cython 0.10.2 problems

[Tim Abbott]

Ondrej,

The sagemath package got REJECTEd with another copyright triviality for me
to fix (this time, a missing complete statement of the CC-BY-SA in
debian/copyright; again, completely uninteresting).

However, I cannot re-upload with this fixed because Sage 3.0.5 does not
build against cython 0.10.2 (new in sid on December 2). The error is
available at: <http://pastebin.com/d54cc0f9a>.

I don't follow cython development very closely; do you think this is a bug
in the new version of cython, or a problem with the Sage code?

This obviously determines whether I should start packaging a newer version
of Sage or whether we just need to fix the cython bug.

-Tim Abbott

[mabshoff]

On Dec 13, 4:24 pm, Tim Abbott <tabb...@MIT.EDU> wrote:
> Ondrej,

Hi Tim,

Every time Cython is updated we have to fix bugs in the Sage library.
Since 3.0.5 we updated Cython twice, so see #3896 and #4476 for
patches that fix issues in the Sage library. I don't have a 3.0.5
around, so I cannot test if they apply cleanly, but it seems like less
work now to apply those fixes than to repackage 3.2.1. Sage 3.2.2 will
be out in a couple days and 3.3 will bring major changes the way,
switch to Sphinx for the documentation, removal of the doc and extcode
repo, so I would wait for that to happen.

>         -Tim Abbott

Cheers,

Michael

[Tim Abbott]

On Sat, 13 Dec 2008, mabshoff wrote:

> Every time Cython is updated we have to fix bugs in the Sage library.
> Since 3.0.5 we updated Cython twice, so see #3896 and #4476 for
> patches that fix issues in the Sage library. I don't have a 3.0.5
> around, so I cannot test if they apply cleanly, but it seems like less
> work now to apply those fixes than to repackage 3.2.1. Sage 3.2.2 will

Thanks, Michael. The patches did apply, though it seems there is at least
one remaining problem in the same vein (see below). Does it ring a bell?

> be out in a couple days and 3.3 will bring major changes the way,
> switch to Sphinx for the documentation, removal of the doc and extcode
> repo, so I would wait for that to happen.

Okay, I'll keep that in mind. When is 3.3 scheduled?

-Tim Abbott

Building sage/libs/linbox/linbox.cpp because it depends on
sage/libs/linbox/linbox.pyx.
python2.5 `which cython` --embed-positions --incref-local-binop
-I/home/tabbott/sagemath-3.0.5dfsg/deve
Error converting Pyrex file to C:
------------------------------------------------------------
...
mod_int **ans,
mod_int **B,
size_t B_nr, size_t B_nc)
cdef unsigned long rank(self) except -1

cdef mod_int det(self) except -1
^
------------------------------------------------------------

/home/tabbott/sagemath-3.0.5dfsg/devel/sage-main/sage/libs/linbox/linbox.pxd:20:20:
C method 'det' is declared but not defined
l/sage-main -o sage/libs/linbox/linbox.cpp sage/libs/linbox/linbox.pyx
sage: Error running cython.
sage: There was an error installing modified sage library code.

[mabshoff]

On Dec 14, 9:35 am, Tim Abbott <tabb...@MIT.EDU> wrote:
> On Sat, 13 Dec 2008, mabshoff wrote:

Hi Tim,

> > Every time Cython is updated we have to fix bugs in the Sage library.
> > Since 3.0.5 we updated Cython twice, so see #3896 and #4476 for
> > patches that fix issues in the Sage library. I don't have a 3.0.5
> > around, so I cannot test if they apply cleanly, but it seems like less
> > work now to apply those fixes than to repackage 3.2.1. Sage 3.2.2 will
>
> Thanks, Michael.  The patches did apply, though it seems there is at least
> one remaining problem in the same vein (see below).  Does it ring a bell?

Not really, I am CCing RobertWB - he might have a clue, but he won't
answer before Monday.

> > be out in a couple days and 3.3 will bring major changes the way,
> > switch to Sphinx for the documentation, removal of the doc and extcode
> > repo, so I would wait for that to happen.
>
> Okay, I'll keep that in mind.  When is 3.3 scheduled?

3.2.2 should be out by Tuesday, i.e. in about 3 days and 3.3 ought to
be rather short, i.e. 4 to 8 days depending on how many problems pop
up.

>         -Tim Abbott

Cheers,

Michael

> Building sage/libs/linbox/linbox.cpp because it depends on
> sage/libs/linbox/linbox.pyx.
> python2.5 `which cython` --embed-positions --incref-local-binop
> -I/home/tabbott/sagemath-3.0.5dfsg/deve
> Error converting Pyrex file to C:
> ------------------------------------------------------------
> ...
>                                  mod_int **ans,
>                                  mod_int **B,
>                                  size_t B_nr, size_t B_nc)
>      cdef unsigned long rank(self) except -1
>
>      cdef mod_int det(self) except -1
>                     ^
> ------------------------------------------------------------
>

> /home/tabbott/sagemath-3.0.5dfsg/devel/sage-main/sage/libs/linbox/linbox.px d:20:20:

[Ondrej Certik]

Hi Tim!

Sorry for the problems with Cython. As Michael has said, this should
be fixed in Sage, rather than downgrading Cython.

