Graham Dumpleton

So far it looks like the issue is that the change to correct behaviour in wrapt is exposing an object of type `types.MappingProxyType` to tensorflow `__getattribute__()` override which it has...

My first guess at what maybe they should be doing is: ``` def __getattribute__(self, name): import types #if (hasattr(type(self), name) # and (isinstance(getattr(type(self), name), property)): if hasattr(type(self), name): obj =...

Another possibility is that for that access type it should not return anything but instead should raise an AttributeError. It is quite possible that higher layers are checking for existence...

To summarise at this point as I don't think I can do anything more. The behaviour of wrapt changed for the C extension when a bug in wrapt was fixed...

BTW, forgot to highlight this, but the trigger for this is that the exception that tensorflow raises: ``` TypeError: this __dict__ descriptor does not support '_DictWrapper' objects ``` uses `TypeError`...

What is it that you are trying to do that requires this? Knowing the underlying reason, rather than what you think may be the solution will help me guide you...

Look at https://github.com/GrahamDumpleton/wrapt/issues/255#issuecomment-1817722900 Didn't occur to me before to add a dummy class attribute so that assignment to instance in constructor would then use the instance rather than wrapped object....

The comments in the StackOverflow issue suggest to me that PySide is doing something unconventional. The comments claim for the `decorator` package: > Your decorator actually returns a function that...

Related reading about really weird stuff PySide does. * https://github.com/pyside/pyside-setup/blob/dev/sources/pyside6/doc/developer/signature_doc.rst * https://github.com/pyside/pyside-setup/tree/dev/sources/shiboken6/shibokenmodule/files.dir/shibokensupport/signature

Sorry for the slow reply, have had a lot going on the last few weeks. The `getattribute()` behaviour is definitely inscrutable at times and I can't even pretend to fully...