refactor(hud): event-driven PostMessage architecture for stability (v0.2.31)#12
Merged
refactor(hud): event-driven PostMessage architecture for stability (v0.2.31)#12
Conversation
… upgrade HUD failure logs to WARNING
- ui.py: Remove redundant threading.Thread(target=self.indicator.start) call in
TrayApp.run(). _ensure_indicator() in __init__ already started the indicator.
The second thread could race with the first before _is_running was set, creating
two concurrent Win32 message pumps on the same HWND and silently killing the HUD.
- hud_renderer.py: Remove orphaned ('VISIBILITY', ...) puts from show() and hide().
ShowWindow() already handles visibility directly; the queue item was never consumed
by _wnd_proc and served no purpose.
- hud_renderer.py: Upgrade 'HUD Run loop exited.' log from INFO to WARNING so HUD
thread death is clearly visible at INFO log level.
- indicator_ui.py: Upgrade HudOverlay fallback log messages from INFO to WARNING so
silent degradation to GdiFallbackWindow is immediately visible in the log.
After sleep/lock/unlock, the layered window bitmap goes stale and the Win32 timer (WM_TIMER) can die. ShowWindow makes the HWND visible but with no fresh content, leaving it transparent. Fix: In show(), queue a REDRAW event to force changed=True on the next timer tick, and re-register the 50ms SetTimer to resurrect it if it died during a long session lock. The previous VISIBILITY queue items served this same purpose (forcing changed=True) but were removed in the prior commit under the assumption they were dead code. Also stores _refresh_rate_ms as an instance variable so show() can access it for the SetTimer call.
Added targeted diagnostics to HudOverlay to pinpoint why the window remains invisible after long sleep/lock cycles: - Log GetLastError() if UpdateLayeredWindow fails (Fix A) - Log HudOverlay.show() calls at DEBUG level (Fix B) - Log WM_TIMER periodic heartbeat every 60s at DEBUG level (Fix C)
- show(): log at INFO with immediate IsWindowVisible() result after ShowWindow to confirm the Win32 call's effect at INFO log level. - hide(): log at INFO so hide/show pairs are traceable in the log. - WM_TIMER heartbeat: upgrade from DEBUG to INFO so we can confirm the message loop is alive in normal (non-verbose) operation.
Root cause confirmed by logs (2026-03-26 10:30): show() called at 10:30:26 hide() called at 10:30:27 (linger A fires, same session_id) hide() called at 10:30:28 (linger B fires, same session_id) Multiple callers of _start_linger_timer() (click-away, CONNECTING state, update_status_icon) all captured the same _session_id=N, so every one of their timers would fire and hide the HUD — even immediately after a valid show() call for the new reconnect. Fix: _start_linger_timer() now increments _session_id BEFORE capturing it as session_at_start. Each linger timer therefore gets a unique, monotonically increasing ID. Any earlier pending timer sees a stale ID and skips the hide(). Only the most recently scheduled linger can fire.
…namic repositioning
…r, hasattr guard)
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.
his production-ready PR integrates the comprehensive HUD stability refactor with recent updates to the application's evaluation framework.
🚀 Key Changes
The core of this PR is the transition from constant polling to a thread-safe Win32 PostMessage architecture.
Dynamic Repositioning: Resolves the "ghost window" bug by recalculating HUD coordinates on every WM_APP_SHOW event, ensuring it stays correctly anchored across sleep/wake cycles.
Thread-Safety: All UI state updates (text, status, provider) now occur on the HUD thread via async messaging, eliminating 64-bit Win32 race conditions.
Zero Idle CPU: The heartbeat timer is now only active when the HUD is visible.
2. Evaluation Framework Updates
Spec Improvements: Updates to conductor/archive/ reflects progress in the Arabic and Multilingual accuracy evaluation specs.
Data Persistence: Includes recent transcription outputs in eval_results/ for baseline comparisons.
3. Documentation & Infrastructure
Architecture: architecture.md has been updated to reflect the new HUD design.
Version Bump: Atomic synchronization of pyproject.toml, version_info.txt, uv.lock, and parrotink.iss to version 0.2.31.
✅ Verification
Passed full Definition of Done Gate (Ruff, MyPy, Pytest).
Verified local build and Inno Setup integration.
Confirmed zero HUD idle CPU footprint.