convert strncpy to memcpy
[alt-git.git] / fast-import.c
blob4d01efcb9d1e41062d76e88a4f928ceb50f79ca8
1 /*
2 (See Documentation/git-fast-import.txt for maintained documentation.)
3 Format of STDIN stream:
5 stream ::= cmd*;
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
20 new_commit ::= 'commit' sp ref_str lf
21 mark?
22 ('author' (sp name)? sp '<' email '>' sp when lf)?
23 'committer' (sp name)? sp '<' email '>' sp when lf
24 commit_msg
25 ('from' sp commit-ish lf)?
26 ('merge' sp commit-ish lf)*
27 (file_change | ls)*
28 lf?;
29 commit_msg ::= data;
31 ls ::= 'ls' sp '"' quoted(path) '"' lf;
33 file_change ::= file_clr
34 | file_del
35 | file_rnm
36 | file_cpy
37 | file_obm
38 | file_inm;
39 file_clr ::= 'deleteall' lf;
40 file_del ::= 'D' sp path_str lf;
41 file_rnm ::= 'R' sp path_str sp path_str lf;
42 file_cpy ::= 'C' sp path_str sp path_str lf;
43 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
44 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
45 data;
46 note_obm ::= 'N' sp (hexsha1 | idnum) sp commit-ish lf;
47 note_inm ::= 'N' sp 'inline' sp commit-ish lf
48 data;
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp commit-ish lf
52 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
53 tag_msg;
54 tag_msg ::= data;
56 reset_branch ::= 'reset' sp ref_str lf
57 ('from' sp commit-ish lf)?
58 lf?;
60 checkpoint ::= 'checkpoint' lf
61 lf?;
63 progress ::= 'progress' sp not_lf* lf
64 lf?;
66 # note: the first idnum in a stream should be 1 and subsequent
67 # idnums should not have gaps between values as this will cause
68 # the stream parser to reserve space for the gapped values. An
69 # idnum can be updated in the future to a new object by issuing
70 # a new mark directive with the old idnum.
72 mark ::= 'mark' sp idnum lf;
73 data ::= (delimited_data | exact_data)
74 lf?;
76 # note: delim may be any string but must not contain lf.
77 # data_line may contain any data but must not be exactly
78 # delim.
79 delimited_data ::= 'data' sp '<<' delim lf
80 (data_line lf)*
81 delim lf;
83 # note: declen indicates the length of binary_data in bytes.
84 # declen does not include the lf preceding the binary data.
86 exact_data ::= 'data' sp declen lf
87 binary_data;
89 # note: quoted strings are C-style quoting supporting \c for
90 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
91 # is the signed byte value in octal. Note that the only
92 # characters which must actually be escaped to protect the
93 # stream formatting is: \, " and LF. Otherwise these values
94 # are UTF8.
96 commit-ish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
97 ref_str ::= ref;
98 sha1exp_str ::= sha1exp;
99 tag_str ::= tag;
100 path_str ::= path | '"' quoted(path) '"' ;
101 mode ::= '100644' | '644'
102 | '100755' | '755'
103 | '120000'
106 declen ::= # unsigned 32 bit value, ascii base10 notation;
107 bigint ::= # unsigned integer value, ascii base10 notation;
108 binary_data ::= # file content, not interpreted;
110 when ::= raw_when | rfc2822_when;
111 raw_when ::= ts sp tz;
112 rfc2822_when ::= # Valid RFC 2822 date and time;
114 sp ::= # ASCII space character;
115 lf ::= # ASCII newline (LF) character;
117 # note: a colon (':') must precede the numerical value assigned to
118 # an idnum. This is to distinguish it from a ref or tag name as
119 # GIT does not permit ':' in ref or tag strings.
121 idnum ::= ':' bigint;
122 path ::= # GIT style file path, e.g. "a/b/c";
123 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
124 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
125 sha1exp ::= # Any valid GIT SHA1 expression;
126 hexsha1 ::= # SHA1 in hexadecimal format;
128 # note: name and email are UTF8 strings, however name must not
129 # contain '<' or lf and email must not contain any of the
130 # following: '<', '>', lf.
132 name ::= # valid GIT author/committer name;
133 email ::= # valid GIT author/committer email;
134 ts ::= # time since the epoch in seconds, ascii base10 notation;
135 tz ::= # GIT style timezone;
137 # note: comments, get-mark, ls-tree, and cat-blob requests may
138 # appear anywhere in the input, except within a data command. Any
139 # form of the data command always escapes the related input from
140 # comment processing.
142 # In case it is not clear, the '#' that starts the comment
143 # must be the first character on that line (an lf
144 # preceded it).
147 get_mark ::= 'get-mark' sp idnum lf;
148 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
149 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
151 comment ::= '#' not_lf* lf;
152 not_lf ::= # Any byte that is not ASCII newline (LF);
155 #include "builtin.h"
156 #include "cache.h"
157 #include "lockfile.h"
158 #include "object.h"
159 #include "blob.h"
160 #include "tree.h"
161 #include "commit.h"
162 #include "delta.h"
163 #include "pack.h"
164 #include "refs.h"
165 #include "csum-file.h"
166 #include "quote.h"
167 #include "exec_cmd.h"
168 #include "dir.h"
170 #define PACK_ID_BITS 16
171 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
172 #define DEPTH_BITS 13
173 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
176 * We abuse the setuid bit on directories to mean "do not delta".
178 #define NO_DELTA S_ISUID
180 struct object_entry {
181 struct pack_idx_entry idx;
182 struct object_entry *next;
183 uint32_t type : TYPE_BITS,
184 pack_id : PACK_ID_BITS,
185 depth : DEPTH_BITS;
188 struct object_entry_pool {
189 struct object_entry_pool *next_pool;
190 struct object_entry *next_free;
191 struct object_entry *end;
192 struct object_entry entries[FLEX_ARRAY]; /* more */
195 struct mark_set {
196 union {
197 struct object_entry *marked[1024];
198 struct mark_set *sets[1024];
199 } data;
200 unsigned int shift;
203 struct last_object {
204 struct strbuf data;
205 off_t offset;
206 unsigned int depth;
207 unsigned no_swap : 1;
210 struct mem_pool {
211 struct mem_pool *next_pool;
212 char *next_free;
213 char *end;
214 uintmax_t space[FLEX_ARRAY]; /* more */
217 struct atom_str {
218 struct atom_str *next_atom;
219 unsigned short str_len;
220 char str_dat[FLEX_ARRAY]; /* more */
223 struct tree_content;
224 struct tree_entry {
225 struct tree_content *tree;
226 struct atom_str *name;
227 struct tree_entry_ms {
228 uint16_t mode;
229 unsigned char sha1[20];
230 } versions[2];
233 struct tree_content {
234 unsigned int entry_capacity; /* must match avail_tree_content */
235 unsigned int entry_count;
236 unsigned int delta_depth;
237 struct tree_entry *entries[FLEX_ARRAY]; /* more */
240 struct avail_tree_content {
241 unsigned int entry_capacity; /* must match tree_content */
242 struct avail_tree_content *next_avail;
245 struct branch {
246 struct branch *table_next_branch;
247 struct branch *active_next_branch;
248 const char *name;
249 struct tree_entry branch_tree;
250 uintmax_t last_commit;
251 uintmax_t num_notes;
252 unsigned active : 1;
253 unsigned delete : 1;
254 unsigned pack_id : PACK_ID_BITS;
255 unsigned char sha1[20];
258 struct tag {
259 struct tag *next_tag;
260 const char *name;
261 unsigned int pack_id;
262 unsigned char sha1[20];
265 struct hash_list {
266 struct hash_list *next;
267 unsigned char sha1[20];
270 typedef enum {
271 WHENSPEC_RAW = 1,
272 WHENSPEC_RFC2822,
273 WHENSPEC_NOW
274 } whenspec_type;
276 struct recent_command {
277 struct recent_command *prev;
278 struct recent_command *next;
279 char *buf;
282 /* Configured limits on output */
283 static unsigned long max_depth = 10;
284 static off_t max_packsize;
285 static int force_update;
286 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
287 static int pack_compression_seen;
289 /* Stats and misc. counters */
290 static uintmax_t alloc_count;
291 static uintmax_t marks_set_count;
292 static uintmax_t object_count_by_type[1 << TYPE_BITS];
293 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
294 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
295 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
296 static unsigned long object_count;
297 static unsigned long branch_count;
298 static unsigned long branch_load_count;
299 static int failure;
300 static FILE *pack_edges;
301 static unsigned int show_stats = 1;
302 static int global_argc;
303 static char **global_argv;
305 /* Memory pools */
306 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
307 static size_t total_allocd;
308 static struct mem_pool *mem_pool;
310 /* Atom management */
311 static unsigned int atom_table_sz = 4451;
312 static unsigned int atom_cnt;
313 static struct atom_str **atom_table;
315 /* The .pack file being generated */
316 static struct pack_idx_option pack_idx_opts;
317 static unsigned int pack_id;
318 static struct sha1file *pack_file;
319 static struct packed_git *pack_data;
320 static struct packed_git **all_packs;
321 static off_t pack_size;
323 /* Table of objects we've written. */
324 static unsigned int object_entry_alloc = 5000;
325 static struct object_entry_pool *blocks;
326 static struct object_entry *object_table[1 << 16];
327 static struct mark_set *marks;
328 static const char *export_marks_file;
329 static const char *import_marks_file;
330 static int import_marks_file_from_stream;
331 static int import_marks_file_ignore_missing;
332 static int relative_marks_paths;
334 /* Our last blob */
335 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
337 /* Tree management */
338 static unsigned int tree_entry_alloc = 1000;
339 static void *avail_tree_entry;
340 static unsigned int avail_tree_table_sz = 100;
341 static struct avail_tree_content **avail_tree_table;
342 static struct strbuf old_tree = STRBUF_INIT;
343 static struct strbuf new_tree = STRBUF_INIT;
345 /* Branch data */
346 static unsigned long max_active_branches = 5;
347 static unsigned long cur_active_branches;
348 static unsigned long branch_table_sz = 1039;
349 static struct branch **branch_table;
350 static struct branch *active_branches;
352 /* Tag data */
353 static struct tag *first_tag;
354 static struct tag *last_tag;
356 /* Input stream parsing */
357 static whenspec_type whenspec = WHENSPEC_RAW;
358 static struct strbuf command_buf = STRBUF_INIT;
359 static int unread_command_buf;
360 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
361 static struct recent_command *cmd_tail = &cmd_hist;
362 static struct recent_command *rc_free;
363 static unsigned int cmd_save = 100;
364 static uintmax_t next_mark;
365 static struct strbuf new_data = STRBUF_INIT;
366 static int seen_data_command;
367 static int require_explicit_termination;
369 /* Signal handling */
370 static volatile sig_atomic_t checkpoint_requested;
372 /* Where to write output of cat-blob commands */
373 static int cat_blob_fd = STDOUT_FILENO;
375 static void parse_argv(void);
376 static void parse_get_mark(const char *p);
377 static void parse_cat_blob(const char *p);
378 static void parse_ls(const char *p, struct branch *b);
380 static void write_branch_report(FILE *rpt, struct branch *b)
382 fprintf(rpt, "%s:\n", b->name);
384 fprintf(rpt, " status :");
385 if (b->active)
386 fputs(" active", rpt);
387 if (b->branch_tree.tree)
388 fputs(" loaded", rpt);
389 if (is_null_sha1(b->branch_tree.versions[1].sha1))
390 fputs(" dirty", rpt);
391 fputc('\n', rpt);
393 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
394 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
395 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
396 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
398 fputs(" last pack : ", rpt);
399 if (b->pack_id < MAX_PACK_ID)
400 fprintf(rpt, "%u", b->pack_id);
401 fputc('\n', rpt);
403 fputc('\n', rpt);
406 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
408 static void write_crash_report(const char *err)
410 char *loc = git_pathdup("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
411 FILE *rpt = fopen(loc, "w");
412 struct branch *b;
413 unsigned long lu;
414 struct recent_command *rc;
416 if (!rpt) {
417 error("can't write crash report %s: %s", loc, strerror(errno));
418 free(loc);
419 return;
422 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
424 fprintf(rpt, "fast-import crash report:\n");
425 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
426 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
427 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_MODE(LOCAL)));
428 fputc('\n', rpt);
430 fputs("fatal: ", rpt);
431 fputs(err, rpt);
432 fputc('\n', rpt);
434 fputc('\n', rpt);
435 fputs("Most Recent Commands Before Crash\n", rpt);
436 fputs("---------------------------------\n", rpt);
437 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
438 if (rc->next == &cmd_hist)
439 fputs("* ", rpt);
440 else
441 fputs(" ", rpt);
442 fputs(rc->buf, rpt);
443 fputc('\n', rpt);
446 fputc('\n', rpt);
447 fputs("Active Branch LRU\n", rpt);
448 fputs("-----------------\n", rpt);
449 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
450 cur_active_branches,
451 max_active_branches);
452 fputc('\n', rpt);
453 fputs(" pos clock name\n", rpt);
454 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
455 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
456 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
457 ++lu, b->last_commit, b->name);
459 fputc('\n', rpt);
460 fputs("Inactive Branches\n", rpt);
461 fputs("-----------------\n", rpt);
462 for (lu = 0; lu < branch_table_sz; lu++) {
463 for (b = branch_table[lu]; b; b = b->table_next_branch)
464 write_branch_report(rpt, b);
467 if (first_tag) {
468 struct tag *tg;
469 fputc('\n', rpt);
470 fputs("Annotated Tags\n", rpt);
471 fputs("--------------\n", rpt);
472 for (tg = first_tag; tg; tg = tg->next_tag) {
473 fputs(sha1_to_hex(tg->sha1), rpt);
474 fputc(' ', rpt);
475 fputs(tg->name, rpt);
476 fputc('\n', rpt);
480 fputc('\n', rpt);
481 fputs("Marks\n", rpt);
482 fputs("-----\n", rpt);
483 if (export_marks_file)
484 fprintf(rpt, " exported to %s\n", export_marks_file);
485 else
486 dump_marks_helper(rpt, 0, marks);
488 fputc('\n', rpt);
489 fputs("-------------------\n", rpt);
490 fputs("END OF CRASH REPORT\n", rpt);
491 fclose(rpt);
492 free(loc);
495 static void end_packfile(void);
496 static void unkeep_all_packs(void);
497 static void dump_marks(void);
499 static NORETURN void die_nicely(const char *err, va_list params)
501 static int zombie;
502 char message[2 * PATH_MAX];
504 vsnprintf(message, sizeof(message), err, params);
505 fputs("fatal: ", stderr);
506 fputs(message, stderr);
507 fputc('\n', stderr);
509 if (!zombie) {
510 zombie = 1;
511 write_crash_report(message);
512 end_packfile();
513 unkeep_all_packs();
514 dump_marks();
516 exit(128);
519 #ifndef SIGUSR1 /* Windows, for example */
521 static void set_checkpoint_signal(void)
525 #else
527 static void checkpoint_signal(int signo)
529 checkpoint_requested = 1;
532 static void set_checkpoint_signal(void)
534 struct sigaction sa;
536 memset(&sa, 0, sizeof(sa));
537 sa.sa_handler = checkpoint_signal;
538 sigemptyset(&sa.sa_mask);
539 sa.sa_flags = SA_RESTART;
540 sigaction(SIGUSR1, &sa, NULL);
543 #endif
545 static void alloc_objects(unsigned int cnt)
547 struct object_entry_pool *b;
549 b = xmalloc(sizeof(struct object_entry_pool)
550 + cnt * sizeof(struct object_entry));
551 b->next_pool = blocks;
552 b->next_free = b->entries;
553 b->end = b->entries + cnt;
554 blocks = b;
555 alloc_count += cnt;
558 static struct object_entry *new_object(unsigned char *sha1)
560 struct object_entry *e;
562 if (blocks->next_free == blocks->end)
563 alloc_objects(object_entry_alloc);
565 e = blocks->next_free++;
566 hashcpy(e->idx.sha1, sha1);
567 return e;
570 static struct object_entry *find_object(unsigned char *sha1)
572 unsigned int h = sha1[0] << 8 | sha1[1];
573 struct object_entry *e;
574 for (e = object_table[h]; e; e = e->next)
575 if (!hashcmp(sha1, e->idx.sha1))
576 return e;
577 return NULL;
580 static struct object_entry *insert_object(unsigned char *sha1)
582 unsigned int h = sha1[0] << 8 | sha1[1];
583 struct object_entry *e = object_table[h];
585 while (e) {
586 if (!hashcmp(sha1, e->idx.sha1))
587 return e;
588 e = e->next;
591 e = new_object(sha1);
592 e->next = object_table[h];
593 e->idx.offset = 0;
594 object_table[h] = e;
595 return e;
598 static unsigned int hc_str(const char *s, size_t len)
600 unsigned int r = 0;
601 while (len-- > 0)
602 r = r * 31 + *s++;
603 return r;
606 static void *pool_alloc(size_t len)
608 struct mem_pool *p;
609 void *r;
611 /* round up to a 'uintmax_t' alignment */
612 if (len & (sizeof(uintmax_t) - 1))
613 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
615 for (p = mem_pool; p; p = p->next_pool)
616 if ((p->end - p->next_free >= len))
617 break;
619 if (!p) {
620 if (len >= (mem_pool_alloc/2)) {
621 total_allocd += len;
622 return xmalloc(len);
624 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
625 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
626 p->next_pool = mem_pool;
627 p->next_free = (char *) p->space;
628 p->end = p->next_free + mem_pool_alloc;
629 mem_pool = p;
632 r = p->next_free;
633 p->next_free += len;
634 return r;
637 static void *pool_calloc(size_t count, size_t size)
639 size_t len = count * size;
640 void *r = pool_alloc(len);
641 memset(r, 0, len);
642 return r;
645 static char *pool_strdup(const char *s)
647 size_t len = strlen(s) + 1;
648 char *r = pool_alloc(len);
649 memcpy(r, s, len);
650 return r;
653 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
655 struct mark_set *s = marks;
656 while ((idnum >> s->shift) >= 1024) {
657 s = pool_calloc(1, sizeof(struct mark_set));
658 s->shift = marks->shift + 10;
659 s->data.sets[0] = marks;
660 marks = s;
662 while (s->shift) {
663 uintmax_t i = idnum >> s->shift;
664 idnum -= i << s->shift;
665 if (!s->data.sets[i]) {
666 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
667 s->data.sets[i]->shift = s->shift - 10;
669 s = s->data.sets[i];
671 if (!s->data.marked[idnum])
672 marks_set_count++;
673 s->data.marked[idnum] = oe;
676 static struct object_entry *find_mark(uintmax_t idnum)
678 uintmax_t orig_idnum = idnum;
679 struct mark_set *s = marks;
680 struct object_entry *oe = NULL;
681 if ((idnum >> s->shift) < 1024) {
682 while (s && s->shift) {
683 uintmax_t i = idnum >> s->shift;
684 idnum -= i << s->shift;
685 s = s->data.sets[i];
687 if (s)
688 oe = s->data.marked[idnum];
690 if (!oe)
691 die("mark :%" PRIuMAX " not declared", orig_idnum);
692 return oe;
695 static struct atom_str *to_atom(const char *s, unsigned short len)
697 unsigned int hc = hc_str(s, len) % atom_table_sz;
698 struct atom_str *c;
700 for (c = atom_table[hc]; c; c = c->next_atom)
701 if (c->str_len == len && !strncmp(s, c->str_dat, len))
702 return c;
704 c = pool_alloc(sizeof(struct atom_str) + len + 1);
705 c->str_len = len;
706 memcpy(c->str_dat, s, len);
707 c->str_dat[len] = 0;
708 c->next_atom = atom_table[hc];
709 atom_table[hc] = c;
710 atom_cnt++;
711 return c;
714 static struct branch *lookup_branch(const char *name)
716 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
717 struct branch *b;
719 for (b = branch_table[hc]; b; b = b->table_next_branch)
720 if (!strcmp(name, b->name))
721 return b;
722 return NULL;
725 static struct branch *new_branch(const char *name)
727 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
728 struct branch *b = lookup_branch(name);
730 if (b)
731 die("Invalid attempt to create duplicate branch: %s", name);
732 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
733 die("Branch name doesn't conform to GIT standards: %s", name);
735 b = pool_calloc(1, sizeof(struct branch));
736 b->name = pool_strdup(name);
737 b->table_next_branch = branch_table[hc];
738 b->branch_tree.versions[0].mode = S_IFDIR;
739 b->branch_tree.versions[1].mode = S_IFDIR;
740 b->num_notes = 0;
741 b->active = 0;
742 b->pack_id = MAX_PACK_ID;
743 branch_table[hc] = b;
744 branch_count++;
745 return b;
748 static unsigned int hc_entries(unsigned int cnt)
750 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
751 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
754 static struct tree_content *new_tree_content(unsigned int cnt)
756 struct avail_tree_content *f, *l = NULL;
757 struct tree_content *t;
758 unsigned int hc = hc_entries(cnt);
760 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
761 if (f->entry_capacity >= cnt)
762 break;
764 if (f) {
765 if (l)
766 l->next_avail = f->next_avail;
767 else
768 avail_tree_table[hc] = f->next_avail;
769 } else {
770 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
771 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
772 f->entry_capacity = cnt;
775 t = (struct tree_content*)f;
776 t->entry_count = 0;
777 t->delta_depth = 0;
778 return t;
781 static void release_tree_entry(struct tree_entry *e);
782 static void release_tree_content(struct tree_content *t)
784 struct avail_tree_content *f = (struct avail_tree_content*)t;
785 unsigned int hc = hc_entries(f->entry_capacity);
786 f->next_avail = avail_tree_table[hc];
787 avail_tree_table[hc] = f;
790 static void release_tree_content_recursive(struct tree_content *t)
792 unsigned int i;
793 for (i = 0; i < t->entry_count; i++)
794 release_tree_entry(t->entries[i]);
795 release_tree_content(t);
798 static struct tree_content *grow_tree_content(
799 struct tree_content *t,
800 int amt)
802 struct tree_content *r = new_tree_content(t->entry_count + amt);
803 r->entry_count = t->entry_count;
804 r->delta_depth = t->delta_depth;
805 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
806 release_tree_content(t);
807 return r;
810 static struct tree_entry *new_tree_entry(void)
812 struct tree_entry *e;
814 if (!avail_tree_entry) {
815 unsigned int n = tree_entry_alloc;
816 total_allocd += n * sizeof(struct tree_entry);
817 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
818 while (n-- > 1) {
819 *((void**)e) = e + 1;
820 e++;
822 *((void**)e) = NULL;
825 e = avail_tree_entry;
826 avail_tree_entry = *((void**)e);
827 return e;
830 static void release_tree_entry(struct tree_entry *e)
832 if (e->tree)
833 release_tree_content_recursive(e->tree);
834 *((void**)e) = avail_tree_entry;
835 avail_tree_entry = e;
838 static struct tree_content *dup_tree_content(struct tree_content *s)
840 struct tree_content *d;
841 struct tree_entry *a, *b;
842 unsigned int i;
844 if (!s)
845 return NULL;
846 d = new_tree_content(s->entry_count);
847 for (i = 0; i < s->entry_count; i++) {
848 a = s->entries[i];
849 b = new_tree_entry();
850 memcpy(b, a, sizeof(*a));
851 if (a->tree && is_null_sha1(b->versions[1].sha1))
852 b->tree = dup_tree_content(a->tree);
853 else
854 b->tree = NULL;
855 d->entries[i] = b;
857 d->entry_count = s->entry_count;
858 d->delta_depth = s->delta_depth;
860 return d;
863 static void start_packfile(void)
865 static char tmp_file[PATH_MAX];
866 struct packed_git *p;
867 int namelen;
868 struct pack_header hdr;
869 int pack_fd;
871 pack_fd = odb_mkstemp(tmp_file, sizeof(tmp_file),
872 "pack/tmp_pack_XXXXXX");
873 namelen = strlen(tmp_file) + 2;
874 p = xcalloc(1, sizeof(*p) + namelen);
875 xsnprintf(p->pack_name, namelen, "%s", tmp_file);
876 p->pack_fd = pack_fd;
877 p->do_not_close = 1;
878 pack_file = sha1fd(pack_fd, p->pack_name);
880 hdr.hdr_signature = htonl(PACK_SIGNATURE);
881 hdr.hdr_version = htonl(2);
882 hdr.hdr_entries = 0;
883 sha1write(pack_file, &hdr, sizeof(hdr));
885 pack_data = p;
886 pack_size = sizeof(hdr);
887 object_count = 0;
889 REALLOC_ARRAY(all_packs, pack_id + 1);
890 all_packs[pack_id] = p;
893 static const char *create_index(void)
895 const char *tmpfile;
