Regression for some architectures (arm64 and mips64el)

[tillea]

Hi, there is a bug report against the Debian packaged version of gemma that the test suite fails on arm64 and mips64el. You can find a full build log for arm64 (please see at the end of the log). The build was fine for gemma version 0.98.0. Kind regards, Andreas.

[pjotrp]

Thanks @tillea. It looks like all tests segfault which is not good! Do you have access to a test system that reproduces that? If so, a stack trace would be helpful. Alternatively we can just aim for AMD64. I doubt anyone is interested in running Gemma on alternate platforms at this point.

[tillea]

On Wed, Dec 19, 2018 at 09:31:40AM -0800, Pjotr Prins wrote:

Thanks @tillea. It looks like all tests segfault which is not good! Yes. Specifically that the minor version bump caused this - so the issue will probably be in the diff to the previous version.

Do you have access to a test system that reproduces that? If so, a stack trace would be helpful. In principle every DD has access to a machine of each architecture. So I can try my luck with this but I'm not sure when I will manage.

Alternatively we can just aim for AMD64. I doubt anyone is interested in running Gemma on alternate platforms at this point. I agree that the current use is mostly AMD64. However, it might be that the issue is a sign for a general problem which just by chance does not occure on AMD64. So it might make some sense to track this down also for AMD64 usage.

Kind regards, Andreas.

[pjotrp]

Alternatively it can be caused by the compiler tool chain. Funny it was working before. That diff should not affect all tests. I'll look if I can do something online somewhere. The Guix project also builds for arm64.

[pjotrp]

Appears to work, no? If not we should try the new build.

[mr-c]

@pjotrp the build is failing, it seems: https://buildd.debian.org/status/package.php?p=gemma

[tillea]

On Tue, Sep 29, 2020 at 05:08:12AM -0700, Pjotr Prins wrote:

Appears to work, no? If not we should try the new build. Do you want me to try the patch or should I wait for a release (which I would prefer)?

[pjotrp]

@pjotrp the build is failing, it seems: https://buildd.debian.org/status/package.php?p=gemma

I think I fixed that error.

[pjotrp]

On Tue, Sep 29, 2020 at 05:08:12AM -0700, Pjotr Prins wrote: Appears to work, no? If not we should try the new build. Do you want me to try the patch or should I wait for a release (which I would prefer)?

Let me do a release in the coming week or so. Needs some real life testing.

[pjotrp]

The travis builds are passing with gcc, clang and on OSX: https://travis-ci.org/github/genetics-statistics/GEMMA

[tillea]

On Tue, Sep 29, 2020 at 06:49:05AM -0700, Pjotr Prins wrote:

Let me do a release in the coming week or so. Needs some real life testing. Perfekt. Andreas.

[tillea]

Ping? What about the release with the fixes for the build and the test suite?

[pjotrp]

What is your schedule. I can probably do next week.

[tillea]

On Tue, Nov 10, 2020 at 04:38:33AM -0800, Pjotr Prins wrote:

What is your schedule.

As always - I need it yesterday. ;-) :-P

I can probably do next week.

That's perfectly fine.

Thanks a lot for your work on several free bioinformatics projects, Andreas.

[pjotrp]

@tillea a new release of GEMMA. Tests pass on Travis and in Guix.

[pjotrp]

The travis CI build now includes ARM64 and passes. I think we can close this issue.

[tillea]

On Sun, Nov 29, 2020 at 01:41:04AM -0800, Pjotr Prins wrote:

The travis CI build now includes ARM64 and passes. I think we can close this issue.

Every possible Debian architecture has now built successfully: https://buildd.debian.org/status/package.php?p=gemma Thanks a lot for your work on this, Andreas.