transport: add client support for object-info
Sometimes, it is beneficial to retrieve information about an object without downloading it entirely. The server-side logic for this functionality was implemented in commit "a2ba162cda (object-info: support for retrieving object info, 2021-04-20)." And the wire format is documented at https://git-scm.com/docs/protocol-v2#_object_info. Introduce client-side support for the object-info capability. Add its own function for object-info separate from existing fetch infrastructure. Currently, the client supports requesting a list of object IDs with the size feature from a v2 server. If the server does not advertise this feature (i.e., transfer.advertiseobjectinfo is set to false), the client returns an error and exit. Note that: 1. the entire request is written into req_buf before being sent to the remote. This approach follows the pattern used in the send_fetch_request() logic within 'fetch-pack.c'. Streaming the request is not addressed in this patch. 2. When the server does not recognize an OID, following the v2 protocol, the server returns "<OID> SP", when this happens, fetch_object_info() sets the corresponding size pointer to NULL so that callers can detect and handle it. Helped-by: Jonathan Tan <jonathantanmy@google.com> Helped-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Calvin Wan <calvinwan@google.com> Signed-off-by: Eric Ju <eric.peijian@gmail.com> Signed-off-by: Pablo Sabater <pabloosabaterr@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Calvin Wan committed
Jul 10, 2026 at 18:41 UTC
82e288a14848918784d5a88a9a2d97a6a5b207a2
9 files changed
+209
-2
Makefile
+1
@@ -1159,6 +1159,7 @@ LIB_OBJS += ewah/ewah_rlw.o
1159
LIB_OBJS += exec-cmd.o
1160
LIB_OBJS += fetch-negotiator.o
1161
LIB_OBJS += fetch-pack.o
1162
+LIB_OBJS += fetch-object-info.o
1163
LIB_OBJS += fmt-merge-msg.o
1164
LIB_OBJS += fsck.o
1165
LIB_OBJS += fsmonitor.o
fetch-object-info.c
new
+109
@@ -0,0 +1,109 @@
1
+#include "git-compat-util.h"
2
+#include "gettext.h"
3
+#include "hex.h"
4
+#include "pkt-line.h"
5
+#include "connect.h"
6
+#include "oid-array.h"
7
+#include "odb.h"
8
+#include "fetch-object-info.h"
9
+#include "string-list.h"
10
+
11
+/* Sends object-info command and its arguments into the request buffer. */
12
+static void send_object_info_request(const int fd_out, struct object_info_args *args)
13
+{
14
+ struct strbuf req_buf = STRBUF_INIT;
15
+
16
+ write_command_and_capabilities(&req_buf, "object-info", args->server_options);
17
+
18
+ if (unsorted_string_list_has_string(args->object_info_options, "size"))
19
+ packet_buf_write(&req_buf, "size");
20
+ else
21
+ BUG("only size should be in object_info_options");
22
+
23
+ if (args->oids)
24
+ for (size_t i = 0; i < args->oids->nr; i++)
25
+ packet_buf_write(&req_buf, "oid %s", oid_to_hex(&args->oids->oid[i]));
26
+
27
+ packet_buf_flush(&req_buf);
28
+ if (write_in_full(fd_out, req_buf.buf, req_buf.len) < 0)
29
+ die_errno(_("unable to write request to remote"));
30
+
31
+ strbuf_release(&req_buf);
32
+}
33
+
34
+static size_t parse_object_size(const char *s, size_t *res)
35
+{
36
+ uintmax_t uim;
37
+
38
+ if (!s[0] || s[strspn(s, "0123456789")])
39
+ return -1;
40
+ errno = 0;
41
+ uim = strtoumax(s, NULL, 10);
42
+ if (errno || uim > SIZE_MAX)
43
+ return -1;
44
+ *res = uim;
45
+ return 0;
46
+}
47
+
48
+int fetch_object_info(const enum protocol_version version, struct object_info_args *args,
49
+ struct packet_reader *reader, struct object_info *object_info_data,
50
+ const int stateless_rpc, const int fd_out)
51
+{
52
+ int size_index = -1;
53
+
54
+ switch (version) {
55
+ case protocol_v2:
56
+ if (!server_supports_v2("object-info"))
57
+ die(_("object-info capability is not enabled on the server"));
58
+ send_object_info_request(fd_out, args);
59
+ break;
60
+ case protocol_v1:
61
+ case protocol_v0:
62
+ die(_("unsupported protocol version. expected v2"));
63
+ case protocol_unknown_version:
64
+ BUG("unknown protocol version");
65
+ }
66
+
67
+ for (size_t i = 0; i < args->object_info_options->nr; i++) {
68
+ if (packet_reader_read(reader) != PACKET_READ_NORMAL) {
