Re: latest merged ppc 32/64 NPTL port

> I saw tst-tls2 (I think it was) fail once with that message.  Doesn't
> it just mean that there was a race where the test finished just as the
> timeout occurred?  That test shouldn't take that long, though.

Ordinarily, that's what it means.  But it would be strange for that to
happen so consistently, hence my suspicion.  Increasing the timeout doesn't
change this behavior.  When I run tst-cond10 myself, it sits without
producing any output.  tst-cond10 --direct (which does not fork to set the
timeout alarm in the parent) works fine (producing lots of output quickly).
Something is amiss.