> I figured as much on this point, which is why I polled you.  (Given that I
> have been allocated primarily to my job for some time, you're a lot more in
> tune with the current state of codegen and ICE bugs in gcc at the moment.)

It isn't much different for me... Just keeping up with the gcc mailing 
lists is a major effort.

I would guess that we probably aren't much more than half way through the 
gcc-3 development cycle and major parts of the compiler are still 
changing: cpp is one; a new, and hopefully more stable ABI for C++ is 
another (perhaps one that won't be binary incompatible with each gcc 
release :-)  I think we should be wary of having two many unknowns in the 
equation at one time.

> I will be looking at
> making this a bit more uniform--particularly, to do proper a.out/ELF
> detection when using config.guess for a native host--and getting those
> changes back to the gcc trunk.

The thing that confused me when I tried the build was that config.guess 
reported an i386-netbsdelf platform, but configure.in still treated it as 
a netbsd-aout.

R.