stable GCC 3.4 configurations

[walt]

On Mon, 21 Mar 2005, Andreas Bartelt wrote:

> Hi,
>
> Andreas Hauser wrote:
> ...
> >
> > Should be stable without CPUTYPE.

> hmmm - perhaps my problem isn't related to the P4 optimizations at all. I have
> the same problems, when I don't specify CPUTYPE, but only define CCVER= gcc34
> in /etc/make.conf. More curiously, gcc -v still gives
> "gcc version 2.95.4 20020320 [DragonFly]" after make buildworld && make
> buildkernel (and installing them via make installworld...). So, I've probably
> done something wrong.

Not necessarily -- I see the same on my gcc34 machine.  /usr/bin/gcc -v will
always give the 2.95 message.  Try /usr/libexec/gcc34/gcc -v and see what you
get.

The magic happens in the /usr/share/mk directory.  If you grep through the
files in that directory for CCVER you can follow the logic.  The 'make'
utility always reads those files before doing any real work.

> btw, without CCVER= gcc34 defined in make.conf, the system builds, installs
> and works normally with the same source tree.

I have no trouble compiling/installing/running DFly with either compiler, but
I have only Athlon machines.  To me this implies that there may be some kind
of problem with the Intel optimizations in gcc34 -- but I'm only guessing.
I have no way to test on my Athlon machines.