Merge branch 'jk/t5561-missing-curl'
[git.git] / fast-import.c
blob99f8f56e8c4fbd20b6e6186e26d89eda1baff0d0
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 "repository.h"
158 #include "config.h"
159 #include "lockfile.h"
160 #include "object.h"
161 #include "blob.h"
162 #include "tree.h"
163 #include "commit.h"
164 #include "delta.h"
165 #include "pack.h"
166 #include "refs.h"
167 #include "csum-file.h"
168 #include "quote.h"
169 #include "dir.h"
170 #include "run-command.h"
171 #include "packfile.h"
172 #include "object-store.h"
174 #define PACK_ID_BITS 16
175 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
176 #define DEPTH_BITS 13
177 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
180 * We abuse the setuid bit on directories to mean "do not delta".
182 #define NO_DELTA S_ISUID
184 struct object_entry {
185 struct pack_idx_entry idx;
186 struct object_entry *next;
187 uint32_t type : TYPE_BITS,
188 pack_id : PACK_ID_BITS,
189 depth : DEPTH_BITS;
192 struct object_entry_pool {
193 struct object_entry_pool *next_pool;
194 struct object_entry *next_free;
195 struct object_entry *end;
196 struct object_entry entries[FLEX_ARRAY]; /* more */
199 struct mark_set {
200 union {
201 struct object_entry *marked[1024];
202 struct mark_set *sets[1024];
203 } data;
204 unsigned int shift;
207 struct last_object {
208 struct strbuf data;
209 off_t offset;
210 unsigned int depth;
211 unsigned no_swap : 1;
214 struct mem_pool {
215 struct mem_pool *next_pool;
216 char *next_free;
217 char *end;
218 uintmax_t space[FLEX_ARRAY]; /* more */
221 struct atom_str {
222 struct atom_str *next_atom;
223 unsigned short str_len;
224 char str_dat[FLEX_ARRAY]; /* more */
227 struct tree_content;
228 struct tree_entry {
229 struct tree_content *tree;
230 struct atom_str *name;
231 struct tree_entry_ms {
232 uint16_t mode;
233 struct object_id oid;
234 } versions[2];
237 struct tree_content {
238 unsigned int entry_capacity; /* must match avail_tree_content */
239 unsigned int entry_count;
240 unsigned int delta_depth;
241 struct tree_entry *entries[FLEX_ARRAY]; /* more */
244 struct avail_tree_content {
245 unsigned int entry_capacity; /* must match tree_content */
246 struct avail_tree_content *next_avail;
249 struct branch {
250 struct branch *table_next_branch;
251 struct branch *active_next_branch;
252 const char *name;
253 struct tree_entry branch_tree;
254 uintmax_t last_commit;
255 uintmax_t num_notes;
256 unsigned active : 1;
257 unsigned delete : 1;
258 unsigned pack_id : PACK_ID_BITS;
259 struct object_id oid;
262 struct tag {
263 struct tag *next_tag;
264 const char *name;
265 unsigned int pack_id;
266 struct object_id oid;
269 struct hash_list {
270 struct hash_list *next;
271 struct object_id oid;
274 typedef enum {
275 WHENSPEC_RAW = 1,
276 WHENSPEC_RFC2822,
277 WHENSPEC_NOW
278 } whenspec_type;
280 struct recent_command {
281 struct recent_command *prev;
282 struct recent_command *next;
283 char *buf;
286 /* Configured limits on output */
287 static unsigned long max_depth = 50;
288 static off_t max_packsize;
289 static int unpack_limit = 100;
290 static int force_update;
292 /* Stats and misc. counters */
293 static uintmax_t alloc_count;
294 static uintmax_t marks_set_count;
295 static uintmax_t object_count_by_type[1 << TYPE_BITS];
296 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
297 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
298 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
299 static unsigned long object_count;
300 static unsigned long branch_count;
301 static unsigned long branch_load_count;
302 static int failure;
303 static FILE *pack_edges;
304 static unsigned int show_stats = 1;
305 static int global_argc;
306 static const char **global_argv;
308 /* Memory pools */
309 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
310 static size_t total_allocd;
311 static struct mem_pool *mem_pool;
313 /* Atom management */
314 static unsigned int atom_table_sz = 4451;
315 static unsigned int atom_cnt;
316 static struct atom_str **atom_table;
318 /* The .pack file being generated */
319 static struct pack_idx_option pack_idx_opts;
320 static unsigned int pack_id;
321 static struct hashfile *pack_file;
322 static struct packed_git *pack_data;
323 static struct packed_git **all_packs;
324 static off_t pack_size;
326 /* Table of objects we've written. */
327 static unsigned int object_entry_alloc = 5000;
328 static struct object_entry_pool *blocks;
329 static struct object_entry *object_table[1 << 16];
330 static struct mark_set *marks;
331 static const char *export_marks_file;
332 static const char *import_marks_file;
333 static int import_marks_file_from_stream;
334 static int import_marks_file_ignore_missing;
335 static int import_marks_file_done;
336 static int relative_marks_paths;
338 /* Our last blob */
339 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
341 /* Tree management */
342 static unsigned int tree_entry_alloc = 1000;
343 static void *avail_tree_entry;
344 static unsigned int avail_tree_table_sz = 100;
345 static struct avail_tree_content **avail_tree_table;
346 static struct strbuf old_tree = STRBUF_INIT;
347 static struct strbuf new_tree = STRBUF_INIT;
349 /* Branch data */
350 static unsigned long max_active_branches = 5;
351 static unsigned long cur_active_branches;
352 static unsigned long branch_table_sz = 1039;
353 static struct branch **branch_table;
354 static struct branch *active_branches;
356 /* Tag data */
357 static struct tag *first_tag;
358 static struct tag *last_tag;
360 /* Input stream parsing */
361 static whenspec_type whenspec = WHENSPEC_RAW;
362 static struct strbuf command_buf = STRBUF_INIT;
363 static int unread_command_buf;
364 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
365 static struct recent_command *cmd_tail = &cmd_hist;
366 static struct recent_command *rc_free;
367 static unsigned int cmd_save = 100;
368 static uintmax_t next_mark;
369 static struct strbuf new_data = STRBUF_INIT;
370 static int seen_data_command;
371 static int require_explicit_termination;
373 /* Signal handling */
374 static volatile sig_atomic_t checkpoint_requested;
376 /* Where to write output of cat-blob commands */
377 static int cat_blob_fd = STDOUT_FILENO;
379 static void parse_argv(void);
380 static void parse_get_mark(const char *p);
381 static void parse_cat_blob(const char *p);
382 static void parse_ls(const char *p, struct branch *b);
384 static void write_branch_report(FILE *rpt, struct branch *b)
386 fprintf(rpt, "%s:\n", b->name);
388 fprintf(rpt, " status :");
389 if (b->active)
390 fputs(" active", rpt);
391 if (b->branch_tree.tree)
392 fputs(" loaded", rpt);
393 if (is_null_oid(&b->branch_tree.versions[1].oid))
394 fputs(" dirty", rpt);
395 fputc('\n', rpt);
397 fprintf(rpt, " tip commit : %s\n", oid_to_hex(&b->oid));
398 fprintf(rpt, " old tree : %s\n",
399 oid_to_hex(&b->branch_tree.versions[0].oid));
400 fprintf(rpt, " cur tree : %s\n",
401 oid_to_hex(&b->branch_tree.versions[1].oid));
402 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
404 fputs(" last pack : ", rpt);
405 if (b->pack_id < MAX_PACK_ID)
406 fprintf(rpt, "%u", b->pack_id);
407 fputc('\n', rpt);
409 fputc('\n', rpt);
412 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
414 static void write_crash_report(const char *err)
416 char *loc = git_pathdup("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
417 FILE *rpt = fopen(loc, "w");
418 struct branch *b;
419 unsigned long lu;
420 struct recent_command *rc;
422 if (!rpt) {
423 error_errno("can't write crash report %s", loc);
424 free(loc);
425 return;
428 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
430 fprintf(rpt, "fast-import crash report:\n");
431 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
432 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
433 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_MODE(ISO8601)));
434 fputc('\n', rpt);
436 fputs("fatal: ", rpt);
437 fputs(err, rpt);
438 fputc('\n', rpt);
440 fputc('\n', rpt);
441 fputs("Most Recent Commands Before Crash\n", rpt);
442 fputs("---------------------------------\n", rpt);
443 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
444 if (rc->next == &cmd_hist)
445 fputs("* ", rpt);
446 else
447 fputs(" ", rpt);
448 fputs(rc->buf, rpt);
449 fputc('\n', rpt);
452 fputc('\n', rpt);
453 fputs("Active Branch LRU\n", rpt);
454 fputs("-----------------\n", rpt);
455 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
456 cur_active_branches,
457 max_active_branches);
458 fputc('\n', rpt);
459 fputs(" pos clock name\n", rpt);
460 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
461 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
462 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
463 ++lu, b->last_commit, b->name);
465 fputc('\n', rpt);
466 fputs("Inactive Branches\n", rpt);
467 fputs("-----------------\n", rpt);
468 for (lu = 0; lu < branch_table_sz; lu++) {
469 for (b = branch_table[lu]; b; b = b->table_next_branch)
470 write_branch_report(rpt, b);
473 if (first_tag) {
474 struct tag *tg;
475 fputc('\n', rpt);
476 fputs("Annotated Tags\n", rpt);
477 fputs("--------------\n", rpt);
478 for (tg = first_tag; tg; tg = tg->next_tag) {
479 fputs(oid_to_hex(&tg->oid), rpt);
480 fputc(' ', rpt);
481 fputs(tg->name, rpt);
482 fputc('\n', rpt);
486 fputc('\n', rpt);
487 fputs("Marks\n", rpt);
488 fputs("-----\n", rpt);
489 if (export_marks_file)
490 fprintf(rpt, " exported to %s\n", export_marks_file);
491 else
492 dump_marks_helper(rpt, 0, marks);
494 fputc('\n', rpt);
495 fputs("-------------------\n", rpt);
496 fputs("END OF CRASH REPORT\n", rpt);
497 fclose(rpt);
498 free(loc);
501 static void end_packfile(void);
502 static void unkeep_all_packs(void);
503 static void dump_marks(void);
505 static NORETURN void die_nicely(const char *err, va_list params)
507 static int zombie;
508 char message[2 * PATH_MAX];
510 vsnprintf(message, sizeof(message), err, params);
511 fputs("fatal: ", stderr);
512 fputs(message, stderr);
513 fputc('\n', stderr);
515 if (!zombie) {
516 zombie = 1;
517 write_crash_report(message);
518 end_packfile();
519 unkeep_all_packs();
520 dump_marks();
522 exit(128);
525 #ifndef SIGUSR1 /* Windows, for example */
527 static void set_checkpoint_signal(void)
531 #else
533 static void checkpoint_signal(int signo)
535 checkpoint_requested = 1;
538 static void set_checkpoint_signal(void)
540 struct sigaction sa;
542 memset(&sa, 0, sizeof(sa));
543 sa.sa_handler = checkpoint_signal;
544 sigemptyset(&sa.sa_mask);
545 sa.sa_flags = SA_RESTART;
546 sigaction(SIGUSR1, &sa, NULL);
549 #endif
551 static void alloc_objects(unsigned int cnt)
553 struct object_entry_pool *b;
