Tom Lane wrote:
> read-modify-write type of operation it uses an exclusive lock, so only
> one can clear its pg_listener entry at a time. The 'waiting' ones you
> are seeing are stacked up behind whichever one has the lock at the
> moment. They shouldn't be waiting for long.
>
I certainly hadn't expected that to be the implementation technique -
isn't it smply that we need
a sngle flag per worker process and can set/test-and-clear with atomic
operations and then a
signal to wake them up?
Anyway - how hard would it be to install triggers on commit and
rollback? Then we could write
our own mechanisms.
James