Tony Nguyen wrote: > Rich, > > Thanks for review the changes. My comments inline. > > Regards, > tony > > Richard Lowe wrote: >> Tony Nguyen wrote: >>> Apologies for not making the webrev public. It's now at >>> >>> http://cr.grommit.com/~tonyn/6537863/webrev/ >> >> I'm not sure if you have already have review from people more SMF >> aware, so I'll stick to the less SMF specific bits. >> >> usr/src/lib/libscf/common/midlevel.c:527 >> You never seem to use 'exist' here. >> > You're right. Leftover from my last change :^). I'll remove 'exist'. >> If I'm understanding the more SMF specifics bit correctly, the change >> forces the general/enabled properly to exist, even if the actual >> enabling is temporary. If this property is generally required, why >> not force its existence on import (and in similar places)? (I suspect >> there's a reason, I just can't find what it is). > We normally create the general/enable property after importing the > manifest, at the time when we need to set the service to either enabled > or disabled. However, there are two cases where we don't have a > general/enable property > > 1. Someone explicitly does 'svccfg -s fmri delprop general/enable' > 2. When we programmatically creates a new instance > > If we temporarily enable/disable(svcadm enable -t ) the instance after > any of the above two scenarios, the command won't work correctly since > general/enable doesn't exist. Does it make sense? >
Ah, yes it does, thanks. I'd considered deleting the prop as a generally unwise thing to do, I hadn't noticed the second situation. -- Rich
