coccinelle: add a rule to make "expression" code use FREE_AND_NULL()
[alt-git.git] / fast-import.c
blobeeab927d5ab9f8bca61e68ef3fe29b565a11fdc1
1 /*
2 (See Documentation/git-fast-import.txt for maintained documentation.)
3 Format of STDIN stream:
5 stream ::= cmd*;
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
20 new_commit ::= 'commit' sp ref_str lf
21 mark?
22 ('author' (sp name)? sp '<' email '>' sp when lf)?
23 'committer' (sp name)? sp '<' email '>' sp when lf
24 commit_msg
25 ('from' sp commit-ish lf)?
26 ('merge' sp commit-ish lf)*
27 (file_change | ls)*
28 lf?;
29 commit_msg ::= data;
31 ls ::= 'ls' sp '"' quoted(path) '"' lf;
33 file_change ::= file_clr
34 | file_del
35 | file_rnm
36 | file_cpy
37 | file_obm
38 | file_inm;
39 file_clr ::= 'deleteall' lf;
40 file_del ::= 'D' sp path_str lf;
41 file_rnm ::= 'R' sp path_str sp path_str lf;
42 file_cpy ::= 'C' sp path_str sp path_str lf;
43 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
44 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
45 data;
46 note_obm ::= 'N' sp (hexsha1 | idnum) sp commit-ish lf;
47 note_inm ::= 'N' sp 'inline' sp commit-ish lf
48 data;
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp commit-ish lf
52 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
53 tag_msg;
54 tag_msg ::= data;
56 reset_branch ::= 'reset' sp ref_str lf
57 ('from' sp commit-ish lf)?
58 lf?;
60 checkpoint ::= 'checkpoint' lf
61 lf?;
63 progress ::= 'progress' sp not_lf* lf
64 lf?;
66 # note: the first idnum in a stream should be 1 and subsequent
67 # idnums should not have gaps between values as this will cause
68 # the stream parser to reserve space for the gapped values. An
69 # idnum can be updated in the future to a new object by issuing
70 # a new mark directive with the old idnum.
72 mark ::= 'mark' sp idnum lf;
73 data ::= (delimited_data | exact_data)
74 lf?;
76 # note: delim may be any string but must not contain lf.
77 # data_line may contain any data but must not be exactly
78 # delim.
79 delimited_data ::= 'data' sp '<<' delim lf
80 (data_line lf)*
81 delim lf;
83 # note: declen indicates the length of binary_data in bytes.
84 # declen does not include the lf preceding the binary data.
86 exact_data ::= 'data' sp declen lf
87 binary_data;
89 # note: quoted strings are C-style quoting supporting \c for
90 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
91 # is the signed byte value in octal. Note that the only
92 # characters which must actually be escaped to protect the
93 # stream formatting is: \, " and LF. Otherwise these values
94 # are UTF8.
96 commit-ish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
97 ref_str ::= ref;
98 sha1exp_str ::= sha1exp;
99 tag_str ::= tag;
100 path_str ::= path | '"' quoted(path) '"' ;
101 mode ::= '100644' | '644'
102 | '100755' | '755'
103 | '120000'
106 declen ::= # unsigned 32 bit value, ascii base10 notation;
107 bigint ::= # unsigned integer value, ascii base10 notation;
108 binary_data ::= # file content, not interpreted;
110 when ::= raw_when | rfc2822_when;
111 raw_when ::= ts sp tz;
112 rfc2822_when ::= # Valid RFC 2822 date and time;
114 sp ::= # ASCII space character;
115 lf ::= # ASCII newline (LF) character;
117 # note: a colon (':') must precede the numerical value assigned to
118 # an idnum. This is to distinguish it from a ref or tag name as
119 # GIT does not permit ':' in ref or tag strings.
121 idnum ::= ':' bigint;
122 path ::= # GIT style file path, e.g. "a/b/c";
123 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
124 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
125 sha1exp ::= # Any valid GIT SHA1 expression;
126 hexsha1 ::= # SHA1 in hexadecimal format;
128 # note: name and email are UTF8 strings, however name must not
129 # contain '<' or lf and email must not contain any of the
130 # following: '<', '>', lf.
132 name ::= # valid GIT author/committer name;
133 email ::= # valid GIT author/committer email;
134 ts ::= # time since the epoch in seconds, ascii base10 notation;
135 tz ::= # GIT style timezone;
137 # note: comments, get-mark, ls-tree, and cat-blob requests may
138 # appear anywhere in the input, except within a data command. Any
139 # form of the data command always escapes the related input from
140 # comment processing.
142 # In case it is not clear, the '#' that starts the comment
143 # must be the first character on that line (an lf
144 # preceded it).
147 get_mark ::= 'get-mark' sp idnum lf;
148 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
149 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
151 comment ::= '#' not_lf* lf;
152 not_lf ::= # Any byte that is not ASCII newline (LF);
155 #include "builtin.h"
156 #include "cache.h"
157 #include "lockfile.h"
158 #include "object.h"
159 #include "blob.h"
160 #include "tree.h"
161 #include "commit.h"
162 #include "delta.h"
163 #include "pack.h"
164 #include "refs.h"
165 #include "csum-file.h"
166 #include "quote.h"
167 #include "dir.h"
168 #include "run-command.h"
170 #define PACK_ID_BITS 16
171 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
172 #define DEPTH_BITS 13
173 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
176 * We abuse the setuid bit on directories to mean "do not delta".
178 #define NO_DELTA S_ISUID
180 struct object_entry {
181 struct pack_idx_entry idx;
182 struct object_entry *next;
183 uint32_t type : TYPE_BITS,
184 pack_id : PACK_ID_BITS,
185 depth : DEPTH_BITS;
188 struct object_entry_pool {
189 struct object_entry_pool *next_pool;
190 struct object_entry *next_free;
191 struct object_entry *end;
192 struct object_entry entries[FLEX_ARRAY]; /* more */
195 struct mark_set {
196 union {
197 struct object_entry *marked[1024];
198 struct mark_set *sets[1024];
199 } data;
200 unsigned int shift;
203 struct last_object {
204 struct strbuf data;
205 off_t offset;
206 unsigned int depth;
207 unsigned no_swap : 1;
210 struct mem_pool {
211 struct mem_pool *next_pool;
212 char *next_free;
213 char *end;
214 uintmax_t space[FLEX_ARRAY]; /* more */
217 struct atom_str {
218 struct atom_str *next_atom;
219 unsigned short str_len;
220 char str_dat[FLEX_ARRAY]; /* more */
223 struct tree_content;
224 struct tree_entry {
225 struct tree_content *tree;
226 struct atom_str *name;
227 struct tree_entry_ms {
228 uint16_t mode;
229 struct object_id oid;
230 } versions[2];
233 struct tree_content {
234 unsigned int entry_capacity; /* must match avail_tree_content */
235 unsigned int entry_count;
236 unsigned int delta_depth;
237 struct tree_entry *entries[FLEX_ARRAY]; /* more */
240 struct avail_tree_content {
241 unsigned int entry_capacity; /* must match tree_content */
242 struct avail_tree_content *next_avail;
245 struct branch {
246 struct branch *table_next_branch;
247 struct branch *active_next_branch;
248 const char *name;
249 struct tree_entry branch_tree;
250 uintmax_t last_commit;
251 uintmax_t num_notes;
252 unsigned active : 1;
253 unsigned delete : 1;
254 unsigned pack_id : PACK_ID_BITS;
255 struct object_id oid;
258 struct tag {
259 struct tag *next_tag;
260 const char *name;
261 unsigned int pack_id;
262 struct object_id oid;
265 struct hash_list {
266 struct hash_list *next;
267 struct object_id oid;
270 typedef enum {
271 WHENSPEC_RAW = 1,
272 WHENSPEC_RFC2822,
273 WHENSPEC_NOW
274 } whenspec_type;
276 struct recent_command {
277 struct recent_command *prev;
278 struct recent_command *next;
279 char *buf;
282 /* Configured limits on output */
283 static unsigned long max_depth = 10;
284 static off_t max_packsize;
285 static int unpack_limit = 100;
286 static int force_update;
288 /* Stats and misc. counters */
289 static uintmax_t alloc_count;
290 static uintmax_t marks_set_count;
291 static uintmax_t object_count_by_type[1 << TYPE_BITS];
292 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
293 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
294 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
295 static unsigned long object_count;
296 static unsigned long branch_count;
297 static unsigned long branch_load_count;
298 static int failure;
299 static FILE *pack_edges;
300 static unsigned int show_stats = 1;
301 static int global_argc;
302 static const char **global_argv;
304 /* Memory pools */
305 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
306 static size_t total_allocd;
307 static struct mem_pool *mem_pool;
309 /* Atom management */
310 static unsigned int atom_table_sz = 4451;
311 static unsigned int atom_cnt;
312 static struct atom_str **atom_table;
314 /* The .pack file being generated */
315 static struct pack_idx_option pack_idx_opts;
316 static unsigned int pack_id;
317 static struct sha1file *pack_file;
318 static struct packed_git *pack_data;
319 static struct packed_git **all_packs;
320 static off_t pack_size;
322 /* Table of objects we've written. */
323 static unsigned int object_entry_alloc = 5000;
324 static struct object_entry_pool *blocks;
325 static struct object_entry *object_table[1 << 16];
326 static struct mark_set *marks;
327 static const char *export_marks_file;
328 static const char *import_marks_file;
329 static int import_marks_file_from_stream;
330 static int import_marks_file_ignore_missing;
331 static int import_marks_file_done;
332 static int relative_marks_paths;
334 /* Our last blob */
335 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
337 /* Tree management */
338 static unsigned int tree_entry_alloc = 1000;
339 static void *avail_tree_entry;
340 static unsigned int avail_tree_table_sz = 100;
341 static struct avail_tree_content **avail_tree_table;
342 static struct strbuf old_tree = STRBUF_INIT;
343 static struct strbuf new_tree = STRBUF_INIT;
345 /* Branch data */
346 static unsigned long max_active_branches = 5;
347 static unsigned long cur_active_branches;
348 static unsigned long branch_table_sz = 1039;
349 static struct branch **branch_table;
350 static struct branch *active_branches;
352 /* Tag data */
353 static struct tag *first_tag;
354 static struct tag *last_tag;
356 /* Input stream parsing */
357 static whenspec_type whenspec = WHENSPEC_RAW;
358 static struct strbuf command_buf = STRBUF_INIT;
359 static int unread_command_buf;
360 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
361 static struct recent_command *cmd_tail = &cmd_hist;
362 static struct recent_command *rc_free;
363 static unsigned int cmd_save = 100;
364 static uintmax_t next_mark;
365 static struct strbuf new_data = STRBUF_INIT;
366 static int seen_data_command;
367 static int require_explicit_termination;
369 /* Signal handling */
370 static volatile sig_atomic_t checkpoint_requested;
372 /* Where to write output of cat-blob commands */
373 static int cat_blob_fd = STDOUT_FILENO;
375 static void parse_argv(void);
376 static void parse_get_mark(const char *p);
377 static void parse_cat_blob(const char *p);
378 static void parse_ls(const char *p, struct branch *b);
380 static void write_branch_report(FILE *rpt, struct branch *b)
382 fprintf(rpt, "%s:\n", b->name);
384 fprintf(rpt, " status :");
385 if (b->active)
386 fputs(" active", rpt);
387 if (b->branch_tree.tree)
388 fputs(" loaded", rpt);
389 if (is_null_oid(&b->branch_tree.versions[1].oid))
390 fputs(" dirty", rpt);
391 fputc('\n', rpt);
393 fprintf(rpt, " tip commit : %s\n", oid_to_hex(&b->oid));
394 fprintf(rpt, " old tree : %s\n",
395 oid_to_hex(&b->branch_tree.versions[0].oid));
396 fprintf(rpt, " cur tree : %s\n",
397 oid_to_hex(&b->branch_tree.versions[1].oid));
398 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
400 fputs(" last pack : ", rpt);
401 if (b->pack_id < MAX_PACK_ID)
402 fprintf(rpt, "%u", b->pack_id);
403 fputc('\n', rpt);
405 fputc('\n', rpt);
408 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
410 static void write_crash_report(const char *err)
412 char *loc = git_pathdup("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
413 FILE *rpt = fopen(loc, "w");
414 struct branch *b;
415 unsigned long lu;
416 struct recent_command *rc;
418 if (!rpt) {
419 error_errno("can't write crash report %s", loc);
420 free(loc);
421 return;
424 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
