08:24 alatiera: hmm, are the gst linux runners registered untagged or not? I haven't seen them pickup untagged jobs and they are fairly idle so I wonder if I forgot to specify run-untagged=true upon registering them or so
08:24 alatiera: need a gitlab admin to check though
08:46 bentiss: alatiera: FWIW, gst-htz-4 is run-untagged=false
08:56 ishitatsuyuki: any documentation on how to run CI container builds locally? need to debug but looks like it's not as straightforward as running a Dockerfile :/
08:57 alatiera: bentiss what about gst-htz 2 and 3
08:58 bentiss: alatiera: same, not allowed to run untagged jobs
08:58 bentiss: alatiera: do you want me to enale them?
08:59 bentiss: enable
08:59 alatiera: the only caveat would be that they run podman unpriv, but otherwise we probably should
09:00 alatiera: what's the status of the other shared runners
09:00 alatiera: wrt priv
09:00 bentiss: they are running privileged, yes
09:01 alatiera: hmm, could make them priv if its a req to make them run untagged jobs to not break anything, could also keep them unpriv and tag more jobs
09:01 alatiera: but yea we should have them be more used ideally :P
09:01 bentiss: alatiera: ideally I'd like to run everybody unpriv, but testing showed that some projects would fail (iirc virgl)
09:02 alatiera: if we have a list we could register 1-2 priv tagged runners and then default the shared ones to unpriv
09:02 bentiss: alatiera: if you have time, maybe you can try forking virgilrenderer, add the gstreamer tag and see if that passes now?
09:02 alatiera: can do yea
09:02 bentiss: great
09:05 kxkamil2: patchwork cannot parse r-b if in body are some > or |
09:05 kxkamil2: for example: > |Dev Name <developer@addr> |
09:06 MrCooper: ishitatsuyuki: for Mesa, see https://docs.mesa3d.org/ci/index.html#building-locally-using-ci-docker-images
09:08 kxkamil2: it stopped parsing body after such line
09:08 ishitatsuyuki: MrCooper: what I want to debug is building the images (one of the patches doesn't apply to deqp after uprev). It seems the images are built in CI using buildah. Would that part of docs still work for this purpose?
09:09 MrCooper: ah, I'm afraid not
09:09 ishitatsuyuki: ok
09:10 MrCooper: images are built using https://gitlab.freedesktop.org/freedesktop/ci-templates
09:10 MrCooper: maybe there's something in the docs there
09:12 bentiss: ishitatsuyuki: internally the tool is using https://gitlab.freedesktop.org/freedesktop/ci-templates/-/blob/master/bootstrap/cbuild, which is a bash script. It's a matter of providing the correct FDO_* environment variables
09:12 ishitatsuyuki: ok, thanks
09:22 alatiera: the build jobs work as expected, but the tests try to mount a tmpfs which fails in unpriv sigh
09:23 alatiera: hmm, wonder if that will work if we switch the driver to fuse-overlayfs
09:25 alatiera: no wait, fuse-overlayfs is the default already, what we would need is the new overlay thingy
10:08 alatiera: I think short term best thing would be to switch/repurpose some of our runners to privileged and tag the projects that need it, and otherwise default to unpriv for shared runners
10:08 alatiera: for the time being I tagged the rest of gstreamer's jobs, so it should ease the load of the other shared runners
10:19 bentiss: alatiera: OK and thanks
10:20 bentiss: alatiera: agree, we really need to rethink how runners are used, but -ENOTIME right now :(
10:20 alatiera: I feel you