fetch: fix duplicate remote parallel fetch bug
PATCH 1/1 fetch: fix duplicate remote parallel fetch bug
Fetching in parallel from a remote group with a duplicated remote results in the following:
error: cannot lock ref '<ref>': is at <oid> but expected <oid>
This doesn't happen in serial since fetching from the same remote that has already been fetched from is a noop. Therefore, remove any duplicated remotes after remote groups are parsed.
--- builtin/fetch.c | 1 + t/t5506-remote-groups.sh | 9 +++++++++ 2 files changed, 10 insertions(+)
builtin/fetch.c +1 -0
@@ -2225,6 +2225,7 @@ int cmd_fetch(int argc, const char **argv, const char *prefix) argv++; } } +string_list_remove_duplicates (&list,0 );if (negotiate_only) {struct oidset acked_commits =OIDSET_INIT ;
t/t5506-remote-groups.sh +9 -0
@@ -99,4 +99,13 @@ test_expect_success 'updating remote name updates that remote' ' ! repo_fetched two ' +test_expect_success 'updating group in parallel with a duplicate remote does not fail (fetch)' ' + mark fetch-group-duplicate && + update_repo one && + git config --add remotes.duplicate one && + git config --add remotes.duplicate one && + git -c fetch.parallel=2 remote update duplicate && + repo_fetched one +' + test_done
Signature · 1 line
2.39.0.246.g2a6d74b583-goog
Quoted 20 lines · Calvin Wan
Fetching in parallel from a remote group with a duplicated remote results in the following:
error: cannot lock ref '<ref>': is at <oid> but expected <oid>
This doesn't happen in serial since fetching from the same remote that has already been fetched from is a noop. Therefore, remove any duplicated remotes after remote groups are parsed.
Signed-off-by Calvin Wan <[email protected]>--- builtin/fetch.c | 1 + t/t5506-remote-groups.sh | 9 +++++++++ 2 files changed, 10 insertions(+)builtin/fetch.c +0
-0
builtin/fetch.c int cmd_fetch(int argc, const char **argv, const char *prefix) · line 2228
Quoted 4 lines
builtin/fetch.c +1
-0@@ -2225,6 +2225,7 @@ int cmd_fetch(int argc, const char **argv, const char *prefix) argv++; } } +string_list_remove_duplicates (&list,0 );
As it always is possible to edit .git/config manually, it is necessary to perform deduplication like this patch does on the consumer side of the list, but do you know if our tool create duplication, or is it entirely something the end-user does manually?
If it is the former, I am wondering if we should also fix such a code path that does so in the first place.
Will queue. Thanks.
Quoted 4 lines
As it always is possible to edit .git/config manually, it is necessary to perform deduplication like this patch does on the consumer side of the list, but do you know if our tool create duplication, or is it entirely something the end-user does manually?
I checked git-remote and there is protection against duplication there, but I'm unsure if there are other places where remotes are being added/renamed. I discovered the bug initially by using git-config.