On Friday 20 June 2008 17:04, walter harms wrote: > > Can you test current svn? If that works, we can just go ahead and backport > > it. > > hi denis, > yep that is what we have done. we just replaced the 10.3 version with the svn > version. > at least no more daemons can be started :) > FYI: while testing we note something odd, when a daemon is killed sometimes > all shells > also gets closed. This is new.
That is a symptom of a particular flavor of ssd which matches processes by their executable's inode number. When you try to kill syslogd, it kills syslogd... and all other processes which are run by same executable. If your syslogd was busyboxed and e.g. telnetd and/or sh were busyboxed too, they will be killed too. ssd was suffering from multiple people having a problem with the way it handles --exec, devising a quick fix accourding to their particular idea what --exec means, fixing it, not updating documentation (or should I say "not creating documenttion" - nobody bothered to actually document what --exec means in the first place), and leaving it at that. That last big fix tried to do away with this practice - both help text and ssd.c now contain somewhat complete explanation how --exec is supposed to operate. > Unfortunately the guy who found the bug is home, so we will continue testing > on monday. > > should we still test the busybox-1.10.1-ssd.patch ? No, please concentrate on svn version. The point is, "what --exec FOO should do" is actually not so simple. It's ambiguous. Please, take a look at current ssd.c comment. If --exec doesn't do what it has to _according to documentation_ - fix it. If documented behaviour is actually happening correctly but you need something different - please give examples and provide rationale why --exec semantic should be changed. -- vda _______________________________________________ busybox mailing list busybox@busybox.net http://busybox.net/cgi-bin/mailman/listinfo/busybox
