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
13 ;
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 committish lf)?
26 ('merge' sp committish 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 committish lf;
47 note_inm ::= 'N' sp 'inline' sp committish lf
48 data;
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp committish 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 committish 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.
71 #
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.
85 #
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.
95 #
96 committish ::= (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'
104 ;
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.
120 #
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.
131 #
132 name ::= # valid GIT author/committer name;
133 email ::= # valid GIT author/committer email;
134 ts ::= # time since the epoch in seconds, ascii base10 notation;
135 tz ::= # GIT style timezone;
137 # note: comments, ls and cat requests may appear anywhere
138 # in the input, except within a data command. Any form
139 # of the data command always escapes the related input
140 # from comment processing.
141 #
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).
145 #
147 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
148 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
150 comment ::= '#' not_lf* lf;
151 not_lf ::= # Any byte that is not ASCII newline (LF);
152 */
154 #include "builtin.h"
155 #include "cache.h"
156 #include "object.h"
157 #include "blob.h"
158 #include "tree.h"
159 #include "commit.h"
160 #include "delta.h"
161 #include "pack.h"
162 #include "refs.h"
163 #include "csum-file.h"
164 #include "quote.h"
165 #include "exec_cmd.h"
166 #include "dir.h"
168 #define PACK_ID_BITS 16
169 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
170 #define DEPTH_BITS 13
171 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
173 /*
174 * We abuse the setuid bit on directories to mean "do not delta".
175 */
176 #define NO_DELTA S_ISUID
178 struct object_entry {
179 struct pack_idx_entry idx;
180 struct object_entry *next;
181 uint32_t type : TYPE_BITS,
182 pack_id : PACK_ID_BITS,
183 depth : DEPTH_BITS;
184 };
186 struct object_entry_pool {
187 struct object_entry_pool *next_pool;
188 struct object_entry *next_free;
189 struct object_entry *end;
190 struct object_entry entries[FLEX_ARRAY]; /* more */
191 };
193 struct mark_set {
194 union {
195 struct object_entry *marked[1024];
196 struct mark_set *sets[1024];
197 } data;
198 unsigned int shift;
199 };
201 struct last_object {
202 struct strbuf data;
203 off_t offset;
204 unsigned int depth;
205 unsigned no_swap : 1;
206 };
208 struct mem_pool {
209 struct mem_pool *next_pool;
210 char *next_free;
211 char *end;
212 uintmax_t space[FLEX_ARRAY]; /* more */
213 };
215 struct atom_str {
216 struct atom_str *next_atom;
217 unsigned short str_len;
218 char str_dat[FLEX_ARRAY]; /* more */
219 };
221 struct tree_content;
222 struct tree_entry {
223 struct tree_content *tree;
224 struct atom_str *name;
225 struct tree_entry_ms {
226 uint16_t mode;
227 unsigned char sha1[20];
228 } versions[2];
229 };
231 struct tree_content {
232 unsigned int entry_capacity; /* must match avail_tree_content */
233 unsigned int entry_count;
234 unsigned int delta_depth;
235 struct tree_entry *entries[FLEX_ARRAY]; /* more */
236 };
238 struct avail_tree_content {
239 unsigned int entry_capacity; /* must match tree_content */
240 struct avail_tree_content *next_avail;
241 };
243 struct branch {
244 struct branch *table_next_branch;
245 struct branch *active_next_branch;
246 const char *name;
247 struct tree_entry branch_tree;
248 uintmax_t last_commit;
249 uintmax_t num_notes;
250 unsigned active : 1;
251 unsigned pack_id : PACK_ID_BITS;
252 unsigned char sha1[20];
253 };
255 struct tag {
256 struct tag *next_tag;
257 const char *name;
258 unsigned int pack_id;
259 unsigned char sha1[20];
260 };
262 struct hash_list {
263 struct hash_list *next;
264 unsigned char sha1[20];
265 };
267 typedef enum {
268 WHENSPEC_RAW = 1,
269 WHENSPEC_RFC2822,
270 WHENSPEC_NOW
271 } whenspec_type;
273 struct recent_command {
274 struct recent_command *prev;
275 struct recent_command *next;
276 char *buf;
277 };
279 /* Configured limits on output */
280 static unsigned long max_depth = 10;
281 static off_t max_packsize;
282 static int force_update;
283 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
284 static int pack_compression_seen;
286 /* Stats and misc. counters */
287 static uintmax_t alloc_count;
288 static uintmax_t marks_set_count;
289 static uintmax_t object_count_by_type[1 << TYPE_BITS];
290 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
291 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
292 static unsigned long object_count;
293 static unsigned long branch_count;
294 static unsigned long branch_load_count;
295 static int failure;
296 static FILE *pack_edges;
297 static unsigned int show_stats = 1;
298 static int global_argc;
299 static const char **global_argv;
301 /* Memory pools */
302 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
303 static size_t total_allocd;
304 static struct mem_pool *mem_pool;
306 /* Atom management */
307 static unsigned int atom_table_sz = 4451;
308 static unsigned int atom_cnt;
309 static struct atom_str **atom_table;
311 /* The .pack file being generated */
312 static struct pack_idx_option pack_idx_opts;
313 static unsigned int pack_id;
314 static struct sha1file *pack_file;
315 static struct packed_git *pack_data;
316 static struct packed_git **all_packs;
317 static off_t pack_size;
319 /* Table of objects we've written. */
320 static unsigned int object_entry_alloc = 5000;
321 static struct object_entry_pool *blocks;
322 static struct object_entry *object_table[1 << 16];
323 static struct mark_set *marks;
324 static const char *export_marks_file;
325 static const char *import_marks_file;
326 static int import_marks_file_from_stream;
327 static int import_marks_file_ignore_missing;
328 static int relative_marks_paths;
330 /* Our last blob */
331 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
333 /* Tree management */
334 static unsigned int tree_entry_alloc = 1000;
335 static void *avail_tree_entry;
336 static unsigned int avail_tree_table_sz = 100;
337 static struct avail_tree_content **avail_tree_table;
338 static struct strbuf old_tree = STRBUF_INIT;
339 static struct strbuf new_tree = STRBUF_INIT;
341 /* Branch data */
342 static unsigned long max_active_branches = 5;
343 static unsigned long cur_active_branches;
344 static unsigned long branch_table_sz = 1039;
345 static struct branch **branch_table;
346 static struct branch *active_branches;
348 /* Tag data */
349 static struct tag *first_tag;
350 static struct tag *last_tag;
352 /* Input stream parsing */
353 static whenspec_type whenspec = WHENSPEC_RAW;
354 static struct strbuf command_buf = STRBUF_INIT;
355 static int unread_command_buf;
356 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
357 static struct recent_command *cmd_tail = &cmd_hist;
358 static struct recent_command *rc_free;
359 static unsigned int cmd_save = 100;
360 static uintmax_t next_mark;
361 static struct strbuf new_data = STRBUF_INIT;
362 static int seen_data_command;
363 static int require_explicit_termination;
365 /* Signal handling */
366 static volatile sig_atomic_t checkpoint_requested;
368 /* Where to write output of cat-blob commands */
369 static int cat_blob_fd = STDOUT_FILENO;
371 static void parse_argv(void);
372 static void parse_cat_blob(void);
373 static void parse_ls(struct branch *b);
375 static void write_branch_report(FILE *rpt, struct branch *b)
376 {
377 fprintf(rpt, "%s:\n", b->name);
379 fprintf(rpt, " status :");
380 if (b->active)
381 fputs(" active", rpt);
382 if (b->branch_tree.tree)
383 fputs(" loaded", rpt);
384 if (is_null_sha1(b->branch_tree.versions[1].sha1))
385 fputs(" dirty", rpt);
386 fputc('\n', rpt);
388 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
389 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
390 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
391 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
393 fputs(" last pack : ", rpt);
394 if (b->pack_id < MAX_PACK_ID)
395 fprintf(rpt, "%u", b->pack_id);
396 fputc('\n', rpt);
398 fputc('\n', rpt);
399 }
401 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
403 static void write_crash_report(const char *err)
404 {
405 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
406 FILE *rpt = fopen(loc, "w");
407 struct branch *b;
408 unsigned long lu;
409 struct recent_command *rc;
411 if (!rpt) {
412 error("can't write crash report %s: %s", loc, strerror(errno));
413 return;
414 }
416 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
418 fprintf(rpt, "fast-import crash report:\n");
419 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
420 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
421 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
422 fputc('\n', rpt);
424 fputs("fatal: ", rpt);
425 fputs(err, rpt);
426 fputc('\n', rpt);
428 fputc('\n', rpt);
429 fputs("Most Recent Commands Before Crash\n", rpt);
430 fputs("---------------------------------\n", rpt);
431 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
432 if (rc->next == &cmd_hist)
433 fputs("* ", rpt);
434 else
435 fputs(" ", rpt);
436 fputs(rc->buf, rpt);
437 fputc('\n', rpt);
438 }
440 fputc('\n', rpt);
441 fputs("Active Branch LRU\n", rpt);
442 fputs("-----------------\n", rpt);
443 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
444 cur_active_branches,
445 max_active_branches);
446 fputc('\n', rpt);
447 fputs(" pos clock name\n", rpt);
448 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
449 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
450 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
451 ++lu, b->last_commit, b->name);
453 fputc('\n', rpt);
454 fputs("Inactive Branches\n", rpt);
455 fputs("-----------------\n", rpt);
456 for (lu = 0; lu < branch_table_sz; lu++) {
457 for (b = branch_table[lu]; b; b = b->table_next_branch)
458 write_branch_report(rpt, b);
459 }
461 if (first_tag) {
462 struct tag *tg;
463 fputc('\n', rpt);
464 fputs("Annotated Tags\n", rpt);
465 fputs("--------------\n", rpt);
466 for (tg = first_tag; tg; tg = tg->next_tag) {
467 fputs(sha1_to_hex(tg->sha1), rpt);
468 fputc(' ', rpt);
469 fputs(tg->name, rpt);
470 fputc('\n', rpt);
471 }
472 }
474 fputc('\n', rpt);
475 fputs("Marks\n", rpt);
476 fputs("-----\n", rpt);
477 if (export_marks_file)
478 fprintf(rpt, " exported to %s\n", export_marks_file);
479 else
480 dump_marks_helper(rpt, 0, marks);
482 fputc('\n', rpt);
483 fputs("-------------------\n", rpt);
484 fputs("END OF CRASH REPORT\n", rpt);
485 fclose(rpt);
486 }
488 static void end_packfile(void);
489 static void unkeep_all_packs(void);
490 static void dump_marks(void);