555 b = xmalloc(sizeof(struct object_entry_pool)
556 + cnt * sizeof(struct object_entry));
557 b->next_pool = blocks;
558 b->next_free = b->entries;
559 b->end = b->entries + cnt;
560 blocks = b;
561 alloc_count += cnt;
564 static struct object_entry *new_object(struct object_id *oid)
566 struct object_entry *e;
568 if (blocks->next_free == blocks->end)
569 alloc_objects(object_entry_alloc);
571 e = blocks->next_free++;
572 oidcpy(&e->idx.oid, oid);
573 return e;
576 static struct object_entry *find_object(struct object_id *oid)
578 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
579 struct object_entry *e;
580 for (e = object_table[h]; e; e = e->next)
581 if (!oidcmp(oid, &e->idx.oid))
582 return e;
583 return NULL;
586 static struct object_entry *insert_object(struct object_id *oid)
588 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
589 struct object_entry *e = object_table[h];
591 while (e) {
592 if (!oidcmp(oid, &e->idx.oid))
593 return e;
594 e = e->next;
597 e = new_object(oid);
598 e->next = object_table[h];
599 e->idx.offset = 0;
600 object_table[h] = e;
601 return e;
604 static void invalidate_pack_id(unsigned int id)
606 unsigned int h;
607 unsigned long lu;
608 struct tag *t;
610 for (h = 0; h < ARRAY_SIZE(object_table); h++) {
611 struct object_entry *e;
613 for (e = object_table[h]; e; e = e->next)
614 if (e->pack_id == id)
615 e->pack_id = MAX_PACK_ID;
618 for (lu = 0; lu < branch_table_sz; lu++) {
619 struct branch *b;
621 for (b = branch_table[lu]; b; b = b->table_next_branch)
622 if (b->pack_id == id)
623 b->pack_id = MAX_PACK_ID;
626 for (t = first_tag; t; t = t->next_tag)
627 if (t->pack_id == id)
628 t->pack_id = MAX_PACK_ID;
631 static unsigned int hc_str(const char *s, size_t len)
633 unsigned int r = 0;
634 while (len-- > 0)
635 r = r * 31 + *s++;
636 return r;
639 static void *pool_alloc(size_t len)
641 struct mem_pool *p;
642 void *r;
644 /* round up to a 'uintmax_t' alignment */
645 if (len & (sizeof(uintmax_t) - 1))
646 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
648 for (p = mem_pool; p; p = p->next_pool)
649 if ((p->end - p->next_free >= len))
650 break;
652 if (!p) {
653 if (len >= (mem_pool_alloc/2)) {
654 total_allocd += len;
655 return xmalloc(len);
657 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
658 p = xmalloc(st_add(sizeof(struct mem_pool), mem_pool_alloc));
659 p->next_pool = mem_pool;
660 p->next_free = (char *) p->space;
661 p->end = p->next_free + mem_pool_alloc;
662 mem_pool = p;
665 r = p->next_free;
666 p->next_free += len;
667 return r;
670 static void *pool_calloc(size_t count, size_t size)
672 size_t len = count * size;
673 void *r = pool_alloc(len);
674 memset(r, 0, len);
675 return r;
678 static char *pool_strdup(const char *s)
680 size_t len = strlen(s) + 1;
681 char *r = pool_alloc(len);
682 memcpy(r, s, len);
683 return r;
686 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
688 struct mark_set *s = marks;
689 while ((idnum >> s->shift) >= 1024) {
690 s = pool_calloc(1, sizeof(struct mark_set));
691 s->shift = marks->shift + 10;
692 s->data.sets[0] = marks;
693 marks = s;
695 while (s->shift) {
696 uintmax_t i = idnum >> s->shift;
697 idnum -= i << s->shift;
698 if (!s->data.sets[i]) {
699 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
700 s->data.sets[i]->shift = s->shift - 10;
702 s = s->data.sets[i];
704 if (!s->data.marked[idnum])
705 marks_set_count++;
706 s->data.marked[idnum] = oe;
709 static struct object_entry *find_mark(uintmax_t idnum)
711 uintmax_t orig_idnum = idnum;
712 struct mark_set *s = marks;
713 struct object_entry *oe = NULL;
714 if ((idnum >> s->shift) < 1024) {
715 while (s && s->shift) {
716 uintmax_t i = idnum >> s->shift;
717 idnum -= i << s->shift;
718 s = s->data.sets[i];
720 if (s)
721 oe = s->data.marked[idnum];
723 if (!oe)
724 die("mark :%" PRIuMAX " not declared", orig_idnum);
725 return oe;
728 static struct atom_str *to_atom(const char *s, unsigned short len)
730 unsigned int hc = hc_str(s, len) % atom_table_sz;
731 struct atom_str *c;
733 for (c = atom_table[hc]; c; c = c->next_atom)
734 if (c->str_len == len && !strncmp(s, c->str_dat, len))
735 return c;
737 c = pool_alloc(sizeof(struct atom_str) + len + 1);
738 c->str_len = len;
739 memcpy(c->str_dat, s, len);
740 c->str_dat[len] = 0;
741 c->next_atom = atom_table[hc];
742 atom_table[hc] = c;
743 atom_cnt++;
744 return c;
747 static struct branch *lookup_branch(const char *name)
749 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
750 struct branch *b;
752 for (b = branch_table[hc]; b; b = b->table_next_branch)
753 if (!strcmp(name, b->name))
754 return b;
755 return NULL;
758 static struct branch *new_branch(const char *name)
760 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
761 struct branch *b = lookup_branch(name);
763 if (b)
764 die("Invalid attempt to create duplicate branch: %s", name);
765 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
766 die("Branch name doesn't conform to GIT standards: %s", name);
768 b = pool_calloc(1, sizeof(struct branch));
769 b->name = pool_strdup(name);
770 b->table_next_branch = branch_table[hc];
771 b->branch_tree.versions[0].mode = S_IFDIR;
772 b->branch_tree.versions[1].mode = S_IFDIR;
773 b->num_notes = 0;
774 b->active = 0;
775 b->pack_id = MAX_PACK_ID;
776 branch_table[hc] = b;
777 branch_count++;
778 return b;
781 static unsigned int hc_entries(unsigned int cnt)
783 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
784 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
787 static struct tree_content *new_tree_content(unsigned int cnt)
789 struct avail_tree_content *f, *l = NULL;
790 struct tree_content *t;
791 unsigned int hc = hc_entries(cnt);
793 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
794 if (f->entry_capacity >= cnt)
795 break;
797 if (f) {
798 if (l)
799 l->next_avail = f->next_avail;
800 else
801 avail_tree_table[hc] = f->next_avail;
802 } else {
803 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
804 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
805 f->entry_capacity = cnt;
808 t = (struct tree_content*)f;
809 t->entry_count = 0;
810 t->delta_depth = 0;
811 return t;
814 static void release_tree_entry(struct tree_entry *e);
815 static void release_tree_content(struct tree_content *t)
817 struct avail_tree_content *f = (struct avail_tree_content*)t;
818 unsigned int hc = hc_entries(f->entry_capacity);
819 f->next_avail = avail_tree_table[hc];
820 avail_tree_table[hc] = f;
823 static void release_tree_content_recursive(struct tree_content *t)
825 unsigned int i;
826 for (i = 0; i < t->entry_count; i++)
827 release_tree_entry(t->entries[i]);
828 release_tree_content(t);
831 static struct tree_content *grow_tree_content(
832 struct tree_content *t,
833 int amt)
835 struct tree_content *r = new_tree_content(t->entry_count + amt);
836 r->entry_count = t->entry_count;
837 r->delta_depth = t->delta_depth;
838 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
839 release_tree_content(t);
840 return r;
843 static struct tree_entry *new_tree_entry(void)
845 struct tree_entry *e;
847 if (!avail_tree_entry) {
848 unsigned int n = tree_entry_alloc;
849 total_allocd += n * sizeof(struct tree_entry);
850 ALLOC_ARRAY(e, n);
851 avail_tree_entry = e;
852 while (n-- > 1) {
853 *((void**)e) = e + 1;
854 e++;
856 *((void**)e) = NULL;
859 e = avail_tree_entry;
860 avail_tree_entry = *((void**)e);
861 return e;
864 static void release_tree_entry(struct tree_entry *e)
866 if (e->tree)
867 release_tree_content_recursive(e->tree);
868 *((void**)e) = avail_tree_entry;
869 avail_tree_entry = e;
872 static struct tree_content *dup_tree_content(struct tree_content *s)
874 struct tree_content *d;
875 struct tree_entry *a, *b;
876 unsigned int i;
878 if (!s)
879 return NULL;
880 d = new_tree_content(s->entry_count);
881 for (i = 0; i < s->entry_count; i++) {
882 a = s->entries[i];
883 b = new_tree_entry();
884 memcpy(b, a, sizeof(*a));
885 if (a->tree && is_null_oid(&b->versions[1].oid))
886 b->tree = dup_tree_content(a->tree);
887 else
888 b->tree = NULL;
889 d->entries[i] = b;
891 d->entry_count = s->entry_count;
892 d->delta_depth = s->delta_depth;
894 return d;
897 static void start_packfile(void)
899 struct strbuf tmp_file = STRBUF_INIT;
900 struct packed_git *p;
901 struct pack_header hdr;
902 int pack_fd;
904 pack_fd = odb_mkstemp(&tmp_file, "pack/tmp_pack_XXXXXX");
905 FLEX_ALLOC_STR(p, pack_name, tmp_file.buf);
906 strbuf_release(&tmp_file);
908 p->pack_fd = pack_fd;
909 p->do_not_close = 1;
910 pack_file = hashfd(pack_fd, p->pack_name);
912 hdr.hdr_signature = htonl(PACK_SIGNATURE);
913 hdr.hdr_version = htonl(2);
914 hdr.hdr_entries = 0;
915 hashwrite(pack_file, &hdr, sizeof(hdr));
917 pack_data = p;
918 pack_size = sizeof(hdr);
919 object_count = 0;
921 REALLOC_ARRAY(all_packs, pack_id + 1);
922 all_packs[pack_id] = p;
925 static const char *create_index(void)
927 const char *tmpfile;
928 struct pack_idx_entry **idx, **c, **last;
929 struct object_entry *e;
930 struct object_entry_pool *o;
932 /* Build the table of object IDs. */
933 ALLOC_ARRAY(idx, object_count);
934 c = idx;
935 for (o = blocks; o; o = o->next_pool)
936 for (e = o->next_free; e-- != o->entries;)
937 if (pack_id == e->pack_id)
938 *c++ = &e->idx;
939 last = idx + object_count;
940 if (c != last)
941 die("internal consistency error creating the index");
943 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
944 free(idx);
945 return tmpfile;
948 static char *keep_pack(const char *curr_index_name)
950 static const char *keep_msg = "fast-import";
951 struct strbuf name = STRBUF_INIT;
952 int keep_fd;
954 odb_pack_name(&name, pack_data->sha1, "keep");
955 keep_fd = odb_pack_keep(name.buf);
956 if (keep_fd < 0)
957 die_errno("cannot create keep file");
958 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
959 if (close(keep_fd))
960 die_errno("failed to write keep file");
962 odb_pack_name(&name, pack_data->sha1, "pack");
963 if (finalize_object_file(pack_data->pack_name, name.buf))
964 die("cannot store pack file");
966 odb_pack_name(&name, pack_data->sha1, "idx");
967 if (finalize_object_file(curr_index_name, name.buf))
968 die("cannot store index file");
969 free((void *)curr_index_name);
970 return strbuf_detach(&name, NULL);
973 static void unkeep_all_packs(void)
975 struct strbuf name = STRBUF_INIT;
976 int k;
978 for (k = 0; k < pack_id; k++) {
979 struct packed_git *p = all_packs[k];
980 odb_pack_name(&name, p->sha1, "keep");
981 unlink_or_warn(name.buf);
983 strbuf_release(&name);
986 static int loosen_small_pack(const struct packed_git *p)
988 struct child_process unpack = CHILD_PROCESS_INIT;
990 if (lseek(p->pack_fd, 0, SEEK_SET) < 0)
991 die_errno("Failed seeking to start of '%s'", p->pack_name);
