[Python-3000-checkins] r59042 - python/branches/py3k/Modules/xxmodule.c

[Brett Cannon]

On Nov 18, 2007 3:16 PM, Christian Heimes <lists at cheimes.de> wrote:
> Brett Cannon wrote:
> > Can't you do this testing in your own checkout without committing the
> > change until you have talked to python-dev about the idea of changing
> > how all types are initialized?
>
> CC to Python Dev
> The discussion is about http://svn.python.org/view?rev=59042&view=rev
>
> I've carefully studied the docs before I've committed the change. The
> problem is a well documented compiler issue on Windows. The tp_base slot
> can't be filled with a type on Windows when the module is compiled with
> distutils.
>
> On Unix the gcc can handle tp_base fine. Since most Python developers
> are using Linux or Mac OS X and the file isn't covered by a unit test
> nobody has noticed the problem
>
> http://docs.python.org/dev/3.0/extending/newtypes.html#subclassing-other-types
>
> "When filling out the PyTypeObject() for the Shoddy type, you see a slot
> for tp_base(). Due to cross platform compiler issues, you can't fill
> that field directly with the PyList_Type(); it can be done later in the
> module's init() function."
>

If the docs point that fact out then it is fine.  I didn't even know about it!

> I've changed it for two reasons:
>
>  * The xxmodule.c file is an example and template for developers. It's
> giving a bad example and it leads to code that doesn't compile on Windows.
>
>  * I'm working on a very basic unit test to test
> distutils.core.Extension. I've chosen xxmodule and xxsubtype for the tests.
>
> I hope I didn't step over a line with the change. I didn't change the
> files w/o consulting the documentation first.

Nope.  With the docs specifying that then it makes sense to change the
example code.

-Brett