Fwd: radutmp bugs

[Alan Curry]

Alan DeKok writes the following:
>
>3APA3A <3APA3A@SECURITY.NNOV.RU> wrote:
>> Can  someone  confirm  this  problem and patch? I can do the patch but I
>> can't test it with SNMP.
>
>  I would STRONGLY discourage use of a global variable 'acctfd' in the
>server.  We want to add the ability to listen on multiple sockets in
>the future, so a global 'acctfd' will be wrong.
>
>  Another solution would be better, b ut I haven't had time to think
>about one yet.

Ideally, session_zap shouldn't need any socket at all. Putting the zap packet
into wire format and sending it to localhost is definitely a kludge. It
should instead build a REQUEST and inject it directly into the top of the
module tree. I just never found a clean way to do that since the code that
retrieves the packet from the network was too closely bound to the code that
begins module processing.