[Debian-science-sagemath] Sage with gap 4.10 status

[Tobias Hansen]

On 1/3/19 7:16 PM, E. Madison Bray wrote:
> On Thu, Jan 3, 2019 at 12:44 PM Tobias Hansen <thansen at debian.org> wrote:
>> On 1/2/19 3:20 PM, E. Madison Bray wrote:
>>> On Wed, Jan 2, 2019 at 1:48 PM Tobias Hansen <thansen at debian.org> wrote:
>>>> On 1/2/19 1:07 PM, E. Madison Bray wrote:
>>>>> On Wed, Jan 2, 2019 at 12:57 PM Tobias Hansen <thansen at debian.org> wrote:
>>>>>> From the gap 4.10 ticket I take that you install
>>>>>>
>>>>>> + pkg/autpgrp-* \
>>>>>> + pkg/alnuth-* \
>>>>>> + pkg/crisp-* \
>>>>>> + pkg/ctbllib \
>>>>>> + pkg/FactInt-* \
>>>>>> + pkg/fga \
>>>>>> + pkg/irredsol-* \
>>>>>> + pkg/laguna-* \
>>>>>> + pkg/polenta-* \
>>>>>> + pkg/polycyclic-* \
>>>>>> + pkg/resclasses-* \
>>>>>> + pkg/sophus-* \
>>>>>> + pkg/tomlib-* \
>>>>>> and atlasrep.
>>>>>>
>>>>>> When I installed more gap libraries (gap-atlasrep, gap-autpgrp, gap-factint, gap-laguna, gap-polycyclic) I got an error during the docbuild (see attached sagemath_docbuild_error.build).
>>>>> Did you also include the patches from
>>>>> https://trac.sagemath.org/ticket/26856 and its dependency
>>>>> https://trac.sagemath.org/ticket/26889 ?
>>>>> This updated support for various additional GAP packages with 4.10,
>>>>> including all the packages you listed above, among others.
>>>> No, I didn't. I am running a test build with these two patches now. Are there any gap libs that change particularly many test results when they are installed? Ideally the sagemath package shouldn't depend on additional gap libs (apart from the package gap-libs).
>>> It's not so much a question of changing test results (although
>>> obviously that is a side-effect) as there are GAP packages that change
>>> the behavior of GAP, sometimes in significant ways.  Therefore code
>>> that uses GAP sometimes may need additional logic to account for
>>> differences in results that occur depending on what packages are
>>> loaded.  It's quite frustrating and confusing.
>> Ok, adding the two patches and depending on the Debian packages gap-primgrp, gap-smallgrp and gap-transgrp I get much less test failures. I uploaded an updated sage 8.4-4.
>>
>> I also updated the package to 8.6.beta1 (in the branch 8.6), skipping 8.5 because it's better to avoid the large gap related patches. I uploaded cython 0.29.2 and cypari2 1.3.1 to experimental. The only other change that sage 8.6.beta1 needs to get through compilation is the patch using_namespace_std.patch for lcalc, which breaks sage 8.4. If anyone wants to test, this needs to be applied locally for now.
> Thanks!
>
>> The main reason 8.6.beta1 has significantly more test failures than 8.4 seem to be many variations of this error:
>>
>> File "src/sage/misc/bindable_class.py", line 180, in sage.misc.bindable_class.BoundClass
> <...snip...>
>>     ApplicationError: Source directory and destination directory cannot be identical
>> If we could fix this we could upload sage 8.6 to experimental.
> I have no idea about that.  I wasn't involved in anything that would
> have affected that.  I'm slightly annoyed that the release manager has
> included so many additional changes in 8.6 that have nothing to do
> with GAP, but to my knowledge they're all just bug fixes.
>
> Could this be due to an incompatibility with the Sphinx version?

Since I skipped sage 8.5 it could also be a change in sage 8.5 that doesn't play well with something in Debian.