492 static NORETURN void die_nicely(const char *err, va_list params)
493 {
494 static int zombie;
495 char message[2 * PATH_MAX];
497 vsnprintf(message, sizeof(message), err, params);
498 fputs("fatal: ", stderr);
499 fputs(message, stderr);
500 fputc('\n', stderr);
502 if (!zombie) {
503 zombie = 1;
504 write_crash_report(message);
505 end_packfile();
506 unkeep_all_packs();
507 dump_marks();
508 }
509 exit(128);
510 }
512 #ifndef SIGUSR1 /* Windows, for example */
514 static void set_checkpoint_signal(void)
515 {
516 }
518 #else
520 static void checkpoint_signal(int signo)
521 {
522 checkpoint_requested = 1;
523 }
525 static void set_checkpoint_signal(void)
526 {
527 struct sigaction sa;
529 memset(&sa, 0, sizeof(sa));
530 sa.sa_handler = checkpoint_signal;
531 sigemptyset(&sa.sa_mask);
532 sa.sa_flags = SA_RESTART;
533 sigaction(SIGUSR1, &sa, NULL);
534 }
536 #endif
538 static void alloc_objects(unsigned int cnt)
539 {
540 struct object_entry_pool *b;
542 b = xmalloc(sizeof(struct object_entry_pool)
543 + cnt * sizeof(struct object_entry));
544 b->next_pool = blocks;
545 b->next_free = b->entries;
546 b->end = b->entries + cnt;
547 blocks = b;
548 alloc_count += cnt;
549 }
551 static struct object_entry *new_object(unsigned char *sha1)
552 {
553 struct object_entry *e;
555 if (blocks->next_free == blocks->end)
556 alloc_objects(object_entry_alloc);
558 e = blocks->next_free++;
559 hashcpy(e->idx.sha1, sha1);
560 return e;
561 }
563 static struct object_entry *find_object(unsigned char *sha1)
564 {
565 unsigned int h = sha1[0] << 8 | sha1[1];
566 struct object_entry *e;
567 for (e = object_table[h]; e; e = e->next)
568 if (!hashcmp(sha1, e->idx.sha1))
569 return e;
570 return NULL;
571 }
573 static struct object_entry *insert_object(unsigned char *sha1)
574 {
575 unsigned int h = sha1[0] << 8 | sha1[1];
576 struct object_entry *e = object_table[h];
578 while (e) {
579 if (!hashcmp(sha1, e->idx.sha1))
580 return e;
581 e = e->next;
582 }
584 e = new_object(sha1);
585 e->next = object_table[h];
586 e->idx.offset = 0;
587 object_table[h] = e;
588 return e;
589 }
591 static unsigned int hc_str(const char *s, size_t len)
592 {
593 unsigned int r = 0;
594 while (len-- > 0)
595 r = r * 31 + *s++;
596 return r;
597 }
599 static void *pool_alloc(size_t len)
600 {
601 struct mem_pool *p;
602 void *r;
604 /* round up to a 'uintmax_t' alignment */
605 if (len & (sizeof(uintmax_t) - 1))
606 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
608 for (p = mem_pool; p; p = p->next_pool)
609 if ((p->end - p->next_free >= len))
610 break;
612 if (!p) {
613 if (len >= (mem_pool_alloc/2)) {
614 total_allocd += len;
615 return xmalloc(len);
616 }
617 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
618 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
619 p->next_pool = mem_pool;
620 p->next_free = (char *) p->space;
621 p->end = p->next_free + mem_pool_alloc;
622 mem_pool = p;
623 }
625 r = p->next_free;
626 p->next_free += len;
627 return r;
628 }
630 static void *pool_calloc(size_t count, size_t size)
631 {
632 size_t len = count * size;
633 void *r = pool_alloc(len);
634 memset(r, 0, len);
635 return r;
636 }
638 static char *pool_strdup(const char *s)
639 {
640 char *r = pool_alloc(strlen(s) + 1);
641 strcpy(r, s);
642 return r;
643 }
645 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
646 {
647 struct mark_set *s = marks;
648 while ((idnum >> s->shift) >= 1024) {
649 s = pool_calloc(1, sizeof(struct mark_set));
650 s->shift = marks->shift + 10;
651 s->data.sets[0] = marks;
652 marks = s;
653 }
654 while (s->shift) {
655 uintmax_t i = idnum >> s->shift;
656 idnum -= i << s->shift;
657 if (!s->data.sets[i]) {
658 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
659 s->data.sets[i]->shift = s->shift - 10;
660 }
661 s = s->data.sets[i];
662 }
663 if (!s->data.marked[idnum])
664 marks_set_count++;
665 s->data.marked[idnum] = oe;
666 }
668 static struct object_entry *find_mark(uintmax_t idnum)
669 {
670 uintmax_t orig_idnum = idnum;
671 struct mark_set *s = marks;
672 struct object_entry *oe = NULL;
673 if ((idnum >> s->shift) < 1024) {
674 while (s && s->shift) {
675 uintmax_t i = idnum >> s->shift;
676 idnum -= i << s->shift;
677 s = s->data.sets[i];
678 }
679 if (s)
680 oe = s->data.marked[idnum];
681 }
682 if (!oe)
683 die("mark :%" PRIuMAX " not declared", orig_idnum);
684 return oe;
685 }
687 static struct atom_str *to_atom(const char *s, unsigned short len)
688 {
689 unsigned int hc = hc_str(s, len) % atom_table_sz;
690 struct atom_str *c;
692 for (c = atom_table[hc]; c; c = c->next_atom)
693 if (c->str_len == len && !strncmp(s, c->str_dat, len))
694 return c;
696 c = pool_alloc(sizeof(struct atom_str) + len + 1);
697 c->str_len = len;
698 strncpy(c->str_dat, s, len);
699 c->str_dat[len] = 0;
700 c->next_atom = atom_table[hc];
701 atom_table[hc] = c;
702 atom_cnt++;
703 return c;
704 }
706 static struct branch *lookup_branch(const char *name)
707 {
708 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
709 struct branch *b;
711 for (b = branch_table[hc]; b; b = b->table_next_branch)
712 if (!strcmp(name, b->name))
713 return b;
714 return NULL;
715 }
717 static struct branch *new_branch(const char *name)
718 {
719 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
720 struct branch *b = lookup_branch(name);
722 if (b)
723 die("Invalid attempt to create duplicate branch: %s", name);
724 switch (check_ref_format(name)) {
725 case 0: break; /* its valid */
726 case CHECK_REF_FORMAT_ONELEVEL:
727 break; /* valid, but too few '/', allow anyway */
728 default:
729 die("Branch name doesn't conform to GIT standards: %s", name);
730 }
732 b = pool_calloc(1, sizeof(struct branch));
733 b->name = pool_strdup(name);
734 b->table_next_branch = branch_table[hc];
735 b->branch_tree.versions[0].mode = S_IFDIR;
736 b->branch_tree.versions[1].mode = S_IFDIR;
737 b->num_notes = 0;
738 b->active = 0;
739 b->pack_id = MAX_PACK_ID;
740 branch_table[hc] = b;
741 branch_count++;
742 return b;
743 }
745 static unsigned int hc_entries(unsigned int cnt)
746 {
747 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
748 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
749 }
751 static struct tree_content *new_tree_content(unsigned int cnt)
752 {
753 struct avail_tree_content *f, *l = NULL;
754 struct tree_content *t;
755 unsigned int hc = hc_entries(cnt);
757 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
758 if (f->entry_capacity >= cnt)
759 break;
761 if (f) {
762 if (l)
763 l->next_avail = f->next_avail;
764 else
765 avail_tree_table[hc] = f->next_avail;
766 } else {
767 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
768 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
769 f->entry_capacity = cnt;
770 }
772 t = (struct tree_content*)f;
773 t->entry_count = 0;
774 t->delta_depth = 0;
775 return t;
776 }
778 static void release_tree_entry(struct tree_entry *e);
779 static void release_tree_content(struct tree_content *t)
780 {
781 struct avail_tree_content *f = (struct avail_tree_content*)t;
782 unsigned int hc = hc_entries(f->entry_capacity);
783 f->next_avail = avail_tree_table[hc];
784 avail_tree_table[hc] = f;
785 }
787 static void release_tree_content_recursive(struct tree_content *t)
788 {
789 unsigned int i;
790 for (i = 0; i < t->entry_count; i++)
791 release_tree_entry(t->entries[i]);
792 release_tree_content(t);
793 }
795 static struct tree_content *grow_tree_content(
796 struct tree_content *t,
797 int amt)
798 {
799 struct tree_content *r = new_tree_content(t->entry_count + amt);
800 r->entry_count = t->entry_count;
801 r->delta_depth = t->delta_depth;
802 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
803 release_tree_content(t);
804 return r;
805 }
807 static struct tree_entry *new_tree_entry(void)
808 {
809 struct tree_entry *e;
811 if (!avail_tree_entry) {
812 unsigned int n = tree_entry_alloc;
813 total_allocd += n * sizeof(struct tree_entry);
814 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
815 while (n-- > 1) {
816 *((void**)e) = e + 1;
817 e++;
818 }
819 *((void**)e) = NULL;
820 }
822 e = avail_tree_entry;
823 avail_tree_entry = *((void**)e);
824 return e;
825 }
827 static void release_tree_entry(struct tree_entry *e)
828 {
829 if (e->tree)
830 release_tree_content_recursive(e->tree);
831 *((void**)e) = avail_tree_entry;
832 avail_tree_entry = e;
833 }
835 static struct tree_content *dup_tree_content(struct tree_content *s)
836 {
837 struct tree_content *d;
838 struct tree_entry *a, *b;
839 unsigned int i;
841 if (!s)
842 return NULL;
843 d = new_tree_content(s->entry_count);
844 for (i = 0; i < s->entry_count; i++) {
845 a = s->entries[i];
846 b = new_tree_entry();
847 memcpy(b, a, sizeof(*a));
848 if (a->tree && is_null_sha1(b->versions[1].sha1))
849 b->tree = dup_tree_content(a->tree);
850 else
851 b->tree = NULL;
852 d->entries[i] = b;
853 }
854 d->entry_count = s->entry_count;
855 d->delta_depth = s->delta_depth;
857 return d;
858 }
860 static void start_packfile(void)
861 {
862 static char tmpfile[PATH_MAX];
863 struct packed_git *p;
864 struct pack_header hdr;
865 int pack_fd;
867 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
868 "pack/tmp_pack_XXXXXX");
869 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
870 strcpy(p->pack_name, tmpfile);
871 p->pack_fd = pack_fd;
872 p->do_not_close = 1;
873 pack_file = sha1fd(pack_fd, p->pack_name);
875 hdr.hdr_signature = htonl(PACK_SIGNATURE);
876 hdr.hdr_version = htonl(2);
877 hdr.hdr_entries = 0;
878 sha1write(pack_file, &hdr, sizeof(hdr));
880 pack_data = p;
881 pack_size = sizeof(hdr);
882 object_count = 0;
884 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
885 all_packs[pack_id] = p;
886 }
888 static const char *create_index(void)
889 {
890 const char *tmpfile;
891 struct pack_idx_entry **idx, **c, **last;
892 struct object_entry *e;
893 struct object_entry_pool *o;
895 /* Build the table of object IDs. */
896 idx = xmalloc(object_count * sizeof(*idx));
897 c = idx;
898 for (o = blocks; o; o = o->next_pool)
899 for (e = o->next_free; e-- != o->entries;)
900 if (pack_id == e->pack_id)
901 *c++ = &e->idx;
902 last = idx + object_count;
903 if (c != last)
904 die("internal consistency error creating the index");
906 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
907 free(idx);
908 return tmpfile;
909 }
911 static char *keep_pack(const char *curr_index_name)
912 {
913 static char name[PATH_MAX];
914 static const char *keep_msg = "fast-import";
915 int keep_fd;
917 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
918 if (keep_fd < 0)
919 die_errno("cannot create keep file");
920 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
921 if (close(keep_fd))
922 die_errno("failed to write keep file");
924 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
925 get_object_directory(), sha1_to_hex(pack_data->sha1));
926 if (move_temp_to_file(pack_data->pack_name, name))
927 die("cannot store pack file");
