Re: tests: n810 arm failing

[Philippe Mathieu-Daudé]

On 2/22/21 7:25 AM, Thomas Huth wrote:
> On 21/02/2021 20.33, Peter Maydell wrote:
>> On Sun, 21 Feb 2021 at 17:26, Philippe Mathieu-Daudé <f4bug@amsat.org>
>> wrote:
>>>
>>> Hi,
>>>
>>> The n810 arm test failed on latest master:
>>>
>>> https://gitlab.com/qemu-project/qemu/-/jobs/1045015424/artifacts/browse/build/tests/results/latest/test-results/26-tests_acceptance_machine_arm_n8x0.py_N8x0Machine.test_n810/
>>>
>>
>> It passed on the gitlab CI run on the exact same commit hash
>> as 'staging' before I pushed the merge, though...
> 
> Looking at the following run, it failed with some timeouts as well:
> 
>  https://gitlab.com/qemu-project/qemu/-/pipelines/259425956
> 
> ... but in completely different spots.

This one is "read ASDF from console but some char got eaten
so ASF was displayed and Avocado is reading infinitely for
ASDF".

> So maybe it was "just" a global
> slow-down in the CI farm? (Which sounds somewhat worrisome, too ... I
> hope gitlab-CI is not going down the same road as travis-ci.org did
> during the past year)

The question here is how can we detect the failure cause? Is it the
CI infra or our test?

Currently Avocado only dump log for "FAILed" tests.

I sent a patch to also dump the log for "INTERRUPTed (by timeout)":
https://www.mail-archive.com/qemu-devel@nongnu.org/msg782782.html

Here we got an ERROR result (SIGTERM) so the log is not dumped.

Regards,

Phil.