Created attachment 7946 [details] PyQt5 test application Hello, I compiled Qt5 with Tslib support (BR2_PACKAGE_QT5BASE_TSLIB=y), and ts_calibrate and ts_test work properly. I also have a /usr/lib/qt/plugins/generic/libqtslibplugin.so, so the compilation part seems to have worked. I created a tiny PyQt5 test application (see attachment), which creates 4 buttons. When I run this qt5test.py, the y-axis of the touchscreen appears inverted (x-axis ok). I've read http://lists.busybox.net/pipermail/buildroot/2015-November/143571.html and from that it seems that my PyQt5 application uses evdev instead of tslib for input. I can use an extra environment variable: ~> QT_QPA_EVDEV_TOUCHSCREEN_PARAMETERS='rotate=180' ./qt5test.py Now the touchscreen is indeed rotated by 180 degrees, i.e. the x-axis is inverted while the y-axis is ok. Further trials with 'rotate=90' or 'rotate=270' changed the touchscreen orientation accordingly, but none is usable to cancel the inversion of the y-axis. What strikes me is that the above is an **evdev** option, despite the fact that I want to use Qt5 with **tslib**. I've read about a "-plugin tslib" command-line option, but that changes nothing with my Python script. So I'm going to need either a way to switch to tslib (preferred), or a QT_QPA_EVDEV_TOUCHSCREEN_PARAMETERS value that just flips the y-axis.^ Any help or info will be really appreciated. Best regards, Tibor Stolz
Hello, I found <http://doc.qt.io/qt-5/embedded-linux.html#touch>, which mentions some more settings to be done by means of environment variables: ~> QT_QPA_EVDEV_TOUCHSCREEN_PARAMETERS='inverty' ./qt5test.py This works for me, I get proper touchscreen input. (Hardware: BeagleBoneBlack with 4D Systems gen4-4DCAPE-43T LCD) There are two options to enable tslib: ~> QT_QPA_FB_TSLIB=1 ./qt5test.py This seems to completely disable touchscreen input. ~> QT_QPA_EGLFS_TSLIB=1 ./qt5test.py This seems to have no effect, i.e. touchscreen input works but is still y-wise inverted. Since there is no need for tslib in my configuration, as y-wise inversion can be done without it, I declare this bug as RESOLVED WORKSFORME. Best regards, Tibor Stolz