(undeterministic) test failures in rng test

[carnil]

Hi Rafael,

While uploading yapet 2.0 version for Debian, the package failed to build on two architectures (amd64 and arm64).

The architecture itself is probably not relevant, because the package builded before uploading into the archive as well on amd64.

Any ideas on a possible cause?

[RafaelOstertag]

Hi

The issue can easily be reproduced by running the test multiple times

while ./rng ; do : ; done

The problem is caused by the move assignment operator not properly cleaning up the source object. I'm rolling release 2.1 which addresses the issue.

[carnil]

Hi Rafael,

On Sat, Oct 27, 2018 at 06:12:19AM -0700, Rafael Ostertag wrote:

Hi

The issue can easily be reproduced by running the test multiple times

while ./rng ; do : ; done

The problem is caused by the move assignment operator not properly cleaning up the source object. I'm rolling release 2.1 which addresses the issue.

Should fill maybe a new issue, but as heads up: I still get (undeterminstically) build failures either on our CI infrastructure (cf. https://salsa.debian.org/carnil/yapet/-/jobs which might suceed after retrying) or on our build infrastructure: https://buildd.debian.org/status/package.php?p=yapet .

[RafaelOstertag]

Yes, I saw them. But so far I'm not able to come up with a method to reproduce them. I'll keep an eye on it, though.

[carnil]

Hi,

On Wed, Oct 31, 2018 at 12:50:20AM -0700, Rafael Ostertag wrote:

Yes, I saw them. But so far I'm not able to come up with a method to reproduce them. I'll keep an eye on it, though.

Alright.

Do you prefer to have a new issue opened for those individually as I notice or will you want to keep it track differently?

Btw, many thanks for your upstream work!

[RafaelOstertag]

Hey,

You're welcome!

I'd prefer if you open a new issue upon notice.