Merge branch 'ta/docfix-index-format-tech'
[git/mjg.git] / fast-import.c
blob2ad4fee07e208a490d0942cccef07112005de413
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, get-mark, ls-tree, and cat-blob requests may
138 # appear anywhere in the input, except within a data command. Any
139 # form of the data command always escapes the related input from
140 # 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 get_mark ::= 'get-mark' sp idnum lf;
148 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
149 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
151 comment ::= '#' not_lf* lf;
152 not_lf ::= # Any byte that is not ASCII newline (LF);
155 #include "builtin.h"
156 #include "cache.h"
157 #include "lockfile.h"
158 #include "object.h"
159 #include "blob.h"
160 #include "tree.h"
161 #include "commit.h"
162 #include "delta.h"
163 #include "pack.h"
164 #include "refs.h"
165 #include "csum-file.h"
166 #include "quote.h"
167 #include "exec_cmd.h"
168 #include "dir.h"
170 #define PACK_ID_BITS 16
171 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
172 #define DEPTH_BITS 13
173 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
176 * We abuse the setuid bit on directories to mean "do not delta".
178 #define NO_DELTA S_ISUID
180 struct object_entry {
181 struct pack_idx_entry idx;
182 struct object_entry *next;
183 uint32_t type : TYPE_BITS,
184 pack_id : PACK_ID_BITS,
185 depth : DEPTH_BITS;
188 struct object_entry_pool {
189 struct object_entry_pool *next_pool;
190 struct object_entry *next_free;
191 struct object_entry *end;
192 struct object_entry entries[FLEX_ARRAY]; /* more */
195 struct mark_set {
196 union {
197 struct object_entry *marked[1024];
198 struct mark_set *sets[1024];
199 } data;
200 unsigned int shift;
203 struct last_object {
204 struct strbuf data;
205 off_t offset;
206 unsigned int depth;
207 unsigned no_swap : 1;
210 struct mem_pool {
211 struct mem_pool *next_pool;
212 char *next_free;
213 char *end;
214 uintmax_t space[FLEX_ARRAY]; /* more */
217 struct atom_str {
218 struct atom_str *next_atom;
219 unsigned short str_len;
220 char str_dat[FLEX_ARRAY]; /* more */
223 struct tree_content;
224 struct tree_entry {
225 struct tree_content *tree;
226 struct atom_str *name;
227 struct tree_entry_ms {
228 uint16_t mode;
229 unsigned char sha1[20];
230 } versions[2];
233 struct tree_content {
234 unsigned int entry_capacity; /* must match avail_tree_content */
235 unsigned int entry_count;
236 unsigned int delta_depth;
237 struct tree_entry *entries[FLEX_ARRAY]; /* more */
240 struct avail_tree_content {
241 unsigned int entry_capacity; /* must match tree_content */
242 struct avail_tree_content *next_avail;
245 struct branch {
246 struct branch *table_next_branch;
247 struct branch *active_next_branch;
248 const char *name;
249 struct tree_entry branch_tree;
250 uintmax_t last_commit;
251 uintmax_t num_notes;
252 unsigned active : 1;
253 unsigned delete : 1;
254 unsigned pack_id : PACK_ID_BITS;
255 unsigned char sha1[20];
258 struct tag {
259 struct tag *next_tag;
260 const char *name;
261 unsigned int pack_id;
262 unsigned char sha1[20];
265 struct hash_list {
266 struct hash_list *next;
267 unsigned char sha1[20];
270 typedef enum {
271 WHENSPEC_RAW = 1,
272 WHENSPEC_RFC2822,
273 WHENSPEC_NOW
274 } whenspec_type;
276 struct recent_command {
277 struct recent_command *prev;
278 struct recent_command *next;
279 char *buf;
282 /* Configured limits on output */
283 static unsigned long max_depth = 10;
284 static off_t max_packsize;
285 static int force_update;
286 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
287 static int pack_compression_seen;
289 /* Stats and misc. counters */
290 static uintmax_t alloc_count;
291 static uintmax_t marks_set_count;
292 static uintmax_t object_count_by_type[1 << TYPE_BITS];
293 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
294 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
295 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
296 static unsigned long object_count;
297 static unsigned long branch_count;
298 static unsigned long branch_load_count;
299 static int failure;
300 static FILE *pack_edges;
301 static unsigned int show_stats = 1;
302 static int global_argc;
303 static char **global_argv;
305 /* Memory pools */
306 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
307 static size_t total_allocd;
308 static struct mem_pool *mem_pool;
310 /* Atom management */
311 static unsigned int atom_table_sz = 4451;
312 static unsigned int atom_cnt;
313 static struct atom_str **atom_table;
315 /* The .pack file being generated */
316 static struct pack_idx_option pack_idx_opts;
317 static unsigned int pack_id;
318 static struct sha1file *pack_file;
319 static struct packed_git *pack_data;
320 static struct packed_git **all_packs;
321 static off_t pack_size;
323 /* Table of objects we've written. */
324 static unsigned int object_entry_alloc = 5000;
325 static struct object_entry_pool *blocks;
326 static struct object_entry *object_table[1 << 16];
327 static struct mark_set *marks;
328 static const char *export_marks_file;
329 static const char *import_marks_file;
330 static int import_marks_file_from_stream;
331 static int import_marks_file_ignore_missing;
332 static int relative_marks_paths;
334 /* Our last blob */
335 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
337 /* Tree management */
338 static unsigned int tree_entry_alloc = 1000;
339 static void *avail_tree_entry;
340 static unsigned int avail_tree_table_sz = 100;
341 static struct avail_tree_content **avail_tree_table;
342 static struct strbuf old_tree = STRBUF_INIT;
343 static struct strbuf new_tree = STRBUF_INIT;
345 /* Branch data */
346 static unsigned long max_active_branches = 5;
347 static unsigned long cur_active_branches;
348 static unsigned long branch_table_sz = 1039;
349 static struct branch **branch_table;
350 static struct branch *active_branches;
352 /* Tag data */
353 static struct tag *first_tag;
354 static struct tag *last_tag;
356 /* Input stream parsing */
357 static whenspec_type whenspec = WHENSPEC_RAW;
358 static struct strbuf command_buf = STRBUF_INIT;
359 static int unread_command_buf;
360 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
361 static struct recent_command *cmd_tail = &cmd_hist;
362 static struct recent_command *rc_free;
363 static unsigned int cmd_save = 100;
364 static uintmax_t next_mark;
365 static struct strbuf new_data = STRBUF_INIT;
366 static int seen_data_command;
367 static int require_explicit_termination;
369 /* Signal handling */
370 static volatile sig_atomic_t checkpoint_requested;
372 /* Where to write output of cat-blob commands */
373 static int cat_blob_fd = STDOUT_FILENO;
375 static void parse_argv(void);
376 static void parse_get_mark(const char *p);
377 static void parse_cat_blob(const char *p);
378 static void parse_ls(const char *p, struct branch *b);
380 static void write_branch_report(FILE *rpt, struct branch *b)
382 fprintf(rpt, "%s:\n", b->name);
384 fprintf(rpt, " status :");
385 if (b->active)
386 fputs(" active", rpt);
387 if (b->branch_tree.tree)
388 fputs(" loaded", rpt);
389 if (is_null_sha1(b->branch_tree.versions[1].sha1))
390 fputs(" dirty", rpt);
391 fputc('\n', rpt);
393 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
394 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
395 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
396 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
398 fputs(" last pack : ", rpt);
399 if (b->pack_id < MAX_PACK_ID)
400 fprintf(rpt, "%u", b->pack_id);
401 fputc('\n', rpt);
403 fputc('\n', rpt);
406 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
408 static void write_crash_report(const char *err)
410 const char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
411 FILE *rpt = fopen(loc, "w");
412 struct branch *b;
413 unsigned long lu;
414 struct recent_command *rc;
416 if (!rpt) {
417 error("can't write crash report %s: %s", loc, strerror(errno));
418 return;
421 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
423 fprintf(rpt, "fast-import crash report:\n");
424 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
425 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
426 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_MODE(LOCAL)));
427 fputc('\n', rpt);
429 fputs("fatal: ", rpt);
430 fputs(err, rpt);
431 fputc('\n', rpt);
433 fputc('\n', rpt);
434 fputs("Most Recent Commands Before Crash\n", rpt);
435 fputs("---------------------------------\n", rpt);
436 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
437 if (rc->next == &cmd_hist)
438 fputs("* ", rpt);
439 else
440 fputs(" ", rpt);
441 fputs(rc->buf, rpt);
442 fputc('\n', rpt);
445 fputc('\n', rpt);
446 fputs("Active Branch LRU\n", rpt);
447 fputs("-----------------\n", rpt);
448 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
449 cur_active_branches,
450 max_active_branches);
451 fputc('\n', rpt);
452 fputs(" pos clock name\n", rpt);
453 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
454 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
455 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
456 ++lu, b->last_commit, b->name);
458 fputc('\n', rpt);
459 fputs("Inactive Branches\n", rpt);
460 fputs("-----------------\n", rpt);
461 for (lu = 0; lu < branch_table_sz; lu++) {
462 for (b = branch_table[lu]; b; b = b->table_next_branch)
463 write_branch_report(rpt, b);
466 if (first_tag) {
467 struct tag *tg;
468 fputc('\n', rpt);
469 fputs("Annotated Tags\n", rpt);
470 fputs("--------------\n", rpt);
471 for (tg = first_tag; tg; tg = tg->next_tag) {
472 fputs(sha1_to_hex(tg->sha1), rpt);
473 fputc(' ', rpt);
474 fputs(tg->name, rpt);
475 fputc('\n', rpt);
479 fputc('\n', rpt);
480 fputs("Marks\n", rpt);
481 fputs("-----\n", rpt);
482 if (export_marks_file)
483 fprintf(rpt, " exported to %s\n", export_marks_file);
484 else
485 dump_marks_helper(rpt, 0, marks);
487 fputc('\n', rpt);
488 fputs("-------------------\n", rpt);
489 fputs("END OF CRASH REPORT\n", rpt);
490 fclose(rpt);
493 static void end_packfile(void);
494 static void unkeep_all_packs(void);
495 static void dump_marks(void);
497 static NORETURN void die_nicely(const char *err, va_list params)
499 static int zombie;
500 char message[2 * PATH_MAX];
502 vsnprintf(message, sizeof(message), err, params);
503 fputs("fatal: ", stderr);
504 fputs(message, stderr);
505 fputc('\n', stderr);
507 if (!zombie) {
508 zombie = 1;
509 write_crash_report(message);
510 end_packfile();
511 unkeep_all_packs();
512 dump_marks();
514 exit(128);
517 #ifndef SIGUSR1 /* Windows, for example */
519 static void set_checkpoint_signal(void)
523 #else
525 static void checkpoint_signal(int signo)
527 checkpoint_requested = 1;
530 static void set_checkpoint_signal(void)
532 struct sigaction sa;
534 memset(&sa, 0, sizeof(sa));
535 sa.sa_handler = checkpoint_signal;
536 sigemptyset(&sa.sa_mask);
537 sa.sa_flags = SA_RESTART;
538 sigaction(SIGUSR1, &sa, NULL);
541 #endif
543 static void alloc_objects(unsigned int cnt)
545 struct object_entry_pool *b;
547 b = xmalloc(sizeof(struct object_entry_pool)
548 + cnt * sizeof(struct object_entry));
549 b->next_pool = blocks;
550 b->next_free = b->entries;
551 b->end = b->entries + cnt;
552 blocks = b;
553 alloc_count += cnt;
556 static struct object_entry *new_object(unsigned char *sha1)
558 struct object_entry *e;
560 if (blocks->next_free == blocks->end)
561 alloc_objects(object_entry_alloc);
563 e = blocks->next_free++;
564 hashcpy(e->idx.sha1, sha1);
565 return e;
568 static struct object_entry *find_object(unsigned char *sha1)
570 unsigned int h = sha1[0] << 8 | sha1[1];
571 struct object_entry *e;
572 for (e = object_table[h]; e; e = e->next)
573 if (!hashcmp(sha1, e->idx.sha1))
574 return e;
575 return NULL;
578 static struct object_entry *insert_object(unsigned char *sha1)
580 unsigned int h = sha1[0] << 8 | sha1[1];
581 struct object_entry *e = object_table[h];
583 while (e) {
584 if (!hashcmp(sha1, e->idx.sha1))
585 return e;
586 e = e->next;
589 e = new_object(sha1);
590 e->next = object_table[h];
591 e->idx.offset = 0;
592 object_table[h] = e;
593 return e;
596 static unsigned int hc_str(const char *s, size_t len)
598 unsigned int r = 0;
599 while (len-- > 0)
600 r = r * 31 + *s++;
601 return r;
604 static void *pool_alloc(size_t len)
