check_everything_connected: assume alternate ref tips are valid
When we receive a remote ref update to sha1 "X", we want to check that we have all of the objects needed by "X". We can assume that our repository is not currently corrupted, and therefore if we have a ref pointing at "Y", we have all of its objects. So we can stop our traversal from "X" as soon as we hit "Y". If we make the same non-corruption assumption about any repositories we use to store alternates, then we can also use their ref tips to shorten the traversal. This is especially useful when cloning with "--reference", as we otherwise do not have any local refs to check against, and have to traverse the whole history, even though the other side may have sent us few or no objects. Here are results for the included perf test (which shows off more or less the maximal savings, getting one new commit and sharing the whole history): Test HEAD^ HEAD -------------------------------------------------------------------- [on git.git] 5600.3: clone --reference 2.94(2.86+0.08) 0.09(0.08+0.01) -96.9% [on linux.git] 5600.3: clone --reference 45.74(45.34+0.41) 0.36(0.30+0.08) -99.2% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King committed
Jul 1, 2019 at 09:18 UTC
39b44ba771a315602fd1fdca2e12dfc31ef9c613
5 files changed
+125
Documentation/rev-list-options.txt
+8
@@ -182,6 +182,14 @@ explicitly.
182
Pretend as if all objects mentioned by reflogs are listed on the
183
command line as `<commit>`.
184
185
+--alternate-refs::
186
+ Pretend as if all objects mentioned as ref tips of alternate
187
+ repositories were listed on the command line. An alternate
188
+ repository is any repository whose object directory is specified
189
+ in `objects/info/alternates`. The set of included objects may
190
+ be modified by `core.alternateRefsCommand`, etc. See
191
+ linkgit:git-config[1].
192
+
193
--single-worktree::
194
By default, all working trees will be examined by the
195
following options when there are more than one (see
connected.c
+1
@@ -80,6 +80,7 @@ int check_connected(oid_iterate_fn fn, void *cb_data,
80
argv_array_push(&rev_list.args, "--all");
81
}
82
argv_array_push(&rev_list.args, "--quiet");
83
+ argv_array_push(&rev_list.args, "--alternate-refs");
84
if (opt->progress)
85
argv_array_pushf(&rev_list.args, "--progress=%s",
86
_("Checking connectivity"));
revision.c
+29
@@ -1552,6 +1552,32 @@ void add_index_objects_to_pending(struct rev_info *revs, unsigned int flags)
1552
free_worktrees(worktrees);
1553
}
1554
1555
+struct add_alternate_refs_data {
1556
+ struct rev_info *revs;
1557
+ unsigned int flags;
1558
+};
1559
+
1560
+static void add_one_alternate_ref(const struct object_id *oid,
1561
+ void *vdata)
1562
+{
1563
+ const char *name = ".alternate";
1564
+ struct add_alternate_refs_data *data = vdata;
1565
+ struct object *obj;
1566
+
1567
+ obj = get_reference(data->revs, name, oid, data->flags);
1568
+ add_rev_cmdline(data->revs, obj, name, REV_CMD_REV, data->flags);
1569
+ add_pending_object(data->revs, obj, name);
1570
+}
1571
+
1572
+static void add_alternate_refs_to_pending(struct rev_info *revs,
1573
+ unsigned int flags)
1574
+{
1575
+ struct add_alternate_refs_data data;
1576
+ data.revs = revs;
1577
+ data.flags = flags;
1578
+ for_each_alternate_ref(add_one_alternate_ref, &data);
1579
+}
1580
+
1581
static int add_parents_only(struct rev_info *revs, const char *arg_, int flags,
1582
int exclude_parent)
1583
{
@@ -1954,6 +1980,7 @@ static int handle_revision_opt(struct rev_info *revs, int argc, const char **arg
1980
!strcmp(arg, "--no-walk") || !strcmp(arg, "--do-walk") ||
