Upcoming Debian freeze

[Samuel Lelièvre]

Dear sage-devel,

The freeze period for the next Debian release starts on
12 January 2019, as discussed at

https://trac.sagemath.org/ticket/22626#comment:118

This conditions what functionality users of Debian and its
derivatives (including Ubuntu) will find in Sage for the next
two years, if they install from their distribution's package
manager.

It might make sense for Sage developers to brainstorm on
priorities for what needs to be integrated by then.

One priority is definitely making progress on updating GAP,
either using GAP 4.10 which should be announced any day
now and will provide its own LibGAP, or working on our own
libGAP based on GAP 4.9.3. Failing that, Sage might end up
*not being in the next Debian release at all*, if I understand
correctly. That would be a shame, given the number of users
of Debian and its derivatives (including Ubuntu).

Can you think of other things that should be given high priority?

Volker, could you give a provisional timeline for the upcoming
Sage release? Having a provisional timeline at the start of the
release cycle, and an update to this provisional timeline near
the end of the release cycle, is helpful for prioritising and for
synchronising the Sage-Windows release.

Kind regards,
Samuel

[David Roe]

On Mon, Nov 5, 2018 at 7:02 AM Samuel Lelièvre <samuel....@gmail.com> wrote:

Dear sage-devel,

The freeze period for the next Debian release starts on
12 January 2019, as discussed at

  https://trac.sagemath.org/ticket/22626#comment:118

This conditions what functionality users of Debian and its
derivatives (including Ubuntu) will find in Sage for the next
two years, if they install from their distribution's package
manager.

It might make sense for Sage developers to brainstorm on
priorities for what needs to be integrated by then.

One priority is definitely making progress on updating GAP,
either using GAP 4.10 which should be announced any day
now and will provide its own LibGAP, or working on our own
libGAP based on GAP 4.9.3. Failing that, Sage might end up
*not being in the next Debian release at all*, if I understand
correctly. That would be a shame, given the number of users
of Debian and its derivatives (including Ubuntu).

Can you think of other things that should be given high priority?

Another obvious thing is the switch to Python 3, which seems to have been making good progress.  Perhaps we could have some work sessions on Zulip focused on Python 3?  Frederic, how feasible would it be to finish the transition by mid January if you got some help?  I'm happy to work on it more than I have been.

David

 

Volker, could you give a provisional timeline for the upcoming
Sage release? Having a provisional timeline at the start of the
release cycle, and an update to this provisional timeline near
the end of the release cycle, is helpful for prioritising and for
synchronising the Sage-Windows release.

Kind regards,
Samuel

--
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+...@googlegroups.com.
To post to this group, send email to sage-...@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.

[Frédéric Chapoton]

Hello,

Being somewhat pessimistic, I would say that chances are very small that sage will be fully python3 compliant by then. For the moment, only 2 or 3 people are actively working on the problem. Erik and Jeroen have apparently turned their attention to other important matters. There remains on the one hand many small annoying issues to be taken care of one by one, and on the other some difficult large-scale issues ("round,ceil,floor in python semantics" and "graphs sort their vertices" and "modular symbols with another basis" to just name 3).

What we should do is just go on, and then sage will be as much compatible as we could. As I already said somewhere, help is welcome.

Frederic

[Daniel Krenn]

On 2018-11-06 10:04, Frédéric Chapoton wrote:
> ("round,ceil,floor in python semantics"

I am curious: What is this about?

Daniel

[Erik Bray]

On Mon, Nov 5, 2018 at 8:12 PM David Roe <roed...@gmail.com> wrote:
>
>
>
> On Mon, Nov 5, 2018 at 7:02 AM Samuel Lelièvre <samuel....@gmail.com> wrote:
>>
>> Dear sage-devel,
>>
>> The freeze period for the next Debian release starts on
>> 12 January 2019, as discussed at
>>
>> https://trac.sagemath.org/ticket/22626#comment:118
>>
>> This conditions what functionality users of Debian and its
>> derivatives (including Ubuntu) will find in Sage for the next
>> two years, if they install from their distribution's package
>> manager.
>>
>> It might make sense for Sage developers to brainstorm on
>> priorities for what needs to be integrated by then.
>>
>> One priority is definitely making progress on updating GAP,
>> either using GAP 4.10 which should be announced any day
>> now and will provide its own LibGAP, or working on our own
>> libGAP based on GAP 4.9.3. Failing that, Sage might end up
>> *not being in the next Debian release at all*, if I understand
>> correctly. That would be a shame, given the number of users
>> of Debian and its derivatives (including Ubuntu).
>>
>> Can you think of other things that should be given high priority?
>
>
> Another obvious thing is the switch to Python 3, which seems to have been making good progress. Perhaps we could have some work sessions on Zulip focused on Python 3? Frederic, how feasible would it be to finish the transition by mid January if you got some help? I'm happy to work on it more than I have been.
> David

While I agree that continuing work on the Python 3 port is high
priority, I agree with Frédéric's assessment that doing that before
the Debian freeze, much less having a Sage release that is fully
Python 3 compatible before the freeze, in infeasible.