426 fprintf(rpt, "fast-import crash report:\n");
427 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
428 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
429 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_MODE(ISO8601)));
430 fputc('\n', rpt);
432 fputs("fatal: ", rpt);
433 fputs(err, rpt);
434 fputc('\n', rpt);
436 fputc('\n', rpt);
437 fputs("Most Recent Commands Before Crash\n", rpt);
438 fputs("---------------------------------\n", rpt);
439 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
440 if (rc->next == &cmd_hist)
441 fputs("* ", rpt);
442 else
443 fputs(" ", rpt);
444 fputs(rc->buf, rpt);
445 fputc('\n', rpt);
448 fputc('\n', rpt);
449 fputs("Active Branch LRU\n", rpt);
450 fputs("-----------------\n", rpt);
451 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
452 cur_active_branches,
453 max_active_branches);
454 fputc('\n', rpt);
455 fputs(" pos clock name\n", rpt);
456 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
457 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
458 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
459 ++lu, b->last_commit, b->name);
461 fputc('\n', rpt);
462 fputs("Inactive Branches\n", rpt);
463 fputs("-----------------\n", rpt);
464 for (lu = 0; lu < branch_table_sz; lu++) {
465 for (b = branch_table[lu]; b; b = b->table_next_branch)
466 write_branch_report(rpt, b);
469 if (first_tag) {
470 struct tag *tg;
471 fputc('\n', rpt);
472 fputs("Annotated Tags\n", rpt);
473 fputs("--------------\n", rpt);
474 for (tg = first_tag; tg; tg = tg->next_tag) {
475 fputs(oid_to_hex(&tg->oid), rpt);
476 fputc(' ', rpt);
477 fputs(tg->name, rpt);
478 fputc('\n', rpt);
482 fputc('\n', rpt);
483 fputs("Marks\n", rpt);
484 fputs("-----\n", rpt);
485 if (export_marks_file)
486 fprintf(rpt, " exported to %s\n", export_marks_file);
487 else
488 dump_marks_helper(rpt, 0, marks);
490 fputc('\n', rpt);
491 fputs("-------------------\n", rpt);
492 fputs("END OF CRASH REPORT\n", rpt);
493 fclose(rpt);
494 free(loc);
497 static void end_packfile(void);
498 static void unkeep_all_packs(void);
499 static void dump_marks(void);
501 static NORETURN void die_nicely(const char *err, va_list params)
503 static int zombie;
504 char message[2 * PATH_MAX];
506 vsnprintf(message, sizeof(message), err, params);
507 fputs("fatal: ", stderr);
508 fputs(message, stderr);
509 fputc('\n', stderr);
511 if (!zombie) {
512 zombie = 1;
513 write_crash_report(message);
514 end_packfile();
515 unkeep_all_packs();
516 dump_marks();
518 exit(128);
521 #ifndef SIGUSR1 /* Windows, for example */
523 static void set_checkpoint_signal(void)
527 #else
529 static void checkpoint_signal(int signo)
531 checkpoint_requested = 1;
534 static void set_checkpoint_signal(void)
536 struct sigaction sa;
538 memset(&sa, 0, sizeof(sa));
539 sa.sa_handler = checkpoint_signal;
540 sigemptyset(&sa.sa_mask);
541 sa.sa_flags = SA_RESTART;
542 sigaction(SIGUSR1, &sa, NULL);
545 #endif
547 static void alloc_objects(unsigned int cnt)
549 struct object_entry_pool *b;
551 b = xmalloc(sizeof(struct object_entry_pool)
552 + cnt * sizeof(struct object_entry));
553 b->next_pool = blocks;
554 b->next_free = b->entries;
555 b->end = b->entries + cnt;
556 blocks = b;
557 alloc_count += cnt;
560 static struct object_entry *new_object(struct object_id *oid)
562 struct object_entry *e;
564 if (blocks->next_free == blocks->end)
565 alloc_objects(object_entry_alloc);
567 e = blocks->next_free++;
568 oidcpy(&e->idx.oid, oid);
569 return e;
572 static struct object_entry *find_object(struct object_id *oid)
574 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
575 struct object_entry *e;
576 for (e = object_table[h]; e; e = e->next)
577 if (!oidcmp(oid, &e->idx.oid))
578 return e;
579 return NULL;
582 static struct object_entry *insert_object(struct object_id *oid)
584 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
585 struct object_entry *e = object_table[h];
587 while (e) {
588 if (!oidcmp(oid, &e->idx.oid))
589 return e;
590 e = e->next;
593 e = new_object(oid);
594 e->next = object_table[h];
595 e->idx.offset = 0;
596 object_table[h] = e;
597 return e;
600 static void invalidate_pack_id(unsigned int id)
602 unsigned int h;
603 unsigned long lu;
604 struct tag *t;
606 for (h = 0; h < ARRAY_SIZE(object_table); h++) {
607 struct object_entry *e;
609 for (e = object_table[h]; e; e = e->next)
610 if (e->pack_id == id)
611 e->pack_id = MAX_PACK_ID;
614 for (lu = 0; lu < branch_table_sz; lu++) {
615 struct branch *b;
617 for (b = branch_table[lu]; b; b = b->table_next_branch)
618 if (b->pack_id == id)
619 b->pack_id = MAX_PACK_ID;
622 for (t = first_tag; t; t = t->next_tag)
623 if (t->pack_id == id)
624 t->pack_id = MAX_PACK_ID;
627 static unsigned int hc_str(const char *s, size_t len)
629 unsigned int r = 0;
630 while (len-- > 0)
631 r = r * 31 + *s++;
632 return r;
635 static void *pool_alloc(size_t len)
637 struct mem_pool *p;
638 void *r;
640 /* round up to a 'uintmax_t' alignment */
641 if (len & (sizeof(uintmax_t) - 1))
642 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
644 for (p = mem_pool; p; p = p->next_pool)
645 if ((p->end - p->next_free >= len))
646 break;
648 if (!p) {
649 if (len >= (mem_pool_alloc/2)) {
650 total_allocd += len;
651 return xmalloc(len);
653 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
654 p = xmalloc(st_add(sizeof(struct mem_pool), mem_pool_alloc));
655 p->next_pool = mem_pool;
656 p->next_free = (char *) p->space;
657 p->end = p->next_free + mem_pool_alloc;
658 mem_pool = p;
661 r = p->next_free;
662 p->next_free += len;
663 return r;
666 static void *pool_calloc(size_t count, size_t size)
668 size_t len = count * size;
669 void *r = pool_alloc(len);
670 memset(r, 0, len);
671 return r;
674 static char *pool_strdup(const char *s)
676 size_t len = strlen(s) + 1;
677 char *r = pool_alloc(len);
678 memcpy(r, s, len);
679 return r;
682 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
684 struct mark_set *s = marks;
685 while ((idnum >> s->shift) >= 1024) {
686 s = pool_calloc(1, sizeof(struct mark_set));
687 s->shift = marks->shift + 10;
688 s->data.sets[0] = marks;
689 marks = s;
691 while (s->shift) {
692 uintmax_t i = idnum >> s->shift;
693 idnum -= i << s->shift;
694 if (!s->data.sets[i]) {
695 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
696 s->data.sets[i]->shift = s->shift - 10;
698 s = s->data.sets[i];
700 if (!s->data.marked[idnum])
701 marks_set_count++;
702 s->data.marked[idnum] = oe;
705 static struct object_entry *find_mark(uintmax_t idnum)
707 uintmax_t orig_idnum = idnum;
708 struct mark_set *s = marks;
709 struct object_entry *oe = NULL;
710 if ((idnum >> s->shift) < 1024) {
711 while (s && s->shift) {
712 uintmax_t i = idnum >> s->shift;
713 idnum -= i << s->shift;
714 s = s->data.sets[i];
716 if (s)
717 oe = s->data.marked[idnum];
719 if (!oe)
720 die("mark :%" PRIuMAX " not declared", orig_idnum);
721 return oe;
724 static struct atom_str *to_atom(const char *s, unsigned short len)
726 unsigned int hc = hc_str(s, len) % atom_table_sz;
727 struct atom_str *c;
729 for (c = atom_table[hc]; c; c = c->next_atom)
730 if (c->str_len == len && !strncmp(s, c->str_dat, len))
731 return c;
733 c = pool_alloc(sizeof(struct atom_str) + len + 1);
734 c->str_len = len;
735 memcpy(c->str_dat, s, len);
736 c->str_dat[len] = 0;
737 c->next_atom = atom_table[hc];
738 atom_table[hc] = c;
739 atom_cnt++;
740 return c;
743 static struct branch *lookup_branch(const char *name)
745 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
746 struct branch *b;
748 for (b = branch_table[hc]; b; b = b->table_next_branch)
749 if (!strcmp(name, b->name))
750 return b;
751 return NULL;
754 static struct branch *new_branch(const char *name)
756 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
757 struct branch *b = lookup_branch(name);
759 if (b)
760 die("Invalid attempt to create duplicate branch: %s", name);
761 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
762 die("Branch name doesn't conform to GIT standards: %s", name);
764 b = pool_calloc(1, sizeof(struct branch));
765 b->name = pool_strdup(name);
766 b->table_next_branch = branch_table[hc];
767 b->branch_tree.versions[0].mode = S_IFDIR;
768 b->branch_tree.versions[1].mode = S_IFDIR;
769 b->num_notes = 0;
770 b->active = 0;
771 b->pack_id = MAX_PACK_ID;
772 branch_table[hc] = b;
773 branch_count++;
774 return b;
777 static unsigned int hc_entries(unsigned int cnt)
779 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
780 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
783 static struct tree_content *new_tree_content(unsigned int cnt)
785 struct avail_tree_content *f, *l = NULL;
786 struct tree_content *t;
787 unsigned int hc = hc_entries(cnt);
789 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
790 if (f->entry_capacity >= cnt)
791 break;
793 if (f) {
794 if (l)
795 l->next_avail = f->next_avail;
796 else
797 avail_tree_table[hc] = f->next_avail;
798 } else {
799 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
800 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
801 f->entry_capacity = cnt;
804 t = (struct tree_content*)f;
805 t->entry_count = 0;
806 t->delta_depth = 0;
807 return t;
810 static void release_tree_entry(struct tree_entry *e);
811 static void release_tree_content(struct tree_content *t)
813 struct avail_tree_content *f = (struct avail_tree_content*)t;
814 unsigned int hc = hc_entries(f->entry_capacity);
815 f->next_avail = avail_tree_table[hc];
816 avail_tree_table[hc] = f;
819 static void release_tree_content_recursive(struct tree_content *t)
821 unsigned int i;
822 for (i = 0; i < t->entry_count; i++)
823 release_tree_entry(t->entries[i]);
824 release_tree_content(t);
827 static struct tree_content *grow_tree_content(
828 struct tree_content *t,
829 int amt)
831 struct tree_content *r = new_tree_content(t->entry_count + amt);
832 r->entry_count = t->entry_count;
833 r->delta_depth = t->delta_depth;
834 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
835 release_tree_content(t);
836 return r;
839 static struct tree_entry *new_tree_entry(void)
841 struct tree_entry *e;
843 if (!avail_tree_entry) {
844 unsigned int n = tree_entry_alloc;
845 total_allocd += n * sizeof(struct tree_entry);
846 ALLOC_ARRAY(e, n);
847 avail_tree_entry = e;
848 while (n-- > 1) {
849 *((void**)e) = e + 1;
850 e++;
852 *((void**)e) = NULL;
855 e = avail_tree_entry;
856 avail_tree_entry = *((void**)e);
857 return e;
860 static void release_tree_entry(struct tree_entry *e)
862 if (e->tree)
863 release_tree_content_recursive(e->tree);
864 *((void**)e) = avail_tree_entry;
865 avail_tree_entry = e;
868 static struct tree_content *dup_tree_content(struct tree_content *s)
870 struct tree_content *d;
871 struct tree_entry *a, *b;
872 unsigned int i;
874 if (!s)
875 return NULL;
876 d = new_tree_content(s->entry_count);
877 for (i = 0; i < s->entry_count; i++) {
878 a = s->entries[i];
879 b = new_tree_entry();
880 memcpy(b, a, sizeof(*a));
881 if (a->tree && is_null_oid(&b->versions[1].oid))
882 b->tree = dup_tree_content(a->tree);
883 else
884 b->tree = NULL;
885 d->entries[i] = b;
887 d->entry_count = s->entry_count;
888 d->delta_depth = s->delta_depth;
890 return d;
893 static void start_packfile(void)
895 struct strbuf tmp_file = STRBUF_INIT;
896 struct packed_git *p;
897 struct pack_header hdr;
898 int pack_fd;
900 pack_fd = odb_mkstemp(&tmp_file, "pack/tmp_pack_XXXXXX");
901 FLEX_ALLOC_STR(p, pack_name, tmp_file.buf);
902 strbuf_release(&tmp_file);
904 p->pack_fd = pack_fd;
905 p->do_not_close = 1;
906 pack_file = sha1fd(pack_fd, p->pack_name);
908 hdr.hdr_signature = htonl(PACK_SIGNATURE);
909 hdr.hdr_version = htonl(2);
910 hdr.hdr_entries = 0;
911 sha1write(pack_file, &hdr, sizeof(hdr));
913 pack_data = p;
914 pack_size = sizeof(hdr);
915 object_count = 0;
917 REALLOC_ARRAY(all_packs, pack_id + 1);
918 all_packs[pack_id] = p;
921 static const char *create_index(void)
923 const char *tmpfile;
924 struct pack_idx_entry **idx, **c, **last;
925 struct object_entry *e;
926 struct object_entry_pool *o;
928 /* Build the table of object IDs. */