1981
!strcmp(arg, "--bisect") || starts_with(arg, "--glob=") ||
1982
!strcmp(arg, "--indexed-objects") ||
1983
+ !strcmp(arg, "--alternate-refs") ||
1984
starts_with(arg, "--exclude=") ||
1985
starts_with(arg, "--branches=") || starts_with(arg, "--tags=") ||
1986
starts_with(arg, "--remotes=") || starts_with(arg, "--no-walk="))
@@ -2440,6 +2467,8 @@ static int handle_revision_pseudo_opt(const char *submodule,
2467
add_reflogs_to_pending(revs, *flags);
2468
} else if (!strcmp(arg, "--indexed-objects")) {
2469
add_index_objects_to_pending(revs, *flags);
2470
+ } else if (!strcmp(arg, "--alternate-refs")) {
2471
+ add_alternate_refs_to_pending(revs, *flags);
2472
} else if (!strcmp(arg, "--not")) {
2473
*flags ^= UNINTERESTING | BOTTOM;
2474
} else if (!strcmp(arg, "--no-walk")) {
t/perf/p5600-clone-reference.sh
new
+27
@@ -0,0 +1,27 @@
1
+#!/bin/sh
2
+
3
+test_description='speed of clone --reference'
4
+. ./perf-lib.sh
5
+
6
+test_perf_default_repo
7
+
8
+test_expect_success 'create shareable repository' '
9
+ git clone --bare . shared.git
10
+'
11
+
12
+test_expect_success 'advance base repository' '
13
+ # Do not use test_commit here; its test_tick will
14
+ # use some ancient hard-coded date. The resulting clock
15
+ # skew will cause pack-objects to traverse in a very
16
+ # sub-optimal order, skewing the results.
17
+ echo content >new-file-that-does-not-exist &&
18
+ git add new-file-that-does-not-exist &&
19
+ git commit -m "new commit"
20
+'
21
+
22
+test_perf 'clone --reference' '
23
+ rm -rf dst.git &&
24
+ git clone --no-local --bare --reference shared.git . dst.git
25
+'
26
+
27
+test_done
t/t5618-alternate-refs.sh
new
+60
@@ -0,0 +1,60 @@
1
+#!/bin/sh
2
+
3
+test_description='test handling of --alternate-refs traversal'
4
+. ./test-lib.sh
5
+
6
+# Avoid test_commit because we want a specific and known set of refs:
7
+#
8
+# base -- one
9
+# \ \
10
+# two -- merged
11
+#
12
+# where "one" and "two" are on separate refs, and "merged" is available only in
13
+# the dependent child repository.
14
+test_expect_success 'set up local refs' '
15
+ git checkout -b one &&
16
+ test_tick &&
17
+ git commit --allow-empty -m base &&
18
+ test_tick &&
19
+ git commit --allow-empty -m one &&
20
+ git checkout -b two HEAD^ &&
21
+ test_tick &&
22
+ git commit --allow-empty -m two
23
+'
24
+
25
+# We'll enter the child repository after it's set up since that's where
26
+# all of the subsequent tests will want to run (and it's easy to forget a
27
+# "-C child" and get nonsense results).
28
+test_expect_success 'set up shared clone' '
29
+ git clone -s . child &&
30
+ cd child &&
31
+ git merge origin/one
32
+'
33
+
34
+test_expect_success 'rev-list --alternate-refs' '
35
+ git rev-list --remotes=origin >expect &&
36
+ git rev-list --alternate-refs >actual &&
37
+ test_cmp expect actual
38
+'
39
+
40
+test_expect_success 'rev-list --not --alternate-refs' '
41
+ git rev-parse HEAD >expect &&
42
+ git rev-list HEAD --not --alternate-refs >actual &&
43
+ test_cmp expect actual
44
+'
45
+
46
+test_expect_success 'limiting with alternateRefsPrefixes' '
47
+ test_config core.alternateRefsPrefixes refs/heads/one &&
48
+ git rev-list origin/one >expect &&
49
+ git rev-list --alternate-refs >actual &&
50
+ test_cmp expect actual
51
+'
52
+
53
+test_expect_success 'log --source shows .alternate marker' '
54
+ git log --oneline --source --remotes=origin >expect.orig &&
55
+ sed "s/origin.* /.alternate /" <expect.orig >expect &&
56
+ git log --oneline --source --alternate-refs >actual &&
57
+ test_cmp expect actual
58
+'
59
+
60
+test_done