896 struct pack_idx_entry **idx, **c, **last;
897 struct object_entry *e;
898 struct object_entry_pool *o;
900 /* Build the table of object IDs. */
901 idx = xmalloc(object_count * sizeof(*idx));
902 c = idx;
903 for (o = blocks; o; o = o->next_pool)
904 for (e = o->next_free; e-- != o->entries;)
905 if (pack_id == e->pack_id)
906 *c++ = &e->idx;
907 last = idx + object_count;
908 if (c != last)
909 die("internal consistency error creating the index");
911 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
912 free(idx);
913 return tmpfile;
916 static char *keep_pack(const char *curr_index_name)
918 static char name[PATH_MAX];
919 static const char *keep_msg = "fast-import";
920 int keep_fd;
922 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
923 if (keep_fd < 0)
924 die_errno("cannot create keep file");
925 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
926 if (close(keep_fd))
927 die_errno("failed to write keep file");
929 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
930 get_object_directory(), sha1_to_hex(pack_data->sha1));
931 if (finalize_object_file(pack_data->pack_name, name))
932 die("cannot store pack file");
934 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
935 get_object_directory(), sha1_to_hex(pack_data->sha1));
936 if (finalize_object_file(curr_index_name, name))
937 die("cannot store index file");
938 free((void *)curr_index_name);
939 return name;
942 static void unkeep_all_packs(void)
944 static char name[PATH_MAX];
945 int k;
947 for (k = 0; k < pack_id; k++) {
948 struct packed_git *p = all_packs[k];
949 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
950 get_object_directory(), sha1_to_hex(p->sha1));
951 unlink_or_warn(name);
955 static void end_packfile(void)
957 static int running;
959 if (running || !pack_data)
960 return;
962 running = 1;
963 clear_delta_base_cache();
964 if (object_count) {
965 struct packed_git *new_p;
966 unsigned char cur_pack_sha1[20];
967 char *idx_name;
968 int i;
969 struct branch *b;
970 struct tag *t;
972 close_pack_windows(pack_data);
973 sha1close(pack_file, cur_pack_sha1, 0);
974 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
975 pack_data->pack_name, object_count,
976 cur_pack_sha1, pack_size);
977 close(pack_data->pack_fd);
978 idx_name = keep_pack(create_index());
980 /* Register the packfile with core git's machinery. */
981 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
982 if (!new_p)
983 die("core git rejected index %s", idx_name);
984 all_packs[pack_id] = new_p;
985 install_packed_git(new_p);
987 /* Print the boundary */
988 if (pack_edges) {
989 fprintf(pack_edges, "%s:", new_p->pack_name);
990 for (i = 0; i < branch_table_sz; i++) {
991 for (b = branch_table[i]; b; b = b->table_next_branch) {
992 if (b->pack_id == pack_id)
993 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
996 for (t = first_tag; t; t = t->next_tag) {
997 if (t->pack_id == pack_id)
998 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
1000 fputc('\n', pack_edges);
1001 fflush(pack_edges);
1004 pack_id++;
1006 else {
1007 close(pack_data->pack_fd);
1008 unlink_or_warn(pack_data->pack_name);
1010 free(pack_data);
1011 pack_data = NULL;
1012 running = 0;
1014 /* We can't carry a delta across packfiles. */
1015 strbuf_release(&last_blob.data);
1016 last_blob.offset = 0;
1017 last_blob.depth = 0;
1020 static void cycle_packfile(void)
1022 end_packfile();
1023 start_packfile();
1026 static int store_object(
1027 enum object_type type,
1028 struct strbuf *dat,
1029 struct last_object *last,
1030 unsigned char *sha1out,
1031 uintmax_t mark)
1033 void *out, *delta;
1034 struct object_entry *e;
1035 unsigned char hdr[96];
1036 unsigned char sha1[20];
1037 unsigned long hdrlen, deltalen;
1038 git_SHA_CTX c;
1039 git_zstream s;
1041 hdrlen = xsnprintf((char *)hdr, sizeof(hdr), "%s %lu",
1042 typename(type), (unsigned long)dat->len) + 1;
1043 git_SHA1_Init(&c);
1044 git_SHA1_Update(&c, hdr, hdrlen);
1045 git_SHA1_Update(&c, dat->buf, dat->len);
1046 git_SHA1_Final(sha1, &c);
1047 if (sha1out)
1048 hashcpy(sha1out, sha1);
1050 e = insert_object(sha1);
1051 if (mark)
1052 insert_mark(mark, e);
1053 if (e->idx.offset) {
1054 duplicate_count_by_type[type]++;
1055 return 1;
1056 } else if (find_sha1_pack(sha1, packed_git)) {
1057 e->type = type;
1058 e->pack_id = MAX_PACK_ID;
1059 e->idx.offset = 1; /* just not zero! */
1060 duplicate_count_by_type[type]++;
1061 return 1;
1064 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1065 delta_count_attempts_by_type[type]++;
1066 delta = diff_delta(last->data.buf, last->data.len,
1067 dat->buf, dat->len,
1068 &deltalen, dat->len - 20);
1069 } else
1070 delta = NULL;
1072 git_deflate_init(&s, pack_compression_level);
1073 if (delta) {
1074 s.next_in = delta;
1075 s.avail_in = deltalen;
1076 } else {
1077 s.next_in = (void *)dat->buf;
1078 s.avail_in = dat->len;
1080 s.avail_out = git_deflate_bound(&s, s.avail_in);
1081 s.next_out = out = xmalloc(s.avail_out);
1082 while (git_deflate(&s, Z_FINISH) == Z_OK)
1083 ; /* nothing */
1084 git_deflate_end(&s);
1086 /* Determine if we should auto-checkpoint. */
1087 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1088 || (pack_size + 60 + s.total_out) < pack_size) {
1090 /* This new object needs to *not* have the current pack_id. */
1091 e->pack_id = pack_id + 1;
1092 cycle_packfile();
1094 /* We cannot carry a delta into the new pack. */
1095 if (delta) {
1096 free(delta);
1097 delta = NULL;
1099 git_deflate_init(&s, pack_compression_level);
1100 s.next_in = (void *)dat->buf;
1101 s.avail_in = dat->len;
1102 s.avail_out = git_deflate_bound(&s, s.avail_in);
1103 s.next_out = out = xrealloc(out, s.avail_out);
1104 while (git_deflate(&s, Z_FINISH) == Z_OK)
1105 ; /* nothing */
1106 git_deflate_end(&s);
1110 e->type = type;
1111 e->pack_id = pack_id;
1112 e->idx.offset = pack_size;
1113 object_count++;
1114 object_count_by_type[type]++;
1116 crc32_begin(pack_file);
1118 if (delta) {
1119 off_t ofs = e->idx.offset - last->offset;
1120 unsigned pos = sizeof(hdr) - 1;
1122 delta_count_by_type[type]++;
1123 e->depth = last->depth + 1;
1125 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1126 sha1write(pack_file, hdr, hdrlen);
1127 pack_size += hdrlen;
1129 hdr[pos] = ofs & 127;
1130 while (ofs >>= 7)
1131 hdr[--pos] = 128 | (--ofs & 127);
1132 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1133 pack_size += sizeof(hdr) - pos;
1134 } else {
1135 e->depth = 0;
1136 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1137 sha1write(pack_file, hdr, hdrlen);
1138 pack_size += hdrlen;
1141 sha1write(pack_file, out, s.total_out);
1142 pack_size += s.total_out;
1144 e->idx.crc32 = crc32_end(pack_file);
1146 free(out);
1147 free(delta);
1148 if (last) {
1149 if (last->no_swap) {
1150 last->data = *dat;
1151 } else {
1152 strbuf_swap(&last->data, dat);
1154 last->offset = e->idx.offset;
1155 last->depth = e->depth;
1157 return 0;
1160 static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1162 if (sha1file_truncate(pack_file, checkpoint))
1163 die_errno("cannot truncate pack to skip duplicate");
1164 pack_size = checkpoint->offset;
1167 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1169 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1170 unsigned char *in_buf = xmalloc(in_sz);
1171 unsigned char *out_buf = xmalloc(out_sz);
1172 struct object_entry *e;
1173 unsigned char sha1[20];
1174 unsigned long hdrlen;
1175 off_t offset;
1176 git_SHA_CTX c;
1177 git_zstream s;
1178 struct sha1file_checkpoint checkpoint;
1179 int status = Z_OK;
1181 /* Determine if we should auto-checkpoint. */
1182 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1183 || (pack_size + 60 + len) < pack_size)
1184 cycle_packfile();
1186 sha1file_checkpoint(pack_file, &checkpoint);
1187 offset = checkpoint.offset;
1189 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1190 if (out_sz <= hdrlen)
1191 die("impossibly large object header");
1193 git_SHA1_Init(&c);
1194 git_SHA1_Update(&c, out_buf, hdrlen);
1196 crc32_begin(pack_file);
1198 git_deflate_init(&s, pack_compression_level);
1200 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1201 if (out_sz <= hdrlen)
1202 die("impossibly large object header");
1204 s.next_out = out_buf + hdrlen;
1205 s.avail_out = out_sz - hdrlen;
1207 while (status != Z_STREAM_END) {
1208 if (0 < len && !s.avail_in) {
1209 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1210 size_t n = fread(in_buf, 1, cnt, stdin);
1211 if (!n && feof(stdin))
1212 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1214 git_SHA1_Update(&c, in_buf, n);
1215 s.next_in = in_buf;
1216 s.avail_in = n;
1217 len -= n;
1220 status = git_deflate(&s, len ? 0 : Z_FINISH);
1222 if (!s.avail_out || status == Z_STREAM_END) {
1223 size_t n = s.next_out - out_buf;
1224 sha1write(pack_file, out_buf, n);
1225 pack_size += n;
1226 s.next_out = out_buf;
1227 s.avail_out = out_sz;
1230 switch (status) {
1231 case Z_OK:
1232 case Z_BUF_ERROR:
1233 case Z_STREAM_END:
1234 continue;
1235 default:
1236 die("unexpected deflate failure: %d", status);
1239 git_deflate_end(&s);
1240 git_SHA1_Final(sha1, &c);
1242 if (sha1out)
1243 hashcpy(sha1out, sha1);
1245 e = insert_object(sha1);
1247 if (mark)
1248 insert_mark(mark, e);
1250 if (e->idx.offset) {
1251 duplicate_count_by_type[OBJ_BLOB]++;
1252 truncate_pack(&checkpoint);
1254 } else if (find_sha1_pack(sha1, packed_git)) {
1255 e->type = OBJ_BLOB;
1256 e->pack_id = MAX_PACK_ID;
1257 e->idx.offset = 1; /* just not zero! */
1258 duplicate_count_by_type[OBJ_BLOB]++;
1259 truncate_pack(&checkpoint);
1261 } else {
1262 e->depth = 0;
1263 e->type = OBJ_BLOB;
1264 e->pack_id = pack_id;
1265 e->idx.offset = offset;
1266 e->idx.crc32 = crc32_end(pack_file);
1267 object_count++;
1268 object_count_by_type[OBJ_BLOB]++;
1271 free(in_buf);
1272 free(out_buf);
1275 /* All calls must be guarded by find_object() or find_mark() to
1276 * ensure the 'struct object_entry' passed was written by this
1277 * process instance. We unpack the entry by the offset, avoiding
1278 * the need for the corresponding .idx file. This unpacking rule
1279 * works because we only use OBJ_REF_DELTA within the packfiles
1280 * created by fast-import.
1282 * oe must not be NULL. Such an oe usually comes from giving
1283 * an unknown SHA-1 to find_object() or an undefined mark to
1284 * find_mark(). Callers must test for this condition and use
1285 * the standard read_sha1_file() when it happens.
1287 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1288 * find_mark(), where the mark was reloaded from an existing marks
1289 * file and is referencing an object that this fast-import process
1290 * instance did not write out to a packfile. Callers must test for
1291 * this condition and use read_sha1_file() instead.
1293 static void *gfi_unpack_entry(
1294 struct object_entry *oe,
1295 unsigned long *sizep)
1297 enum object_type type;
1298 struct packed_git *p = all_packs[oe->pack_id];
1299 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1300 /* The object is stored in the packfile we are writing to
1301 * and we have modified it since the last time we scanned
1302 * back to read a previously written object. If an old
1303 * window covered [p->pack_size, p->pack_size + 20) its
1304 * data is stale and is not valid. Closing all windows
1305 * and updating the packfile length ensures we can read
1306 * the newly written data.
1308 close_pack_windows(p);
1309 sha1flush(pack_file);
1311 /* We have to offer 20 bytes additional on the end of
1312 * the packfile as the core unpacker code assumes the
1313 * footer is present at the file end and must promise
1314 * at least 20 bytes within any window it maps. But
1315 * we don't actually create the footer here.
