Re: [PATCH] fix out of bound access on hard_regno_nregs

[bernds_cb1 at t-online dot de (Bernd Schmidt)]

Graham Stott wrote:
> All,
>
> int
> true_regnum (rtx x)
> {
>   ...
>
>   if (GET_CODE (x) == SUBREG)
>     {
>       int base = true_regnum (SUBREG_REG (x));
>       if (base >= 0 && base < FIRST_PSEUDO_REGISTER)
>         return base + subreg_regno_offset (REGNO (SUBREG_REG (x)),
>                                            GET_MODE (SUBREG_REG (x)),
>                                            SUBREG_BYTE (x), GET_MODE (x));
>     }
>   return -1;
> }
>
>
> On 2nd reading it looks like we do have a bug in true_regnum.
>
> Assume that X is a SUBREG whos reg is a pseudo_register that has been 
> renumbered
> then it looks like we should be passing BASE and not REGNO (SUBREG_REG 
> (x)) onto
> subreg_regno_offset.
>
> Thoughts?

Looks like you're right.  This seems to have been broken by the 
subreg-byte patches (before that we had "base + SUBREG_WORD (...)").


Bernd