On Mon, Jun 03, 2019 at 01:32:41PM -0400, Stephen John Smoogen wrote: > On Mon, 3 Jun 2019 at 13:17, Chris Murphy <lists@xxxxxxxxxxxxxxxxx> wrote: > > > On Mon, Jun 3, 2019 at 11:07 AM Adam Williamson > > <adamwill@xxxxxxxxxxxxxxxxx> wrote: > > > > > > On Sun, 2019-06-02 at 21:35 -0600, Chris Murphy wrote: > > > > Perhaps related, on an upgraded Fedora 30 system I see > > > > ghostscript-fonts-5.50-37.fc27.noarch, which does not appear on any > > > > clean installed systems, and also can't be installed (Error: Unable to > > > > find a match ). That tells me it's been dropped or is obsolete, so is > > > > it normal for such packages to persist through upgrades? > > > > > > Sure, very common. Packages are frequently retired and not formally > > > obsoleted by anything else: in this case, if you have them installed, > > > they'll stay installed until some dependency issue crops up and you > > > have to remove them manually (or use --allowerasing) to clear it up. > > > > And does gnome-software do --allowerasing, or equivalent? Or other > > behavior? > > > > > > > > Some people don't see any problem with this, personally it drives me > > > crazy and I wish it were policy that *every* retired package must be > > > obsoleted. But it isn't. > > > > Not obsoleting retired packages is arguably inconsistent with the > > Workstation PRD: > > > > "Upgrading the system multiple times through the upgrade process > > should give a result that is the same as an original install of Fedora > > Workstation." > > > > I understand that is a goal, not a policy or release criterion. > > > > > I expect it is an impossible goal as it is making the installer guess that > you didn't want to keep that version of wumpus from RHL6.2 which works > still but isn't in the repository. And forcing an obsolete has knock-on > effects. At best I can obsolete the version of emacs-freebird which came > from Fedora up until release N, but if the person has a version they > compile themselves.. then a centralized obsoletes has a good chance of > removing it unless the packager did exactly the right things in the > obsoletes and the other version of emacs-freebird also did the right > things. I expect that instead you end up with a lot of pissed off people... > which no one has the emotional labour to deal with. It's possible that this used to be true, i.e. that the number of people who compiled *their own versions* of packages that are already in the distro was non-trivial. Nowadays, I'm pretty sure this happens very very rarely. People install external and self-compiled packages mostly when they cannot get something from the distro. I don't have any formal numbers for this, but based on the bug reports that come in before and around every release, it's quite obvious that the number of users negatively impacted by non-obsoleted distro packages (*) dwarves the number of people who have an external package, and that in turn is still higher than those that have an external package with a lower nevra. (*) Fedora removes many many packages and subpackages on every release. So every non-minimal installation of Fedora will have such stale-but-not-obsoleted package after *every* upgrade. It's only a question of time until some so-version changes and such packages cause upgrades to break. Normal users are (correctly) wary of --allowerasing which is risky and requires a good understanding of packaging and recovery methods. I understand why this policy drives adamw crazy. I think we're shooting our users in the foot. Zbyszek _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx
