From stefan_ml at behnel.de  Sun Dec  4 03:34:50 2016
From: stefan_ml at behnel.de (Stefan Behnel)
Date: Sun, 4 Dec 2016 09:34:50 +0100
Subject: [Cython] Cython and PyPy
In-Reply-To: <2fbe319e-7c6b-1ff8-c502-9ed03529047b@gmail.com>
References: <2fbe319e-7c6b-1ff8-c502-9ed03529047b@gmail.com>
Message-ID: <2792bfb9-a1d5-7902-8fe2-b49060992d10@behnel.de>

Hi Matti!

Matti Picus schrieb am 24.11.2016 um 06:48:
> I am trying to get cython to work better with PyPy. I am sort of documenting my 
> progress and failures at 
> https://bitbucket.org/pypy/pypy/wiki/edit/cpyext_2_-_cython

Thanks for working on that.


> So far I can run the test suite on a nightly PyPy2 
> http://buildbot.pypy.org/nightly/trunk, but only when using --no-refnanny. On 
> the missing-tp_new branch of PyPy (trying to fix the datetime problems), running 
> only the c backend I get something like
> 
> Ran 4632 tests in 931.486s FAILED (failures=80, errors=18, skipped=1)
> 
> This mail is a heads-up, I will hopefully issue some pull requests soon. Also, I 
> have some questions, mainly about the test runner:
> 
> - Shouldn't the "skipped" field include the number of tests in pypy_bugs.txt?

That would be nice, yes, but it still wouldn't count entirely right,
because it could only count the test files, not the number of tests inside
of such a file, which it can't know without compiling the file first.


> - How can I get pdb to work during a single test run to try to work out the 
> internals of cython? In nose or pytest I can add the -s option, I could not find 
> an equivalent.

Difficult. The Python debugger doesn't play very nicely with Cython code,
because it cannot step through compiled code, look at C variables, etc.

If you only care about Cython itself, I'd use the normal "pdb.set_trace()"
pattern. You wouldn't want to step through all of the compiler pipeline
anyway, as it involves a lot of deep, generic syntax tree traversal code.

To figure out the internals, you might find some hints here:

https://github.com/cython/cython/wiki/HackerGuide

It's a bit oldish, but not really outdated.


> - Is there a marker for test start/test stop in the test report? I would like to 
> use awk or grep to try to analyse the multiple failures into groups

Tests start with "runTest [something]", but there's not end marker.


> - The XML backend seems to miss some of the stdout/stderr messages. Is there 
> more documentation of test running options somewhere?

I'm sure you've found "--help", but I don't think there's more than that. I
agree that the XML test output is suboptimal, but haven't looked into it
any recently. It's certainly fixable.

Stefan


From robertwb at gmail.com  Fri Dec  9 01:57:46 2016
From: robertwb at gmail.com (Robert Bradshaw)
Date: Thu, 8 Dec 2016 22:57:46 -0800
Subject: [Cython] Cython 0.25.2 released
In-Reply-To: <CADiQ+QBYBMoShge_B1ABtqt_GFzZkOi6tFv38i1wCMXLVBL9tg@mail.gmail.com>
References: <CADiQ+QBYBMoShge_B1ABtqt_GFzZkOi6tFv38i1wCMXLVBL9tg@mail.gmail.com>
Message-ID: <CADiQ+QDCgVxp5aB6vu5_OAWiyqAFFwecgbeoOgThgqNgBpS4dQ@mail.gmail.com>

On Dec 8, 2016 10:29 PM, "Robert Bradshaw" <robertwb at gmail.com> wrote:

I pushed a bugfix release to PyPi this afternoon. This is a minor
release with the following fixes:

- Fixes several issues with C++ template deduction.
- Fixes a issue with bound method type inference (Github issue #551).
- Fixes a bug with cascaded tuple assignment (Github issue #1523).
- Fixes bug with powers of pure real complex numbers (Github issue #1538).
- Fixes bug with namespace resolution of enum values (Github issue #1533).
- Fixes bug with references and exception handling (Github issue #1519).
- Fixed or silenced many Clang warnings.

It also silences the build deprecation warnings on Python 3. It should
be fully backwards compatible with 0.25.1.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cython-devel/attachments/20161208/493ff1f5/attachment.html>