fast-import: check committer name more strictly
[git/mingw.git] / fast-import.c
blob967d70c7004ba59df41ddb6154fef8f7cb2925a9
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 unsigned int pack_id;
308 static struct sha1file *pack_file;
309 static struct packed_git *pack_data;
310 static struct packed_git **all_packs;
311 static off_t pack_size;
313 /* Table of objects we've written. */
314 static unsigned int object_entry_alloc = 5000;
315 static struct object_entry_pool *blocks;
316 static struct object_entry *object_table[1 << 16];
317 static struct mark_set *marks;
318 static const char *export_marks_file;
319 static const char *import_marks_file;
320 static int import_marks_file_from_stream;
321 static int import_marks_file_ignore_missing;
322 static int relative_marks_paths;
324 /* Our last blob */
325 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
327 /* Tree management */
328 static unsigned int tree_entry_alloc = 1000;
329 static void *avail_tree_entry;
330 static unsigned int avail_tree_table_sz = 100;
331 static struct avail_tree_content **avail_tree_table;
332 static struct strbuf old_tree = STRBUF_INIT;
333 static struct strbuf new_tree = STRBUF_INIT;
335 /* Branch data */
336 static unsigned long max_active_branches = 5;
337 static unsigned long cur_active_branches;
338 static unsigned long branch_table_sz = 1039;
339 static struct branch **branch_table;
340 static struct branch *active_branches;
342 /* Tag data */
343 static struct tag *first_tag;
344 static struct tag *last_tag;
346 /* Input stream parsing */
347 static whenspec_type whenspec = WHENSPEC_RAW;
348 static struct strbuf command_buf = STRBUF_INIT;
349 static int unread_command_buf;
350 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
351 static struct recent_command *cmd_tail = &cmd_hist;
352 static struct recent_command *rc_free;
353 static unsigned int cmd_save = 100;
354 static uintmax_t next_mark;
355 static struct strbuf new_data = STRBUF_INIT;
356 static int seen_data_command;
358 /* Signal handling */
359 static volatile sig_atomic_t checkpoint_requested;
361 /* Where to write output of cat-blob commands */
362 static int cat_blob_fd = STDOUT_FILENO;
364 static void parse_argv(void);
365 static void parse_cat_blob(void);
366 static void parse_ls(struct branch *b);
368 static void write_branch_report(FILE *rpt, struct branch *b)
370 fprintf(rpt, "%s:\n", b->name);
372 fprintf(rpt, " status :");
373 if (b->active)
374 fputs(" active", rpt);
375 if (b->branch_tree.tree)
376 fputs(" loaded", rpt);
377 if (is_null_sha1(b->branch_tree.versions[1].sha1))
378 fputs(" dirty", rpt);
379 fputc('\n', rpt);
381 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
382 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
383 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
384 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
386 fputs(" last pack : ", rpt);
387 if (b->pack_id < MAX_PACK_ID)
388 fprintf(rpt, "%u", b->pack_id);
389 fputc('\n', rpt);
391 fputc('\n', rpt);
394 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
396 static void write_crash_report(const char *err)
398 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
399 FILE *rpt = fopen(loc, "w");
400 struct branch *b;
401 unsigned long lu;
402 struct recent_command *rc;
404 if (!rpt) {
405 error("can't write crash report %s: %s", loc, strerror(errno));
406 return;
409 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
411 fprintf(rpt, "fast-import crash report:\n");
412 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
413 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
414 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
415 fputc('\n', rpt);
417 fputs("fatal: ", rpt);
418 fputs(err, rpt);
419 fputc('\n', rpt);
421 fputc('\n', rpt);
422 fputs("Most Recent Commands Before Crash\n", rpt);
423 fputs("---------------------------------\n", rpt);
424 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
425 if (rc->next == &cmd_hist)
426 fputs("* ", rpt);
427 else
428 fputs(" ", rpt);
429 fputs(rc->buf, rpt);
430 fputc('\n', rpt);
433 fputc('\n', rpt);
434 fputs("Active Branch LRU\n", rpt);
435 fputs("-----------------\n", rpt);
436 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
437 cur_active_branches,
438 max_active_branches);
439 fputc('\n', rpt);
440 fputs(" pos clock name\n", rpt);
441 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
442 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
443 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
444 ++lu, b->last_commit, b->name);
446 fputc('\n', rpt);
447 fputs("Inactive Branches\n", rpt);
448 fputs("-----------------\n", rpt);
449 for (lu = 0; lu < branch_table_sz; lu++) {
450 for (b = branch_table[lu]; b; b = b->table_next_branch)
451 write_branch_report(rpt, b);
454 if (first_tag) {
455 struct tag *tg;
456 fputc('\n', rpt);
457 fputs("Annotated Tags\n", rpt);
458 fputs("--------------\n", rpt);
459 for (tg = first_tag; tg; tg = tg->next_tag) {
460 fputs(sha1_to_hex(tg->sha1), rpt);
461 fputc(' ', rpt);
462 fputs(tg->name, rpt);
463 fputc('\n', rpt);
467 fputc('\n', rpt);
468 fputs("Marks\n", rpt);
469 fputs("-----\n", rpt);
470 if (export_marks_file)
471 fprintf(rpt, " exported to %s\n", export_marks_file);
472 else
473 dump_marks_helper(rpt, 0, marks);
475 fputc('\n', rpt);
476 fputs("-------------------\n", rpt);
477 fputs("END OF CRASH REPORT\n", rpt);
478 fclose(rpt);
481 static void end_packfile(void);
482 static void unkeep_all_packs(void);
483 static void dump_marks(void);
485 static NORETURN void die_nicely(const char *err, va_list params)
487 static int zombie;
488 char message[2 * PATH_MAX];
490 vsnprintf(message, sizeof(message), err, params);
491 fputs("fatal: ", stderr);
492 fputs(message, stderr);
493 fputc('\n', stderr);
495 if (!zombie) {
496 zombie = 1;
497 write_crash_report(message);
498 end_packfile();
499 unkeep_all_packs();
500 dump_marks();
502 exit(128);
505 #ifndef SIGUSR1 /* Windows, for example */
507 static void set_checkpoint_signal(void)
511 #else
513 static void checkpoint_signal(int signo)
515 checkpoint_requested = 1;
518 static void set_checkpoint_signal(void)
520 struct sigaction sa;
522 memset(&sa, 0, sizeof(sa));
523 sa.sa_handler = checkpoint_signal;
524 sigemptyset(&sa.sa_mask);
525 sa.sa_flags = SA_RESTART;
526 sigaction(SIGUSR1, &sa, NULL);
529 #endif
531 static void alloc_objects(unsigned int cnt)
533 struct object_entry_pool *b;
535 b = xmalloc(sizeof(struct object_entry_pool)
536 + cnt * sizeof(struct object_entry));
537 b->next_pool = blocks;
538 b->next_free = b->entries;
539 b->end = b->entries + cnt;
540 blocks = b;
541 alloc_count += cnt;
544 static struct object_entry *new_object(unsigned char *sha1)
546 struct object_entry *e;
548 if (blocks->next_free == blocks->end)
549 alloc_objects(object_entry_alloc);
551 e = blocks->next_free++;
552 hashcpy(e->idx.sha1, sha1);
553 return e;
556 static struct object_entry *find_object(unsigned char *sha1)
558 unsigned int h = sha1[0] << 8 | sha1[1];
559 struct object_entry *e;
560 for (e = object_table[h]; e; e = e->next)
561 if (!hashcmp(sha1, e->idx.sha1))
562 return e;
563 return NULL;
566 static struct object_entry *insert_object(unsigned char *sha1)
568 unsigned int h = sha1[0] << 8 | sha1[1];
569 struct object_entry *e = object_table[h];
571 while (e) {
572 if (!hashcmp(sha1, e->idx.sha1))
573 return e;
574 e = e->next;
577 e = new_object(sha1);
578 e->next = object_table[h];
579 e->idx.offset = 0;
580 object_table[h] = e;
581 return e;
584 static unsigned int hc_str(const char *s, size_t len)
586 unsigned int r = 0;
587 while (len-- > 0)
588 r = r * 31 + *s++;
589 return r;
592 static void *pool_alloc(size_t len)
594 struct mem_pool *p;
595 void *r;
597 /* round up to a 'uintmax_t' alignment */
598 if (len & (sizeof(uintmax_t) - 1))
599 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
601 for (p = mem_pool; p; p = p->next_pool)
602 if ((p->end - p->next_free >= len))
603 break;
605 if (!p) {
606 if (len >= (mem_pool_alloc/2)) {
607 total_allocd += len;
608 return xmalloc(len);
610 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
611 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
612 p->next_pool = mem_pool;
613 p->next_free = (char *) p->space;
614 p->end = p->next_free + mem_pool_alloc;
615 mem_pool = p;
618 r = p->next_free;
619 p->next_free += len;
620 return r;
623 static void *pool_calloc(size_t count, size_t size)
625 size_t len = count * size;
626 void *r = pool_alloc(len);
627 memset(r, 0, len);
628 return r;
631 static char *pool_strdup(const char *s)
633 char *r = pool_alloc(strlen(s) + 1);
634 strcpy(r, s);
635 return r;
638 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
640 struct mark_set *s = marks;
641 while ((idnum >> s->shift) >= 1024) {
642 s = pool_calloc(1, sizeof(struct mark_set));
643 s->shift = marks->shift + 10;
644 s->data.sets[0] = marks;
645 marks = s;
647 while (s->shift) {
648 uintmax_t i = idnum >> s->shift;
649 idnum -= i << s->shift;
650 if (!s->data.sets[i]) {
651 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
652 s->data.sets[i]->shift = s->shift - 10;
654 s = s->data.sets[i];
656 if (!s->data.marked[idnum])
657 marks_set_count++;
658 s->data.marked[idnum] = oe;
661 static struct object_entry *find_mark(uintmax_t idnum)
663 uintmax_t orig_idnum = idnum;
664 struct mark_set *s = marks;
665 struct object_entry *oe = NULL;
666 if ((idnum >> s->shift) < 1024) {
667 while (s && s->shift) {
668 uintmax_t i = idnum >> s->shift;
669 idnum -= i << s->shift;
670 s = s->data.sets[i];
672 if (s)
673 oe = s->data.marked[idnum];
675 if (!oe)
676 die("mark :%" PRIuMAX " not declared", orig_idnum);
677 return oe;
680 static struct atom_str *to_atom(const char *s, unsigned short len)
682 unsigned int hc = hc_str(s, len) % atom_table_sz;
683 struct atom_str *c;
685 for (c = atom_table[hc]; c; c = c->next_atom)
686 if (c->str_len == len && !strncmp(s, c->str_dat, len))
687 return c;
689 c = pool_alloc(sizeof(struct atom_str) + len + 1);
690 c->str_len = len;
691 strncpy(c->str_dat, s, len);
692 c->str_dat[len] = 0;
693 c->next_atom = atom_table[hc];
694 atom_table[hc] = c;
695 atom_cnt++;
696 return c;
699 static struct branch *lookup_branch(const char *name)
701 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
702 struct branch *b;
704 for (b = branch_table[hc]; b; b = b->table_next_branch)
705 if (!strcmp(name, b->name))
706 return b;
707 return NULL;
710 static struct branch *new_branch(const char *name)
712 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
713 struct branch *b = lookup_branch(name);
715 if (b)
716 die("Invalid attempt to create duplicate branch: %s", name);
717 switch (check_ref_format(name)) {
718 case 0: break; /* its valid */
719 case CHECK_REF_FORMAT_ONELEVEL:
720 break; /* valid, but too few '/', allow anyway */
721 default:
722 die("Branch name doesn't conform to GIT standards: %s", name);
725 b = pool_calloc(1, sizeof(struct branch));
726 b->name = pool_strdup(name);
727 b->table_next_branch = branch_table[hc];
728 b->branch_tree.versions[0].mode = S_IFDIR;
729 b->branch_tree.versions[1].mode = S_IFDIR;
730 b->num_notes = 0;
731 b->active = 0;
732 b->pack_id = MAX_PACK_ID;
733 branch_table[hc] = b;
734 branch_count++;
735 return b;
738 static unsigned int hc_entries(unsigned int cnt)
