tests: rename work-tree tests to *work-tree*
[git/gitweb.git] / fast-import.c
blobece009662379dd26a523cb082b50c11109c96b14
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 "lockfile.h"
157 #include "object.h"
158 #include "blob.h"
159 #include "tree.h"
160 #include "commit.h"
161 #include "delta.h"
162 #include "pack.h"
163 #include "refs.h"
164 #include "csum-file.h"
165 #include "quote.h"
166 #include "exec_cmd.h"
167 #include "dir.h"
169 #define PACK_ID_BITS 16
170 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
171 #define DEPTH_BITS 13
172 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
175 * We abuse the setuid bit on directories to mean "do not delta".
177 #define NO_DELTA S_ISUID
179 struct object_entry {
180 struct pack_idx_entry idx;
181 struct object_entry *next;
182 uint32_t type : TYPE_BITS,
183 pack_id : PACK_ID_BITS,
184 depth : DEPTH_BITS;
187 struct object_entry_pool {
188 struct object_entry_pool *next_pool;
189 struct object_entry *next_free;
190 struct object_entry *end;
191 struct object_entry entries[FLEX_ARRAY]; /* more */
194 struct mark_set {
195 union {
196 struct object_entry *marked[1024];
197 struct mark_set *sets[1024];
198 } data;
199 unsigned int shift;
202 struct last_object {
203 struct strbuf data;
204 off_t offset;
205 unsigned int depth;
206 unsigned no_swap : 1;
209 struct mem_pool {
210 struct mem_pool *next_pool;
211 char *next_free;
212 char *end;
213 uintmax_t space[FLEX_ARRAY]; /* more */
216 struct atom_str {
217 struct atom_str *next_atom;
218 unsigned short str_len;
219 char str_dat[FLEX_ARRAY]; /* more */
222 struct tree_content;
223 struct tree_entry {
224 struct tree_content *tree;
225 struct atom_str *name;
226 struct tree_entry_ms {
227 uint16_t mode;
228 unsigned char sha1[20];
229 } versions[2];
232 struct tree_content {
233 unsigned int entry_capacity; /* must match avail_tree_content */
234 unsigned int entry_count;
235 unsigned int delta_depth;
236 struct tree_entry *entries[FLEX_ARRAY]; /* more */
239 struct avail_tree_content {
240 unsigned int entry_capacity; /* must match tree_content */
241 struct avail_tree_content *next_avail;
244 struct branch {
245 struct branch *table_next_branch;
246 struct branch *active_next_branch;
247 const char *name;
248 struct tree_entry branch_tree;
249 uintmax_t last_commit;
250 uintmax_t num_notes;
251 unsigned active : 1;
252 unsigned delete : 1;
253 unsigned pack_id : PACK_ID_BITS;
254 unsigned char sha1[20];
257 struct tag {
258 struct tag *next_tag;
259 const char *name;
260 unsigned int pack_id;
261 unsigned char sha1[20];
264 struct hash_list {
265 struct hash_list *next;
266 unsigned char sha1[20];
269 typedef enum {
270 WHENSPEC_RAW = 1,
271 WHENSPEC_RFC2822,
272 WHENSPEC_NOW
273 } whenspec_type;
275 struct recent_command {
276 struct recent_command *prev;
277 struct recent_command *next;
278 char *buf;
281 /* Configured limits on output */
282 static unsigned long max_depth = 10;
283 static off_t max_packsize;
284 static int force_update;
285 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
286 static int pack_compression_seen;
288 /* Stats and misc. counters */
289 static uintmax_t alloc_count;
290 static uintmax_t marks_set_count;
291 static uintmax_t object_count_by_type[1 << TYPE_BITS];
292 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
293 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
294 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
295 static unsigned long object_count;
296 static unsigned long branch_count;
297 static unsigned long branch_load_count;
298 static int failure;
299 static FILE *pack_edges;
300 static unsigned int show_stats = 1;
301 static int global_argc;
302 static char **global_argv;
304 /* Memory pools */
305 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
306 static size_t total_allocd;
307 static struct mem_pool *mem_pool;
309 /* Atom management */
310 static unsigned int atom_table_sz = 4451;
311 static unsigned int atom_cnt;
312 static struct atom_str **atom_table;
314 /* The .pack file being generated */
315 static struct pack_idx_option pack_idx_opts;
316 static unsigned int pack_id;
317 static struct sha1file *pack_file;
318 static struct packed_git *pack_data;
319 static struct packed_git **all_packs;
320 static off_t pack_size;
322 /* Table of objects we've written. */
323 static unsigned int object_entry_alloc = 5000;
324 static struct object_entry_pool *blocks;
325 static struct object_entry *object_table[1 << 16];
326 static struct mark_set *marks;
327 static const char *export_marks_file;
328 static const char *import_marks_file;
329 static int import_marks_file_from_stream;
330 static int import_marks_file_ignore_missing;
331 static int relative_marks_paths;
333 /* Our last blob */
334 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
336 /* Tree management */
337 static unsigned int tree_entry_alloc = 1000;
338 static void *avail_tree_entry;
339 static unsigned int avail_tree_table_sz = 100;
340 static struct avail_tree_content **avail_tree_table;
341 static struct strbuf old_tree = STRBUF_INIT;
342 static struct strbuf new_tree = STRBUF_INIT;
344 /* Branch data */
345 static unsigned long max_active_branches = 5;
346 static unsigned long cur_active_branches;
347 static unsigned long branch_table_sz = 1039;
348 static struct branch **branch_table;
349 static struct branch *active_branches;
351 /* Tag data */
352 static struct tag *first_tag;
353 static struct tag *last_tag;
355 /* Input stream parsing */
356 static whenspec_type whenspec = WHENSPEC_RAW;
357 static struct strbuf command_buf = STRBUF_INIT;
358 static int unread_command_buf;
359 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
360 static struct recent_command *cmd_tail = &cmd_hist;
361 static struct recent_command *rc_free;
362 static unsigned int cmd_save = 100;
363 static uintmax_t next_mark;
364 static struct strbuf new_data = STRBUF_INIT;
365 static int seen_data_command;
366 static int require_explicit_termination;
368 /* Signal handling */
369 static volatile sig_atomic_t checkpoint_requested;
371 /* Where to write output of cat-blob commands */
372 static int cat_blob_fd = STDOUT_FILENO;
374 static void parse_argv(void);
375 static void parse_cat_blob(const char *p);
376 static void parse_ls(const char *p, struct branch *b);
378 static void write_branch_report(FILE *rpt, struct branch *b)
380 fprintf(rpt, "%s:\n", b->name);
382 fprintf(rpt, " status :");
383 if (b->active)
384 fputs(" active", rpt);
385 if (b->branch_tree.tree)
386 fputs(" loaded", rpt);
387 if (is_null_sha1(b->branch_tree.versions[1].sha1))
388 fputs(" dirty", rpt);
389 fputc('\n', rpt);
391 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
392 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
393 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
394 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
396 fputs(" last pack : ", rpt);
397 if (b->pack_id < MAX_PACK_ID)
398 fprintf(rpt, "%u", b->pack_id);
399 fputc('\n', rpt);
401 fputc('\n', rpt);
404 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
406 static void write_crash_report(const char *err)
408 const char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
409 FILE *rpt = fopen(loc, "w");
410 struct branch *b;
411 unsigned long lu;
412 struct recent_command *rc;
414 if (!rpt) {
415 error("can't write crash report %s: %s", loc, strerror(errno));
416 return;
419 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
421 fprintf(rpt, "fast-import crash report:\n");
422 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
423 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
424 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
425 fputc('\n', rpt);
427 fputs("fatal: ", rpt);
428 fputs(err, rpt);
429 fputc('\n', rpt);
431 fputc('\n', rpt);
432 fputs("Most Recent Commands Before Crash\n", rpt);
433 fputs("---------------------------------\n", rpt);
434 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
435 if (rc->next == &cmd_hist)
436 fputs("* ", rpt);
437 else
438 fputs(" ", rpt);
439 fputs(rc->buf, rpt);
440 fputc('\n', rpt);
443 fputc('\n', rpt);
444 fputs("Active Branch LRU\n", rpt);
445 fputs("-----------------\n", rpt);
446 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
447 cur_active_branches,
448 max_active_branches);
449 fputc('\n', rpt);
450 fputs(" pos clock name\n", rpt);
451 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
452 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
453 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
454 ++lu, b->last_commit, b->name);
456 fputc('\n', rpt);
457 fputs("Inactive Branches\n", rpt);
458 fputs("-----------------\n", rpt);
459 for (lu = 0; lu < branch_table_sz; lu++) {
460 for (b = branch_table[lu]; b; b = b->table_next_branch)
461 write_branch_report(rpt, b);
464 if (first_tag) {
465 struct tag *tg;
466 fputc('\n', rpt);
467 fputs("Annotated Tags\n", rpt);
468 fputs("--------------\n", rpt);
469 for (tg = first_tag; tg; tg = tg->next_tag) {
470 fputs(sha1_to_hex(tg->sha1), rpt);
471 fputc(' ', rpt);
472 fputs(tg->name, rpt);
473 fputc('\n', rpt);
477 fputc('\n', rpt);
478 fputs("Marks\n", rpt);
479 fputs("-----\n", rpt);
480 if (export_marks_file)
481 fprintf(rpt, " exported to %s\n", export_marks_file);
482 else
483 dump_marks_helper(rpt, 0, marks);
485 fputc('\n', rpt);
486 fputs("-------------------\n", rpt);
487 fputs("END OF CRASH REPORT\n", rpt);
488 fclose(rpt);
491 static void end_packfile(void);
492 static void unkeep_all_packs(void);
493 static void dump_marks(void);
495 static NORETURN void die_nicely(const char *err, va_list params)
497 static int zombie;
498 char message[2 * PATH_MAX];
500 vsnprintf(message, sizeof(message), err, params);
501 fputs("fatal: ", stderr);
502 fputs(message, stderr);
503 fputc('\n', stderr);
505 if (!zombie) {
506 zombie = 1;
507 write_crash_report(message);
508 end_packfile();
509 unkeep_all_packs();
510 dump_marks();
512 exit(128);
515 #ifndef SIGUSR1 /* Windows, for example */
517 static void set_checkpoint_signal(void)
521 #else
523 static void checkpoint_signal(int signo)
525 checkpoint_requested = 1;
528 static void set_checkpoint_signal(void)
530 struct sigaction sa;
532 memset(&sa, 0, sizeof(sa));
533 sa.sa_handler = checkpoint_signal;
534 sigemptyset(&sa.sa_mask);
535 sa.sa_flags = SA_RESTART;
536 sigaction(SIGUSR1, &sa, NULL);
539 #endif
541 static void alloc_objects(unsigned int cnt)
543 struct object_entry_pool *b;
545 b = xmalloc(sizeof(struct object_entry_pool)
546 + cnt * sizeof(struct object_entry));
547 b->next_pool = blocks;
548 b->next_free = b->entries;
549 b->end = b->entries + cnt;
550 blocks = b;
551 alloc_count += cnt;
554 static struct object_entry *new_object(unsigned char *sha1)
556 struct object_entry *e;
558 if (blocks->next_free == blocks->end)
559 alloc_objects(object_entry_alloc);
561 e = blocks->next_free++;
562 hashcpy(e->idx.sha1, sha1);
563 return e;
566 static struct object_entry *find_object(unsigned char *sha1)
568 unsigned int h = sha1[0] << 8 | sha1[1];
569 struct object_entry *e;
570 for (e = object_table[h]; e; e = e->next)
571 if (!hashcmp(sha1, e->idx.sha1))
572 return e;
573 return NULL;
576 static struct object_entry *insert_object(unsigned char *sha1)
578 unsigned int h = sha1[0] << 8 | sha1[1];
579 struct object_entry *e = object_table[h];
581 while (e) {
582 if (!hashcmp(sha1, e->idx.sha1))
583 return e;
584 e = e->next;
587 e = new_object(sha1);
588 e->next = object_table[h];
589 e->idx.offset = 0;
590 object_table[h] = e;
591 return e;
594 static unsigned int hc_str(const char *s, size_t len)
596 unsigned int r = 0;
597 while (len-- > 0)
598 r = r * 31 + *s++;
599 return r;
602 static void *pool_alloc(size_t len)
604 struct mem_pool *p;
605 void *r;
607 /* round up to a 'uintmax_t' alignment */
608 if (len & (sizeof(uintmax_t) - 1))
609 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
611 for (p = mem_pool; p; p = p->next_pool)
612 if ((p->end - p->next_free >= len))
613 break;
615 if (!p) {
616 if (len >= (mem_pool_alloc/2)) {
617 total_allocd += len;
618 return xmalloc(len);