993 unpack.in = p->pack_fd;
994 unpack.git_cmd = 1;
995 unpack.stdout_to_stderr = 1;
996 argv_array_push(&unpack.args, "unpack-objects");
997 if (!show_stats)
998 argv_array_push(&unpack.args, "-q");
1000 return run_command(&unpack);
1003 static void end_packfile(void)
1005 static int running;
1007 if (running || !pack_data)
1008 return;
1010 running = 1;
1011 clear_delta_base_cache();
1012 if (object_count) {
1013 struct packed_git *new_p;
1014 struct object_id cur_pack_oid;
1015 char *idx_name;
1016 int i;
1017 struct branch *b;
1018 struct tag *t;
1020 close_pack_windows(pack_data);
1021 hashclose(pack_file, cur_pack_oid.hash, 0);
1022 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
1023 pack_data->pack_name, object_count,
1024 cur_pack_oid.hash, pack_size);
1026 if (object_count <= unpack_limit) {
1027 if (!loosen_small_pack(pack_data)) {
1028 invalidate_pack_id(pack_id);
1029 goto discard_pack;
1033 close(pack_data->pack_fd);
1034 idx_name = keep_pack(create_index());
1036 /* Register the packfile with core git's machinery. */
1037 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
1038 if (!new_p)
1039 die("core git rejected index %s", idx_name);
1040 all_packs[pack_id] = new_p;
1041 install_packed_git(the_repository, new_p);
1042 free(idx_name);
1044 /* Print the boundary */
1045 if (pack_edges) {
1046 fprintf(pack_edges, "%s:", new_p->pack_name);
1047 for (i = 0; i < branch_table_sz; i++) {
1048 for (b = branch_table[i]; b; b = b->table_next_branch) {
1049 if (b->pack_id == pack_id)
1050 fprintf(pack_edges, " %s",
1051 oid_to_hex(&b->oid));
1054 for (t = first_tag; t; t = t->next_tag) {
1055 if (t->pack_id == pack_id)
1056 fprintf(pack_edges, " %s",
1057 oid_to_hex(&t->oid));
1059 fputc('\n', pack_edges);
1060 fflush(pack_edges);
1063 pack_id++;
1065 else {
1066 discard_pack:
1067 close(pack_data->pack_fd);
1068 unlink_or_warn(pack_data->pack_name);
1070 FREE_AND_NULL(pack_data);
1071 running = 0;
1073 /* We can't carry a delta across packfiles. */
1074 strbuf_release(&last_blob.data);
1075 last_blob.offset = 0;
1076 last_blob.depth = 0;
1079 static void cycle_packfile(void)
1081 end_packfile();
1082 start_packfile();
1085 static int store_object(
1086 enum object_type type,
1087 struct strbuf *dat,
1088 struct last_object *last,
1089 struct object_id *oidout,
1090 uintmax_t mark)
1092 void *out, *delta;
1093 struct object_entry *e;
1094 unsigned char hdr[96];
1095 struct object_id oid;
1096 unsigned long hdrlen, deltalen;
1097 git_hash_ctx c;
1098 git_zstream s;
1100 hdrlen = xsnprintf((char *)hdr, sizeof(hdr), "%s %lu",
1101 type_name(type), (unsigned long)dat->len) + 1;
1102 the_hash_algo->init_fn(&c);
1103 the_hash_algo->update_fn(&c, hdr, hdrlen);
1104 the_hash_algo->update_fn(&c, dat->buf, dat->len);
1105 the_hash_algo->final_fn(oid.hash, &c);
1106 if (oidout)
1107 oidcpy(oidout, &oid);
1109 e = insert_object(&oid);
1110 if (mark)
1111 insert_mark(mark, e);
1112 if (e->idx.offset) {
1113 duplicate_count_by_type[type]++;
1114 return 1;
1115 } else if (find_sha1_pack(oid.hash,
1116 get_packed_git(the_repository))) {
1117 e->type = type;
1118 e->pack_id = MAX_PACK_ID;
1119 e->idx.offset = 1; /* just not zero! */
1120 duplicate_count_by_type[type]++;
1121 return 1;
1124 if (last && last->data.buf && last->depth < max_depth
1125 && dat->len > the_hash_algo->rawsz) {
1127 delta_count_attempts_by_type[type]++;
1128 delta = diff_delta(last->data.buf, last->data.len,
1129 dat->buf, dat->len,
1130 &deltalen, dat->len - the_hash_algo->rawsz);
1131 } else
1132 delta = NULL;
1134 git_deflate_init(&s, pack_compression_level);
1135 if (delta) {
1136 s.next_in = delta;
1137 s.avail_in = deltalen;
1138 } else {
1139 s.next_in = (void *)dat->buf;
1140 s.avail_in = dat->len;
1142 s.avail_out = git_deflate_bound(&s, s.avail_in);
1143 s.next_out = out = xmalloc(s.avail_out);
1144 while (git_deflate(&s, Z_FINISH) == Z_OK)
1145 ; /* nothing */
1146 git_deflate_end(&s);
1148 /* Determine if we should auto-checkpoint. */
1149 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1150 || (pack_size + 60 + s.total_out) < pack_size) {
1152 /* This new object needs to *not* have the current pack_id. */
1153 e->pack_id = pack_id + 1;
1154 cycle_packfile();
1156 /* We cannot carry a delta into the new pack. */
1157 if (delta) {
1158 FREE_AND_NULL(delta);
1160 git_deflate_init(&s, pack_compression_level);
1161 s.next_in = (void *)dat->buf;
1162 s.avail_in = dat->len;
1163 s.avail_out = git_deflate_bound(&s, s.avail_in);
1164 s.next_out = out = xrealloc(out, s.avail_out);
1165 while (git_deflate(&s, Z_FINISH) == Z_OK)
1166 ; /* nothing */
1167 git_deflate_end(&s);
1171 e->type = type;
1172 e->pack_id = pack_id;
1173 e->idx.offset = pack_size;
1174 object_count++;
1175 object_count_by_type[type]++;
1177 crc32_begin(pack_file);
1179 if (delta) {
1180 off_t ofs = e->idx.offset - last->offset;
1181 unsigned pos = sizeof(hdr) - 1;
1183 delta_count_by_type[type]++;
1184 e->depth = last->depth + 1;
1186 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1187 OBJ_OFS_DELTA, deltalen);
1188 hashwrite(pack_file, hdr, hdrlen);
1189 pack_size += hdrlen;
1191 hdr[pos] = ofs & 127;
1192 while (ofs >>= 7)
1193 hdr[--pos] = 128 | (--ofs & 127);
1194 hashwrite(pack_file, hdr + pos, sizeof(hdr) - pos);
1195 pack_size += sizeof(hdr) - pos;
1196 } else {
1197 e->depth = 0;
1198 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1199 type, dat->len);
1200 hashwrite(pack_file, hdr, hdrlen);
1201 pack_size += hdrlen;
1204 hashwrite(pack_file, out, s.total_out);
1205 pack_size += s.total_out;
1207 e->idx.crc32 = crc32_end(pack_file);
1209 free(out);
1210 free(delta);
1211 if (last) {
1212 if (last->no_swap) {
1213 last->data = *dat;
1214 } else {
1215 strbuf_swap(&last->data, dat);
1217 last->offset = e->idx.offset;
1218 last->depth = e->depth;
1220 return 0;
1223 static void truncate_pack(struct hashfile_checkpoint *checkpoint)
1225 if (hashfile_truncate(pack_file, checkpoint))
1226 die_errno("cannot truncate pack to skip duplicate");
1227 pack_size = checkpoint->offset;
1230 static void stream_blob(uintmax_t len, struct object_id *oidout, uintmax_t mark)
1232 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1233 unsigned char *in_buf = xmalloc(in_sz);
1234 unsigned char *out_buf = xmalloc(out_sz);
1235 struct object_entry *e;
1236 struct object_id oid;
1237 unsigned long hdrlen;
1238 off_t offset;
1239 git_hash_ctx c;
1240 git_zstream s;
1241 struct hashfile_checkpoint checkpoint;
1242 int status = Z_OK;
1244 /* Determine if we should auto-checkpoint. */
1245 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1246 || (pack_size + 60 + len) < pack_size)
1247 cycle_packfile();
1249 hashfile_checkpoint(pack_file, &checkpoint);
1250 offset = checkpoint.offset;
1252 hdrlen = xsnprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1254 the_hash_algo->init_fn(&c);
1255 the_hash_algo->update_fn(&c, out_buf, hdrlen);
1257 crc32_begin(pack_file);
1259 git_deflate_init(&s, pack_compression_level);
1261 hdrlen = encode_in_pack_object_header(out_buf, out_sz, OBJ_BLOB, len);
1263 s.next_out = out_buf + hdrlen;
1264 s.avail_out = out_sz - hdrlen;
1266 while (status != Z_STREAM_END) {
1267 if (0 < len && !s.avail_in) {
1268 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1269 size_t n = fread(in_buf, 1, cnt, stdin);
1270 if (!n && feof(stdin))
1271 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1273 the_hash_algo->update_fn(&c, in_buf, n);
1274 s.next_in = in_buf;
1275 s.avail_in = n;
1276 len -= n;
1279 status = git_deflate(&s, len ? 0 : Z_FINISH);
1281 if (!s.avail_out || status == Z_STREAM_END) {
1282 size_t n = s.next_out - out_buf;
1283 hashwrite(pack_file, out_buf, n);
1284 pack_size += n;
1285 s.next_out = out_buf;
1286 s.avail_out = out_sz;
1289 switch (status) {
1290 case Z_OK:
1291 case Z_BUF_ERROR:
1292 case Z_STREAM_END:
1293 continue;
1294 default:
1295 die("unexpected deflate failure: %d", status);
1298 git_deflate_end(&s);
1299 the_hash_algo->final_fn(oid.hash, &c);
1301 if (oidout)
1302 oidcpy(oidout, &oid);
1304 e = insert_object(&oid);
1306 if (mark)
1307 insert_mark(mark, e);
1309 if (e->idx.offset) {
1310 duplicate_count_by_type[OBJ_BLOB]++;
1311 truncate_pack(&checkpoint);
1313 } else if (find_sha1_pack(oid.hash,
1314 get_packed_git(the_repository))) {
1315 e->type = OBJ_BLOB;
1316 e->pack_id = MAX_PACK_ID;
1317 e->idx.offset = 1; /* just not zero! */
1318 duplicate_count_by_type[OBJ_BLOB]++;
1319 truncate_pack(&checkpoint);
1321 } else {
1322 e->depth = 0;
1323 e->type = OBJ_BLOB;
1324 e->pack_id = pack_id;
1325 e->idx.offset = offset;
1326 e->idx.crc32 = crc32_end(pack_file);
1327 object_count++;
1328 object_count_by_type[OBJ_BLOB]++;
1331 free(in_buf);
1332 free(out_buf);
1335 /* All calls must be guarded by find_object() or find_mark() to
1336 * ensure the 'struct object_entry' passed was written by this
1337 * process instance. We unpack the entry by the offset, avoiding
1338 * the need for the corresponding .idx file. This unpacking rule
1339 * works because we only use OBJ_REF_DELTA within the packfiles
1340 * created by fast-import.
1342 * oe must not be NULL. Such an oe usually comes from giving
1343 * an unknown SHA-1 to find_object() or an undefined mark to
1344 * find_mark(). Callers must test for this condition and use
1345 * the standard read_sha1_file() when it happens.
1347 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1348 * find_mark(), where the mark was reloaded from an existing marks
1349 * file and is referencing an object that this fast-import process
1350 * instance did not write out to a packfile. Callers must test for
1351 * this condition and use read_sha1_file() instead.
1353 static void *gfi_unpack_entry(
1354 struct object_entry *oe,
1355 unsigned long *sizep)
1357 enum object_type type;
1358 struct packed_git *p = all_packs[oe->pack_id];
1359 if (p == pack_data && p->pack_size < (pack_size + the_hash_algo->rawsz)) {
1360 /* The object is stored in the packfile we are writing to
1361 * and we have modified it since the last time we scanned
1362 * back to read a previously written object. If an old
1363 * window covered [p->pack_size, p->pack_size + rawsz) its
1364 * data is stale and is not valid. Closing all windows
1365 * and updating the packfile length ensures we can read
1366 * the newly written data.
1368 close_pack_windows(p);
1369 hashflush(pack_file);
1371 /* We have to offer rawsz bytes additional on the end of
1372 * the packfile as the core unpacker code assumes the
1373 * footer is present at the file end and must promise
1374 * at least rawsz bytes within any window it maps. But
1375 * we don't actually create the footer here.
