Display tweaks: no animations, BT profile text, 20s slideshow — CI failing #3

Open
opened 2026-08-19 22:06:27 +02:00 by ma3s7r0 · 0 comments
Owner

What

  • Left (nice_epaper): Luna animation off (CONFIG_NICE_OLED_WIDGET_WPM_LUNA=n); active BT profile now shown as text next to the connection icon instead of a barely-visible 3px dot row; the WPM widget's non-Luna gauge/needle fallback dropped (upstream bug — calls cos/sin with no usable declaration on this freestanding libc, so it never compiled until Luna was actually disabled).
  • Right (nice_view_custom): slideshow dwell changed from a hardcoded 10 minutes to 20 seconds. CONFIG_CUSTOM_ANIMATION_SPEED in eyelash_sofle_right.conf was dead — the code never read it.
  • Both display shields are vendored git modules fetched via west (not in-repo), so the fixes live as patches/*.patch, applied at build time in build.sh, flake.nix, and .github/workflows/build-user-config.yml.
  • README corrected: Forgejo Actions does run here (30+ successful runs), contrary to the old "no CI" claim — only the cross-repo draw.yml is actually blocked.

State

  • Branch display-tweaks (commit 2d360cd227f7d50f6c4d686d12bd9c32ddb738bb), pushed to origin, matches local working copy exactly (verified with jj st / jj log -r 'remote_bookmarks()').
  • Local Nix builds pass for all four targets (left, right, studioLeft, settingsReset) — verified with nix build .#<target> --no-link --print-build-logs.
  • Local Docker path (./build.sh) was not exercised this session — the patch-loop was added there by inspection/analogy with the Nix path, not build-tested.

Blocker

Forgejo CI run #116 failed in 1m43s — too fast to be a real firmware build failure (those take 5–12 min elsewhere in the run history), so it likely died in the new "Apply local display patches" step or an earlier matrix step. Log not yet inspected: no Forgejo API token is configured in this environment, and the forgejo MCP tool set has no raw job-log endpoint. Two ways to unblock:

  1. User opens the run URL above and pastes the failing step's output, or
  2. User generates a token (Forgejo web UI → Settings → Applications) and runs tea login add --url https://git.cl12.de --token <token> (tea available via nix run nixpkgs#tea) so logs can be pulled directly.

Next

  • Get the failing step's log (see Blocker) and fix the CI patch step.
  • Build-test ./build.sh at least once (untested this session).
  • Once CI is green, merge display-tweaks into main (needs confirmation first — never merge to main without checking with the user).
## What - **Left** (`nice_epaper`): Luna animation off (`CONFIG_NICE_OLED_WIDGET_WPM_LUNA=n`); active BT profile now shown as text next to the connection icon instead of a barely-visible 3px dot row; the WPM widget's non-Luna gauge/needle fallback dropped (upstream bug — calls `cos`/`sin` with no usable declaration on this freestanding libc, so it never compiled until Luna was actually disabled). - **Right** (`nice_view_custom`): slideshow dwell changed from a hardcoded 10 minutes to 20 seconds. `CONFIG_CUSTOM_ANIMATION_SPEED` in `eyelash_sofle_right.conf` was dead — the code never read it. - Both display shields are vendored git modules fetched via `west` (not in-repo), so the fixes live as `patches/*.patch`, applied at build time in `build.sh`, `flake.nix`, and `.github/workflows/build-user-config.yml`. - README corrected: Forgejo Actions *does* run here (30+ successful runs), contrary to the old "no CI" claim — only the cross-repo `draw.yml` is actually blocked. ## State - Branch `display-tweaks` (commit `2d360cd227f7d50f6c4d686d12bd9c32ddb738bb`), pushed to origin, matches local working copy exactly (verified with `jj st` / `jj log -r 'remote_bookmarks()'`). - Local **Nix** builds pass for all four targets (`left`, `right`, `studioLeft`, `settingsReset`) — verified with `nix build .#<target> --no-link --print-build-logs`. - Local **Docker** path (`./build.sh`) was *not* exercised this session — the patch-loop was added there by inspection/analogy with the Nix path, not build-tested. ## Blocker Forgejo CI run [#116](https://git.cl12.de/ma3s7r0/zmk-sofle-eyelash-config/actions/runs/31) failed in 1m43s — too fast to be a real firmware build failure (those take 5–12 min elsewhere in the run history), so it likely died in the new "Apply local display patches" step or an earlier matrix step. **Log not yet inspected**: no Forgejo API token is configured in this environment, and the `forgejo` MCP tool set has no raw job-log endpoint. Two ways to unblock: 1. User opens the run URL above and pastes the failing step's output, or 2. User generates a token (Forgejo web UI → Settings → Applications) and runs `tea login add --url https://git.cl12.de --token <token>` (`tea` available via `nix run nixpkgs#tea`) so logs can be pulled directly. ## Next - Get the failing step's log (see Blocker) and fix the CI patch step. - Build-test `./build.sh` at least once (untested this session). - Once CI is green, merge `display-tweaks` into `main` (needs confirmation first — never merge to main without checking with the user).
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
ma3s7r0/zmk-sofle-eyelash-config#3
No description provided.