| 6 Jan 2025 |
 zowoq | Considering the budgets of the orgs involved I think it would make more sense for the nixos org to pay for an aarch64-linux ofborg builder that can also be used as a community builder. | 03:31:28 |
 emily | it sounded like infra team didn't want to have to deal with it, but agreed that the foundation paying for it makes sense | 03:36:28 |
| emily | I think nix-community could open a ticket on the foundation GitHub for that? (i forget the exact process) | 03:36:53 |
| zowoq | And I don't want to deal with ofborg. | 03:39:08 |
| emily | maybe we can get ofborg team to admin it š
| 03:39:53 |
 GaƩtan Lepage | In reply to @zowoq:matrix.org
We're still short of the target but Jonas Chevalier is going to cover the difference for now.
New machine is up: aarch64-build-box.nix-community.org
I gave 1kā¬ yesterday, I hope it helps a bit. For now at least. | 06:42:44 |
| zowoq | Helps a lot, thank you. | 06:47:45 |
| GaƩtan Lepage | Thanks a lot to you for maintaining all this infra :) | 07:09:54 |
 Perchun Pak [don't ping; dm instead] | can someone update the description to point to the new box please?
https://github.com/NixOS/aarch64-build-box | 11:57:36 |
 liberodark | Hi | 13:05:12 |
| liberodark | The builder will be available this weekend if you are still interested.
I could put it up with a dedicated ipv4 and v6. | 13:05:13 |
| liberodark | Best Regards | 13:05:15 |
| 7 Jan 2025 |
|  ruro joined the room. | 03:42:19 |
|  @siraben:matrix.org left the room. | 03:42:31 |
|  Renato Trevisan set a profile picture. | 21:58:29 |
 hexa | GaƩtan Lepage: load1 at 390 | 23:55:12 |
| hexa | cmon | 23:55:13 |
| hexa | and oom as well | 23:55:26 |
| hexa |
23:56:08 up 1 day 11:47, 2 users, load average: 351.43, 344.19, 207.12
| 23:56:13 |
| 8 Jan 2025 |
| GaƩtan Lepage | On ARM ? | 06:53:53 |
| GaƩtan Lepage | I'm fighting against a single package (pytensor) which is building perfectly fine on x86 but for some reason filling RAM + swap on the aarch box | 06:55:07 |
| GaƩtan Lepage | That's why I was asking for overwritting pytest's num jobs. | 06:55:54 |
| GaƩtan Lepage | I have been able to cache it for now by manually ssh-ing on the box and building it with --cores 16 but it sometimes gets rebuilt by nixpkgs-review. | 06:59:31 |
| GaƩtan Lepage | Or are you talking about darwin ? | 07:01:56 |
| hexa | aarch64-linux | 13:46:11 |
| hexa | needed to get a firefox out and twice my build job was killed for no good reason | 13:46:39 |
| GaƩtan Lepage | Ok, I think it's numpyro then | 13:47:52 |
| GaƩtan Lepage | I open a PR to prevent it from bothering us again: https://github.com/NixOS/nixpkgs/pull/372111 | 13:48:11 |
| emily | that looks like it will use 8 cores even if --cores 1? | 13:53:24 |
| GaƩtan Lepage | I was wondering if we could do something like min(1, NIX_BUILD_CORES) but I fear it's not possible.
Do you have a better solution emily ? | 13:57:29 |