1377 p->pack_size = pack_size + the_hash_algo->rawsz;
1379 return unpack_entry(p, oe->idx.offset, &type, sizep);
1382 static const char *get_mode(const char *str, uint16_t *modep)
1384 unsigned char c;
1385 uint16_t mode = 0;
1387 while ((c = *str++) != ' ') {
1388 if (c < '0' || c > '7')
1389 return NULL;
1390 mode = (mode << 3) + (c - '0');
1392 *modep = mode;
1393 return str;
1396 static void load_tree(struct tree_entry *root)
1398 struct object_id *oid = &root->versions[1].oid;
1399 struct object_entry *myoe;
1400 struct tree_content *t;
1401 unsigned long size;
1402 char *buf;
1403 const char *c;
1405 root->tree = t = new_tree_content(8);
1406 if (is_null_oid(oid))
1407 return;
1409 myoe = find_object(oid);
1410 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1411 if (myoe->type != OBJ_TREE)
1412 die("Not a tree: %s", oid_to_hex(oid));
1413 t->delta_depth = myoe->depth;
1414 buf = gfi_unpack_entry(myoe, &size);
1415 if (!buf)
1416 die("Can't load tree %s", oid_to_hex(oid));
1417 } else {
1418 enum object_type type;
1419 buf = read_object_file(oid, &type, &size);
1420 if (!buf || type != OBJ_TREE)
1421 die("Can't load tree %s", oid_to_hex(oid));
1424 c = buf;
1425 while (c != (buf + size)) {
1426 struct tree_entry *e = new_tree_entry();
1428 if (t->entry_count == t->entry_capacity)
1429 root->tree = t = grow_tree_content(t, t->entry_count);
1430 t->entries[t->entry_count++] = e;
1432 e->tree = NULL;
1433 c = get_mode(c, &e->versions[1].mode);
1434 if (!c)
1435 die("Corrupt mode in %s", oid_to_hex(oid));
1436 e->versions[0].mode = e->versions[1].mode;
1437 e->name = to_atom(c, strlen(c));
1438 c += e->name->str_len + 1;
1439 hashcpy(e->versions[0].oid.hash, (unsigned char *)c);
1440 hashcpy(e->versions[1].oid.hash, (unsigned char *)c);
1441 c += GIT_SHA1_RAWSZ;
1443 free(buf);
1446 static int tecmp0 (const void *_a, const void *_b)
1448 struct tree_entry *a = *((struct tree_entry**)_a);
1449 struct tree_entry *b = *((struct tree_entry**)_b);
1450 return base_name_compare(
1451 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1452 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1455 static int tecmp1 (const void *_a, const void *_b)
1457 struct tree_entry *a = *((struct tree_entry**)_a);
1458 struct tree_entry *b = *((struct tree_entry**)_b);
1459 return base_name_compare(
1460 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1461 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1464 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1466 size_t maxlen = 0;
1467 unsigned int i;
1469 if (!v)
1470 QSORT(t->entries, t->entry_count, tecmp0);
1471 else
1472 QSORT(t->entries, t->entry_count, tecmp1);
1474 for (i = 0; i < t->entry_count; i++) {
1475 if (t->entries[i]->versions[v].mode)
1476 maxlen += t->entries[i]->name->str_len + 34;
1479 strbuf_reset(b);
1480 strbuf_grow(b, maxlen);
1481 for (i = 0; i < t->entry_count; i++) {
1482 struct tree_entry *e = t->entries[i];
1483 if (!e->versions[v].mode)
1484 continue;
1485 strbuf_addf(b, "%o %s%c",
1486 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1487 e->name->str_dat, '\0');
1488 strbuf_add(b, e->versions[v].oid.hash, GIT_SHA1_RAWSZ);
1492 static void store_tree(struct tree_entry *root)
1494 struct tree_content *t;
1495 unsigned int i, j, del;
1496 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1497 struct object_entry *le = NULL;
1499 if (!is_null_oid(&root->versions[1].oid))
1500 return;
1502 if (!root->tree)
1503 load_tree(root);
1504 t = root->tree;
1506 for (i = 0; i < t->entry_count; i++) {
1507 if (t->entries[i]->tree)
1508 store_tree(t->entries[i]);
1511 if (!(root->versions[0].mode & NO_DELTA))
1512 le = find_object(&root->versions[0].oid);
1513 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1514 mktree(t, 0, &old_tree);
1515 lo.data = old_tree;
1516 lo.offset = le->idx.offset;
1517 lo.depth = t->delta_depth;
1520 mktree(t, 1, &new_tree);
1521 store_object(OBJ_TREE, &new_tree, &lo, &root->versions[1].oid, 0);
1523 t->delta_depth = lo.depth;
1524 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1525 struct tree_entry *e = t->entries[i];
1526 if (e->versions[1].mode) {
1527 e->versions[0].mode = e->versions[1].mode;
1528 oidcpy(&e->versions[0].oid, &e->versions[1].oid);
1529 t->entries[j++] = e;
1530 } else {
1531 release_tree_entry(e);
1532 del++;
1535 t->entry_count -= del;
1538 static void tree_content_replace(
1539 struct tree_entry *root,
1540 const struct object_id *oid,
1541 const uint16_t mode,
1542 struct tree_content *newtree)
1544 if (!S_ISDIR(mode))
1545 die("Root cannot be a non-directory");
1546 oidclr(&root->versions[0].oid);
1547 oidcpy(&root->versions[1].oid, oid);
1548 if (root->tree)
1549 release_tree_content_recursive(root->tree);
1550 root->tree = newtree;
1553 static int tree_content_set(
1554 struct tree_entry *root,
1555 const char *p,
1556 const struct object_id *oid,
1557 const uint16_t mode,
1558 struct tree_content *subtree)
1560 struct tree_content *t;
1561 const char *slash1;
1562 unsigned int i, n;
1563 struct tree_entry *e;
1565 slash1 = strchrnul(p, '/');
1566 n = slash1 - p;
1567 if (!n)
1568 die("Empty path component found in input");
1569 if (!*slash1 && !S_ISDIR(mode) && subtree)
1570 die("Non-directories cannot have subtrees");
1572 if (!root->tree)
1573 load_tree(root);
1574 t = root->tree;
1575 for (i = 0; i < t->entry_count; i++) {
1576 e = t->entries[i];
1577 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1578 if (!*slash1) {
1579 if (!S_ISDIR(mode)
1580 && e->versions[1].mode == mode
1581 && !oidcmp(&e->versions[1].oid, oid))
1582 return 0;
1583 e->versions[1].mode = mode;
1584 oidcpy(&e->versions[1].oid, oid);
1585 if (e->tree)
1586 release_tree_content_recursive(e->tree);
1587 e->tree = subtree;
1590 * We need to leave e->versions[0].sha1 alone
1591 * to avoid modifying the preimage tree used
1592 * when writing out the parent directory.
1593 * But after replacing the subdir with a
1594 * completely different one, it's not a good
1595 * delta base any more, and besides, we've
1596 * thrown away the tree entries needed to
1597 * make a delta against it.
1599 * So let's just explicitly disable deltas
1600 * for the subtree.
1602 if (S_ISDIR(e->versions[0].mode))
1603 e->versions[0].mode |= NO_DELTA;
1605 oidclr(&root->versions[1].oid);
1606 return 1;
1608 if (!S_ISDIR(e->versions[1].mode)) {
1609 e->tree = new_tree_content(8);
1610 e->versions[1].mode = S_IFDIR;
1612 if (!e->tree)
1613 load_tree(e);
1614 if (tree_content_set(e, slash1 + 1, oid, mode, subtree)) {
1615 oidclr(&root->versions[1].oid);
1616 return 1;
1618 return 0;
1622 if (t->entry_count == t->entry_capacity)
1623 root->tree = t = grow_tree_content(t, t->entry_count);
1624 e = new_tree_entry();
1625 e->name = to_atom(p, n);
1626 e->versions[0].mode = 0;
1627 oidclr(&e->versions[0].oid);
1628 t->entries[t->entry_count++] = e;
1629 if (*slash1) {
1630 e->tree = new_tree_content(8);
1631 e->versions[1].mode = S_IFDIR;
1632 tree_content_set(e, slash1 + 1, oid, mode, subtree);
1633 } else {
1634 e->tree = subtree;
1635 e->versions[1].mode = mode;
1636 oidcpy(&e->versions[1].oid, oid);
1638 oidclr(&root->versions[1].oid);
1639 return 1;
1642 static int tree_content_remove(
1643 struct tree_entry *root,
1644 const char *p,
1645 struct tree_entry *backup_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;
1656 if (!root->tree)
1657 load_tree(root);
1659 if (!*p && allow_root) {
1660 e = root;
1661 goto del_entry;
1664 t = root->tree;
1665 for (i = 0; i < t->entry_count; i++) {
1666 e = t->entries[i];
1667 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1668 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1670 * If p names a file in some subdirectory, and a
1671 * file or symlink matching the name of the
1672 * parent directory of p exists, then p cannot
1673 * exist and need not be deleted.
1675 return 1;
1676 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1677 goto del_entry;
1678 if (!e->tree)
1679 load_tree(e);
1680 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1681 for (n = 0; n < e->tree->entry_count; n++) {
1682 if (e->tree->entries[n]->versions[1].mode) {
1683 oidclr(&root->versions[1].oid);
1684 return 1;
1687 backup_leaf = NULL;
1688 goto del_entry;
1690 return 0;
1693 return 0;
1695 del_entry:
1696 if (backup_leaf)
1697 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1698 else if (e->tree)
1699 release_tree_content_recursive(e->tree);
1700 e->tree = NULL;
1701 e->versions[1].mode = 0;
1702 oidclr(&e->versions[1].oid);
1703 oidclr(&root->versions[1].oid);
1704 return 1;
1707 static int tree_content_get(
1708 struct tree_entry *root,
1709 const char *p,
1710 struct tree_entry *leaf,
1711 int allow_root)
1713 struct tree_content *t;
1714 const char *slash1;
1715 unsigned int i, n;
1716 struct tree_entry *e;
1718 slash1 = strchrnul(p, '/');
1719 n = slash1 - p;
1720 if (!n && !allow_root)
1721 die("Empty path component found in input");
1723 if (!root->tree)
1724 load_tree(root);
1726 if (!n) {
1727 e = root;
1728 goto found_entry;
1731 t = root->tree;
1732 for (i = 0; i < t->entry_count; i++) {
1733 e = t->entries[i];
1734 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1735 if (!*slash1)
1736 goto found_entry;
1737 if (!S_ISDIR(e->versions[1].mode))
1738 return 0;
1739 if (!e->tree)
1740 load_tree(e);
1741 return tree_content_get(e, slash1 + 1, leaf, 0);
1744 return 0;
1746 found_entry:
1747 memcpy(leaf, e, sizeof(*leaf));
1748 if (e->tree && is_null_oid(&e->versions[1].oid))
1749 leaf->tree = dup_tree_content(e->tree);
1750 else
1751 leaf->tree = NULL;
1752 return 1;
1755 static int update_branch(struct branch *b)
1757 static const char *msg = "fast-import";
1758 struct ref_transaction *transaction;
1759 struct object_id old_oid;
1760 struct strbuf err = STRBUF_INIT;
1762 if (is_null_oid(&b->oid)) {
1763 if (b->delete)
1764 delete_ref(NULL, b->name, NULL, 0);
1765 return 0;
1767 if (read_ref(b->name, &old_oid))
1768 oidclr(&old_oid);
1769 if (!force_update && !is_null_oid(&old_oid)) {
1770 struct commit *old_cmit, *new_cmit;
1772 old_cmit = lookup_commit_reference_gently(&old_oid, 0);
1773 new_cmit = lookup_commit_reference_gently(&b->oid, 0);
1774 if (!old_cmit || !new_cmit)
1775 return error("Branch %s is missing commits.", b->name);
1777 if (!in_merge_bases(old_cmit, new_cmit)) {
1778 warning("Not updating %s"
1779 " (new tip %s does not contain %s)",
1780 b->name, oid_to_hex(&b->oid),
1781 oid_to_hex(&old_oid));
1782 return -1;
1785 transaction = ref_transaction_begin(&err);
1786 if (!transaction ||
1787 ref_transaction_update(transaction, b->name, &b->oid, &old_oid,
1788 0, msg, &err) ||
1789 ref_transaction_commit(transaction, &err)) {
1790 ref_transaction_free(transaction);
1791 error("%s", err.buf);
1792 strbuf_release(&err);
1793 return -1;
1795 ref_transaction_free(transaction);
1796 strbuf_release(&err);
1797 return 0;
1800 static void dump_branches(void)
1802 unsigned int i;
1803 struct branch *b;
1805 for (i = 0; i < branch_table_sz; i++) {
1806 for (b = branch_table[i]; b; b = b->table_next_branch)
1807 failure |= update_branch(b);
1811 static void dump_tags(void)
1813 static const char *msg = "fast-import";
1814 struct tag *t;
1815 struct strbuf ref_name = STRBUF_INIT;
1816 struct strbuf err = STRBUF_INIT;
1817 struct ref_transaction *transaction;
1819 transaction = ref_transaction_begin(&err);
1820 if (!transaction) {
1821 failure |= error("%s", err.buf);
1822 goto cleanup;
1824 for (t = first_tag; t; t = t->next_tag) {
1825 strbuf_reset(&ref_name);
1826 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1828 if (ref_transaction_update(transaction, ref_name.buf,
1829 &t->oid, NULL, 0, msg, &err)) {
1830 failure |= error("%s", err.buf);
1831 goto cleanup;
1834 if (ref_transaction_commit(transaction, &err))
1835 failure |= error("%s", err.buf);
1837 cleanup:
1838 ref_transaction_free(transaction);
1839 strbuf_release(&ref_name);
1840 strbuf_release(&err);
1843 static void dump_marks_helper(FILE *f,
1844 uintmax_t base,
1845 struct mark_set *m)
1847 uintmax_t k;
1848 if (m->shift) {
1849 for (k = 0; k < 1024; k++) {
1850 if (m->data.sets[k])
1851 dump_marks_helper(f, base + (k << m->shift),
1852 m->data.sets[k]);
1854 } else {
1855 for (k = 0; k < 1024; k++) {
1856 if (m->data.marked[k])
1857 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1858 oid_to_hex(&m->data.marked[k]->idx.oid));
1863 static void dump_marks(void)
