<cole-h>
We turned on darwin for everybody, but we only have one darwin machine :P
<gchristensen>
not a great position :(
<hexa->
ah ok
<hexa->
yeah, I didn't look at the arches
<hexa->
silly me
<cole-h>
:)
<gchristensen>
thank you for noticing
<cole-h>
It's probably more pronounced right now because of ZHF and other release things
<hexa->
sure
<hexa->
home-assistant, home-assistant.passthru.tests on x86_64-darwin Queued — Waiting to run this check...
<hexa->
any reason ofborg tries to run those tests, when the package is set platforms = platforms.linux;
<cole-h>
it has to try and evaluate it before it can say "oh, this is unsupported"
<gchristensen>
yeah, it dispatches all the builds and it evals each attr locally and partitions based on the platform
<hexa->
oh, I kinda expected the evaluator to know supported platforms and dispatch from there :)
<MichaelRaskin>
Next you would say centralised evaluation with remote builders based on architecture?
<gchristensen>
yeah, it could. I think it was done this way b/c the `build` command doesn't evaluate and just spews to all archs
<hexa->
no, next I would say: I'm dumb and I don't understand how any of this works!
<hexa->
right, makes sense
<gchristensen>
oh the other thing is the naive case sends the build instruction to a "builds" exchange and the architectures make a queue thath subscribes to the builds exchange
<MichaelRaskin>
Yeah, and deployment model of builders has not been exactly constant over the lifetime of OfBorg
<gchristensen>
yea
<gchristensen>
it wasn't until mac / ACLs was added that I think the "filtered" distribution became a thing
<gchristensen>
cole-h: can you send a PR making the new disable_... field for ACLs optional?
<cole-h>
Is it not? Pretty sure it should deserialize with a default if it's not present
<gchristensen>
May 18 16:37:25 aarch64.nixos.community grahamcofborg-builder-1-start[10419]: thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: Error("missing field `disable_trusted_users`", line: 7, column: 5)', src/config.rs:163:32
<andi->
Or maybe nixpkgs is just slowly taking over the world
<gchristensen>
it'd be useful if ofborg published more metrics like "builds started" and "builds finished" but the current design of metrics is a tire fire
<gchristensen>
we can assume they're all network accessible now on port 9xxx
<andi->
The ofBorg CI currently sets nix_path for the cachix installer, is that a requirement? I thought all the deps were actually pinned in the nixpkgs.json. I am starting to guess what the difference between my local setup and CI is :(
orivej has joined #nixos-borg
orivej has quit [Ping timeout: 260 seconds]
orivej_ has joined #nixos-borg
<gchristensen>
should not be a requirement
supersandro2000 is now known as Guest2135
Guest2135 has quit [Killed (orwell.freenode.net (Nickname regained by services))]
supersandro2000 has joined #nixos-borg
<andi->
Could one of you run nix-shell --pure --run checkPhase on my nixpkgs bump PR locally? This is a fascinating rust compiler bug...
<cole-h>
works for me
<cole-h>
I mean, tests fail because I have experimental features enabled but
<andi->
grml, if it would fail that would be much better :D