1317 p->pack_size = pack_size + 20;
1319 return unpack_entry(p, oe->idx.offset, &type, sizep);
1322 static const char *get_mode(const char *str, uint16_t *modep)
1324 unsigned char c;
1325 uint16_t mode = 0;
1327 while ((c = *str++) != ' ') {
1328 if (c < '0' || c > '7')
1329 return NULL;
1330 mode = (mode << 3) + (c - '0');
1332 *modep = mode;
1333 return str;
1336 static void load_tree(struct tree_entry *root)
1338 unsigned char *sha1 = root->versions[1].sha1;
1339 struct object_entry *myoe;
1340 struct tree_content *t;
1341 unsigned long size;
1342 char *buf;
1343 const char *c;
1345 root->tree = t = new_tree_content(8);
1346 if (is_null_sha1(sha1))
1347 return;
1349 myoe = find_object(sha1);
1350 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1351 if (myoe->type != OBJ_TREE)
1352 die("Not a tree: %s", sha1_to_hex(sha1));
1353 t->delta_depth = myoe->depth;
1354 buf = gfi_unpack_entry(myoe, &size);
1355 if (!buf)
1356 die("Can't load tree %s", sha1_to_hex(sha1));
1357 } else {
1358 enum object_type type;
1359 buf = read_sha1_file(sha1, &type, &size);
1360 if (!buf || type != OBJ_TREE)
1361 die("Can't load tree %s", sha1_to_hex(sha1));
1364 c = buf;
1365 while (c != (buf + size)) {
1366 struct tree_entry *e = new_tree_entry();
1368 if (t->entry_count == t->entry_capacity)
1369 root->tree = t = grow_tree_content(t, t->entry_count);
1370 t->entries[t->entry_count++] = e;
1372 e->tree = NULL;
1373 c = get_mode(c, &e->versions[1].mode);
1374 if (!c)
1375 die("Corrupt mode in %s", sha1_to_hex(sha1));
1376 e->versions[0].mode = e->versions[1].mode;
1377 e->name = to_atom(c, strlen(c));
1378 c += e->name->str_len + 1;
1379 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1380 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1381 c += 20;
1383 free(buf);
1386 static int tecmp0 (const void *_a, const void *_b)
1388 struct tree_entry *a = *((struct tree_entry**)_a);
1389 struct tree_entry *b = *((struct tree_entry**)_b);
1390 return base_name_compare(
1391 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1392 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1395 static int tecmp1 (const void *_a, const void *_b)
1397 struct tree_entry *a = *((struct tree_entry**)_a);
1398 struct tree_entry *b = *((struct tree_entry**)_b);
1399 return base_name_compare(
1400 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1401 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1404 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1406 size_t maxlen = 0;
1407 unsigned int i;
1409 if (!v)
1410 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1411 else
1412 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1414 for (i = 0; i < t->entry_count; i++) {
1415 if (t->entries[i]->versions[v].mode)
1416 maxlen += t->entries[i]->name->str_len + 34;
1419 strbuf_reset(b);
1420 strbuf_grow(b, maxlen);
1421 for (i = 0; i < t->entry_count; i++) {
1422 struct tree_entry *e = t->entries[i];
1423 if (!e->versions[v].mode)
1424 continue;
1425 strbuf_addf(b, "%o %s%c",
1426 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1427 e->name->str_dat, '\0');
1428 strbuf_add(b, e->versions[v].sha1, 20);
1432 static void store_tree(struct tree_entry *root)
1434 struct tree_content *t;
1435 unsigned int i, j, del;
1436 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1437 struct object_entry *le = NULL;
1439 if (!is_null_sha1(root->versions[1].sha1))
1440 return;
1442 if (!root->tree)
1443 load_tree(root);
1444 t = root->tree;
1446 for (i = 0; i < t->entry_count; i++) {
1447 if (t->entries[i]->tree)
1448 store_tree(t->entries[i]);
1451 if (!(root->versions[0].mode & NO_DELTA))
1452 le = find_object(root->versions[0].sha1);
1453 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1454 mktree(t, 0, &old_tree);
1455 lo.data = old_tree;
1456 lo.offset = le->idx.offset;
1457 lo.depth = t->delta_depth;
1460 mktree(t, 1, &new_tree);
1461 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1463 t->delta_depth = lo.depth;
1464 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1465 struct tree_entry *e = t->entries[i];
1466 if (e->versions[1].mode) {
1467 e->versions[0].mode = e->versions[1].mode;
1468 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1469 t->entries[j++] = e;
1470 } else {
1471 release_tree_entry(e);
1472 del++;
1475 t->entry_count -= del;
1478 static void tree_content_replace(
1479 struct tree_entry *root,
1480 const unsigned char *sha1,
1481 const uint16_t mode,
1482 struct tree_content *newtree)
1484 if (!S_ISDIR(mode))
1485 die("Root cannot be a non-directory");
1486 hashclr(root->versions[0].sha1);
1487 hashcpy(root->versions[1].sha1, sha1);
1488 if (root->tree)
1489 release_tree_content_recursive(root->tree);
1490 root->tree = newtree;
1493 static int tree_content_set(
1494 struct tree_entry *root,
1495 const char *p,
1496 const unsigned char *sha1,
1497 const uint16_t mode,
1498 struct tree_content *subtree)
1500 struct tree_content *t;
1501 const char *slash1;
1502 unsigned int i, n;
1503 struct tree_entry *e;
1505 slash1 = strchrnul(p, '/');
1506 n = slash1 - p;
1507 if (!n)
1508 die("Empty path component found in input");
1509 if (!*slash1 && !S_ISDIR(mode) && subtree)
1510 die("Non-directories cannot have subtrees");
1512 if (!root->tree)
1513 load_tree(root);
1514 t = root->tree;
1515 for (i = 0; i < t->entry_count; i++) {
1516 e = t->entries[i];
1517 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1518 if (!*slash1) {
1519 if (!S_ISDIR(mode)
1520 && e->versions[1].mode == mode
1521 && !hashcmp(e->versions[1].sha1, sha1))
1522 return 0;
1523 e->versions[1].mode = mode;
1524 hashcpy(e->versions[1].sha1, sha1);
1525 if (e->tree)
1526 release_tree_content_recursive(e->tree);
1527 e->tree = subtree;
1530 * We need to leave e->versions[0].sha1 alone
1531 * to avoid modifying the preimage tree used
1532 * when writing out the parent directory.
1533 * But after replacing the subdir with a
1534 * completely different one, it's not a good
1535 * delta base any more, and besides, we've
1536 * thrown away the tree entries needed to
1537 * make a delta against it.
1539 * So let's just explicitly disable deltas
1540 * for the subtree.
1542 if (S_ISDIR(e->versions[0].mode))
1543 e->versions[0].mode |= NO_DELTA;
1545 hashclr(root->versions[1].sha1);
1546 return 1;
1548 if (!S_ISDIR(e->versions[1].mode)) {
1549 e->tree = new_tree_content(8);
1550 e->versions[1].mode = S_IFDIR;
1552 if (!e->tree)
1553 load_tree(e);
1554 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1555 hashclr(root->versions[1].sha1);
1556 return 1;
1558 return 0;
1562 if (t->entry_count == t->entry_capacity)
1563 root->tree = t = grow_tree_content(t, t->entry_count);
1564 e = new_tree_entry();
1565 e->name = to_atom(p, n);
1566 e->versions[0].mode = 0;
1567 hashclr(e->versions[0].sha1);
1568 t->entries[t->entry_count++] = e;
1569 if (*slash1) {
1570 e->tree = new_tree_content(8);
1571 e->versions[1].mode = S_IFDIR;
1572 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1573 } else {
1574 e->tree = subtree;
1575 e->versions[1].mode = mode;
1576 hashcpy(e->versions[1].sha1, sha1);
1578 hashclr(root->versions[1].sha1);
1579 return 1;
1582 static int tree_content_remove(
1583 struct tree_entry *root,
1584 const char *p,
1585 struct tree_entry *backup_leaf,
1586 int allow_root)
1588 struct tree_content *t;
1589 const char *slash1;
1590 unsigned int i, n;
1591 struct tree_entry *e;
1593 slash1 = strchrnul(p, '/');
1594 n = slash1 - p;
1596 if (!root->tree)
1597 load_tree(root);
1599 if (!*p && allow_root) {
1600 e = root;
1601 goto del_entry;
1604 t = root->tree;
1605 for (i = 0; i < t->entry_count; i++) {
1606 e = t->entries[i];
1607 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1608 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1610 * If p names a file in some subdirectory, and a
1611 * file or symlink matching the name of the
1612 * parent directory of p exists, then p cannot
1613 * exist and need not be deleted.
1615 return 1;
1616 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1617 goto del_entry;
1618 if (!e->tree)
1619 load_tree(e);
1620 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1621 for (n = 0; n < e->tree->entry_count; n++) {
1622 if (e->tree->entries[n]->versions[1].mode) {
1623 hashclr(root->versions[1].sha1);
1624 return 1;
1627 backup_leaf = NULL;
1628 goto del_entry;
1630 return 0;
1633 return 0;
1635 del_entry:
1636 if (backup_leaf)
1637 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1638 else if (e->tree)
1639 release_tree_content_recursive(e->tree);
1640 e->tree = NULL;
1641 e->versions[1].mode = 0;
1642 hashclr(e->versions[1].sha1);
1643 hashclr(root->versions[1].sha1);
1644 return 1;
1647 static int tree_content_get(
1648 struct tree_entry *root,
1649 const char *p,
1650 struct tree_entry *leaf,
1651 int allow_root)
1653 struct tree_content *t;
1654 const char *slash1;
1655 unsigned int i, n;
1656 struct tree_entry *e;
1658 slash1 = strchrnul(p, '/');
1659 n = slash1 - p;
1660 if (!n && !allow_root)
1661 die("Empty path component found in input");
1663 if (!root->tree)
1664 load_tree(root);
1666 if (!n) {
1667 e = root;
1668 goto found_entry;
1671 t = root->tree;
1672 for (i = 0; i < t->entry_count; i++) {
1673 e = t->entries[i];
1674 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1675 if (!*slash1)
1676 goto found_entry;
1677 if (!S_ISDIR(e->versions[1].mode))
1678 return 0;
1679 if (!e->tree)
1680 load_tree(e);
1681 return tree_content_get(e, slash1 + 1, leaf, 0);
1684 return 0;
1686 found_entry:
1687 memcpy(leaf, e, sizeof(*leaf));
1688 if (e->tree && is_null_sha1(e->versions[1].sha1))
1689 leaf->tree = dup_tree_content(e->tree);
1690 else
1691 leaf->tree = NULL;
1692 return 1;
1695 static int update_branch(struct branch *b)
1697 static const char *msg = "fast-import";
1698 struct ref_transaction *transaction;
1699 unsigned char old_sha1[20];
1700 struct strbuf err = STRBUF_INIT;
1702 if (is_null_sha1(b->sha1)) {
1703 if (b->delete)
1704 delete_ref(b->name, NULL, 0);
1705 return 0;
1707 if (read_ref(b->name, old_sha1))
1708 hashclr(old_sha1);
1709 if (!force_update && !is_null_sha1(old_sha1)) {
1710 struct commit *old_cmit, *new_cmit;
1712 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1713 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1714 if (!old_cmit || !new_cmit)
1715 return error("Branch %s is missing commits.", b->name);
1717 if (!in_merge_bases(old_cmit, new_cmit)) {
1718 warning("Not updating %s"
1719 " (new tip %s does not contain %s)",
1720 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1721 return -1;
1724 transaction = ref_transaction_begin(&err);
1725 if (!transaction ||
1726 ref_transaction_update(transaction, b->name, b->sha1, old_sha1,
1727 0, msg, &err) ||
1728 ref_transaction_commit(transaction, &err)) {
1729 ref_transaction_free(transaction);
1730 error("%s", err.buf);
1731 strbuf_release(&err);
1732 return -1;
1734 ref_transaction_free(transaction);
1735 strbuf_release(&err);
1736 return 0;
1739 static void dump_branches(void)
1741 unsigned int i;
1742 struct branch *b;
1744 for (i = 0; i < branch_table_sz; i++) {
1745 for (b = branch_table[i]; b; b = b->table_next_branch)
1746 failure |= update_branch(b);
1750 static void dump_tags(void)
1752 static const char *msg = "fast-import";
1753 struct tag *t;
1754 struct strbuf ref_name = STRBUF_INIT;
1755 struct strbuf err = STRBUF_INIT;
1756 struct ref_transaction *transaction;
1758 transaction = ref_transaction_begin(&err);
1759 if (!transaction) {
1760 failure |= error("%s", err.buf);
1761 goto cleanup;
1763 for (t = first_tag; t; t = t->next_tag) {
1764 strbuf_reset(&ref_name);
1765 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1767 if (ref_transaction_update(transaction, ref_name.buf,
1768 t->sha1, NULL, 0, msg, &err)) {
1769 failure |= error("%s", err.buf);
1770 goto cleanup;
1773 if (ref_transaction_commit(transaction, &err))
1774 failure |= error("%s", err.buf);
1776 cleanup:
1777 ref_transaction_free(transaction);
1778 strbuf_release(&ref_name);
1779 strbuf_release(&err);
1782 static void dump_marks_helper(FILE *f,
1783 uintmax_t base,
