Yazelix diagnostics
:: yzx status --json
{"schema_version":1,"name":"Yazelix Nova","version":"1.1.0","package":"full","config_home":"/home/dan/.config/yazelix","state_dir":"/home/dan/.local/share/yazelix","shell":"nu","editor_command":"yzx-hx","editor":"/nix/store/27ck90gplyc5hj5vl1qn7lg5fp63pq0q-yzx-hx/bin/yzx-hx","agent_command":"auto","inside_zellij":true}
Details
The stable branch hard-asserts zellijBuildBase.version == "0.44.3", but nixpkgs unstable now ships 0.45.1. Anyone using inputs.nixpkgs.follows = "nixpkgs" in their yazelix input will hit an evaluation failure.
The workaround is to remove the follows statement, but this forces a full separate closure build against yazelix's bundled June nixpkgs, including the kernel, glibc, and everything downstream. On my machine this triggered ~40GB of rebuilds just to get my system to evaluate again.
More broadly, I've now had two issues where the stable branch produces a completely broken config for me without warning. I think the CI tests may not be covering follows-against-recent-unstable, or configs with non-trivial option sets enabled. A test that actually attempts a build with inputs.nixpkgs.follows pointing to a current unstable pin, and with a more representative set of options enabled, would catch both of these before they hit users.
I can't actually build the system now because I have other dependencies that relied on the latest version of nixpkgs unstable.
Yazelix diagnostics
:: yzx status --json
{"schema_version":1,"name":"Yazelix Nova","version":"1.1.0","package":"full","config_home":"/home/dan/.config/yazelix","state_dir":"/home/dan/.local/share/yazelix","shell":"nu","editor_command":"yzx-hx","editor":"/nix/store/27ck90gplyc5hj5vl1qn7lg5fp63pq0q-yzx-hx/bin/yzx-hx","agent_command":"auto","inside_zellij":true}
Details
The stable branch hard-asserts zellijBuildBase.version == "0.44.3", but nixpkgs unstable now ships 0.45.1. Anyone using inputs.nixpkgs.follows = "nixpkgs" in their yazelix input will hit an evaluation failure.
The workaround is to remove the follows statement, but this forces a full separate closure build against yazelix's bundled June nixpkgs, including the kernel, glibc, and everything downstream. On my machine this triggered ~40GB of rebuilds just to get my system to evaluate again.
More broadly, I've now had two issues where the stable branch produces a completely broken config for me without warning. I think the CI tests may not be covering follows-against-recent-unstable, or configs with non-trivial option sets enabled. A test that actually attempts a build with inputs.nixpkgs.follows pointing to a current unstable pin, and with a more representative set of options enabled, would catch both of these before they hit users.
I can't actually build the system now because I have other dependencies that relied on the latest version of nixpkgs unstable.