[PATCH] Re: 3 new fails with builtin patch

[Mark Mitchell]

> Incidently, Mark, I think we all agreed that new testcases should go
> under g++.dg, not g++.old-deja.  g++.dg don't unilateraly set any
> optimization level, I think that is a good path (we would have caught
> this regression).

I agree.  I was just applying Roger's patch.  I could (should?) have
asked Roger to rework his tests for the g++.dg testsuite, but I wanted
to close the PR quickly so we could go on to other things.  In hindsight,
that was a mistake.

Thanks,

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com