620 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
621 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
622 p->next_pool = mem_pool;
623 p->next_free = (char *) p->space;
624 p->end = p->next_free + mem_pool_alloc;
625 mem_pool = p;
628 r = p->next_free;
629 p->next_free += len;
630 return r;
633 static void *pool_calloc(size_t count, size_t size)
635 size_t len = count * size;
636 void *r = pool_alloc(len);
637 memset(r, 0, len);
638 return r;
641 static char *pool_strdup(const char *s)
643 char *r = pool_alloc(strlen(s) + 1);
644 strcpy(r, s);
645 return r;
648 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
650 struct mark_set *s = marks;
651 while ((idnum >> s->shift) >= 1024) {
652 s = pool_calloc(1, sizeof(struct mark_set));
653 s->shift = marks->shift + 10;
654 s->data.sets[0] = marks;
655 marks = s;
657 while (s->shift) {
658 uintmax_t i = idnum >> s->shift;
659 idnum -= i << s->shift;
660 if (!s->data.sets[i]) {
661 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
662 s->data.sets[i]->shift = s->shift - 10;
664 s = s->data.sets[i];
666 if (!s->data.marked[idnum])
667 marks_set_count++;
668 s->data.marked[idnum] = oe;
671 static struct object_entry *find_mark(uintmax_t idnum)
673 uintmax_t orig_idnum = idnum;
674 struct mark_set *s = marks;
675 struct object_entry *oe = NULL;
676 if ((idnum >> s->shift) < 1024) {
677 while (s && s->shift) {
678 uintmax_t i = idnum >> s->shift;
679 idnum -= i << s->shift;
680 s = s->data.sets[i];
682 if (s)
683 oe = s->data.marked[idnum];
685 if (!oe)
686 die("mark :%" PRIuMAX " not declared", orig_idnum);
687 return oe;
690 static struct atom_str *to_atom(const char *s, unsigned short len)
692 unsigned int hc = hc_str(s, len) % atom_table_sz;
693 struct atom_str *c;
695 for (c = atom_table[hc]; c; c = c->next_atom)
696 if (c->str_len == len && !strncmp(s, c->str_dat, len))
697 return c;
699 c = pool_alloc(sizeof(struct atom_str) + len + 1);
700 c->str_len = len;
701 strncpy(c->str_dat, s, len);
702 c->str_dat[len] = 0;
703 c->next_atom = atom_table[hc];
704 atom_table[hc] = c;
705 atom_cnt++;
706 return c;
709 static struct branch *lookup_branch(const char *name)
711 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
712 struct branch *b;
714 for (b = branch_table[hc]; b; b = b->table_next_branch)
715 if (!strcmp(name, b->name))
716 return b;
717 return NULL;
720 static struct branch *new_branch(const char *name)
722 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
723 struct branch *b = lookup_branch(name);
725 if (b)
726 die("Invalid attempt to create duplicate branch: %s", name);
727 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
728 die("Branch name doesn't conform to GIT standards: %s", name);
730 b = pool_calloc(1, sizeof(struct branch));
731 b->name = pool_strdup(name);
732 b->table_next_branch = branch_table[hc];
733 b->branch_tree.versions[0].mode = S_IFDIR;
734 b->branch_tree.versions[1].mode = S_IFDIR;
735 b->num_notes = 0;
736 b->active = 0;
737 b->pack_id = MAX_PACK_ID;
738 branch_table[hc] = b;
739 branch_count++;
740 return b;
743 static unsigned int hc_entries(unsigned int cnt)
745 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
746 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
749 static struct tree_content *new_tree_content(unsigned int cnt)
751 struct avail_tree_content *f, *l = NULL;
752 struct tree_content *t;
753 unsigned int hc = hc_entries(cnt);
755 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
756 if (f->entry_capacity >= cnt)
757 break;
759 if (f) {
760 if (l)
761 l->next_avail = f->next_avail;
762 else
763 avail_tree_table[hc] = f->next_avail;
764 } else {
765 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
766 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
767 f->entry_capacity = cnt;
770 t = (struct tree_content*)f;
771 t->entry_count = 0;
772 t->delta_depth = 0;
773 return t;
776 static void release_tree_entry(struct tree_entry *e);
777 static void release_tree_content(struct tree_content *t)
779 struct avail_tree_content *f = (struct avail_tree_content*)t;
780 unsigned int hc = hc_entries(f->entry_capacity);
781 f->next_avail = avail_tree_table[hc];
782 avail_tree_table[hc] = f;
785 static void release_tree_content_recursive(struct tree_content *t)
787 unsigned int i;
788 for (i = 0; i < t->entry_count; i++)
789 release_tree_entry(t->entries[i]);
790 release_tree_content(t);
793 static struct tree_content *grow_tree_content(
794 struct tree_content *t,
795 int amt)
797 struct tree_content *r = new_tree_content(t->entry_count + amt);
798 r->entry_count = t->entry_count;
799 r->delta_depth = t->delta_depth;
800 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
801 release_tree_content(t);
802 return r;
805 static struct tree_entry *new_tree_entry(void)
807 struct tree_entry *e;
809 if (!avail_tree_entry) {
810 unsigned int n = tree_entry_alloc;
811 total_allocd += n * sizeof(struct tree_entry);
812 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
813 while (n-- > 1) {
814 *((void**)e) = e + 1;
815 e++;
817 *((void**)e) = NULL;
820 e = avail_tree_entry;
821 avail_tree_entry = *((void**)e);
822 return e;
825 static void release_tree_entry(struct tree_entry *e)
827 if (e->tree)
828 release_tree_content_recursive(e->tree);
829 *((void**)e) = avail_tree_entry;
830 avail_tree_entry = e;
833 static struct tree_content *dup_tree_content(struct tree_content *s)
835 struct tree_content *d;
836 struct tree_entry *a, *b;
837 unsigned int i;
839 if (!s)
840 return NULL;
841 d = new_tree_content(s->entry_count);
842 for (i = 0; i < s->entry_count; i++) {
843 a = s->entries[i];
844 b = new_tree_entry();
845 memcpy(b, a, sizeof(*a));
846 if (a->tree && is_null_sha1(b->versions[1].sha1))
847 b->tree = dup_tree_content(a->tree);
848 else
849 b->tree = NULL;
850 d->entries[i] = b;
852 d->entry_count = s->entry_count;
853 d->delta_depth = s->delta_depth;
855 return d;
858 static void start_packfile(void)
860 static char tmp_file[PATH_MAX];
861 struct packed_git *p;
862 struct pack_header hdr;
863 int pack_fd;
865 pack_fd = odb_mkstemp(tmp_file, sizeof(tmp_file),
866 "pack/tmp_pack_XXXXXX");
867 p = xcalloc(1, sizeof(*p) + strlen(tmp_file) + 2);
868 strcpy(p->pack_name, tmp_file);
869 p->pack_fd = pack_fd;
870 p->do_not_close = 1;
871 pack_file = sha1fd(pack_fd, p->pack_name);
873 hdr.hdr_signature = htonl(PACK_SIGNATURE);
874 hdr.hdr_version = htonl(2);
875 hdr.hdr_entries = 0;
876 sha1write(pack_file, &hdr, sizeof(hdr));
878 pack_data = p;
879 pack_size = sizeof(hdr);
880 object_count = 0;
882 REALLOC_ARRAY(all_packs, pack_id + 1);
883 all_packs[pack_id] = p;
886 static const char *create_index(void)
888 const char *tmpfile;
889 struct pack_idx_entry **idx, **c, **last;
890 struct object_entry *e;
891 struct object_entry_pool *o;
893 /* Build the table of object IDs. */
894 idx = xmalloc(object_count * sizeof(*idx));
895 c = idx;
896 for (o = blocks; o; o = o->next_pool)
897 for (e = o->next_free; e-- != o->entries;)
898 if (pack_id == e->pack_id)
899 *c++ = &e->idx;
900 last = idx + object_count;
901 if (c != last)
902 die("internal consistency error creating the index");
904 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
905 free(idx);
906 return tmpfile;
909 static char *keep_pack(const char *curr_index_name)
911 static char name[PATH_MAX];
912 static const char *keep_msg = "fast-import";
913 int keep_fd;
915 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
916 if (keep_fd < 0)
917 die_errno("cannot create keep file");
918 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
919 if (close(keep_fd))
920 die_errno("failed to write keep file");
922 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
923 get_object_directory(), sha1_to_hex(pack_data->sha1));
924 if (move_temp_to_file(pack_data->pack_name, name))
925 die("cannot store pack file");
927 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
928 get_object_directory(), sha1_to_hex(pack_data->sha1));
929 if (move_temp_to_file(curr_index_name, name))
930 die("cannot store index file");
931 free((void *)curr_index_name);
932 return name;
935 static void unkeep_all_packs(void)
937 static char name[PATH_MAX];
938 int k;
940 for (k = 0; k < pack_id; k++) {
941 struct packed_git *p = all_packs[k];
942 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
943 get_object_directory(), sha1_to_hex(p->sha1));
944 unlink_or_warn(name);
948 static void end_packfile(void)
950 static int running;
952 if (running || !pack_data)
953 return;
955 running = 1;
956 clear_delta_base_cache();
957 if (object_count) {
958 struct packed_git *new_p;
959 unsigned char cur_pack_sha1[20];
960 char *idx_name;
961 int i;
962 struct branch *b;
963 struct tag *t;
965 close_pack_windows(pack_data);
966 sha1close(pack_file, cur_pack_sha1, 0);
967 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
968 pack_data->pack_name, object_count,
969 cur_pack_sha1, pack_size);
970 close(pack_data->pack_fd);
971 idx_name = keep_pack(create_index());
973 /* Register the packfile with core git's machinery. */
974 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
975 if (!new_p)
976 die("core git rejected index %s", idx_name);
977 all_packs[pack_id] = new_p;
978 install_packed_git(new_p);
980 /* Print the boundary */
981 if (pack_edges) {
982 fprintf(pack_edges, "%s:", new_p->pack_name);
983 for (i = 0; i < branch_table_sz; i++) {
984 for (b = branch_table[i]; b; b = b->table_next_branch) {
985 if (b->pack_id == pack_id)
986 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
989 for (t = first_tag; t; t = t->next_tag) {
990 if (t->pack_id == pack_id)
991 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
993 fputc('\n', pack_edges);
994 fflush(pack_edges);
997 pack_id++;
999 else {
1000 close(pack_data->pack_fd);
1001 unlink_or_warn(pack_data->pack_name);
1003 free(pack_data);
1004 pack_data = NULL;
1005 running = 0;
1007 /* We can't carry a delta across packfiles. */
1008 strbuf_release(&last_blob.data);
1009 last_blob.offset = 0;
1010 last_blob.depth = 0;
1013 static void cycle_packfile(void)
1015 end_packfile();
1016 start_packfile();
1019 static int store_object(
1020 enum object_type type,
1021 struct strbuf *dat,
1022 struct last_object *last,
1023 unsigned char *sha1out,
1024 uintmax_t mark)
1026 void *out, *delta;
1027 struct object_entry *e;
1028 unsigned char hdr[96];
1029 unsigned char sha1[20];
1030 unsigned long hdrlen, deltalen;
1031 git_SHA_CTX c;
1032 git_zstream s;
1034 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1035 (unsigned long)dat->len) + 1;
1036 git_SHA1_Init(&c);
1037 git_SHA1_Update(&c, hdr, hdrlen);
1038 git_SHA1_Update(&c, dat->buf, dat->len);
1039 git_SHA1_Final(sha1, &c);
1040 if (sha1out)
1041 hashcpy(sha1out, sha1);
1043 e = insert_object(sha1);
1044 if (mark)
1045 insert_mark(mark, e);
1046 if (e->idx.offset) {
1047 duplicate_count_by_type[type]++;
1048 return 1;
1049 } else if (find_sha1_pack(sha1, packed_git)) {
1050 e->type = type;
1051 e->pack_id = MAX_PACK_ID;
1052 e->idx.offset = 1; /* just not zero! */
1053 duplicate_count_by_type[type]++;
1054 return 1;
1057 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1058 delta_count_attempts_by_type[type]++;
1059 delta = diff_delta(last->data.buf, last->data.len,
1060 dat->buf, dat->len,
1061 &deltalen, dat->len - 20);
1062 } else
1063 delta = NULL;
1065 git_deflate_init(&s, pack_compression_level);
1066 if (delta) {
1067 s.next_in = delta;
1068 s.avail_in = deltalen;
1069 } else {
1070 s.next_in = (void *)dat->buf;
1071 s.avail_in = dat->len;
1073 s.avail_out = git_deflate_bound(&s, s.avail_in);
1074 s.next_out = out = xmalloc(s.avail_out);
1075 while (git_deflate(&s, Z_FINISH) == Z_OK)
1076 ; /* nothing */
1077 git_deflate_end(&s);
1079 /* Determine if we should auto-checkpoint. */
1080 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1081 || (pack_size + 60 + s.total_out) < pack_size) {
1083 /* This new object needs to *not* have the current pack_id. */
1084 e->pack_id = pack_id + 1;
1085 cycle_packfile();
1087 /* We cannot carry a delta into the new pack. */
1088 if (delta) {
1089 free(delta);
1090 delta = NULL;
1092 git_deflate_init(&s, pack_compression_level);
