unicode_width.h: update the width tables to Unicode 9.0
[alt-git.git] / fast-import.c
blobd0bd285a16d0161b50d0b43e0315f8fa003e0e32
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 if (!pack_data)
951 return;
953 clear_delta_base_cache();
954 if (object_count) {
955 struct packed_git *new_p;
956 unsigned char cur_pack_sha1[20];
957 char *idx_name;
958 int i;
959 struct branch *b;
960 struct tag *t;
962 close_pack_windows(pack_data);
963 sha1close(pack_file, cur_pack_sha1, 0);
964 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
965 pack_data->pack_name, object_count,
966 cur_pack_sha1, pack_size);
967 close(pack_data->pack_fd);
968 idx_name = keep_pack(create_index());
970 /* Register the packfile with core git's machinery. */
971 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
972 if (!new_p)
973 die("core git rejected index %s", idx_name);
974 all_packs[pack_id] = new_p;
975 install_packed_git(new_p);
977 /* Print the boundary */
978 if (pack_edges) {
979 fprintf(pack_edges, "%s:", new_p->pack_name);
980 for (i = 0; i < branch_table_sz; i++) {
981 for (b = branch_table[i]; b; b = b->table_next_branch) {
982 if (b->pack_id == pack_id)
983 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
986 for (t = first_tag; t; t = t->next_tag) {
987 if (t->pack_id == pack_id)
988 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
990 fputc('\n', pack_edges);
991 fflush(pack_edges);
994 pack_id++;
996 else {
997 close(pack_data->pack_fd);
998 unlink_or_warn(pack_data->pack_name);
1000 free(pack_data);
1001 pack_data = NULL;
1003 /* We can't carry a delta across packfiles. */
1004 strbuf_release(&last_blob.data);
1005 last_blob.offset = 0;
1006 last_blob.depth = 0;
1009 static void cycle_packfile(void)
1011 end_packfile();
1012 start_packfile();
1015 static int store_object(
1016 enum object_type type,
1017 struct strbuf *dat,
1018 struct last_object *last,
1019 unsigned char *sha1out,
1020 uintmax_t mark)
1022 void *out, *delta;
1023 struct object_entry *e;
1024 unsigned char hdr[96];
1025 unsigned char sha1[20];
1026 unsigned long hdrlen, deltalen;
1027 git_SHA_CTX c;
1028 git_zstream s;
1030 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1031 (unsigned long)dat->len) + 1;
1032 git_SHA1_Init(&c);
1033 git_SHA1_Update(&c, hdr, hdrlen);
1034 git_SHA1_Update(&c, dat->buf, dat->len);
1035 git_SHA1_Final(sha1, &c);
1036 if (sha1out)
1037 hashcpy(sha1out, sha1);
1039 e = insert_object(sha1);
1040 if (mark)
1041 insert_mark(mark, e);
1042 if (e->idx.offset) {
1043 duplicate_count_by_type[type]++;
1044 return 1;
1045 } else if (find_sha1_pack(sha1, packed_git)) {
1046 e->type = type;
1047 e->pack_id = MAX_PACK_ID;
1048 e->idx.offset = 1; /* just not zero! */
1049 duplicate_count_by_type[type]++;
1050 return 1;
1053 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1054 delta_count_attempts_by_type[type]++;
1055 delta = diff_delta(last->data.buf, last->data.len,
1056 dat->buf, dat->len,
1057 &deltalen, dat->len - 20);
1058 } else
1059 delta = NULL;
1061 memset(&s, 0, sizeof(s));
1062 git_deflate_init(&s, pack_compression_level);
1063 if (delta) {
1064 s.next_in = delta;
1065 s.avail_in = deltalen;
1066 } else {
1067 s.next_in = (void *)dat->buf;
1068 s.avail_in = dat->len;
1070 s.avail_out = git_deflate_bound(&s, s.avail_in);
1071 s.next_out = out = xmalloc(s.avail_out);
1072 while (git_deflate(&s, Z_FINISH) == Z_OK)
1073 ; /* nothing */
1074 git_deflate_end(&s);
1076 /* Determine if we should auto-checkpoint. */
1077 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1078 || (pack_size + 60 + s.total_out) < pack_size) {
1080 /* This new object needs to *not* have the current pack_id. */
1081 e->pack_id = pack_id + 1;
1082 cycle_packfile();
1084 /* We cannot carry a delta into the new pack. */
1085 if (delta) {
1086 free(delta);
1087 delta = NULL;
1089 memset(&s, 0, sizeof(s));
1090 git_deflate_init(&s, pack_compression_level);
1091 s.next_in = (void *)dat->buf;
1092 s.avail_in = dat->len;
1093 s.avail_out = git_deflate_bound(&s, s.avail_in);
1094 s.next_out = out = xrealloc(out, s.avail_out);
1095 while (git_deflate(&s, Z_FINISH) == Z_OK)
1096 ; /* nothing */
1097 git_deflate_end(&s);
1101 e->type = type;
1102 e->pack_id = pack_id;
1103 e->idx.offset = pack_size;
1104 object_count++;
1105 object_count_by_type[type]++;
1107 crc32_begin(pack_file);
1109 if (delta) {
1110 off_t ofs = e->idx.offset - last->offset;
1111 unsigned pos = sizeof(hdr) - 1;
1113 delta_count_by_type[type]++;
1114 e->depth = last->depth + 1;
1116 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1117 sha1write(pack_file, hdr, hdrlen);
1118 pack_size += hdrlen;
1120 hdr[pos] = ofs & 127;
1121 while (ofs >>= 7)
1122 hdr[--pos] = 128 | (--ofs & 127);
1123 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1124 pack_size += sizeof(hdr) - pos;
1125 } else {
1126 e->depth = 0;
1127 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1128 sha1write(pack_file, hdr, hdrlen);
1129 pack_size += hdrlen;
1132 sha1write(pack_file, out, s.total_out);
1133 pack_size += s.total_out;
1135 e->idx.crc32 = crc32_end(pack_file);
1137 free(out);
1138 free(delta);
1139 if (last) {
1140 if (last->no_swap) {
1141 last->data = *dat;
1142 } else {
1143 strbuf_swap(&last->data, dat);
1145 last->offset = e->idx.offset;
1146 last->depth = e->depth;
1148 return 0;
1151 static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1153 if (sha1file_truncate(pack_file, checkpoint))
1154 die_errno("cannot truncate pack to skip duplicate");
1155 pack_size = checkpoint->offset;
1158 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1160 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1161 unsigned char *in_buf = xmalloc(in_sz);
1162 unsigned char *out_buf = xmalloc(out_sz);
1163 struct object_entry *e;
1164 unsigned char sha1[20];
1165 unsigned long hdrlen;
1166 off_t offset;
1167 git_SHA_CTX c;
1168 git_zstream s;
1169 struct sha1file_checkpoint checkpoint;
1170 int status = Z_OK;
1172 /* Determine if we should auto-checkpoint. */
1173 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1174 || (pack_size + 60 + len) < pack_size)
1175 cycle_packfile();
1177 sha1file_checkpoint(pack_file, &checkpoint);
1178 offset = checkpoint.offset;
1180 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1181 if (out_sz <= hdrlen)
1182 die("impossibly large object header");
1184 git_SHA1_Init(&c);
1185 git_SHA1_Update(&c, out_buf, hdrlen);
1187 crc32_begin(pack_file);
1189 memset(&s, 0, sizeof(s));
1190 git_deflate_init(&s, pack_compression_level);
1192 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1193 if (out_sz <= hdrlen)
1194 die("impossibly large object header");
1196 s.next_out = out_buf + hdrlen;
1197 s.avail_out = out_sz - hdrlen;
1199 while (status != Z_STREAM_END) {
1200 if (0 < len && !s.avail_in) {
1201 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1202 size_t n = fread(in_buf, 1, cnt, stdin);
1203 if (!n && feof(stdin))
1204 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1206 git_SHA1_Update(&c, in_buf, n);
1207 s.next_in = in_buf;
1208 s.avail_in = n;
1209 len -= n;
1212 status = git_deflate(&s, len ? 0 : Z_FINISH);
1214 if (!s.avail_out || status == Z_STREAM_END) {
1215 size_t n = s.next_out - out_buf;
1216 sha1write(pack_file, out_buf, n);
1217 pack_size += n;
1218 s.next_out = out_buf;
1219 s.avail_out = out_sz;
1222 switch (status) {
1223 case Z_OK:
1224 case Z_BUF_ERROR:
1225 case Z_STREAM_END:
1226 continue;
1227 default:
1228 die("unexpected deflate failure: %d", status);
1231 git_deflate_end(&s);
1232 git_SHA1_Final(sha1, &c);
1234 if (sha1out)
1235 hashcpy(sha1out, sha1);
1237 e = insert_object(sha1);
1239 if (mark)
1240 insert_mark(mark, e);
1242 if (e->idx.offset) {
1243 duplicate_count_by_type[OBJ_BLOB]++;
1244 truncate_pack(&checkpoint);
1246 } else if (find_sha1_pack(sha1, packed_git)) {
1247 e->type = OBJ_BLOB;
1248 e->pack_id = MAX_PACK_ID;
1249 e->idx.offset = 1; /* just not zero! */
1250 duplicate_count_by_type[OBJ_BLOB]++;
1251 truncate_pack(&checkpoint);
1253 } else {
1254 e->depth = 0;
1255 e->type = OBJ_BLOB;
1256 e->pack_id = pack_id;
1257 e->idx.offset = offset;
1258 e->idx.crc32 = crc32_end(pack_file);
1259 object_count++;
1260 object_count_by_type[OBJ_BLOB]++;
1263 free(in_buf);
1264 free(out_buf);
1267 /* All calls must be guarded by find_object() or find_mark() to
1268 * ensure the 'struct object_entry' passed was written by this
1269 * process instance. We unpack the entry by the offset, avoiding
1270 * the need for the corresponding .idx file. This unpacking rule
1271 * works because we only use OBJ_REF_DELTA within the packfiles
1272 * created by fast-import.