1784 struct mark_set *m)
1786 uintmax_t k;
1787 if (m->shift) {
1788 for (k = 0; k < 1024; k++) {
1789 if (m->data.sets[k])
1790 dump_marks_helper(f, base + (k << m->shift),
1791 m->data.sets[k]);
1793 } else {
1794 for (k = 0; k < 1024; k++) {
1795 if (m->data.marked[k])
1796 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1797 sha1_to_hex(m->data.marked[k]->idx.sha1));
1802 static void dump_marks(void)
1804 static struct lock_file mark_lock;
1805 FILE *f;
1807 if (!export_marks_file)
1808 return;
1810 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1811 failure |= error("Unable to write marks file %s: %s",
1812 export_marks_file, strerror(errno));
1813 return;
1816 f = fdopen_lock_file(&mark_lock, "w");
1817 if (!f) {
1818 int saved_errno = errno;
1819 rollback_lock_file(&mark_lock);
1820 failure |= error("Unable to write marks file %s: %s",
1821 export_marks_file, strerror(saved_errno));
1822 return;
1825 dump_marks_helper(f, 0, marks);
1826 if (commit_lock_file(&mark_lock)) {
1827 failure |= error("Unable to commit marks file %s: %s",
1828 export_marks_file, strerror(errno));
1829 return;
1833 static void read_marks(void)
1835 char line[512];
1836 FILE *f = fopen(import_marks_file, "r");
1837 if (f)
1839 else if (import_marks_file_ignore_missing && errno == ENOENT)
1840 return; /* Marks file does not exist */
1841 else
1842 die_errno("cannot read '%s'", import_marks_file);
1843 while (fgets(line, sizeof(line), f)) {
1844 uintmax_t mark;
1845 char *end;
1846 unsigned char sha1[20];
1847 struct object_entry *e;
1849 end = strchr(line, '\n');
1850 if (line[0] != ':' || !end)
1851 die("corrupt mark line: %s", line);
1852 *end = 0;
1853 mark = strtoumax(line + 1, &end, 10);
1854 if (!mark || end == line + 1
1855 || *end != ' ' || get_sha1_hex(end + 1, sha1))
1856 die("corrupt mark line: %s", line);
1857 e = find_object(sha1);
1858 if (!e) {
1859 enum object_type type = sha1_object_info(sha1, NULL);
1860 if (type < 0)
1861 die("object not found: %s", sha1_to_hex(sha1));
1862 e = insert_object(sha1);
1863 e->type = type;
1864 e->pack_id = MAX_PACK_ID;
1865 e->idx.offset = 1; /* just not zero! */
1867 insert_mark(mark, e);
1869 fclose(f);
1873 static int read_next_command(void)
1875 static int stdin_eof = 0;
1877 if (stdin_eof) {
1878 unread_command_buf = 0;
1879 return EOF;
1882 for (;;) {
1883 const char *p;
1885 if (unread_command_buf) {
1886 unread_command_buf = 0;
1887 } else {
1888 struct recent_command *rc;
1890 strbuf_detach(&command_buf, NULL);
1891 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1892 if (stdin_eof)
1893 return EOF;
1895 if (!seen_data_command
1896 && !starts_with(command_buf.buf, "feature ")
1897 && !starts_with(command_buf.buf, "option ")) {
1898 parse_argv();
1901 rc = rc_free;
1902 if (rc)
1903 rc_free = rc->next;
1904 else {
1905 rc = cmd_hist.next;
1906 cmd_hist.next = rc->next;
1907 cmd_hist.next->prev = &cmd_hist;
1908 free(rc->buf);
1911 rc->buf = command_buf.buf;
1912 rc->prev = cmd_tail;
1913 rc->next = cmd_hist.prev;
1914 rc->prev->next = rc;
1915 cmd_tail = rc;
1917 if (skip_prefix(command_buf.buf, "get-mark ", &p)) {
1918 parse_get_mark(p);
1919 continue;
1921 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1922 parse_cat_blob(p);
1923 continue;
1925 if (command_buf.buf[0] == '#')
1926 continue;
1927 return 0;
1931 static void skip_optional_lf(void)
1933 int term_char = fgetc(stdin);
1934 if (term_char != '\n' && term_char != EOF)
1935 ungetc(term_char, stdin);
1938 static void parse_mark(void)
1940 const char *v;
1941 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1942 next_mark = strtoumax(v, NULL, 10);
1943 read_next_command();
1945 else
1946 next_mark = 0;
1949 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1951 const char *data;
1952 strbuf_reset(sb);
1954 if (!skip_prefix(command_buf.buf, "data ", &data))
1955 die("Expected 'data n' command, found: %s", command_buf.buf);
1957 if (skip_prefix(data, "<<", &data)) {
1958 char *term = xstrdup(data);
1959 size_t term_len = command_buf.len - (data - command_buf.buf);
1961 strbuf_detach(&command_buf, NULL);
1962 for (;;) {
1963 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1964 die("EOF in data (terminator '%s' not found)", term);
1965 if (term_len == command_buf.len
1966 && !strcmp(term, command_buf.buf))
1967 break;
1968 strbuf_addbuf(sb, &command_buf);
1969 strbuf_addch(sb, '\n');
1971 free(term);
1973 else {
1974 uintmax_t len = strtoumax(data, NULL, 10);
1975 size_t n = 0, length = (size_t)len;
1977 if (limit && limit < len) {
1978 *len_res = len;
1979 return 0;
1981 if (length < len)
1982 die("data is too large to use in this context");
1984 while (n < length) {
1985 size_t s = strbuf_fread(sb, length - n, stdin);
1986 if (!s && feof(stdin))
1987 die("EOF in data (%lu bytes remaining)",
1988 (unsigned long)(length - n));
1989 n += s;
1993 skip_optional_lf();
1994 return 1;
1997 static int validate_raw_date(const char *src, struct strbuf *result)
1999 const char *orig_src = src;
2000 char *endp;
2001 unsigned long num;
2003 errno = 0;
2005 num = strtoul(src, &endp, 10);
2006 /* NEEDSWORK: perhaps check for reasonable values? */
2007 if (errno || endp == src || *endp != ' ')
2008 return -1;
2010 src = endp + 1;
2011 if (*src != '-' && *src != '+')
2012 return -1;
2014 num = strtoul(src + 1, &endp, 10);
2015 if (errno || endp == src + 1 || *endp || 1400 < num)
2016 return -1;
2018 strbuf_addstr(result, orig_src);
2019 return 0;
2022 static char *parse_ident(const char *buf)
2024 const char *ltgt;
2025 size_t name_len;
2026 struct strbuf ident = STRBUF_INIT;
2028 /* ensure there is a space delimiter even if there is no name */
2029 if (*buf == '<')
2030 --buf;
2032 ltgt = buf + strcspn(buf, "<>");
2033 if (*ltgt != '<')
2034 die("Missing < in ident string: %s", buf);
2035 if (ltgt != buf && ltgt[-1] != ' ')
2036 die("Missing space before < in ident string: %s", buf);
2037 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2038 if (*ltgt != '>')
2039 die("Missing > in ident string: %s", buf);
2040 ltgt++;
2041 if (*ltgt != ' ')
2042 die("Missing space after > in ident string: %s", buf);
2043 ltgt++;
2044 name_len = ltgt - buf;
2045 strbuf_add(&ident, buf, name_len);
2047 switch (whenspec) {
2048 case WHENSPEC_RAW:
2049 if (validate_raw_date(ltgt, &ident) < 0)
2050 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2051 break;
2052 case WHENSPEC_RFC2822:
2053 if (parse_date(ltgt, &ident) < 0)
2054 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2055 break;
2056 case WHENSPEC_NOW:
2057 if (strcmp("now", ltgt))
2058 die("Date in ident must be 'now': %s", buf);
2059 datestamp(&ident);
2060 break;
2063 return strbuf_detach(&ident, NULL);
2066 static void parse_and_store_blob(
2067 struct last_object *last,
2068 unsigned char *sha1out,
2069 uintmax_t mark)
2071 static struct strbuf buf = STRBUF_INIT;
2072 uintmax_t len;
2074 if (parse_data(&buf, big_file_threshold, &len))
2075 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2076 else {
2077 if (last) {
2078 strbuf_release(&last->data);
2079 last->offset = 0;
2080 last->depth = 0;
2082 stream_blob(len, sha1out, mark);
2083 skip_optional_lf();
2087 static void parse_new_blob(void)
2089 read_next_command();
2090 parse_mark();
2091 parse_and_store_blob(&last_blob, NULL, next_mark);
2094 static void unload_one_branch(void)
2096 while (cur_active_branches
2097 && cur_active_branches >= max_active_branches) {
2098 uintmax_t min_commit = ULONG_MAX;
2099 struct branch *e, *l = NULL, *p = NULL;
2101 for (e = active_branches; e; e = e->active_next_branch) {
2102 if (e->last_commit < min_commit) {
2103 p = l;
2104 min_commit = e->last_commit;
2106 l = e;
2109 if (p) {
2110 e = p->active_next_branch;
2111 p->active_next_branch = e->active_next_branch;
2112 } else {
2113 e = active_branches;
2114 active_branches = e->active_next_branch;
2116 e->active = 0;
2117 e->active_next_branch = NULL;
2118 if (e->branch_tree.tree) {
2119 release_tree_content_recursive(e->branch_tree.tree);
2120 e->branch_tree.tree = NULL;
2122 cur_active_branches--;
2126 static void load_branch(struct branch *b)
2128 load_tree(&b->branch_tree);
2129 if (!b->active) {
2130 b->active = 1;
2131 b->active_next_branch = active_branches;
2132 active_branches = b;
2133 cur_active_branches++;
2134 branch_load_count++;
2138 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2140 unsigned char fanout = 0;
2141 while ((num_notes >>= 8))
2142 fanout++;
2143 return fanout;
2146 static void construct_path_with_fanout(const char *hex_sha1,
2147 unsigned char fanout, char *path)
2149 unsigned int i = 0, j = 0;
2150 if (fanout >= 20)
2151 die("Too large fanout (%u)", fanout);
2152 while (fanout) {
2153 path[i++] = hex_sha1[j++];
2154 path[i++] = hex_sha1[j++];
2155 path[i++] = '/';
2156 fanout--;
2158 memcpy(path + i, hex_sha1 + j, 40 - j);
2159 path[i + 40 - j] = '\0';
2162 static uintmax_t do_change_note_fanout(
2163 struct tree_entry *orig_root, struct tree_entry *root,
2164 char *hex_sha1, unsigned int hex_sha1_len,
2165 char *fullpath, unsigned int fullpath_len,
2166 unsigned char fanout)
2168 struct tree_content *t = root->tree;
2169 struct tree_entry *e, leaf;
2170 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2171 uintmax_t num_notes = 0;
2172 unsigned char sha1[20];
2173 char realpath[60];
2175 for (i = 0; t && i < t->entry_count; i++) {
2176 e = t->entries[i];
2177 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2178 tmp_fullpath_len = fullpath_len;
2181 * We're interested in EITHER existing note entries (entries
2182 * with exactly 40 hex chars in path, not including directory
2183 * separators), OR directory entries that may contain note
2184 * entries (with < 40 hex chars in path).
2185 * Also, each path component in a note entry must be a multiple
2186 * of 2 chars.
2188 if (!e->versions[1].mode ||
2189 tmp_hex_sha1_len > 40 ||
2190 e->name->str_len % 2)
2191 continue;
2193 /* This _may_ be a note entry, or a subdir containing notes */
2194 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2195 e->name->str_len);
2196 if (tmp_fullpath_len)
2197 fullpath[tmp_fullpath_len++] = '/';
2198 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2199 e->name->str_len);
2200 tmp_fullpath_len += e->name->str_len;
2201 fullpath[tmp_fullpath_len] = '\0';
2203 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2204 /* This is a note entry */
2205 if (fanout == 0xff) {
2206 /* Counting mode, no rename */
2207 num_notes++;
2208 continue;
2210 construct_path_with_fanout(hex_sha1, fanout, realpath);
2211 if (!strcmp(fullpath, realpath)) {
2212 /* Note entry is in correct location */
2213 num_notes++;
2214 continue;
2217 /* Rename fullpath to realpath */
2218 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2219 die("Failed to remove path %s", fullpath);
2220 tree_content_set(orig_root, realpath,
2221 leaf.versions[1].sha1,
2222 leaf.versions[1].mode,
2223 leaf.tree);
2224 } else if (S_ISDIR(e->versions[1].mode)) {
2225 /* This is a subdir that may contain note entries */
2226 if (!e->tree)
2227 load_tree(e);
2228 num_notes += do_change_note_fanout(orig_root, e,
2229 hex_sha1, tmp_hex_sha1_len,
2230 fullpath, tmp_fullpath_len, fanout);
2233 /* The above may have reallocated the current tree_content */
2234 t = root->tree;
2236 return num_notes;
2239 static uintmax_t change_note_fanout(struct tree_entry *root,
2240 unsigned char fanout)
2242 char hex_sha1[40], path[60];
2243 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2247 * Given a pointer into a string, parse a mark reference:
2249 * idnum ::= ':' bigint;
2251 * Return the first character after the value in *endptr.
2253 * Complain if the following character is not what is expected,
2254 * either a space or end of the string.
2256 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2258 uintmax_t mark;
2260 assert(*p == ':');
2261 p++;
2262 mark = strtoumax(p, endptr, 10);
2263 if (*endptr == p)
2264 die("No value after ':' in mark: %s", command_buf.buf);
2265 return mark;
2269 * Parse the mark reference, and complain if this is not the end of
2270 * the string.
