> > For packages with maintainer scripts that do "invoke-rc.d restart" or > > equivalent, the multiuser runlevel directories must contain either an > > S or a K symlink. This will remain true even when #243159 is > > implemented. > > Must this be so? As long as the no-S-and-K-links situation is officially > undefined, it wouldn't do any harm to let invoke-rc.d actually do a > "try-restart" when called with "restart" and no links exists, would it?
Your previous message proposed that "invoke-rc.d restart" (or "invoke-rc.d stop" followed by "invoke-rc.d start") do "stop" to a service that is configured as "floating". Now you are proposing that "invoke-rc.d restart" do a "try-restart" in that case. This is different, so I am confused. I have already written that having "invoke-rc.d restart" do "stop" to a service would be OK and might even be preferable to having it do "restart" (which is what it currently does). Neither action is ideal, and that is why the "floating" state can't be regarded as supported, currently. On the other hand, we can't have "invoke-rc.d restart" (or the equivalent) do "try-restart" because this requires support for the "try-restart" method in the initscript and there is no way for invoke-rc.d to know whether or not the "try-restart" method is implemented in a particular initscript. -- Thomas Hood -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
