Sorry Aaron, the unixsocket code is not for 2.0 imo. Though it's pretty straightforward code, and appears to work fine, there are still several rough edges:
- too much code replication between create_unix_socket and create_inet_socket - PerformChildTask is rather PF_INET centric atm (which leads no weird ip adresses being logged atm). - the unix-socket approach is currently exclusive. Either unixsockets *or* inetsockets. Done right we want to offer both at the same time however. This means the accept(2) in PerformChildTask should be non-blocking and select/poll both the unix socket and the inet socket in a loop. And that implies changing several structs so we can hold and track both sockets at once. And yes Geo, we also will need the self-pipe pattern to deal with sigchld :-) All this will need more time to mature, since I don't think we want to introduce quick-and-dirty fixes in 2.0. That said, I don't have any problems with the backlog patch. Aaron Stone wrote: > On Wed, May 11, 2005, [EMAIL PROTECTED] said: > > >>A BUGNOTE has been added to this bug. >>====================================================================== >>http://www.dbmail.org/mantis/bug_view_advanced_page.php?bug_id=0000202 >>====================================================================== > > >>Summary: wish: user configurable listen backlog >>Description: >>In some cases is much better to set listen backlog to low number >>and refuse other connections instead of having client socket waiting in >>backlog queue. This refusing helps client programs to switch faster >>to another configured daemon. > > > > The patch looks great. I say we go for it. I like that it's been done for > all three daemons, too. Paul, would you mind doing likewise for SOCKET? > > Aaron > _______________________________________________ > Dbmail-dev mailing list > Dbmail-dev@dbmail.org > http://twister.fastxs.net/mailman/listinfo/dbmail-dev > -- ________________________________________________________________ Paul Stevens paul at nfg.nl NET FACILITIES GROUP GPG/PGP: 1024D/11F8CD31 The Netherlands________________________________http://www.nfg.nl
