[RFC][PATCH] krb5 => libverto main loop

[Greg Hudson]

On Thu, 2011-08-25 at 12:53 -0400, Luke Howard wrote:
> That said, the convention on Darwin is that loadable modules have
> the .so suffix, and one would want to have a pretty good reason before
> defying this.

verto modules are linkable shared libraries as well as loadable ones;
applications can link to a verto module directly and call
verto_new_libev() or similar.  So, it makes sense for the linkable
shared library suffix to take precedence, since the loadable suffix
isn't as hardcoded.

The upshot is the same as I originally said, though: libverto needs to
expect its library pathname and module pathnames to contain ".dylib"
instead of ".so".

On another note, I did some testing (mostly to validate my libverto
integration) and discovered another problem when using the private libev
module: kadmind is accepting one TCP connection and then closing the
listener socket.  I'm not sure why.  The code in accept_tcp_connection()
around where it calls add_tcp_read_fd() looks broken (it grabs the
private data from ev instead of newev, and on failure it deletes the
listener socket event instead of cleaning up the socket), but that
brokenness doesn't account for the problem.