On Mon, Jul 02, 2007 at 03:00:02PM +0100, Matthias Scheler wrote:
> On Mon, Jul 02, 2007 at 03:21:23PM +0200, Joerg Sonnenberger wrote:
> > > I'm not objecting to a better solution. But just removing it will
> > > result in e.g. "milter-greylist" to build the "bind9" package
> > > for no good reason. And that's hardly an improvment from the
> > > users point of view.
> > 
> > Attached is a first stab at a more generic solution. The operating
> > systems with a thread-safe resolver needs to be added, I don't think we
> > can detect that automatically.
> 
> "milter-greylist" checks for res_ninit().

...which is a sign for a reentrant resolver, not a thread-safe one.
I'd patch milter-greylist to always use the library we are providing and
ignore the check result.

Joerg