606 struct mem_pool *p;
607 void *r;
609 /* round up to a 'uintmax_t' alignment */
610 if (len & (sizeof(uintmax_t) - 1))
611 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
613 for (p = mem_pool; p; p = p->next_pool)
614 if ((p->end - p->next_free >= len))
615 break;
617 if (!p) {
618 if (len >= (mem_pool_alloc/2)) {
619 total_allocd += len;
620 return xmalloc(len);
622 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
623 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
624 p->next_pool = mem_pool;
625 p->next_free = (char *) p->space;
626 p->end = p->next_free + mem_pool_alloc;
627 mem_pool = p;
630 r = p->next_free;
631 p->next_free += len;
632 return r;
635 static void *pool_calloc(size_t count, size_t size)
637 size_t len = count * size;
638 void *r = pool_alloc(len);
639 memset(r, 0, len);
640 return r;
643 static char *pool_strdup(const char *s)
645 char *r = pool_alloc(strlen(s) + 1);
646 strcpy(r, s);
647 return r;
650 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
652 struct mark_set *s = marks;
653 while ((idnum >> s->shift) >= 1024) {
654 s = pool_calloc(1, sizeof(struct mark_set));
655 s->shift = marks->shift + 10;
656 s->data.sets[0] = marks;
657 marks = s;
659 while (s->shift) {
660 uintmax_t i = idnum >> s->shift;
661 idnum -= i << s->shift;
662 if (!s->data.sets[i]) {
663 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
664 s->data.sets[i]->shift = s->shift - 10;
666 s = s->data.sets[i];
668 if (!s->data.marked[idnum])
669 marks_set_count++;
670 s->data.marked[idnum] = oe;
673 static struct object_entry *find_mark(uintmax_t idnum)
675 uintmax_t orig_idnum = idnum;
676 struct mark_set *s = marks;
677 struct object_entry *oe = NULL;
678 if ((idnum >> s->shift) < 1024) {
679 while (s && s->shift) {
680 uintmax_t i = idnum >> s->shift;
681 idnum -= i << s->shift;
682 s = s->data.sets[i];
684 if (s)
685 oe = s->data.marked[idnum];
687 if (!oe)
688 die("mark :%" PRIuMAX " not declared", orig_idnum);
689 return oe;
692 static struct atom_str *to_atom(const char *s, unsigned short len)
694 unsigned int hc = hc_str(s, len) % atom_table_sz;
695 struct atom_str *c;
697 for (c = atom_table[hc]; c; c = c->next_atom)
698 if (c->str_len == len && !strncmp(s, c->str_dat, len))
699 return c;
701 c = pool_alloc(sizeof(struct atom_str) + len + 1);
702 c->str_len = len;
703 strncpy(c->str_dat, s, len);
704 c->str_dat[len] = 0;
705 c->next_atom = atom_table[hc];
706 atom_table[hc] = c;
707 atom_cnt++;
708 return c;
711 static struct branch *lookup_branch(const char *name)
713 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
714 struct branch *b;
716 for (b = branch_table[hc]; b; b = b->table_next_branch)
717 if (!strcmp(name, b->name))
718 return b;
719 return NULL;
722 static struct branch *new_branch(const char *name)
724 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
725 struct branch *b = lookup_branch(name);
727 if (b)
728 die("Invalid attempt to create duplicate branch: %s", name);
729 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
730 die("Branch name doesn't conform to GIT standards: %s", name);
732 b = pool_calloc(1, sizeof(struct branch));
733 b->name = pool_strdup(name);
734 b->table_next_branch = branch_table[hc];
735 b->branch_tree.versions[0].mode = S_IFDIR;
736 b->branch_tree.versions[1].mode = S_IFDIR;
737 b->num_notes = 0;
738 b->active = 0;
739 b->pack_id = MAX_PACK_ID;
740 branch_table[hc] = b;
741 branch_count++;
742 return b;
745 static unsigned int hc_entries(unsigned int cnt)
747 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
748 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
751 static struct tree_content *new_tree_content(unsigned int cnt)
753 struct avail_tree_content *f, *l = NULL;
754 struct tree_content *t;
755 unsigned int hc = hc_entries(cnt);
757 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
758 if (f->entry_capacity >= cnt)
759 break;
761 if (f) {
762 if (l)
763 l->next_avail = f->next_avail;
764 else
765 avail_tree_table[hc] = f->next_avail;
766 } else {
767 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
768 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
769 f->entry_capacity = cnt;
772 t = (struct tree_content*)f;
773 t->entry_count = 0;
774 t->delta_depth = 0;
775 return t;
778 static void release_tree_entry(struct tree_entry *e);
779 static void release_tree_content(struct tree_content *t)
781 struct avail_tree_content *f = (struct avail_tree_content*)t;
782 unsigned int hc = hc_entries(f->entry_capacity);
783 f->next_avail = avail_tree_table[hc];
784 avail_tree_table[hc] = f;
787 static void release_tree_content_recursive(struct tree_content *t)
789 unsigned int i;
790 for (i = 0; i < t->entry_count; i++)
791 release_tree_entry(t->entries[i]);
792 release_tree_content(t);
795 static struct tree_content *grow_tree_content(
796 struct tree_content *t,
797 int amt)
799 struct tree_content *r = new_tree_content(t->entry_count + amt);
800 r->entry_count = t->entry_count;
801 r->delta_depth = t->delta_depth;
802 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
803 release_tree_content(t);
804 return r;
807 static struct tree_entry *new_tree_entry(void)
809 struct tree_entry *e;
811 if (!avail_tree_entry) {
812 unsigned int n = tree_entry_alloc;
813 total_allocd += n * sizeof(struct tree_entry);
814 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
815 while (n-- > 1) {
816 *((void**)e) = e + 1;
817 e++;
819 *((void**)e) = NULL;
822 e = avail_tree_entry;
823 avail_tree_entry = *((void**)e);
824 return e;
827 static void release_tree_entry(struct tree_entry *e)
829 if (e->tree)
830 release_tree_content_recursive(e->tree);
831 *((void**)e) = avail_tree_entry;
832 avail_tree_entry = e;
835 static struct tree_content *dup_tree_content(struct tree_content *s)
837 struct tree_content *d;
838 struct tree_entry *a, *b;
839 unsigned int i;
841 if (!s)
842 return NULL;
843 d = new_tree_content(s->entry_count);
844 for (i = 0; i < s->entry_count; i++) {
845 a = s->entries[i];
846 b = new_tree_entry();
847 memcpy(b, a, sizeof(*a));
848 if (a->tree && is_null_sha1(b->versions[1].sha1))
849 b->tree = dup_tree_content(a->tree);
850 else
851 b->tree = NULL;
852 d->entries[i] = b;
854 d->entry_count = s->entry_count;
855 d->delta_depth = s->delta_depth;
857 return d;
860 static void start_packfile(void)
862 static char tmp_file[PATH_MAX];
863 struct packed_git *p;
864 struct pack_header hdr;
865 int pack_fd;
867 pack_fd = odb_mkstemp(tmp_file, sizeof(tmp_file),
868 "pack/tmp_pack_XXXXXX");
869 p = xcalloc(1, sizeof(*p) + strlen(tmp_file) + 2);
870 strcpy(p->pack_name, tmp_file);
871 p->pack_fd = pack_fd;
872 p->do_not_close = 1;
873 pack_file = sha1fd(pack_fd, p->pack_name);
875 hdr.hdr_signature = htonl(PACK_SIGNATURE);
876 hdr.hdr_version = htonl(2);
877 hdr.hdr_entries = 0;
878 sha1write(pack_file, &hdr, sizeof(hdr));
880 pack_data = p;
881 pack_size = sizeof(hdr);
882 object_count = 0;
884 REALLOC_ARRAY(all_packs, pack_id + 1);
885 all_packs[pack_id] = p;
888 static const char *create_index(void)
890 const char *tmpfile;
891 struct pack_idx_entry **idx, **c, **last;
892 struct object_entry *e;
893 struct object_entry_pool *o;
895 /* Build the table of object IDs. */
896 idx = xmalloc(object_count * sizeof(*idx));
897 c = idx;
898 for (o = blocks; o; o = o->next_pool)
899 for (e = o->next_free; e-- != o->entries;)
900 if (pack_id == e->pack_id)
901 *c++ = &e->idx;
902 last = idx + object_count;
903 if (c != last)
904 die("internal consistency error creating the index");
906 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
907 free(idx);
908 return tmpfile;
911 static char *keep_pack(const char *curr_index_name)
913 static char name[PATH_MAX];
914 static const char *keep_msg = "fast-import";
915 int keep_fd;
917 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
918 if (keep_fd < 0)
919 die_errno("cannot create keep file");
920 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
921 if (close(keep_fd))
922 die_errno("failed to write keep file");
924 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
925 get_object_directory(), sha1_to_hex(pack_data->sha1));
926 if (move_temp_to_file(pack_data->pack_name, name))
927 die("cannot store pack file");
929 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
930 get_object_directory(), sha1_to_hex(pack_data->sha1));
931 if (move_temp_to_file(curr_index_name, name))
932 die("cannot store index file");
933 free((void *)curr_index_name);
934 return name;
937 static void unkeep_all_packs(void)
939 static char name[PATH_MAX];
940 int k;
942 for (k = 0; k < pack_id; k++) {
943 struct packed_git *p = all_packs[k];
944 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
945 get_object_directory(), sha1_to_hex(p->sha1));
946 unlink_or_warn(name);
950 static void end_packfile(void)
952 static int running;
954 if (running || !pack_data)
955 return;
957 running = 1;
958 clear_delta_base_cache();
959 if (object_count) {
960 struct packed_git *new_p;
961 unsigned char cur_pack_sha1[20];
962 char *idx_name;
963 int i;
964 struct branch *b;
965 struct tag *t;
967 close_pack_windows(pack_data);
968 sha1close(pack_file, cur_pack_sha1, 0);
969 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
970 pack_data->pack_name, object_count,
971 cur_pack_sha1, pack_size);
972 close(pack_data->pack_fd);
973 idx_name = keep_pack(create_index());
975 /* Register the packfile with core git's machinery. */
976 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
977 if (!new_p)
978 die("core git rejected index %s", idx_name);
979 all_packs[pack_id] = new_p;
980 install_packed_git(new_p);
982 /* Print the boundary */
983 if (pack_edges) {
984 fprintf(pack_edges, "%s:", new_p->pack_name);
985 for (i = 0; i < branch_table_sz; i++) {
986 for (b = branch_table[i]; b; b = b->table_next_branch) {
987 if (b->pack_id == pack_id)
988 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
991 for (t = first_tag; t; t = t->next_tag) {
992 if (t->pack_id == pack_id)
993 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
995 fputc('\n', pack_edges);
996 fflush(pack_edges);
999 pack_id++;
1001 else {
1002 close(pack_data->pack_fd);
1003 unlink_or_warn(pack_data->pack_name);
1005 free(pack_data);
1006 pack_data = NULL;
1007 running = 0;
1009 /* We can't carry a delta across packfiles. */
1010 strbuf_release(&last_blob.data);
1011 last_blob.offset = 0;
1012 last_blob.depth = 0;
1015 static void cycle_packfile(void)
1017 end_packfile();
1018 start_packfile();
1021 static int store_object(
1022 enum object_type type,
1023 struct strbuf *dat,
1024 struct last_object *last,
1025 unsigned char *sha1out,
1026 uintmax_t mark)
1028 void *out, *delta;
1029 struct object_entry *e;
1030 unsigned char hdr[96];
1031 unsigned char sha1[20];
1032 unsigned long hdrlen, deltalen;
1033 git_SHA_CTX c;
1034 git_zstream s;
1036 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1037 (unsigned long)dat->len) + 1;
1038 git_SHA1_Init(&c);
1039 git_SHA1_Update(&c, hdr, hdrlen);
1040 git_SHA1_Update(&c, dat->buf, dat->len);
1041 git_SHA1_Final(sha1, &c);
1042 if (sha1out)
1043 hashcpy(sha1out, sha1);
1045 e = insert_object(sha1);
1046 if (mark)
1047 insert_mark(mark, e);
1048 if (e->idx.offset) {
1049 duplicate_count_by_type[type]++;
1050 return 1;
1051 } else if (find_sha1_pack(sha1, packed_git)) {
1052 e->type = type;
1053 e->pack_id = MAX_PACK_ID;
1054 e->idx.offset = 1; /* just not zero! */
1055 duplicate_count_by_type[type]++;
1056 return 1;
1059 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1060 delta_count_attempts_by_type[type]++;
1061 delta = diff_delta(last->data.buf, last->data.len,
1062 dat->buf, dat->len,
1063 &deltalen, dat->len - 20);
1064 } else
1065 delta = NULL;
1067 git_deflate_init(&s, pack_compression_level);
1068 if (delta) {
1069 s.next_in = delta;
1070 s.avail_in = deltalen;
1071 } else {
1072 s.next_in = (void *)dat->buf;
1073 s.avail_in = dat->len;
1075 s.avail_out = git_deflate_bound(&s, s.avail_in);
1076 s.next_out = out = xmalloc(s.avail_out);
1077 while (git_deflate(&s, Z_FINISH) == Z_OK)
1078 ; /* nothing */
1079 git_deflate_end(&s);
1081 /* Determine if we should auto-checkpoint. */
1082 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1083 || (pack_size + 60 + s.total_out) < pack_size) {
1085 /* This new object needs to *not* have the current pack_id. */
