ChanServ changed the topic of #freedesktop to: https://www.freedesktop.org infrastructure and online services || for questions about freedesktop.org projects, please see each project's contact || for discussions about specifications, please use https://gitlab.freedesktop.org/xdg or xdg@lists.freedesktop.org
shashanks_ has joined #freedesktop
shashanks has quit [Ping timeout: 480 seconds]
MrBonkers has quit [Quit: Ping timeout (120 seconds)]
MrBonkers has joined #freedesktop
m5zs7k_ has quit []
m5zs7k has joined #freedesktop
scrumplex has joined #freedesktop
scrumplex_ has quit [Ping timeout: 480 seconds]
ximion has quit [Remote host closed the connection]
haaninjo has quit [Quit: Ex-Chat]
guludo has quit [Ping timeout: 480 seconds]
sauce has quit [Ping timeout: 480 seconds]
sauce has joined #freedesktop
nephyrin_ has joined #freedesktop
nephyrin has quit [Ping timeout: 480 seconds]
Kayden has quit [Ping timeout: 480 seconds]
alanc has quit [Remote host closed the connection]
alanc has joined #freedesktop
Kayden has joined #freedesktop
georgc has quit [Quit: Leaving]
gchini has joined #freedesktop
sima has joined #freedesktop
jsa1 has joined #freedesktop
jsa1 has quit [Remote host closed the connection]
tzimmermann has joined #freedesktop
sghuge has quit [Remote host closed the connection]
sghuge has joined #freedesktop
AbleBacon has quit [Read error: Connection reset by peer]
ity1 has joined #freedesktop
ity has quit [Ping timeout: 480 seconds]
guludo has joined #freedesktop
julian1 has joined #freedesktop
julian1 has quit [Quit: WeeChat 4.5.2]
haaninjo has joined #freedesktop
anddam has joined #freedesktop
<DragoonAethis>
bentiss: re: igt-gpu-tools tags, nuke them all from the orbit
<DragoonAethis>
I tried and it didn't work from the GitLab UI
guludo has quit [Quit: WeeChat 4.5.2]
guludo has joined #freedesktop
<bentiss>
DragoonAethis: shouldn't I keep a couple of them?
<DragoonAethis>
bentiss: we don't use them for anything outside of the built-in GitLab CI tests, not sure if anybody uses that
<DragoonAethis>
IMO fine to nuke it all
<bentiss>
the dockerfile-* are from 2021
scrumplex has quit [Quit: Quassel - Signing Off]
<mupuf>
bentiss, DragoonAethis: I'm using them occasionally, but all I need is to keep HEAD. Don't we have an expiration policy for tags in the project?
<DragoonAethis>
mupuf: don't know, I don't even have admin perms on that project ._.
<DragoonAethis>
All this stuff is from before my time
<mupuf>
bentiss: there is a bit of a misunderstanding here. The containers being rebuilt are basically just doing the compilation IIRC. The base containers don't get rebuilt every time
<mupuf>
Agreed, and that's because there are hundreds of 5 MB containers
<mupuf>
and that is ridiculous
<bentiss>
yep
<bentiss>
if you don't need to reuse the container, why keepingit in the registry, just install the stuff and run the tests
<mupuf>
the point is to do the build once then reuse it in other parts of the pipeline. Rather than using artifacts, containers are used because that's the logical thing to do and ultimately ends up exactly in the same spot: the s3 server
<bentiss>
except no. It's a different S3 server with different rules
<mupuf>
the difference here is that artifacts have a default expiration of 30 days which is not something done on the repo
<bentiss>
and deleting things on the container registry is harder (both technically and ethically) than on the artifact server
<bentiss>
mupuf: also the "containers:igt" job is run at the end of the pipeline, and never used, so what's this is about?
<mupuf>
looking into the pipeline again, it's been quite some time since I looked at it
<mupuf>
but on the topic of ethicality of removing container images from registries, I don't see the difference with losing artifacts. I however see a big difference in how to reproduce a test environment: a single podman cmdline vs downloading the build artifacts, extracting, then adding a -v in your command line
<mupuf>
IIRC, this is exactly why the container was built. Not for CI (yet), but to allow developers to re-use the test environment on their dev machine.
<mupuf>
but yeah, to be honest, I think the whole pipeline needs to be re-designed
<mupuf>
Intel dropped the ball hard on moving IGT from the mailing list to being a gitlab project, and whole thing remains in limbo
<bentiss>
that Dockerfile takes $CI_REGISTRY_IMAGE/build-fedora:commit-$CI_COMMIT_SHA, shove in it installdir/opt/igt /opt/igt and .gitlab-ci/docker-help.sh /usr/local/bin/docker-help.sh and done, this is just a curl/git pull away
<mupuf>
you mean that the container itself could be doing it at boot time?
<bentiss>
yes, or you can have a script that does that and people use it when they need
<bentiss>
you don't need to store the content of git in a container
<mupuf>
amen to that
<mupuf>
I think we can all agree that the current state of affair is just dumb and wasteful
<DragoonAethis>
Yup
<bentiss>
where is that installdir coming from???
<bentiss>
build:tests-fedora apparently
<mupuf>
yep, I was about to say
<bentiss>
ok, well, yeah, could be just a script away
scrumplex has joined #freedesktop
<mupuf>
DragoonAethis: mind disabling the job?
<DragoonAethis>
mupuf: I don't have perms on this repo at all
<DragoonAethis>
Well, I have developer perms, so I can merge stuff, but not look at settings
<mupuf>
I can give you that, but that's not how you disable a job
<mupuf>
just add a . in from of `containers:igt`
<mupuf>
in .gitlab-ci.yml
<bentiss>
the thing is if I delete the commit-* tags, skopeo will delete all of them, including the original dockerfile-<sha256>, because they all share the same digest, and you delete a manifest by its digest
<mupuf>
but to be honest, the whole thing needs fixing :s While the base containers are not rebuilt every time, they still push commit- tags for no reason
<bentiss>
so I'll probably not touch anything in the build-* repos, except for the igt one
<mupuf>
agreed. Let's first fix the problem then deleting all the commit-* ones
<mupuf>
welp, I guess I'll have to volunteer myself to rewrite this mess
<bentiss>
\o/
<DragoonAethis>
yay for mupuf <3
lsd|2 has joined #freedesktop
<mupuf>
why do I still freaking care...
<bentiss>
IIRC we had in a past a ci-templates user bassed on the sha256 of a file in the repo
<bentiss>
this is pretty much what they need here
<mupuf>
pretty much, yeah
<mupuf>
and daniels is somewhat moving to that too in Mesa :D
<sima>
mupuf, my apologies, I've hacked it all together before ci-templates and then I guess it never moved further :-/
<sima>
or at least not much
<mupuf>
sima: no worries, I had my hands in it too
<mupuf>
but you know what's the worst thing here? I will have to send freaking emails to fix this shit and I am not pleased with it
<DragoonAethis>
mupuf: why, don't you enjoy just sending your work straight into /dev/null without kicking people for 2 weeks straight?
<mupuf>
well, that's not much different from sending an MR on gitlab ;) The difference I'll need to set up git send-email again.
<bentiss>
FWIW, I'm currently deleting all manifests under drm/igt-gpu-tools/igt/ older than a day
<bentiss>
(~2800)
<mupuf>
bentiss: sounds like the best course of action, yeah
<mupuf>
and I could enable a similar rule in the settings
<bentiss>
FWIW, the build-* are just taking space in the db for nothing, but not in storage, so we can leave this like that
jsa1 has joined #freedesktop
<mupuf>
sima: what's the plan for the IGT and DRM migration to GitLab?