1274 * oe must not be NULL. Such an oe usually comes from giving
1275 * an unknown SHA-1 to find_object() or an undefined mark to
1276 * find_mark(). Callers must test for this condition and use
1277 * the standard read_sha1_file() when it happens.
1279 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1280 * find_mark(), where the mark was reloaded from an existing marks
1281 * file and is referencing an object that this fast-import process
1282 * instance did not write out to a packfile. Callers must test for
1283 * this condition and use read_sha1_file() instead.
1285 static void *gfi_unpack_entry(
1286 struct object_entry *oe,
1287 unsigned long *sizep)
1289 enum object_type type;
1290 struct packed_git *p = all_packs[oe->pack_id];
1291 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1292 /* The object is stored in the packfile we are writing to
1293 * and we have modified it since the last time we scanned
1294 * back to read a previously written object. If an old
1295 * window covered [p->pack_size, p->pack_size + 20) its
1296 * data is stale and is not valid. Closing all windows
1297 * and updating the packfile length ensures we can read
1298 * the newly written data.
1300 close_pack_windows(p);
1301 sha1flush(pack_file);
1303 /* We have to offer 20 bytes additional on the end of
1304 * the packfile as the core unpacker code assumes the
1305 * footer is present at the file end and must promise
1306 * at least 20 bytes within any window it maps. But
1307 * we don't actually create the footer here.
1309 p->pack_size = pack_size + 20;
1311 return unpack_entry(p, oe->idx.offset, &type, sizep);
1314 static const char *get_mode(const char *str, uint16_t *modep)
1316 unsigned char c;
1317 uint16_t mode = 0;
1319 while ((c = *str++) != ' ') {
1320 if (c < '0' || c > '7')
1321 return NULL;
1322 mode = (mode << 3) + (c - '0');
1324 *modep = mode;
1325 return str;
1328 static void load_tree(struct tree_entry *root)
1330 unsigned char *sha1 = root->versions[1].sha1;
1331 struct object_entry *myoe;
1332 struct tree_content *t;
1333 unsigned long size;
1334 char *buf;
1335 const char *c;
1337 root->tree = t = new_tree_content(8);
1338 if (is_null_sha1(sha1))
1339 return;
1341 myoe = find_object(sha1);
1342 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1343 if (myoe->type != OBJ_TREE)
1344 die("Not a tree: %s", sha1_to_hex(sha1));
1345 t->delta_depth = myoe->depth;
1346 buf = gfi_unpack_entry(myoe, &size);
1347 if (!buf)
1348 die("Can't load tree %s", sha1_to_hex(sha1));
1349 } else {
1350 enum object_type type;
1351 buf = read_sha1_file(sha1, &type, &size);
1352 if (!buf || type != OBJ_TREE)
1353 die("Can't load tree %s", sha1_to_hex(sha1));
1356 c = buf;
1357 while (c != (buf + size)) {
1358 struct tree_entry *e = new_tree_entry();
1360 if (t->entry_count == t->entry_capacity)
1361 root->tree = t = grow_tree_content(t, t->entry_count);
1362 t->entries[t->entry_count++] = e;
1364 e->tree = NULL;
1365 c = get_mode(c, &e->versions[1].mode);
1366 if (!c)
1367 die("Corrupt mode in %s", sha1_to_hex(sha1));
1368 e->versions[0].mode = e->versions[1].mode;
1369 e->name = to_atom(c, strlen(c));
1370 c += e->name->str_len + 1;
1371 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1372 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1373 c += 20;
1375 free(buf);
1378 static int tecmp0 (const void *_a, const void *_b)
1380 struct tree_entry *a = *((struct tree_entry**)_a);
1381 struct tree_entry *b = *((struct tree_entry**)_b);
1382 return base_name_compare(
1383 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1384 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1387 static int tecmp1 (const void *_a, const void *_b)
1389 struct tree_entry *a = *((struct tree_entry**)_a);
1390 struct tree_entry *b = *((struct tree_entry**)_b);
1391 return base_name_compare(
1392 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1393 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1396 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1398 size_t maxlen = 0;
1399 unsigned int i;
1401 if (!v)
1402 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1403 else
1404 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1406 for (i = 0; i < t->entry_count; i++) {
1407 if (t->entries[i]->versions[v].mode)
1408 maxlen += t->entries[i]->name->str_len + 34;
1411 strbuf_reset(b);
1412 strbuf_grow(b, maxlen);
1413 for (i = 0; i < t->entry_count; i++) {
1414 struct tree_entry *e = t->entries[i];
1415 if (!e->versions[v].mode)
1416 continue;
1417 strbuf_addf(b, "%o %s%c",
1418 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1419 e->name->str_dat, '\0');
1420 strbuf_add(b, e->versions[v].sha1, 20);
1424 static void store_tree(struct tree_entry *root)
1426 struct tree_content *t;
1427 unsigned int i, j, del;
1428 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1429 struct object_entry *le = NULL;
1431 if (!is_null_sha1(root->versions[1].sha1))
1432 return;
1434 if (!root->tree)
1435 load_tree(root);
1436 t = root->tree;
1438 for (i = 0; i < t->entry_count; i++) {
1439 if (t->entries[i]->tree)
1440 store_tree(t->entries[i]);
1443 if (!(root->versions[0].mode & NO_DELTA))
1444 le = find_object(root->versions[0].sha1);
1445 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1446 mktree(t, 0, &old_tree);
1447 lo.data = old_tree;
1448 lo.offset = le->idx.offset;
1449 lo.depth = t->delta_depth;
1452 mktree(t, 1, &new_tree);
1453 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1455 t->delta_depth = lo.depth;
1456 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1457 struct tree_entry *e = t->entries[i];
1458 if (e->versions[1].mode) {
1459 e->versions[0].mode = e->versions[1].mode;
1460 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1461 t->entries[j++] = e;
1462 } else {
1463 release_tree_entry(e);
1464 del++;
1467 t->entry_count -= del;
1470 static void tree_content_replace(
1471 struct tree_entry *root,
1472 const unsigned char *sha1,
1473 const uint16_t mode,
1474 struct tree_content *newtree)
1476 if (!S_ISDIR(mode))
1477 die("Root cannot be a non-directory");
1478 hashclr(root->versions[0].sha1);
1479 hashcpy(root->versions[1].sha1, sha1);
1480 if (root->tree)
1481 release_tree_content_recursive(root->tree);
1482 root->tree = newtree;
1485 static int tree_content_set(
1486 struct tree_entry *root,
1487 const char *p,
1488 const unsigned char *sha1,
1489 const uint16_t mode,
1490 struct tree_content *subtree)
1492 struct tree_content *t;
1493 const char *slash1;
1494 unsigned int i, n;
1495 struct tree_entry *e;
1497 slash1 = strchrnul(p, '/');
1498 n = slash1 - p;
1499 if (!n)
1500 die("Empty path component found in input");
1501 if (!*slash1 && !S_ISDIR(mode) && subtree)
1502 die("Non-directories cannot have subtrees");
1504 if (!root->tree)
1505 load_tree(root);
1506 t = root->tree;
1507 for (i = 0; i < t->entry_count; i++) {
1508 e = t->entries[i];
1509 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1510 if (!*slash1) {
1511 if (!S_ISDIR(mode)
1512 && e->versions[1].mode == mode
1513 && !hashcmp(e->versions[1].sha1, sha1))
1514 return 0;
1515 e->versions[1].mode = mode;
1516 hashcpy(e->versions[1].sha1, sha1);
1517 if (e->tree)
1518 release_tree_content_recursive(e->tree);
1519 e->tree = subtree;
1522 * We need to leave e->versions[0].sha1 alone
1523 * to avoid modifying the preimage tree used
1524 * when writing out the parent directory.
1525 * But after replacing the subdir with a
1526 * completely different one, it's not a good
1527 * delta base any more, and besides, we've
1528 * thrown away the tree entries needed to
1529 * make a delta against it.
1531 * So let's just explicitly disable deltas
1532 * for the subtree.
1534 if (S_ISDIR(e->versions[0].mode))
1535 e->versions[0].mode |= NO_DELTA;
1537 hashclr(root->versions[1].sha1);
1538 return 1;
1540 if (!S_ISDIR(e->versions[1].mode)) {
1541 e->tree = new_tree_content(8);
1542 e->versions[1].mode = S_IFDIR;
1544 if (!e->tree)
1545 load_tree(e);
1546 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1547 hashclr(root->versions[1].sha1);
1548 return 1;
1550 return 0;
1554 if (t->entry_count == t->entry_capacity)
1555 root->tree = t = grow_tree_content(t, t->entry_count);
1556 e = new_tree_entry();
1557 e->name = to_atom(p, n);
1558 e->versions[0].mode = 0;
1559 hashclr(e->versions[0].sha1);
1560 t->entries[t->entry_count++] = e;
1561 if (*slash1) {
1562 e->tree = new_tree_content(8);
1563 e->versions[1].mode = S_IFDIR;
1564 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1565 } else {
1566 e->tree = subtree;
1567 e->versions[1].mode = mode;
1568 hashcpy(e->versions[1].sha1, sha1);
1570 hashclr(root->versions[1].sha1);
1571 return 1;
1574 static int tree_content_remove(
1575 struct tree_entry *root,
1576 const char *p,
1577 struct tree_entry *backup_leaf,
1578 int allow_root)
1580 struct tree_content *t;
1581 const char *slash1;
1582 unsigned int i, n;
1583 struct tree_entry *e;
1585 slash1 = strchrnul(p, '/');
1586 n = slash1 - p;
1588 if (!root->tree)
1589 load_tree(root);
1591 if (!*p && allow_root) {
1592 e = root;
1593 goto del_entry;
1596 t = root->tree;
1597 for (i = 0; i < t->entry_count; i++) {
1598 e = t->entries[i];
1599 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1600 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1602 * If p names a file in some subdirectory, and a
1603 * file or symlink matching the name of the
1604 * parent directory of p exists, then p cannot
1605 * exist and need not be deleted.
