[Debian-science-sagemath] [sage-devel] Upcoming Debian freeze

[Thierry]

Hi,

On Mon, Jan 07, 2019 at 02:38:40PM +0100, E. Madison Bray wrote:
> On Sun, Jan 6, 2019 at 11:28 AM Tobias Hansen <thansen at debian.org> wrote:
> >
> > On 1/5/19 1:18 PM, Thierry wrote:
> > > Hi,
> > >
> > > could Debian maintainers please explicitely tell us on the sage-devel
> > > mailing-list what should be done soon so that 8.6 could enter forthcoming
> > > buster release with recent dependencies (gap 4.10, etc) ? Also, what are
> > > the deadlines ?
> > >
> > > In particular, i noticed that networkx in buster will be version 2.2 [1],
> > > so should #26326 be set as blocker ?
> > >
> > > I noticed that sagemath disapeared from testing [2], is there a way we
> > > could help fixing that with 8.6 ?
> > >
> > > Ciao,
> > > Thierry
> > >
> > > [1] https://packages.debian.org/buster/python-networkx
> > > [2] https://packages.debian.org/cgi-bin/search_packages.pl?keywords=sagemath&searchon=names&subword=1&version=all&release=all
> > >
> > Hi,
> >
> > I think the main bottleneck at the moment is manpower for working directly on the sagemath package. I'm on vacation until January 13. Right now the sagemath 8.4 (with patched in support for gap 4.10 and networkx 2.2 etc) builds in Debian unstable and should be on the way to migrate to testing. [1] Since it was built with numpy 1.16 it can only migrate to testing when that migrates. Since I'm short on time I could just make the package build, but did not have much time to look at the doctests that are still failing. See [2] for an overview and links to the build logs. If someone could check for important failures (and fixes), that would be appreciated. If people could test the package and report bugs that would be helpful too.
> >
> > The freeze is described in [3]. The transition freeze on January 12 does probably not affect us much, but the soft freeze on February 12 means that sagemath must be in testing before that date. Preferably sagemath 8.6. I already updated the package to 8.6.beta1 in git and could upload 8.6.* to Debian experimental soon to make build logs available.
> >
> > Directly on the sage side I think the only thing you can do to help is to release 8.6 in time (maybe around January 17?) so that it can be uploaded to unstable soon and migrate to Debian testing before the soft freeze on February 12. Testing migration is not very predictable. New bug reports or uploads of other packages can cause unexpected dalays.
> 
> Could somebody help explain exactly what the difference is between the
> "transition freeze" and the "soft freeze"?

In his email, Tobias pointed the link (after the signature):

[3] https://release.debian.org/buster/freeze_policy.html

> I was under the impression
> that we were working more up against the transition freeze on January
> 12.  Is the idea there just to not make any major package version
> changes that break other packages?  For Sage that's relatively easy
> since it's near the bottom of the dependency tree.  8.6.rc0 is out
> now, so 8.6 should be out in time for that.

My understanding of what he said is (please tell me if i am wrong) :

- currently sage disapeared from testing because of dependencies and
  doctests issues.

- we can help by making a 8.6 that is consistent with buster regarding
  dependencies version (e.g. gap 4.10, networkx 2.2, numpy 1.16), so that
  they won't have to patch sage themselves to let things work.

- if at the soft freeze deadline (February 12th), sage did not succeed to
  enter to testing, it will not be part of Debian buster.

- for this, sage 8.6 has to be first in unstable for some time

- so our (sage-devel) deadline to have a nice 8.6 with good dependencies
  is around January 17th, so that sage-debian devs can use it and let it
  enter in unstable and the stretch before the soft freeze.

After a very quick look at the Debian build logs, regarding numpy, some
errors are due to te fact that 1.16 is not out yet (hence not in sage),
but Debian uses its release candidate.

Gap 4.10 is merged, networkx 2.2 is on the way (#26326), and we have to
wait for upstream for numpy 1.16 (stay tuned at #27000). Perhaps could we
start to make a branch there with 1.16.rc2 so that we can already work on
the issues and be faster when 1.16 will be out ?

Tobias, what would be the other problematic packages or isues ? On what
should we focus ?

Ciao,
Thierry


> I'm building Sage on Debian unstable now and will look into some of
> the remaining doctest failures in the meantime.
> 
> -- 
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscribe at googlegroups.com.
> To post to this group, send email to sage-devel at googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.