740 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
741 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
744 static struct tree_content *new_tree_content(unsigned int cnt)
746 struct avail_tree_content *f, *l = NULL;
747 struct tree_content *t;
748 unsigned int hc = hc_entries(cnt);
750 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
751 if (f->entry_capacity >= cnt)
752 break;
754 if (f) {
755 if (l)
756 l->next_avail = f->next_avail;
757 else
758 avail_tree_table[hc] = f->next_avail;
759 } else {
760 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
761 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
762 f->entry_capacity = cnt;
765 t = (struct tree_content*)f;
766 t->entry_count = 0;
767 t->delta_depth = 0;
768 return t;
771 static void release_tree_entry(struct tree_entry *e);
772 static void release_tree_content(struct tree_content *t)
774 struct avail_tree_content *f = (struct avail_tree_content*)t;
775 unsigned int hc = hc_entries(f->entry_capacity);
776 f->next_avail = avail_tree_table[hc];
777 avail_tree_table[hc] = f;
780 static void release_tree_content_recursive(struct tree_content *t)
782 unsigned int i;
783 for (i = 0; i < t->entry_count; i++)
784 release_tree_entry(t->entries[i]);
785 release_tree_content(t);
788 static struct tree_content *grow_tree_content(
789 struct tree_content *t,
790 int amt)
792 struct tree_content *r = new_tree_content(t->entry_count + amt);
793 r->entry_count = t->entry_count;
794 r->delta_depth = t->delta_depth;
795 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
796 release_tree_content(t);
797 return r;
800 static struct tree_entry *new_tree_entry(void)
802 struct tree_entry *e;
804 if (!avail_tree_entry) {
805 unsigned int n = tree_entry_alloc;
806 total_allocd += n * sizeof(struct tree_entry);
807 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
808 while (n-- > 1) {
809 *((void**)e) = e + 1;
810 e++;
812 *((void**)e) = NULL;
815 e = avail_tree_entry;
816 avail_tree_entry = *((void**)e);
817 return e;
820 static void release_tree_entry(struct tree_entry *e)
822 if (e->tree)
823 release_tree_content_recursive(e->tree);
824 *((void**)e) = avail_tree_entry;
825 avail_tree_entry = e;
828 static struct tree_content *dup_tree_content(struct tree_content *s)
830 struct tree_content *d;
831 struct tree_entry *a, *b;
832 unsigned int i;
834 if (!s)
835 return NULL;
836 d = new_tree_content(s->entry_count);
837 for (i = 0; i < s->entry_count; i++) {
838 a = s->entries[i];
839 b = new_tree_entry();
840 memcpy(b, a, sizeof(*a));
841 if (a->tree && is_null_sha1(b->versions[1].sha1))
842 b->tree = dup_tree_content(a->tree);
843 else
844 b->tree = NULL;
845 d->entries[i] = b;
847 d->entry_count = s->entry_count;
848 d->delta_depth = s->delta_depth;
850 return d;
853 static void start_packfile(void)
855 static char tmpfile[PATH_MAX];
856 struct packed_git *p;
857 struct pack_header hdr;
858 int pack_fd;
860 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
861 "pack/tmp_pack_XXXXXX");
862 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
863 strcpy(p->pack_name, tmpfile);
864 p->pack_fd = pack_fd;
865 p->do_not_close = 1;
866 pack_file = sha1fd(pack_fd, p->pack_name);
868 hdr.hdr_signature = htonl(PACK_SIGNATURE);
869 hdr.hdr_version = htonl(2);
870 hdr.hdr_entries = 0;
871 sha1write(pack_file, &hdr, sizeof(hdr));
873 pack_data = p;
874 pack_size = sizeof(hdr);
875 object_count = 0;
877 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
878 all_packs[pack_id] = p;
881 static const char *create_index(void)
883 const char *tmpfile;
884 struct pack_idx_entry **idx, **c, **last;
885 struct object_entry *e;
886 struct object_entry_pool *o;
888 /* Build the table of object IDs. */
889 idx = xmalloc(object_count * sizeof(*idx));
890 c = idx;
891 for (o = blocks; o; o = o->next_pool)
892 for (e = o->next_free; e-- != o->entries;)
893 if (pack_id == e->pack_id)
894 *c++ = &e->idx;
895 last = idx + object_count;
896 if (c != last)
897 die("internal consistency error creating the index");
899 tmpfile = write_idx_file(NULL, idx, object_count, pack_data->sha1);
900 free(idx);
901 return tmpfile;
904 static char *keep_pack(const char *curr_index_name)
906 static char name[PATH_MAX];
907 static const char *keep_msg = "fast-import";
908 int keep_fd;
910 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
911 if (keep_fd < 0)
912 die_errno("cannot create keep file");
913 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
914 if (close(keep_fd))
915 die_errno("failed to write keep file");
917 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
918 get_object_directory(), sha1_to_hex(pack_data->sha1));
919 if (move_temp_to_file(pack_data->pack_name, name))
920 die("cannot store pack file");
922 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
923 get_object_directory(), sha1_to_hex(pack_data->sha1));
924 if (move_temp_to_file(curr_index_name, name))
925 die("cannot store index file");
926 free((void *)curr_index_name);
927 return name;
930 static void unkeep_all_packs(void)
932 static char name[PATH_MAX];
933 int k;
935 for (k = 0; k < pack_id; k++) {
936 struct packed_git *p = all_packs[k];
937 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
938 get_object_directory(), sha1_to_hex(p->sha1));
939 unlink_or_warn(name);
943 static void end_packfile(void)
945 struct packed_git *old_p = pack_data, *new_p;
947 clear_delta_base_cache();
948 if (object_count) {
949 unsigned char cur_pack_sha1[20];
950 char *idx_name;
951 int i;
952 struct branch *b;
953 struct tag *t;
955 close_pack_windows(pack_data);
956 sha1close(pack_file, cur_pack_sha1, 0);
957 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
958 pack_data->pack_name, object_count,
959 cur_pack_sha1, pack_size);
960 close(pack_data->pack_fd);
961 idx_name = keep_pack(create_index());
963 /* Register the packfile with core git's machinery. */
964 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
965 if (!new_p)
966 die("core git rejected index %s", idx_name);
967 all_packs[pack_id] = new_p;
968 install_packed_git(new_p);
970 /* Print the boundary */
971 if (pack_edges) {
972 fprintf(pack_edges, "%s:", new_p->pack_name);
973 for (i = 0; i < branch_table_sz; i++) {
974 for (b = branch_table[i]; b; b = b->table_next_branch) {
975 if (b->pack_id == pack_id)
976 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
979 for (t = first_tag; t; t = t->next_tag) {
980 if (t->pack_id == pack_id)
981 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
983 fputc('\n', pack_edges);
984 fflush(pack_edges);
987 pack_id++;
989 else {
990 close(old_p->pack_fd);
991 unlink_or_warn(old_p->pack_name);
993 free(old_p);
995 /* We can't carry a delta across packfiles. */
996 strbuf_release(&last_blob.data);
997 last_blob.offset = 0;
998 last_blob.depth = 0;
1001 static void cycle_packfile(void)
1003 end_packfile();
1004 start_packfile();
1007 static int store_object(
1008 enum object_type type,
1009 struct strbuf *dat,
1010 struct last_object *last,
1011 unsigned char *sha1out,
1012 uintmax_t mark)
1014 void *out, *delta;
1015 struct object_entry *e;
1016 unsigned char hdr[96];
1017 unsigned char sha1[20];
1018 unsigned long hdrlen, deltalen;
1019 git_SHA_CTX c;
1020 z_stream s;
1022 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1023 (unsigned long)dat->len) + 1;
1024 git_SHA1_Init(&c);
1025 git_SHA1_Update(&c, hdr, hdrlen);
1026 git_SHA1_Update(&c, dat->buf, dat->len);
1027 git_SHA1_Final(sha1, &c);
1028 if (sha1out)
1029 hashcpy(sha1out, sha1);
1031 e = insert_object(sha1);
1032 if (mark)
1033 insert_mark(mark, e);
1034 if (e->idx.offset) {
1035 duplicate_count_by_type[type]++;
1036 return 1;
1037 } else if (find_sha1_pack(sha1, packed_git)) {
1038 e->type = type;
1039 e->pack_id = MAX_PACK_ID;
1040 e->idx.offset = 1; /* just not zero! */
1041 duplicate_count_by_type[type]++;
1042 return 1;
1045 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1046 delta = diff_delta(last->data.buf, last->data.len,
1047 dat->buf, dat->len,
1048 &deltalen, dat->len - 20);
1049 } else
1050 delta = NULL;
1052 memset(&s, 0, sizeof(s));
1053 deflateInit(&s, pack_compression_level);
1054 if (delta) {
1055 s.next_in = delta;
1056 s.avail_in = deltalen;
1057 } else {
1058 s.next_in = (void *)dat->buf;
1059 s.avail_in = dat->len;
1061 s.avail_out = deflateBound(&s, s.avail_in);
1062 s.next_out = out = xmalloc(s.avail_out);
1063 while (deflate(&s, Z_FINISH) == Z_OK)
1064 /* nothing */;
1065 deflateEnd(&s);
1067 /* Determine if we should auto-checkpoint. */
1068 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1069 || (pack_size + 60 + s.total_out) < pack_size) {
1071 /* This new object needs to *not* have the current pack_id. */
1072 e->pack_id = pack_id + 1;
1073 cycle_packfile();
1075 /* We cannot carry a delta into the new pack. */
1076 if (delta) {
1077 free(delta);
1078 delta = NULL;
1080 memset(&s, 0, sizeof(s));
1081 deflateInit(&s, pack_compression_level);
1082 s.next_in = (void *)dat->buf;
1083 s.avail_in = dat->len;
1084 s.avail_out = deflateBound(&s, s.avail_in);
1085 s.next_out = out = xrealloc(out, s.avail_out);
1086 while (deflate(&s, Z_FINISH) == Z_OK)
1087 /* nothing */;
1088 deflateEnd(&s);
1092 e->type = type;
1093 e->pack_id = pack_id;
1094 e->idx.offset = pack_size;
1095 object_count++;
1096 object_count_by_type[type]++;
1098 crc32_begin(pack_file);
1100 if (delta) {
1101 off_t ofs = e->idx.offset - last->offset;
1102 unsigned pos = sizeof(hdr) - 1;
1104 delta_count_by_type[type]++;
1105 e->depth = last->depth + 1;
1107 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1108 sha1write(pack_file, hdr, hdrlen);
1109 pack_size += hdrlen;
1111 hdr[pos] = ofs & 127;
1112 while (ofs >>= 7)
1113 hdr[--pos] = 128 | (--ofs & 127);
1114 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1115 pack_size += sizeof(hdr) - pos;
1116 } else {
1117 e->depth = 0;
1118 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1119 sha1write(pack_file, hdr, hdrlen);
1120 pack_size += hdrlen;
1123 sha1write(pack_file, out, s.total_out);
1124 pack_size += s.total_out;
1126 e->idx.crc32 = crc32_end(pack_file);
1128 free(out);
1129 free(delta);
1130 if (last) {
1131 if (last->no_swap) {
1132 last->data = *dat;
1133 } else {
1134 strbuf_swap(&last->data, dat);
1136 last->offset = e->idx.offset;
1137 last->depth = e->depth;
1139 return 0;
1142 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1144 if (ftruncate(pack_data->pack_fd, to)
1145 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1146 die_errno("cannot truncate pack to skip duplicate");
1147 pack_size = to;
1149 /* yes this is a layering violation */
1150 pack_file->total = to;
1151 pack_file->offset = 0;
1152 pack_file->ctx = *ctx;
1155 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1157 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1158 unsigned char *in_buf = xmalloc(in_sz);
1159 unsigned char *out_buf = xmalloc(out_sz);
1160 struct object_entry *e;
1161 unsigned char sha1[20];
1162 unsigned long hdrlen;
1163 off_t offset;
1164 git_SHA_CTX c;
1165 git_SHA_CTX pack_file_ctx;
1166 z_stream s;
1167 int status = Z_OK;
1169 /* Determine if we should auto-checkpoint. */
1170 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1171 || (pack_size + 60 + len) < pack_size)
1172 cycle_packfile();
1174 offset = pack_size;
