pre_inc/pre_dec/PUSH_ROUNDING inconsistency

[Richard Henderson]

On Sat, Jul 29, 2000 at 12:17:00AM +0200, Jan Hubicka wrote:
> It looks to me, that current POST_MODIFY patches are exactly what we
> need for way out of the PUSH_ROUNDING troubles.  Does sound sensible
> to you patch to emit_push_insn that will just use POST_MODIFY instead
> of PRE_DEC in case the PUSH_ROUNDING != GET_MODE_SIZE?

Or rather PRE_MODIFY you mean?  *shrug*  Perhaps.  The only
issue I see is that PUSH_ROUNDING is actually lying about
what the machine is doing, and that of course is causing the
problems.


r~