Post AWcDFbyHV83MqUmxwe by mic92@toot.matereal.eu
(DIR) More posts by mic92@toot.matereal.eu
(DIR) Post #AWcDFbyHV83MqUmxwe by mic92@toot.matereal.eu
2023-06-12T12:17:25Z
1 likes, 0 repeats
We are at this point where you need a 10GB swap file to evaluate nixpkgs for nix-index-database on github runners who are equipped with 8GB RAM. Maybe it's time to stop adding things to nixpkgs until those performance issues are under control.
(DIR) Post #AWdrjfemYNqAHvW4nY by sandro@c3d2.social
2023-06-12T17:45:15Z
1 likes, 0 repeats
@mic92 maybe we should generate a flame graph and see what functions are the worst offenders. Maybe we can find some low hanging fruits.
(DIR) Post #AWdrlbDe1XR16Vk69w by mic92@toot.matereal.eu
2023-06-12T18:15:43Z
0 likes, 0 repeats
@sandro the naive flamegraphs of nixpkgs are actually quite hard to interpret. ninjatrappeur built some with a patched nix that are better... Maybe there is something. Also I have the feeling it's rather a "death by a thousand papercuts" kind of problem.
(DIR) Post #AWdrlboVoSvQwqzXcm by sandro@c3d2.social
2023-06-12T19:15:03Z
1 likes, 1 repeats
@mic92 nixos eval time with no rebuilds is dominated by one systemd assert function https://github.com/NixOS/nixpkgs/pull/193706