Mike Gerdts wrote: >On 8/10/07, Ethan Quach <Ethan.Quach at sun.com> wrote: > > >>So are we talking about /var/sadm on the running system getting lost or the >>/var/sadm in your wanboot miniroot image mounted at /mnt ? If the latter, >>then there might be an explanation (or at least a known issue that explains >>it). In the miniroot image, ./var is symlinked to ./tmp/root/var and >>there is a >>known issue in the pkg tools where if the pkg delivers something into >>./var, it >>breaks that symlink. >> >> > >Taking a look at a broken wanboot image (not sure if this is the same >one or not) > ># find var >var >var/svc >var/svc/manifest >var/svc/manifest/system >var/sadm >var/sadm/pkg ># find tmp/root/var | wc -l > 3346 ># cat tmp/root/var/sadm/patch/.patchaddLock >3643 > >Note that the var symlink was broken and I was left with a small >skeleton directory there. The likely culprit was patchadd, not pkgrm. > >Mike > > > Hi If this was a miniroot, then you need toset PKG_ABI_SYMLINKS=true before adding/removing packages/patches that affect /var, in particular those that deliver manifests, as these all deliver to /var. But this does not affect regular patching to a live system or luupgrade -t. note that patching/packaging into a miniroot is very tricky in general, also need to be aware that packages that deliver manifest methods might also affect a miniroot booting, for instance applying a KU to a miniroot even using PKG_ABI_SYMLINKS=true will result in an unbootable miniroot, unless one modifies the smf repository to reset freshly delivered methods start/exec back to true. In general adding patches/packages to a miniroot is very messy, I logged some bugs on this subject ( trying to get the tools to determine PKG_ABI_SYMLINK as opposed to end user, and also in relation to the SMF issues arounf redelivery of methods that overwrite those that are no op in the miniroot )
Enda Enda