1086 e->pack_id = pack_id + 1;
1087 cycle_packfile();
1089 /* We cannot carry a delta into the new pack. */
1090 if (delta) {
1091 free(delta);
1092 delta = NULL;
1094 git_deflate_init(&s, pack_compression_level);
1095 s.next_in = (void *)dat->buf;
1096 s.avail_in = dat->len;
1097 s.avail_out = git_deflate_bound(&s, s.avail_in);
1098 s.next_out = out = xrealloc(out, s.avail_out);
1099 while (git_deflate(&s, Z_FINISH) == Z_OK)
1100 ; /* nothing */
1101 git_deflate_end(&s);
1105 e->type = type;
1106 e->pack_id = pack_id;
1107 e->idx.offset = pack_size;
1108 object_count++;
1109 object_count_by_type[type]++;
1111 crc32_begin(pack_file);
1113 if (delta) {
1114 off_t ofs = e->idx.offset - last->offset;
1115 unsigned pos = sizeof(hdr) - 1;
1117 delta_count_by_type[type]++;
1118 e->depth = last->depth + 1;
1120 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1121 sha1write(pack_file, hdr, hdrlen);
1122 pack_size += hdrlen;
1124 hdr[pos] = ofs & 127;
1125 while (ofs >>= 7)
1126 hdr[--pos] = 128 | (--ofs & 127);
1127 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1128 pack_size += sizeof(hdr) - pos;
1129 } else {
1130 e->depth = 0;
1131 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1132 sha1write(pack_file, hdr, hdrlen);
1133 pack_size += hdrlen;
1136 sha1write(pack_file, out, s.total_out);
1137 pack_size += s.total_out;
1139 e->idx.crc32 = crc32_end(pack_file);
1141 free(out);
1142 free(delta);
1143 if (last) {
1144 if (last->no_swap) {
1145 last->data = *dat;
1146 } else {
1147 strbuf_swap(&last->data, dat);
1149 last->offset = e->idx.offset;
1150 last->depth = e->depth;
1152 return 0;
1155 static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1157 if (sha1file_truncate(pack_file, checkpoint))
1158 die_errno("cannot truncate pack to skip duplicate");
1159 pack_size = checkpoint->offset;
1162 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1164 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1165 unsigned char *in_buf = xmalloc(in_sz);
1166 unsigned char *out_buf = xmalloc(out_sz);
1167 struct object_entry *e;
1168 unsigned char sha1[20];
1169 unsigned long hdrlen;
1170 off_t offset;
1171 git_SHA_CTX c;
1172 git_zstream s;
1173 struct sha1file_checkpoint checkpoint;
1174 int status = Z_OK;
1176 /* Determine if we should auto-checkpoint. */
1177 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1178 || (pack_size + 60 + len) < pack_size)
1179 cycle_packfile();
1181 sha1file_checkpoint(pack_file, &checkpoint);
1182 offset = checkpoint.offset;
1184 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1185 if (out_sz <= hdrlen)
1186 die("impossibly large object header");
1188 git_SHA1_Init(&c);
1189 git_SHA1_Update(&c, out_buf, hdrlen);
1191 crc32_begin(pack_file);
1193 git_deflate_init(&s, pack_compression_level);
1195 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1196 if (out_sz <= hdrlen)
1197 die("impossibly large object header");
1199 s.next_out = out_buf + hdrlen;
1200 s.avail_out = out_sz - hdrlen;
1202 while (status != Z_STREAM_END) {
1203 if (0 < len && !s.avail_in) {
1204 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1205 size_t n = fread(in_buf, 1, cnt, stdin);
1206 if (!n && feof(stdin))
1207 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1209 git_SHA1_Update(&c, in_buf, n);
1210 s.next_in = in_buf;
1211 s.avail_in = n;
1212 len -= n;
1215 status = git_deflate(&s, len ? 0 : Z_FINISH);
1217 if (!s.avail_out || status == Z_STREAM_END) {
1218 size_t n = s.next_out - out_buf;
1219 sha1write(pack_file, out_buf, n);
1220 pack_size += n;
1221 s.next_out = out_buf;
1222 s.avail_out = out_sz;
1225 switch (status) {
1226 case Z_OK:
1227 case Z_BUF_ERROR:
1228 case Z_STREAM_END:
1229 continue;
1230 default:
1231 die("unexpected deflate failure: %d", status);
1234 git_deflate_end(&s);
1235 git_SHA1_Final(sha1, &c);
1237 if (sha1out)
1238 hashcpy(sha1out, sha1);
1240 e = insert_object(sha1);
1242 if (mark)
1243 insert_mark(mark, e);
1245 if (e->idx.offset) {
1246 duplicate_count_by_type[OBJ_BLOB]++;
1247 truncate_pack(&checkpoint);
1249 } else if (find_sha1_pack(sha1, packed_git)) {
1250 e->type = OBJ_BLOB;
1251 e->pack_id = MAX_PACK_ID;
1252 e->idx.offset = 1; /* just not zero! */
1253 duplicate_count_by_type[OBJ_BLOB]++;
1254 truncate_pack(&checkpoint);
1256 } else {
1257 e->depth = 0;
1258 e->type = OBJ_BLOB;
1259 e->pack_id = pack_id;
1260 e->idx.offset = offset;
1261 e->idx.crc32 = crc32_end(pack_file);
1262 object_count++;
1263 object_count_by_type[OBJ_BLOB]++;
1266 free(in_buf);
1267 free(out_buf);
1270 /* All calls must be guarded by find_object() or find_mark() to
1271 * ensure the 'struct object_entry' passed was written by this
1272 * process instance. We unpack the entry by the offset, avoiding
1273 * the need for the corresponding .idx file. This unpacking rule
1274 * works because we only use OBJ_REF_DELTA within the packfiles
1275 * created by fast-import.
1277 * oe must not be NULL. Such an oe usually comes from giving
1278 * an unknown SHA-1 to find_object() or an undefined mark to
1279 * find_mark(). Callers must test for this condition and use
1280 * the standard read_sha1_file() when it happens.
1282 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1283 * find_mark(), where the mark was reloaded from an existing marks
1284 * file and is referencing an object that this fast-import process
1285 * instance did not write out to a packfile. Callers must test for
1286 * this condition and use read_sha1_file() instead.
1288 static void *gfi_unpack_entry(
1289 struct object_entry *oe,
1290 unsigned long *sizep)
1292 enum object_type type;
1293 struct packed_git *p = all_packs[oe->pack_id];
1294 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1295 /* The object is stored in the packfile we are writing to
1296 * and we have modified it since the last time we scanned
1297 * back to read a previously written object. If an old
1298 * window covered [p->pack_size, p->pack_size + 20) its
1299 * data is stale and is not valid. Closing all windows
1300 * and updating the packfile length ensures we can read
1301 * the newly written data.
1303 close_pack_windows(p);
1304 sha1flush(pack_file);
1306 /* We have to offer 20 bytes additional on the end of
1307 * the packfile as the core unpacker code assumes the
1308 * footer is present at the file end and must promise
1309 * at least 20 bytes within any window it maps. But
1310 * we don't actually create the footer here.
1312 p->pack_size = pack_size + 20;
1314 return unpack_entry(p, oe->idx.offset, &type, sizep);
1317 static const char *get_mode(const char *str, uint16_t *modep)
1319 unsigned char c;
1320 uint16_t mode = 0;
1322 while ((c = *str++) != ' ') {
1323 if (c < '0' || c > '7')
1324 return NULL;
1325 mode = (mode << 3) + (c - '0');
1327 *modep = mode;
1328 return str;
1331 static void load_tree(struct tree_entry *root)
1333 unsigned char *sha1 = root->versions[1].sha1;
1334 struct object_entry *myoe;
1335 struct tree_content *t;
1336 unsigned long size;
1337 char *buf;
1338 const char *c;
1340 root->tree = t = new_tree_content(8);
1341 if (is_null_sha1(sha1))
1342 return;
1344 myoe = find_object(sha1);
1345 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1346 if (myoe->type != OBJ_TREE)
1347 die("Not a tree: %s", sha1_to_hex(sha1));
1348 t->delta_depth = myoe->depth;
1349 buf = gfi_unpack_entry(myoe, &size);
1350 if (!buf)
1351 die("Can't load tree %s", sha1_to_hex(sha1));
1352 } else {
1353 enum object_type type;
1354 buf = read_sha1_file(sha1, &type, &size);
1355 if (!buf || type != OBJ_TREE)
1356 die("Can't load tree %s", sha1_to_hex(sha1));
1359 c = buf;
1360 while (c != (buf + size)) {
1361 struct tree_entry *e = new_tree_entry();
1363 if (t->entry_count == t->entry_capacity)
1364 root->tree = t = grow_tree_content(t, t->entry_count);
1365 t->entries[t->entry_count++] = e;
1367 e->tree = NULL;
1368 c = get_mode(c, &e->versions[1].mode);
1369 if (!c)
1370 die("Corrupt mode in %s", sha1_to_hex(sha1));
1371 e->versions[0].mode = e->versions[1].mode;
1372 e->name = to_atom(c, strlen(c));
1373 c += e->name->str_len + 1;
1374 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1375 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1376 c += 20;
1378 free(buf);
1381 static int tecmp0 (const void *_a, const void *_b)
1383 struct tree_entry *a = *((struct tree_entry**)_a);
1384 struct tree_entry *b = *((struct tree_entry**)_b);
1385 return base_name_compare(
1386 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1387 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1390 static int tecmp1 (const void *_a, const void *_b)
1392 struct tree_entry *a = *((struct tree_entry**)_a);
1393 struct tree_entry *b = *((struct tree_entry**)_b);
1394 return base_name_compare(
1395 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1396 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1399 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1401 size_t maxlen = 0;
1402 unsigned int i;
1404 if (!v)
1405 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1406 else
1407 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1409 for (i = 0; i < t->entry_count; i++) {
1410 if (t->entries[i]->versions[v].mode)
1411 maxlen += t->entries[i]->name->str_len + 34;
1414 strbuf_reset(b);
1415 strbuf_grow(b, maxlen);
1416 for (i = 0; i < t->entry_count; i++) {
1417 struct tree_entry *e = t->entries[i];
1418 if (!e->versions[v].mode)
1419 continue;
1420 strbuf_addf(b, "%o %s%c",
1421 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1422 e->name->str_dat, '\0');
1423 strbuf_add(b, e->versions[v].sha1, 20);
1427 static void store_tree(struct tree_entry *root)
1429 struct tree_content *t;
1430 unsigned int i, j, del;
1431 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1432 struct object_entry *le = NULL;
1434 if (!is_null_sha1(root->versions[1].sha1))
1435 return;
1437 if (!root->tree)
1438 load_tree(root);
1439 t = root->tree;
1441 for (i = 0; i < t->entry_count; i++) {
1442 if (t->entries[i]->tree)
1443 store_tree(t->entries[i]);
1446 if (!(root->versions[0].mode & NO_DELTA))
1447 le = find_object(root->versions[0].sha1);
1448 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1449 mktree(t, 0, &old_tree);
1450 lo.data = old_tree;
1451 lo.offset = le->idx.offset;
1452 lo.depth = t->delta_depth;
1455 mktree(t, 1, &new_tree);
1456 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1458 t->delta_depth = lo.depth;
1459 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1460 struct tree_entry *e = t->entries[i];
1461 if (e->versions[1].mode) {
1462 e->versions[0].mode = e->versions[1].mode;
1463 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1464 t->entries[j++] = e;
1465 } else {
1466 release_tree_entry(e);
1467 del++;
1470 t->entry_count -= del;
1473 static void tree_content_replace(
1474 struct tree_entry *root,
1475 const unsigned char *sha1,
1476 const uint16_t mode,
1477 struct tree_content *newtree)
1479 if (!S_ISDIR(mode))
1480 die("Root cannot be a non-directory");
1481 hashclr(root->versions[0].sha1);
1482 hashcpy(root->versions[1].sha1, sha1);
1483 if (root->tree)
1484 release_tree_content_recursive(root->tree);
1485 root->tree = newtree;
1488 static int tree_content_set(
1489 struct tree_entry *root,
1490 const char *p,
1491 const unsigned char *sha1,
1492 const uint16_t mode,
1493 struct tree_content *subtree)
1495 struct tree_content *t;
1496 const char *slash1;
1497 unsigned int i, n;
1498 struct tree_entry *e;
1500 slash1 = strchrnul(p, '/');
1501 n = slash1 - p;
1502 if (!n)
1503 die("Empty path component found in input");
1504 if (!*slash1 && !S_ISDIR(mode) && subtree)
1505 die("Non-directories cannot have subtrees");
1507 if (!root->tree)
1508 load_tree(root);
1509 t = root->tree;
1510 for (i = 0; i < t->entry_count; i++) {
1511 e = t->entries[i];
1512 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1513 if (!*slash1) {
1514 if (!S_ISDIR(mode)
1515 && e->versions[1].mode == mode
1516 && !hashcmp(e->versions[1].sha1, sha1))
1517 return 0;
1518 e->versions[1].mode = mode;
1519 hashcpy(e->versions[1].sha1, sha1);
1520 if (e->tree)
1521 release_tree_content_recursive(e->tree);
1522 e->tree = subtree;
1525 * We need to leave e->versions[0].sha1 alone
1526 * to avoid modifying the preimage tree used
1527 * when writing out the parent directory.
