On Wed, Nov 28, 2001 at 03:07:27PM +0000, Nick Clifton wrote: > In the particular case of the arm-elf build these the RTX pointed to > by recog_data.operand_loc[i] was the REG inside the SUBREG held in > recog_data.operand[i]. (Is this a bug, I am not sure ?) Yes, this is a bug. *operand_loc[i] == operand[i]. r~