Interesting.  I dug into the pyport.h file and found that it has a single include - pyconfig.h

This .h file is constructed by ./configure and supposed copied into the python Include/ directory.  In my case, it was not copied into the correct location, and pyport.h was failing to have the properly architecture configured version of variable definitions.  Specifically, LONG_BIT was defined for 32-bit by default and not 64-bit, as is my OS.  Placing the file in the right location seemed to allow boost to build without any apparent errors.

Thank you for your assistance.

-Kirk







On Thu, Nov 19, 2009 at 11:35 PM, Vladimir Prus <ghost@cs.msu.su> wrote:
On Friday 20 November 2009 02:30:50 rkdelisle@gmail.com wrote:

> I'm trying to build boost on a 64-bit CentOS 5.4 install. I have Python
> 2.6.4 built and installed at /opt/Python_2.6.4/, and I've appended the
> user-config.jam file with:
>
> using python : 2.6 : /opt/Python_2.6.4/python : /opt/Python_2.6.4/Include :
> /opt/Python_2.6.4/Lib ;
>
> The standard system Python is 2.4.1 but the tools I'm using require 2.6, so
> I've built this version and installed it independently of the system
> version 2.4.1 to avoid any conflicts.
>
> As I'm sure you've already imagined, I get the error:
>
> LONG_BIT definition appears wrong for platform (bad gcc/glibc config?)
>
> I see that this is a long standing bug, but I have yet to find a fix. I've
> tried various CCFLAGS, CXXFLAGS, etc. to push for a 64-bit compile or a
> 32-bit compile (-m64 or -m32).
>
> The offending file is pyport.h - is there a 64-bit friendly version that I
> don't know about?

Does this error happen if you write a test program that includes just that file,
and try to compile it by hand? If so, I suggest you raise the bug with CentOS folks.

Did you try to look at the code that emits the error? What is it checking for?


- Volodya
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build