[Debian-science-sagemath] sagemath and Python 3

[Dima Pasechnik]

On Sun, Sep 8, 2019 at 11:41 AM Tobias Hansen <thansen at debian.org> wrote:
>
> On 9/6/19 2:10 PM, Tobias Hansen wrote:
> > On 9/6/19 1:30 PM, Dima Pasechnik wrote:
> >> On Fri, Sep 6, 2019 at 12:23 PM Tobias Hansen <thansen at debian.org> wrote:
> >>> On 9/4/19 8:55 PM, Julien Puydt wrote:
> >>>> Hi,
> >>>>
> >>>> Le 04/09/2019 à 18:25, Tobias Hansen a écrit :
> >>>>
> >>>>> It seems that the only dependency that needs to be changed is pynac. It is currently linked to libpython2.7 and this causes a segfault in sage. It can be switched to Python 3 by changing the Build-Dependency to python3-dev and passing PYTHON=python3 to configure. However, that seems to require a soname change (especially because the current one was also used in buster), so what do we do, custom soname until the next time upstream bumps it?
> >>>> I choose to ask upstream: https://github.com/pynac/pynac/issues/346 !
> >>>>
> >>>> JP
> >>> I pushed the changes to the git repo. I could use some help figuring out the reasons for the ~126 failing doctests.
> >>>
> >>> Another issue is that the build log is 25 MB big because the docbuild prints this message all over:
> >>>
> >>> RemovedInSphinx20Warning: formatargspec() is now deprecated.  Please use sphinx.util.inspect.Signature instead.
> >>>
> >> Aren't you far behind the develop branch? Note that Sage 8.8 is far
> >> from being fully ported to Python 3.
> >> Sage 8.9.beta9 is much closer. (and this warning is gone too)
> >>
> >> Dima
> >
> > Ok, thanks for the hint. I think I will just try the beta then.
> >
> > The (almost) finished Python 3 port is coming just at the right time. The Python 2 removal in Debian has begun and sage is blocking lots of Python 2 packages from being removed. And even though this was not supposed to happen, the first Python 2 dependencies of sage were already removed from Debian.
> >
> > Best,
> >
> > Tobias
>
>
> Updating sagemath to 8.9.beta9 improved things. There are now 48 failing tests. The log is at [1]. Does anything look familiar to someone?
>
on a stock 8.9.beta9 with Python3 one gets roughly the following from ptestlong:
----------------------------------------------------------------------
sage -t --long --warn-long 53.7
src/sage/combinat/root_system/weyl_group.py  # 7 doctests failed
sage -t --long --warn-long 53.7 src/sage/libs/eclib/interface.py  # 1
doctest failed
sage -t --long --warn-long 53.7
src/sage/rings/polynomial/polynomial_rational_flint.pyx  # 1 doctest
failed
sage -t --long --warn-long 53.7 src/sage/repl/attach.py  # 4 doctests failed
sage -t --long --warn-long 53.7 src/sage/libs/singular/function.pyx  #
1 doctest failed
sage -t --long --warn-long 53.7
src/sage/numerical/backends/generic_backend.pyx  # 2 doctests failed
----------------------------------------------------------------------

So 10-15 tests are failing due to Sage's bugs, and the rest are likely
Debian's own - probably different package versions of some components
lead to this.


> Apart from the pynac rebuild this requires updated versions of givaro, fflas-ffpack and linbox.
>
> We should also update three.js to r100. 3d plots don't work, probably because of the version mismatch.
>
> Best,
>
> Tobias
>
> [1] https://salsa.debian.org/science-team/sagemath/snippets/324
>
>
>
> _______________________________________________
> Debian-science-sagemath mailing list
> Debian-science-sagemath at alioth-lists.debian.net
> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-science-sagemath