Fortunately I don't think that's so important for the Debian
case--it's not like they'll be dropping the Python 2 package
ecosystem, are they?

It sounds like the GAP stuff might be the most important. If we can
get Sage + GAP 4.10 with the official libGAP working, the we should
prioritize making a SageMath 8.5 release for that alone if nothing
else.

[Erik Bray]

Also, Samuel, could you forward this to debian-science-sage mailing
list?On Mon, Nov 5, 2018 at 1:02 PM Samuel Lelièvre

[Frédéric Chapoton]

Well, python3 builtin "round" calls the method "__round__" of the objects, which does not exist for sage objects so far and which is expected to do something different from what sage does with its methods .round

See ticket https://trac.sagemath.org/ticket/25827

F

[Erik Bray]

On Tue, Nov 6, 2018 at 10:41 AM Frédéric Chapoton <fchap...@gmail.com> wrote:
>
> Well, python3 builtin "round" calls the method "__round__" of the objects, which does not exist for sage objects so far and which is expected to do something different from what sage does with its methods .round
>
> See ticket https://trac.sagemath.org/ticket/25827

I don't think it's a very difficult or important problem. Just one
that needs to be sorted out in a sane and consistent manner and for
some reason we haven't been able to get a useful discussion about it
going despite your efforts to bring it up :(

> Le mardi 6 novembre 2018 10:31:02 UTC+1, Daniel Krenn a écrit :
>>
>> On 2018-11-06 10:04, Frédéric Chapoton wrote:
>> > ("round,ceil,floor in python semantics"
>>
>> I am curious: What is this about?
>>
>> Daniel
>

[Thierry]

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

[Tobias Hansen]

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.

Best,

Tobias


[1] https://tracker.debian.org/pkg/sagemath

[2] https://people.debian.org/~thansen/sage-test-status.html

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

[E. Madison Bray]

Could somebody help explain exactly what the difference is between the
"transition freeze" and the "soft freeze"? 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.

I'm building Sage on Debian unstable now and will look into some of
the remaining doctest failures in the meantime.

[E. Madison Bray]

Sage 8.6 will also require an update of cysignals to 1.8.1. Is that
in the works? How would I ago about updating that package?

[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 <tha...@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.
>

[E. Madison Bray]

On Mon, Jan 7, 2019 at 4:20 PM Jerome BENOIT <calc...@rezozer.net> wrote:
>
> Hello,

>
> On 07/01/2019 18:07, E. Madison Bray wrote:
> > Sage 8.6 will also require an update of cysignals to 1.8.1. Is that
> > in the works? How would I ago about updating that package?
>

> I will take care of the packaging of cygsignals 1.8.1 the next week-end.
> Basically before the return of Tobias, so it should be fine.

Thanks!

> Note that the packaging of cygsygnals is tricky, so I would prefer to manage it.

Out of curiosity, what's tricky about it? I noticed that there's a
python-cysignals-base, and python-cysignals-pari (the latter being the
one required for Sage). Does it have to do with that? I wonder if
there's anything that could make it easier (it really *shouldn't* be
hard...)

> I will also try to package singular 4.1.1-p4 .
>
> Concerning GAP, is there any missing key GAP package that is not yet packaged by Debian ?

Not to my knowledge, but I'm not an expert on GAP packages.

[E. Madison Bray]

Getting back on the general topic....

On Mon, Jan 7, 2019 at 3:09 PM Thierry <sage-goo...@lma.metelu.net> wrote:
> 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 ?

As I commented on #27000, the main problem with Numpy 1.16 seems to
just be deprecation warnings that originate from matplotlib (use of
np.asscalar()). As far as I can tell there aren't any Numpy
DeprecationWarnings originating from Sage itself. I could be wrong,
but I don't see any other obvious problems resulting from using Sage
with Numpy 1.16.

The bug is already fixed in matplotlib
(https://github.com/matplotlib/matplotlib/pull/12478) and has been
backported to v2.2.4, so at most all we need is a patch in the
matplotlib package.

[Dima Pasechnik]

Perhaps we can also include in 8.6 the update to sagetex which is
currently under review on
https://trac.sagemath.org/ticket/27024 and has Debian patches merged,
and appears to work on py3 too
thanks to many other patches...

[Dima Pasechnik]

sagetex 3.1 does not correspond to anything on the upstream repo. I
pulled it as it was broken in various ways, don't use it.
Please switch to https://github.com/sagemath/sagetex/releases/tag/3.2

Thanks,
Dima

On Thu, Jan 10, 2019 at 2:07 AM Jerome BENOIT <calc...@rezozer.net> wrote:
>
> Hello,
>

> On 09/01/2019 21:40, Dima Pasechnik wrote:
> > Perhaps we can also include in 8.6 the update to sagetex which is
> > currently under review on
> > https://trac.sagemath.org/ticket/27024 and has Debian patches merged,
> > and appears to work on py3 too
> > thanks to many other patches...
>

> SageTeX is managed as a different package.
>
> Whatever, sagetex 3.1 is already in Debian. I was not aware of version 3.2 thanks.
>
> It is not clear to me whether or not it make sense to make Python 3 version of it
> given that SageMath seems to run with Python.
>
>
>
> Cheers,
> Jerome
>

[Dima Pasechnik]

Could we merge the sagetex update on #27024 into 8.6, still?
As we see, Debian is already picking up this update.

