[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] transient failure in the test-qht tests
From: |
Peter Maydell |
Subject: |
Re: [Qemu-devel] transient failure in the test-qht tests |
Date: |
Thu, 25 Aug 2016 00:52:36 +0100 |
On 25 August 2016 at 00:44, Emilio G. Cota <address@hidden> wrote:
> On Wed, Aug 24, 2016 at 21:39:01 +0100, Peter Maydell wrote:
>> So I encountered this test failure running 'make check' on
>> 32-bit ARM:
>>
>> MALLOC_PERTURB_=${MALLOC_PERTURB_:-$((RANDOM % 255 + 1))} gtester -k
>> --verbose -m=quick tests/test-qht
>> TEST: tests/test-qht... (pid=15763)
>> /qht/mode/default: OK
>> /qht/mode/resize: FAIL
>> GTester: last random seed: R02S08efd89fe4d862dd0191c13d5ce4d76e
>> (pid=16462)
>> FAIL: tests/test-qht
>>
>> The test suite passed on a rerun.
>>
>> Any ideas?
>
> I wonder whether malloc perturb had to do with the failure, because
> -ENOMEM is unlikely (I only see a few MB of peak mem usage for qht-test)
>
> However, I just ran qht-test under valgrind on an i686 machine, and it comes
> clean.
>
> I also brute-forced this to see if a particular perturb value would
> make it fail:
> for i in $(seq 0 255); do \
> echo $i && \
> MALLOC_PERTURB_=$i gtester -k --verbose -m=quick tests/test-qht \
> --seed=R02S08efd89fe4d862dd0191c13d5ce4d76e || break; \
> done
>
> I get no failures on both i686 and x86_64, with and without that --seed flag.
>
> Is there any chance of getting a core dump for the failure you encountered?
Unfortunately not, the test config doesn't save core dumps. In any case
I assume from the output that the test didn't actually dump core, it
just failed (without saying anything about why).
thanks
-- PMM