929 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
930 get_object_directory(), sha1_to_hex(pack_data->sha1));
931 if (move_temp_to_file(curr_index_name, name))
932 die("cannot store index file");
933 free((void *)curr_index_name);
934 return name;
935 }
937 static void unkeep_all_packs(void)
938 {
939 static char name[PATH_MAX];
940 int k;
942 for (k = 0; k < pack_id; k++) {
943 struct packed_git *p = all_packs[k];
944 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
945 get_object_directory(), sha1_to_hex(p->sha1));
946 unlink_or_warn(name);
947 }
948 }
950 static void end_packfile(void)
951 {
952 struct packed_git *old_p = pack_data, *new_p;
954 clear_delta_base_cache();
955 if (object_count) {
956 unsigned char cur_pack_sha1[20];
957 char *idx_name;
958 int i;
959 struct branch *b;
960 struct tag *t;
962 close_pack_windows(pack_data);
963 sha1close(pack_file, cur_pack_sha1, 0);
964 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
965 pack_data->pack_name, object_count,
966 cur_pack_sha1, pack_size);
967 close(pack_data->pack_fd);
968 idx_name = keep_pack(create_index());
970 /* Register the packfile with core git's machinery. */
971 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
972 if (!new_p)
973 die("core git rejected index %s", idx_name);
974 all_packs[pack_id] = new_p;
975 install_packed_git(new_p);
977 /* Print the boundary */
978 if (pack_edges) {
979 fprintf(pack_edges, "%s:", new_p->pack_name);
980 for (i = 0; i < branch_table_sz; i++) {
981 for (b = branch_table[i]; b; b = b->table_next_branch) {
982 if (b->pack_id == pack_id)
983 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
984 }
985 }
986 for (t = first_tag; t; t = t->next_tag) {
987 if (t->pack_id == pack_id)
988 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
989 }
990 fputc('\n', pack_edges);
991 fflush(pack_edges);
992 }
994 pack_id++;
995 }
996 else {
997 close(old_p->pack_fd);
998 unlink_or_warn(old_p->pack_name);
999 }
1000 free(old_p);
1002 /* We can't carry a delta across packfiles. */
1003 strbuf_release(&last_blob.data);
1004 last_blob.offset = 0;
1005 last_blob.depth = 0;
1006 }
1008 static void cycle_packfile(void)
1009 {
1010 end_packfile();
1011 start_packfile();
1012 }
1014 static int store_object(
1015 enum object_type type,
1016 struct strbuf *dat,
1017 struct last_object *last,
1018 unsigned char *sha1out,
1019 uintmax_t mark)
1020 {
1021 void *out, *delta;
1022 struct object_entry *e;
1023 unsigned char hdr[96];
1024 unsigned char sha1[20];
1025 unsigned long hdrlen, deltalen;
1026 git_SHA_CTX c;
1027 git_zstream s;
1029 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1030 (unsigned long)dat->len) + 1;
1031 git_SHA1_Init(&c);
1032 git_SHA1_Update(&c, hdr, hdrlen);
1033 git_SHA1_Update(&c, dat->buf, dat->len);
1034 git_SHA1_Final(sha1, &c);
1035 if (sha1out)
1036 hashcpy(sha1out, sha1);
1038 e = insert_object(sha1);
1039 if (mark)
1040 insert_mark(mark, e);
1041 if (e->idx.offset) {
1042 duplicate_count_by_type[type]++;
1043 return 1;
1044 } else if (find_sha1_pack(sha1, packed_git)) {
1045 e->type = type;
1046 e->pack_id = MAX_PACK_ID;
1047 e->idx.offset = 1; /* just not zero! */
1048 duplicate_count_by_type[type]++;
1049 return 1;
1050 }
1052 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1053 delta = diff_delta(last->data.buf, last->data.len,
1054 dat->buf, dat->len,
1055 &deltalen, dat->len - 20);
1056 } else
1057 delta = NULL;
1059 memset(&s, 0, sizeof(s));
1060 git_deflate_init(&s, pack_compression_level);
1061 if (delta) {
1062 s.next_in = delta;
1063 s.avail_in = deltalen;
1064 } else {
1065 s.next_in = (void *)dat->buf;
1066 s.avail_in = dat->len;
1067 }
1068 s.avail_out = git_deflate_bound(&s, s.avail_in);
1069 s.next_out = out = xmalloc(s.avail_out);
1070 while (git_deflate(&s, Z_FINISH) == Z_OK)
1071 ; /* nothing */
1072 git_deflate_end(&s);
1074 /* Determine if we should auto-checkpoint. */
1075 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1076 || (pack_size + 60 + s.total_out) < pack_size) {
1078 /* This new object needs to *not* have the current pack_id. */
1079 e->pack_id = pack_id + 1;
1080 cycle_packfile();
1082 /* We cannot carry a delta into the new pack. */
1083 if (delta) {
1084 free(delta);
1085 delta = NULL;
1087 memset(&s, 0, sizeof(s));
1088 git_deflate_init(&s, pack_compression_level);
1089 s.next_in = (void *)dat->buf;
1090 s.avail_in = dat->len;
1091 s.avail_out = git_deflate_bound(&s, s.avail_in);
1092 s.next_out = out = xrealloc(out, s.avail_out);
1093 while (git_deflate(&s, Z_FINISH) == Z_OK)
1094 ; /* nothing */
1095 git_deflate_end(&s);
1096 }
1097 }
1099 e->type = type;
1100 e->pack_id = pack_id;
1101 e->idx.offset = pack_size;
1102 object_count++;
1103 object_count_by_type[type]++;
1105 crc32_begin(pack_file);
1107 if (delta) {
1108 off_t ofs = e->idx.offset - last->offset;
1109 unsigned pos = sizeof(hdr) - 1;
1111 delta_count_by_type[type]++;
1112 e->depth = last->depth + 1;
1114 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1115 sha1write(pack_file, hdr, hdrlen);
1116 pack_size += hdrlen;
1118 hdr[pos] = ofs & 127;
1119 while (ofs >>= 7)
1120 hdr[--pos] = 128 | (--ofs & 127);
1121 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1122 pack_size += sizeof(hdr) - pos;
1123 } else {
1124 e->depth = 0;
1125 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1126 sha1write(pack_file, hdr, hdrlen);
1127 pack_size += hdrlen;
1128 }
1130 sha1write(pack_file, out, s.total_out);
1131 pack_size += s.total_out;
1133 e->idx.crc32 = crc32_end(pack_file);
1135 free(out);
1136 free(delta);
1137 if (last) {
1138 if (last->no_swap) {
1139 last->data = *dat;
1140 } else {
1141 strbuf_swap(&last->data, dat);
1142 }
1143 last->offset = e->idx.offset;
1144 last->depth = e->depth;
1145 }
1146 return 0;
1147 }
1149 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1150 {
1151 if (ftruncate(pack_data->pack_fd, to)
1152 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1153 die_errno("cannot truncate pack to skip duplicate");
1154 pack_size = to;
1156 /* yes this is a layering violation */
1157 pack_file->total = to;
1158 pack_file->offset = 0;
1159 pack_file->ctx = *ctx;
1160 }
1162 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1163 {
1164 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1165 unsigned char *in_buf = xmalloc(in_sz);
1166 unsigned char *out_buf = xmalloc(out_sz);
1167 struct object_entry *e;
1168 unsigned char sha1[20];
1169 unsigned long hdrlen;
1170 off_t offset;
1171 git_SHA_CTX c;
1172 git_SHA_CTX pack_file_ctx;
1173 git_zstream s;
1174 int status = Z_OK;
1176 /* Determine if we should auto-checkpoint. */
1177 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1178 || (pack_size + 60 + len) < pack_size)
1179 cycle_packfile();
1181 offset = pack_size;
1183 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1184 sha1flush(pack_file);
1185 pack_file_ctx = pack_file->ctx;
1187 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1188 if (out_sz <= hdrlen)
1189 die("impossibly large object header");
1191 git_SHA1_Init(&c);
1192 git_SHA1_Update(&c, out_buf, hdrlen);
1194 crc32_begin(pack_file);
1196 memset(&s, 0, sizeof(s));
1197 git_deflate_init(&s, pack_compression_level);
1199 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1200 if (out_sz <= hdrlen)
1201 die("impossibly large object header");
1203 s.next_out = out_buf + hdrlen;
1204 s.avail_out = out_sz - hdrlen;
1206 while (status != Z_STREAM_END) {
1207 if (0 < len && !s.avail_in) {
1208 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1209 size_t n = fread(in_buf, 1, cnt, stdin);
1210 if (!n && feof(stdin))
1211 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1213 git_SHA1_Update(&c, in_buf, n);
1214 s.next_in = in_buf;
1215 s.avail_in = n;
1216 len -= n;
1217 }
1219 status = git_deflate(&s, len ? 0 : Z_FINISH);
1221 if (!s.avail_out || status == Z_STREAM_END) {
1222 size_t n = s.next_out - out_buf;
1223 sha1write(pack_file, out_buf, n);
1224 pack_size += n;
1225 s.next_out = out_buf;
1226 s.avail_out = out_sz;
1227 }
1229 switch (status) {
1230 case Z_OK:
1231 case Z_BUF_ERROR:
1232 case Z_STREAM_END:
1233 continue;
1234 default:
1235 die("unexpected deflate failure: %d", status);
1236 }
1237 }
1238 git_deflate_end(&s);
1239 git_SHA1_Final(sha1, &c);
1241 if (sha1out)
1242 hashcpy(sha1out, sha1);
1244 e = insert_object(sha1);
1246 if (mark)
1247 insert_mark(mark, e);
1249 if (e->idx.offset) {
1250 duplicate_count_by_type[OBJ_BLOB]++;
1251 truncate_pack(offset, &pack_file_ctx);
1253 } else if (find_sha1_pack(sha1, packed_git)) {
1254 e->type = OBJ_BLOB;
1255 e->pack_id = MAX_PACK_ID;
1256 e->idx.offset = 1; /* just not zero! */
1257 duplicate_count_by_type[OBJ_BLOB]++;
1258 truncate_pack(offset, &pack_file_ctx);
1260 } else {
1261 e->depth = 0;
1262 e->type = OBJ_BLOB;
1263 e->pack_id = pack_id;
1264 e->idx.offset = offset;
1265 e->idx.crc32 = crc32_end(pack_file);
1266 object_count++;
1267 object_count_by_type[OBJ_BLOB]++;
1268 }
1270 free(in_buf);
1271 free(out_buf);
1272 }
1274 /* All calls must be guarded by find_object() or find_mark() to
1275 * ensure the 'struct object_entry' passed was written by this
1276 * process instance. We unpack the entry by the offset, avoiding
1277 * the need for the corresponding .idx file. This unpacking rule
1278 * works because we only use OBJ_REF_DELTA within the packfiles
1279 * created by fast-import.
1280 *
1281 * oe must not be NULL. Such an oe usually comes from giving
1282 * an unknown SHA-1 to find_object() or an undefined mark to
1283 * find_mark(). Callers must test for this condition and use
1284 * the standard read_sha1_file() when it happens.
1285 *
1286 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1287 * find_mark(), where the mark was reloaded from an existing marks
1288 * file and is referencing an object that this fast-import process
1289 * instance did not write out to a packfile. Callers must test for
1290 * this condition and use read_sha1_file() instead.
1291 */
1292 static void *gfi_unpack_entry(
1293 struct object_entry *oe,
1294 unsigned long *sizep)
1295 {
1296 enum object_type type;
1297 struct packed_git *p = all_packs[oe->pack_id];
1298 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1299 /* The object is stored in the packfile we are writing to
1300 * and we have modified it since the last time we scanned
1301 * back to read a previously written object. If an old
1302 * window covered [p->pack_size, p->pack_size + 20) its
1303 * data is stale and is not valid. Closing all windows
1304 * and updating the packfile length ensures we can read
1305 * the newly written data.
1306 */
1307 close_pack_windows(p);
1308 sha1flush(pack_file);
1310 /* We have to offer 20 bytes additional on the end of
1311 * the packfile as the core unpacker code assumes the
1312 * footer is present at the file end and must promise
1313 * at least 20 bytes within any window it maps. But
1314 * we don't actually create the footer here.
