Raw
1 /*
2 * GIT - The information manager from hell
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 */
6
7 #define USE_THE_REPOSITORY_VARIABLE
8 #define DISABLE_SIGN_COMPARE_WARNINGS
9
10 #include "builtin.h"
11 #include "config.h"
12 #include "convert.h"
13 #include "diff.h"
14 #include "environment.h"
15 #include "gettext.h"
16 #include "hex.h"
17 #include "ident.h"
18 #include "list-objects-filter-options.h"
19 #include "parse-options.h"
20 #include "userdiff.h"
21 #include "oid-array.h"
22 #include "packfile.h"
23 #include "pack-bitmap.h"
24 #include "object-file.h"
25 #include "object-name.h"
26 #include "odb.h"
27 #include "odb/streaming.h"
28 #include "replace-object.h"
29 #include "promisor-remote.h"
30 #include "mailmap.h"
31 #include "write-or-die.h"
32
33 enum batch_mode {
34 BATCH_MODE_CONTENTS,
35 BATCH_MODE_INFO,
36 BATCH_MODE_QUEUE_AND_DISPATCH,
37 };
38
39 struct batch_options {
40 struct list_objects_filter_options objects_filter;
41 int enabled;
42 int follow_symlinks;
43 enum batch_mode batch_mode;
44 int buffer_output;
45 int all_objects;
46 int unordered;
47 int transform_mode; /* may be 'w' or 'c' for --filters or --textconv */
48 char input_delim;
49 char output_delim;
50 const char *format;
51 };
52
53 static const char *force_path;
54
55 static struct string_list mailmap = STRING_LIST_INIT_NODUP;
56 static int use_mailmap;
57
58 static char *replace_idents_using_mailmap(char *, size_t *);
59
60 /*
61 * The mailmap is initialized with .strdup_strings set to 0,
62 * but read_mailmap() sets the bit to 1 (this is true even when
63 * not a single mailmap entry is read), so it can be used for
64 * lazy loading.
65 */
66 static void load_mailmap(void)
67 {
68 if (mailmap.strdup_strings)
69 return;
70
71 read_mailmap(the_repository, &mailmap);
72 }
73
74 static char *replace_idents_using_mailmap(char *object_buf, size_t *size)
75 {
76 struct strbuf sb = STRBUF_INIT;
77 const char *headers[] = { "author ", "committer ", "tagger ", NULL };
78
79 strbuf_attach(&sb, object_buf, *size, *size + 1);
80 apply_mailmap_to_header(&sb, headers, &mailmap);
81 *size = sb.len;
82 return strbuf_detach(&sb, NULL);
83 }
84
85 static int filter_object(const char *path, unsigned mode,
86 const struct object_id *oid,
87 char **buf, unsigned long *size)
88 {
89 enum object_type type;
90
91 *buf = odb_read_object(the_repository->objects, oid, &type, size);
92 if (!*buf)
93 return error(_("cannot read object %s '%s'"),
94 oid_to_hex(oid), path);
95 if ((type == OBJ_BLOB) && S_ISREG(mode)) {
96 struct strbuf strbuf = STRBUF_INIT;
97 struct checkout_metadata meta;
98
99 init_checkout_metadata(&meta, NULL, NULL, oid);
100 if (convert_to_working_tree(the_repository->index, path, *buf, *size, &strbuf, &meta)) {
101 free(*buf);
102 *size = strbuf.len;
103 *buf = strbuf_detach(&strbuf, NULL);
104 }
105 }
106
107 return 0;
108 }
109
110 static int stream_blob(const struct object_id *oid)
111 {
112 if (odb_stream_blob_to_fd(the_repository->objects, 1, oid, NULL, 0))
113 die("unable to stream %s to stdout", oid_to_hex(oid));
114 return 0;
115 }
116
117 static int cat_one_file(int opt, const char *exp_type, const char *obj_name)
118 {
119 int ret;
120 struct object_id oid;
121 enum object_type type;
122 char *buf;
123 unsigned long size;
124 struct object_context obj_context = {0};
125 struct object_info oi = OBJECT_INFO_INIT;
126 unsigned flags = OBJECT_INFO_LOOKUP_REPLACE;
127 unsigned get_oid_flags =
128 GET_OID_RECORD_PATH |
129 GET_OID_ONLY_TO_DIE |
130 GET_OID_HASH_ANY;
131 const char *path = force_path;
132 const int opt_cw = (opt == 'c' || opt == 'w');
133 if (!path && opt_cw)
134 get_oid_flags |= GET_OID_REQUIRE_PATH;
135
136 if (get_oid_with_context(the_repository, obj_name, get_oid_flags, &oid,
137 &obj_context))
138 die("Not a valid object name %s", obj_name);
139
140 if (!path)
141 path = obj_context.path;
142 if (obj_context.mode == S_IFINVALID)
143 obj_context.mode = 0100644;
144
145 buf = NULL;
146 switch (opt) {
147 case 't':
148 oi.typep = &type;
149 if (odb_read_object_info_extended(the_repository->objects, &oid, &oi, flags) < 0)
150 die("git cat-file: could not get object info");
151 printf("%s\n", type_name(type));
152 ret = 0;
153 goto cleanup;
154
155 case 's':
156 oi.sizep = &size;
157
158 if (use_mailmap) {
