| 18 Jun 2026 |
 hexa | I assume that would bring correlation of nar and narinfo closer together again | 08:53:55 |
 Vladimír Čunát | I'm not happy with nixos-26.05-small blockage either. | 08:54:01 |
| Vladimír Čunát | (caused by abortions which don't go away on restarts) | 08:54:18 |
 Mic92 | working on that part | 08:54:37 |
| Mic92 | I have to figure out if this is indeed pre-signing and not also somehow related with the switch from xz to zstd | 08:55:57 |
|  erents joined the room. | 09:04:34 |
| Mic92 | I guess I take the day off to fix this stuff | 09:10:38 |
|  Myria joined the room. | 09:14:04 |
| Mic92 | https://github.com/NixOS/infra/pull/1085 Have looked into the aborts.... | 09:35:23 |
| Mic92 | I will deploy this now and watch | 09:35:30 |
| Mic92 | and in the meantime look into the binary cache | 09:35:40 |
| Mic92 | * https://github.com/NixOS/infra/pull/1085 Had looked into the aborts.... | 09:37:19 |
| Mic92 | My working theory is that the same issue is also causing the NAR mis-matches | 09:50:12 |
| Mic92 | * My working theory is that the same issue I fixed in there, is also causing the NAR mis-matches | 09:50:48 |
| Mic92 | Arian: did you just reset a single file in the cache or more than that? | 09:56:44 |
| Mic92 | because this seems to be reset now? https://github.com/NixOS/hydra/issues/1804 | 09:57:22 |
|  sampointon joined the room. | 10:05:29 |
 Arian | I didn't do anything. Wrong person tagged? | 10:05:58 |
| hexa | Not me either fwiw | 10:06:21 |
|  Grégor joined the room. | 10:26:44 |
| Vladimír Čunát | I suppose we'll now keep running into various consequences, e.g.
https://hydra.nixos.org/build/331666283/nixlog/262 | 10:29:02 |
| Mic92 | I am scanning the cache | 10:30:09 |
| Mic92 | To find all candidates | 10:30:15 |
| Vladimír Čunát | I keep hating that we have to "build" all constituents to pass the job (they can be huge closures), but that's mostly an independent issue. | 10:30:38 |
| Mic92 | Just checking the version won't be enough if it also affects new store path | 10:30:47 |
| Mic92 | * Just checking the bucket version won't be enough if it also affects new store path | 10:30:59 |
| Mic92 | It will help with recovery a bit though | 10:31:07 |
| Vladimír Čunát | * I keep hating that we have to "build" by moving all constituents to a builder - in order to pass the job (they can be huge closures), but that's mostly an independent issue. | 10:31:05 |
| Vladimír Čunát | Non-FOD derivations should get obsolete within a month or two. | 10:32:20 |
| Vladimír Čunát | * Non-FOD derivations should get "obsolete" within a month or two. | 10:32:44 |