1528 * But after replacing the subdir with a
1529 * completely different one, it's not a good
1530 * delta base any more, and besides, we've
1531 * thrown away the tree entries needed to
1532 * make a delta against it.
1534 * So let's just explicitly disable deltas
1535 * for the subtree.
1537 if (S_ISDIR(e->versions[0].mode))
1538 e->versions[0].mode |= NO_DELTA;
1540 hashclr(root->versions[1].sha1);
1541 return 1;
1543 if (!S_ISDIR(e->versions[1].mode)) {
1544 e->tree = new_tree_content(8);
1545 e->versions[1].mode = S_IFDIR;
1547 if (!e->tree)
1548 load_tree(e);
1549 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1550 hashclr(root->versions[1].sha1);
1551 return 1;
1553 return 0;
1557 if (t->entry_count == t->entry_capacity)
1558 root->tree = t = grow_tree_content(t, t->entry_count);
1559 e = new_tree_entry();
1560 e->name = to_atom(p, n);
1561 e->versions[0].mode = 0;
1562 hashclr(e->versions[0].sha1);
1563 t->entries[t->entry_count++] = e;
1564 if (*slash1) {
1565 e->tree = new_tree_content(8);
1566 e->versions[1].mode = S_IFDIR;
1567 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1568 } else {
1569 e->tree = subtree;
1570 e->versions[1].mode = mode;
1571 hashcpy(e->versions[1].sha1, sha1);
1573 hashclr(root->versions[1].sha1);
1574 return 1;
1577 static int tree_content_remove(
1578 struct tree_entry *root,
1579 const char *p,
1580 struct tree_entry *backup_leaf,
1581 int allow_root)
1583 struct tree_content *t;
1584 const char *slash1;
1585 unsigned int i, n;
1586 struct tree_entry *e;
1588 slash1 = strchrnul(p, '/');
1589 n = slash1 - p;
1591 if (!root->tree)
1592 load_tree(root);
1594 if (!*p && allow_root) {
1595 e = root;
1596 goto del_entry;
1599 t = root->tree;
1600 for (i = 0; i < t->entry_count; i++) {
1601 e = t->entries[i];
1602 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1603 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1605 * If p names a file in some subdirectory, and a
1606 * file or symlink matching the name of the
1607 * parent directory of p exists, then p cannot
1608 * exist and need not be deleted.
1610 return 1;
1611 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1612 goto del_entry;
1613 if (!e->tree)
1614 load_tree(e);
1615 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1616 for (n = 0; n < e->tree->entry_count; n++) {
1617 if (e->tree->entries[n]->versions[1].mode) {
1618 hashclr(root->versions[1].sha1);
1619 return 1;
1622 backup_leaf = NULL;
1623 goto del_entry;
1625 return 0;
1628 return 0;
1630 del_entry:
1631 if (backup_leaf)
1632 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1633 else if (e->tree)
1634 release_tree_content_recursive(e->tree);
1635 e->tree = NULL;
1636 e->versions[1].mode = 0;
1637 hashclr(e->versions[1].sha1);
1638 hashclr(root->versions[1].sha1);
1639 return 1;
1642 static int tree_content_get(
1643 struct tree_entry *root,
1644 const char *p,
1645 struct tree_entry *leaf,
1646 int allow_root)
1648 struct tree_content *t;
1649 const char *slash1;
1650 unsigned int i, n;
1651 struct tree_entry *e;
1653 slash1 = strchrnul(p, '/');
1654 n = slash1 - p;
1655 if (!n && !allow_root)
1656 die("Empty path component found in input");
1658 if (!root->tree)
1659 load_tree(root);
1661 if (!n) {
1662 e = root;
1663 goto found_entry;
1666 t = root->tree;
1667 for (i = 0; i < t->entry_count; i++) {
1668 e = t->entries[i];
1669 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1670 if (!*slash1)
1671 goto found_entry;
1672 if (!S_ISDIR(e->versions[1].mode))
1673 return 0;
1674 if (!e->tree)
1675 load_tree(e);
1676 return tree_content_get(e, slash1 + 1, leaf, 0);
1679 return 0;
1681 found_entry:
1682 memcpy(leaf, e, sizeof(*leaf));
1683 if (e->tree && is_null_sha1(e->versions[1].sha1))
1684 leaf->tree = dup_tree_content(e->tree);
1685 else
1686 leaf->tree = NULL;
1687 return 1;
1690 static int update_branch(struct branch *b)
1692 static const char *msg = "fast-import";
1693 struct ref_transaction *transaction;
1694 unsigned char old_sha1[20];
1695 struct strbuf err = STRBUF_INIT;
1697 if (is_null_sha1(b->sha1)) {
1698 if (b->delete)
1699 delete_ref(b->name, NULL, 0);
1700 return 0;
1702 if (read_ref(b->name, old_sha1))
1703 hashclr(old_sha1);
1704 if (!force_update && !is_null_sha1(old_sha1)) {
1705 struct commit *old_cmit, *new_cmit;
1707 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1708 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1709 if (!old_cmit || !new_cmit)
1710 return error("Branch %s is missing commits.", b->name);
1712 if (!in_merge_bases(old_cmit, new_cmit)) {
1713 warning("Not updating %s"
1714 " (new tip %s does not contain %s)",
1715 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1716 return -1;
1719 transaction = ref_transaction_begin(&err);
1720 if (!transaction ||
1721 ref_transaction_update(transaction, b->name, b->sha1, old_sha1,
1722 0, msg, &err) ||
1723 ref_transaction_commit(transaction, &err)) {
1724 ref_transaction_free(transaction);
1725 error("%s", err.buf);
1726 strbuf_release(&err);
1727 return -1;
1729 ref_transaction_free(transaction);
1730 strbuf_release(&err);
1731 return 0;
1734 static void dump_branches(void)
1736 unsigned int i;
1737 struct branch *b;
1739 for (i = 0; i < branch_table_sz; i++) {
1740 for (b = branch_table[i]; b; b = b->table_next_branch)
1741 failure |= update_branch(b);
1745 static void dump_tags(void)
1747 static const char *msg = "fast-import";
1748 struct tag *t;
1749 struct strbuf ref_name = STRBUF_INIT;
1750 struct strbuf err = STRBUF_INIT;
1751 struct ref_transaction *transaction;
1753 transaction = ref_transaction_begin(&err);
1754 if (!transaction) {
1755 failure |= error("%s", err.buf);
1756 goto cleanup;
1758 for (t = first_tag; t; t = t->next_tag) {
1759 strbuf_reset(&ref_name);
1760 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1762 if (ref_transaction_update(transaction, ref_name.buf,
1763 t->sha1, NULL, 0, msg, &err)) {
1764 failure |= error("%s", err.buf);
1765 goto cleanup;
1768 if (ref_transaction_commit(transaction, &err))
1769 failure |= error("%s", err.buf);
1771 cleanup:
1772 ref_transaction_free(transaction);
1773 strbuf_release(&ref_name);
1774 strbuf_release(&err);
1777 static void dump_marks_helper(FILE *f,
1778 uintmax_t base,
1779 struct mark_set *m)
1781 uintmax_t k;
1782 if (m->shift) {
1783 for (k = 0; k < 1024; k++) {
1784 if (m->data.sets[k])
1785 dump_marks_helper(f, base + (k << m->shift),
1786 m->data.sets[k]);
1788 } else {
1789 for (k = 0; k < 1024; k++) {
1790 if (m->data.marked[k])
1791 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1792 sha1_to_hex(m->data.marked[k]->idx.sha1));
1797 static void dump_marks(void)
1799 static struct lock_file mark_lock;
1800 FILE *f;
1802 if (!export_marks_file)
1803 return;
1805 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1806 failure |= error("Unable to write marks file %s: %s",
1807 export_marks_file, strerror(errno));
1808 return;
1811 f = fdopen_lock_file(&mark_lock, "w");
1812 if (!f) {
1813 int saved_errno = errno;
1814 rollback_lock_file(&mark_lock);
1815 failure |= error("Unable to write marks file %s: %s",
1816 export_marks_file, strerror(saved_errno));
1817 return;
1820 dump_marks_helper(f, 0, marks);
1821 if (commit_lock_file(&mark_lock)) {
1822 failure |= error("Unable to commit marks file %s: %s",
1823 export_marks_file, strerror(errno));
1824 return;
1828 static void read_marks(void)
1830 char line[512];
1831 FILE *f = fopen(import_marks_file, "r");
1832 if (f)
1834 else if (import_marks_file_ignore_missing && errno == ENOENT)
1835 return; /* Marks file does not exist */
1836 else
1837 die_errno("cannot read '%s'", import_marks_file);
1838 while (fgets(line, sizeof(line), f)) {
1839 uintmax_t mark;
1840 char *end;
1841 unsigned char sha1[20];
1842 struct object_entry *e;
1844 end = strchr(line, '\n');
1845 if (line[0] != ':' || !end)
1846 die("corrupt mark line: %s", line);
1847 *end = 0;
1848 mark = strtoumax(line + 1, &end, 10);
1849 if (!mark || end == line + 1
1850 || *end != ' ' || get_sha1_hex(end + 1, sha1))
1851 die("corrupt mark line: %s", line);
1852 e = find_object(sha1);
1853 if (!e) {
1854 enum object_type type = sha1_object_info(sha1, NULL);
1855 if (type < 0)
1856 die("object not found: %s", sha1_to_hex(sha1));
1857 e = insert_object(sha1);
1858 e->type = type;
1859 e->pack_id = MAX_PACK_ID;
1860 e->idx.offset = 1; /* just not zero! */
1862 insert_mark(mark, e);
1864 fclose(f);
1868 static int read_next_command(void)
1870 static int stdin_eof = 0;
1872 if (stdin_eof) {
1873 unread_command_buf = 0;
1874 return EOF;
1877 for (;;) {
1878 const char *p;
1880 if (unread_command_buf) {
1881 unread_command_buf = 0;
1882 } else {
1883 struct recent_command *rc;
1885 strbuf_detach(&command_buf, NULL);
1886 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1887 if (stdin_eof)
1888 return EOF;
1890 if (!seen_data_command
1891 && !starts_with(command_buf.buf, "feature ")
1892 && !starts_with(command_buf.buf, "option ")) {
1893 parse_argv();
1896 rc = rc_free;
1897 if (rc)
1898 rc_free = rc->next;
1899 else {
1900 rc = cmd_hist.next;
1901 cmd_hist.next = rc->next;
1902 cmd_hist.next->prev = &cmd_hist;
1903 free(rc->buf);
1906 rc->buf = command_buf.buf;
1907 rc->prev = cmd_tail;
1908 rc->next = cmd_hist.prev;
1909 rc->prev->next = rc;
1910 cmd_tail = rc;
