>I promised to tell you when my first cross-compiled kernel is accepted by the
>booter. The first one is running now as good (or as bad)

Great!

> - gcc-2.7.2 could not compile some files (e.g. kern_synch.c). Exchanging
>   2.7.2's 'cc1' with the one from 2.6.3 solved the problem.

That's odd.  I'm running 2.7.2 on puma right now and compiling kernels
happily (although the one I compiled with -O6 last night didn't get
through the ADB probe sucessfully while running).  Have you considered
reporting this behavior to gcc-bug or bug-gcc or whatever the proper
address is?

> - GNU's 'cpp' had trouble with the comments in the assembler sources. I had
>   to change all assembler comments to C comments ('|' -> '//'). Previously
>   all the objects made from a assembler source were empty.

The std. NetBSD cpp is actually a script that invokes the real cpp.  You
might try using that--it usually invokes /usr/libexec/cpp in traditional
mode, among other things (if I remember right).

> - Following does not work with GNU's cpp:
>      #define foo(x) _/**/x
>      foo (bla)
>   this results not in '_bla' as expected but in '_  bla' (Note the spaces).

It should work in -traditional mode, I think.

> - broken definition of MACHINE and MACHINS_ARCH forced a lot of undefined
>   symbols.

Ah, yes...  I can imagine.

-allen

-- 
Allen Briggs - end killing - allen.briggs@bev.net ** MacBSD == NetBSD/mac68k **
   Where does all my time go?  <a href="http://www.netbsd.org/">Guess.</a>