Grant,
I figure I best describe what exactly my application is doing in case my interpretations are wrong. After due initialization the application calls tobii_get_display_area() to get screen corners coordinates and derive tracker-to-world transformation from these. Without an initial calibration the call reports 4×4 screen at z = 0 (i.e. top left corner is at [-2, 2, 0]). If I run ‘tobii_config’ applet and complete the screen calibration the screen position report changes to some sane though not too precise numbers. My application is mostly interested in relative motion rather than in absolute head position so some inaccuracy is no problem. The screen position report remains the same after the application and even ‘tobiiusbserviced’ service restart however if I disconnect and connect again the tracker or power cycle the whole hosting system the screen position report reverts to the initial 4×4 state which leads me to believe the calibration is actually stored in tracker’s volatile memory.
There’s no error reported by the API (every call is checked). I didn’t look deep into report quality but visually it seems OK, I mean at least the report appears to follow head movements as expected. I didn’t try on Windows yet but I might in future, some code maintenance will be needed for that.
Is ‘tobiiusbserviced’ service expected to provide screen calibration persistency across power cycles somehow? Admittedly I might have skipped something while adapting the original Debian package for my system. Does it expect some extras (i.e. running GVfs) which may be lacking on my rather bare bone installation?