Re: emba.gnu.org overload

[Michael Albinus]

Mattias Engdegård <mattiase@acm.org> writes:

Hi Mattias,

>> For example, four different jobs for native-compilation builds are run,
>> build-native-comp-speed{0,1,2} and test-native-comp-speed0. That was
>> fine for development of this feature, but perhaps not so important these
>> days. We might ask Andrea Corallo about.
>> 
>> Also the platform tests build-image-{filenotify-gio,gnustep} and
>> test-{filenotify-gio,gnustep} could be reduced. We don't need them in
>> both emacs-29 and master pipelines; for the time being we might disable
>> them in master.
>
> Then let's drop all of these and see how far that takes us.
> Once enough load has been shed for everything to run in a sustainable way we 
> can think about reinstating some of it.

I've applied some changes:

- Scheduled jobs run 2 times a day only.
- Jobs build-image-filenotify-gio and test-filenotify-gio do not run
  anymore scheduled (they can still be started manually).
- Jobs build-native-comp-speed1 and build-native-comp-speed2 are
  commented out.

>> UCS Normalize Test Part1, rule 2...7% 
>> UCS Normalize Test Part1, rule 2...8% 
>> timeout: the monitored command dumped core
>> --8<---------------cut here---------------end--------------->8---
>> 
>> Doesn't this mean rather a bug we need to capture?
>
> Who knows, I assumed it was just the result of it being killed by the
> time-out. The whole test takes some 20 min on an old laptop here,
> recently reduced from 100 min. Practically all time is spent in the
> "UCS Normalize Test Part1, rule 2..." phase.

Glenn added a skip for the respective test case on hydra, a while
ago. So I've extended this to skip also on emba.

Furthermore, our second runner on emba-runner.gnu.org seems to be
disabled since 12 days. I've never learned how to (re-)activate
it. Toon, could you pls check?

Best regards, Michael.