On Wed, Aug 13, 2008 at 8:39 PM, Rob Browning <[EMAIL PROTECTED]> wrote: > "Chris Lawrence" <[EMAIL PROTECTED]> writes: > >>> I suppose this should probably be serious, since it prevents >>> /etc/init.d/at from stopping the server. I think it may also kill an >>> attempt to run "/etc/init.d/atd restart". >>> >>> I'm also cc'ing this to lsb-base, since as I mentioned, I'm not sure >>> which package is actually at fault. >> >> Looks like it might be a problem with start-stop-daemon when a pidfile >> isn't specified... it matches on --name and your init script is named >> "atd". > > If that's the case (and if I understand the situation), I wonder if > Debian policy should forbid packages from using --name. Otherwise it > seems like the init.d start stop scripts, package control scripts, > etc. could end up killing random, unrelated user-created executables > -- not a particularly appealing result.
start-stop-daemon uses /proc/$$/stat; this gives the script name as name of the executable (instead of *sh, which is what you'd expect would happen from ps output). I think using --exec instead of --name would work (change line 116 of /lib/lsb/init-functions), but that might break other packages, particularly daemons that are written in interpreted languages. I could hack something with pidof like in the status_of_proc function which might work better in the general case. Chris -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
