[libvirt] libvirt-tck failures after sparse stream support

Hi Michal, I noticed some of the libvirt-tck tests fail with 3.4.0 RC2. I bisected the failure to commit 07c2399c01. The failing tests are 400-vol-download.t, 405-vol-download-all.t, and 410-vol-download-nonblock.t. Looking at the tests, they all create a pool and vol, write some data to the vol, create an MD5 digest of a subset of vol data, download the subset of the vol, and finally compare MD5 digest of dowloaded vol vs original. If the digests are different the test fails. For reference, here is the first test http://libvirt.org/git/?p=libvirt-tck.git;a=blob;f=scripts/storage/400-vol-d... All three tests randomly fail with commit 07c2399c01. When the tests fail, it is always the case that 'length' was > 0. I.e. I have not seen the tests fail with offset=0, length=0 or offset=1024*1024, length=0. Regards, Jim

On 06/02/2017 01:10 AM, Jim Fehlig wrote:
Hi Michal,
I noticed some of the libvirt-tck tests fail with 3.4.0 RC2. I bisected the failure to commit 07c2399c01. The failing tests are 400-vol-download.t, 405-vol-download-all.t, and 410-vol-download-nonblock.t. Looking at the tests, they all create a pool and vol, write some data to the vol, create an MD5 digest of a subset of vol data, download the subset of the vol, and finally compare MD5 digest of dowloaded vol vs original. If the digests are different the test fails. For reference, here is the first test
http://libvirt.org/git/?p=libvirt-tck.git;a=blob;f=scripts/storage/400-vol-d...
All three tests randomly fail with commit 07c2399c01. When the tests fail, it is always the case that 'length' was > 0. I.e. I have not seen the tests fail with offset=0, length=0 or offset=1024*1024, length=0.
Ah, yeah. I think this is the same as: https://www.redhat.com/archives/libvir-list/2017-May/msg01241.html I think I might have a fix for that which I'm testing now. Will post shortly to the list. Michal
participants (2)
-
Jim Fehlig
-
Michal Privoznik