Numerous Libjava Test Failures for gcc 3.1, not for gcc 3.0

[Jeffrey Oldham]

With help very quickly provided by Jeff Sturm, Tom Tromey, and
Alexandre Petit-Bianco, I was able to eliminate the numerous gcc 3.1
libjava regression test failures that I have recently seen.

The problem was that the tests were using an old libgcc_s.so.0 in the
installation directory, not the libgcc_s.so.0 built during
bootstrapping.  Removing everything from the installation directory
permitted the executable to run correctly.

Is there a reason that using a library not in the build directory
would be preferred over one that is in the build directory?

Thanks again for helping me through the difficulties.

Thanks,
Jeffrey D. Oldham
oldham@codesourcery.com