1607 return 1;
1608 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1609 goto del_entry;
1610 if (!e->tree)
1611 load_tree(e);
1612 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1613 for (n = 0; n < e->tree->entry_count; n++) {
1614 if (e->tree->entries[n]->versions[1].mode) {
1615 hashclr(root->versions[1].sha1);
1616 return 1;
1619 backup_leaf = NULL;
1620 goto del_entry;
1622 return 0;
1625 return 0;
1627 del_entry:
1628 if (backup_leaf)
1629 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1630 else if (e->tree)
1631 release_tree_content_recursive(e->tree);
1632 e->tree = NULL;
1633 e->versions[1].mode = 0;
1634 hashclr(e->versions[1].sha1);
1635 hashclr(root->versions[1].sha1);
1636 return 1;
1639 static int tree_content_get(
1640 struct tree_entry *root,
1641 const char *p,
1642 struct tree_entry *leaf,
1643 int allow_root)
1645 struct tree_content *t;
1646 const char *slash1;
1647 unsigned int i, n;
1648 struct tree_entry *e;
1650 slash1 = strchrnul(p, '/');
1651 n = slash1 - p;
1652 if (!n && !allow_root)
1653 die("Empty path component found in input");
1655 if (!root->tree)
1656 load_tree(root);
1658 if (!n) {
1659 e = root;
1660 goto found_entry;
1663 t = root->tree;
1664 for (i = 0; i < t->entry_count; i++) {
1665 e = t->entries[i];
1666 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1667 if (!*slash1)
1668 goto found_entry;
1669 if (!S_ISDIR(e->versions[1].mode))
1670 return 0;
1671 if (!e->tree)
1672 load_tree(e);
1673 return tree_content_get(e, slash1 + 1, leaf, 0);
1676 return 0;
1678 found_entry:
1679 memcpy(leaf, e, sizeof(*leaf));
1680 if (e->tree && is_null_sha1(e->versions[1].sha1))
1681 leaf->tree = dup_tree_content(e->tree);
1682 else
1683 leaf->tree = NULL;
1684 return 1;
1687 static int update_branch(struct branch *b)
1689 static const char *msg = "fast-import";
1690 struct ref_transaction *transaction;
1691 unsigned char old_sha1[20];
1692 struct strbuf err = STRBUF_INIT;
1694 if (read_ref(b->name, old_sha1))
1695 hashclr(old_sha1);
1696 if (is_null_sha1(b->sha1)) {
1697 if (b->delete)
1698 delete_ref(b->name, old_sha1, 0);
1699 return 0;
1701 if (!force_update && !is_null_sha1(old_sha1)) {
1702 struct commit *old_cmit, *new_cmit;
1704 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1705 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1706 if (!old_cmit || !new_cmit)
1707 return error("Branch %s is missing commits.", b->name);
1709 if (!in_merge_bases(old_cmit, new_cmit)) {
1710 warning("Not updating %s"
1711 " (new tip %s does not contain %s)",
1712 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1713 return -1;
1716 transaction = ref_transaction_begin(&err);
1717 if (!transaction ||
1718 ref_transaction_update(transaction, b->name, b->sha1, old_sha1,
1719 0, 1, msg, &err) ||
1720 ref_transaction_commit(transaction, &err)) {
1721 ref_transaction_free(transaction);
1722 error("%s", err.buf);
1723 strbuf_release(&err);
1724 return -1;
1726 ref_transaction_free(transaction);
1727 strbuf_release(&err);
1728 return 0;
1731 static void dump_branches(void)
1733 unsigned int i;
1734 struct branch *b;
1736 for (i = 0; i < branch_table_sz; i++) {
1737 for (b = branch_table[i]; b; b = b->table_next_branch)
1738 failure |= update_branch(b);
1742 static void dump_tags(void)
1744 static const char *msg = "fast-import";
1745 struct tag *t;
1746 struct strbuf ref_name = STRBUF_INIT;
1747 struct strbuf err = STRBUF_INIT;
1748 struct ref_transaction *transaction;
1750 transaction = ref_transaction_begin(&err);
1751 if (!transaction) {
1752 failure |= error("%s", err.buf);
1753 goto cleanup;
1755 for (t = first_tag; t; t = t->next_tag) {
1756 strbuf_reset(&ref_name);
1757 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1759 if (ref_transaction_update(transaction, ref_name.buf, t->sha1,
1760 NULL, 0, 0, msg, &err)) {
1761 failure |= error("%s", err.buf);
1762 goto cleanup;
1765 if (ref_transaction_commit(transaction, &err))
1766 failure |= error("%s", err.buf);
1768 cleanup:
1769 ref_transaction_free(transaction);
1770 strbuf_release(&ref_name);
1771 strbuf_release(&err);
1774 static void dump_marks_helper(FILE *f,
1775 uintmax_t base,
1776 struct mark_set *m)
1778 uintmax_t k;
1779 if (m->shift) {
1780 for (k = 0; k < 1024; k++) {
1781 if (m->data.sets[k])
1782 dump_marks_helper(f, base + (k << m->shift),
1783 m->data.sets[k]);
1785 } else {
1786 for (k = 0; k < 1024; k++) {
1787 if (m->data.marked[k])
1788 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1789 sha1_to_hex(m->data.marked[k]->idx.sha1));
1794 static void dump_marks(void)
1796 static struct lock_file mark_lock;
1797 FILE *f;
1799 if (!export_marks_file)
1800 return;
1802 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1803 failure |= error("Unable to write marks file %s: %s",
1804 export_marks_file, strerror(errno));
1805 return;
1808 f = fdopen_lock_file(&mark_lock, "w");
1809 if (!f) {
1810 int saved_errno = errno;
1811 rollback_lock_file(&mark_lock);
1812 failure |= error("Unable to write marks file %s: %s",
1813 export_marks_file, strerror(saved_errno));
1814 return;
1817 dump_marks_helper(f, 0, marks);
1818 if (commit_lock_file(&mark_lock)) {
1819 failure |= error("Unable to commit marks file %s: %s",
1820 export_marks_file, strerror(errno));
1821 return;
1825 static void read_marks(void)
1827 char line[512];
1828 FILE *f = fopen(import_marks_file, "r");
1829 if (f)
1831 else if (import_marks_file_ignore_missing && errno == ENOENT)
1832 return; /* Marks file does not exist */
1833 else
1834 die_errno("cannot read '%s'", import_marks_file);
1835 while (fgets(line, sizeof(line), f)) {
1836 uintmax_t mark;
1837 char *end;
1838 unsigned char sha1[20];
1839 struct object_entry *e;
1841 end = strchr(line, '\n');
1842 if (line[0] != ':' || !end)
1843 die("corrupt mark line: %s", line);
1844 *end = 0;
1845 mark = strtoumax(line + 1, &end, 10);
1846 if (!mark || end == line + 1
1847 || *end != ' ' || get_sha1_hex(end + 1, sha1))
1848 die("corrupt mark line: %s", line);
1849 e = find_object(sha1);
1850 if (!e) {
1851 enum object_type type = sha1_object_info(sha1, NULL);
1852 if (type < 0)
1853 die("object not found: %s", sha1_to_hex(sha1));
1854 e = insert_object(sha1);
1855 e->type = type;
1856 e->pack_id = MAX_PACK_ID;
1857 e->idx.offset = 1; /* just not zero! */
1859 insert_mark(mark, e);
1861 fclose(f);
1865 static int read_next_command(void)
1867 static int stdin_eof = 0;
1869 if (stdin_eof) {
1870 unread_command_buf = 0;
1871 return EOF;
1874 for (;;) {
1875 const char *p;
1877 if (unread_command_buf) {
1878 unread_command_buf = 0;
1879 } else {
1880 struct recent_command *rc;
1882 strbuf_detach(&command_buf, NULL);
1883 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1884 if (stdin_eof)
1885 return EOF;
1887 if (!seen_data_command
1888 && !starts_with(command_buf.buf, "feature ")
1889 && !starts_with(command_buf.buf, "option ")) {
1890 parse_argv();
1893 rc = rc_free;
1894 if (rc)
1895 rc_free = rc->next;
1896 else {
1897 rc = cmd_hist.next;
1898 cmd_hist.next = rc->next;
1899 cmd_hist.next->prev = &cmd_hist;
1900 free(rc->buf);
1903 rc->buf = command_buf.buf;
1904 rc->prev = cmd_tail;
1905 rc->next = cmd_hist.prev;
1906 rc->prev->next = rc;
1907 cmd_tail = rc;
1909 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1910 parse_cat_blob(p);
1911 continue;
1913 if (command_buf.buf[0] == '#')
1914 continue;
1915 return 0;
1919 static void skip_optional_lf(void)
1921 int term_char = fgetc(stdin);
1922 if (term_char != '\n' && term_char != EOF)
1923 ungetc(term_char, stdin);
1926 static void parse_mark(void)
1928 const char *v;
1929 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1930 next_mark = strtoumax(v, NULL, 10);
1931 read_next_command();
1933 else
1934 next_mark = 0;
1937 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1939 const char *data;
1940 strbuf_reset(sb);
1942 if (!skip_prefix(command_buf.buf, "data ", &data))
1943 die("Expected 'data n' command, found: %s", command_buf.buf);
1945 if (skip_prefix(data, "<<", &data)) {
1946 char *term = xstrdup(data);
