Versioning mess proved!!!

[Jack Howarth]

Franz,
        I have been rebuilding the RedHat SRPMS on my linuxppc machine
under glibc 2.1.94/gcc 2.95.3/Linux 2.4-test9 and am finally stablized
more or less. Everything I have seen reinforces our previous observations.
That is when you rebuild a program under glibc 2.1.94 with a mixture
of glibc 2.1.3 and glibc 2.1.94 built libs that the resulting program works
fine. However when you rebuild any of the glibc 2.1.3 built libs linked to
that program, the program will then segfault under those new libs.
It might be helpful if we can come up with a simple shell script using
readelf that can identify any 'problem' binaries on our linuxppc machines
by the use of PLT for __gmon_start.
                                                        Jack
ps I have been working under the assumption that only programs that
are linked to glibc 2.1.3 built libs can be consider broken (because of
the use of PLT in those cases) and that if I end up with a system
purely of binaries/libs built against glibc 2.1.94 that whatever fix
we adopt will be unlikely to break compatibility with those binaries.
Hopefully I haven't been wasting my time...