1912 if (skip_prefix(command_buf.buf, "get-mark ", &p)) {
1913 parse_get_mark(p);
1914 continue;
1916 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1917 parse_cat_blob(p);
1918 continue;
1920 if (command_buf.buf[0] == '#')
1921 continue;
1922 return 0;
1926 static void skip_optional_lf(void)
1928 int term_char = fgetc(stdin);
1929 if (term_char != '\n' && term_char != EOF)
1930 ungetc(term_char, stdin);
1933 static void parse_mark(void)
1935 const char *v;
1936 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1937 next_mark = strtoumax(v, NULL, 10);
1938 read_next_command();
1940 else
1941 next_mark = 0;
1944 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1946 const char *data;
1947 strbuf_reset(sb);
1949 if (!skip_prefix(command_buf.buf, "data ", &data))
1950 die("Expected 'data n' command, found: %s", command_buf.buf);
1952 if (skip_prefix(data, "<<", &data)) {
1953 char *term = xstrdup(data);
1954 size_t term_len = command_buf.len - (data - command_buf.buf);
1956 strbuf_detach(&command_buf, NULL);
1957 for (;;) {
1958 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1959 die("EOF in data (terminator '%s' not found)", term);
1960 if (term_len == command_buf.len
1961 && !strcmp(term, command_buf.buf))
1962 break;
1963 strbuf_addbuf(sb, &command_buf);
1964 strbuf_addch(sb, '\n');
1966 free(term);
1968 else {
1969 uintmax_t len = strtoumax(data, NULL, 10);
1970 size_t n = 0, length = (size_t)len;
1972 if (limit && limit < len) {
1973 *len_res = len;
1974 return 0;
1976 if (length < len)
1977 die("data is too large to use in this context");
1979 while (n < length) {
1980 size_t s = strbuf_fread(sb, length - n, stdin);
1981 if (!s && feof(stdin))
1982 die("EOF in data (%lu bytes remaining)",
1983 (unsigned long)(length - n));
1984 n += s;
1988 skip_optional_lf();
1989 return 1;
1992 static int validate_raw_date(const char *src, struct strbuf *result)
1994 const char *orig_src = src;
1995 char *endp;
1996 unsigned long num;
1998 errno = 0;
2000 num = strtoul(src, &endp, 10);
2001 /* NEEDSWORK: perhaps check for reasonable values? */
2002 if (errno || endp == src || *endp != ' ')
2003 return -1;
2005 src = endp + 1;
2006 if (*src != '-' && *src != '+')
2007 return -1;
2009 num = strtoul(src + 1, &endp, 10);
2010 if (errno || endp == src + 1 || *endp || 1400 < num)
2011 return -1;
2013 strbuf_addstr(result, orig_src);
2014 return 0;
2017 static char *parse_ident(const char *buf)
2019 const char *ltgt;
2020 size_t name_len;
2021 struct strbuf ident = STRBUF_INIT;
2023 /* ensure there is a space delimiter even if there is no name */
2024 if (*buf == '<')
2025 --buf;
2027 ltgt = buf + strcspn(buf, "<>");
2028 if (*ltgt != '<')
2029 die("Missing < in ident string: %s", buf);
2030 if (ltgt != buf && ltgt[-1] != ' ')
2031 die("Missing space before < in ident string: %s", buf);
2032 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2033 if (*ltgt != '>')
2034 die("Missing > in ident string: %s", buf);
2035 ltgt++;
2036 if (*ltgt != ' ')
2037 die("Missing space after > in ident string: %s", buf);
2038 ltgt++;
2039 name_len = ltgt - buf;
2040 strbuf_add(&ident, buf, name_len);
2042 switch (whenspec) {
2043 case WHENSPEC_RAW:
2044 if (validate_raw_date(ltgt, &ident) < 0)
2045 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2046 break;
2047 case WHENSPEC_RFC2822:
2048 if (parse_date(ltgt, &ident) < 0)
2049 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2050 break;
2051 case WHENSPEC_NOW:
2052 if (strcmp("now", ltgt))
2053 die("Date in ident must be 'now': %s", buf);
2054 datestamp(&ident);
2055 break;
2058 return strbuf_detach(&ident, NULL);
2061 static void parse_and_store_blob(
2062 struct last_object *last,
2063 unsigned char *sha1out,
2064 uintmax_t mark)
2066 static struct strbuf buf = STRBUF_INIT;
2067 uintmax_t len;
2069 if (parse_data(&buf, big_file_threshold, &len))
2070 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2071 else {
2072 if (last) {
2073 strbuf_release(&last->data);
2074 last->offset = 0;
2075 last->depth = 0;
2077 stream_blob(len, sha1out, mark);
2078 skip_optional_lf();
2082 static void parse_new_blob(void)
2084 read_next_command();
2085 parse_mark();
2086 parse_and_store_blob(&last_blob, NULL, next_mark);
2089 static void unload_one_branch(void)
2091 while (cur_active_branches
2092 && cur_active_branches >= max_active_branches) {
2093 uintmax_t min_commit = ULONG_MAX;
2094 struct branch *e, *l = NULL, *p = NULL;
2096 for (e = active_branches; e; e = e->active_next_branch) {
2097 if (e->last_commit < min_commit) {
2098 p = l;
2099 min_commit = e->last_commit;
2101 l = e;
2104 if (p) {
2105 e = p->active_next_branch;
2106 p->active_next_branch = e->active_next_branch;
2107 } else {
2108 e = active_branches;
2109 active_branches = e->active_next_branch;
2111 e->active = 0;
2112 e->active_next_branch = NULL;
2113 if (e->branch_tree.tree) {
2114 release_tree_content_recursive(e->branch_tree.tree);
2115 e->branch_tree.tree = NULL;
2117 cur_active_branches--;
2121 static void load_branch(struct branch *b)
2123 load_tree(&b->branch_tree);
2124 if (!b->active) {
2125 b->active = 1;
2126 b->active_next_branch = active_branches;
2127 active_branches = b;
2128 cur_active_branches++;
2129 branch_load_count++;
2133 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2135 unsigned char fanout = 0;
2136 while ((num_notes >>= 8))
2137 fanout++;
2138 return fanout;
2141 static void construct_path_with_fanout(const char *hex_sha1,
2142 unsigned char fanout, char *path)
2144 unsigned int i = 0, j = 0;
2145 if (fanout >= 20)
2146 die("Too large fanout (%u)", fanout);
2147 while (fanout) {
2148 path[i++] = hex_sha1[j++];
2149 path[i++] = hex_sha1[j++];
2150 path[i++] = '/';
2151 fanout--;
2153 memcpy(path + i, hex_sha1 + j, 40 - j);
2154 path[i + 40 - j] = '\0';
2157 static uintmax_t do_change_note_fanout(
2158 struct tree_entry *orig_root, struct tree_entry *root,
2159 char *hex_sha1, unsigned int hex_sha1_len,
2160 char *fullpath, unsigned int fullpath_len,
2161 unsigned char fanout)
2163 struct tree_content *t = root->tree;
2164 struct tree_entry *e, leaf;
2165 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2166 uintmax_t num_notes = 0;
2167 unsigned char sha1[20];
2168 char realpath[60];
2170 for (i = 0; t && i < t->entry_count; i++) {
2171 e = t->entries[i];
2172 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2173 tmp_fullpath_len = fullpath_len;
2176 * We're interested in EITHER existing note entries (entries
2177 * with exactly 40 hex chars in path, not including directory
2178 * separators), OR directory entries that may contain note
2179 * entries (with < 40 hex chars in path).
2180 * Also, each path component in a note entry must be a multiple
2181 * of 2 chars.
2183 if (!e->versions[1].mode ||
2184 tmp_hex_sha1_len > 40 ||
2185 e->name->str_len % 2)
2186 continue;
2188 /* This _may_ be a note entry, or a subdir containing notes */
2189 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2190 e->name->str_len);
2191 if (tmp_fullpath_len)
2192 fullpath[tmp_fullpath_len++] = '/';
2193 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2194 e->name->str_len);
2195 tmp_fullpath_len += e->name->str_len;
2196 fullpath[tmp_fullpath_len] = '\0';
2198 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2199 /* This is a note entry */
2200 if (fanout == 0xff) {
2201 /* Counting mode, no rename */
2202 num_notes++;
2203 continue;
2205 construct_path_with_fanout(hex_sha1, fanout, realpath);
2206 if (!strcmp(fullpath, realpath)) {
2207 /* Note entry is in correct location */
2208 num_notes++;
2209 continue;
2212 /* Rename fullpath to realpath */
2213 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2214 die("Failed to remove path %s", fullpath);
2215 tree_content_set(orig_root, realpath,
2216 leaf.versions[1].sha1,
2217 leaf.versions[1].mode,
2218 leaf.tree);
2219 } else if (S_ISDIR(e->versions[1].mode)) {
2220 /* This is a subdir that may contain note entries */
2221 if (!e->tree)
2222 load_tree(e);
2223 num_notes += do_change_note_fanout(orig_root, e,
2224 hex_sha1, tmp_hex_sha1_len,
2225 fullpath, tmp_fullpath_len, fanout);
2228 /* The above may have reallocated the current tree_content */
2229 t = root->tree;
2231 return num_notes;
2234 static uintmax_t change_note_fanout(struct tree_entry *root,
2235 unsigned char fanout)
2237 char hex_sha1[40], path[60];
2238 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2242 * Given a pointer into a string, parse a mark reference:
2244 * idnum ::= ':' bigint;
2246 * Return the first character after the value in *endptr.
2248 * Complain if the following character is not what is expected,
2249 * either a space or end of the string.
2251 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2253 uintmax_t mark;
2255 assert(*p == ':');
2256 p++;
2257 mark = strtoumax(p, endptr, 10);
2258 if (*endptr == p)
2259 die("No value after ':' in mark: %s", command_buf.buf);
2260 return mark;
2264 * Parse the mark reference, and complain if this is not the end of
2265 * the string.
2267 static uintmax_t parse_mark_ref_eol(const char *p)
2269 char *end;
2270 uintmax_t mark;
2272 mark = parse_mark_ref(p, &end);
2273 if (*end != '\0')
2274 die("Garbage after mark: %s", command_buf.buf);
2275 return mark;
2279 * Parse the mark reference, demanding a trailing space. Return a
2280 * pointer to the space.