1315 */
1316 p->pack_size = pack_size + 20;
1317 }
1318 return unpack_entry(p, oe->idx.offset, &type, sizep);
1319 }
1321 static const char *get_mode(const char *str, uint16_t *modep)
1322 {
1323 unsigned char c;
1324 uint16_t mode = 0;
1326 while ((c = *str++) != ' ') {
1327 if (c < '0' || c > '7')
1328 return NULL;
1329 mode = (mode << 3) + (c - '0');
1330 }
1331 *modep = mode;
1332 return str;
1333 }
1335 static void load_tree(struct tree_entry *root)
1336 {
1337 unsigned char *sha1 = root->versions[1].sha1;
1338 struct object_entry *myoe;
1339 struct tree_content *t;
1340 unsigned long size;
1341 char *buf;
1342 const char *c;
1344 root->tree = t = new_tree_content(8);
1345 if (is_null_sha1(sha1))
1346 return;
1348 myoe = find_object(sha1);
1349 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1350 if (myoe->type != OBJ_TREE)
1351 die("Not a tree: %s", sha1_to_hex(sha1));
1352 t->delta_depth = myoe->depth;
1353 buf = gfi_unpack_entry(myoe, &size);
1354 if (!buf)
1355 die("Can't load tree %s", sha1_to_hex(sha1));
1356 } else {
1357 enum object_type type;
1358 buf = read_sha1_file(sha1, &type, &size);
1359 if (!buf || type != OBJ_TREE)
1360 die("Can't load tree %s", sha1_to_hex(sha1));
1361 }
1363 c = buf;
1364 while (c != (buf + size)) {
1365 struct tree_entry *e = new_tree_entry();
1367 if (t->entry_count == t->entry_capacity)
1368 root->tree = t = grow_tree_content(t, t->entry_count);
1369 t->entries[t->entry_count++] = e;
1371 e->tree = NULL;
1372 c = get_mode(c, &e->versions[1].mode);
1373 if (!c)
1374 die("Corrupt mode in %s", sha1_to_hex(sha1));
1375 e->versions[0].mode = e->versions[1].mode;
1376 e->name = to_atom(c, strlen(c));
1377 c += e->name->str_len + 1;
1378 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1379 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1380 c += 20;
1381 }
1382 free(buf);
1383 }
1385 static int tecmp0 (const void *_a, const void *_b)
1386 {
1387 struct tree_entry *a = *((struct tree_entry**)_a);
1388 struct tree_entry *b = *((struct tree_entry**)_b);
1389 return base_name_compare(
1390 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1391 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1392 }
1394 static int tecmp1 (const void *_a, const void *_b)
1395 {
1396 struct tree_entry *a = *((struct tree_entry**)_a);
1397 struct tree_entry *b = *((struct tree_entry**)_b);
1398 return base_name_compare(
1399 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1400 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1401 }
1403 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1404 {
1405 size_t maxlen = 0;
1406 unsigned int i;
1408 if (!v)
1409 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1410 else
1411 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1413 for (i = 0; i < t->entry_count; i++) {
1414 if (t->entries[i]->versions[v].mode)
1415 maxlen += t->entries[i]->name->str_len + 34;
1416 }
1418 strbuf_reset(b);
1419 strbuf_grow(b, maxlen);
1420 for (i = 0; i < t->entry_count; i++) {
1421 struct tree_entry *e = t->entries[i];
1422 if (!e->versions[v].mode)
1423 continue;
1424 strbuf_addf(b, "%o %s%c",
1425 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1426 e->name->str_dat, '\0');
1427 strbuf_add(b, e->versions[v].sha1, 20);
1428 }
1429 }
1431 static void store_tree(struct tree_entry *root)
1432 {
1433 struct tree_content *t = root->tree;
1434 unsigned int i, j, del;
1435 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1436 struct object_entry *le = NULL;
1438 if (!is_null_sha1(root->versions[1].sha1))
1439 return;
1441 for (i = 0; i < t->entry_count; i++) {
1442 if (t->entries[i]->tree)
1443 store_tree(t->entries[i]);
1444 }
1446 if (!(root->versions[0].mode & NO_DELTA))
1447 le = find_object(root->versions[0].sha1);
1448 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1449 mktree(t, 0, &old_tree);
1450 lo.data = old_tree;
1451 lo.offset = le->idx.offset;
1452 lo.depth = t->delta_depth;
1453 }
1455 mktree(t, 1, &new_tree);
1456 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1458 t->delta_depth = lo.depth;
1459 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1460 struct tree_entry *e = t->entries[i];
1461 if (e->versions[1].mode) {
1462 e->versions[0].mode = e->versions[1].mode;
1463 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1464 t->entries[j++] = e;
1465 } else {
1466 release_tree_entry(e);
1467 del++;
1468 }
1469 }
1470 t->entry_count -= del;
1471 }
1473 static void tree_content_replace(
1474 struct tree_entry *root,
1475 const unsigned char *sha1,
1476 const uint16_t mode,
1477 struct tree_content *newtree)
1478 {
1479 if (!S_ISDIR(mode))
1480 die("Root cannot be a non-directory");
1481 hashclr(root->versions[0].sha1);
1482 hashcpy(root->versions[1].sha1, sha1);
1483 if (root->tree)
1484 release_tree_content_recursive(root->tree);
1485 root->tree = newtree;
1486 }
1488 static int tree_content_set(
1489 struct tree_entry *root,
1490 const char *p,
1491 const unsigned char *sha1,
1492 const uint16_t mode,
1493 struct tree_content *subtree)
1494 {
1495 struct tree_content *t;
1496 const char *slash1;
1497 unsigned int i, n;
1498 struct tree_entry *e;
1500 slash1 = strchr(p, '/');
1501 if (slash1)
1502 n = slash1 - p;
1503 else
1504 n = strlen(p);
1505 if (!n)
1506 die("Empty path component found in input");
1507 if (!slash1 && !S_ISDIR(mode) && subtree)
1508 die("Non-directories cannot have subtrees");
1510 if (!root->tree)
1511 load_tree(root);
1512 t = root->tree;
1513 for (i = 0; i < t->entry_count; i++) {
1514 e = t->entries[i];
1515 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1516 if (!slash1) {
1517 if (!S_ISDIR(mode)
1518 && e->versions[1].mode == mode
1519 && !hashcmp(e->versions[1].sha1, sha1))
1520 return 0;
1521 e->versions[1].mode = mode;
1522 hashcpy(e->versions[1].sha1, sha1);
1523 if (e->tree)
1524 release_tree_content_recursive(e->tree);
1525 e->tree = subtree;
1527 /*
1528 * We need to leave e->versions[0].sha1 alone
1529 * to avoid modifying the preimage tree used
1530 * when writing out the parent directory.
1531 * But after replacing the subdir with a
1532 * completely different one, it's not a good
1533 * delta base any more, and besides, we've
1534 * thrown away the tree entries needed to
1535 * make a delta against it.
1536 *
1537 * So let's just explicitly disable deltas
1538 * for the subtree.
1539 */
1540 if (S_ISDIR(e->versions[0].mode))
1541 e->versions[0].mode |= NO_DELTA;
1543 hashclr(root->versions[1].sha1);
1544 return 1;
1545 }
1546 if (!S_ISDIR(e->versions[1].mode)) {
1547 e->tree = new_tree_content(8);
1548 e->versions[1].mode = S_IFDIR;
1549 }
1550 if (!e->tree)
1551 load_tree(e);
1552 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1553 hashclr(root->versions[1].sha1);
1554 return 1;
1555 }
1556 return 0;
1557 }
1558 }
1560 if (t->entry_count == t->entry_capacity)
1561 root->tree = t = grow_tree_content(t, t->entry_count);
1562 e = new_tree_entry();
1563 e->name = to_atom(p, n);
1564 e->versions[0].mode = 0;
1565 hashclr(e->versions[0].sha1);
1566 t->entries[t->entry_count++] = e;
1567 if (slash1) {
1568 e->tree = new_tree_content(8);
1569 e->versions[1].mode = S_IFDIR;
1570 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1571 } else {
1572 e->tree = subtree;
1573 e->versions[1].mode = mode;
1574 hashcpy(e->versions[1].sha1, sha1);
1575 }
1576 hashclr(root->versions[1].sha1);
1577 return 1;
1578 }
1580 static int tree_content_remove(
1581 struct tree_entry *root,
1582 const char *p,
1583 struct tree_entry *backup_leaf)
1584 {
1585 struct tree_content *t;
1586 const char *slash1;
1587 unsigned int i, n;
1588 struct tree_entry *e;
1590 slash1 = strchr(p, '/');
1591 if (slash1)
1592 n = slash1 - p;
1593 else
1594 n = strlen(p);
1596 if (!root->tree)
1597 load_tree(root);
1598 t = root->tree;
1599 for (i = 0; i < t->entry_count; i++) {
1600 e = t->entries[i];
1601 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1602 if (slash1 && !S_ISDIR(e->versions[1].mode))
1603 /*
1604 * If p names a file in some subdirectory, and a
1605 * file or symlink matching the name of the
1606 * parent directory of p exists, then p cannot
1607 * exist and need not be deleted.
1608 */
1609 return 1;
1610 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1611 goto del_entry;
1612 if (!e->tree)
1613 load_tree(e);
1614 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1615 for (n = 0; n < e->tree->entry_count; n++) {
1616 if (e->tree->entries[n]->versions[1].mode) {
1617 hashclr(root->versions[1].sha1);
1618 return 1;
1619 }
1620 }
1621 backup_leaf = NULL;
1622 goto del_entry;
1623 }
1624 return 0;
1625 }
1626 }
1627 return 0;
1629 del_entry:
1630 if (backup_leaf)
1631 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1632 else if (e->tree)
1633 release_tree_content_recursive(e->tree);
1634 e->tree = NULL;
1635 e->versions[1].mode = 0;
1636 hashclr(e->versions[1].sha1);
1637 hashclr(root->versions[1].sha1);
1638 return 1;
1639 }
1641 static int tree_content_get(
1642 struct tree_entry *root,
1643 const char *p,
1644 struct tree_entry *leaf)
1645 {
1646 struct tree_content *t;
1647 const char *slash1;
1648 unsigned int i, n;
1649 struct tree_entry *e;
1651 slash1 = strchr(p, '/');
1652 if (slash1)
1653 n = slash1 - p;
1654 else
1655 n = strlen(p);
1657 if (!root->tree)
1658 load_tree(root);
1659 t = root->tree;
1660 for (i = 0; i < t->entry_count; i++) {
1661 e = t->entries[i];
1662 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1663 if (!slash1) {
1664 memcpy(leaf, e, sizeof(*leaf));
1665 if (e->tree && is_null_sha1(e->versions[1].sha1))
1666 leaf->tree = dup_tree_content(e->tree);
1667 else
1668 leaf->tree = NULL;
1669 return 1;
1670 }
1671 if (!S_ISDIR(e->versions[1].mode))
1672 return 0;
1673 if (!e->tree)
1674 load_tree(e);
1675 return tree_content_get(e, slash1 + 1, leaf);
1676 }
1677 }
1678 return 0;
1679 }
1681 static int update_branch(struct branch *b)
1682 {
1683 static const char *msg = "fast-import";
1684 struct ref_lock *lock;
1685 unsigned char old_sha1[20];
1687 if (is_null_sha1(b->sha1))
1688 return 0;
1689 if (read_ref(b->name, old_sha1))
1690 hashclr(old_sha1);
1691 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1692 if (!lock)
1693 return error("Unable to lock %s", b->name);
1694 if (!force_update && !is_null_sha1(old_sha1)) {
1695 struct commit *old_cmit, *new_cmit;
1697 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1698 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1699 if (!old_cmit || !new_cmit) {
1700 unlock_ref(lock);
1701 return error("Branch %s is missing commits.", b->name);
1702 }
1704 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1705 unlock_ref(lock);
1706 warning("Not updating %s"
1707 " (new tip %s does not contain %s)",
1708 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1709 return -1;
1710 }
1711 }
1712 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1713 return error("Unable to update %s", b->name);
1714 return 0;
1715 }
1717 static void dump_branches(void)
1718 {
1719 unsigned int i;
1720 struct branch *b;
1722 for (i = 0; i < branch_table_sz; i++) {
1723 for (b = branch_table[i]; b; b = b->table_next_branch)
1724 failure |= update_branch(b);
1725 }
1726 }
1728 static void dump_tags(void)
1729 {
1730 static const char *msg = "fast-import";
1731 struct tag *t;
1732 struct ref_lock *lock;
1733 char ref_name[PATH_MAX];
1735 for (t = first_tag; t; t = t->next_tag) {
1736 sprintf(ref_name, "tags/%s", t->name);
1737 lock = lock_ref_sha1(ref_name, NULL);
1738 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1739 failure |= error("Unable to update %s", ref_name);
1740 }
1741 }
1743 static void dump_marks_helper(FILE *f,
1744 uintmax_t base,
1745 struct mark_set *m)
1746 {
1747 uintmax_t k;
1748 if (m->shift) {
1749 for (k = 0; k < 1024; k++) {
1750 if (m->data.sets[k])
1751 dump_marks_helper(f, base + (k << m->shift),
1752 m->data.sets[k]);
1753 }
1754 } else {
1755 for (k = 0; k < 1024; k++) {
1756 if (m->data.marked[k])
1757 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1758 sha1_to_hex(m->data.marked[k]->idx.sha1));
1759 }
1760 }
1761 }
1763 static void dump_marks(void)
1764 {
1765 static struct lock_file mark_lock;
1766 int mark_fd;
1767 FILE *f;
1769 if (!export_marks_file)
1770 return;
1772 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1773 if (mark_fd < 0) {
1774 failure |= error("Unable to write marks file %s: %s",
1775 export_marks_file, strerror(errno));
1776 return;
1777 }
1779 f = fdopen(mark_fd, "w");
1780 if (!f) {
1781 int saved_errno = errno;
1782 rollback_lock_file(&mark_lock);
1783 failure |= error("Unable to write marks file %s: %s",
1784 export_marks_file, strerror(saved_errno));
1785 return;
1786 }
1788 /*
1789 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1790 * Assign -1 to the lock file descriptor so that commit_lock_file()
1791 * won't try to close() it.