1176 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1177 sha1flush(pack_file);
1178 pack_file_ctx = pack_file->ctx;
1180 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1181 if (out_sz <= hdrlen)
1182 die("impossibly large object header");
1184 git_SHA1_Init(&c);
1185 git_SHA1_Update(&c, out_buf, hdrlen);
1187 crc32_begin(pack_file);
1189 memset(&s, 0, sizeof(s));
1190 deflateInit(&s, pack_compression_level);
1192 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1193 if (out_sz <= hdrlen)
1194 die("impossibly large object header");
1196 s.next_out = out_buf + hdrlen;
1197 s.avail_out = out_sz - hdrlen;
1199 while (status != Z_STREAM_END) {
1200 if (0 < len && !s.avail_in) {
1201 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1202 size_t n = fread(in_buf, 1, cnt, stdin);
1203 if (!n && feof(stdin))
1204 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1206 git_SHA1_Update(&c, in_buf, n);
1207 s.next_in = in_buf;
1208 s.avail_in = n;
1209 len -= n;
1212 status = deflate(&s, len ? 0 : Z_FINISH);
1214 if (!s.avail_out || status == Z_STREAM_END) {
1215 size_t n = s.next_out - out_buf;
1216 sha1write(pack_file, out_buf, n);
1217 pack_size += n;
1218 s.next_out = out_buf;
1219 s.avail_out = out_sz;
1222 switch (status) {
1223 case Z_OK:
1224 case Z_BUF_ERROR:
1225 case Z_STREAM_END:
1226 continue;
1227 default:
1228 die("unexpected deflate failure: %d", status);
1231 deflateEnd(&s);
1232 git_SHA1_Final(sha1, &c);
1234 if (sha1out)
1235 hashcpy(sha1out, sha1);
1237 e = insert_object(sha1);
1239 if (mark)
1240 insert_mark(mark, e);
1242 if (e->idx.offset) {
1243 duplicate_count_by_type[OBJ_BLOB]++;
1244 truncate_pack(offset, &pack_file_ctx);
1246 } else if (find_sha1_pack(sha1, packed_git)) {
1247 e->type = OBJ_BLOB;
1248 e->pack_id = MAX_PACK_ID;
1249 e->idx.offset = 1; /* just not zero! */
1250 duplicate_count_by_type[OBJ_BLOB]++;
1251 truncate_pack(offset, &pack_file_ctx);
1253 } else {
1254 e->depth = 0;
1255 e->type = OBJ_BLOB;
1256 e->pack_id = pack_id;
1257 e->idx.offset = offset;
1258 e->idx.crc32 = crc32_end(pack_file);
1259 object_count++;
1260 object_count_by_type[OBJ_BLOB]++;
1263 free(in_buf);
1264 free(out_buf);
1267 /* All calls must be guarded by find_object() or find_mark() to
1268 * ensure the 'struct object_entry' passed was written by this
1269 * process instance. We unpack the entry by the offset, avoiding
1270 * the need for the corresponding .idx file. This unpacking rule
1271 * works because we only use OBJ_REF_DELTA within the packfiles
1272 * created by fast-import.
1274 * oe must not be NULL. Such an oe usually comes from giving
1275 * an unknown SHA-1 to find_object() or an undefined mark to
1276 * find_mark(). Callers must test for this condition and use
1277 * the standard read_sha1_file() when it happens.
1279 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1280 * find_mark(), where the mark was reloaded from an existing marks
1281 * file and is referencing an object that this fast-import process
1282 * instance did not write out to a packfile. Callers must test for
1283 * this condition and use read_sha1_file() instead.
1285 static void *gfi_unpack_entry(
1286 struct object_entry *oe,
1287 unsigned long *sizep)
1289 enum object_type type;
1290 struct packed_git *p = all_packs[oe->pack_id];
1291 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1292 /* The object is stored in the packfile we are writing to
1293 * and we have modified it since the last time we scanned
1294 * back to read a previously written object. If an old
1295 * window covered [p->pack_size, p->pack_size + 20) its
1296 * data is stale and is not valid. Closing all windows
1297 * and updating the packfile length ensures we can read
1298 * the newly written data.
1300 close_pack_windows(p);
1301 sha1flush(pack_file);
1303 /* We have to offer 20 bytes additional on the end of
1304 * the packfile as the core unpacker code assumes the
1305 * footer is present at the file end and must promise
1306 * at least 20 bytes within any window it maps. But
1307 * we don't actually create the footer here.
1309 p->pack_size = pack_size + 20;
1311 return unpack_entry(p, oe->idx.offset, &type, sizep);
1314 static const char *get_mode(const char *str, uint16_t *modep)
1316 unsigned char c;
1317 uint16_t mode = 0;
1319 while ((c = *str++) != ' ') {
1320 if (c < '0' || c > '7')
1321 return NULL;
1322 mode = (mode << 3) + (c - '0');
1324 *modep = mode;
1325 return str;
1328 static void load_tree(struct tree_entry *root)
1330 unsigned char *sha1 = root->versions[1].sha1;
1331 struct object_entry *myoe;
1332 struct tree_content *t;
1333 unsigned long size;
1334 char *buf;
1335 const char *c;
1337 root->tree = t = new_tree_content(8);
1338 if (is_null_sha1(sha1))
1339 return;
1341 myoe = find_object(sha1);
1342 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1343 if (myoe->type != OBJ_TREE)
1344 die("Not a tree: %s", sha1_to_hex(sha1));
1345 t->delta_depth = myoe->depth;
1346 buf = gfi_unpack_entry(myoe, &size);
1347 if (!buf)
1348 die("Can't load tree %s", sha1_to_hex(sha1));
1349 } else {
1350 enum object_type type;
1351 buf = read_sha1_file(sha1, &type, &size);
1352 if (!buf || type != OBJ_TREE)
1353 die("Can't load tree %s", sha1_to_hex(sha1));
1356 c = buf;
1357 while (c != (buf + size)) {
1358 struct tree_entry *e = new_tree_entry();
1360 if (t->entry_count == t->entry_capacity)
1361 root->tree = t = grow_tree_content(t, t->entry_count);
1362 t->entries[t->entry_count++] = e;
1364 e->tree = NULL;
1365 c = get_mode(c, &e->versions[1].mode);
1366 if (!c)
1367 die("Corrupt mode in %s", sha1_to_hex(sha1));
1368 e->versions[0].mode = e->versions[1].mode;
1369 e->name = to_atom(c, strlen(c));
1370 c += e->name->str_len + 1;
1371 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1372 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1373 c += 20;
1375 free(buf);
1378 static int tecmp0 (const void *_a, const void *_b)
1380 struct tree_entry *a = *((struct tree_entry**)_a);
1381 struct tree_entry *b = *((struct tree_entry**)_b);
1382 return base_name_compare(
1383 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1384 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1387 static int tecmp1 (const void *_a, const void *_b)
1389 struct tree_entry *a = *((struct tree_entry**)_a);
1390 struct tree_entry *b = *((struct tree_entry**)_b);
1391 return base_name_compare(
1392 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1393 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1396 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1398 size_t maxlen = 0;
1399 unsigned int i;
1401 if (!v)
1402 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1403 else
1404 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1406 for (i = 0; i < t->entry_count; i++) {
1407 if (t->entries[i]->versions[v].mode)
1408 maxlen += t->entries[i]->name->str_len + 34;
1411 strbuf_reset(b);
1412 strbuf_grow(b, maxlen);
1413 for (i = 0; i < t->entry_count; i++) {
1414 struct tree_entry *e = t->entries[i];
1415 if (!e->versions[v].mode)
1416 continue;
1417 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1418 e->name->str_dat, '\0');
1419 strbuf_add(b, e->versions[v].sha1, 20);
1423 static void store_tree(struct tree_entry *root)
1425 struct tree_content *t = root->tree;
1426 unsigned int i, j, del;
1427 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1428 struct object_entry *le;
1430 if (!is_null_sha1(root->versions[1].sha1))
1431 return;
1433 for (i = 0; i < t->entry_count; i++) {
1434 if (t->entries[i]->tree)
1435 store_tree(t->entries[i]);
1438 le = find_object(root->versions[0].sha1);
1439 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1440 mktree(t, 0, &old_tree);
1441 lo.data = old_tree;
1442 lo.offset = le->idx.offset;
1443 lo.depth = t->delta_depth;
1446 mktree(t, 1, &new_tree);
1447 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1449 t->delta_depth = lo.depth;
1450 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1451 struct tree_entry *e = t->entries[i];
1452 if (e->versions[1].mode) {
1453 e->versions[0].mode = e->versions[1].mode;
1454 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1455 t->entries[j++] = e;
1456 } else {
1457 release_tree_entry(e);
1458 del++;
1461 t->entry_count -= del;
1464 static void tree_content_replace(
1465 struct tree_entry *root,
1466 const unsigned char *sha1,
1467 const uint16_t mode,
1468 struct tree_content *newtree)
1470 if (!S_ISDIR(mode))
1471 die("Root cannot be a non-directory");
1472 hashcpy(root->versions[1].sha1, sha1);
1473 if (root->tree)
1474 release_tree_content_recursive(root->tree);
1475 root->tree = newtree;
1478 static int tree_content_set(
1479 struct tree_entry *root,
1480 const char *p,
1481 const unsigned char *sha1,
1482 const uint16_t mode,
1483 struct tree_content *subtree)
1485 struct tree_content *t;
1486 const char *slash1;
1487 unsigned int i, n;
1488 struct tree_entry *e;
1490 slash1 = strchr(p, '/');
1491 if (slash1)
1492 n = slash1 - p;
1493 else
1494 n = strlen(p);
1495 if (!n)
1496 die("Empty path component found in input");
1497 if (!slash1 && !S_ISDIR(mode) && subtree)
1498 die("Non-directories cannot have subtrees");
1500 if (!root->tree)
1501 load_tree(root);
1502 t = root->tree;
1503 for (i = 0; i < t->entry_count; i++) {
1504 e = t->entries[i];
1505 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1506 if (!slash1) {
1507 if (!S_ISDIR(mode)
1508 && e->versions[1].mode == mode
1509 && !hashcmp(e->versions[1].sha1, sha1))
1510 return 0;
1511 e->versions[1].mode = mode;
1512 hashcpy(e->versions[1].sha1, sha1);
1513 if (e->tree)
1514 release_tree_content_recursive(e->tree);
1515 e->tree = subtree;
1516 hashclr(root->versions[1].sha1);
1517 return 1;
1519 if (!S_ISDIR(e->versions[1].mode)) {
1520 e->tree = new_tree_content(8);
1521 e->versions[1].mode = S_IFDIR;
1523 if (!e->tree)
1524 load_tree(e);
1525 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1526 hashclr(root->versions[1].sha1);
1527 return 1;
1529 return 0;
1533 if (t->entry_count == t->entry_capacity)
1534 root->tree = t = grow_tree_content(t, t->entry_count);
1535 e = new_tree_entry();
1536 e->name = to_atom(p, n);
1537 e->versions[0].mode = 0;
1538 hashclr(e->versions[0].sha1);
1539 t->entries[t->entry_count++] = e;
1540 if (slash1) {
1541 e->tree = new_tree_content(8);
1542 e->versions[1].mode = S_IFDIR;
1543 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1544 } else {
1545 e->tree = subtree;
1546 e->versions[1].mode = mode;
1547 hashcpy(e->versions[1].sha1, sha1);
1549 hashclr(root->versions[1].sha1);
1550 return 1;
1553 static int tree_content_remove(
1554 struct tree_entry *root,
1555 const char *p,
1556 struct tree_entry *backup_leaf)
1558 struct tree_content *t;
1559 const char *slash1;
1560 unsigned int i, n;
1561 struct tree_entry *e;
1563 slash1 = strchr(p, '/');
1564 if (slash1)
1565 n = slash1 - p;
1566 else
1567 n = strlen(p);
1569 if (!root->tree)
1570 load_tree(root);
1571 t = root->tree;
1572 for (i = 0; i < t->entry_count; i++) {
1573 e = t->entries[i];
1574 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1575 if (slash1 && !S_ISDIR(e->versions[1].mode))
1577 * If p names a file in some subdirectory, and a
1578 * file or symlink matching the name of the
1579 * parent directory of p exists, then p cannot
1580 * exist and need not be deleted.