2272 static uintmax_t parse_mark_ref_eol(const char *p)
2274 char *end;
2275 uintmax_t mark;
2277 mark = parse_mark_ref(p, &end);
2278 if (*end != '\0')
2279 die("Garbage after mark: %s", command_buf.buf);
2280 return mark;
2284 * Parse the mark reference, demanding a trailing space. Return a
2285 * pointer to the space.
2287 static uintmax_t parse_mark_ref_space(const char **p)
2289 uintmax_t mark;
2290 char *end;
2292 mark = parse_mark_ref(*p, &end);
2293 if (*end++ != ' ')
2294 die("Missing space after mark: %s", command_buf.buf);
2295 *p = end;
2296 return mark;
2299 static void file_change_m(const char *p, struct branch *b)
2301 static struct strbuf uq = STRBUF_INIT;
2302 const char *endp;
2303 struct object_entry *oe;
2304 unsigned char sha1[20];
2305 uint16_t mode, inline_data = 0;
2307 p = get_mode(p, &mode);
2308 if (!p)
2309 die("Corrupt mode: %s", command_buf.buf);
2310 switch (mode) {
2311 case 0644:
2312 case 0755:
2313 mode |= S_IFREG;
2314 case S_IFREG | 0644:
2315 case S_IFREG | 0755:
2316 case S_IFLNK:
2317 case S_IFDIR:
2318 case S_IFGITLINK:
2319 /* ok */
2320 break;
2321 default:
2322 die("Corrupt mode: %s", command_buf.buf);
2325 if (*p == ':') {
2326 oe = find_mark(parse_mark_ref_space(&p));
2327 hashcpy(sha1, oe->idx.sha1);
2328 } else if (skip_prefix(p, "inline ", &p)) {
2329 inline_data = 1;
2330 oe = NULL; /* not used with inline_data, but makes gcc happy */
2331 } else {
2332 if (get_sha1_hex(p, sha1))
2333 die("Invalid dataref: %s", command_buf.buf);
2334 oe = find_object(sha1);
2335 p += 40;
2336 if (*p++ != ' ')
2337 die("Missing space after SHA1: %s", command_buf.buf);
2340 strbuf_reset(&uq);
2341 if (!unquote_c_style(&uq, p, &endp)) {
2342 if (*endp)
2343 die("Garbage after path in: %s", command_buf.buf);
2344 p = uq.buf;
2347 /* Git does not track empty, non-toplevel directories. */
2348 if (S_ISDIR(mode) && !hashcmp(sha1, EMPTY_TREE_SHA1_BIN) && *p) {
2349 tree_content_remove(&b->branch_tree, p, NULL, 0);
2350 return;
2353 if (S_ISGITLINK(mode)) {
2354 if (inline_data)
2355 die("Git links cannot be specified 'inline': %s",
2356 command_buf.buf);
2357 else if (oe) {
2358 if (oe->type != OBJ_COMMIT)
2359 die("Not a commit (actually a %s): %s",
2360 typename(oe->type), command_buf.buf);
2363 * Accept the sha1 without checking; it expected to be in
2364 * another repository.
2366 } else if (inline_data) {
2367 if (S_ISDIR(mode))
2368 die("Directories cannot be specified 'inline': %s",
2369 command_buf.buf);
2370 if (p != uq.buf) {
2371 strbuf_addstr(&uq, p);
2372 p = uq.buf;
2374 read_next_command();
2375 parse_and_store_blob(&last_blob, sha1, 0);
2376 } else {
2377 enum object_type expected = S_ISDIR(mode) ?
2378 OBJ_TREE: OBJ_BLOB;
2379 enum object_type type = oe ? oe->type :
2380 sha1_object_info(sha1, NULL);
2381 if (type < 0)
2382 die("%s not found: %s",
2383 S_ISDIR(mode) ? "Tree" : "Blob",
2384 command_buf.buf);
2385 if (type != expected)
2386 die("Not a %s (actually a %s): %s",
2387 typename(expected), typename(type),
2388 command_buf.buf);
2391 if (!*p) {
2392 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2393 return;
2395 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2398 static void file_change_d(const char *p, struct branch *b)
2400 static struct strbuf uq = STRBUF_INIT;
2401 const char *endp;
2403 strbuf_reset(&uq);
2404 if (!unquote_c_style(&uq, p, &endp)) {
2405 if (*endp)
2406 die("Garbage after path in: %s", command_buf.buf);
2407 p = uq.buf;
2409 tree_content_remove(&b->branch_tree, p, NULL, 1);
2412 static void file_change_cr(const char *s, struct branch *b, int rename)
2414 const char *d;
2415 static struct strbuf s_uq = STRBUF_INIT;
2416 static struct strbuf d_uq = STRBUF_INIT;
2417 const char *endp;
2418 struct tree_entry leaf;
2420 strbuf_reset(&s_uq);
2421 if (!unquote_c_style(&s_uq, s, &endp)) {
2422 if (*endp != ' ')
2423 die("Missing space after source: %s", command_buf.buf);
2424 } else {
2425 endp = strchr(s, ' ');
2426 if (!endp)
2427 die("Missing space after source: %s", command_buf.buf);
2428 strbuf_add(&s_uq, s, endp - s);
2430 s = s_uq.buf;
2432 endp++;
2433 if (!*endp)
2434 die("Missing dest: %s", command_buf.buf);
2436 d = endp;
2437 strbuf_reset(&d_uq);
2438 if (!unquote_c_style(&d_uq, d, &endp)) {
2439 if (*endp)
2440 die("Garbage after dest in: %s", command_buf.buf);
2441 d = d_uq.buf;
2444 memset(&leaf, 0, sizeof(leaf));
2445 if (rename)
2446 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2447 else
2448 tree_content_get(&b->branch_tree, s, &leaf, 1);
2449 if (!leaf.versions[1].mode)
2450 die("Path %s not in branch", s);
2451 if (!*d) { /* C "path/to/subdir" "" */
2452 tree_content_replace(&b->branch_tree,
2453 leaf.versions[1].sha1,
2454 leaf.versions[1].mode,
2455 leaf.tree);
2456 return;
2458 tree_content_set(&b->branch_tree, d,
2459 leaf.versions[1].sha1,
2460 leaf.versions[1].mode,
2461 leaf.tree);
2464 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2466 static struct strbuf uq = STRBUF_INIT;
2467 struct object_entry *oe;
2468 struct branch *s;
2469 unsigned char sha1[20], commit_sha1[20];
2470 char path[60];
2471 uint16_t inline_data = 0;
2472 unsigned char new_fanout;
2475 * When loading a branch, we don't traverse its tree to count the real
2476 * number of notes (too expensive to do this for all non-note refs).
2477 * This means that recently loaded notes refs might incorrectly have
2478 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2480 * Fix this by traversing the tree and counting the number of notes
2481 * when b->num_notes == 0. If the notes tree is truly empty, the
2482 * calculation should not take long.
2484 if (b->num_notes == 0 && *old_fanout == 0) {
2485 /* Invoke change_note_fanout() in "counting mode". */
2486 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2487 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2490 /* Now parse the notemodify command. */
2491 /* <dataref> or 'inline' */
2492 if (*p == ':') {
2493 oe = find_mark(parse_mark_ref_space(&p));
2494 hashcpy(sha1, oe->idx.sha1);
2495 } else if (skip_prefix(p, "inline ", &p)) {
2496 inline_data = 1;
2497 oe = NULL; /* not used with inline_data, but makes gcc happy */
2498 } else {
2499 if (get_sha1_hex(p, sha1))
2500 die("Invalid dataref: %s", command_buf.buf);
2501 oe = find_object(sha1);
2502 p += 40;
2503 if (*p++ != ' ')
2504 die("Missing space after SHA1: %s", command_buf.buf);
2507 /* <commit-ish> */
2508 s = lookup_branch(p);
2509 if (s) {
2510 if (is_null_sha1(s->sha1))
2511 die("Can't add a note on empty branch.");
2512 hashcpy(commit_sha1, s->sha1);
2513 } else if (*p == ':') {
2514 uintmax_t commit_mark = parse_mark_ref_eol(p);
2515 struct object_entry *commit_oe = find_mark(commit_mark);
2516 if (commit_oe->type != OBJ_COMMIT)
2517 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2518 hashcpy(commit_sha1, commit_oe->idx.sha1);
2519 } else if (!get_sha1(p, commit_sha1)) {
2520 unsigned long size;
2521 char *buf = read_object_with_reference(commit_sha1,
2522 commit_type, &size, commit_sha1);
2523 if (!buf || size < 46)
2524 die("Not a valid commit: %s", p);
2525 free(buf);
2526 } else
2527 die("Invalid ref name or SHA1 expression: %s", p);
2529 if (inline_data) {
2530 if (p != uq.buf) {
2531 strbuf_addstr(&uq, p);
2532 p = uq.buf;
2534 read_next_command();
2535 parse_and_store_blob(&last_blob, sha1, 0);
2536 } else if (oe) {
2537 if (oe->type != OBJ_BLOB)
2538 die("Not a blob (actually a %s): %s",
2539 typename(oe->type), command_buf.buf);
2540 } else if (!is_null_sha1(sha1)) {
2541 enum object_type type = sha1_object_info(sha1, NULL);
2542 if (type < 0)
2543 die("Blob not found: %s", command_buf.buf);
2544 if (type != OBJ_BLOB)
2545 die("Not a blob (actually a %s): %s",
2546 typename(type), command_buf.buf);
2549 construct_path_with_fanout(sha1_to_hex(commit_sha1), *old_fanout, path);
2550 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2551 b->num_notes--;
2553 if (is_null_sha1(sha1))
2554 return; /* nothing to insert */
2556 b->num_notes++;
2557 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2558 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2559 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2562 static void file_change_deleteall(struct branch *b)
2564 release_tree_content_recursive(b->branch_tree.tree);
2565 hashclr(b->branch_tree.versions[0].sha1);
2566 hashclr(b->branch_tree.versions[1].sha1);
2567 load_tree(&b->branch_tree);
2568 b->num_notes = 0;
2571 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2573 if (!buf || size < 46)
2574 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2575 if (memcmp("tree ", buf, 5)
2576 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2577 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2578 hashcpy(b->branch_tree.versions[0].sha1,
2579 b->branch_tree.versions[1].sha1);
2582 static void parse_from_existing(struct branch *b)
2584 if (is_null_sha1(b->sha1)) {
2585 hashclr(b->branch_tree.versions[0].sha1);
2586 hashclr(b->branch_tree.versions[1].sha1);
2587 } else {
2588 unsigned long size;
2589 char *buf;
2591 buf = read_object_with_reference(b->sha1,
2592 commit_type, &size, b->sha1);
2593 parse_from_commit(b, buf, size);
2594 free(buf);
2598 static int parse_from(struct branch *b)
2600 const char *from;
2601 struct branch *s;
2602 unsigned char sha1[20];
2604 if (!skip_prefix(command_buf.buf, "from ", &from))
2605 return 0;
2607 hashcpy(sha1, b->branch_tree.versions[1].sha1);
2609 s = lookup_branch(from);
2610 if (b == s)
2611 die("Can't create a branch from itself: %s", b->name);
2612 else if (s) {
2613 unsigned char *t = s->branch_tree.versions[1].sha1;
2614 hashcpy(b->sha1, s->sha1);
2615 hashcpy(b->branch_tree.versions[0].sha1, t);
2616 hashcpy(b->branch_tree.versions[1].sha1, t);
2617 } else if (*from == ':') {
2618 uintmax_t idnum = parse_mark_ref_eol(from);
2619 struct object_entry *oe = find_mark(idnum);
2620 if (oe->type != OBJ_COMMIT)
2621 die("Mark :%" PRIuMAX " not a commit", idnum);
2622 if (hashcmp(b->sha1, oe->idx.sha1)) {
2623 hashcpy(b->sha1, oe->idx.sha1);
2624 if (oe->pack_id != MAX_PACK_ID) {
2625 unsigned long size;
2626 char *buf = gfi_unpack_entry(oe, &size);
2627 parse_from_commit(b, buf, size);
2628 free(buf);
2629 } else
2630 parse_from_existing(b);
2632 } else if (!get_sha1(from, b->sha1)) {
2633 parse_from_existing(b);
2634 if (is_null_sha1(b->sha1))
2635 b->delete = 1;
2637 else
2638 die("Invalid ref name or SHA1 expression: %s", from);
2640 if (b->branch_tree.tree && hashcmp(sha1, b->branch_tree.versions[1].sha1)) {
2641 release_tree_content_recursive(b->branch_tree.tree);
2642 b->branch_tree.tree = NULL;
2645 read_next_command();
2646 return 1;
2649 static struct hash_list *parse_merge(unsigned int *count)
2651 struct hash_list *list = NULL, **tail = &list, *n;
2652 const char *from;
2653 struct branch *s;
2655 *count = 0;
2656 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2657 n = xmalloc(sizeof(*n));
2658 s = lookup_branch(from);
2659 if (s)
2660 hashcpy(n->sha1, s->sha1);
2661 else if (*from == ':') {
2662 uintmax_t idnum = parse_mark_ref_eol(from);
2663 struct object_entry *oe = find_mark(idnum);
2664 if (oe->type != OBJ_COMMIT)
2665 die("Mark :%" PRIuMAX " not a commit", idnum);
2666 hashcpy(n->sha1, oe->idx.sha1);
2667 } else if (!get_sha1(from, n->sha1)) {
2668 unsigned long size;
2669 char *buf = read_object_with_reference(n->sha1,
2670 commit_type, &size, n->sha1);
2671 if (!buf || size < 46)
2672 die("Not a valid commit: %s", from);
2673 free(buf);
2674 } else
2675 die("Invalid ref name or SHA1 expression: %s", from);