1792 */
1793 mark_lock.fd = -1;
1795 dump_marks_helper(f, 0, marks);
1796 if (ferror(f) || fclose(f)) {
1797 int saved_errno = errno;
1798 rollback_lock_file(&mark_lock);
1799 failure |= error("Unable to write marks file %s: %s",
1800 export_marks_file, strerror(saved_errno));
1801 return;
1802 }
1804 if (commit_lock_file(&mark_lock)) {
1805 int saved_errno = errno;
1806 rollback_lock_file(&mark_lock);
1807 failure |= error("Unable to commit marks file %s: %s",
1808 export_marks_file, strerror(saved_errno));
1809 return;
1810 }
1811 }
1813 static void read_marks(void)
1814 {
1815 char line[512];
1816 FILE *f = fopen(import_marks_file, "r");
1817 if (f)
1818 ;
1819 else if (import_marks_file_ignore_missing && errno == ENOENT)
1820 return; /* Marks file does not exist */
1821 else
1822 die_errno("cannot read '%s'", import_marks_file);
1823 while (fgets(line, sizeof(line), f)) {
1824 uintmax_t mark;
1825 char *end;
1826 unsigned char sha1[20];
1827 struct object_entry *e;
1829 end = strchr(line, '\n');
1830 if (line[0] != ':' || !end)
1831 die("corrupt mark line: %s", line);
1832 *end = 0;
1833 mark = strtoumax(line + 1, &end, 10);
1834 if (!mark || end == line + 1
1835 || *end != ' ' || get_sha1(end + 1, sha1))
1836 die("corrupt mark line: %s", line);
1837 e = find_object(sha1);
1838 if (!e) {
1839 enum object_type type = sha1_object_info(sha1, NULL);
1840 if (type < 0)
1841 die("object not found: %s", sha1_to_hex(sha1));
1842 e = insert_object(sha1);
1843 e->type = type;
1844 e->pack_id = MAX_PACK_ID;
1845 e->idx.offset = 1; /* just not zero! */
1846 }
1847 insert_mark(mark, e);
1848 }
1849 fclose(f);
1850 }
1853 static int read_next_command(void)
1854 {
1855 static int stdin_eof = 0;
1857 if (stdin_eof) {
1858 unread_command_buf = 0;
1859 return EOF;
1860 }
1862 for (;;) {
1863 if (unread_command_buf) {
1864 unread_command_buf = 0;
1865 } else {
1866 struct recent_command *rc;
1868 strbuf_detach(&command_buf, NULL);
1869 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1870 if (stdin_eof)
1871 return EOF;
1873 if (!seen_data_command
1874 && prefixcmp(command_buf.buf, "feature ")
1875 && prefixcmp(command_buf.buf, "option ")) {
1876 parse_argv();
1877 }
1879 rc = rc_free;
1880 if (rc)
1881 rc_free = rc->next;
1882 else {
1883 rc = cmd_hist.next;
1884 cmd_hist.next = rc->next;
1885 cmd_hist.next->prev = &cmd_hist;
1886 free(rc->buf);
1887 }
1889 rc->buf = command_buf.buf;
1890 rc->prev = cmd_tail;
1891 rc->next = cmd_hist.prev;
1892 rc->prev->next = rc;
1893 cmd_tail = rc;
1894 }
1895 if (!prefixcmp(command_buf.buf, "cat-blob ")) {
1896 parse_cat_blob();
1897 continue;
1898 }
1899 if (command_buf.buf[0] == '#')
1900 continue;
1901 return 0;
1902 }
1903 }
1905 static void skip_optional_lf(void)
1906 {
1907 int term_char = fgetc(stdin);
1908 if (term_char != '\n' && term_char != EOF)
1909 ungetc(term_char, stdin);
1910 }
1912 static void parse_mark(void)
1913 {
1914 if (!prefixcmp(command_buf.buf, "mark :")) {
1915 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1916 read_next_command();
1917 }
1918 else
1919 next_mark = 0;
1920 }
1922 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1923 {
1924 strbuf_reset(sb);
1926 if (prefixcmp(command_buf.buf, "data "))
1927 die("Expected 'data n' command, found: %s", command_buf.buf);
1929 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1930 char *term = xstrdup(command_buf.buf + 5 + 2);
1931 size_t term_len = command_buf.len - 5 - 2;
1933 strbuf_detach(&command_buf, NULL);
1934 for (;;) {
1935 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1936 die("EOF in data (terminator '%s' not found)", term);
1937 if (term_len == command_buf.len
1938 && !strcmp(term, command_buf.buf))
1939 break;
1940 strbuf_addbuf(sb, &command_buf);
1941 strbuf_addch(sb, '\n');
1942 }
1943 free(term);
1944 }
1945 else {
1946 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1947 size_t n = 0, length = (size_t)len;
1949 if (limit && limit < len) {
1950 *len_res = len;
1951 return 0;
1952 }
1953 if (length < len)
1954 die("data is too large to use in this context");
1956 while (n < length) {
1957 size_t s = strbuf_fread(sb, length - n, stdin);
1958 if (!s && feof(stdin))
1959 die("EOF in data (%lu bytes remaining)",
1960 (unsigned long)(length - n));
1961 n += s;
1962 }
1963 }
1965 skip_optional_lf();
1966 return 1;
1967 }
1969 static int validate_raw_date(const char *src, char *result, int maxlen)
1970 {
1971 const char *orig_src = src;
1972 char *endp;
1973 unsigned long num;
1975 errno = 0;
1977 num = strtoul(src, &endp, 10);
1978 /* NEEDSWORK: perhaps check for reasonable values? */
1979 if (errno || endp == src || *endp != ' ')
1980 return -1;
1982 src = endp + 1;
1983 if (*src != '-' && *src != '+')
1984 return -1;
1986 num = strtoul(src + 1, &endp, 10);
1987 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1988 1400 < num)
1989 return -1;
1991 strcpy(result, orig_src);
1992 return 0;
1993 }
1995 static char *parse_ident(const char *buf)
1996 {
1997 const char *ltgt;
1998 size_t name_len;
1999 char *ident;
2001 /* ensure there is a space delimiter even if there is no name */
2002 if (*buf == '<')
2003 --buf;
2005 ltgt = buf + strcspn(buf, "<>");
2006 if (*ltgt != '<')
2007 die("Missing < in ident string: %s", buf);
2008 if (ltgt != buf && ltgt[-1] != ' ')
2009 die("Missing space before < in ident string: %s", buf);
2010 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2011 if (*ltgt != '>')
2012 die("Missing > in ident string: %s", buf);
2013 ltgt++;
2014 if (*ltgt != ' ')
2015 die("Missing space after > in ident string: %s", buf);
2016 ltgt++;
2017 name_len = ltgt - buf;
2018 ident = xmalloc(name_len + 24);
2019 strncpy(ident, buf, name_len);
2021 switch (whenspec) {
2022 case WHENSPEC_RAW:
2023 if (validate_raw_date(ltgt, ident + name_len, 24) < 0)
2024 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2025 break;
2026 case WHENSPEC_RFC2822:
2027 if (parse_date(ltgt, ident + name_len, 24) < 0)
2028 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2029 break;
2030 case WHENSPEC_NOW:
2031 if (strcmp("now", ltgt))
2032 die("Date in ident must be 'now': %s", buf);
2033 datestamp(ident + name_len, 24);
2034 break;
2035 }
2037 return ident;
2038 }
2040 static void parse_and_store_blob(
2041 struct last_object *last,
2042 unsigned char *sha1out,
2043 uintmax_t mark)
2044 {
2045 static struct strbuf buf = STRBUF_INIT;
2046 uintmax_t len;
2048 if (parse_data(&buf, big_file_threshold, &len))
2049 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2050 else {
2051 if (last) {
2052 strbuf_release(&last->data);
2053 last->offset = 0;
2054 last->depth = 0;
2055 }
2056 stream_blob(len, sha1out, mark);
2057 skip_optional_lf();
2058 }
2059 }
2061 static void parse_new_blob(void)
2062 {
2063 read_next_command();
2064 parse_mark();
2065 parse_and_store_blob(&last_blob, NULL, next_mark);
2066 }
2068 static void unload_one_branch(void)
2069 {
2070 while (cur_active_branches
2071 && cur_active_branches >= max_active_branches) {
2072 uintmax_t min_commit = ULONG_MAX;
2073 struct branch *e, *l = NULL, *p = NULL;
2075 for (e = active_branches; e; e = e->active_next_branch) {
2076 if (e->last_commit < min_commit) {
2077 p = l;
2078 min_commit = e->last_commit;
2079 }
2080 l = e;
2081 }
2083 if (p) {
2084 e = p->active_next_branch;
2085 p->active_next_branch = e->active_next_branch;
2086 } else {
2087 e = active_branches;
2088 active_branches = e->active_next_branch;
2089 }
2090 e->active = 0;
2091 e->active_next_branch = NULL;
2092 if (e->branch_tree.tree) {
2093 release_tree_content_recursive(e->branch_tree.tree);
2094 e->branch_tree.tree = NULL;
2095 }
2096 cur_active_branches--;
2097 }
2098 }
2100 static void load_branch(struct branch *b)
2101 {
2102 load_tree(&b->branch_tree);
2103 if (!b->active) {
2104 b->active = 1;
2105 b->active_next_branch = active_branches;
2106 active_branches = b;
2107 cur_active_branches++;
2108 branch_load_count++;
2109 }
2110 }
2112 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2113 {
2114 unsigned char fanout = 0;
2115 while ((num_notes >>= 8))
2116 fanout++;
2117 return fanout;
2118 }
2120 static void construct_path_with_fanout(const char *hex_sha1,
2121 unsigned char fanout, char *path)
2122 {
2123 unsigned int i = 0, j = 0;
2124 if (fanout >= 20)
2125 die("Too large fanout (%u)", fanout);
2126 while (fanout) {
2127 path[i++] = hex_sha1[j++];
2128 path[i++] = hex_sha1[j++];
2129 path[i++] = '/';
2130 fanout--;
2131 }
2132 memcpy(path + i, hex_sha1 + j, 40 - j);
2133 path[i + 40 - j] = '\0';
2134 }
2136 static uintmax_t do_change_note_fanout(
2137 struct tree_entry *orig_root, struct tree_entry *root,
2138 char *hex_sha1, unsigned int hex_sha1_len,
2139 char *fullpath, unsigned int fullpath_len,
2140 unsigned char fanout)
2141 {
2142 struct tree_content *t = root->tree;
2143 struct tree_entry *e, leaf;
2144 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2145 uintmax_t num_notes = 0;
2146 unsigned char sha1[20];
2147 char realpath[60];
2149 for (i = 0; t && i < t->entry_count; i++) {
2150 e = t->entries[i];
2151 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2152 tmp_fullpath_len = fullpath_len;
2154 /*
2155 * We're interested in EITHER existing note entries (entries
2156 * with exactly 40 hex chars in path, not including directory
2157 * separators), OR directory entries that may contain note
2158 * entries (with < 40 hex chars in path).
2159 * Also, each path component in a note entry must be a multiple
2160 * of 2 chars.
2161 */
2162 if (!e->versions[1].mode ||
2163 tmp_hex_sha1_len > 40 ||
2164 e->name->str_len % 2)
2165 continue;
2167 /* This _may_ be a note entry, or a subdir containing notes */
2168 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2169 e->name->str_len);
2170 if (tmp_fullpath_len)
2171 fullpath[tmp_fullpath_len++] = '/';
2172 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2173 e->name->str_len);
2174 tmp_fullpath_len += e->name->str_len;
2175 fullpath[tmp_fullpath_len] = '\0';
2177 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2178 /* This is a note entry */
2179 construct_path_with_fanout(hex_sha1, fanout, realpath);
2180 if (!strcmp(fullpath, realpath)) {
2181 /* Note entry is in correct location */
2182 num_notes++;
2183 continue;
2184 }
2186 /* Rename fullpath to realpath */
2187 if (!tree_content_remove(orig_root, fullpath, &leaf))
2188 die("Failed to remove path %s", fullpath);
2189 tree_content_set(orig_root, realpath,
2190 leaf.versions[1].sha1,
2191 leaf.versions[1].mode,
2192 leaf.tree);
2193 } else if (S_ISDIR(e->versions[1].mode)) {
2194 /* This is a subdir that may contain note entries */
2195 if (!e->tree)
2196 load_tree(e);
2197 num_notes += do_change_note_fanout(orig_root, e,
2198 hex_sha1, tmp_hex_sha1_len,
2199 fullpath, tmp_fullpath_len, fanout);
2200 }
2202 /* The above may have reallocated the current tree_content */
2203 t = root->tree;
2204 }
2205 return num_notes;
2206 }
2208 static uintmax_t change_note_fanout(struct tree_entry *root,
2209 unsigned char fanout)
2210 {
2211 char hex_sha1[40], path[60];
2212 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2213 }
2215 static void file_change_m(struct branch *b)
2216 {
2217 const char *p = command_buf.buf + 2;
2218 static struct strbuf uq = STRBUF_INIT;
2219 const char *endp;
2220 struct object_entry *oe = oe;
2221 unsigned char sha1[20];
2222 uint16_t mode, inline_data = 0;
2224 p = get_mode(p, &mode);
2225 if (!p)
2226 die("Corrupt mode: %s", command_buf.buf);
2227 switch (mode) {
2228 case 0644:
2229 case 0755:
2230 mode |= S_IFREG;
2231 case S_IFREG | 0644:
2232 case S_IFREG | 0755:
2233 case S_IFLNK:
2234 case S_IFDIR:
2235 case S_IFGITLINK:
2236 /* ok */
2237 break;
2238 default:
2239 die("Corrupt mode: %s", command_buf.buf);
2240 }
2242 if (*p == ':') {
2243 char *x;
2244 oe = find_mark(strtoumax(p + 1, &x, 10));
2245 hashcpy(sha1, oe->idx.sha1);
2246 p = x;
2247 } else if (!prefixcmp(p, "inline")) {
2248 inline_data = 1;
2249 p += 6;
2250 } else {
2251 if (get_sha1_hex(p, sha1))
2252 die("Invalid SHA1: %s", command_buf.buf);
2253 oe = find_object(sha1);
2254 p += 40;
2255 }
2256 if (*p++ != ' ')
2257 die("Missing space after SHA1: %s", command_buf.buf);
2259 strbuf_reset(&uq);
2260 if (!unquote_c_style(&uq, p, &endp)) {
2261 if (*endp)
2262 die("Garbage after path in: %s", command_buf.buf);
2263 p = uq.buf;
2264 }
2266 /* Git does not track empty, non-toplevel directories. */
2267 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2268 tree_content_remove(&b->branch_tree, p, NULL);
2269 return;
2270 }
2272 if (S_ISGITLINK(mode)) {
2273 if (inline_data)
2274 die("Git links cannot be specified 'inline': %s",
2275 command_buf.buf);
2276 else if (oe) {
2277 if (oe->type != OBJ_COMMIT)
2278 die("Not a commit (actually a %s): %s",
2279 typename(oe->type), command_buf.buf);
2280 }
2281 /*
2282 * Accept the sha1 without checking; it expected to be in
2283 * another repository.