159 oi.typep = &type;
160 oi.contentp = (void**)&buf;
161 }
162
163 if (odb_read_object_info_extended(the_repository->objects, &oid, &oi, flags) < 0)
164 die("git cat-file: could not get object info");
165
166 if (use_mailmap && (type == OBJ_COMMIT || type == OBJ_TAG)) {
167 size_t s = size;
168 buf = replace_idents_using_mailmap(buf, &s);
169 size = cast_size_t_to_ulong(s);
170 }
171
172 printf("%"PRIuMAX"\n", (uintmax_t)size);
173 ret = 0;
174 goto cleanup;
175
176 case 'e':
177 ret = !odb_has_object(the_repository->objects, &oid,
178 ODB_HAS_OBJECT_RECHECK_PACKED | ODB_HAS_OBJECT_FETCH_PROMISOR);
179 goto cleanup;
180
181 case 'w':
182
183 if (filter_object(path, obj_context.mode,
184 &oid, &buf, &size)) {
185 ret = -1;
186 goto cleanup;
187 }
188 break;
189
190 case 'c':
191 if (textconv_object(the_repository, path, obj_context.mode,
192 &oid, 1, &buf, &size))
193 break;
194 /* else fallthrough */
195
196 case 'p':
197 type = odb_read_object_info(the_repository->objects, &oid, NULL);
198 if (type < 0)
199 die("Not a valid object name %s", obj_name);
200
201 /* custom pretty-print here */
202 if (type == OBJ_TREE) {
203 const char *ls_args[3] = { NULL };
204 ls_args[0] = "ls-tree";
205 ls_args[1] = obj_name;
206 ret = cmd_ls_tree(2, ls_args, NULL, the_repository);
207 goto cleanup;
208 }
209
210 if (type == OBJ_BLOB) {
211 ret = stream_blob(&oid);
212 goto cleanup;
213 }
214 buf = odb_read_object(the_repository->objects, &oid,
215 &type, &size);
216 if (!buf)
217 die("Cannot read object %s", obj_name);
218
219 if (use_mailmap) {
220 size_t s = size;
221 buf = replace_idents_using_mailmap(buf, &s);
222 size = cast_size_t_to_ulong(s);
223 }
224
225 /* otherwise just spit out the data */
226 break;
227
228 case 0:
229 {
230 enum object_type exp_type_id = type_from_string(exp_type);
231
232 if (exp_type_id == OBJ_BLOB) {
233 struct object_id blob_oid;
234 if (odb_read_object_info(the_repository->objects,
235 &oid, NULL) == OBJ_TAG) {
236 char *buffer = odb_read_object(the_repository->objects,
237 &oid, &type, &size);
238 const char *target;
239
240 if (!buffer)
241 die(_("unable to read %s"), oid_to_hex(&oid));
242
243 if (!skip_prefix(buffer, "object ", &target) ||
244 get_oid_hex_algop(target, &blob_oid,
245 &hash_algos[oid.algo]))
246 die("%s not a valid tag", oid_to_hex(&oid));
247 free(buffer);
248 } else
249 oidcpy(&blob_oid, &oid);
250
251 if (odb_read_object_info(the_repository->objects,
252 &blob_oid, NULL) == OBJ_BLOB) {
253 ret = stream_blob(&blob_oid);
254 goto cleanup;
255 }
256 /*
257 * we attempted to dereference a tag to a blob
258 * and failed; there may be new dereference
259 * mechanisms this code is not aware of.
260 * fall-back to the usual case.
261 */
262 }
263 buf = odb_read_object_peeled(the_repository->objects, &oid,
264 exp_type_id, &size, NULL);
265
266 if (use_mailmap) {
267 size_t s = size;
268 buf = replace_idents_using_mailmap(buf, &s);
269 size = cast_size_t_to_ulong(s);
270 }
271 break;
272 }
273 default:
274 die("git cat-file: unknown option: %s", exp_type);
275 }
276
277 if (!buf)
278 die("git cat-file %s: bad file", obj_name);
279
280 write_or_die(1, buf, size);
281 ret = 0;
282 cleanup:
283 free(buf);
284 object_context_release(&obj_context);
285 return ret;
286 }
287
288 struct expand_data {
289 struct object_id oid;
290 enum object_type type;
291 unsigned long size;
292 unsigned short mode;
293 off_t disk_size;
294 const char *rest;
295 struct object_id delta_base_oid;
296
297 /*
298 * If mark_query is true, we do not expand anything, but rather
299 * just mark the object_info with items we wish to query.
300 */
301 int mark_query;
302
303 /*
304 * Whether to split the input on whitespace before feeding it to
305 * get_sha1; this is decided during the mark_query phase based on
306 * whether we have a %(rest) token in our format.
307 */
308 int split_on_whitespace;
309
310 /*
311 * After a mark_query run, this object_info is set up to be
312 * passed to odb_read_object_info_extended. It will point to the data
313 * elements above, so you can retrieve the response from there.
314 */
315 struct object_info info;
316
317 /*
318 * This flag will be true if the requested batch format and options
319 * don't require us to call oid_object_info, which can then be
320 * optimized out.
