[Lsb-infrastructure] odd failure building tet on two platforms

[Wichmann, Mats D]

After recent changes, tet-harness is not autobuilding
on two platforms, s390 and ppc64; on the other five it
does build.  The failure is in the Python API, as follows:

make[1]: Entering directory
`/nethome/autobuild/tmp/rpm-build/BUILD/tet3-lite-3.7/contrib/python_api
'
/opt/lsb/bin/lsbcc -D_POSIX_SOURCE -DTET_LITE -O -fPIC
-I/usr/include/python2.4 -c pytet_wrap.c
warning: dangerous include path /usr/include/python2.4
In file included from /usr/include/python2.4/Python.h:55,
                 from pytet_wrap.c:13:
/usr/include/python2.4/pyport.h:612:2: error: #error "LONG_BIT
definition appears wrong for platform (bad gcc/glibc config?)."


pyport.h defines some things if they're not defined,
and eventually checks

#if LONG_BIT != 8 * SIZEOF_LONG

neither of these constants are defined by LSB headers,
so somehow they're being pulled from native headers
the wrong way in the autobuilders (if LONG_BIT wasn't
defined at all, pyport would have defined it as
8 * SIZEOF_LONG, and thus pass the preprocessor check).

what I don't get is why this has just broken, it's been
working for ages as far as I know...