1865 static struct lock_file mark_lock;
1866 FILE *f;
1868 if (!export_marks_file || (import_marks_file && !import_marks_file_done))
1869 return;
1871 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1872 failure |= error_errno("Unable to write marks file %s",
1873 export_marks_file);
1874 return;
1877 f = fdopen_lock_file(&mark_lock, "w");
1878 if (!f) {
1879 int saved_errno = errno;
1880 rollback_lock_file(&mark_lock);
1881 failure |= error("Unable to write marks file %s: %s",
1882 export_marks_file, strerror(saved_errno));
1883 return;
1886 dump_marks_helper(f, 0, marks);
1887 if (commit_lock_file(&mark_lock)) {
1888 failure |= error_errno("Unable to write file %s",
1889 export_marks_file);
1890 return;
1894 static void read_marks(void)
1896 char line[512];
1897 FILE *f = fopen(import_marks_file, "r");
1898 if (f)
1900 else if (import_marks_file_ignore_missing && errno == ENOENT)
1901 goto done; /* Marks file does not exist */
1902 else
1903 die_errno("cannot read '%s'", import_marks_file);
1904 while (fgets(line, sizeof(line), f)) {
1905 uintmax_t mark;
1906 char *end;
1907 struct object_id oid;
1908 struct object_entry *e;
1910 end = strchr(line, '\n');
1911 if (line[0] != ':' || !end)
1912 die("corrupt mark line: %s", line);
1913 *end = 0;
1914 mark = strtoumax(line + 1, &end, 10);
1915 if (!mark || end == line + 1
1916 || *end != ' ' || get_oid_hex(end + 1, &oid))
1917 die("corrupt mark line: %s", line);
1918 e = find_object(&oid);
1919 if (!e) {
1920 enum object_type type = oid_object_info(&oid, NULL);
1921 if (type < 0)
1922 die("object not found: %s", oid_to_hex(&oid));
1923 e = insert_object(&oid);
1924 e->type = type;
1925 e->pack_id = MAX_PACK_ID;
1926 e->idx.offset = 1; /* just not zero! */
1928 insert_mark(mark, e);
1930 fclose(f);
1931 done:
1932 import_marks_file_done = 1;
1936 static int read_next_command(void)
1938 static int stdin_eof = 0;
1940 if (stdin_eof) {
1941 unread_command_buf = 0;
1942 return EOF;
1945 for (;;) {
1946 const char *p;
1948 if (unread_command_buf) {
1949 unread_command_buf = 0;
1950 } else {
1951 struct recent_command *rc;
1953 strbuf_detach(&command_buf, NULL);
1954 stdin_eof = strbuf_getline_lf(&command_buf, stdin);
1955 if (stdin_eof)
1956 return EOF;
1958 if (!seen_data_command
1959 && !starts_with(command_buf.buf, "feature ")
1960 && !starts_with(command_buf.buf, "option ")) {
1961 parse_argv();
1964 rc = rc_free;
1965 if (rc)
1966 rc_free = rc->next;
1967 else {
1968 rc = cmd_hist.next;
1969 cmd_hist.next = rc->next;
1970 cmd_hist.next->prev = &cmd_hist;
1971 free(rc->buf);
1974 rc->buf = command_buf.buf;
1975 rc->prev = cmd_tail;
1976 rc->next = cmd_hist.prev;
1977 rc->prev->next = rc;
1978 cmd_tail = rc;
1980 if (skip_prefix(command_buf.buf, "get-mark ", &p)) {
1981 parse_get_mark(p);
1982 continue;
1984 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1985 parse_cat_blob(p);
1986 continue;
1988 if (command_buf.buf[0] == '#')
1989 continue;
1990 return 0;
1994 static void skip_optional_lf(void)
1996 int term_char = fgetc(stdin);
1997 if (term_char != '\n' && term_char != EOF)
1998 ungetc(term_char, stdin);
2001 static void parse_mark(void)
2003 const char *v;
2004 if (skip_prefix(command_buf.buf, "mark :", &v)) {
2005 next_mark = strtoumax(v, NULL, 10);
2006 read_next_command();
2008 else
2009 next_mark = 0;
2012 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
2014 const char *data;
2015 strbuf_reset(sb);
2017 if (!skip_prefix(command_buf.buf, "data ", &data))
2018 die("Expected 'data n' command, found: %s", command_buf.buf);
2020 if (skip_prefix(data, "<<", &data)) {
2021 char *term = xstrdup(data);
2022 size_t term_len = command_buf.len - (data - command_buf.buf);
2024 strbuf_detach(&command_buf, NULL);
2025 for (;;) {
2026 if (strbuf_getline_lf(&command_buf, stdin) == EOF)
2027 die("EOF in data (terminator '%s' not found)", term);
2028 if (term_len == command_buf.len
2029 && !strcmp(term, command_buf.buf))
2030 break;
2031 strbuf_addbuf(sb, &command_buf);
2032 strbuf_addch(sb, '\n');
2034 free(term);
2036 else {
2037 uintmax_t len = strtoumax(data, NULL, 10);
2038 size_t n = 0, length = (size_t)len;
2040 if (limit && limit < len) {
2041 *len_res = len;
2042 return 0;
2044 if (length < len)
2045 die("data is too large to use in this context");
2047 while (n < length) {
2048 size_t s = strbuf_fread(sb, length - n, stdin);
2049 if (!s && feof(stdin))
2050 die("EOF in data (%lu bytes remaining)",
2051 (unsigned long)(length - n));
2052 n += s;
2056 skip_optional_lf();
2057 return 1;
2060 static int validate_raw_date(const char *src, struct strbuf *result)
2062 const char *orig_src = src;
2063 char *endp;
2064 unsigned long num;
2066 errno = 0;
2068 num = strtoul(src, &endp, 10);
2069 /* NEEDSWORK: perhaps check for reasonable values? */
2070 if (errno || endp == src || *endp != ' ')
2071 return -1;
2073 src = endp + 1;
2074 if (*src != '-' && *src != '+')
2075 return -1;
2077 num = strtoul(src + 1, &endp, 10);
2078 if (errno || endp == src + 1 || *endp || 1400 < num)
2079 return -1;
2081 strbuf_addstr(result, orig_src);
2082 return 0;
2085 static char *parse_ident(const char *buf)
2087 const char *ltgt;
2088 size_t name_len;
2089 struct strbuf ident = STRBUF_INIT;
2091 /* ensure there is a space delimiter even if there is no name */
2092 if (*buf == '<')
2093 --buf;
2095 ltgt = buf + strcspn(buf, "<>");
2096 if (*ltgt != '<')
2097 die("Missing < in ident string: %s", buf);
2098 if (ltgt != buf && ltgt[-1] != ' ')
2099 die("Missing space before < in ident string: %s", buf);
2100 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2101 if (*ltgt != '>')
2102 die("Missing > in ident string: %s", buf);
2103 ltgt++;
2104 if (*ltgt != ' ')
2105 die("Missing space after > in ident string: %s", buf);
2106 ltgt++;
2107 name_len = ltgt - buf;
2108 strbuf_add(&ident, buf, name_len);
2110 switch (whenspec) {
2111 case WHENSPEC_RAW:
2112 if (validate_raw_date(ltgt, &ident) < 0)
2113 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2114 break;
2115 case WHENSPEC_RFC2822:
2116 if (parse_date(ltgt, &ident) < 0)
2117 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2118 break;
2119 case WHENSPEC_NOW:
2120 if (strcmp("now", ltgt))
2121 die("Date in ident must be 'now': %s", buf);
2122 datestamp(&ident);
2123 break;
2126 return strbuf_detach(&ident, NULL);
2129 static void parse_and_store_blob(
2130 struct last_object *last,
2131 struct object_id *oidout,
2132 uintmax_t mark)
2134 static struct strbuf buf = STRBUF_INIT;
2135 uintmax_t len;
2137 if (parse_data(&buf, big_file_threshold, &len))
2138 store_object(OBJ_BLOB, &buf, last, oidout, mark);
2139 else {
2140 if (last) {
2141 strbuf_release(&last->data);
2142 last->offset = 0;
2143 last->depth = 0;
2145 stream_blob(len, oidout, mark);
2146 skip_optional_lf();
2150 static void parse_new_blob(void)
2152 read_next_command();
2153 parse_mark();
2154 parse_and_store_blob(&last_blob, NULL, next_mark);
2157 static void unload_one_branch(void)
2159 while (cur_active_branches
2160 && cur_active_branches >= max_active_branches) {
2161 uintmax_t min_commit = ULONG_MAX;
2162 struct branch *e, *l = NULL, *p = NULL;
2164 for (e = active_branches; e; e = e->active_next_branch) {
2165 if (e->last_commit < min_commit) {
2166 p = l;
2167 min_commit = e->last_commit;
2169 l = e;
2172 if (p) {
2173 e = p->active_next_branch;
2174 p->active_next_branch = e->active_next_branch;
2175 } else {
2176 e = active_branches;
2177 active_branches = e->active_next_branch;
2179 e->active = 0;
2180 e->active_next_branch = NULL;
2181 if (e->branch_tree.tree) {
2182 release_tree_content_recursive(e->branch_tree.tree);
2183 e->branch_tree.tree = NULL;
2185 cur_active_branches--;
2189 static void load_branch(struct branch *b)
2191 load_tree(&b->branch_tree);
2192 if (!b->active) {
2193 b->active = 1;
2194 b->active_next_branch = active_branches;
2195 active_branches = b;
2196 cur_active_branches++;
2197 branch_load_count++;
2201 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2203 unsigned char fanout = 0;
2204 while ((num_notes >>= 8))
2205 fanout++;
2206 return fanout;
2209 static void construct_path_with_fanout(const char *hex_sha1,
2210 unsigned char fanout, char *path)
2212 unsigned int i = 0, j = 0;
2213 if (fanout >= the_hash_algo->rawsz)
2214 die("Too large fanout (%u)", fanout);
2215 while (fanout) {
2216 path[i++] = hex_sha1[j++];
2217 path[i++] = hex_sha1[j++];
2218 path[i++] = '/';
2219 fanout--;
2221 memcpy(path + i, hex_sha1 + j, the_hash_algo->hexsz - j);
2222 path[i + the_hash_algo->hexsz - j] = '\0';
2225 static uintmax_t do_change_note_fanout(
2226 struct tree_entry *orig_root, struct tree_entry *root,
2227 char *hex_oid, unsigned int hex_oid_len,
2228 char *fullpath, unsigned int fullpath_len,
2229 unsigned char fanout)
2231 struct tree_content *t;
2232 struct tree_entry *e, leaf;
2233 unsigned int i, tmp_hex_oid_len, tmp_fullpath_len;
2234 uintmax_t num_notes = 0;
2235 struct object_id oid;
2236 char realpath[60];
2238 if (!root->tree)
2239 load_tree(root);
2240 t = root->tree;
2242 for (i = 0; t && i < t->entry_count; i++) {
2243 e = t->entries[i];
2244 tmp_hex_oid_len = hex_oid_len + e->name->str_len;
2245 tmp_fullpath_len = fullpath_len;
2248 * We're interested in EITHER existing note entries (entries
2249 * with exactly 40 hex chars in path, not including directory
2250 * separators), OR directory entries that may contain note
2251 * entries (with < 40 hex chars in path).
2252 * Also, each path component in a note entry must be a multiple
2253 * of 2 chars.
2255 if (!e->versions[1].mode ||
2256 tmp_hex_oid_len > GIT_SHA1_HEXSZ ||
2257 e->name->str_len % 2)
2258 continue;
2260 /* This _may_ be a note entry, or a subdir containing notes */
2261 memcpy(hex_oid + hex_oid_len, e->name->str_dat,
2262 e->name->str_len);
2263 if (tmp_fullpath_len)
2264 fullpath[tmp_fullpath_len++] = '/';
2265 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2266 e->name->str_len);
2267 tmp_fullpath_len += e->name->str_len;
2268 fullpath[tmp_fullpath_len] = '\0';
2270 if (tmp_hex_oid_len == GIT_SHA1_HEXSZ && !get_oid_hex(hex_oid, &oid)) {
2271 /* This is a note entry */
2272 if (fanout == 0xff) {
2273 /* Counting mode, no rename */
2274 num_notes++;
2275 continue;
2277 construct_path_with_fanout(hex_oid, fanout, realpath);
2278 if (!strcmp(fullpath, realpath)) {
2279 /* Note entry is in correct location */
2280 num_notes++;
2281 continue;
2284 /* Rename fullpath to realpath */
2285 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2286 die("Failed to remove path %s", fullpath);
2287 tree_content_set(orig_root, realpath,
2288 &leaf.versions[1].oid,
2289 leaf.versions[1].mode,
2290 leaf.tree);
2291 } else if (S_ISDIR(e->versions[1].mode)) {
2292 /* This is a subdir that may contain note entries */
2293 num_notes += do_change_note_fanout(orig_root, e,
2294 hex_oid, tmp_hex_oid_len,
2295 fullpath, tmp_fullpath_len, fanout);
2298 /* The above may have reallocated the current tree_content */
2299 t = root->tree;
2301 return num_notes;
2304 static uintmax_t change_note_fanout(struct tree_entry *root,
2305 unsigned char fanout)
2308 * The size of path is due to one slash between every two hex digits,
2309 * plus the terminating NUL. Note that there is no slash at the end, so
2310 * the number of slashes is one less than half the number of hex
2311 * characters.
2313 char hex_oid[GIT_MAX_HEXSZ], path[GIT_MAX_HEXSZ + (GIT_MAX_HEXSZ / 2) - 1 + 1];
2314 return do_change_note_fanout(root, root, hex_oid, 0, path, 0, fanout);
2318 * Given a pointer into a string, parse a mark reference:
2320 * idnum ::= ':' bigint;
2322 * Return the first character after the value in *endptr.
2324 * Complain if the following character is not what is expected,
2325 * either a space or end of the string.
2327 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2329 uintmax_t mark;
2331 assert(*p == ':');
2332 p++;
2333 mark = strtoumax(p, endptr, 10);
2334 if (*endptr == p)
2335 die("No value after ':' in mark: %s", command_buf.buf);
2336 return mark;
2340 * Parse the mark reference, and complain if this is not the end of
2341 * the string.
2343 static uintmax_t parse_mark_ref_eol(const char *p)
2345 char *end;
2346 uintmax_t mark;
2348 mark = parse_mark_ref(p, &end);
2349 if (*end != '\0')
2350 die("Garbage after mark: %s", command_buf.buf);
2351 return mark;
2355 * Parse the mark reference, demanding a trailing space. Return a
2356 * pointer to the space.
