| 19 Feb 2026 |
 emily | right now every channel setup will try to rebuild the world once -next merges 🫠| 15:53:44 |
 VladimÃr ÄŒunát | * We also have some broken NixOS tests, even -small channel blockers. | 15:53:51 |
| emily | are those the installer tests? | 15:53:59 |
| emily | because this would also plausibly break those | 15:54:02 |
| emily | because how you get your Nixpkgs becomes a cache buster | 15:54:10 |
| VladimÃr ÄŒunát | Oh right, maybe this PR will fix them. | 15:54:17 |
| VladimÃr ÄŒunát | Makes sense. | 15:54:21 |
| emily | I'm not sure if lib.trivial.version even matches between the -next and master Hydra jobs, so there's a funny scenario where we still get a world rebuild on master… | 15:55:21 |
| emily | * I'm not sure if lib.trivial.version even matches between the -next and master Hydra jobsets, so there's a funny scenario where we still get a world rebuild on master… | 15:55:31 |
| emily | (but worst case we can just hardcode "26.11pre-git" for now) | 15:55:33 |
| VladimÃr ÄŒunát | It fixed nixosTests.containers-imperative apparently. | 16:03:48 |
| VladimÃr ÄŒunát | So maybe it's all fine now. | 16:03:55 |
| VladimÃr ÄŒunát | (I've fixed rsyslog build in the meantime.) | 16:04:04 |
| VladimÃr ÄŒunát | * So maybe it's all fine now, at least after we merge that PR. | 16:04:35 |
 K900 | FYI: https://github.com/NixOS/nixpkgs/pull/492157 | 16:16:37 |
| K900 | Will merge once I'm sure it's 0 rebuilds | 16:16:48 |
| K900 | Also libphonenumber is busted with boost 1.89 | 16:19:00 |
| K900 | But 1.88 works | 16:19:02 |
| K900 | Will pin unless anyone wants to dig into it | 16:19:07 |
| emily | is there actually a formal distinction between dynamic attributes in rec and any use of with, btw? (not that I wouldn't be happy to see both rec and with purged from the language) | 16:22:28 |
| K900 | Formal, no | 16:23:22 |
| K900 | But it gets REALLY funny with rec | 16:23:32 |
| emily | rec { ${if cond then "ambiguous" else null} = …; test = ambiguous; }.test seems indistinguishable from with { ${if cond then "ambiguous" else null} = … }; ambiguous | 16:23:43 |
| emily | hmm, except the former always complains about ambiguous not being bound | 16:26:17 |
| emily | which, yeah exposes ${"ambiguous"} being treated specially | 16:26:34 |
| emily | since rec { ${"ambiguous"} = 123; test = ambiguous; }.test does work | 16:26:56 |
| K900 | Yes, try ${"a" + "mbiguous"} | 16:27:10 |
| emily | arguably rec with conditional attributes should let you refer to them and just turn off scope checking like with does 🙃 | 16:27:15 |
| K900 | To bypass that optimization | 16:27:14 |
| emily | right, in which case it just complains that ambiguous is unbound | 16:28:15 |
