lament/pantheon

builder changes, and niri fixes #100

Merged

lament merged 1 commit from niri-fixes into main

2026-07-19 21:20:23 -05:00

lament commented

2026-07-19 21:08:55 -05:00

Owner

builders-use-substitutes = true fleet-wide - brigid is the build path, so let it substitute build inputs from its own caches instead of clients shipping closures over WG
- set in buildMachines.nix via mkMerge, alongside the existing builder-only keep-outputs/keep-derivations block
stylix.targets.niri.enable = false on lament - stylix auto-enabled its niri target the moment niri came on and had HM generate its own ~/.config/niri/config.kdl (themed, but zero binds and no spawn-at-startup), silently clobbering the raw config so nothing worked with no parse error
- phase 0 keeps niri config RAW for live hot-reload iteration, so stylix must not own it; re-enable (or hand theming to noctalia) when niri codifies into HM in phase 1
programs.niri.package -> niri-unstable on ishtar - noctalia v5 themes niri by writing noctalia.kdl + appending include "noctalia.kdl" (focus/border colors, recent-windows), and those nodes only parse on niri git main, not niri-flake's stable 25.08 (the mismatch silently broke every config reload)
- referenced as inputs.niri.packages.${pkgs.system}.niri-unstable directly - the niri overlay isn't applied here, so pkgs.niri-unstable doesn't exist
- trade-off: niri now builds from git-pinned deps (smithay et al.) instead of a locked release; still cached on niri.cachix.org
left for later: the raw config.kdl also needed a noctalia v5 IPC fix (noctalia msg panel-toggle ..., not v4's noctalia-shell ipc call ...) plus the monitor-layout swap - both live on ishtar's /home, not in-repo until phase 1

- `builders-use-substitutes = true` fleet-wide - `brigid` is the build path, so let it substitute build inputs from its own caches instead of clients shipping closures over WG - set in `buildMachines.nix` via `mkMerge`, alongside the existing builder-only `keep-outputs`/`keep-derivations` block - `stylix.targets.niri.enable = false` on `lament` - stylix auto-enabled its niri target the moment niri came on and had HM generate its own `~/.config/niri/config.kdl` (themed, but zero binds and no `spawn-at-startup`), silently clobbering the raw config so nothing worked with no parse error - phase 0 keeps niri config RAW for live hot-reload iteration, so stylix must not own it; re-enable (or hand theming to noctalia) when niri codifies into HM in phase 1 - `programs.niri.package` -> `niri-unstable` on `ishtar` - noctalia v5 themes niri by writing `noctalia.kdl` + appending `include "noctalia.kdl"` (focus/border colors, `recent-windows`), and those nodes only parse on niri git main, not niri-flake's stable 25.08 (the mismatch silently broke every config reload) - referenced as `inputs.niri.packages.${pkgs.system}.niri-unstable` directly - the niri overlay isn't applied here, so `pkgs.niri-unstable` doesn't exist - trade-off: niri now builds from git-pinned deps (`smithay` et al.) instead of a locked release; still cached on `niri.cachix.org` - left for later: the raw `config.kdl` also needed a noctalia **v5** IPC fix (`noctalia msg panel-toggle ...`, not v4's `noctalia-shell ipc call ...`) plus the monitor-layout swap - both live on `ishtar`'s `/home`, not in-repo until phase 1

lament added 1 commit

2026-07-19 21:08:55 -05:00

builder changes, and niri fixes

CI / check (pull_request) Successful in 1m13s

Details

CI / deploy (pull_request) Has been skipped

Details

CI / dry-activate (pull_request) Has been cancelled

Details

CI / build (pull_request) Has been cancelled

Details

8aa8754a46