Subject: Re: Cross compiling under MkLinux
To: Michael R. Zucca <mrz5149@cs.rit.edu>
From: Colin Wood <ender@is.rice.edu>
List: port-mac68k
Date: 03/30/1997 17:01:14
> While setting up a cross-compiler with 2.7.2.2 I've been having trouble
> compiling the kernel. I don't know if this is a 2.7.2.2 problem but I couldn't
> get gas to assemble the output of asm2gas .s files. I'm not sure if this is
> a problem with asm2gas or what but it seems to choke on labels like:
> 
> 1:
> (I found this in m68k/fpsp/netbsd.s)
> 
> and it also seems to have a problem with the floating point notation:
> 
> #:xxxxxx
> 
> I can't understand this since we're obviously using gnu binutils and gcc
> under NetBSD but the silly thing won't go under my 2.7.2.2 compile!
> 
> Does anybody know if there have been signifigant changes to gas since the
> 2.7.2 release? (Well, I suppose binutuils and gcc use different versions but
> you get the idea)

I think that NetBSD has made rather significant changes to gcc-2.7.2, so 
it's not quite stock gcc anymore.  You might want to ask this question on 
tech-toolchain or current-users.  I think I've seen it come up lately.

> The makefile also has difficulty compiling genassym. First of all, it tries to
> link it with the cross-linker using $(CC) instead of $(HOSTED-CC). Second
> of all it tries to do as using the cross tools because even though HOST-CC
> is absolute pathed out to the local compiler the main PATH points to the cross
> tools. I had to compile the silly thing by hand. Can't we make this a shell
> script?
> 
> However, the gas problem seems to be holding me up most of all. The cross gcc
> seems to be making code OK.

Have you checked out the cross-compiler HOWTO?  It might contain some 
useful info on this (I don't quite know since I haven't looked at in a 
long while).

> Wow, what a speed improvement, though! It took my PowerCenter 132 about 1 minute
> to make depend while my IIvx took hours! I would *really* love to get this
> going so that when I get 3 free seconds to finish the intvid kernels I can
> work on code instead of waiting for recompiles!

Pretty nice.  Have you considered converting your code into an LKM?  Then 
you could just modunload an old one and modload the new one.

Later.

-- 
Colin Wood                                      ender@is.rice.edu
Consultant                                        Rice University
Information Technology Services                       Houston, TX