<pie_>
Ericson2314: a summary is that inspecting nixpkgs' generated kernel versions doesnt make sense in containers, IIUC, because the containers dont actually influence the kernel but instead get it from the environment
<pie_>
so to properly fix this we would need to pull in some mutable state from somewhere
<pie_>
im askin you beacause i think you did a lot of stuff with system or whatever the parameter was called, because of the cross-compile infra, so i figured you might know something about where this would need to be plugged or who to talk to
<pie_>
i don't know my way around this at all but id like to try to help getting the ball rolling if nothing else
evanjs has quit [Read error: Connection reset by peer]
evanjs has joined #nixos-dev
devhell has joined #nixos-dev
m1cr0man has quit [Ping timeout: 246 seconds]
m1cr0man has joined #nixos-dev
devhell has quit [Quit: leaving]
<ScottHDev>
Hello, I've got a CI problem with my nixos test. The test runs fine on my NixOS Laptop with kvm enabled. But on a github workflow without kvm, I get a strange error. It seems like the files installed by environment.systemPackages aren't accessible within the testScript, here is the corresponding CI run :
supersandro2000 has quit [Remote host closed the connection]
supersandro2000 has joined #nixos-dev
<endocrimes>
o/ - Partner and I noticed that Go bumps are decreasingly reviewed/merged - does the Go team need extra folks? (we both depend on Go in nix heavily and are fine with bumping + testing stuff)
<eyJhb>
I think anything helps, and if you have the time please do review the PRs, and leave a comment saying it works/does not work
<eyJhb>
Regarding merge rights, I have no clue what the requirements are for that :p
<qyliss>
Ericson2314: had a quick look at couldn't figure it out. Probably won't have much time today but feel free to remind me in a few days if you're still stuck.
<pie_>
Ericson2314: not sure what the question is - yes if you mean "kernel version checks"
<Ericson2314>
pie_: i mean is that PR sufficient to address your concerns?
jonringer has joined #nixos-dev
<Ericson2314>
kernel version for non-kernel things is really about asserting what syscalls should be available, which applies in containers too
<pie_>
yeah but i mean the nixpkgs running in the container doesnt actually know what the kernel is
<pie_>
maybe its just me but there seemed to be reluctance do just drop a toggle in
copumpkin has joined #nixos-dev
<pie_>
its certainly simpler than working on better detection stuff, but the better detection stuff might be warranted anyway - or triggerable by some central override for the kernel version or something, idk
<sterni>
Ericson2314: I've managed to get the binutils thing to work! #119625
<Ericson2314>
pie_: were we somehow auto-detecting kernel version?
<pie_>
well the immediate issue is that the kernel.version attribute didnt even exist in this case
<pie_>
Ericson2314: no, the assumption was that kernel.version is availible
<pie_>
which makes sense if you are the one setting the kernel
<pie_>
but it doesnt make sense if the kernel is provided to you
<sterni>
really happy it went without any major hickups since I've spend ages waiting for stdenv builds
<pie_>
by the environment <pie_> but it doesnt make sense if the kernel is provided to you
<Ericson2314>
that PR solves that
<Ericson2314>
sterni: great!
<Ericson2314>
right i think nixos vs non nixos is more interesting than containers vs no container
<Ericson2314>
sterni: I think we once had an issue of always passing target caused permanent prefixing
<sterni>
Ericson2314: seems like now binutils refuses to prefix if host == target
<sterni>
I mean it doesn't really matter if we always pass target or not actually, but I thought I'd clean it up at the same time given the TODO comment :)
<pie_>
sure why not <Ericson2314> right i think nixos vs non nixos is more interesting than containers vs no container
<Ericson2314>
sterni I am making a suggestion to this point
<Ericson2314>
sterni: OK I left the comment
<Ericson2314>
feel free to reword or rewrap the text
<Ericson2314>
I just sort guessed at line length
<sterni>
wait where? either github is having a hickup or you haven't posted it yet
<Ericson2314>
sterni: oh oops
<Ericson2314>
refresh
<Ericson2314>
also maybe we want to do the same for gcc and other things that had that same TODO?
<sterni>
now :)
<Ericson2314>
I original intended the TODO to mean always prefix with the target config
copumpkin has quit [Remote host closed the connection]
<Ericson2314>
but that is a much harder thing
<Ericson2314>
doing it your way we can knock off all those TODOs without issue :)
<sterni>
ah yeah, it makes a lot of sense always forcing the targetPrefix, you're right
<Ericson2314>
sterni: for your build failure, we might need to force cross compilation mode
<infinisil>
lovesegfault: Oh yeah sorry I'm really short on time recently, so yeah dismissing my review sounds good, I trust in the other people reviewing it :)
<lovesegfault>
Got it, just wanted to make sure I didn't upset you by marking it stale :)
<elvishjerricco>
Sweet. I have a very usable initrd rescue environment. Tested by switching nixos.qcow2 with an image with a luks volume. Boot failed because /dev/vda isn't a file system, entered the root password, ran cryptsetup open and mount manually, systemctl default, and got myself booted
v0|d has quit [Remote host closed the connection]
<eyJhb>
Ehm. If I have two packages that needs to be updated together, then how do I title the PR?