This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Optimize stdarg functions with void * or char * va_list

On Fri, 1 Oct 2004 11:11:24 -0400, Daniel Berlin <> wrote:
> On Oct 1, 2004, at 11:01 AM, Andi Kleen wrote:
> > Alan Modra <> writes:
> >
> >> On Fri, Oct 01, 2004 at 03:47:24AM -0400, Jakub Jelinek wrote:
> >>> On Richard's request I have reverted this optimization (although I'm
> >>> seeing
> >>
> >> What for?!  I really like this patch for powerpc.
> >
> > FWIW i also liked it really on x86-64.
> I don't believe this patch should be reverted.
> I believe (and maybe this isn't a popular opinion, who knows) that our
> maintainers and our release manager are responsible for enforcing the
> stage rules, and trust that they will do so as they believe is
> appropriate for gcc.
> So that if they believe something is appropriate for stage3, they will
> okay it (which RTH did).  If they don't, they will say so.
> If maintainers disagree, there is a discussion about it.
> Maybe i'm just being naive, though, as this is how i thought things
> worked.

I do trust maintainers to be able to decide if a patch is safe or not,
but adding a new optimization pass doesn't look like stage 3 material.
 Not that I wouldn't like to have f.i. structure aliasing in gcc 4.0,
but it didn't made it in time.  If we want to get more features in for
4.0 we should go back to stage 2 (or 1) without releasing, not simply
"bypassing" stage 3 by maintainers approving non-stage-3 material.

Just my 2â-Cents,

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]