Hi Jochen Am 26.08.26 um 11:37 schrieb Jochen Sprickerhof:
* Michael Biebl <[email protected]> [2026-08-26 11:34]:Tbh, I'm tempted to simply disable 'make check' when it is run as part of a reproducible build. Is there a reliable way to detect this from within debian/rules?No and there will not be. We try to reproduce the exact build environment, so I would be really interested why it fails only on some computers. Do you have any hint how to debug this?
I had filed https://github.com/rsyslog/rsyslog/issues/6293 a while ago (which was closed)
As Rainer (upstream) wrote, some of the tests are inherently racy as they rely on timing. I guess the question is, what conditions must be met to trigger those issues. Is it high cpu load making the build slow (but then again, we have far slower architectures than amd64 on our buildds), slow I/O (your remark regarding ext4 vs tmpfs hints that way) or some other issue. I think this this particular test relies on inotify (via imfile [0]), see Mode (default inotify). Maybe inotify is unstable within the reproduce environment (just a wild guess)
The package build rather reliably on our buildds and on salsa. Regards, Michael [0] https://docs.rsyslog.com/doc/configuration/modules/imfile.html
OpenPGP_signature.asc
Description: OpenPGP digital signature