1947 size_t term_len = command_buf.len - (data - command_buf.buf);
1949 strbuf_detach(&command_buf, NULL);
1950 for (;;) {
1951 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1952 die("EOF in data (terminator '%s' not found)", term);
1953 if (term_len == command_buf.len
1954 && !strcmp(term, command_buf.buf))
1955 break;
1956 strbuf_addbuf(sb, &command_buf);
1957 strbuf_addch(sb, '\n');
1959 free(term);
1961 else {
1962 uintmax_t len = strtoumax(data, NULL, 10);
1963 size_t n = 0, length = (size_t)len;
1965 if (limit && limit < len) {
1966 *len_res = len;
1967 return 0;
1969 if (length < len)
1970 die("data is too large to use in this context");
1972 while (n < length) {
1973 size_t s = strbuf_fread(sb, length - n, stdin);
1974 if (!s && feof(stdin))
1975 die("EOF in data (%lu bytes remaining)",
1976 (unsigned long)(length - n));
1977 n += s;
1981 skip_optional_lf();
1982 return 1;
1985 static int validate_raw_date(const char *src, struct strbuf *result)
1987 const char *orig_src = src;
1988 char *endp;
1989 unsigned long num;
1991 errno = 0;
1993 num = strtoul(src, &endp, 10);
1994 /* NEEDSWORK: perhaps check for reasonable values? */
1995 if (errno || endp == src || *endp != ' ')
1996 return -1;
1998 src = endp + 1;
1999 if (*src != '-' && *src != '+')
2000 return -1;
2002 num = strtoul(src + 1, &endp, 10);
2003 if (errno || endp == src + 1 || *endp || 1400 < num)
2004 return -1;
2006 strbuf_addstr(result, orig_src);
2007 return 0;
2010 static char *parse_ident(const char *buf)
2012 const char *ltgt;
2013 size_t name_len;
2014 struct strbuf ident = STRBUF_INIT;
2016 /* ensure there is a space delimiter even if there is no name */
2017 if (*buf == '<')
2018 --buf;
2020 ltgt = buf + strcspn(buf, "<>");
2021 if (*ltgt != '<')
2022 die("Missing < in ident string: %s", buf);
2023 if (ltgt != buf && ltgt[-1] != ' ')
2024 die("Missing space before < in ident string: %s", buf);
2025 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2026 if (*ltgt != '>')
2027 die("Missing > in ident string: %s", buf);
2028 ltgt++;
2029 if (*ltgt != ' ')
2030 die("Missing space after > in ident string: %s", buf);
2031 ltgt++;
2032 name_len = ltgt - buf;
2033 strbuf_add(&ident, buf, name_len);
2035 switch (whenspec) {
2036 case WHENSPEC_RAW:
2037 if (validate_raw_date(ltgt, &ident) < 0)
2038 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2039 break;
2040 case WHENSPEC_RFC2822:
2041 if (parse_date(ltgt, &ident) < 0)
2042 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2043 break;
2044 case WHENSPEC_NOW:
2045 if (strcmp("now", ltgt))
2046 die("Date in ident must be 'now': %s", buf);
2047 datestamp(&ident);
2048 break;
2051 return strbuf_detach(&ident, NULL);
2054 static void parse_and_store_blob(
2055 struct last_object *last,
2056 unsigned char *sha1out,
2057 uintmax_t mark)
2059 static struct strbuf buf = STRBUF_INIT;
2060 uintmax_t len;
2062 if (parse_data(&buf, big_file_threshold, &len))
2063 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2064 else {
2065 if (last) {
2066 strbuf_release(&last->data);
2067 last->offset = 0;
2068 last->depth = 0;
2070 stream_blob(len, sha1out, mark);
2071 skip_optional_lf();
2075 static void parse_new_blob(void)
2077 read_next_command();
2078 parse_mark();
2079 parse_and_store_blob(&last_blob, NULL, next_mark);
2082 static void unload_one_branch(void)
2084 while (cur_active_branches
2085 && cur_active_branches >= max_active_branches) {
2086 uintmax_t min_commit = ULONG_MAX;
2087 struct branch *e, *l = NULL, *p = NULL;
2089 for (e = active_branches; e; e = e->active_next_branch) {
2090 if (e->last_commit < min_commit) {
2091 p = l;
2092 min_commit = e->last_commit;
2094 l = e;
2097 if (p) {
2098 e = p->active_next_branch;
2099 p->active_next_branch = e->active_next_branch;
2100 } else {
2101 e = active_branches;
2102 active_branches = e->active_next_branch;
2104 e->active = 0;
2105 e->active_next_branch = NULL;
2106 if (e->branch_tree.tree) {
2107 release_tree_content_recursive(e->branch_tree.tree);
2108 e->branch_tree.tree = NULL;
2110 cur_active_branches--;
2114 static void load_branch(struct branch *b)
2116 load_tree(&b->branch_tree);
2117 if (!b->active) {
2118 b->active = 1;
2119 b->active_next_branch = active_branches;
2120 active_branches = b;
2121 cur_active_branches++;
2122 branch_load_count++;
2126 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2128 unsigned char fanout = 0;
2129 while ((num_notes >>= 8))
2130 fanout++;
2131 return fanout;
2134 static void construct_path_with_fanout(const char *hex_sha1,
2135 unsigned char fanout, char *path)
2137 unsigned int i = 0, j = 0;
2138 if (fanout >= 20)
2139 die("Too large fanout (%u)", fanout);
2140 while (fanout) {
2141 path[i++] = hex_sha1[j++];
2142 path[i++] = hex_sha1[j++];
2143 path[i++] = '/';
2144 fanout--;
2146 memcpy(path + i, hex_sha1 + j, 40 - j);
2147 path[i + 40 - j] = '\0';
2150 static uintmax_t do_change_note_fanout(
2151 struct tree_entry *orig_root, struct tree_entry *root,
2152 char *hex_sha1, unsigned int hex_sha1_len,
2153 char *fullpath, unsigned int fullpath_len,
2154 unsigned char fanout)
2156 struct tree_content *t = root->tree;
2157 struct tree_entry *e, leaf;
2158 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2159 uintmax_t num_notes = 0;
2160 unsigned char sha1[20];
2161 char realpath[60];
2163 for (i = 0; t && i < t->entry_count; i++) {
2164 e = t->entries[i];
2165 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2166 tmp_fullpath_len = fullpath_len;
2169 * We're interested in EITHER existing note entries (entries
2170 * with exactly 40 hex chars in path, not including directory
2171 * separators), OR directory entries that may contain note
2172 * entries (with < 40 hex chars in path).
2173 * Also, each path component in a note entry must be a multiple
2174 * of 2 chars.
2176 if (!e->versions[1].mode ||
2177 tmp_hex_sha1_len > 40 ||
2178 e->name->str_len % 2)
2179 continue;
2181 /* This _may_ be a note entry, or a subdir containing notes */
2182 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2183 e->name->str_len);
2184 if (tmp_fullpath_len)
2185 fullpath[tmp_fullpath_len++] = '/';
2186 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2187 e->name->str_len);
2188 tmp_fullpath_len += e->name->str_len;
2189 fullpath[tmp_fullpath_len] = '\0';
2191 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2192 /* This is a note entry */
2193 if (fanout == 0xff) {
2194 /* Counting mode, no rename */
2195 num_notes++;
2196 continue;
2198 construct_path_with_fanout(hex_sha1, fanout, realpath);
2199 if (!strcmp(fullpath, realpath)) {
2200 /* Note entry is in correct location */
2201 num_notes++;
2202 continue;
2205 /* Rename fullpath to realpath */
2206 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2207 die("Failed to remove path %s", fullpath);
2208 tree_content_set(orig_root, realpath,
2209 leaf.versions[1].sha1,
2210 leaf.versions[1].mode,
2211 leaf.tree);
2212 } else if (S_ISDIR(e->versions[1].mode)) {
2213 /* This is a subdir that may contain note entries */
2214 if (!e->tree)
2215 load_tree(e);
2216 num_notes += do_change_note_fanout(orig_root, e,
2217 hex_sha1, tmp_hex_sha1_len,
2218 fullpath, tmp_fullpath_len, fanout);
2221 /* The above may have reallocated the current tree_content */
2222 t = root->tree;
2224 return num_notes;
2227 static uintmax_t change_note_fanout(struct tree_entry *root,
2228 unsigned char fanout)
2230 char hex_sha1[40], path[60];
2231 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2235 * Given a pointer into a string, parse a mark reference:
2237 * idnum ::= ':' bigint;
2239 * Return the first character after the value in *endptr.
2241 * Complain if the following character is not what is expected,
2242 * either a space or end of the string.
2244 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2246 uintmax_t mark;
2248 assert(*p == ':');
2249 p++;
2250 mark = strtoumax(p, endptr, 10);
2251 if (*endptr == p)
2252 die("No value after ':' in mark: %s", command_buf.buf);
2253 return mark;
2257 * Parse the mark reference, and complain if this is not the end of
2258 * the string.
