The console's own settings never left the console, and the layer under a card missed its corners #362

Merged

enricobuehler merged 2 commits from worktree-android-console-extra-keys into main 2026-08-20 20:03:48 +00:00

Conversation 0 Commits 2 Files Changed 3

+111 -21

enricobuehler commented 2026-08-20 19:52:36 +00:00

Owner

Copy Link

Copy Source

Reported on glass: the "Controller-optimized UI" switch is there, turning it off does nothing — the console stays up.

The switch

pf_client_core::trust::Settings::extra is #[serde(flatten)], so the android.* console settings are top-level keys of the settings document, beside width and codec. There is no "extra" object — trust.rs's own settings_unknown_keys_survive_round_trip already asserts a serialized default contains no such key.

ConsoleJson wrote and read them nested under one. Serde put that whole object into the map under the literal key "extra", so:

• no console row ever found android.gamepad_ui_enabled — each showed its compiled default, whatever the touch UI had stored;
• the value the console saved came back to Kotlin as the one Kotlin had just sent, so applySettings saw no change, raised no onSettingsChange, and App never recomputed gamepadUiActive.

Hence a switch you can flip with nothing behind it.

It hit every Android-only row, not just this one: low latency, phone rumble, phone gyro, SC2 capture, DualSense capture, and the Show it mode picker. None of them round-tripped from the console.

A store written by the nesting build carries the dead wrapper; it is dropped on the next write rather than echoed for the life of the install.

Why nothing caught it

Both halves agreed on the same wrong nesting, so a write-then-read round-trip passes while the feature is broken — theRoundTripKeepsEveryAndroidRow in the new test still passes with the fix reverted. The other three cases assert the shape (android.… present at the top level, no "extra" wrapper, a console-shaped save read back) and all three fail without it.

The layer under the cards

Also reported: the layer behind every card, the one that reads as an outline, does not follow the card's corners.

That is theme::focus_halo. It grows the card's rect by 4 design units on every side but drew it with the card's own radius. A shape grown by d keeps its corners parallel to the original's only if its radius grows by d too — otherwise the two arcs stop sharing a centre. So the halo came out squarer than the card: clean along the edges, visibly off at the four corners.

Same rule applied to panel_highlight, which pulls in half a unit and kept the full radius. drop_shadow only offsets, so its geometry was already right, and the collections plate uses RRect::with_outset, which adjusts the radii itself.

Every card in the console goes through these two helpers — home tiles, library grid, coverflow, collections deck.

Verification

• 98 app unit tests green; the new ConsoleSettingsExtraTest fails 3 of 4 with the fix stashed.
• cargo fmt --check -p pf-console-ui clean.
• Native compiles for arm64-v8a, armeabi-v7a and x86_64 (:kit:cargoNdkRelease).

⏳ On glass owed — nobody has yet watched the console actually drop to the touch UI, and the halo change is a visual call worth eyeballing.

Reported on glass: the "Controller-optimized UI" switch is there, turning it off does nothing — the console stays up. ## The switch `pf_client_core::trust::Settings::extra` is `#[serde(flatten)]`, so the `android.*` console settings are **top-level** keys of the settings document, beside `width` and `codec`. There is no `"extra"` object — `trust.rs`'s own `settings_unknown_keys_survive_round_trip` already asserts a serialized default contains no such key. `ConsoleJson` wrote and read them nested under one. Serde put that whole object into the map under the literal key `"extra"`, so: - no console row ever found `android.gamepad_ui_enabled` — each showed its compiled default, whatever the touch UI had stored; - the value the console saved came back to Kotlin as the one Kotlin had just sent, so `applySettings` saw no change, raised no `onSettingsChange`, and `App` never recomputed `gamepadUiActive`. Hence a switch you can flip with nothing behind it. It hit **every** Android-only row, not just this one: low latency, phone rumble, phone gyro, SC2 capture, DualSense capture, and the `Show it` mode picker. None of them round-tripped from the console. A store written by the nesting build carries the dead wrapper; it is dropped on the next write rather than echoed for the life of the install. ### Why nothing caught it Both halves agreed on the same wrong nesting, so a write-then-read round-trip passes while the feature is broken — `theRoundTripKeepsEveryAndroidRow` in the new test still passes with the fix reverted. The other three cases assert the **shape** (`android.…` present at the top level, no `"extra"` wrapper, a console-shaped save read back) and all three fail without it. ## The layer under the cards Also reported: the layer behind every card, the one that reads as an outline, does not follow the card's corners. That is `theme::focus_halo`. It grows the card's rect by 4 design units on every side but drew it with the card's *own* radius. A shape grown by `d` keeps its corners parallel to the original's only if its radius grows by `d` too — otherwise the two arcs stop sharing a centre. So the halo came out squarer than the card: clean along the edges, visibly off at the four corners. Same rule applied to `panel_highlight`, which pulls in half a unit and kept the full radius. `drop_shadow` only offsets, so its geometry was already right, and the collections plate uses `RRect::with_outset`, which adjusts the radii itself. Every card in the console goes through these two helpers — home tiles, library grid, coverflow, collections deck. ## Verification - 98 app unit tests green; the new `ConsoleSettingsExtraTest` fails 3 of 4 with the fix stashed. - `cargo fmt --check -p pf-console-ui` clean. - Native compiles for arm64-v8a, armeabi-v7a and x86_64 (`:kit:cargoNdkRelease`). ⏳ **On glass owed** — nobody has yet watched the console actually drop to the touch UI, and the halo change is a visual call worth eyeballing.

