Bug#844924: pygobject: FTBFS: Tests failures

[Emilio Pozuelo Monfort]

On 20/11/16 02:58, Michael Biebl wrote:
> Am 20.11.2016 um 02:45 schrieb Emilio Pozuelo Monfort:
>> On 20/11/16 01:44, Michael Biebl wrote:
>>> Am 19.11.2016 um 08:03 schrieb Lucas Nussbaum:
>>>> Source: pygobject
>>>> Version: 3.22.0-1
>>>> Severity: serious
>>>> Tags: stretch sid
>>>> User: debian-qa at lists.debian.org
>>>> Usertags: qa-ftbfs-20161118 qa-ftbfs
>>>> Justification: FTBFS on amd64
>>>>
>>>> Hi,
>>>>
>>>> During a rebuild of all packages in sid, your package failed to build on
>>>> amd64.
>>>>
>>>> Relevant part (hopefully):
>>>>> if 'generic-c-marshaller' in GObject.features:
>>>>> ^
>>>>> pygtkcompat/pygtkcompat.py:102:1: E305 expected 2 blank lines after class or function definition, found 1
>>>>> _unset = object()
>>>>> ^
>>>
>>> Looks like pep8 got stricter (again).
>>> There were some new upstream releases just recently.
>>> I got some build failures in other packages as well.
>>>
>>> Somehow I think it would be good to avoid such uploads so close to the
>>> freeze or better coordinate uploads of pep8.
>>>
>>> Would it be possible to re-compile reverse build-dependencies of pep8
>>> before a new major upstream release? Somehow this feels like it should
>>> be treated like a (library) transtion.
>>
>> I don't think so.
> 
> No to what exactly?

Sorry for the confusion. I don't think that this needs to be treated as a
library transition.

If rdeps abort on pep8 errors, then maybe they shouldn't do that. Just like
using -Werror is calling for trouble. Or they need to leave with the occasional
FTBFS bug.

Cheers,
Emilio