Merge branch 'ak/git-done-help-cleanup' into maint
[git.git] / fast-import.c
blobaac2c24f74e496d11b66e1cc2806c55555f71496
1 /*
2 (See Documentation/git-fast-import.txt for maintained documentation.)
3 Format of STDIN stream:
5 stream ::= cmd*;
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
20 new_commit ::= 'commit' sp ref_str lf
21 mark?
22 ('author' (sp name)? sp '<' email '>' sp when lf)?
23 'committer' (sp name)? sp '<' email '>' sp when lf
24 commit_msg
25 ('from' sp commit-ish lf)?
26 ('merge' sp commit-ish lf)*
27 (file_change | ls)*
28 lf?;
29 commit_msg ::= data;
31 ls ::= 'ls' sp '"' quoted(path) '"' lf;
33 file_change ::= file_clr
34 | file_del
35 | file_rnm
36 | file_cpy
37 | file_obm
38 | file_inm;
39 file_clr ::= 'deleteall' lf;
40 file_del ::= 'D' sp path_str lf;
41 file_rnm ::= 'R' sp path_str sp path_str lf;
42 file_cpy ::= 'C' sp path_str sp path_str lf;
43 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
44 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
45 data;
46 note_obm ::= 'N' sp (hexsha1 | idnum) sp commit-ish lf;
47 note_inm ::= 'N' sp 'inline' sp commit-ish lf
48 data;
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp commit-ish lf
52 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
53 tag_msg;
54 tag_msg ::= data;
56 reset_branch ::= 'reset' sp ref_str lf
57 ('from' sp commit-ish lf)?
58 lf?;
60 checkpoint ::= 'checkpoint' lf
61 lf?;
63 progress ::= 'progress' sp not_lf* lf
64 lf?;
66 # note: the first idnum in a stream should be 1 and subsequent
67 # idnums should not have gaps between values as this will cause
68 # the stream parser to reserve space for the gapped values. An
69 # idnum can be updated in the future to a new object by issuing
70 # a new mark directive with the old idnum.
72 mark ::= 'mark' sp idnum lf;
73 data ::= (delimited_data | exact_data)
74 lf?;
76 # note: delim may be any string but must not contain lf.
77 # data_line may contain any data but must not be exactly
78 # delim.
79 delimited_data ::= 'data' sp '<<' delim lf
80 (data_line lf)*
81 delim lf;
83 # note: declen indicates the length of binary_data in bytes.
84 # declen does not include the lf preceding the binary data.
86 exact_data ::= 'data' sp declen lf
87 binary_data;
89 # note: quoted strings are C-style quoting supporting \c for
90 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
91 # is the signed byte value in octal. Note that the only
92 # characters which must actually be escaped to protect the
93 # stream formatting is: \, " and LF. Otherwise these values
94 # are UTF8.
96 commit-ish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
97 ref_str ::= ref;
98 sha1exp_str ::= sha1exp;
99 tag_str ::= tag;
100 path_str ::= path | '"' quoted(path) '"' ;
101 mode ::= '100644' | '644'
102 | '100755' | '755'
103 | '120000'
106 declen ::= # unsigned 32 bit value, ascii base10 notation;
107 bigint ::= # unsigned integer value, ascii base10 notation;
108 binary_data ::= # file content, not interpreted;
110 when ::= raw_when | rfc2822_when;
111 raw_when ::= ts sp tz;
112 rfc2822_when ::= # Valid RFC 2822 date and time;
114 sp ::= # ASCII space character;
115 lf ::= # ASCII newline (LF) character;
117 # note: a colon (':') must precede the numerical value assigned to
118 # an idnum. This is to distinguish it from a ref or tag name as
119 # GIT does not permit ':' in ref or tag strings.
121 idnum ::= ':' bigint;
122 path ::= # GIT style file path, e.g. "a/b/c";
123 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
124 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
125 sha1exp ::= # Any valid GIT SHA1 expression;
126 hexsha1 ::= # SHA1 in hexadecimal format;
128 # note: name and email are UTF8 strings, however name must not
129 # contain '<' or lf and email must not contain any of the
130 # following: '<', '>', lf.
132 name ::= # valid GIT author/committer name;
133 email ::= # valid GIT author/committer email;
134 ts ::= # time since the epoch in seconds, ascii base10 notation;
135 tz ::= # GIT style timezone;
137 # note: comments, ls and cat requests may appear anywhere
138 # in the input, except within a data command. Any form
139 # of the data command always escapes the related input
140 # from comment processing.
142 # In case it is not clear, the '#' that starts the comment
143 # must be the first character on that line (an lf
144 # preceded it).
147 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
148 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
150 comment ::= '#' not_lf* lf;
151 not_lf ::= # Any byte that is not ASCII newline (LF);
154 #include "builtin.h"
155 #include "cache.h"
156 #include "lockfile.h"
157 #include "object.h"
158 #include "blob.h"
159 #include "tree.h"
160 #include "commit.h"
161 #include "delta.h"
162 #include "pack.h"
163 #include "refs.h"
164 #include "csum-file.h"
165 #include "quote.h"
166 #include "exec_cmd.h"
167 #include "dir.h"
169 #define PACK_ID_BITS 16
170 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
171 #define DEPTH_BITS 13
172 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
175 * We abuse the setuid bit on directories to mean "do not delta".
177 #define NO_DELTA S_ISUID
179 struct object_entry {
180 struct pack_idx_entry idx;
181 struct object_entry *next;
182 uint32_t type : TYPE_BITS,
183 pack_id : PACK_ID_BITS,
184 depth : DEPTH_BITS;
187 struct object_entry_pool {
188 struct object_entry_pool *next_pool;
189 struct object_entry *next_free;
190 struct object_entry *end;
191 struct object_entry entries[FLEX_ARRAY]; /* more */
194 struct mark_set {
195 union {
196 struct object_entry *marked[1024];
197 struct mark_set *sets[1024];
198 } data;
199 unsigned int shift;
202 struct last_object {
203 struct strbuf data;
204 off_t offset;
205 unsigned int depth;
206 unsigned no_swap : 1;
209 struct mem_pool {
210 struct mem_pool *next_pool;
211 char *next_free;
212 char *end;
213 uintmax_t space[FLEX_ARRAY]; /* more */
216 struct atom_str {
217 struct atom_str *next_atom;
218 unsigned short str_len;
219 char str_dat[FLEX_ARRAY]; /* more */
222 struct tree_content;
223 struct tree_entry {
224 struct tree_content *tree;
225 struct atom_str *name;
226 struct tree_entry_ms {
227 uint16_t mode;
228 unsigned char sha1[20];
229 } versions[2];
232 struct tree_content {
233 unsigned int entry_capacity; /* must match avail_tree_content */
234 unsigned int entry_count;
235 unsigned int delta_depth;
236 struct tree_entry *entries[FLEX_ARRAY]; /* more */
239 struct avail_tree_content {
240 unsigned int entry_capacity; /* must match tree_content */
241 struct avail_tree_content *next_avail;
244 struct branch {
245 struct branch *table_next_branch;
246 struct branch *active_next_branch;
247 const char *name;
248 struct tree_entry branch_tree;
249 uintmax_t last_commit;
250 uintmax_t num_notes;
251 unsigned active : 1;
252 unsigned delete : 1;
253 unsigned pack_id : PACK_ID_BITS;
254 unsigned char sha1[20];
257 struct tag {
258 struct tag *next_tag;
259 const char *name;
260 unsigned int pack_id;
261 unsigned char sha1[20];
264 struct hash_list {
265 struct hash_list *next;
266 unsigned char sha1[20];
269 typedef enum {
270 WHENSPEC_RAW = 1,
271 WHENSPEC_RFC2822,
272 WHENSPEC_NOW
273 } whenspec_type;
275 struct recent_command {
276 struct recent_command *prev;
277 struct recent_command *next;
278 char *buf;
281 /* Configured limits on output */
282 static unsigned long max_depth = 10;
283 static off_t max_packsize;
284 static int force_update;
285 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
286 static int pack_compression_seen;
288 /* Stats and misc. counters */
289 static uintmax_t alloc_count;
290 static uintmax_t marks_set_count;
291 static uintmax_t object_count_by_type[1 << TYPE_BITS];
292 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
293 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
294 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
295 static unsigned long object_count;
296 static unsigned long branch_count;
297 static unsigned long branch_load_count;
298 static int failure;
299 static FILE *pack_edges;
300 static unsigned int show_stats = 1;
301 static int global_argc;
302 static char **global_argv;
304 /* Memory pools */
305 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
306 static size_t total_allocd;
307 static struct mem_pool *mem_pool;
309 /* Atom management */
310 static unsigned int atom_table_sz = 4451;
311 static unsigned int atom_cnt;
312 static struct atom_str **atom_table;
314 /* The .pack file being generated */
315 static struct pack_idx_option pack_idx_opts;
316 static unsigned int pack_id;
317 static struct sha1file *pack_file;
318 static struct packed_git *pack_data;
319 static struct packed_git **all_packs;
320 static off_t pack_size;
322 /* Table of objects we've written. */
323 static unsigned int object_entry_alloc = 5000;
324 static struct object_entry_pool *blocks;
325 static struct object_entry *object_table[1 << 16];
326 static struct mark_set *marks;
327 static const char *export_marks_file;
328 static const char *import_marks_file;
329 static int import_marks_file_from_stream;
330 static int import_marks_file_ignore_missing;
331 static int relative_marks_paths;
333 /* Our last blob */
334 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
336 /* Tree management */
337 static unsigned int tree_entry_alloc = 1000;
338 static void *avail_tree_entry;
339 static unsigned int avail_tree_table_sz = 100;
340 static struct avail_tree_content **avail_tree_table;
341 static struct strbuf old_tree = STRBUF_INIT;
342 static struct strbuf new_tree = STRBUF_INIT;
344 /* Branch data */
345 static unsigned long max_active_branches = 5;
346 static unsigned long cur_active_branches;
347 static unsigned long branch_table_sz = 1039;
348 static struct branch **branch_table;
349 static struct branch *active_branches;
351 /* Tag data */
352 static struct tag *first_tag;
353 static struct tag *last_tag;
355 /* Input stream parsing */
356 static whenspec_type whenspec = WHENSPEC_RAW;
357 static struct strbuf command_buf = STRBUF_INIT;
358 static int unread_command_buf;
359 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
360 static struct recent_command *cmd_tail = &cmd_hist;
361 static struct recent_command *rc_free;
362 static unsigned int cmd_save = 100;
363 static uintmax_t next_mark;
364 static struct strbuf new_data = STRBUF_INIT;
365 static int seen_data_command;
366 static int require_explicit_termination;
368 /* Signal handling */
369 static volatile sig_atomic_t checkpoint_requested;
371 /* Where to write output of cat-blob commands */
372 static int cat_blob_fd = STDOUT_FILENO;
374 static void parse_argv(void);
375 static void parse_cat_blob(const char *p);
376 static void parse_ls(const char *p, struct branch *b);
378 static void write_branch_report(FILE *rpt, struct branch *b)
380 fprintf(rpt, "%s:\n", b->name);
382 fprintf(rpt, " status :");
383 if (b->active)
384 fputs(" active", rpt);
385 if (b->branch_tree.tree)
386 fputs(" loaded", rpt);
387 if (is_null_sha1(b->branch_tree.versions[1].sha1))
388 fputs(" dirty", rpt);
389 fputc('\n', rpt);
391 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
392 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
393 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
394 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
396 fputs(" last pack : ", rpt);
397 if (b->pack_id < MAX_PACK_ID)
398 fprintf(rpt, "%u", b->pack_id);
399 fputc('\n', rpt);
401 fputc('\n', rpt);
404 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
406 static void write_crash_report(const char *err)
408 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
