Re: Remove HeapTuple and Buffer dependency for predicate locking functions

OK, I kept only the small comment change from that little fixup patch,
and pushed this.

> I had proposed as alternative way in initial email and also later,
> didn't receive comment on that, so re-posting.

> typedef bool (*AMCheckForSerializableConflictOutCallback) (void *arg);
...
> Just due to void* callback argument aspect I didn't prefer that
> solution and felt AM performing checks and calling
> CheckForSerializableConflictOut() seems better.  Please let me know
> how you feel about this.

Yeah.  We could always come back to this idea if it looks better once
we have more experience with new table AMs.