929 ALLOC_ARRAY(idx, object_count);
930 c = idx;
931 for (o = blocks; o; o = o->next_pool)
932 for (e = o->next_free; e-- != o->entries;)
933 if (pack_id == e->pack_id)
934 *c++ = &e->idx;
935 last = idx + object_count;
936 if (c != last)
937 die("internal consistency error creating the index");
939 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
940 free(idx);
941 return tmpfile;
944 static char *keep_pack(const char *curr_index_name)
946 static const char *keep_msg = "fast-import";
947 struct strbuf name = STRBUF_INIT;
948 int keep_fd;
950 odb_pack_name(&name, pack_data->sha1, "keep");
951 keep_fd = odb_pack_keep(name.buf);
952 if (keep_fd < 0)
953 die_errno("cannot create keep file");
954 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
955 if (close(keep_fd))
956 die_errno("failed to write keep file");
958 odb_pack_name(&name, pack_data->sha1, "pack");
959 if (finalize_object_file(pack_data->pack_name, name.buf))
960 die("cannot store pack file");
962 odb_pack_name(&name, pack_data->sha1, "idx");
963 if (finalize_object_file(curr_index_name, name.buf))
964 die("cannot store index file");
965 free((void *)curr_index_name);
966 return strbuf_detach(&name, NULL);
969 static void unkeep_all_packs(void)
971 struct strbuf name = STRBUF_INIT;
972 int k;
974 for (k = 0; k < pack_id; k++) {
975 struct packed_git *p = all_packs[k];
976 odb_pack_name(&name, p->sha1, "keep");
977 unlink_or_warn(name.buf);
979 strbuf_release(&name);
982 static int loosen_small_pack(const struct packed_git *p)
984 struct child_process unpack = CHILD_PROCESS_INIT;
986 if (lseek(p->pack_fd, 0, SEEK_SET) < 0)
987 die_errno("Failed seeking to start of '%s'", p->pack_name);
989 unpack.in = p->pack_fd;
990 unpack.git_cmd = 1;
991 unpack.stdout_to_stderr = 1;
992 argv_array_push(&unpack.args, "unpack-objects");
993 if (!show_stats)
994 argv_array_push(&unpack.args, "-q");
996 return run_command(&unpack);
999 static void end_packfile(void)
1001 static int running;
1003 if (running || !pack_data)
1004 return;
1006 running = 1;
1007 clear_delta_base_cache();
1008 if (object_count) {
1009 struct packed_git *new_p;
1010 struct object_id cur_pack_oid;
1011 char *idx_name;
1012 int i;
1013 struct branch *b;
1014 struct tag *t;
1016 close_pack_windows(pack_data);
1017 sha1close(pack_file, cur_pack_oid.hash, 0);
1018 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
1019 pack_data->pack_name, object_count,
1020 cur_pack_oid.hash, pack_size);
1022 if (object_count <= unpack_limit) {
1023 if (!loosen_small_pack(pack_data)) {
1024 invalidate_pack_id(pack_id);
1025 goto discard_pack;
1029 close(pack_data->pack_fd);
1030 idx_name = keep_pack(create_index());
1032 /* Register the packfile with core git's machinery. */
1033 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
1034 if (!new_p)
1035 die("core git rejected index %s", idx_name);
1036 all_packs[pack_id] = new_p;
1037 install_packed_git(new_p);
1038 free(idx_name);
1040 /* Print the boundary */
1041 if (pack_edges) {
1042 fprintf(pack_edges, "%s:", new_p->pack_name);
1043 for (i = 0; i < branch_table_sz; i++) {
1044 for (b = branch_table[i]; b; b = b->table_next_branch) {
1045 if (b->pack_id == pack_id)
1046 fprintf(pack_edges, " %s",
1047 oid_to_hex(&b->oid));
1050 for (t = first_tag; t; t = t->next_tag) {
1051 if (t->pack_id == pack_id)
1052 fprintf(pack_edges, " %s",
1053 oid_to_hex(&t->oid));
1055 fputc('\n', pack_edges);
1056 fflush(pack_edges);
1059 pack_id++;
1061 else {
1062 discard_pack:
1063 close(pack_data->pack_fd);
1064 unlink_or_warn(pack_data->pack_name);
1066 FREE_AND_NULL(pack_data);
1067 running = 0;
1069 /* We can't carry a delta across packfiles. */
1070 strbuf_release(&last_blob.data);
1071 last_blob.offset = 0;
1072 last_blob.depth = 0;
1075 static void cycle_packfile(void)
1077 end_packfile();
1078 start_packfile();
1081 static int store_object(
1082 enum object_type type,
1083 struct strbuf *dat,
1084 struct last_object *last,
1085 struct object_id *oidout,
1086 uintmax_t mark)
1088 void *out, *delta;
1089 struct object_entry *e;
1090 unsigned char hdr[96];
1091 struct object_id oid;
1092 unsigned long hdrlen, deltalen;
1093 git_SHA_CTX c;
1094 git_zstream s;
1096 hdrlen = xsnprintf((char *)hdr, sizeof(hdr), "%s %lu",
1097 typename(type), (unsigned long)dat->len) + 1;
1098 git_SHA1_Init(&c);
1099 git_SHA1_Update(&c, hdr, hdrlen);
1100 git_SHA1_Update(&c, dat->buf, dat->len);
1101 git_SHA1_Final(oid.hash, &c);
1102 if (oidout)
1103 oidcpy(oidout, &oid);
1105 e = insert_object(&oid);
1106 if (mark)
1107 insert_mark(mark, e);
1108 if (e->idx.offset) {
1109 duplicate_count_by_type[type]++;
1110 return 1;
1111 } else if (find_sha1_pack(oid.hash, packed_git)) {
1112 e->type = type;
1113 e->pack_id = MAX_PACK_ID;
1114 e->idx.offset = 1; /* just not zero! */
1115 duplicate_count_by_type[type]++;
1116 return 1;
1119 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1120 delta_count_attempts_by_type[type]++;
1121 delta = diff_delta(last->data.buf, last->data.len,
1122 dat->buf, dat->len,
1123 &deltalen, dat->len - 20);
1124 } else
1125 delta = NULL;
1127 git_deflate_init(&s, pack_compression_level);
1128 if (delta) {
1129 s.next_in = delta;
1130 s.avail_in = deltalen;
1131 } else {
1132 s.next_in = (void *)dat->buf;
1133 s.avail_in = dat->len;
1135 s.avail_out = git_deflate_bound(&s, s.avail_in);
1136 s.next_out = out = xmalloc(s.avail_out);
1137 while (git_deflate(&s, Z_FINISH) == Z_OK)
1138 ; /* nothing */
1139 git_deflate_end(&s);
1141 /* Determine if we should auto-checkpoint. */
1142 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1143 || (pack_size + 60 + s.total_out) < pack_size) {
1145 /* This new object needs to *not* have the current pack_id. */
1146 e->pack_id = pack_id + 1;
1147 cycle_packfile();
1149 /* We cannot carry a delta into the new pack. */
1150 if (delta) {
1151 FREE_AND_NULL(delta);
1153 git_deflate_init(&s, pack_compression_level);
1154 s.next_in = (void *)dat->buf;
1155 s.avail_in = dat->len;
1156 s.avail_out = git_deflate_bound(&s, s.avail_in);
1157 s.next_out = out = xrealloc(out, s.avail_out);
1158 while (git_deflate(&s, Z_FINISH) == Z_OK)
1159 ; /* nothing */
1160 git_deflate_end(&s);
1164 e->type = type;
1165 e->pack_id = pack_id;
1166 e->idx.offset = pack_size;
1167 object_count++;
1168 object_count_by_type[type]++;
1170 crc32_begin(pack_file);
1172 if (delta) {
1173 off_t ofs = e->idx.offset - last->offset;
1174 unsigned pos = sizeof(hdr) - 1;
1176 delta_count_by_type[type]++;
1177 e->depth = last->depth + 1;
1179 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1180 OBJ_OFS_DELTA, deltalen);
1181 sha1write(pack_file, hdr, hdrlen);
1182 pack_size += hdrlen;
1184 hdr[pos] = ofs & 127;
1185 while (ofs >>= 7)
1186 hdr[--pos] = 128 | (--ofs & 127);
1187 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1188 pack_size += sizeof(hdr) - pos;
1189 } else {
1190 e->depth = 0;
1191 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1192 type, dat->len);
1193 sha1write(pack_file, hdr, hdrlen);
1194 pack_size += hdrlen;
1197 sha1write(pack_file, out, s.total_out);
1198 pack_size += s.total_out;
1200 e->idx.crc32 = crc32_end(pack_file);
1202 free(out);
1203 free(delta);
1204 if (last) {
1205 if (last->no_swap) {
1206 last->data = *dat;
1207 } else {
1208 strbuf_swap(&last->data, dat);
1210 last->offset = e->idx.offset;
1211 last->depth = e->depth;
1213 return 0;
1216 static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1218 if (sha1file_truncate(pack_file, checkpoint))
1219 die_errno("cannot truncate pack to skip duplicate");
1220 pack_size = checkpoint->offset;
1223 static void stream_blob(uintmax_t len, struct object_id *oidout, uintmax_t mark)
1225 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1226 unsigned char *in_buf = xmalloc(in_sz);
1227 unsigned char *out_buf = xmalloc(out_sz);
1228 struct object_entry *e;
1229 struct object_id oid;
1230 unsigned long hdrlen;
1231 off_t offset;
1232 git_SHA_CTX c;
1233 git_zstream s;
1234 struct sha1file_checkpoint checkpoint;
1235 int status = Z_OK;
1237 /* Determine if we should auto-checkpoint. */
1238 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1239 || (pack_size + 60 + len) < pack_size)
1240 cycle_packfile();
1242 sha1file_checkpoint(pack_file, &checkpoint);
1243 offset = checkpoint.offset;
1245 hdrlen = xsnprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1247 git_SHA1_Init(&c);
1248 git_SHA1_Update(&c, out_buf, hdrlen);
1250 crc32_begin(pack_file);
1252 git_deflate_init(&s, pack_compression_level);
1254 hdrlen = encode_in_pack_object_header(out_buf, out_sz, OBJ_BLOB, len);
1256 s.next_out = out_buf + hdrlen;
1257 s.avail_out = out_sz - hdrlen;
1259 while (status != Z_STREAM_END) {
1260 if (0 < len && !s.avail_in) {
1261 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1262 size_t n = fread(in_buf, 1, cnt, stdin);
1263 if (!n && feof(stdin))
1264 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1266 git_SHA1_Update(&c, in_buf, n);
1267 s.next_in = in_buf;
1268 s.avail_in = n;
1269 len -= n;
1272 status = git_deflate(&s, len ? 0 : Z_FINISH);
1274 if (!s.avail_out || status == Z_STREAM_END) {
1275 size_t n = s.next_out - out_buf;
1276 sha1write(pack_file, out_buf, n);
1277 pack_size += n;
1278 s.next_out = out_buf;
1279 s.avail_out = out_sz;
1282 switch (status) {
1283 case Z_OK:
1284 case Z_BUF_ERROR:
1285 case Z_STREAM_END:
1286 continue;
1287 default:
1288 die("unexpected deflate failure: %d", status);
1291 git_deflate_end(&s);
1292 git_SHA1_Final(oid.hash, &c);
1294 if (oidout)
1295 oidcpy(oidout, &oid);
1297 e = insert_object(&oid);
1299 if (mark)
1300 insert_mark(mark, e);
1302 if (e->idx.offset) {
1303 duplicate_count_by_type[OBJ_BLOB]++;
1304 truncate_pack(&checkpoint);
1306 } else if (find_sha1_pack(oid.hash, packed_git)) {
1307 e->type = OBJ_BLOB;
1308 e->pack_id = MAX_PACK_ID;
1309 e->idx.offset = 1; /* just not zero! */
1310 duplicate_count_by_type[OBJ_BLOB]++;
1311 truncate_pack(&checkpoint);
1313 } else {
1314 e->depth = 0;
1315 e->type = OBJ_BLOB;
1316 e->pack_id = pack_id;
1317 e->idx.offset = offset;
1318 e->idx.crc32 = crc32_end(pack_file);
1319 object_count++;
1320 object_count_by_type[OBJ_BLOB]++;
1323 free(in_buf);
1324 free(out_buf);
1327 /* All calls must be guarded by find_object() or find_mark() to
1328 * ensure the 'struct object_entry' passed was written by this
1329 * process instance. We unpack the entry by the offset, avoiding
1330 * the need for the corresponding .idx file. This unpacking rule
1331 * works because we only use OBJ_REF_DELTA within the packfiles
1332 * created by fast-import.
1334 * oe must not be NULL. Such an oe usually comes from giving
1335 * an unknown SHA-1 to find_object() or an undefined mark to
1336 * find_mark(). Callers must test for this condition and use
1337 * the standard read_sha1_file() when it happens.
1339 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1340 * find_mark(), where the mark was reloaded from an existing marks
1341 * file and is referencing an object that this fast-import process
1342 * instance did not write out to a packfile. Callers must test for
1343 * this condition and use read_sha1_file() instead.
1345 static void *gfi_unpack_entry(
1346 struct object_entry *oe,
1347 unsigned long *sizep)
1349 enum object_type type;
1350 struct packed_git *p = all_packs[oe->pack_id];
1351 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1352 /* The object is stored in the packfile we are writing to
1353 * and we have modified it since the last time we scanned
1354 * back to read a previously written object. If an old
1355 * window covered [p->pack_size, p->pack_size + 20) its
1356 * data is stale and is not valid. Closing all windows
1357 * and updating the packfile length ensures we can read
1358 * the newly written data.
1360 close_pack_windows(p);
1361 sha1flush(pack_file);
1363 /* We have to offer 20 bytes additional on the end of
1364 * the packfile as the core unpacker code assumes the
1365 * footer is present at the file end and must promise
1366 * at least 20 bytes within any window it maps. But
1367 * we don't actually create the footer here.
