[JDEV] Server feature request

[Dave Smith]

In the next iteration of server development, MIO will be modified such
that it will not will use SIGUSR2 for this. Having debugging turned on
via SIGUSR2 is intriguing and we'll look into that for the next
iteration. Probably, it would simply be used as a boolean -- if
debugging is off SIGUSR2 would turn it on and vice-versa. This way we
don't need to have so many signals just to turn debugging on and off.

Diz


On Thu, May 10, 2001 at 10:53:25PM -0500, temas wrote:
> SIGUSR2 is used by mio as a mechanism to break the loop so new sockets
> can be added and stuff like that.
> 
> --temas
> 
> On 10 May 2001 18:18:23 -0600, David Waite wrote:
> > kadokev at msg.net wrote:
> > 
> > > > Those are the definitions of those signals in BIND though. They have
> > > > different (already used) meaning within Jabber - they are used by pth
> > > > (and I believe MIO). They are simply user-defined signals.
> > >
> > > Right, exactly.
> > >
> > > So we pick two new signals to serve this purpose in Jabber, the concept is
> > > still the same. So we use SIGINFO to turn on debugging, and SIGWINCH to turn
> > > it off, or any two other signals that are not already taken.
> > >
> > > It would also be useful to have the ability to turn debugging on and off by
> > > sending an admin message, but the signal method is more important, as it would
> > > be very useful to debugging cases where I cannot connect/login at all, and thuse
> > > cannot send any messages...
> > 
> > I'd welcome "any two other signals which are not already taken." The system or parent
> > process sends those signals. I do not believe SIGINFO is defined on i386, its also
> > mapped to SIGPWR which is an exit on some other platforms, SIGWINCH is sent whenever
> > the terminal size changes - it sure wouldn't help me debug if the program behavior
> > changed when I resized the terminal I was getting output in :-).
> > 
> > If those signals weren't used internally, I agree it would be a really nice feature
> > to add (you actually only need one signal as a toggle, really). What is SIGUSR2
> > really used for?
> > 
> > > Kevin
> > >
> > > > -David Waite
> > > > kadokev at msg.net wrote:
> > > > >Server feature request, the ability to turn debugging on/off by sending a
> > > > >Unix signal to the jabberd process, something like this:
> > > > >
> > > > >     SIGUSR1   Turns on debugging; each SIGUSR1 increments debug level.
> > > > >               (SIGEMTon older systems without SIGUSR1.)
> > > > >
> > > > >     SIGUSR2   Turns off debugging completely.   (SIGFPEon older systems
> > > > >               without SIGUSR2.)
> > > > >
> > > > >The astute observer will notice that the two paragraphs above are verbatim from
> > > > >the documentation for BIND :-)
> > > ...
> > > > >>Yes!! It's been a long time I was thinking about it!
> > > > >>In certain cases it is just necessary...
> > > > >>
> > > > >>/MAciek
> > > > >>
> > > > >>On Thu, 10 May 2001 11:45:35 -0600 Malone, Pat wrote:
> > > > >>
> > > > >>>In trying to work through a nasty client problem today, I had 'if only'
> > > > >>>thought pass through my head.  So, here is a wish item for the jabberd
> > > > >>>developer gods:
> > > > >>>
> > > > >>> Enable the ability to send a sig of some kind to jabberd to toggle
> > > > >>>debugging on and off, a.k.a. the way debug can be toggled with named.
> > > > >>>
> > > > >>>It sure would be nice to toggle debug on when a problem shows up without
> > > > >>>having to kill the server.
> > > _______________________________________________
> > > jdev mailing list
> > > jdev at jabber.org
> > > http://mailman.jabber.org/listinfo/jdev
> > 
> > _______________________________________________
> > jdev mailing list
> > jdev at jabber.org
> > http://mailman.jabber.org/listinfo/jdev
> 
> _______________________________________________
> jdev mailing list
> jdev at jabber.org
> http://mailman.jabber.org/listinfo/jdev