Mistaken change in GCC (fwd)

[Zack Weinberg]

On Wed, Nov 22, 2000 at 04:55:19PM -0800, Joe Buck wrote:
> > On Wed, Nov 22, 2000 at 04:43:45PM -0700, Jeffrey A Law wrote:
> > > 
> > >   In message < 20001122151553.J820@wolery.stanford.edu >you write:
> > >   > All I am saying is that I think we could better support emacs by
> > >   > lending enough assistance to the emacs project to make it work with
> > >   > the present preprocessor.  I'm fully prepared to do this.
> > > But that doesn't help existing copies of emacs that are already floating
> > > around.  It would only help new emacs releases.
> > 
> > How is this different from any of the other ways that new gcc releases
> > break existing software?  (Strict aliasing comes to mind.)
> 
> In the case of strict aliasing, we fixed the problem by making
> -fno-strict-alias the default in response to user outcry.

In 2.95, yes.  -fstrict-aliasing is still the default in the
development tree and I haven't heard it will be off in 3.0.  None of
the proposals for helping programmers with code that breaks under
-fstrict-aliasing have been implemented, either.

zw