Do you think there's still a case for having this around?

We always decrement the PC first thing in handle_inferior_event,
before reiniting the cache.

Of all other cases I can think of that end up calling normal_stop,
if we end up here without a call to reinit_frame_cache, this hack
is wrong/insuficient anyway.

While I'm at it, I can eliminate deprecated_update_frame_pc_hack
and deprecated_update_frame_base_hack, by inlining them in their
only caller left.

Tested on x86_64-linux.

-- 
Pedro Alves