[Python-modules-team] New version of python-numpy breaks autopkgtests of python-scipy in testing

[Julian Taylor]

I'll look into it tomorrow, scipy is due for an update anyway.

On 12.05.2018 21:41, Paul Gevers wrote:
> Hi Sandro,
> 
> On 12-05-18 21:29, Sandro Tosi wrote:
>> the np.random.dirichlet errors seem a bug in how scipy calls that function:
>> https://docs.scipy.org/doc/numpy-1.14.0/reference/generated/numpy.random.dirichlet.html
> 
> Can you file a bug against python-scipy for this then?
> 
>> the other issues with numerical values are hard to understand from my POV -
>> can you get upstream involved?
> 
> Not me, but I hope some of the people in my original TO. I have no
> knowledge at all about these python packages or their upstream, that is
> why I sent out the mail to start with.
> 
> Paul
> 
>> On Tue, May 8, 2018 at 1:51 PM Paul Gevers <elbrus at debian.org> wrote:
>>
>>> Dear maintainers,
>>
>>> [This e-mail is automatically sent. V2 (20180508)]
>>
>>> As recently announced [1] Debian is now running autopkgtests in testing
>>> to check if the migration of a new source package causes regressions. It
>>> does this with the binary packages of the new version of the source
>>> package from unstable.
>>
>>> With a recent upload of python-numpy the autopkgtest of python-scipy
>>> started to fail in testing [2]. This is currently delaying the migration
>>> of python-numpy version 1:1.14.3-2 [3].
>>
>>> This e-mail is meant to trigger prompt direct communication between the
>>> maintainers of the involved packages as one party has insight in what
>>> changed and the other party insight in what is being tested. Please
>>> therefore get in touch with each other with your ideas about what the
>>> causes of the problem might be, proposed patches, etc. A regression in a
>>> reverse dependency can be due to one of the following reasons (of course
>>> not complete):
>>> * new bug in the candidate package (fix the package)
>>> * bug in the test case that only gets triggered due to the update (fix
>>>    the reverse dependency, but see below)
>>> * out-of-date reference date in the test case that captures a former bug
>>>    in the candidate package (fix the reverse dependency, but see below)
>>> * deprecation of functionality that is used in the reverse dependency
>>>    and/or its test case (discussion needed)
>>> Triaging tips are being collected on the Debian Wiki [4].
>>
>>> Unfortunately sometimes a regression is only intermittent. Ideally this
>>> should be fixed, but it may be OK to just have the autopkgtest retried
>>> (a link is available in the excuses [3]).
>>
>>> There are cases where it is required to have multiple packages migrate
>>> together to have the test cases pass, e.g. when there was a bug in a
>>> regressing test case of a reverse dependency and that got fixed. In that
>>> case the test cases need to be triggered with both packages from
>>> unstable (reply to this e-mail and/or contact the ci-team [5]) or just
>>> wait until the aging time is over (if the fixed reverse dependency
>>> migrates before that time, the failed test can be retriggered [3]).
>>
>>> Of course no system is perfect. In case a framework issue is suspected,
>>> don't hesitate to raise the issue via BTS or to the ci-team [5] (reply to
>>> me is also fine for initial cross-check).
>>
>>> To avoid stepping on peoples toes, this e-mail does not automatically
>>> generate a bug in the BTS, but it is highly recommended to forward this
>>> e-mail there (psuedo-header boilerplate below [6,7]) in case it is
>>> clear which package should solve this regression.
>>
>>> It can be appropriate to file an RC bug against the depended-on package,
>>> if the regression amounts to an RC bug in the depending package, and to
>>> keep it open while the matter is investigated. That will prevent
>>> migration of an RC regression.
>>
>>> If the maintainers of the depending package don't have available effort
>>> to fix a problem, it is appropriate for the maintainers of the
>>> depended-on package to consider an NMU of the depending package. Any
>>> such an NMU should take place in accordance with the normal NMU rules.
>>
>>> Neither of the above steps should be seen as hostile; they are part of
>>> trying to work together to keep Debian in tip-top shape.
>>
>>> If you find that you are not able to agree between you about the right
>>> next steps, bug severities, etc., please try to find a neutral third
>>> party to help you mediate and/or provide a third opinion. Failing that
>>> your best bet is probably to post to debian-devel.
>>
>>> [1] https://lists.debian.org/debian-devel-announce/2018/05/msg00001.html
>>> [2] https://ci.debian.net/packages/p/python-scipy/testing/amd64/
>>> [3] https://qa.debian.org/excuses.php?package=python-numpy
>>> [4] https://wiki.debian.org/ContinuousIntegration/TriagingTips
>>> [5] #debci on oftc or debian-ci at lists.debian.org
>>> [6] python-numpy has an issue
>>> ============
>>> Source: python-numpy
>>> Version: 1:1.14.3-2
>>> Severity: normal or higher
>>> Control: affects -1 src:python-scipy
>>> User: debian-ci at lists.debian.org
>>> Usertags: breaks
>>> ============
>>> [7] python-scipy has an issue
>>> ============
>>> Source: python-scipy
>>> Version: 0.19.1-2
>>> Severity: normal or higher
>>> Control: affects -1 src:python-numpy
>>> User: debian-ci at lists.debian.org
>>> Usertags: needs-update
>>> ============
>>
>>
>>
>