Merge branch 'di/fast-import-ident'
[git/gitweb.git] / fast-import.c
blob6d491b92fefe5845fe53bf6995baa077e78a93be
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 committish lf)?
26 ('merge' sp committish 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 committish lf;
47 note_inm ::= 'N' sp 'inline' sp committish lf
48 data;
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp committish 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 committish 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 committish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
97 ref_str ::= ref;
98 sha1exp_str ::= sha1exp;
99 tag_str ::= tag;
100 path_str ::= path | '"' quoted(path) '"' ;
101 mode ::= '100644' | '644'
102 | '100755' | '755'
103 | '120000'
106 declen ::= # unsigned 32 bit value, ascii base10 notation;
107 bigint ::= # unsigned integer value, ascii base10 notation;
108 binary_data ::= # file content, not interpreted;
110 when ::= raw_when | rfc2822_when;
111 raw_when ::= ts sp tz;
112 rfc2822_when ::= # Valid RFC 2822 date and time;
114 sp ::= # ASCII space character;
115 lf ::= # ASCII newline (LF) character;
117 # note: a colon (':') must precede the numerical value assigned to
118 # an idnum. This is to distinguish it from a ref or tag name as
119 # GIT does not permit ':' in ref or tag strings.
121 idnum ::= ':' bigint;
122 path ::= # GIT style file path, e.g. "a/b/c";
123 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
124 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
125 sha1exp ::= # Any valid GIT SHA1 expression;
126 hexsha1 ::= # SHA1 in hexadecimal format;
128 # note: name and email are UTF8 strings, however name must not
129 # contain '<' or lf and email must not contain any of the
130 # following: '<', '>', lf.
132 name ::= # valid GIT author/committer name;
133 email ::= # valid GIT author/committer email;
134 ts ::= # time since the epoch in seconds, ascii base10 notation;
135 tz ::= # GIT style timezone;
137 # note: comments, ls and cat requests may appear anywhere
138 # in the input, except within a data command. Any form
139 # of the data command always escapes the related input
140 # from comment processing.
142 # In case it is not clear, the '#' that starts the comment
143 # must be the first character on that line (an lf
144 # preceded it).
147 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
148 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
150 comment ::= '#' not_lf* lf;
151 not_lf ::= # Any byte that is not ASCII newline (LF);
154 #include "builtin.h"
155 #include "cache.h"
156 #include "object.h"
157 #include "blob.h"
158 #include "tree.h"
159 #include "commit.h"
160 #include "delta.h"
161 #include "pack.h"
162 #include "refs.h"
163 #include "csum-file.h"
164 #include "quote.h"
165 #include "exec_cmd.h"
166 #include "dir.h"
168 #define PACK_ID_BITS 16
169 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
170 #define DEPTH_BITS 13
171 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
173 struct object_entry {
174 struct pack_idx_entry idx;
175 struct object_entry *next;
176 uint32_t type : TYPE_BITS,
177 pack_id : PACK_ID_BITS,
178 depth : DEPTH_BITS;
181 struct object_entry_pool {
182 struct object_entry_pool *next_pool;
183 struct object_entry *next_free;
184 struct object_entry *end;
185 struct object_entry entries[FLEX_ARRAY]; /* more */
188 struct mark_set {
189 union {
190 struct object_entry *marked[1024];
191 struct mark_set *sets[1024];
192 } data;
193 unsigned int shift;
196 struct last_object {
197 struct strbuf data;
198 off_t offset;
199 unsigned int depth;
200 unsigned no_swap : 1;
203 struct mem_pool {
204 struct mem_pool *next_pool;
205 char *next_free;
206 char *end;
207 uintmax_t space[FLEX_ARRAY]; /* more */
210 struct atom_str {
211 struct atom_str *next_atom;
212 unsigned short str_len;
213 char str_dat[FLEX_ARRAY]; /* more */
216 struct tree_content;
217 struct tree_entry {
218 struct tree_content *tree;
219 struct atom_str *name;
220 struct tree_entry_ms {
221 uint16_t mode;
222 unsigned char sha1[20];
223 } versions[2];
226 struct tree_content {
227 unsigned int entry_capacity; /* must match avail_tree_content */
228 unsigned int entry_count;
229 unsigned int delta_depth;
230 struct tree_entry *entries[FLEX_ARRAY]; /* more */
233 struct avail_tree_content {
234 unsigned int entry_capacity; /* must match tree_content */
235 struct avail_tree_content *next_avail;
238 struct branch {
239 struct branch *table_next_branch;
240 struct branch *active_next_branch;
241 const char *name;
242 struct tree_entry branch_tree;
243 uintmax_t last_commit;
244 uintmax_t num_notes;
245 unsigned active : 1;
246 unsigned pack_id : PACK_ID_BITS;
247 unsigned char sha1[20];
250 struct tag {
251 struct tag *next_tag;
252 const char *name;
253 unsigned int pack_id;
254 unsigned char sha1[20];
257 struct hash_list {
258 struct hash_list *next;
259 unsigned char sha1[20];
262 typedef enum {
263 WHENSPEC_RAW = 1,
264 WHENSPEC_RFC2822,
265 WHENSPEC_NOW
266 } whenspec_type;
268 struct recent_command {
269 struct recent_command *prev;
270 struct recent_command *next;
271 char *buf;
274 /* Configured limits on output */
275 static unsigned long max_depth = 10;
276 static off_t max_packsize;
277 static int force_update;
278 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
279 static int pack_compression_seen;
281 /* Stats and misc. counters */
282 static uintmax_t alloc_count;
283 static uintmax_t marks_set_count;
284 static uintmax_t object_count_by_type[1 << TYPE_BITS];
285 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
286 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
287 static unsigned long object_count;
288 static unsigned long branch_count;
289 static unsigned long branch_load_count;
290 static int failure;
291 static FILE *pack_edges;
292 static unsigned int show_stats = 1;
293 static int global_argc;
294 static const char **global_argv;
296 /* Memory pools */
297 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
298 static size_t total_allocd;
299 static struct mem_pool *mem_pool;
301 /* Atom management */
302 static unsigned int atom_table_sz = 4451;
303 static unsigned int atom_cnt;
304 static struct atom_str **atom_table;
306 /* The .pack file being generated */
307 static struct pack_idx_option pack_idx_opts;
308 static unsigned int pack_id;
309 static struct sha1file *pack_file;
310 static struct packed_git *pack_data;
311 static struct packed_git **all_packs;
312 static off_t pack_size;
314 /* Table of objects we've written. */
315 static unsigned int object_entry_alloc = 5000;
316 static struct object_entry_pool *blocks;
317 static struct object_entry *object_table[1 << 16];
318 static struct mark_set *marks;
319 static const char *export_marks_file;
320 static const char *import_marks_file;
321 static int import_marks_file_from_stream;
322 static int import_marks_file_ignore_missing;
323 static int relative_marks_paths;
325 /* Our last blob */
326 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
328 /* Tree management */
329 static unsigned int tree_entry_alloc = 1000;
330 static void *avail_tree_entry;
331 static unsigned int avail_tree_table_sz = 100;
332 static struct avail_tree_content **avail_tree_table;
333 static struct strbuf old_tree = STRBUF_INIT;
334 static struct strbuf new_tree = STRBUF_INIT;
336 /* Branch data */
337 static unsigned long max_active_branches = 5;
338 static unsigned long cur_active_branches;
339 static unsigned long branch_table_sz = 1039;
340 static struct branch **branch_table;
341 static struct branch *active_branches;
343 /* Tag data */
344 static struct tag *first_tag;
345 static struct tag *last_tag;
347 /* Input stream parsing */
348 static whenspec_type whenspec = WHENSPEC_RAW;
349 static struct strbuf command_buf = STRBUF_INIT;
350 static int unread_command_buf;
351 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
352 static struct recent_command *cmd_tail = &cmd_hist;
353 static struct recent_command *rc_free;
354 static unsigned int cmd_save = 100;
355 static uintmax_t next_mark;
356 static struct strbuf new_data = STRBUF_INIT;
357 static int seen_data_command;
358 static int require_explicit_termination;
360 /* Signal handling */
361 static volatile sig_atomic_t checkpoint_requested;
363 /* Where to write output of cat-blob commands */
364 static int cat_blob_fd = STDOUT_FILENO;
366 static void parse_argv(void);
367 static void parse_cat_blob(void);
368 static void parse_ls(struct branch *b);
370 static void write_branch_report(FILE *rpt, struct branch *b)
372 fprintf(rpt, "%s:\n", b->name);
374 fprintf(rpt, " status :");
375 if (b->active)
376 fputs(" active", rpt);
377 if (b->branch_tree.tree)
378 fputs(" loaded", rpt);
379 if (is_null_sha1(b->branch_tree.versions[1].sha1))
380 fputs(" dirty", rpt);
381 fputc('\n', rpt);
383 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
384 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
385 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
386 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
388 fputs(" last pack : ", rpt);
389 if (b->pack_id < MAX_PACK_ID)
390 fprintf(rpt, "%u", b->pack_id);
391 fputc('\n', rpt);
393 fputc('\n', rpt);
396 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
398 static void write_crash_report(const char *err)
400 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
401 FILE *rpt = fopen(loc, "w");
402 struct branch *b;
403 unsigned long lu;
404 struct recent_command *rc;
406 if (!rpt) {
407 error("can't write crash report %s: %s", loc, strerror(errno));
408 return;
411 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
413 fprintf(rpt, "fast-import crash report:\n");
414 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
415 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
416 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
417 fputc('\n', rpt);
419 fputs("fatal: ", rpt);
420 fputs(err, rpt);
421 fputc('\n', rpt);
423 fputc('\n', rpt);
424 fputs("Most Recent Commands Before Crash\n", rpt);
425 fputs("---------------------------------\n", rpt);
426 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
427 if (rc->next == &cmd_hist)
428 fputs("* ", rpt);
429 else
430 fputs(" ", rpt);
431 fputs(rc->buf, rpt);
432 fputc('\n', rpt);
435 fputc('\n', rpt);
436 fputs("Active Branch LRU\n", rpt);
437 fputs("-----------------\n", rpt);
438 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
439 cur_active_branches,
440 max_active_branches);
441 fputc('\n', rpt);
442 fputs(" pos clock name\n", rpt);
443 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
444 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
445 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
446 ++lu, b->last_commit, b->name);
448 fputc('\n', rpt);
449 fputs("Inactive Branches\n", rpt);
450 fputs("-----------------\n", rpt);
451 for (lu = 0; lu < branch_table_sz; lu++) {
