OK, I've edited this update to the 32.1.2-1.fc32.1 build, which AFAICS fixes the Recommends but doesn't change much else. 32.1.2 seems identical to 32.1.1 upstream, not sure what's going on with that.
@luya please do not change the build in the update again, as I am going to pull this build into RC 4 and we then must keep it at the same version unless we respin again and change it. If we ship RC 4, we must push the exact version of every package it contains stable, so if an update is edited to include a different build in the meantime, that gives us a problem. Thanks!
@zpytela yeah, I saw that too, the interesting question is why that happened I guess; I don't do anything particularly odd to these base images, their state should be a fairly "normal" one. Perhaps a missing dependency or ordering issue or something, somewhere?
So, just a note here: the openQA tests for this initially failed, the test of using GNOME Software to update the system failed as Software got stuck during startup. The journal (see /var/log tarball) showed quite a lot of AVCs.
The base disk image the update tests were using yesterday was quite old (nearly two weeks old, the cutoff for an automatic rebuild), so the update was going from selinux-policy -31.fc32 straight to -35.fc32 , it wasn't updating from -32 (which is currently in stable). This may have had something to do with it.
I regenerated the base disk image manually and ran the test again and it passed, so now the test shows a pass (well, a soft failure on a known bug not to do with this update). But I thought I'd mention the issue just in case anyone wants to take a look at the logs and see if they can see what happened.
the openQA test failures here are spurious, the tests don't cope well with builds that have no x86_64 or noarch packages. This is actually annoyingly difficult to fix, so if you don't mind I'm just going to leave it at least for now :)
FEDORA-2020-8bc645d09f has been created as an alternative to this: it is only 3.24.16 with the specific fixes for the Evolution crash backported, rather than the whole of 3.24.17. Can folks test that instead and check it doesn't produce any of the bugs reported here? Thanks!