Re: Strange VM/HW dichotomy building Ruby packages

[Kevin Fenzi]

On Mon, 29 Feb 2016 14:49:10 +0100
Vít Ondruch  wrote:

> If you observe the build logs carefully, there were some segfaults for
> example:
> 
> https://koji.fedoraproject.org/koji/taskinfo?taskID=13171578
> 
> This particular one is PPC, so it might be platform specific issue,
> who knows 

Yeah, all the failures were on ppc64le... 

I guess I'd say file a bug on the rhel7 ruby and see if the folks there
can duplicate. In the mean time you should be able to (as you have
seen) just rebuild until it hits a non ppc64le builder. 

kevin



pgp9Z470395Bo.pgp
Description: OpenPGP digital signature
--
devel mailing list
devel@lists.fedoraproject.org
http://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: Strange VM/HW dichotomy building Ruby packages

[Vít Ondruch]

If you observe the build logs carefully, there were some segfaults for
example:

https://koji.fedoraproject.org/koji/taskinfo?taskID=13171578

This particular one is PPC, so it might be platform specific issue, who
knows 


Vít



Dne 29.2.2016 v 06:22 Greg Hellings napsal(a):
> I've been struggling this evening with building a particular package
> into the EPEL7 repositories. At first I thought I was just going
> bonkers, but I think I've narrowed down a particular behavior's cause.
> At least the only cause of it that I can see.
>
> The package in question is rubygem-activesupport and the build is
> supposed to be a noarch, as the package is pure Ruby. Building locally
> in a mock environment has worked without fail all evening long, but
> building in Koji was giving me some issues. After eliminating
> mismatched packages from the two environments by tagging one of my
> prior builds into the buildroot and limiting minimum dependency
> version, I was still seeing a mismatch. My local builds were still
> fine, but the builds in Koji kept failing, both straight builds and
> scratch builds.
>
> Links to the failed scratch builds:
> https://koji.fedoraproject.org/koji/taskinfo?taskID=13171824
> https://koji.fedoraproject.org/koji/taskinfo?taskID=13171578
>
> And failed source build:
> https://koji.fedoraproject.org/koji/taskinfo?taskID=13171784
>
> And my succeeded scratch build:
> https://koji.fedoraproject.org/koji/taskinfo?taskID=13171627
>
> And the successful source build:
> https://koji.fedoraproject.org/koji/taskinfo?taskID=13171904
>
> Note that nothing of consequence changed in the spec file between any
> of these builds. In fact, during the two source builds NOTHING changed
> in git. However, the only difference that I can see is that all of the
> failed builds occurred on either VMs or PPC machines and the
> successful builds all took place on hardware boxes.
>
> I'm not sure if this is a koji thing, a Fedora/EPEL thing, or maybe
> even a core Ruby thing.
>
> Has anyone else experienced similar strange behaviors?
>
> --Greg
> --
> devel mailing list
> devel@lists.fedoraproject.org
> http://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org
--
devel mailing list
devel@lists.fedoraproject.org
http://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Strange VM/HW dichotomy building Ruby packages

[Greg Hellings]

I've been struggling this evening with building a particular package
into the EPEL7 repositories. At first I thought I was just going
bonkers, but I think I've narrowed down a particular behavior's cause.
At least the only cause of it that I can see.

The package in question is rubygem-activesupport and the build is
supposed to be a noarch, as the package is pure Ruby. Building locally
in a mock environment has worked without fail all evening long, but
building in Koji was giving me some issues. After eliminating
mismatched packages from the two environments by tagging one of my
prior builds into the buildroot and limiting minimum dependency
version, I was still seeing a mismatch. My local builds were still
fine, but the builds in Koji kept failing, both straight builds and
scratch builds.

Links to the failed scratch builds:
https://koji.fedoraproject.org/koji/taskinfo?taskID=13171824
https://koji.fedoraproject.org/koji/taskinfo?taskID=13171578

And failed source build:
https://koji.fedoraproject.org/koji/taskinfo?taskID=13171784

And my succeeded scratch build:
https://koji.fedoraproject.org/koji/taskinfo?taskID=13171627

And the successful source build:
https://koji.fedoraproject.org/koji/taskinfo?taskID=13171904

Note that nothing of consequence changed in the spec file between any
of these builds. In fact, during the two source builds NOTHING changed
in git. However, the only difference that I can see is that all of the
failed builds occurred on either VMs or PPC machines and the
successful builds all took place on hardware boxes.

I'm not sure if this is a koji thing, a Fedora/EPEL thing, or maybe
even a core Ruby thing.

Has anyone else experienced similar strange behaviors?

--Greg
--
devel mailing list
devel@lists.fedoraproject.org
http://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org