2677 n->next = NULL;
2678 *tail = n;
2679 tail = &n->next;
2681 (*count)++;
2682 read_next_command();
2684 return list;
2687 static void parse_new_commit(const char *arg)
2689 static struct strbuf msg = STRBUF_INIT;
2690 struct branch *b;
2691 char *author = NULL;
2692 char *committer = NULL;
2693 struct hash_list *merge_list = NULL;
2694 unsigned int merge_count;
2695 unsigned char prev_fanout, new_fanout;
2696 const char *v;
2698 b = lookup_branch(arg);
2699 if (!b)
2700 b = new_branch(arg);
2702 read_next_command();
2703 parse_mark();
2704 if (skip_prefix(command_buf.buf, "author ", &v)) {
2705 author = parse_ident(v);
2706 read_next_command();
2708 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2709 committer = parse_ident(v);
2710 read_next_command();
2712 if (!committer)
2713 die("Expected committer but didn't get one");
2714 parse_data(&msg, 0, NULL);
2715 read_next_command();
2716 parse_from(b);
2717 merge_list = parse_merge(&merge_count);
2719 /* ensure the branch is active/loaded */
2720 if (!b->branch_tree.tree || !max_active_branches) {
2721 unload_one_branch();
2722 load_branch(b);
2725 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2727 /* file_change* */
2728 while (command_buf.len > 0) {
2729 if (skip_prefix(command_buf.buf, "M ", &v))
2730 file_change_m(v, b);
2731 else if (skip_prefix(command_buf.buf, "D ", &v))
2732 file_change_d(v, b);
2733 else if (skip_prefix(command_buf.buf, "R ", &v))
2734 file_change_cr(v, b, 1);
2735 else if (skip_prefix(command_buf.buf, "C ", &v))
2736 file_change_cr(v, b, 0);
2737 else if (skip_prefix(command_buf.buf, "N ", &v))
2738 note_change_n(v, b, &prev_fanout);
2739 else if (!strcmp("deleteall", command_buf.buf))
2740 file_change_deleteall(b);
2741 else if (skip_prefix(command_buf.buf, "ls ", &v))
2742 parse_ls(v, b);
2743 else {
2744 unread_command_buf = 1;
2745 break;
2747 if (read_next_command() == EOF)
2748 break;
2751 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2752 if (new_fanout != prev_fanout)
2753 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2755 /* build the tree and the commit */
2756 store_tree(&b->branch_tree);
2757 hashcpy(b->branch_tree.versions[0].sha1,
2758 b->branch_tree.versions[1].sha1);
2760 strbuf_reset(&new_data);
2761 strbuf_addf(&new_data, "tree %s\n",
2762 sha1_to_hex(b->branch_tree.versions[1].sha1));
2763 if (!is_null_sha1(b->sha1))
2764 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2765 while (merge_list) {
2766 struct hash_list *next = merge_list->next;
2767 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2768 free(merge_list);
2769 merge_list = next;
2771 strbuf_addf(&new_data,
2772 "author %s\n"
2773 "committer %s\n"
2774 "\n",
2775 author ? author : committer, committer);
2776 strbuf_addbuf(&new_data, &msg);
2777 free(author);
2778 free(committer);
2780 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2781 b->pack_id = pack_id;
2782 b->last_commit = object_count_by_type[OBJ_COMMIT];
2785 static void parse_new_tag(const char *arg)
2787 static struct strbuf msg = STRBUF_INIT;
2788 const char *from;
2789 char *tagger;
2790 struct branch *s;
2791 struct tag *t;
2792 uintmax_t from_mark = 0;
2793 unsigned char sha1[20];
2794 enum object_type type;
2795 const char *v;
2797 t = pool_alloc(sizeof(struct tag));
2798 memset(t, 0, sizeof(struct tag));
2799 t->name = pool_strdup(arg);
2800 if (last_tag)
2801 last_tag->next_tag = t;
2802 else
2803 first_tag = t;
2804 last_tag = t;
2805 read_next_command();
2807 /* from ... */
2808 if (!skip_prefix(command_buf.buf, "from ", &from))
2809 die("Expected from command, got %s", command_buf.buf);
2810 s = lookup_branch(from);
2811 if (s) {
2812 if (is_null_sha1(s->sha1))
2813 die("Can't tag an empty branch.");
2814 hashcpy(sha1, s->sha1);
2815 type = OBJ_COMMIT;
2816 } else if (*from == ':') {
2817 struct object_entry *oe;
2818 from_mark = parse_mark_ref_eol(from);
2819 oe = find_mark(from_mark);
2820 type = oe->type;
2821 hashcpy(sha1, oe->idx.sha1);
2822 } else if (!get_sha1(from, sha1)) {
2823 struct object_entry *oe = find_object(sha1);
2824 if (!oe) {
2825 type = sha1_object_info(sha1, NULL);
2826 if (type < 0)
2827 die("Not a valid object: %s", from);
2828 } else
2829 type = oe->type;
2830 } else
2831 die("Invalid ref name or SHA1 expression: %s", from);
2832 read_next_command();
2834 /* tagger ... */
2835 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2836 tagger = parse_ident(v);
2837 read_next_command();
2838 } else
2839 tagger = NULL;
2841 /* tag payload/message */
2842 parse_data(&msg, 0, NULL);
2844 /* build the tag object */
2845 strbuf_reset(&new_data);
2847 strbuf_addf(&new_data,
2848 "object %s\n"
2849 "type %s\n"
2850 "tag %s\n",
2851 sha1_to_hex(sha1), typename(type), t->name);
2852 if (tagger)
2853 strbuf_addf(&new_data,
2854 "tagger %s\n", tagger);
2855 strbuf_addch(&new_data, '\n');
2856 strbuf_addbuf(&new_data, &msg);
2857 free(tagger);
2859 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2860 t->pack_id = MAX_PACK_ID;
2861 else
2862 t->pack_id = pack_id;
2865 static void parse_reset_branch(const char *arg)
2867 struct branch *b;
2869 b = lookup_branch(arg);
2870 if (b) {
2871 hashclr(b->sha1);
2872 hashclr(b->branch_tree.versions[0].sha1);
2873 hashclr(b->branch_tree.versions[1].sha1);
2874 if (b->branch_tree.tree) {
2875 release_tree_content_recursive(b->branch_tree.tree);
2876 b->branch_tree.tree = NULL;
2879 else
2880 b = new_branch(arg);
2881 read_next_command();
2882 parse_from(b);
2883 if (command_buf.len > 0)
2884 unread_command_buf = 1;
2887 static void cat_blob_write(const char *buf, unsigned long size)
2889 if (write_in_full(cat_blob_fd, buf, size) != size)
2890 die_errno("Write to frontend failed");
2893 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2895 struct strbuf line = STRBUF_INIT;
2896 unsigned long size;
2897 enum object_type type = 0;
2898 char *buf;
2900 if (!oe || oe->pack_id == MAX_PACK_ID) {
2901 buf = read_sha1_file(sha1, &type, &size);
2902 } else {
2903 type = oe->type;
2904 buf = gfi_unpack_entry(oe, &size);
2908 * Output based on batch_one_object() from cat-file.c.
2910 if (type <= 0) {
2911 strbuf_reset(&line);
2912 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2913 cat_blob_write(line.buf, line.len);
2914 strbuf_release(&line);
2915 free(buf);
2916 return;
2918 if (!buf)
2919 die("Can't read object %s", sha1_to_hex(sha1));
2920 if (type != OBJ_BLOB)
2921 die("Object %s is a %s but a blob was expected.",
2922 sha1_to_hex(sha1), typename(type));
2923 strbuf_reset(&line);
2924 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2925 typename(type), size);
2926 cat_blob_write(line.buf, line.len);
2927 strbuf_release(&line);
2928 cat_blob_write(buf, size);
2929 cat_blob_write("\n", 1);
2930 if (oe && oe->pack_id == pack_id) {
2931 last_blob.offset = oe->idx.offset;
2932 strbuf_attach(&last_blob.data, buf, size, size);
2933 last_blob.depth = oe->depth;
2934 } else
2935 free(buf);
2938 static void parse_get_mark(const char *p)
2940 struct object_entry *oe = oe;
2941 char output[42];
2943 /* get-mark SP <object> LF */
2944 if (*p != ':')
2945 die("Not a mark: %s", p);
2947 oe = find_mark(parse_mark_ref_eol(p));
2948 if (!oe)
2949 die("Unknown mark: %s", command_buf.buf);
2951 snprintf(output, sizeof(output), "%s\n", sha1_to_hex(oe->idx.sha1));
2952 cat_blob_write(output, 41);
2955 static void parse_cat_blob(const char *p)
2957 struct object_entry *oe = oe;
2958 unsigned char sha1[20];
2960 /* cat-blob SP <object> LF */
2961 if (*p == ':') {
2962 oe = find_mark(parse_mark_ref_eol(p));
2963 if (!oe)
2964 die("Unknown mark: %s", command_buf.buf);
2965 hashcpy(sha1, oe->idx.sha1);
2966 } else {
2967 if (get_sha1_hex(p, sha1))
2968 die("Invalid dataref: %s", command_buf.buf);
2969 if (p[40])
2970 die("Garbage after SHA1: %s", command_buf.buf);
2971 oe = find_object(sha1);
2974 cat_blob(oe, sha1);
2977 static struct object_entry *dereference(struct object_entry *oe,
2978 unsigned char sha1[20])
2980 unsigned long size;
2981 char *buf = NULL;
2982 if (!oe) {
2983 enum object_type type = sha1_object_info(sha1, NULL);
2984 if (type < 0)
2985 die("object not found: %s", sha1_to_hex(sha1));
2986 /* cache it! */
2987 oe = insert_object(sha1);
2988 oe->type = type;
2989 oe->pack_id = MAX_PACK_ID;
2990 oe->idx.offset = 1;
2992 switch (oe->type) {
2993 case OBJ_TREE: /* easy case. */
2994 return oe;
2995 case OBJ_COMMIT:
2996 case OBJ_TAG:
2997 break;
2998 default:
2999 die("Not a tree-ish: %s", command_buf.buf);
3002 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
3003 buf = gfi_unpack_entry(oe, &size);
3004 } else {
3005 enum object_type unused;
3006 buf = read_sha1_file(sha1, &unused, &size);
3008 if (!buf)
3009 die("Can't load object %s", sha1_to_hex(sha1));
3011 /* Peel one layer. */
3012 switch (oe->type) {
3013 case OBJ_TAG:
3014 if (size < 40 + strlen("object ") ||
3015 get_sha1_hex(buf + strlen("object "), sha1))
3016 die("Invalid SHA1 in tag: %s", command_buf.buf);
3017 break;
3018 case OBJ_COMMIT:
3019 if (size < 40 + strlen("tree ") ||
3020 get_sha1_hex(buf + strlen("tree "), sha1))
3021 die("Invalid SHA1 in commit: %s", command_buf.buf);
3024 free(buf);
3025 return find_object(sha1);
3028 static struct object_entry *parse_treeish_dataref(const char **p)
3030 unsigned char sha1[20];
3031 struct object_entry *e;
3033 if (**p == ':') { /* <mark> */
3034 e = find_mark(parse_mark_ref_space(p));
3035 if (!e)
3036 die("Unknown mark: %s", command_buf.buf);
3037 hashcpy(sha1, e->idx.sha1);
3038 } else { /* <sha1> */
3039 if (get_sha1_hex(*p, sha1))
3040 die("Invalid dataref: %s", command_buf.buf);
3041 e = find_object(sha1);
3042 *p += 40;
3043 if (*(*p)++ != ' ')
3044 die("Missing space after tree-ish: %s", command_buf.buf);
3047 while (!e || e->type != OBJ_TREE)
3048 e = dereference(e, sha1);
3049 return e;
3052 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3054 static struct strbuf line = STRBUF_INIT;
3056 /* See show_tree(). */
3057 const char *type =
3058 S_ISGITLINK(mode) ? commit_type :
3059 S_ISDIR(mode) ? tree_type :
3060 blob_type;
3062 if (!mode) {
3063 /* missing SP path LF */
3064 strbuf_reset(&line);
3065 strbuf_addstr(&line, "missing ");
3066 quote_c_style(path, &line, NULL, 0);
3067 strbuf_addch(&line, '\n');
3068 } else {
3069 /* mode SP type SP object_name TAB path LF */
3070 strbuf_reset(&line);
3071 strbuf_addf(&line, "%06o %s %s\t",
3072 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3073 quote_c_style(path, &line, NULL, 0);
3074 strbuf_addch(&line, '\n');
3076 cat_blob_write(line.buf, line.len);
3079 static void parse_ls(const char *p, struct branch *b)
3081 struct tree_entry *root = NULL;
3082 struct tree_entry leaf = {NULL};
3084 /* ls SP (<tree-ish> SP)? <path> */
3085 if (*p == '"') {
3086 if (!b)
3087 die("Not in a commit: %s", command_buf.buf);
3088 root = &b->branch_tree;
3089 } else {
3090 struct object_entry *e = parse_treeish_dataref(&p);
3091 root = new_tree_entry();
3092 hashcpy(root->versions[1].sha1, e->idx.sha1);
3093 if (!is_null_sha1(root->versions[1].sha1))
3094 root->versions[1].mode = S_IFDIR;
3095 load_tree(root);
3097 if (*p == '"') {
3098 static struct strbuf uq = STRBUF_INIT;
3099 const char *endp;
3100 strbuf_reset(&uq);
3101 if (unquote_c_style(&uq, p, &endp))
3102 die("Invalid path: %s", command_buf.buf);
3103 if (*endp)
3104 die("Garbage after path in: %s", command_buf.buf);
3105 p = uq.buf;
3107 tree_content_get(root, p, &leaf, 1);
3109 * A directory in preparation would have a sha1 of zero
3110 * until it is saved. Save, for simplicity.