1582 return 1;
1583 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1584 goto del_entry;
1585 if (!e->tree)
1586 load_tree(e);
1587 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1588 for (n = 0; n < e->tree->entry_count; n++) {
1589 if (e->tree->entries[n]->versions[1].mode) {
1590 hashclr(root->versions[1].sha1);
1591 return 1;
1594 backup_leaf = NULL;
1595 goto del_entry;
1597 return 0;
1600 return 0;
1602 del_entry:
1603 if (backup_leaf)
1604 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1605 else if (e->tree)
1606 release_tree_content_recursive(e->tree);
1607 e->tree = NULL;
1608 e->versions[1].mode = 0;
1609 hashclr(e->versions[1].sha1);
1610 hashclr(root->versions[1].sha1);
1611 return 1;
1614 static int tree_content_get(
1615 struct tree_entry *root,
1616 const char *p,
1617 struct tree_entry *leaf)
1619 struct tree_content *t;
1620 const char *slash1;
1621 unsigned int i, n;
1622 struct tree_entry *e;
1624 slash1 = strchr(p, '/');
1625 if (slash1)
1626 n = slash1 - p;
1627 else
1628 n = strlen(p);
1630 if (!root->tree)
1631 load_tree(root);
1632 t = root->tree;
1633 for (i = 0; i < t->entry_count; i++) {
1634 e = t->entries[i];
1635 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1636 if (!slash1) {
1637 memcpy(leaf, e, sizeof(*leaf));
1638 if (e->tree && is_null_sha1(e->versions[1].sha1))
1639 leaf->tree = dup_tree_content(e->tree);
1640 else
1641 leaf->tree = NULL;
1642 return 1;
1644 if (!S_ISDIR(e->versions[1].mode))
1645 return 0;
1646 if (!e->tree)
1647 load_tree(e);
1648 return tree_content_get(e, slash1 + 1, leaf);
1651 return 0;
1654 static int update_branch(struct branch *b)
1656 static const char *msg = "fast-import";
1657 struct ref_lock *lock;
1658 unsigned char old_sha1[20];
1660 if (is_null_sha1(b->sha1))
1661 return 0;
1662 if (read_ref(b->name, old_sha1))
1663 hashclr(old_sha1);
1664 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1665 if (!lock)
1666 return error("Unable to lock %s", b->name);
1667 if (!force_update && !is_null_sha1(old_sha1)) {
1668 struct commit *old_cmit, *new_cmit;
1670 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1671 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1672 if (!old_cmit || !new_cmit) {
1673 unlock_ref(lock);
1674 return error("Branch %s is missing commits.", b->name);
1677 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1678 unlock_ref(lock);
1679 warning("Not updating %s"
1680 " (new tip %s does not contain %s)",
1681 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1682 return -1;
1685 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1686 return error("Unable to update %s", b->name);
1687 return 0;
1690 static void dump_branches(void)
1692 unsigned int i;
1693 struct branch *b;
1695 for (i = 0; i < branch_table_sz; i++) {
1696 for (b = branch_table[i]; b; b = b->table_next_branch)
1697 failure |= update_branch(b);
1701 static void dump_tags(void)
1703 static const char *msg = "fast-import";
1704 struct tag *t;
1705 struct ref_lock *lock;
1706 char ref_name[PATH_MAX];
1708 for (t = first_tag; t; t = t->next_tag) {
1709 sprintf(ref_name, "tags/%s", t->name);
1710 lock = lock_ref_sha1(ref_name, NULL);
1711 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1712 failure |= error("Unable to update %s", ref_name);
1716 static void dump_marks_helper(FILE *f,
1717 uintmax_t base,
1718 struct mark_set *m)
1720 uintmax_t k;
1721 if (m->shift) {
1722 for (k = 0; k < 1024; k++) {
1723 if (m->data.sets[k])
1724 dump_marks_helper(f, base + (k << m->shift),
1725 m->data.sets[k]);
1727 } else {
1728 for (k = 0; k < 1024; k++) {
1729 if (m->data.marked[k])
1730 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1731 sha1_to_hex(m->data.marked[k]->idx.sha1));
1736 static void dump_marks(void)
1738 static struct lock_file mark_lock;
1739 int mark_fd;
1740 FILE *f;
1742 if (!export_marks_file)
1743 return;
1745 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1746 if (mark_fd < 0) {
1747 failure |= error("Unable to write marks file %s: %s",
1748 export_marks_file, strerror(errno));
1749 return;
1752 f = fdopen(mark_fd, "w");
1753 if (!f) {
1754 int saved_errno = errno;
1755 rollback_lock_file(&mark_lock);
1756 failure |= error("Unable to write marks file %s: %s",
1757 export_marks_file, strerror(saved_errno));
1758 return;
1762 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1763 * Assign -1 to the lock file descriptor so that commit_lock_file()
1764 * won't try to close() it.
1766 mark_lock.fd = -1;
1768 dump_marks_helper(f, 0, marks);
1769 if (ferror(f) || fclose(f)) {
1770 int saved_errno = errno;
1771 rollback_lock_file(&mark_lock);
1772 failure |= error("Unable to write marks file %s: %s",
1773 export_marks_file, strerror(saved_errno));
1774 return;
1777 if (commit_lock_file(&mark_lock)) {
1778 int saved_errno = errno;
1779 rollback_lock_file(&mark_lock);
1780 failure |= error("Unable to commit marks file %s: %s",
1781 export_marks_file, strerror(saved_errno));
1782 return;
1786 static void read_marks(void)
1788 char line[512];
1789 FILE *f = fopen(import_marks_file, "r");
1790 if (f)
1792 else if (import_marks_file_ignore_missing && errno == ENOENT)
1793 return; /* Marks file does not exist */
1794 else
1795 die_errno("cannot read '%s'", import_marks_file);
1796 while (fgets(line, sizeof(line), f)) {
1797 uintmax_t mark;
1798 char *end;
1799 unsigned char sha1[20];
1800 struct object_entry *e;
1802 end = strchr(line, '\n');
1803 if (line[0] != ':' || !end)
1804 die("corrupt mark line: %s", line);
1805 *end = 0;
1806 mark = strtoumax(line + 1, &end, 10);
1807 if (!mark || end == line + 1
1808 || *end != ' ' || get_sha1(end + 1, sha1))
1809 die("corrupt mark line: %s", line);
1810 e = find_object(sha1);
1811 if (!e) {
1812 enum object_type type = sha1_object_info(sha1, NULL);
1813 if (type < 0)
1814 die("object not found: %s", sha1_to_hex(sha1));
1815 e = insert_object(sha1);
1816 e->type = type;
1817 e->pack_id = MAX_PACK_ID;
1818 e->idx.offset = 1; /* just not zero! */
1820 insert_mark(mark, e);
1822 fclose(f);
1826 static int read_next_command(void)
1828 static int stdin_eof = 0;
1830 if (stdin_eof) {
1831 unread_command_buf = 0;
1832 return EOF;
1835 for (;;) {
1836 if (unread_command_buf) {
1837 unread_command_buf = 0;
1838 } else {
1839 struct recent_command *rc;
1841 strbuf_detach(&command_buf, NULL);
1842 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1843 if (stdin_eof)
1844 return EOF;
1846 if (!seen_data_command
1847 && prefixcmp(command_buf.buf, "feature ")
1848 && prefixcmp(command_buf.buf, "option ")) {
1849 parse_argv();
1852 rc = rc_free;
1853 if (rc)
1854 rc_free = rc->next;
1855 else {
1856 rc = cmd_hist.next;
1857 cmd_hist.next = rc->next;
1858 cmd_hist.next->prev = &cmd_hist;
1859 free(rc->buf);
1862 rc->buf = command_buf.buf;
1863 rc->prev = cmd_tail;
1864 rc->next = cmd_hist.prev;
1865 rc->prev->next = rc;
1866 cmd_tail = rc;
1868 if (!prefixcmp(command_buf.buf, "cat-blob ")) {
1869 parse_cat_blob();
1870 continue;
1872 if (command_buf.buf[0] == '#')
1873 continue;
1874 return 0;
1878 static void skip_optional_lf(void)
1880 int term_char = fgetc(stdin);
1881 if (term_char != '\n' && term_char != EOF)
1882 ungetc(term_char, stdin);
1885 static void parse_mark(void)
1887 if (!prefixcmp(command_buf.buf, "mark :")) {
1888 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1889 read_next_command();
1891 else
1892 next_mark = 0;
1895 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1897 strbuf_reset(sb);
1899 if (prefixcmp(command_buf.buf, "data "))
1900 die("Expected 'data n' command, found: %s", command_buf.buf);
1902 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1903 char *term = xstrdup(command_buf.buf + 5 + 2);
1904 size_t term_len = command_buf.len - 5 - 2;
1906 strbuf_detach(&command_buf, NULL);
1907 for (;;) {
1908 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1909 die("EOF in data (terminator '%s' not found)", term);
1910 if (term_len == command_buf.len
1911 && !strcmp(term, command_buf.buf))
1912 break;
1913 strbuf_addbuf(sb, &command_buf);
1914 strbuf_addch(sb, '\n');
1916 free(term);
1918 else {
1919 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1920 size_t n = 0, length = (size_t)len;
1922 if (limit && limit < len) {
1923 *len_res = len;
1924 return 0;
1926 if (length < len)
1927 die("data is too large to use in this context");
1929 while (n < length) {
1930 size_t s = strbuf_fread(sb, length - n, stdin);
1931 if (!s && feof(stdin))
1932 die("EOF in data (%lu bytes remaining)",
1933 (unsigned long)(length - n));
1934 n += s;
1938 skip_optional_lf();
1939 return 1;
