Upgrade Cython to 0.15.1

[robertwb]

Lots of improvements and bug fixes, including generators.

See ​http://wiki.cython.org/ReleaseNotes-0.15.1 for details.

---

New spkg: ​http://sage.math.washington.edu/home/jdemeyer/spkg/cython-0.15.1.spkg

Apply 11761-cython-0.15-rebased.patch​ and 11761_review.patch​ to the Sage library.

[leif]

Replying to robertwb:

... including generators.

Yeah!

[robertwb]

Spkg up at ​http://sage.math.washington.edu/home/robertwb/cython/cython-0.15.spkg

[robertwb]

Note that 0.15.1 is around the corner, but it's not as big of a jump.

[leif]

Would 0.15.1 require additional changes to Sage?

[jason]

Just FYI, 0.15.1 is now released.

[leif]

Replying to jason:

Just FYI, 0.15.1 is now released.

Oh, did you tell Robert?

(Although he said "it's not as big of a jump".)

[jason]

Replying to leif:

Replying to jason:

Just FYI, 0.15.1 is now released.

Oh, did you tell Robert?

Oh, I'm sure he knows, since he's the BFDL for Cython, if there is such a position ;).

(Although he said "it's not as big of a jump".)

[leif]

Replying to jason:

Replying to leif:

Replying to jason:

Just FYI, 0.15.1 is now released.

Oh, did you tell Robert?

Oh, I'm sure he knows, since he's the BFDL for Cython, if there is such a position ;).

I was just kidding.

(Wonder whether he updates the spkg... :P)

[leif]

:)

[robertwb]

No further patches needed for 0.15.1.

[jdemeyer]

Diff for the new Cython spkg, for review only

[jdemeyer]

Minor comment: spkg-install is not executable (I know it is made executable upon building, but better fix this in the spkg anyway).

[jdemeyer]

Why is the directory tests missing from the Cython source? This should either be fixed, or be documented.

[jdemeyer]

What should be done with:

cdef mpz_t denom
mpz_init_set_si(denom, 1)

This code is correct but gives a local variable 'denom' referenced before assignment warning. (there are many of these in sage/matrix/misc.pyx for example)

[jdemeyer]

This new Cython really gives a lot more warnings than the old one, in many cases for code which is perfectly fine. For me, this isn't really a problem, but it is certainly annoying.

[robertwb]

I've fixed the executable bit on spkg-install and added the tests to the spkg. (I remember someone complaining way back that they were too big, but I don't think it's an issue and we've regularly shipped them before, and it's easier to just have them there.) The spkg at the above link has been refreshed.

The warnings are primarily due to the fact that mpz_t et al are declared as void* rather than arrays of size 1. ​http://hg.sagemath.org/sage-main/file/2a2abbcad325/sage/libs/gmp/types.pxd#l1 Once this is fixed the warnings should go away (and they are correct with respect to the current declarations). All the other warnings I've noticed are correct too, I simply haven't had the time to fix/update Sage's code.

Your review patch looks fine. You obviously looked at this, is the window closed for 4.7.2?

[jdemeyer]

Replying to robertwb:

Your review patch looks fine.

In fact, it doesn't. I removed one line too much, fixed now.

You obviously looked at this, is the window closed for 4.7.2?

This will certainly not be merged in sage-4.7.2.

[jdemeyer]

Replying to robertwb:

The warnings are primarily due to the fact that mpz_t et al are declared as void* rather than arrays of size 1. ​http://hg.sagemath.org/sage-main/file/2a2abbcad325/sage/libs/gmp/types.pxd#l1 Once this is fixed the warnings should go away (and they are correct with respect to the current declarations).

If there is an easy way to fix this, then we should do this. I'm saying we should fix every single warning but if it is possible to declare mpz_t such that there are no warnings, we should certainly do this.

[jdemeyer]

Of course, that should have been "I'm not saying we should fix every single warning"

[jdemeyer]

Replying to robertwb:

I've fixed the executable bit on spkg-install and added the tests to the spkg.

One more thing: changing permissions is a change which should be committed. So could you please do a hg commit and re-upload the spkg? Otherwise, the package and patch looks good.

[jdemeyer]

Milestone sage-4.7.3 deleted

[robertwb]

Spkg updated with executable bit set commit.

[robertwb]

Yes, I agree these warnings should be fixed. #11992 .

[jdemeyer]

Replying to robertwb:

Spkg updated with executable bit set commit.

Looks you forgot to update the spkg, it is exactly the same as before (last modified on 1 november):

$ ls -l /home/robertwb/cython/spkg/cython-0.15.1.spkg
-rw-r--r-- 1 robertwb robertwb 1420704 2011-11-01 06:57 /home/robertwb/cython/spkg/cython-0.15.1.spkg

Never mind, I will make the change.

[robertwb]

Oops, I put it at

robertwb@sage:~/cython$ ls -l /home/robertwb/cython/cython-0.15.1.spkg 
-rw-r--r-- 1 robertwb robertwb 1412114 2011-11-04 22:51 /home/robertwb/cython/cython-0.15.1.spkg

That's what comes of getting too little sleep. Go ahead and use that one (or if you've already done the trivial change, even better.)

[fbissey]

Sorry to be late to the party. I get an error building the optional mpc extension in 4.8.alpha2 and I suspect cython-0.15.1 is the problem

sage/rings/complex_mpc.c: In function '__pyx_pf_4sage_5rings_11complex_mpc_15MPComplexNumber_16__float__':
sage/rings/complex_mpc.c:9720:48: error: cast specifies array type

I am guessing the complex_mpc.pyx code needs to be updated but I don't really know how.