fix: eighteen validation issues across the plugin fleet

[ChuckBuilds]

Closes 18 of the open validation issues in one branch, so CodeRabbit sees one
diff instead of twenty. Eight commits, each a self-contained theme — reviewing
commit-by-commit is the sane order.

Companion core PR: ChuckBuilds/LEDMatrix#539 (needed for the golden refresh
below; nothing else here depends on it).

What is in it

fix(tests) — #467. Twenty test files across sixteen plugins probed
/home/rackpi/projects/LEDMatrix first, then guessed a sibling checkout, and
inserted whichever hit at sys.path[0] — beating the PYTHONPATH that
run_plugin_tests.py --core sets. The flag whose whole purpose is choosing a
core was ignored whenever either candidate existed, so tests ran against the
wrong checkout and reported pass or fail with equal confidence. They now read
LEDMATRIX_CORE, and the home directory is gone from the repo.

fix(sports) — #414, #447.%-m, %-d and %-I are glibc extensions;
strftime raises ValueError on Windows and musl. Fourteen call sites across
ten plugins. In baseball-scoreboard's scroll renderer the enclosing except
swallowed it, dropped the start time and drew a raw ISO date instead — which is
what made three upcoming goldens the only cards failing
test_scroll_card_renders.py off glibc. Output is byte-identical on glibc.

fix(scoreboards) — #427, #429, #426, #422, #432, #435, #373. Seven
settings-level defects of one shape: the schema and the web UI promise
something the runtime never reads, or reads with a different default. The two
worth calling out:

• soccer's leagues.<slug>.display_options was read by nothing — 30 settings in
  the UI doing nothing, and the opposite precedence to every sibling. The
  league block wins now. Caveat the issue did not have: the web UI writes a
  schema default into every league block, so an install that set the root
  show_records/show_ranking away from its default and left the league
  blocks alone will see the league value win after this. The README says so.
• afl and nrl's scroll_settings block was unreachable from either name
  (additionalProperties: false forbids writing scroll_mode, which is what
  the code read). Making it reachable also required moving scroll_speed's
  default 1.0 → 50.0, because the block had been shadowing the core default —
  without that, every afl/nrl panel would have slowed to a crawl on upgrade.

On #373's background_service.max_workers: removing it would have been wrong.
The service is a process-wide singleton, so a per-plugin worker count cannot
mean what it says — the first plugin to start decides for everyone. It is wired
through instead, its default corrected 3 → 1 to match the behaviour every
install has actually had, and its description now says so.

fix(static-image, flights) — #372, #400.date_based rotation
implemented (day-of-year, the indexing of-the-day uses). skyaware_url no
longer defaults to 192.168.86.30 — one receiver on one home network, which
every unconfigured install polled on its own LAN, and which cost the harness
~40s of connect timeouts per run.

fix(fonts, stocks) — #387, #395. Six more plugins retry a .bdf at the
size its PIXEL_SIZE header declares instead of silently substituting another
face (twelve of thirteen offered sizes failed before). ledmatrix-stocks gets
the test/harness.json it never had — all eight sizes were passing against a
blank frame, so nothing about its ticker layout, fonts, colours or inline chart
was being checked.

feat!: remove seventeen settings — #381, #377, #418, #406. Per your call.
Breaking: all six plugins set additionalProperties: false, so a config.json
still carrying a removed key fails validation until it is saved again — the
plugin still loads, the UI flags the config. Majors on all six.

fix(render) — #383 only. Both documentation renderers picked a caption
face from a list of host font paths, and the pomodoro one listed no Windows
path at all, so a Windows rerun fell through to PIL's default face and rewrote
three of four contact sheets on a plugin nobody had changed. Both use
ImageFont.load_default(size=...) now — bundled Aileron, no host lookup, no
new files. Pomodoro's four images are regenerated and reproduce byte-for-byte
across runs. The 217 composites under docs/assets/ are deliberately left
alone: their captions change on the next regeneration of each, better done per
plugin than as one commit.

chore(store) — 20 manifests bumped, plugins.json regenerated.

#371, #375, #378, #391 are diagnosed but not closed here

The cause is the text layout engine, not Pillow. ImageFont.truetype picks it
at load time — Raqm where the host Pillow has libraqm, Basic otherwise — and
the two round fractional glyph advances differently. PressStart2P at 8px has
whole-pixel advances so both agree; 4x6-font.ttf at 6px does not. The four
drifting plugins are exactly the four drawing body text in the 4x6 face, and
#371 recorded the tell without naming it: of-the-day's PressStart2P title is
pixel-identical while its 4x6 body is not.

The core half is ChuckBuilds/LEDMatrix#539,
which pins the engine and is green.

I had the golden refresh in this branch and took it back out, because this
runner has libraqm and my machine does not — so CI is the only place that can
settle it, and it said the refresh was not finished. Even with the engine
pinned plugin-side, geochron still drifted here by 135px. Regenerating goldens
from Windows is guessing; they want a --update-golden run on this runner,
after #539 lands. The four plugins are byte-identical to main in this branch,
so nothing here blocks on that.

Checks

• scripts/check_plugin.py — 32/32 plugins, 0 failures, against #539.
• scripts/check_module_collisions.py — clean across 43 plugins.
• scripts/check_manifest_version_fields.py — clean across the 23 bumped.
• scripts/run_plugin_tests.py --all — 245 passed, 1 skipped, 2 failed.

