Other testsuite failures

[Tom Tromey]

>>>>> "Mark" == Mark Wielaard <mark@klomp.org> writes:

Mark> FAIL: PR1343 compilation from bytecode
Mark> FAIL: PR1343 -O compilation from bytecode

I think this is the libtool bug with `$' that nobody has gotten around
to fixing.

Mark> FAIL: SyncTest execution - gij test
Mark> WARNING: program timed out.
Mark> FAIL: SyncTest execution - gij test

I've seen this too.  I haven't investigated.

Mark> FAIL: pr8823 execution - gij test
Mark> FAIL: pr8823 compilation from bytecode
Mark> FAIL: pr8823 execution - gij test
Mark> FAIL: pr8823 -O compilation from bytecode

This is the Miranda method check.
We'll have to xfail it for compilation at least, since gcj has a lot
of bugs in this area.  I tried fixing them this weekend but ran into a
morass -- the problem is more pervasive than I'd hoped.

Mark> On powerpc-unknown-linux-gnu the following additional failures occur:
Mark> FAIL: Array_3 execution - source compiled test
Mark> FAIL: Array_3 execution - bytecode->native test
Mark> FAIL: Array_3 -O execution - source compiled test
Mark> FAIL: Array_3 -O execution - bytecode->native test

This is a long-time failure.  The problem has something to do with
null pointer exceptions.  Perhaps we'll need to amend the test suite
to allow per-target xfails.

Tom