enricobuehler added 2 commits 2026-08-20 19:52:39 +00:00

fix(clients/android): the console's Android rows were nested where serde flattens them ... cbd3d02817

Turning "Controller-optimized UI" off in the console did nothing: the console
stayed up, because the setting never left the console.

`trust::Settings::extra` is `#[serde(flatten)]`, so the `android.*` keys are
TOP-LEVEL keys of the settings document, beside `width` and `codec`.
`ConsoleJson` wrote and read them nested under an `"extra"` object instead.
Serde put that whole object into the map under the literal key `"extra"`, so
no console row ever found `android.gamepad_ui_enabled` — the row showed its
own default, and the value the console saved came back to Kotlin as the one
Kotlin had just sent. `applySettings` then saw no change, raised no callback,
and `App` never recomputed `gamepadUiActive`.

Every Android-only row rode the same broken path: low latency, phone
rumble/gyro, SC2 and DualSense capture, and the console-UI mode picker.

A store written by the nesting build carries the dead wrapper; it is dropped
on the next write rather than echoed for the life of the install.

The new test pins the shape from both sides. A round-trip alone could not have
caught this — both halves agreed on the same wrong nesting, which is exactly
how it survived review.

fix(console-ui): the layer under a card takes the card's corner plus its own outset ... c9a76287d8

The focus halo grows the card's rect by 4 design units on every side but drew
it with the card's own corner radius. A shape grown by `d` keeps its corners
parallel to the original's only if its radius grows by `d` too — otherwise the
two arcs stop sharing a centre. So the halo came out squarer than the card it
sits under: clean along the edges, visibly misaligned at the four corners,
where it read as a badly drawn outline rather than as light spilling out.

Same rule applied to `panel_highlight`, which pulls in half a unit and kept the
full radius. `drop_shadow` only offsets, so its geometry was already right, and
the collections plate uses `RRect::with_outset`, which adjusts the radii itself.

Every card in the console goes through these two helpers — the home tiles, the
library grid, the coverflow, the collections deck.

enricobuehler merged commit 1c60e641b3 into main 2026-08-20 20:03:48 +00:00

enricobuehler referenced this issue from a commit 2026-08-20 20:37:13 +00:00

release: 0.31.1 — fold in #362, which makes the console's off switch work

enricobuehler referenced this pull request 2026-08-20 20:37:28 +00:00

0.31.1 notes: fold in #362, which is what makes the console's off switch work #364

enricobuehler referenced this issue from a commit 2026-08-20 20:38:05 +00:00

Merge pull request '0.31.1 notes: fold in #362, which is what makes the console's off switch work' (#364) from worktree-release-0311-foldin into main

enricobuehler referenced this issue from a commit 2026-08-22 22:54:42 +00:00

release: 0.31.3 — version bump, notes, CHANGELOG, Play notes

Sign in to join this conversation.

Reviewers

No Reviewers

Labels

No items

No labels

Milestone

No items

No Milestone

Projects

Clear projects

No projects

Assignees

Clear assignees

FernandoMX FlipperDan HitFrostbite bluelightspecial enricobuehler (Enrico Bühler) kszlim luxus nannku neoney

No Assignees

1 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: unom/punktfunk#362