"Nigel J. Andrews" <nandrews(at)investsystems(dot)co(dot)uk> writes:
> Why does dropping a table for some dependencies give notice that the
> table is depended upon and for other dependencies that the primary key
> index for the table is depended upon?
It looks like a foreign-key constraint is marked as depending on *both*
the foreign table and the foreign table's pkey index. Which one gets
reported depends on the luck of the draw of traversal order in the
dependency tree.
I'm not sure whether this marking is overkill, though offhand it seems
it might be.
regards, tom lane