00:36 <DavidHeidelberg[m]> so, yeah, the libdrm should be in /install; but that's not the critical thing. It's in /usr/local; but ... sometimes deqp pick the one in /usr/lib... so I just ugly removed it and with bookworm it'll be clean implementation.

00:38 <DavidHeidelberg[m]> also we should be removing the old installed mesa, but that's is cared for by LD_LIBRARY_PATH; only disadvantage is that it still takes some MiB in images

02:26 <Wallbraker> daniels: Oh? How does that work?

12:18 <DavidHeidelberg[m]> bentiss: can you put me into group of users allowed to run CI, please? :)

12:30 <bentiss> DavidHeidelberg[m]: you should bebalready. sorry on the phone right now and far from a laptop atm

12:31 <bentiss> DavidHeidelberg[m]: if it's n9t working get yourself added to mesa/ci-ok

12:32 <DavidHeidelberg[m]> np, it stopped working like yesterday, before it was fine

12:32 <bentiss> weird

12:32 <bentiss> i haven't touched it

12:32 <DavidHeidelberg[m]> requested access to mesa/ci-ok, thanks bentiss

12:33 <bentiss> worse case, the token used for the runner gating expired. but i don't think i put an expiration date

12:34 <bentiss> i will be able tonchexk after the hike tonight

12:36 <bentiss> well my wife told me it's more of a walk than a hike but the pointbis no internet soonish

12:36 <Wallbraker> Sometime Thur-Fri-Sat I was no longer able to run CI pipelines in my personal repo for Monado project even tho I'm in the "monado" group, which I hope has the same rules as "mesa" in terms of Ci-Ok ness.

12:36 <Wallbraker> So probably something changed.

12:39 <DavidHeidelberg[m]> bentiss: enjoy :D

13:05 <__tim> I've also seen that the user-in-group check stopped working with various cairo MRs

13:06 <__tim> like within the last two days or so

16:42 <bentiss> It should be fixed now. For reference, I have a global (instance) variable with the guest token from the main ci-ok group

16:42 <bentiss> and it expired because I probably did not clear the expiration field

16:48 <bentiss> Alright, added a note on the script itself in case I am not around if this happens again due to a mistake in the group handling: https://gitlab.freedesktop.org/freedesktop/helm-gitlab-infra/-/commit/ed7b18567a28df2a2c93eb3f23bff6f8ac8568a8

16:49 <__tim> thanks :)

19:31 <Wallbraker> Worked for me as well, thanks.

19:32 <Wallbraker> Is there a way we can replicate that user check in gitlab CI files as well? Currently we have rules in the gitlab files not run on branches outside of the main project. But for users in the CI-Ok group/user, we would still like to run it. So we don't have to open a MR for every experiment we do?

19:40 <__tim> the reason for these rules is that CI resources are not free and we usually don't want to run full pipelines for every push to a wip branch (but users can trigger the trigger if needed if it's behind a trigger)

19:41 <__tim> (generally speaking, I don't know your specific project/situation)