Fix PR optimization/7520

[Richard Henderson]

On Thu, Sep 19, 2002 at 11:58:13PM +0200, Eric Botcazou wrote:
> > If eax is live after this instruction in this block,
> 
> That's the case (see my first message). That's what I meant by "when the
> clobbered register is live after the insn".
> 
> > with no other uses, that is wrong.
> 
> So the principle of the patch (preventing flow2 from putting its REG_DEAD
> note) is correct ?

No.  The REG_DEAD note is not the problem.  The liveness
propagation happened elsewhere.


r~