409 FILE *rpt = fopen(loc, "w");
410 struct branch *b;
411 unsigned long lu;
412 struct recent_command *rc;
414 if (!rpt) {
415 error("can't write crash report %s: %s", loc, strerror(errno));
416 return;
419 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
421 fprintf(rpt, "fast-import crash report:\n");
422 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
423 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
424 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
425 fputc('\n', rpt);
427 fputs("fatal: ", rpt);
428 fputs(err, rpt);
429 fputc('\n', rpt);
431 fputc('\n', rpt);
432 fputs("Most Recent Commands Before Crash\n", rpt);
433 fputs("---------------------------------\n", rpt);
434 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
435 if (rc->next == &cmd_hist)
436 fputs("* ", rpt);
437 else
438 fputs(" ", rpt);
439 fputs(rc->buf, rpt);
440 fputc('\n', rpt);
443 fputc('\n', rpt);
444 fputs("Active Branch LRU\n", rpt);
445 fputs("-----------------\n", rpt);
446 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
447 cur_active_branches,
448 max_active_branches);
449 fputc('\n', rpt);
450 fputs(" pos clock name\n", rpt);
451 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
452 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
453 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
454 ++lu, b->last_commit, b->name);
456 fputc('\n', rpt);
457 fputs("Inactive Branches\n", rpt);
458 fputs("-----------------\n", rpt);
459 for (lu = 0; lu < branch_table_sz; lu++) {
460 for (b = branch_table[lu]; b; b = b->table_next_branch)
461 write_branch_report(rpt, b);
464 if (first_tag) {
465 struct tag *tg;
466 fputc('\n', rpt);
467 fputs("Annotated Tags\n", rpt);
468 fputs("--------------\n", rpt);
469 for (tg = first_tag; tg; tg = tg->next_tag) {
470 fputs(sha1_to_hex(tg->sha1), rpt);
471 fputc(' ', rpt);
472 fputs(tg->name, rpt);
473 fputc('\n', rpt);
477 fputc('\n', rpt);
478 fputs("Marks\n", rpt);
479 fputs("-----\n", rpt);
480 if (export_marks_file)
481 fprintf(rpt, " exported to %s\n", export_marks_file);
482 else
483 dump_marks_helper(rpt, 0, marks);
485 fputc('\n', rpt);
486 fputs("-------------------\n", rpt);
487 fputs("END OF CRASH REPORT\n", rpt);
488 fclose(rpt);
491 static void end_packfile(void);
492 static void unkeep_all_packs(void);
493 static void dump_marks(void);
495 static NORETURN void die_nicely(const char *err, va_list params)
497 static int zombie;
498 char message[2 * PATH_MAX];
500 vsnprintf(message, sizeof(message), err, params);
501 fputs("fatal: ", stderr);
502 fputs(message, stderr);
503 fputc('\n', stderr);
505 if (!zombie) {
506 zombie = 1;
507 write_crash_report(message);
508 end_packfile();
509 unkeep_all_packs();
510 dump_marks();
512 exit(128);
515 #ifndef SIGUSR1 /* Windows, for example */
517 static void set_checkpoint_signal(void)
521 #else
523 static void checkpoint_signal(int signo)
525 checkpoint_requested = 1;
528 static void set_checkpoint_signal(void)
530 struct sigaction sa;
532 memset(&sa, 0, sizeof(sa));
533 sa.sa_handler = checkpoint_signal;
534 sigemptyset(&sa.sa_mask);
535 sa.sa_flags = SA_RESTART;
536 sigaction(SIGUSR1, &sa, NULL);
539 #endif
541 static void alloc_objects(unsigned int cnt)
543 struct object_entry_pool *b;
545 b = xmalloc(sizeof(struct object_entry_pool)
546 + cnt * sizeof(struct object_entry));
547 b->next_pool = blocks;
548 b->next_free = b->entries;
549 b->end = b->entries + cnt;
550 blocks = b;
551 alloc_count += cnt;
554 static struct object_entry *new_object(unsigned char *sha1)
556 struct object_entry *e;
558 if (blocks->next_free == blocks->end)
559 alloc_objects(object_entry_alloc);
561 e = blocks->next_free++;
562 hashcpy(e->idx.sha1, sha1);
563 return e;
566 static struct object_entry *find_object(unsigned char *sha1)
568 unsigned int h = sha1[0] << 8 | sha1[1];
569 struct object_entry *e;
570 for (e = object_table[h]; e; e = e->next)
571 if (!hashcmp(sha1, e->idx.sha1))
572 return e;
573 return NULL;
576 static struct object_entry *insert_object(unsigned char *sha1)
578 unsigned int h = sha1[0] << 8 | sha1[1];
579 struct object_entry *e = object_table[h];
581 while (e) {
582 if (!hashcmp(sha1, e->idx.sha1))
583 return e;
584 e = e->next;
587 e = new_object(sha1);
588 e->next = object_table[h];
589 e->idx.offset = 0;
590 object_table[h] = e;
591 return e;
594 static unsigned int hc_str(const char *s, size_t len)
596 unsigned int r = 0;
597 while (len-- > 0)
598 r = r * 31 + *s++;
599 return r;
602 static void *pool_alloc(size_t len)
604 struct mem_pool *p;
605 void *r;
607 /* round up to a 'uintmax_t' alignment */
608 if (len & (sizeof(uintmax_t) - 1))
609 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
611 for (p = mem_pool; p; p = p->next_pool)
612 if ((p->end - p->next_free >= len))
613 break;
615 if (!p) {
616 if (len >= (mem_pool_alloc/2)) {
617 total_allocd += len;
618 return xmalloc(len);
620 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
621 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
622 p->next_pool = mem_pool;
623 p->next_free = (char *) p->space;
624 p->end = p->next_free + mem_pool_alloc;
625 mem_pool = p;
628 r = p->next_free;
629 p->next_free += len;
630 return r;
633 static void *pool_calloc(size_t count, size_t size)
635 size_t len = count * size;
636 void *r = pool_alloc(len);
637 memset(r, 0, len);
638 return r;
641 static char *pool_strdup(const char *s)
643 char *r = pool_alloc(strlen(s) + 1);
644 strcpy(r, s);
645 return r;
648 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
650 struct mark_set *s = marks;
651 while ((idnum >> s->shift) >= 1024) {
652 s = pool_calloc(1, sizeof(struct mark_set));
653 s->shift = marks->shift + 10;
654 s->data.sets[0] = marks;
655 marks = s;
657 while (s->shift) {
658 uintmax_t i = idnum >> s->shift;
659 idnum -= i << s->shift;
660 if (!s->data.sets[i]) {
661 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
662 s->data.sets[i]->shift = s->shift - 10;
664 s = s->data.sets[i];
666 if (!s->data.marked[idnum])
667 marks_set_count++;
668 s->data.marked[idnum] = oe;
671 static struct object_entry *find_mark(uintmax_t idnum)
673 uintmax_t orig_idnum = idnum;
674 struct mark_set *s = marks;
675 struct object_entry *oe = NULL;
676 if ((idnum >> s->shift) < 1024) {
677 while (s && s->shift) {
678 uintmax_t i = idnum >> s->shift;
679 idnum -= i << s->shift;
680 s = s->data.sets[i];
682 if (s)
683 oe = s->data.marked[idnum];
685 if (!oe)
686 die("mark :%" PRIuMAX " not declared", orig_idnum);
687 return oe;
690 static struct atom_str *to_atom(const char *s, unsigned short len)
692 unsigned int hc = hc_str(s, len) % atom_table_sz;
693 struct atom_str *c;
695 for (c = atom_table[hc]; c; c = c->next_atom)
696 if (c->str_len == len && !strncmp(s, c->str_dat, len))
697 return c;
699 c = pool_alloc(sizeof(struct atom_str) + len + 1);
700 c->str_len = len;
701 strncpy(c->str_dat, s, len);
702 c->str_dat[len] = 0;
703 c->next_atom = atom_table[hc];
704 atom_table[hc] = c;
705 atom_cnt++;
706 return c;
709 static struct branch *lookup_branch(const char *name)
711 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
712 struct branch *b;
714 for (b = branch_table[hc]; b; b = b->table_next_branch)
715 if (!strcmp(name, b->name))
716 return b;
717 return NULL;
720 static struct branch *new_branch(const char *name)
722 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
723 struct branch *b = lookup_branch(name);
725 if (b)
726 die("Invalid attempt to create duplicate branch: %s", name);
727 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
728 die("Branch name doesn't conform to GIT standards: %s", name);
730 b = pool_calloc(1, sizeof(struct branch));
731 b->name = pool_strdup(name);
732 b->table_next_branch = branch_table[hc];
733 b->branch_tree.versions[0].mode = S_IFDIR;
734 b->branch_tree.versions[1].mode = S_IFDIR;
735 b->num_notes = 0;
736 b->active = 0;
737 b->pack_id = MAX_PACK_ID;
738 branch_table[hc] = b;
739 branch_count++;
740 return b;
743 static unsigned int hc_entries(unsigned int cnt)
745 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
746 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
749 static struct tree_content *new_tree_content(unsigned int cnt)
751 struct avail_tree_content *f, *l = NULL;
752 struct tree_content *t;
753 unsigned int hc = hc_entries(cnt);
755 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
756 if (f->entry_capacity >= cnt)
757 break;
759 if (f) {
760 if (l)
761 l->next_avail = f->next_avail;
762 else
763 avail_tree_table[hc] = f->next_avail;
764 } else {
765 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
766 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
767 f->entry_capacity = cnt;
770 t = (struct tree_content*)f;
771 t->entry_count = 0;
772 t->delta_depth = 0;
773 return t;
776 static void release_tree_entry(struct tree_entry *e);
777 static void release_tree_content(struct tree_content *t)
779 struct avail_tree_content *f = (struct avail_tree_content*)t;
780 unsigned int hc = hc_entries(f->entry_capacity);
781 f->next_avail = avail_tree_table[hc];
782 avail_tree_table[hc] = f;
785 static void release_tree_content_recursive(struct tree_content *t)
787 unsigned int i;
788 for (i = 0; i < t->entry_count; i++)
789 release_tree_entry(t->entries[i]);
790 release_tree_content(t);
793 static struct tree_content *grow_tree_content(
794 struct tree_content *t,
795 int amt)
797 struct tree_content *r = new_tree_content(t->entry_count + amt);
798 r->entry_count = t->entry_count;
799 r->delta_depth = t->delta_depth;
800 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
801 release_tree_content(t);
802 return r;
805 static struct tree_entry *new_tree_entry(void)
807 struct tree_entry *e;
809 if (!avail_tree_entry) {
810 unsigned int n = tree_entry_alloc;
811 total_allocd += n * sizeof(struct tree_entry);
812 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
813 while (n-- > 1) {
814 *((void**)e) = e + 1;
815 e++;
817 *((void**)e) = NULL;
820 e = avail_tree_entry;
821 avail_tree_entry = *((void**)e);
822 return e;
825 static void release_tree_entry(struct tree_entry *e)
827 if (e->tree)
828 release_tree_content_recursive(e->tree);
829 *((void**)e) = avail_tree_entry;
830 avail_tree_entry = e;
833 static struct tree_content *dup_tree_content(struct tree_content *s)
835 struct tree_content *d;
836 struct tree_entry *a, *b;
837 unsigned int i;
839 if (!s)
840 return NULL;
841 d = new_tree_content(s->entry_count);
842 for (i = 0; i < s->entry_count; i++) {
843 a = s->entries[i];
844 b = new_tree_entry();
845 memcpy(b, a, sizeof(*a));
846 if (a->tree && is_null_sha1(b->versions[1].sha1))
847 b->tree = dup_tree_content(a->tree);
848 else
849 b->tree = NULL;
850 d->entries[i] = b;
852 d->entry_count = s->entry_count;
853 d->delta_depth = s->delta_depth;
855 return d;
858 static void start_packfile(void)
860 static char tmp_file[PATH_MAX];
861 struct packed_git *p;
862 struct pack_header hdr;
863 int pack_fd;
865 pack_fd = odb_mkstemp(tmp_file, sizeof(tmp_file),
866 "pack/tmp_pack_XXXXXX");
867 p = xcalloc(1, sizeof(*p) + strlen(tmp_file) + 2);
868 strcpy(p->pack_name, tmp_file);
869 p->pack_fd = pack_fd;
870 p->do_not_close = 1;
871 pack_file = sha1fd(pack_fd, p->pack_name);
873 hdr.hdr_signature = htonl(PACK_SIGNATURE);
874 hdr.hdr_version = htonl(2);
875 hdr.hdr_entries = 0;
876 sha1write(pack_file, &hdr, sizeof(hdr));
878 pack_data = p;
879 pack_size = sizeof(hdr);
880 object_count = 0;
882 REALLOC_ARRAY(all_packs, pack_id + 1);
883 all_packs[pack_id] = p;
886 static const char *create_index(void)
888 const char *tmpfile;
889 struct pack_idx_entry **idx, **c, **last;
890 struct object_entry *e;
891 struct object_entry_pool *o;