69
+ check_stateless_delimiter(stateless_rpc, reader,
70
+ "stateless delimiter expected");
71
+ return -1;
72
+ }
73
+
74
+ if (!string_list_has_string(args->object_info_options, reader->line))
75
+ return -1;
76
+
77
+ if (!strcmp(reader->line, "size")) {
78
+ size_index = i;
79
+ for (size_t j = 0; j < args->oids->nr; j++)
80
+ object_info_data[j].sizep = xcalloc(1, sizeof(*object_info_data[j].sizep));
81
+ } else {
82
+ BUG("only size is supported");
83
+ }
84
+ }
85
+
86
+ for (size_t i = 0; packet_reader_read(reader) == PACKET_READ_NORMAL && i < args->oids->nr; i++) {
87
+ struct string_list object_info_values = STRING_LIST_INIT_DUP;
88
+
89
+ string_list_split(&object_info_values, reader->line, " ", -1);
90
+ if (size_index >= 0) {
91
+ if (!strcmp(object_info_values.items[1 + size_index].string, "")) {
92
+ FREE_AND_NULL(object_info_data[i].sizep);
93
+ string_list_clear(&object_info_values, 0);
94
+ continue;
95
+ }
96
+
97
+ if (parse_object_size(object_info_values.items[1 + size_index].string,
98
+ object_info_data[i].sizep))
99
+ die("object-info: ref %s has invalid size %s",
100
+ object_info_values.items[0].string,
101
+ object_info_values.items[1 + size_index].string);
102
+ }
103
+
104
+ string_list_clear(&object_info_values, 0);
105
+ }
106
+ check_stateless_delimiter(stateless_rpc, reader, "stateless delimiter expected");
107
+
108
+ return 0;
109
+}
fetch-object-info.h
new
+22
@@ -0,0 +1,22 @@
1
+#ifndef FETCH_OBJECT_INFO_H
2
+#define FETCH_OBJECT_INFO_H
3
+
4
+#include "pkt-line.h"
5
+#include "protocol.h"
6
+#include "odb.h"
7
+
8
+struct object_info_args {
9
+ struct string_list *object_info_options;
10
+ const struct string_list *server_options;
11
+ struct oid_array *oids;
12
+};
13
+
14
+/*
15
+ * Sends git-cat-file object-info command into the request buf and read the
16
+ * results from packets.
17
+ */
18
+int fetch_object_info(enum protocol_version version, struct object_info_args *args,
19
+ struct packet_reader *reader, struct object_info *object_info_data,
20
+ int stateless_rpc, int fd_out);
21
+
22
+#endif /* FETCH_OBJECT_INFO_H */
fetch-pack.h
+1
@@ -16,6 +16,7 @@ struct fetch_pack_args {
16
const struct string_list *deepen_not;
17
struct list_objects_filter_options filter_options;
18
const struct string_list *server_options;
19
+ struct object_info *object_info_data;
20
21
/*
22
* If not NULL, during packfile negotiation, fetch-pack will send "have"
meson.build
+1
@@ -347,6 +347,7 @@ libgit_sources = [
347
'exec-cmd.c',
348
'fetch-negotiator.c',
349
'fetch-pack.c',
350
+ 'fetch-object-info.c',
351
'fmt-merge-msg.c',
352
'fsck.c',
353
'fsmonitor.c',
transport-helper.c
+11
-2
@@ -727,8 +727,7 @@ static int fetch_refs(struct transport *transport,
727
728
/*
729
* If we reach here, then the server, the client, and/or the transport
730
- * helper does not support protocol v2. --negotiate-only requires
731
- * protocol v2.
730
+ * helper does not support protocol v2. --negotiate-only.
731
*/
732
if (data->transport_options.acked_commits) {
733
warning(_("--negotiate-only requires protocol v2"));
@@ -784,6 +783,15 @@ static int fetch_refs(struct transport *transport,
783
return -1;
784
}
785
786
+static int fetch_object_info_helper(struct transport *transport)
787
+{
788
+ get_helper(transport);
789
+ if (process_connect(transport, 0))
790
+ return transport->vtable->fetch_object_info(transport);
791
+
792
+ die(_("object-info requires protocol v2"));
793
+}
794
+
795
struct push_update_ref_state {
796
struct ref *hint;
797
struct ref_push_report *report;
@@ -1330,6 +1338,7 @@ static struct transport_vtable vtable = {
1338
.get_refs_list = get_refs_list,
1339
.get_bundle_uri = get_bundle_uri,
1340
.fetch_refs = fetch_refs,
1341
+ .fetch_object_info = fetch_object_info_helper,
1342
.push_refs = push_refs,
1343
.connect = connect_helper,
1344
.disconnect = release_helper
transport-internal.h
+8
@@ -45,6 +45,14 @@ struct transport_vtable {
45
**/
46
int (*fetch_refs)(struct transport *transport, int refs_nr, struct ref **refs);
47
48
+ /*
49
+ * Fetch object info (only size currently) from remote without
50
+ * downloading the objects.
