| 22 Jan 2026 |
 raboof | We follow stackage LTS https://nixos.org/manual/nixpkgs/stable/#ghc-deprecation-policy | 22:03:50 |
| raboof | Squinting at past ghc and stackage release dates it looks like stackage LTS being released with ghc 9.12 will still be some months away, but I haven't followed closely | 22:13:34 |
| raboof | So not for 26.05 but probably for 26.11 I'd wager | 22:18:21 |
| 23 Jan 2026 |
 abacef | What do yall think about verifying nixos arm reproducibility? | 23:49:46 |
| 24 Jan 2026 |
 dish [Fox/It/She] | its definitely a good idea | 00:23:34 |
| dish [Fox/It/She] | but would there necessarily be differences w.r.t. x86 there? Are there any cases you've found where ARM is reproducible and x86 isn't, or vice versa? | 00:24:12 |
| abacef | Debian packages have differences per architecture https://reproduce.debian.net/ | 01:28:01 |
 Julien | Would be good im | 12:23:25 |
| Julien | * | 12:23:29 |
| raboof | sounds like a good idea to do, shall I define a graphical-iso-runtime-aarch64 jobset? | 12:29:25 |
| Julien | Does anyone have hardware to verify it ? | 14:44:46 |
| raboof | I have a pi5 that I could put to use (though it'll probably be slow and fail on some of the larger builds). I still need to finish/troubleshoot setting it up though - for some reason it seems the async-nix-post-build-hook is hanging. | 16:54:21 |
| raboof | * I have a pi5 that I could put to use (though it'll probably be slow and fail on some of the larger builds). | 17:04:17 |
| abacef | I have some Azure credit I can get an arm server | 20:03:50 |
| 25 Jan 2026 |
 atemu12 | Never tried it but oracle has a free tier that includes an ARM VM with 24GB | 00:32:24 |
| abacef | I try to get free Oracle ARM VMs with my account sometimes but every time it says they dont have capacity in the single region where my free account is locked into. | 10:39:34 |
| atemu12 | Ah, welp; you get what you pay for I suppose.. | 10:41:18 |
| raboof | https://reproducibility.nixos.social/evaluations/6/88d3861acdd3 slowly getting populated - more rebuilders definitely welcome ;) . | 11:04:42 |
| raboof | reviews on https://github.com/nix-community/lila/pulls as well btw | 11:04:46 |
| 26 Jan 2026 |
| abacef | nixos on azure is not easy | 21:39:00 |
| 27 Jan 2026 |
|  whispers [& it/fae] changed their display name from whispers (it/fae) to whispers [& it/fae]. | 02:51:44 |
| 28 Jan 2026 |
| abacef | raboof: when I do nix shell for lila on arm, I get an error that the lila flake does not provide an attribute for the aarch64-linux packages | 17:38:41 |
| abacef | I just guessed that I had to change the system variable to aarch64-linux from the lila readme | 17:40:03 |
| Julien | I think there is a branch to reintroduce aarch64 outputs but I don’t know if it works | 17:40:50 |
| 31 Jan 2026 |
|  martijn joined the room. | 15:36:56 |
| 1 Feb 2026 |
| raboof | yes I used the branch at https://github.com/nix-community/lila/pull/117 - sadly there's something wrong with the nix tests when run through qemu (https://github.com/NixOS/nix/issues/13045) | 10:24:19 |
| raboof | I represented Nix at the reproducible-builds BOF at FOSDEM. People are (rightfully) pretty excited about the progress on full-source bootstrap (both https://chaos.social/@nzbr/115973847897716839 and https://github.com/NixOS/nixpkgs/pull/479322) | 10:39:35 |
|  Fernando Rodrigues changed their profile picture. | 10:42:15 |
| raboof | it's good to see there's recent progress on https://lore.kernel.org/all/20260113-module-hashes-v4-0-0b932db9b56b@weissschuh.net/ - currently we typically disable module signing in NixOS (because we have no great way to achieve reproducible signatures), this could be an attractive way to allow reproducibly locking down modules. Supporting out-of-tree modules would still need some thought/work, though. | 10:48:41 |
| raboof | I'm still sympathetic to this goal, but I learned the Debian project (who have spent a lot of energy fixing build path differences causing unreproducibilities) have gone in the opposite direction: they have essentially given up on making build paths not influence the build and are moving to fixating the build path instead. Since they don't do that lightly, when we start varying the build path on Nix Linux, I think we should at minimum have an escape hatch to opt select derivations out of that variation. | 10:53:56 |
