| 17 Oct 2025 |
 fzakaria | Can we make local-nar-cache have a default ? | 00:12:17 |
| fzakaria | supported for BinaryStore
https://cache.nixos.org/?local-nar-cache=...
| 00:12:30 |
| fzakaria | it's the reason why it reads the NAR all in memory | 00:15:14 |
| fzakaria | wrote it here https://github.com/NixOS/nix/issues/14284 | 00:19:44 |
| fzakaria | (wasn't even sure why can't just write to the /nix/store | 00:19:53 |
| fzakaria | * (wasn't even sure why can't just write to the /nix/store) | 00:19:56 |
 John Ericson | I did change some of that stuff, but not in a way that I think would affect fresh buidls | 02:14:02 |
| John Ericson | * I did change some of that stuff, but not in a way that I think would affect fresh builds | 02:14:05 |
| John Ericson | only perhaps substitutions of CA derivations or something like that | 02:14:16 |
| John Ericson | * only perhaps substitutions of CA derivations or something obscure like that | 02:14:22 |
| fzakaria | Another fix in the bag | 03:06:22 |
| fzakaria | https://github.com/NixOS/nix/pull/14285 | 03:06:22 |
| fzakaria | Redacted or Malformed Event | 04:07:48 |
| fzakaria | Redacted or Malformed Event | 04:18:54 |
|  Avdhut7009 joined the room. | 05:07:27 |
 raboof | this is indeed not about 'regular' builds but about rebuilds - filed https://github.com/NixOS/nix/issues/14287 | 07:27:57 |
 Sergei Zimmerman (xokdvium) | In reply to @raboof:matrix.org
this is indeed not about 'regular' builds but about rebuilds - filed https://github.com/NixOS/nix/issues/14287 I think there’s been some changes around builtOutputs that might not keep those for bmCheck mode. | 07:34:33 |
| Sergei Zimmerman (xokdvium) | Must have been the DerivationGoal per-output split | 07:35:13 |
| Sergei Zimmerman (xokdvium) | Or was it called just in the DerivationBuildingGoal? | 07:36:17 |
| raboof | jup indeed looks like it broke in https://github.com/NixOS/nix/pull/13294 | 13:30:15 |
| Sergei Zimmerman (xokdvium) | In reply to @raboof:matrix.org
jup indeed looks like it broke in https://github.com/NixOS/nix/pull/13294 That itself was a fix for checking when not all outputs are valid | 13:38:44 |
| Sergei Zimmerman (xokdvium) | So when some outputs have been substituted previously | 13:39:16 |
| raboof | yeah I won't pretend I understand what exactly is going on, just sharing my observation after bisecting :) | 13:45:27 |
| John Ericson | raboof: ah OK I think I see what is going on | 15:24:41 |
| John Ericson | there is no need to re-register the path info in the check case, but we still want to re-sign it | 15:25:00 |
| John Ericson | (assuming it didn't change) | 15:25:05 |
| John Ericson | raboof: so a tricky thing with --check is that it doesn't register outputs that were not registered before | 15:46:05 |
| John Ericson | I was surprised to learn that, and a few of my refactors were dedicated to making sure I didn't break that | 15:46:47 |
| John Ericson | as a I split things up | 15:46:56 |
| John Ericson | that is why DerivationBuilder doesn't register or return any path infos in the --check case | 15:47:27 |