1093 s.next_in = (void *)dat->buf;
1094 s.avail_in = dat->len;
1095 s.avail_out = git_deflate_bound(&s, s.avail_in);
1096 s.next_out = out = xrealloc(out, s.avail_out);
1097 while (git_deflate(&s, Z_FINISH) == Z_OK)
1098 ; /* nothing */
1099 git_deflate_end(&s);
1103 e->type = type;
1104 e->pack_id = pack_id;
1105 e->idx.offset = pack_size;
1106 object_count++;
1107 object_count_by_type[type]++;
1109 crc32_begin(pack_file);
1111 if (delta) {
1112 off_t ofs = e->idx.offset - last->offset;
1113 unsigned pos = sizeof(hdr) - 1;
1115 delta_count_by_type[type]++;
1116 e->depth = last->depth + 1;
1118 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1119 sha1write(pack_file, hdr, hdrlen);
1120 pack_size += hdrlen;
1122 hdr[pos] = ofs & 127;
1123 while (ofs >>= 7)
1124 hdr[--pos] = 128 | (--ofs & 127);
1125 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1126 pack_size += sizeof(hdr) - pos;
1127 } else {
1128 e->depth = 0;
1129 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1130 sha1write(pack_file, hdr, hdrlen);
1131 pack_size += hdrlen;
1134 sha1write(pack_file, out, s.total_out);
1135 pack_size += s.total_out;
1137 e->idx.crc32 = crc32_end(pack_file);
1139 free(out);
1140 free(delta);
1141 if (last) {
1142 if (last->no_swap) {
1143 last->data = *dat;
1144 } else {
1145 strbuf_swap(&last->data, dat);
1147 last->offset = e->idx.offset;
1148 last->depth = e->depth;
1150 return 0;
1153 static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1155 if (sha1file_truncate(pack_file, checkpoint))
1156 die_errno("cannot truncate pack to skip duplicate");
1157 pack_size = checkpoint->offset;
1160 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1162 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1163 unsigned char *in_buf = xmalloc(in_sz);
1164 unsigned char *out_buf = xmalloc(out_sz);
1165 struct object_entry *e;
1166 unsigned char sha1[20];
1167 unsigned long hdrlen;
1168 off_t offset;
1169 git_SHA_CTX c;
1170 git_zstream s;
1171 struct sha1file_checkpoint checkpoint;
1172 int status = Z_OK;
1174 /* Determine if we should auto-checkpoint. */
1175 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1176 || (pack_size + 60 + len) < pack_size)
1177 cycle_packfile();
1179 sha1file_checkpoint(pack_file, &checkpoint);
1180 offset = checkpoint.offset;
1182 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1183 if (out_sz <= hdrlen)
1184 die("impossibly large object header");
1186 git_SHA1_Init(&c);
1187 git_SHA1_Update(&c, out_buf, hdrlen);
1189 crc32_begin(pack_file);
1191 git_deflate_init(&s, pack_compression_level);
1193 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1194 if (out_sz <= hdrlen)
1195 die("impossibly large object header");
1197 s.next_out = out_buf + hdrlen;
1198 s.avail_out = out_sz - hdrlen;
1200 while (status != Z_STREAM_END) {
1201 if (0 < len && !s.avail_in) {
1202 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1203 size_t n = fread(in_buf, 1, cnt, stdin);
1204 if (!n && feof(stdin))
1205 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1207 git_SHA1_Update(&c, in_buf, n);
1208 s.next_in = in_buf;
1209 s.avail_in = n;
1210 len -= n;
1213 status = git_deflate(&s, len ? 0 : Z_FINISH);
1215 if (!s.avail_out || status == Z_STREAM_END) {
1216 size_t n = s.next_out - out_buf;
1217 sha1write(pack_file, out_buf, n);
1218 pack_size += n;
1219 s.next_out = out_buf;
1220 s.avail_out = out_sz;
1223 switch (status) {
1224 case Z_OK:
1225 case Z_BUF_ERROR:
1226 case Z_STREAM_END:
1227 continue;
1228 default:
1229 die("unexpected deflate failure: %d", status);
1232 git_deflate_end(&s);
1233 git_SHA1_Final(sha1, &c);
1235 if (sha1out)
1236 hashcpy(sha1out, sha1);
1238 e = insert_object(sha1);
1240 if (mark)
1241 insert_mark(mark, e);
1243 if (e->idx.offset) {
1244 duplicate_count_by_type[OBJ_BLOB]++;
1245 truncate_pack(&checkpoint);
1247 } else if (find_sha1_pack(sha1, packed_git)) {
1248 e->type = OBJ_BLOB;
1249 e->pack_id = MAX_PACK_ID;
1250 e->idx.offset = 1; /* just not zero! */
1251 duplicate_count_by_type[OBJ_BLOB]++;
1252 truncate_pack(&checkpoint);
1254 } else {
1255 e->depth = 0;
1256 e->type = OBJ_BLOB;
1257 e->pack_id = pack_id;
1258 e->idx.offset = offset;
1259 e->idx.crc32 = crc32_end(pack_file);
1260 object_count++;
1261 object_count_by_type[OBJ_BLOB]++;
1264 free(in_buf);
1265 free(out_buf);
1268 /* All calls must be guarded by find_object() or find_mark() to
1269 * ensure the 'struct object_entry' passed was written by this
1270 * process instance. We unpack the entry by the offset, avoiding
1271 * the need for the corresponding .idx file. This unpacking rule
1272 * works because we only use OBJ_REF_DELTA within the packfiles
1273 * created by fast-import.
1275 * oe must not be NULL. Such an oe usually comes from giving
1276 * an unknown SHA-1 to find_object() or an undefined mark to
1277 * find_mark(). Callers must test for this condition and use
1278 * the standard read_sha1_file() when it happens.
1280 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1281 * find_mark(), where the mark was reloaded from an existing marks
1282 * file and is referencing an object that this fast-import process
1283 * instance did not write out to a packfile. Callers must test for
1284 * this condition and use read_sha1_file() instead.
1286 static void *gfi_unpack_entry(
1287 struct object_entry *oe,
1288 unsigned long *sizep)
1290 enum object_type type;
1291 struct packed_git *p = all_packs[oe->pack_id];
1292 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1293 /* The object is stored in the packfile we are writing to
1294 * and we have modified it since the last time we scanned
1295 * back to read a previously written object. If an old
1296 * window covered [p->pack_size, p->pack_size + 20) its
1297 * data is stale and is not valid. Closing all windows
1298 * and updating the packfile length ensures we can read
1299 * the newly written data.
1301 close_pack_windows(p);
1302 sha1flush(pack_file);
1304 /* We have to offer 20 bytes additional on the end of
1305 * the packfile as the core unpacker code assumes the
1306 * footer is present at the file end and must promise
1307 * at least 20 bytes within any window it maps. But
1308 * we don't actually create the footer here.
1310 p->pack_size = pack_size + 20;
1312 return unpack_entry(p, oe->idx.offset, &type, sizep);
1315 static const char *get_mode(const char *str, uint16_t *modep)
1317 unsigned char c;
1318 uint16_t mode = 0;
1320 while ((c = *str++) != ' ') {
1321 if (c < '0' || c > '7')
1322 return NULL;
1323 mode = (mode << 3) + (c - '0');
1325 *modep = mode;
1326 return str;
1329 static void load_tree(struct tree_entry *root)
1331 unsigned char *sha1 = root->versions[1].sha1;
1332 struct object_entry *myoe;
1333 struct tree_content *t;
1334 unsigned long size;
1335 char *buf;
1336 const char *c;
1338 root->tree = t = new_tree_content(8);
1339 if (is_null_sha1(sha1))
1340 return;
1342 myoe = find_object(sha1);
1343 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1344 if (myoe->type != OBJ_TREE)
1345 die("Not a tree: %s", sha1_to_hex(sha1));
1346 t->delta_depth = myoe->depth;
1347 buf = gfi_unpack_entry(myoe, &size);
1348 if (!buf)
1349 die("Can't load tree %s", sha1_to_hex(sha1));
1350 } else {
1351 enum object_type type;
1352 buf = read_sha1_file(sha1, &type, &size);
1353 if (!buf || type != OBJ_TREE)
1354 die("Can't load tree %s", sha1_to_hex(sha1));
1357 c = buf;
1358 while (c != (buf + size)) {
1359 struct tree_entry *e = new_tree_entry();
1361 if (t->entry_count == t->entry_capacity)
1362 root->tree = t = grow_tree_content(t, t->entry_count);
1363 t->entries[t->entry_count++] = e;
1365 e->tree = NULL;
1366 c = get_mode(c, &e->versions[1].mode);
1367 if (!c)
1368 die("Corrupt mode in %s", sha1_to_hex(sha1));
1369 e->versions[0].mode = e->versions[1].mode;
1370 e->name = to_atom(c, strlen(c));
1371 c += e->name->str_len + 1;
1372 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1373 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1374 c += 20;
1376 free(buf);
1379 static int tecmp0 (const void *_a, const void *_b)
1381 struct tree_entry *a = *((struct tree_entry**)_a);
1382 struct tree_entry *b = *((struct tree_entry**)_b);
1383 return base_name_compare(
1384 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1385 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1388 static int tecmp1 (const void *_a, const void *_b)
1390 struct tree_entry *a = *((struct tree_entry**)_a);
1391 struct tree_entry *b = *((struct tree_entry**)_b);
1392 return base_name_compare(
1393 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1394 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1397 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1399 size_t maxlen = 0;
1400 unsigned int i;
1402 if (!v)
1403 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1404 else
1405 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1407 for (i = 0; i < t->entry_count; i++) {
1408 if (t->entries[i]->versions[v].mode)
1409 maxlen += t->entries[i]->name->str_len + 34;
1412 strbuf_reset(b);
1413 strbuf_grow(b, maxlen);
1414 for (i = 0; i < t->entry_count; i++) {
1415 struct tree_entry *e = t->entries[i];
1416 if (!e->versions[v].mode)
1417 continue;
1418 strbuf_addf(b, "%o %s%c",
1419 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1420 e->name->str_dat, '\0');
1421 strbuf_add(b, e->versions[v].sha1, 20);
1425 static void store_tree(struct tree_entry *root)
1427 struct tree_content *t;
1428 unsigned int i, j, del;
1429 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1430 struct object_entry *le = NULL;
1432 if (!is_null_sha1(root->versions[1].sha1))
1433 return;
1435 if (!root->tree)
1436 load_tree(root);
1437 t = root->tree;
1439 for (i = 0; i < t->entry_count; i++) {
1440 if (t->entries[i]->tree)
1441 store_tree(t->entries[i]);
1444 if (!(root->versions[0].mode & NO_DELTA))
1445 le = find_object(root->versions[0].sha1);
1446 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1447 mktree(t, 0, &old_tree);
1448 lo.data = old_tree;
1449 lo.offset = le->idx.offset;
1450 lo.depth = t->delta_depth;
1453 mktree(t, 1, &new_tree);
1454 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1456 t->delta_depth = lo.depth;
1457 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1458 struct tree_entry *e = t->entries[i];
1459 if (e->versions[1].mode) {
1460 e->versions[0].mode = e->versions[1].mode;
1461 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1462 t->entries[j++] = e;
1463 } else {
1464 release_tree_entry(e);
1465 del++;
1468 t->entry_count -= del;
1471 static void tree_content_replace(
1472 struct tree_entry *root,
1473 const unsigned char *sha1,
1474 const uint16_t mode,
1475 struct tree_content *newtree)
1477 if (!S_ISDIR(mode))
1478 die("Root cannot be a non-directory");
1479 hashclr(root->versions[0].sha1);
1480 hashcpy(root->versions[1].sha1, sha1);
1481 if (root->tree)
1482 release_tree_content_recursive(root->tree);
1483 root->tree = newtree;
1486 static int tree_content_set(
1487 struct tree_entry *root,
1488 const char *p,
1489 const unsigned char *sha1,
1490 const uint16_t mode,
1491 struct tree_content *subtree)
1493 struct tree_content *t;
1494 const char *slash1;
1495 unsigned int i, n;
1496 struct tree_entry *e;
1498 slash1 = strchrnul(p, '/');
1499 n = slash1 - p;
1500 if (!n)
1501 die("Empty path component found in input");
1502 if (!*slash1 && !S_ISDIR(mode) && subtree)
1503 die("Non-directories cannot have subtrees");
1505 if (!root->tree)
1506 load_tree(root);
1507 t = root->tree;
1508 for (i = 0; i < t->entry_count; i++) {
1509 e = t->entries[i];
1510 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1511 if (!*slash1) {
1512 if (!S_ISDIR(mode)
1513 && e->versions[1].mode == mode
1514 && !hashcmp(e->versions[1].sha1, sha1))
1515 return 0;
1516 e->versions[1].mode = mode;
1517 hashcpy(e->versions[1].sha1, sha1);
1518 if (e->tree)
1519 release_tree_content_recursive(e->tree);
1520 e->tree = subtree;
1523 * We need to leave e->versions[0].sha1 alone
1524 * to avoid modifying the preimage tree used
1525 * when writing out the parent directory.