2358 static uintmax_t parse_mark_ref_space(const char **p)
2360 uintmax_t mark;
2361 char *end;
2363 mark = parse_mark_ref(*p, &end);
2364 if (*end++ != ' ')
2365 die("Missing space after mark: %s", command_buf.buf);
2366 *p = end;
2367 return mark;
2370 static void file_change_m(const char *p, struct branch *b)
2372 static struct strbuf uq = STRBUF_INIT;
2373 const char *endp;
2374 struct object_entry *oe;
2375 struct object_id oid;
2376 uint16_t mode, inline_data = 0;
2378 p = get_mode(p, &mode);
2379 if (!p)
2380 die("Corrupt mode: %s", command_buf.buf);
2381 switch (mode) {
2382 case 0644:
2383 case 0755:
2384 mode |= S_IFREG;
2385 case S_IFREG | 0644:
2386 case S_IFREG | 0755:
2387 case S_IFLNK:
2388 case S_IFDIR:
2389 case S_IFGITLINK:
2390 /* ok */
2391 break;
2392 default:
2393 die("Corrupt mode: %s", command_buf.buf);
2396 if (*p == ':') {
2397 oe = find_mark(parse_mark_ref_space(&p));
2398 oidcpy(&oid, &oe->idx.oid);
2399 } else if (skip_prefix(p, "inline ", &p)) {
2400 inline_data = 1;
2401 oe = NULL; /* not used with inline_data, but makes gcc happy */
2402 } else {
2403 if (parse_oid_hex(p, &oid, &p))
2404 die("Invalid dataref: %s", command_buf.buf);
2405 oe = find_object(&oid);
2406 if (*p++ != ' ')
2407 die("Missing space after SHA1: %s", command_buf.buf);
2410 strbuf_reset(&uq);
2411 if (!unquote_c_style(&uq, p, &endp)) {
2412 if (*endp)
2413 die("Garbage after path in: %s", command_buf.buf);
2414 p = uq.buf;
2417 /* Git does not track empty, non-toplevel directories. */
2418 if (S_ISDIR(mode) && is_empty_tree_oid(&oid) && *p) {
2419 tree_content_remove(&b->branch_tree, p, NULL, 0);
2420 return;
2423 if (S_ISGITLINK(mode)) {
2424 if (inline_data)
2425 die("Git links cannot be specified 'inline': %s",
2426 command_buf.buf);
2427 else if (oe) {
2428 if (oe->type != OBJ_COMMIT)
2429 die("Not a commit (actually a %s): %s",
2430 type_name(oe->type), command_buf.buf);
2433 * Accept the sha1 without checking; it expected to be in
2434 * another repository.
2436 } else if (inline_data) {
2437 if (S_ISDIR(mode))
2438 die("Directories cannot be specified 'inline': %s",
2439 command_buf.buf);
2440 if (p != uq.buf) {
2441 strbuf_addstr(&uq, p);
2442 p = uq.buf;
2444 read_next_command();
2445 parse_and_store_blob(&last_blob, &oid, 0);
2446 } else {
2447 enum object_type expected = S_ISDIR(mode) ?
2448 OBJ_TREE: OBJ_BLOB;
2449 enum object_type type = oe ? oe->type :
2450 oid_object_info(&oid, NULL);
2451 if (type < 0)
2452 die("%s not found: %s",
2453 S_ISDIR(mode) ? "Tree" : "Blob",
2454 command_buf.buf);
2455 if (type != expected)
2456 die("Not a %s (actually a %s): %s",
2457 type_name(expected), type_name(type),
2458 command_buf.buf);
2461 if (!*p) {
2462 tree_content_replace(&b->branch_tree, &oid, mode, NULL);
2463 return;
2465 tree_content_set(&b->branch_tree, p, &oid, mode, NULL);
2468 static void file_change_d(const char *p, struct branch *b)
2470 static struct strbuf uq = STRBUF_INIT;
2471 const char *endp;
2473 strbuf_reset(&uq);
2474 if (!unquote_c_style(&uq, p, &endp)) {
2475 if (*endp)
2476 die("Garbage after path in: %s", command_buf.buf);
2477 p = uq.buf;
2479 tree_content_remove(&b->branch_tree, p, NULL, 1);
2482 static void file_change_cr(const char *s, struct branch *b, int rename)
2484 const char *d;
2485 static struct strbuf s_uq = STRBUF_INIT;
2486 static struct strbuf d_uq = STRBUF_INIT;
2487 const char *endp;
2488 struct tree_entry leaf;
2490 strbuf_reset(&s_uq);
2491 if (!unquote_c_style(&s_uq, s, &endp)) {
2492 if (*endp != ' ')
2493 die("Missing space after source: %s", command_buf.buf);
2494 } else {
2495 endp = strchr(s, ' ');
2496 if (!endp)
2497 die("Missing space after source: %s", command_buf.buf);
2498 strbuf_add(&s_uq, s, endp - s);
2500 s = s_uq.buf;
2502 endp++;
2503 if (!*endp)
2504 die("Missing dest: %s", command_buf.buf);
2506 d = endp;
2507 strbuf_reset(&d_uq);
2508 if (!unquote_c_style(&d_uq, d, &endp)) {
2509 if (*endp)
2510 die("Garbage after dest in: %s", command_buf.buf);
2511 d = d_uq.buf;
2514 memset(&leaf, 0, sizeof(leaf));
2515 if (rename)
2516 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2517 else
2518 tree_content_get(&b->branch_tree, s, &leaf, 1);
2519 if (!leaf.versions[1].mode)
2520 die("Path %s not in branch", s);
2521 if (!*d) { /* C "path/to/subdir" "" */
2522 tree_content_replace(&b->branch_tree,
2523 &leaf.versions[1].oid,
2524 leaf.versions[1].mode,
2525 leaf.tree);
2526 return;
2528 tree_content_set(&b->branch_tree, d,
2529 &leaf.versions[1].oid,
2530 leaf.versions[1].mode,
2531 leaf.tree);
2534 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2536 static struct strbuf uq = STRBUF_INIT;
2537 struct object_entry *oe;
2538 struct branch *s;
2539 struct object_id oid, commit_oid;
2540 char path[60];
2541 uint16_t inline_data = 0;
2542 unsigned char new_fanout;
2545 * When loading a branch, we don't traverse its tree to count the real
2546 * number of notes (too expensive to do this for all non-note refs).
2547 * This means that recently loaded notes refs might incorrectly have
2548 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2550 * Fix this by traversing the tree and counting the number of notes
2551 * when b->num_notes == 0. If the notes tree is truly empty, the
2552 * calculation should not take long.
2554 if (b->num_notes == 0 && *old_fanout == 0) {
2555 /* Invoke change_note_fanout() in "counting mode". */
2556 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2557 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2560 /* Now parse the notemodify command. */
2561 /* <dataref> or 'inline' */
2562 if (*p == ':') {
2563 oe = find_mark(parse_mark_ref_space(&p));
2564 oidcpy(&oid, &oe->idx.oid);
2565 } else if (skip_prefix(p, "inline ", &p)) {
2566 inline_data = 1;
2567 oe = NULL; /* not used with inline_data, but makes gcc happy */
2568 } else {
2569 if (parse_oid_hex(p, &oid, &p))
2570 die("Invalid dataref: %s", command_buf.buf);
2571 oe = find_object(&oid);
2572 if (*p++ != ' ')
2573 die("Missing space after SHA1: %s", command_buf.buf);
2576 /* <commit-ish> */
2577 s = lookup_branch(p);
2578 if (s) {
2579 if (is_null_oid(&s->oid))
2580 die("Can't add a note on empty branch.");
2581 oidcpy(&commit_oid, &s->oid);
2582 } else if (*p == ':') {
2583 uintmax_t commit_mark = parse_mark_ref_eol(p);
2584 struct object_entry *commit_oe = find_mark(commit_mark);
2585 if (commit_oe->type != OBJ_COMMIT)
2586 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2587 oidcpy(&commit_oid, &commit_oe->idx.oid);
2588 } else if (!get_oid(p, &commit_oid)) {
2589 unsigned long size;
2590 char *buf = read_object_with_reference(&commit_oid,
2591 commit_type, &size,
2592 &commit_oid);
2593 if (!buf || size < 46)
2594 die("Not a valid commit: %s", p);
2595 free(buf);
2596 } else
2597 die("Invalid ref name or SHA1 expression: %s", p);
2599 if (inline_data) {
2600 if (p != uq.buf) {
2601 strbuf_addstr(&uq, p);
2602 p = uq.buf;
2604 read_next_command();
2605 parse_and_store_blob(&last_blob, &oid, 0);
2606 } else if (oe) {
2607 if (oe->type != OBJ_BLOB)
2608 die("Not a blob (actually a %s): %s",
2609 type_name(oe->type), command_buf.buf);
2610 } else if (!is_null_oid(&oid)) {
2611 enum object_type type = oid_object_info(&oid, NULL);
2612 if (type < 0)
2613 die("Blob not found: %s", command_buf.buf);
2614 if (type != OBJ_BLOB)
2615 die("Not a blob (actually a %s): %s",
2616 type_name(type), command_buf.buf);
2619 construct_path_with_fanout(oid_to_hex(&commit_oid), *old_fanout, path);
2620 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2621 b->num_notes--;
2623 if (is_null_oid(&oid))
2624 return; /* nothing to insert */
2626 b->num_notes++;
2627 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2628 construct_path_with_fanout(oid_to_hex(&commit_oid), new_fanout, path);
2629 tree_content_set(&b->branch_tree, path, &oid, S_IFREG | 0644, NULL);
2632 static void file_change_deleteall(struct branch *b)
2634 release_tree_content_recursive(b->branch_tree.tree);
2635 oidclr(&b->branch_tree.versions[0].oid);
2636 oidclr(&b->branch_tree.versions[1].oid);
2637 load_tree(&b->branch_tree);
2638 b->num_notes = 0;
2641 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2643 if (!buf || size < GIT_SHA1_HEXSZ + 6)
2644 die("Not a valid commit: %s", oid_to_hex(&b->oid));
2645 if (memcmp("tree ", buf, 5)
2646 || get_oid_hex(buf + 5, &b->branch_tree.versions[1].oid))
2647 die("The commit %s is corrupt", oid_to_hex(&b->oid));
2648 oidcpy(&b->branch_tree.versions[0].oid,
2649 &b->branch_tree.versions[1].oid);
2652 static void parse_from_existing(struct branch *b)
2654 if (is_null_oid(&b->oid)) {
2655 oidclr(&b->branch_tree.versions[0].oid);
2656 oidclr(&b->branch_tree.versions[1].oid);
2657 } else {
2658 unsigned long size;
2659 char *buf;
2661 buf = read_object_with_reference(&b->oid, commit_type, &size,
2662 &b->oid);
2663 parse_from_commit(b, buf, size);
2664 free(buf);
2668 static int parse_from(struct branch *b)
2670 const char *from;
2671 struct branch *s;
2672 struct object_id oid;
2674 if (!skip_prefix(command_buf.buf, "from ", &from))
2675 return 0;
2677 oidcpy(&oid, &b->branch_tree.versions[1].oid);
2679 s = lookup_branch(from);
2680 if (b == s)
2681 die("Can't create a branch from itself: %s", b->name);
2682 else if (s) {
2683 struct object_id *t = &s->branch_tree.versions[1].oid;
2684 oidcpy(&b->oid, &s->oid);
2685 oidcpy(&b->branch_tree.versions[0].oid, t);
2686 oidcpy(&b->branch_tree.versions[1].oid, t);
2687 } else if (*from == ':') {
2688 uintmax_t idnum = parse_mark_ref_eol(from);
2689 struct object_entry *oe = find_mark(idnum);
2690 if (oe->type != OBJ_COMMIT)
2691 die("Mark :%" PRIuMAX " not a commit", idnum);
2692 if (oidcmp(&b->oid, &oe->idx.oid)) {
2693 oidcpy(&b->oid, &oe->idx.oid);
2694 if (oe->pack_id != MAX_PACK_ID) {
2695 unsigned long size;
2696 char *buf = gfi_unpack_entry(oe, &size);
2697 parse_from_commit(b, buf, size);
2698 free(buf);
2699 } else
2700 parse_from_existing(b);
2702 } else if (!get_oid(from, &b->oid)) {
2703 parse_from_existing(b);
2704 if (is_null_oid(&b->oid))
2705 b->delete = 1;
2707 else
2708 die("Invalid ref name or SHA1 expression: %s", from);
2710 if (b->branch_tree.tree && oidcmp(&oid, &b->branch_tree.versions[1].oid)) {
2711 release_tree_content_recursive(b->branch_tree.tree);
2712 b->branch_tree.tree = NULL;
2715 read_next_command();
2716 return 1;
2719 static struct hash_list *parse_merge(unsigned int *count)
2721 struct hash_list *list = NULL, **tail = &list, *n;
2722 const char *from;
2723 struct branch *s;
2725 *count = 0;
2726 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2727 n = xmalloc(sizeof(*n));
2728 s = lookup_branch(from);
2729 if (s)
2730 oidcpy(&n->oid, &s->oid);
2731 else if (*from == ':') {
2732 uintmax_t idnum = parse_mark_ref_eol(from);
2733 struct object_entry *oe = find_mark(idnum);
2734 if (oe->type != OBJ_COMMIT)
2735 die("Mark :%" PRIuMAX " not a commit", idnum);
2736 oidcpy(&n->oid, &oe->idx.oid);
2737 } else if (!get_oid(from, &n->oid)) {
2738 unsigned long size;
2739 char *buf = read_object_with_reference(&n->oid,
2740 commit_type,
2741 &size, &n->oid);
2742 if (!buf || size < 46)
2743 die("Not a valid commit: %s", from);
2744 free(buf);
2745 } else
2746 die("Invalid ref name or SHA1 expression: %s", from);
2748 n->next = NULL;
2749 *tail = n;
2750 tail = &n->next;
2752 (*count)++;
2753 read_next_command();
2755 return list;
2758 static void parse_new_commit(const char *arg)