321 */
322 unsigned skip_object_info : 1;
323 };
324 #define EXPAND_DATA_INIT { .mode = S_IFINVALID }
325
326 static int is_atom(const char *atom, const char *s, int slen)
327 {
328 int alen = strlen(atom);
329 return alen == slen && !memcmp(atom, s, alen);
330 }
331
332 static int expand_atom(struct strbuf *sb, const char *atom, int len,
333 struct expand_data *data)
334 {
335 if (is_atom("objectname", atom, len)) {
336 if (!data->mark_query)
337 strbuf_add_oid_hex(sb, &data->oid);
338 } else if (is_atom("objecttype", atom, len)) {
339 if (data->mark_query)
340 data->info.typep = &data->type;
341 else
342 strbuf_addstr(sb, type_name(data->type));
343 } else if (is_atom("objectsize", atom, len)) {
344 if (data->mark_query)
345 data->info.sizep = &data->size;
346 else
347 strbuf_add_uint(sb, data->size);
348 } else if (is_atom("objectsize:disk", atom, len)) {
349 if (data->mark_query)
350 data->info.disk_sizep = &data->disk_size;
351 else
352 strbuf_add_uint(sb, data->disk_size);
353 } else if (is_atom("rest", atom, len)) {
354 if (data->mark_query)
355 data->split_on_whitespace = 1;
356 else if (data->rest)
357 strbuf_addstr(sb, data->rest);
358 } else if (is_atom("deltabase", atom, len)) {
359 if (data->mark_query)
360 data->info.delta_base_oid = &data->delta_base_oid;
361 else
362 strbuf_add_oid_hex(sb, &data->delta_base_oid);
363 } else if (is_atom("objectmode", atom, len)) {
364 if (!data->mark_query && !(S_IFINVALID == data->mode))
365 strbuf_addf(sb, "%06o", data->mode);
366 } else
367 return 0;
368 return 1;
369 }
370
371 static void expand_format(struct strbuf *sb, const char *start,
372 struct expand_data *data)
373 {
374 while (strbuf_expand_step(sb, &start)) {
375 const char *end;
376
377 if (skip_prefix(start, "%", &start) || *start != '(')
378 strbuf_addch(sb, '%');
379 else if ((end = strchr(start + 1, ')')) &&
380 expand_atom(sb, start + 1, end - start - 1, data))
381 start = end + 1;
382 else
383 strbuf_expand_bad_format(start, "cat-file");
384 }
385 }
386
387 static void batch_write(struct batch_options *opt, const void *data, int len)
388 {
389 if (opt->buffer_output) {
390 if (fwrite(data, 1, len, stdout) != len)
391 die_errno("unable to write to stdout");
392 } else
393 write_or_die(1, data, len);
394 }
395
396 static void print_object_or_die(struct batch_options *opt, struct expand_data *data)
397 {
398 const struct object_id *oid = &data->oid;
399
400 assert(data->info.typep);
401
402 if (data->type == OBJ_BLOB) {
403 if (opt->buffer_output)
404 fflush(stdout);
405 if (opt->transform_mode) {
406 char *contents;
407 unsigned long size;
408
409 if (!data->rest)
410 die("missing path for '%s'", oid_to_hex(oid));
411
412 if (opt->transform_mode == 'w') {
413 if (filter_object(data->rest, 0100644, oid,
414 &contents, &size))
415 die("could not convert '%s' %s",
416 oid_to_hex(oid), data->rest);
417 } else if (opt->transform_mode == 'c') {
418 enum object_type type;
419 if (!textconv_object(the_repository,
420 data->rest, 0100644, oid,
421 1, &contents, &size))
422 contents = odb_read_object(the_repository->objects,
423 oid, &type, &size);
424 if (!contents)
425 die("could not convert '%s' %s",
426 oid_to_hex(oid), data->rest);
427 } else
428 BUG("invalid transform_mode: %c", opt->transform_mode);
429 batch_write(opt, contents, size);
430 free(contents);
431 } else {
432 stream_blob(oid);
433 }
434 }
435 else {
436 enum object_type type;
437 unsigned long size;
438 void *contents;
439
440 contents = odb_read_object(the_repository->objects, oid,
441 &type, &size);
442 if (!contents)
443 die("object %s disappeared", oid_to_hex(oid));
444
445 if (use_mailmap) {
446 size_t s = size;
447 contents = replace_idents_using_mailmap(contents, &s);
448 size = cast_size_t_to_ulong(s);
449 }
450
451 if (type != data->type)
452 die("object %s changed type!?", oid_to_hex(oid));
453 if (data->info.sizep && size != data->size && !use_mailmap)
454 die("object %s changed size!?", oid_to_hex(oid));
455
456 batch_write(opt, contents, size);
457 free(contents);
458 }
459 }
460
461 static void print_default_format(struct strbuf *scratch, struct expand_data *data,
462 struct batch_options *opt)
463 {
464 strbuf_addf(scratch, "%s %s %"PRIuMAX"%c", oid_to_hex(&data->oid),
465 type_name(data->type),
466 (uintmax_t)data->size, opt->output_delim);
467 }
468
469 static void report_object_status(struct batch_options *opt,
470 const char *obj_name,
471 const struct object_id *oid,
472 const char *status)
473 {
474 printf("%s %s%c", obj_name ? obj_name : oid_to_hex(oid),
475 status, opt->output_delim);
476 fflush(stdout);
477 }
478
479 /*
480 * If "pack" is non-NULL, then "offset" is the byte offset within the pack from
481 * which the object may be accessed (though note that we may also rely on
482 * data->oid, too). If "pack" is NULL, then offset is ignored.