1526 * But after replacing the subdir with a
1527 * completely different one, it's not a good
1528 * delta base any more, and besides, we've
1529 * thrown away the tree entries needed to
1530 * make a delta against it.
1532 * So let's just explicitly disable deltas
1533 * for the subtree.
1535 if (S_ISDIR(e->versions[0].mode))
1536 e->versions[0].mode |= NO_DELTA;
1538 hashclr(root->versions[1].sha1);
1539 return 1;
1541 if (!S_ISDIR(e->versions[1].mode)) {
1542 e->tree = new_tree_content(8);
1543 e->versions[1].mode = S_IFDIR;
1545 if (!e->tree)
1546 load_tree(e);
1547 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1548 hashclr(root->versions[1].sha1);
1549 return 1;
1551 return 0;
1555 if (t->entry_count == t->entry_capacity)
1556 root->tree = t = grow_tree_content(t, t->entry_count);
1557 e = new_tree_entry();
1558 e->name = to_atom(p, n);
1559 e->versions[0].mode = 0;
1560 hashclr(e->versions[0].sha1);
1561 t->entries[t->entry_count++] = e;
1562 if (*slash1) {
1563 e->tree = new_tree_content(8);
1564 e->versions[1].mode = S_IFDIR;
1565 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1566 } else {
1567 e->tree = subtree;
1568 e->versions[1].mode = mode;
1569 hashcpy(e->versions[1].sha1, sha1);
1571 hashclr(root->versions[1].sha1);
1572 return 1;
1575 static int tree_content_remove(
1576 struct tree_entry *root,
1577 const char *p,
1578 struct tree_entry *backup_leaf,
1579 int allow_root)
1581 struct tree_content *t;
1582 const char *slash1;
1583 unsigned int i, n;
1584 struct tree_entry *e;
1586 slash1 = strchrnul(p, '/');
1587 n = slash1 - p;
1589 if (!root->tree)
1590 load_tree(root);
1592 if (!*p && allow_root) {
1593 e = root;
1594 goto del_entry;
1597 t = root->tree;
1598 for (i = 0; i < t->entry_count; i++) {
1599 e = t->entries[i];
1600 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1601 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1603 * If p names a file in some subdirectory, and a
1604 * file or symlink matching the name of the
1605 * parent directory of p exists, then p cannot
1606 * exist and need not be deleted.
1608 return 1;
1609 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1610 goto del_entry;
1611 if (!e->tree)
1612 load_tree(e);
1613 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1614 for (n = 0; n < e->tree->entry_count; n++) {
1615 if (e->tree->entries[n]->versions[1].mode) {
1616 hashclr(root->versions[1].sha1);
1617 return 1;
1620 backup_leaf = NULL;
1621 goto del_entry;
1623 return 0;
1626 return 0;
1628 del_entry:
1629 if (backup_leaf)
1630 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1631 else if (e->tree)
1632 release_tree_content_recursive(e->tree);
1633 e->tree = NULL;
1634 e->versions[1].mode = 0;
1635 hashclr(e->versions[1].sha1);
1636 hashclr(root->versions[1].sha1);
1637 return 1;
1640 static int tree_content_get(
1641 struct tree_entry *root,
1642 const char *p,
1643 struct tree_entry *leaf,
1644 int allow_root)
1646 struct tree_content *t;
1647 const char *slash1;
1648 unsigned int i, n;
1649 struct tree_entry *e;
1651 slash1 = strchrnul(p, '/');
1652 n = slash1 - p;
1653 if (!n && !allow_root)
1654 die("Empty path component found in input");
1656 if (!root->tree)
1657 load_tree(root);
1659 if (!n) {
1660 e = root;
1661 goto found_entry;
1664 t = root->tree;
1665 for (i = 0; i < t->entry_count; i++) {
1666 e = t->entries[i];
1667 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1668 if (!*slash1)
1669 goto found_entry;
1670 if (!S_ISDIR(e->versions[1].mode))
1671 return 0;
1672 if (!e->tree)
1673 load_tree(e);
1674 return tree_content_get(e, slash1 + 1, leaf, 0);
1677 return 0;
1679 found_entry:
1680 memcpy(leaf, e, sizeof(*leaf));
1681 if (e->tree && is_null_sha1(e->versions[1].sha1))
1682 leaf->tree = dup_tree_content(e->tree);
1683 else
1684 leaf->tree = NULL;
1685 return 1;
1688 static int update_branch(struct branch *b)
1690 static const char *msg = "fast-import";
1691 struct ref_transaction *transaction;
1692 unsigned char old_sha1[20];
1693 struct strbuf err = STRBUF_INIT;
1695 if (read_ref(b->name, old_sha1))
1696 hashclr(old_sha1);
1697 if (is_null_sha1(b->sha1)) {
1698 if (b->delete)
1699 delete_ref(b->name, old_sha1, 0);
1700 return 0;
1702 if (!force_update && !is_null_sha1(old_sha1)) {
1703 struct commit *old_cmit, *new_cmit;
1705 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1706 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1707 if (!old_cmit || !new_cmit)
1708 return error("Branch %s is missing commits.", b->name);
1710 if (!in_merge_bases(old_cmit, new_cmit)) {
1711 warning("Not updating %s"
1712 " (new tip %s does not contain %s)",
1713 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1714 return -1;
1717 transaction = ref_transaction_begin(&err);
1718 if (!transaction ||
1719 ref_transaction_update(transaction, b->name, b->sha1, old_sha1,
1720 0, msg, &err) ||
1721 ref_transaction_commit(transaction, &err)) {
1722 ref_transaction_free(transaction);
1723 error("%s", err.buf);
1724 strbuf_release(&err);
1725 return -1;
1727 ref_transaction_free(transaction);
1728 strbuf_release(&err);
1729 return 0;
1732 static void dump_branches(void)
1734 unsigned int i;
1735 struct branch *b;
1737 for (i = 0; i < branch_table_sz; i++) {
1738 for (b = branch_table[i]; b; b = b->table_next_branch)
1739 failure |= update_branch(b);
1743 static void dump_tags(void)
1745 static const char *msg = "fast-import";
1746 struct tag *t;
1747 struct strbuf ref_name = STRBUF_INIT;
1748 struct strbuf err = STRBUF_INIT;
1749 struct ref_transaction *transaction;
1751 transaction = ref_transaction_begin(&err);
1752 if (!transaction) {
1753 failure |= error("%s", err.buf);
1754 goto cleanup;
1756 for (t = first_tag; t; t = t->next_tag) {
1757 strbuf_reset(&ref_name);
1758 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1760 if (ref_transaction_update(transaction, ref_name.buf,
1761 t->sha1, NULL, 0, msg, &err)) {
1762 failure |= error("%s", err.buf);
1763 goto cleanup;
1766 if (ref_transaction_commit(transaction, &err))
1767 failure |= error("%s", err.buf);
1769 cleanup:
1770 ref_transaction_free(transaction);
1771 strbuf_release(&ref_name);
1772 strbuf_release(&err);
1775 static void dump_marks_helper(FILE *f,
1776 uintmax_t base,
1777 struct mark_set *m)
1779 uintmax_t k;
1780 if (m->shift) {
1781 for (k = 0; k < 1024; k++) {
1782 if (m->data.sets[k])
1783 dump_marks_helper(f, base + (k << m->shift),
1784 m->data.sets[k]);
1786 } else {
1787 for (k = 0; k < 1024; k++) {
1788 if (m->data.marked[k])
1789 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1790 sha1_to_hex(m->data.marked[k]->idx.sha1));
1795 static void dump_marks(void)
1797 static struct lock_file mark_lock;
1798 FILE *f;
1800 if (!export_marks_file)
1801 return;
1803 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1804 failure |= error("Unable to write marks file %s: %s",
1805 export_marks_file, strerror(errno));
1806 return;
1809 f = fdopen_lock_file(&mark_lock, "w");
1810 if (!f) {
1811 int saved_errno = errno;
1812 rollback_lock_file(&mark_lock);
1813 failure |= error("Unable to write marks file %s: %s",
1814 export_marks_file, strerror(saved_errno));
1815 return;
1818 dump_marks_helper(f, 0, marks);
1819 if (commit_lock_file(&mark_lock)) {
1820 failure |= error("Unable to commit marks file %s: %s",
1821 export_marks_file, strerror(errno));
1822 return;
1826 static void read_marks(void)
1828 char line[512];
1829 FILE *f = fopen(import_marks_file, "r");
1830 if (f)
1832 else if (import_marks_file_ignore_missing && errno == ENOENT)
1833 return; /* Marks file does not exist */
1834 else
1835 die_errno("cannot read '%s'", import_marks_file);
1836 while (fgets(line, sizeof(line), f)) {
1837 uintmax_t mark;
1838 char *end;
1839 unsigned char sha1[20];
1840 struct object_entry *e;
1842 end = strchr(line, '\n');
1843 if (line[0] != ':' || !end)
1844 die("corrupt mark line: %s", line);
1845 *end = 0;
1846 mark = strtoumax(line + 1, &end, 10);
1847 if (!mark || end == line + 1
1848 || *end != ' ' || get_sha1_hex(end + 1, sha1))
1849 die("corrupt mark line: %s", line);
1850 e = find_object(sha1);
1851 if (!e) {
1852 enum object_type type = sha1_object_info(sha1, NULL);
1853 if (type < 0)
1854 die("object not found: %s", sha1_to_hex(sha1));
1855 e = insert_object(sha1);
1856 e->type = type;
1857 e->pack_id = MAX_PACK_ID;
1858 e->idx.offset = 1; /* just not zero! */
1860 insert_mark(mark, e);
1862 fclose(f);
1866 static int read_next_command(void)
1868 static int stdin_eof = 0;
1870 if (stdin_eof) {
1871 unread_command_buf = 0;
1872 return EOF;
1875 for (;;) {
1876 const char *p;
1878 if (unread_command_buf) {
1879 unread_command_buf = 0;
1880 } else {
1881 struct recent_command *rc;
1883 strbuf_detach(&command_buf, NULL);
1884 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1885 if (stdin_eof)
1886 return EOF;
1888 if (!seen_data_command
1889 && !starts_with(command_buf.buf, "feature ")
1890 && !starts_with(command_buf.buf, "option ")) {
1891 parse_argv();
1894 rc = rc_free;
1895 if (rc)
1896 rc_free = rc->next;
1897 else {
1898 rc = cmd_hist.next;
1899 cmd_hist.next = rc->next;
1900 cmd_hist.next->prev = &cmd_hist;
1901 free(rc->buf);
1904 rc->buf = command_buf.buf;
1905 rc->prev = cmd_tail;
1906 rc->next = cmd_hist.prev;
1907 rc->prev->next = rc;
