Repository navigation
Conversation
A full day of remote-driven dictation surfaced a set of issues in the shipping V2.0.0 build; this is the set of fixes for them. Performance: - Remove PaintWorkspaceTexture; a scroll repaint went from 227 ms to 1.74 ms. - Cache the log tail, the process snapshot, the bridge and capture health JSON, and the session marker scan in PollActivity; route label changes through SetLabelText/SetLabelColor and pause the poll during WM_ENTERSIZEMOVE. A poll went from 22 ms to about 10.5 ms, and the UI thread from 5.4% to 2.4% of one core. Behaviour: - Stop the Live HUD from presenting itself: liveHudAutoShow defaults to false and the settings page exposes it as its own checkbox. The tray item still shows it on demand. - Keep the bridge state and health files under the configured UserData root and record last_raw_press for diagnostics. - Judge onboarding step 4 freshness by the completion receipt timestamp so the step can be finished (issue richlearntodo-debug#9). - Sync the confirmed shortcut revision to the bridge after the gesture layer saves. - Stop the voice wake false positive that armed on a stale observation (issue richlearntodo-debug#10). - Write the focus lock line only when the state actually changes instead of once per poll. Verified: BUILD_DEVELOPMENT.ps1, node scripts/validate.js, the V2 focused feature suite, and the host self-test all pass; the frozen capture stays at B62DE035A9CAD0A16B97F6935C6E4DE0BF2B73C61B180595482D852C0582E683.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
背景
我把这套遥控器当主力输入工具在用（每天大量口述），V2.0.0 用了几天后攒下一批问题，已经在 #9、#10 里报过。这个 PR 是把改好的版本回馈给上游。
改动基于
main（1b56c2f)之上，1 个提交，6 个文件。改了什么
性能（重点）
PaintWorkspaceTexture：一次滚动重绘 227 ms → 1.74 ms。PollActivity做缓存：日志尾、进程表快照、桥与采集的健康 JSON、会话标记扫描都改成按版本号缓存；文本/颜色更新走SetLabelText/SetLabelColor；拖动窗口（WM_ENTERSIZEMOVE）期间暂停轮询。单次轮询 22 ms → 约 10.5 ms，界面线程 5.4% → 2.4% 单核。空闲时整机占用随之下降。行为
liveHudAutoShow，默认 false（对老用户保持"关闭"这个安全默认），设置页"交互反馈"里可开启；托盘菜单里的"显示 Live HUD"不受影响，仍可手动呼出。last_raw_press便于诊断。没有动的东西（按仓库 AGENTS.md 的约束）
VibeMicAtvvCapture.exe校验值保持B62DE035A9CAD0A16B97F6935C6E4DE0BF2B73C61B180595482D852C0582E683不变。验证
真机信息：Windows 11（build 26200）+ 小米 RC003，ATVV，用遥控器麦克风，hold-to-talk；音频出口是 VB-CABLE（
CABLE Input）。说明
docs/images下的六个截图我没有重新生成（没跑capture-ui-screenshots.ps1）。需要的话我可以补一版。English summary
Fixes a batch of issues that a daily-use session surfaced in V2.0.0 (originally reported in #9 and #10): idle CPU load, a Live HUD that popped up unasked, onboarding step 4 that could not be completed after a restart, a voice-wake false positive, and repeated focus log lines.
Behaviour is unchanged when the new option is off. No changes to Raw Input, the keyboard hook, device filtering, gesture constants, or the frozen Capture component; the new configuration field defaults to off, so existing configurations keep working.