893 /* Build the table of object IDs. */
894 idx = xmalloc(object_count * sizeof(*idx));
895 c = idx;
896 for (o = blocks; o; o = o->next_pool)
897 for (e = o->next_free; e-- != o->entries;)
898 if (pack_id == e->pack_id)
899 *c++ = &e->idx;
900 last = idx + object_count;
901 if (c != last)
902 die("internal consistency error creating the index");
904 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
905 free(idx);
906 return tmpfile;
909 static char *keep_pack(const char *curr_index_name)
911 static char name[PATH_MAX];
912 static const char *keep_msg = "fast-import";
913 int keep_fd;
915 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
916 if (keep_fd < 0)
917 die_errno("cannot create keep file");
918 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
919 if (close(keep_fd))
920 die_errno("failed to write keep file");
922 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
923 get_object_directory(), sha1_to_hex(pack_data->sha1));
924 if (move_temp_to_file(pack_data->pack_name, name))
925 die("cannot store pack file");
927 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
928 get_object_directory(), sha1_to_hex(pack_data->sha1));
929 if (move_temp_to_file(curr_index_name, name))
930 die("cannot store index file");
931 free((void *)curr_index_name);
932 return name;
935 static void unkeep_all_packs(void)
937 static char name[PATH_MAX];
938 int k;
940 for (k = 0; k < pack_id; k++) {
941 struct packed_git *p = all_packs[k];
942 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
943 get_object_directory(), sha1_to_hex(p->sha1));
944 unlink_or_warn(name);
948 static void end_packfile(void)
950 static int running;
952 if (running || !pack_data)
953 return;
955 running = 1;
956 clear_delta_base_cache();
957 if (object_count) {
958 struct packed_git *new_p;
959 unsigned char cur_pack_sha1[20];
960 char *idx_name;
961 int i;
962 struct branch *b;
963 struct tag *t;
965 close_pack_windows(pack_data);
966 sha1close(pack_file, cur_pack_sha1, 0);
967 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
968 pack_data->pack_name, object_count,
969 cur_pack_sha1, pack_size);
970 close(pack_data->pack_fd);
971 idx_name = keep_pack(create_index());
973 /* Register the packfile with core git's machinery. */
974 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
975 if (!new_p)
976 die("core git rejected index %s", idx_name);
977 all_packs[pack_id] = new_p;
978 install_packed_git(new_p);
980 /* Print the boundary */
981 if (pack_edges) {
982 fprintf(pack_edges, "%s:", new_p->pack_name);
983 for (i = 0; i < branch_table_sz; i++) {
984 for (b = branch_table[i]; b; b = b->table_next_branch) {
985 if (b->pack_id == pack_id)
986 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
989 for (t = first_tag; t; t = t->next_tag) {
990 if (t->pack_id == pack_id)
991 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
993 fputc('\n', pack_edges);
994 fflush(pack_edges);
997 pack_id++;
999 else {
1000 close(pack_data->pack_fd);
1001 unlink_or_warn(pack_data->pack_name);
1003 free(pack_data);
1004 pack_data = NULL;
1005 running = 0;
1007 /* We can't carry a delta across packfiles. */
1008 strbuf_release(&last_blob.data);
1009 last_blob.offset = 0;
1010 last_blob.depth = 0;
1013 static void cycle_packfile(void)
1015 end_packfile();
1016 start_packfile();
1019 static int store_object(
1020 enum object_type type,
1021 struct strbuf *dat,
1022 struct last_object *last,
1023 unsigned char *sha1out,
1024 uintmax_t mark)
1026 void *out, *delta;
1027 struct object_entry *e;
1028 unsigned char hdr[96];
1029 unsigned char sha1[20];
1030 unsigned long hdrlen, deltalen;
1031 git_SHA_CTX c;
1032 git_zstream s;
1034 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1035 (unsigned long)dat->len) + 1;
1036 git_SHA1_Init(&c);
1037 git_SHA1_Update(&c, hdr, hdrlen);
1038 git_SHA1_Update(&c, dat->buf, dat->len);
1039 git_SHA1_Final(sha1, &c);
1040 if (sha1out)
1041 hashcpy(sha1out, sha1);
1043 e = insert_object(sha1);
1044 if (mark)
1045 insert_mark(mark, e);
1046 if (e->idx.offset) {
1047 duplicate_count_by_type[type]++;
1048 return 1;
1049 } else if (find_sha1_pack(sha1, packed_git)) {
1050 e->type = type;
1051 e->pack_id = MAX_PACK_ID;
1052 e->idx.offset = 1; /* just not zero! */
1053 duplicate_count_by_type[type]++;
1054 return 1;
1057 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1058 delta_count_attempts_by_type[type]++;
1059 delta = diff_delta(last->data.buf, last->data.len,
1060 dat->buf, dat->len,
1061 &deltalen, dat->len - 20);
1062 } else
1063 delta = NULL;
1065 memset(&s, 0, sizeof(s));
1066 git_deflate_init(&s, pack_compression_level);
1067 if (delta) {
1068 s.next_in = delta;
1069 s.avail_in = deltalen;
1070 } else {
1071 s.next_in = (void *)dat->buf;
1072 s.avail_in = dat->len;
1074 s.avail_out = git_deflate_bound(&s, s.avail_in);
1075 s.next_out = out = xmalloc(s.avail_out);
1076 while (git_deflate(&s, Z_FINISH) == Z_OK)
1077 ; /* nothing */
1078 git_deflate_end(&s);
1080 /* Determine if we should auto-checkpoint. */
1081 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1082 || (pack_size + 60 + s.total_out) < pack_size) {
1084 /* This new object needs to *not* have the current pack_id. */
1085 e->pack_id = pack_id + 1;
1086 cycle_packfile();
1088 /* We cannot carry a delta into the new pack. */
1089 if (delta) {
1090 free(delta);
1091 delta = NULL;
1093 memset(&s, 0, sizeof(s));
1094 git_deflate_init(&s, pack_compression_level);
1095 s.next_in = (void *)dat->buf;
1096 s.avail_in = dat->len;
1097 s.avail_out = git_deflate_bound(&s, s.avail_in);
1098 s.next_out = out = xrealloc(out, s.avail_out);
1099 while (git_deflate(&s, Z_FINISH) == Z_OK)
1100 ; /* nothing */
1101 git_deflate_end(&s);
1105 e->type = type;
1106 e->pack_id = pack_id;
1107 e->idx.offset = pack_size;
1108 object_count++;
1109 object_count_by_type[type]++;
1111 crc32_begin(pack_file);
1113 if (delta) {
1114 off_t ofs = e->idx.offset - last->offset;
1115 unsigned pos = sizeof(hdr) - 1;
1117 delta_count_by_type[type]++;
1118 e->depth = last->depth + 1;
1120 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1121 sha1write(pack_file, hdr, hdrlen);
1122 pack_size += hdrlen;
1124 hdr[pos] = ofs & 127;
1125 while (ofs >>= 7)
1126 hdr[--pos] = 128 | (--ofs & 127);
1127 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1128 pack_size += sizeof(hdr) - pos;
1129 } else {
1130 e->depth = 0;
1131 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1132 sha1write(pack_file, hdr, hdrlen);
1133 pack_size += hdrlen;
1136 sha1write(pack_file, out, s.total_out);
1137 pack_size += s.total_out;
1139 e->idx.crc32 = crc32_end(pack_file);
1141 free(out);
1142 free(delta);
1143 if (last) {
1144 if (last->no_swap) {
1145 last->data = *dat;
1146 } else {
1147 strbuf_swap(&last->data, dat);
1149 last->offset = e->idx.offset;
1150 last->depth = e->depth;
1152 return 0;
1155 static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1157 if (sha1file_truncate(pack_file, checkpoint))
1158 die_errno("cannot truncate pack to skip duplicate");
1159 pack_size = checkpoint->offset;
1162 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1164 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1165 unsigned char *in_buf = xmalloc(in_sz);
1166 unsigned char *out_buf = xmalloc(out_sz);
1167 struct object_entry *e;
1168 unsigned char sha1[20];
1169 unsigned long hdrlen;
1170 off_t offset;
1171 git_SHA_CTX c;
1172 git_zstream s;
1173 struct sha1file_checkpoint checkpoint;
1174 int status = Z_OK;
1176 /* Determine if we should auto-checkpoint. */
1177 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1178 || (pack_size + 60 + len) < pack_size)
1179 cycle_packfile();
1181 sha1file_checkpoint(pack_file, &checkpoint);
1182 offset = checkpoint.offset;
1184 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1185 if (out_sz <= hdrlen)
1186 die("impossibly large object header");
1188 git_SHA1_Init(&c);
1189 git_SHA1_Update(&c, out_buf, hdrlen);
1191 crc32_begin(pack_file);
1193 memset(&s, 0, sizeof(s));
1194 git_deflate_init(&s, pack_compression_level);
1196 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1197 if (out_sz <= hdrlen)
1198 die("impossibly large object header");
1200 s.next_out = out_buf + hdrlen;
1201 s.avail_out = out_sz - hdrlen;
1203 while (status != Z_STREAM_END) {
1204 if (0 < len && !s.avail_in) {
1205 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1206 size_t n = fread(in_buf, 1, cnt, stdin);
1207 if (!n && feof(stdin))
1208 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1210 git_SHA1_Update(&c, in_buf, n);
1211 s.next_in = in_buf;
1212 s.avail_in = n;
1213 len -= n;
1216 status = git_deflate(&s, len ? 0 : Z_FINISH);
1218 if (!s.avail_out || status == Z_STREAM_END) {
1219 size_t n = s.next_out - out_buf;
1220 sha1write(pack_file, out_buf, n);
1221 pack_size += n;
1222 s.next_out = out_buf;
1223 s.avail_out = out_sz;
1226 switch (status) {
1227 case Z_OK:
1228 case Z_BUF_ERROR:
1229 case Z_STREAM_END:
1230 continue;
1231 default:
1232 die("unexpected deflate failure: %d", status);
1235 git_deflate_end(&s);
1236 git_SHA1_Final(sha1, &c);
1238 if (sha1out)
1239 hashcpy(sha1out, sha1);
1241 e = insert_object(sha1);
1243 if (mark)
1244 insert_mark(mark, e);
1246 if (e->idx.offset) {
1247 duplicate_count_by_type[OBJ_BLOB]++;
1248 truncate_pack(&checkpoint);
1250 } else if (find_sha1_pack(sha1, packed_git)) {
1251 e->type = OBJ_BLOB;
1252 e->pack_id = MAX_PACK_ID;
1253 e->idx.offset = 1; /* just not zero! */
1254 duplicate_count_by_type[OBJ_BLOB]++;
1255 truncate_pack(&checkpoint);
1257 } else {
1258 e->depth = 0;
1259 e->type = OBJ_BLOB;
1260 e->pack_id = pack_id;
1261 e->idx.offset = offset;
1262 e->idx.crc32 = crc32_end(pack_file);
1263 object_count++;
1264 object_count_by_type[OBJ_BLOB]++;
1267 free(in_buf);
1268 free(out_buf);
1271 /* All calls must be guarded by find_object() or find_mark() to
1272 * ensure the 'struct object_entry' passed was written by this
1273 * process instance. We unpack the entry by the offset, avoiding
1274 * the need for the corresponding .idx file. This unpacking rule
1275 * works because we only use OBJ_REF_DELTA within the packfiles
1276 * created by fast-import.
1278 * oe must not be NULL. Such an oe usually comes from giving
1279 * an unknown SHA-1 to find_object() or an undefined mark to
1280 * find_mark(). Callers must test for this condition and use
1281 * the standard read_sha1_file() when it happens.
1283 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1284 * find_mark(), where the mark was reloaded from an existing marks
1285 * file and is referencing an object that this fast-import process
1286 * instance did not write out to a packfile. Callers must test for
1287 * this condition and use read_sha1_file() instead.
1289 static void *gfi_unpack_entry(
1290 struct object_entry *oe,
1291 unsigned long *sizep)
1293 enum object_type type;
1294 struct packed_git *p = all_packs[oe->pack_id];
1295 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1296 /* The object is stored in the packfile we are writing to
1297 * and we have modified it since the last time we scanned
1298 * back to read a previously written object. If an old
1299 * window covered [p->pack_size, p->pack_size + 20) its
1300 * data is stale and is not valid. Closing all windows
1301 * and updating the packfile length ensures we can read
1302 * the newly written data.
1304 close_pack_windows(p);
1305 sha1flush(pack_file);
1307 /* We have to offer 20 bytes additional on the end of
1308 * the packfile as the core unpacker code assumes the
1309 * footer is present at the file end and must promise
1310 * at least 20 bytes within any window it maps. But
1311 * we don't actually create the footer here.