2260 static uintmax_t parse_mark_ref_eol(const char *p)
2262 char *end;
2263 uintmax_t mark;
2265 mark = parse_mark_ref(p, &end);
2266 if (*end != '\0')
2267 die("Garbage after mark: %s", command_buf.buf);
2268 return mark;
2272 * Parse the mark reference, demanding a trailing space. Return a
2273 * pointer to the space.
2275 static uintmax_t parse_mark_ref_space(const char **p)
2277 uintmax_t mark;
2278 char *end;
2280 mark = parse_mark_ref(*p, &end);
2281 if (*end++ != ' ')
2282 die("Missing space after mark: %s", command_buf.buf);
2283 *p = end;
2284 return mark;
2287 static void file_change_m(const char *p, struct branch *b)
2289 static struct strbuf uq = STRBUF_INIT;
2290 const char *endp;
2291 struct object_entry *oe;
2292 unsigned char sha1[20];
2293 uint16_t mode, inline_data = 0;
2295 p = get_mode(p, &mode);
2296 if (!p)
2297 die("Corrupt mode: %s", command_buf.buf);
2298 switch (mode) {
2299 case 0644:
2300 case 0755:
2301 mode |= S_IFREG;
2302 case S_IFREG | 0644:
2303 case S_IFREG | 0755:
2304 case S_IFLNK:
2305 case S_IFDIR:
2306 case S_IFGITLINK:
2307 /* ok */
2308 break;
2309 default:
2310 die("Corrupt mode: %s", command_buf.buf);
2313 if (*p == ':') {
2314 oe = find_mark(parse_mark_ref_space(&p));
2315 hashcpy(sha1, oe->idx.sha1);
2316 } else if (skip_prefix(p, "inline ", &p)) {
2317 inline_data = 1;
2318 oe = NULL; /* not used with inline_data, but makes gcc happy */
2319 } else {
2320 if (get_sha1_hex(p, sha1))
2321 die("Invalid dataref: %s", command_buf.buf);
2322 oe = find_object(sha1);
2323 p += 40;
2324 if (*p++ != ' ')
2325 die("Missing space after SHA1: %s", command_buf.buf);
2328 strbuf_reset(&uq);
2329 if (!unquote_c_style(&uq, p, &endp)) {
2330 if (*endp)
2331 die("Garbage after path in: %s", command_buf.buf);
2332 p = uq.buf;
2335 /* Git does not track empty, non-toplevel directories. */
2336 if (S_ISDIR(mode) && !hashcmp(sha1, EMPTY_TREE_SHA1_BIN) && *p) {
2337 tree_content_remove(&b->branch_tree, p, NULL, 0);
2338 return;
2341 if (S_ISGITLINK(mode)) {
2342 if (inline_data)
2343 die("Git links cannot be specified 'inline': %s",
2344 command_buf.buf);
2345 else if (oe) {
2346 if (oe->type != OBJ_COMMIT)
2347 die("Not a commit (actually a %s): %s",
2348 typename(oe->type), command_buf.buf);
2351 * Accept the sha1 without checking; it expected to be in
2352 * another repository.
2354 } else if (inline_data) {
2355 if (S_ISDIR(mode))
2356 die("Directories cannot be specified 'inline': %s",
2357 command_buf.buf);
2358 if (p != uq.buf) {
2359 strbuf_addstr(&uq, p);
2360 p = uq.buf;
2362 read_next_command();
2363 parse_and_store_blob(&last_blob, sha1, 0);
2364 } else {
2365 enum object_type expected = S_ISDIR(mode) ?
2366 OBJ_TREE: OBJ_BLOB;
2367 enum object_type type = oe ? oe->type :
2368 sha1_object_info(sha1, NULL);
2369 if (type < 0)
2370 die("%s not found: %s",
2371 S_ISDIR(mode) ? "Tree" : "Blob",
2372 command_buf.buf);
2373 if (type != expected)
2374 die("Not a %s (actually a %s): %s",
2375 typename(expected), typename(type),
2376 command_buf.buf);
2379 if (!*p) {
2380 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2381 return;
2383 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2386 static void file_change_d(const char *p, struct branch *b)
2388 static struct strbuf uq = STRBUF_INIT;
2389 const char *endp;
2391 strbuf_reset(&uq);
2392 if (!unquote_c_style(&uq, p, &endp)) {
2393 if (*endp)
2394 die("Garbage after path in: %s", command_buf.buf);
2395 p = uq.buf;
2397 tree_content_remove(&b->branch_tree, p, NULL, 1);
2400 static void file_change_cr(const char *s, struct branch *b, int rename)
2402 const char *d;
2403 static struct strbuf s_uq = STRBUF_INIT;
2404 static struct strbuf d_uq = STRBUF_INIT;
2405 const char *endp;
2406 struct tree_entry leaf;
2408 strbuf_reset(&s_uq);
2409 if (!unquote_c_style(&s_uq, s, &endp)) {
2410 if (*endp != ' ')
2411 die("Missing space after source: %s", command_buf.buf);
2412 } else {
2413 endp = strchr(s, ' ');
2414 if (!endp)
2415 die("Missing space after source: %s", command_buf.buf);
2416 strbuf_add(&s_uq, s, endp - s);
2418 s = s_uq.buf;
2420 endp++;
2421 if (!*endp)
2422 die("Missing dest: %s", command_buf.buf);
2424 d = endp;
2425 strbuf_reset(&d_uq);
2426 if (!unquote_c_style(&d_uq, d, &endp)) {
2427 if (*endp)
2428 die("Garbage after dest in: %s", command_buf.buf);
2429 d = d_uq.buf;
2432 memset(&leaf, 0, sizeof(leaf));
2433 if (rename)
2434 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2435 else
2436 tree_content_get(&b->branch_tree, s, &leaf, 1);
2437 if (!leaf.versions[1].mode)
2438 die("Path %s not in branch", s);
2439 if (!*d) { /* C "path/to/subdir" "" */
2440 tree_content_replace(&b->branch_tree,
2441 leaf.versions[1].sha1,
2442 leaf.versions[1].mode,
2443 leaf.tree);
2444 return;
2446 tree_content_set(&b->branch_tree, d,
2447 leaf.versions[1].sha1,
2448 leaf.versions[1].mode,
2449 leaf.tree);
2452 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2454 static struct strbuf uq = STRBUF_INIT;
2455 struct object_entry *oe;
2456 struct branch *s;
2457 unsigned char sha1[20], commit_sha1[20];
2458 char path[60];
2459 uint16_t inline_data = 0;
2460 unsigned char new_fanout;
2463 * When loading a branch, we don't traverse its tree to count the real
2464 * number of notes (too expensive to do this for all non-note refs).
2465 * This means that recently loaded notes refs might incorrectly have
2466 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2468 * Fix this by traversing the tree and counting the number of notes
2469 * when b->num_notes == 0. If the notes tree is truly empty, the
2470 * calculation should not take long.
