John Naylor <john(dot)naylor(at)postgresql(dot)org> writes:
> Add assert checking to pg_leftmost_one_pos32() and friends
I can see that this was worth writing for testing purposes, but
is it really worth carrying permanently? Even in a debug build,
the ratio of cycles expended to chances of finding a problem seems
mighty poor, and you've done a lot of damage to the readability
of these functions too.
Maybe we could condition the duplicate computation on some
additional not-commonly-defined macro? That'd do little for the
readability issue though.
regards, tom lane