On Mon, Mar 04, 2002 at 09:10:56PM -0500, I wrote: > Actually, it was a different sig handler from the same manpage, IIRC. >
Errr... now I remember. I think that the sighandler for SIGCHLD blocks any new SIGCHLDs from occuring, meaning (as I interpret it) SIGCHLD doesn't get re-executed. The sighandler with -m causes all children to be cleaned-up after rather than just 1, as it was berfore v. 1.23 of spamd. > Don't use -m on BSD might be wise anyways, since essentially, my patch > simply allows connections to pile up without accepting them, until we have > the resources to handle them. I believe SOMAXCONN is 5 on BSD, so if you get > 11 mails with -m 5, you get some connections refused. > > On Mon, Mar 04, 2002 at 06:03:23PM -0800, Craig Hughes wrote: > > I had the perlipc manpage signal handler in there in the first place, > > and it had to be replaced with the one you just re-replaced. Definitely > > would be good to have a number of BSDers hammer on this patch before I > > roll it in. Otherwise, I could roll it in with a giant "Don't use -m on > > BSD" warning. > > -- Duncan Findlay _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