1908 cmd_tail = rc;
1910 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1911 parse_cat_blob(p);
1912 continue;
1914 if (command_buf.buf[0] == '#')
1915 continue;
1916 return 0;
1920 static void skip_optional_lf(void)
1922 int term_char = fgetc(stdin);
1923 if (term_char != '\n' && term_char != EOF)
1924 ungetc(term_char, stdin);
1927 static void parse_mark(void)
1929 const char *v;
1930 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1931 next_mark = strtoumax(v, NULL, 10);
1932 read_next_command();
1934 else
1935 next_mark = 0;
1938 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1940 const char *data;
1941 strbuf_reset(sb);
1943 if (!skip_prefix(command_buf.buf, "data ", &data))
1944 die("Expected 'data n' command, found: %s", command_buf.buf);
1946 if (skip_prefix(data, "<<", &data)) {
1947 char *term = xstrdup(data);
1948 size_t term_len = command_buf.len - (data - command_buf.buf);
1950 strbuf_detach(&command_buf, NULL);
1951 for (;;) {
1952 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1953 die("EOF in data (terminator '%s' not found)", term);
1954 if (term_len == command_buf.len
1955 && !strcmp(term, command_buf.buf))
1956 break;
1957 strbuf_addbuf(sb, &command_buf);
1958 strbuf_addch(sb, '\n');
1960 free(term);
1962 else {
1963 uintmax_t len = strtoumax(data, NULL, 10);
1964 size_t n = 0, length = (size_t)len;
1966 if (limit && limit < len) {
1967 *len_res = len;
1968 return 0;
1970 if (length < len)
1971 die("data is too large to use in this context");
1973 while (n < length) {
1974 size_t s = strbuf_fread(sb, length - n, stdin);
1975 if (!s && feof(stdin))
1976 die("EOF in data (%lu bytes remaining)",
1977 (unsigned long)(length - n));
1978 n += s;
1982 skip_optional_lf();
1983 return 1;
1986 static int validate_raw_date(const char *src, struct strbuf *result)
1988 const char *orig_src = src;
1989 char *endp;
1990 unsigned long num;
1992 errno = 0;
1994 num = strtoul(src, &endp, 10);
1995 /* NEEDSWORK: perhaps check for reasonable values? */
1996 if (errno || endp == src || *endp != ' ')
1997 return -1;
1999 src = endp + 1;
2000 if (*src != '-' && *src != '+')
2001 return -1;
2003 num = strtoul(src + 1, &endp, 10);
2004 if (errno || endp == src + 1 || *endp || 1400 < num)
2005 return -1;
2007 strbuf_addstr(result, orig_src);
2008 return 0;
2011 static char *parse_ident(const char *buf)
2013 const char *ltgt;
2014 size_t name_len;
2015 struct strbuf ident = STRBUF_INIT;
2017 /* ensure there is a space delimiter even if there is no name */
2018 if (*buf == '<')
2019 --buf;
2021 ltgt = buf + strcspn(buf, "<>");
2022 if (*ltgt != '<')
2023 die("Missing < in ident string: %s", buf);
2024 if (ltgt != buf && ltgt[-1] != ' ')
2025 die("Missing space before < in ident string: %s", buf);
2026 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2027 if (*ltgt != '>')
2028 die("Missing > in ident string: %s", buf);
2029 ltgt++;
2030 if (*ltgt != ' ')
2031 die("Missing space after > in ident string: %s", buf);
2032 ltgt++;
2033 name_len = ltgt - buf;
2034 strbuf_add(&ident, buf, name_len);
2036 switch (whenspec) {
2037 case WHENSPEC_RAW:
2038 if (validate_raw_date(ltgt, &ident) < 0)
2039 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2040 break;
2041 case WHENSPEC_RFC2822:
2042 if (parse_date(ltgt, &ident) < 0)
2043 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2044 break;
2045 case WHENSPEC_NOW:
2046 if (strcmp("now", ltgt))
2047 die("Date in ident must be 'now': %s", buf);
2048 datestamp(&ident);
2049 break;
2052 return strbuf_detach(&ident, NULL);
2055 static void parse_and_store_blob(
2056 struct last_object *last,
2057 unsigned char *sha1out,
2058 uintmax_t mark)
2060 static struct strbuf buf = STRBUF_INIT;
2061 uintmax_t len;
2063 if (parse_data(&buf, big_file_threshold, &len))
2064 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2065 else {
2066 if (last) {
2067 strbuf_release(&last->data);
2068 last->offset = 0;
2069 last->depth = 0;
2071 stream_blob(len, sha1out, mark);
2072 skip_optional_lf();
2076 static void parse_new_blob(void)
2078 read_next_command();
2079 parse_mark();
2080 parse_and_store_blob(&last_blob, NULL, next_mark);
2083 static void unload_one_branch(void)
2085 while (cur_active_branches
2086 && cur_active_branches >= max_active_branches) {
2087 uintmax_t min_commit = ULONG_MAX;
2088 struct branch *e, *l = NULL, *p = NULL;
2090 for (e = active_branches; e; e = e->active_next_branch) {
2091 if (e->last_commit < min_commit) {
2092 p = l;
2093 min_commit = e->last_commit;
2095 l = e;
2098 if (p) {
2099 e = p->active_next_branch;
2100 p->active_next_branch = e->active_next_branch;
2101 } else {
2102 e = active_branches;
2103 active_branches = e->active_next_branch;
2105 e->active = 0;
2106 e->active_next_branch = NULL;
2107 if (e->branch_tree.tree) {
2108 release_tree_content_recursive(e->branch_tree.tree);
2109 e->branch_tree.tree = NULL;
2111 cur_active_branches--;
2115 static void load_branch(struct branch *b)
2117 load_tree(&b->branch_tree);
2118 if (!b->active) {
2119 b->active = 1;
2120 b->active_next_branch = active_branches;
2121 active_branches = b;
2122 cur_active_branches++;
2123 branch_load_count++;
2127 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2129 unsigned char fanout = 0;
2130 while ((num_notes >>= 8))
2131 fanout++;
2132 return fanout;
2135 static void construct_path_with_fanout(const char *hex_sha1,
2136 unsigned char fanout, char *path)
2138 unsigned int i = 0, j = 0;
2139 if (fanout >= 20)
2140 die("Too large fanout (%u)", fanout);
2141 while (fanout) {
2142 path[i++] = hex_sha1[j++];
2143 path[i++] = hex_sha1[j++];
2144 path[i++] = '/';
2145 fanout--;
2147 memcpy(path + i, hex_sha1 + j, 40 - j);
2148 path[i + 40 - j] = '\0';
2151 static uintmax_t do_change_note_fanout(
2152 struct tree_entry *orig_root, struct tree_entry *root,
2153 char *hex_sha1, unsigned int hex_sha1_len,
2154 char *fullpath, unsigned int fullpath_len,
2155 unsigned char fanout)
2157 struct tree_content *t = root->tree;
2158 struct tree_entry *e, leaf;
2159 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2160 uintmax_t num_notes = 0;
2161 unsigned char sha1[20];
2162 char realpath[60];
2164 for (i = 0; t && i < t->entry_count; i++) {
2165 e = t->entries[i];
2166 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2167 tmp_fullpath_len = fullpath_len;
2170 * We're interested in EITHER existing note entries (entries
2171 * with exactly 40 hex chars in path, not including directory
2172 * separators), OR directory entries that may contain note
2173 * entries (with < 40 hex chars in path).
2174 * Also, each path component in a note entry must be a multiple
2175 * of 2 chars.
2177 if (!e->versions[1].mode ||
2178 tmp_hex_sha1_len > 40 ||
2179 e->name->str_len % 2)
2180 continue;
2182 /* This _may_ be a note entry, or a subdir containing notes */
2183 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2184 e->name->str_len);
2185 if (tmp_fullpath_len)
2186 fullpath[tmp_fullpath_len++] = '/';
2187 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2188 e->name->str_len);
2189 tmp_fullpath_len += e->name->str_len;
2190 fullpath[tmp_fullpath_len] = '\0';
2192 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2193 /* This is a note entry */
2194 if (fanout == 0xff) {
2195 /* Counting mode, no rename */
2196 num_notes++;
2197 continue;
2199 construct_path_with_fanout(hex_sha1, fanout, realpath);
2200 if (!strcmp(fullpath, realpath)) {
2201 /* Note entry is in correct location */
2202 num_notes++;
2203 continue;
2206 /* Rename fullpath to realpath */
2207 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2208 die("Failed to remove path %s", fullpath);
2209 tree_content_set(orig_root, realpath,
2210 leaf.versions[1].sha1,
2211 leaf.versions[1].mode,
2212 leaf.tree);
2213 } else if (S_ISDIR(e->versions[1].mode)) {
2214 /* This is a subdir that may contain note entries */
2215 if (!e->tree)
2216 load_tree(e);
2217 num_notes += do_change_note_fanout(orig_root, e,
2218 hex_sha1, tmp_hex_sha1_len,
2219 fullpath, tmp_fullpath_len, fanout);
2222 /* The above may have reallocated the current tree_content */
2223 t = root->tree;
2225 return num_notes;
2228 static uintmax_t change_note_fanout(struct tree_entry *root,
2229 unsigned char fanout)
2231 char hex_sha1[40], path[60];
2232 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2236 * Given a pointer into a string, parse a mark reference:
2238 * idnum ::= ':' bigint;
2240 * Return the first character after the value in *endptr.
2242 * Complain if the following character is not what is expected,
2243 * either a space or end of the string.
2245 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2247 uintmax_t mark;
2249 assert(*p == ':');
2250 p++;
2251 mark = strtoumax(p, endptr, 10);
2252 if (*endptr == p)
2253 die("No value after ':' in mark: %s", command_buf.buf);
2254 return mark;
2258 * Parse the mark reference, and complain if this is not the end of
2259 * the string.
2261 static uintmax_t parse_mark_ref_eol(const char *p)
2263 char *end;
2264 uintmax_t mark;
2266 mark = parse_mark_ref(p, &end);
2267 if (*end != '\0')
2268 die("Garbage after mark: %s", command_buf.buf);
2269 return mark;
2273 * Parse the mark reference, demanding a trailing space. Return a
2274 * pointer to the space.