1942 static int validate_raw_date(const char *src, char *result, int maxlen)
1944 const char *orig_src = src;
1945 char *endp;
1946 unsigned long num;
1948 errno = 0;
1950 num = strtoul(src, &endp, 10);
1951 /* NEEDSWORK: perhaps check for reasonable values? */
1952 if (errno || endp == src || *endp != ' ')
1953 return -1;
1955 src = endp + 1;
1956 if (*src != '-' && *src != '+')
1957 return -1;
1959 num = strtoul(src + 1, &endp, 10);
1960 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1961 1400 < num)
1962 return -1;
1964 strcpy(result, orig_src);
1965 return 0;
1968 static char *parse_ident(const char *buf)
1970 const char *ltgt;
1971 size_t name_len;
1972 char *ident;
1974 /* ensure there is a space delimiter even if there is no name */
1975 if (*buf == '<')
1976 --buf;
1978 ltgt = buf + strcspn(buf, "<>");
1979 if (*ltgt != '<')
1980 die("Missing < in ident string: %s", buf);
1981 if (ltgt != buf && ltgt[-1] != ' ')
1982 die("Missing space before < in ident string: %s", buf);
1983 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
1984 if (*ltgt != '>')
1985 die("Missing > in ident string: %s", buf);
1986 ltgt++;
1987 if (*ltgt != ' ')
1988 die("Missing space after > in ident string: %s", buf);
1989 ltgt++;
1990 name_len = ltgt - buf;
1991 ident = xmalloc(name_len + 24);
1992 strncpy(ident, buf, name_len);
1994 switch (whenspec) {
1995 case WHENSPEC_RAW:
1996 if (validate_raw_date(ltgt, ident + name_len, 24) < 0)
1997 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
1998 break;
1999 case WHENSPEC_RFC2822:
2000 if (parse_date(ltgt, ident + name_len, 24) < 0)
2001 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2002 break;
2003 case WHENSPEC_NOW:
2004 if (strcmp("now", ltgt))
2005 die("Date in ident must be 'now': %s", buf);
2006 datestamp(ident + name_len, 24);
2007 break;
2010 return ident;
2013 static void parse_and_store_blob(
2014 struct last_object *last,
2015 unsigned char *sha1out,
2016 uintmax_t mark)
2018 static struct strbuf buf = STRBUF_INIT;
2019 uintmax_t len;
2021 if (parse_data(&buf, big_file_threshold, &len))
2022 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2023 else {
2024 if (last) {
2025 strbuf_release(&last->data);
2026 last->offset = 0;
2027 last->depth = 0;
2029 stream_blob(len, sha1out, mark);
2030 skip_optional_lf();
2034 static void parse_new_blob(void)
2036 read_next_command();
2037 parse_mark();
2038 parse_and_store_blob(&last_blob, NULL, next_mark);
2041 static void unload_one_branch(void)
2043 while (cur_active_branches
2044 && cur_active_branches >= max_active_branches) {
2045 uintmax_t min_commit = ULONG_MAX;
2046 struct branch *e, *l = NULL, *p = NULL;
2048 for (e = active_branches; e; e = e->active_next_branch) {
2049 if (e->last_commit < min_commit) {
2050 p = l;
2051 min_commit = e->last_commit;
2053 l = e;
2056 if (p) {
2057 e = p->active_next_branch;
2058 p->active_next_branch = e->active_next_branch;
2059 } else {
2060 e = active_branches;
2061 active_branches = e->active_next_branch;
2063 e->active = 0;
2064 e->active_next_branch = NULL;
2065 if (e->branch_tree.tree) {
2066 release_tree_content_recursive(e->branch_tree.tree);
2067 e->branch_tree.tree = NULL;
2069 cur_active_branches--;
2073 static void load_branch(struct branch *b)
2075 load_tree(&b->branch_tree);
2076 if (!b->active) {
2077 b->active = 1;
2078 b->active_next_branch = active_branches;
2079 active_branches = b;
2080 cur_active_branches++;
2081 branch_load_count++;
2085 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2087 unsigned char fanout = 0;
2088 while ((num_notes >>= 8))
2089 fanout++;
2090 return fanout;
2093 static void construct_path_with_fanout(const char *hex_sha1,
2094 unsigned char fanout, char *path)
2096 unsigned int i = 0, j = 0;
2097 if (fanout >= 20)
2098 die("Too large fanout (%u)", fanout);
2099 while (fanout) {
2100 path[i++] = hex_sha1[j++];
2101 path[i++] = hex_sha1[j++];
2102 path[i++] = '/';
2103 fanout--;
2105 memcpy(path + i, hex_sha1 + j, 40 - j);
2106 path[i + 40 - j] = '\0';
2109 static uintmax_t do_change_note_fanout(
2110 struct tree_entry *orig_root, struct tree_entry *root,
2111 char *hex_sha1, unsigned int hex_sha1_len,
2112 char *fullpath, unsigned int fullpath_len,
2113 unsigned char fanout)
2115 struct tree_content *t = root->tree;
2116 struct tree_entry *e, leaf;
2117 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2118 uintmax_t num_notes = 0;
2119 unsigned char sha1[20];
2120 char realpath[60];
2122 for (i = 0; t && i < t->entry_count; i++) {
2123 e = t->entries[i];
2124 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2125 tmp_fullpath_len = fullpath_len;
2128 * We're interested in EITHER existing note entries (entries
2129 * with exactly 40 hex chars in path, not including directory
2130 * separators), OR directory entries that may contain note
2131 * entries (with < 40 hex chars in path).
2132 * Also, each path component in a note entry must be a multiple
2133 * of 2 chars.
2135 if (!e->versions[1].mode ||
2136 tmp_hex_sha1_len > 40 ||
2137 e->name->str_len % 2)
2138 continue;
2140 /* This _may_ be a note entry, or a subdir containing notes */
2141 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2142 e->name->str_len);
2143 if (tmp_fullpath_len)
2144 fullpath[tmp_fullpath_len++] = '/';
2145 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2146 e->name->str_len);
2147 tmp_fullpath_len += e->name->str_len;
2148 fullpath[tmp_fullpath_len] = '\0';
2150 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2151 /* This is a note entry */
2152 construct_path_with_fanout(hex_sha1, fanout, realpath);
2153 if (!strcmp(fullpath, realpath)) {
2154 /* Note entry is in correct location */
2155 num_notes++;
2156 continue;
2159 /* Rename fullpath to realpath */
2160 if (!tree_content_remove(orig_root, fullpath, &leaf))
2161 die("Failed to remove path %s", fullpath);
2162 tree_content_set(orig_root, realpath,
2163 leaf.versions[1].sha1,
2164 leaf.versions[1].mode,
2165 leaf.tree);
2166 } else if (S_ISDIR(e->versions[1].mode)) {
2167 /* This is a subdir that may contain note entries */
2168 if (!e->tree)
2169 load_tree(e);
2170 num_notes += do_change_note_fanout(orig_root, e,
2171 hex_sha1, tmp_hex_sha1_len,
2172 fullpath, tmp_fullpath_len, fanout);
2175 /* The above may have reallocated the current tree_content */
2176 t = root->tree;
2178 return num_notes;
2181 static uintmax_t change_note_fanout(struct tree_entry *root,
2182 unsigned char fanout)
2184 char hex_sha1[40], path[60];
2185 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2188 static void file_change_m(struct branch *b)
2190 const char *p = command_buf.buf + 2;
2191 static struct strbuf uq = STRBUF_INIT;
2192 const char *endp;
2193 struct object_entry *oe = oe;
2194 unsigned char sha1[20];
2195 uint16_t mode, inline_data = 0;
2197 p = get_mode(p, &mode);
2198 if (!p)
2199 die("Corrupt mode: %s", command_buf.buf);
2200 switch (mode) {
2201 case 0644:
2202 case 0755:
2203 mode |= S_IFREG;
2204 case S_IFREG | 0644:
2205 case S_IFREG | 0755:
2206 case S_IFLNK:
2207 case S_IFDIR:
2208 case S_IFGITLINK:
2209 /* ok */
2210 break;
2211 default:
2212 die("Corrupt mode: %s", command_buf.buf);
2215 if (*p == ':') {
2216 char *x;
2217 oe = find_mark(strtoumax(p + 1, &x, 10));
2218 hashcpy(sha1, oe->idx.sha1);
2219 p = x;
2220 } else if (!prefixcmp(p, "inline")) {
2221 inline_data = 1;
2222 p += 6;
2223 } else {
2224 if (get_sha1_hex(p, sha1))
2225 die("Invalid SHA1: %s", command_buf.buf);
2226 oe = find_object(sha1);
2227 p += 40;
2229 if (*p++ != ' ')
2230 die("Missing space after SHA1: %s", command_buf.buf);
2232 strbuf_reset(&uq);
2233 if (!unquote_c_style(&uq, p, &endp)) {
2234 if (*endp)
2235 die("Garbage after path in: %s", command_buf.buf);
2236 p = uq.buf;
2239 /* Git does not track empty, non-toplevel directories. */
2240 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2241 tree_content_remove(&b->branch_tree, p, NULL);
2242 return;
2245 if (S_ISGITLINK(mode)) {
2246 if (inline_data)
2247 die("Git links cannot be specified 'inline': %s",
2248 command_buf.buf);
2249 else if (oe) {
2250 if (oe->type != OBJ_COMMIT)
2251 die("Not a commit (actually a %s): %s",
2252 typename(oe->type), command_buf.buf);
2255 * Accept the sha1 without checking; it expected to be in
2256 * another repository.
2258 } else if (inline_data) {
2259 if (S_ISDIR(mode))
2260 die("Directories cannot be specified 'inline': %s",
2261 command_buf.buf);
2262 if (p != uq.buf) {
2263 strbuf_addstr(&uq, p);
2264 p = uq.buf;
2266 read_next_command();
2267 parse_and_store_blob(&last_blob, sha1, 0);
2268 } else {
2269 enum object_type expected = S_ISDIR(mode) ?
