The openQA failure here looks like a genuine one: the test of a basic database deployment fails. After doing postgresql-setup --initdb, the test does su postgres -c "/usr/bin/createuser openqa", and this fails with This account is currently not available.

This is very likely caused by the attempt to switch to sysusers, I guess.

This failure is awkward because it's been "inherited" by the icu 83.2 update - FEDORA-2023-18495e9c7c - and also, other updates are now failing the database server test because the existing stable postgresql is not rebuilt for perl 5.38. We really need a postgresql that's rebuilt for perl 5.38 but does not have the sysusers problem.

Update, the failure happens because the change to sysusers was done in such a way that the postgres user no longer has a login shell, whereas previously it did. The openQA test's use of su postgres -c thus fails.

We could just adapt the test to work with postgres not having a login shell, but a cursory internet search indicates there are lots of docs, stackoverflow answers, reddit replies etc. which suggest commands that only work if postgres has a shell. Removing postgres's shell would be a substantial change that should probably be carefully planned and communicated, and it was not the intent of the sysusers migration, so it seems better at least for now just to give postgres its shell back. We can always, later, consider removing it in a more intentional and clearly communicated way, if we want to do that.

So for now, I'm doing new builds with the shell put back (thanks to @travier 's patch), -7 as a standalone update not rebuilt for the new ICU, and -8 rebuilt for the new ICU and which will be edited into the ICU update.