452 for (b = branch_table[lu]; b; b = b->table_next_branch)
453 write_branch_report(rpt, b);
456 if (first_tag) {
457 struct tag *tg;
458 fputc('\n', rpt);
459 fputs("Annotated Tags\n", rpt);
460 fputs("--------------\n", rpt);
461 for (tg = first_tag; tg; tg = tg->next_tag) {
462 fputs(sha1_to_hex(tg->sha1), rpt);
463 fputc(' ', rpt);
464 fputs(tg->name, rpt);
465 fputc('\n', rpt);
469 fputc('\n', rpt);
470 fputs("Marks\n", rpt);
471 fputs("-----\n", rpt);
472 if (export_marks_file)
473 fprintf(rpt, " exported to %s\n", export_marks_file);
474 else
475 dump_marks_helper(rpt, 0, marks);
477 fputc('\n', rpt);
478 fputs("-------------------\n", rpt);
479 fputs("END OF CRASH REPORT\n", rpt);
480 fclose(rpt);
483 static void end_packfile(void);
484 static void unkeep_all_packs(void);
485 static void dump_marks(void);
487 static NORETURN void die_nicely(const char *err, va_list params)
489 static int zombie;
490 char message[2 * PATH_MAX];
492 vsnprintf(message, sizeof(message), err, params);
493 fputs("fatal: ", stderr);
494 fputs(message, stderr);
495 fputc('\n', stderr);
497 if (!zombie) {
498 zombie = 1;
499 write_crash_report(message);
500 end_packfile();
501 unkeep_all_packs();
502 dump_marks();
504 exit(128);
507 #ifndef SIGUSR1 /* Windows, for example */
509 static void set_checkpoint_signal(void)
513 #else
515 static void checkpoint_signal(int signo)
517 checkpoint_requested = 1;
520 static void set_checkpoint_signal(void)
522 struct sigaction sa;
524 memset(&sa, 0, sizeof(sa));
525 sa.sa_handler = checkpoint_signal;
526 sigemptyset(&sa.sa_mask);
527 sa.sa_flags = SA_RESTART;
528 sigaction(SIGUSR1, &sa, NULL);
531 #endif
533 static void alloc_objects(unsigned int cnt)
535 struct object_entry_pool *b;
537 b = xmalloc(sizeof(struct object_entry_pool)
538 + cnt * sizeof(struct object_entry));
539 b->next_pool = blocks;
540 b->next_free = b->entries;
541 b->end = b->entries + cnt;
542 blocks = b;
543 alloc_count += cnt;
546 static struct object_entry *new_object(unsigned char *sha1)
548 struct object_entry *e;
550 if (blocks->next_free == blocks->end)
551 alloc_objects(object_entry_alloc);
553 e = blocks->next_free++;
554 hashcpy(e->idx.sha1, sha1);
555 return e;
558 static struct object_entry *find_object(unsigned char *sha1)
560 unsigned int h = sha1[0] << 8 | sha1[1];
561 struct object_entry *e;
562 for (e = object_table[h]; e; e = e->next)
563 if (!hashcmp(sha1, e->idx.sha1))
564 return e;
565 return NULL;
568 static struct object_entry *insert_object(unsigned char *sha1)
570 unsigned int h = sha1[0] << 8 | sha1[1];
571 struct object_entry *e = object_table[h];
573 while (e) {
574 if (!hashcmp(sha1, e->idx.sha1))
575 return e;
576 e = e->next;
579 e = new_object(sha1);
580 e->next = object_table[h];
581 e->idx.offset = 0;
582 object_table[h] = e;
583 return e;
586 static unsigned int hc_str(const char *s, size_t len)
588 unsigned int r = 0;
589 while (len-- > 0)
590 r = r * 31 + *s++;
591 return r;
594 static void *pool_alloc(size_t len)
596 struct mem_pool *p;
597 void *r;
599 /* round up to a 'uintmax_t' alignment */
600 if (len & (sizeof(uintmax_t) - 1))
601 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
603 for (p = mem_pool; p; p = p->next_pool)
604 if ((p->end - p->next_free >= len))
605 break;
607 if (!p) {
608 if (len >= (mem_pool_alloc/2)) {
609 total_allocd += len;
610 return xmalloc(len);
612 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
613 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
614 p->next_pool = mem_pool;
615 p->next_free = (char *) p->space;
616 p->end = p->next_free + mem_pool_alloc;
617 mem_pool = p;
620 r = p->next_free;
621 p->next_free += len;
622 return r;
625 static void *pool_calloc(size_t count, size_t size)
627 size_t len = count * size;
628 void *r = pool_alloc(len);
629 memset(r, 0, len);
630 return r;
633 static char *pool_strdup(const char *s)
635 char *r = pool_alloc(strlen(s) + 1);
636 strcpy(r, s);
637 return r;
640 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
642 struct mark_set *s = marks;
643 while ((idnum >> s->shift) >= 1024) {
644 s = pool_calloc(1, sizeof(struct mark_set));
645 s->shift = marks->shift + 10;
646 s->data.sets[0] = marks;
647 marks = s;
649 while (s->shift) {
650 uintmax_t i = idnum >> s->shift;
651 idnum -= i << s->shift;
652 if (!s->data.sets[i]) {
653 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
654 s->data.sets[i]->shift = s->shift - 10;
656 s = s->data.sets[i];
658 if (!s->data.marked[idnum])
659 marks_set_count++;
660 s->data.marked[idnum] = oe;
663 static struct object_entry *find_mark(uintmax_t idnum)
665 uintmax_t orig_idnum = idnum;
666 struct mark_set *s = marks;
667 struct object_entry *oe = NULL;
668 if ((idnum >> s->shift) < 1024) {
669 while (s && s->shift) {
670 uintmax_t i = idnum >> s->shift;
671 idnum -= i << s->shift;
672 s = s->data.sets[i];
674 if (s)
675 oe = s->data.marked[idnum];
677 if (!oe)
678 die("mark :%" PRIuMAX " not declared", orig_idnum);
679 return oe;
682 static struct atom_str *to_atom(const char *s, unsigned short len)
684 unsigned int hc = hc_str(s, len) % atom_table_sz;
685 struct atom_str *c;
687 for (c = atom_table[hc]; c; c = c->next_atom)
688 if (c->str_len == len && !strncmp(s, c->str_dat, len))
689 return c;
691 c = pool_alloc(sizeof(struct atom_str) + len + 1);
692 c->str_len = len;
693 strncpy(c->str_dat, s, len);
694 c->str_dat[len] = 0;
695 c->next_atom = atom_table[hc];
696 atom_table[hc] = c;
697 atom_cnt++;
698 return c;
701 static struct branch *lookup_branch(const char *name)
703 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
704 struct branch *b;
706 for (b = branch_table[hc]; b; b = b->table_next_branch)
707 if (!strcmp(name, b->name))
708 return b;
709 return NULL;
712 static struct branch *new_branch(const char *name)
714 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
715 struct branch *b = lookup_branch(name);
717 if (b)
718 die("Invalid attempt to create duplicate branch: %s", name);
719 switch (check_ref_format(name)) {
720 case 0: break; /* its valid */
721 case CHECK_REF_FORMAT_ONELEVEL:
722 break; /* valid, but too few '/', allow anyway */
723 default:
724 die("Branch name doesn't conform to GIT standards: %s", name);
727 b = pool_calloc(1, sizeof(struct branch));
728 b->name = pool_strdup(name);
729 b->table_next_branch = branch_table[hc];
730 b->branch_tree.versions[0].mode = S_IFDIR;
731 b->branch_tree.versions[1].mode = S_IFDIR;
732 b->num_notes = 0;
733 b->active = 0;
734 b->pack_id = MAX_PACK_ID;
735 branch_table[hc] = b;
736 branch_count++;
737 return b;
740 static unsigned int hc_entries(unsigned int cnt)
742 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
743 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
746 static struct tree_content *new_tree_content(unsigned int cnt)
748 struct avail_tree_content *f, *l = NULL;
749 struct tree_content *t;
750 unsigned int hc = hc_entries(cnt);
752 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
753 if (f->entry_capacity >= cnt)
754 break;
756 if (f) {
757 if (l)
758 l->next_avail = f->next_avail;
759 else
760 avail_tree_table[hc] = f->next_avail;
761 } else {
762 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
763 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
764 f->entry_capacity = cnt;
767 t = (struct tree_content*)f;
768 t->entry_count = 0;
769 t->delta_depth = 0;
770 return t;
773 static void release_tree_entry(struct tree_entry *e);
774 static void release_tree_content(struct tree_content *t)
776 struct avail_tree_content *f = (struct avail_tree_content*)t;
777 unsigned int hc = hc_entries(f->entry_capacity);
778 f->next_avail = avail_tree_table[hc];
779 avail_tree_table[hc] = f;
782 static void release_tree_content_recursive(struct tree_content *t)
784 unsigned int i;
785 for (i = 0; i < t->entry_count; i++)
786 release_tree_entry(t->entries[i]);
787 release_tree_content(t);
790 static struct tree_content *grow_tree_content(
791 struct tree_content *t,
792 int amt)
794 struct tree_content *r = new_tree_content(t->entry_count + amt);
795 r->entry_count = t->entry_count;
796 r->delta_depth = t->delta_depth;
797 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
798 release_tree_content(t);
799 return r;
802 static struct tree_entry *new_tree_entry(void)
804 struct tree_entry *e;
806 if (!avail_tree_entry) {
807 unsigned int n = tree_entry_alloc;
808 total_allocd += n * sizeof(struct tree_entry);
809 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
810 while (n-- > 1) {
811 *((void**)e) = e + 1;
812 e++;
814 *((void**)e) = NULL;
817 e = avail_tree_entry;
818 avail_tree_entry = *((void**)e);
819 return e;
822 static void release_tree_entry(struct tree_entry *e)
824 if (e->tree)
825 release_tree_content_recursive(e->tree);
826 *((void**)e) = avail_tree_entry;
827 avail_tree_entry = e;
830 static struct tree_content *dup_tree_content(struct tree_content *s)
832 struct tree_content *d;
833 struct tree_entry *a, *b;
834 unsigned int i;
836 if (!s)
837 return NULL;
838 d = new_tree_content(s->entry_count);
839 for (i = 0; i < s->entry_count; i++) {
840 a = s->entries[i];
841 b = new_tree_entry();
842 memcpy(b, a, sizeof(*a));
843 if (a->tree && is_null_sha1(b->versions[1].sha1))
844 b->tree = dup_tree_content(a->tree);
845 else
846 b->tree = NULL;
847 d->entries[i] = b;
849 d->entry_count = s->entry_count;
850 d->delta_depth = s->delta_depth;
852 return d;
855 static void start_packfile(void)
857 static char tmpfile[PATH_MAX];
858 struct packed_git *p;
859 struct pack_header hdr;
860 int pack_fd;
862 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
863 "pack/tmp_pack_XXXXXX");
864 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
865 strcpy(p->pack_name, tmpfile);
866 p->pack_fd = pack_fd;
867 p->do_not_close = 1;
868 pack_file = sha1fd(pack_fd, p->pack_name);
870 hdr.hdr_signature = htonl(PACK_SIGNATURE);
871 hdr.hdr_version = htonl(2);
872 hdr.hdr_entries = 0;
873 sha1write(pack_file, &hdr, sizeof(hdr));
875 pack_data = p;
876 pack_size = sizeof(hdr);
877 object_count = 0;
879 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
880 all_packs[pack_id] = p;
883 static const char *create_index(void)
885 const char *tmpfile;
886 struct pack_idx_entry **idx, **c, **last;
887 struct object_entry *e;
888 struct object_entry_pool *o;
890 /* Build the table of object IDs. */
891 idx = xmalloc(object_count * sizeof(*idx));
892 c = idx;
893 for (o = blocks; o; o = o->next_pool)
894 for (e = o->next_free; e-- != o->entries;)
895 if (pack_id == e->pack_id)
896 *c++ = &e->idx;
897 last = idx + object_count;
898 if (c != last)
899 die("internal consistency error creating the index");
901 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
902 free(idx);
903 return tmpfile;
906 static char *keep_pack(const char *curr_index_name)
908 static char name[PATH_MAX];
909 static const char *keep_msg = "fast-import";
910 int keep_fd;
912 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
913 if (keep_fd < 0)
914 die_errno("cannot create keep file");
915 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
916 if (close(keep_fd))
917 die_errno("failed to write keep file");
919 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
920 get_object_directory(), sha1_to_hex(pack_data->sha1));
921 if (move_temp_to_file(pack_data->pack_name, name))
922 die("cannot store pack file");
924 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
925 get_object_directory(), sha1_to_hex(pack_data->sha1));
926 if (move_temp_to_file(curr_index_name, name))
927 die("cannot store index file");
928 free((void *)curr_index_name);
929 return name;
932 static void unkeep_all_packs(void)
934 static char name[PATH_MAX];
935 int k;
937 for (k = 0; k < pack_id; k++) {
938 struct packed_git *p = all_packs[k];
939 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
940 get_object_directory(), sha1_to_hex(p->sha1));
941 unlink_or_warn(name);
945 static void end_packfile(void)
947 struct packed_git *old_p = pack_data, *new_p;
949 clear_delta_base_cache();
950 if (object_count) {
951 unsigned char cur_pack_sha1[20];
952 char *idx_name;
953 int i;
954 struct branch *b;
955 struct tag *t;
957 close_pack_windows(pack_data);
958 sha1close(pack_file, cur_pack_sha1, 0);
959 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
960 pack_data->pack_name, object_count,
961 cur_pack_sha1, pack_size);
962 close(pack_data->pack_fd);
963 idx_name = keep_pack(create_index());
965 /* Register the packfile with core git's machinery. */
966 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
967 if (!new_p)
968 die("core git rejected index %s", idx_name);
969 all_packs[pack_id] = new_p;
970 install_packed_git(new_p);
972 /* Print the boundary */
973 if (pack_edges) {
974 fprintf(pack_edges, "%s:", new_p->pack_name);
975 for (i = 0; i < branch_table_sz; i++) {
976 for (b = branch_table[i]; b; b = b->table_next_branch) {
977 if (b->pack_id == pack_id)
978 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
981 for (t = first_tag; t; t = t->next_tag) {
982 if (t->pack_id == pack_id)
983 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
985 fputc('\n', pack_edges);
986 fflush(pack_edges);
989 pack_id++;
991 else {
992 close(old_p->pack_fd);
993 unlink_or_warn(old_p->pack_name);
995 free(old_p);
997 /* We can't carry a delta across packfiles. */
998 strbuf_release(&last_blob.data);
999 last_blob.offset = 0;
1000 last_blob.depth = 0;
1003 static void cycle_packfile(void)
1005 end_packfile();
1006 start_packfile();
1009 static int store_object(
1010 enum object_type type,
1011 struct strbuf *dat,
1012 struct last_object *last,
1013 unsigned char *sha1out,
1014 uintmax_t mark)
1016 void *out, *delta;
1017 struct object_entry *e;
1018 unsigned char hdr[96];
1019 unsigned char sha1[20];
1020 unsigned long hdrlen, deltalen;
1021 git_SHA_CTX c;
1022 git_zstream s;
1024 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1025 (unsigned long)dat->len) + 1;
1026 git_SHA1_Init(&c);
1027 git_SHA1_Update(&c, hdr, hdrlen);
1028 git_SHA1_Update(&c, dat->buf, dat->len);
1029 git_SHA1_Final(sha1, &c);
1030 if (sha1out)
1031 hashcpy(sha1out, sha1);
1033 e = insert_object(sha1);
1034 if (mark)
1035 insert_mark(mark, e);
1036 if (e->idx.offset) {
1037 duplicate_count_by_type[type]++;
1038 return 1;
1039 } else if (find_sha1_pack(sha1, packed_git)) {
1040 e->type = type;
1041 e->pack_id = MAX_PACK_ID;
1042 e->idx.offset = 1; /* just not zero! */
1043 duplicate_count_by_type[type]++;
1044 return 1;
1047 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1048 delta = diff_delta(last->data.buf, last->data.len,
1049 dat->buf, dat->len,
1050 &deltalen, dat->len - 20);
1051 } else
1052 delta = NULL;
1054 memset(&s, 0, sizeof(s));
1055 git_deflate_init(&s, pack_compression_level);
1056 if (delta) {
1057 s.next_in = delta;
1058 s.avail_in = deltalen;
1059 } else {
1060 s.next_in = (void *)dat->buf;
1061 s.avail_in = dat->len;
1063 s.avail_out = git_deflate_bound(&s, s.avail_in);
1064 s.next_out = out = xmalloc(s.avail_out);
1065 while (git_deflate(&s, Z_FINISH) == Z_OK)
1066 ; /* nothing */
1067 git_deflate_end(&s);
1069 /* Determine if we should auto-checkpoint. */
1070 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1071 || (pack_size + 60 + s.total_out) < pack_size) {
1073 /* This new object needs to *not* have the current pack_id. */
1074 e->pack_id = pack_id + 1;
1075 cycle_packfile();
1077 /* We cannot carry a delta into the new pack. */
1078 if (delta) {
1079 free(delta);
1080 delta = NULL;
1082 memset(&s, 0, sizeof(s));
1083 git_deflate_init(&s, pack_compression_level);
1084 s.next_in = (void *)dat->buf;
1085 s.avail_in = dat->len;
1086 s.avail_out = git_deflate_bound(&s, s.avail_in);
1087 s.next_out = out = xrealloc(out, s.avail_out);
1088 while (git_deflate(&s, Z_FINISH) == Z_OK)
1089 ; /* nothing */
1090 git_deflate_end(&s);
1094 e->type = type;
1095 e->pack_id = pack_id;
1096 e->idx.offset = pack_size;
1097 object_count++;
1098 object_count_by_type[type]++;
1100 crc32_begin(pack_file);
1102 if (delta) {
1103 off_t ofs = e->idx.offset - last->offset;
1104 unsigned pos = sizeof(hdr) - 1;
1106 delta_count_by_type[type]++;
1107 e->depth = last->depth + 1;
1109 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1110 sha1write(pack_file, hdr, hdrlen);
1111 pack_size += hdrlen;
1113 hdr[pos] = ofs & 127;
1114 while (ofs >>= 7)
1115 hdr[--pos] = 128 | (--ofs & 127);
1116 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1117 pack_size += sizeof(hdr) - pos;
1118 } else {
1119 e->depth = 0;
1120 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1121 sha1write(pack_file, hdr, hdrlen);
1122 pack_size += hdrlen;
1125 sha1write(pack_file, out, s.total_out);
1126 pack_size += s.total_out;
1128 e->idx.crc32 = crc32_end(pack_file);
1130 free(out);
1131 free(delta);
1132 if (last) {
1133 if (last->no_swap) {
1134 last->data = *dat;
1135 } else {
1136 strbuf_swap(&last->data, dat);
1138 last->offset = e->idx.offset;
1139 last->depth = e->depth;
1141 return 0;
1144 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1146 if (ftruncate(pack_data->pack_fd, to)
1147 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1148 die_errno("cannot truncate pack to skip duplicate");
1149 pack_size = to;
1151 /* yes this is a layering violation */
1152 pack_file->total = to;
1153 pack_file->offset = 0;
1154 pack_file->ctx = *ctx;
1157 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1159 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1160 unsigned char *in_buf = xmalloc(in_sz);
1161 unsigned char *out_buf = xmalloc(out_sz);
1162 struct object_entry *e;
1163 unsigned char sha1[20];
1164 unsigned long hdrlen;
1165 off_t offset;
1166 git_SHA_CTX c;
1167 git_SHA_CTX pack_file_ctx;
1168 git_zstream s;
1169 int status = Z_OK;
1171 /* Determine if we should auto-checkpoint. */
1172 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1173 || (pack_size + 60 + len) < pack_size)
1174 cycle_packfile();
1176 offset = pack_size;
1178 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1179 sha1flush(pack_file);
1180 pack_file_ctx = pack_file->ctx;
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 memset(&s, 0, sizeof(s));
1192 git_deflate_init(&s, pack_compression_level);
1194 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1195 if (out_sz <= hdrlen)
1196 die("impossibly large object header");
1198 s.next_out = out_buf + hdrlen;
1199 s.avail_out = out_sz - hdrlen;
1201 while (status != Z_STREAM_END) {
1202 if (0 < len && !s.avail_in) {
1203 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1204 size_t n = fread(in_buf, 1, cnt, stdin);
1205 if (!n && feof(stdin))
1206 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1208 git_SHA1_Update(&c, in_buf, n);
1209 s.next_in = in_buf;
1210 s.avail_in = n;
1211 len -= n;
1214 status = git_deflate(&s, len ? 0 : Z_FINISH);
1216 if (!s.avail_out || status == Z_STREAM_END) {
1217 size_t n = s.next_out - out_buf;
1218 sha1write(pack_file, out_buf, n);
1219 pack_size += n;
1220 s.next_out = out_buf;
1221 s.avail_out = out_sz;
1224 switch (status) {
1225 case Z_OK:
1226 case Z_BUF_ERROR:
1227 case Z_STREAM_END:
1228 continue;
1229 default:
1230 die("unexpected deflate failure: %d", status);
1233 git_deflate_end(&s);
1234 git_SHA1_Final(sha1, &c);
1236 if (sha1out)
1237 hashcpy(sha1out, sha1);
1239 e = insert_object(sha1);
1241 if (mark)
1242 insert_mark(mark, e);
1244 if (e->idx.offset) {
1245 duplicate_count_by_type[OBJ_BLOB]++;
1246 truncate_pack(offset, &pack_file_ctx);
1248 } else if (find_sha1_pack(sha1, packed_git)) {
1249 e->type = OBJ_BLOB;
1250 e->pack_id = MAX_PACK_ID;
1251 e->idx.offset = 1; /* just not zero! */
1252 duplicate_count_by_type[OBJ_BLOB]++;
1253 truncate_pack(offset, &pack_file_ctx);
1255 } else {
1256 e->depth = 0;
1257 e->type = OBJ_BLOB;
1258 e->pack_id = pack_id;
1259 e->idx.offset = offset;
1260 e->idx.crc32 = crc32_end(pack_file);
1261 object_count++;
1262 object_count_by_type[OBJ_BLOB]++;
1265 free(in_buf);
1266 free(out_buf);
1269 /* All calls must be guarded by find_object() or find_mark() to
1270 * ensure the 'struct object_entry' passed was written by this
1271 * process instance. We unpack the entry by the offset, avoiding
1272 * the need for the corresponding .idx file. This unpacking rule
1273 * works because we only use OBJ_REF_DELTA within the packfiles
1274 * created by fast-import.
1276 * oe must not be NULL. Such an oe usually comes from giving
1277 * an unknown SHA-1 to find_object() or an undefined mark to
1278 * find_mark(). Callers must test for this condition and use
1279 * the standard read_sha1_file() when it happens.
1281 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1282 * find_mark(), where the mark was reloaded from an existing marks
1283 * file and is referencing an object that this fast-import process
1284 * instance did not write out to a packfile. Callers must test for
1285 * this condition and use read_sha1_file() instead.
1287 static void *gfi_unpack_entry(
1288 struct object_entry *oe,
1289 unsigned long *sizep)
1291 enum object_type type;
1292 struct packed_git *p = all_packs[oe->pack_id];
1293 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1294 /* The object is stored in the packfile we are writing to
1295 * and we have modified it since the last time we scanned
1296 * back to read a previously written object. If an old
1297 * window covered [p->pack_size, p->pack_size + 20) its
1298 * data is stale and is not valid. Closing all windows
1299 * and updating the packfile length ensures we can read
1300 * the newly written data.
1302 close_pack_windows(p);
1303 sha1flush(pack_file);
1305 /* We have to offer 20 bytes additional on the end of
1306 * the packfile as the core unpacker code assumes the
1307 * footer is present at the file end and must promise
1308 * at least 20 bytes within any window it maps. But
1309 * we don't actually create the footer here.
