Re: [rfc, frame] Always check for unsaved PC

[Andreas Schwab <schwab@suse.de>]

Daniel Jacobowitz <drow@false.org> writes:

> @@ -1111,6 +1141,42 @@ get_prev_frame_1 (struct frame_info *thi
>        return NULL;
>      }
>  
> +  /* Check that this and the next frame do not unwind the PC register
> +     to the same memory location.  If they do, then even though they
> +     have different frame IDs, the new frame will be bogus; two
> +     functions can't share a register save slot for the PC.  This can
> +     happen when the prologue analyzer finds a stack adjustment, but
> +     no PC save.  This check does assume that the "PC register" is
> +     roughly a traditional PC, even if the gdbarch_unwind_pc method
> +     frobs it.  */
> +  if (this_frame->level > 0
> +      && get_frame_type (this_frame) == NORMAL_FRAME
> +      && get_frame_type (this_frame->next) == NORMAL_FRAME)
> +    {
> +      int optimized, realnum;
> +      enum lval_type lval, nlval;
> +      CORE_ADDR addr, naddr;
> +
> +      frame_register_unwind_location (this_frame, PC_REGNUM, &optimized,
> +				      &lval, &addr, &realnum);
> +      frame_register_unwind_location (get_next_frame (this_frame), PC_REGNUM,
> +				      &optimized, &nlval, &naddr, &realnum);
> +

This is broken.  You can't use PC_REGNUM unconditionally without checking
whether that register actually exists.  The ia64 does not have such a
register.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."