!kyXJonZuBXCGzVwuSn:nixos.org

Systems Programming

305 Members
Kernel, stdenv, low-level hacking, patchelf, … 78 Servers

Load older messages


SenderMessageTime
23 Nov 2023
@k900:0upti.meK900 ⚡️I think what they actually want is a way to quickly test a kernel from a local dirty tree on a system16:00:58
@thefossguy:matrix.orgPratham Patel (you can mention me)Yes, but those make targets were mere examples of "having packaging in upstream".16:01:11
@k900:0upti.meK900 ⚡️Which is annoyingly difficult on NixOS, but a new kernel build system target will not help16:01:16
@thefossguy:matrix.orgPratham Patel (you can mention me)My goal of proposing this is simply testing a dirty tree on NixOS :)16:01:38
@thefossguy:matrix.orgPratham Patel (you can mention me)Which is what would help in the bcachefs testing for Kent16:02:11
@joerg:thalheim.ioMic92I usally build a kernel less nixos image and than just pass the kernel directly to qemu along with this image.16:04:00
@thefossguy:matrix.orgPratham Patel (you can mention me)Wouldn't something like this work? Yes, I know that we will need to use the package in a VM/nixos-config for it to be actually useful but this is more of a proof-of-concept. flake.nix: https://rpa.st/RLSQ linux.nix: https://rpa.st/TVZA16:26:29
@thefossguy:matrix.orgPratham Patel (you can mention me)(I'll sleep, have finals tomorrow.)16:26:54
@k900:0upti.meK900 ⚡️It probably would, but that doesn't solve your original problem16:39:45
@woobilicious:matrix.orgwoobiliciousturns out my linking issues are related to a patch on bcachefs master....fuck knows how that happens.23:54:15
@woobilicious:matrix.orgwoobiliciousnone of them use or seem to touch pr_debug()23:54:34
@woobilicious:matrix.orgwoobiliciousand to make matters worse if you compile with a dirty build it seems to work.23:55:50
24 Nov 2023
@janik0:matrix.orgJanik (they/them) joined the room.16:44:56
@elvishjerricco:matrix.orgElvishJerricco joined the room.21:29:52
@joepie91:pixie.townjoepie91 🏳️‍🌈 joined the room.21:53:02
@c0ba1t:matrix.orgCobalt joined the room.21:54:42
25 Nov 2023
@thefossguy:matrix.orgPratham Patel (you can mention me)Something that I can't put a finger on is what config should this "possible in-tree solution" use (by default)? tinyconfig (broken on arm because this doesn't build DTBs)? defconfig? or what mainline.nix has + maintainer's overrides?03:44:08
@thefossguy:matrix.orgPratham Patel (you can mention me)I can see tinyconfig being useful with a single override for the maintainer's module to test if the changes even compile and then the nixos config + maintainer's override (if not enabled already) as the one they build to test on physical hardware.03:46:05
26 Nov 2023
@thefossguy:matrix.orgPratham Patel (you can mention me)

Is anyone willing to explain me how a kernel is actually built in for NixOS? At the moment, I'm looking at three files and here's what I understand so far:

  1. mainline.nix: it all starts here, where kernel-specific items such as version etc are defined
  2. generic.nix: this defines buildLinux which [I think] is the builder for building kernels
  3. common-config.nix: specifies config options that are answered during make nconfig
  4. manual-config: this seems to have the build logic for everything from creating a config to installing it in $out
15:08:57
@k900:0upti.meK900 ⚡️That is mostly accurate, yes15:39:24
@thefossguy:matrix.orgPratham Patel (you can mention me)Is there something else that I've not listed?15:50:02
@thefossguy:matrix.orgPratham Patel (you can mention me)Also in context to "mostly accurate", what did I assume wrong?15:50:27

There are no newer messages yet.


Back to Room ListRoom Version: 6