1369 p->pack_size = pack_size + 20;
1371 return unpack_entry(p, oe->idx.offset, &type, sizep);
1374 static const char *get_mode(const char *str, uint16_t *modep)
1376 unsigned char c;
1377 uint16_t mode = 0;
1379 while ((c = *str++) != ' ') {
1380 if (c < '0' || c > '7')
1381 return NULL;
1382 mode = (mode << 3) + (c - '0');
1384 *modep = mode;
1385 return str;
1388 static void load_tree(struct tree_entry *root)
1390 struct object_id *oid = &root->versions[1].oid;
1391 struct object_entry *myoe;
1392 struct tree_content *t;
1393 unsigned long size;
1394 char *buf;
1395 const char *c;
1397 root->tree = t = new_tree_content(8);
1398 if (is_null_oid(oid))
1399 return;
1401 myoe = find_object(oid);
1402 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1403 if (myoe->type != OBJ_TREE)
1404 die("Not a tree: %s", oid_to_hex(oid));
1405 t->delta_depth = myoe->depth;
1406 buf = gfi_unpack_entry(myoe, &size);
1407 if (!buf)
1408 die("Can't load tree %s", oid_to_hex(oid));
1409 } else {
1410 enum object_type type;
1411 buf = read_sha1_file(oid->hash, &type, &size);
1412 if (!buf || type != OBJ_TREE)
1413 die("Can't load tree %s", oid_to_hex(oid));
1416 c = buf;
1417 while (c != (buf + size)) {
1418 struct tree_entry *e = new_tree_entry();
1420 if (t->entry_count == t->entry_capacity)
1421 root->tree = t = grow_tree_content(t, t->entry_count);
1422 t->entries[t->entry_count++] = e;
1424 e->tree = NULL;
1425 c = get_mode(c, &e->versions[1].mode);
1426 if (!c)
1427 die("Corrupt mode in %s", oid_to_hex(oid));
1428 e->versions[0].mode = e->versions[1].mode;
1429 e->name = to_atom(c, strlen(c));
1430 c += e->name->str_len + 1;
1431 hashcpy(e->versions[0].oid.hash, (unsigned char *)c);
1432 hashcpy(e->versions[1].oid.hash, (unsigned char *)c);
1433 c += GIT_SHA1_RAWSZ;
1435 free(buf);
1438 static int tecmp0 (const void *_a, const void *_b)
1440 struct tree_entry *a = *((struct tree_entry**)_a);
1441 struct tree_entry *b = *((struct tree_entry**)_b);
1442 return base_name_compare(
1443 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1444 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1447 static int tecmp1 (const void *_a, const void *_b)
1449 struct tree_entry *a = *((struct tree_entry**)_a);
1450 struct tree_entry *b = *((struct tree_entry**)_b);
1451 return base_name_compare(
1452 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1453 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1456 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1458 size_t maxlen = 0;
1459 unsigned int i;
1461 if (!v)
1462 QSORT(t->entries, t->entry_count, tecmp0);
1463 else
1464 QSORT(t->entries, t->entry_count, tecmp1);
1466 for (i = 0; i < t->entry_count; i++) {
1467 if (t->entries[i]->versions[v].mode)
1468 maxlen += t->entries[i]->name->str_len + 34;
1471 strbuf_reset(b);
1472 strbuf_grow(b, maxlen);
1473 for (i = 0; i < t->entry_count; i++) {
1474 struct tree_entry *e = t->entries[i];
1475 if (!e->versions[v].mode)
1476 continue;
1477 strbuf_addf(b, "%o %s%c",
1478 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1479 e->name->str_dat, '\0');
1480 strbuf_add(b, e->versions[v].oid.hash, GIT_SHA1_RAWSZ);
1484 static void store_tree(struct tree_entry *root)
1486 struct tree_content *t;
1487 unsigned int i, j, del;
1488 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1489 struct object_entry *le = NULL;
1491 if (!is_null_oid(&root->versions[1].oid))
1492 return;
1494 if (!root->tree)
1495 load_tree(root);
1496 t = root->tree;
1498 for (i = 0; i < t->entry_count; i++) {
1499 if (t->entries[i]->tree)
1500 store_tree(t->entries[i]);
1503 if (!(root->versions[0].mode & NO_DELTA))
1504 le = find_object(&root->versions[0].oid);
1505 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1506 mktree(t, 0, &old_tree);
1507 lo.data = old_tree;
1508 lo.offset = le->idx.offset;
1509 lo.depth = t->delta_depth;
1512 mktree(t, 1, &new_tree);
1513 store_object(OBJ_TREE, &new_tree, &lo, &root->versions[1].oid, 0);
1515 t->delta_depth = lo.depth;
1516 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1517 struct tree_entry *e = t->entries[i];
1518 if (e->versions[1].mode) {
1519 e->versions[0].mode = e->versions[1].mode;
1520 oidcpy(&e->versions[0].oid, &e->versions[1].oid);
1521 t->entries[j++] = e;
1522 } else {
1523 release_tree_entry(e);
1524 del++;
1527 t->entry_count -= del;
1530 static void tree_content_replace(
1531 struct tree_entry *root,
1532 const struct object_id *oid,
1533 const uint16_t mode,
1534 struct tree_content *newtree)
1536 if (!S_ISDIR(mode))
1537 die("Root cannot be a non-directory");
1538 oidclr(&root->versions[0].oid);
1539 oidcpy(&root->versions[1].oid, oid);
1540 if (root->tree)
1541 release_tree_content_recursive(root->tree);
1542 root->tree = newtree;
1545 static int tree_content_set(
1546 struct tree_entry *root,
1547 const char *p,
1548 const struct object_id *oid,
1549 const uint16_t mode,
1550 struct tree_content *subtree)
1552 struct tree_content *t;
1553 const char *slash1;
1554 unsigned int i, n;
1555 struct tree_entry *e;
1557 slash1 = strchrnul(p, '/');
1558 n = slash1 - p;
1559 if (!n)
1560 die("Empty path component found in input");
1561 if (!*slash1 && !S_ISDIR(mode) && subtree)
1562 die("Non-directories cannot have subtrees");
1564 if (!root->tree)
1565 load_tree(root);
1566 t = root->tree;
1567 for (i = 0; i < t->entry_count; i++) {
1568 e = t->entries[i];
1569 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1570 if (!*slash1) {
1571 if (!S_ISDIR(mode)
1572 && e->versions[1].mode == mode
1573 && !oidcmp(&e->versions[1].oid, oid))
1574 return 0;
1575 e->versions[1].mode = mode;
1576 oidcpy(&e->versions[1].oid, oid);
1577 if (e->tree)
1578 release_tree_content_recursive(e->tree);
1579 e->tree = subtree;
1582 * We need to leave e->versions[0].sha1 alone
1583 * to avoid modifying the preimage tree used
1584 * when writing out the parent directory.
1585 * But after replacing the subdir with a
1586 * completely different one, it's not a good
1587 * delta base any more, and besides, we've
1588 * thrown away the tree entries needed to
1589 * make a delta against it.
1591 * So let's just explicitly disable deltas
1592 * for the subtree.
1594 if (S_ISDIR(e->versions[0].mode))
1595 e->versions[0].mode |= NO_DELTA;
1597 oidclr(&root->versions[1].oid);
1598 return 1;
1600 if (!S_ISDIR(e->versions[1].mode)) {
1601 e->tree = new_tree_content(8);
1602 e->versions[1].mode = S_IFDIR;
1604 if (!e->tree)
1605 load_tree(e);
1606 if (tree_content_set(e, slash1 + 1, oid, mode, subtree)) {
1607 oidclr(&root->versions[1].oid);
1608 return 1;
1610 return 0;
1614 if (t->entry_count == t->entry_capacity)
1615 root->tree = t = grow_tree_content(t, t->entry_count);
1616 e = new_tree_entry();
1617 e->name = to_atom(p, n);
1618 e->versions[0].mode = 0;
1619 oidclr(&e->versions[0].oid);
1620 t->entries[t->entry_count++] = e;
1621 if (*slash1) {
1622 e->tree = new_tree_content(8);
1623 e->versions[1].mode = S_IFDIR;
1624 tree_content_set(e, slash1 + 1, oid, mode, subtree);
1625 } else {
1626 e->tree = subtree;
1627 e->versions[1].mode = mode;
1628 oidcpy(&e->versions[1].oid, oid);
1630 oidclr(&root->versions[1].oid);
1631 return 1;
1634 static int tree_content_remove(
1635 struct tree_entry *root,
1636 const char *p,
1637 struct tree_entry *backup_leaf,
1638 int allow_root)
1640 struct tree_content *t;
1641 const char *slash1;
1642 unsigned int i, n;
1643 struct tree_entry *e;
1645 slash1 = strchrnul(p, '/');
1646 n = slash1 - p;
1648 if (!root->tree)
1649 load_tree(root);
1651 if (!*p && allow_root) {
1652 e = root;
1653 goto del_entry;
1656 t = root->tree;
1657 for (i = 0; i < t->entry_count; i++) {
1658 e = t->entries[i];
1659 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1660 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1662 * If p names a file in some subdirectory, and a
1663 * file or symlink matching the name of the
1664 * parent directory of p exists, then p cannot
1665 * exist and need not be deleted.
1667 return 1;
1668 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1669 goto del_entry;
1670 if (!e->tree)
1671 load_tree(e);
1672 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1673 for (n = 0; n < e->tree->entry_count; n++) {
1674 if (e->tree->entries[n]->versions[1].mode) {
1675 oidclr(&root->versions[1].oid);
1676 return 1;
1679 backup_leaf = NULL;
1680 goto del_entry;
1682 return 0;
1685 return 0;
1687 del_entry:
1688 if (backup_leaf)
1689 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1690 else if (e->tree)
1691 release_tree_content_recursive(e->tree);
1692 e->tree = NULL;
1693 e->versions[1].mode = 0;
1694 oidclr(&e->versions[1].oid);
1695 oidclr(&root->versions[1].oid);
1696 return 1;
1699 static int tree_content_get(
1700 struct tree_entry *root,
1701 const char *p,
1702 struct tree_entry *leaf,
1703 int allow_root)
1705 struct tree_content *t;
1706 const char *slash1;
1707 unsigned int i, n;
1708 struct tree_entry *e;
1710 slash1 = strchrnul(p, '/');
1711 n = slash1 - p;
1712 if (!n && !allow_root)
1713 die("Empty path component found in input");
1715 if (!root->tree)
1716 load_tree(root);
1718 if (!n) {
1719 e = root;
1720 goto found_entry;
1723 t = root->tree;
1724 for (i = 0; i < t->entry_count; i++) {
1725 e = t->entries[i];
1726 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1727 if (!*slash1)
1728 goto found_entry;
1729 if (!S_ISDIR(e->versions[1].mode))
1730 return 0;
1731 if (!e->tree)
1732 load_tree(e);
1733 return tree_content_get(e, slash1 + 1, leaf, 0);
1736 return 0;
1738 found_entry:
1739 memcpy(leaf, e, sizeof(*leaf));
1740 if (e->tree && is_null_oid(&e->versions[1].oid))
1741 leaf->tree = dup_tree_content(e->tree);
1742 else
1743 leaf->tree = NULL;
1744 return 1;
1747 static int update_branch(struct branch *b)
1749 static const char *msg = "fast-import";
1750 struct ref_transaction *transaction;
1751 struct object_id old_oid;
1752 struct strbuf err = STRBUF_INIT;
1754 if (is_null_oid(&b->oid)) {
1755 if (b->delete)
1756 delete_ref(NULL, b->name, NULL, 0);
1757 return 0;
1759 if (read_ref(b->name, old_oid.hash))
1760 oidclr(&old_oid);
1761 if (!force_update && !is_null_oid(&old_oid)) {
1762 struct commit *old_cmit, *new_cmit;
1764 old_cmit = lookup_commit_reference_gently(&old_oid, 0);
1765 new_cmit = lookup_commit_reference_gently(&b->oid, 0);
1766 if (!old_cmit || !new_cmit)
1767 return error("Branch %s is missing commits.", b->name);
1769 if (!in_merge_bases(old_cmit, new_cmit)) {
1770 warning("Not updating %s"
1771 " (new tip %s does not contain %s)",
1772 b->name, oid_to_hex(&b->oid),
1773 oid_to_hex(&old_oid));
1774 return -1;
1777 transaction = ref_transaction_begin(&err);
1778 if (!transaction ||
1779 ref_transaction_update(transaction, b->name, b->oid.hash, old_oid.hash,
1780 0, msg, &err) ||
1781 ref_transaction_commit(transaction, &err)) {
1782 ref_transaction_free(transaction);
1783 error("%s", err.buf);
1784 strbuf_release(&err);
1785 return -1;
1787 ref_transaction_free(transaction);
1788 strbuf_release(&err);
1789 return 0;
1792 static void dump_branches(void)
1794 unsigned int i;
1795 struct branch *b;
1797 for (i = 0; i < branch_table_sz; i++) {
1798 for (b = branch_table[i]; b; b = b->table_next_branch)
1799 failure |= update_branch(b);
1803 static void dump_tags(void)
1805 static const char *msg = "fast-import";
1806 struct tag *t;
1807 struct strbuf ref_name = STRBUF_INIT;
1808 struct strbuf err = STRBUF_INIT;
1809 struct ref_transaction *transaction;
1811 transaction = ref_transaction_begin(&err);
1812 if (!transaction) {
1813 failure |= error("%s", err.buf);
1814 goto cleanup;
1816 for (t = first_tag; t; t = t->next_tag) {
1817 strbuf_reset(&ref_name);
1818 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1820 if (ref_transaction_update(transaction, ref_name.buf,
1821 t->oid.hash, NULL, 0, msg, &err)) {
1822 failure |= error("%s", err.buf);
1823 goto cleanup;
1826 if (ref_transaction_commit(transaction, &err))