Thanks,
Dima

On Thu, Jan 10, 2019 at 3:41 PM Jerome BENOIT <calc...@rezozer.net> wrote:

>
>
>
> On 10/01/2019 10:31, Dima Pasechnik wrote:
> > sagetex 3.1 does not correspond to anything on the upstream repo. I
> > pulled it as it was broken in various ways, don't use it.
> > Please switch to https://github.com/sagemath/sagetex/releases/tag/3.2
>

> SageTeX 3.2 on its way to Sid
>
> Jerome

>
> >
> > Thanks,
> > Dima
> >
> > On Thu, Jan 10, 2019 at 2:07 AM Jerome BENOIT <calc...@rezozer.net> wrote:
> >>
> >> Hello,
> >>
> >> On 09/01/2019 21:40, Dima Pasechnik wrote:
> >>> Perhaps we can also include in 8.6 the update to sagetex which is
> >>> currently under review on
> >>> https://trac.sagemath.org/ticket/27024 and has Debian patches merged,
> >>> and appears to work on py3 too
> >>> thanks to many other patches...
> >>
> >> SageTeX is managed as a different package.
> >>
> >> Whatever, sagetex 3.1 is already in Debian. I was not aware of version 3.2 thanks.
> >>
> >> It is not clear to me whether or not it make sense to make Python 3 version of it
> >> given that SageMath seems to run with Python.
> >>
> >>
> >>
> >> Cheers,
> >> Jerome
> >>
>

> --
> Jerome BENOIT | calculus+at-rezozer^dot*net
> https://qa.debian.org/developer.php?login=calc...@rezozer.net
> AE28 AE15 710D FF1D 87E5 A762 3F92 19A6 7F36 C68B
>

[Jeroen Demeyer]

On 2019-01-13 15:34, Dima Pasechnik wrote:
> As we see, Debian is already picking up this update.

If Debian *already* has it anyway, what's the point of upgrading it in
SageMath? Especially if it doesn't require any changes to the Sage library.

[Dima Pasechnik]

to save them grief of having to deal with being ahead of Sage master
at some points?

[Samuel Lelievre]

Mon 2019-01-07 14:09:18 UTC, Thierry:

>

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

>

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

>

> [...]

>

> Tobias, what would be the other problematic packages or issues?

> What should we focus on?

There is now a "Sage Buster roadmap":

https://wiki.debian.org/DebianScience/Sage#Buster_roadmap

[E. Madison Bray]

FWIW I added some notes next to the bullet point about gap-io.
There's a link there to a patch I proposed a few weeks ago which
should fix the root problem, at least on Debian.

[Dima Pasechnik]

Erik,
we have a ticket for GAP's IO package, updating your patch:
https://trac.sagemath.org/ticket/26930
- ready for review

[Dima Pasechnik]

In case it matters, I have just uploaded Sagetex 3.2 to CTAN.
Not sure whether Debian does any synchronisation of Sagetex@CTAN with
Sagtex @ sage - hopefully it does not break anything for you.

On Thu, Jan 10, 2019 at 3:41 PM Jerome BENOIT <calc...@rezozer.net> wrote:
>
>
>
> On 10/01/2019 10:31, Dima Pasechnik wrote:

> > sagetex 3.1 does not correspond to anything on the upstream repo. I
> > pulled it as it was broken in various ways, don't use it.
> > Please switch to https://github.com/sagemath/sagetex/releases/tag/3.2
>

> SageTeX 3.2 on its way to Sid
>
> Jerome
>
> >

> > Thanks,
> > Dima
> >
> > On Thu, Jan 10, 2019 at 2:07 AM Jerome BENOIT <calc...@rezozer.net> wrote:
> >>
> >> Hello,
> >>
> >> On 09/01/2019 21:40, Dima Pasechnik wrote:
> >>> Perhaps we can also include in 8.6 the update to sagetex which is
> >>> currently under review on
> >>> https://trac.sagemath.org/ticket/27024 and has Debian patches merged,
> >>> and appears to work on py3 too
> >>> thanks to many other patches...
> >>
> >> SageTeX is managed as a different package.
> >>
> >> Whatever, sagetex 3.1 is already in Debian. I was not aware of version 3.2 thanks.
> >>
> >> It is not clear to me whether or not it make sense to make Python 3 version of it
> >> given that SageMath seems to run with Python.
> >>
> >>
> >>
> >> Cheers,
> >> Jerome
> >>
>