Bug#913837: transition: hdf5

[Gilles Filippini <pini@debian.org>]

Sebastiaan Couwenberg a écrit le 01/12/2018 à 17:23 :
> On 12/1/18 5:16 PM, Gilles Filippini wrote:
>> On Fri, 30 Nov 2018 15:53:29 +0100 Gilles Filippini <pini@debian.org> wrote:
>>> On 2018-11-30 14:31, Emilio Pozuelo Monfort wrote:
>>>> On 17/11/2018 11:54, Emilio Pozuelo Monfort wrote:
>>>>> Control: tags -1 confirmed
>>>>>
>>>>> On 15/11/2018 21:37, Gilles Filippini wrote:
>>>>>> Package: release.debian.org
>>>>>> Severity: normal
>>>>>> User: release.debian.org@packages.debian.org
>>>>>> Usertags: transition
>>>>>>
>>>>>> Hi Release Team,
>>>>>>
>>>>>> I hereby request a transition slot for hdf5 1.10.4 currently in 
>>>>>> experimental.
>>>>>>
>>>>>> Ben file:
>>>>>>
>>>>>> title = "hdf5";
>>>>>> is_affected = .depends ~ /libhdf5/ | .build-depends ~ /hdf5/;
>>>>>> is_good = .depends ~ 
>>>>>> /libhdf5-103|libhdf5-openmpi-103|libhdf5-mpich-103/;
>>>>>> is_bad = .depends ~ 
>>>>>> /libhdf5-100|libhdf5-openmpi-100|libhdf5-mpich-100/;
>>>>>>
>>>>>> I've checked the build of all the reverse dependencies against this 
>>>>>> release, and from the 110+ of them, only those - which are not in 
>>>>>> testing - aren't binnmu ready:
>>>>>
>>>>> Go ahead.
>>>>
>>>> gnudatelanguage's autopkgtests fail with the new hdf5/netcdf, which 
>>>> block
>>>> netcdf's testing migration. Can you take a look?
>>>>
>>>> https://ci.debian.net/packages/g/gnudatalanguage/testing/amd64/
>>>
>>> I'll have a look this w-e, but I'm not very optimistic because these 
>>> failures are segmentation faults.
>>> Since gnudatalanguage has a not so high popcon, an option would be to 
>>> temporarily remove it from testing, to gain time to investigate the 
>>> problem.
>>
>> I really don't know how to debug this.
> 
> You shouldn't have to, you're not the maintainer of that package.
> 
>> What I can say is that it isn't related to the current HDF5 1.10.4 transition: I've rebuilt gnudatalanguage against HDF5 1.10.0-patch1+docs-4+b2 from testing (with netcdf and grib-api rebuilt as well) and the very same failure occurs, at least for TEST_POINT_LUN:
> 
> gnudatalanguage has been problematic in a prior netcdf transition too.
> 
> I would just file an RC bug for the failing tests and let the maintainer
> take it from there. If they are unable to fix the issue, the package
> will be autoremoved from testing.

I've just opened #915207.

Thanks,

_g.