Nicolas Williams wrote: > On Mon, Nov 24, 2008 at 04:28:09PM -0800, David Powell wrote: >> There isn't a "run this service from the package you're removing >> before the package is removed" mechanism. > > No? Why not synchronous enable followed by synchronous disable? The > only problem: how to tell that service that what's happening is that a > pkg related to it is being removed. That could be done by dropping some > file into a deaddrop for that service, but that's not going to work when > the pkg being removed is the one that delivered that service (also, > better than a deaddrop: provide an API by which the service methods > could query what's up).
Such a mechanism doesn't exist today. It could be implemented, but that doesn't mean that the IPS team should or will. The primarily problem is it just moves the packaging script from one place to another. The only difference between IPS + this mechanism and System V packages is this makes your package more sensitive to the environment in which it is uninstalled: if you can't start and stop the service before you remove the package, it doesn't work. While the constraints the IPS team have put into place for IPS are really annoying, the reason for those constraints is to make delivery of software *less* sensitive to the install environment. Whether or not these constraints are reasonable is a separate question, one which Bart and Jordan are debating in the other branch of this thread. Dave _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
