Re: GCC 3.0 Release Criteria

[Joe Buck <jbuck at possibly dot synopsys dot com>]

> I cannot compile pooma applications with g++ 2.95.2 on irix without
> -fsquangle.  If use of -fsquangle is unacceptable for regression testing,
> then application testing c++ on irix is trivial - nothing builds with g++
> 2.95.2 so there can be no regressions.  I was hoping to do better than this,
> and am seeking guidance.

I don't know whether gas works on Irix or not.  If it does, that's one
solution.  If not, then it seems that you could compare -fsquangle/2.95.2
to -fnew-abi/3.0.  While this is a bit ugly, if -fsquangle produces a
wrong result with 2.95.2 you'll see no regression, if it produces a
correct result we are certainly interested if 3.0 does not.

I'm pretty sure that in the msg below Martin was talking about testing
-fsquangle for the new code.

> > -----Original Message-----
> > From:	Martin v. Loewis [SMTP:martin@loewis.home.cs.tu-berlin.de]
> > Sent:	Wednesday, 10 May 2000 17:06
> > To:	David.Billinghurst@riotinto.com.au
> > Cc:	gcc@gcc.gnu.org
> > Subject:	Re: GCC 3.0 Release Criteria
> > 
> > > AFAICT, I can only compile pooma on irix if I rebuild c++ and libstdc++
> > with
> > > -fsquangle. Is this then an appropriate application for regression
> > testing?
> > 
> > The new C++ ABI defines a new mangling scheme, which is even more
> > compact than -fsquangling, according to initial tests. So no,
> > -fsquangling won't be used for regression testing - it appears that
> > this option will never be used by default in a released compiler.
> > 
> > Regards,
> > Martin
>