1313 p->pack_size = pack_size + 20;
1315 return unpack_entry(p, oe->idx.offset, &type, sizep);
1318 static const char *get_mode(const char *str, uint16_t *modep)
1320 unsigned char c;
1321 uint16_t mode = 0;
1323 while ((c = *str++) != ' ') {
1324 if (c < '0' || c > '7')
1325 return NULL;
1326 mode = (mode << 3) + (c - '0');
1328 *modep = mode;
1329 return str;
1332 static void load_tree(struct tree_entry *root)
1334 unsigned char *sha1 = root->versions[1].sha1;
1335 struct object_entry *myoe;
1336 struct tree_content *t;
1337 unsigned long size;
1338 char *buf;
1339 const char *c;
1341 root->tree = t = new_tree_content(8);
1342 if (is_null_sha1(sha1))
1343 return;
1345 myoe = find_object(sha1);
1346 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1347 if (myoe->type != OBJ_TREE)
1348 die("Not a tree: %s", sha1_to_hex(sha1));
1349 t->delta_depth = myoe->depth;
1350 buf = gfi_unpack_entry(myoe, &size);
1351 if (!buf)
1352 die("Can't load tree %s", sha1_to_hex(sha1));
1353 } else {
1354 enum object_type type;
1355 buf = read_sha1_file(sha1, &type, &size);
1356 if (!buf || type != OBJ_TREE)
1357 die("Can't load tree %s", sha1_to_hex(sha1));
1360 c = buf;
1361 while (c != (buf + size)) {
1362 struct tree_entry *e = new_tree_entry();
1364 if (t->entry_count == t->entry_capacity)
1365 root->tree = t = grow_tree_content(t, t->entry_count);
1366 t->entries[t->entry_count++] = e;
1368 e->tree = NULL;
1369 c = get_mode(c, &e->versions[1].mode);
1370 if (!c)
1371 die("Corrupt mode in %s", sha1_to_hex(sha1));
1372 e->versions[0].mode = e->versions[1].mode;
1373 e->name = to_atom(c, strlen(c));
1374 c += e->name->str_len + 1;
1375 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1376 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1377 c += 20;
1379 free(buf);
1382 static int tecmp0 (const void *_a, const void *_b)
1384 struct tree_entry *a = *((struct tree_entry**)_a);
1385 struct tree_entry *b = *((struct tree_entry**)_b);
1386 return base_name_compare(
1387 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1388 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1391 static int tecmp1 (const void *_a, const void *_b)
1393 struct tree_entry *a = *((struct tree_entry**)_a);
1394 struct tree_entry *b = *((struct tree_entry**)_b);
1395 return base_name_compare(
1396 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1397 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1400 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1402 size_t maxlen = 0;
1403 unsigned int i;
1405 if (!v)
1406 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1407 else
1408 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1410 for (i = 0; i < t->entry_count; i++) {
1411 if (t->entries[i]->versions[v].mode)
1412 maxlen += t->entries[i]->name->str_len + 34;
1415 strbuf_reset(b);
1416 strbuf_grow(b, maxlen);
1417 for (i = 0; i < t->entry_count; i++) {
1418 struct tree_entry *e = t->entries[i];
1419 if (!e->versions[v].mode)
1420 continue;
1421 strbuf_addf(b, "%o %s%c",
1422 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1423 e->name->str_dat, '\0');
1424 strbuf_add(b, e->versions[v].sha1, 20);
1428 static void store_tree(struct tree_entry *root)
1430 struct tree_content *t;
1431 unsigned int i, j, del;
1432 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1433 struct object_entry *le = NULL;
1435 if (!is_null_sha1(root->versions[1].sha1))
1436 return;
1438 if (!root->tree)
1439 load_tree(root);
1440 t = root->tree;
1442 for (i = 0; i < t->entry_count; i++) {
1443 if (t->entries[i]->tree)
1444 store_tree(t->entries[i]);
1447 if (!(root->versions[0].mode & NO_DELTA))
1448 le = find_object(root->versions[0].sha1);
1449 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1450 mktree(t, 0, &old_tree);
1451 lo.data = old_tree;
1452 lo.offset = le->idx.offset;
1453 lo.depth = t->delta_depth;
1456 mktree(t, 1, &new_tree);
1457 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1459 t->delta_depth = lo.depth;
1460 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1461 struct tree_entry *e = t->entries[i];
1462 if (e->versions[1].mode) {
1463 e->versions[0].mode = e->versions[1].mode;
1464 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1465 t->entries[j++] = e;
1466 } else {
1467 release_tree_entry(e);
1468 del++;
1471 t->entry_count -= del;
1474 static void tree_content_replace(
1475 struct tree_entry *root,
1476 const unsigned char *sha1,
1477 const uint16_t mode,
1478 struct tree_content *newtree)
1480 if (!S_ISDIR(mode))
1481 die("Root cannot be a non-directory");
1482 hashclr(root->versions[0].sha1);
1483 hashcpy(root->versions[1].sha1, sha1);
1484 if (root->tree)
1485 release_tree_content_recursive(root->tree);
1486 root->tree = newtree;
1489 static int tree_content_set(
1490 struct tree_entry *root,
1491 const char *p,
1492 const unsigned char *sha1,
1493 const uint16_t mode,
1494 struct tree_content *subtree)
1496 struct tree_content *t;
1497 const char *slash1;
1498 unsigned int i, n;
1499 struct tree_entry *e;
1501 slash1 = strchrnul(p, '/');
1502 n = slash1 - p;
1503 if (!n)
1504 die("Empty path component found in input");
1505 if (!*slash1 && !S_ISDIR(mode) && subtree)
1506 die("Non-directories cannot have subtrees");
1508 if (!root->tree)
1509 load_tree(root);
1510 t = root->tree;
1511 for (i = 0; i < t->entry_count; i++) {
1512 e = t->entries[i];
1513 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1514 if (!*slash1) {
1515 if (!S_ISDIR(mode)
1516 && e->versions[1].mode == mode
1517 && !hashcmp(e->versions[1].sha1, sha1))
1518 return 0;
1519 e->versions[1].mode = mode;
1520 hashcpy(e->versions[1].sha1, sha1);
1521 if (e->tree)
1522 release_tree_content_recursive(e->tree);
1523 e->tree = subtree;
1526 * We need to leave e->versions[0].sha1 alone
1527 * to avoid modifying the preimage tree used
1528 * when writing out the parent directory.
1529 * But after replacing the subdir with a
1530 * completely different one, it's not a good
1531 * delta base any more, and besides, we've
1532 * thrown away the tree entries needed to
1533 * make a delta against it.
1535 * So let's just explicitly disable deltas
1536 * for the subtree.
1538 if (S_ISDIR(e->versions[0].mode))
1539 e->versions[0].mode |= NO_DELTA;
1541 hashclr(root->versions[1].sha1);
1542 return 1;
1544 if (!S_ISDIR(e->versions[1].mode)) {
1545 e->tree = new_tree_content(8);
1546 e->versions[1].mode = S_IFDIR;
1548 if (!e->tree)
1549 load_tree(e);
1550 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1551 hashclr(root->versions[1].sha1);
1552 return 1;
1554 return 0;
1558 if (t->entry_count == t->entry_capacity)
1559 root->tree = t = grow_tree_content(t, t->entry_count);
1560 e = new_tree_entry();
1561 e->name = to_atom(p, n);
1562 e->versions[0].mode = 0;
1563 hashclr(e->versions[0].sha1);
1564 t->entries[t->entry_count++] = e;
1565 if (*slash1) {
1566 e->tree = new_tree_content(8);
1567 e->versions[1].mode = S_IFDIR;
1568 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1569 } else {
1570 e->tree = subtree;
1571 e->versions[1].mode = mode;
1572 hashcpy(e->versions[1].sha1, sha1);
1574 hashclr(root->versions[1].sha1);
1575 return 1;
1578 static int tree_content_remove(
1579 struct tree_entry *root,
1580 const char *p,
1581 struct tree_entry *backup_leaf,
1582 int allow_root)
1584 struct tree_content *t;
1585 const char *slash1;
1586 unsigned int i, n;
1587 struct tree_entry *e;
1589 slash1 = strchrnul(p, '/');
1590 n = slash1 - p;
1592 if (!root->tree)
1593 load_tree(root);
1595 if (!*p && allow_root) {
1596 e = root;
1597 goto del_entry;
1600 t = root->tree;
1601 for (i = 0; i < t->entry_count; i++) {
1602 e = t->entries[i];
1603 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1604 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1606 * If p names a file in some subdirectory, and a
1607 * file or symlink matching the name of the
1608 * parent directory of p exists, then p cannot
1609 * exist and need not be deleted.
1611 return 1;
1612 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1613 goto del_entry;
1614 if (!e->tree)
1615 load_tree(e);
1616 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1617 for (n = 0; n < e->tree->entry_count; n++) {
1618 if (e->tree->entries[n]->versions[1].mode) {
1619 hashclr(root->versions[1].sha1);
1620 return 1;
1623 backup_leaf = NULL;
1624 goto del_entry;
1626 return 0;
1629 return 0;
1631 del_entry:
1632 if (backup_leaf)
1633 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1634 else if (e->tree)
1635 release_tree_content_recursive(e->tree);
1636 e->tree = NULL;
1637 e->versions[1].mode = 0;
1638 hashclr(e->versions[1].sha1);
1639 hashclr(root->versions[1].sha1);
1640 return 1;
1643 static int tree_content_get(
1644 struct tree_entry *root,
1645 const char *p,
1646 struct tree_entry *leaf,
1647 int allow_root)
1649 struct tree_content *t;
1650 const char *slash1;
1651 unsigned int i, n;
1652 struct tree_entry *e;
1654 slash1 = strchrnul(p, '/');
1655 n = slash1 - p;
1656 if (!n && !allow_root)
1657 die("Empty path component found in input");
1659 if (!root->tree)
1660 load_tree(root);
1662 if (!n) {
1663 e = root;
1664 goto found_entry;
1667 t = root->tree;
1668 for (i = 0; i < t->entry_count; i++) {
1669 e = t->entries[i];
1670 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1671 if (!*slash1)
1672 goto found_entry;
1673 if (!S_ISDIR(e->versions[1].mode))
1674 return 0;
1675 if (!e->tree)
1676 load_tree(e);
1677 return tree_content_get(e, slash1 + 1, leaf, 0);
1680 return 0;
1682 found_entry:
1683 memcpy(leaf, e, sizeof(*leaf));
1684 if (e->tree && is_null_sha1(e->versions[1].sha1))
1685 leaf->tree = dup_tree_content(e->tree);
1686 else
1687 leaf->tree = NULL;
1688 return 1;
1691 static int update_branch(struct branch *b)
1693 static const char *msg = "fast-import";
1694 struct ref_transaction *transaction;
1695 unsigned char old_sha1[20];
1696 struct strbuf err = STRBUF_INIT;
1698 if (read_ref(b->name, old_sha1))
1699 hashclr(old_sha1);
1700 if (is_null_sha1(b->sha1)) {
1701 if (b->delete)
1702 delete_ref(b->name, old_sha1, 0);
1703 return 0;
1705 if (!force_update && !is_null_sha1(old_sha1)) {
1706 struct commit *old_cmit, *new_cmit;
1708 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1709 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1710 if (!old_cmit || !new_cmit)
1711 return error("Branch %s is missing commits.", b->name);
1713 if (!in_merge_bases(old_cmit, new_cmit)) {
1714 warning("Not updating %s"
1715 " (new tip %s does not contain %s)",
1716 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1717 return -1;
1720 transaction = ref_transaction_begin(&err);
1721 if (!transaction ||
1722 ref_transaction_update(transaction, b->name, b->sha1, old_sha1,
1723 0, 1, msg, &err) ||
1724 ref_transaction_commit(transaction, &err)) {
1725 ref_transaction_free(transaction);
1726 error("%s", err.buf);
1727 strbuf_release(&err);
1728 return -1;
1730 ref_transaction_free(transaction);
1731 strbuf_release(&err);
1732 return 0;
1735 static void dump_branches(void)
1737 unsigned int i;
1738 struct branch *b;
1740 for (i = 0; i < branch_table_sz; i++) {
1741 for (b = branch_table[i]; b; b = b->table_next_branch)
1742 failure |= update_branch(b);
1746 static void dump_tags(void)
1748 static const char *msg = "fast-import";
1749 struct tag *t;
1750 struct strbuf ref_name = STRBUF_INIT;
1751 struct strbuf err = STRBUF_INIT;
1752 struct ref_transaction *transaction;
1754 transaction = ref_transaction_begin(&err);
1755 if (!transaction) {
1756 failure |= error("%s", err.buf);
1757 goto cleanup;
1759 for (t = first_tag; t; t = t->next_tag) {
1760 strbuf_reset(&ref_name);
1761 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1763 if (ref_transaction_update(transaction, ref_name.buf, t->sha1,
1764 NULL, 0, 0, msg, &err)) {
1765 failure |= error("%s", err.buf);
1766 goto cleanup;
1769 if (ref_transaction_commit(transaction, &err))
