openjdk team mailing list archive

[Paul Gevers <elbrus@xxxxxxxxxx>]

Hi Matthias

On 23-10-2020 13:31, Matthias Klose wrote:
> On 10/7/20 10:33 PM, Paul Gevers wrote:
>> Hi Matthias,
>>
>> On 21-09-2020 17:52, Paul Gevers wrote:
>>>> Apparently
>>>> the very same tests don't time out on the buildds.
>>>
>>> I don't know what timeouts apply to buildds, but indeed I think they are
>>> much higher to cope with *building* some extremely big packages.
>>
>> Do you know how much time these tests would take? If so, I wonder if we
>> should make it possible for individual packages to have a longer
>> timeout. It would be work on the debci/ci.d.n side of things, but not
>> impossible of course.
> 
> For an upper bound, use the build time: The very same tests are run during the
> build.

Ack. Side note: I see a hell load of errors and failures in the build
log, is that expected? And some builds took more than a day, is that
reasonable to do for these tests (if they're ignored anyways, your tests
are marked flaky)?

> The other question is, if it's sensible to run the upstream test suite as an
> autopkg test, if it's already run during the build.  Ideally it should only run
> the autopkg tests which test on integration issues with run time dependencies,
> but this would require analyzing some 10000 tests and
> 
> For now I just disabled the jdk tests as an autopkg test. So please clear the
> test results, to run it again, and let it migrate.

Also the hotpot test times out, so only having the jdk tests disabled
doesn't really help at this stage.

Paul

Attachment: signature.asc
Description: OpenPGP digital signature