Build bugs in Python 2.2.1?

Mats Wichmann mats at laplaza.nospam.org
Wed Aug 21 21:17:38 CEST 2002


On Tue, 20 Aug 2002 17:54:34 -0400, "Mahrt, Dallas"
<dallasm at aiinet.com> wrote:

:One aspect of the build process that has been avoided in this discussion is
:the problems 
:that the setup.py/distutils process causes for cross-compiling. Because the
:build process 
:depends on the host and build targets to be the same, cross-compilation
:cannot happened with 
:a './configure <opts>; make; make install'.

Even in something that's not strictly a cross-compile environment,
this causes problems.  To be concrete, where this has bitten me - and
not just for Python - the LSB project (www.linuxbase.org) attempts to
carve out a core piece of Linux that can be guaranteed consistent
across all (conforming) Linux systems.  It makes an attactive target
to an application developer as they can bundle a binary that's likely
to work across multiple distros.  To conform, though, an application
must limit itself to using interfaces in the spec, and the configure
assumption "if it's found, it can be used" just plain doesn't hold
true.

Python actually does very well in this regard; there's only one issue
I have to work around today, and that's the (perhaps questionable)
assumption that if a header file <netpacket/packet.h> is found then
the netpacket stuff can be used, and I can't quite figure out what you
could test for that would be more accurate even in this case...

(After the configure, I have to patch the generated pyconfig.h to undo
the problem).


Mats Wichmann




More information about the Python-list mailing list