483 */
484 static void batch_object_write(const char *obj_name,
485 struct strbuf *scratch,
486 struct batch_options *opt,
487 struct expand_data *data,
488 struct packed_git *pack,
489 off_t offset)
490 {
491 if (!data->skip_object_info) {
492 int ret;
493
494 if (use_mailmap ||
495 opt->objects_filter.choice == LOFC_BLOB_NONE ||
496 opt->objects_filter.choice == LOFC_BLOB_LIMIT ||
497 opt->objects_filter.choice == LOFC_OBJECT_TYPE)
498 data->info.typep = &data->type;
499 if (opt->objects_filter.choice == LOFC_BLOB_LIMIT)
500 data->info.sizep = &data->size;
501
502 if (pack)
503 ret = packed_object_info(pack, offset, &data->info);
504 else
505 ret = odb_read_object_info_extended(the_repository->objects,
506 &data->oid, &data->info,
507 OBJECT_INFO_LOOKUP_REPLACE);
508 if (ret < 0) {
509 if (data->mode == S_IFGITLINK)
510 report_object_status(opt, NULL, &data->oid, "submodule");
511 else
512 report_object_status(opt, obj_name, &data->oid, "missing");
513 return;
514 }
515
516 switch (opt->objects_filter.choice) {
517 case LOFC_DISABLED:
518 break;
519 case LOFC_BLOB_NONE:
520 if (data->type == OBJ_BLOB) {
521 if (!opt->all_objects)
522 report_object_status(opt, obj_name,
523 &data->oid, "excluded");
524 return;
525 }
526 break;
527 case LOFC_BLOB_LIMIT:
528 if (data->type == OBJ_BLOB &&
529 data->size >= opt->objects_filter.blob_limit_value) {
530 if (!opt->all_objects)
531 report_object_status(opt, obj_name,
532 &data->oid, "excluded");
533 return;
534 }
535 break;
536 case LOFC_OBJECT_TYPE:
537 if (data->type != opt->objects_filter.object_type) {
538 if (!opt->all_objects)
539 report_object_status(opt, obj_name,
540 &data->oid, "excluded");
541 return;
542 }
543 break;
544 default:
545 BUG("unsupported objects filter");
546 }
547
548 if (use_mailmap && (data->type == OBJ_COMMIT || data->type == OBJ_TAG)) {
549 size_t s = data->size;
550 char *buf = NULL;
551
552 buf = odb_read_object(the_repository->objects, &data->oid,
553 &data->type, &data->size);
554 if (!buf)
555 die(_("unable to read %s"), oid_to_hex(&data->oid));
556 buf = replace_idents_using_mailmap(buf, &s);
557 data->size = cast_size_t_to_ulong(s);
558
559 free(buf);
560 }
561 }
562
563 strbuf_reset(scratch);
564
565 if (!opt->format) {
566 print_default_format(scratch, data, opt);
567 } else {
568 expand_format(scratch, opt->format, data);
569 strbuf_addch(scratch, opt->output_delim);
570 }
571
572 batch_write(opt, scratch->buf, scratch->len);
573
574 if (opt->batch_mode == BATCH_MODE_CONTENTS) {
575 print_object_or_die(opt, data);
576 batch_write(opt, &opt->output_delim, 1);
577 }
578 }
579
580 static void batch_one_object(const char *obj_name,
581 struct strbuf *scratch,
582 struct batch_options *opt,
583 struct expand_data *data)
584 {
585 struct object_context ctx = {0};
586 int flags =
587 GET_OID_HASH_ANY |
588 (opt->follow_symlinks ? GET_OID_FOLLOW_SYMLINKS : 0);
589 enum get_oid_result result;
590
591 result = get_oid_with_context(the_repository, obj_name,
592 flags, &data->oid, &ctx);
593 if (result != FOUND) {
594 switch (result) {
595 case MISSING_OBJECT:
596 report_object_status(opt, obj_name, &data->oid, "missing");
597 break;
598 case SHORT_NAME_AMBIGUOUS:
599 report_object_status(opt, obj_name, &data->oid, "ambiguous");
600 break;
601 case DANGLING_SYMLINK:
602 printf("dangling %"PRIuMAX"%c%s%c",
603 (uintmax_t)strlen(obj_name),
604 opt->output_delim, obj_name, opt->output_delim);
605 break;
606 case SYMLINK_LOOP:
607 printf("loop %"PRIuMAX"%c%s%c",
608 (uintmax_t)strlen(obj_name),
609 opt->output_delim, obj_name, opt->output_delim);
610 break;
611 case NOT_DIR:
612 printf("notdir %"PRIuMAX"%c%s%c",
613 (uintmax_t)strlen(obj_name),
614 opt->output_delim, obj_name, opt->output_delim);
615 break;
616 default:
617 BUG("unknown get_sha1_with_context result %d\n",
618 result);
619 break;
620 }
621 fflush(stdout);
622
623 goto out;
624 }
625
626 if (ctx.mode == 0) {
627 printf("symlink %"PRIuMAX"%c%s%c",
628 (uintmax_t)ctx.symlink_path.len,
629 opt->output_delim, ctx.symlink_path.buf, opt->output_delim);
630 fflush(stdout);
631 goto out;
632 }
633
634 data->mode = ctx.mode;
635 batch_object_write(obj_name, scratch, opt, data, NULL, 0);
636
637 out:
638 object_context_release(&ctx);
639 }
640
641 struct object_cb_data {
642 struct batch_options *opt;
643 struct expand_data *expand;
644 struct oidset *seen;
645 struct strbuf *scratch;
646 };
647
648 static int batch_object_cb(const struct object_id *oid, void *vdata)
649 {
650 struct object_cb_data *data = vdata;
651 oidcpy(&data->expand->oid, oid);
652 batch_object_write(NULL, data->scratch, data->opt, data->expand,
653 NULL, 0);
654 return 0;
655 }
656
657 static int collect_object(const struct object_id *oid,
658 struct packed_git *pack UNUSED,
659 off_t offset UNUSED,
660 void *data)
661 {
662 oid_array_append(data, oid);
663 return 0;
664 }
665
666 static int batch_unordered_object(const struct object_id *oid,
667 struct packed_git *pack,
668 off_t offset,
669 void *vdata)
670 {
671 struct object_cb_data *data = vdata;
672
673 if (oidset_insert(data->seen, oid))
674 return 0;
675
676 oidcpy(&data->expand->oid, oid);
677 batch_object_write(NULL, data->scratch, data->opt, data->expand,
678 pack, offset);
679 return 0;
680 }
681
682 typedef void (*parse_cmd_fn_t)(struct batch_options *, const char *,
683 struct strbuf *, struct expand_data *);
684
685 struct queued_cmd {
686 parse_cmd_fn_t fn;