2282 static uintmax_t parse_mark_ref_space(const char **p)
2284 uintmax_t mark;
2285 char *end;
2287 mark = parse_mark_ref(*p, &end);
2288 if (*end++ != ' ')
2289 die("Missing space after mark: %s", command_buf.buf);
2290 *p = end;
2291 return mark;
2294 static void file_change_m(const char *p, struct branch *b)
2296 static struct strbuf uq = STRBUF_INIT;
2297 const char *endp;
2298 struct object_entry *oe;
2299 unsigned char sha1[20];
2300 uint16_t mode, inline_data = 0;
2302 p = get_mode(p, &mode);
2303 if (!p)
2304 die("Corrupt mode: %s", command_buf.buf);
2305 switch (mode) {
2306 case 0644:
2307 case 0755:
2308 mode |= S_IFREG;
2309 case S_IFREG | 0644:
2310 case S_IFREG | 0755:
2311 case S_IFLNK:
2312 case S_IFDIR:
2313 case S_IFGITLINK:
2314 /* ok */
2315 break;
2316 default:
2317 die("Corrupt mode: %s", command_buf.buf);
2320 if (*p == ':') {
2321 oe = find_mark(parse_mark_ref_space(&p));
2322 hashcpy(sha1, oe->idx.sha1);
2323 } else if (skip_prefix(p, "inline ", &p)) {
2324 inline_data = 1;
2325 oe = NULL; /* not used with inline_data, but makes gcc happy */
2326 } else {
2327 if (get_sha1_hex(p, sha1))
2328 die("Invalid dataref: %s", command_buf.buf);
2329 oe = find_object(sha1);
2330 p += 40;
2331 if (*p++ != ' ')
2332 die("Missing space after SHA1: %s", command_buf.buf);
2335 strbuf_reset(&uq);
2336 if (!unquote_c_style(&uq, p, &endp)) {
2337 if (*endp)
2338 die("Garbage after path in: %s", command_buf.buf);
2339 p = uq.buf;
2342 /* Git does not track empty, non-toplevel directories. */
2343 if (S_ISDIR(mode) && !hashcmp(sha1, EMPTY_TREE_SHA1_BIN) && *p) {
2344 tree_content_remove(&b->branch_tree, p, NULL, 0);
2345 return;
2348 if (S_ISGITLINK(mode)) {
2349 if (inline_data)
2350 die("Git links cannot be specified 'inline': %s",
2351 command_buf.buf);
2352 else if (oe) {
2353 if (oe->type != OBJ_COMMIT)
2354 die("Not a commit (actually a %s): %s",
2355 typename(oe->type), command_buf.buf);
2358 * Accept the sha1 without checking; it expected to be in
2359 * another repository.
2361 } else if (inline_data) {
2362 if (S_ISDIR(mode))
2363 die("Directories cannot be specified 'inline': %s",
2364 command_buf.buf);
2365 if (p != uq.buf) {
2366 strbuf_addstr(&uq, p);
2367 p = uq.buf;
2369 read_next_command();
2370 parse_and_store_blob(&last_blob, sha1, 0);
2371 } else {
2372 enum object_type expected = S_ISDIR(mode) ?
2373 OBJ_TREE: OBJ_BLOB;
2374 enum object_type type = oe ? oe->type :
2375 sha1_object_info(sha1, NULL);
2376 if (type < 0)
2377 die("%s not found: %s",
2378 S_ISDIR(mode) ? "Tree" : "Blob",
2379 command_buf.buf);
2380 if (type != expected)
2381 die("Not a %s (actually a %s): %s",
2382 typename(expected), typename(type),
2383 command_buf.buf);
2386 if (!*p) {
2387 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2388 return;
2390 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2393 static void file_change_d(const char *p, struct branch *b)
2395 static struct strbuf uq = STRBUF_INIT;
2396 const char *endp;
2398 strbuf_reset(&uq);
2399 if (!unquote_c_style(&uq, p, &endp)) {
2400 if (*endp)
2401 die("Garbage after path in: %s", command_buf.buf);
2402 p = uq.buf;
2404 tree_content_remove(&b->branch_tree, p, NULL, 1);
2407 static void file_change_cr(const char *s, struct branch *b, int rename)
2409 const char *d;
2410 static struct strbuf s_uq = STRBUF_INIT;
2411 static struct strbuf d_uq = STRBUF_INIT;
2412 const char *endp;
2413 struct tree_entry leaf;
2415 strbuf_reset(&s_uq);
2416 if (!unquote_c_style(&s_uq, s, &endp)) {
2417 if (*endp != ' ')
2418 die("Missing space after source: %s", command_buf.buf);
2419 } else {
2420 endp = strchr(s, ' ');
2421 if (!endp)
2422 die("Missing space after source: %s", command_buf.buf);
2423 strbuf_add(&s_uq, s, endp - s);
2425 s = s_uq.buf;
2427 endp++;
2428 if (!*endp)
2429 die("Missing dest: %s", command_buf.buf);
2431 d = endp;
2432 strbuf_reset(&d_uq);
2433 if (!unquote_c_style(&d_uq, d, &endp)) {
2434 if (*endp)
2435 die("Garbage after dest in: %s", command_buf.buf);
2436 d = d_uq.buf;
2439 memset(&leaf, 0, sizeof(leaf));
2440 if (rename)
2441 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2442 else
2443 tree_content_get(&b->branch_tree, s, &leaf, 1);
2444 if (!leaf.versions[1].mode)
2445 die("Path %s not in branch", s);
2446 if (!*d) { /* C "path/to/subdir" "" */
2447 tree_content_replace(&b->branch_tree,
2448 leaf.versions[1].sha1,
2449 leaf.versions[1].mode,
2450 leaf.tree);
2451 return;
2453 tree_content_set(&b->branch_tree, d,
2454 leaf.versions[1].sha1,
2455 leaf.versions[1].mode,
2456 leaf.tree);
2459 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2461 static struct strbuf uq = STRBUF_INIT;
2462 struct object_entry *oe;
2463 struct branch *s;
2464 unsigned char sha1[20], commit_sha1[20];
2465 char path[60];
2466 uint16_t inline_data = 0;
2467 unsigned char new_fanout;
2470 * When loading a branch, we don't traverse its tree to count the real
2471 * number of notes (too expensive to do this for all non-note refs).
2472 * This means that recently loaded notes refs might incorrectly have
2473 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2475 * Fix this by traversing the tree and counting the number of notes
2476 * when b->num_notes == 0. If the notes tree is truly empty, the
2477 * calculation should not take long.
2479 if (b->num_notes == 0 && *old_fanout == 0) {
2480 /* Invoke change_note_fanout() in "counting mode". */
2481 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2482 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2485 /* Now parse the notemodify command. */
2486 /* <dataref> or 'inline' */
2487 if (*p == ':') {
2488 oe = find_mark(parse_mark_ref_space(&p));
2489 hashcpy(sha1, oe->idx.sha1);
2490 } else if (skip_prefix(p, "inline ", &p)) {
2491 inline_data = 1;
2492 oe = NULL; /* not used with inline_data, but makes gcc happy */
2493 } else {
2494 if (get_sha1_hex(p, sha1))
2495 die("Invalid dataref: %s", command_buf.buf);
2496 oe = find_object(sha1);
2497 p += 40;
2498 if (*p++ != ' ')
2499 die("Missing space after SHA1: %s", command_buf.buf);
2502 /* <commit-ish> */
2503 s = lookup_branch(p);
2504 if (s) {
2505 if (is_null_sha1(s->sha1))
2506 die("Can't add a note on empty branch.");
2507 hashcpy(commit_sha1, s->sha1);
2508 } else if (*p == ':') {
2509 uintmax_t commit_mark = parse_mark_ref_eol(p);
2510 struct object_entry *commit_oe = find_mark(commit_mark);
2511 if (commit_oe->type != OBJ_COMMIT)
2512 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2513 hashcpy(commit_sha1, commit_oe->idx.sha1);
2514 } else if (!get_sha1(p, commit_sha1)) {
2515 unsigned long size;
2516 char *buf = read_object_with_reference(commit_sha1,
2517 commit_type, &size, commit_sha1);
2518 if (!buf || size < 46)
2519 die("Not a valid commit: %s", p);
2520 free(buf);
2521 } else
2522 die("Invalid ref name or SHA1 expression: %s", p);
2524 if (inline_data) {
2525 if (p != uq.buf) {
2526 strbuf_addstr(&uq, p);
2527 p = uq.buf;
2529 read_next_command();
2530 parse_and_store_blob(&last_blob, sha1, 0);
2531 } else if (oe) {
2532 if (oe->type != OBJ_BLOB)
2533 die("Not a blob (actually a %s): %s",
2534 typename(oe->type), command_buf.buf);
2535 } else if (!is_null_sha1(sha1)) {
2536 enum object_type type = sha1_object_info(sha1, NULL);
2537 if (type < 0)
2538 die("Blob not found: %s", command_buf.buf);
2539 if (type != OBJ_BLOB)
2540 die("Not a blob (actually a %s): %s",
2541 typename(type), command_buf.buf);
2544 construct_path_with_fanout(sha1_to_hex(commit_sha1), *old_fanout, path);
2545 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2546 b->num_notes--;
2548 if (is_null_sha1(sha1))
2549 return; /* nothing to insert */
2551 b->num_notes++;
2552 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2553 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2554 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2557 static void file_change_deleteall(struct branch *b)
2559 release_tree_content_recursive(b->branch_tree.tree);
2560 hashclr(b->branch_tree.versions[0].sha1);
2561 hashclr(b->branch_tree.versions[1].sha1);
2562 load_tree(&b->branch_tree);
2563 b->num_notes = 0;
2566 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2568 if (!buf || size < 46)
2569 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2570 if (memcmp("tree ", buf, 5)
2571 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2572 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2573 hashcpy(b->branch_tree.versions[0].sha1,
2574 b->branch_tree.versions[1].sha1);
2577 static void parse_from_existing(struct branch *b)
2579 if (is_null_sha1(b->sha1)) {
2580 hashclr(b->branch_tree.versions[0].sha1);
2581 hashclr(b->branch_tree.versions[1].sha1);
2582 } else {
2583 unsigned long size;
2584 char *buf;
2586 buf = read_object_with_reference(b->sha1,
2587 commit_type, &size, b->sha1);
2588 parse_from_commit(b, buf, size);
2589 free(buf);
2593 static int parse_from(struct branch *b)
2595 const char *from;
2596 struct branch *s;
2597 unsigned char sha1[20];
2599 if (!skip_prefix(command_buf.buf, "from ", &from))
2600 return 0;
2602 hashcpy(sha1, b->branch_tree.versions[1].sha1);
2604 s = lookup_branch(from);
2605 if (b == s)
2606 die("Can't create a branch from itself: %s", b->name);
2607 else if (s) {
2608 unsigned char *t = s->branch_tree.versions[1].sha1;
2609 hashcpy(b->sha1, s->sha1);
2610 hashcpy(b->branch_tree.versions[0].sha1, t);
2611 hashcpy(b->branch_tree.versions[1].sha1, t);
2612 } else if (*from == ':') {
2613 uintmax_t idnum = parse_mark_ref_eol(from);
2614 struct object_entry *oe = find_mark(idnum);
2615 if (oe->type != OBJ_COMMIT)
2616 die("Mark :%" PRIuMAX " not a commit", idnum);
2617 if (hashcmp(b->sha1, oe->idx.sha1)) {
2618 hashcpy(b->sha1, oe->idx.sha1);
2619 if (oe->pack_id != MAX_PACK_ID) {
2620 unsigned long size;
2621 char *buf = gfi_unpack_entry(oe, &size);
2622 parse_from_commit(b, buf, size);
2623 free(buf);
2624 } else
2625 parse_from_existing(b);
2627 } else if (!get_sha1(from, b->sha1)) {
2628 parse_from_existing(b);
2629 if (is_null_sha1(b->sha1))
2630 b->delete = 1;
2632 else
2633 die("Invalid ref name or SHA1 expression: %s", from);
2635 if (b->branch_tree.tree && hashcmp(sha1, b->branch_tree.versions[1].sha1)) {
2636 release_tree_content_recursive(b->branch_tree.tree);
2637 b->branch_tree.tree = NULL;
2640 read_next_command();
2641 return 1;
2644 static struct hash_list *parse_merge(unsigned int *count)
2646 struct hash_list *list = NULL, **tail = &list, *n;
2647 const char *from;
2648 struct branch *s;
2650 *count = 0;
2651 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2652 n = xmalloc(sizeof(*n));
2653 s = lookup_branch(from);
2654 if (s)
2655 hashcpy(n->sha1, s->sha1);
2656 else if (*from == ':') {
2657 uintmax_t idnum = parse_mark_ref_eol(from);
2658 struct object_entry *oe = find_mark(idnum);
2659 if (oe->type != OBJ_COMMIT)
2660 die("Mark :%" PRIuMAX " not a commit", idnum);
2661 hashcpy(n->sha1, oe->idx.sha1);
2662 } else if (!get_sha1(from, n->sha1)) {
2663 unsigned long size;
2664 char *buf = read_object_with_reference(n->sha1,
2665 commit_type, &size, n->sha1);
2666 if (!buf || size < 46)
2667 die("Not a valid commit: %s", from);
2668 free(buf);
2669 } else
2670 die("Invalid ref name or SHA1 expression: %s", from);
2672 n->next = NULL;
2673 *tail = n;
2674 tail = &n->next;
2676 (*count)++;
2677 read_next_command();
2679 return list;
2682 static void parse_new_commit(const char *arg)
