| 20 Oct 2021 |
 grahamc (he/him) | :joy: | 19:12:20 |
 das_j | you could do it in haskell for no reason :D | 19:13:17 |
| grahamc (he/him) | I could see justifiable reasons | 19:13:28 |
 tomberek | I've used the "remote-exec" as a deployment alternative before. With lifecycle options it's effective even as a continuous provisioning. Pointing to hydra output via Terraform is nice. | 19:14:01 |
| grahamc (he/him) | like if you're putting a lot of work in to observability around it | 19:14:03 |
| grahamc (he/him) | https://github.com/DeterminateSystems/terraform-provider-hydra/issues/72 | 19:15:23 |
 cransom | In reply to @grahamc:nixos.org
data "hydra_job" "system" {
project = "app"
jobset = "production"
job = "webserver"
}
resource "aws_instance" "web" {
# ...
provisioner "remote-exec" {
inline = [
"nix-env --profile /nix/var/nix/profiles/system --set '${data.hydra_job.system.outputs.out.path}'",
"/nix/var/nix/profiles/system/bin/switch-to-configuration switch",
]
}
}
that is smooth. we do the ami baking, but the next speed hack, i was contemplating was for the deploy to upload a user-data for cloud-init that does a nix-store -r <system path>; <system path>/bin/switch-to-configuration switch and then either launch or reboot machines. | 20:17:32 |
|  hpfr joined the room. | 21:12:52 |
| hpfr | maybe this is a better place to ask?
https://nixos.org/channels/nixos-unstable is behind https://status.nixos.org, which I didn't even think was possible. shouldn't the status page wait until the channel itself actually updates?
| 21:13:21 |
| hpfr | nixos-unstable still has not advanced... | 23:42:56 |
 Ryan Burns | maybe related https://github.com/NixOS/nixos-org-configurations/issues/169 | 23:46:27 |
| 21 Oct 2021 |
| hpfr | thanks, good to know it is recent | 00:36:30 |
| tomberek | what's the "merge these derivations with symlinks, with some default ordering imposed" function? | 00:58:10 |
| tomberek | buildEnv... maybe | 00:59:44 |
 lukegb (he/him) | Caught exception in Hydra::Controller::Build->restart "Can't locate object method "search" via package "Hydra::Model::DB::Builds" at /nix/store/4yh4dda4w8f4w7gr5mvvc576b157nq50-hydra-0.1.20211020.30e5001/libexec/hydra/lib/Hydra/Helper/Nix.pm line 511." | 06:29:45 |
| lukegb (he/him) | wondering if removing the prototype from restartBuilds changed the way that args are being passed to it... | 06:40:33 |
| lukegb (he/him) | in particular, I think it's flipped it to being evaluated in a list context rather than a scalar one sigh | 06:42:13 |
|  phaer joined the room. | 12:52:11 |
| grahamc (he/him) | I didn't think prototypes actually changed semantics, but just sometimes introduced compiler errors if you broke it | 13:17:40 |
| grahamc (he/him) | what page did you get that error? | 13:17:49 |
| grahamc (he/him) | oh that is actually pretty obvious from the message :) | 13:19:22 |
| grahamc (he/him) | * oh that is actually pretty obvious from the message :) | 13:19:24 |
| grahamc (he/him) | I just tried it locally and it worked fine ... hmm | 13:21:23 |
| grahamc (he/him) | ah this is restarting a specific build. | 13:26:31 |
| grahamc (he/him) | it is definitely getting a ResultSet | 13:28:38 |
| grahamc (he/him) | hrm! | 13:32:32 |
| 22 Oct 2021 |
| lukegb (he/him) | I think that prototypes do change semantics? because iirc function arguments like to induce list context rather than scalar context | 00:53:05 |
| lukegb (he/him) | so having a ($$) prototype would force scalar context evaluation but without you might end up in list context | 00:53:24 |
| grahamc (he/him) | so I ended up writing a fix for it but fell short on writing a test for it, and I really don't like sending PRs without tests | 15:29:19 |
| grahamc (he/him) | but I should just buck up and do the test or PR | 15:29:37 |
