I just checked in a change (borrowed from the egcs repository; the bug
had already been fixed) that fixes the compilation error in if_we.c,
and the one in GNU malloc on the m68k.

I've also checked in a bunch of changes to fix ARM-specific problems,
and enabled egcs by default on the ARM.

Steve Woodford has indicated that he's doing a full build on a m68k,
so we should know soon whether that works.

Someone needs to test whether these changes fix the mysterious (and
not reproducible) if_le lossage on the SPARC.

I'll see if I can get the Mac booting again to test it on PowerPC.