1827 failure |= error("%s", err.buf);
1829 cleanup:
1830 ref_transaction_free(transaction);
1831 strbuf_release(&ref_name);
1832 strbuf_release(&err);
1835 static void dump_marks_helper(FILE *f,
1836 uintmax_t base,
1837 struct mark_set *m)
1839 uintmax_t k;
1840 if (m->shift) {
1841 for (k = 0; k < 1024; k++) {
1842 if (m->data.sets[k])
1843 dump_marks_helper(f, base + (k << m->shift),
1844 m->data.sets[k]);
1846 } else {
1847 for (k = 0; k < 1024; k++) {
1848 if (m->data.marked[k])
1849 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1850 oid_to_hex(&m->data.marked[k]->idx.oid));
1855 static void dump_marks(void)
1857 static struct lock_file mark_lock;
1858 FILE *f;
1860 if (!export_marks_file || (import_marks_file && !import_marks_file_done))
1861 return;
1863 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1864 failure |= error_errno("Unable to write marks file %s",
1865 export_marks_file);
1866 return;
1869 f = fdopen_lock_file(&mark_lock, "w");
1870 if (!f) {
1871 int saved_errno = errno;
1872 rollback_lock_file(&mark_lock);
1873 failure |= error("Unable to write marks file %s: %s",
1874 export_marks_file, strerror(saved_errno));
1875 return;
1878 dump_marks_helper(f, 0, marks);
1879 if (commit_lock_file(&mark_lock)) {
1880 failure |= error_errno("Unable to write file %s",
1881 export_marks_file);
1882 return;
1886 static void read_marks(void)
1888 char line[512];
1889 FILE *f = fopen(import_marks_file, "r");
1890 if (f)
1892 else if (import_marks_file_ignore_missing && errno == ENOENT)
1893 goto done; /* Marks file does not exist */
1894 else
1895 die_errno("cannot read '%s'", import_marks_file);
1896 while (fgets(line, sizeof(line), f)) {
1897 uintmax_t mark;
1898 char *end;
1899 struct object_id oid;
1900 struct object_entry *e;
1902 end = strchr(line, '\n');
1903 if (line[0] != ':' || !end)
1904 die("corrupt mark line: %s", line);
1905 *end = 0;
1906 mark = strtoumax(line + 1, &end, 10);
1907 if (!mark || end == line + 1
1908 || *end != ' ' || get_oid_hex(end + 1, &oid))
1909 die("corrupt mark line: %s", line);
1910 e = find_object(&oid);
1911 if (!e) {
1912 enum object_type type = sha1_object_info(oid.hash, NULL);
1913 if (type < 0)
1914 die("object not found: %s", oid_to_hex(&oid));
1915 e = insert_object(&oid);
1916 e->type = type;
1917 e->pack_id = MAX_PACK_ID;
1918 e->idx.offset = 1; /* just not zero! */
1920 insert_mark(mark, e);
1922 fclose(f);
1923 done:
1924 import_marks_file_done = 1;
1928 static int read_next_command(void)
1930 static int stdin_eof = 0;
1932 if (stdin_eof) {
1933 unread_command_buf = 0;
1934 return EOF;
1937 for (;;) {
1938 const char *p;
1940 if (unread_command_buf) {
1941 unread_command_buf = 0;
1942 } else {
1943 struct recent_command *rc;
1945 strbuf_detach(&command_buf, NULL);
1946 stdin_eof = strbuf_getline_lf(&command_buf, stdin);
1947 if (stdin_eof)
1948 return EOF;
1950 if (!seen_data_command
1951 && !starts_with(command_buf.buf, "feature ")
1952 && !starts_with(command_buf.buf, "option ")) {
1953 parse_argv();
1956 rc = rc_free;
1957 if (rc)
1958 rc_free = rc->next;
1959 else {
1960 rc = cmd_hist.next;
1961 cmd_hist.next = rc->next;
1962 cmd_hist.next->prev = &cmd_hist;
1963 free(rc->buf);
1966 rc->buf = command_buf.buf;
1967 rc->prev = cmd_tail;
1968 rc->next = cmd_hist.prev;
1969 rc->prev->next = rc;
1970 cmd_tail = rc;
1972 if (skip_prefix(command_buf.buf, "get-mark ", &p)) {
1973 parse_get_mark(p);
1974 continue;
1976 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1977 parse_cat_blob(p);
1978 continue;
1980 if (command_buf.buf[0] == '#')
1981 continue;
1982 return 0;
1986 static void skip_optional_lf(void)
1988 int term_char = fgetc(stdin);
1989 if (term_char != '\n' && term_char != EOF)
1990 ungetc(term_char, stdin);
1993 static void parse_mark(void)
1995 const char *v;
1996 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1997 next_mark = strtoumax(v, NULL, 10);
1998 read_next_command();
2000 else
2001 next_mark = 0;
2004 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
2006 const char *data;
2007 strbuf_reset(sb);
2009 if (!skip_prefix(command_buf.buf, "data ", &data))
2010 die("Expected 'data n' command, found: %s", command_buf.buf);
2012 if (skip_prefix(data, "<<", &data)) {
2013 char *term = xstrdup(data);
2014 size_t term_len = command_buf.len - (data - command_buf.buf);
2016 strbuf_detach(&command_buf, NULL);
2017 for (;;) {
2018 if (strbuf_getline_lf(&command_buf, stdin) == EOF)
2019 die("EOF in data (terminator '%s' not found)", term);
2020 if (term_len == command_buf.len
2021 && !strcmp(term, command_buf.buf))
2022 break;
2023 strbuf_addbuf(sb, &command_buf);
2024 strbuf_addch(sb, '\n');
2026 free(term);
2028 else {
2029 uintmax_t len = strtoumax(data, NULL, 10);
2030 size_t n = 0, length = (size_t)len;
2032 if (limit && limit < len) {
2033 *len_res = len;
2034 return 0;
2036 if (length < len)
2037 die("data is too large to use in this context");
2039 while (n < length) {
2040 size_t s = strbuf_fread(sb, length - n, stdin);
2041 if (!s && feof(stdin))
2042 die("EOF in data (%lu bytes remaining)",
2043 (unsigned long)(length - n));
2044 n += s;
2048 skip_optional_lf();
2049 return 1;
2052 static int validate_raw_date(const char *src, struct strbuf *result)
2054 const char *orig_src = src;
2055 char *endp;
2056 unsigned long num;
2058 errno = 0;
2060 num = strtoul(src, &endp, 10);
2061 /* NEEDSWORK: perhaps check for reasonable values? */
2062 if (errno || endp == src || *endp != ' ')
2063 return -1;
2065 src = endp + 1;
2066 if (*src != '-' && *src != '+')
2067 return -1;
2069 num = strtoul(src + 1, &endp, 10);
2070 if (errno || endp == src + 1 || *endp || 1400 < num)
2071 return -1;
2073 strbuf_addstr(result, orig_src);
2074 return 0;
2077 static char *parse_ident(const char *buf)
2079 const char *ltgt;
2080 size_t name_len;
2081 struct strbuf ident = STRBUF_INIT;
2083 /* ensure there is a space delimiter even if there is no name */
2084 if (*buf == '<')
2085 --buf;
2087 ltgt = buf + strcspn(buf, "<>");
2088 if (*ltgt != '<')
2089 die("Missing < in ident string: %s", buf);
2090 if (ltgt != buf && ltgt[-1] != ' ')
2091 die("Missing space before < in ident string: %s", buf);
2092 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2093 if (*ltgt != '>')
2094 die("Missing > in ident string: %s", buf);
2095 ltgt++;
2096 if (*ltgt != ' ')
2097 die("Missing space after > in ident string: %s", buf);
2098 ltgt++;
2099 name_len = ltgt - buf;
2100 strbuf_add(&ident, buf, name_len);
2102 switch (whenspec) {
2103 case WHENSPEC_RAW:
2104 if (validate_raw_date(ltgt, &ident) < 0)
2105 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2106 break;
2107 case WHENSPEC_RFC2822:
2108 if (parse_date(ltgt, &ident) < 0)
2109 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2110 break;
2111 case WHENSPEC_NOW:
2112 if (strcmp("now", ltgt))
2113 die("Date in ident must be 'now': %s", buf);
2114 datestamp(&ident);
2115 break;
2118 return strbuf_detach(&ident, NULL);
2121 static void parse_and_store_blob(
2122 struct last_object *last,
2123 struct object_id *oidout,
2124 uintmax_t mark)
2126 static struct strbuf buf = STRBUF_INIT;
2127 uintmax_t len;
2129 if (parse_data(&buf, big_file_threshold, &len))
2130 store_object(OBJ_BLOB, &buf, last, oidout, mark);
2131 else {
2132 if (last) {
2133 strbuf_release(&last->data);
2134 last->offset = 0;
2135 last->depth = 0;
2137 stream_blob(len, oidout, mark);
2138 skip_optional_lf();
2142 static void parse_new_blob(void)
2144 read_next_command();
2145 parse_mark();
2146 parse_and_store_blob(&last_blob, NULL, next_mark);
2149 static void unload_one_branch(void)
2151 while (cur_active_branches
2152 && cur_active_branches >= max_active_branches) {
2153 uintmax_t min_commit = ULONG_MAX;
2154 struct branch *e, *l = NULL, *p = NULL;
2156 for (e = active_branches; e; e = e->active_next_branch) {
2157 if (e->last_commit < min_commit) {
2158 p = l;
2159 min_commit = e->last_commit;
2161 l = e;
2164 if (p) {
2165 e = p->active_next_branch;
2166 p->active_next_branch = e->active_next_branch;
2167 } else {
2168 e = active_branches;
2169 active_branches = e->active_next_branch;
2171 e->active = 0;
2172 e->active_next_branch = NULL;
2173 if (e->branch_tree.tree) {
2174 release_tree_content_recursive(e->branch_tree.tree);
2175 e->branch_tree.tree = NULL;
2177 cur_active_branches--;
2181 static void load_branch(struct branch *b)
2183 load_tree(&b->branch_tree);
2184 if (!b->active) {
2185 b->active = 1;
2186 b->active_next_branch = active_branches;
2187 active_branches = b;
2188 cur_active_branches++;
2189 branch_load_count++;
2193 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2195 unsigned char fanout = 0;
2196 while ((num_notes >>= 8))
2197 fanout++;
2198 return fanout;
2201 static void construct_path_with_fanout(const char *hex_sha1,
2202 unsigned char fanout, char *path)
2204 unsigned int i = 0, j = 0;
2205 if (fanout >= 20)
2206 die("Too large fanout (%u)", fanout);
2207 while (fanout) {
2208 path[i++] = hex_sha1[j++];
2209 path[i++] = hex_sha1[j++];
2210 path[i++] = '/';
2211 fanout--;
2213 memcpy(path + i, hex_sha1 + j, GIT_SHA1_HEXSZ - j);
2214 path[i + GIT_SHA1_HEXSZ - j] = '\0';
2217 static uintmax_t do_change_note_fanout(
2218 struct tree_entry *orig_root, struct tree_entry *root,
2219 char *hex_oid, unsigned int hex_oid_len,
2220 char *fullpath, unsigned int fullpath_len,
2221 unsigned char fanout)
2223 struct tree_content *t;
2224 struct tree_entry *e, leaf;
2225 unsigned int i, tmp_hex_oid_len, tmp_fullpath_len;
2226 uintmax_t num_notes = 0;
2227 struct object_id oid;
2228 char realpath[60];
2230 if (!root->tree)
2231 load_tree(root);
2232 t = root->tree;
2234 for (i = 0; t && i < t->entry_count; i++) {
2235 e = t->entries[i];
2236 tmp_hex_oid_len = hex_oid_len + e->name->str_len;
2237 tmp_fullpath_len = fullpath_len;
2240 * We're interested in EITHER existing note entries (entries
2241 * with exactly 40 hex chars in path, not including directory
2242 * separators), OR directory entries that may contain note
2243 * entries (with < 40 hex chars in path).
2244 * Also, each path component in a note entry must be a multiple
2245 * of 2 chars.
2247 if (!e->versions[1].mode ||
2248 tmp_hex_oid_len > GIT_SHA1_HEXSZ ||
2249 e->name->str_len % 2)
2250 continue;
2252 /* This _may_ be a note entry, or a subdir containing notes */
2253 memcpy(hex_oid + hex_oid_len, e->name->str_dat,
2254 e->name->str_len);
2255 if (tmp_fullpath_len)
2256 fullpath[tmp_fullpath_len++] = '/';
2257 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2258 e->name->str_len);
2259 tmp_fullpath_len += e->name->str_len;
2260 fullpath[tmp_fullpath_len] = '\0';
2262 if (tmp_hex_oid_len == GIT_SHA1_HEXSZ && !get_oid_hex(hex_oid, &oid)) {
2263 /* This is a note entry */
2264 if (fanout == 0xff) {
2265 /* Counting mode, no rename */
2266 num_notes++;
2267 continue;
2269 construct_path_with_fanout(hex_oid, fanout, realpath);
2270 if (!strcmp(fullpath, realpath)) {
2271 /* Note entry is in correct location */
2272 num_notes++;
2273 continue;
2276 /* Rename fullpath to realpath */
2277 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2278 die("Failed to remove path %s", fullpath);
2279 tree_content_set(orig_root, realpath,
2280 &leaf.versions[1].oid,
2281 leaf.versions[1].mode,
2282 leaf.tree);
2283 } else if (S_ISDIR(e->versions[1].mode)) {
2284 /* This is a subdir that may contain note entries */
2285 num_notes += do_change_note_fanout(orig_root, e,
2286 hex_oid, tmp_hex_oid_len,
2287 fullpath, tmp_fullpath_len, fanout);
2290 /* The above may have reallocated the current tree_content */
2291 t = root->tree;
2293 return num_notes;
2296 static uintmax_t change_note_fanout(struct tree_entry *root,
2297 unsigned char fanout)
2300 * The size of path is due to one slash between every two hex digits,
2301 * plus the terminating NUL. Note that there is no slash at the end, so
2302 * the number of slashes is one less than half the number of hex
2303 * characters.