1770 failure |= error("%s", err.buf);
1772 cleanup:
1773 ref_transaction_free(transaction);
1774 strbuf_release(&ref_name);
1775 strbuf_release(&err);
1778 static void dump_marks_helper(FILE *f,
1779 uintmax_t base,
1780 struct mark_set *m)
1782 uintmax_t k;
1783 if (m->shift) {
1784 for (k = 0; k < 1024; k++) {
1785 if (m->data.sets[k])
1786 dump_marks_helper(f, base + (k << m->shift),
1787 m->data.sets[k]);
1789 } else {
1790 for (k = 0; k < 1024; k++) {
1791 if (m->data.marked[k])
1792 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1793 sha1_to_hex(m->data.marked[k]->idx.sha1));
1798 static void dump_marks(void)
1800 static struct lock_file mark_lock;
1801 FILE *f;
1803 if (!export_marks_file)
1804 return;
1806 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1807 failure |= error("Unable to write marks file %s: %s",
1808 export_marks_file, strerror(errno));
1809 return;
1812 f = fdopen_lock_file(&mark_lock, "w");
1813 if (!f) {
1814 int saved_errno = errno;
1815 rollback_lock_file(&mark_lock);
1816 failure |= error("Unable to write marks file %s: %s",
1817 export_marks_file, strerror(saved_errno));
1818 return;
1821 dump_marks_helper(f, 0, marks);
1822 if (commit_lock_file(&mark_lock)) {
1823 failure |= error("Unable to commit marks file %s: %s",
1824 export_marks_file, strerror(errno));
1825 return;
1829 static void read_marks(void)
1831 char line[512];
1832 FILE *f = fopen(import_marks_file, "r");
1833 if (f)
1835 else if (import_marks_file_ignore_missing && errno == ENOENT)
1836 return; /* Marks file does not exist */
1837 else
1838 die_errno("cannot read '%s'", import_marks_file);
1839 while (fgets(line, sizeof(line), f)) {
1840 uintmax_t mark;
1841 char *end;
1842 unsigned char sha1[20];
1843 struct object_entry *e;
1845 end = strchr(line, '\n');
1846 if (line[0] != ':' || !end)
1847 die("corrupt mark line: %s", line);
1848 *end = 0;
1849 mark = strtoumax(line + 1, &end, 10);
1850 if (!mark || end == line + 1
1851 || *end != ' ' || get_sha1_hex(end + 1, sha1))
1852 die("corrupt mark line: %s", line);
1853 e = find_object(sha1);
1854 if (!e) {
1855 enum object_type type = sha1_object_info(sha1, NULL);
1856 if (type < 0)
1857 die("object not found: %s", sha1_to_hex(sha1));
1858 e = insert_object(sha1);
1859 e->type = type;
1860 e->pack_id = MAX_PACK_ID;
1861 e->idx.offset = 1; /* just not zero! */
1863 insert_mark(mark, e);
1865 fclose(f);
1869 static int read_next_command(void)
1871 static int stdin_eof = 0;
1873 if (stdin_eof) {
1874 unread_command_buf = 0;
1875 return EOF;
1878 for (;;) {
1879 const char *p;
1881 if (unread_command_buf) {
1882 unread_command_buf = 0;
1883 } else {
1884 struct recent_command *rc;
1886 strbuf_detach(&command_buf, NULL);
1887 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1888 if (stdin_eof)
1889 return EOF;
1891 if (!seen_data_command
1892 && !starts_with(command_buf.buf, "feature ")
1893 && !starts_with(command_buf.buf, "option ")) {
1894 parse_argv();
1897 rc = rc_free;
1898 if (rc)
1899 rc_free = rc->next;
1900 else {
1901 rc = cmd_hist.next;
1902 cmd_hist.next = rc->next;
1903 cmd_hist.next->prev = &cmd_hist;
1904 free(rc->buf);
1907 rc->buf = command_buf.buf;
1908 rc->prev = cmd_tail;
1909 rc->next = cmd_hist.prev;
1910 rc->prev->next = rc;
1911 cmd_tail = rc;
1913 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1914 parse_cat_blob(p);
1915 continue;
1917 if (command_buf.buf[0] == '#')
1918 continue;
1919 return 0;
1923 static void skip_optional_lf(void)
1925 int term_char = fgetc(stdin);
1926 if (term_char != '\n' && term_char != EOF)
1927 ungetc(term_char, stdin);
1930 static void parse_mark(void)
1932 const char *v;
1933 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1934 next_mark = strtoumax(v, NULL, 10);
1935 read_next_command();
1937 else
1938 next_mark = 0;
1941 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1943 const char *data;
1944 strbuf_reset(sb);
1946 if (!skip_prefix(command_buf.buf, "data ", &data))
1947 die("Expected 'data n' command, found: %s", command_buf.buf);
1949 if (skip_prefix(data, "<<", &data)) {
1950 char *term = xstrdup(data);
1951 size_t term_len = command_buf.len - (data - command_buf.buf);
1953 strbuf_detach(&command_buf, NULL);
1954 for (;;) {
1955 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1956 die("EOF in data (terminator '%s' not found)", term);
1957 if (term_len == command_buf.len
1958 && !strcmp(term, command_buf.buf))
1959 break;
1960 strbuf_addbuf(sb, &command_buf);
1961 strbuf_addch(sb, '\n');
1963 free(term);
1965 else {
1966 uintmax_t len = strtoumax(data, NULL, 10);
1967 size_t n = 0, length = (size_t)len;
1969 if (limit && limit < len) {
1970 *len_res = len;
1971 return 0;
1973 if (length < len)
1974 die("data is too large to use in this context");
1976 while (n < length) {
1977 size_t s = strbuf_fread(sb, length - n, stdin);
1978 if (!s && feof(stdin))
1979 die("EOF in data (%lu bytes remaining)",
1980 (unsigned long)(length - n));
1981 n += s;
1985 skip_optional_lf();
1986 return 1;
1989 static int validate_raw_date(const char *src, struct strbuf *result)
1991 const char *orig_src = src;
1992 char *endp;
1993 unsigned long num;
1995 errno = 0;
1997 num = strtoul(src, &endp, 10);
1998 /* NEEDSWORK: perhaps check for reasonable values? */
1999 if (errno || endp == src || *endp != ' ')
2000 return -1;
2002 src = endp + 1;
2003 if (*src != '-' && *src != '+')
2004 return -1;
2006 num = strtoul(src + 1, &endp, 10);
2007 if (errno || endp == src + 1 || *endp || 1400 < num)
2008 return -1;
2010 strbuf_addstr(result, orig_src);
2011 return 0;
2014 static char *parse_ident(const char *buf)
2016 const char *ltgt;
2017 size_t name_len;
2018 struct strbuf ident = STRBUF_INIT;
2020 /* ensure there is a space delimiter even if there is no name */
2021 if (*buf == '<')
2022 --buf;
2024 ltgt = buf + strcspn(buf, "<>");
2025 if (*ltgt != '<')
2026 die("Missing < in ident string: %s", buf);
2027 if (ltgt != buf && ltgt[-1] != ' ')
2028 die("Missing space before < in ident string: %s", buf);
2029 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2030 if (*ltgt != '>')
2031 die("Missing > in ident string: %s", buf);
2032 ltgt++;
2033 if (*ltgt != ' ')
2034 die("Missing space after > in ident string: %s", buf);
2035 ltgt++;
2036 name_len = ltgt - buf;
2037 strbuf_add(&ident, buf, name_len);
2039 switch (whenspec) {
2040 case WHENSPEC_RAW:
2041 if (validate_raw_date(ltgt, &ident) < 0)
2042 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2043 break;
2044 case WHENSPEC_RFC2822:
2045 if (parse_date(ltgt, &ident) < 0)
2046 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2047 break;
2048 case WHENSPEC_NOW:
2049 if (strcmp("now", ltgt))
2050 die("Date in ident must be 'now': %s", buf);
2051 datestamp(&ident);
2052 break;
2055 return strbuf_detach(&ident, NULL);
2058 static void parse_and_store_blob(
2059 struct last_object *last,
2060 unsigned char *sha1out,
2061 uintmax_t mark)
2063 static struct strbuf buf = STRBUF_INIT;
2064 uintmax_t len;
2066 if (parse_data(&buf, big_file_threshold, &len))
2067 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2068 else {
2069 if (last) {
2070 strbuf_release(&last->data);
2071 last->offset = 0;
2072 last->depth = 0;
2074 stream_blob(len, sha1out, mark);
2075 skip_optional_lf();
2079 static void parse_new_blob(void)
2081 read_next_command();
2082 parse_mark();
2083 parse_and_store_blob(&last_blob, NULL, next_mark);
2086 static void unload_one_branch(void)
2088 while (cur_active_branches
2089 && cur_active_branches >= max_active_branches) {
2090 uintmax_t min_commit = ULONG_MAX;
2091 struct branch *e, *l = NULL, *p = NULL;
2093 for (e = active_branches; e; e = e->active_next_branch) {
2094 if (e->last_commit < min_commit) {
2095 p = l;
2096 min_commit = e->last_commit;
2098 l = e;
2101 if (p) {
2102 e = p->active_next_branch;
2103 p->active_next_branch = e->active_next_branch;
2104 } else {
2105 e = active_branches;
2106 active_branches = e->active_next_branch;
2108 e->active = 0;
2109 e->active_next_branch = NULL;
2110 if (e->branch_tree.tree) {
2111 release_tree_content_recursive(e->branch_tree.tree);
2112 e->branch_tree.tree = NULL;
2114 cur_active_branches--;
2118 static void load_branch(struct branch *b)
2120 load_tree(&b->branch_tree);
2121 if (!b->active) {
2122 b->active = 1;
2123 b->active_next_branch = active_branches;
2124 active_branches = b;
2125 cur_active_branches++;
2126 branch_load_count++;
2130 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2132 unsigned char fanout = 0;
2133 while ((num_notes >>= 8))
2134 fanout++;
2135 return fanout;
2138 static void construct_path_with_fanout(const char *hex_sha1,
2139 unsigned char fanout, char *path)
2141 unsigned int i = 0, j = 0;
2142 if (fanout >= 20)
2143 die("Too large fanout (%u)", fanout);
2144 while (fanout) {
2145 path[i++] = hex_sha1[j++];
2146 path[i++] = hex_sha1[j++];
2147 path[i++] = '/';
2148 fanout--;
2150 memcpy(path + i, hex_sha1 + j, 40 - j);
2151 path[i + 40 - j] = '\0';
2154 static uintmax_t do_change_note_fanout(
2155 struct tree_entry *orig_root, struct tree_entry *root,
2156 char *hex_sha1, unsigned int hex_sha1_len,
2157 char *fullpath, unsigned int fullpath_len,
2158 unsigned char fanout)
2160 struct tree_content *t = root->tree;
2161 struct tree_entry *e, leaf;
2162 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2163 uintmax_t num_notes = 0;
2164 unsigned char sha1[20];
2165 char realpath[60];
2167 for (i = 0; t && i < t->entry_count; i++) {
2168 e = t->entries[i];
2169 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2170 tmp_fullpath_len = fullpath_len;
2173 * We're interested in EITHER existing note entries (entries
2174 * with exactly 40 hex chars in path, not including directory
2175 * separators), OR directory entries that may contain note
2176 * entries (with < 40 hex chars in path).
2177 * Also, each path component in a note entry must be a multiple
2178 * of 2 chars.
2180 if (!e->versions[1].mode ||
2181 tmp_hex_sha1_len > 40 ||
2182 e->name->str_len % 2)
2183 continue;
2185 /* This _may_ be a note entry, or a subdir containing notes */
2186 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2187 e->name->str_len);
2188 if (tmp_fullpath_len)
2189 fullpath[tmp_fullpath_len++] = '/';
2190 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2191 e->name->str_len);
2192 tmp_fullpath_len += e->name->str_len;
2193 fullpath[tmp_fullpath_len] = '\0';
2195 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2196 /* This is a note entry */
2197 if (fanout == 0xff) {
2198 /* Counting mode, no rename */
2199 num_notes++;
2200 continue;
2202 construct_path_with_fanout(hex_sha1, fanout, realpath);
2203 if (!strcmp(fullpath, realpath)) {
2204 /* Note entry is in correct location */
2205 num_notes++;
2206 continue;
2209 /* Rename fullpath to realpath */
2210 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2211 die("Failed to remove path %s", fullpath);
2212 tree_content_set(orig_root, realpath,
2213 leaf.versions[1].sha1,
2214 leaf.versions[1].mode,
2215 leaf.tree);
2216 } else if (S_ISDIR(e->versions[1].mode)) {
2217 /* This is a subdir that may contain note entries */
2218 if (!e->tree)
2219 load_tree(e);
2220 num_notes += do_change_note_fanout(orig_root, e,
2221 hex_sha1, tmp_hex_sha1_len,
2222 fullpath, tmp_fullpath_len, fanout);
2225 /* The above may have reallocated the current tree_content */
2226 t = root->tree;
2228 return num_notes;
2231 static uintmax_t change_note_fanout(struct tree_entry *root,
2232 unsigned char fanout)
2234 char hex_sha1[40], path[60];
2235 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2239 * Given a pointer into a string, parse a mark reference:
2241 * idnum ::= ':' bigint;
2243 * Return the first character after the value in *endptr.
2245 * Complain if the following character is not what is expected,
2246 * either a space or end of the string.
