git-commit-vandalism/t/t5810-proto-disable-local.sh
Ævar Arnfjörð Bjarmason c65d18cb52 push: free_refs() the "local_refs" in set_refspecs()
Fix a memory leak that's been with us since this code was added in
ca02465b41 (push: use remote.$name.push as a refmap, 2013-12-03).

The "remote = remote_get(...)" added in the same commit would seem to
leak based only on the context here, but that function is a wrapper
for sticking the remotes we fetch into "the_repository->remote_state".

See fd3cb0501e (remote: move static variables into per-repository
struct, 2021-11-17) for the addition of code in repository.c that
free's the "remote" allocated here.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2023-02-06 15:34:40 -08:00

40 lines
1.0 KiB
Bash
Executable File

#!/bin/sh
test_description='test disabling of local paths in clone/fetch'
TEST_PASSES_SANITIZE_LEAK=true
. ./test-lib.sh
. "$TEST_DIRECTORY/lib-proto-disable.sh"
test_expect_success 'setup repository to clone' '
test_commit one
'
test_proto "file://" file "file://$PWD"
test_proto "path" file .
test_expect_success 'setup repo with dash' '
git init --bare repo.git &&
git push repo.git HEAD &&
mv repo.git "$PWD/-repo.git"
'
# This will fail even without our rejection because upload-pack will
# complain about the bogus option. So let's make sure that GIT_TRACE
# doesn't show us even running upload-pack.
#
# We must also be sure to use "fetch" and not "clone" here, as the latter
# actually canonicalizes our input into an absolute path (which is fine
# to allow).
test_expect_success 'repo names starting with dash are rejected' '
rm -f trace.out &&
test_must_fail env GIT_TRACE="$PWD/trace.out" git fetch -- -repo.git &&
! grep upload-pack trace.out
'
test_expect_success 'full paths still work' '
git fetch "$PWD/-repo.git"
'
test_done