Re: Role Attribute Bitmask Catalog Representation

FWIW I've been giving this patch a look and and adjusting some coding
details here and there.  Do you intend to commit it yourself?  You're
not listed as reviewer or committer for it in the commitfest app, FWIW.

One thing I don't very much like is that check_role_attribute() receives
a RoleAttr but nowhere it checks that only one bit is set in
'attribute'.  From the coding, this routine would return true if just
one of those bits is set, which is probably not what is intended.  Now I
realize that current callers all pass a bitmask with a single bit set,
but I think it'd be better to have some protection against that, for
possible future coding mistakes.

-- 
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services