[Debian-med-packaging] Bug#986256: [RFS] simka: test failure common denominator ci-worker13

[Étienne Mollier]

Hi Paul, Hi Nilesh,

when looking up that issue today, I noticed on the CI web
page[1], that all tests were failing on one given host, which is
ci-worker13.  I would thus suspect, that the failure is
triggered by some hardware specific configuration: maybe core
count (ci-worker13 reports builds with 48 parallel jobs), memory
size, or something more exotic.

[1]: https://ci.debian.net/packages/s/simka/testing/amd64/

Looking then further, I think I'm on a good trail on the high
cores count hypothesis.  I can reproduce some severe slowdowns
using qemu with 48 cpus overcommit, or even less.  Actually I
could bissect clunky execution speed of the test case past 8 cpu
cores allocated.  As far as I can tell, passing CI runs only
occurred on 2 cpu cores count machines, so they seem consistent
with my own observations.

I brought a change on simka run-unit-test on Salsa[2], to cap
the execution of simka to 8 cores in the context of the test and
address the issue.  simka 1.5.3-3 is available for review, and
possibly upload, feel also free to grant dm rights if you feel
appropriate.  Note that CI pipeline is on hold, but I expect the
blhc test to fail due to missing compilation flags; other CI
tests did seem to pass on my end.

[2]: https://salsa.debian.org/med-team/simka/

Hopefully the slowdown is specific to the test case, but this
might need to be brought upstream.  Thanks for having caught it!

Kind Regards,
-- 
Étienne Mollier <etienne.mollier at mailoo.org>
Fingerprint:  8f91 b227 c7d6 f2b1 948c  8236 793c f67e 8f0d 11da
Sent from /dev/pts/2, please excuse my verbosity.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/debian-med-packaging/attachments/20210405/bfa73798/attachment.sig>