687 char *line;
688 };
689
690 static void parse_cmd_contents(struct batch_options *opt,
691 const char *line,
692 struct strbuf *output,
693 struct expand_data *data)
694 {
695 opt->batch_mode = BATCH_MODE_CONTENTS;
696 batch_one_object(line, output, opt, data);
697 }
698
699 static void parse_cmd_info(struct batch_options *opt,
700 const char *line,
701 struct strbuf *output,
702 struct expand_data *data)
703 {
704 opt->batch_mode = BATCH_MODE_INFO;
705 batch_one_object(line, output, opt, data);
706 }
707
708 static void parse_cmd_mailmap(struct batch_options *opt UNUSED,
709 const char *line,
710 struct strbuf *output UNUSED,
711 struct expand_data *data UNUSED)
712 {
713 use_mailmap = git_parse_maybe_bool(line);
714
715 if (use_mailmap < 0)
716 die(_("mailmap: invalid boolean '%s'"), line);
717
718 if (use_mailmap)
719 load_mailmap();
720 }
721
722 static void dispatch_calls(struct batch_options *opt,
723 struct strbuf *output,
724 struct expand_data *data,
725 struct queued_cmd *cmd,
726 int nr)
727 {
728 int i;
729
730 if (!opt->buffer_output)
731 die(_("flush is only for --buffer mode"));
732
733 for (i = 0; i < nr; i++)
734 cmd[i].fn(opt, cmd[i].line, output, data);
735
736 fflush(stdout);
737 }
738
739 static void free_cmds(struct queued_cmd *cmd, size_t *nr)
740 {
741 size_t i;
742
743 for (i = 0; i < *nr; i++)
744 FREE_AND_NULL(cmd[i].line);
745
746 *nr = 0;
747 }
748
749
750 static const struct parse_cmd {
751 const char *name;
752 parse_cmd_fn_t fn;
753 unsigned takes_args;
754 } commands[] = {
755 { "contents", parse_cmd_contents, 1 },
756 { "info", parse_cmd_info, 1 },
757 { "flush", NULL, 0 },
758 { "mailmap", parse_cmd_mailmap, 1 },
759 };
760
761 static void batch_objects_command(struct batch_options *opt,
762 struct strbuf *output,
763 struct expand_data *data)
764 {
765 struct strbuf input = STRBUF_INIT;
766 struct queued_cmd *queued_cmd = NULL;
767 size_t alloc = 0, nr = 0;
768
769 while (strbuf_getdelim_strip_crlf(&input, stdin, opt->input_delim) != EOF) {
770 int i;
771 const struct parse_cmd *cmd = NULL;
772 const char *p = NULL, *cmd_end;
773 struct queued_cmd call = {0};
774
775 if (!input.len)
776 die(_("empty command in input"));
777 if (isspace(*input.buf))
778 die(_("whitespace before command: '%s'"), input.buf);
779
780 for (i = 0; i < ARRAY_SIZE(commands); i++) {
781 if (!skip_prefix(input.buf, commands[i].name, &cmd_end))
782 continue;
783
784 cmd = &commands[i];
785 if (cmd->takes_args) {
786 if (*cmd_end != ' ')
787 die(_("%s requires arguments"),
788 commands[i].name);
789
790 p = cmd_end + 1;
791 } else if (*cmd_end) {
792 die(_("%s takes no arguments"),
793 commands[i].name);
794 }
795
796 break;
797 }
798
799 if (!cmd)
800 die(_("unknown command: '%s'"), input.buf);
801
802 if (!strcmp(cmd->name, "flush")) {
803 dispatch_calls(opt, output, data, queued_cmd, nr);
804 free_cmds(queued_cmd, &nr);
805 } else if (!opt->buffer_output) {
806 cmd->fn(opt, p, output, data);
807 } else {
808 ALLOC_GROW(queued_cmd, nr + 1, alloc);
809 call.fn = cmd->fn;
810 call.line = xstrdup_or_null(p);
811 queued_cmd[nr++] = call;
812 }
813 }
814
815 if (opt->buffer_output &&
816 nr &&
817 !git_env_bool("GIT_TEST_CAT_FILE_NO_FLUSH_ON_EXIT", 0)) {
818 dispatch_calls(opt, output, data, queued_cmd, nr);
819 free_cmds(queued_cmd, &nr);
820 }
821
822 free_cmds(queued_cmd, &nr);
823 free(queued_cmd);
824 strbuf_release(&input);
825 }
826
827 #define DEFAULT_FORMAT "%(objectname) %(objecttype) %(objectsize)"
828
829 typedef int (*for_each_object_fn)(const struct object_id *oid, struct packed_git *pack,
830 off_t offset, void *data);
831
832 struct for_each_object_payload {
833 for_each_object_fn callback;
834 void *payload;
835 };
836
837 static int batch_one_object_oi(const struct object_id *oid,
838 struct object_info *oi,
839 void *_payload)
840 {
841 struct for_each_object_payload *payload = _payload;
842 if (oi && oi->whence == OI_PACKED)
843 return payload->callback(oid, oi->u.packed.pack, oi->u.packed.offset,
844 payload->payload);
845 return payload->callback(oid, NULL, 0, payload->payload);
846 }
847
848 static int batch_one_object_packed(const struct object_id *oid,
849 struct packed_git *pack,
850 uint32_t pos,
851 void *_payload)
852 {
853 struct for_each_object_payload *payload = _payload;
854 return payload->callback(oid, pack, nth_packed_object_offset(pack, pos),
855 payload->payload);
856 }
857
858 static int batch_one_object_bitmapped(const struct object_id *oid,
859 enum object_type type UNUSED,
860 int flags UNUSED,
861 uint32_t hash UNUSED,
862 struct packed_git *pack,
863 off_t offset,
864 void *_payload)
865 {
866 struct for_each_object_payload *payload = _payload;
867 return payload->callback(oid, pack, offset, payload->payload);
868 }
869
870 static void batch_each_object(struct batch_options *opt,
871 for_each_object_fn callback,
872 unsigned flags,
873 void *_payload)
874 {
875 struct for_each_object_payload payload = {
876 .callback = callback,
877 .payload = _payload,
878 };
879 struct odb_for_each_object_options opts = {
880 .flags = flags,
881 };
882 struct bitmap_index *bitmap = NULL;
883 struct odb_source *source;
884
885 /*
886 * TODO: we still need to tap into implementation details of the object
887 * database sources. Ideally, we should extend `odb_for_each_object()`
888 * to handle object filters itself so that we can move the filtering
889 * logic into the individual sources.
