idle_timeout processing in the parent smbd?

[James Peach]

On Jun 4, 2007, at 9:22 AM, Volker Lendecke wrote:

> On Mon, Jun 04, 2007 at 08:44:47AM -0700, James Peach wrote:
>> How can you distinguish between an event that represents real work  
>> and
>> an event that represents internal housekeeping?
>>
>>> What to others think about an idle event API that says
>>> "please run me every minute if you can, but every ten
>>> minutes at least" or something like this?
>>
>> This seems like something that's generically useful. The hardest part
>> aboout idle events is when the definition of idle depends on another
>> subsystem. for example, it's quite easy for an LDAP subsystem to  
>> track
>> the last time it was called and idle out that way. For exit on idle,
>> however, the idle event needs to be able to introspect other events
>> and figure out the last time they were idle.
>
> Well, probably in that particular case the best thing would
> be to just exit once the number of children drops to zero.
> Why wait at all?

I'm perfectly happy with that approach (as per the patch I suggested  
earlier). I'll commit something along those lines.

--
James Peach | jpeach at samba.org