1311 p->pack_size = pack_size + 20;
1313 return unpack_entry(p, oe->idx.offset, &type, sizep);
1316 static const char *get_mode(const char *str, uint16_t *modep)
1318 unsigned char c;
1319 uint16_t mode = 0;
1321 while ((c = *str++) != ' ') {
1322 if (c < '0' || c > '7')
1323 return NULL;
1324 mode = (mode << 3) + (c - '0');
1326 *modep = mode;
1327 return str;
1330 static void load_tree(struct tree_entry *root)
1332 unsigned char *sha1 = root->versions[1].sha1;
1333 struct object_entry *myoe;
1334 struct tree_content *t;
1335 unsigned long size;
1336 char *buf;
1337 const char *c;
1339 root->tree = t = new_tree_content(8);
1340 if (is_null_sha1(sha1))
1341 return;
1343 myoe = find_object(sha1);
1344 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1345 if (myoe->type != OBJ_TREE)
1346 die("Not a tree: %s", sha1_to_hex(sha1));
1347 t->delta_depth = myoe->depth;
1348 buf = gfi_unpack_entry(myoe, &size);
1349 if (!buf)
1350 die("Can't load tree %s", sha1_to_hex(sha1));
1351 } else {
1352 enum object_type type;
1353 buf = read_sha1_file(sha1, &type, &size);
1354 if (!buf || type != OBJ_TREE)
1355 die("Can't load tree %s", sha1_to_hex(sha1));
1358 c = buf;
1359 while (c != (buf + size)) {
1360 struct tree_entry *e = new_tree_entry();
1362 if (t->entry_count == t->entry_capacity)
1363 root->tree = t = grow_tree_content(t, t->entry_count);
1364 t->entries[t->entry_count++] = e;
1366 e->tree = NULL;
1367 c = get_mode(c, &e->versions[1].mode);
1368 if (!c)
1369 die("Corrupt mode in %s", sha1_to_hex(sha1));
1370 e->versions[0].mode = e->versions[1].mode;
1371 e->name = to_atom(c, strlen(c));
1372 c += e->name->str_len + 1;
1373 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1374 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1375 c += 20;
1377 free(buf);
1380 static int tecmp0 (const void *_a, const void *_b)
1382 struct tree_entry *a = *((struct tree_entry**)_a);
1383 struct tree_entry *b = *((struct tree_entry**)_b);
1384 return base_name_compare(
1385 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1386 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1389 static int tecmp1 (const void *_a, const void *_b)
1391 struct tree_entry *a = *((struct tree_entry**)_a);
1392 struct tree_entry *b = *((struct tree_entry**)_b);
1393 return base_name_compare(
1394 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1395 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1398 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1400 size_t maxlen = 0;
1401 unsigned int i;
1403 if (!v)
1404 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1405 else
1406 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1408 for (i = 0; i < t->entry_count; i++) {
1409 if (t->entries[i]->versions[v].mode)
1410 maxlen += t->entries[i]->name->str_len + 34;
1413 strbuf_reset(b);
1414 strbuf_grow(b, maxlen);
1415 for (i = 0; i < t->entry_count; i++) {
1416 struct tree_entry *e = t->entries[i];
1417 if (!e->versions[v].mode)
1418 continue;
1419 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
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 = root->tree;
1428 unsigned int i, j, del;
1429 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1430 struct object_entry *le;
1432 if (!is_null_sha1(root->versions[1].sha1))
1433 return;
1435 for (i = 0; i < t->entry_count; i++) {
1436 if (t->entries[i]->tree)
1437 store_tree(t->entries[i]);
1440 le = find_object(root->versions[0].sha1);
1441 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1442 mktree(t, 0, &old_tree);
1443 lo.data = old_tree;
1444 lo.offset = le->idx.offset;
1445 lo.depth = t->delta_depth;
1448 mktree(t, 1, &new_tree);
1449 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1451 t->delta_depth = lo.depth;
1452 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1453 struct tree_entry *e = t->entries[i];
1454 if (e->versions[1].mode) {
1455 e->versions[0].mode = e->versions[1].mode;
1456 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1457 t->entries[j++] = e;
1458 } else {
1459 release_tree_entry(e);
1460 del++;
1463 t->entry_count -= del;
1466 static void tree_content_replace(
1467 struct tree_entry *root,
1468 const unsigned char *sha1,
1469 const uint16_t mode,
1470 struct tree_content *newtree)
1472 if (!S_ISDIR(mode))
1473 die("Root cannot be a non-directory");
1474 hashcpy(root->versions[1].sha1, sha1);
1475 if (root->tree)
1476 release_tree_content_recursive(root->tree);
1477 root->tree = newtree;
1480 static int tree_content_set(
1481 struct tree_entry *root,
1482 const char *p,
1483 const unsigned char *sha1,
1484 const uint16_t mode,
1485 struct tree_content *subtree)
1487 struct tree_content *t;
1488 const char *slash1;
1489 unsigned int i, n;
1490 struct tree_entry *e;
1492 slash1 = strchr(p, '/');
1493 if (slash1)
1494 n = slash1 - p;
1495 else
1496 n = strlen(p);
1497 if (!n)
1498 die("Empty path component found in input");
1499 if (!slash1 && !S_ISDIR(mode) && subtree)
1500 die("Non-directories cannot have subtrees");
1502 if (!root->tree)
1503 load_tree(root);
1504 t = root->tree;
1505 for (i = 0; i < t->entry_count; i++) {
1506 e = t->entries[i];
1507 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1508 if (!slash1) {
1509 if (!S_ISDIR(mode)
1510 && e->versions[1].mode == mode
1511 && !hashcmp(e->versions[1].sha1, sha1))
1512 return 0;
1513 e->versions[1].mode = mode;
1514 hashcpy(e->versions[1].sha1, sha1);
1515 if (e->tree)
1516 release_tree_content_recursive(e->tree);
1517 e->tree = subtree;
1518 hashclr(root->versions[1].sha1);
1519 return 1;
1521 if (!S_ISDIR(e->versions[1].mode)) {
1522 e->tree = new_tree_content(8);
1523 e->versions[1].mode = S_IFDIR;
1525 if (!e->tree)
1526 load_tree(e);
1527 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1528 hashclr(root->versions[1].sha1);
1529 return 1;
1531 return 0;
1535 if (t->entry_count == t->entry_capacity)
1536 root->tree = t = grow_tree_content(t, t->entry_count);
1537 e = new_tree_entry();
1538 e->name = to_atom(p, n);
1539 e->versions[0].mode = 0;
1540 hashclr(e->versions[0].sha1);
1541 t->entries[t->entry_count++] = e;
1542 if (slash1) {
1543 e->tree = new_tree_content(8);
1544 e->versions[1].mode = S_IFDIR;
1545 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1546 } else {
1547 e->tree = subtree;
1548 e->versions[1].mode = mode;
1549 hashcpy(e->versions[1].sha1, sha1);
1551 hashclr(root->versions[1].sha1);
1552 return 1;
1555 static int tree_content_remove(
1556 struct tree_entry *root,
1557 const char *p,
1558 struct tree_entry *backup_leaf)
1560 struct tree_content *t;
1561 const char *slash1;
1562 unsigned int i, n;
1563 struct tree_entry *e;
1565 slash1 = strchr(p, '/');
1566 if (slash1)
1567 n = slash1 - p;
1568 else
1569 n = strlen(p);
1571 if (!root->tree)
1572 load_tree(root);
1573 t = root->tree;
1574 for (i = 0; i < t->entry_count; i++) {
1575 e = t->entries[i];
1576 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1577 if (slash1 && !S_ISDIR(e->versions[1].mode))
1579 * If p names a file in some subdirectory, and a
1580 * file or symlink matching the name of the
1581 * parent directory of p exists, then p cannot
1582 * exist and need not be deleted.
1584 return 1;
1585 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1586 goto del_entry;
1587 if (!e->tree)
1588 load_tree(e);
1589 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1590 for (n = 0; n < e->tree->entry_count; n++) {
1591 if (e->tree->entries[n]->versions[1].mode) {
1592 hashclr(root->versions[1].sha1);
1593 return 1;
1596 backup_leaf = NULL;
1597 goto del_entry;
1599 return 0;
1602 return 0;
1604 del_entry:
1605 if (backup_leaf)
1606 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1607 else if (e->tree)
1608 release_tree_content_recursive(e->tree);
1609 e->tree = NULL;
1610 e->versions[1].mode = 0;
1611 hashclr(e->versions[1].sha1);
1612 hashclr(root->versions[1].sha1);
1613 return 1;
1616 static int tree_content_get(
1617 struct tree_entry *root,
1618 const char *p,
1619 struct tree_entry *leaf)
1621 struct tree_content *t;
1622 const char *slash1;
1623 unsigned int i, n;
1624 struct tree_entry *e;
1626 slash1 = strchr(p, '/');
1627 if (slash1)
1628 n = slash1 - p;
1629 else
1630 n = strlen(p);
1632 if (!root->tree)
1633 load_tree(root);
1634 t = root->tree;
1635 for (i = 0; i < t->entry_count; i++) {
1636 e = t->entries[i];
1637 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1638 if (!slash1) {
1639 memcpy(leaf, e, sizeof(*leaf));
1640 if (e->tree && is_null_sha1(e->versions[1].sha1))
1641 leaf->tree = dup_tree_content(e->tree);
1642 else
1643 leaf->tree = NULL;
1644 return 1;
1646 if (!S_ISDIR(e->versions[1].mode))
1647 return 0;
1648 if (!e->tree)
1649 load_tree(e);
1650 return tree_content_get(e, slash1 + 1, leaf);
1653 return 0;
1656 static int update_branch(struct branch *b)
1658 static const char *msg = "fast-import";
1659 struct ref_lock *lock;
1660 unsigned char old_sha1[20];
1662 if (is_null_sha1(b->sha1))
1663 return 0;
1664 if (read_ref(b->name, old_sha1))
1665 hashclr(old_sha1);
1666 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1667 if (!lock)
1668 return error("Unable to lock %s", b->name);
1669 if (!force_update && !is_null_sha1(old_sha1)) {
1670 struct commit *old_cmit, *new_cmit;
1672 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1673 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1674 if (!old_cmit || !new_cmit) {
1675 unlock_ref(lock);
1676 return error("Branch %s is missing commits.", b->name);
1679 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1680 unlock_ref(lock);
1681 warning("Not updating %s"
1682 " (new tip %s does not contain %s)",
1683 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1684 return -1;
1687 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1688 return error("Unable to update %s", b->name);
1689 return 0;
1692 static void dump_branches(void)
1694 unsigned int i;
1695 struct branch *b;
1697 for (i = 0; i < branch_table_sz; i++) {
1698 for (b = branch_table[i]; b; b = b->table_next_branch)
1699 failure |= update_branch(b);
1703 static void dump_tags(void)
1705 static const char *msg = "fast-import";
1706 struct tag *t;
1707 struct ref_lock *lock;
1708 char ref_name[PATH_MAX];
1710 for (t = first_tag; t; t = t->next_tag) {
1711 sprintf(ref_name, "tags/%s", t->name);
1712 lock = lock_ref_sha1(ref_name, NULL);
1713 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1714 failure |= error("Unable to update %s", ref_name);
1718 static void dump_marks_helper(FILE *f,
1719 uintmax_t base,
1720 struct mark_set *m)
1722 uintmax_t k;
1723 if (m->shift) {
1724 for (k = 0; k < 1024; k++) {
1725 if (m->data.sets[k])
1726 dump_marks_helper(f, base + (k << m->shift),
1727 m->data.sets[k]);
1729 } else {
1730 for (k = 0; k < 1024; k++) {
1731 if (m->data.marked[k])
1732 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1733 sha1_to_hex(m->data.marked[k]->idx.sha1));
1738 static void dump_marks(void)
1740 static struct lock_file mark_lock;
1741 int mark_fd;
1742 FILE *f;
1744 if (!export_marks_file)
1745 return;
1747 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1748 if (mark_fd < 0) {
1749 failure |= error("Unable to write marks file %s: %s",
1750 export_marks_file, strerror(errno));
1751 return;
1754 f = fdopen(mark_fd, "w");
1755 if (!f) {
1756 int saved_errno = errno;
1757 rollback_lock_file(&mark_lock);
1758 failure |= error("Unable to write marks file %s: %s",
1759 export_marks_file, strerror(saved_errno));
1760 return;
1764 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1765 * Assign -1 to the lock file descriptor so that commit_lock_file()
1766 * won't try to close() it.