2284 */
2285 } else if (inline_data) {
2286 if (S_ISDIR(mode))
2287 die("Directories cannot be specified 'inline': %s",
2288 command_buf.buf);
2289 if (p != uq.buf) {
2290 strbuf_addstr(&uq, p);
2291 p = uq.buf;
2292 }
2293 read_next_command();
2294 parse_and_store_blob(&last_blob, sha1, 0);
2295 } else {
2296 enum object_type expected = S_ISDIR(mode) ?
2297 OBJ_TREE: OBJ_BLOB;
2298 enum object_type type = oe ? oe->type :
2299 sha1_object_info(sha1, NULL);
2300 if (type < 0)
2301 die("%s not found: %s",
2302 S_ISDIR(mode) ? "Tree" : "Blob",
2303 command_buf.buf);
2304 if (type != expected)
2305 die("Not a %s (actually a %s): %s",
2306 typename(expected), typename(type),
2307 command_buf.buf);
2308 }
2310 if (!*p) {
2311 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2312 return;
2313 }
2314 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2315 }
2317 static void file_change_d(struct branch *b)
2318 {
2319 const char *p = command_buf.buf + 2;
2320 static struct strbuf uq = STRBUF_INIT;
2321 const char *endp;
2323 strbuf_reset(&uq);
2324 if (!unquote_c_style(&uq, p, &endp)) {
2325 if (*endp)
2326 die("Garbage after path in: %s", command_buf.buf);
2327 p = uq.buf;
2328 }
2329 tree_content_remove(&b->branch_tree, p, NULL);
2330 }
2332 static void file_change_cr(struct branch *b, int rename)
2333 {
2334 const char *s, *d;
2335 static struct strbuf s_uq = STRBUF_INIT;
2336 static struct strbuf d_uq = STRBUF_INIT;
2337 const char *endp;
2338 struct tree_entry leaf;
2340 s = command_buf.buf + 2;
2341 strbuf_reset(&s_uq);
2342 if (!unquote_c_style(&s_uq, s, &endp)) {
2343 if (*endp != ' ')
2344 die("Missing space after source: %s", command_buf.buf);
2345 } else {
2346 endp = strchr(s, ' ');
2347 if (!endp)
2348 die("Missing space after source: %s", command_buf.buf);
2349 strbuf_add(&s_uq, s, endp - s);
2350 }
2351 s = s_uq.buf;
2353 endp++;
2354 if (!*endp)
2355 die("Missing dest: %s", command_buf.buf);
2357 d = endp;
2358 strbuf_reset(&d_uq);
2359 if (!unquote_c_style(&d_uq, d, &endp)) {
2360 if (*endp)
2361 die("Garbage after dest in: %s", command_buf.buf);
2362 d = d_uq.buf;
2363 }
2365 memset(&leaf, 0, sizeof(leaf));
2366 if (rename)
2367 tree_content_remove(&b->branch_tree, s, &leaf);
2368 else
2369 tree_content_get(&b->branch_tree, s, &leaf);
2370 if (!leaf.versions[1].mode)
2371 die("Path %s not in branch", s);
2372 if (!*d) { /* C "path/to/subdir" "" */
2373 tree_content_replace(&b->branch_tree,
2374 leaf.versions[1].sha1,
2375 leaf.versions[1].mode,
2376 leaf.tree);
2377 return;
2378 }
2379 tree_content_set(&b->branch_tree, d,
2380 leaf.versions[1].sha1,
2381 leaf.versions[1].mode,
2382 leaf.tree);
2383 }
2385 static void note_change_n(struct branch *b, unsigned char old_fanout)
2386 {
2387 const char *p = command_buf.buf + 2;
2388 static struct strbuf uq = STRBUF_INIT;
2389 struct object_entry *oe = oe;
2390 struct branch *s;
2391 unsigned char sha1[20], commit_sha1[20];
2392 char path[60];
2393 uint16_t inline_data = 0;
2394 unsigned char new_fanout;
2396 /* <dataref> or 'inline' */
2397 if (*p == ':') {
2398 char *x;
2399 oe = find_mark(strtoumax(p + 1, &x, 10));
2400 hashcpy(sha1, oe->idx.sha1);
2401 p = x;
2402 } else if (!prefixcmp(p, "inline")) {
2403 inline_data = 1;
2404 p += 6;
2405 } else {
2406 if (get_sha1_hex(p, sha1))
2407 die("Invalid SHA1: %s", command_buf.buf);
2408 oe = find_object(sha1);
2409 p += 40;
2410 }
2411 if (*p++ != ' ')
2412 die("Missing space after SHA1: %s", command_buf.buf);
2414 /* <committish> */
2415 s = lookup_branch(p);
2416 if (s) {
2417 hashcpy(commit_sha1, s->sha1);
2418 } else if (*p == ':') {
2419 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2420 struct object_entry *commit_oe = find_mark(commit_mark);
2421 if (commit_oe->type != OBJ_COMMIT)
2422 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2423 hashcpy(commit_sha1, commit_oe->idx.sha1);
2424 } else if (!get_sha1(p, commit_sha1)) {
2425 unsigned long size;
2426 char *buf = read_object_with_reference(commit_sha1,
2427 commit_type, &size, commit_sha1);
2428 if (!buf || size < 46)
2429 die("Not a valid commit: %s", p);
2430 free(buf);
2431 } else
2432 die("Invalid ref name or SHA1 expression: %s", p);
2434 if (inline_data) {
2435 if (p != uq.buf) {
2436 strbuf_addstr(&uq, p);
2437 p = uq.buf;
2438 }
2439 read_next_command();
2440 parse_and_store_blob(&last_blob, sha1, 0);
2441 } else if (oe) {
2442 if (oe->type != OBJ_BLOB)
2443 die("Not a blob (actually a %s): %s",
2444 typename(oe->type), command_buf.buf);
2445 } else if (!is_null_sha1(sha1)) {
2446 enum object_type type = sha1_object_info(sha1, NULL);
2447 if (type < 0)
2448 die("Blob not found: %s", command_buf.buf);
2449 if (type != OBJ_BLOB)
2450 die("Not a blob (actually a %s): %s",
2451 typename(type), command_buf.buf);
2452 }
2454 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2455 if (tree_content_remove(&b->branch_tree, path, NULL))
2456 b->num_notes--;
2458 if (is_null_sha1(sha1))
2459 return; /* nothing to insert */
2461 b->num_notes++;
2462 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2463 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2464 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2465 }
2467 static void file_change_deleteall(struct branch *b)
2468 {
2469 release_tree_content_recursive(b->branch_tree.tree);
2470 hashclr(b->branch_tree.versions[0].sha1);
2471 hashclr(b->branch_tree.versions[1].sha1);
2472 load_tree(&b->branch_tree);
2473 b->num_notes = 0;
2474 }
2476 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2477 {
2478 if (!buf || size < 46)
2479 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2480 if (memcmp("tree ", buf, 5)
2481 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2482 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2483 hashcpy(b->branch_tree.versions[0].sha1,
2484 b->branch_tree.versions[1].sha1);
2485 }
2487 static void parse_from_existing(struct branch *b)
2488 {
2489 if (is_null_sha1(b->sha1)) {
2490 hashclr(b->branch_tree.versions[0].sha1);
2491 hashclr(b->branch_tree.versions[1].sha1);
2492 } else {
2493 unsigned long size;
2494 char *buf;
2496 buf = read_object_with_reference(b->sha1,
2497 commit_type, &size, b->sha1);
2498 parse_from_commit(b, buf, size);
2499 free(buf);
2500 }
2501 }
2503 static int parse_from(struct branch *b)
2504 {
2505 const char *from;
2506 struct branch *s;
2508 if (prefixcmp(command_buf.buf, "from "))
2509 return 0;
2511 if (b->branch_tree.tree) {
2512 release_tree_content_recursive(b->branch_tree.tree);
2513 b->branch_tree.tree = NULL;
2514 }
2516 from = strchr(command_buf.buf, ' ') + 1;
2517 s = lookup_branch(from);
2518 if (b == s)
2519 die("Can't create a branch from itself: %s", b->name);
2520 else if (s) {
2521 unsigned char *t = s->branch_tree.versions[1].sha1;
2522 hashcpy(b->sha1, s->sha1);
2523 hashcpy(b->branch_tree.versions[0].sha1, t);
2524 hashcpy(b->branch_tree.versions[1].sha1, t);
2525 } else if (*from == ':') {
2526 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2527 struct object_entry *oe = find_mark(idnum);
2528 if (oe->type != OBJ_COMMIT)
2529 die("Mark :%" PRIuMAX " not a commit", idnum);
2530 hashcpy(b->sha1, oe->idx.sha1);
2531 if (oe->pack_id != MAX_PACK_ID) {
2532 unsigned long size;
2533 char *buf = gfi_unpack_entry(oe, &size);
2534 parse_from_commit(b, buf, size);
2535 free(buf);
2536 } else
2537 parse_from_existing(b);
2538 } else if (!get_sha1(from, b->sha1))
2539 parse_from_existing(b);
2540 else
2541 die("Invalid ref name or SHA1 expression: %s", from);
2543 read_next_command();
2544 return 1;
2545 }
2547 static struct hash_list *parse_merge(unsigned int *count)
2548 {
2549 struct hash_list *list = NULL, *n, *e = e;
2550 const char *from;
2551 struct branch *s;
2553 *count = 0;
2554 while (!prefixcmp(command_buf.buf, "merge ")) {
2555 from = strchr(command_buf.buf, ' ') + 1;
2556 n = xmalloc(sizeof(*n));
2557 s = lookup_branch(from);
2558 if (s)
2559 hashcpy(n->sha1, s->sha1);
2560 else if (*from == ':') {
2561 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2562 struct object_entry *oe = find_mark(idnum);
2563 if (oe->type != OBJ_COMMIT)
2564 die("Mark :%" PRIuMAX " not a commit", idnum);
2565 hashcpy(n->sha1, oe->idx.sha1);
2566 } else if (!get_sha1(from, n->sha1)) {
2567 unsigned long size;
2568 char *buf = read_object_with_reference(n->sha1,
2569 commit_type, &size, n->sha1);
2570 if (!buf || size < 46)
2571 die("Not a valid commit: %s", from);
2572 free(buf);
2573 } else
2574 die("Invalid ref name or SHA1 expression: %s", from);
2576 n->next = NULL;
2577 if (list)
2578 e->next = n;
2579 else
2580 list = n;
2581 e = n;
2582 (*count)++;
2583 read_next_command();
2584 }
2585 return list;
2586 }
2588 static void parse_new_commit(void)
2589 {
2590 static struct strbuf msg = STRBUF_INIT;
2591 struct branch *b;
2592 char *sp;
2593 char *author = NULL;
2594 char *committer = NULL;
2595 struct hash_list *merge_list = NULL;
2596 unsigned int merge_count;
2597 unsigned char prev_fanout, new_fanout;
2599 /* Obtain the branch name from the rest of our command */
2600 sp = strchr(command_buf.buf, ' ') + 1;
2601 b = lookup_branch(sp);
2602 if (!b)
2603 b = new_branch(sp);
2605 read_next_command();
2606 parse_mark();
2607 if (!prefixcmp(command_buf.buf, "author ")) {
2608 author = parse_ident(command_buf.buf + 7);
2609 read_next_command();
2610 }
2611 if (!prefixcmp(command_buf.buf, "committer ")) {
2612 committer = parse_ident(command_buf.buf + 10);
2613 read_next_command();
2614 }
2615 if (!committer)
2616 die("Expected committer but didn't get one");
2617 parse_data(&msg, 0, NULL);
2618 read_next_command();
2619 parse_from(b);
2620 merge_list = parse_merge(&merge_count);
2622 /* ensure the branch is active/loaded */
2623 if (!b->branch_tree.tree || !max_active_branches) {
2624 unload_one_branch();
2625 load_branch(b);
2626 }
2628 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2630 /* file_change* */
2631 while (command_buf.len > 0) {
2632 if (!prefixcmp(command_buf.buf, "M "))
2633 file_change_m(b);
2634 else if (!prefixcmp(command_buf.buf, "D "))
2635 file_change_d(b);
2636 else if (!prefixcmp(command_buf.buf, "R "))
2637 file_change_cr(b, 1);