890 */
891 odb_prepare_alternates(the_repository->objects);
892 for (source = the_repository->objects->sources; source; source = source->next) {
893 struct odb_source_files *files = odb_source_files_downcast(source);
894 int ret = odb_source_for_each_object(&files->loose->base, NULL, batch_one_object_oi,
895 &payload, &opts);
896 if (ret)
897 break;
898 }
899
900 if (opt->objects_filter.choice != LOFC_DISABLED &&
901 (bitmap = prepare_bitmap_git(the_repository)) &&
902 !for_each_bitmapped_object(bitmap, &opt->objects_filter,
903 batch_one_object_bitmapped, &payload)) {
904 struct packed_git *pack;
905
906 repo_for_each_pack(the_repository, pack) {
907 if (bitmap_index_contains_pack(bitmap, pack) ||
908 open_pack_index(pack))
909 continue;
910 for_each_object_in_pack(pack, batch_one_object_packed,
911 &payload, flags);
912 }
913 } else {
914 struct object_info oi = { 0 };
915
916 for (source = the_repository->objects->sources; source; source = source->next) {
917 struct odb_source_files *files = odb_source_files_downcast(source);
918 int ret = packfile_store_for_each_object(files->packed, &oi,
919 batch_one_object_oi, &payload, &opts);
920 if (ret)
921 break;
922 }
923 }
924
925 free_bitmap_index(bitmap);
926 }
927
928 static int batch_objects(struct batch_options *opt)
929 {
930 struct strbuf input = STRBUF_INIT;
931 struct strbuf output = STRBUF_INIT;
932 struct expand_data data = EXPAND_DATA_INIT;
933 struct repo_config_values *cfg = repo_config_values(the_repository);
934 int save_warning;
935 int retval = 0;
936
937 /*
938 * Expand once with our special mark_query flag, which will prime the
939 * object_info to be handed to odb_read_object_info_extended for each
940 * object.
941 */
942 data.mark_query = 1;
943 expand_format(&output,
944 opt->format ? opt->format : DEFAULT_FORMAT,
945 &data);
946 data.mark_query = 0;
947 strbuf_release(&output);
948 if (opt->transform_mode)
949 data.split_on_whitespace = 1;
950
951 if (opt->format && !strcmp(opt->format, DEFAULT_FORMAT))
952 opt->format = NULL;
953 /*
954 * If we are printing out the object, then always fill in the type,
955 * since we will want to decide whether or not to stream.
956 */
957 if (opt->batch_mode == BATCH_MODE_CONTENTS)
958 data.info.typep = &data.type;
959
960 if (opt->all_objects) {
961 struct object_cb_data cb;
962 struct object_info empty = OBJECT_INFO_INIT;
963
964 if (!memcmp(&data.info, &empty, sizeof(empty)) &&
965 opt->objects_filter.choice == LOFC_DISABLED)
966 data.skip_object_info = 1;
967
968 if (repo_has_promisor_remote(the_repository))
969 warning("This repository uses promisor remotes. Some objects may not be loaded.");
970
971 disable_replace_refs();
972
973 cb.opt = opt;
974 cb.expand = &data;
975 cb.scratch = &output;
976
977 if (opt->unordered) {
978 struct oidset seen = OIDSET_INIT;
979
980 cb.seen = &seen;
981
982 batch_each_object(opt, batch_unordered_object,
983 ODB_FOR_EACH_OBJECT_PACK_ORDER, &cb);
984
985 oidset_clear(&seen);
986 } else {
987 struct oid_array sa = OID_ARRAY_INIT;
988
989 batch_each_object(opt, collect_object, 0, &sa);
990 oid_array_for_each_unique(&sa, batch_object_cb, &cb);
991
992 oid_array_clear(&sa);
993 }
994
995 strbuf_release(&output);
996 return 0;
997 }
998
999 /*
1000 * We are going to call get_sha1 on a potentially very large number of
1001 * objects. In most large cases, these will be actual object sha1s. The
1002 * cost to double-check that each one is not also a ref (just so we can
1003 * warn) ends up dwarfing the actual cost of the object lookups
1004 * themselves. We can work around it by just turning off the warning.
1005 */
1006 save_warning = cfg->warn_on_object_refname_ambiguity;
1007 cfg->warn_on_object_refname_ambiguity = 0;
1008
1009 if (opt->batch_mode == BATCH_MODE_QUEUE_AND_DISPATCH) {
1010 batch_objects_command(opt, &output, &data);
1011 goto cleanup;
1012 }
1013
1014 while (strbuf_getdelim_strip_crlf(&input, stdin, opt->input_delim) != EOF) {
1015 if (data.split_on_whitespace) {
1016 /*
1017 * Split at first whitespace, tying off the beginning
1018 * of the string and saving the remainder (or NULL) in
1019 * data.rest.
