>  > By "configury magic" I thought you meant bits like gnu/lib/libgcc/m68000.mk,
>  > which are what I haven't checked in yet.
> 
> Ah.  I just noticed that GCC doesn't default to the correct target mask
> for m68000 ... I'm going to fix that now, and integrate the change into
> my gcc-current m68k-elf patches.

Oh yeah, right, I forgot about my config/m68k/netbsd-m68000-elf.h.
FWIW here's what I use:

[snip]
#include <m68k/netbsd-elf.h>

/* plain 68010 */
#undef	TARGET_DEFAULT
#define	TARGET_DEFAULT	(0)

/* Don't try using XFmode.  */
#undef	LONG_DOUBLE_TYPE_SIZE
#define	LONG_DOUBLE_TYPE_SIZE 64

/* Define __HAVE_68881 in preprocessor only if -m68881 is specified.
   This will control the use of inline 68881 insns in certain macros.
   Also inform the program which CPU this is for.  */
#undef CPP_SPEC
#define CPP_SPEC \
"%{m68881:-D__HAVE_68881__ }%{mfpa:-D__HAVE_FPA__ }\
%{!ansi:%{m68020:-Dmc68020}%{mc68020:-Dmc68020}%{!mc68020:%{!m68020:-Dmc68010}}} \
%{m68020:-D__mc68020__ -D__mc68020}%{mc68020:-D__mc68020__ -D__mc68020}%{!mc68020:%{!m68020:-D__mc68010__ -D__mc68010}}"

/* Invoke as very simply. */
#undef	ASM_SPEC
#define	ASM_SPEC \
 "%| %{fpic:-k} %{fPIC:-k -K} %{m68020:-mc68020}%{mc68020:-mc68020}%{m5200:-m5200}%{mcpu32:-mcpu32} %{m68881:-m68881}%{mc68881:-mc68881}"
[snip]

-- 
Matt Fredette