But I suggest to reupload something, so that they can finally approve
it. If it's trivial to fix it.

Ondrej

[mabshoff]

On Dec 15, 9:59 am, "Ondrej Certik" <ond...@certik.cz> wrote:
> Hi Tim!

Hi,

<SNIP>

> > This obviously determines whether I should start packaging a newer version
> > of Sage or whether we just need to fix the cython bug.
>
> Sorry for the problems with Cython. As Michael has said, this should
> be fixed in Sage, rather than downgrading Cython.
>
> But I suggest to reupload something, so that they can finally approve
> it. If it's trivial to fix it.

By the way: There should be a new stable Cython release (0.10.3) in
the very short term that fixes some Cython internal leak issues that
can greatly affect Sage. AFAIK it hasn't even been announced on the
Cython-dev list yet, but see #4798 for some details.

> Ondrej

Cheers,

Michael

[Tim Abbott]

On Mon, 15 Dec 2008, Ondrej Certik wrote:

> Sorry for the problems with Cython. As Michael has said, this should
> be fixed in Sage, rather than downgrading Cython.

Yeah, I wasn't thinking of a downgrade; the only question was about the
unlikely event of this being a new bug in cython that needed fixing. It
sounds like that's not the case.

> But I suggest to reupload something, so that they can finally approve
> it. If it's trivial to fix it.

Well, one thing that I can do is temporarily switch Sage to not use the
system cython, and instead build the cython shipped with Sage and use
that. This would be ugly, but would allow a quick re-upload.

-Tim Abbott

[mabshoff]

FYI:

On Dec 14, 9:35 am, Tim Abbott <tabb...@MIT.EDU> wrote:

> On Sat, 13 Dec 2008, mabshoff wrote:

<SNIP>

> Building sage/libs/linbox/linbox.cpp because it depends on
> sage/libs/linbox/linbox.pyx.
> python2.5 `which cython` --embed-positions --incref-local-binop
> -I/home/tabbott/sagemath-3.0.5dfsg/deve
> Error converting Pyrex file to C:
> ------------------------------------------------------------
> ...
>                                  mod_int **ans,
>                                  mod_int **B,
>                                  size_t B_nr, size_t B_nc)
>      cdef unsigned long rank(self) except -1
>
>      cdef mod_int det(self) except -1
>                     ^
> ------------------------------------------------------------
>

> /home/tabbott/sagemath-3.0.5dfsg/devel/sage-main/sage/libs/linbox/linbox.px d:20:20:

> C method 'det' is declared but not defined
> l/sage-main -o sage/libs/linbox/linbox.cpp sage/libs/linbox/linbox.pyx
> sage: Error running cython.
> sage: There was an error installing modified sage library code.

Robert Bradshaw wrote:
> On Dec 14, 2008, at 10:08 AM, mabshoff wrote:
>

<SNIP>

> This is function definition cruft and needs to be simply deleted. See http://trac.cython.org/cython_trac/ticket/80
>
> - Robert
>

Thanks, I will forward this to the list.

Cheers,

Michael

[Ondrej Certik]

I would suggest to use the shipped one cython, so that you can upload.
Once it gets to Debian, it can be sorted out/improved.

Ondrej

[Tim Abbott]

On Mon, 15 Dec 2008, Ondrej Certik wrote:

> I would suggest to use the shipped one cython, so that you can upload.
> Once it gets to Debian, it can be sorted out/improved.

Since I kept running into problems trying to backport patches to support
cython 0.10.2, I'm executing this plan.

-Tim Abbott

[Tim Abbott]

On Sun, 14 Dec 2008, mabshoff wrote:

> 3.2.2 should be out by Tuesday, i.e. in about 3 days and 3.3 ought to
> be rather short, i.e. 4 to 8 days depending on how many problems pop
> up.

OK. I'd appreciate it if some of the various minor cleanup tickets that I
opened when getting Sage 3.0.5 release ready were fixed sometime during
that interval, in particular:

3686
3687
3688
3689
3690
3730
3766
3769
3770

I think all of these are really easy, and it'd be wonderful to not have to
repeat the hacks I did to work around these in 3.0.5 again. At the very
least, fixing them will close 1% of the open tickets in the bug tracker.

-Tim Abbott

[mabshoff]

On Dec 18, 9:03 pm, Tim Abbott <tabb...@MIT.EDU> wrote:
> On Sun, 14 Dec 2008, mabshoff wrote:
> > 3.2.2 should be out by Tuesday, i.e. in about 3 days and 3.3 ought to
> > be rather short, i.e. 4 to 8 days depending on how many problems pop
> > up.

Hi Tim,

> OK.  I'd appreciate it if some of the various minor cleanup tickets that I
> opened when getting Sage 3.0.5 release ready were fixed sometime during
> that interval, in particular:
>
> 3686
> 3687
> 3688
> 3689
> 3690
> 3730
> 3766
> 3769
> 3770
>
> I think all of these are really easy, and it'd be wonderful to not have to
> repeat the hacks I did to work around these in 3.0.5 again.  At the very
> least, fixing them will close 1% of the open tickets in the bug tracker.

I see what I will be able to do.

>         -Tim Abbott

Cheers,

Michael