Re: -O vs -O2, revisited

Adrian Chadd wrote:

> > Hmm.. so you are saying that the same bug does not exists if the exact
> > same GCC version is running on Linux? Odd.. I imagined the code
> > generation was mostly the same across the OS:es running on the same
> > CPU..
>
> You'd think .. :-)

And I still do. It is more of a matter that none of the distributions
uses the same GCC.. (all GCC versions in Linux distributions are rather
heavily patched, and more or less all starting from different GCC
versions)
 
> > Have someone contacted the FreeBSD/OpenBSD maintainers to get the
> > compiler fixed?
>
> Well, FreeBSD doesn't build any part of the system with -O2 or
> above.

So FreeBSD has completely lost the trust in GCC optimization then? Sad
as it does make a difference..
 
> The trouble is that even if someone wanted to chase down the bug(s)
> and provide patches, the files would come off the vendor branch and
> it would become a pain to import new versions.

Is FreeBSD is using it's CVS tree for GCC? Not a tarball+patches like
everyone else? Sounds like a bad move to put such a massive piece of
foreign code in a CVS tree..

--
Henrik