2276 static uintmax_t parse_mark_ref_space(const char **p)
2278 uintmax_t mark;
2279 char *end;
2281 mark = parse_mark_ref(*p, &end);
2282 if (*end++ != ' ')
2283 die("Missing space after mark: %s", command_buf.buf);
2284 *p = end;
2285 return mark;
2288 static void file_change_m(const char *p, struct branch *b)
2290 static struct strbuf uq = STRBUF_INIT;
2291 const char *endp;
2292 struct object_entry *oe;
2293 unsigned char sha1[20];
2294 uint16_t mode, inline_data = 0;
2296 p = get_mode(p, &mode);
2297 if (!p)
2298 die("Corrupt mode: %s", command_buf.buf);
2299 switch (mode) {
2300 case 0644:
2301 case 0755:
2302 mode |= S_IFREG;
2303 case S_IFREG | 0644:
2304 case S_IFREG | 0755:
2305 case S_IFLNK:
2306 case S_IFDIR:
2307 case S_IFGITLINK:
2308 /* ok */
2309 break;
2310 default:
2311 die("Corrupt mode: %s", command_buf.buf);
2314 if (*p == ':') {
2315 oe = find_mark(parse_mark_ref_space(&p));
2316 hashcpy(sha1, oe->idx.sha1);
2317 } else if (skip_prefix(p, "inline ", &p)) {
2318 inline_data = 1;
2319 oe = NULL; /* not used with inline_data, but makes gcc happy */
2320 } else {
2321 if (get_sha1_hex(p, sha1))
2322 die("Invalid dataref: %s", command_buf.buf);
2323 oe = find_object(sha1);
2324 p += 40;
2325 if (*p++ != ' ')
2326 die("Missing space after SHA1: %s", command_buf.buf);
2329 strbuf_reset(&uq);
2330 if (!unquote_c_style(&uq, p, &endp)) {
2331 if (*endp)
2332 die("Garbage after path in: %s", command_buf.buf);
2333 p = uq.buf;
2336 /* Git does not track empty, non-toplevel directories. */
2337 if (S_ISDIR(mode) && !hashcmp(sha1, EMPTY_TREE_SHA1_BIN) && *p) {
2338 tree_content_remove(&b->branch_tree, p, NULL, 0);
2339 return;
2342 if (S_ISGITLINK(mode)) {
2343 if (inline_data)
2344 die("Git links cannot be specified 'inline': %s",
2345 command_buf.buf);
2346 else if (oe) {
2347 if (oe->type != OBJ_COMMIT)
2348 die("Not a commit (actually a %s): %s",
2349 typename(oe->type), command_buf.buf);
2352 * Accept the sha1 without checking; it expected to be in
2353 * another repository.
2355 } else if (inline_data) {
2356 if (S_ISDIR(mode))
2357 die("Directories cannot be specified 'inline': %s",
2358 command_buf.buf);
2359 if (p != uq.buf) {
2360 strbuf_addstr(&uq, p);
2361 p = uq.buf;
2363 read_next_command();
2364 parse_and_store_blob(&last_blob, sha1, 0);
2365 } else {
2366 enum object_type expected = S_ISDIR(mode) ?
2367 OBJ_TREE: OBJ_BLOB;
2368 enum object_type type = oe ? oe->type :
2369 sha1_object_info(sha1, NULL);
2370 if (type < 0)
2371 die("%s not found: %s",
2372 S_ISDIR(mode) ? "Tree" : "Blob",
2373 command_buf.buf);
2374 if (type != expected)
2375 die("Not a %s (actually a %s): %s",
2376 typename(expected), typename(type),
2377 command_buf.buf);
2380 if (!*p) {
2381 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2382 return;
2384 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2387 static void file_change_d(const char *p, struct branch *b)
2389 static struct strbuf uq = STRBUF_INIT;
2390 const char *endp;
2392 strbuf_reset(&uq);
2393 if (!unquote_c_style(&uq, p, &endp)) {
2394 if (*endp)
2395 die("Garbage after path in: %s", command_buf.buf);
2396 p = uq.buf;
2398 tree_content_remove(&b->branch_tree, p, NULL, 1);
2401 static void file_change_cr(const char *s, struct branch *b, int rename)
2403 const char *d;
2404 static struct strbuf s_uq = STRBUF_INIT;
2405 static struct strbuf d_uq = STRBUF_INIT;
2406 const char *endp;
2407 struct tree_entry leaf;
2409 strbuf_reset(&s_uq);
2410 if (!unquote_c_style(&s_uq, s, &endp)) {
2411 if (*endp != ' ')
2412 die("Missing space after source: %s", command_buf.buf);
2413 } else {
2414 endp = strchr(s, ' ');
2415 if (!endp)
2416 die("Missing space after source: %s", command_buf.buf);
2417 strbuf_add(&s_uq, s, endp - s);
2419 s = s_uq.buf;
2421 endp++;
2422 if (!*endp)
2423 die("Missing dest: %s", command_buf.buf);
2425 d = endp;
2426 strbuf_reset(&d_uq);
2427 if (!unquote_c_style(&d_uq, d, &endp)) {
2428 if (*endp)
2429 die("Garbage after dest in: %s", command_buf.buf);
2430 d = d_uq.buf;
2433 memset(&leaf, 0, sizeof(leaf));
2434 if (rename)
2435 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2436 else
2437 tree_content_get(&b->branch_tree, s, &leaf, 1);
2438 if (!leaf.versions[1].mode)
2439 die("Path %s not in branch", s);
2440 if (!*d) { /* C "path/to/subdir" "" */
2441 tree_content_replace(&b->branch_tree,
2442 leaf.versions[1].sha1,
2443 leaf.versions[1].mode,
2444 leaf.tree);
2445 return;
2447 tree_content_set(&b->branch_tree, d,
2448 leaf.versions[1].sha1,
2449 leaf.versions[1].mode,
2450 leaf.tree);
2453 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2455 static struct strbuf uq = STRBUF_INIT;
2456 struct object_entry *oe;
2457 struct branch *s;
2458 unsigned char sha1[20], commit_sha1[20];
2459 char path[60];
2460 uint16_t inline_data = 0;
2461 unsigned char new_fanout;
2464 * When loading a branch, we don't traverse its tree to count the real
2465 * number of notes (too expensive to do this for all non-note refs).
2466 * This means that recently loaded notes refs might incorrectly have
2467 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2469 * Fix this by traversing the tree and counting the number of notes
2470 * when b->num_notes == 0. If the notes tree is truly empty, the
2471 * calculation should not take long.
2473 if (b->num_notes == 0 && *old_fanout == 0) {
2474 /* Invoke change_note_fanout() in "counting mode". */
2475 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2476 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2479 /* Now parse the notemodify command. */
2480 /* <dataref> or 'inline' */
2481 if (*p == ':') {
2482 oe = find_mark(parse_mark_ref_space(&p));
2483 hashcpy(sha1, oe->idx.sha1);
2484 } else if (skip_prefix(p, "inline ", &p)) {
2485 inline_data = 1;
2486 oe = NULL; /* not used with inline_data, but makes gcc happy */
2487 } else {
2488 if (get_sha1_hex(p, sha1))
2489 die("Invalid dataref: %s", command_buf.buf);
2490 oe = find_object(sha1);
2491 p += 40;
2492 if (*p++ != ' ')
2493 die("Missing space after SHA1: %s", command_buf.buf);
2496 /* <commit-ish> */
2497 s = lookup_branch(p);
2498 if (s) {
2499 if (is_null_sha1(s->sha1))
2500 die("Can't add a note on empty branch.");
2501 hashcpy(commit_sha1, s->sha1);
2502 } else if (*p == ':') {
2503 uintmax_t commit_mark = parse_mark_ref_eol(p);
2504 struct object_entry *commit_oe = find_mark(commit_mark);
2505 if (commit_oe->type != OBJ_COMMIT)
2506 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2507 hashcpy(commit_sha1, commit_oe->idx.sha1);
2508 } else if (!get_sha1(p, commit_sha1)) {
2509 unsigned long size;
2510 char *buf = read_object_with_reference(commit_sha1,
2511 commit_type, &size, commit_sha1);
2512 if (!buf || size < 46)
2513 die("Not a valid commit: %s", p);
2514 free(buf);
2515 } else
2516 die("Invalid ref name or SHA1 expression: %s", p);
2518 if (inline_data) {
2519 if (p != uq.buf) {
2520 strbuf_addstr(&uq, p);
2521 p = uq.buf;
2523 read_next_command();
2524 parse_and_store_blob(&last_blob, sha1, 0);
2525 } else if (oe) {
2526 if (oe->type != OBJ_BLOB)
2527 die("Not a blob (actually a %s): %s",
2528 typename(oe->type), command_buf.buf);
2529 } else if (!is_null_sha1(sha1)) {
2530 enum object_type type = sha1_object_info(sha1, NULL);
2531 if (type < 0)
2532 die("Blob not found: %s", command_buf.buf);
2533 if (type != OBJ_BLOB)
2534 die("Not a blob (actually a %s): %s",
2535 typename(type), command_buf.buf);
2538 construct_path_with_fanout(sha1_to_hex(commit_sha1), *old_fanout, path);
2539 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2540 b->num_notes--;
2542 if (is_null_sha1(sha1))
2543 return; /* nothing to insert */
2545 b->num_notes++;
2546 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2547 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2548 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2551 static void file_change_deleteall(struct branch *b)
2553 release_tree_content_recursive(b->branch_tree.tree);
2554 hashclr(b->branch_tree.versions[0].sha1);
2555 hashclr(b->branch_tree.versions[1].sha1);
2556 load_tree(&b->branch_tree);
2557 b->num_notes = 0;
2560 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2562 if (!buf || size < 46)
2563 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2564 if (memcmp("tree ", buf, 5)
2565 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2566 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2567 hashcpy(b->branch_tree.versions[0].sha1,
2568 b->branch_tree.versions[1].sha1);
2571 static void parse_from_existing(struct branch *b)
2573 if (is_null_sha1(b->sha1)) {
2574 hashclr(b->branch_tree.versions[0].sha1);
2575 hashclr(b->branch_tree.versions[1].sha1);
2576 } else {
2577 unsigned long size;
2578 char *buf;
2580 buf = read_object_with_reference(b->sha1,
2581 commit_type, &size, b->sha1);
2582 parse_from_commit(b, buf, size);
2583 free(buf);
2587 static int parse_from(struct branch *b)
2589 const char *from;
2590 struct branch *s;
2591 unsigned char sha1[20];
2593 if (!skip_prefix(command_buf.buf, "from ", &from))
2594 return 0;
2596 hashcpy(sha1, b->branch_tree.versions[1].sha1);
2598 s = lookup_branch(from);
2599 if (b == s)
2600 die("Can't create a branch from itself: %s", b->name);
2601 else if (s) {
2602 unsigned char *t = s->branch_tree.versions[1].sha1;
2603 hashcpy(b->sha1, s->sha1);
2604 hashcpy(b->branch_tree.versions[0].sha1, t);
2605 hashcpy(b->branch_tree.versions[1].sha1, t);
2606 } else if (*from == ':') {
2607 uintmax_t idnum = parse_mark_ref_eol(from);
2608 struct object_entry *oe = find_mark(idnum);
2609 if (oe->type != OBJ_COMMIT)
2610 die("Mark :%" PRIuMAX " not a commit", idnum);
2611 if (hashcmp(b->sha1, oe->idx.sha1)) {
2612 hashcpy(b->sha1, oe->idx.sha1);
2613 if (oe->pack_id != MAX_PACK_ID) {
2614 unsigned long size;
2615 char *buf = gfi_unpack_entry(oe, &size);
2616 parse_from_commit(b, buf, size);
2617 free(buf);
2618 } else
2619 parse_from_existing(b);
2621 } else if (!get_sha1(from, b->sha1)) {
2622 parse_from_existing(b);
2623 if (is_null_sha1(b->sha1))
2624 b->delete = 1;
2626 else
2627 die("Invalid ref name or SHA1 expression: %s", from);
2629 if (b->branch_tree.tree && hashcmp(sha1, b->branch_tree.versions[1].sha1)) {
2630 release_tree_content_recursive(b->branch_tree.tree);
2631 b->branch_tree.tree = NULL;
2634 read_next_command();
2635 return 1;
2638 static struct hash_list *parse_merge(unsigned int *count)
2640 struct hash_list *list = NULL, **tail = &list, *n;
2641 const char *from;
2642 struct branch *s;
2644 *count = 0;
2645 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2646 n = xmalloc(sizeof(*n));
2647 s = lookup_branch(from);
2648 if (s)
2649 hashcpy(n->sha1, s->sha1);
2650 else if (*from == ':') {
2651 uintmax_t idnum = parse_mark_ref_eol(from);
2652 struct object_entry *oe = find_mark(idnum);
2653 if (oe->type != OBJ_COMMIT)
2654 die("Mark :%" PRIuMAX " not a commit", idnum);
2655 hashcpy(n->sha1, oe->idx.sha1);
2656 } else if (!get_sha1(from, n->sha1)) {
2657 unsigned long size;
2658 char *buf = read_object_with_reference(n->sha1,
2659 commit_type, &size, n->sha1);
2660 if (!buf || size < 46)
2661 die("Not a valid commit: %s", from);
2662 free(buf);
2663 } else
