Arne Juul writes:

>In the past I have found that on NetBSD, it's easiest to use the
>standard gcc distribution (2.7.2.2) plus g77, and configure with some
>other prefix (/usr/local or whatever).  However, for this to work
>on current NetBSD/pmax you need to apply the changes to the gcc
>configuration files that is in the NetBSD tree (so you get
>-mabicalls default, -mhalf-pic default and so on).  It's probably not
>easy to extract those changes from the NetBSD tree, which has
>lots of diffs for changed directory structure.

Arne,

It's actually pretty easy to estract the changes if you know the
config file naming conventions.  The NetBSD-specific gcc config
replaces some config files, and adds some NetBSD-specific patches;
AFAIk none of these are required for the mips port.

What you need to do is extract the NetBSD replacement config files
from /usr/src/gnu/usr.bin/gcc/arch/mips, and install those *over* the
standard GCC config files in <FSF-gcc-src>/config/mips.

AFAIK, the relevant files are:
	NetBSD file	maps to		FSF GCC file
	netbsd.h			netbsd.h
	mips.h				mips.h
	xm-netbsd.h			xm-netbsd.h

To build a g77 for NetBSD/mips, these are the files you should update
from the stock gcc-2.7.2.2 versions to the NetBSD versions.

for the remaining files in Netbsd's gcc/arch/mips:

The NetBSD and FSF versions of xm-mips.h and mips.c are identical.

The NetBSD tm.h, tconfig.h, and config.h are only relevant to NetBSD;
the NetBSD make/include files use them where the FSF distribution uses
an config-script-generated symlink.  (I think this is true for all
ports.)

I have e-mailed updated NetBSD config files to the FSF bug address at
least twice. As far as I can tell, FSF and/or Cygnus is ignoring them.
That's quite strange; I used to be a gcc developer back in the 80s...