| 24 Aug 2022 |
 colemickens | Elvish, do you have any intution for how hard it would be to modify the netboot stuff to load the squashfs on the fly frm the initrd? Or rather, if there are any gotchas about how early stages work that would make it trickier than just downloading it ahead of time? | 03:20:15 |
| colemickens | to clarify, rather than packing the squashfs into the initrd, the initrd would be slim and would wget it on the fly at boot | 03:20:44 |
| colemickens | hm, it might be easier to just have it boot a slim config and then nix-store -r a bigger config and switch into it... | 03:40:03 |
|  @tinybronca:sibnsk.net changed their display name from tinybronca to underpantsgnome. | 23:20:10 |
| 26 Aug 2022 |
 @elvishjerricco:matrix.org | So one mildly interesting idea relating to using the x-initrd.mount option is that apparently systemd won't unmount those FSes in stage 2 (fstab_is_extrinsic, among other things, prevents a mount from being unmounted by umount.target/shutdown.target). So that's actually one systemd patch we could delete from nixpkgs. | 00:48:04 |
| 28 Aug 2022 |
 flokli | ElvishJerricco: yeah, i wrote about this in https://github.com/NixOS/nixpkgs/issues/80038#issuecomment-699660256 a while ago - see the text below 0002*.patch | 22:43:01 |
| 30 Aug 2022 |
 @linus:schreibt.jetzt | finally getting back to https://github.com/NixOS/nixpkgs/pull/183314/files | 14:14:26 |
| @linus:schreibt.jetzt | I'm writing the release notes entry, but not sure whether it counts as a backwards incompatibility | 14:14:43 |
| @elvishjerricco:matrix.org | Linux Hackerman: Isn't it technically an improvement to compatibility? | 14:37:38 |
| @linus:schreibt.jetzt | not really, it's an improvement to initramfs size at the cost of some compatibility. | 14:39:20 |
| @linus:schreibt.jetzt | I'll put it in backwards incompatible changes. | 14:39:31 |
| @elvishjerricco:matrix.org | Linux Hackerman: I guess what I mean is that swraid was disabled by default with systemd stage 1 before, but that PR changes it so that whether it's disabled by default is a function of your stateVersion | 14:41:37 |
|  @aruzeta:matrix.org joined the room. | 14:41:42 |
| @linus:schreibt.jetzt | it does also affect non-systemd-stage1 though | 14:42:48 |
| @elvishjerricco:matrix.org | Yes but it only affects people with stateVersion >= 22.11, which theoretically don't exist | 14:43:17 |
| @linus:schreibt.jetzt | maybe I'm reaching too hard here, but it does break "install on mdraid, writing your own config, without using nixos-generate-config" | 14:46:51 |
| @elvishjerricco:matrix.org | *including writing stateVersion >= 22.11 | 14:47:50 |
| @elvishjerricco:matrix.org | but yea, I see your point | 14:47:55 |
| @linus:schreibt.jetzt | we've done the best we can to limit the amount of cases where it happens, and I'm pretty happy with that, but I think I feel better with that little corner of incompatibility being documented | 14:48:58 |
| @elvishjerricco:matrix.org | yep, makes sense | 14:49:13 |
| @elvishjerricco:matrix.org | regardless, I do still think that PR should cover LVM and bcache since we've changed whether those are enabled by default with systemd stage 1. | 14:50:36 |
| @linus:schreibt.jetzt | sure | 14:50:54 |
| @linus:schreibt.jetzt | though... bcache? Haven't seen that yet, is that enabled by default? | 14:51:44 |
| @elvishjerricco:matrix.org | Maybe I'm misremembering but I thought that the scripted initrd has support for bcache by default, and that systemd stage 1 doesn't. | 14:52:43 |
| @linus:schreibt.jetzt | wow you're right | 14:53:19 |
| @linus:schreibt.jetzt | it's in systemPackages unconditionally by default | 14:54:04 |
| @linus:schreibt.jetzt | I would have thought that bcache is obscure enough that by default no support would be included, but apparently I underestimated how old bcache is lol | 14:54:40 |
 @andreas.schraegle:helsinki-systems.de | bcache, not bcachefs, right? | 14:55:37 |
| @linus:schreibt.jetzt | yes | 14:55:39 |
| @linus:schreibt.jetzt | but yeah that really shouldn't be supported by default | 14:55:59 |
