This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: C constant expressions proposals
Mark Mitchell wrote:
> > Heh, this is that discussion we've already had before.
> >
> > I'm not going to fan the flames here.
On Wed, Aug 18, 2004 at 04:00:32PM -0400, Robert Dewar wrote:
> Can you point to the thread? I can't imagine any valid
> arguments on the other (your) side.
Something has to give, because users have desires that are mutually
inconsistent:
1) get the same warnings regardless of optimization level (even -O0)
2) get high-quality warnings
3) -O0 should run fast (or at least not become still slower)
We can't have all three: to have 1) and 2), then dataflow analysis,
plus other simplifications, will need to run all the time, so -O0 would
have to run passes that aren't run now. Without dataflow analysis,
the uninitialized-variable analysis can't be as good even as it is now.