Burst filtering (same PV detection) doesn't work well with aliases

[ralphlange]

Currently, the messages that are sent from the IOC to the caPutLogTasks use a pointer to the field (from the DBADDR structure) as a unique key for the burst filtering.

That leads to writes that use different names (aliases) of the same record being treated by the burst filter as writes going to the same PV.

I do not think this is the correct behavior. Writes to different PV names should be seen as different operations.
(Otherwise, two moving sliders writing to different aliases of the same record would be seen as a single operation, one of which would possibly not be logged.)

In the CA Gateway, there is no DBADDR structure. The address of the PV name string inside the gdd container is used as key, which is a bug as it effectively prevents burst filtering.

A hash of the PV name might be a better-suited key that works in both use cases.

[ralphlange]

If concurrent write bursts from two different machines or users should appear as separate caPutLog messages, the hash should actually be on pv + user + host.

Note that this would call for separate burst filtering by hash...