The two failures are pre-existing on main and reproduce identically
there, unchanged by this branch:

• baseball-scoreboard/test_odds_placement.py — 2 checks, both odds y-placement
  (stays at the top when the O/U is short: y=8).
• ledmatrix-weather/test_almanac_layout.py — 1 check
  (time-mode 128x64: expected full am/pm on 128-wide, got compact).

CI on this branch is green — safety, collisions and Codacy all pass.

I have not touched them; say the word if you want them in scope.

Not closed

• soccer-scoreboard: manifest declares soccer_live/recent/upcoming but runtime registers 21 league-qualified modes — none of the declared names exist #461, Three plugins are missing from the root README's Available Plugins listing #385, odds-ticker: display_options.scroll_speed/scroll_delay are silently ignored; frame-based mode is unreachable #408, ledmatrix-elections: test_mode fixture load fails on Windows (open() without an encoding) #440, ufc-scoreboard: every upcoming fight card is blank in switch mode (KeyError on a font that does not exist) #437, olympics: countdown goes negative after the Games, and the label is clipped below 256px #410 — already fixed on main; I verified
  each against the current tree rather than assuming. Worth closing.
• Starlark Plugin #291 — fixed in the core by #535 (fix(starlark): restore the API routes #330 dropped), which is on core main now.
• [of-the-day] Golden images drift on all eight harness sizes #371, [geochron] Overflow at 128x32 and golden drift on six sizes #375, [christmas-countdown] Golden images drift on seven of eight harness sizes #378, ledmatrix-weather: almanac goldens drift on main at four panel sizes #391 — diagnosed above; the core half is #539 and the golden refresh wants a Linux run.
• get_font_height() returns 0 for the 5x7 BDF face, so BDF-based row stacking collapses #397 — the fix is in the core, in #539. I checked the three other plugins
  that issue flagged: calendar measures with textbbox itself, clock-simple
  and ledmatrix-flights never hand a raw freetype.Face to
  get_font_height(). ledmatrix-music was the only one exposed and was
  already guarded.

🤖 Generated with Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 9a4d74b4-1081-4e8e-85be-e756977920da

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[codacy-production]

Up to standards ✅

🟢 Issues0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics38 complexity

Metric	Results
Complexity	38

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

[ChuckBuilds]

The safety-harness failure is the expected ordering, plus one real finding

CI's harness checks the branch out against core main, which does not yet
carry the layout-engine pin from ChuckBuilds/LEDMatrix#539.
So the four refreshed golden sets — christmas-countdown, geochron,
ledmatrix-weather, of-the-day, and no others — drift there. #539 has to merge
first; nothing else in this PR depends on it.

That run was also the first real test of the diagnosis, because this runner has
libraqm and my machine does not. It confirmed the layout engine as the cause
and surfaced a second, smaller one:

• christmas-countdown, of-the-day, weather draw through the core's
  draw_text, so their font loads are core-owned. Unpinned core → drift.
  Pinned core → they match. Nothing more to do here.
• geochron loads its own face, so this branch already pinned it — and it
  still drifted in CI by 52–135px at max Δ=60–83. Low deltas mean edge
  coverage, not glyph position: the sidebar was drawn anti-aliased, and
  coverage varies with the host FreeType build even at the same Pillow
  version.

Fixed in f906a93: the sidebar draws 1-bit for the duration of the readout,
restoring the previous fontmode so nothing else on that shared canvas
changes. That is what an LED grid wants anyway — a half-lit pixel is a dim
pixel, not a smooth edge. The sidebar golden is now exactly three colours,
background plus the two text colours, which leaves nothing for a FreeType
build difference to disagree about. Goldens regenerated.

I would expect geochron to go green on this run even against the unpinned core,
and the other three to follow once #539 lands. If geochron still drifts here,
that is worth knowing and I will keep digging.

[ChuckBuilds]

One more caught on a merge-safety pass — 91b7982

Re-reading #422 rather than trusting green CI, because "a dead config block
becomes live" means every key in it becomes an override, and one of them
disagreed with what the code had been resolving.

Both afl and nrl set gap_between_games: 24 in scroll_settings_defaults(),
overriding the core's 48. The schema block declared 48. So the fix for #422
would have silently doubled the gap between game cards on every afl and nrl
install — same trap as scroll_speed, which I only caught because its factor
was 50x rather than 2x. No test covers it and the harness renders frame 0,
where the gap does not show, so CI would not have told anyone.

The schema declares 24 now, and every key in the block agrees with the value
the resolver produced while it was unreachable:

key	schema	resolved before
scroll_speed	50.0	50.0	core default
scroll_delay	0.01	0.01	core default
gap_between_games	24	24	plugin override — this commit
show_league_separators	true	true	core default
dynamic_duration	true	true	core default
game_card_width	128	computed	already guarded as "unchosen"

game_card_width needed nothing: both plugins already treat a value equal to
the schema default as unchosen and substitute their computed width — the same
guard this needed and did not have.

Reachability is behaviour-neutral for a default config now, which is what #422
should have been all along. READMEs corrected, harness and unit tests green for
both plugins.