2664 die("Invalid ref name or SHA1 expression: %s", from);
2666 n->next = NULL;
2667 *tail = n;
2668 tail = &n->next;
2670 (*count)++;
2671 read_next_command();
2673 return list;
2676 static void parse_new_commit(const char *arg)
2678 static struct strbuf msg = STRBUF_INIT;
2679 struct branch *b;
2680 char *author = NULL;
2681 char *committer = NULL;
2682 struct hash_list *merge_list = NULL;
2683 unsigned int merge_count;
2684 unsigned char prev_fanout, new_fanout;
2685 const char *v;
2687 b = lookup_branch(arg);
2688 if (!b)
2689 b = new_branch(arg);
2691 read_next_command();
2692 parse_mark();
2693 if (skip_prefix(command_buf.buf, "author ", &v)) {
2694 author = parse_ident(v);
2695 read_next_command();
2697 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2698 committer = parse_ident(v);
2699 read_next_command();
2701 if (!committer)
2702 die("Expected committer but didn't get one");
2703 parse_data(&msg, 0, NULL);
2704 read_next_command();
2705 parse_from(b);
2706 merge_list = parse_merge(&merge_count);
2708 /* ensure the branch is active/loaded */
2709 if (!b->branch_tree.tree || !max_active_branches) {
2710 unload_one_branch();
2711 load_branch(b);
2714 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2716 /* file_change* */
2717 while (command_buf.len > 0) {
2718 if (skip_prefix(command_buf.buf, "M ", &v))
2719 file_change_m(v, b);
2720 else if (skip_prefix(command_buf.buf, "D ", &v))
2721 file_change_d(v, b);
2722 else if (skip_prefix(command_buf.buf, "R ", &v))
2723 file_change_cr(v, b, 1);
2724 else if (skip_prefix(command_buf.buf, "C ", &v))
2725 file_change_cr(v, b, 0);
2726 else if (skip_prefix(command_buf.buf, "N ", &v))
2727 note_change_n(v, b, &prev_fanout);
2728 else if (!strcmp("deleteall", command_buf.buf))
2729 file_change_deleteall(b);
2730 else if (skip_prefix(command_buf.buf, "ls ", &v))
2731 parse_ls(v, b);
2732 else {
2733 unread_command_buf = 1;
2734 break;
2736 if (read_next_command() == EOF)
2737 break;
2740 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2741 if (new_fanout != prev_fanout)
2742 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2744 /* build the tree and the commit */
2745 store_tree(&b->branch_tree);
2746 hashcpy(b->branch_tree.versions[0].sha1,
2747 b->branch_tree.versions[1].sha1);
2749 strbuf_reset(&new_data);
2750 strbuf_addf(&new_data, "tree %s\n",
2751 sha1_to_hex(b->branch_tree.versions[1].sha1));
2752 if (!is_null_sha1(b->sha1))
2753 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2754 while (merge_list) {
2755 struct hash_list *next = merge_list->next;
2756 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2757 free(merge_list);
2758 merge_list = next;
2760 strbuf_addf(&new_data,
2761 "author %s\n"
2762 "committer %s\n"
2763 "\n",
2764 author ? author : committer, committer);
2765 strbuf_addbuf(&new_data, &msg);
2766 free(author);
2767 free(committer);
2769 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2770 b->pack_id = pack_id;
2771 b->last_commit = object_count_by_type[OBJ_COMMIT];
2774 static void parse_new_tag(const char *arg)
2776 static struct strbuf msg = STRBUF_INIT;
2777 const char *from;
2778 char *tagger;
2779 struct branch *s;
2780 struct tag *t;
2781 uintmax_t from_mark = 0;
2782 unsigned char sha1[20];
2783 enum object_type type;
2784 const char *v;
2786 t = pool_alloc(sizeof(struct tag));
2787 memset(t, 0, sizeof(struct tag));
2788 t->name = pool_strdup(arg);
2789 if (last_tag)
2790 last_tag->next_tag = t;
2791 else
2792 first_tag = t;
2793 last_tag = t;
2794 read_next_command();
2796 /* from ... */
2797 if (!skip_prefix(command_buf.buf, "from ", &from))
2798 die("Expected from command, got %s", command_buf.buf);
2799 s = lookup_branch(from);
2800 if (s) {
2801 if (is_null_sha1(s->sha1))
2802 die("Can't tag an empty branch.");
2803 hashcpy(sha1, s->sha1);
2804 type = OBJ_COMMIT;
2805 } else if (*from == ':') {
2806 struct object_entry *oe;
2807 from_mark = parse_mark_ref_eol(from);
2808 oe = find_mark(from_mark);
2809 type = oe->type;
2810 hashcpy(sha1, oe->idx.sha1);
2811 } else if (!get_sha1(from, sha1)) {
2812 struct object_entry *oe = find_object(sha1);
2813 if (!oe) {
2814 type = sha1_object_info(sha1, NULL);
2815 if (type < 0)
2816 die("Not a valid object: %s", from);
2817 } else
2818 type = oe->type;
2819 } else
2820 die("Invalid ref name or SHA1 expression: %s", from);
2821 read_next_command();
2823 /* tagger ... */
2824 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2825 tagger = parse_ident(v);
2826 read_next_command();
2827 } else
2828 tagger = NULL;
2830 /* tag payload/message */
2831 parse_data(&msg, 0, NULL);
2833 /* build the tag object */
2834 strbuf_reset(&new_data);
2836 strbuf_addf(&new_data,
2837 "object %s\n"
2838 "type %s\n"
2839 "tag %s\n",
2840 sha1_to_hex(sha1), typename(type), t->name);
2841 if (tagger)
2842 strbuf_addf(&new_data,
2843 "tagger %s\n", tagger);
2844 strbuf_addch(&new_data, '\n');
2845 strbuf_addbuf(&new_data, &msg);
2846 free(tagger);
2848 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2849 t->pack_id = MAX_PACK_ID;
2850 else
2851 t->pack_id = pack_id;
2854 static void parse_reset_branch(const char *arg)
2856 struct branch *b;
2858 b = lookup_branch(arg);
2859 if (b) {
2860 hashclr(b->sha1);
2861 hashclr(b->branch_tree.versions[0].sha1);
2862 hashclr(b->branch_tree.versions[1].sha1);
2863 if (b->branch_tree.tree) {
2864 release_tree_content_recursive(b->branch_tree.tree);
2865 b->branch_tree.tree = NULL;
2868 else
2869 b = new_branch(arg);
2870 read_next_command();
2871 parse_from(b);
2872 if (command_buf.len > 0)
2873 unread_command_buf = 1;
2876 static void cat_blob_write(const char *buf, unsigned long size)
2878 if (write_in_full(cat_blob_fd, buf, size) != size)
2879 die_errno("Write to frontend failed");
2882 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2884 struct strbuf line = STRBUF_INIT;
2885 unsigned long size;
2886 enum object_type type = 0;
2887 char *buf;
2889 if (!oe || oe->pack_id == MAX_PACK_ID) {
2890 buf = read_sha1_file(sha1, &type, &size);
2891 } else {
2892 type = oe->type;
2893 buf = gfi_unpack_entry(oe, &size);
2897 * Output based on batch_one_object() from cat-file.c.
2899 if (type <= 0) {
2900 strbuf_reset(&line);
2901 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2902 cat_blob_write(line.buf, line.len);
2903 strbuf_release(&line);
2904 free(buf);
2905 return;
2907 if (!buf)
2908 die("Can't read object %s", sha1_to_hex(sha1));
2909 if (type != OBJ_BLOB)
2910 die("Object %s is a %s but a blob was expected.",
2911 sha1_to_hex(sha1), typename(type));
2912 strbuf_reset(&line);
2913 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2914 typename(type), size);
2915 cat_blob_write(line.buf, line.len);
2916 strbuf_release(&line);
2917 cat_blob_write(buf, size);
2918 cat_blob_write("\n", 1);
2919 if (oe && oe->pack_id == pack_id) {
2920 last_blob.offset = oe->idx.offset;
2921 strbuf_attach(&last_blob.data, buf, size, size);
2922 last_blob.depth = oe->depth;
2923 } else
2924 free(buf);
2927 static void parse_cat_blob(const char *p)
2929 struct object_entry *oe = oe;
2930 unsigned char sha1[20];
2932 /* cat-blob SP <object> LF */
2933 if (*p == ':') {
2934 oe = find_mark(parse_mark_ref_eol(p));
2935 if (!oe)
2936 die("Unknown mark: %s", command_buf.buf);
2937 hashcpy(sha1, oe->idx.sha1);
2938 } else {
2939 if (get_sha1_hex(p, sha1))
2940 die("Invalid dataref: %s", command_buf.buf);
2941 if (p[40])
2942 die("Garbage after SHA1: %s", command_buf.buf);
2943 oe = find_object(sha1);
2946 cat_blob(oe, sha1);
2949 static struct object_entry *dereference(struct object_entry *oe,
2950 unsigned char sha1[20])
2952 unsigned long size;
2953 char *buf = NULL;
2954 if (!oe) {
2955 enum object_type type = sha1_object_info(sha1, NULL);
2956 if (type < 0)
2957 die("object not found: %s", sha1_to_hex(sha1));
2958 /* cache it! */
2959 oe = insert_object(sha1);
2960 oe->type = type;
2961 oe->pack_id = MAX_PACK_ID;
2962 oe->idx.offset = 1;
2964 switch (oe->type) {
2965 case OBJ_TREE: /* easy case. */
2966 return oe;
2967 case OBJ_COMMIT:
2968 case OBJ_TAG:
2969 break;
2970 default:
2971 die("Not a tree-ish: %s", command_buf.buf);
2974 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2975 buf = gfi_unpack_entry(oe, &size);
2976 } else {
2977 enum object_type unused;
2978 buf = read_sha1_file(sha1, &unused, &size);
2980 if (!buf)
2981 die("Can't load object %s", sha1_to_hex(sha1));
2983 /* Peel one layer. */
2984 switch (oe->type) {
2985 case OBJ_TAG:
2986 if (size < 40 + strlen("object ") ||
2987 get_sha1_hex(buf + strlen("object "), sha1))
2988 die("Invalid SHA1 in tag: %s", command_buf.buf);
2989 break;
2990 case OBJ_COMMIT:
2991 if (size < 40 + strlen("tree ") ||
2992 get_sha1_hex(buf + strlen("tree "), sha1))
2993 die("Invalid SHA1 in commit: %s", command_buf.buf);
2996 free(buf);
2997 return find_object(sha1);
3000 static struct object_entry *parse_treeish_dataref(const char **p)
3002 unsigned char sha1[20];
3003 struct object_entry *e;
3005 if (**p == ':') { /* <mark> */
3006 e = find_mark(parse_mark_ref_space(p));
3007 if (!e)
3008 die("Unknown mark: %s", command_buf.buf);
3009 hashcpy(sha1, e->idx.sha1);
3010 } else { /* <sha1> */
3011 if (get_sha1_hex(*p, sha1))
3012 die("Invalid dataref: %s", command_buf.buf);
3013 e = find_object(sha1);
3014 *p += 40;
3015 if (*(*p)++ != ' ')
3016 die("Missing space after tree-ish: %s", command_buf.buf);
3019 while (!e || e->type != OBJ_TREE)
3020 e = dereference(e, sha1);
3021 return e;
3024 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3026 static struct strbuf line = STRBUF_INIT;
3028 /* See show_tree(). */
3029 const char *type =
3030 S_ISGITLINK(mode) ? commit_type :
3031 S_ISDIR(mode) ? tree_type :
3032 blob_type;
3034 if (!mode) {
3035 /* missing SP path LF */
3036 strbuf_reset(&line);
3037 strbuf_addstr(&line, "missing ");
3038 quote_c_style(path, &line, NULL, 0);
3039 strbuf_addch(&line, '\n');
3040 } else {
3041 /* mode SP type SP object_name TAB path LF */
3042 strbuf_reset(&line);
3043 strbuf_addf(&line, "%06o %s %s\t",
3044 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3045 quote_c_style(path, &line, NULL, 0);
3046 strbuf_addch(&line, '\n');
3048 cat_blob_write(line.buf, line.len);
3051 static void parse_ls(const char *p, struct branch *b)
3053 struct tree_entry *root = NULL;
3054 struct tree_entry leaf = {NULL};
3056 /* ls SP (<tree-ish> SP)? <path> */
3057 if (*p == '"') {
3058 if (!b)
3059 die("Not in a commit: %s", command_buf.buf);
3060 root = &b->branch_tree;
3061 } else {
3062 struct object_entry *e = parse_treeish_dataref(&p);
3063 root = new_tree_entry();
3064 hashcpy(root->versions[1].sha1, e->idx.sha1);
3065 if (!is_null_sha1(root->versions[1].sha1))
3066 root->versions[1].mode = S_IFDIR;
3067 load_tree(root);
3069 if (*p == '"') {
3070 static struct strbuf uq = STRBUF_INIT;
3071 const char *endp;
3072 strbuf_reset(&uq);
3073 if (unquote_c_style(&uq, p, &endp))
3074 die("Invalid path: %s", command_buf.buf);
3075 if (*endp)
3076 die("Garbage after path in: %s", command_buf.buf);
3077 p = uq.buf;
3079 tree_content_get(root, p, &leaf, 1);
3081 * A directory in preparation would have a sha1 of zero
3082 * until it is saved. Save, for simplicity.
