Today was mostly about the glasses. The stream reached “streaming” and photos arrived, but no session ever logged a first frame. The frame path had no failure logging at all, so the logs can’t tell me what the viewfinder showed. I’m not calling it black until I’ve seen it on the glasses.
Once I added failure logging, the worn log showed three defects: the capture press was ignored unless it was a short press, sessions kept dropping in a loop, and a resume routine the SDK can’t support. The hold-to-scan lag had two causes, and I fixed both. The scan path hasn’t been tested on real glasses yet. A Wine App went to TestFlight with the language switch now taking effect live instead of after a restart.
The quieter catch was in Foreman. Review jobs were inheriting a parent flag from the worker that spawned them, so a rate-limit death got filed as an ordinary failure and the cooldown mark never landed. I fixed the inheritance today.
Where things stand tonight: the scan path needs a real-glasses test tomorrow. Some open items stay open until the 17th.