At Mon, 12 Mar 2001 07:04:41 -0800,
Matt Thomas wrote:
> 
> At 12:38 PM 3/12/2001 +0000, Ben Harris wrote:
> >On Mon, 12 Mar 2001, IWAMOTO Toshihiro wrote:
> >
> > > As hpcarm is finally approaching to multiuser and it should use
> > > sys/arch/hpc/hpc/disksubr.c, I'd like to move disksubr* references
> > > in files.arm to files.arm26 and files.arm32.
> > > (disksubr*.c will continue to stay under arch/arm/arm.)
> >
> >That seems like a reasonable idea to me.
> 
> Umm, then put sys/arch/hpc/hpc/disksubr.c is files.hpcarm and it will
> override the disksubr.c in files.arm.  If you don't want any of the
> disksubr.c in files.arm, can't you just put a !hpc on them?

Having !hpc in files.arm would look messy.  Imagine files.m68k with
!amiga, !mac68k and !x68k.

> > > Also, I'd like to change #ifdef __arm32__'s in
> > > src/sbin/disklabel/disklabel.c to #ifdef FILECORE and define macros in
> > > its Makefile based on ${MACHINE} value.
> >
> >Hrm.  This seems wrong to me (though so is leaving tests based on
> >__arm32__).  We'd really like all binaries to be the same across ARM
> >systems if at all possible.  Would it be feasible to switch at run-time?
> 
> This seems wrong to me as well.  The tests should be based on __arm__
> and use the value of MACHINE to determine if the code should be used.
> (Note that the shark and the netwinder don't need FILECORE support as well.)

I'm not sure I want hpcarm userland set different from arm32.
At the moment I'll try runtime switching.

--
IWAMOTO Toshihiro