On Tue, Jan 20, 2009 at 08:38:30PM +0100, Tino Keitel wrote: > On Tue, Jan 20, 2009 at 17:54:20 +0100, Rémi Cardona wrote: > > hald already has a pipe set up between the launcher process and the > > daemon process. The launcher process exits when it receives a "message" > > on the pipe from the daemon. Maybe that message could be delayed until > > HAL is _really_ up and running. > > If I got this right, this implies that HAL still detaches from the > terminal after startup, which is a bad design for a init system wich > tries to implement service dependencies (you loose controll over the > service if it detaches from the controlling terminal). However, the > init system I use (runit) has the concept of a service specific check > script. The service is reported as fully started if this check script > returns success.
You are right on designing a reliable system. HAL is wrong on this point. But at here we've just put this wrong into use, to make HAL do it's init at the background and, in the foreground, we run forward to start other services. Most new systems ships 2 or more processors so this trick _do_ saves boot time. -- Li, Yan _______________________________________________ xorg mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/xorg
