)) {
3234 option_date_format(arg);
3235 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3236 option_import_marks(arg, from_stream, 0);
3237 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3238 option_import_marks(arg, from_stream, 1);
3239 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3240 option_export_marks(arg);
3241 } else if (!strcmp(feature, "cat-blob")) {
3242 ; /* Don't die - this feature is supported */
3243 } else if (!strcmp(feature, "relative-marks")) {
3244 relative_marks_paths = 1;
3245 } else if (!strcmp(feature, "no-relative-marks")) {
3246 relative_marks_paths = 0;
3247 } else if (!strcmp(feature, "done")) {
3248 require_explicit_termination = 1;
3249 } else if (!strcmp(feature, "force")) {
3250 force_update = 1;
3251 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3252 ; /* do nothing; we have the feature */
3253 } else {
3254 return 0;
3257 return 1;
3260 static void parse_feature(const char *feature)
3262 if (seen_data_command)
3263 die("Got feature command '%s' after data command", feature);
3265 if (parse_one_feature(feature, 1))
3266 return;
3268 die("This version of fast-import does not support feature %s.", feature);
3271 static void parse_option(const char *option)
3273 if (seen_data_command)
3274 die("Got option command '%s' after data command", option);
3276 if (parse_one_option(option))
3277 return;
3279 die("This version of fast-import does not support option: %s", option);
3282 static int git_pack_config(const char *k, const char *v, void *cb)
3284 if (!strcmp(k, "pack.depth")) {
3285 max_depth = git_config_int(k, v);
3286 if (max_depth > MAX_DEPTH)
3287 max_depth = MAX_DEPTH;
3288 return 0;
3290 if (!strcmp(k, "pack.compression")) {
3291 int level = git_config_int(k, v);
3292 if (level == -1)
3293 level = Z_DEFAULT_COMPRESSION;
3294 else if (level < 0 || level > Z_BEST_COMPRESSION)
3295 die("bad pack compression level %d", level);
3296 pack_compression_level = level;
3297 pack_compression_seen = 1;
3298 return 0;
3300 if (!strcmp(k, "pack.indexversion")) {
3301 pack_idx_opts.version = git_config_int(k, v);
3302 if (pack_idx_opts.version > 2)
3303 die("bad pack.indexversion=%"PRIu32,
3304 pack_idx_opts.version);
3305 return 0;
3307 if (!strcmp(k, "pack.packsizelimit")) {
3308 max_packsize = git_config_ulong(k, v);
3309 return 0;
3311 return git_default_config(k, v, cb);
3314 static const char fast_import_usage[] =
3315 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3317 static void parse_argv(void)
3319 unsigned int i;
3321 for (i = 1; i < global_argc; i++) {
3322 const char *a = global_argv[i];
3324 if (*a != '-' || !strcmp(a, "--"))
3325 break;
3327 if (!skip_prefix(a, "--", &a))
3328 die("unknown option %s", a);
3330 if (parse_one_option(a))
3331 continue;
3333 if (parse_one_feature(a, 0))
3334 continue;
3336 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3337 option_cat_blob_fd(a);
3338 continue;
3341 die("unknown option --%s", a);
3343 if (i != global_argc)
3344 usage(fast_import_usage);
3346 seen_data_command = 1;
3347 if (import_marks_file)
3348 read_marks();
3351 int main(int argc, char **argv)
3353 unsigned int i;
3355 git_extract_argv0_path(argv[0]);
3357 git_setup_gettext();
3359 if (argc == 2 && !strcmp(argv[1], "-h"))
3360 usage(fast_import_usage);
3362 setup_git_directory();
3363 reset_pack_idx_option(&pack_idx_opts);
3364 git_config(git_pack_config, NULL);
3365 if (!pack_compression_seen && core_compression_seen)
3366 pack_compression_level = core_compression_level;
3368 alloc_objects(object_entry_alloc);
3369 strbuf_init(&command_buf, 0);
3370 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3371 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3372 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3373 marks = pool_calloc(1, sizeof(struct mark_set));
3375 global_argc = argc;
3376 global_argv = argv;
3378 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3379 for (i = 0; i < (cmd_save - 1); i++)
3380 rc_free[i].next = &rc_free[i + 1];
3381 rc_free[cmd_save - 1].next = NULL;
3383 prepare_packed_git();
3384 start_packfile();
3385 set_die_routine(die_nicely);
3386 set_checkpoint_signal();
3387 while (read_next_command() != EOF) {
3388 const char *v;
3389 if (!strcmp("blob", command_buf.buf))
3390 parse_new_blob();
3391 else if (skip_prefix(command_buf.buf, "ls ", &v))
3392 parse_ls(v, NULL);
3393 else if (skip_prefix(command_buf.buf, "commit ", &v))
3394 parse_new_commit(v);
3395 else if (skip_prefix(command_buf.buf, "tag ", &v))
3396 parse_new_tag(v);
3397 else if (skip_prefix(command_buf.buf, "reset ", &v))
3398 parse_reset_branch(v);
3399 else if (!strcmp("checkpoint", command_buf.buf))
3400 parse_checkpoint();
3401 else if (!strcmp("done", command_buf.buf))
3402 break;
3403 else if (starts_with(command_buf.buf, "progress "))
3404 parse_progress();
3405 else if (skip_prefix(command_buf.buf, "feature ", &v))
3406 parse_feature(v);
3407 else if (skip_prefix(command_buf.buf, "option git ", &v))
3408 parse_option(v);
3409 else if (starts_with(command_buf.buf, "option "))
3410 /* ignore non-git options*/;
3411 else
3412 die("Unsupported command: %s", command_buf.buf);
3414 if (checkpoint_requested)
3415 checkpoint();
3418 /* argv hasn't been parsed yet, do so */
3419 if (!seen_data_command)
3420 parse_argv();
3422 if (require_explicit_termination && feof(stdin))
3423 die("stream ends early");
3425 end_packfile();
3427 dump_branches();
3428 dump_tags();
3429 unkeep_all_packs();
3430 dump_marks();
3432 if (pack_edges)
3433 fclose(pack_edges);
3435 if (show_stats) {
3436 uintmax_t total_count = 0, duplicate_count = 0;
3437 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3438 total_count += object_count_by_type[i];
3439 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3440 duplicate_count += duplicate_count_by_type[i];
3442 fprintf(stderr, "%s statistics:\n", argv[0]);
3443 fprintf(stderr, "---------------------------------------------------------------------\n");
3444 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3445 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3446 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]);
3447 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]);
3448 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]);
3449 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]);
3450 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3451 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3452 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3453 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3454 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3455 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3456 fprintf(stderr, "---------------------------------------------------------------------\n");
3457 pack_report();
3458 fprintf(stderr, "---------------------------------------------------------------------\n");
3459 fprintf(stderr, "\n");
3462 return failure ? 1 : 0;