QTransform::transposed() creates a transposed matrix, but doesn't set the dirty flag.

Because of this, a subsequent call to type() returns the type of the original transformation, which will mostly be wrong:

QTransform t;
t.translate(5, -5); // t.type() == TxTranslate
QTransform tt = t.transposed();
Q_ASSERT(tt.type() == QTransform::TxProject); // fails

In transposing, dx := m31 will be swapped with m13 (and dy := m32 with m23): The transposed form of a translating QTransform will be a projection.