| 28 Jun 2026 |
 hexa | we can try rolling that ban back for a while and see if it escalates again | 10:48:29 |
 Eymeric | If I'm the only one complaining, maybe it isn't worth the trouble. | 10:49:56 |
| hexa | we'll still match indvidiual clients that way, just not sure how great that works and willing to find out | 10:51:27 |
| hexa | Eymeric applied | 10:55:38 |
| Eymeric | ok for me | 10:56:14 |
| Eymeric | 
Download image.png | 10:56:16 |
| hexa | good | 10:56:31 |
|  guillaume joined the room. | 11:02:16 |
|  letgamer changed their display name from Alex Stephan to letgamer. | 11:08:16 |
|  Xy joined the room. | 22:16:17 |
| 29 Jun 2026 |
 dramforever | re the still happening gc race, could this possibly be relevant? https://github.com/DeterminateSystems/nix-src/pull/482/changes/734a205d58a559a622ebc98256f4af4d968426b7 | 06:43:49 |
| dramforever | i haven't had energy to do the rework on the massive changes in https://github.com/NixOS/nix/pull/15719 but it probably wouldn't really hurt to do the for() addTempRoot in copyPaths first | 06:44:47 |
| dramforever | * i haven't had energy to do the rework on the massive changes in https://github.com/NixOS/nix/pull/15719 but it probably wouldn't really hurt to do the for() addTempRoot in copyPaths first, in eelco's commit linked above | 06:45:18 |
 Vladimír Čunát | Yes, I keep seeing GC issues, e.g. ISO from eval 17h ago:
xorriso : FAILURE : Cannot determine attributes of source file '/nix/store/wny29c159hcwmcdijcxkb4rgy0mlzsvc-initrd-linux-7.1.2/initrd' : No such file or directory
| 08:06:17 |
| Vladimír Čunát | I'm trying to make at least the most important stuff work by manual retries. (channel blockers and whatever I notice around) | 08:07:35 |
| Vladimír Čunát | But it's risky to have this over longer term, especially as many packages have optional features and will just succeed if they miss them. | 08:08:14 |
| Vladimír Čunát | Sigh, 3/4 build slots of hopeful-rivest are stuck after finishing the build (for 22h now). The build logs look like the finished successfully, but the result is not in cache.nixos.org | 08:28:29 |
| Vladimír Čunát | Lots of other too long builds on https://hydra.nixos.org/machines | 08:28:36 |
| Vladimír Čunát | Roughly 1 day long. | 08:28:53 |
| Vladimír Čunát | * Sigh, 3/4 build slots of hopeful-rivest are stuck after finishing the build (for 22h now). The build logs look like they finished successfully, but the result is not in cache.nixos.org | 08:29:56 |
| Vladimír Čunát | I guess that's why the "running jobs" metric has been lower than usual in the past several hours. | 08:30:47 |
| Vladimír Čunát | Now I deployed lockfile updates, which restarted the queue-runner, so this state should go away. | 08:50:43 |
 Sergei Zimmerman (xokdvium) | In reply to @dramforever:matrix.org
re the still happening gc race, could this possibly be relevant? https://github.com/DeterminateSystems/nix-src/pull/482/changes/734a205d58a559a622ebc98256f4af4d968426b7 No, not really. The queue runner roots stuff itself and interacts with the daemon through harmonia | 09:11:48 |
| Sergei Zimmerman (xokdvium) | I think the last bits of FFI to nix are gone? | 09:12:05 |
| Sergei Zimmerman (xokdvium) | In reply to @xokdvium:matrix.org
No, not really. The queue runner roots stuff itself and interacts with the daemon through harmonia (And I know roughly nothing in that codebase) | 09:14:01 |
| Vladimír Čunát | * Now I deployed lockfile updates, which restarted the queue-runner, so this state should go away. | 09:58:58 |
| Vladimír Čunát | Yes, a few minutes ago again, a new ISO fail. The graphical ISOs seem to be quite a reliable test for this, as they have loads of dependencies and run for quite a few minutes (so likely hit the 1h GC interval). | 11:19:13 |
| Vladimír Čunát | * Yes, a few minutes ago again, a new ISO fail. The graphical ISOs seem to be quite a reliable test for this, as they have loads of dependencies and run for quite a few minutes (so likely hit the hourly GC runs). | 11:19:25 |
| Vladimír Čunát | * Yes, a few minutes ago again (step 97 here, a new ISO fail. The graphical ISOs seem to be quite a reliable test for this, as they have loads of dependencies and run for quite a few minutes (so likely hit the hourly GC runs). | 11:20:09 |
| Vladimír Čunát | * Yes, a few minutes ago again (step 97 here), a new ISO fail. The graphical ISOs seem to be quite a reliable test for this, as they have loads of dependencies and run for quite a few minutes (so likely hit the hourly GC runs). | 11:20:18 |
