On Wed, 2 Nov 2005 08:17:39 +0100
Martin Husemann <martin@duskware.de> wrote:

> On Wed, Nov 02, 2005 at 04:53:22PM +1000, George Michaelson wrote:
> > and (b)
> > we wind up being stuck on old compilers which has *HUGE* downside
> > costs when packages people use in real-world deployments start
> > depending on the newer compiler features (and that has happened in
> > the past to pkgsrc)
> 
> This coin has a third side: last time we adopted gcc 3.3 early, and
> had *lots* of trouble to fix all the pkgs that had strict-alias
> violations.

Do you expect that to be a repeated experience, or is it one-time? It
sounds to me like a one-time compliance issue.

> 
> Martin
> P.S.: tech-kern is not apropriate for this discussion, IMHO
> 

When I checked archives, it was on tech-kern, because of problems with
compiling kernels with gcc4 at that time. That was the relevancy. Its
probably moot right now, I agree with a sentiment that compilers maybe
should be in pkgsrc a while before they are tested for system
dependencies. 

 -G