OK. So a more substantial patch this time.
The original 8 bit H8/300 part is horrifically old. They probably haven't been made in 20 years or longer. While the tester builds multilibs for that part, we don't test them. Dropping support for the original H8/300 part ultimately allows for some useful simplifications in the remaining code and reduces the amount of work necessary for the cc0->CC_REG transition. We still support the H8/300H and newer parts and default to H8/300H code generation. This patch does _some_ simplification, but there's more to follow... This doesn't drop the "normal" multilibs, but we might in the near future. The "normal" multilibs allow building H8/300H code that runs with 16bit pointers. There's just not much to be gained from a simplification standpoint in dropping "normal" support. I've verified with this patch doesn't alter the code generated for the chip variants we still support -- at least across libgcc and newlib. Committing to the trunk. Jeff
