[Stackless] 'Normal' sys.path being used instead of Stackless one

Ben Sizer kylotan at gmail.com
Wed Jan 21 02:06:31 CET 2009

2009/1/21 Ben Sizer <kylotan at gmail.com>:
> 2009/1/19 Richard Tew <richard.m.tew at gmail.com>:
>> Sure.  But if socket isn't compiling, I personally consider it worth
>> getting it to compile and working out why it didn't as I consider its
>> failure to compile unexpected and unacceptable.
> It looks like Python expects the Windows build environment to not have
> inet_pton or inet_ntop. It does, at least on Visual C++ 2008 Express
> Edition, but they're not in the WS2_32.DLL. An examination of the
> constants in the relevant header makes it look like Visual C++ is
> building with Vista as the minimum target, as presumably the Vista
> version of the DLL has those functions. Sadly I can't find out why it
> is building for Vista (I'm on XP SP2) but will look into it further.

Sorry to reply to myself, but I figure it might be of use to anybody
else who encounters this - setting _WIN32_WINNT=_WIN32_WINNT_WINXP (or
some other pre-Vista constant in the same header file) as a
preprocessor definition will fix the build and allow for a successful
build of the _socket.pyd file.

Strangely this doesn't appear to have come up before on
comp.lang.python or here on stackless.com, even though I've just used
the default installation of the most common compiler most Windows
users would be using for this.

Ben Sizer

More information about the Stackless mailing list