1768 mark_lock.fd = -1;
1770 dump_marks_helper(f, 0, marks);
1771 if (ferror(f) || fclose(f)) {
1772 int saved_errno = errno;
1773 rollback_lock_file(&mark_lock);
1774 failure |= error("Unable to write marks file %s: %s",
1775 export_marks_file, strerror(saved_errno));
1776 return;
1779 if (commit_lock_file(&mark_lock)) {
1780 int saved_errno = errno;
1781 rollback_lock_file(&mark_lock);
1782 failure |= error("Unable to commit marks file %s: %s",
1783 export_marks_file, strerror(saved_errno));
1784 return;
1788 static void read_marks(void)
1790 char line[512];
1791 FILE *f = fopen(import_marks_file, "r");
1792 if (f)
1794 else if (import_marks_file_ignore_missing && errno == ENOENT)
1795 return; /* Marks file does not exist */
1796 else
1797 die_errno("cannot read '%s'", import_marks_file);
1798 while (fgets(line, sizeof(line), f)) {
1799 uintmax_t mark;
1800 char *end;
1801 unsigned char sha1[20];
1802 struct object_entry *e;
1804 end = strchr(line, '\n');
1805 if (line[0] != ':' || !end)
1806 die("corrupt mark line: %s", line);
1807 *end = 0;
1808 mark = strtoumax(line + 1, &end, 10);
1809 if (!mark || end == line + 1
1810 || *end != ' ' || get_sha1(end + 1, sha1))
1811 die("corrupt mark line: %s", line);
1812 e = find_object(sha1);
1813 if (!e) {
1814 enum object_type type = sha1_object_info(sha1, NULL);
1815 if (type < 0)
1816 die("object not found: %s", sha1_to_hex(sha1));
1817 e = insert_object(sha1);
1818 e->type = type;
1819 e->pack_id = MAX_PACK_ID;
1820 e->idx.offset = 1; /* just not zero! */
1822 insert_mark(mark, e);
1824 fclose(f);
1828 static int read_next_command(void)
1830 static int stdin_eof = 0;
1832 if (stdin_eof) {
1833 unread_command_buf = 0;
1834 return EOF;
1837 for (;;) {
1838 if (unread_command_buf) {
1839 unread_command_buf = 0;
1840 } else {
1841 struct recent_command *rc;
1843 strbuf_detach(&command_buf, NULL);
1844 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1845 if (stdin_eof)
1846 return EOF;
1848 if (!seen_data_command
1849 && prefixcmp(command_buf.buf, "feature ")
1850 && prefixcmp(command_buf.buf, "option ")) {
1851 parse_argv();
1854 rc = rc_free;
1855 if (rc)
1856 rc_free = rc->next;
1857 else {
1858 rc = cmd_hist.next;
1859 cmd_hist.next = rc->next;
1860 cmd_hist.next->prev = &cmd_hist;
1861 free(rc->buf);
1864 rc->buf = command_buf.buf;
1865 rc->prev = cmd_tail;
1866 rc->next = cmd_hist.prev;
1867 rc->prev->next = rc;
1868 cmd_tail = rc;
1870 if (!prefixcmp(command_buf.buf, "cat-blob ")) {
1871 parse_cat_blob();
1872 continue;
1874 if (command_buf.buf[0] == '#')
1875 continue;
1876 return 0;
1880 static void skip_optional_lf(void)
1882 int term_char = fgetc(stdin);
1883 if (term_char != '\n' && term_char != EOF)
1884 ungetc(term_char, stdin);
1887 static void parse_mark(void)
1889 if (!prefixcmp(command_buf.buf, "mark :")) {
1890 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1891 read_next_command();
1893 else
1894 next_mark = 0;
1897 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1899 strbuf_reset(sb);
1901 if (prefixcmp(command_buf.buf, "data "))
1902 die("Expected 'data n' command, found: %s", command_buf.buf);
1904 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1905 char *term = xstrdup(command_buf.buf + 5 + 2);
1906 size_t term_len = command_buf.len - 5 - 2;
1908 strbuf_detach(&command_buf, NULL);
1909 for (;;) {
1910 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1911 die("EOF in data (terminator '%s' not found)", term);
1912 if (term_len == command_buf.len
1913 && !strcmp(term, command_buf.buf))
1914 break;
1915 strbuf_addbuf(sb, &command_buf);
1916 strbuf_addch(sb, '\n');
1918 free(term);
1920 else {
1921 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1922 size_t n = 0, length = (size_t)len;
1924 if (limit && limit < len) {
1925 *len_res = len;
1926 return 0;
1928 if (length < len)
1929 die("data is too large to use in this context");
1931 while (n < length) {
1932 size_t s = strbuf_fread(sb, length - n, stdin);
1933 if (!s && feof(stdin))
1934 die("EOF in data (%lu bytes remaining)",
1935 (unsigned long)(length - n));
1936 n += s;
1940 skip_optional_lf();
1941 return 1;
1944 static int validate_raw_date(const char *src, char *result, int maxlen)
1946 const char *orig_src = src;
1947 char *endp;
1948 unsigned long num;
1950 errno = 0;
1952 num = strtoul(src, &endp, 10);
1953 /* NEEDSWORK: perhaps check for reasonable values? */
1954 if (errno || endp == src || *endp != ' ')
1955 return -1;
1957 src = endp + 1;
1958 if (*src != '-' && *src != '+')
1959 return -1;
1961 num = strtoul(src + 1, &endp, 10);
1962 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1963 1400 < num)
1964 return -1;
1966 strcpy(result, orig_src);
1967 return 0;
1970 static char *parse_ident(const char *buf)
1972 const char *ltgt;
1973 size_t name_len;
1974 char *ident;
1976 /* ensure there is a space delimiter even if there is no name */
1977 if (*buf == '<')
1978 --buf;
1980 ltgt = buf + strcspn(buf, "<>");
1981 if (*ltgt != '<')
1982 die("Missing < in ident string: %s", buf);
1983 if (ltgt != buf && ltgt[-1] != ' ')
1984 die("Missing space before < in ident string: %s", buf);
1985 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
1986 if (*ltgt != '>')
1987 die("Missing > in ident string: %s", buf);
1988 ltgt++;
1989 if (*ltgt != ' ')
1990 die("Missing space after > in ident string: %s", buf);
1991 ltgt++;
1992 name_len = ltgt - buf;
1993 ident = xmalloc(name_len + 24);
1994 strncpy(ident, buf, name_len);
1996 switch (whenspec) {
1997 case WHENSPEC_RAW:
1998 if (validate_raw_date(ltgt, ident + name_len, 24) < 0)
1999 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2000 break;
2001 case WHENSPEC_RFC2822:
2002 if (parse_date(ltgt, ident + name_len, 24) < 0)
2003 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2004 break;
2005 case WHENSPEC_NOW:
2006 if (strcmp("now", ltgt))
2007 die("Date in ident must be 'now': %s", buf);
2008 datestamp(ident + name_len, 24);
2009 break;
2012 return ident;
2015 static void parse_and_store_blob(
2016 struct last_object *last,
2017 unsigned char *sha1out,
2018 uintmax_t mark)
2020 static struct strbuf buf = STRBUF_INIT;
2021 uintmax_t len;
2023 if (parse_data(&buf, big_file_threshold, &len))
2024 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2025 else {
2026 if (last) {
2027 strbuf_release(&last->data);
2028 last->offset = 0;
2029 last->depth = 0;
2031 stream_blob(len, sha1out, mark);
2032 skip_optional_lf();
2036 static void parse_new_blob(void)
2038 read_next_command();
2039 parse_mark();
2040 parse_and_store_blob(&last_blob, NULL, next_mark);
2043 static void unload_one_branch(void)
2045 while (cur_active_branches
2046 && cur_active_branches >= max_active_branches) {
2047 uintmax_t min_commit = ULONG_MAX;
2048 struct branch *e, *l = NULL, *p = NULL;
2050 for (e = active_branches; e; e = e->active_next_branch) {
2051 if (e->last_commit < min_commit) {
2052 p = l;
2053 min_commit = e->last_commit;
2055 l = e;
2058 if (p) {
2059 e = p->active_next_branch;
2060 p->active_next_branch = e->active_next_branch;
2061 } else {
2062 e = active_branches;
2063 active_branches = e->active_next_branch;
2065 e->active = 0;
2066 e->active_next_branch = NULL;
2067 if (e->branch_tree.tree) {
2068 release_tree_content_recursive(e->branch_tree.tree);
2069 e->branch_tree.tree = NULL;
2071 cur_active_branches--;
2075 static void load_branch(struct branch *b)
2077 load_tree(&b->branch_tree);
2078 if (!b->active) {
2079 b->active = 1;
2080 b->active_next_branch = active_branches;
2081 active_branches = b;
2082 cur_active_branches++;
2083 branch_load_count++;
2087 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2089 unsigned char fanout = 0;
2090 while ((num_notes >>= 8))
2091 fanout++;
2092 return fanout;
2095 static void construct_path_with_fanout(const char *hex_sha1,
2096 unsigned char fanout, char *path)
2098 unsigned int i = 0, j = 0;
2099 if (fanout >= 20)
2100 die("Too large fanout (%u)", fanout);
2101 while (fanout) {
2102 path[i++] = hex_sha1[j++];
2103 path[i++] = hex_sha1[j++];
2104 path[i++] = '/';
2105 fanout--;
2107 memcpy(path + i, hex_sha1 + j, 40 - j);
2108 path[i + 40 - j] = '\0';
2111 static uintmax_t do_change_note_fanout(
2112 struct tree_entry *orig_root, struct tree_entry *root,
2113 char *hex_sha1, unsigned int hex_sha1_len,
2114 char *fullpath, unsigned int fullpath_len,
2115 unsigned char fanout)
2117 struct tree_content *t = root->tree;
2118 struct tree_entry *e, leaf;
2119 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2120 uintmax_t num_notes = 0;
2121 unsigned char sha1[20];
2122 char realpath[60];
2124 for (i = 0; t && i < t->entry_count; i++) {
2125 e = t->entries[i];
2126 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2127 tmp_fullpath_len = fullpath_len;
2130 * We're interested in EITHER existing note entries (entries
2131 * with exactly 40 hex chars in path, not including directory
2132 * separators), OR directory entries that may contain note
2133 * entries (with < 40 hex chars in path).
2134 * Also, each path component in a note entry must be a multiple
2135 * of 2 chars.
2137 if (!e->versions[1].mode ||
2138 tmp_hex_sha1_len > 40 ||
2139 e->name->str_len % 2)
2140 continue;
2142 /* This _may_ be a note entry, or a subdir containing notes */
2143 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2144 e->name->str_len);
2145 if (tmp_fullpath_len)
2146 fullpath[tmp_fullpath_len++] = '/';
2147 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2148 e->name->str_len);
2149 tmp_fullpath_len += e->name->str_len;
2150 fullpath[tmp_fullpath_len] = '\0';
2152 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2153 /* This is a note entry */
2154 construct_path_with_fanout(hex_sha1, fanout, realpath);
2155 if (!strcmp(fullpath, realpath)) {
2156 /* Note entry is in correct location */
2157 num_notes++;
2158 continue;
2161 /* Rename fullpath to realpath */
2162 if (!tree_content_remove(orig_root, fullpath, &leaf))
2163 die("Failed to remove path %s", fullpath);
2164 tree_content_set(orig_root, realpath,
2165 leaf.versions[1].sha1,
2166 leaf.versions[1].mode,
2167 leaf.tree);
2168 } else if (S_ISDIR(e->versions[1].mode)) {
2169 /* This is a subdir that may contain note entries */
2170 if (!e->tree)
2171 load_tree(e);
2172 num_notes += do_change_note_fanout(orig_root, e,
2173 hex_sha1, tmp_hex_sha1_len,
2174 fullpath, tmp_fullpath_len, fanout);
2177 /* The above may have reallocated the current tree_content */
2178 t = root->tree;
2180 return num_notes;
2183 static uintmax_t change_note_fanout(struct tree_entry *root,
2184 unsigned char fanout)
2186 char hex_sha1[40], path[60];
2187 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2190 static void file_change_m(struct branch *b)
2192 const char *p = command_buf.buf + 2;
2193 static struct strbuf uq = STRBUF_INIT;
2194 const char *endp;
2195 struct object_entry *oe = oe;
2196 unsigned char sha1[20];
2197 uint16_t mode, inline_data = 0;
2199 p = get_mode(p, &mode);
2200 if (!p)
2201 die("Corrupt mode: %s", command_buf.buf);
2202 switch (mode) {
2203 case 0644:
2204 case 0755:
2205 mode |= S_IFREG;
2206 case S_IFREG | 0644:
2207 case S_IFREG | 0755:
2208 case S_IFLNK:
2209 case S_IFDIR:
2210 case S_IFGITLINK:
2211 /* ok */
2212 break;
2213 default:
2214 die("Corrupt mode: %s", command_buf.buf);
2217 if (*p == ':') {
2218 char *x;
2219 oe = find_mark(strtoumax(p + 1, &x, 10));
2220 hashcpy(sha1, oe->idx.sha1);
2221 p = x;
2222 } else if (!prefixcmp(p, "inline")) {
2223 inline_data = 1;
2224 p += 6;
2225 } else {
2226 if (get_sha1_hex(p, sha1))
2227 die("Invalid SHA1: %s", command_buf.buf);
2228 oe = find_object(sha1);
2229 p += 40;
2231 if (*p++ != ' ')
2232 die("Missing space after SHA1: %s", command_buf.buf);
2234 strbuf_reset(&uq);
2235 if (!unquote_c_style(&uq, p, &endp)) {
2236 if (*endp)
2237 die("Garbage after path in: %s", command_buf.buf);
2238 p = uq.buf;
2241 /* Git does not track empty, non-toplevel directories. */
2242 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2243 tree_content_remove(&b->branch_tree, p, NULL);
2244 return;
2247 if (S_ISGITLINK(mode)) {
2248 if (inline_data)
2249 die("Git links cannot be specified 'inline': %s",
2250 command_buf.buf);
2251 else if (oe) {
2252 if (oe->type != OBJ_COMMIT)
2253 die("Not a commit (actually a %s): %s",
2254 typename(oe->type), command_buf.buf);
2257 * Accept the sha1 without checking; it expected to be in
2258 * another repository.