2472 if (b->num_notes == 0 && *old_fanout == 0) {
2473 /* Invoke change_note_fanout() in "counting mode". */
2474 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2475 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2478 /* Now parse the notemodify command. */
2479 /* <dataref> or 'inline' */
2480 if (*p == ':') {
2481 oe = find_mark(parse_mark_ref_space(&p));
2482 hashcpy(sha1, oe->idx.sha1);
2483 } else if (skip_prefix(p, "inline ", &p)) {
2484 inline_data = 1;
2485 oe = NULL; /* not used with inline_data, but makes gcc happy */
2486 } else {
2487 if (get_sha1_hex(p, sha1))
2488 die("Invalid dataref: %s", command_buf.buf);
2489 oe = find_object(sha1);
2490 p += 40;
2491 if (*p++ != ' ')
2492 die("Missing space after SHA1: %s", command_buf.buf);
2495 /* <commit-ish> */
2496 s = lookup_branch(p);
2497 if (s) {
2498 if (is_null_sha1(s->sha1))
2499 die("Can't add a note on empty branch.");
2500 hashcpy(commit_sha1, s->sha1);
2501 } else if (*p == ':') {
2502 uintmax_t commit_mark = parse_mark_ref_eol(p);
2503 struct object_entry *commit_oe = find_mark(commit_mark);
2504 if (commit_oe->type != OBJ_COMMIT)
2505 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2506 hashcpy(commit_sha1, commit_oe->idx.sha1);
2507 } else if (!get_sha1(p, commit_sha1)) {
2508 unsigned long size;
2509 char *buf = read_object_with_reference(commit_sha1,
2510 commit_type, &size, commit_sha1);
2511 if (!buf || size < 46)
2512 die("Not a valid commit: %s", p);
2513 free(buf);
2514 } else
2515 die("Invalid ref name or SHA1 expression: %s", p);
2517 if (inline_data) {
2518 if (p != uq.buf) {
2519 strbuf_addstr(&uq, p);
2520 p = uq.buf;
2522 read_next_command();
2523 parse_and_store_blob(&last_blob, sha1, 0);
2524 } else if (oe) {
2525 if (oe->type != OBJ_BLOB)
2526 die("Not a blob (actually a %s): %s",
2527 typename(oe->type), command_buf.buf);
2528 } else if (!is_null_sha1(sha1)) {
2529 enum object_type type = sha1_object_info(sha1, NULL);
2530 if (type < 0)
2531 die("Blob not found: %s", command_buf.buf);
2532 if (type != OBJ_BLOB)
2533 die("Not a blob (actually a %s): %s",
2534 typename(type), command_buf.buf);
2537 construct_path_with_fanout(sha1_to_hex(commit_sha1), *old_fanout, path);
2538 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2539 b->num_notes--;
2541 if (is_null_sha1(sha1))
2542 return; /* nothing to insert */
2544 b->num_notes++;
2545 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2546 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2547 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2550 static void file_change_deleteall(struct branch *b)
2552 release_tree_content_recursive(b->branch_tree.tree);
2553 hashclr(b->branch_tree.versions[0].sha1);
2554 hashclr(b->branch_tree.versions[1].sha1);
2555 load_tree(&b->branch_tree);
2556 b->num_notes = 0;
2559 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2561 if (!buf || size < 46)
2562 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2563 if (memcmp("tree ", buf, 5)
2564 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2565 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2566 hashcpy(b->branch_tree.versions[0].sha1,
2567 b->branch_tree.versions[1].sha1);
2570 static void parse_from_existing(struct branch *b)
2572 if (is_null_sha1(b->sha1)) {
2573 hashclr(b->branch_tree.versions[0].sha1);
2574 hashclr(b->branch_tree.versions[1].sha1);
2575 } else {
2576 unsigned long size;
2577 char *buf;
2579 buf = read_object_with_reference(b->sha1,
2580 commit_type, &size, b->sha1);
2581 parse_from_commit(b, buf, size);
2582 free(buf);
2586 static int parse_from(struct branch *b)
2588 const char *from;
2589 struct branch *s;
2591 if (!skip_prefix(command_buf.buf, "from ", &from))
2592 return 0;
2594 if (b->branch_tree.tree) {
2595 release_tree_content_recursive(b->branch_tree.tree);
2596 b->branch_tree.tree = NULL;
2599 s = lookup_branch(from);
2600 if (b == s)
2601 die("Can't create a branch from itself: %s", b->name);
2602 else if (s) {
2603 unsigned char *t = s->branch_tree.versions[1].sha1;
2604 hashcpy(b->sha1, s->sha1);
2605 hashcpy(b->branch_tree.versions[0].sha1, t);
2606 hashcpy(b->branch_tree.versions[1].sha1, t);
2607 } else if (*from == ':') {
2608 uintmax_t idnum = parse_mark_ref_eol(from);
2609 struct object_entry *oe = find_mark(idnum);
2610 if (oe->type != OBJ_COMMIT)
2611 die("Mark :%" PRIuMAX " not a commit", idnum);
2612 hashcpy(b->sha1, oe->idx.sha1);
2613 if (oe->pack_id != MAX_PACK_ID) {
2614 unsigned long size;
2615 char *buf = gfi_unpack_entry(oe, &size);
2616 parse_from_commit(b, buf, size);
2617 free(buf);
2618 } else
2619 parse_from_existing(b);
2620 } else if (!get_sha1(from, b->sha1)) {
2621 parse_from_existing(b);
2622 if (is_null_sha1(b->sha1))
2623 b->delete = 1;
2625 else
2626 die("Invalid ref name or SHA1 expression: %s", from);
2628 read_next_command();
2629 return 1;
2632 static struct hash_list *parse_merge(unsigned int *count)
2634 struct hash_list *list = NULL, **tail = &list, *n;
2635 const char *from;
2636 struct branch *s;
2638 *count = 0;
2639 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2640 n = xmalloc(sizeof(*n));
2641 s = lookup_branch(from);
2642 if (s)
2643 hashcpy(n->sha1, s->sha1);
2644 else if (*from == ':') {
2645 uintmax_t idnum = parse_mark_ref_eol(from);
2646 struct object_entry *oe = find_mark(idnum);
2647 if (oe->type != OBJ_COMMIT)
2648 die("Mark :%" PRIuMAX " not a commit", idnum);
2649 hashcpy(n->sha1, oe->idx.sha1);
2650 } else if (!get_sha1(from, n->sha1)) {
2651 unsigned long size;
2652 char *buf = read_object_with_reference(n->sha1,
2653 commit_type, &size, n->sha1);
2654 if (!buf || size < 46)
2655 die("Not a valid commit: %s", from);
2656 free(buf);
2657 } else
2658 die("Invalid ref name or SHA1 expression: %s", from);
2660 n->next = NULL;
2661 *tail = n;
2662 tail = &n->next;
2664 (*count)++;
2665 read_next_command();
2667 return list;
2670 static void parse_new_commit(const char *arg)
2672 static struct strbuf msg = STRBUF_INIT;
2673 struct branch *b;
2674 char *author = NULL;
2675 char *committer = NULL;
2676 struct hash_list *merge_list = NULL;
2677 unsigned int merge_count;
2678 unsigned char prev_fanout, new_fanout;
2679 const char *v;
2681 b = lookup_branch(arg);
2682 if (!b)
2683 b = new_branch(arg);
2685 read_next_command();
2686 parse_mark();
2687 if (skip_prefix(command_buf.buf, "author ", &v)) {
2688 author = parse_ident(v);
2689 read_next_command();
2691 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2692 committer = parse_ident(v);
2693 read_next_command();
2695 if (!committer)
2696 die("Expected committer but didn't get one");
2697 parse_data(&msg, 0, NULL);
2698 read_next_command();
2699 parse_from(b);
2700 merge_list = parse_merge(&merge_count);
2702 /* ensure the branch is active/loaded */
2703 if (!b->branch_tree.tree || !max_active_branches) {
2704 unload_one_branch();
2705 load_branch(b);
2708 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2710 /* file_change* */
2711 while (command_buf.len > 0) {
2712 if (skip_prefix(command_buf.buf, "M ", &v))
2713 file_change_m(v, b);
2714 else if (skip_prefix(command_buf.buf, "D ", &v))
2715 file_change_d(v, b);
2716 else if (skip_prefix(command_buf.buf, "R ", &v))
2717 file_change_cr(v, b, 1);
2718 else if (skip_prefix(command_buf.buf, "C ", &v))
2719 file_change_cr(v, b, 0);
2720 else if (skip_prefix(command_buf.buf, "N ", &v))
2721 note_change_n(v, b, &prev_fanout);
2722 else if (!strcmp("deleteall", command_buf.buf))
2723 file_change_deleteall(b);
2724 else if (skip_prefix(command_buf.buf, "ls ", &v))
2725 parse_ls(v, b);
2726 else {
2727 unread_command_buf = 1;
2728 break;
2730 if (read_next_command() == EOF)
2731 break;
2734 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2735 if (new_fanout != prev_fanout)
2736 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2738 /* build the tree and the commit */
2739 store_tree(&b->branch_tree);
2740 hashcpy(b->branch_tree.versions[0].sha1,
2741 b->branch_tree.versions[1].sha1);
2743 strbuf_reset(&new_data);
2744 strbuf_addf(&new_data, "tree %s\n",
2745 sha1_to_hex(b->branch_tree.versions[1].sha1));
2746 if (!is_null_sha1(b->sha1))
2747 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2748 while (merge_list) {
2749 struct hash_list *next = merge_list->next;
2750 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2751 free(merge_list);
2752 merge_list = next;
2754 strbuf_addf(&new_data,
2755 "author %s\n"
2756 "committer %s\n"
2757 "\n",
2758 author ? author : committer, committer);
2759 strbuf_addbuf(&new_data, &msg);
2760 free(author);
2761 free(committer);
2763 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2764 b->pack_id = pack_id;
2765 b->last_commit = object_count_by_type[OBJ_COMMIT];
2768 static void parse_new_tag(const char *arg)
2770 static struct strbuf msg = STRBUF_INIT;
2771 const char *from;
2772 char *tagger;
2773 struct branch *s;
2774 struct tag *t;
2775 uintmax_t from_mark = 0;
2776 unsigned char sha1[20];
2777 enum object_type type;
2778 const char *v;
2780 t = pool_alloc(sizeof(struct tag));
2781 memset(t, 0, sizeof(struct tag));
2782 t->name = pool_strdup(arg);
2783 if (last_tag)
2784 last_tag->next_tag = t;
2785 else
2786 first_tag = t;
2787 last_tag = t;
2788 read_next_command();
2790 /* from ... */
2791 if (!skip_prefix(command_buf.buf, "from ", &from))
2792 die("Expected from command, got %s", command_buf.buf);
2793 s = lookup_branch(from);
2794 if (s) {
2795 if (is_null_sha1(s->sha1))
2796 die("Can't tag an empty branch.");
2797 hashcpy(sha1, s->sha1);
2798 type = OBJ_COMMIT;
2799 } else if (*from == ':') {
2800 struct object_entry *oe;
2801 from_mark = parse_mark_ref_eol(from);
2802 oe = find_mark(from_mark);
2803 type = oe->type;
2804 hashcpy(sha1, oe->idx.sha1);
2805 } else if (!get_sha1(from, sha1)) {
2806 struct object_entry *oe = find_object(sha1);
2807 if (!oe) {
2808 type = sha1_object_info(sha1, NULL);
2809 if (type < 0)
2810 die("Not a valid object: %s", from);
2811 } else
2812 type = oe->type;
2813 } else
2814 die("Invalid ref name or SHA1 expression: %s", from);
2815 read_next_command();
2817 /* tagger ... */
2818 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2819 tagger = parse_ident(v);
2820 read_next_command();
2821 } else
2822 tagger = NULL;
2824 /* tag payload/message */
2825 parse_data(&msg, 0, NULL);
2827 /* build the tag object */
2828 strbuf_reset(&new_data);
2830 strbuf_addf(&new_data,
2831 "object %s\n"
2832 "type %s\n"
2833 "tag %s\n",
2834 sha1_to_hex(sha1), typename(type), t->name);
2835 if (tagger)
2836 strbuf_addf(&new_data,
2837 "tagger %s\n", tagger);
2838 strbuf_addch(&new_data, '\n');
2839 strbuf_addbuf(&new_data, &msg);
2840 free(tagger);
2842 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2843 t->pack_id = MAX_PACK_ID;
2844 else
2845 t->pack_id = pack_id;
2848 static void parse_reset_branch(const char *arg)
2850 struct branch *b;
2852 b = lookup_branch(arg);
2853 if (b) {
2854 hashclr(b->sha1);
2855 hashclr(b->branch_tree.versions[0].sha1);
2856 hashclr(b->branch_tree.versions[1].sha1);
2857 if (b->branch_tree.tree) {
2858 release_tree_content_recursive(b->branch_tree.tree);
2859 b->branch_tree.tree = NULL;
2862 else
2863 b = new_branch(arg);
2864 read_next_command();
2865 parse_from(b);
2866 if (command_buf.len > 0)
2867 unread_command_buf = 1;
2870 static void cat_blob_write(const char *buf, unsigned long size)
2872 if (write_in_full(cat_blob_fd, buf, size) != size)
2873 die_errno("Write to frontend failed");
2876 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2878 struct strbuf line = STRBUF_INIT;
2879 unsigned long size;
2880 enum object_type type = 0;
2881 char *buf;
2883 if (!oe || oe->pack_id == MAX_PACK_ID) {
2884 buf = read_sha1_file(sha1, &type, &size);
2885 } else {
2886 type = oe->type;
2887 buf = gfi_unpack_entry(oe, &size);
2891 * Output based on batch_one_object() from cat-file.c.