2248 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2250 uintmax_t mark;
2252 assert(*p == ':');
2253 p++;
2254 mark = strtoumax(p, endptr, 10);
2255 if (*endptr == p)
2256 die("No value after ':' in mark: %s", command_buf.buf);
2257 return mark;
2261 * Parse the mark reference, and complain if this is not the end of
2262 * the string.
2264 static uintmax_t parse_mark_ref_eol(const char *p)
2266 char *end;
2267 uintmax_t mark;
2269 mark = parse_mark_ref(p, &end);
2270 if (*end != '\0')
2271 die("Garbage after mark: %s", command_buf.buf);
2272 return mark;
2276 * Parse the mark reference, demanding a trailing space. Return a
2277 * pointer to the space.
2279 static uintmax_t parse_mark_ref_space(const char **p)
2281 uintmax_t mark;
2282 char *end;
2284 mark = parse_mark_ref(*p, &end);
2285 if (*end++ != ' ')
2286 die("Missing space after mark: %s", command_buf.buf);
2287 *p = end;
2288 return mark;
2291 static void file_change_m(const char *p, struct branch *b)
2293 static struct strbuf uq = STRBUF_INIT;
2294 const char *endp;
2295 struct object_entry *oe;
2296 unsigned char sha1[20];
2297 uint16_t mode, inline_data = 0;
2299 p = get_mode(p, &mode);
2300 if (!p)
2301 die("Corrupt mode: %s", command_buf.buf);
2302 switch (mode) {
2303 case 0644:
2304 case 0755:
2305 mode |= S_IFREG;
2306 case S_IFREG | 0644:
2307 case S_IFREG | 0755:
2308 case S_IFLNK:
2309 case S_IFDIR:
2310 case S_IFGITLINK:
2311 /* ok */
2312 break;
2313 default:
2314 die("Corrupt mode: %s", command_buf.buf);
2317 if (*p == ':') {
2318 oe = find_mark(parse_mark_ref_space(&p));
2319 hashcpy(sha1, oe->idx.sha1);
2320 } else if (skip_prefix(p, "inline ", &p)) {
2321 inline_data = 1;
2322 oe = NULL; /* not used with inline_data, but makes gcc happy */
2323 } else {
2324 if (get_sha1_hex(p, sha1))
2325 die("Invalid dataref: %s", command_buf.buf);
2326 oe = find_object(sha1);
2327 p += 40;
2328 if (*p++ != ' ')
2329 die("Missing space after SHA1: %s", command_buf.buf);
2332 strbuf_reset(&uq);
2333 if (!unquote_c_style(&uq, p, &endp)) {
2334 if (*endp)
2335 die("Garbage after path in: %s", command_buf.buf);
2336 p = uq.buf;
2339 /* Git does not track empty, non-toplevel directories. */
2340 if (S_ISDIR(mode) && !hashcmp(sha1, EMPTY_TREE_SHA1_BIN) && *p) {
2341 tree_content_remove(&b->branch_tree, p, NULL, 0);
2342 return;
2345 if (S_ISGITLINK(mode)) {
2346 if (inline_data)
2347 die("Git links cannot be specified 'inline': %s",
2348 command_buf.buf);
2349 else if (oe) {
2350 if (oe->type != OBJ_COMMIT)
2351 die("Not a commit (actually a %s): %s",
2352 typename(oe->type), command_buf.buf);
2355 * Accept the sha1 without checking; it expected to be in
2356 * another repository.
2358 } else if (inline_data) {
2359 if (S_ISDIR(mode))
2360 die("Directories cannot be specified 'inline': %s",
2361 command_buf.buf);
2362 if (p != uq.buf) {
2363 strbuf_addstr(&uq, p);
2364 p = uq.buf;
2366 read_next_command();
2367 parse_and_store_blob(&last_blob, sha1, 0);
2368 } else {
2369 enum object_type expected = S_ISDIR(mode) ?
2370 OBJ_TREE: OBJ_BLOB;
2371 enum object_type type = oe ? oe->type :
2372 sha1_object_info(sha1, NULL);
2373 if (type < 0)
2374 die("%s not found: %s",
2375 S_ISDIR(mode) ? "Tree" : "Blob",
2376 command_buf.buf);
2377 if (type != expected)
2378 die("Not a %s (actually a %s): %s",
2379 typename(expected), typename(type),
2380 command_buf.buf);
2383 if (!*p) {
2384 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2385 return;
2387 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2390 static void file_change_d(const char *p, struct branch *b)
2392 static struct strbuf uq = STRBUF_INIT;
2393 const char *endp;
2395 strbuf_reset(&uq);
2396 if (!unquote_c_style(&uq, p, &endp)) {
2397 if (*endp)
2398 die("Garbage after path in: %s", command_buf.buf);
2399 p = uq.buf;
2401 tree_content_remove(&b->branch_tree, p, NULL, 1);
2404 static void file_change_cr(const char *s, struct branch *b, int rename)
2406 const char *d;
2407 static struct strbuf s_uq = STRBUF_INIT;
2408 static struct strbuf d_uq = STRBUF_INIT;
2409 const char *endp;
2410 struct tree_entry leaf;
2412 strbuf_reset(&s_uq);
2413 if (!unquote_c_style(&s_uq, s, &endp)) {
2414 if (*endp != ' ')
2415 die("Missing space after source: %s", command_buf.buf);
2416 } else {
2417 endp = strchr(s, ' ');
2418 if (!endp)
2419 die("Missing space after source: %s", command_buf.buf);
2420 strbuf_add(&s_uq, s, endp - s);
2422 s = s_uq.buf;
2424 endp++;
2425 if (!*endp)
2426 die("Missing dest: %s", command_buf.buf);
2428 d = endp;
2429 strbuf_reset(&d_uq);
2430 if (!unquote_c_style(&d_uq, d, &endp)) {
2431 if (*endp)
2432 die("Garbage after dest in: %s", command_buf.buf);
2433 d = d_uq.buf;
2436 memset(&leaf, 0, sizeof(leaf));
2437 if (rename)
2438 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2439 else
2440 tree_content_get(&b->branch_tree, s, &leaf, 1);
2441 if (!leaf.versions[1].mode)
2442 die("Path %s not in branch", s);
2443 if (!*d) { /* C "path/to/subdir" "" */
2444 tree_content_replace(&b->branch_tree,
2445 leaf.versions[1].sha1,
2446 leaf.versions[1].mode,
2447 leaf.tree);
2448 return;
2450 tree_content_set(&b->branch_tree, d,
2451 leaf.versions[1].sha1,
2452 leaf.versions[1].mode,
2453 leaf.tree);
2456 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2458 static struct strbuf uq = STRBUF_INIT;
2459 struct object_entry *oe;
2460 struct branch *s;
2461 unsigned char sha1[20], commit_sha1[20];
2462 char path[60];
2463 uint16_t inline_data = 0;
2464 unsigned char new_fanout;
2467 * When loading a branch, we don't traverse its tree to count the real
2468 * number of notes (too expensive to do this for all non-note refs).
2469 * This means that recently loaded notes refs might incorrectly have
2470 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2472 * Fix this by traversing the tree and counting the number of notes
2473 * when b->num_notes == 0. If the notes tree is truly empty, the
2474 * calculation should not take long.
2476 if (b->num_notes == 0 && *old_fanout == 0) {
2477 /* Invoke change_note_fanout() in "counting mode". */
2478 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2479 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2482 /* Now parse the notemodify command. */
2483 /* <dataref> or 'inline' */
2484 if (*p == ':') {
2485 oe = find_mark(parse_mark_ref_space(&p));
2486 hashcpy(sha1, oe->idx.sha1);
2487 } else if (skip_prefix(p, "inline ", &p)) {
2488 inline_data = 1;
2489 oe = NULL; /* not used with inline_data, but makes gcc happy */
2490 } else {
2491 if (get_sha1_hex(p, sha1))
2492 die("Invalid dataref: %s", command_buf.buf);
2493 oe = find_object(sha1);
2494 p += 40;
2495 if (*p++ != ' ')
2496 die("Missing space after SHA1: %s", command_buf.buf);
2499 /* <commit-ish> */
2500 s = lookup_branch(p);
2501 if (s) {
2502 if (is_null_sha1(s->sha1))
2503 die("Can't add a note on empty branch.");
2504 hashcpy(commit_sha1, s->sha1);
2505 } else if (*p == ':') {
2506 uintmax_t commit_mark = parse_mark_ref_eol(p);
2507 struct object_entry *commit_oe = find_mark(commit_mark);
2508 if (commit_oe->type != OBJ_COMMIT)
2509 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2510 hashcpy(commit_sha1, commit_oe->idx.sha1);
2511 } else if (!get_sha1(p, commit_sha1)) {
2512 unsigned long size;
2513 char *buf = read_object_with_reference(commit_sha1,
2514 commit_type, &size, commit_sha1);
2515 if (!buf || size < 46)
2516 die("Not a valid commit: %s", p);
2517 free(buf);
2518 } else
2519 die("Invalid ref name or SHA1 expression: %s", p);
2521 if (inline_data) {
2522 if (p != uq.buf) {
2523 strbuf_addstr(&uq, p);
2524 p = uq.buf;
2526 read_next_command();
2527 parse_and_store_blob(&last_blob, sha1, 0);
2528 } else if (oe) {
2529 if (oe->type != OBJ_BLOB)
2530 die("Not a blob (actually a %s): %s",
2531 typename(oe->type), command_buf.buf);
2532 } else if (!is_null_sha1(sha1)) {
2533 enum object_type type = sha1_object_info(sha1, NULL);
2534 if (type < 0)
2535 die("Blob not found: %s", command_buf.buf);
2536 if (type != OBJ_BLOB)
2537 die("Not a blob (actually a %s): %s",
2538 typename(type), command_buf.buf);
2541 construct_path_with_fanout(sha1_to_hex(commit_sha1), *old_fanout, path);
2542 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2543 b->num_notes--;
2545 if (is_null_sha1(sha1))
2546 return; /* nothing to insert */
2548 b->num_notes++;
2549 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2550 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2551 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2554 static void file_change_deleteall(struct branch *b)
2556 release_tree_content_recursive(b->branch_tree.tree);
2557 hashclr(b->branch_tree.versions[0].sha1);
2558 hashclr(b->branch_tree.versions[1].sha1);
2559 load_tree(&b->branch_tree);
2560 b->num_notes = 0;
2563 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2565 if (!buf || size < 46)
2566 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2567 if (memcmp("tree ", buf, 5)
2568 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2569 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2570 hashcpy(b->branch_tree.versions[0].sha1,
2571 b->branch_tree.versions[1].sha1);
2574 static void parse_from_existing(struct branch *b)
2576 if (is_null_sha1(b->sha1)) {
2577 hashclr(b->branch_tree.versions[0].sha1);
2578 hashclr(b->branch_tree.versions[1].sha1);
2579 } else {
2580 unsigned long size;
2581 char *buf;
2583 buf = read_object_with_reference(b->sha1,
2584 commit_type, &size, b->sha1);
2585 parse_from_commit(b, buf, size);
2586 free(buf);
2590 static int parse_from(struct branch *b)
2592 const char *from;
2593 struct branch *s;
2595 if (!skip_prefix(command_buf.buf, "from ", &from))
2596 return 0;
2598 if (b->branch_tree.tree) {
2599 release_tree_content_recursive(b->branch_tree.tree);
2600 b->branch_tree.tree = NULL;
2603 s = lookup_branch(from);
2604 if (b == s)
2605 die("Can't create a branch from itself: %s", b->name);
2606 else if (s) {
2607 unsigned char *t = s->branch_tree.versions[1].sha1;
2608 hashcpy(b->sha1, s->sha1);
2609 hashcpy(b->branch_tree.versions[0].sha1, t);
2610 hashcpy(b->branch_tree.versions[1].sha1, t);
2611 } else if (*from == ':') {
2612 uintmax_t idnum = parse_mark_ref_eol(from);
2613 struct object_entry *oe = find_mark(idnum);
2614 if (oe->type != OBJ_COMMIT)
2615 die("Mark :%" PRIuMAX " not a commit", idnum);
2616 hashcpy(b->sha1, oe->idx.sha1);
2617 if (oe->pack_id != MAX_PACK_ID) {
2618 unsigned long size;
2619 char *buf = gfi_unpack_entry(oe, &size);
2620 parse_from_commit(b, buf, size);
2621 free(buf);
2622 } else
2623 parse_from_existing(b);
2624 } else if (!get_sha1(from, b->sha1)) {
2625 parse_from_existing(b);
2626 if (is_null_sha1(b->sha1))
2627 b->delete = 1;
2629 else
2630 die("Invalid ref name or SHA1 expression: %s", from);
2632 read_next_command();
2633 return 1;
2636 static struct hash_list *parse_merge(unsigned int *count)
2638 struct hash_list *list = NULL, **tail = &list, *n;
2639 const char *from;
2640 struct branch *s;
2642 *count = 0;
2643 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2644 n = xmalloc(sizeof(*n));
2645 s = lookup_branch(from);
2646 if (s)
2647 hashcpy(n->sha1, s->sha1);
2648 else if (*from == ':') {
2649 uintmax_t idnum = parse_mark_ref_eol(from);
2650 struct object_entry *oe = find_mark(idnum);
2651 if (oe->type != OBJ_COMMIT)
2652 die("Mark :%" PRIuMAX " not a commit", idnum);
2653 hashcpy(n->sha1, oe->idx.sha1);
