gcc internal failure during optimization delete_trivially_dead_insns

[Jeff Law]

On 04/10/2015 07:41 PM, ftwilliam wrote:
> Hi Jeff,
>
> My apology again; please use instead this patch.
You've got an insn with a bogus register number:

(insn 38 39 24 7 (set (reg:SI 27)
         (reg:SI -1)) 18 {*movsi}
      (nil))
$3 = void


-1 is not a valid register number.

This probably due to EH_RETURN_DATA_REGNO not being properly defined for 
your port.


jeff