"David G. Johnston" <david.g.johnston@gmail.com> writes:
> Is this covering the case of executing at the end of an outer SQL command
> (thus not deferred) that contains volatile DML functions that temporarily
> change current_user within the function?
Not quite. I think that a non-deferred AFTER trigger would ordinarily
run as the same user that was active when we queued the event, earlier
in the same statement --- but it's possible that some function that
runs in between would change the active role in a non-temporary way.
Doing that will now have different effects than it did before.
regards, tom lane
