racluster request

[rick]

>
> Hey Rick,

> I've added the 'cont' flag to the racluster() configuration strategy.
> So, if a configuration line has a cont at the end it just keeps going
> to the next line in the file.

thankyou indeed. That will help somewhat reducing passes over data :) and
without breaking existing applications :)

>
> Just a first step, but could you test it a few times to see if it  
> provides
> any of the functionality that you were looking for?
>

indeed it does. This seems to work well  (after fixing my config file
several times as it was getting very late and i wasn't thinking well :))

This seems to be functioning well in tests on ipv4.. it works in v6 aswell
tho there is a slight issue with it i'm trying to track down. For some
reason one of my filter="..." gets in there and nothing else does.. or the
other ones get in there and it doesn't.. depending on whether it is first or
not.. if it is in the middle of the bunch that work then it also gets
ignored.. quite bizaar.. i'm trying to locate it and / or determine a better
description of the problem :)

To test this with ipv6 i've had to fix the ipv6 bits a bit.... more on this
in the other more ipv6 related issue mail thread in a tick :)

speaking of threads.. i've noticed signals used in the code along with
pthreads.. this is a bad combination as the outcome and behaviour of signals
in threads is not clearly defined by posix and the outcomes are possibly
more annoying to deal with than determining which of poll()/select() to use
for the case of the SIGALARMs in there at least..