2270 OBJ_TREE: OBJ_BLOB;
2271 enum object_type type = oe ? oe->type :
2272 sha1_object_info(sha1, NULL);
2273 if (type < 0)
2274 die("%s not found: %s",
2275 S_ISDIR(mode) ? "Tree" : "Blob",
2276 command_buf.buf);
2277 if (type != expected)
2278 die("Not a %s (actually a %s): %s",
2279 typename(expected), typename(type),
2280 command_buf.buf);
2283 if (!*p) {
2284 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2285 return;
2287 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2290 static void file_change_d(struct branch *b)
2292 const char *p = command_buf.buf + 2;
2293 static struct strbuf uq = STRBUF_INIT;
2294 const char *endp;
2296 strbuf_reset(&uq);
2297 if (!unquote_c_style(&uq, p, &endp)) {
2298 if (*endp)
2299 die("Garbage after path in: %s", command_buf.buf);
2300 p = uq.buf;
2302 tree_content_remove(&b->branch_tree, p, NULL);
2305 static void file_change_cr(struct branch *b, int rename)
2307 const char *s, *d;
2308 static struct strbuf s_uq = STRBUF_INIT;
2309 static struct strbuf d_uq = STRBUF_INIT;
2310 const char *endp;
2311 struct tree_entry leaf;
2313 s = command_buf.buf + 2;
2314 strbuf_reset(&s_uq);
2315 if (!unquote_c_style(&s_uq, s, &endp)) {
2316 if (*endp != ' ')
2317 die("Missing space after source: %s", command_buf.buf);
2318 } else {
2319 endp = strchr(s, ' ');
2320 if (!endp)
2321 die("Missing space after source: %s", command_buf.buf);
2322 strbuf_add(&s_uq, s, endp - s);
2324 s = s_uq.buf;
2326 endp++;
2327 if (!*endp)
2328 die("Missing dest: %s", command_buf.buf);
2330 d = endp;
2331 strbuf_reset(&d_uq);
2332 if (!unquote_c_style(&d_uq, d, &endp)) {
2333 if (*endp)
2334 die("Garbage after dest in: %s", command_buf.buf);
2335 d = d_uq.buf;
2338 memset(&leaf, 0, sizeof(leaf));
2339 if (rename)
2340 tree_content_remove(&b->branch_tree, s, &leaf);
2341 else
2342 tree_content_get(&b->branch_tree, s, &leaf);
2343 if (!leaf.versions[1].mode)
2344 die("Path %s not in branch", s);
2345 if (!*d) { /* C "path/to/subdir" "" */
2346 tree_content_replace(&b->branch_tree,
2347 leaf.versions[1].sha1,
2348 leaf.versions[1].mode,
2349 leaf.tree);
2350 return;
2352 tree_content_set(&b->branch_tree, d,
2353 leaf.versions[1].sha1,
2354 leaf.versions[1].mode,
2355 leaf.tree);
2358 static void note_change_n(struct branch *b, unsigned char old_fanout)
2360 const char *p = command_buf.buf + 2;
2361 static struct strbuf uq = STRBUF_INIT;
2362 struct object_entry *oe = oe;
2363 struct branch *s;
2364 unsigned char sha1[20], commit_sha1[20];
2365 char path[60];
2366 uint16_t inline_data = 0;
2367 unsigned char new_fanout;
2369 /* <dataref> or 'inline' */
2370 if (*p == ':') {
2371 char *x;
2372 oe = find_mark(strtoumax(p + 1, &x, 10));
2373 hashcpy(sha1, oe->idx.sha1);
2374 p = x;
2375 } else if (!prefixcmp(p, "inline")) {
2376 inline_data = 1;
2377 p += 6;
2378 } else {
2379 if (get_sha1_hex(p, sha1))
2380 die("Invalid SHA1: %s", command_buf.buf);
2381 oe = find_object(sha1);
2382 p += 40;
2384 if (*p++ != ' ')
2385 die("Missing space after SHA1: %s", command_buf.buf);
2387 /* <committish> */
2388 s = lookup_branch(p);
2389 if (s) {
2390 hashcpy(commit_sha1, s->sha1);
2391 } else if (*p == ':') {
2392 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2393 struct object_entry *commit_oe = find_mark(commit_mark);
2394 if (commit_oe->type != OBJ_COMMIT)
2395 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2396 hashcpy(commit_sha1, commit_oe->idx.sha1);
2397 } else if (!get_sha1(p, commit_sha1)) {
2398 unsigned long size;
2399 char *buf = read_object_with_reference(commit_sha1,
2400 commit_type, &size, commit_sha1);
2401 if (!buf || size < 46)
2402 die("Not a valid commit: %s", p);
2403 free(buf);
2404 } else
2405 die("Invalid ref name or SHA1 expression: %s", p);
2407 if (inline_data) {
2408 if (p != uq.buf) {
2409 strbuf_addstr(&uq, p);
2410 p = uq.buf;
2412 read_next_command();
2413 parse_and_store_blob(&last_blob, sha1, 0);
2414 } else if (oe) {
2415 if (oe->type != OBJ_BLOB)
2416 die("Not a blob (actually a %s): %s",
2417 typename(oe->type), command_buf.buf);
2418 } else if (!is_null_sha1(sha1)) {
2419 enum object_type type = sha1_object_info(sha1, NULL);
2420 if (type < 0)
2421 die("Blob not found: %s", command_buf.buf);
2422 if (type != OBJ_BLOB)
2423 die("Not a blob (actually a %s): %s",
2424 typename(type), command_buf.buf);
2427 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2428 if (tree_content_remove(&b->branch_tree, path, NULL))
2429 b->num_notes--;
2431 if (is_null_sha1(sha1))
2432 return; /* nothing to insert */
2434 b->num_notes++;
2435 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2436 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2437 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2440 static void file_change_deleteall(struct branch *b)
2442 release_tree_content_recursive(b->branch_tree.tree);
2443 hashclr(b->branch_tree.versions[0].sha1);
2444 hashclr(b->branch_tree.versions[1].sha1);
2445 load_tree(&b->branch_tree);
2446 b->num_notes = 0;
2449 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2451 if (!buf || size < 46)
2452 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2453 if (memcmp("tree ", buf, 5)
2454 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2455 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2456 hashcpy(b->branch_tree.versions[0].sha1,
2457 b->branch_tree.versions[1].sha1);
2460 static void parse_from_existing(struct branch *b)
2462 if (is_null_sha1(b->sha1)) {
2463 hashclr(b->branch_tree.versions[0].sha1);
2464 hashclr(b->branch_tree.versions[1].sha1);
2465 } else {
2466 unsigned long size;
2467 char *buf;
2469 buf = read_object_with_reference(b->sha1,
2470 commit_type, &size, b->sha1);
2471 parse_from_commit(b, buf, size);
2472 free(buf);
2476 static int parse_from(struct branch *b)
2478 const char *from;
2479 struct branch *s;
2481 if (prefixcmp(command_buf.buf, "from "))
2482 return 0;
2484 if (b->branch_tree.tree) {
2485 release_tree_content_recursive(b->branch_tree.tree);
2486 b->branch_tree.tree = NULL;
2489 from = strchr(command_buf.buf, ' ') + 1;
2490 s = lookup_branch(from);
2491 if (b == s)
2492 die("Can't create a branch from itself: %s", b->name);
2493 else if (s) {
2494 unsigned char *t = s->branch_tree.versions[1].sha1;
2495 hashcpy(b->sha1, s->sha1);
2496 hashcpy(b->branch_tree.versions[0].sha1, t);
2497 hashcpy(b->branch_tree.versions[1].sha1, t);
2498 } else if (*from == ':') {
2499 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2500 struct object_entry *oe = find_mark(idnum);
2501 if (oe->type != OBJ_COMMIT)
2502 die("Mark :%" PRIuMAX " not a commit", idnum);
2503 hashcpy(b->sha1, oe->idx.sha1);
2504 if (oe->pack_id != MAX_PACK_ID) {
2505 unsigned long size;
2506 char *buf = gfi_unpack_entry(oe, &size);
2507 parse_from_commit(b, buf, size);
2508 free(buf);
2509 } else
2510 parse_from_existing(b);
2511 } else if (!get_sha1(from, b->sha1))
2512 parse_from_existing(b);
2513 else
2514 die("Invalid ref name or SHA1 expression: %s", from);
2516 read_next_command();
2517 return 1;
2520 static struct hash_list *parse_merge(unsigned int *count)
2522 struct hash_list *list = NULL, *n, *e = e;
2523 const char *from;
2524 struct branch *s;
2526 *count = 0;
2527 while (!prefixcmp(command_buf.buf, "merge ")) {
2528 from = strchr(command_buf.buf, ' ') + 1;
2529 n = xmalloc(sizeof(*n));
2530 s = lookup_branch(from);
2531 if (s)
2532 hashcpy(n->sha1, s->sha1);
2533 else if (*from == ':') {
2534 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2535 struct object_entry *oe = find_mark(idnum);
2536 if (oe->type != OBJ_COMMIT)
2537 die("Mark :%" PRIuMAX " not a commit", idnum);
2538 hashcpy(n->sha1, oe->idx.sha1);
2539 } else if (!get_sha1(from, n->sha1)) {
2540 unsigned long size;
2541 char *buf = read_object_with_reference(n->sha1,
2542 commit_type, &size, n->sha1);
2543 if (!buf || size < 46)
2544 die("Not a valid commit: %s", from);
2545 free(buf);
2546 } else
2547 die("Invalid ref name or SHA1 expression: %s", from);
2549 n->next = NULL;
2550 if (list)
2551 e->next = n;
2552 else
2553 list = n;
2554 e = n;
2555 (*count)++;
2556 read_next_command();
2558 return list;
2561 static void parse_new_commit(void)
2563 static struct strbuf msg = STRBUF_INIT;
2564 struct branch *b;
2565 char *sp;
2566 char *author = NULL;
2567 char *committer = NULL;
2568 struct hash_list *merge_list = NULL;
2569 unsigned int merge_count;
2570 unsigned char prev_fanout, new_fanout;
2572 /* Obtain the branch name from the rest of our command */
2573 sp = strchr(command_buf.buf, ' ') + 1;
2574 b = lookup_branch(sp);
2575 if (!b)
2576 b = new_branch(sp);
2578 read_next_command();
2579 parse_mark();
2580 if (!prefixcmp(command_buf.buf, "author ")) {
2581 author = parse_ident(command_buf.buf + 7);
2582 read_next_command();
2584 if (!prefixcmp(command_buf.buf, "committer ")) {
2585 committer = parse_ident(command_buf.buf + 10);
2586 read_next_command();
2588 if (!committer)
2589 die("Expected committer but didn't get one");
2590 parse_data(&msg, 0, NULL);
2591 read_next_command();
2592 parse_from(b);
2593 merge_list = parse_merge(&merge_count);
2595 /* ensure the branch is active/loaded */
2596 if (!b->branch_tree.tree || !max_active_branches) {
2597 unload_one_branch();
2598 load_branch(b);
2601 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2603 /* file_change* */
2604 while (command_buf.len > 0) {
2605 if (!prefixcmp(command_buf.buf, "M "))
2606 file_change_m(b);
2607 else if (!prefixcmp(command_buf.buf, "D "))
2608 file_change_d(b);
2609 else if (!prefixcmp(command_buf.buf, "R "))
2610 file_change_cr(b, 1);
2611 else if (!prefixcmp(command_buf.buf, "C "))
2612 file_change_cr(b, 0);
2613 else if (!prefixcmp(command_buf.buf, "N "))
2614 note_change_n(b, prev_fanout);
