Jeff Janes <jeff(dot)janes(at)gmail(dot)com> writes:
> When I drop the optional triconsistent support function for a gin index
> operator family, the system continues to use the function to support index
> queries anyway, until I close the session and restart it, at which point it
> reverts to using the binary consistent function.
> Is this a bug in the cache invalidation mechanism? Or am I just expecting
> too much of it?
There is no invalidation support for mucking with operator class contents.
regards, tom lane