2305 char hex_oid[GIT_MAX_HEXSZ], path[GIT_MAX_HEXSZ + (GIT_MAX_HEXSZ / 2) - 1 + 1];
2306 return do_change_note_fanout(root, root, hex_oid, 0, path, 0, fanout);
2310 * Given a pointer into a string, parse a mark reference:
2312 * idnum ::= ':' bigint;
2314 * Return the first character after the value in *endptr.
2316 * Complain if the following character is not what is expected,
2317 * either a space or end of the string.
2319 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2321 uintmax_t mark;
2323 assert(*p == ':');
2324 p++;
2325 mark = strtoumax(p, endptr, 10);
2326 if (*endptr == p)
2327 die("No value after ':' in mark: %s", command_buf.buf);
2328 return mark;
2332 * Parse the mark reference, and complain if this is not the end of
2333 * the string.
2335 static uintmax_t parse_mark_ref_eol(const char *p)
2337 char *end;
2338 uintmax_t mark;
2340 mark = parse_mark_ref(p, &end);
2341 if (*end != '\0')
2342 die("Garbage after mark: %s", command_buf.buf);
2343 return mark;
2347 * Parse the mark reference, demanding a trailing space. Return a
2348 * pointer to the space.
2350 static uintmax_t parse_mark_ref_space(const char **p)
2352 uintmax_t mark;
2353 char *end;
2355 mark = parse_mark_ref(*p, &end);
2356 if (*end++ != ' ')
2357 die("Missing space after mark: %s", command_buf.buf);
2358 *p = end;
2359 return mark;
2362 static void file_change_m(const char *p, struct branch *b)
2364 static struct strbuf uq = STRBUF_INIT;
2365 const char *endp;
2366 struct object_entry *oe;
2367 struct object_id oid;
2368 uint16_t mode, inline_data = 0;
2370 p = get_mode(p, &mode);
2371 if (!p)
2372 die("Corrupt mode: %s", command_buf.buf);
2373 switch (mode) {
2374 case 0644:
2375 case 0755:
2376 mode |= S_IFREG;
2377 case S_IFREG | 0644:
2378 case S_IFREG | 0755:
2379 case S_IFLNK:
2380 case S_IFDIR:
2381 case S_IFGITLINK:
2382 /* ok */
2383 break;
2384 default:
2385 die("Corrupt mode: %s", command_buf.buf);
2388 if (*p == ':') {
2389 oe = find_mark(parse_mark_ref_space(&p));
2390 oidcpy(&oid, &oe->idx.oid);
2391 } else if (skip_prefix(p, "inline ", &p)) {
2392 inline_data = 1;
2393 oe = NULL; /* not used with inline_data, but makes gcc happy */
2394 } else {
2395 if (parse_oid_hex(p, &oid, &p))
2396 die("Invalid dataref: %s", command_buf.buf);
2397 oe = find_object(&oid);
2398 if (*p++ != ' ')
2399 die("Missing space after SHA1: %s", command_buf.buf);
2402 strbuf_reset(&uq);
2403 if (!unquote_c_style(&uq, p, &endp)) {
2404 if (*endp)
2405 die("Garbage after path in: %s", command_buf.buf);
2406 p = uq.buf;
2409 /* Git does not track empty, non-toplevel directories. */
2410 if (S_ISDIR(mode) && is_empty_tree_oid(&oid) && *p) {
2411 tree_content_remove(&b->branch_tree, p, NULL, 0);
2412 return;
2415 if (S_ISGITLINK(mode)) {
2416 if (inline_data)
2417 die("Git links cannot be specified 'inline': %s",
2418 command_buf.buf);
2419 else if (oe) {
2420 if (oe->type != OBJ_COMMIT)
2421 die("Not a commit (actually a %s): %s",
2422 typename(oe->type), command_buf.buf);
2425 * Accept the sha1 without checking; it expected to be in
2426 * another repository.
2428 } else if (inline_data) {
2429 if (S_ISDIR(mode))
2430 die("Directories cannot be specified 'inline': %s",
2431 command_buf.buf);
2432 if (p != uq.buf) {
2433 strbuf_addstr(&uq, p);
2434 p = uq.buf;
2436 read_next_command();
2437 parse_and_store_blob(&last_blob, &oid, 0);
2438 } else {
2439 enum object_type expected = S_ISDIR(mode) ?
2440 OBJ_TREE: OBJ_BLOB;
2441 enum object_type type = oe ? oe->type :
2442 sha1_object_info(oid.hash, NULL);
2443 if (type < 0)
2444 die("%s not found: %s",
2445 S_ISDIR(mode) ? "Tree" : "Blob",
2446 command_buf.buf);
2447 if (type != expected)
2448 die("Not a %s (actually a %s): %s",
2449 typename(expected), typename(type),
2450 command_buf.buf);
2453 if (!*p) {
2454 tree_content_replace(&b->branch_tree, &oid, mode, NULL);
2455 return;
2457 tree_content_set(&b->branch_tree, p, &oid, mode, NULL);
2460 static void file_change_d(const char *p, struct branch *b)
2462 static struct strbuf uq = STRBUF_INIT;
2463 const char *endp;
2465 strbuf_reset(&uq);
2466 if (!unquote_c_style(&uq, p, &endp)) {
2467 if (*endp)
2468 die("Garbage after path in: %s", command_buf.buf);
2469 p = uq.buf;
2471 tree_content_remove(&b->branch_tree, p, NULL, 1);
2474 static void file_change_cr(const char *s, struct branch *b, int rename)
2476 const char *d;
2477 static struct strbuf s_uq = STRBUF_INIT;
2478 static struct strbuf d_uq = STRBUF_INIT;
2479 const char *endp;
2480 struct tree_entry leaf;
2482 strbuf_reset(&s_uq);
2483 if (!unquote_c_style(&s_uq, s, &endp)) {
2484 if (*endp != ' ')
2485 die("Missing space after source: %s", command_buf.buf);
2486 } else {
2487 endp = strchr(s, ' ');
2488 if (!endp)
2489 die("Missing space after source: %s", command_buf.buf);
2490 strbuf_add(&s_uq, s, endp - s);
2492 s = s_uq.buf;
2494 endp++;
2495 if (!*endp)
2496 die("Missing dest: %s", command_buf.buf);
2498 d = endp;
2499 strbuf_reset(&d_uq);
2500 if (!unquote_c_style(&d_uq, d, &endp)) {
2501 if (*endp)
2502 die("Garbage after dest in: %s", command_buf.buf);
2503 d = d_uq.buf;
2506 memset(&leaf, 0, sizeof(leaf));
2507 if (rename)
2508 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2509 else
2510 tree_content_get(&b->branch_tree, s, &leaf, 1);
2511 if (!leaf.versions[1].mode)
2512 die("Path %s not in branch", s);
2513 if (!*d) { /* C "path/to/subdir" "" */
2514 tree_content_replace(&b->branch_tree,
2515 &leaf.versions[1].oid,
2516 leaf.versions[1].mode,
2517 leaf.tree);
2518 return;
2520 tree_content_set(&b->branch_tree, d,
2521 &leaf.versions[1].oid,
2522 leaf.versions[1].mode,
2523 leaf.tree);
2526 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2528 static struct strbuf uq = STRBUF_INIT;
2529 struct object_entry *oe;
2530 struct branch *s;
2531 struct object_id oid, commit_oid;
2532 char path[60];
2533 uint16_t inline_data = 0;
2534 unsigned char new_fanout;
2537 * When loading a branch, we don't traverse its tree to count the real
2538 * number of notes (too expensive to do this for all non-note refs).
2539 * This means that recently loaded notes refs might incorrectly have
2540 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2542 * Fix this by traversing the tree and counting the number of notes
2543 * when b->num_notes == 0. If the notes tree is truly empty, the
2544 * calculation should not take long.
2546 if (b->num_notes == 0 && *old_fanout == 0) {
2547 /* Invoke change_note_fanout() in "counting mode". */
2548 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2549 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2552 /* Now parse the notemodify command. */
2553 /* <dataref> or 'inline' */
2554 if (*p == ':') {
2555 oe = find_mark(parse_mark_ref_space(&p));
2556 oidcpy(&oid, &oe->idx.oid);
2557 } else if (skip_prefix(p, "inline ", &p)) {
2558 inline_data = 1;
2559 oe = NULL; /* not used with inline_data, but makes gcc happy */
2560 } else {
2561 if (parse_oid_hex(p, &oid, &p))
2562 die("Invalid dataref: %s", command_buf.buf);
2563 oe = find_object(&oid);
2564 if (*p++ != ' ')
2565 die("Missing space after SHA1: %s", command_buf.buf);
2568 /* <commit-ish> */
2569 s = lookup_branch(p);
2570 if (s) {
2571 if (is_null_oid(&s->oid))
2572 die("Can't add a note on empty branch.");
2573 oidcpy(&commit_oid, &s->oid);
2574 } else if (*p == ':') {
2575 uintmax_t commit_mark = parse_mark_ref_eol(p);
2576 struct object_entry *commit_oe = find_mark(commit_mark);
2577 if (commit_oe->type != OBJ_COMMIT)
2578 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2579 oidcpy(&commit_oid, &commit_oe->idx.oid);
2580 } else if (!get_oid(p, &commit_oid)) {
2581 unsigned long size;
2582 char *buf = read_object_with_reference(commit_oid.hash,
2583 commit_type, &size, commit_oid.hash);
2584 if (!buf || size < 46)
2585 die("Not a valid commit: %s", p);
2586 free(buf);
2587 } else
2588 die("Invalid ref name or SHA1 expression: %s", p);
2590 if (inline_data) {
2591 if (p != uq.buf) {
2592 strbuf_addstr(&uq, p);
2593 p = uq.buf;
2595 read_next_command();
2596 parse_and_store_blob(&last_blob, &oid, 0);
2597 } else if (oe) {
2598 if (oe->type != OBJ_BLOB)
2599 die("Not a blob (actually a %s): %s",
2600 typename(oe->type), command_buf.buf);
2601 } else if (!is_null_oid(&oid)) {
2602 enum object_type type = sha1_object_info(oid.hash, NULL);
2603 if (type < 0)
2604 die("Blob not found: %s", command_buf.buf);
2605 if (type != OBJ_BLOB)
2606 die("Not a blob (actually a %s): %s",
2607 typename(type), command_buf.buf);
2610 construct_path_with_fanout(oid_to_hex(&commit_oid), *old_fanout, path);
2611 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2612 b->num_notes--;
2614 if (is_null_oid(&oid))
2615 return; /* nothing to insert */
2617 b->num_notes++;
2618 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2619 construct_path_with_fanout(oid_to_hex(&commit_oid), new_fanout, path);
2620 tree_content_set(&b->branch_tree, path, &oid, S_IFREG | 0644, NULL);
2623 static void file_change_deleteall(struct branch *b)
2625 release_tree_content_recursive(b->branch_tree.tree);
2626 oidclr(&b->branch_tree.versions[0].oid);
2627 oidclr(&b->branch_tree.versions[1].oid);
2628 load_tree(&b->branch_tree);
2629 b->num_notes = 0;
2632 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2634 if (!buf || size < GIT_SHA1_HEXSZ + 6)
2635 die("Not a valid commit: %s", oid_to_hex(&b->oid));
2636 if (memcmp("tree ", buf, 5)
2637 || get_oid_hex(buf + 5, &b->branch_tree.versions[1].oid))
2638 die("The commit %s is corrupt", oid_to_hex(&b->oid));
2639 oidcpy(&b->branch_tree.versions[0].oid,
2640 &b->branch_tree.versions[1].oid);
2643 static void parse_from_existing(struct branch *b)
2645 if (is_null_oid(&b->oid)) {
2646 oidclr(&b->branch_tree.versions[0].oid);
2647 oidclr(&b->branch_tree.versions[1].oid);
2648 } else {
2649 unsigned long size;
2650 char *buf;
2652 buf = read_object_with_reference(b->oid.hash,
2653 commit_type, &size,
2654 b->oid.hash);
2655 parse_from_commit(b, buf, size);
2656 free(buf);
2660 static int parse_from(struct branch *b)
2662 const char *from;
2663 struct branch *s;
2664 struct object_id oid;
2666 if (!skip_prefix(command_buf.buf, "from ", &from))
2667 return 0;
2669 oidcpy(&oid, &b->branch_tree.versions[1].oid);
2671 s = lookup_branch(from);
2672 if (b == s)
2673 die("Can't create a branch from itself: %s", b->name);
2674 else if (s) {
2675 struct object_id *t = &s->branch_tree.versions[1].oid;
2676 oidcpy(&b->oid, &s->oid);
2677 oidcpy(&b->branch_tree.versions[0].oid, t);
2678 oidcpy(&b->branch_tree.versions[1].oid, t);
2679 } else if (*from == ':') {
2680 uintmax_t idnum = parse_mark_ref_eol(from);
2681 struct object_entry *oe = find_mark(idnum);
2682 if (oe->type != OBJ_COMMIT)
2683 die("Mark :%" PRIuMAX " not a commit", idnum);
2684 if (oidcmp(&b->oid, &oe->idx.oid)) {
2685 oidcpy(&b->oid, &oe->idx.oid);
2686 if (oe->pack_id != MAX_PACK_ID) {
2687 unsigned long size;
2688 char *buf = gfi_unpack_entry(oe, &size);
2689 parse_from_commit(b, buf, size);
2690 free(buf);
2691 } else
2692 parse_from_existing(b);
2694 } else if (!get_oid(from, &b->oid)) {
2695 parse_from_existing(b);
2696 if (is_null_oid(&b->oid))
2697 b->delete = 1;
2699 else
2700 die("Invalid ref name or SHA1 expression: %s", from);
2702 if (b->branch_tree.tree && oidcmp(&oid, &b->branch_tree.versions[1].oid)) {
2703 release_tree_content_recursive(b->branch_tree.tree);
2704 b->branch_tree.tree = NULL;
2707 read_next_command();
2708 return 1;
2711 static struct hash_list *parse_merge(unsigned int *count)
2713 struct hash_list *list = NULL, **tail = &list, *n;
2714 const char *from;
2715 struct branch *s;
2717 *count = 0;
2718 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2719 n = xmalloc(sizeof(*n));
2720 s = lookup_branch(from);
2721 if (s)
2722 oidcpy(&n->oid, &s->oid);
2723 else if (*from == ':') {
2724 uintmax_t idnum = parse_mark_ref_eol(from);
2725 struct object_entry *oe = find_mark(idnum);
2726 if (oe->type != OBJ_COMMIT)
2727 die("Mark :%" PRIuMAX " not a commit", idnum);