1020 */
1021 char *p = strpbrk(input.buf, " \t");
1022 if (p) {
1023 while (*p && strchr(" \t", *p))
1024 *p++ = '\0';
1025 }
1026 data.rest = p;
1027 }
1028
1029 batch_one_object(input.buf, &output, opt, &data);
1030 }
1031
1032 cleanup:
1033 strbuf_release(&input);
1034 strbuf_release(&output);
1035 cfg->warn_on_object_refname_ambiguity = save_warning;
1036 return retval;
1037 }
1038
1039 static int git_cat_file_config(const char *var, const char *value,
1040 const struct config_context *ctx, void *cb)
1041 {
1042 if (userdiff_config(var, value) < 0)
1043 return -1;
1044
1045 return git_default_config(var, value, ctx, cb);
1046 }
1047
1048 static int batch_option_callback(const struct option *opt,
1049 const char *arg,
1050 int unset)
1051 {
1052 struct batch_options *bo = opt->value;
1053
1054 BUG_ON_OPT_NEG(unset);
1055
1056 if (bo->enabled) {
1057 return error(_("only one batch option may be specified"));
1058 }
1059
1060 bo->enabled = 1;
1061
1062 if (!strcmp(opt->long_name, "batch"))
1063 bo->batch_mode = BATCH_MODE_CONTENTS;
1064 else if (!strcmp(opt->long_name, "batch-check"))
1065 bo->batch_mode = BATCH_MODE_INFO;
1066 else if (!strcmp(opt->long_name, "batch-command"))
1067 bo->batch_mode = BATCH_MODE_QUEUE_AND_DISPATCH;
1068 else
1069 BUG("%s given to batch-option-callback", opt->long_name);
1070
1071 bo->format = arg;
1072
1073 return 0;
1074 }
1075
1076 int cmd_cat_file(int argc,
1077 const char **argv,
1078 const char *prefix,
1079 struct repository *repo UNUSED)
1080 {
1081 int opt = 0;
1082 int opt_cw = 0;
1083 int opt_epts = 0;
1084 const char *exp_type = NULL, *obj_name = NULL;
1085 struct batch_options batch = {
1086 .objects_filter = LIST_OBJECTS_FILTER_INIT,
1087 };
1088 int unknown_type = 0;
1089 int input_nul_terminated = 0;
1090 int nul_terminated = 0;
1091 int ret;
1092
1093 const char * const builtin_catfile_usage[] = {
1094 N_("git cat-file <type> <object>"),
1095 N_("git cat-file (-e | -p | -t | -s) <object>"),
1096 N_("git cat-file (--textconv | --filters)\n"
1097 " [<rev>:<path|tree-ish> | --path=<path|tree-ish> <rev>]"),
1098 N_("git cat-file (--batch | --batch-check | --batch-command) [--batch-all-objects]\n"
1099 " [--buffer] [--follow-symlinks] [--unordered]\n"
1100 " [--textconv | --filters] [-Z]"),
1101 NULL
1102 };
1103 const struct option options[] = {
1104 /* Simple queries */
1105 OPT_GROUP(N_("Check object existence or emit object contents")),
1106 OPT_CMDMODE('e', NULL, &opt,
1107 N_("check if <object> exists"), 'e'),
1108 OPT_CMDMODE('p', NULL, &opt, N_("pretty-print <object> content"), 'p'),
1109
1110 OPT_GROUP(N_("Emit [broken] object attributes")),
1111 OPT_CMDMODE('t', NULL, &opt, N_("show object type (one of 'blob', 'tree', 'commit', 'tag', ...)"), 't'),
1112 OPT_CMDMODE('s', NULL, &opt, N_("show object size"), 's'),
1113 OPT_HIDDEN_BOOL(0, "allow-unknown-type", &unknown_type,
1114 N_("historical option -- no-op")),
1115 OPT_BOOL(0, "use-mailmap", &use_mailmap, N_("use mail map file")),
1116 OPT_ALIAS(0, "mailmap", "use-mailmap"),
1117 /* Batch mode */
1118 OPT_GROUP(N_("Batch objects requested on stdin (or --batch-all-objects)")),
1119 OPT_CALLBACK_F(0, "batch", &batch, N_("format"),
1120 N_("show full <object> or <rev> contents"),
1121 PARSE_OPT_OPTARG | PARSE_OPT_NONEG,
1122 batch_option_callback),
1123 OPT_CALLBACK_F(0, "batch-check", &batch, N_("format"),
1124 N_("like --batch, but don't emit <contents>"),
1125 PARSE_OPT_OPTARG | PARSE_OPT_NONEG,
1126 batch_option_callback),
1127 OPT_BOOL_F('z', NULL, &input_nul_terminated, N_("stdin is NUL-terminated"),
1128 PARSE_OPT_HIDDEN),
1129 OPT_BOOL('Z', NULL, &nul_terminated, N_("stdin and stdout is NUL-terminated")),
1130 OPT_CALLBACK_F(0, "batch-command", &batch, N_("format"),
1131 N_("read commands from stdin"),
1132 PARSE_OPT_OPTARG | PARSE_OPT_NONEG,
1133 batch_option_callback),
1134 OPT_CMDMODE(0, "batch-all-objects", &opt,
1135 N_("with --batch[-check]: ignores stdin, batches all known objects"), 'b'),
1136 /* Batch-specific options */
1137 OPT_GROUP(N_("Change or optimize batch output")),
1138 OPT_BOOL(0, "buffer", &batch.buffer_output, N_("buffer --batch output")),
1139 OPT_BOOL(0, "follow-symlinks", &batch.follow_symlinks,
1140 N_("follow in-tree symlinks")),
1141 OPT_BOOL(0, "unordered", &batch.unordered,
1142 N_("do not order objects before emitting them")),
