Important: Please read the Qt Code of Conduct -

Qt 5.12.2+ sensors can't be used for Android because of QTBUG-77423

  • Hi,
    I updated from Qt 5.12.1 to 5.15.2 and found this error.

    All versions from Qt 5.12.2 upwards are buggy regarding the QRotationReading x,y,z values. Probably the calculation from Quaternion to Euler is somewhere missing or wrong. I can test and confirm it on my app. So all of these versions are not usable for Android!

    It seem so that this bug is fixed with Qt 5.15.3 but this is a commercial lisense, I've only a very low budget and can't spend 500€/year for this bug.

    Here is written that more then 300 bugfixes will come with Qt 6.1, but this version will only appear at the end of April, this is too late for me. ​And it's not clear how stable this version will be.

    Unfortunately in Qt 5.12.1 there is no Android AAB (Android App Bundles) functionality.

    Does someone has an idea what else I can do? Thanks...

  • Based on that QTBUG link you shared, I was able to find a related gerrit patch in the Qt gerrit, but it was abandoned:

    At the end of the gerrit discussion for that patch, the author states:

    I'm closing this patch in favour of
    in hopes of getting it out earlier. 

    ... and I just learned that:

    The Qt 5.15 LTS Commercial source codes are available in the tqtc shadow repositories and their lts branches. The lts branches have a common prefix, "tqtc/lts-". For example, "tqtc/lts-5.15".

    (source of quote: )

    This definitely matches what you (you, @PowerNow ) already stated, namely: the fix is only published into a commercial edition.

    However... I guess one option (though possibly difficult) is to apply the gerrit patch that was made public and compile your own Qt framework with that patch included.

    A similar option (but one that would avoid re-compiling the Qt framework itself) is maybe that you take the wrong x,y,z values that you already have access to... and in whatever slot or handler-function where you see these wrong {x,y,z}s, you could perhaps apply the same logic from the patch but just apply it in your own code?


    void AndroidRotation::dataReceived(const ASensorEvent &event)
        // From android documentation, the rotation sensor values are:
        // values[0]: x*sin(θ/2)
        // values[1]: y*sin(θ/2)
        // values[2]: z*sin(θ/2)
        // values[3]: cos(θ/2)
        // The mathematics below is adapted from
        // and
        // (getRotationMatrixFromVector() followed by getOrientation())
        qreal angles[3];
        qreal q1 = qreal([0]);
        qreal q2 = qreal([1]);
        qreal q3 = qreal([2]);
        qreal q0 = qreal([3]);
        qreal sq_q1 = 2 * q1 * q1;
        qreal sq_q2 = 2 * q2 * q2;
        qreal sq_q3 = 2 * q3 * q3;
        qreal q1_q2 = 2 * q1 * q2;
        qreal q3_q0 = 2 * q3 * q0;
        qreal q1_q3 = 2 * q1 * q3;
        qreal q2_q0 = 2 * q2 * q0;
        qreal q2_q3 = 2 * q2 * q3;
        qreal q1_q0 = 2 * q1 * q0;
        angles[0] = std::atan2((q1_q2 - q3_q0), (1 - sq_q1 - sq_q3));
        angles[1] = std::asin(-(q2_q3 + q1_q0));
        angles[2] = std::atan2(-(q1_q3 - q2_q0), (1 - sq_q1 - sq_q2));
        qreal rz = -qRadiansToDegrees(angles[0]);
        qreal rx = -qRadiansToDegrees(angles[1]);
        qreal ry =  qRadiansToDegrees(angles[2]);
        if (sensor()->skipDuplicates() && qFuzzyCompare(m_reader.x(), rx) &&
                qFuzzyCompare(m_reader.y(), ry) &&
                qFuzzyCompare(m_reader.z(), rz)) {
        m_reader.setTimestamp(uint64_t(event.timestamp / 1000));
        m_reader.setFromEuler(rx, ry, rz);

  • Yeah, but there is a further critical point as described in the link :

    Also the readings are become rough after the update (~2-3 readings per second), before they was much smoother (~20-30 readings per second). And dataRate doesn't help at all...

    I observe this also on some smartphones. With this low dataRate it's unfortunately not possible to realize smooth movements.

  • Wow. I don't know what else to say.

    I am forced to just agree with another comment from that bug ticket:

    Matti Vuori added a comment - 21 Feb '20 12:48

    This is a disgrace. First, a change was made that doesn't work, obviously hasn't ever worked and never will. Clearly it hasn't been tested before including it it 5.12.4. Then there's no attempt to revert it or repair it. Or even discuss it.

Log in to reply