Subject: Re: CVS commit: src/gnu/usr.bin/gdb53/arch/sh3el
To: matthew green <mrg@eterna.com.au>
From: Valeriy E. Ushakov <uwe@ptc.spbu.ru>
List: source-changes
Date: 10/28/2003 05:22:44
On Tue, Oct 28, 2003 at 13:04:08 +1100, matthew green wrote:

>    Module Name:	src
>    Committed By:	uwe
>    Date:		Tue Oct 28 01:58:23 UTC 2003
>    
>    Modified Files:
>    	src/gnu/usr.bin/gdb53/arch/sh3el: defs.mk
>    
>    Log Message:
>    Default to the little-endian bfd_elf32_shlnbsd_vec.
> 
> 
> you probably should also change the default in (i assume)
> gnu/dist/gdb/gdb/configure.tgt.

No, this is bfd's.

Apparently, I made a mistake with original configure or there's some
internal inconsistency in platform naming.  I happen to have the old
build tree around and it lists --host=shl-unknown-netbsdelf1.6ZC
(apparently guessed by configure) which bfd knows about, but matches
as sh*-*-netbsdelf* and defaults to big-endian.

The kosher host to configure for seems to be "sh3le--netbsdelf" (sic:
"le").  The difference for bfd is only the DEFAULT_VECTOR, and it
choses the correct vector from debugged programs just fine anyway.

NB: bfd and opcode are in disagreement about platform naming.  bfd
will compile sh5 support only if explicitly asked for a sh5/sh64.
OTOH, opcode will compile sh5 support for 'unqualified' sh- and shl-
as well and will fail to link.  So the '3' in that sh3le is important
to prevent this lossage.  Oh, and note that it's "le", not "el".

So now that I found the name that gets correctly sifted through this
twisty maze of shell case statements, I really don't feel like
touching any of their pervert logic.

SY, Uwe
-- 
uwe@ptc.spbu.ru                         |       Zu Grunde kommen
http://www.ptc.spbu.ru/~uwe/            |       Ist zu Grunde gehen