1143 /* Textconv options, stand-ole*/
1144 OPT_GROUP(N_("Emit object (blob or tree) with conversion or filter (stand-alone, or with batch)")),
1145 OPT_CMDMODE(0, "textconv", &opt,
1146 N_("run textconv on object's content"), 'c'),
1147 OPT_CMDMODE(0, "filters", &opt,
1148 N_("run filters on object's content"), 'w'),
1149 OPT_STRING(0, "path", &force_path, N_("blob|tree"),
1150 N_("use a <path> for (--textconv | --filters); Not with 'batch'")),
1151 OPT_PARSE_LIST_OBJECTS_FILTER(&batch.objects_filter),
1152 OPT_END()
1153 };
1154
1155 repo_config(the_repository, git_cat_file_config, NULL);
1156
1157 batch.buffer_output = -1;
1158
1159 argc = parse_options(argc, argv, prefix, options, builtin_catfile_usage, 0);
1160 opt_cw = (opt == 'c' || opt == 'w');
1161 opt_epts = (opt == 'e' || opt == 'p' || opt == 't' || opt == 's');
1162
1163 if (use_mailmap)
1164 load_mailmap();
1165
1166 switch (batch.objects_filter.choice) {
1167 case LOFC_DISABLED:
1168 break;
1169 case LOFC_BLOB_NONE:
1170 case LOFC_BLOB_LIMIT:
1171 case LOFC_OBJECT_TYPE:
1172 if (!batch.enabled)
1173 usage(_("objects filter only supported in batch mode"));
1174 break;
1175 default:
1176 usagef(_("objects filter not supported: '%s'"),
1177 list_object_filter_config_name(batch.objects_filter.choice));
1178 }
1179
1180 /* --batch-all-objects? */
1181 if (opt == 'b')
1182 batch.all_objects = 1;
1183
1184 /* Option compatibility */
1185 if (force_path && !opt_cw)
1186 usage_msg_optf(_("'%s=<%s>' needs '%s' or '%s'"),
1187 builtin_catfile_usage, options,
1188 "--path", _("path|tree-ish"), "--filters",
1189 "--textconv");
1190
1191 /* Option compatibility with batch mode */
1192 if (batch.enabled)
1193 ;
1194 else if (batch.follow_symlinks)
1195 usage_msg_optf(_("'%s' requires a batch mode"), builtin_catfile_usage,
1196 options, "--follow-symlinks");
1197 else if (batch.buffer_output >= 0)
1198 usage_msg_optf(_("'%s' requires a batch mode"), builtin_catfile_usage,
1199 options, "--buffer");
1200 else if (batch.all_objects)
1201 usage_msg_optf(_("'%s' requires a batch mode"), builtin_catfile_usage,
1202 options, "--batch-all-objects");
1203 else if (input_nul_terminated)
1204 usage_msg_optf(_("'%s' requires a batch mode"), builtin_catfile_usage,
1205 options, "-z");
1206 else if (nul_terminated)
1207 usage_msg_optf(_("'%s' requires a batch mode"), builtin_catfile_usage,
1208 options, "-Z");
1209
1210 batch.input_delim = batch.output_delim = '\n';
1211 if (input_nul_terminated)
1212 batch.input_delim = '\0';
1213 if (nul_terminated)
1214 batch.input_delim = batch.output_delim = '\0';
1215
1216 /* Batch defaults */
1217 if (batch.buffer_output < 0)
1218 batch.buffer_output = batch.all_objects;
1219
1220 prepare_repo_settings(the_repository);
1221 the_repository->settings.command_requires_full_index = 0;
1222
1223 /* Return early if we're in batch mode? */
1224 if (batch.enabled) {
1225 if (opt_cw)
1226 batch.transform_mode = opt;
1227 else if (opt && opt != 'b')
1228 usage_msg_optf(_("'-%c' is incompatible with batch mode"),
1229 builtin_catfile_usage, options, opt);
1230 else if (argc)
1231 usage_msg_opt(_("batch modes take no arguments"),
1232 builtin_catfile_usage, options);
1233
1234 ret = batch_objects(&batch);
1235 goto out;
1236 }
1237
1238 if (opt) {
1239 if (!argc && opt == 'c')
1240 usage_msg_optf(_("<rev> required with '%s'"),
1241 builtin_catfile_usage, options,
1242 "--textconv");
1243 else if (!argc && opt == 'w')
1244 usage_msg_optf(_("<rev> required with '%s'"),
1245 builtin_catfile_usage, options,
1246 "--filters");
1247 else if (!argc && opt_epts)
1248 usage_msg_optf(_("<object> required with '-%c'"),
1249 builtin_catfile_usage, options, opt);
1250 else if (argc == 1)
1251 obj_name = argv[0];
1252 else
1253 usage_msg_opt(_("too many arguments"), builtin_catfile_usage,
1254 options);
1255 } else if (!argc) {
1256 usage_with_options(builtin_catfile_usage, options);
1257 } else if (argc != 2) {
1258 usage_msg_optf(_("only two arguments allowed in <type> <object> mode, not %d"),
1259 builtin_catfile_usage, options, argc);
1260 } else if (argc) {
1261 exp_type = argv[0];
1262 obj_name = argv[1];
1263 }
1264
1265 ret = cat_one_file(opt, exp_type, obj_name);
1266
1267 out:
1268 list_objects_filter_release(&batch.objects_filter);
1269 return ret;
1270 }