Skip to content

Loopr 1.11

Choose a tag to compare

@PortableDiag PortableDiag released this 17 Aug 01:18
· 15 commits to main since this release

A floating window that goes black is now noticed even when the video is still decoding perfectly.

1.10 watched the frames the player handed to the window. That turns out to be the wrong question: those frames keep flowing into a window that has stopped putting anything on screen — so the black window 1.10 was built to catch was exactly the one it could not see.

  • Each window now also watches frames actually reaching the screen, so the two signals together cover both halves of the failure.
  • New recovery step for that case. Handing the surface back doesn't help a window whose drawing is itself dead, so before giving up Loopr takes the window down and puts it straight back up — rebuilt from scratch, video carrying on from where it was.
  • The diagnostic now says which signal tripped: signal=composite for a dead picture under a healthy player, signal=frames for playback itself stopping. That is the detail that names the underlying bug.
    adb shell setprop log.tag.LooprQueue DEBUG
    adb logcat -d -s LooprQueue | grep -E 'stalled|recovered|rebuilt'
    

Still a safety net, not a root-cause fix. The failure has never been reproduced away from the reporting device — eight scenarios were run against it on an API 35 emulator this session (three windows through a full-screen video, seven trim-memory levels, screen off/on, rotation, Settings on top, the notification shade, and a soak) and none of them produced it. What this release does is close the blind spot that stopped 1.10 from seeing it, and make the next occurrence say what it was.

Verified not to fire on a healthy window across all of the above, on the signed release build from a clean install. All 6 instrumented tests pass.