| 20 Mar 2026 |
|  irrelevancy changed their display name from irrelevancyisthygoal One Confused Being to irrelevancy. | 06:46:39 |
| 21 Mar 2026 |
 Vladimír Čunát | maximum-snail.mac.nixos.org was stuck, all job slots in "sending inputs" for several hours. Rebooted now, seems to be building again. | 05:55:09 |
 hexa | that means it ran out of disk space | 09:23:41 |
| Vladimír Čunát | As it tends to happen recently, what's the best action when I see it again? (assuming we can't easily avoid it) | 09:30:21 |
| hexa | reboot the builder or restart the queue-runner | 09:34:32 |
| hexa | I chose the former because it is less disruptive | 09:34:40 |
| Vladimír Čunát | I want to have long uninterrupted queue-runner periods. | 09:35:59 |
| Vladimír Čunát | Big builds suffer, especially chromium and derivatives on aarch64-linux. The aarch64-linux:big-parallel queue doesn't manage to catch up too well lately. | 09:36:49 |
| Vladimír Čunát | (a chromium often runs for half a day in there) | 09:37:23 |
| Vladimír Čunát | * (a chromium build often runs for half a day in there) | 09:37:30 |
| Vladimír Čunát | * (a single chromium build often runs for half a day in there) | 09:37:48 |
| hexa | yeah, so another machine would be useful | 09:41:20 |
| Vladimír Čunát | A dedicated machine for all the chromiums and electrons 😆 | 09:49:36 |
|  Mindavi joined the room. | 17:26:41 |
| 24 Mar 2026 |
 leona | something seems to be stuck with these 4 new tests: https://hydra.nixos.org/eval/1823783#tabs-unfinished. They don't get build for >1 day, even when they are at the moved to the front of the queue | 09:19:11 |
| leona | * something seems to be stuck with these 4 new tests: https://hydra.nixos.org/eval/1823783#tabs-unfinished. They don't get build for ~2 days, even when they are at the moved to the front of the queue | 09:20:28 |
| Vladimír Čunát | I've noticed, but I don't know what's wrong in there. | 09:29:36 |
| Vladimír Čunát | A couple days ago I just aborted them to let the channel advance and then restarted+bumped them. | 09:30:21 |
| Vladimír Čunát | Wait... I think I see why. | 09:31:18 |
| Vladimír Čunát | # Containers use systemd-nspawn, which requires pid 0 inside of the sandbox.
| 09:33:10 |
| Vladimír Čunát | which adds uid-range into requiredSystemFeatures | 09:33:34 |
| Vladimír Čunát | PR https://github.com/NixOS/nixpkgs/pull/478109 | 09:33:56 |
 Jeremy Fleischman (jfly) | @vcunat:matrix.org, how does https://github.com/NixOS/infra/pull/986 look? | 11:05:36 |
| hexa | diff --git a/builders/common/nix.nix b/builders/common/nix.nix
index ce13643..79249e8 100644
--- a/builders/common/nix.nix
+++ b/builders/common/nix.nix
@@ -22,8 +22,11 @@
settings = {
accept-flake-config = false;
+ auto-allocate-uids = true;
builders-use-substitutes = true;
extra-experimental-features = [
+ "auto-allocate-uids"
+ "cgroups"
"nix-command"
"no-url-literals"
"flakes"
@@ -32,6 +35,7 @@
"kvm"
"nixos-test"
"benchmark" # we may restrict this in the central /etc/nix/machines anyway
+ "uid-range"
];
trusted-users = [
"build"
| 11:06:17 |
| hexa | same changes that I did, but inline | 11:06:36 |
| hexa | still, I'm wary of cgroups in cppnix | 11:06:47 |
| Jeremy Fleischman (jfly) | Or that :) | 11:06:47 |
| hexa | elated-minsky exposes uid-range feature now | 11:07:03 |
| Jeremy Fleischman (jfly) | I thought you said we used them, but I couldn't find evidence of that | 11:07:26 |
| Jeremy Fleischman (jfly) | * | 11:07:45 |