2893 if (type <= 0) {
2894 strbuf_reset(&line);
2895 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2896 cat_blob_write(line.buf, line.len);
2897 strbuf_release(&line);
2898 free(buf);
2899 return;
2901 if (!buf)
2902 die("Can't read object %s", sha1_to_hex(sha1));
2903 if (type != OBJ_BLOB)
2904 die("Object %s is a %s but a blob was expected.",
2905 sha1_to_hex(sha1), typename(type));
2906 strbuf_reset(&line);
2907 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2908 typename(type), size);
2909 cat_blob_write(line.buf, line.len);
2910 strbuf_release(&line);
2911 cat_blob_write(buf, size);
2912 cat_blob_write("\n", 1);
2913 if (oe && oe->pack_id == pack_id) {
2914 last_blob.offset = oe->idx.offset;
2915 strbuf_attach(&last_blob.data, buf, size, size);
2916 last_blob.depth = oe->depth;
2917 } else
2918 free(buf);
2921 static void parse_cat_blob(const char *p)
2923 struct object_entry *oe = oe;
2924 unsigned char sha1[20];
2926 /* cat-blob SP <object> LF */
2927 if (*p == ':') {
2928 oe = find_mark(parse_mark_ref_eol(p));
2929 if (!oe)
2930 die("Unknown mark: %s", command_buf.buf);
2931 hashcpy(sha1, oe->idx.sha1);
2932 } else {
2933 if (get_sha1_hex(p, sha1))
2934 die("Invalid dataref: %s", command_buf.buf);
2935 if (p[40])
2936 die("Garbage after SHA1: %s", command_buf.buf);
2937 oe = find_object(sha1);
2940 cat_blob(oe, sha1);
2943 static struct object_entry *dereference(struct object_entry *oe,
2944 unsigned char sha1[20])
2946 unsigned long size;
2947 char *buf = NULL;
2948 if (!oe) {
2949 enum object_type type = sha1_object_info(sha1, NULL);
2950 if (type < 0)
2951 die("object not found: %s", sha1_to_hex(sha1));
2952 /* cache it! */
2953 oe = insert_object(sha1);
2954 oe->type = type;
2955 oe->pack_id = MAX_PACK_ID;
2956 oe->idx.offset = 1;
2958 switch (oe->type) {
2959 case OBJ_TREE: /* easy case. */
2960 return oe;
2961 case OBJ_COMMIT:
2962 case OBJ_TAG:
2963 break;
2964 default:
2965 die("Not a tree-ish: %s", command_buf.buf);
2968 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2969 buf = gfi_unpack_entry(oe, &size);
2970 } else {
2971 enum object_type unused;
2972 buf = read_sha1_file(sha1, &unused, &size);
2974 if (!buf)
2975 die("Can't load object %s", sha1_to_hex(sha1));
2977 /* Peel one layer. */
2978 switch (oe->type) {
2979 case OBJ_TAG:
2980 if (size < 40 + strlen("object ") ||
2981 get_sha1_hex(buf + strlen("object "), sha1))
2982 die("Invalid SHA1 in tag: %s", command_buf.buf);
2983 break;
2984 case OBJ_COMMIT:
2985 if (size < 40 + strlen("tree ") ||
2986 get_sha1_hex(buf + strlen("tree "), sha1))
2987 die("Invalid SHA1 in commit: %s", command_buf.buf);
2990 free(buf);
2991 return find_object(sha1);
2994 static struct object_entry *parse_treeish_dataref(const char **p)
2996 unsigned char sha1[20];
2997 struct object_entry *e;
2999 if (**p == ':') { /* <mark> */
3000 e = find_mark(parse_mark_ref_space(p));
3001 if (!e)
3002 die("Unknown mark: %s", command_buf.buf);
3003 hashcpy(sha1, e->idx.sha1);
3004 } else { /* <sha1> */
3005 if (get_sha1_hex(*p, sha1))
3006 die("Invalid dataref: %s", command_buf.buf);
3007 e = find_object(sha1);
3008 *p += 40;
3009 if (*(*p)++ != ' ')
3010 die("Missing space after tree-ish: %s", command_buf.buf);
3013 while (!e || e->type != OBJ_TREE)
3014 e = dereference(e, sha1);
3015 return e;
3018 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3020 static struct strbuf line = STRBUF_INIT;
3022 /* See show_tree(). */
3023 const char *type =
3024 S_ISGITLINK(mode) ? commit_type :
3025 S_ISDIR(mode) ? tree_type :
3026 blob_type;
3028 if (!mode) {
3029 /* missing SP path LF */
3030 strbuf_reset(&line);
3031 strbuf_addstr(&line, "missing ");
3032 quote_c_style(path, &line, NULL, 0);
3033 strbuf_addch(&line, '\n');
3034 } else {
3035 /* mode SP type SP object_name TAB path LF */
3036 strbuf_reset(&line);
3037 strbuf_addf(&line, "%06o %s %s\t",
3038 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3039 quote_c_style(path, &line, NULL, 0);
3040 strbuf_addch(&line, '\n');
3042 cat_blob_write(line.buf, line.len);
3045 static void parse_ls(const char *p, struct branch *b)
3047 struct tree_entry *root = NULL;
3048 struct tree_entry leaf = {NULL};
3050 /* ls SP (<tree-ish> SP)? <path> */
3051 if (*p == '"') {
3052 if (!b)
3053 die("Not in a commit: %s", command_buf.buf);
3054 root = &b->branch_tree;
3055 } else {
3056 struct object_entry *e = parse_treeish_dataref(&p);
3057 root = new_tree_entry();
3058 hashcpy(root->versions[1].sha1, e->idx.sha1);
3059 if (!is_null_sha1(root->versions[1].sha1))
3060 root->versions[1].mode = S_IFDIR;
3061 load_tree(root);
3063 if (*p == '"') {
3064 static struct strbuf uq = STRBUF_INIT;
3065 const char *endp;
3066 strbuf_reset(&uq);
3067 if (unquote_c_style(&uq, p, &endp))
3068 die("Invalid path: %s", command_buf.buf);
3069 if (*endp)
3070 die("Garbage after path in: %s", command_buf.buf);
3071 p = uq.buf;
3073 tree_content_get(root, p, &leaf, 1);
3075 * A directory in preparation would have a sha1 of zero
3076 * until it is saved. Save, for simplicity.
