More /u conflicts

[Richard Henderson]

On Wed, Feb 14, 2001 at 06:48:11PM -0500, Richard Kenner wrote:
> *Now* we have two stores into a /u value, the second one being originally
> inserted by regmove.

Regmove attempts to check for the /u bit and not perform
such invalid optimizations.  Apparently it missed a case.

> I think the only way to deal with this is to view two /u references as
> always conflicting.

I think you've just uncovered a bug in regmove.  Even if
the pseudo _did_ get a hard register, we have invalid rtl
coming out of regmove: /u on a reg means it is assigned
not more than once.

> Note also that this does not show up without -g, meaning that register
> allocation or something depends on -g and that's very bad news!

Eek!


r~