Re: [Qemu-devel] iotest 026 -nocache is broken

[Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>]

On 22.03.2016 21:06, Paolo Bonzini wrote:
>
> On 22/03/2016 19:00, Vladimir Sementsov-Ogievskiy wrote:
>> Hi all!
>>
>> Just a notice
>>
>> for commit (near to master branch top)
>> commit 3666a97f78704b941c360dc917acb14c8774eca7
>> Author: Eric Blake <eblake@redhat.com>
>> Date:   Thu Mar 17 16:48:40 2016 -0600
>>
>>      qapi: Use anonymous bases in QMP flat unions
>>
>> ./configure --target-list=x86_64-softmmu
>>
>> ...
>>
>> ./check -T -nocache -qcow2 026
>>
>> gives a lot of
>> +Failed to flush the L2 table cache: No space left on device
>> +Failed to flush the refcount block cache: No space left on device
>>
>> and some other differences.
>>
>> this affects ./tests/check-block.sh
>>
>> I've failed to find commit, for which this test was working..
>>
>> it fails even for the first commit, where 026.out.nocache was introduced:
>>
>> commit 8f94b077877151de93a63c73f796897309568ddb
>> Author: Kevin Wolf <kwolf@redhat.com>
>> Date:   Wed Sep 4 13:16:05 2013 +0200
>>
>>      qemu-iotests: Fixed test case 026
>>
>>
>> May be I'm doing something wrong, but I can't imagine what..
> This test fails with different error messages depending on whether you
> use nocache or writethrough.

Yes, and therefore there are two output files: 026.out and 
026.out.nocache. 026,out is ok, and ./check -qcow2 026 is ok, (and 
./check -qcow2 -writetrhough 026 is ok too), result is compared with 
026.out. But for -nocache it is compared with 026.out.nocache which is 
wrong.


>
> Paolo


-- 
Best regards,
Vladimir