While trying to update one of my packages (rdkit), I am hitting an issue with its tests suite, so apparently every non-x86_64 build is failing.
rawhide example: https://koji.fedoraproject.org/koji/taskinfo?taskID=19487191 EPEL example: https://koji.fedoraproject.org/koji/taskinfo?taskID=19486424
So, since this is a pretty large codebase (that I am not familiar with) I am going to ask upstream if they can reproduce the issue, but before doing so I'm wondering if anyone here can think of any reason why our builders can determine such a behavior.
I know this kind of a long shot, but any idea to start debugging the issue will be appreciated.
Cheers
G.
On Thu, May 11, 2017 at 2:51 PM, Gianluca Sforna giallu@gmail.com wrote:
While trying to update one of my packages (rdkit), I am hitting an issue with its tests suite, so apparently every non-x86_64 build is failing.
Well if i686 fails too there's a problem....
rawhide example: https://koji.fedoraproject.org/koji/taskinfo?taskID=19487191 EPEL example: https://koji.fedoraproject.org/koji/taskinfo?taskID=19486424
So, since this is a pretty large codebase (that I am not familiar with) I am going to ask upstream if they can reproduce the issue, but before doing so I'm wondering if anyone here can think of any reason why our builders can determine such a behavior.
I know this kind of a long shot, but any idea to start debugging the issue will be appreciated.
So it seems some arches have a single test failure test3D, while others have a few more.
Looking at the history it build on all the arches with the 2016.09.4 release so it looks to me upstream has regressed between then and 2017.03.1 so I'd probably engage with upstream to see what's changed.
On Thu, 11 May 2017 15:51:17 +0200 Gianluca Sforna giallu@gmail.com wrote:
While trying to update one of my packages (rdkit), I am hitting an issue with its tests suite, so apparently every non-x86_64 build is failing.
rawhide example: https://koji.fedoraproject.org/koji/taskinfo?taskID=19487191 EPEL example: https://koji.fedoraproject.org/koji/taskinfo?taskID=19486424
So, since this is a pretty large codebase (that I am not familiar with) I am going to ask upstream if they can reproduce the issue, but before doing so I'm wondering if anyone here can think of any reason why our builders can determine such a behavior.
I know this kind of a long shot, but any idea to start debugging the issue will be appreciated.
after a very brief look I expect few kinds of issues - math precision (test3D) - big endian issues (visible on ppc64 and s390x) - others (i686)
Dan