[Gluster-devel] $B1 in tests

Ravishankar N ravishankar at redhat.com
Fri Aug 8 05:40:23 UTC 2014


On 08/08/2014 11:01 AM, Avra Sengupta wrote:
> On 08/08/2014 10:07 AM, Ravishankar N wrote:
>> On 08/08/2014 08:52 AM, Emmanuel Dreyfus wrote:
>>> Emmanuel Dreyfus <manu at netbsd.org> wrote:
>>>
>>>> And what about basic/mgmt_v3-locks.t ?
>>>> It uses $B1, $B2, $B3, $H1, $H2 and $H3 without ever defining them.
>>>> Perhaps this is why that specific test randomly fails?
>>> I was wrong, this is defined in cluster.rc:
>>> eval "H$i=${CLUSTER_PFX}.$i";
>>> (...)
>>> eval "B$i=$B0/$i";
>>>
>>> This suggests this is just a s/B1/B0/ bug in self-heal.t
>>>
>> Right, I just went through the bugs and found bugs/bug-1095097.t 
>> seems to be using B1 without using the cluster framework and cleanup 
>> does not handle it. You end up with a /brick1 on your root filesystem 
>> with glusterfs extended attributes set.
>>
>> Avra, perhaps you should modify the test case to this:
>> diff --git a/tests/bugs/bug-1095097.t b/tests/bugs/bug-1095097.t
>>
>> -TEST $CLI volume create $V0 $H0:$B0/brick1 $H0:$B1/brick1;
>> +TEST $CLI volume create $V0 $H0:$B0/brick1 $H0:$B0/brick3;
>>
>>
> The change sounds fine. Is there a bug for this? If you are sending a 
> fix for self-heal.t, I suggest you make the required change as part of 
> the same patch.
>
self-heal patch (http://review.gluster.org/#/c/8431/) already merged :-)
I suppose you could use the bug id  of the original testcase to submit it.


More information about the Gluster-devel mailing list