ChanServ changed the topic of #freedesktop to:
greeb has quit [Quit: bye bye]
___nick___ has quit []
___nick___ has joined #freedesktop
ximion1 has joined #freedesktop
ximion has quit [Ping timeout: 480 seconds]
ximion1 has quit []
pendingchaos_ has joined #freedesktop
pendingchaos has quit [Ping timeout: 480 seconds]
<thaller> bentiss, the selected container is "fedora:35", which on my machine resolves to registry.fedoraproject.org/fedora:35. Here (https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/jobs/13868823/raw) we see:
<thaller> +++ buildah from --isolation=chroot fedora:35
<thaller> Copying blob sha256:fc972fbab1a5944103b867e535959227bbc28e43e19827d51668d3169547a4ba
<thaller> Getting image source signatures
<thaller> Copying config sha256:b2f5884f8581640b952e750e6673c8ea71a97e709c567a8aaed6648753e1c111
<thaller> where fc972fbab1a5944103b867e535959227bbc28e43e19827d51668d3169547a4ba seems to be an image of Rawhide from July.
<thaller> is the a ci-templates project that uses f35? I clicked through the list of projects on the website (mesa,libinput,weston), and it seems that is nowhere else used (yet)...
<thaller> I would guess, other projects will run into the same issue, once they bump the fedora version to 35...
<bentiss> thaller: your machine is set to first look at registry.fedora.org instead of dockerhub, which the fdo runners are not (they are debian based)
<bentiss> thaller: and the upstream tip of ci-templates forces the use of registry.fedora.org, so it's worth a shot upgrading the sha of ci-templates
DasChaos[m] has left #freedesktop [#freedesktop]
<thaller> ok, I tested to add F35 container on libinput. That doesn't have the problem: https://gitlab.freedesktop.org/thaller/libinput/-/pipelines/406312 ... so it's an NM problem
<thaller> bentiss, ok, will try to update ci-templates...
danvet has joined #freedesktop
<bentiss> thaller: one option could be to hit "clear runner cache" on the CI/CD -> pipelines page
<bentiss> because libinput is not enforcing the registry on fedora but pulls a later image
<bentiss> I wonder if we should not force a podman clear cache before pulling the image
<thaller> bentiss, updating ci-templates fixes f35/f36, but breaks centos7, centos8, f28, f29... https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/pipelines/406334
<thaller> are these no longer supported?
<bentiss> thaller: they should be supported, though I see you are using specific tags...
<bentiss> let me check if the centos registry has them or something similar
<thaller> bootstrap/cbuild does:
<thaller> centos)
<thaller> if [[ $version == '7' ]]
<thaller> DNF_EXTRA_ARGS=""
<thaller> DNF="yum"
<thaller> then
<thaller> else
<thaller> but version seems to be "7.9.2009"
<thaller> that explains the failure on Centos7...
<bentiss> thaller: looks like registry.centos.org is dead actually, it hasn't been updated since Dec 2020
<bentiss> "skopeo inspect docker://registry.centos.org/centos:8" gives a handful of tags
<bentiss> when "skopeo inspect docker://quay.io/centos/centos:8" gives all the tags you want
<thaller> bentiss, isn't hte problem:
<thaller> + buildah run --isolation chroot dnf upgrade -y --setopt=install_weak_deps=False
<thaller> error reading build container "dnf": container not known
<thaller> (at least for Centos7...)
<bentiss> thaller: you're having 2 issues actually
<bentiss> the first one is what I described: "Error initializing source docker://registry.centos.org/centos:7.5.1804: Error reading manifest 7.5.1804 in registry.centos.org/centos: manifest unknown: manifest unknown"
<bentiss> which means that buildah can not create the container, and $buildcntr is empty
<bentiss> so then, when you run "buildah run" it thinks the container is named dnf
<bentiss> the second problem is what you described, we should be using yum, and not dnf
<thaller> (note that building CentOS / f2[89] with the earlier ci-templates version still works)
<thaller> for the second problem, the fix is easy. I will create a merge request now.
<bentiss> thaller:on it
<thaller> bentiss, thank you!!!!
<bentiss> thaller: there is a chance centos7 has an alias for dnf too
<bentiss> because I don't think we had something different for centos 7
<thaller> bentiss, I think no such alias exists, and it also wouldn't support "-setopt=install_weak_dep" ...
* bentiss wonderswhat magic trick we used before
<bentiss> thaller: yep, saw it just now :)
<thaller> tested with: version=7; [[ "$version" == [567]* || "$version" == centos[567]* ]] && echo ja
<thaller> oh my. Spelling error in the subject line. Let me fix that :)
<bentiss> thaller: you'll need to run make for the sanity check to pass, and ideally if you could rebase on top of !121 that would be nice
<thaller> ah. Sorry... will do.
<bentiss> cause you'll need both
<thaller> done.
<bentiss> thaller: thanks.
<bentiss> now we just need to wait for the CI to pass
<thaller> bentiss, about f28/f29, I guess those containers are no longer at docker://quay.io/fedora/fedora . Does that mean, I need to drop them?
<thaller> which would be possible... at some point the need to be dropped.
<bentiss> thaller: Fedora 28 officially enters End Of Life (EOL) status effective May 28, 2019 -> I would drop them
<thaller> right. The point would be to build in an ancient environment... only to see that we can. The practical use may be limited.