2728 oidcpy(&n->oid, &oe->idx.oid);
2729 } else if (!get_oid(from, &n->oid)) {
2730 unsigned long size;
2731 char *buf = read_object_with_reference(n->oid.hash,
2732 commit_type, &size, n->oid.hash);
2733 if (!buf || size < 46)
2734 die("Not a valid commit: %s", from);
2735 free(buf);
2736 } else
2737 die("Invalid ref name or SHA1 expression: %s", from);
2739 n->next = NULL;
2740 *tail = n;
2741 tail = &n->next;
2743 (*count)++;
2744 read_next_command();
2746 return list;
2749 static void parse_new_commit(const char *arg)
2751 static struct strbuf msg = STRBUF_INIT;
2752 struct branch *b;
2753 char *author = NULL;
2754 char *committer = NULL;
2755 struct hash_list *merge_list = NULL;
2756 unsigned int merge_count;
2757 unsigned char prev_fanout, new_fanout;
2758 const char *v;
2760 b = lookup_branch(arg);
2761 if (!b)
2762 b = new_branch(arg);
2764 read_next_command();
2765 parse_mark();
2766 if (skip_prefix(command_buf.buf, "author ", &v)) {
2767 author = parse_ident(v);
2768 read_next_command();
2770 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2771 committer = parse_ident(v);
2772 read_next_command();
2774 if (!committer)
2775 die("Expected committer but didn't get one");
2776 parse_data(&msg, 0, NULL);
2777 read_next_command();
2778 parse_from(b);
2779 merge_list = parse_merge(&merge_count);
2781 /* ensure the branch is active/loaded */
2782 if (!b->branch_tree.tree || !max_active_branches) {
2783 unload_one_branch();
2784 load_branch(b);
2787 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2789 /* file_change* */
2790 while (command_buf.len > 0) {
2791 if (skip_prefix(command_buf.buf, "M ", &v))
2792 file_change_m(v, b);
2793 else if (skip_prefix(command_buf.buf, "D ", &v))
2794 file_change_d(v, b);
2795 else if (skip_prefix(command_buf.buf, "R ", &v))
2796 file_change_cr(v, b, 1);
2797 else if (skip_prefix(command_buf.buf, "C ", &v))
2798 file_change_cr(v, b, 0);
2799 else if (skip_prefix(command_buf.buf, "N ", &v))
2800 note_change_n(v, b, &prev_fanout);
2801 else if (!strcmp("deleteall", command_buf.buf))
2802 file_change_deleteall(b);
2803 else if (skip_prefix(command_buf.buf, "ls ", &v))
2804 parse_ls(v, b);
2805 else {
2806 unread_command_buf = 1;
2807 break;
2809 if (read_next_command() == EOF)
2810 break;
2813 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2814 if (new_fanout != prev_fanout)
2815 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2817 /* build the tree and the commit */
2818 store_tree(&b->branch_tree);
2819 oidcpy(&b->branch_tree.versions[0].oid,
2820 &b->branch_tree.versions[1].oid);
2822 strbuf_reset(&new_data);
2823 strbuf_addf(&new_data, "tree %s\n",
2824 oid_to_hex(&b->branch_tree.versions[1].oid));
2825 if (!is_null_oid(&b->oid))
2826 strbuf_addf(&new_data, "parent %s\n",
2827 oid_to_hex(&b->oid));
2828 while (merge_list) {
2829 struct hash_list *next = merge_list->next;
2830 strbuf_addf(&new_data, "parent %s\n",
2831 oid_to_hex(&merge_list->oid));
2832 free(merge_list);
2833 merge_list = next;
2835 strbuf_addf(&new_data,
2836 "author %s\n"
2837 "committer %s\n"
2838 "\n",
2839 author ? author : committer, committer);
2840 strbuf_addbuf(&new_data, &msg);
2841 free(author);
2842 free(committer);
2844 if (!store_object(OBJ_COMMIT, &new_data, NULL, &b->oid, next_mark))
2845 b->pack_id = pack_id;
2846 b->last_commit = object_count_by_type[OBJ_COMMIT];
2849 static void parse_new_tag(const char *arg)
2851 static struct strbuf msg = STRBUF_INIT;
2852 const char *from;
2853 char *tagger;
2854 struct branch *s;
2855 struct tag *t;
2856 uintmax_t from_mark = 0;
2857 struct object_id oid;
2858 enum object_type type;
2859 const char *v;
2861 t = pool_alloc(sizeof(struct tag));
2862 memset(t, 0, sizeof(struct tag));
2863 t->name = pool_strdup(arg);
2864 if (last_tag)
2865 last_tag->next_tag = t;
2866 else
2867 first_tag = t;
2868 last_tag = t;
2869 read_next_command();
2871 /* from ... */
2872 if (!skip_prefix(command_buf.buf, "from ", &from))
2873 die("Expected from command, got %s", command_buf.buf);
2874 s = lookup_branch(from);
2875 if (s) {
2876 if (is_null_oid(&s->oid))
2877 die("Can't tag an empty branch.");
2878 oidcpy(&oid, &s->oid);
2879 type = OBJ_COMMIT;
2880 } else if (*from == ':') {
2881 struct object_entry *oe;
2882 from_mark = parse_mark_ref_eol(from);
2883 oe = find_mark(from_mark);
2884 type = oe->type;
2885 oidcpy(&oid, &oe->idx.oid);
2886 } else if (!get_oid(from, &oid)) {
2887 struct object_entry *oe = find_object(&oid);
2888 if (!oe) {
2889 type = sha1_object_info(oid.hash, NULL);
2890 if (type < 0)
2891 die("Not a valid object: %s", from);
2892 } else
2893 type = oe->type;
2894 } else
2895 die("Invalid ref name or SHA1 expression: %s", from);
2896 read_next_command();
2898 /* tagger ... */
2899 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2900 tagger = parse_ident(v);
2901 read_next_command();
2902 } else
2903 tagger = NULL;
2905 /* tag payload/message */
2906 parse_data(&msg, 0, NULL);
2908 /* build the tag object */
2909 strbuf_reset(&new_data);
2911 strbuf_addf(&new_data,
2912 "object %s\n"
2913 "type %s\n"
2914 "tag %s\n",
2915 oid_to_hex(&oid), typename(type), t->name);
2916 if (tagger)
2917 strbuf_addf(&new_data,
2918 "tagger %s\n", tagger);
2919 strbuf_addch(&new_data, '\n');
2920 strbuf_addbuf(&new_data, &msg);
2921 free(tagger);
2923 if (store_object(OBJ_TAG, &new_data, NULL, &t->oid, 0))
2924 t->pack_id = MAX_PACK_ID;
2925 else
2926 t->pack_id = pack_id;
2929 static void parse_reset_branch(const char *arg)
2931 struct branch *b;
2933 b = lookup_branch(arg);
2934 if (b) {
2935 oidclr(&b->oid);
2936 oidclr(&b->branch_tree.versions[0].oid);
2937 oidclr(&b->branch_tree.versions[1].oid);
2938 if (b->branch_tree.tree) {
2939 release_tree_content_recursive(b->branch_tree.tree);
2940 b->branch_tree.tree = NULL;
2943 else
2944 b = new_branch(arg);
2945 read_next_command();
2946 parse_from(b);
2947 if (command_buf.len > 0)
2948 unread_command_buf = 1;
2951 static void cat_blob_write(const char *buf, unsigned long size)
2953 if (write_in_full(cat_blob_fd, buf, size) != size)
2954 die_errno("Write to frontend failed");
2957 static void cat_blob(struct object_entry *oe, struct object_id *oid)
2959 struct strbuf line = STRBUF_INIT;
2960 unsigned long size;
2961 enum object_type type = 0;
2962 char *buf;
2964 if (!oe || oe->pack_id == MAX_PACK_ID) {
2965 buf = read_sha1_file(oid->hash, &type, &size);
2966 } else {
2967 type = oe->type;
2968 buf = gfi_unpack_entry(oe, &size);
2972 * Output based on batch_one_object() from cat-file.c.
2974 if (type <= 0) {
2975 strbuf_reset(&line);
2976 strbuf_addf(&line, "%s missing\n", oid_to_hex(oid));
2977 cat_blob_write(line.buf, line.len);
2978 strbuf_release(&line);
2979 free(buf);
2980 return;
2982 if (!buf)
2983 die("Can't read object %s", oid_to_hex(oid));
2984 if (type != OBJ_BLOB)
2985 die("Object %s is a %s but a blob was expected.",
2986 oid_to_hex(oid), typename(type));
2987 strbuf_reset(&line);
2988 strbuf_addf(&line, "%s %s %lu\n", oid_to_hex(oid),
2989 typename(type), size);
2990 cat_blob_write(line.buf, line.len);
2991 strbuf_release(&line);
2992 cat_blob_write(buf, size);
2993 cat_blob_write("\n", 1);
2994 if (oe && oe->pack_id == pack_id) {
2995 last_blob.offset = oe->idx.offset;
2996 strbuf_attach(&last_blob.data, buf, size, size);
2997 last_blob.depth = oe->depth;
2998 } else
2999 free(buf);
3002 static void parse_get_mark(const char *p)
3004 struct object_entry *oe = oe;
3005 char output[GIT_MAX_HEXSZ + 2];
3007 /* get-mark SP <object> LF */
3008 if (*p != ':')
3009 die("Not a mark: %s", p);
3011 oe = find_mark(parse_mark_ref_eol(p));
3012 if (!oe)
3013 die("Unknown mark: %s", command_buf.buf);
3015 xsnprintf(output, sizeof(output), "%s\n", oid_to_hex(&oe->idx.oid));
3016 cat_blob_write(output, GIT_SHA1_HEXSZ + 1);
3019 static void parse_cat_blob(const char *p)
3021 struct object_entry *oe = oe;
3022 struct object_id oid;
3024 /* cat-blob SP <object> LF */
3025 if (*p == ':') {
3026 oe = find_mark(parse_mark_ref_eol(p));
3027 if (!oe)
3028 die("Unknown mark: %s", command_buf.buf);
3029 oidcpy(&oid, &oe->idx.oid);
3030 } else {
3031 if (parse_oid_hex(p, &oid, &p))
3032 die("Invalid dataref: %s", command_buf.buf);
3033 if (*p)
3034 die("Garbage after SHA1: %s", command_buf.buf);
3035 oe = find_object(&oid);
3038 cat_blob(oe, &oid);
3041 static struct object_entry *dereference(struct object_entry *oe,
3042 struct object_id *oid)
3044 unsigned long size;
3045 char *buf = NULL;
3046 if (!oe) {
3047 enum object_type type = sha1_object_info(oid->hash, NULL);
3048 if (type < 0)
3049 die("object not found: %s", oid_to_hex(oid));
3050 /* cache it! */
3051 oe = insert_object(oid);
3052 oe->type = type;
3053 oe->pack_id = MAX_PACK_ID;
3054 oe->idx.offset = 1;
3056 switch (oe->type) {
3057 case OBJ_TREE: /* easy case. */
3058 return oe;
3059 case OBJ_COMMIT:
3060 case OBJ_TAG:
3061 break;
3062 default:
3063 die("Not a tree-ish: %s", command_buf.buf);
3066 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
3067 buf = gfi_unpack_entry(oe, &size);
3068 } else {
3069 enum object_type unused;
3070 buf = read_sha1_file(oid->hash, &unused, &size);
3072 if (!buf)
3073 die("Can't load object %s", oid_to_hex(oid));
3075 /* Peel one layer. */
3076 switch (oe->type) {
3077 case OBJ_TAG:
3078 if (size < GIT_SHA1_HEXSZ + strlen("object ") ||
3079 get_oid_hex(buf + strlen("object "), oid))
3080 die("Invalid SHA1 in tag: %s", command_buf.buf);
3081 break;
3082 case OBJ_COMMIT:
3083 if (size < GIT_SHA1_HEXSZ + strlen("tree ") ||
3084 get_oid_hex(buf + strlen("tree "), oid))
3085 die("Invalid SHA1 in commit: %s", command_buf.buf);
3088 free(buf);
3089 return find_object(oid);
3092 static struct object_entry *parse_treeish_dataref(const char **p)
3094 struct object_id oid;
3095 struct object_entry *e;
3097 if (**p == ':') { /* <mark> */
3098 e = find_mark(parse_mark_ref_space(p));
3099 if (!e)
3100 die("Unknown mark: %s", command_buf.buf);
3101 oidcpy(&oid, &e->idx.oid);
3102 } else { /* <sha1> */
3103 if (parse_oid_hex(*p, &oid, p))
3104 die("Invalid dataref: %s", command_buf.buf);
3105 e = find_object(&oid);
3106 if (*(*p)++ != ' ')
3107 die("Missing space after tree-ish: %s", command_buf.buf);
3110 while (!e || e->type != OBJ_TREE)
3111 e = dereference(e, &oid);
3112 return e;
3115 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3117 static struct strbuf line = STRBUF_INIT;
3119 /* See show_tree(). */
3120 const char *type =
3121 S_ISGITLINK(mode) ? commit_type :
3122 S_ISDIR(mode) ? tree_type :
3123 blob_type;
3125 if (!mode) {
3126 /* missing SP path LF */
3127 strbuf_reset(&line);
3128 strbuf_addstr(&line, "missing ");
3129 quote_c_style(path, &line, NULL, 0);
3130 strbuf_addch(&line, '\n');
3131 } else {
3132 /* mode SP type SP object_name TAB path LF */
3133 strbuf_reset(&line);
3134 strbuf_addf(&line, "%06o %s %s\t",
3135 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3136 quote_c_style(path, &line, NULL, 0);
3137 strbuf_addch(&line, '\n');
3139 cat_blob_write(line.buf, line.len);
3142 static void parse_ls(const char *p, struct branch *b)
3144 struct tree_entry *root = NULL;
3145 struct tree_entry leaf = {NULL};
3147 /* ls SP (<tree-ish> SP)? <path> */
3148 if (*p == '"') {
3149 if (!b)
3150 die("Not in a commit: %s", command_buf.buf);
3151 root = &b->branch_tree;
3152 } else {
3153 struct object_entry *e = parse_treeish_dataref(&p);
3154 root = new_tree_entry();
3155 oidcpy(&root->versions[1].oid, &e->idx.oid);
3156 if (!is_null_oid(&root->versions[1].oid))
3157 root->versions[1].mode = S_IFDIR;
3158 load_tree(root);
3160 if (*p == '"') {
3161 static struct strbuf uq = STRBUF_INIT;
3162 const char *endp;
3163 strbuf_reset(&uq);
3164 if (unquote_c_style(&uq, p, &endp))
3165 die("Invalid path: %s", command_buf.buf);
3166 if (*endp)
3167 die("Garbage after path in: %s", command_buf.buf);
3168 p = uq.buf;
3170 tree_content_get(root, p, &leaf, 1);
3172 * A directory in preparation would have a sha1 of zero
3173 * until it is saved. Save, for simplicity.
