| 12 Dec 2024 |
 @hexa:lossy.network | I agree that it is a hack | 15:17:36 |
| @hexa:lossy.network | substract 899 from your rebuild count and you're golden | 15:17:43 |
| @hexa:lossy.network | that'll throw us back years in terms of maintainability | 15:18:04 |
| @hexa:lossy.network | rebuild count and nixpkgs-review look at the same list | 15:18:36 |
| @hexa:lossy.network | so we can't have one without the other | 15:18:40 |
| @hexa:lossy.network | and folding all tests back in will make the core package much more annoying again | 15:19:23 |
| @hexa:lossy.network | not recursing into the attrset also means we don't get hydra results anymore | 15:19:46 |
 dotlambda | We could add a top-level attr tests that is recursed into but subtracted from the rebuild count | 15:20:16 |
| @hexa:lossy.network | to be fair, they are rebuilds, and unpacking the home-assistant source 900 times has a certain cost to it | 15:20:56 |
| @hexa:lossy.network | and creating a precedent for top-level attributes that are free is probably not smart | 15:21:39 |
| dotlambda | In reply to @hexa:lossy.network
to be fair, they are rebuilds, and unpacking the home-assistant source 900 times has a certain cost to it I don't think many people mind the cost the builds incur on Hydra. | 15:22:26 |
| dotlambda | In reply to @hexa:lossy.network
and creating a precedent for top-level attributes that are free is probably not smart Wdym? | 15:22:34 |
| @hexa:lossy.network | emacs people are also saying their package set rebuilds are essentially free 😄 | 15:24:48 |
| dotlambda | One alternative to having a top-level attribute at all would be adding the respective component tests to all reverse dependencies' passthru.tests | 15:29:34 |
| @hexa:lossy.network | yeah, I thought about that as well, but it generally fluctuates too much | 15:30:18 |
| @hexa:lossy.network | they're swapping out dependencies for the same component every other release and we fail to clean up unused deps already | 15:30:59 |
| dotlambda | We'd have write a script that keeps the passthru.tests up to date | 15:31:27 |
| dotlambda | * We'd have to write a script that keeps the
passthru.tests up to date | 15:31:51 |
| @hexa:lossy.network | sorry for being the naysayer in this conversation fwiw 😄 | 15:31:55 |
| dotlambda | * We'd have to write a script that keeps the `passthru.tests` up to date | 15:32:14 |
| @hexa:lossy.network | there is two things to loose:
a) hydra for bisects
b) nixpkgs-review for dependency bumps | 15:33:19 |
| @hexa:lossy.network | I'd much rather loose a) than b) | 15:33:49 |
| @hexa:lossy.network | * there is two things to loose:
a) hydra builds for per-component bisects
b) nixpkgs-review for dependency bumps | 15:34:11 |
| dotlambda | I don't know what you mean by a) | 15:34:28 |
| @hexa:lossy.network | 
Download image.png | 15:35:19 |
| @hexa:lossy.network | knowning when something broke | 15:35:31 |
| @hexa:lossy.network | * knowing when something broke | 15:35:36 |
| @hexa:lossy.network | in terms of good/bad bisect starting points | 15:36:08 |
| @hexa:lossy.network | nixpkgs-review currently does not take passthru.tests into account | 15:36:51 |
| dotlambda | Well, that seems like an obvious thing to give up but also nobody minds it being there | 15:37:07 |
