On 2018-04-06 17:32, Eric Blake wrote: > On 04/06/2018 10:17 AM, Max Reitz wrote: >> The new test case for compressed clusters (added in commit >> abd3622cc03cf41ed542126a540385f30a4c0175) requires two refcount bits to >> succeed, so we need to skip the test when refcount_bits=1 was requested. > > We already questioned whether the changes to 122 were appropriate, given > that it didn't even use qemu-img convert. Should we instead split this > into two tests (keeping 122 as it was before, and making the new > compressed refcount munging tests be separate), where only the new test > needs this filter?
Can do, why not. Max >> Signed-off-by: Max Reitz <mre...@redhat.com> >> --- >> tests/qemu-iotests/122 | 4 ++++ >> 1 file changed, 4 insertions(+) >> >> diff --git a/tests/qemu-iotests/122 b/tests/qemu-iotests/122 >> index 6cf4fcb866..95e7c51e72 100755 >> --- a/tests/qemu-iotests/122 >> +++ b/tests/qemu-iotests/122 >> @@ -42,6 +42,10 @@ _supported_fmt qcow2 >> _supported_proto file >> _supported_os Linux >> >> +# Repairing the compressed image requires qemu-img check to store a >> +# refcount up to 3, which requires at least two refcount bits. >> +_unsupported_imgopts 'refcount_bits=1[^0-9]' >> + > > This makes sense, but I'm withholding R-b until we decide if splitting > the tests is wiser. >
Description: OpenPGP digital signature