[Debichem-devel] Bug#1058454: Bug#1058454: prody: autopkgtest regression on i386

[Santiago Vila]

On Fri, Nov 28, 2025 at 12:16:57PM +0200, Andrius Merkys wrote:

> > I did not mark this bug as fixed at the time, because implicitly it
> > would mean it's also fixed in unstable, but I could not check that
> > because the package also failed to build in unstable for other reasons.
> 
> Thanks for letting me know. I have updated the package to the newest
> upstream release (2.6.1) and can no longer reproduce the problem (checked on
> i386 porterbox). I do not know what made the issue disappear and will not
> have enough time to investigate it, but the important thing is that it is
> resolved, and autopkgtests will let us know if it reoccurs.

My bet would be different floating point code from glibc, but I agree
that as far as it works, figuring out why exactly it works now is more
like a task for bug archeologists (which I sometimes become when I'm
curious, but in no way it's something that should be put on other's
todo lists).

> [...]
> > (I'm not adding it myself at this moment because I don't like to see
> > red pipelines...)
> 
> As said, feel free to add it if you see the benefit.

I am really a late adopter of Salsa CI, and I guess there are several
different ways to approach it.

My idea of using Salsa CI is to achieve a pipeline which is all green,
excluding if necessary some tests which might not work, and from that
point forward, try not to regress.

I will try adding a salsa-ci.yml to prody and see what happens (and
will commit if I see it works well enough).

In the meantime, please feel free to upload if you have 2.6.1 ready.

Thanks.