On Thu, Apr 30, 2009 at 08:19:23AM -0700, Danek Duvall wrote: > On Thu, Apr 30, 2009 at 01:59:24PM +0200, Ramshankar wrote: > > Bart and I are debating this. The drivers are intentionally not loaded, > since I was (and am still) paranoid about having just-delivered code run at > package install time. We may compromise and have the post-execute phase of > the driver install load all drivers that had been added or updated, so that > at least all bits are on the system before the potential panic.
Forcing a reboot when it shouldn't really be needed (e.g., after adding drivers) is one of those things for which many people used to rightly denigrate one of our competitors' offerings (which used to require a reboot just to change DNS resolver config, once upon a time). But I agree with you that preventing a panic while pkg(1) isn't done with its package installation work is very important. Deferring driver loading to the very end, or, perhaps better yet, adding a new pkg(1) sub-command to run all the post-package installation aspects of actions strikes me as a very good idea. By making driver loading a separate step you also get to detect when a reboot really is needed, such as when a driver was updated and cannot be unloaded. At that point the user can be informed of the need to reboot for driver updates to take effect. > > If I understood right, even actuators don't "enable" a service on > > installation right? > > Correct. So not only does pkg(1) force a reboot when adding drivers, but it also does that when adding software that relies on SMF for self-assembly? See above. Nico -- _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