3084 if (S_ISDIR(leaf.versions[1].mode))
3085 store_tree(&leaf);
3087 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3088 if (leaf.tree)
3089 release_tree_content_recursive(leaf.tree);
3090 if (!b || root != &b->branch_tree)
3091 release_tree_entry(root);
3094 static void checkpoint(void)
3096 checkpoint_requested = 0;
3097 if (object_count) {
3098 cycle_packfile();
3099 dump_branches();
3100 dump_tags();
3101 dump_marks();
3105 static void parse_checkpoint(void)
3107 checkpoint_requested = 1;
3108 skip_optional_lf();
3111 static void parse_progress(void)
3113 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3114 fputc('\n', stdout);
3115 fflush(stdout);
3116 skip_optional_lf();
3119 static char* make_fast_import_path(const char *path)
3121 if (!relative_marks_paths || is_absolute_path(path))
3122 return xstrdup(path);
3123 return xstrdup(git_path("info/fast-import/%s", path));
3126 static void option_import_marks(const char *marks,
3127 int from_stream, int ignore_missing)
3129 if (import_marks_file) {
3130 if (from_stream)
3131 die("Only one import-marks command allowed per stream");
3133 /* read previous mark file */
3134 if(!import_marks_file_from_stream)
3135 read_marks();
3138 import_marks_file = make_fast_import_path(marks);
3139 safe_create_leading_directories_const(import_marks_file);
3140 import_marks_file_from_stream = from_stream;
3141 import_marks_file_ignore_missing = ignore_missing;
3144 static void option_date_format(const char *fmt)
3146 if (!strcmp(fmt, "raw"))
3147 whenspec = WHENSPEC_RAW;
3148 else if (!strcmp(fmt, "rfc2822"))
3149 whenspec = WHENSPEC_RFC2822;
3150 else if (!strcmp(fmt, "now"))
3151 whenspec = WHENSPEC_NOW;
3152 else
3153 die("unknown --date-format argument %s", fmt);
3156 static unsigned long ulong_arg(const char *option, const char *arg)
3158 char *endptr;
3159 unsigned long rv = strtoul(arg, &endptr, 0);
3160 if (strchr(arg, '-') || endptr == arg || *endptr)
3161 die("%s: argument must be a non-negative integer", option);
3162 return rv;
3165 static void option_depth(const char *depth)
3167 max_depth = ulong_arg("--depth", depth);
3168 if (max_depth > MAX_DEPTH)
3169 die("--depth cannot exceed %u", MAX_DEPTH);
3172 static void option_active_branches(const char *branches)
3174 max_active_branches = ulong_arg("--active-branches", branches);
3177 static void option_export_marks(const char *marks)
3179 export_marks_file = make_fast_import_path(marks);
3180 safe_create_leading_directories_const(export_marks_file);
3183 static void option_cat_blob_fd(const char *fd)
3185 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3186 if (n > (unsigned long) INT_MAX)
3187 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3188 cat_blob_fd = (int) n;
3191 static void option_export_pack_edges(const char *edges)
3193 if (pack_edges)
3194 fclose(pack_edges);
3195 pack_edges = fopen(edges, "a");
3196 if (!pack_edges)
3197 die_errno("Cannot open '%s'", edges);
3200 static int parse_one_option(const char *option)
3202 if (skip_prefix(option, "max-pack-size=", &option)) {
3203 unsigned long v;
3204 if (!git_parse_ulong(option, &v))
3205 return 0;
3206 if (v < 8192) {
3207 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3208 v *= 1024 * 1024;
3209 } else if (v < 1024 * 1024) {
3210 warning("minimum max-pack-size is 1 MiB");
3211 v = 1024 * 1024;
3213 max_packsize = v;
3214 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3215 unsigned long v;
3216 if (!git_parse_ulong(option, &v))
3217 return 0;
3218 big_file_threshold = v;
3219 } else if (skip_prefix(option, "depth=", &option)) {
3220 option_depth(option);
3221 } else if (skip_prefix(option, "active-branches=", &option)) {
3222 option_active_branches(option);
3223 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3224 option_export_pack_edges(option);
3225 } else if (starts_with(option, "quiet")) {
3226 show_stats = 0;
3227 } else if (starts_with(option, "stats")) {
3228 show_stats = 1;
3229 } else {
3230 return 0;
3233 return 1;
3236 static int parse_one_feature(const char *feature, int from_stream)
3238 const char *arg;
3240 if (skip_prefix(feature, "date-format=", &arg)) {
3241 option_date_format(arg);
3242 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3243 option_import_marks(arg, from_stream, 0);
3244 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3245 option_import_marks(arg, from_stream, 1);
3246 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3247 option_export_marks(arg);
3248 } else if (!strcmp(feature, "cat-blob")) {
3249 ; /* Don't die - this feature is supported */
3250 } else if (!strcmp(feature, "relative-marks")) {
3251 relative_marks_paths = 1;
3252 } else if (!strcmp(feature, "no-relative-marks")) {
3253 relative_marks_paths = 0;
3254 } else if (!strcmp(feature, "done")) {
3255 require_explicit_termination = 1;
3256 } else if (!strcmp(feature, "force")) {
3257 force_update = 1;
3258 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3259 ; /* do nothing; we have the feature */
3260 } else {
3261 return 0;
3264 return 1;
3267 static void parse_feature(const char *feature)
3269 if (seen_data_command)
3270 die("Got feature command '%s' after data command", feature);
3272 if (parse_one_feature(feature, 1))
3273 return;
3275 die("This version of fast-import does not support feature %s.", feature);
3278 static void parse_option(const char *option)
3280 if (seen_data_command)
3281 die("Got option command '%s' after data command", option);
3283 if (parse_one_option(option))
3284 return;
3286 die("This version of fast-import does not support option: %s", option);
3289 static void git_pack_config(void)
3291 int indexversion_value;
3292 unsigned long packsizelimit_value;
3294 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3295 if (max_depth > MAX_DEPTH)
3296 max_depth = MAX_DEPTH;
3298 if (!git_config_get_int("pack.compression", &pack_compression_level)) {
3299 if (pack_compression_level == -1)
3300 pack_compression_level = Z_DEFAULT_COMPRESSION;
3301 else if (pack_compression_level < 0 ||
3302 pack_compression_level > Z_BEST_COMPRESSION)
3303 git_die_config("pack.compression",
3304 "bad pack compression level %d", pack_compression_level);
3305 pack_compression_seen = 1;
3307 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3308 pack_idx_opts.version = indexversion_value;
3309 if (pack_idx_opts.version > 2)
3310 git_die_config("pack.indexversion",
3311 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3313 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3314 max_packsize = packsizelimit_value;
3316 git_config(git_default_config, NULL);
3319 static const char fast_import_usage[] =
3320 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3322 static void parse_argv(void)
3324 unsigned int i;
3326 for (i = 1; i < global_argc; i++) {
3327 const char *a = global_argv[i];
3329 if (*a != '-' || !strcmp(a, "--"))
3330 break;
3332 if (!skip_prefix(a, "--", &a))
3333 die("unknown option %s", a);
3335 if (parse_one_option(a))
3336 continue;
3338 if (parse_one_feature(a, 0))
3339 continue;
3341 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3342 option_cat_blob_fd(a);
3343 continue;
3346 die("unknown option --%s", a);
3348 if (i != global_argc)
3349 usage(fast_import_usage);
3351 seen_data_command = 1;
3352 if (import_marks_file)
3353 read_marks();
3356 int main(int argc, char **argv)
3358 unsigned int i;
3360 git_extract_argv0_path(argv[0]);
3362 git_setup_gettext();
3364 if (argc == 2 && !strcmp(argv[1], "-h"))
3365 usage(fast_import_usage);
3367 setup_git_directory();
3368 reset_pack_idx_option(&pack_idx_opts);
3369 git_pack_config();
3370 if (!pack_compression_seen && core_compression_seen)
3371 pack_compression_level = core_compression_level;
3373 alloc_objects(object_entry_alloc);
3374 strbuf_init(&command_buf, 0);
3375 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3376 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3377 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3378 marks = pool_calloc(1, sizeof(struct mark_set));
3380 global_argc = argc;
3381 global_argv = argv;
3383 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3384 for (i = 0; i < (cmd_save - 1); i++)
3385 rc_free[i].next = &rc_free[i + 1];
3386 rc_free[cmd_save - 1].next = NULL;
3388 prepare_packed_git();
3389 start_packfile();
3390 set_die_routine(die_nicely);
3391 set_checkpoint_signal();
3392 while (read_next_command() != EOF) {
3393 const char *v;
3394 if (!strcmp("blob", command_buf.buf))
3395 parse_new_blob();
3396 else if (skip_prefix(command_buf.buf, "ls ", &v))
3397 parse_ls(v, NULL);
3398 else if (skip_prefix(command_buf.buf, "commit ", &v))
3399 parse_new_commit(v);
3400 else if (skip_prefix(command_buf.buf, "tag ", &v))
3401 parse_new_tag(v);
3402 else if (skip_prefix(command_buf.buf, "reset ", &v))
3403 parse_reset_branch(v);
3404 else if (!strcmp("checkpoint", command_buf.buf))
3405 parse_checkpoint();
3406 else if (!strcmp("done", command_buf.buf))
3407 break;
3408 else if (starts_with(command_buf.buf, "progress "))
3409 parse_progress();
3410 else if (skip_prefix(command_buf.buf, "feature ", &v))
3411 parse_feature(v);
3412 else if (skip_prefix(command_buf.buf, "option git ", &v))
3413 parse_option(v);
3414 else if (starts_with(command_buf.buf, "option "))
3415 /* ignore non-git options*/;
3416 else
3417 die("Unsupported command: %s", command_buf.buf);
3419 if (checkpoint_requested)
3420 checkpoint();
3423 /* argv hasn't been parsed yet, do so */
3424 if (!seen_data_command)
3425 parse_argv();
3427 if (require_explicit_termination && feof(stdin))
3428 die("stream ends early");
3430 end_packfile();
3432 dump_branches();
3433 dump_tags();
3434 unkeep_all_packs();
3435 dump_marks();
3437 if (pack_edges)
3438 fclose(pack_edges);
3440 if (show_stats) {
3441 uintmax_t total_count = 0, duplicate_count = 0;
3442 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3443 total_count += object_count_by_type[i];
3444 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3445 duplicate_count += duplicate_count_by_type[i];
3447 fprintf(stderr, "%s statistics:\n", argv[0]);
3448 fprintf(stderr, "---------------------------------------------------------------------\n");
3449 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3450 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3451 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]);
3452 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]);
3453 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]);
3454 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]);
3455 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3456 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3457 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3458 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3459 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3460 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3461 fprintf(stderr, "---------------------------------------------------------------------\n");
3462 pack_report();
3463 fprintf(stderr, "---------------------------------------------------------------------\n");
3464 fprintf(stderr, "\n");
3467 return failure ? 1 : 0;