2260 } else if (inline_data) {
2261 if (S_ISDIR(mode))
2262 die("Directories cannot be specified 'inline': %s",
2263 command_buf.buf);
2264 if (p != uq.buf) {
2265 strbuf_addstr(&uq, p);
2266 p = uq.buf;
2268 read_next_command();
2269 parse_and_store_blob(&last_blob, sha1, 0);
2270 } else {
2271 enum object_type expected = S_ISDIR(mode) ?
2272 OBJ_TREE: OBJ_BLOB;
2273 enum object_type type = oe ? oe->type :
2274 sha1_object_info(sha1, NULL);
2275 if (type < 0)
2276 die("%s not found: %s",
2277 S_ISDIR(mode) ? "Tree" : "Blob",
2278 command_buf.buf);
2279 if (type != expected)
2280 die("Not a %s (actually a %s): %s",
2281 typename(expected), typename(type),
2282 command_buf.buf);
2285 if (!*p) {
2286 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2287 return;
2289 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2292 static void file_change_d(struct branch *b)
2294 const char *p = command_buf.buf + 2;
2295 static struct strbuf uq = STRBUF_INIT;
2296 const char *endp;
2298 strbuf_reset(&uq);
2299 if (!unquote_c_style(&uq, p, &endp)) {
2300 if (*endp)
2301 die("Garbage after path in: %s", command_buf.buf);
2302 p = uq.buf;
2304 tree_content_remove(&b->branch_tree, p, NULL);
2307 static void file_change_cr(struct branch *b, int rename)
2309 const char *s, *d;
2310 static struct strbuf s_uq = STRBUF_INIT;
2311 static struct strbuf d_uq = STRBUF_INIT;
2312 const char *endp;
2313 struct tree_entry leaf;
2315 s = command_buf.buf + 2;
2316 strbuf_reset(&s_uq);
2317 if (!unquote_c_style(&s_uq, s, &endp)) {
2318 if (*endp != ' ')
2319 die("Missing space after source: %s", command_buf.buf);
2320 } else {
2321 endp = strchr(s, ' ');
2322 if (!endp)
2323 die("Missing space after source: %s", command_buf.buf);
2324 strbuf_add(&s_uq, s, endp - s);
2326 s = s_uq.buf;
2328 endp++;
2329 if (!*endp)
2330 die("Missing dest: %s", command_buf.buf);
2332 d = endp;
2333 strbuf_reset(&d_uq);
2334 if (!unquote_c_style(&d_uq, d, &endp)) {
2335 if (*endp)
2336 die("Garbage after dest in: %s", command_buf.buf);
2337 d = d_uq.buf;
2340 memset(&leaf, 0, sizeof(leaf));
2341 if (rename)
2342 tree_content_remove(&b->branch_tree, s, &leaf);
2343 else
2344 tree_content_get(&b->branch_tree, s, &leaf);
2345 if (!leaf.versions[1].mode)
2346 die("Path %s not in branch", s);
2347 if (!*d) { /* C "path/to/subdir" "" */
2348 tree_content_replace(&b->branch_tree,
2349 leaf.versions[1].sha1,
2350 leaf.versions[1].mode,
2351 leaf.tree);
2352 return;
2354 tree_content_set(&b->branch_tree, d,
2355 leaf.versions[1].sha1,
2356 leaf.versions[1].mode,
2357 leaf.tree);
2360 static void note_change_n(struct branch *b, unsigned char old_fanout)
2362 const char *p = command_buf.buf + 2;
2363 static struct strbuf uq = STRBUF_INIT;
2364 struct object_entry *oe = oe;
2365 struct branch *s;
2366 unsigned char sha1[20], commit_sha1[20];
2367 char path[60];
2368 uint16_t inline_data = 0;
2369 unsigned char new_fanout;
2371 /* <dataref> or 'inline' */
2372 if (*p == ':') {
2373 char *x;
2374 oe = find_mark(strtoumax(p + 1, &x, 10));
2375 hashcpy(sha1, oe->idx.sha1);
2376 p = x;
2377 } else if (!prefixcmp(p, "inline")) {
2378 inline_data = 1;
2379 p += 6;
2380 } else {
2381 if (get_sha1_hex(p, sha1))
2382 die("Invalid SHA1: %s", command_buf.buf);
2383 oe = find_object(sha1);
2384 p += 40;
2386 if (*p++ != ' ')
2387 die("Missing space after SHA1: %s", command_buf.buf);
2389 /* <committish> */
2390 s = lookup_branch(p);
2391 if (s) {
2392 hashcpy(commit_sha1, s->sha1);
2393 } else if (*p == ':') {
2394 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2395 struct object_entry *commit_oe = find_mark(commit_mark);
2396 if (commit_oe->type != OBJ_COMMIT)
2397 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2398 hashcpy(commit_sha1, commit_oe->idx.sha1);
2399 } else if (!get_sha1(p, commit_sha1)) {
2400 unsigned long size;
2401 char *buf = read_object_with_reference(commit_sha1,
2402 commit_type, &size, commit_sha1);
2403 if (!buf || size < 46)
2404 die("Not a valid commit: %s", p);
2405 free(buf);
2406 } else
2407 die("Invalid ref name or SHA1 expression: %s", p);
2409 if (inline_data) {
2410 if (p != uq.buf) {
2411 strbuf_addstr(&uq, p);
2412 p = uq.buf;
2414 read_next_command();
2415 parse_and_store_blob(&last_blob, sha1, 0);
2416 } else if (oe) {
2417 if (oe->type != OBJ_BLOB)
2418 die("Not a blob (actually a %s): %s",
2419 typename(oe->type), command_buf.buf);
2420 } else if (!is_null_sha1(sha1)) {
2421 enum object_type type = sha1_object_info(sha1, NULL);
2422 if (type < 0)
2423 die("Blob not found: %s", command_buf.buf);
2424 if (type != OBJ_BLOB)
2425 die("Not a blob (actually a %s): %s",
2426 typename(type), command_buf.buf);
2429 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2430 if (tree_content_remove(&b->branch_tree, path, NULL))
2431 b->num_notes--;
2433 if (is_null_sha1(sha1))
2434 return; /* nothing to insert */
2436 b->num_notes++;
2437 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2438 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2439 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2442 static void file_change_deleteall(struct branch *b)
2444 release_tree_content_recursive(b->branch_tree.tree);
2445 hashclr(b->branch_tree.versions[0].sha1);
2446 hashclr(b->branch_tree.versions[1].sha1);
2447 load_tree(&b->branch_tree);
2448 b->num_notes = 0;
2451 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2453 if (!buf || size < 46)
2454 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2455 if (memcmp("tree ", buf, 5)
2456 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2457 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2458 hashcpy(b->branch_tree.versions[0].sha1,
2459 b->branch_tree.versions[1].sha1);
2462 static void parse_from_existing(struct branch *b)
2464 if (is_null_sha1(b->sha1)) {
2465 hashclr(b->branch_tree.versions[0].sha1);
2466 hashclr(b->branch_tree.versions[1].sha1);
2467 } else {
2468 unsigned long size;
2469 char *buf;
2471 buf = read_object_with_reference(b->sha1,
2472 commit_type, &size, b->sha1);
2473 parse_from_commit(b, buf, size);
2474 free(buf);
2478 static int parse_from(struct branch *b)
2480 const char *from;
2481 struct branch *s;
2483 if (prefixcmp(command_buf.buf, "from "))
2484 return 0;
2486 if (b->branch_tree.tree) {
2487 release_tree_content_recursive(b->branch_tree.tree);
2488 b->branch_tree.tree = NULL;
2491 from = strchr(command_buf.buf, ' ') + 1;
2492 s = lookup_branch(from);
2493 if (b == s)
2494 die("Can't create a branch from itself: %s", b->name);
2495 else if (s) {
2496 unsigned char *t = s->branch_tree.versions[1].sha1;
2497 hashcpy(b->sha1, s->sha1);
2498 hashcpy(b->branch_tree.versions[0].sha1, t);
2499 hashcpy(b->branch_tree.versions[1].sha1, t);
2500 } else if (*from == ':') {
2501 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2502 struct object_entry *oe = find_mark(idnum);
2503 if (oe->type != OBJ_COMMIT)
2504 die("Mark :%" PRIuMAX " not a commit", idnum);
2505 hashcpy(b->sha1, oe->idx.sha1);
2506 if (oe->pack_id != MAX_PACK_ID) {
2507 unsigned long size;
2508 char *buf = gfi_unpack_entry(oe, &size);
2509 parse_from_commit(b, buf, size);
2510 free(buf);
2511 } else
2512 parse_from_existing(b);
2513 } else if (!get_sha1(from, b->sha1))
2514 parse_from_existing(b);
2515 else
2516 die("Invalid ref name or SHA1 expression: %s", from);
2518 read_next_command();
2519 return 1;
2522 static struct hash_list *parse_merge(unsigned int *count)
2524 struct hash_list *list = NULL, *n, *e = e;
2525 const char *from;
2526 struct branch *s;
2528 *count = 0;
2529 while (!prefixcmp(command_buf.buf, "merge ")) {
2530 from = strchr(command_buf.buf, ' ') + 1;
2531 n = xmalloc(sizeof(*n));
2532 s = lookup_branch(from);
2533 if (s)
2534 hashcpy(n->sha1, s->sha1);
2535 else if (*from == ':') {
2536 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2537 struct object_entry *oe = find_mark(idnum);
2538 if (oe->type != OBJ_COMMIT)
2539 die("Mark :%" PRIuMAX " not a commit", idnum);
2540 hashcpy(n->sha1, oe->idx.sha1);
2541 } else if (!get_sha1(from, n->sha1)) {
2542 unsigned long size;
2543 char *buf = read_object_with_reference(n->sha1,
2544 commit_type, &size, n->sha1);
2545 if (!buf || size < 46)
2546 die("Not a valid commit: %s", from);
2547 free(buf);
2548 } else
2549 die("Invalid ref name or SHA1 expression: %s", from);
2551 n->next = NULL;
2552 if (list)
2553 e->next = n;
2554 else
2555 list = n;
2556 e = n;
2557 (*count)++;
2558 read_next_command();
2560 return list;
2563 static void parse_new_commit(void)
2565 static struct strbuf msg = STRBUF_INIT;
2566 struct branch *b;
2567 char *sp;
2568 char *author = NULL;
2569 char *committer = NULL;
2570 struct hash_list *merge_list = NULL;
2571 unsigned int merge_count;
2572 unsigned char prev_fanout, new_fanout;
2574 /* Obtain the branch name from the rest of our command */
2575 sp = strchr(command_buf.buf, ' ') + 1;
2576 b = lookup_branch(sp);
2577 if (!b)
2578 b = new_branch(sp);
2580 read_next_command();
2581 parse_mark();
2582 if (!prefixcmp(command_buf.buf, "author ")) {
2583 author = parse_ident(command_buf.buf + 7);
2584 read_next_command();
2586 if (!prefixcmp(command_buf.buf, "committer ")) {
2587 committer = parse_ident(command_buf.buf + 10);
2588 read_next_command();
2590 if (!committer)
2591 die("Expected committer but didn't get one");
2592 parse_data(&msg, 0, NULL);
2593 read_next_command();
2594 parse_from(b);
2595 merge_list = parse_merge(&merge_count);
2597 /* ensure the branch is active/loaded */
2598 if (!b->branch_tree.tree || !max_active_branches) {
2599 unload_one_branch();
2600 load_branch(b);
2603 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2605 /* file_change* */
2606 while (command_buf.len > 0) {
2607 if (!prefixcmp(command_buf.buf, "M "))
2608 file_change_m(b);
2609 else if (!prefixcmp(command_buf.buf, "D "))
2610 file_change_d(b);
2611 else if (!prefixcmp(command_buf.buf, "R "))