2615 else if (!strcmp("deleteall", command_buf.buf))
2616 file_change_deleteall(b);
2617 else if (!prefixcmp(command_buf.buf, "ls "))
2618 parse_ls(b);
2619 else {
2620 unread_command_buf = 1;
2621 break;
2623 if (read_next_command() == EOF)
2624 break;
2627 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2628 if (new_fanout != prev_fanout)
2629 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2631 /* build the tree and the commit */
2632 store_tree(&b->branch_tree);
2633 hashcpy(b->branch_tree.versions[0].sha1,
2634 b->branch_tree.versions[1].sha1);
2636 strbuf_reset(&new_data);
2637 strbuf_addf(&new_data, "tree %s\n",
2638 sha1_to_hex(b->branch_tree.versions[1].sha1));
2639 if (!is_null_sha1(b->sha1))
2640 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2641 while (merge_list) {
2642 struct hash_list *next = merge_list->next;
2643 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2644 free(merge_list);
2645 merge_list = next;
2647 strbuf_addf(&new_data,
2648 "author %s\n"
2649 "committer %s\n"
2650 "\n",
2651 author ? author : committer, committer);
2652 strbuf_addbuf(&new_data, &msg);
2653 free(author);
2654 free(committer);
2656 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2657 b->pack_id = pack_id;
2658 b->last_commit = object_count_by_type[OBJ_COMMIT];
2661 static void parse_new_tag(void)
2663 static struct strbuf msg = STRBUF_INIT;
2664 char *sp;
2665 const char *from;
2666 char *tagger;
2667 struct branch *s;
2668 struct tag *t;
2669 uintmax_t from_mark = 0;
2670 unsigned char sha1[20];
2671 enum object_type type;
2673 /* Obtain the new tag name from the rest of our command */
2674 sp = strchr(command_buf.buf, ' ') + 1;
2675 t = pool_alloc(sizeof(struct tag));
2676 t->next_tag = NULL;
2677 t->name = pool_strdup(sp);
2678 if (last_tag)
2679 last_tag->next_tag = t;
2680 else
2681 first_tag = t;
2682 last_tag = t;
2683 read_next_command();
2685 /* from ... */
2686 if (prefixcmp(command_buf.buf, "from "))
2687 die("Expected from command, got %s", command_buf.buf);
2688 from = strchr(command_buf.buf, ' ') + 1;
2689 s = lookup_branch(from);
2690 if (s) {
2691 hashcpy(sha1, s->sha1);
2692 type = OBJ_COMMIT;
2693 } else if (*from == ':') {
2694 struct object_entry *oe;
2695 from_mark = strtoumax(from + 1, NULL, 10);
2696 oe = find_mark(from_mark);
2697 type = oe->type;
2698 hashcpy(sha1, oe->idx.sha1);
2699 } else if (!get_sha1(from, sha1)) {
2700 unsigned long size;
2701 char *buf;
2703 buf = read_sha1_file(sha1, &type, &size);
2704 if (!buf || size < 46)
2705 die("Not a valid commit: %s", from);
2706 free(buf);
2707 } else
2708 die("Invalid ref name or SHA1 expression: %s", from);
2709 read_next_command();
2711 /* tagger ... */
2712 if (!prefixcmp(command_buf.buf, "tagger ")) {
2713 tagger = parse_ident(command_buf.buf + 7);
2714 read_next_command();
2715 } else
2716 tagger = NULL;
2718 /* tag payload/message */
2719 parse_data(&msg, 0, NULL);
2721 /* build the tag object */
2722 strbuf_reset(&new_data);
2724 strbuf_addf(&new_data,
2725 "object %s\n"
2726 "type %s\n"
2727 "tag %s\n",
2728 sha1_to_hex(sha1), typename(type), t->name);
2729 if (tagger)
2730 strbuf_addf(&new_data,
2731 "tagger %s\n", tagger);
2732 strbuf_addch(&new_data, '\n');
2733 strbuf_addbuf(&new_data, &msg);
2734 free(tagger);
2736 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2737 t->pack_id = MAX_PACK_ID;
2738 else
2739 t->pack_id = pack_id;
2742 static void parse_reset_branch(void)
2744 struct branch *b;
2745 char *sp;
2747 /* Obtain the branch name from the rest of our command */
2748 sp = strchr(command_buf.buf, ' ') + 1;
2749 b = lookup_branch(sp);
2750 if (b) {
2751 hashclr(b->sha1);
2752 hashclr(b->branch_tree.versions[0].sha1);
2753 hashclr(b->branch_tree.versions[1].sha1);
2754 if (b->branch_tree.tree) {
2755 release_tree_content_recursive(b->branch_tree.tree);
2756 b->branch_tree.tree = NULL;
2759 else
2760 b = new_branch(sp);
2761 read_next_command();
2762 parse_from(b);
2763 if (command_buf.len > 0)
2764 unread_command_buf = 1;
2767 static void cat_blob_write(const char *buf, unsigned long size)
2769 if (write_in_full(cat_blob_fd, buf, size) != size)
2770 die_errno("Write to frontend failed");
2773 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2775 struct strbuf line = STRBUF_INIT;
2776 unsigned long size;
2777 enum object_type type = 0;
2778 char *buf;
2780 if (!oe || oe->pack_id == MAX_PACK_ID) {
2781 buf = read_sha1_file(sha1, &type, &size);
2782 } else {
2783 type = oe->type;
2784 buf = gfi_unpack_entry(oe, &size);
2788 * Output based on batch_one_object() from cat-file.c.
2790 if (type <= 0) {
2791 strbuf_reset(&line);
2792 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2793 cat_blob_write(line.buf, line.len);
2794 strbuf_release(&line);
2795 free(buf);
2796 return;
2798 if (!buf)
2799 die("Can't read object %s", sha1_to_hex(sha1));
2800 if (type != OBJ_BLOB)
2801 die("Object %s is a %s but a blob was expected.",
2802 sha1_to_hex(sha1), typename(type));
2803 strbuf_reset(&line);
2804 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2805 typename(type), size);
2806 cat_blob_write(line.buf, line.len);
2807 strbuf_release(&line);
2808 cat_blob_write(buf, size);
2809 cat_blob_write("\n", 1);
2810 free(buf);
2813 static void parse_cat_blob(void)
2815 const char *p;
2816 struct object_entry *oe = oe;
2817 unsigned char sha1[20];
2819 /* cat-blob SP <object> LF */
2820 p = command_buf.buf + strlen("cat-blob ");
2821 if (*p == ':') {
2822 char *x;
2823 oe = find_mark(strtoumax(p + 1, &x, 10));
2824 if (x == p + 1)
2825 die("Invalid mark: %s", command_buf.buf);
2826 if (!oe)
2827 die("Unknown mark: %s", command_buf.buf);
2828 if (*x)
2829 die("Garbage after mark: %s", command_buf.buf);
2830 hashcpy(sha1, oe->idx.sha1);
2831 } else {
2832 if (get_sha1_hex(p, sha1))
2833 die("Invalid SHA1: %s", command_buf.buf);
2834 if (p[40])
2835 die("Garbage after SHA1: %s", command_buf.buf);
2836 oe = find_object(sha1);
2839 cat_blob(oe, sha1);
2842 static struct object_entry *dereference(struct object_entry *oe,
2843 unsigned char sha1[20])
2845 unsigned long size;
2846 char *buf = NULL;
2847 if (!oe) {
2848 enum object_type type = sha1_object_info(sha1, NULL);
2849 if (type < 0)
2850 die("object not found: %s", sha1_to_hex(sha1));
2851 /* cache it! */
2852 oe = insert_object(sha1);
2853 oe->type = type;
2854 oe->pack_id = MAX_PACK_ID;
2855 oe->idx.offset = 1;
2857 switch (oe->type) {
2858 case OBJ_TREE: /* easy case. */
2859 return oe;
2860 case OBJ_COMMIT:
2861 case OBJ_TAG:
2862 break;
2863 default:
2864 die("Not a treeish: %s", command_buf.buf);
2867 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2868 buf = gfi_unpack_entry(oe, &size);
2869 } else {
2870 enum object_type unused;
2871 buf = read_sha1_file(sha1, &unused, &size);
2873 if (!buf)
2874 die("Can't load object %s", sha1_to_hex(sha1));
2876 /* Peel one layer. */
2877 switch (oe->type) {
2878 case OBJ_TAG:
2879 if (size < 40 + strlen("object ") ||
2880 get_sha1_hex(buf + strlen("object "), sha1))
2881 die("Invalid SHA1 in tag: %s", command_buf.buf);
2882 break;
2883 case OBJ_COMMIT:
2884 if (size < 40 + strlen("tree ") ||
2885 get_sha1_hex(buf + strlen("tree "), sha1))
2886 die("Invalid SHA1 in commit: %s", command_buf.buf);
2889 free(buf);
2890 return find_object(sha1);
2893 static struct object_entry *parse_treeish_dataref(const char **p)
2895 unsigned char sha1[20];
2896 struct object_entry *e;
2898 if (**p == ':') { /* <mark> */
2899 char *endptr;
2900 e = find_mark(strtoumax(*p + 1, &endptr, 10));
2901 if (endptr == *p + 1)
2902 die("Invalid mark: %s", command_buf.buf);
2903 if (!e)
2904 die("Unknown mark: %s", command_buf.buf);
2905 *p = endptr;
2906 hashcpy(sha1, e->idx.sha1);
2907 } else { /* <sha1> */
2908 if (get_sha1_hex(*p, sha1))
2909 die("Invalid SHA1: %s", command_buf.buf);
2910 e = find_object(sha1);
2911 *p += 40;
2914 while (!e || e->type != OBJ_TREE)
2915 e = dereference(e, sha1);
2916 return e;
2919 static void print_ls(int mode, const unsigned char *sha1, const char *path)
2921 static struct strbuf line = STRBUF_INIT;
2923 /* See show_tree(). */
2924 const char *type =
2925 S_ISGITLINK(mode) ? commit_type :
2926 S_ISDIR(mode) ? tree_type :
2927 blob_type;
2929 if (!mode) {
2930 /* missing SP path LF */
2931 strbuf_reset(&line);
2932 strbuf_addstr(&line, "missing ");
2933 quote_c_style(path, &line, NULL, 0);
2934 strbuf_addch(&line, '\n');
2935 } else {
2936 /* mode SP type SP object_name TAB path LF */
2937 strbuf_reset(&line);
2938 strbuf_addf(&line, "%06o %s %s\t",
2939 mode, type, sha1_to_hex(sha1));
2940 quote_c_style(path, &line, NULL, 0);
2941 strbuf_addch(&line, '\n');
2943 cat_blob_write(line.buf, line.len);
2946 static void parse_ls(struct branch *b)
2948 const char *p;
2949 struct tree_entry *root = NULL;
2950 struct tree_entry leaf = {NULL};
2952 /* ls SP (<treeish> SP)? <path> */
2953 p = command_buf.buf + strlen("ls ");
2954 if (*p == '"') {
2955 if (!b)
2956 die("Not in a commit: %s", command_buf.buf);
2957 root = &b->branch_tree;
2958 } else {
2959 struct object_entry *e = parse_treeish_dataref(&p);
2960 root = new_tree_entry();
2961 hashcpy(root->versions[1].sha1, e->idx.sha1);
2962 load_tree(root);
2963 if (*p++ != ' ')
2964 die("Missing space after tree-ish: %s", command_buf.buf);
2966 if (*p == '"') {
2967 static struct strbuf uq = STRBUF_INIT;
2968 const char *endp;
2969 strbuf_reset(&uq);
2970 if (unquote_c_style(&uq, p, &endp))
2971 die("Invalid path: %s", command_buf.buf);
2972 if (*endp)
2973 die("Garbage after path in: %s", command_buf.buf);
2974 p = uq.buf;
2976 tree_content_get(root, p, &leaf);
2978 * A directory in preparation would have a sha1 of zero
2979 * until it is saved. Save, for simplicity.
