several messages (regarding gcc)

[Sean McNeil]

On Sat, 2004-07-03 at 13:41, Gerald Pfeifer wrote:
> On Mon, 28 Jun 2004, Gary Jennejohn wrote:
> > On my amd64 box running a freshly cvsup'ed and installed system (world
> > and kernel in 32 and 64 bit mode) I see:
> >
> > 32bit mode - ``make all'' succeeds
> > 64bit mode - gentype fails with signal 10, just like you've observed
> On Mon, 28 Jun 2004, Sean McNeil wrote:
> > This is sort of expected (at least by me).  Support for amd64 is not
> > very mature in 3.3 and I've run into several issues with it.  3.4 should
> > work a lot better.  Hopefully, 3.4 will be merged into FreeBSD soon.
> 
> Thanks for your feedback.  I agree, and thus marked the gcc33 port BROKEN
> on amd64:
> 
>   Mark BROKEN on amd64.  It looks like the system compiler miscompiles
>   this, and most people on this platform will use either the system
>   compiler or gcc34 anyways.
> 
> There is a very stable variant of GCC 3.3 available on the hammer-branch
> in GCC CVS, which got a lot of feeding and caring from David O'Brien and
> colleagues plus colleagues of mine (in our respective day jobs), but when
> it comes to official GCC releases, it's really 3.4 and later that rock on
> amd64.

Yes, I'm sorry for not qualifying that statement: gcc 3.3 ports has poor
amd64 support.  The gcc 3.3 in core FreeBSD has been modified by the
persons mentioned so that it is very stable on amd64.

As you say, gcc 3.4 has much better amd64 support and closer ANSI
compliance.  Because of this, I have found a few packages in ports that
will not compile with the gcc 3.4 port (CC=gcc34, CXX=g++34).  These
tend to be trivial fixes.  The only concern I have is that gcc 3.4
produced a non-working openldap-sasl-server-2.1.30.  The compiler in
core -CURRENT made a functional one.

Cheers,
Sean