Re: NPTL -- not happening?
On Tue, 2005-01-18 at 19:45, Benjamin Herrenschmidt wrote:
> On Tue, 2005-01-18 at 18:09 +0100, Sven Luther wrote:
> > Don't be stupid, only a few people really need NPTL stuff, and you can always
> > follow testing/unstable once sarge is released.
> This is a bogus argument actually :) It translates basically that "only
> a few people need more performant, more conformant and less buggy code"
Alternately, it translates as "nobody uses release anyway".
> > > BTW, gcc-3.4 is working quite well for me. Not that
> > > it should be needed though; I'd have expected glibc
> > > to be getting at thread-local data via functions.
> > Probably. There is gcc-4.0 also which i don't know the release schedule about.
> 3.4 branch is very stable and well behaved on ppc, it's definitely what
> I would recommend for now, _even_ for sarge in fact.
> 4.0 is still not ready to be used by anything but gcc hackers imho :)
Not even for potential bug reporters? Will a kernel boot
if compiled with gcc 4?
At times, I write some pretty wild code using gcc extensions.
I'd be happy to run my code through the latest gcc if it were
available through the regular Debian package mechanism.
I use computed goto, the C99 restrict keyword, a variety of
function attributes, __constructor__, and more. I even have
a test suite I can run to make sure things work OK.