[Python-modules-team] Bug#848758: Latest upgrade of Numpy breaks tests of other packages (Was: Bug#848758: python-skbio: FTBFS: Test failures)

[Sandro Tosi]

On Wed, Dec 21, 2016 at 5:59 AM, Andreas Tille <andreas at an3as.eu> wrote:
> Hi Adrian and Anton,
>
> On Wed, Dec 21, 2016 at 11:40:16AM +0200, Adrian Bunk wrote:
>> On Tue, Dec 20, 2016 at 09:01:31AM -0500, Sandro Tosi wrote:
>> > > it seems the recent upgrade of numpy has broken some tests in other
>> > > packages like for instance this one in python-skbio.  I wonder whether
>> > > you are either able to suggest patches to get the tests working with the
>> > > new interface of numpy again or whether it might be sensible to revert
>> > > the upgrade of python-numpy to the previous version considering that we
>> > > are quite close to the freeze.
>> >
>> > There is still a month to the freeze for the existing package, so
>> > let's not panic already. I'd suggest contacting the upstream of the
>> > projects with failing test and ask for support
>>
>> For python-skbio it is *really* time to panic *right now*.
>>
>> python-skbio is currently not in testing.

the last upload (0.5.1-1) was done on Nov 19 and skbio hasnt been in
testing for more than 11 months. if it hasnt reached testing by now,
it must have had several other issues.

>>
>> For python-skbio to have any chance of being part of stretch, the
>> latest possible date for an upload of a package without any RC bugs
>> is Christmas Day (and then not touch the package for 10 days).
>>
>> If this issue is not resolved by Christmas, your Christmas present for
>> the python-skbio maintainers is that their package will definitely not
>> be part of stretch.

sarcasm wont help you make your point.

>
> Thanks for confirming that I was not actually panicing. ;-)
>
> Sandro, would you reconsider my suggestion to revert this
> mini-transition?  I do not see a realistic chance that upstream will
> come up with fixes in the next two days.

how did you reach this conclusion? before i reported this bug upstream
(why didnt you do it?) they probably werent even aware of the issue.

I think you have multiple actions you can do:

* actually investigate what the issue is, did you try to do it? you
might also find out the fix is easy and end up submitting a patch
upstream
* engage with upstream to get this fixed by them and backport the
patch to the current package (if they prefer not to release a new
version now)
* skip that test / dont make the build fail if there is a test failure
(for now!). Test suites are made to catch errors, and in this case it
worked! expecting the test suite to pass every time is either
unrealistic or an indication the suite is too shallow. grab
information about the failure and report them upstream, but in the
meantime (since it's a known issue) you can decide to skip the failure

you say you are in a hurry, and yet you decided to waste a whole day
in this blame-game, i'm sure there is much more you can do to fix
skbio.

-- 
Sandro "morph" Tosi
My website: http://sandrotosi.me/
Me at Debian: http://wiki.debian.org/SandroTosi
G+: https://plus.google.com/u/0/+SandroTosi