On Wed, Jan 14, 2009 at 07:25:27PM -0800, David Bustos wrote: > Quoth Nicolas Williams on Wed, Jan 14, 2009 at 09:08:09PM -0600: > > I guess what I'm saying is: when the repository isn't > > /etc/svc/repository.db then the deletemanifest sub-command should always > > fail. > > Oh, that hadn't occurred to me as necessary. What's the motivation?
That the repository you're editing need not be directly related to /var/svc/manifest on the system where you're editing that repo. You don't want to damage a system just because you want to delete an FMRI and associated manifest on another system's repo using svccfg(1M). Perhaps what should happen is this: ... delmanifest(...) { ... if (!is_system_repo(repo)) record_delmanifest(repo, fmri); else delete_fmri_and_manifest(repo, fmri); ... } And at startup svc.config should check if the repo is the system repo, and if it is then perform delayed delmanifest actions recorded by record_delmanifest(). Nico --