| 8 Jan 2026 |
 emily | Wolfgang had a WIP PR to fix that with the queue if anyone wants to take it up FWIW | 16:09:39 |
| emily | I forget what it was missing | 16:10:07 |
| emily | there's a bit of subtlety about what should count as drift and how to signal it | 16:10:18 |
| emily | seems like probably OS upgrade fallout? | 16:37:01 |
| emily | I don't really feel qualified to diagnose cursed Nix bugs, though… | 16:37:20 |
| emily | could perhaps just skip the tests. | 16:37:39 |
| emily | https://github.com/NixOS/nixpkgs/issues/476794#issue-3778962273 lists shebang as failing, but it is not failing in the linked log. | 16:39:53 |
| emily | so at least one part of this is flakiness? | 16:39:57 |
| emily | ++(nix-shell.sh:63) nix-shell --pure -p foo bar --run 'echo "$(foo) $(bar)"'
these 2 derivations will be built:
/nix/var/nix/builds/nix-68111-2661222595/nix-test/main/nix-shell/store/pvbnf5qhij1hp7b6l5iyvsg54w1781pl-bar.drv
/nix/var/nix/builds/nix-68111-2661222595/nix-test/main/nix-shell/store/z8944igxbz7hsf1dgj5i0gijhb78ja3g-foo.drv
building '/nix/var/nix/builds/nix-68111-2661222595/nix-test/main/nix-shell/store/pvbnf5qhij1hp7b6l5iyvsg54w1781pl-bar.drv'...
building '/nix/var/nix/builds/nix-68111-2661222595/nix-test/main/nix-shell/store/z8944igxbz7hsf1dgj5i0gijhb78ja3g-foo.drv'...
/nix/var/nix/builds/nix-68111-2661222595/nix-shell-99938-0/rc: line 1: rm: command not found
+(nix-shell.sh:63) output=' '
+(nix-shell.sh:64) '[' ' ' = 'foo bar' ']'
| 16:40:52 |
| emily | it's a bit hard to diagnose without seeing what actually happened there. | 16:40:58 |
| emily | presumably https://github.com/NixOS/nix/pull/14778 isn't immediately relevant if the shebang test stopped failing on the most recent retry. | 16:41:31 |
| emily | the one that's still failing is main / nix-shell, which is not the ca / nix-shell that was previously failing | 16:41:44 |
 vcunat | Probably. And lix also fails. The reasons probably overlap.
https://hydra.nixos.org/build/318503096 | 16:41:51 |
| emily | so 0 overlap with the listed failures. so it's just very flaky? | 16:41:51 |
| emily | there's one test failing and it's not either of the two listed as failing in the issue report | 16:42:07 |
| emily | I mean, I can hit the restart button… | 16:42:49 |
| vcunat | Feel free. | 16:43:41 |
| vcunat | But the Nix build has been retried several times already. | 16:43:51 |
| vcunat | * But the Nix build has been retried several times already before reporting it. | 16:44:01 |
| vcunat | Lots of llvm builds on darwin now fail this test
LLVM-Unit :: TargetParser/./TargetParserTests/HostTest/getMacOSHostVersion
| 20:15:03 |
| vcunat | (a few thousand jobs killed from that on staging-next-25.11) | 20:15:41 |
| vcunat | And darwin channel would be blocked by cabal which depends on django version which isn't supported anymore (eval error now). | 20:19:13 |
| vcunat | * And darwin channel would be blocked by cabal2nix which depends on django version which isn't supported anymore (eval error now). | 20:19:29 |
 hexa | cabal2nix depends on django???? | 20:42:31 |
| hexa | error: django-4.2.27 not supported for interpreter python3.13
note: trace involved the following derivations:
derivation 'cabal2nix-2.20.1'
derivation 'nix-prefetch-scripts'
derivation 'nix-prefetch-bzr'
derivation 'python3.13-breezy-3.3.12'
derivation 'python3.13-launchpadlib-2.1.0'
derivation 'python3.13-httplib2-0.22.0'
derivation 'python3.13-pytest-randomly-3.13.0'
derivation 'python3.13-factory-boy-3.3.3'
| 20:49:48 |
| hexa | I think you backported the thing | 20:49:55 |
| hexa | which I said I needed to look into but did not | 20:50:10 |
| hexa | Redacted or Malformed Event | 20:51:03 |
| hexa | the reason the disable cannot be backported is that django 4.2 is the default python on 25.11 | 21:12:35 |
| hexa | and I repeat that it is my mistake that I misjudged the version constraints | 21:12:59 |