2612 file_change_cr(b, 1);
2613 else if (!prefixcmp(command_buf.buf, "C "))
2614 file_change_cr(b, 0);
2615 else if (!prefixcmp(command_buf.buf, "N "))
2616 note_change_n(b, prev_fanout);
2617 else if (!strcmp("deleteall", command_buf.buf))
2618 file_change_deleteall(b);
2619 else if (!prefixcmp(command_buf.buf, "ls "))
2620 parse_ls(b);
2621 else {
2622 unread_command_buf = 1;
2623 break;
2625 if (read_next_command() == EOF)
2626 break;
2629 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2630 if (new_fanout != prev_fanout)
2631 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2633 /* build the tree and the commit */
2634 store_tree(&b->branch_tree);
2635 hashcpy(b->branch_tree.versions[0].sha1,
2636 b->branch_tree.versions[1].sha1);
2638 strbuf_reset(&new_data);
2639 strbuf_addf(&new_data, "tree %s\n",
2640 sha1_to_hex(b->branch_tree.versions[1].sha1));
2641 if (!is_null_sha1(b->sha1))
2642 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2643 while (merge_list) {
2644 struct hash_list *next = merge_list->next;
2645 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2646 free(merge_list);
2647 merge_list = next;
2649 strbuf_addf(&new_data,
2650 "author %s\n"
2651 "committer %s\n"
2652 "\n",
2653 author ? author : committer, committer);
2654 strbuf_addbuf(&new_data, &msg);
2655 free(author);
2656 free(committer);
2658 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2659 b->pack_id = pack_id;
2660 b->last_commit = object_count_by_type[OBJ_COMMIT];
2663 static void parse_new_tag(void)
2665 static struct strbuf msg = STRBUF_INIT;
2666 char *sp;
2667 const char *from;
2668 char *tagger;
2669 struct branch *s;
2670 struct tag *t;
2671 uintmax_t from_mark = 0;
2672 unsigned char sha1[20];
2673 enum object_type type;
2675 /* Obtain the new tag name from the rest of our command */
2676 sp = strchr(command_buf.buf, ' ') + 1;
2677 t = pool_alloc(sizeof(struct tag));
2678 t->next_tag = NULL;
2679 t->name = pool_strdup(sp);
2680 if (last_tag)
2681 last_tag->next_tag = t;
2682 else
2683 first_tag = t;
2684 last_tag = t;
2685 read_next_command();
2687 /* from ... */
2688 if (prefixcmp(command_buf.buf, "from "))
2689 die("Expected from command, got %s", command_buf.buf);
2690 from = strchr(command_buf.buf, ' ') + 1;
2691 s = lookup_branch(from);
2692 if (s) {
2693 hashcpy(sha1, s->sha1);
2694 type = OBJ_COMMIT;
2695 } else if (*from == ':') {
2696 struct object_entry *oe;
2697 from_mark = strtoumax(from + 1, NULL, 10);
2698 oe = find_mark(from_mark);
2699 type = oe->type;
2700 hashcpy(sha1, oe->idx.sha1);
2701 } else if (!get_sha1(from, sha1)) {
2702 unsigned long size;
2703 char *buf;
2705 buf = read_sha1_file(sha1, &type, &size);
2706 if (!buf || size < 46)
2707 die("Not a valid commit: %s", from);
2708 free(buf);
2709 } else
2710 die("Invalid ref name or SHA1 expression: %s", from);
2711 read_next_command();
2713 /* tagger ... */
2714 if (!prefixcmp(command_buf.buf, "tagger ")) {
2715 tagger = parse_ident(command_buf.buf + 7);
2716 read_next_command();
2717 } else
2718 tagger = NULL;
2720 /* tag payload/message */
2721 parse_data(&msg, 0, NULL);
2723 /* build the tag object */
2724 strbuf_reset(&new_data);
2726 strbuf_addf(&new_data,
2727 "object %s\n"
2728 "type %s\n"
2729 "tag %s\n",
2730 sha1_to_hex(sha1), typename(type), t->name);
2731 if (tagger)
2732 strbuf_addf(&new_data,
2733 "tagger %s\n", tagger);
2734 strbuf_addch(&new_data, '\n');
2735 strbuf_addbuf(&new_data, &msg);
2736 free(tagger);
2738 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2739 t->pack_id = MAX_PACK_ID;
2740 else
2741 t->pack_id = pack_id;
2744 static void parse_reset_branch(void)
2746 struct branch *b;
2747 char *sp;
2749 /* Obtain the branch name from the rest of our command */
2750 sp = strchr(command_buf.buf, ' ') + 1;
2751 b = lookup_branch(sp);
2752 if (b) {
2753 hashclr(b->sha1);
2754 hashclr(b->branch_tree.versions[0].sha1);
2755 hashclr(b->branch_tree.versions[1].sha1);
2756 if (b->branch_tree.tree) {
2757 release_tree_content_recursive(b->branch_tree.tree);
2758 b->branch_tree.tree = NULL;
2761 else
2762 b = new_branch(sp);
2763 read_next_command();
2764 parse_from(b);
2765 if (command_buf.len > 0)
2766 unread_command_buf = 1;
2769 static void cat_blob_write(const char *buf, unsigned long size)
2771 if (write_in_full(cat_blob_fd, buf, size) != size)
2772 die_errno("Write to frontend failed");
2775 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2777 struct strbuf line = STRBUF_INIT;
2778 unsigned long size;
2779 enum object_type type = 0;
2780 char *buf;
2782 if (!oe || oe->pack_id == MAX_PACK_ID) {
2783 buf = read_sha1_file(sha1, &type, &size);
2784 } else {
2785 type = oe->type;
2786 buf = gfi_unpack_entry(oe, &size);
2790 * Output based on batch_one_object() from cat-file.c.
2792 if (type <= 0) {
2793 strbuf_reset(&line);
2794 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2795 cat_blob_write(line.buf, line.len);
2796 strbuf_release(&line);
2797 free(buf);
2798 return;
2800 if (!buf)
2801 die("Can't read object %s", sha1_to_hex(sha1));
2802 if (type != OBJ_BLOB)
2803 die("Object %s is a %s but a blob was expected.",
2804 sha1_to_hex(sha1), typename(type));
2805 strbuf_reset(&line);
2806 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2807 typename(type), size);
2808 cat_blob_write(line.buf, line.len);
2809 strbuf_release(&line);
2810 cat_blob_write(buf, size);
2811 cat_blob_write("\n", 1);
2812 free(buf);
2815 static void parse_cat_blob(void)
2817 const char *p;
2818 struct object_entry *oe = oe;
2819 unsigned char sha1[20];
2821 /* cat-blob SP <object> LF */
2822 p = command_buf.buf + strlen("cat-blob ");
2823 if (*p == ':') {
2824 char *x;
2825 oe = find_mark(strtoumax(p + 1, &x, 10));
2826 if (x == p + 1)
2827 die("Invalid mark: %s", command_buf.buf);
2828 if (!oe)
2829 die("Unknown mark: %s", command_buf.buf);
2830 if (*x)
2831 die("Garbage after mark: %s", command_buf.buf);
2832 hashcpy(sha1, oe->idx.sha1);
2833 } else {
2834 if (get_sha1_hex(p, sha1))
2835 die("Invalid SHA1: %s", command_buf.buf);
2836 if (p[40])
2837 die("Garbage after SHA1: %s", command_buf.buf);
2838 oe = find_object(sha1);
2841 cat_blob(oe, sha1);
2844 static struct object_entry *dereference(struct object_entry *oe,
2845 unsigned char sha1[20])
2847 unsigned long size;
2848 char *buf = NULL;
2849 if (!oe) {
2850 enum object_type type = sha1_object_info(sha1, NULL);
2851 if (type < 0)
2852 die("object not found: %s", sha1_to_hex(sha1));
2853 /* cache it! */
2854 oe = insert_object(sha1);
2855 oe->type = type;
2856 oe->pack_id = MAX_PACK_ID;
2857 oe->idx.offset = 1;
2859 switch (oe->type) {
2860 case OBJ_TREE: /* easy case. */
2861 return oe;
2862 case OBJ_COMMIT:
2863 case OBJ_TAG:
2864 break;
2865 default:
2866 die("Not a treeish: %s", command_buf.buf);
2869 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2870 buf = gfi_unpack_entry(oe, &size);
2871 } else {
2872 enum object_type unused;
2873 buf = read_sha1_file(sha1, &unused, &size);
2875 if (!buf)
2876 die("Can't load object %s", sha1_to_hex(sha1));
2878 /* Peel one layer. */
2879 switch (oe->type) {
2880 case OBJ_TAG:
2881 if (size < 40 + strlen("object ") ||
2882 get_sha1_hex(buf + strlen("object "), sha1))
2883 die("Invalid SHA1 in tag: %s", command_buf.buf);
2884 break;
2885 case OBJ_COMMIT:
2886 if (size < 40 + strlen("tree ") ||
2887 get_sha1_hex(buf + strlen("tree "), sha1))
2888 die("Invalid SHA1 in commit: %s", command_buf.buf);
2891 free(buf);
2892 return find_object(sha1);
2895 static struct object_entry *parse_treeish_dataref(const char **p)
2897 unsigned char sha1[20];
2898 struct object_entry *e;
2900 if (**p == ':') { /* <mark> */
2901 char *endptr;
2902 e = find_mark(strtoumax(*p + 1, &endptr, 10));
2903 if (endptr == *p + 1)
2904 die("Invalid mark: %s", command_buf.buf);
2905 if (!e)
2906 die("Unknown mark: %s", command_buf.buf);
2907 *p = endptr;
2908 hashcpy(sha1, e->idx.sha1);
2909 } else { /* <sha1> */
2910 if (get_sha1_hex(*p, sha1))
2911 die("Invalid SHA1: %s", command_buf.buf);
2912 e = find_object(sha1);
2913 *p += 40;
2916 while (!e || e->type != OBJ_TREE)
2917 e = dereference(e, sha1);
2918 return e;
2921 static void print_ls(int mode, const unsigned char *sha1, const char *path)
2923 static struct strbuf line = STRBUF_INIT;
2925 /* See show_tree(). */
2926 const char *type =
2927 S_ISGITLINK(mode) ? commit_type :
2928 S_ISDIR(mode) ? tree_type :
2929 blob_type;
2931 if (!mode) {
2932 /* missing SP path LF */
2933 strbuf_reset(&line);
2934 strbuf_addstr(&line, "missing ");
2935 quote_c_style(path, &line, NULL, 0);
2936 strbuf_addch(&line, '\n');
2937 } else {
2938 /* mode SP type SP object_name TAB path LF */
2939 strbuf_reset(&line);
2940 strbuf_addf(&line, "%06o %s %s\t",
2941 mode, type, sha1_to_hex(sha1));
2942 quote_c_style(path, &line, NULL, 0);
2943 strbuf_addch(&line, '\n');
2945 cat_blob_write(line.buf, line.len);
2948 static void parse_ls(struct branch *b)
2950 const char *p;
2951 struct tree_entry *root = NULL;
2952 struct tree_entry leaf = {NULL};
2954 /* ls SP (<treeish> SP)? <path> */
2955 p = command_buf.buf + strlen("ls ");
2956 if (*p == '"') {
2957 if (!b)
2958 die("Not in a commit: %s", command_buf.buf);
2959 root = &b->branch_tree;
2960 } else {
2961 struct object_entry *e = parse_treeish_dataref(&p);
2962 root = new_tree_entry();
2963 hashcpy(root->versions[1].sha1, e->idx.sha1);
2964 load_tree(root);
2965 if (*p++ != ' ')
2966 die("Missing space after tree-ish: %s", command_buf.buf);
2968 if (*p == '"') {
2969 static struct strbuf uq = STRBUF_INIT;
2970 const char *endp;
2971 strbuf_reset(&uq);
2972 if (unquote_c_style(&uq, p, &endp))
2973 die("Invalid path: %s", command_buf.buf);
2974 if (*endp)
2975 die("Garbage after path in: %s", command_buf.buf);
2976 p = uq.buf;
2978 tree_content_get(root, p, &leaf);
2980 * A directory in preparation would have a sha1 of zero
2981 * until it is saved. Save, for simplicity.
