Enda O'Connor wrote: > Dennis Clarke wrote: >>> Enda O'Connor wrote: >>> >>>> Bart Smaalders wrote: >>>> >>>>> Enda O'Connor ( Sun Micro Systems Ireland) wrote: >>>>> >>>>> >>>>>> Hi >>>>>> How does IPS rid itself of the various scripts ( postinstall say ) >>>>>> that >>>>>> SVR4 uses to provision your system post pkgadd, ie say add_drv or >>>>>> so on. >>>>>> >>>>>> >>>>>> >>>>> We've just started along this path. We're basically breaking tasks >>>>> performed in post-install scripts into two bunches: those that need >>>>> to happen before boot/during install, and those that can wait after >>>>> restart (whether service or OS). The tasks that are done immediately >>>>> are called actions. Right now we have file, directory, link, >>>>> hardlink, >>>>> license, and driver actions. The latter action takes care of calling >>>>> add_drv/update_drv as needed, for example. >>>>> >>>>> There will be more actions; clearly ones to take care of the *attr >>>>> files, creating users/groups, etc are needed. >>>>> >>>>> We're also going to provide standard mechanisms to handle post-restart >>>>> services; these are likely to be more flexible than the actions and >>>>> are easier for everyone to code, since they only have to handle a >>>>> single run-time context. >>>>> >>>>> One of the great benefits of eliminating scripting is that we can >>>>> validate the behavior of the actions on zones; it should be >>>>> possible to update disabled zones or cloned zone filesystems w/o >>>>> security issues that require the use of scratch zones today. >>>>> >>>>> - Bart >>>>> >>>>> >>>> Hi Bart/Stephen >>>> sounds interesting, what would the post restart service depend on etc ( >>>> ie filesystem/local ), and who would you see writing them, package >>>> developers? >>>> >>> Well, it seems as if this might be handled by adding config methods >>> to the services smf scripts; the config methods would run when the >>> service's dependencies are satisfied. Stephen has done a lot more >>> thinking on this, I'm sure. >>> >>> >> >> I spend a lot of time sitting on the sidelines and I don't say much. >> Mostly >> because I fear that I don't fully understand the problem.[1] One of the >> things that crossed my mind was the slow migration that we have >> already seen >> from the old SVR4 init scripts to the SMF infrastructure. I have my DNS >> servers run this way with hand crafted SMF scripts on Solaris 10. It >> seems >> very reasonable to me that any future package system would extend this >> thinking towards package install both pre and post. >> >> So, as verbosely as possible, I am simply saying that I like the idea >> but I >> don't see *how* you are going to handle the actions that need to be >> taken. >> >> - >> Dennis Clarke >> >> [1] I sync with the pkg code tree regularly and I have built it on S10 >> with >> the addition of Python 2.5.1 from Blastwave. I have even taken a crack >> at a >> build on Solaris 8 where libmd raises an issue. But I don't have a solid >> grasp on all the issues yet. >> >> >> > Hi Bart/Stephen > Yes, I like the idea as well, but I suspect it might be harder than > anticipated. But that only my intuition, no hard data there. > Some of the things I have seen over the years in package scripts can be > bizarre to say the least. > > With some direction I'd be happy to have a crack at helping out here in > any capacity you say fit. > > Enda
Excellent! Grab a clone of the pkg gate, take a look at the list of bugs and have at it... there's really _no_ shortage of work :-). - Bart -- Bart Smaalders Solaris Kernel Performance [EMAIL PROTECTED] http://blogs.sun.com/barts _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
