#nixos-systemd on 2020-08-11

09:15 <flokli> eww: https://github.com/NixOS/nixpkgs/pull/94679#issuecomment-671830974

09:15 <flokli> anyone up for a bisect on what introduced this?

10:11 <Mic92> flokli: does it use c++ in any form?

10:12 <Mic92> And why is this dependency not allowed?

10:12 <flokli> https://github.com/NixOS/nixpkgs/issues/95163

10:12 <{^_^}> #95163 (by flokli, 18 seconds ago, open): initrd currently doesn't build, blocking channels

10:18 <Mic92> flokli: I can see that it is at the very end of the elf header

10:18 <Mic92> mhm actually no the end

10:51 <flokli> Yeah

10:51 <flokli> Not sure what it's doing there

10:54 <Mic92> flokli dynstr is usually used by the dynamic linker to encode symbols. But I don't see how this string is used.

11:00 <Mic92> I don't see this string when I compile systemd from source

11:20 <flokli> Mic92: mhh?

11:21 <flokli> At least the bisect showed https://github.com/NixOS/nixpkgs/commit/d8ac41f86394365f8319d8b37fd25e6a7ba7d904 as a possible culprit

11:23 <Mic92> flokli: oh, now that I see this patch: https://github.com/NixOS/nixpkgs/commit/d8ac41f86394365f8319d8b37fd25e6a7ba7d904#diff-4515662399150972fd68dc36a6ef559bL8

11:23 <Mic92> I can see where the bug is coming form

11:23 <Mic92> I can also see a lot of XXXX in the binary

11:23 <flokli> Yes

11:23 <flokli> And the /nix/ is cut off as well

11:24 <flokli> Can't we just use patchelf there?

11:24 <Mic92> flokli: you can removeReferencesTo here

11:24 <Mic92> But I am not sure what the bug is about.

11:30 <flokli> the bug is about initrd assembly failing since the meson bump

11:32 <Mic92> flokli: the length of rpath is not calculated correctly I thin

11:33 <Mic92> flokli: so the hostfix could be even sed

14:52 <Mic92> flokli: Is there actually a reason why lvm/dmsetup are included in initrd by default?

14:52 <Mic92> I don't think I use any of them

15:12 <ajs124> Mic92: hasn't it always been like that? Iirc our initrd is kind of suboptimal when it comes to that kind of stuff anyways, because it's hard to include arbitraty "phases" optionally, right?

15:13 <Mic92> ajs124: is it tough? I see that some programs are included optionally

15:15 <ajs124> I got that impression from reading e.g. #46541, but it's been a while, so I might be misremembering.

15:15 <{^_^}> https://github.com/NixOS/nixpkgs/pull/46541 (by YorikSar, 1 year ago, open): nixos/initrd: Add thin-provisioning-tools binaries

15:15 <flokli> look at ac8a2ae57700e65e11396027993aab17b7187b51

15:17 <Mic92> ajs124: It's funny that it was 2018-me asking the same question

15:17 <flokli> That PR is somewhat related - I wonder if we should only provide lvm and mdadm in initrd if it's enabled in the nixos configuration (which should happen via nixos-generate-config)

15:18 <flokli> and provide some release notes of course

15:18 <ajs124> lol. Yeah, the impression I'm always getting is that the initrd does too much for some people and too little for others. Maybe even the same people, just for different environments.