| 1 Aug 2025 |
 emily | the Linux "installation scripts" are very silly and pointless | 12:47:44 |
| emily | I am literally replacing them with cp | 12:47:48 |
| emily | oh this is recent | 12:48:14 |
| emily | I don't understand why this says "so that we don't have to maintain some postInstall logic for LoongArch." but then deletes nothing | 12:48:38 |
| emily | Use the generic script/install.sh to perform the make install operation.
This will automatically generate the initrd file and modify the grub.cfg
without manual intervention (The previous kernel image, config file and
System.map will also be generated), similar to other architectures.
imagine running a kernel make install and having it modify your bootloader configuration
| 12:49:16 |
 dramforever | idk i haven't checked the actual pr | 12:49:57 |
| dramforever | i think linux make install is just a very concerning thing in general | 12:51:10 |
| dramforever | like it has different behavior depending on if your current distro is rpm or deb or something | 12:51:33 |
| emily | yeah and there are arch-specific scripts that all do more or less the same thing but in subtly different annoying ways | 12:52:18 |
| emily | have you seen our preInstall | 12:52:34 |
| emily | I am deleting all of that | 12:52:43 |
 Alyssa Ross | I tried to refactor it a long time ago to always use KBUILD_IMAGE but it got reverted because it broke MIPS or POWER or something | 12:53:13 |
| emily | my WIP just uses cp | 12:53:31 |
| emily | because if you trace down all the installation scripts all they do is cp $(KBUILD_IMAGE) | 12:53:42 |
| emily | like I don't even make install, at all | 12:53:52 |
| emily | do you have a link to that though | 12:54:31 |
| emily | I remember seeing some good-looking kernel PR you did that got amjoseph'd | 12:54:44 |
| emily | maybe same one | 12:54:48 |
|  @brisingr05:matrix.org joined the room. | 18:05:44 |
| 2 Aug 2025 |
|  @saiko:knifepoint.net changed their profile picture. | 00:27:55 |
 @aloisw:julia0815.de | Are lines like this https://github.com/NixOS/nixpkgs/blob/59dd2e527d4ccc18fd9f8f773ccd0f2b4f43b980/pkgs/by-name/li/libimagequant/package.nix#L40 correct or should it be from buildPackages? | 09:14:21 |
| @aloisw:julia0815.de | I noticed that libimagequant != pkgsStatic.buildPackages.libimagequant, but am not sure if this is intentional. | 09:14:53 |
| emily | assume Rust cross / setEnv stuff is screwy by default | 09:25:35 |
| emily | has never steered me wrong | 09:25:45 |
| @aloisw:julia0815.de | Screwy in what way though, is the question here. FWIW putting buildPackages restores equality but I'm not sure if this is correct here. | 09:28:26 |
| emily | I believe it is because the corresponding use in the normal hooks will be one level up | 09:28:57 |
| @aloisw:julia0815.de | Thank you for validating my conjecture then, I will prepare a treewide fix then. | 09:32:47 |
| emily | a hook may be in order if you have the energy | 09:34:32 |
| @aloisw:julia0815.de | Not sure if that's worth it for the 6 packages that do it? | 09:39:42 |
| emily | 6 opportunities to mess up cross! | 09:46:27 |
