Major 3.1 branch Java regressions

[Alexandre Oliva]

On May 16, 2002, "David S. Miller" <davem@redhat.com> wrote:

> Java fails to be able to compile any of the testsuite on
> the 3.1 branch now

Indeed.  Not introduced by my patch, for sure.  I can't tell whether
the failure was already there when I checked it in or if it was
introduced afterwards, because all I did was to mechanically compare
pre- and post-results, and there were no differences.  I didn't save
the results, though.  Anyway, I've just verified that reverting my
patch doesn't fix the problem, though.

> The only functional changes from the 3.1 release to the current branch
> are c++ frontend changes.

Which can't possibly affect gcj's ability to find libgcj.jar, which is
the reason for the failures I've seen.

My best guess is that this may be some latent bug in the libjava
testsuite, exposed due to the change in the version string or
something alike.  However, I've just attempted to prove this by
reverting the GCC version to its state as of this Monday's snapshot,
and it didn't help either, so I'm at a loss as to what may have caused
it :-(

I've verified, however, that, while the gcj compile-and-link commands
run by `make check-target-libjava' in <target>/libjava/testsuite fail
when run in that directory, they succeed when run in
<target>/libjava.

Like you, I couldn't find anything in the diffs since the last
snapshot that helped figure out the reason for this difference :-(

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer