[Python-modules-team] Bug#923826: python-blessed: FTBFS randomly (failing tests)

[Santiago Vila]

On Sun, May 05, 2019 at 02:07:32AM +0200, Pierre-Elliott Bécue wrote:
> > I've put several failed build logs here for reference:
> > 
> > https://people.debian.org/~sanvila/build-logs/python-blessed/
> > 
> > If you need a test machine to reproduce this, please contact me
> > privately and I could provide ssh access to a machine where it happens
> > (caveat: randomly and maybe with low probability).
> > 
> > If this is really a bug in one of the build-depends, please use reassign and affects,
> > so that this is still visible in the BTS web page for this package.
> 
> Dear Santiago,
> 
> Thanks for your bug report.
> 
> I wonder what kind of build system you use. Could you give me a little
> more intel on your environment? The status seems to be 512 when the bug
> occurs, and this status is not connected to anything expected.

I'm using sbuild + schroot on virtual machines from Scaleway.

On instances of type START1-XS (with 1 GB RAM and 1 CPU), I have not
seen it to fail yet.

On instances of type START1-S (with 2 GB RAM and 2 CPUs), the failure
rate is around 5%. This is enough to be quite confident that the
error will happen if one tries enough times, and enough to report it
as a bug, but probably not enough to make the bug RC.

I believe the root of the problem should be some kind of race
condition, maybe related to the fact that the machine has 2 CPUs, but
I don't really know.

So, the only thing I can do about this is to offer ssh access to a
machine where this randomness happens (please contact me privately for
details).

Thanks.