# Lappy v0.3.3 - one touch, one action

The v0.3.2 device test exposed repeated actions from a single short-to-medium held touch. Faster polling could expose brief empty contact reports as releases, allowing the same finger to start another gesture. v0.3.3 retains the fast background touch worker but requires 60 ms of clear contact before accepting a normal lift. Position updates remain immediate while pressed. A stale sample or I2C failure cancels the contact and requires a confirmed lift before that finger can start another gesture. Release coordinates retain the last real position.

This applies to the input stream, rather than rate-limiting individual buttons. Ordinary separate taps and swipes remain supported. Diagnostics now report raw contact edges and accepted press/release counts so the device check can distinguish noisy reports from multiple accepted gestures. Native LVGL tests exercise the actual Drag ARM ALL/DISARM ALL control through one-second held gestures with intermittent empty reports and require one action per gesture.

The hardware test also identified costly full-frame rotation. Rotation now uses 8x8 tiles to reduce strided PSRAM accesses. The result is byte-identical to the previous pixel mapping: independent tests cover full frames and 1,000 arbitrary dirty rectangles in both orientations. The LCD command/transfer sequence, source artwork, 16 fps animation source and timing algorithms are unchanged. This is a memory-access optimization, not an assertion that every hardware scene achieves 16 fps.

All host regressions, native interactions and retained-widget benchmarks, production display/touch harnesses, target compilation, linked BLE audits and renderer comparisons pass. 178/179 screens are identical to v0.3.2; only the version caption differs. Nine native course files, saved-data formats, timing/BLE/storage services, power policy, bootloader and SDK/partition configuration are preserved.

A correction candidate was installed by an application-only write on the connected V2 board, with independent flash verification and preservation of bootloader/table/NVS/PHY/existing crash evidence. A full 16 MiB pre-update backup is retained privately. The owner requested testing without Dragy, so live receiver behavior and physical timing accuracy are not accepted by this USB bench. Hands-on touch confirmation and the measured scene/performance scope belong in the matching hardware report; software tests alone do not establish that all device functions work perfectly. Existing v0.3.2 release files remain immutable.