2684 static struct strbuf msg = STRBUF_INIT;
2685 struct branch *b;
2686 char *author = NULL;
2687 char *committer = NULL;
2688 struct hash_list *merge_list = NULL;
2689 unsigned int merge_count;
2690 unsigned char prev_fanout, new_fanout;
2691 const char *v;
2693 b = lookup_branch(arg);
2694 if (!b)
2695 b = new_branch(arg);
2697 read_next_command();
2698 parse_mark();
2699 if (skip_prefix(command_buf.buf, "author ", &v)) {
2700 author = parse_ident(v);
2701 read_next_command();
2703 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2704 committer = parse_ident(v);
2705 read_next_command();
2707 if (!committer)
2708 die("Expected committer but didn't get one");
2709 parse_data(&msg, 0, NULL);
2710 read_next_command();
2711 parse_from(b);
2712 merge_list = parse_merge(&merge_count);
2714 /* ensure the branch is active/loaded */
2715 if (!b->branch_tree.tree || !max_active_branches) {
2716 unload_one_branch();
2717 load_branch(b);
2720 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2722 /* file_change* */
2723 while (command_buf.len > 0) {
2724 if (skip_prefix(command_buf.buf, "M ", &v))
2725 file_change_m(v, b);
2726 else if (skip_prefix(command_buf.buf, "D ", &v))
2727 file_change_d(v, b);
2728 else if (skip_prefix(command_buf.buf, "R ", &v))
2729 file_change_cr(v, b, 1);
2730 else if (skip_prefix(command_buf.buf, "C ", &v))
2731 file_change_cr(v, b, 0);
2732 else if (skip_prefix(command_buf.buf, "N ", &v))
2733 note_change_n(v, b, &prev_fanout);
2734 else if (!strcmp("deleteall", command_buf.buf))
2735 file_change_deleteall(b);
2736 else if (skip_prefix(command_buf.buf, "ls ", &v))
2737 parse_ls(v, b);
2738 else {
2739 unread_command_buf = 1;
2740 break;
2742 if (read_next_command() == EOF)
2743 break;
2746 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2747 if (new_fanout != prev_fanout)
2748 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2750 /* build the tree and the commit */
2751 store_tree(&b->branch_tree);
2752 hashcpy(b->branch_tree.versions[0].sha1,
2753 b->branch_tree.versions[1].sha1);
2755 strbuf_reset(&new_data);
2756 strbuf_addf(&new_data, "tree %s\n",
2757 sha1_to_hex(b->branch_tree.versions[1].sha1));
2758 if (!is_null_sha1(b->sha1))
2759 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2760 while (merge_list) {
2761 struct hash_list *next = merge_list->next;
2762 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2763 free(merge_list);
2764 merge_list = next;
2766 strbuf_addf(&new_data,
2767 "author %s\n"
2768 "committer %s\n"
2769 "\n",
2770 author ? author : committer, committer);
2771 strbuf_addbuf(&new_data, &msg);
2772 free(author);
2773 free(committer);
2775 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2776 b->pack_id = pack_id;
2777 b->last_commit = object_count_by_type[OBJ_COMMIT];
2780 static void parse_new_tag(const char *arg)
2782 static struct strbuf msg = STRBUF_INIT;
2783 const char *from;
2784 char *tagger;
2785 struct branch *s;
2786 struct tag *t;
2787 uintmax_t from_mark = 0;
2788 unsigned char sha1[20];
2789 enum object_type type;
2790 const char *v;
2792 t = pool_alloc(sizeof(struct tag));
2793 memset(t, 0, sizeof(struct tag));
2794 t->name = pool_strdup(arg);
2795 if (last_tag)
2796 last_tag->next_tag = t;
2797 else
2798 first_tag = t;
2799 last_tag = t;
2800 read_next_command();
2802 /* from ... */
2803 if (!skip_prefix(command_buf.buf, "from ", &from))
2804 die("Expected from command, got %s", command_buf.buf);
2805 s = lookup_branch(from);
2806 if (s) {
2807 if (is_null_sha1(s->sha1))
2808 die("Can't tag an empty branch.");
2809 hashcpy(sha1, s->sha1);
2810 type = OBJ_COMMIT;
2811 } else if (*from == ':') {
2812 struct object_entry *oe;
2813 from_mark = parse_mark_ref_eol(from);
2814 oe = find_mark(from_mark);
2815 type = oe->type;
2816 hashcpy(sha1, oe->idx.sha1);
2817 } else if (!get_sha1(from, sha1)) {
2818 struct object_entry *oe = find_object(sha1);
2819 if (!oe) {
2820 type = sha1_object_info(sha1, NULL);
2821 if (type < 0)
2822 die("Not a valid object: %s", from);
2823 } else
2824 type = oe->type;
2825 } else
2826 die("Invalid ref name or SHA1 expression: %s", from);
2827 read_next_command();
2829 /* tagger ... */
2830 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2831 tagger = parse_ident(v);
2832 read_next_command();
2833 } else
2834 tagger = NULL;
2836 /* tag payload/message */
2837 parse_data(&msg, 0, NULL);
2839 /* build the tag object */
2840 strbuf_reset(&new_data);
2842 strbuf_addf(&new_data,
2843 "object %s\n"
2844 "type %s\n"
2845 "tag %s\n",
2846 sha1_to_hex(sha1), typename(type), t->name);
2847 if (tagger)
2848 strbuf_addf(&new_data,
2849 "tagger %s\n", tagger);
2850 strbuf_addch(&new_data, '\n');
2851 strbuf_addbuf(&new_data, &msg);
2852 free(tagger);
2854 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2855 t->pack_id = MAX_PACK_ID;
2856 else
2857 t->pack_id = pack_id;
2860 static void parse_reset_branch(const char *arg)
2862 struct branch *b;
2864 b = lookup_branch(arg);
2865 if (b) {
2866 hashclr(b->sha1);
2867 hashclr(b->branch_tree.versions[0].sha1);
2868 hashclr(b->branch_tree.versions[1].sha1);
2869 if (b->branch_tree.tree) {
2870 release_tree_content_recursive(b->branch_tree.tree);
2871 b->branch_tree.tree = NULL;
2874 else
2875 b = new_branch(arg);
2876 read_next_command();
2877 parse_from(b);
2878 if (command_buf.len > 0)
2879 unread_command_buf = 1;
2882 static void cat_blob_write(const char *buf, unsigned long size)
2884 if (write_in_full(cat_blob_fd, buf, size) != size)
2885 die_errno("Write to frontend failed");
2888 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2890 struct strbuf line = STRBUF_INIT;
2891 unsigned long size;
2892 enum object_type type = 0;
2893 char *buf;
2895 if (!oe || oe->pack_id == MAX_PACK_ID) {
2896 buf = read_sha1_file(sha1, &type, &size);
2897 } else {
2898 type = oe->type;
2899 buf = gfi_unpack_entry(oe, &size);
2903 * Output based on batch_one_object() from cat-file.c.
2905 if (type <= 0) {
2906 strbuf_reset(&line);
2907 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2908 cat_blob_write(line.buf, line.len);
2909 strbuf_release(&line);
2910 free(buf);
2911 return;
2913 if (!buf)
2914 die("Can't read object %s", sha1_to_hex(sha1));
2915 if (type != OBJ_BLOB)
2916 die("Object %s is a %s but a blob was expected.",
2917 sha1_to_hex(sha1), typename(type));
2918 strbuf_reset(&line);
2919 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2920 typename(type), size);
2921 cat_blob_write(line.buf, line.len);
2922 strbuf_release(&line);
2923 cat_blob_write(buf, size);
2924 cat_blob_write("\n", 1);
2925 if (oe && oe->pack_id == pack_id) {
2926 last_blob.offset = oe->idx.offset;
2927 strbuf_attach(&last_blob.data, buf, size, size);
2928 last_blob.depth = oe->depth;
2929 } else
2930 free(buf);
2933 static void parse_get_mark(const char *p)
2935 struct object_entry *oe = oe;
2936 char output[42];
2938 /* get-mark SP <object> LF */
2939 if (*p != ':')
2940 die("Not a mark: %s", p);
2942 oe = find_mark(parse_mark_ref_eol(p));
2943 if (!oe)
2944 die("Unknown mark: %s", command_buf.buf);
2946 snprintf(output, sizeof(output), "%s\n", sha1_to_hex(oe->idx.sha1));
2947 cat_blob_write(output, 41);
2950 static void parse_cat_blob(const char *p)
2952 struct object_entry *oe = oe;
2953 unsigned char sha1[20];
2955 /* cat-blob SP <object> LF */
2956 if (*p == ':') {
2957 oe = find_mark(parse_mark_ref_eol(p));
2958 if (!oe)
2959 die("Unknown mark: %s", command_buf.buf);
2960 hashcpy(sha1, oe->idx.sha1);
2961 } else {
2962 if (get_sha1_hex(p, sha1))
2963 die("Invalid dataref: %s", command_buf.buf);
2964 if (p[40])
2965 die("Garbage after SHA1: %s", command_buf.buf);
2966 oe = find_object(sha1);
2969 cat_blob(oe, sha1);
2972 static struct object_entry *dereference(struct object_entry *oe,
2973 unsigned char sha1[20])
2975 unsigned long size;
2976 char *buf = NULL;
2977 if (!oe) {
2978 enum object_type type = sha1_object_info(sha1, NULL);
2979 if (type < 0)
2980 die("object not found: %s", sha1_to_hex(sha1));
2981 /* cache it! */
2982 oe = insert_object(sha1);
2983 oe->type = type;
2984 oe->pack_id = MAX_PACK_ID;
2985 oe->idx.offset = 1;
2987 switch (oe->type) {
2988 case OBJ_TREE: /* easy case. */
2989 return oe;
2990 case OBJ_COMMIT:
2991 case OBJ_TAG:
2992 break;
2993 default:
2994 die("Not a tree-ish: %s", command_buf.buf);
2997 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2998 buf = gfi_unpack_entry(oe, &size);
2999 } else {
3000 enum object_type unused;
3001 buf = read_sha1_file(sha1, &unused, &size);
3003 if (!buf)
3004 die("Can't load object %s", sha1_to_hex(sha1));
3006 /* Peel one layer. */
3007 switch (oe->type) {
3008 case OBJ_TAG:
3009 if (size < 40 + strlen("object ") ||
3010 get_sha1_hex(buf + strlen("object "), sha1))
3011 die("Invalid SHA1 in tag: %s", command_buf.buf);
3012 break;
3013 case OBJ_COMMIT:
3014 if (size < 40 + strlen("tree ") ||
3015 get_sha1_hex(buf + strlen("tree "), sha1))
3016 die("Invalid SHA1 in commit: %s", command_buf.buf);
3019 free(buf);
3020 return find_object(sha1);
3023 static struct object_entry *parse_treeish_dataref(const char **p)
3025 unsigned char sha1[20];
3026 struct object_entry *e;
3028 if (**p == ':') { /* <mark> */
3029 e = find_mark(parse_mark_ref_space(p));
3030 if (!e)
3031 die("Unknown mark: %s", command_buf.buf);
3032 hashcpy(sha1, e->idx.sha1);
3033 } else { /* <sha1> */
3034 if (get_sha1_hex(*p, sha1))
3035 die("Invalid dataref: %s", command_buf.buf);
3036 e = find_object(sha1);
3037 *p += 40;
3038 if (*(*p)++ != ' ')
3039 die("Missing space after tree-ish: %s", command_buf.buf);
3042 while (!e || e->type != OBJ_TREE)
3043 e = dereference(e, sha1);
3044 return e;
3047 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3049 static struct strbuf line = STRBUF_INIT;
3051 /* See show_tree(). */
3052 const char *type =
3053 S_ISGITLINK(mode) ? commit_type :
3054 S_ISDIR(mode) ? tree_type :
3055 blob_type;
3057 if (!mode) {
3058 /* missing SP path LF */
3059 strbuf_reset(&line);
3060 strbuf_addstr(&line, "missing ");
3061 quote_c_style(path, &line, NULL, 0);
3062 strbuf_addch(&line, '\n');
3063 } else {
3064 /* mode SP type SP object_name TAB path LF */
3065 strbuf_reset(&line);
3066 strbuf_addf(&line, "%06o %s %s\t",
3067 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3068 quote_c_style(path, &line, NULL, 0);
3069 strbuf_addch(&line, '\n');
3071 cat_blob_write(line.buf, line.len);
3074 static void parse_ls(const char *p, struct branch *b)
3076 struct tree_entry *root = NULL;
3077 struct tree_entry leaf = {NULL};
3079 /* ls SP (<tree-ish> SP)? <path> */
3080 if (*p == '"') {
3081 if (!b)
3082 die("Not in a commit: %s", command_buf.buf);
3083 root = &b->branch_tree;
3084 } else {
3085 struct object_entry *e = parse_treeish_dataref(&p);
3086 root = new_tree_entry();
3087 hashcpy(root->versions[1].sha1, e->idx.sha1);
3088 if (!is_null_sha1(root->versions[1].sha1))
3089 root->versions[1].mode = S_IFDIR;
3090 load_tree(root);
3092 if (*p == '"') {
3093 static struct strbuf uq = STRBUF_INIT;
3094 const char *endp;
3095 strbuf_reset(&uq);
3096 if (unquote_c_style(&uq, p, &endp))
3097 die("Invalid path: %s", command_buf.buf);
3098 if (*endp)
3099 die("Garbage after path in: %s", command_buf.buf);
3100 p = uq.buf;
3102 tree_content_get(root, p, &leaf, 1);
3104 * A directory in preparation would have a sha1 of zero
3105 * until it is saved. Save, for simplicity.
