Upgrade Cython to 0.14.1

[jason]

Version 0.14.1 was recently released, and has some bugfixes and nice features.

Use the following spkg and patch:

1. ​http://boxen.math.washington.edu/home/jdemeyer/spkg/cython-0.14.1.p3.spkg
2. 10493-cython-0.14.1.patch​

[robertwb]

We're in pretty good shape. There's a couple of minor patches to the library that need to go in, but other than that upgrading to 0.14 should be pretty straightforward. See, e.g. ​https://sage.math.washington.edu:8091/hudson/view/ExtLibs/job/sage-tests/

[jdemeyer]

Is the __getattr__ patch (in the 0.13 Cython spkg) still needed?

[robertwb]

No, that is now handled via a flag. There are several bugfixes in 0.14.1 that I'd like to get in, which should be out shortly (before 4.6.2 is at release candidate for sure, and I knew we missed the window for 4.6.1) which is why I haven't done anything on this ticket yet.

[vbraun]

Updated package with just the __getattr__ patches removed as they are now in the upstream source:

​http://www.stp.dias.ie/~vbraun/Sage/spkg/cython-0.14.spkg

[robertwb]

Thanks, but that is not sufficient, as we need to pass the appropriate flags.

The most recent spkg can be found at ​https://sage.math.washington.edu:8091/hudson/job/sage-build/lastSuccessfulBuild/artifact/cython-devel.spkg (though the Sage library requires a couple of patches as well).

[jason]

Updating the description to upgrade to 0.14.1, which is now the most recent release. Has anyone worked on an 0.14.1 spkg?

[robertwb]

Spkg at ​http://sage.math.washington.edu/home/robertwb/cython/cython-0.14.1.spkg

[jason]

Replying to robertwb:

There's a couple of minor patches to the library that need to go in

Is this still the case with the 0.14.1 spkg?

[fbissey]

I think that's what the patch is all about. I applied it to 4.6.2.alpha4 and it happily built and sage -testall did not show any problems for me. But may be the following warnings should be addressed

warning: sage/combinat/combinat_cython.pyx:14:0: 'stdlib' is deprecated, use 'libc.stdlib'
warning: sage/ext/interpreters/wrapper_rdf.pxd:3:0: 'python_object' is deprecated, use 'cpython'
warning: sage/rings/polynomial/polynomial_real_mpfr_dense.pyx:7:0: 'python_int' is deprecated, use 'cpython'
warning: sage/rings/polynomial/polynomial_real_mpfr_dense.pyx:8:0: 'python_float' is deprecated, use 'cpython'
warning: sage/ext/interpreters/wrapper_cdf.pxd:3:0: 'python_object' is deprecated, use 'cpython'
warning: sage/ext/interpreters/wrapper_rdf.pxd:3:0: 'python_object' is deprecated, use 'cpython'
warning: sage/ext/interpreters/wrapper_py.pxd:3:0: 'python_object' is deprecated, use 'cpython'
warning: sage/ext/interpreters/wrapper_rr.pxd:3:0: 'python_object' is deprecated, use 'cpython'
warning: sage/ext/interpreters/wrapper_el.pxd:3:0: 'python_object' is deprecated, use 'cpython'

And possibly

warning: sage/modular/modsym/p1list.pyx:703:17: cdef variable 'i' declared after it is used

Finally I am wondering about these

warning: sage/matrix/matrix0.pyx:53:22: Function signature does not match previous declaration
warning: sage/matrix/matrix_window.pyx:9:17: Function signature does not match previous declaration
warning: sage/matrix/matrix_sparse.pyx:18:33: Function signature does not match previous declaration

But all these were present before upgrading to cython-0.14.1. We could use the opportunity to fix some of them before they bite us later, especially the deprecated ones.

[fbissey]

Forgot these two:

warning: sage/structure/parent.pyx:124:4: dict already a builtin Cython type
warning: sage/symbolic/pynac.pyx:589:5: Function 'py_get_parent_char' previously declared as 'extern'

[robertwb]

Yes, the (attached) patch is still needed. There's some significant cleanup that I want to do, including fixing these warnings, but I think that should be a separate ticket.

[fbissey]

Well if you open a separate ticket for these warnings I am quite happy to give this a positive review. I have built it on linux-86, linux-amd64 and OS X 10.5. It all looks good to me.

[robertwb]

Ticket for warnings at #10764.

[jdemeyer]

After building sage-4.6.2.rc0 and applying the spkg and patch from this ticket:

$ ./sage -ba-force
[...]
gcc -fno-strict-aliasing -g -O2 -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -fPIC -I/usr/local/src/sage-4.6.2.rc0/local//include -I/usr/local/src/sage-4.6.2.rc0/local//include/csage -I/usr/local/src/sage-4.6.2.rc0/devel//sage/sage/ext -I/usr/local/src/sage-4.6.2.rc0/local/include/python2.6 -c sage/combinat/combinat_cython.c -o build/temp.linux-x86_64-2.6/sage/combinat/combinat_cython.o -w
sage/combinat/combinat_cython.c: In function '__pyx_pf_4sage_8combinat_15combinat_cython__stirling_number2':
sage/combinat/combinat_cython.c:2431:13: error: incompatible types when assigning to type 'mpz_t' from type 'struct __mpz_struct *'
error: command 'gcc' failed with exit status 1
sage: There was an error installing modified sage library code.

This is using

$ gcc --version
gcc-4.4.3 (Gentoo 4.4.3-r2 p1.2) 4.4.3
Copyright (C) 2010 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

[jason]

Why is the deprecation warning in polynomial_compiled.pyx commented out in this patch? If it is not being used, shouldn't it just be removed?

[robertwb]

I'm not sure why the gcc error--I'll take a look at that. (I'm guessing that's some new code?)

As for commenting out the deprecation warning, the issue is that decorators aren't implemented for Cython special functions. They used to be silently ignored, but now an error is raised. When they're supported, we should put this back in.

[robertwb]

Somehow one of the fixes for ​http://trac.cython.org/cython_trac/ticket/654 was merged into the release at the last minute, which is incompatible with how we declare many of the *_t types. I'll post a new spkg disabling this "fix."

[robertwb]

I've created a new spkg. All tests pass.

[fbissey]

Replying to robertwb:

Somehow one of the fixes for ​http://trac.cython.org/cython_trac/ticket/654 was merged into the release at the last minute, which is incompatible with how we declare many of the *_t types. I'll post a new spkg disabling this "fix."

How come you and I never got affected by it?

[robertwb]

I'm not sure, but maybe the last spkg was the rc, and the fix for this issue was not supposed to go in this release, but we messed up and it did.

[jdemeyer]

It works for me now, so I'm putting back the positive review.

[robertwb]

Thanks.

[jdemeyer]

Please fix: the SPKG.txt must be updated, there is no mention of this ticket.

[robertwb]

The SPKG.txt has been updated, see ​http://sage.math.washington.edu/home/robertwb/cython/cython-0.14.1.p2.spkg

[jdemeyer]

I made some further changes to SPKG.txt: ​http://boxen.math.washington.edu/home/jdemeyer/spkg/cython-0.14.1.p3.spkg

[robertwb]

I'm happy with your spkg.txt.