| 22 Apr 2026 |
 Rick (Mindavi) | In reply to @mynacol:mynacol.xyz
And I don't understand the hydra website. https://hydra.nixos.org/build/326614946 and https://hydra.nixos.org/build/326734886 succeeded but the log shows errors; and https://hydra.nixos.org/build/326734884 and https://hydra.nixos.org/build/326734883 failed (timed out), but in the logs it succeeded. Is that a known Hydra bug/inconsistency? Rebuilds keep the old logs as well, is that what you maybe see? | 08:44:26 |
 mynacol | I just clicked on the log of e.g. https://hydra.nixos.org/build/326734886 and it shows failures despite succeeding. Could very well be an old log, but then how do I get/see the new one? | 08:46:28 |
 hexa | Huh, odd | 08:47:00 |
| 23 Apr 2026 |
 Arian | Moving this from private channel to public because the incident is now public
https://www.githubstatus.com/incidents/zsg1lk7w13cf
I think we should disable merge queues in nixpkgs as a pre-caution. Just had a repo corrupt itself at work due to this incident.
We don't have squash merges in nixpkgs so maybe we're not affected but it'd prefer to be cautious.
@lassulus:lassul.us do you have permission bits to disable this on the repo? | 20:17:40 |
 Winter | not sure i love that idea | 20:23:50 |
| Winter | cc mdaniels5757 | 20:24:21 |
| Winter |
| 20:24:27 |
| Arian | I'm hoping we're not affected because of us not having squash/rebase enabled ... But who knows | 20:26:01 |
 emily | I don't think there's reason for us to act if they've explicitly narrowed it down that way | 20:28:40 |
| emily | (though admittedly they also don't say "corruption", do you have any details on that?) | 20:29:08 |
| emily | I expect any issue along those lines is far less likely with merge commits since it's not rewriting commits | 20:29:21 |
| emily | i.e. the way they've narrowed it down makes me strongly suspect it's an issue in their reimplementation of git replay type functionality | 20:29:42 |
| emily | (I hope enough people have git clone --mirrors of Nixpkgs that we would be fine even if something did go wrong at least… and also that GitHub has backups… 😬) | 20:30:30 |
| Arian | Squash merge the merge queue generated contained random reverts of a bunch of commits. Turned a 12 line 1 file diff into a 2500 line 207 file diff | 20:31:07 |
| 24 Apr 2026 |
 Vladimír Čunát | They reported it as resolved, 6-7h ago now. So I guess we're good. | 04:15:41 |
|  Corngood joined the room. | 13:29:27 |
| Corngood | I have what seems to be an intermittent build failure in dotnet-sdk_10: https://hydra.nixos.org/build/326664066
Any chance I can get it restarted? It's been reported upstream by another distro, but I've had no luck reproducing it myself. | 16:34:45 |
| Vladimír Čunát | Restarted on nixpkgs side: https://hydra.nixos.org/build/326672215 | 16:36:40 |
| Vladimír Čunát | (but there's a newer build on a later nixpkgs commit which succeeded, too) | 16:36:57 |
| Corngood | I saw that there was a newer successful one on nixpkgs/unstable, but the latest one showing on nixos/unstable was failed. Did I just need to wait for another eval? | 16:48:55 |
| Vladimír Čunát | No, the one in the last eval was a different one: https://hydra.nixos.org/build/326789264 | 16:57:29 |
| Vladimír Čunát | I guess you only looked at finished jobs. | 16:57:44 |
| Corngood | Ah yeah maybe. Sorry about that. | 17:47:41 |
| hexa | error: users cannot be create over SSH without Full Disk Access, aborting activation
The user hydra-queue-builder could not be create as `darwin-rebuild` was not executed with Full Disk Access over SSH.
You can either:
grant Full Disk Access to all programs run over SSH
or
run `darwin-rebuild` in a graphical session.
The option "Allow full disk access for remote users" can be found by
navigating to System Settings > General > Sharing > Remote Login
and then pressing on the i icon next to the switch.
| 20:24:57 |
| hexa | Arian seeing this on norwegian-blue | 20:25:07 |
| hexa | can we fix this via MDM? | 20:25:26 |
| hexa | 
Download | 20:29:46 |
| hexa | Arian I don't see the creds for the relevant apple account in vault | 20:31:33 |
| hexa | ok, could accept terms on behalf of the account | 20:39:50 |
 lassulus | Is this taken care of or do I need to think about? :) | 20:44:43 |
