I have seen the FAQ https://wiki.qt.io/Qt_for_Python#Frequently_Asked_Questions which states
Take into account that if you want to generate bindings for a Qt/C++ project, the linking to the Qt shared libraries will be missing, and you will need to do this by hand. We recommend to build PySide2 from scratch to have everything properly linked.
Suppose you want to use shiboken2 to wrap a c++ library which itself uses qt and it even uses qt classes in its interface. In the python world you should be able to
pip install PySide2 my_cpp_lib
and afterwards python scripts shall be able to
import my_cpp_lib
from PySide2.QtCore import QObject
my_cool_object = QObject()
my_cpp_lib.my_cool_api(my_cool_object)
It is pretty unclear to me how to achieve that setup. From the FAQ I deduce that it would be good to compile PySide2 from source in this case (because I want to use shiboken2_generator to wrap my_cpp_lib). However, I still want to be compatible to the standard PySide2 package (of course there will be a version dependency between my_cpp_lib and PySide2).
Is linking against the QT shared object/dll of PySide2 a bad idea? If so, why? what would be better instead?
If the qt library should be packaged together with the my_cpp_lib python package, then how is it possible to still use PySide without conflicts (because there are essential two different shared objects)?