Nixos-rebuild fails on a weird error after bumping nixpkgs

Help
1

I have updated my nixpkgs unstable revision from 9f4128e00b0ae8ec65918efeba59db998750ead6 (2024-07-03) to cb9a96f23c491c081b38eab96d22fa958043c9fa (2024-08-04) and performing a rebuild after this fails with the following log output:

output for /build/i3-4.23/testcases/t/533-randr15.t:
Successfully bound to /tmp/.X11-unix/X0
SIGCHLD, waitpid() = 0 at /build/i3-4.23/build/i3test.pm line 971, <$ipc_child> line 272.
ok 1 - outputs are __i3 and the fake monitor
ok 2 - Fake output at 3840x2160+0+0
SIGCHLD, waitpid() = 0 at /build/i3-4.23/build/i3test.pm line 971, <$ipc_child> line 272.
ok 3 - one bar configured
SIGCHLD, waitpid() = 0 at /build/i3-4.23/build/i3test.pm line 971, <$ipc_child> line 272.
ok 4 - bar_config output name is normalized
Exiting i3 cleanly...
test.inject_randr15: EOF
RRGetMonitors reply!
injecting reply
RRGetOutputInfo reply!
injecting reply
SIGCHLD, waitpid() = 0 at /build/i3-4.23/build/i3test.pm line 971, <$ipc_child> line 272.
ok 5 - outputs are __i3 and default
xrandr: Failed to get size of gamma for output default
SIGCHLD, waitpid() = 0 at /build/i3-4.23/build/i3test.pm line 971, <$ipc_child> line 272.
ok 6 - tick nonce received
ok 7 - Received 1 workspace::init event
ok 8 - Workspace initialized in up2414q
SIGCHLD, waitpid() = 0 at /build/i3-4.23/build/i3test.pm line 971, <$ipc_child> line 272.
ok 9 - outputs are __i3, default and up2414q
Exiting i3 cleanly...
1..9

===== End of test log =====

Right before this log output, there are 2 things building: qttools and qtwayland. I’m not sure if that’s related to the error; however, on a slight tangent, why are those 2 being built anyway? Is the binary cache missing?

1 Like
2

Which derivation is failing here?

3

I wish I had an answer; I am trying to figure that out myself.

However, the final error output has been slightly flaky. For example, here’s the output from when I tried a similar nixpkgs bump a few days ago:

image
image1416Ă—412 152 KB

Notice that there is an “ignored” error of some directory that is not empty. This rings similar to this issue: Temporary build directories not cleaned up because they are not empty · Issue #5207 · NixOS/nix · GitHub which didn’t really have a conclusive solution. Nevertheless, since it says it is “ignored,” I am not sure if this is directly related to the issue I’m facing.

4

There you go, it’s i3.

What modifications have you done to the i3 drv? The regular upstream one builds just fine.

1 Like
5

My first suspicion was i3; and I do have an overlay for it, but I tried removing it which made no difference and the rebuild still failed. Perhaps I did some silly error, I’ll try the rebuild again later today excluding the overlay.

6

I had an overlay, and after my initial suspicion of i3, I commented out the line adding the overlay; but I overlooked the fact that I’ve set my alias for nixos-rebuild switch to only consider the config at Git HEAD, so any modified files that haven’t been added to the Git tree will be ignored rendering the comment useless. I’ve hit myself in the foot.

It works now.

Though I looked through the i3 code that my patch edits and it’s unchanged. But alas, some external behaviour would’ve definitely changed.

Thanks for pointing me in the right direction!