3107 if (S_ISDIR(leaf.versions[1].mode))
3108 store_tree(&leaf);
3110 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3111 if (leaf.tree)
3112 release_tree_content_recursive(leaf.tree);
3113 if (!b || root != &b->branch_tree)
3114 release_tree_entry(root);
3117 static void checkpoint(void)
3119 checkpoint_requested = 0;
3120 if (object_count) {
3121 cycle_packfile();
3122 dump_branches();
3123 dump_tags();
3124 dump_marks();
3128 static void parse_checkpoint(void)
3130 checkpoint_requested = 1;
3131 skip_optional_lf();
3134 static void parse_progress(void)
3136 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3137 fputc('\n', stdout);
3138 fflush(stdout);
3139 skip_optional_lf();
3142 static char* make_fast_import_path(const char *path)
3144 if (!relative_marks_paths || is_absolute_path(path))
3145 return xstrdup(path);
3146 return xstrdup(git_path("info/fast-import/%s", path));
3149 static void option_import_marks(const char *marks,
3150 int from_stream, int ignore_missing)
3152 if (import_marks_file) {
3153 if (from_stream)
3154 die("Only one import-marks command allowed per stream");
3156 /* read previous mark file */
3157 if(!import_marks_file_from_stream)
3158 read_marks();
3161 import_marks_file = make_fast_import_path(marks);
3162 safe_create_leading_directories_const(import_marks_file);
3163 import_marks_file_from_stream = from_stream;
3164 import_marks_file_ignore_missing = ignore_missing;
3167 static void option_date_format(const char *fmt)
3169 if (!strcmp(fmt, "raw"))
3170 whenspec = WHENSPEC_RAW;
3171 else if (!strcmp(fmt, "rfc2822"))
3172 whenspec = WHENSPEC_RFC2822;
3173 else if (!strcmp(fmt, "now"))
3174 whenspec = WHENSPEC_NOW;
3175 else
3176 die("unknown --date-format argument %s", fmt);
3179 static unsigned long ulong_arg(const char *option, const char *arg)
3181 char *endptr;
3182 unsigned long rv = strtoul(arg, &endptr, 0);
3183 if (strchr(arg, '-') || endptr == arg || *endptr)
3184 die("%s: argument must be a non-negative integer", option);
3185 return rv;
3188 static void option_depth(const char *depth)
3190 max_depth = ulong_arg("--depth", depth);
3191 if (max_depth > MAX_DEPTH)
3192 die("--depth cannot exceed %u", MAX_DEPTH);
3195 static void option_active_branches(const char *branches)
3197 max_active_branches = ulong_arg("--active-branches", branches);
3200 static void option_export_marks(const char *marks)
3202 export_marks_file = make_fast_import_path(marks);
3203 safe_create_leading_directories_const(export_marks_file);
3206 static void option_cat_blob_fd(const char *fd)
3208 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3209 if (n > (unsigned long) INT_MAX)
3210 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3211 cat_blob_fd = (int) n;
3214 static void option_export_pack_edges(const char *edges)
3216 if (pack_edges)
3217 fclose(pack_edges);
3218 pack_edges = fopen(edges, "a");
3219 if (!pack_edges)
3220 die_errno("Cannot open '%s'", edges);
3223 static int parse_one_option(const char *option)
3225 if (skip_prefix(option, "max-pack-size=", &option)) {
3226 unsigned long v;
3227 if (!git_parse_ulong(option, &v))
3228 return 0;
3229 if (v < 8192) {
3230 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3231 v *= 1024 * 1024;
3232 } else if (v < 1024 * 1024) {
3233 warning("minimum max-pack-size is 1 MiB");
3234 v = 1024 * 1024;
3236 max_packsize = v;
3237 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3238 unsigned long v;
3239 if (!git_parse_ulong(option, &v))
3240 return 0;
3241 big_file_threshold = v;
3242 } else if (skip_prefix(option, "depth=", &option)) {
3243 option_depth(option);
3244 } else if (skip_prefix(option, "active-branches=", &option)) {
3245 option_active_branches(option);
3246 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3247 option_export_pack_edges(option);
3248 } else if (starts_with(option, "quiet")) {
3249 show_stats = 0;
3250 } else if (starts_with(option, "stats")) {
3251 show_stats = 1;
3252 } else {
3253 return 0;
3256 return 1;
3259 static int parse_one_feature(const char *feature, int from_stream)
3261 const char *arg;
3263 if (skip_prefix(feature, "date-format=", &arg)) {
3264 option_date_format(arg);
3265 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3266 option_import_marks(arg, from_stream, 0);
3267 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3268 option_import_marks(arg, from_stream, 1);
3269 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3270 option_export_marks(arg);
3271 } else if (!strcmp(feature, "get-mark")) {
3272 ; /* Don't die - this feature is supported */
3273 } else if (!strcmp(feature, "cat-blob")) {
3274 ; /* Don't die - this feature is supported */
3275 } else if (!strcmp(feature, "relative-marks")) {
3276 relative_marks_paths = 1;
3277 } else if (!strcmp(feature, "no-relative-marks")) {
3278 relative_marks_paths = 0;
3279 } else if (!strcmp(feature, "done")) {
3280 require_explicit_termination = 1;
3281 } else if (!strcmp(feature, "force")) {
3282 force_update = 1;
3283 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3284 ; /* do nothing; we have the feature */
3285 } else {
3286 return 0;
3289 return 1;
3292 static void parse_feature(const char *feature)
3294 if (seen_data_command)
3295 die("Got feature command '%s' after data command", feature);
3297 if (parse_one_feature(feature, 1))
3298 return;
3300 die("This version of fast-import does not support feature %s.", feature);
3303 static void parse_option(const char *option)
3305 if (seen_data_command)
3306 die("Got option command '%s' after data command", option);
3308 if (parse_one_option(option))
3309 return;
3311 die("This version of fast-import does not support option: %s", option);
3314 static void git_pack_config(void)
3316 int indexversion_value;
3317 unsigned long packsizelimit_value;
3319 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3320 if (max_depth > MAX_DEPTH)
3321 max_depth = MAX_DEPTH;
3323 if (!git_config_get_int("pack.compression", &pack_compression_level)) {
3324 if (pack_compression_level == -1)
3325 pack_compression_level = Z_DEFAULT_COMPRESSION;
3326 else if (pack_compression_level < 0 ||
3327 pack_compression_level > Z_BEST_COMPRESSION)
3328 git_die_config("pack.compression",
3329 "bad pack compression level %d", pack_compression_level);
3330 pack_compression_seen = 1;
3332 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3333 pack_idx_opts.version = indexversion_value;
3334 if (pack_idx_opts.version > 2)
3335 git_die_config("pack.indexversion",
3336 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3338 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3339 max_packsize = packsizelimit_value;
3341 git_config(git_default_config, NULL);
3344 static const char fast_import_usage[] =
3345 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3347 static void parse_argv(void)
3349 unsigned int i;
3351 for (i = 1; i < global_argc; i++) {
3352 const char *a = global_argv[i];
3354 if (*a != '-' || !strcmp(a, "--"))
3355 break;
3357 if (!skip_prefix(a, "--", &a))
3358 die("unknown option %s", a);
3360 if (parse_one_option(a))
3361 continue;
3363 if (parse_one_feature(a, 0))
3364 continue;
3366 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3367 option_cat_blob_fd(a);
3368 continue;
3371 die("unknown option --%s", a);
3373 if (i != global_argc)
3374 usage(fast_import_usage);
3376 seen_data_command = 1;
3377 if (import_marks_file)
3378 read_marks();
3381 int main(int argc, char **argv)
3383 unsigned int i;
3385 git_extract_argv0_path(argv[0]);
3387 git_setup_gettext();
3389 if (argc == 2 && !strcmp(argv[1], "-h"))
3390 usage(fast_import_usage);
3392 setup_git_directory();
3393 reset_pack_idx_option(&pack_idx_opts);
3394 git_pack_config();
3395 if (!pack_compression_seen && core_compression_seen)
3396 pack_compression_level = core_compression_level;
3398 alloc_objects(object_entry_alloc);
3399 strbuf_init(&command_buf, 0);
3400 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3401 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3402 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3403 marks = pool_calloc(1, sizeof(struct mark_set));
3405 global_argc = argc;
3406 global_argv = argv;
3408 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3409 for (i = 0; i < (cmd_save - 1); i++)
3410 rc_free[i].next = &rc_free[i + 1];
3411 rc_free[cmd_save - 1].next = NULL;
3413 prepare_packed_git();
3414 start_packfile();
3415 set_die_routine(die_nicely);
3416 set_checkpoint_signal();
3417 while (read_next_command() != EOF) {
3418 const char *v;
3419 if (!strcmp("blob", command_buf.buf))
3420 parse_new_blob();
3421 else if (skip_prefix(command_buf.buf, "ls ", &v))
3422 parse_ls(v, NULL);
3423 else if (skip_prefix(command_buf.buf, "commit ", &v))
3424 parse_new_commit(v);
3425 else if (skip_prefix(command_buf.buf, "tag ", &v))
3426 parse_new_tag(v);
3427 else if (skip_prefix(command_buf.buf, "reset ", &v))
3428 parse_reset_branch(v);
3429 else if (!strcmp("checkpoint", command_buf.buf))
3430 parse_checkpoint();
3431 else if (!strcmp("done", command_buf.buf))
3432 break;
3433 else if (starts_with(command_buf.buf, "progress "))
3434 parse_progress();
3435 else if (skip_prefix(command_buf.buf, "feature ", &v))
3436 parse_feature(v);
3437 else if (skip_prefix(command_buf.buf, "option git ", &v))
3438 parse_option(v);
3439 else if (starts_with(command_buf.buf, "option "))
3440 /* ignore non-git options*/;
3441 else
3442 die("Unsupported command: %s", command_buf.buf);
3444 if (checkpoint_requested)
3445 checkpoint();
3448 /* argv hasn't been parsed yet, do so */
3449 if (!seen_data_command)
3450 parse_argv();
3452 if (require_explicit_termination && feof(stdin))
3453 die("stream ends early");
3455 end_packfile();
3457 dump_branches();
3458 dump_tags();
3459 unkeep_all_packs();
3460 dump_marks();
3462 if (pack_edges)
3463 fclose(pack_edges);
3465 if (show_stats) {
3466 uintmax_t total_count = 0, duplicate_count = 0;
3467 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3468 total_count += object_count_by_type[i];
3469 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3470 duplicate_count += duplicate_count_by_type[i];
3472 fprintf(stderr, "%s statistics:\n", argv[0]);
3473 fprintf(stderr, "---------------------------------------------------------------------\n");
3474 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3475 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3476 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]);
3477 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]);
3478 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]);
3479 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]);
3480 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3481 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3482 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3483 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3484 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3485 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3486 fprintf(stderr, "---------------------------------------------------------------------\n");
3487 pack_report();
3488 fprintf(stderr, "---------------------------------------------------------------------\n");
3489 fprintf(stderr, "\n");
3492 return failure ? 1 : 0;