Re: RFA: ``set mips abi''

[Daniel Jacobowitz <drow@mvista.com>]

On Tue, Jun 18, 2002 at 06:25:57PM -0400, Andrew Cagney wrote:
> >
> >If we didn't already have and need the enum all over that file, I'd
> >agree with you.  I don't see a point in all the extra globals.  But
> >hey, I don't mind.
> 
> Yes, I noticed that.  I was thinking of just changing the global_...() 
> function to be:
> 
> 	if (mips_abi == mips_abi_o32)
> 	  return MIPS_ABI_O32;
> 	else if (...)
> 	..
> 	else
> 	  internal error()
> 
> so that the rest didn't need to be changed.  It is how it has been done 
> in the past.

Then you've got a function and an enum to keep in sync.  That's even
more complicated than an array and an enum, in my opinion... not an
advantage at all.  I guess the internal_error will kick you if you do
that... I'll clean this up before I commit it.

> >>(Having an enum mechanism that bound a number to a name would be nice).
> 
> >You can do it very easily with designated initializers, but they are
> >not adequately portable.  You can do it very easily building the array
> >at runtime but why bother?  Keeping two lists in sync is not the most
> >complicated thing in the world.
> 
> (what's a designated initializer?)

I think I'm mixing terms here; I'm thinking of two language extensions
in GCC and c99.  The pertinent one is:

void *array[] = {
  [1] NULL,
  [2] some_void_ptr,
  [3] NULL,
};

I think you can use enum values as the tags but I'm not quite sure.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer