Moinak Ghosh wrote: > Karen Tung wrote: > >> [...] >> >> I know that postrun can complete successfully in the proto area. But >> the proto area >> has a lot more GNOME packages than the microroot. Adding the few GNOME >> packages >> to the microroot already caused its size to grow dramatically. I can >> not imagine >> having to include all the GNOME packages in the microroot. >> >> So, my questions is: do you know exactly which GNOME package is required to >> setup whatever it is necessary for the desktop to start? >> >> In the Live Media and the Slim Prototype code, the postrun stuff is run in >> the proto area. Then, the microroot is setup with the list of files in >> the "mkbootcd.files.minimal". I examined the mkbootcd.files.minimal file >> carefully, and I don't see any obvious GNOME related stuff get copied over >> to the microroot. Does that mean the postrun for GNOME does not have to >> happen in the microroot? Perhaps if somebody can explain exactly how >> the postrun stuff setup the GNOME desktop in the Live Media case, what kind >> of GNOME desktop startup setup needs to be present in the microroot, >> then, we will >> be able to better debug this problem with getting GNOME to start with >> the Distro Constructor code. >> >> > > I think this is one limitation of the approach being taken to build the > microroot. > Doing pkgadds into the microroot will cause these problems and cause > microroot > to grow drastically over a period of time. > > Conceptually we are building a single installed image and creating a > microroot is > just an architectural necessity to get a LiveCD to work. As such only a > single > proto area should be built with all postinstall/postrun done only once - > these should > not have to be repeated for the microroot. That will also cause multiple > package > databases for the full proto and for the microroot. Two package > databases do not > make sense in the LiveCD environment and will cause trouble for the Transfer > Module. > > Having said that it is also not tenable to retain the microroot approach > being used > currently in the LiveKit. Maintaining a list of files does not really > make sense in > the long run. So I think a middle path needs to be taken. > > All we need to build the microroot is a list of files. Given a list of > packages that > should go into the microroot it should be possible to extract the list > of files and > cpio them from the proto area rather than doing the pkgadds, > postinstalls and all > the corresponding bloat. So we avoid the file list and avoid problems > associated > with pkgadds. > Moinak,
This sounds like you're talking about "marking" files within a package. So if you specified I want package X then this special package add (or whatever) would look within package X to see which files are marked for the miniroot and only copy those. These would be copied from the proto area and not the package repository. Is that because some of the files in the proto area have been modified? Or just because they are easily accessible in the proto area? Jean > Regards, > Moinak. > > >> Thanks, >> >> --Karen >> _______________________________________________ >> caiman-discuss mailing list >> caiman-discuss at opensolaris.org >> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >> >> > > _______________________________________________ > caiman-discuss mailing list > caiman-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >
