Patch causes segv

[Dale Johannesen]

On Thursday, October 24, 2002, at 04:26  PM, David Edelsohn wrote:

>>>>>> Dale Johannesen writes:
>
> Dale> Might work, I didn't try it.  It's looking at a field named 
> set_in_loop
> Dale> and (without deep analysis) I suspect that needs to be set 
> correctly to
> Dale> get the right answer.
>
> 	The size of the array is in regs->num.  Check if
>
> REGNO (x) < regs->num
>
> If the pseudo has not been recorded yet, then you cannot check 
> set_in_loop.
> If the register number is larger than the array size, then the pseudo 
> was
> recently created inside the loop, and the pseudo has to be written 
> before
> it is read, so assume set_in_loop is true.

With this, or with my fix, 176.gcc builds successfully.  However it 
crashes
at runtime.  It also does this if your change is commented out, so it's
something else.  I'll see what I can find....