Re: Interest in a SECURITY DEFINER function current_user stack access mechanism?

On Wed, Oct 18, 2017 at 02:13:29PM -0700, David G. Johnston wrote:
> > More useful than this, for me, would be a way to get the top-most user.
>
> That would be "session_user"?

It's not quite since there's a difference between SET SESSION
AUTHORIZATION and SET SESSION ROLE.

But yes, it's what I'm using now.

> > Introducing the concept of a stack at the SQL level here seems, at
> > > first glance, to be over-complicating things.
> >
> > Because of the current implementation of invocation of SECURITY DEFINER
> > functions, a stack is trivial to build, since it's a list of nodes
> > allocated on the C stack in fmgr_security_definer().
>
> Not saying its difficult (or not) to code in C; but exposing that to SQL
> seems like a big step.

Really? Why? I mean, there's an implicit function invocation stack
already. Reifying some bits of the function call stack is useful. I
can't think of how this particular reification would be dangerous or set
a bad precedent.

Hmmm, oh, I forgot about GET DIAGNOSTICS! The stack is already exposed
to SQL. Maybe we could add a CURRENT_USER item to GET STACKED
DIAGNOSTICS or to the PG_CONTEXT.

> If I was in position to dive deeper I wouldn't foreclose on the stack idea
> but I'd be inclined to see if something else could be made to work with
> reasonable effort.

I would think that the more general approach, if easy enough to
implement, would be better. I can (and will) live with using
session_user instead of current_user, for now. But I'm willing to
contribute a patch.

Nico
--