Re: Fixing expected regression-tester complaints for "new" fortrancross failures

[Geoff Keating <geoffk at geoffk dot org>]

> Date: Mon, 8 Apr 2002 21:15:05 -0400 (EDT)
> From: Hans-Peter Nilsson <hp@bitrange.com>
> cc: <gcc-patches@gcc.gnu.org>, <gcc-regression@gcc.gnu.org>
> 
> On Mon, 8 Apr 2002, Geoff Keating wrote:
> > > Date: Mon, 8 Apr 2002 19:50:54 -0400 (EDT)
> > > From: Hans-Peter Nilsson <hp@bitrange.com>
> > > cc: <gcc-patches@gcc.gnu.org>, <gcc-regression@gcc.gnu.org>
> > >
> > > Refer to
> > > <URL:http://gcc.gnu.org/ml/gcc-patches/2002-04/msg00381.html>.
> > > Should I add powerpc-*-* to the xfail target list in
> > > <URL:http://gcc.gnu.org/ml/gcc-patches/2002-04/msg00425.html>?
> > > Or do those tests need to pass to shut it up?
> >
> > So long as the tests you're adding have new names, it's unnecessary to
> > do anything special.  The tester will never have seen the tests pass
> > and so won't consider their failures to be regressions.
> 
> Misunderstanding: the failures are new, the tests are old
> (evident at the URLs).  For most of the tests, it's now a
> compile-time failure where it was a run-time failure before, but
> there are some tests that now only fail at compile-time.  Read
> the URLs for more.

In that case, probably XFAIL is the best choice.  The treats XFAIL the
same way as if the test wasn't there---it is neither a fail nor a pass.

-- 
- Geoffrey Keating <geoffk@geoffk.org> <geoffk@redhat.com>