On Thu, Sep 12, 2002 at 10:19:47AM +0100, Ted_Rule@flextech.co.uk wrote:
Ah.. perhaps the light begins to clear.
I think I must have misinterpreted what the keep-alive option really does.
I had thought it avoided re-opening a listener on the same socket on SIGHUP.
From your reply, I'm now assme that all it does is ensure that 'ephemeral' connections accept()'ed from a listener are left open across a SIGHUP even if the listener is taken away.
Is this nearer the truth?
yes, absolutely.
In which case , as you say, my patch is irrelevant to my real desire to avoid a re-open of any given listener ( of potentially any flavour ) across a SIGHUP, and the previous post about repairing the 'broken' glibc/syslog() call across a re-opened Unix DGRAM socket potentially provides an easier fix than a rewrite of syslog-ng.
And/or avoiding use of Unix DGRAM sockets at all on glibc platforms until such a fix is in place.
It's not a complete rewrite, but not a trivial addition. I rather spend time on my syslog-ng 2 tree. PS: I might release a snapshot of syslog-ng 2, if you like -- Bazsi PGP info: KeyID 9AF8D0A9 Fingerprint CD27 CFB0 802C 0944 9CFD 804E C82C 8EB1