Comments

1137 Comments

This is failing the openQA 'build a network installer then install from it' test. Building the image works, booting it doesn't. It seems to be some kind of space issue.

AFAICS the packages in this update are mostly a bit bigger than they used to be - none of them is a lot bigger, they're all just a little bit bigger.

Looking into it it seems like the root filesystem in F30 installer images when we build them these days is actually almost full. I downloaded the image from another recent passed F30 test and booted it locally, and df shows 860K free on the root filesystem. I also downloaded the image from this update and the image from an F31 update to compare. The image from this update shows 0 free space, the F31 image shows over 100M.

So, I dunno when or why the root filesystem in F30 installer images got so big (or why F31 ones didn't), but it seems like it's balanced right on a knife edge and the small size changes in this update are just enough to push it over the edge into not working. I don't think it makes sense to call it a bug in this update, so for now I'll just tell openQA to use a slightly larger root FS size for F30 updates and re-run the tests and they should pass. If I get time I'll try and figure out why the root fs when we build an F30 image right now is bigger than for F31...

@pbrezina updates-testing is enabled on F32 installs by default, but it's still required for pending updates to follow the policy and work with what's current in 'stable' (i.e. not u-t). I'll re-run the tests for this now. thanks!

Yeah, I was working on the same theory but our main test runner crashed before I could confirm it last night. I've added the F32 pki-base update to the 'workarounds' list for now and re-run the affected test for all affected updates, including this one, it's running or queued now.

but I didn't write or submit or even test the latest fix yet :P do you trust sgallagh and kparal? that's the question...

karma

openQA tests look good now.

BTW, the tests when the update was edited because of this bodhi bug :(. I will re-trigger manually.

@plautrba for the record, yes, it is definitely this update. The openQA tests isolate the packages from a specific update, they do not test all of updates-testing. And the same test is passing on other 32 updates, it only failed on this update. Those facts combined mean this update definitely causes the failure.

This went in Beta, so we should push it stable at this point.

karma

we put this in the Beta compose request, so we should push it stable at this point.

BZ#1812579 Unretired ots

@lnie verified the fix in Bugzilla.

This should have been edited into the nss update, not submitted separately. Interdependent packages are meant to go together in the same update, or go in series. nss has not yet been pushed stable due to the Beta freeze.

In practice it's likely we won't have issues here (the updates will be pushed together or nss will be pushed first), but please do things properly. This is why the openQA tests fail, it is catching that this update does not install without the nss update.

karma

Breaks Cockpit, same as the F32 update.

karma

This seems to break Cockpit. The Cockpit login screen just shows "Internal error in login process". System journal shows several AVCs, these seem most likely related:

Mar 11 13:45:41 ibm-p8-kvm-03-guest-02.virt.pnr.lab.eng.rdu2.redhat.com audit[1044]: AVC avc:  denied  { setsched } for  pid=1044 comm="cockpit-ws" scontext=system_u:system_r:cockpit_ws_t:s0 tcontext=system_u:system_r:cockpit_ws_t:s0 tclass=process permissive=0
Mar 11 13:46:23 ibm-p8-kvm-03-guest-02.virt.pnr.lab.eng.rdu2.redhat.com audit[1170]: AVC avc:  denied  { create } for  pid=1170 comm="cockpit-session" scontext=system_u:system_r:cockpit_session_t:s0 tcontext=system_u:system_r:cockpit_session_t:s0 tclass=netlink_selinux_socket permissive=0
Mar 11 13:46:23 ibm-p8-kvm-03-guest-02.virt.pnr.lab.eng.rdu2.redhat.com cockpit-ws[1170]: avc:  can't open netlink socket: 13 (Permission denied)
Mar 11 13:46:23 ibm-p8-kvm-03-guest-02.virt.pnr.lab.eng.rdu2.redhat.com cockpit-ws[1170]: cockpit-session: avc.c:74: avc_context_to_sid_raw: Assertion `avc_running' failed.

Doesn't seem to break anything in basic openQA tests at least.

@vontonkin ibus stuff should be good with the updated gnome-shell and ibus builds I edited in, can you please test? thanks!

Seems to be working in normal use on my desktop too, and passed x86_64 openQA tests.

BZ#1807661 Display corruption on aarch64 virtual machines

openQA upgrade tests are succeeding with this, so gonna +1 it.

openQA upgrade tests are succeeding with this, so gonna +1 it.

I'm afraid I don't, I'm just the test monkey :) ab might.