Re: [commit] Add BE/LE floating-point type & format tables

[Andrew Cagney <cagney@gnu.org>]

>    Date: Sun, 08 Aug 2004 17:24:26 -0400
>    From: Andrew Cagney <cagney@gnu.org>
> 
>    My original post gave one example, here's another:
> 
>    static struct type *
>    mips_float_register_type (void)
>    {
>       if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG)
> 	return builtin_type_ieee_single_big;
>       else
> 	return builtin_type_ieee_single_little;
>    }
> 
>    Is all over the place.  This:
> 
> 	   builtin_type_ieee_single[gdbarch_byte_order (gdbarch)]
> 
>    is shorter and correct.
> 
> Ah OK.  But is this going to get used for some other target than MIPS?
> It seems to me that you're introducing a lot of machinery just to make
> life a bit easier there.

I've also fixed bugs where that if (...) logic was wrong.  I've a 
greater confidence in the correctness of this code.

 > Oh, and your code really needs to do some
> bounds checking too, so I really doubt whether it is much simpler.

?  The occasional assert (gdbarch_byte_order == BFD_ENDIAN_BIG || 
BFD_ENDIAN_LITTLE) might be interesting, but beyond that?



>    > Anyway, could we rename floatformat_ia64_quad to
>    > floatformat_ieee_quad.  There are quite a few systems out there that
>    > use or will be using the same format, and I believe i will be in a new
>    > revision of the official IEEE standard for it.
>    > 
> 
>    local to gdb?
> 
> Whatever.

Andrew