2654 } else if (!get_sha1(from, n->sha1)) {
2655 unsigned long size;
2656 char *buf = read_object_with_reference(n->sha1,
2657 commit_type, &size, n->sha1);
2658 if (!buf || size < 46)
2659 die("Not a valid commit: %s", from);
2660 free(buf);
2661 } else
2662 die("Invalid ref name or SHA1 expression: %s", from);
2664 n->next = NULL;
2665 *tail = n;
2666 tail = &n->next;
2668 (*count)++;
2669 read_next_command();
2671 return list;
2674 static void parse_new_commit(const char *arg)
2676 static struct strbuf msg = STRBUF_INIT;
2677 struct branch *b;
2678 char *author = NULL;
2679 char *committer = NULL;
2680 struct hash_list *merge_list = NULL;
2681 unsigned int merge_count;
2682 unsigned char prev_fanout, new_fanout;
2683 const char *v;
2685 b = lookup_branch(arg);
2686 if (!b)
2687 b = new_branch(arg);
2689 read_next_command();
2690 parse_mark();
2691 if (skip_prefix(command_buf.buf, "author ", &v)) {
2692 author = parse_ident(v);
2693 read_next_command();
2695 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2696 committer = parse_ident(v);
2697 read_next_command();
2699 if (!committer)
2700 die("Expected committer but didn't get one");
2701 parse_data(&msg, 0, NULL);
2702 read_next_command();
2703 parse_from(b);
2704 merge_list = parse_merge(&merge_count);
2706 /* ensure the branch is active/loaded */
2707 if (!b->branch_tree.tree || !max_active_branches) {
2708 unload_one_branch();
2709 load_branch(b);
2712 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2714 /* file_change* */
2715 while (command_buf.len > 0) {
2716 if (skip_prefix(command_buf.buf, "M ", &v))
2717 file_change_m(v, b);
2718 else if (skip_prefix(command_buf.buf, "D ", &v))
2719 file_change_d(v, b);
2720 else if (skip_prefix(command_buf.buf, "R ", &v))
2721 file_change_cr(v, b, 1);
2722 else if (skip_prefix(command_buf.buf, "C ", &v))
2723 file_change_cr(v, b, 0);
2724 else if (skip_prefix(command_buf.buf, "N ", &v))
2725 note_change_n(v, b, &prev_fanout);
2726 else if (!strcmp("deleteall", command_buf.buf))
2727 file_change_deleteall(b);
2728 else if (skip_prefix(command_buf.buf, "ls ", &v))
2729 parse_ls(v, b);
2730 else {
2731 unread_command_buf = 1;
2732 break;
2734 if (read_next_command() == EOF)
2735 break;
2738 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2739 if (new_fanout != prev_fanout)
2740 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2742 /* build the tree and the commit */
2743 store_tree(&b->branch_tree);
2744 hashcpy(b->branch_tree.versions[0].sha1,
2745 b->branch_tree.versions[1].sha1);
2747 strbuf_reset(&new_data);
2748 strbuf_addf(&new_data, "tree %s\n",
2749 sha1_to_hex(b->branch_tree.versions[1].sha1));
2750 if (!is_null_sha1(b->sha1))
2751 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2752 while (merge_list) {
2753 struct hash_list *next = merge_list->next;
2754 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2755 free(merge_list);
2756 merge_list = next;
2758 strbuf_addf(&new_data,
2759 "author %s\n"
2760 "committer %s\n"
2761 "\n",
2762 author ? author : committer, committer);
2763 strbuf_addbuf(&new_data, &msg);
2764 free(author);
2765 free(committer);
2767 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2768 b->pack_id = pack_id;
2769 b->last_commit = object_count_by_type[OBJ_COMMIT];
2772 static void parse_new_tag(const char *arg)
2774 static struct strbuf msg = STRBUF_INIT;
2775 const char *from;
2776 char *tagger;
2777 struct branch *s;
2778 struct tag *t;
2779 uintmax_t from_mark = 0;
2780 unsigned char sha1[20];
2781 enum object_type type;
2782 const char *v;
2784 t = pool_alloc(sizeof(struct tag));
2785 memset(t, 0, sizeof(struct tag));
2786 t->name = pool_strdup(arg);
2787 if (last_tag)
2788 last_tag->next_tag = t;
2789 else
2790 first_tag = t;
2791 last_tag = t;
2792 read_next_command();
2794 /* from ... */
2795 if (!skip_prefix(command_buf.buf, "from ", &from))
2796 die("Expected from command, got %s", command_buf.buf);
2797 s = lookup_branch(from);
2798 if (s) {
2799 if (is_null_sha1(s->sha1))
2800 die("Can't tag an empty branch.");
2801 hashcpy(sha1, s->sha1);
2802 type = OBJ_COMMIT;
2803 } else if (*from == ':') {
2804 struct object_entry *oe;
2805 from_mark = parse_mark_ref_eol(from);
2806 oe = find_mark(from_mark);
2807 type = oe->type;
2808 hashcpy(sha1, oe->idx.sha1);
2809 } else if (!get_sha1(from, sha1)) {
2810 struct object_entry *oe = find_object(sha1);
2811 if (!oe) {
2812 type = sha1_object_info(sha1, NULL);
2813 if (type < 0)
2814 die("Not a valid object: %s", from);
2815 } else
2816 type = oe->type;
2817 } else
2818 die("Invalid ref name or SHA1 expression: %s", from);
2819 read_next_command();
2821 /* tagger ... */
2822 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2823 tagger = parse_ident(v);
2824 read_next_command();
2825 } else
2826 tagger = NULL;
2828 /* tag payload/message */
2829 parse_data(&msg, 0, NULL);
2831 /* build the tag object */
2832 strbuf_reset(&new_data);
2834 strbuf_addf(&new_data,
2835 "object %s\n"
2836 "type %s\n"
2837 "tag %s\n",
2838 sha1_to_hex(sha1), typename(type), t->name);
2839 if (tagger)
2840 strbuf_addf(&new_data,
2841 "tagger %s\n", tagger);
2842 strbuf_addch(&new_data, '\n');
2843 strbuf_addbuf(&new_data, &msg);
2844 free(tagger);
2846 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2847 t->pack_id = MAX_PACK_ID;
2848 else
2849 t->pack_id = pack_id;
2852 static void parse_reset_branch(const char *arg)
2854 struct branch *b;
2856 b = lookup_branch(arg);
2857 if (b) {
2858 hashclr(b->sha1);
2859 hashclr(b->branch_tree.versions[0].sha1);
2860 hashclr(b->branch_tree.versions[1].sha1);
2861 if (b->branch_tree.tree) {
2862 release_tree_content_recursive(b->branch_tree.tree);
2863 b->branch_tree.tree = NULL;
2866 else
2867 b = new_branch(arg);
2868 read_next_command();
2869 parse_from(b);
2870 if (command_buf.len > 0)
2871 unread_command_buf = 1;
2874 static void cat_blob_write(const char *buf, unsigned long size)
2876 if (write_in_full(cat_blob_fd, buf, size) != size)
2877 die_errno("Write to frontend failed");
2880 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2882 struct strbuf line = STRBUF_INIT;
2883 unsigned long size;
2884 enum object_type type = 0;
2885 char *buf;
2887 if (!oe || oe->pack_id == MAX_PACK_ID) {
2888 buf = read_sha1_file(sha1, &type, &size);
2889 } else {
2890 type = oe->type;
2891 buf = gfi_unpack_entry(oe, &size);
2895 * Output based on batch_one_object() from cat-file.c.
2897 if (type <= 0) {
2898 strbuf_reset(&line);
2899 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2900 cat_blob_write(line.buf, line.len);
2901 strbuf_release(&line);
2902 free(buf);
2903 return;
2905 if (!buf)
2906 die("Can't read object %s", sha1_to_hex(sha1));
2907 if (type != OBJ_BLOB)
2908 die("Object %s is a %s but a blob was expected.",
2909 sha1_to_hex(sha1), typename(type));
2910 strbuf_reset(&line);
2911 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2912 typename(type), size);
2913 cat_blob_write(line.buf, line.len);
2914 strbuf_release(&line);
2915 cat_blob_write(buf, size);
2916 cat_blob_write("\n", 1);
2917 if (oe && oe->pack_id == pack_id) {
2918 last_blob.offset = oe->idx.offset;
2919 strbuf_attach(&last_blob.data, buf, size, size);
2920 last_blob.depth = oe->depth;
2921 } else
2922 free(buf);
2925 static void parse_cat_blob(const char *p)
2927 struct object_entry *oe = oe;
2928 unsigned char sha1[20];
2930 /* cat-blob SP <object> LF */
2931 if (*p == ':') {
2932 oe = find_mark(parse_mark_ref_eol(p));
2933 if (!oe)
2934 die("Unknown mark: %s", command_buf.buf);
2935 hashcpy(sha1, oe->idx.sha1);
2936 } else {
2937 if (get_sha1_hex(p, sha1))
2938 die("Invalid dataref: %s", command_buf.buf);
2939 if (p[40])
2940 die("Garbage after SHA1: %s", command_buf.buf);
2941 oe = find_object(sha1);
2944 cat_blob(oe, sha1);
2947 static struct object_entry *dereference(struct object_entry *oe,
2948 unsigned char sha1[20])
2950 unsigned long size;
2951 char *buf = NULL;
2952 if (!oe) {
2953 enum object_type type = sha1_object_info(sha1, NULL);
2954 if (type < 0)
2955 die("object not found: %s", sha1_to_hex(sha1));
2956 /* cache it! */
2957 oe = insert_object(sha1);
2958 oe->type = type;
2959 oe->pack_id = MAX_PACK_ID;
2960 oe->idx.offset = 1;
2962 switch (oe->type) {
2963 case OBJ_TREE: /* easy case. */
2964 return oe;
2965 case OBJ_COMMIT:
2966 case OBJ_TAG:
2967 break;
2968 default:
2969 die("Not a tree-ish: %s", command_buf.buf);
2972 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2973 buf = gfi_unpack_entry(oe, &size);
2974 } else {
2975 enum object_type unused;
2976 buf = read_sha1_file(sha1, &unused, &size);
2978 if (!buf)
2979 die("Can't load object %s", sha1_to_hex(sha1));
2981 /* Peel one layer. */
2982 switch (oe->type) {
2983 case OBJ_TAG:
2984 if (size < 40 + strlen("object ") ||
2985 get_sha1_hex(buf + strlen("object "), sha1))
2986 die("Invalid SHA1 in tag: %s", command_buf.buf);
2987 break;
2988 case OBJ_COMMIT:
2989 if (size < 40 + strlen("tree ") ||
2990 get_sha1_hex(buf + strlen("tree "), sha1))
2991 die("Invalid SHA1 in commit: %s", command_buf.buf);
2994 free(buf);
2995 return find_object(sha1);
2998 static struct object_entry *parse_treeish_dataref(const char **p)
3000 unsigned char sha1[20];
3001 struct object_entry *e;
3003 if (**p == ':') { /* <mark> */
3004 e = find_mark(parse_mark_ref_space(p));
3005 if (!e)
3006 die("Unknown mark: %s", command_buf.buf);
3007 hashcpy(sha1, e->idx.sha1);
3008 } else { /* <sha1> */
3009 if (get_sha1_hex(*p, sha1))
3010 die("Invalid dataref: %s", command_buf.buf);
3011 e = find_object(sha1);
3012 *p += 40;
3013 if (*(*p)++ != ' ')
3014 die("Missing space after tree-ish: %s", command_buf.buf);
3017 while (!e || e->type != OBJ_TREE)
3018 e = dereference(e, sha1);
3019 return e;
3022 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3024 static struct strbuf line = STRBUF_INIT;
3026 /* See show_tree(). */
3027 const char *type =
3028 S_ISGITLINK(mode) ? commit_type :
3029 S_ISDIR(mode) ? tree_type :
3030 blob_type;
3032 if (!mode) {
3033 /* missing SP path LF */
3034 strbuf_reset(&line);
3035 strbuf_addstr(&line, "missing ");
3036 quote_c_style(path, &line, NULL, 0);
3037 strbuf_addch(&line, '\n');
3038 } else {
3039 /* mode SP type SP object_name TAB path LF */
3040 strbuf_reset(&line);
3041 strbuf_addf(&line, "%06o %s %s\t",
3042 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3043 quote_c_style(path, &line, NULL, 0);
3044 strbuf_addch(&line, '\n');
3046 cat_blob_write(line.buf, line.len);
3049 static void parse_ls(const char *p, struct branch *b)
3051 struct tree_entry *root = NULL;
3052 struct tree_entry leaf = {NULL};
3054 /* ls SP (<tree-ish> SP)? <path> */
3055 if (*p == '"') {
3056 if (!b)
3057 die("Not in a commit: %s", command_buf.buf);
3058 root = &b->branch_tree;
3059 } else {
3060 struct object_entry *e = parse_treeish_dataref(&p);
3061 root = new_tree_entry();
3062 hashcpy(root->versions[1].sha1, e->idx.sha1);
3063 if (!is_null_sha1(root->versions[1].sha1))
3064 root->versions[1].mode = S_IFDIR;
3065 load_tree(root);
3067 if (*p == '"') {
3068 static struct strbuf uq = STRBUF_INIT;
3069 const char *endp;
3070 strbuf_reset(&uq);
3071 if (unquote_c_style(&uq, p, &endp))
3072 die("Invalid path: %s", command_buf.buf);
3073 if (*endp)
3074 die("Garbage after path in: %s", command_buf.buf);
3075 p = uq.buf;
3077 tree_content_get(root, p, &leaf, 1);
3079 * A directory in preparation would have a sha1 of zero
3080 * until it is saved. Save, for simplicity.