2638 else if (!prefixcmp(command_buf.buf, "C "))
2639 file_change_cr(b, 0);
2640 else if (!prefixcmp(command_buf.buf, "N "))
2641 note_change_n(b, prev_fanout);
2642 else if (!strcmp("deleteall", command_buf.buf))
2643 file_change_deleteall(b);
2644 else if (!prefixcmp(command_buf.buf, "ls "))
2645 parse_ls(b);
2646 else {
2647 unread_command_buf = 1;
2648 break;
2649 }
2650 if (read_next_command() == EOF)
2651 break;
2652 }
2654 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2655 if (new_fanout != prev_fanout)
2656 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2658 /* build the tree and the commit */
2659 store_tree(&b->branch_tree);
2660 hashcpy(b->branch_tree.versions[0].sha1,
2661 b->branch_tree.versions[1].sha1);
2663 strbuf_reset(&new_data);
2664 strbuf_addf(&new_data, "tree %s\n",
2665 sha1_to_hex(b->branch_tree.versions[1].sha1));
2666 if (!is_null_sha1(b->sha1))
2667 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2668 while (merge_list) {
2669 struct hash_list *next = merge_list->next;
2670 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2671 free(merge_list);
2672 merge_list = next;
2673 }
2674 strbuf_addf(&new_data,
2675 "author %s\n"
2676 "committer %s\n"
2677 "\n",
2678 author ? author : committer, committer);
2679 strbuf_addbuf(&new_data, &msg);
2680 free(author);
2681 free(committer);
2683 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2684 b->pack_id = pack_id;
2685 b->last_commit = object_count_by_type[OBJ_COMMIT];
2686 }
2688 static void parse_new_tag(void)
2689 {
2690 static struct strbuf msg = STRBUF_INIT;
2691 char *sp;
2692 const char *from;
2693 char *tagger;
2694 struct branch *s;
2695 struct tag *t;
2696 uintmax_t from_mark = 0;
2697 unsigned char sha1[20];
2698 enum object_type type;
2700 /* Obtain the new tag name from the rest of our command */
2701 sp = strchr(command_buf.buf, ' ') + 1;
2702 t = pool_alloc(sizeof(struct tag));
2703 t->next_tag = NULL;
2704 t->name = pool_strdup(sp);
2705 if (last_tag)
2706 last_tag->next_tag = t;
2707 else
2708 first_tag = t;
2709 last_tag = t;
2710 read_next_command();
2712 /* from ... */
2713 if (prefixcmp(command_buf.buf, "from "))
2714 die("Expected from command, got %s", command_buf.buf);
2715 from = strchr(command_buf.buf, ' ') + 1;
2716 s = lookup_branch(from);
2717 if (s) {
2718 hashcpy(sha1, s->sha1);
2719 type = OBJ_COMMIT;
2720 } else if (*from == ':') {
2721 struct object_entry *oe;
2722 from_mark = strtoumax(from + 1, NULL, 10);
2723 oe = find_mark(from_mark);
2724 type = oe->type;
2725 hashcpy(sha1, oe->idx.sha1);
2726 } else if (!get_sha1(from, sha1)) {
2727 unsigned long size;
2728 char *buf;
2730 buf = read_sha1_file(sha1, &type, &size);
2731 if (!buf || size < 46)
2732 die("Not a valid commit: %s", from);
2733 free(buf);
2734 } else
2735 die("Invalid ref name or SHA1 expression: %s", from);
2736 read_next_command();
2738 /* tagger ... */
2739 if (!prefixcmp(command_buf.buf, "tagger ")) {
2740 tagger = parse_ident(command_buf.buf + 7);
2741 read_next_command();
2742 } else
2743 tagger = NULL;
2745 /* tag payload/message */
2746 parse_data(&msg, 0, NULL);
2748 /* build the tag object */
2749 strbuf_reset(&new_data);
2751 strbuf_addf(&new_data,
2752 "object %s\n"
2753 "type %s\n"
2754 "tag %s\n",
2755 sha1_to_hex(sha1), typename(type), t->name);
2756 if (tagger)
2757 strbuf_addf(&new_data,
2758 "tagger %s\n", tagger);
2759 strbuf_addch(&new_data, '\n');
2760 strbuf_addbuf(&new_data, &msg);
2761 free(tagger);
2763 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2764 t->pack_id = MAX_PACK_ID;
2765 else
2766 t->pack_id = pack_id;
2767 }
2769 static void parse_reset_branch(void)
2770 {
2771 struct branch *b;
2772 char *sp;
2774 /* Obtain the branch name from the rest of our command */
2775 sp = strchr(command_buf.buf, ' ') + 1;
2776 b = lookup_branch(sp);
2777 if (b) {
2778 hashclr(b->sha1);
2779 hashclr(b->branch_tree.versions[0].sha1);
2780 hashclr(b->branch_tree.versions[1].sha1);
2781 if (b->branch_tree.tree) {
2782 release_tree_content_recursive(b->branch_tree.tree);
2783 b->branch_tree.tree = NULL;
2784 }
2785 }
2786 else
2787 b = new_branch(sp);
2788 read_next_command();
2789 parse_from(b);
2790 if (command_buf.len > 0)
2791 unread_command_buf = 1;
2792 }
2794 static void cat_blob_write(const char *buf, unsigned long size)
2795 {
2796 if (write_in_full(cat_blob_fd, buf, size) != size)
2797 die_errno("Write to frontend failed");
2798 }
2800 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2801 {
2802 struct strbuf line = STRBUF_INIT;
2803 unsigned long size;
2804 enum object_type type = 0;
2805 char *buf;
2807 if (!oe || oe->pack_id == MAX_PACK_ID) {
2808 buf = read_sha1_file(sha1, &type, &size);
2809 } else {
2810 type = oe->type;
2811 buf = gfi_unpack_entry(oe, &size);
2812 }
2814 /*
2815 * Output based on batch_one_object() from cat-file.c.
2816 */
2817 if (type <= 0) {
2818 strbuf_reset(&line);
2819 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2820 cat_blob_write(line.buf, line.len);
2821 strbuf_release(&line);
2822 free(buf);
2823 return;
2824 }
2825 if (!buf)
2826 die("Can't read object %s", sha1_to_hex(sha1));
2827 if (type != OBJ_BLOB)
2828 die("Object %s is a %s but a blob was expected.",
2829 sha1_to_hex(sha1), typename(type));
2830 strbuf_reset(&line);
2831 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2832 typename(type), size);
2833 cat_blob_write(line.buf, line.len);
2834 strbuf_release(&line);
2835 cat_blob_write(buf, size);
2836 cat_blob_write("\n", 1);
2837 free(buf);
2838 }
2840 static void parse_cat_blob(void)
2841 {
2842 const char *p;
2843 struct object_entry *oe = oe;
2844 unsigned char sha1[20];
2846 /* cat-blob SP <object> LF */
2847 p = command_buf.buf + strlen("cat-blob ");
2848 if (*p == ':') {
2849 char *x;
2850 oe = find_mark(strtoumax(p + 1, &x, 10));
2851 if (x == p + 1)
2852 die("Invalid mark: %s", command_buf.buf);
2853 if (!oe)
2854 die("Unknown mark: %s", command_buf.buf);
2855 if (*x)
2856 die("Garbage after mark: %s", command_buf.buf);
2857 hashcpy(sha1, oe->idx.sha1);
2858 } else {
2859 if (get_sha1_hex(p, sha1))
2860 die("Invalid SHA1: %s", command_buf.buf);
2861 if (p[40])
2862 die("Garbage after SHA1: %s", command_buf.buf);
2863 oe = find_object(sha1);
2864 }
2866 cat_blob(oe, sha1);
2867 }
2869 static struct object_entry *dereference(struct object_entry *oe,
2870 unsigned char sha1[20])
2871 {
2872 unsigned long size;
2873 char *buf = NULL;
2874 if (!oe) {
2875 enum object_type type = sha1_object_info(sha1, NULL);
2876 if (type < 0)
2877 die("object not found: %s", sha1_to_hex(sha1));
2878 /* cache it! */
2879 oe = insert_object(sha1);
2880 oe->type = type;
2881 oe->pack_id = MAX_PACK_ID;
2882 oe->idx.offset = 1;
2883 }
2884 switch (oe->type) {
2885 case OBJ_TREE: /* easy case. */
2886 return oe;
2887 case OBJ_COMMIT:
2888 case OBJ_TAG:
2889 break;
2890 default:
2891 die("Not a treeish: %s", command_buf.buf);
2892 }
2894 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2895 buf = gfi_unpack_entry(oe, &size);
2896 } else {
2897 enum object_type unused;
2898 buf = read_sha1_file(sha1, &unused, &size);
2899 }
2900 if (!buf)
2901 die("Can't load object %s", sha1_to_hex(sha1));
2903 /* Peel one layer. */
2904 switch (oe->type) {
2905 case OBJ_TAG:
2906 if (size < 40 + strlen("object ") ||
2907 get_sha1_hex(buf + strlen("object "), sha1))
2908 die("Invalid SHA1 in tag: %s", command_buf.buf);
2909 break;
2910 case OBJ_COMMIT:
2911 if (size < 40 + strlen("tree ") ||
2912 get_sha1_hex(buf + strlen("tree "), sha1))
2913 die("Invalid SHA1 in commit: %s", command_buf.buf);
2914 }
2916 free(buf);
2917 return find_object(sha1);
2918 }
2920 static struct object_entry *parse_treeish_dataref(const char **p)
2921 {
2922 unsigned char sha1[20];
2923 struct object_entry *e;
2925 if (**p == ':') { /* <mark> */
2926 char *endptr;
2927 e = find_mark(strtoumax(*p + 1, &endptr, 10));
2928 if (endptr == *p + 1)
2929 die("Invalid mark: %s", command_buf.buf);
2930 if (!e)
2931 die("Unknown mark: %s", command_buf.buf);
2932 *p = endptr;
2933 hashcpy(sha1, e->idx.sha1);
2934 } else { /* <sha1> */
2935 if (get_sha1_hex(*p, sha1))
2936 die("Invalid SHA1: %s", command_buf.buf);
2937 e = find_object(sha1);
2938 *p += 40;
2939 }
2941 while (!e || e->type != OBJ_TREE)
2942 e = dereference(e, sha1);
2943 return e;
2944 }
2946 static void print_ls(int mode, const unsigned char *sha1, const char *path)
2947 {
2948 static struct strbuf line = STRBUF_INIT;
2950 /* See show_tree(). */
2951 const char *type =
2952 S_ISGITLINK(mode) ? commit_type :
2953 S_ISDIR(mode) ? tree_type :
2954 blob_type;
2956 if (!mode) {
2957 /* missing SP path LF */
2958 strbuf_reset(&line);
2959 strbuf_addstr(&line, "missing ");
2960 quote_c_style(path, &line, NULL, 0);
2961 strbuf_addch(&line, '\n');
2962 } else {
2963 /* mode SP type SP object_name TAB path LF */
2964 strbuf_reset(&line);
2965 strbuf_addf(&line, "%06o %s %s\t",
2966 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
2967 quote_c_style(path, &line, NULL, 0);
2968 strbuf_addch(&line, '\n');
2969 }
2970 cat_blob_write(line.buf, line.len);
2971 }
2973 static void parse_ls(struct branch *b)
2974 {
2975 const char *p;
2976 struct tree_entry *root = NULL;
2977 struct tree_entry leaf = {NULL};
2979 /* ls SP (<treeish> SP)? <path> */
2980 p = command_buf.buf + strlen("ls ");
2981 if (*p == '"') {
2982 if (!b)
2983 die("Not in a commit: %s", command_buf.buf);
2984 root = &b->branch_tree;
2985 } else {
2986 struct object_entry *e = parse_treeish_dataref(&p);
2987 root = new_tree_entry();
2988 hashcpy(root->versions[1].sha1, e->idx.sha1);
2989 load_tree(root);
2990 if (*p++ != ' ')
2991 die("Missing space after tree-ish: %s", command_buf.buf);
2992 }
2993 if (*p == '"') {
2994 static struct strbuf uq = STRBUF_INIT;
2995 const char *endp;
2996 strbuf_reset(&uq);
2997 if (unquote_c_style(&uq, p, &endp))
2998 die("Invalid path: %s", command_buf.buf);
2999 if (*endp)
3000 die("Garbage after path in: %s", command_buf.buf);
3001 p = uq.buf;
3002 }
3003 tree_content_get(root, p, &leaf);
3004 /*
3005 * A directory in preparation would have a sha1 of zero
3006 * until it is saved. Save, for simplicity.
