Fix random hashing problems with ARRAY_REF and COMPONENT_REF

[Roger Sayle]

On Tue, 20 Jul 2004, Richard Kenner wrote:
>     Wouldn't it be preferable to handle these cases in the recursive calls to
>     COMPONENT_REF and ARRAY_REF instead of slowing down the critical path for
>     every invocation of operand_equal_p?
>
> Yup.  That's why the comment says it's temporary.


My apologies for giving you a hard time.  As pointed out by David
Edelsohn, this thread is a continuation of "Failure to bootstrap for
trunk on i686-pc-linux-gnu" on the gcc list, and your kludge had
been preapproved by Mark (both of which you failed to mention in
your post).

Perhaps this might serve as yet another reminder that, in 2004,
you really should be using a threaded mail client.

Sorry for the noise.

Roger
--