3112 if (S_ISDIR(leaf.versions[1].mode))
3113 store_tree(&leaf);
3115 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3116 if (leaf.tree)
3117 release_tree_content_recursive(leaf.tree);
3118 if (!b || root != &b->branch_tree)
3119 release_tree_entry(root);
3122 static void checkpoint(void)
3124 checkpoint_requested = 0;
3125 if (object_count) {
3126 cycle_packfile();
3127 dump_branches();
3128 dump_tags();
3129 dump_marks();
3133 static void parse_checkpoint(void)
3135 checkpoint_requested = 1;
3136 skip_optional_lf();
3139 static void parse_progress(void)
3141 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3142 fputc('\n', stdout);
3143 fflush(stdout);
3144 skip_optional_lf();
3147 static char* make_fast_import_path(const char *path)
3149 if (!relative_marks_paths || is_absolute_path(path))
3150 return xstrdup(path);
3151 return xstrdup(git_path("info/fast-import/%s", path));
3154 static void option_import_marks(const char *marks,
3155 int from_stream, int ignore_missing)
3157 if (import_marks_file) {
3158 if (from_stream)
3159 die("Only one import-marks command allowed per stream");
3161 /* read previous mark file */
3162 if(!import_marks_file_from_stream)
3163 read_marks();
3166 import_marks_file = make_fast_import_path(marks);
3167 safe_create_leading_directories_const(import_marks_file);
3168 import_marks_file_from_stream = from_stream;
3169 import_marks_file_ignore_missing = ignore_missing;
3172 static void option_date_format(const char *fmt)
3174 if (!strcmp(fmt, "raw"))
3175 whenspec = WHENSPEC_RAW;
3176 else if (!strcmp(fmt, "rfc2822"))
3177 whenspec = WHENSPEC_RFC2822;
3178 else if (!strcmp(fmt, "now"))
3179 whenspec = WHENSPEC_NOW;
3180 else
3181 die("unknown --date-format argument %s", fmt);
3184 static unsigned long ulong_arg(const char *option, const char *arg)
3186 char *endptr;
3187 unsigned long rv = strtoul(arg, &endptr, 0);
3188 if (strchr(arg, '-') || endptr == arg || *endptr)
3189 die("%s: argument must be a non-negative integer", option);
3190 return rv;
3193 static void option_depth(const char *depth)
3195 max_depth = ulong_arg("--depth", depth);
3196 if (max_depth > MAX_DEPTH)
3197 die("--depth cannot exceed %u", MAX_DEPTH);
3200 static void option_active_branches(const char *branches)
3202 max_active_branches = ulong_arg("--active-branches", branches);
3205 static void option_export_marks(const char *marks)
3207 export_marks_file = make_fast_import_path(marks);
3208 safe_create_leading_directories_const(export_marks_file);
3211 static void option_cat_blob_fd(const char *fd)
3213 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3214 if (n > (unsigned long) INT_MAX)
3215 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3216 cat_blob_fd = (int) n;
3219 static void option_export_pack_edges(const char *edges)
3221 if (pack_edges)
3222 fclose(pack_edges);
3223 pack_edges = fopen(edges, "a");
3224 if (!pack_edges)
3225 die_errno("Cannot open '%s'", edges);
3228 static int parse_one_option(const char *option)
3230 if (skip_prefix(option, "max-pack-size=", &option)) {
3231 unsigned long v;
3232 if (!git_parse_ulong(option, &v))
3233 return 0;
3234 if (v < 8192) {
3235 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3236 v *= 1024 * 1024;
3237 } else if (v < 1024 * 1024) {
3238 warning("minimum max-pack-size is 1 MiB");
3239 v = 1024 * 1024;
3241 max_packsize = v;
3242 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3243 unsigned long v;
3244 if (!git_parse_ulong(option, &v))
3245 return 0;
3246 big_file_threshold = v;
3247 } else if (skip_prefix(option, "depth=", &option)) {
3248 option_depth(option);
3249 } else if (skip_prefix(option, "active-branches=", &option)) {
3250 option_active_branches(option);
3251 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3252 option_export_pack_edges(option);
3253 } else if (starts_with(option, "quiet")) {
3254 show_stats = 0;
3255 } else if (starts_with(option, "stats")) {
3256 show_stats = 1;
3257 } else {
3258 return 0;
3261 return 1;
3264 static int parse_one_feature(const char *feature, int from_stream)
3266 const char *arg;
3268 if (skip_prefix(feature, "date-format=", &arg)) {
3269 option_date_format(arg);
3270 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3271 option_import_marks(arg, from_stream, 0);
3272 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3273 option_import_marks(arg, from_stream, 1);
3274 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3275 option_export_marks(arg);
3276 } else if (!strcmp(feature, "get-mark")) {
3277 ; /* Don't die - this feature is supported */
3278 } else if (!strcmp(feature, "cat-blob")) {
3279 ; /* Don't die - this feature is supported */
3280 } else if (!strcmp(feature, "relative-marks")) {
3281 relative_marks_paths = 1;
3282 } else if (!strcmp(feature, "no-relative-marks")) {
3283 relative_marks_paths = 0;
3284 } else if (!strcmp(feature, "done")) {
3285 require_explicit_termination = 1;
3286 } else if (!strcmp(feature, "force")) {
3287 force_update = 1;
3288 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3289 ; /* do nothing; we have the feature */
3290 } else {
3291 return 0;
3294 return 1;
3297 static void parse_feature(const char *feature)
3299 if (seen_data_command)
3300 die("Got feature command '%s' after data command", feature);
3302 if (parse_one_feature(feature, 1))
3303 return;
3305 die("This version of fast-import does not support feature %s.", feature);
3308 static void parse_option(const char *option)
3310 if (seen_data_command)
3311 die("Got option command '%s' after data command", option);
3313 if (parse_one_option(option))
3314 return;
3316 die("This version of fast-import does not support option: %s", option);
3319 static void git_pack_config(void)
3321 int indexversion_value;
3322 unsigned long packsizelimit_value;
3324 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3325 if (max_depth > MAX_DEPTH)
3326 max_depth = MAX_DEPTH;
3328 if (!git_config_get_int("pack.compression", &pack_compression_level)) {
3329 if (pack_compression_level == -1)
3330 pack_compression_level = Z_DEFAULT_COMPRESSION;
3331 else if (pack_compression_level < 0 ||
3332 pack_compression_level > Z_BEST_COMPRESSION)
3333 git_die_config("pack.compression",
3334 "bad pack compression level %d", pack_compression_level);
3335 pack_compression_seen = 1;
3337 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3338 pack_idx_opts.version = indexversion_value;
3339 if (pack_idx_opts.version > 2)
3340 git_die_config("pack.indexversion",
3341 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3343 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3344 max_packsize = packsizelimit_value;
3346 git_config(git_default_config, NULL);
3349 static const char fast_import_usage[] =
3350 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3352 static void parse_argv(void)
3354 unsigned int i;
3356 for (i = 1; i < global_argc; i++) {
3357 const char *a = global_argv[i];
3359 if (*a != '-' || !strcmp(a, "--"))
3360 break;
3362 if (!skip_prefix(a, "--", &a))
3363 die("unknown option %s", a);
3365 if (parse_one_option(a))
3366 continue;
3368 if (parse_one_feature(a, 0))
3369 continue;
3371 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3372 option_cat_blob_fd(a);
3373 continue;
3376 die("unknown option --%s", a);
3378 if (i != global_argc)
3379 usage(fast_import_usage);
3381 seen_data_command = 1;
3382 if (import_marks_file)
3383 read_marks();
3386 int main(int argc, char **argv)
3388 unsigned int i;
3390 git_extract_argv0_path(argv[0]);
3392 git_setup_gettext();
3394 if (argc == 2 && !strcmp(argv[1], "-h"))
3395 usage(fast_import_usage);
3397 setup_git_directory();
3398 reset_pack_idx_option(&pack_idx_opts);
3399 git_pack_config();
3400 if (!pack_compression_seen && core_compression_seen)
3401 pack_compression_level = core_compression_level;
3403 alloc_objects(object_entry_alloc);
3404 strbuf_init(&command_buf, 0);
3405 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3406 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3407 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3408 marks = pool_calloc(1, sizeof(struct mark_set));
3410 global_argc = argc;
3411 global_argv = argv;
3413 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3414 for (i = 0; i < (cmd_save - 1); i++)
3415 rc_free[i].next = &rc_free[i + 1];
3416 rc_free[cmd_save - 1].next = NULL;
3418 prepare_packed_git();
3419 start_packfile();
3420 set_die_routine(die_nicely);
3421 set_checkpoint_signal();
3422 while (read_next_command() != EOF) {
3423 const char *v;
3424 if (!strcmp("blob", command_buf.buf))
3425 parse_new_blob();
3426 else if (skip_prefix(command_buf.buf, "ls ", &v))
3427 parse_ls(v, NULL);
3428 else if (skip_prefix(command_buf.buf, "commit ", &v))
3429 parse_new_commit(v);
3430 else if (skip_prefix(command_buf.buf, "tag ", &v))
3431 parse_new_tag(v);
3432 else if (skip_prefix(command_buf.buf, "reset ", &v))
3433 parse_reset_branch(v);
3434 else if (!strcmp("checkpoint", command_buf.buf))
3435 parse_checkpoint();
3436 else if (!strcmp("done", command_buf.buf))
3437 break;
3438 else if (starts_with(command_buf.buf, "progress "))
3439 parse_progress();
3440 else if (skip_prefix(command_buf.buf, "feature ", &v))
3441 parse_feature(v);
3442 else if (skip_prefix(command_buf.buf, "option git ", &v))
3443 parse_option(v);
3444 else if (starts_with(command_buf.buf, "option "))
3445 /* ignore non-git options*/;
3446 else
3447 die("Unsupported command: %s", command_buf.buf);
3449 if (checkpoint_requested)
3450 checkpoint();
3453 /* argv hasn't been parsed yet, do so */
3454 if (!seen_data_command)
3455 parse_argv();
3457 if (require_explicit_termination && feof(stdin))
3458 die("stream ends early");
3460 end_packfile();
3462 dump_branches();
3463 dump_tags();
3464 unkeep_all_packs();
3465 dump_marks();
3467 if (pack_edges)
3468 fclose(pack_edges);
3470 if (show_stats) {
3471 uintmax_t total_count = 0, duplicate_count = 0;
3472 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3473 total_count += object_count_by_type[i];
3474 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3475 duplicate_count += duplicate_count_by_type[i];
3477 fprintf(stderr, "%s statistics:\n", argv[0]);
3478 fprintf(stderr, "---------------------------------------------------------------------\n");
3479 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3480 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3481 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]);
3482 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]);
3483 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]);
3484 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]);
3485 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3486 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3487 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3488 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3489 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3490 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3491 fprintf(stderr, "---------------------------------------------------------------------\n");
3492 pack_report();
3493 fprintf(stderr, "---------------------------------------------------------------------\n");
3494 fprintf(stderr, "\n");
3497 return failure ? 1 : 0;