2760 static struct strbuf msg = STRBUF_INIT;
2761 struct branch *b;
2762 char *author = NULL;
2763 char *committer = NULL;
2764 struct hash_list *merge_list = NULL;
2765 unsigned int merge_count;
2766 unsigned char prev_fanout, new_fanout;
2767 const char *v;
2769 b = lookup_branch(arg);
2770 if (!b)
2771 b = new_branch(arg);
2773 read_next_command();
2774 parse_mark();
2775 if (skip_prefix(command_buf.buf, "author ", &v)) {
2776 author = parse_ident(v);
2777 read_next_command();
2779 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2780 committer = parse_ident(v);
2781 read_next_command();
2783 if (!committer)
2784 die("Expected committer but didn't get one");
2785 parse_data(&msg, 0, NULL);
2786 read_next_command();
2787 parse_from(b);
2788 merge_list = parse_merge(&merge_count);
2790 /* ensure the branch is active/loaded */
2791 if (!b->branch_tree.tree || !max_active_branches) {
2792 unload_one_branch();
2793 load_branch(b);
2796 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2798 /* file_change* */
2799 while (command_buf.len > 0) {
2800 if (skip_prefix(command_buf.buf, "M ", &v))
2801 file_change_m(v, b);
2802 else if (skip_prefix(command_buf.buf, "D ", &v))
2803 file_change_d(v, b);
2804 else if (skip_prefix(command_buf.buf, "R ", &v))
2805 file_change_cr(v, b, 1);
2806 else if (skip_prefix(command_buf.buf, "C ", &v))
2807 file_change_cr(v, b, 0);
2808 else if (skip_prefix(command_buf.buf, "N ", &v))
2809 note_change_n(v, b, &prev_fanout);
2810 else if (!strcmp("deleteall", command_buf.buf))
2811 file_change_deleteall(b);
2812 else if (skip_prefix(command_buf.buf, "ls ", &v))
2813 parse_ls(v, b);
2814 else {
2815 unread_command_buf = 1;
2816 break;
2818 if (read_next_command() == EOF)
2819 break;
2822 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2823 if (new_fanout != prev_fanout)
2824 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2826 /* build the tree and the commit */
2827 store_tree(&b->branch_tree);
2828 oidcpy(&b->branch_tree.versions[0].oid,
2829 &b->branch_tree.versions[1].oid);
2831 strbuf_reset(&new_data);
2832 strbuf_addf(&new_data, "tree %s\n",
2833 oid_to_hex(&b->branch_tree.versions[1].oid));
2834 if (!is_null_oid(&b->oid))
2835 strbuf_addf(&new_data, "parent %s\n",
2836 oid_to_hex(&b->oid));
2837 while (merge_list) {
2838 struct hash_list *next = merge_list->next;
2839 strbuf_addf(&new_data, "parent %s\n",
2840 oid_to_hex(&merge_list->oid));
2841 free(merge_list);
2842 merge_list = next;
2844 strbuf_addf(&new_data,
2845 "author %s\n"
2846 "committer %s\n"
2847 "\n",
2848 author ? author : committer, committer);
2849 strbuf_addbuf(&new_data, &msg);
2850 free(author);
2851 free(committer);
2853 if (!store_object(OBJ_COMMIT, &new_data, NULL, &b->oid, next_mark))
2854 b->pack_id = pack_id;
2855 b->last_commit = object_count_by_type[OBJ_COMMIT];
2858 static void parse_new_tag(const char *arg)
2860 static struct strbuf msg = STRBUF_INIT;
2861 const char *from;
2862 char *tagger;
2863 struct branch *s;
2864 struct tag *t;
2865 uintmax_t from_mark = 0;
2866 struct object_id oid;
2867 enum object_type type;
2868 const char *v;
2870 t = pool_alloc(sizeof(struct tag));
2871 memset(t, 0, sizeof(struct tag));
2872 t->name = pool_strdup(arg);
2873 if (last_tag)
2874 last_tag->next_tag = t;
2875 else
2876 first_tag = t;
2877 last_tag = t;
2878 read_next_command();
2880 /* from ... */
2881 if (!skip_prefix(command_buf.buf, "from ", &from))
2882 die("Expected from command, got %s", command_buf.buf);
2883 s = lookup_branch(from);
2884 if (s) {
2885 if (is_null_oid(&s->oid))
2886 die("Can't tag an empty branch.");
2887 oidcpy(&oid, &s->oid);
2888 type = OBJ_COMMIT;
2889 } else if (*from == ':') {
2890 struct object_entry *oe;
2891 from_mark = parse_mark_ref_eol(from);
2892 oe = find_mark(from_mark);
2893 type = oe->type;
2894 oidcpy(&oid, &oe->idx.oid);
2895 } else if (!get_oid(from, &oid)) {
2896 struct object_entry *oe = find_object(&oid);
2897 if (!oe) {
2898 type = oid_object_info(&oid, NULL);
2899 if (type < 0)
2900 die("Not a valid object: %s", from);
2901 } else
2902 type = oe->type;
2903 } else
2904 die("Invalid ref name or SHA1 expression: %s", from);
2905 read_next_command();
2907 /* tagger ... */
2908 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2909 tagger = parse_ident(v);
2910 read_next_command();
2911 } else
2912 tagger = NULL;
2914 /* tag payload/message */
2915 parse_data(&msg, 0, NULL);
2917 /* build the tag object */
2918 strbuf_reset(&new_data);
2920 strbuf_addf(&new_data,
2921 "object %s\n"
2922 "type %s\n"
2923 "tag %s\n",
2924 oid_to_hex(&oid), type_name(type), t->name);
2925 if (tagger)
2926 strbuf_addf(&new_data,
2927 "tagger %s\n", tagger);
2928 strbuf_addch(&new_data, '\n');
2929 strbuf_addbuf(&new_data, &msg);
2930 free(tagger);
2932 if (store_object(OBJ_TAG, &new_data, NULL, &t->oid, 0))
2933 t->pack_id = MAX_PACK_ID;
2934 else
2935 t->pack_id = pack_id;
2938 static void parse_reset_branch(const char *arg)
2940 struct branch *b;
2942 b = lookup_branch(arg);
2943 if (b) {
2944 oidclr(&b->oid);
2945 oidclr(&b->branch_tree.versions[0].oid);
2946 oidclr(&b->branch_tree.versions[1].oid);
2947 if (b->branch_tree.tree) {
2948 release_tree_content_recursive(b->branch_tree.tree);
2949 b->branch_tree.tree = NULL;
2952 else
2953 b = new_branch(arg);
2954 read_next_command();
2955 parse_from(b);
2956 if (command_buf.len > 0)
2957 unread_command_buf = 1;
2960 static void cat_blob_write(const char *buf, unsigned long size)
2962 if (write_in_full(cat_blob_fd, buf, size) < 0)
2963 die_errno("Write to frontend failed");
2966 static void cat_blob(struct object_entry *oe, struct object_id *oid)
2968 struct strbuf line = STRBUF_INIT;
2969 unsigned long size;
2970 enum object_type type = 0;
2971 char *buf;
2973 if (!oe || oe->pack_id == MAX_PACK_ID) {
2974 buf = read_object_file(oid, &type, &size);
2975 } else {
2976 type = oe->type;
2977 buf = gfi_unpack_entry(oe, &size);
2981 * Output based on batch_one_object() from cat-file.c.
2983 if (type <= 0) {
2984 strbuf_reset(&line);
2985 strbuf_addf(&line, "%s missing\n", oid_to_hex(oid));
2986 cat_blob_write(line.buf, line.len);
2987 strbuf_release(&line);
2988 free(buf);
2989 return;
2991 if (!buf)
2992 die("Can't read object %s", oid_to_hex(oid));
2993 if (type != OBJ_BLOB)
2994 die("Object %s is a %s but a blob was expected.",
2995 oid_to_hex(oid), type_name(type));
2996 strbuf_reset(&line);
2997 strbuf_addf(&line, "%s %s %lu\n", oid_to_hex(oid),
2998 type_name(type), size);
2999 cat_blob_write(line.buf, line.len);
3000 strbuf_release(&line);
3001 cat_blob_write(buf, size);
3002 cat_blob_write("\n", 1);
3003 if (oe && oe->pack_id == pack_id) {
3004 last_blob.offset = oe->idx.offset;
3005 strbuf_attach(&last_blob.data, buf, size, size);
3006 last_blob.depth = oe->depth;
3007 } else
3008 free(buf);
3011 static void parse_get_mark(const char *p)
3013 struct object_entry *oe;
3014 char output[GIT_MAX_HEXSZ + 2];
3016 /* get-mark SP <object> LF */
3017 if (*p != ':')
3018 die("Not a mark: %s", p);
3020 oe = find_mark(parse_mark_ref_eol(p));
3021 if (!oe)
3022 die("Unknown mark: %s", command_buf.buf);
3024 xsnprintf(output, sizeof(output), "%s\n", oid_to_hex(&oe->idx.oid));
3025 cat_blob_write(output, GIT_SHA1_HEXSZ + 1);
3028 static void parse_cat_blob(const char *p)
3030 struct object_entry *oe;
3031 struct object_id oid;
3033 /* cat-blob SP <object> LF */
3034 if (*p == ':') {
3035 oe = find_mark(parse_mark_ref_eol(p));
3036 if (!oe)
3037 die("Unknown mark: %s", command_buf.buf);
3038 oidcpy(&oid, &oe->idx.oid);
3039 } else {
3040 if (parse_oid_hex(p, &oid, &p))
3041 die("Invalid dataref: %s", command_buf.buf);
3042 if (*p)
3043 die("Garbage after SHA1: %s", command_buf.buf);
3044 oe = find_object(&oid);
3047 cat_blob(oe, &oid);
3050 static struct object_entry *dereference(struct object_entry *oe,
3051 struct object_id *oid)
3053 unsigned long size;
3054 char *buf = NULL;
3055 if (!oe) {
3056 enum object_type type = oid_object_info(oid, NULL);
3057 if (type < 0)
3058 die("object not found: %s", oid_to_hex(oid));
3059 /* cache it! */
3060 oe = insert_object(oid);
3061 oe->type = type;
3062 oe->pack_id = MAX_PACK_ID;
3063 oe->idx.offset = 1;
3065 switch (oe->type) {
3066 case OBJ_TREE: /* easy case. */
3067 return oe;
3068 case OBJ_COMMIT:
3069 case OBJ_TAG:
3070 break;
3071 default:
3072 die("Not a tree-ish: %s", command_buf.buf);
3075 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
3076 buf = gfi_unpack_entry(oe, &size);
3077 } else {
3078 enum object_type unused;
3079 buf = read_object_file(oid, &unused, &size);
3081 if (!buf)
3082 die("Can't load object %s", oid_to_hex(oid));
3084 /* Peel one layer. */
3085 switch (oe->type) {
3086 case OBJ_TAG:
3087 if (size < GIT_SHA1_HEXSZ + strlen("object ") ||
3088 get_oid_hex(buf + strlen("object "), oid))
3089 die("Invalid SHA1 in tag: %s", command_buf.buf);
3090 break;
3091 case OBJ_COMMIT:
3092 if (size < GIT_SHA1_HEXSZ + strlen("tree ") ||
3093 get_oid_hex(buf + strlen("tree "), oid))
3094 die("Invalid SHA1 in commit: %s", command_buf.buf);
3097 free(buf);
3098 return find_object(oid);
3101 static struct object_entry *parse_treeish_dataref(const char **p)
3103 struct object_id oid;
3104 struct object_entry *e;
3106 if (**p == ':') { /* <mark> */
3107 e = find_mark(parse_mark_ref_space(p));
3108 if (!e)
3109 die("Unknown mark: %s", command_buf.buf);
3110 oidcpy(&oid, &e->idx.oid);
3111 } else { /* <sha1> */
3112 if (parse_oid_hex(*p, &oid, p))
3113 die("Invalid dataref: %s", command_buf.buf);
3114 e = find_object(&oid);
3115 if (*(*p)++ != ' ')
3116 die("Missing space after tree-ish: %s", command_buf.buf);
3119 while (!e || e->type != OBJ_TREE)
3120 e = dereference(e, &oid);
3121 return e;
3124 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3126 static struct strbuf line = STRBUF_INIT;
3128 /* See show_tree(). */
3129 const char *type =
3130 S_ISGITLINK(mode) ? commit_type :
3131 S_ISDIR(mode) ? tree_type :
3132 blob_type;
3134 if (!mode) {
3135 /* missing SP path LF */
3136 strbuf_reset(&line);
3137 strbuf_addstr(&line, "missing ");
3138 quote_c_style(path, &line, NULL, 0);
3139 strbuf_addch(&line, '\n');
3140 } else {
3141 /* mode SP type SP object_name TAB path LF */
3142 strbuf_reset(&line);
3143 strbuf_addf(&line, "%06o %s %s\t",
3144 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3145 quote_c_style(path, &line, NULL, 0);
3146 strbuf_addch(&line, '\n');
3148 cat_blob_write(line.buf, line.len);
3151 static void parse_ls(const char *p, struct branch *b)
3153 struct tree_entry *root = NULL;
3154 struct tree_entry leaf = {NULL};
3156 /* ls SP (<tree-ish> SP)? <path> */
3157 if (*p == '"') {
3158 if (!b)
3159 die("Not in a commit: %s", command_buf.buf);
3160 root = &b->branch_tree;
3161 } else {
3162 struct object_entry *e = parse_treeish_dataref(&p);
3163 root = new_tree_entry();
3164 oidcpy(&root->versions[1].oid, &e->idx.oid);
3165 if (!is_null_oid(&root->versions[1].oid))
3166 root->versions[1].mode = S_IFDIR;
3167 load_tree(root);
3169 if (*p == '"') {
3170 static struct strbuf uq = STRBUF_INIT;
3171 const char *endp;
3172 strbuf_reset(&uq);
3173 if (unquote_c_style(&uq, p, &endp))
3174 die("Invalid path: %s", command_buf.buf);
3175 if (*endp)
3176 die("Garbage after path in: %s", command_buf.buf);
3177 p = uq.buf;
3179 tree_content_get(root, p, &leaf, 1);
3181 * A directory in preparation would have a sha1 of zero
3182 * until it is saved. Save, for simplicity.