<bentiss> and 29 is also almost 2 years EOL state
<bentiss> I doubt people who are still on f29 will use an upstream NM
<bentiss> and for the ancient environment, you have ubuntu 16.04
<bentiss> :)
<thaller> sure. It's more about the glib versions that are shipped there. Officially we never bumped the required version (yet).
<thaller> will drop them.
<bentiss> thanks
<bentiss> thaller: BTW, not sure if this is your target, but we now support full freebsd through qemu on ci-templates
<bentiss> don't know if NM supports it, but if so, it could be something interesting to add to the CI
<thaller> bentiss, NM is Linux-only. It would be very nice to change that, but a huge amount of work, so not gonna happen...
<bentiss> thaller: ok, no worries
<thaller> thank you!!
<bentiss> thaller: cool, and no worries :)
pendingchaos_ is now known as pendingchaos
ximion has joined #freedesktop
rbrune has quit [Quit: Leaving]
ximion has quit []
rbrune has joined #freedesktop
seabass has joined #freedesktop
seabass has quit []
seabass has joined #freedesktop
seabass[m] has left #freedesktop [#freedesktop]
zhxuxu has quit [Remote host closed the connection]
<seabass> __tim: Thanks! For those who might also want to use Emacs/Forge with the freedesktop GitLab, here's the code I put in ~/.emacs to make it work:
<seabass> (require 'forge)
<seabass> (setq forge-alist (append forge-alist(list '("gitlab.freedesktop.org" "gitlab.freedesktop.org/api/v4" "gitlab.freedesktop.org" forge-gitlab-repository))))
<seabass> And, after that, you add a line to ~/.authinfo like:
<seabass> machine gitlab.freedesktop.org/api/v4 login seabass^forge password M9VBuaGoLE7Jupx4
<seabass> where you replace seabass with your own GitLab username and M9VBuaGoLE7Jupx4 with your API token
<seabass> Works really nicely now! :)
thaytan has quit [Ping timeout: 480 seconds]
alanc has quit [Remote host closed the connection]
alanc has joined #freedesktop
thaytan has joined #freedesktop
<pq> https://gitlab.freedesktop.org/wayland/weston/-/jobs/13896111 is telling me: ERROR: Preparation failed: failed to pull image "registry.freedesktop.org/wayland/weston/debian/armv7:2021-09-14.2" with specified policies [always]: error pulling image configuration: unexpected EOF (manager.go:205:1s)
<pq> even after a couple retries
<MrCooper> bentiss daniels: ^ docker kicking time?
<daniels> yeah, just done
<MrCooper> ta
khilman has joined #freedesktop
Haaninjo has joined #freedesktop
<bentiss> daniels: BTW, I was wondering if we should not give anholt_ access to the cluster so we get somebody else capable of running this one line when we are asleep
<bentiss> plus he'll be able to check on marge
<bentiss> anholt_: would that be of interest for you? kicking the registry cache is a one command to type
<daniels> bentiss: I think he already had access to GCP, would be totally fine with new cluster too
<bentiss> daniels: OK, once I'll get his answer, I'll set up the credentials if needed
khilman has left #freedesktop [#freedesktop]
ngcortes has joined #freedesktop
<anholt_> bentiss: I do have some gcp access, but I would need a bit a playbook to work from, as I don't see myself developing my own understanding of the gitlab setup.
<bentiss> anholt_: the idea is to give you wireguard access to the cluster, the credentials, and the command to kick the registry when it goes down
<bentiss> no development requiredd
<bentiss> and FWIW, we don have anything left on gcp :)
* bentiss <- diner, I'll set this up after
<bentiss> dinner actually, not the fancy thing :(
<anholt_> oh, right
<anholt_> nothing on gcp any more anyway
<anholt_> bentiss: https://gitlab.freedesktop.org/mesa/mesa/-/jobs/13905021 registry pulls failing again
<bentiss> anholt_: good timing then :)
<bentiss> anholt_: can I use one of the ssh keys on gitlab to temporary add you on one of the servers?
<bentiss> anholt_: or you can give me a public ssh key, that will do too
<bentiss> (actually I don't have the full pub key on gitlab)
<bentiss> daniels: OK, I gave anholt the keys
<anholt_> I promise to try to do the minimum and only what bentiss has told me how to, in all of our interests.
<bentiss> having someone who can just kick the registry cache when we are asleep is already tremendensly helpful
<daniels> it is very strange how peaky it is - sometimes it fails multiple times per day, sometimes it's fine for weeks on the bounce ...
<daniels> but the logs show no particular insight when it is flaky
Haaninjo has quit [Quit: Ex-Chat]
<DragoonAethis> Hey everyone, I just finished publishing all videos from XDC 2021: https://www.youtube.com/playlist?list=PLe6I3NKr-I4LwrhsAG1XVPW0EFL8ZJ_6W - some titles were cut to make them fit into the YouTube title length limit, so if you'd like to make any changes, please let me know. Enjoy!
<DragoonAethis> (media.ccc.de videos are coming in a few days)
<emersion> thanks!
ximion has joined #freedesktop
seabass has quit [Quit: leaving]
danvet has quit [Ping timeout: 480 seconds]
Seirdy has quit [Quit: exiting 3.2]
Seirdy has joined #freedesktop
xexaxo has quit [Read error: Connection reset by peer]
xexaxo has joined #freedesktop
ngcortes has quit [Remote host closed the connection]