SUMMARY STEPS TO REPRODUCE 1. Launch PHD2, no equipment connection 2. Launch Ekos with Telescope and CCD Simulators, PHD2 guider 3. Remain on NCP, start guiding OBSERVED RESULT Guiding at NCP fails because PHD2 doesn't see the guide star move enough. If the guide star doesn't move at all (0 pixel offset), PHD2 will abort calibration without sending a message. In that situation, Ekos remains waiting for the end of the procedure. PHD2 is in looping mode when it abandons calibration without notice. Therefore Ekos cannot resume guiding anymore because guiding frames are being received. The UI prevents stopping because guiding is not actually active. The state machine is lost and situation cannot be recovered, even with a disconnect/reconnect. This issue, while very specific, shows that there are holes in the guiding faut tolerance. EXPECTED RESULT There should be a timeout to waiting for the calibration procedure to end. By default, not considering frame download, it takes PHD2/Simulators ~25s to calibrate when everything works, and ~75s to fail if the star doesn't move enough (there is no failure if the star doesn't move at all). Ekos should poll for PHD2 state and execute proper state machine recovery. SOFTWARE/OS VERSIONS Windows: macOS: Linux/KDE Plasma: Ubuntu 18.04 (available in About System) KDE Plasma Version: KDE Frameworks Version: Qt Version: 5.12.5 ADDITIONAL INFORMATION
I can confirm this issue but it has only just started happening with a very recent build.
This is fixed by 11130020, using the calibration timeout. This behaviour is tested in test_ekos_guide, using Simulators and Polaris as guide star. The Simulator does not move Polaris far enough for PHD2 to detect a movement, and PHD2 switches from calibrating to looping without sending a notification (failed guarantee). The calibration timeout of the Guide Module catches this situation. Note the workaround also works for guide stars that are hot pixels.