Re: sh64-elf (SH5) port: directory opcodes

[Hans-Peter Nilsson <hp at bitrange dot com>]

On Fri, 8 Feb 2002, Andrew Cagney wrote:
> > On 5 Feb 2002, Alexandre Oliva wrote:
> >
> >> On Feb  5, 2002, Andrew Cagney <ac131313@cygnus.com> wrote:
> >
> >> > Anyway, if you always include support for all the SH variants, does
> >> > anything break?
> >
> >>
> >> Probably not, since --enable-targets=all works.
> >
> >
> > Though you'd bloat binutils for people with sh[1-4] only.  Some
> > SH targets run native in limited systems, I've heard.  Can we
> > ignore the bloat issue?
>
>
> Is this a generic problem?  Most of the small systems - MIPS, mn10300,
> sh, ... would all need to be tuned for small systems.

Perhaps, at least if they would suddenly attract other code,
especially code forcing BFD64.  For now, I don't see it has
happened for other targets.  For example, i386 targets haven't
included the x86-64 bfd vector in their configs.

> > Maybe let sh-elf imply bfd+opcodes for sh[1-5] and leave
> > sh[1-4][hl]* the way it is?
> >
> > (Including opcodes but not bfd seems useless.  You can't get a
> > sh5 bfd, so you can't (without tricks) invoke the disassembler
> > AFAICT.)
>
>
> By default having all of them is significantly better (I think).  It
> would also better integrate into GDB.

Is that an agreement to keep it in "sh" and keeping it out of
sh[1-4]{le,be} ?

brgds, H-P