51
+ *
52
+ * Uses object-info capability of v2 protocol.
53
+ */
54
+ int (*fetch_object_info)(struct transport *transport);
55
+
56
/**
57
* Push the objects and refs. Send the necessary objects, and
58
* then, for any refs where peer_ref is set and
transport.c
+46
@@ -1,3 +1,4 @@
1
+#include "compat/posix.h"
2
#define USE_THE_REPOSITORY_VARIABLE
3
4
#include "git-compat-util.h"
@@ -9,6 +10,7 @@
10
#include "hook.h"
11
#include "pkt-line.h"
12
#include "fetch-pack.h"
13
+#include "fetch-object-info.h"
14
#include "remote.h"
15
#include "connect.h"
16
#include "send-pack.h"
@@ -432,6 +434,48 @@ static int get_bundle_uri(struct transport *transport)
434
transport->bundles, stateless_rpc);
435
}
436
437
+static int fetch_object_info_via_pack(struct transport *transport)
438
+{
439
+ int ret = 0;
440
+ struct git_transport_data *data = transport->data;
441
+ struct packet_reader reader;
442
+ struct object_info_args args = { 0 };
443
+
444
+ args.server_options = transport->server_options;
445
+ args.oids = transport->smart_options->object_info_oids;
446
+ args.object_info_options = transport->smart_options->object_info_options;
447
+ string_list_sort(args.object_info_options);
448
+
449
+ connect_setup(transport, 0);
450
+ packet_reader_init(&reader, data->fd[0], NULL, 0,
451
+ PACKET_READ_CHOMP_NEWLINE |
452
+ PACKET_READ_GENTLE_ON_EOF |
453
+ PACKET_READ_DIE_ON_ERR_PACKET);
454
+
455
+ data->version = discover_version(&reader);
456
+ transport->hash_algo = reader.hash_algo;
457
+
458
+ ret = fetch_object_info(data->version, &args, &reader,
459
+ data->options.object_info_data,
460
+ transport->stateless_rpc, data->fd[1]);
461
+
462
+ close(data->fd[0]);
463
+ if (data->fd[1] >= 0)
464
+ close(data->fd[1]);
465
+ if (finish_connect(data->conn))
466
+ ret = -1;
467
+ data->conn = NULL;
468
+
469
+ return ret;
470
+}
471
+
472
+int transport_fetch_object_info(struct transport *transport)
473
+{
474
+ if (!transport->vtable->fetch_object_info)
475
+ die(_("remote does not support object-info"));
476
+ return transport->vtable->fetch_object_info(transport);
477
+}
478
+
479
static int fetch_refs_via_pack(struct transport *transport,
480
int nr_heads, struct ref **to_fetch)
481
{
@@ -1004,6 +1048,7 @@ static struct transport_vtable taken_over_vtable = {
1048
.get_refs_list = get_refs_via_connect,
1049
.get_bundle_uri = get_bundle_uri,
1050
.fetch_refs = fetch_refs_via_pack,
1051
+ .fetch_object_info = fetch_object_info_via_pack,
1052
.push_refs = git_transport_push,
1053
.disconnect = disconnect_git
1054
};
@@ -1169,6 +1214,7 @@ static struct transport_vtable builtin_smart_vtable = {
1214
.get_refs_list = get_refs_via_connect,
1215
.get_bundle_uri = get_bundle_uri,
1216
.fetch_refs = fetch_refs_via_pack,
1217
+ .fetch_object_info = fetch_object_info_via_pack,
1218
.push_refs = git_transport_push,
1219
.connect = connect_git,
1220
.disconnect = disconnect_git
transport.h
+10
@@ -6,6 +6,7 @@
6
#include "list-objects-filter-options.h"
7
#include "string-list.h"
8
#include "connect.h"
9
+#include "odb.h"
10
11
struct git_transport_options {
12
unsigned thin : 1;
@@ -55,6 +56,10 @@ struct git_transport_options {
56
* common commits to this oidset instead of fetching any packfiles.
57
*/
58
struct oidset *acked_commits;
59
+
60
+ struct oid_array *object_info_oids;
61
+ struct object_info *object_info_data;
62
+ struct string_list *object_info_options;
63
};
64
65
enum transport_family {
@@ -309,6 +314,11 @@ int transport_get_remote_bundle_uri(struct transport *transport);
314
const struct git_hash_algo *transport_get_hash_algo(struct transport *transport);
315
int transport_fetch_refs(struct transport *transport, struct ref *refs);
316
317
+/*
318
+ * Fetch the object info from remote
319
+ */
320
+int transport_fetch_object_info(struct transport *transport);
321
+
322
/*
323
* If this flag is set, unlocking will avoid to call non-async-signal-safe
324
* functions. This will necessarily leave behind some data structures which