2983 if (S_ISDIR(leaf.versions[1].mode))
2984 store_tree(&leaf);
2986 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
2987 if (!b || root != &b->branch_tree)
2988 release_tree_entry(root);
2991 static void checkpoint(void)
2993 checkpoint_requested = 0;
2994 if (object_count) {
2995 cycle_packfile();
2996 dump_branches();
2997 dump_tags();
2998 dump_marks();
3002 static void parse_checkpoint(void)
3004 checkpoint_requested = 1;
3005 skip_optional_lf();
3008 static void parse_progress(void)
3010 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3011 fputc('\n', stdout);
3012 fflush(stdout);
3013 skip_optional_lf();
3016 static char* make_fast_import_path(const char *path)
3018 struct strbuf abs_path = STRBUF_INIT;
3020 if (!relative_marks_paths || is_absolute_path(path))
3021 return xstrdup(path);
3022 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
3023 return strbuf_detach(&abs_path, NULL);
3026 static void option_import_marks(const char *marks,
3027 int from_stream, int ignore_missing)
3029 if (import_marks_file) {
3030 if (from_stream)
3031 die("Only one import-marks command allowed per stream");
3033 /* read previous mark file */
3034 if(!import_marks_file_from_stream)
3035 read_marks();
3038 import_marks_file = make_fast_import_path(marks);
3039 safe_create_leading_directories_const(import_marks_file);
3040 import_marks_file_from_stream = from_stream;
3041 import_marks_file_ignore_missing = ignore_missing;
3044 static void option_date_format(const char *fmt)
3046 if (!strcmp(fmt, "raw"))
3047 whenspec = WHENSPEC_RAW;
3048 else if (!strcmp(fmt, "rfc2822"))
3049 whenspec = WHENSPEC_RFC2822;
3050 else if (!strcmp(fmt, "now"))
3051 whenspec = WHENSPEC_NOW;
3052 else
3053 die("unknown --date-format argument %s", fmt);
3056 static unsigned long ulong_arg(const char *option, const char *arg)
3058 char *endptr;
3059 unsigned long rv = strtoul(arg, &endptr, 0);
3060 if (strchr(arg, '-') || endptr == arg || *endptr)
3061 die("%s: argument must be a non-negative integer", option);
3062 return rv;
3065 static void option_depth(const char *depth)
3067 max_depth = ulong_arg("--depth", depth);
3068 if (max_depth > MAX_DEPTH)
3069 die("--depth cannot exceed %u", MAX_DEPTH);
3072 static void option_active_branches(const char *branches)
3074 max_active_branches = ulong_arg("--active-branches", branches);
3077 static void option_export_marks(const char *marks)
3079 export_marks_file = make_fast_import_path(marks);
3080 safe_create_leading_directories_const(export_marks_file);
3083 static void option_cat_blob_fd(const char *fd)
3085 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3086 if (n > (unsigned long) INT_MAX)
3087 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3088 cat_blob_fd = (int) n;
3091 static void option_export_pack_edges(const char *edges)
3093 if (pack_edges)
3094 fclose(pack_edges);
3095 pack_edges = fopen(edges, "a");
3096 if (!pack_edges)
3097 die_errno("Cannot open '%s'", edges);
3100 static int parse_one_option(const char *option)
3102 if (!prefixcmp(option, "max-pack-size=")) {
3103 unsigned long v;
3104 if (!git_parse_ulong(option + 14, &v))
3105 return 0;
3106 if (v < 8192) {
3107 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3108 v *= 1024 * 1024;
3109 } else if (v < 1024 * 1024) {
3110 warning("minimum max-pack-size is 1 MiB");
3111 v = 1024 * 1024;
3113 max_packsize = v;
3114 } else if (!prefixcmp(option, "big-file-threshold=")) {
3115 unsigned long v;
3116 if (!git_parse_ulong(option + 19, &v))
3117 return 0;
3118 big_file_threshold = v;
3119 } else if (!prefixcmp(option, "depth=")) {
3120 option_depth(option + 6);
3121 } else if (!prefixcmp(option, "active-branches=")) {
3122 option_active_branches(option + 16);
3123 } else if (!prefixcmp(option, "export-pack-edges=")) {
3124 option_export_pack_edges(option + 18);
3125 } else if (!prefixcmp(option, "quiet")) {
3126 show_stats = 0;
3127 } else if (!prefixcmp(option, "stats")) {
3128 show_stats = 1;
3129 } else {
3130 return 0;
3133 return 1;
3136 static int parse_one_feature(const char *feature, int from_stream)
3138 if (!prefixcmp(feature, "date-format=")) {
3139 option_date_format(feature + 12);
3140 } else if (!prefixcmp(feature, "import-marks=")) {
3141 option_import_marks(feature + 13, from_stream, 0);
3142 } else if (!prefixcmp(feature, "import-marks-if-exists=")) {
3143 option_import_marks(feature + strlen("import-marks-if-exists="),
3144 from_stream, 1);
3145 } else if (!prefixcmp(feature, "export-marks=")) {
3146 option_export_marks(feature + 13);
3147 } else if (!strcmp(feature, "cat-blob")) {
3148 ; /* Don't die - this feature is supported */
3149 } else if (!strcmp(feature, "relative-marks")) {
3150 relative_marks_paths = 1;
3151 } else if (!strcmp(feature, "no-relative-marks")) {
3152 relative_marks_paths = 0;
3153 } else if (!strcmp(feature, "done")) {
3154 require_explicit_termination = 1;
3155 } else if (!strcmp(feature, "force")) {
3156 force_update = 1;
3157 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3158 ; /* do nothing; we have the feature */
3159 } else {
3160 return 0;
3163 return 1;
3166 static void parse_feature(void)
3168 char *feature = command_buf.buf + 8;
3170 if (seen_data_command)
3171 die("Got feature command '%s' after data command", feature);
3173 if (parse_one_feature(feature, 1))
3174 return;
3176 die("This version of fast-import does not support feature %s.", feature);
3179 static void parse_option(void)
3181 char *option = command_buf.buf + 11;
3183 if (seen_data_command)
3184 die("Got option command '%s' after data command", option);
3186 if (parse_one_option(option))
3187 return;
3189 die("This version of fast-import does not support option: %s", option);
3192 static int git_pack_config(const char *k, const char *v, void *cb)
3194 if (!strcmp(k, "pack.depth")) {
3195 max_depth = git_config_int(k, v);
3196 if (max_depth > MAX_DEPTH)
3197 max_depth = MAX_DEPTH;
3198 return 0;
3200 if (!strcmp(k, "pack.compression")) {
3201 int level = git_config_int(k, v);
3202 if (level == -1)
3203 level = Z_DEFAULT_COMPRESSION;
3204 else if (level < 0 || level > Z_BEST_COMPRESSION)
3205 die("bad pack compression level %d", level);
3206 pack_compression_level = level;
3207 pack_compression_seen = 1;
3208 return 0;
3210 if (!strcmp(k, "pack.indexversion")) {
3211 pack_idx_opts.version = git_config_int(k, v);
3212 if (pack_idx_opts.version > 2)
3213 die("bad pack.indexversion=%"PRIu32,
3214 pack_idx_opts.version);
3215 return 0;
3217 if (!strcmp(k, "pack.packsizelimit")) {
3218 max_packsize = git_config_ulong(k, v);
3219 return 0;
3221 return git_default_config(k, v, cb);
3224 static const char fast_import_usage[] =
3225 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3227 static void parse_argv(void)
3229 unsigned int i;
3231 for (i = 1; i < global_argc; i++) {
3232 const char *a = global_argv[i];
3234 if (*a != '-' || !strcmp(a, "--"))
3235 break;
3237 if (parse_one_option(a + 2))
3238 continue;
3240 if (parse_one_feature(a + 2, 0))
3241 continue;
3243 if (!prefixcmp(a + 2, "cat-blob-fd=")) {
3244 option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
3245 continue;
3248 die("unknown option %s", a);
3250 if (i != global_argc)
3251 usage(fast_import_usage);
3253 seen_data_command = 1;
3254 if (import_marks_file)
3255 read_marks();
3258 int main(int argc, const char **argv)
3260 unsigned int i;
3262 git_extract_argv0_path(argv[0]);
3264 if (argc == 2 && !strcmp(argv[1], "-h"))
3265 usage(fast_import_usage);
3267 setup_git_directory();
3268 reset_pack_idx_option(&pack_idx_opts);
3269 git_config(git_pack_config, NULL);
3270 if (!pack_compression_seen && core_compression_seen)
3271 pack_compression_level = core_compression_level;
3273 alloc_objects(object_entry_alloc);
3274 strbuf_init(&command_buf, 0);
3275 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3276 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3277 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3278 marks = pool_calloc(1, sizeof(struct mark_set));
3280 global_argc = argc;
3281 global_argv = argv;
3283 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3284 for (i = 0; i < (cmd_save - 1); i++)
3285 rc_free[i].next = &rc_free[i + 1];
3286 rc_free[cmd_save - 1].next = NULL;
3288 prepare_packed_git();
3289 start_packfile();
3290 set_die_routine(die_nicely);
3291 set_checkpoint_signal();
3292 while (read_next_command() != EOF) {
3293 if (!strcmp("blob", command_buf.buf))
3294 parse_new_blob();
3295 else if (!prefixcmp(command_buf.buf, "ls "))
3296 parse_ls(NULL);
3297 else if (!prefixcmp(command_buf.buf, "commit "))
3298 parse_new_commit();
3299 else if (!prefixcmp(command_buf.buf, "tag "))
3300 parse_new_tag();
3301 else if (!prefixcmp(command_buf.buf, "reset "))
3302 parse_reset_branch();
3303 else if (!strcmp("checkpoint", command_buf.buf))
3304 parse_checkpoint();
3305 else if (!strcmp("done", command_buf.buf))
3306 break;
3307 else if (!prefixcmp(command_buf.buf, "progress "))
3308 parse_progress();
3309 else if (!prefixcmp(command_buf.buf, "feature "))
3310 parse_feature();
3311 else if (!prefixcmp(command_buf.buf, "option git "))
3312 parse_option();
3313 else if (!prefixcmp(command_buf.buf, "option "))
3314 /* ignore non-git options*/;
3315 else
3316 die("Unsupported command: %s", command_buf.buf);
3318 if (checkpoint_requested)
3319 checkpoint();
3322 /* argv hasn't been parsed yet, do so */
3323 if (!seen_data_command)
3324 parse_argv();
3326 if (require_explicit_termination && feof(stdin))
3327 die("stream ends early");
3329 end_packfile();
3331 dump_branches();
3332 dump_tags();
3333 unkeep_all_packs();
3334 dump_marks();
3336 if (pack_edges)
3337 fclose(pack_edges);
3339 if (show_stats) {
3340 uintmax_t total_count = 0, duplicate_count = 0;
3341 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3342 total_count += object_count_by_type[i];
3343 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3344 duplicate_count += duplicate_count_by_type[i];
3346 fprintf(stderr, "%s statistics:\n", argv[0]);
3347 fprintf(stderr, "---------------------------------------------------------------------\n");
3348 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3349 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3350 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
3351 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
3352 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
3353 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
3354 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3355 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3356 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3357 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3358 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3359 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3360 fprintf(stderr, "---------------------------------------------------------------------\n");
3361 pack_report();
3362 fprintf(stderr, "---------------------------------------------------------------------\n");
3363 fprintf(stderr, "\n");
3366 return failure ? 1 : 0;