3007 */
3008 if (S_ISDIR(leaf.versions[1].mode))
3009 store_tree(&leaf);
3011 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3012 if (!b || root != &b->branch_tree)
3013 release_tree_entry(root);
3014 }
3016 static void checkpoint(void)
3017 {
3018 checkpoint_requested = 0;
3019 if (object_count) {
3020 cycle_packfile();
3021 dump_branches();
3022 dump_tags();
3023 dump_marks();
3024 }
3025 }
3027 static void parse_checkpoint(void)
3028 {
3029 checkpoint_requested = 1;
3030 skip_optional_lf();
3031 }
3033 static void parse_progress(void)
3034 {
3035 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3036 fputc('\n', stdout);
3037 fflush(stdout);
3038 skip_optional_lf();
3039 }
3041 static char* make_fast_import_path(const char *path)
3042 {
3043 struct strbuf abs_path = STRBUF_INIT;
3045 if (!relative_marks_paths || is_absolute_path(path))
3046 return xstrdup(path);
3047 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
3048 return strbuf_detach(&abs_path, NULL);
3049 }
3051 static void option_import_marks(const char *marks,
3052 int from_stream, int ignore_missing)
3053 {
3054 if (import_marks_file) {
3055 if (from_stream)
3056 die("Only one import-marks command allowed per stream");
3058 /* read previous mark file */
3059 if(!import_marks_file_from_stream)
3060 read_marks();
3061 }
3063 import_marks_file = make_fast_import_path(marks);
3064 safe_create_leading_directories_const(import_marks_file);
3065 import_marks_file_from_stream = from_stream;
3066 import_marks_file_ignore_missing = ignore_missing;
3067 }
3069 static void option_date_format(const char *fmt)
3070 {
3071 if (!strcmp(fmt, "raw"))
3072 whenspec = WHENSPEC_RAW;
3073 else if (!strcmp(fmt, "rfc2822"))
3074 whenspec = WHENSPEC_RFC2822;
3075 else if (!strcmp(fmt, "now"))
3076 whenspec = WHENSPEC_NOW;
3077 else
3078 die("unknown --date-format argument %s", fmt);
3079 }
3081 static unsigned long ulong_arg(const char *option, const char *arg)
3082 {
3083 char *endptr;
3084 unsigned long rv = strtoul(arg, &endptr, 0);
3085 if (strchr(arg, '-') || endptr == arg || *endptr)
3086 die("%s: argument must be a non-negative integer", option);
3087 return rv;
3088 }
3090 static void option_depth(const char *depth)
3091 {
3092 max_depth = ulong_arg("--depth", depth);
3093 if (max_depth > MAX_DEPTH)
3094 die("--depth cannot exceed %u", MAX_DEPTH);
3095 }
3097 static void option_active_branches(const char *branches)
3098 {
3099 max_active_branches = ulong_arg("--active-branches", branches);
3100 }
3102 static void option_export_marks(const char *marks)
3103 {
3104 export_marks_file = make_fast_import_path(marks);
3105 safe_create_leading_directories_const(export_marks_file);
3106 }
3108 static void option_cat_blob_fd(const char *fd)
3109 {
3110 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3111 if (n > (unsigned long) INT_MAX)
3112 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3113 cat_blob_fd = (int) n;
3114 }
3116 static void option_export_pack_edges(const char *edges)
3117 {
3118 if (pack_edges)
3119 fclose(pack_edges);
3120 pack_edges = fopen(edges, "a");
3121 if (!pack_edges)
3122 die_errno("Cannot open '%s'", edges);
3123 }
3125 static int parse_one_option(const char *option)
3126 {
3127 if (!prefixcmp(option, "max-pack-size=")) {
3128 unsigned long v;
3129 if (!git_parse_ulong(option + 14, &v))
3130 return 0;
3131 if (v < 8192) {
3132 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3133 v *= 1024 * 1024;
3134 } else if (v < 1024 * 1024) {
3135 warning("minimum max-pack-size is 1 MiB");
3136 v = 1024 * 1024;
3137 }
3138 max_packsize = v;
3139 } else if (!prefixcmp(option, "big-file-threshold=")) {
3140 unsigned long v;
3141 if (!git_parse_ulong(option + 19, &v))
3142 return 0;
3143 big_file_threshold = v;
3144 } else if (!prefixcmp(option, "depth=")) {
3145 option_depth(option + 6);
3146 } else if (!prefixcmp(option, "active-branches=")) {
3147 option_active_branches(option + 16);
3148 } else if (!prefixcmp(option, "export-pack-edges=")) {
3149 option_export_pack_edges(option + 18);
3150 } else if (!prefixcmp(option, "quiet")) {
3151 show_stats = 0;
3152 } else if (!prefixcmp(option, "stats")) {
3153 show_stats = 1;
3154 } else {
3155 return 0;
3156 }
3158 return 1;
3159 }
3161 static int parse_one_feature(const char *feature, int from_stream)
3162 {
3163 if (!prefixcmp(feature, "date-format=")) {
3164 option_date_format(feature + 12);
3165 } else if (!prefixcmp(feature, "import-marks=")) {
3166 option_import_marks(feature + 13, from_stream, 0);
3167 } else if (!prefixcmp(feature, "import-marks-if-exists=")) {
3168 option_import_marks(feature + strlen("import-marks-if-exists="),
3169 from_stream, 1);
3170 } else if (!prefixcmp(feature, "export-marks=")) {
3171 option_export_marks(feature + 13);
3172 } else if (!strcmp(feature, "cat-blob")) {
3173 ; /* Don't die - this feature is supported */
3174 } else if (!strcmp(feature, "relative-marks")) {
3175 relative_marks_paths = 1;
3176 } else if (!strcmp(feature, "no-relative-marks")) {
3177 relative_marks_paths = 0;
3178 } else if (!strcmp(feature, "done")) {
3179 require_explicit_termination = 1;
3180 } else if (!strcmp(feature, "force")) {
3181 force_update = 1;
3182 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3183 ; /* do nothing; we have the feature */
3184 } else {
3185 return 0;
3186 }
3188 return 1;
3189 }
3191 static void parse_feature(void)
3192 {
3193 char *feature = command_buf.buf + 8;
3195 if (seen_data_command)
3196 die("Got feature command '%s' after data command", feature);
3198 if (parse_one_feature(feature, 1))
3199 return;
3201 die("This version of fast-import does not support feature %s.", feature);
3202 }
3204 static void parse_option(void)
3205 {
3206 char *option = command_buf.buf + 11;
3208 if (seen_data_command)
3209 die("Got option command '%s' after data command", option);
3211 if (parse_one_option(option))
3212 return;
3214 die("This version of fast-import does not support option: %s", option);
3215 }
3217 static int git_pack_config(const char *k, const char *v, void *cb)
3218 {
3219 if (!strcmp(k, "pack.depth")) {
3220 max_depth = git_config_int(k, v);
3221 if (max_depth > MAX_DEPTH)
3222 max_depth = MAX_DEPTH;
3223 return 0;
3224 }
3225 if (!strcmp(k, "pack.compression")) {
3226 int level = git_config_int(k, v);
3227 if (level == -1)
3228 level = Z_DEFAULT_COMPRESSION;
3229 else if (level < 0 || level > Z_BEST_COMPRESSION)
3230 die("bad pack compression level %d", level);
3231 pack_compression_level = level;
3232 pack_compression_seen = 1;
3233 return 0;
3234 }
3235 if (!strcmp(k, "pack.indexversion")) {
3236 pack_idx_opts.version = git_config_int(k, v);
3237 if (pack_idx_opts.version > 2)
3238 die("bad pack.indexversion=%"PRIu32,
3239 pack_idx_opts.version);
3240 return 0;
3241 }
3242 if (!strcmp(k, "pack.packsizelimit")) {
3243 max_packsize = git_config_ulong(k, v);
3244 return 0;
3245 }
3246 return git_default_config(k, v, cb);
3247 }
3249 static const char fast_import_usage[] =
3250 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3252 static void parse_argv(void)
3253 {
3254 unsigned int i;
3256 for (i = 1; i < global_argc; i++) {
3257 const char *a = global_argv[i];
3259 if (*a != '-' || !strcmp(a, "--"))
3260 break;
3262 if (parse_one_option(a + 2))
3263 continue;
3265 if (parse_one_feature(a + 2, 0))
3266 continue;
3268 if (!prefixcmp(a + 2, "cat-blob-fd=")) {
3269 option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
3270 continue;
3271 }
3273 die("unknown option %s", a);
3274 }
3275 if (i != global_argc)
3276 usage(fast_import_usage);
3278 seen_data_command = 1;
3279 if (import_marks_file)
3280 read_marks();
3281 }
3283 int main(int argc, const char **argv)
3284 {
3285 unsigned int i;
3287 git_extract_argv0_path(argv[0]);
3289 if (argc == 2 && !strcmp(argv[1], "-h"))
3290 usage(fast_import_usage);
3292 setup_git_directory();
3293 reset_pack_idx_option(&pack_idx_opts);
3294 git_config(git_pack_config, NULL);
3295 if (!pack_compression_seen && core_compression_seen)
3296 pack_compression_level = core_compression_level;
3298 alloc_objects(object_entry_alloc);
3299 strbuf_init(&command_buf, 0);
3300 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3301 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3302 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3303 marks = pool_calloc(1, sizeof(struct mark_set));
3305 global_argc = argc;
3306 global_argv = argv;
3308 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3309 for (i = 0; i < (cmd_save - 1); i++)
3310 rc_free[i].next = &rc_free[i + 1];
3311 rc_free[cmd_save - 1].next = NULL;
3313 prepare_packed_git();
3314 start_packfile();
3315 set_die_routine(die_nicely);
3316 set_checkpoint_signal();
3317 while (read_next_command() != EOF) {
3318 if (!strcmp("blob", command_buf.buf))
3319 parse_new_blob();
3320 else if (!prefixcmp(command_buf.buf, "ls "))
3321 parse_ls(NULL);
3322 else if (!prefixcmp(command_buf.buf, "commit "))
3323 parse_new_commit();
3324 else if (!prefixcmp(command_buf.buf, "tag "))
3325 parse_new_tag();
3326 else if (!prefixcmp(command_buf.buf, "reset "))
3327 parse_reset_branch();
3328 else if (!strcmp("checkpoint", command_buf.buf))
3329 parse_checkpoint();
3330 else if (!strcmp("done", command_buf.buf))
3331 break;
3332 else if (!prefixcmp(command_buf.buf, "progress "))
3333 parse_progress();
3334 else if (!prefixcmp(command_buf.buf, "feature "))
3335 parse_feature();
3336 else if (!prefixcmp(command_buf.buf, "option git "))
3337 parse_option();
3338 else if (!prefixcmp(command_buf.buf, "option "))
3339 /* ignore non-git options*/;
3340 else
3341 die("Unsupported command: %s", command_buf.buf);
3343 if (checkpoint_requested)
3344 checkpoint();
3345 }
3347 /* argv hasn't been parsed yet, do so */
3348 if (!seen_data_command)
3349 parse_argv();
3351 if (require_explicit_termination && feof(stdin))
3352 die("stream ends early");
3354 end_packfile();
3356 dump_branches();
3357 dump_tags();
3358 unkeep_all_packs();
3359 dump_marks();
3361 if (pack_edges)
3362 fclose(pack_edges);
3364 if (show_stats) {
3365 uintmax_t total_count = 0, duplicate_count = 0;
3366 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3367 total_count += object_count_by_type[i];
3368 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3369 duplicate_count += duplicate_count_by_type[i];
3371 fprintf(stderr, "%s statistics:\n", argv[0]);
3372 fprintf(stderr, "---------------------------------------------------------------------\n");
3373 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3374 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3375 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
3376 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
3377 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
3378 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
3379 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3380 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3381 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3382 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3383 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3384 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3385 fprintf(stderr, "---------------------------------------------------------------------\n");
3386 pack_report();
3387 fprintf(stderr, "---------------------------------------------------------------------\n");
3388 fprintf(stderr, "\n");
3389 }
3391 return failure ? 1 : 0;
3392 }