3078 if (S_ISDIR(leaf.versions[1].mode))
3079 store_tree(&leaf);
3081 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3082 if (leaf.tree)
3083 release_tree_content_recursive(leaf.tree);
3084 if (!b || root != &b->branch_tree)
3085 release_tree_entry(root);
3088 static void checkpoint(void)
3090 checkpoint_requested = 0;
3091 if (object_count) {
3092 cycle_packfile();
3093 dump_branches();
3094 dump_tags();
3095 dump_marks();
3099 static void parse_checkpoint(void)
3101 checkpoint_requested = 1;
3102 skip_optional_lf();
3105 static void parse_progress(void)
3107 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3108 fputc('\n', stdout);
3109 fflush(stdout);
3110 skip_optional_lf();
3113 static char* make_fast_import_path(const char *path)
3115 struct strbuf abs_path = STRBUF_INIT;
3117 if (!relative_marks_paths || is_absolute_path(path))
3118 return xstrdup(path);
3119 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
3120 return strbuf_detach(&abs_path, NULL);
3123 static void option_import_marks(const char *marks,
3124 int from_stream, int ignore_missing)
3126 if (import_marks_file) {
3127 if (from_stream)
3128 die("Only one import-marks command allowed per stream");
3130 /* read previous mark file */
3131 if(!import_marks_file_from_stream)
3132 read_marks();
3135 import_marks_file = make_fast_import_path(marks);
3136 safe_create_leading_directories_const(import_marks_file);
3137 import_marks_file_from_stream = from_stream;
3138 import_marks_file_ignore_missing = ignore_missing;
3141 static void option_date_format(const char *fmt)
3143 if (!strcmp(fmt, "raw"))
3144 whenspec = WHENSPEC_RAW;
3145 else if (!strcmp(fmt, "rfc2822"))
3146 whenspec = WHENSPEC_RFC2822;
3147 else if (!strcmp(fmt, "now"))
3148 whenspec = WHENSPEC_NOW;
3149 else
3150 die("unknown --date-format argument %s", fmt);
3153 static unsigned long ulong_arg(const char *option, const char *arg)
3155 char *endptr;
3156 unsigned long rv = strtoul(arg, &endptr, 0);
3157 if (strchr(arg, '-') || endptr == arg || *endptr)
3158 die("%s: argument must be a non-negative integer", option);
3159 return rv;
3162 static void option_depth(const char *depth)
3164 max_depth = ulong_arg("--depth", depth);
3165 if (max_depth > MAX_DEPTH)
3166 die("--depth cannot exceed %u", MAX_DEPTH);
3169 static void option_active_branches(const char *branches)
3171 max_active_branches = ulong_arg("--active-branches", branches);
3174 static void option_export_marks(const char *marks)
3176 export_marks_file = make_fast_import_path(marks);
3177 safe_create_leading_directories_const(export_marks_file);
3180 static void option_cat_blob_fd(const char *fd)
3182 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3183 if (n > (unsigned long) INT_MAX)
3184 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3185 cat_blob_fd = (int) n;
3188 static void option_export_pack_edges(const char *edges)
3190 if (pack_edges)
3191 fclose(pack_edges);
3192 pack_edges = fopen(edges, "a");
3193 if (!pack_edges)
3194 die_errno("Cannot open '%s'", edges);
3197 static int parse_one_option(const char *option)
3199 if (skip_prefix(option, "max-pack-size=", &option)) {
3200 unsigned long v;
3201 if (!git_parse_ulong(option, &v))
3202 return 0;
3203 if (v < 8192) {
3204 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3205 v *= 1024 * 1024;
3206 } else if (v < 1024 * 1024) {
3207 warning("minimum max-pack-size is 1 MiB");
3208 v = 1024 * 1024;
3210 max_packsize = v;
3211 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3212 unsigned long v;
3213 if (!git_parse_ulong(option, &v))
3214 return 0;
3215 big_file_threshold = v;
3216 } else if (skip_prefix(option, "depth=", &option)) {
3217 option_depth(option);
3218 } else if (skip_prefix(option, "active-branches=", &option)) {
3219 option_active_branches(option);
3220 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3221 option_export_pack_edges(option);
3222 } else if (starts_with(option, "quiet")) {
3223 show_stats = 0;
3224 } else if (starts_with(option, "stats")) {
3225 show_stats = 1;
3226 } else {
3227 return 0;
3230 return 1;
3233 static int parse_one_feature(const char *feature, int from_stream)
3235 const char *arg;
3237 if (skip_prefix(feature, "date-format=", &arg)) {
3238 option_date_format(arg);
3239 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3240 option_import_marks(arg, from_stream, 0);
3241 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3242 option_import_marks(arg, from_stream, 1);
3243 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3244 option_export_marks(arg);
3245 } else if (!strcmp(feature, "cat-blob")) {
3246 ; /* Don't die - this feature is supported */
3247 } else if (!strcmp(feature, "relative-marks")) {
3248 relative_marks_paths = 1;
3249 } else if (!strcmp(feature, "no-relative-marks")) {
3250 relative_marks_paths = 0;
3251 } else if (!strcmp(feature, "done")) {
3252 require_explicit_termination = 1;
3253 } else if (!strcmp(feature, "force")) {
3254 force_update = 1;
3255 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3256 ; /* do nothing; we have the feature */
3257 } else {
3258 return 0;
3261 return 1;
3264 static void parse_feature(const char *feature)
3266 if (seen_data_command)
3267 die("Got feature command '%s' after data command", feature);
3269 if (parse_one_feature(feature, 1))
3270 return;
3272 die("This version of fast-import does not support feature %s.", feature);
3275 static void parse_option(const char *option)
3277 if (seen_data_command)
3278 die("Got option command '%s' after data command", option);
3280 if (parse_one_option(option))
3281 return;
3283 die("This version of fast-import does not support option: %s", option);
3286 static void git_pack_config(void)
3288 int indexversion_value;
3289 unsigned long packsizelimit_value;
3291 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3292 if (max_depth > MAX_DEPTH)
3293 max_depth = MAX_DEPTH;
3295 if (!git_config_get_int("pack.compression", &pack_compression_level)) {
3296 if (pack_compression_level == -1)
3297 pack_compression_level = Z_DEFAULT_COMPRESSION;
3298 else if (pack_compression_level < 0 ||
3299 pack_compression_level > Z_BEST_COMPRESSION)
3300 git_die_config("pack.compression",
3301 "bad pack compression level %d", pack_compression_level);
3302 pack_compression_seen = 1;
3304 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3305 pack_idx_opts.version = indexversion_value;
3306 if (pack_idx_opts.version > 2)
3307 git_die_config("pack.indexversion",
3308 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3310 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3311 max_packsize = packsizelimit_value;
3313 git_config(git_default_config, NULL);
3316 static const char fast_import_usage[] =
3317 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3319 static void parse_argv(void)
3321 unsigned int i;
3323 for (i = 1; i < global_argc; i++) {
3324 const char *a = global_argv[i];
3326 if (*a != '-' || !strcmp(a, "--"))
3327 break;
3329 if (!skip_prefix(a, "--", &a))
3330 die("unknown option %s", a);
3332 if (parse_one_option(a))
3333 continue;
3335 if (parse_one_feature(a, 0))
3336 continue;
3338 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3339 option_cat_blob_fd(a);
3340 continue;
3343 die("unknown option --%s", a);
3345 if (i != global_argc)
3346 usage(fast_import_usage);
3348 seen_data_command = 1;
3349 if (import_marks_file)
3350 read_marks();
3353 int main(int argc, char **argv)
3355 unsigned int i;
3357 git_extract_argv0_path(argv[0]);
3359 git_setup_gettext();
3361 if (argc == 2 && !strcmp(argv[1], "-h"))
3362 usage(fast_import_usage);
3364 setup_git_directory();
3365 reset_pack_idx_option(&pack_idx_opts);
3366 git_pack_config();
3367 if (!pack_compression_seen && core_compression_seen)
3368 pack_compression_level = core_compression_level;
3370 alloc_objects(object_entry_alloc);
3371 strbuf_init(&command_buf, 0);
3372 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3373 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3374 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3375 marks = pool_calloc(1, sizeof(struct mark_set));
3377 global_argc = argc;
3378 global_argv = argv;
3380 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3381 for (i = 0; i < (cmd_save - 1); i++)
3382 rc_free[i].next = &rc_free[i + 1];
3383 rc_free[cmd_save - 1].next = NULL;
3385 prepare_packed_git();
3386 start_packfile();
3387 set_die_routine(die_nicely);
3388 set_checkpoint_signal();
3389 while (read_next_command() != EOF) {
3390 const char *v;
3391 if (!strcmp("blob", command_buf.buf))
3392 parse_new_blob();
3393 else if (skip_prefix(command_buf.buf, "ls ", &v))
3394 parse_ls(v, NULL);
3395 else if (skip_prefix(command_buf.buf, "commit ", &v))
3396 parse_new_commit(v);
3397 else if (skip_prefix(command_buf.buf, "tag ", &v))
3398 parse_new_tag(v);
3399 else if (skip_prefix(command_buf.buf, "reset ", &v))
3400 parse_reset_branch(v);
3401 else if (!strcmp("checkpoint", command_buf.buf))
3402 parse_checkpoint();
3403 else if (!strcmp("done", command_buf.buf))
3404 break;
3405 else if (starts_with(command_buf.buf, "progress "))
3406 parse_progress();
3407 else if (skip_prefix(command_buf.buf, "feature ", &v))
3408 parse_feature(v);
3409 else if (skip_prefix(command_buf.buf, "option git ", &v))
3410 parse_option(v);
3411 else if (starts_with(command_buf.buf, "option "))
3412 /* ignore non-git options*/;
3413 else
3414 die("Unsupported command: %s", command_buf.buf);
3416 if (checkpoint_requested)
3417 checkpoint();
3420 /* argv hasn't been parsed yet, do so */
3421 if (!seen_data_command)
3422 parse_argv();
3424 if (require_explicit_termination && feof(stdin))
3425 die("stream ends early");
3427 end_packfile();
3429 dump_branches();
3430 dump_tags();
3431 unkeep_all_packs();
3432 dump_marks();
3434 if (pack_edges)
3435 fclose(pack_edges);
3437 if (show_stats) {
3438 uintmax_t total_count = 0, duplicate_count = 0;
3439 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3440 total_count += object_count_by_type[i];
3441 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3442 duplicate_count += duplicate_count_by_type[i];
3444 fprintf(stderr, "%s statistics:\n", argv[0]);
3445 fprintf(stderr, "---------------------------------------------------------------------\n");
3446 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3447 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3448 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]);
3449 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]);
3450 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]);
3451 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]);
3452 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3453 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3454 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3455 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3456 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3457 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3458 fprintf(stderr, "---------------------------------------------------------------------\n");
3459 pack_report();
3460 fprintf(stderr, "---------------------------------------------------------------------\n");
3461 fprintf(stderr, "\n");
3464 return failure ? 1 : 0;