3184 if (S_ISDIR(leaf.versions[1].mode))
3185 store_tree(&leaf);
3187 print_ls(leaf.versions[1].mode, leaf.versions[1].oid.hash, p);
3188 if (leaf.tree)
3189 release_tree_content_recursive(leaf.tree);
3190 if (!b || root != &b->branch_tree)
3191 release_tree_entry(root);
3194 static void checkpoint(void)
3196 checkpoint_requested = 0;
3197 if (object_count) {
3198 cycle_packfile();
3200 dump_branches();
3201 dump_tags();
3202 dump_marks();
3205 static void parse_checkpoint(void)
3207 checkpoint_requested = 1;
3208 skip_optional_lf();
3211 static void parse_progress(void)
3213 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3214 fputc('\n', stdout);
3215 fflush(stdout);
3216 skip_optional_lf();
3219 static char* make_fast_import_path(const char *path)
3221 if (!relative_marks_paths || is_absolute_path(path))
3222 return xstrdup(path);
3223 return git_pathdup("info/fast-import/%s", path);
3226 static void option_import_marks(const char *marks,
3227 int from_stream, int ignore_missing)
3229 if (import_marks_file) {
3230 if (from_stream)
3231 die("Only one import-marks command allowed per stream");
3233 /* read previous mark file */
3234 if(!import_marks_file_from_stream)
3235 read_marks();
3238 import_marks_file = make_fast_import_path(marks);
3239 safe_create_leading_directories_const(import_marks_file);
3240 import_marks_file_from_stream = from_stream;
3241 import_marks_file_ignore_missing = ignore_missing;
3244 static void option_date_format(const char *fmt)
3246 if (!strcmp(fmt, "raw"))
3247 whenspec = WHENSPEC_RAW;
3248 else if (!strcmp(fmt, "rfc2822"))
3249 whenspec = WHENSPEC_RFC2822;
3250 else if (!strcmp(fmt, "now"))
3251 whenspec = WHENSPEC_NOW;
3252 else
3253 die("unknown --date-format argument %s", fmt);
3256 static unsigned long ulong_arg(const char *option, const char *arg)
3258 char *endptr;
3259 unsigned long rv = strtoul(arg, &endptr, 0);
3260 if (strchr(arg, '-') || endptr == arg || *endptr)
3261 die("%s: argument must be a non-negative integer", option);
3262 return rv;
3265 static void option_depth(const char *depth)
3267 max_depth = ulong_arg("--depth", depth);
3268 if (max_depth > MAX_DEPTH)
3269 die("--depth cannot exceed %u", MAX_DEPTH);
3272 static void option_active_branches(const char *branches)
3274 max_active_branches = ulong_arg("--active-branches", branches);
3277 static void option_export_marks(const char *marks)
3279 export_marks_file = make_fast_import_path(marks);
3280 safe_create_leading_directories_const(export_marks_file);
3283 static void option_cat_blob_fd(const char *fd)
3285 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3286 if (n > (unsigned long) INT_MAX)
3287 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3288 cat_blob_fd = (int) n;
3291 static void option_export_pack_edges(const char *edges)
3293 if (pack_edges)
3294 fclose(pack_edges);
3295 pack_edges = xfopen(edges, "a");
3298 static int parse_one_option(const char *option)
3300 if (skip_prefix(option, "max-pack-size=", &option)) {
3301 unsigned long v;
3302 if (!git_parse_ulong(option, &v))
3303 return 0;
3304 if (v < 8192) {
3305 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3306 v *= 1024 * 1024;
3307 } else if (v < 1024 * 1024) {
3308 warning("minimum max-pack-size is 1 MiB");
3309 v = 1024 * 1024;
3311 max_packsize = v;
3312 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3313 unsigned long v;
3314 if (!git_parse_ulong(option, &v))
3315 return 0;
3316 big_file_threshold = v;
3317 } else if (skip_prefix(option, "depth=", &option)) {
3318 option_depth(option);
3319 } else if (skip_prefix(option, "active-branches=", &option)) {
3320 option_active_branches(option);
3321 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3322 option_export_pack_edges(option);
3323 } else if (starts_with(option, "quiet")) {
3324 show_stats = 0;
3325 } else if (starts_with(option, "stats")) {
3326 show_stats = 1;
3327 } else {
3328 return 0;
3331 return 1;
3334 static int parse_one_feature(const char *feature, int from_stream)
3336 const char *arg;
3338 if (skip_prefix(feature, "date-format=", &arg)) {
3339 option_date_format(arg);
3340 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3341 option_import_marks(arg, from_stream, 0);
3342 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3343 option_import_marks(arg, from_stream, 1);
3344 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3345 option_export_marks(arg);
3346 } else if (!strcmp(feature, "get-mark")) {
3347 ; /* Don't die - this feature is supported */
3348 } else if (!strcmp(feature, "cat-blob")) {
3349 ; /* Don't die - this feature is supported */
3350 } else if (!strcmp(feature, "relative-marks")) {
3351 relative_marks_paths = 1;
3352 } else if (!strcmp(feature, "no-relative-marks")) {
3353 relative_marks_paths = 0;
3354 } else if (!strcmp(feature, "done")) {
3355 require_explicit_termination = 1;
3356 } else if (!strcmp(feature, "force")) {
3357 force_update = 1;
3358 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3359 ; /* do nothing; we have the feature */
3360 } else {
3361 return 0;
3364 return 1;
3367 static void parse_feature(const char *feature)
3369 if (seen_data_command)
3370 die("Got feature command '%s' after data command", feature);
3372 if (parse_one_feature(feature, 1))
3373 return;
3375 die("This version of fast-import does not support feature %s.", feature);
3378 static void parse_option(const char *option)
3380 if (seen_data_command)
3381 die("Got option command '%s' after data command", option);
3383 if (parse_one_option(option))
3384 return;
3386 die("This version of fast-import does not support option: %s", option);
3389 static void git_pack_config(void)
3391 int indexversion_value;
3392 int limit;
3393 unsigned long packsizelimit_value;
3395 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3396 if (max_depth > MAX_DEPTH)
3397 max_depth = MAX_DEPTH;
3399 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3400 pack_idx_opts.version = indexversion_value;
3401 if (pack_idx_opts.version > 2)
3402 git_die_config("pack.indexversion",
3403 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3405 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3406 max_packsize = packsizelimit_value;
3408 if (!git_config_get_int("fastimport.unpacklimit", &limit))
3409 unpack_limit = limit;
3410 else if (!git_config_get_int("transfer.unpacklimit", &limit))
3411 unpack_limit = limit;
3413 git_config(git_default_config, NULL);
3416 static const char fast_import_usage[] =
3417 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3419 static void parse_argv(void)
3421 unsigned int i;
3423 for (i = 1; i < global_argc; i++) {
3424 const char *a = global_argv[i];
3426 if (*a != '-' || !strcmp(a, "--"))
3427 break;
3429 if (!skip_prefix(a, "--", &a))
3430 die("unknown option %s", a);
3432 if (parse_one_option(a))
3433 continue;
3435 if (parse_one_feature(a, 0))
3436 continue;
3438 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3439 option_cat_blob_fd(a);
3440 continue;
3443 die("unknown option --%s", a);
3445 if (i != global_argc)
3446 usage(fast_import_usage);
3448 seen_data_command = 1;
3449 if (import_marks_file)
3450 read_marks();
3453 int cmd_main(int argc, const char **argv)
3455 unsigned int i;
3457 if (argc == 2 && !strcmp(argv[1], "-h"))
3458 usage(fast_import_usage);
3460 setup_git_directory();
3461 reset_pack_idx_option(&pack_idx_opts);
3462 git_pack_config();
3464 alloc_objects(object_entry_alloc);
3465 strbuf_init(&command_buf, 0);
3466 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3467 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3468 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3469 marks = pool_calloc(1, sizeof(struct mark_set));
3471 global_argc = argc;
3472 global_argv = argv;
3474 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3475 for (i = 0; i < (cmd_save - 1); i++)
3476 rc_free[i].next = &rc_free[i + 1];
3477 rc_free[cmd_save - 1].next = NULL;
3479 start_packfile();
3480 set_die_routine(die_nicely);
3481 set_checkpoint_signal();
3482 while (read_next_command() != EOF) {
3483 const char *v;
3484 if (!strcmp("blob", command_buf.buf))
3485 parse_new_blob();
3486 else if (skip_prefix(command_buf.buf, "ls ", &v))
3487 parse_ls(v, NULL);
3488 else if (skip_prefix(command_buf.buf, "commit ", &v))
3489 parse_new_commit(v);
3490 else if (skip_prefix(command_buf.buf, "tag ", &v))
3491 parse_new_tag(v);
3492 else if (skip_prefix(command_buf.buf, "reset ", &v))
3493 parse_reset_branch(v);
3494 else if (!strcmp("checkpoint", command_buf.buf))
3495 parse_checkpoint();
3496 else if (!strcmp("done", command_buf.buf))
3497 break;
3498 else if (starts_with(command_buf.buf, "progress "))
3499 parse_progress();
3500 else if (skip_prefix(command_buf.buf, "feature ", &v))
3501 parse_feature(v);
3502 else if (skip_prefix(command_buf.buf, "option git ", &v))
3503 parse_option(v);
3504 else if (starts_with(command_buf.buf, "option "))
3505 /* ignore non-git options*/;
3506 else
3507 die("Unsupported command: %s", command_buf.buf);
3509 if (checkpoint_requested)
3510 checkpoint();
3513 /* argv hasn't been parsed yet, do so */
3514 if (!seen_data_command)
3515 parse_argv();
3517 if (require_explicit_termination && feof(stdin))
3518 die("stream ends early");
3520 end_packfile();
3522 dump_branches();
3523 dump_tags();
3524 unkeep_all_packs();
3525 dump_marks();
3527 if (pack_edges)
3528 fclose(pack_edges);
3530 if (show_stats) {
3531 uintmax_t total_count = 0, duplicate_count = 0;
3532 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3533 total_count += object_count_by_type[i];
3534 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3535 duplicate_count += duplicate_count_by_type[i];
3537 fprintf(stderr, "%s statistics:\n", argv[0]);
3538 fprintf(stderr, "---------------------------------------------------------------------\n");
3539 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3540 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3541 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]);
3542 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]);
3543 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]);
3544 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]);
3545 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3546 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3547 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3548 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3549 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3550 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3551 fprintf(stderr, "---------------------------------------------------------------------\n");
3552 pack_report();
3553 fprintf(stderr, "---------------------------------------------------------------------\n");
3554 fprintf(stderr, "\n");
3557 return failure ? 1 : 0;