2981 if (S_ISDIR(leaf.versions[1].mode))
2982 store_tree(&leaf);
2984 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
2985 if (!b || root != &b->branch_tree)
2986 release_tree_entry(root);
2989 static void checkpoint(void)
2991 checkpoint_requested = 0;
2992 if (object_count) {
2993 cycle_packfile();
2994 dump_branches();
2995 dump_tags();
2996 dump_marks();
3000 static void parse_checkpoint(void)
3002 checkpoint_requested = 1;
3003 skip_optional_lf();
3006 static void parse_progress(void)
3008 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3009 fputc('\n', stdout);
3010 fflush(stdout);
3011 skip_optional_lf();
3014 static char* make_fast_import_path(const char *path)
3016 struct strbuf abs_path = STRBUF_INIT;
3018 if (!relative_marks_paths || is_absolute_path(path))
3019 return xstrdup(path);
3020 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
3021 return strbuf_detach(&abs_path, NULL);
3024 static void option_import_marks(const char *marks,
3025 int from_stream, int ignore_missing)
3027 if (import_marks_file) {
3028 if (from_stream)
3029 die("Only one import-marks command allowed per stream");
3031 /* read previous mark file */
3032 if(!import_marks_file_from_stream)
3033 read_marks();
3036 import_marks_file = make_fast_import_path(marks);
3037 safe_create_leading_directories_const(import_marks_file);
3038 import_marks_file_from_stream = from_stream;
3039 import_marks_file_ignore_missing = ignore_missing;
3042 static void option_date_format(const char *fmt)
3044 if (!strcmp(fmt, "raw"))
3045 whenspec = WHENSPEC_RAW;
3046 else if (!strcmp(fmt, "rfc2822"))
3047 whenspec = WHENSPEC_RFC2822;
3048 else if (!strcmp(fmt, "now"))
3049 whenspec = WHENSPEC_NOW;
3050 else
3051 die("unknown --date-format argument %s", fmt);
3054 static unsigned long ulong_arg(const char *option, const char *arg)
3056 char *endptr;
3057 unsigned long rv = strtoul(arg, &endptr, 0);
3058 if (strchr(arg, '-') || endptr == arg || *endptr)
3059 die("%s: argument must be a non-negative integer", option);
3060 return rv;
3063 static void option_depth(const char *depth)
3065 max_depth = ulong_arg("--depth", depth);
3066 if (max_depth > MAX_DEPTH)
3067 die("--depth cannot exceed %u", MAX_DEPTH);
3070 static void option_active_branches(const char *branches)
3072 max_active_branches = ulong_arg("--active-branches", branches);
3075 static void option_export_marks(const char *marks)
3077 export_marks_file = make_fast_import_path(marks);
3078 safe_create_leading_directories_const(export_marks_file);
3081 static void option_cat_blob_fd(const char *fd)
3083 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3084 if (n > (unsigned long) INT_MAX)
3085 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3086 cat_blob_fd = (int) n;
3089 static void option_export_pack_edges(const char *edges)
3091 if (pack_edges)
3092 fclose(pack_edges);
3093 pack_edges = fopen(edges, "a");
3094 if (!pack_edges)
3095 die_errno("Cannot open '%s'", edges);
3098 static int parse_one_option(const char *option)
3100 if (!prefixcmp(option, "max-pack-size=")) {
3101 unsigned long v;
3102 if (!git_parse_ulong(option + 14, &v))
3103 return 0;
3104 if (v < 8192) {
3105 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3106 v *= 1024 * 1024;
3107 } else if (v < 1024 * 1024) {
3108 warning("minimum max-pack-size is 1 MiB");
3109 v = 1024 * 1024;
3111 max_packsize = v;
3112 } else if (!prefixcmp(option, "big-file-threshold=")) {
3113 unsigned long v;
3114 if (!git_parse_ulong(option + 19, &v))
3115 return 0;
3116 big_file_threshold = v;
3117 } else if (!prefixcmp(option, "depth=")) {
3118 option_depth(option + 6);
3119 } else if (!prefixcmp(option, "active-branches=")) {
3120 option_active_branches(option + 16);
3121 } else if (!prefixcmp(option, "export-pack-edges=")) {
3122 option_export_pack_edges(option + 18);
3123 } else if (!prefixcmp(option, "quiet")) {
3124 show_stats = 0;
3125 } else if (!prefixcmp(option, "stats")) {
3126 show_stats = 1;
3127 } else {
3128 return 0;
3131 return 1;
3134 static int parse_one_feature(const char *feature, int from_stream)
3136 if (!prefixcmp(feature, "date-format=")) {
3137 option_date_format(feature + 12);
3138 } else if (!prefixcmp(feature, "import-marks=")) {
3139 option_import_marks(feature + 13, from_stream, 0);
3140 } else if (!prefixcmp(feature, "import-marks-if-exists=")) {
3141 option_import_marks(feature + strlen("import-marks-if-exists="),
3142 from_stream, 1);
3143 } else if (!prefixcmp(feature, "export-marks=")) {
3144 option_export_marks(feature + 13);
3145 } else if (!strcmp(feature, "cat-blob")) {
3146 ; /* Don't die - this feature is supported */
3147 } else if (!strcmp(feature, "relative-marks")) {
3148 relative_marks_paths = 1;
3149 } else if (!strcmp(feature, "no-relative-marks")) {
3150 relative_marks_paths = 0;
3151 } else if (!strcmp(feature, "force")) {
3152 force_update = 1;
3153 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3154 ; /* do nothing; we have the feature */
3155 } else {
3156 return 0;
3159 return 1;
3162 static void parse_feature(void)
3164 char *feature = command_buf.buf + 8;
3166 if (seen_data_command)
3167 die("Got feature command '%s' after data command", feature);
3169 if (parse_one_feature(feature, 1))
3170 return;
3172 die("This version of fast-import does not support feature %s.", feature);
3175 static void parse_option(void)
3177 char *option = command_buf.buf + 11;
3179 if (seen_data_command)
3180 die("Got option command '%s' after data command", option);
3182 if (parse_one_option(option))
3183 return;
3185 die("This version of fast-import does not support option: %s", option);
3188 static int git_pack_config(const char *k, const char *v, void *cb)
3190 if (!strcmp(k, "pack.depth")) {
3191 max_depth = git_config_int(k, v);
3192 if (max_depth > MAX_DEPTH)
3193 max_depth = MAX_DEPTH;
3194 return 0;
3196 if (!strcmp(k, "pack.compression")) {
3197 int level = git_config_int(k, v);
3198 if (level == -1)
3199 level = Z_DEFAULT_COMPRESSION;
3200 else if (level < 0 || level > Z_BEST_COMPRESSION)
3201 die("bad pack compression level %d", level);
3202 pack_compression_level = level;
3203 pack_compression_seen = 1;
3204 return 0;
3206 if (!strcmp(k, "pack.indexversion")) {
3207 pack_idx_default_version = git_config_int(k, v);
3208 if (pack_idx_default_version > 2)
3209 die("bad pack.indexversion=%"PRIu32,
3210 pack_idx_default_version);
3211 return 0;
3213 if (!strcmp(k, "pack.packsizelimit")) {
3214 max_packsize = git_config_ulong(k, v);
3215 return 0;
3217 return git_default_config(k, v, cb);
3220 static const char fast_import_usage[] =
3221 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3223 static void parse_argv(void)
3225 unsigned int i;
3227 for (i = 1; i < global_argc; i++) {
3228 const char *a = global_argv[i];
3230 if (*a != '-' || !strcmp(a, "--"))
3231 break;
3233 if (parse_one_option(a + 2))
3234 continue;
3236 if (parse_one_feature(a + 2, 0))
3237 continue;
3239 if (!prefixcmp(a + 2, "cat-blob-fd=")) {
3240 option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
3241 continue;
3244 die("unknown option %s", a);
3246 if (i != global_argc)
3247 usage(fast_import_usage);
3249 seen_data_command = 1;
3250 if (import_marks_file)
3251 read_marks();
3254 int main(int argc, const char **argv)
3256 unsigned int i;
3258 git_extract_argv0_path(argv[0]);
3260 if (argc == 2 && !strcmp(argv[1], "-h"))
3261 usage(fast_import_usage);
3263 setup_git_directory();
3264 git_config(git_pack_config, NULL);
3265 if (!pack_compression_seen && core_compression_seen)
3266 pack_compression_level = core_compression_level;
3268 alloc_objects(object_entry_alloc);
3269 strbuf_init(&command_buf, 0);
3270 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3271 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3272 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3273 marks = pool_calloc(1, sizeof(struct mark_set));
3275 global_argc = argc;
3276 global_argv = argv;
3278 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3279 for (i = 0; i < (cmd_save - 1); i++)
3280 rc_free[i].next = &rc_free[i + 1];
3281 rc_free[cmd_save - 1].next = NULL;
3283 prepare_packed_git();
3284 start_packfile();
3285 set_die_routine(die_nicely);
3286 set_checkpoint_signal();
3287 while (read_next_command() != EOF) {
3288 if (!strcmp("blob", command_buf.buf))
3289 parse_new_blob();
3290 else if (!prefixcmp(command_buf.buf, "ls "))
3291 parse_ls(NULL);
3292 else if (!prefixcmp(command_buf.buf, "commit "))
3293 parse_new_commit();
3294 else if (!prefixcmp(command_buf.buf, "tag "))
3295 parse_new_tag();
3296 else if (!prefixcmp(command_buf.buf, "reset "))
3297 parse_reset_branch();
3298 else if (!strcmp("checkpoint", command_buf.buf))
3299 parse_checkpoint();
3300 else if (!prefixcmp(command_buf.buf, "progress "))
3301 parse_progress();
3302 else if (!prefixcmp(command_buf.buf, "feature "))
3303 parse_feature();
3304 else if (!prefixcmp(command_buf.buf, "option git "))
3305 parse_option();
3306 else if (!prefixcmp(command_buf.buf, "option "))
3307 /* ignore non-git options*/;
3308 else
3309 die("Unsupported command: %s", command_buf.buf);
3311 if (checkpoint_requested)
3312 checkpoint();
3315 /* argv hasn't been parsed yet, do so */
3316 if (!seen_data_command)
3317 parse_argv();
3319 end_packfile();
3321 dump_branches();
3322 dump_tags();
3323 unkeep_all_packs();
3324 dump_marks();
3326 if (pack_edges)
3327 fclose(pack_edges);
3329 if (show_stats) {
3330 uintmax_t total_count = 0, duplicate_count = 0;
3331 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3332 total_count += object_count_by_type[i];
3333 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3334 duplicate_count += duplicate_count_by_type[i];
3336 fprintf(stderr, "%s statistics:\n", argv[0]);
3337 fprintf(stderr, "---------------------------------------------------------------------\n");
3338 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3339 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3340 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]);
3341 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]);
3342 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]);
3343 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]);
3344 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3345 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3346 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3347 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3348 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3349 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3350 fprintf(stderr, "---------------------------------------------------------------------\n");
3351 pack_report();
3352 fprintf(stderr, "---------------------------------------------------------------------\n");
3353 fprintf(stderr, "\n");
3356 return failure ? 1 : 0;