| 29 Jun 2026 |
 flokli | ah, summary, part of evaluation # | 12:38:30 |
 Vladimír Čunát | I usually look at https://hydra.nixos.org/jobset/nixos/unstable | 12:38:56 |
| Vladimír Čunát | (the grey column are numbers of queued per eval) | 12:39:26 |
| Vladimír Čunát | Also note that status.nixos.org timestamps are for the moment of advancing, not for the stamp of the commit it advances to (which is my preferred view, with commit-date as a fairly good estimate of the definition which I'd prefer the most) | 12:41:32 |
| Vladimír Čunát | * Also note that status.nixos.org timestamps are for the moment of advancing, not for some stamp of the commit it advances to (which is my preferred view, with commit-date as a fairly good estimate of the definition which I'd prefer the most) | 12:41:46 |
| Vladimír Čunát | e.g. nixos-unstable is on a 3 days old commit now, almost precisely. | 12:42:40 |
| Vladimír Čunát | * e.g. nixos-unstable is on a 3 days old commit now, almost precisely.
(you can click the commit on status.nixos.org anyway) | 12:43:41 |
| Vladimír Čunát | * e.g. nixos-unstable is on a 3 days old commit now, almost precisely.
(you can click the commit on status.nixos.org anyway to see that) | 12:43:48 |
 Mic92 | are we good now in that regard? | 12:59:06 |
 hexa | too early to say | 13:00:43 |
| hexa | downgrading sleepy-brown from mandatory to supported big-parallel | 19:19:41 |
| hexa | at least then the machine won't idle | 19:19:47 |
| hexa | the macs are regularly idle because they still show a high load for a minute after builds have ended | 19:49:43 |
| hexa | that's stupid and we should look into that | 19:49:56 |
| Vladimír Čunát | I thought the "load" is consistent across POSIX. | 19:50:30 |
| Vladimír Čunát | (though for Linux we use PSI instead, I hope?) | 19:50:40 |
| Vladimír Čunát | * (Though for Linux we use PSI instead, I hope?) | 19:50:45 |
| hexa | yes, psi on linux | 19:50:46 |
| hexa | which is the much more useful metric | 19:50:52 |
| hexa | and averages over 10s, not 60s | 19:50:59 |
| hexa | 10/60/300 | 19:51:17 |
| Vladimír Čunát | Perhaps it will work OK after some tuning. We used to have make's -l parameter. | 19:51:38 |
| hexa | we can only tune machineFreeFn on the queue-runner | 19:52:29 |
| 30 Jun 2026 |
 Arian | I found a corrupt gnugrep nar I think?
https://cache.nixos.org/srk76h64a1ja4z7njh5lxcds7m9hnc7y.narinfo
Says file size 248729 bytes
But when i download i get 248412 bytes
The FileHash also doesn't match
Interestingly the NarHash does.. | 00:23:31 |
| Arian | Or at least the narinfo is corrupt | 00:25:51 |
 Ihar Hrachyshka | let me know if wrong channel...
if a package (thunderbird-unwrapped) store path matches a successful build on hydra, is it reasonable to expect it to be available in cache? (I ask because for 26.05 it is not there, getting MISS on narinfo - which forces thunderbird build on update.)
| 02:02:42 |
| hexa | should be, but I can see it is not | 02:21:46 |
| hexa | https://hydra.nixos.org/job/nixpkgs/nixpkgs-26.05-darwin/thunderbird-unwrapped.aarch64-darwin | 02:22:00 |
| hexa | my best guess is that this is due to a bug in the new queue-runner, which was fairly new when that build happened 8 days ago | 02:22:40 |
| hexa | I wonder more why 152.0 wasn't backported | 02:22:51 |