3175 if (S_ISDIR(leaf.versions[1].mode))
3176 store_tree(&leaf);
3178 print_ls(leaf.versions[1].mode, leaf.versions[1].oid.hash, p);
3179 if (leaf.tree)
3180 release_tree_content_recursive(leaf.tree);
3181 if (!b || root != &b->branch_tree)
3182 release_tree_entry(root);
3185 static void checkpoint(void)
3187 checkpoint_requested = 0;
3188 if (object_count) {
3189 cycle_packfile();
3190 dump_branches();
3191 dump_tags();
3192 dump_marks();
3196 static void parse_checkpoint(void)
3198 checkpoint_requested = 1;
3199 skip_optional_lf();
3202 static void parse_progress(void)
3204 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3205 fputc('\n', stdout);
3206 fflush(stdout);
3207 skip_optional_lf();
3210 static char* make_fast_import_path(const char *path)
3212 if (!relative_marks_paths || is_absolute_path(path))
3213 return xstrdup(path);
3214 return git_pathdup("info/fast-import/%s", path);
3217 static void option_import_marks(const char *marks,
3218 int from_stream, int ignore_missing)
3220 if (import_marks_file) {
3221 if (from_stream)
3222 die("Only one import-marks command allowed per stream");
3224 /* read previous mark file */
3225 if(!import_marks_file_from_stream)
3226 read_marks();
3229 import_marks_file = make_fast_import_path(marks);
3230 safe_create_leading_directories_const(import_marks_file);
3231 import_marks_file_from_stream = from_stream;
3232 import_marks_file_ignore_missing = ignore_missing;
3235 static void option_date_format(const char *fmt)
3237 if (!strcmp(fmt, "raw"))
3238 whenspec = WHENSPEC_RAW;
3239 else if (!strcmp(fmt, "rfc2822"))
3240 whenspec = WHENSPEC_RFC2822;
3241 else if (!strcmp(fmt, "now"))
3242 whenspec = WHENSPEC_NOW;
3243 else
3244 die("unknown --date-format argument %s", fmt);
3247 static unsigned long ulong_arg(const char *option, const char *arg)
3249 char *endptr;
3250 unsigned long rv = strtoul(arg, &endptr, 0);
3251 if (strchr(arg, '-') || endptr == arg || *endptr)
3252 die("%s: argument must be a non-negative integer", option);
3253 return rv;
3256 static void option_depth(const char *depth)
3258 max_depth = ulong_arg("--depth", depth);
3259 if (max_depth > MAX_DEPTH)
3260 die("--depth cannot exceed %u", MAX_DEPTH);
3263 static void option_active_branches(const char *branches)
3265 max_active_branches = ulong_arg("--active-branches", branches);
3268 static void option_export_marks(const char *marks)
3270 export_marks_file = make_fast_import_path(marks);
3271 safe_create_leading_directories_const(export_marks_file);
3274 static void option_cat_blob_fd(const char *fd)
3276 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3277 if (n > (unsigned long) INT_MAX)
3278 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3279 cat_blob_fd = (int) n;
3282 static void option_export_pack_edges(const char *edges)
3284 if (pack_edges)
3285 fclose(pack_edges);
3286 pack_edges = xfopen(edges, "a");
3289 static int parse_one_option(const char *option)
3291 if (skip_prefix(option, "max-pack-size=", &option)) {
3292 unsigned long v;
3293 if (!git_parse_ulong(option, &v))
3294 return 0;
3295 if (v < 8192) {
3296 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3297 v *= 1024 * 1024;
3298 } else if (v < 1024 * 1024) {
3299 warning("minimum max-pack-size is 1 MiB");
3300 v = 1024 * 1024;
3302 max_packsize = v;
3303 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3304 unsigned long v;
3305 if (!git_parse_ulong(option, &v))
3306 return 0;
3307 big_file_threshold = v;
3308 } else if (skip_prefix(option, "depth=", &option)) {
3309 option_depth(option);
3310 } else if (skip_prefix(option, "active-branches=", &option)) {
3311 option_active_branches(option);
3312 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3313 option_export_pack_edges(option);
3314 } else if (starts_with(option, "quiet")) {
3315 show_stats = 0;
3316 } else if (starts_with(option, "stats")) {
3317 show_stats = 1;
3318 } else {
3319 return 0;
3322 return 1;
3325 static int parse_one_feature(const char *feature, int from_stream)
3327 const char *arg;
3329 if (skip_prefix(feature, "date-format=", &arg)) {
3330 option_date_format(arg);
3331 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3332 option_import_marks(arg, from_stream, 0);
3333 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3334 option_import_marks(arg, from_stream, 1);
3335 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3336 option_export_marks(arg);
3337 } else if (!strcmp(feature, "get-mark")) {
3338 ; /* Don't die - this feature is supported */
3339 } else if (!strcmp(feature, "cat-blob")) {
3340 ; /* Don't die - this feature is supported */
3341 } else if (!strcmp(feature, "relative-marks")) {
3342 relative_marks_paths = 1;
3343 } else if (!strcmp(feature, "no-relative-marks")) {
3344 relative_marks_paths = 0;
3345 } else if (!strcmp(feature, "done")) {
3346 require_explicit_termination = 1;
3347 } else if (!strcmp(feature, "force")) {
3348 force_update = 1;
3349 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3350 ; /* do nothing; we have the feature */
3351 } else {
3352 return 0;
3355 return 1;
3358 static void parse_feature(const char *feature)
3360 if (seen_data_command)
3361 die("Got feature command '%s' after data command", feature);
3363 if (parse_one_feature(feature, 1))
3364 return;
3366 die("This version of fast-import does not support feature %s.", feature);
3369 static void parse_option(const char *option)
3371 if (seen_data_command)
3372 die("Got option command '%s' after data command", option);
3374 if (parse_one_option(option))
3375 return;
3377 die("This version of fast-import does not support option: %s", option);
3380 static void git_pack_config(void)
3382 int indexversion_value;
3383 int limit;
3384 unsigned long packsizelimit_value;
3386 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3387 if (max_depth > MAX_DEPTH)
3388 max_depth = MAX_DEPTH;
3390 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3391 pack_idx_opts.version = indexversion_value;
3392 if (pack_idx_opts.version > 2)
3393 git_die_config("pack.indexversion",
3394 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3396 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3397 max_packsize = packsizelimit_value;
3399 if (!git_config_get_int("fastimport.unpacklimit", &limit))
3400 unpack_limit = limit;
3401 else if (!git_config_get_int("transfer.unpacklimit", &limit))
3402 unpack_limit = limit;
3404 git_config(git_default_config, NULL);
3407 static const char fast_import_usage[] =
3408 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3410 static void parse_argv(void)
3412 unsigned int i;
3414 for (i = 1; i < global_argc; i++) {
3415 const char *a = global_argv[i];
3417 if (*a != '-' || !strcmp(a, "--"))
3418 break;
3420 if (!skip_prefix(a, "--", &a))
3421 die("unknown option %s", a);
3423 if (parse_one_option(a))
3424 continue;
3426 if (parse_one_feature(a, 0))
3427 continue;
3429 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3430 option_cat_blob_fd(a);
3431 continue;
3434 die("unknown option --%s", a);
3436 if (i != global_argc)
3437 usage(fast_import_usage);
3439 seen_data_command = 1;
3440 if (import_marks_file)
3441 read_marks();
3444 int cmd_main(int argc, const char **argv)
3446 unsigned int i;
3448 if (argc == 2 && !strcmp(argv[1], "-h"))
3449 usage(fast_import_usage);
3451 setup_git_directory();
3452 reset_pack_idx_option(&pack_idx_opts);
3453 git_pack_config();
3455 alloc_objects(object_entry_alloc);
3456 strbuf_init(&command_buf, 0);
3457 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3458 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3459 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3460 marks = pool_calloc(1, sizeof(struct mark_set));
3462 global_argc = argc;
3463 global_argv = argv;
3465 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3466 for (i = 0; i < (cmd_save - 1); i++)
3467 rc_free[i].next = &rc_free[i + 1];
3468 rc_free[cmd_save - 1].next = NULL;
3470 prepare_packed_git();
3471 start_packfile();
3472 set_die_routine(die_nicely);
3473 set_checkpoint_signal();
3474 while (read_next_command() != EOF) {
3475 const char *v;
3476 if (!strcmp("blob", command_buf.buf))
3477 parse_new_blob();
3478 else if (skip_prefix(command_buf.buf, "ls ", &v))
3479 parse_ls(v, NULL);
3480 else if (skip_prefix(command_buf.buf, "commit ", &v))
3481 parse_new_commit(v);
3482 else if (skip_prefix(command_buf.buf, "tag ", &v))
3483 parse_new_tag(v);
3484 else if (skip_prefix(command_buf.buf, "reset ", &v))
3485 parse_reset_branch(v);
3486 else if (!strcmp("checkpoint", command_buf.buf))
3487 parse_checkpoint();
3488 else if (!strcmp("done", command_buf.buf))
3489 break;
3490 else if (starts_with(command_buf.buf, "progress "))
3491 parse_progress();
3492 else if (skip_prefix(command_buf.buf, "feature ", &v))
3493 parse_feature(v);
3494 else if (skip_prefix(command_buf.buf, "option git ", &v))
3495 parse_option(v);
3496 else if (starts_with(command_buf.buf, "option "))
3497 /* ignore non-git options*/;
3498 else
3499 die("Unsupported command: %s", command_buf.buf);
3501 if (checkpoint_requested)
3502 checkpoint();
3505 /* argv hasn't been parsed yet, do so */
3506 if (!seen_data_command)
3507 parse_argv();
3509 if (require_explicit_termination && feof(stdin))
3510 die("stream ends early");
3512 end_packfile();
3514 dump_branches();
3515 dump_tags();
3516 unkeep_all_packs();
3517 dump_marks();
3519 if (pack_edges)
3520 fclose(pack_edges);
3522 if (show_stats) {
3523 uintmax_t total_count = 0, duplicate_count = 0;
3524 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3525 total_count += object_count_by_type[i];
3526 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3527 duplicate_count += duplicate_count_by_type[i];
3529 fprintf(stderr, "%s statistics:\n", argv[0]);
3530 fprintf(stderr, "---------------------------------------------------------------------\n");
3531 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3532 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3533 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB], delta_count_attempts_by_type[OBJ_BLOB]);
3534 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE], delta_count_attempts_by_type[OBJ_TREE]);
3535 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT], delta_count_attempts_by_type[OBJ_COMMIT]);
3536 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG], delta_count_attempts_by_type[OBJ_TAG]);
3537 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3538 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3539 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3540 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3541 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3542 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3543 fprintf(stderr, "---------------------------------------------------------------------\n");
3544 pack_report();
3545 fprintf(stderr, "---------------------------------------------------------------------\n");
3546 fprintf(stderr, "\n");
3549 return failure ? 1 : 0;