maya%NetBSD.org@localhost writes: > I believe that the change to tolerate /var/db/pkg silently has made it > possible to update setups without harm, and makes upgrade scripts less > urgent. It has helped greatly. But there is still the PKG_DBDIR value in the mk.conf world, which gets passed in via -K to various operations; I see it in make replace (why it's there, I'm not sure, but that's for another day). I feel that I can't prove there are no troubles from having a mismatch from mk.conf and fileysystem, and that having users force-point their pkgsrc/mk world and tools world to their actual database dirs is going to take a lot of problems -- some of which I probably haven't even thought of -- off the table. In addition, lots of people have somewhat broken setups from manual migration, and this script checks that the two databases are in a pair, there is only one set, and that the paths in the refcount database are ok. Some of this, including refcount paths, can be fixed automatically. So part of the point is to repair people's setups. For someone in the situation of not having moved, and having tools with your fix, all the "fix" operation will do is add PKG_DBDIR=/var/db/pkg to /etc/pkg_install.conf, /usr/pkg/sbin/pkg_install.conf and mk.conf. Basically, I am not comfortable cutting 2020Q4 without a check/fix script for users.
Attachment:
signature.asc
Description: PGP signature