3082 if (S_ISDIR(leaf.versions[1].mode))
3083 store_tree(&leaf);
3085 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3086 if (leaf.tree)
3087 release_tree_content_recursive(leaf.tree);
3088 if (!b || root != &b->branch_tree)
3089 release_tree_entry(root);
3092 static void checkpoint(void)
3094 checkpoint_requested = 0;
3095 if (object_count) {
3096 cycle_packfile();
3097 dump_branches();
3098 dump_tags();
3099 dump_marks();
3103 static void parse_checkpoint(void)
3105 checkpoint_requested = 1;
3106 skip_optional_lf();
3109 static void parse_progress(void)
3111 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3112 fputc('\n', stdout);
3113 fflush(stdout);
3114 skip_optional_lf();
3117 static char* make_fast_import_path(const char *path)
3119 struct strbuf abs_path = STRBUF_INIT;
3121 if (!relative_marks_paths || is_absolute_path(path))
3122 return xstrdup(path);
3123 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
3124 return strbuf_detach(&abs_path, NULL);
3127 static void option_import_marks(const char *marks,
3128 int from_stream, int ignore_missing)
3130 if (import_marks_file) {
3131 if (from_stream)
3132 die("Only one import-marks command allowed per stream");
3134 /* read previous mark file */
3135 if(!import_marks_file_from_stream)
3136 read_marks();
3139 import_marks_file = make_fast_import_path(marks);
3140 safe_create_leading_directories_const(import_marks_file);
3141 import_marks_file_from_stream = from_stream;
3142 import_marks_file_ignore_missing = ignore_missing;
3145 static void option_date_format(const char *fmt)
3147 if (!strcmp(fmt, "raw"))
3148 whenspec = WHENSPEC_RAW;
3149 else if (!strcmp(fmt, "rfc2822"))
3150 whenspec = WHENSPEC_RFC2822;
3151 else if (!strcmp(fmt, "now"))
3152 whenspec = WHENSPEC_NOW;
3153 else
3154 die("unknown --date-format argument %s", fmt);
3157 static unsigned long ulong_arg(const char *option, const char *arg)
3159 char *endptr;
3160 unsigned long rv = strtoul(arg, &endptr, 0);
3161 if (strchr(arg, '-') || endptr == arg || *endptr)
3162 die("%s: argument must be a non-negative integer", option);
3163 return rv;
3166 static void option_depth(const char *depth)
3168 max_depth = ulong_arg("--depth", depth);
3169 if (max_depth > MAX_DEPTH)
3170 die("--depth cannot exceed %u", MAX_DEPTH);
3173 static void option_active_branches(const char *branches)
3175 max_active_branches = ulong_arg("--active-branches", branches);
3178 static void option_export_marks(const char *marks)
3180 export_marks_file = make_fast_import_path(marks);
3181 safe_create_leading_directories_const(export_marks_file);
3184 static void option_cat_blob_fd(const char *fd)
3186 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3187 if (n > (unsigned long) INT_MAX)
3188 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3189 cat_blob_fd = (int) n;
3192 static void option_export_pack_edges(const char *edges)
3194 if (pack_edges)
3195 fclose(pack_edges);
3196 pack_edges = fopen(edges, "a");
3197 if (!pack_edges)
3198 die_errno("Cannot open '%s'", edges);
3201 static int parse_one_option(const char *option)
3203 if (skip_prefix(option, "max-pack-size=", &option)) {
3204 unsigned long v;
3205 if (!git_parse_ulong(option, &v))
3206 return 0;
3207 if (v < 8192) {
3208 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3209 v *= 1024 * 1024;
3210 } else if (v < 1024 * 1024) {
3211 warning("minimum max-pack-size is 1 MiB");
3212 v = 1024 * 1024;
3214 max_packsize = v;
3215 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3216 unsigned long v;
3217 if (!git_parse_ulong(option, &v))
3218 return 0;
3219 big_file_threshold = v;
3220 } else if (skip_prefix(option, "depth=", &option)) {
3221 option_depth(option);
3222 } else if (skip_prefix(option, "active-branches=", &option)) {
3223 option_active_branches(option);
3224 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3225 option_export_pack_edges(option);
3226 } else if (starts_with(option, "quiet")) {
3227 show_stats = 0;
3228 } else if (starts_with(option, "stats")) {
3229 show_stats = 1;
3230 } else {
3231 return 0;
3234 return 1;
3237 static int parse_one_feature(const char *feature, int from_stream)
3239 const char *arg;
3241 if (skip_prefix(feature, "date-format=", &arg)) {
3242 option_date_format(arg);
3243 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3244 option_import_marks(arg, from_stream, 0);
3245 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3246 option_import_marks(arg, from_stream, 1);
3247 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3248 option_export_marks(arg);
3249 } else if (!strcmp(feature, "cat-blob")) {
3250 ; /* Don't die - this feature is supported */
3251 } else if (!strcmp(feature, "relative-marks")) {
3252 relative_marks_paths = 1;
3253 } else if (!strcmp(feature, "no-relative-marks")) {
3254 relative_marks_paths = 0;
3255 } else if (!strcmp(feature, "done")) {
3256 require_explicit_termination = 1;
3257 } else if (!strcmp(feature, "force")) {
3258 force_update = 1;
3259 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3260 ; /* do nothing; we have the feature */
3261 } else {
3262 return 0;
3265 return 1;
3268 static void parse_feature(const char *feature)
3270 if (seen_data_command)
3271 die("Got feature command '%s' after data command", feature);
3273 if (parse_one_feature(feature, 1))
3274 return;
3276 die("This version of fast-import does not support feature %s.", feature);
3279 static void parse_option(const char *option)
3281 if (seen_data_command)
3282 die("Got option command '%s' after data command", option);
3284 if (parse_one_option(option))
3285 return;
3287 die("This version of fast-import does not support option: %s", option);
3290 static void git_pack_config(void)
3292 int indexversion_value;
3293 unsigned long packsizelimit_value;
3295 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3296 if (max_depth > MAX_DEPTH)
3297 max_depth = MAX_DEPTH;
3299 if (!git_config_get_int("pack.compression", &pack_compression_level)) {
3300 if (pack_compression_level == -1)
3301 pack_compression_level = Z_DEFAULT_COMPRESSION;
3302 else if (pack_compression_level < 0 ||
3303 pack_compression_level > Z_BEST_COMPRESSION)
3304 git_die_config("pack.compression",
3305 "bad pack compression level %d", pack_compression_level);
3306 pack_compression_seen = 1;
3308 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3309 pack_idx_opts.version = indexversion_value;
3310 if (pack_idx_opts.version > 2)
3311 git_die_config("pack.indexversion",
3312 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3314 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3315 max_packsize = packsizelimit_value;
3317 git_config(git_default_config, NULL);
3320 static const char fast_import_usage[] =
3321 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3323 static void parse_argv(void)
3325 unsigned int i;
3327 for (i = 1; i < global_argc; i++) {
3328 const char *a = global_argv[i];
3330 if (*a != '-' || !strcmp(a, "--"))
3331 break;
3333 if (!skip_prefix(a, "--", &a))
3334 die("unknown option %s", a);
3336 if (parse_one_option(a))
3337 continue;
3339 if (parse_one_feature(a, 0))
3340 continue;
3342 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3343 option_cat_blob_fd(a);
3344 continue;
3347 die("unknown option --%s", a);
3349 if (i != global_argc)
3350 usage(fast_import_usage);
3352 seen_data_command = 1;
3353 if (import_marks_file)
3354 read_marks();
3357 int main(int argc, char **argv)
3359 unsigned int i;
3361 git_extract_argv0_path(argv[0]);
3363 git_setup_gettext();
3365 if (argc == 2 && !strcmp(argv[1], "-h"))
3366 usage(fast_import_usage);
3368 setup_git_directory();
3369 reset_pack_idx_option(&pack_idx_opts);
3370 git_pack_config();
3371 if (!pack_compression_seen && core_compression_seen)
3372 pack_compression_level = core_compression_level;
3374 alloc_objects(object_entry_alloc);
3375 strbuf_init(&command_buf, 0);
3376 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3377 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3378 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3379 marks = pool_calloc(1, sizeof(struct mark_set));
3381 global_argc = argc;
3382 global_argv = argv;
3384 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3385 for (i = 0; i < (cmd_save - 1); i++)
3386 rc_free[i].next = &rc_free[i + 1];
3387 rc_free[cmd_save - 1].next = NULL;
3389 prepare_packed_git();
3390 start_packfile();
3391 set_die_routine(die_nicely);
3392 set_checkpoint_signal();
3393 while (read_next_command() != EOF) {
3394 const char *v;
3395 if (!strcmp("blob", command_buf.buf))
3396 parse_new_blob();
3397 else if (skip_prefix(command_buf.buf, "ls ", &v))
3398 parse_ls(v, NULL);
3399 else if (skip_prefix(command_buf.buf, "commit ", &v))
3400 parse_new_commit(v);
3401 else if (skip_prefix(command_buf.buf, "tag ", &v))
3402 parse_new_tag(v);
3403 else if (skip_prefix(command_buf.buf, "reset ", &v))
3404 parse_reset_branch(v);
3405 else if (!strcmp("checkpoint", command_buf.buf))
3406 parse_checkpoint();
3407 else if (!strcmp("done", command_buf.buf))
3408 break;
3409 else if (starts_with(command_buf.buf, "progress "))
3410 parse_progress();
3411 else if (skip_prefix(command_buf.buf, "feature ", &v))
3412 parse_feature(v);
3413 else if (skip_prefix(command_buf.buf, "option git ", &v))
3414 parse_option(v);
3415 else if (starts_with(command_buf.buf, "option "))
3416 /* ignore non-git options*/;
3417 else
3418 die("Unsupported command: %s", command_buf.buf);
3420 if (checkpoint_requested)
3421 checkpoint();
3424 /* argv hasn't been parsed yet, do so */
3425 if (!seen_data_command)
3426 parse_argv();
3428 if (require_explicit_termination && feof(stdin))
3429 die("stream ends early");
3431 end_packfile();
3433 dump_branches();
3434 dump_tags();
3435 unkeep_all_packs();
3436 dump_marks();
3438 if (pack_edges)
3439 fclose(pack_edges);
3441 if (show_stats) {
3442 uintmax_t total_count = 0, duplicate_count = 0;
3443 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3444 total_count += object_count_by_type[i];
3445 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3446 duplicate_count += duplicate_count_by_type[i];
3448 fprintf(stderr, "%s statistics:\n", argv[0]);
3449 fprintf(stderr, "---------------------------------------------------------------------\n");
3450 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3451 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3452 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB], delta_count_attempts_by_type[OBJ_BLOB]);
3453 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE], delta_count_attempts_by_type[OBJ_TREE]);
3454 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT], delta_count_attempts_by_type[OBJ_COMMIT]);
3455 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG], delta_count_attempts_by_type[OBJ_TAG]);
3456 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3457 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3458 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3459 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3460 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3461 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3462 fprintf(stderr, "---------------------------------------------------------------------\n");
3463 pack_report();
3464 fprintf(stderr, "---------------------------------------------------------------------\n");
3465 fprintf(stderr, "\n");
3468 return failure ? 1 : 0;