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