PyType_GetModule() may return wrong or unexpected result when a type has
the feature flag Py_TPFLAGS_BASETYPE. The base type flag permits subclassing of an extension class. The function will return NULL when a type is subclasses in Python space or a different module object when the type is subclasses and initialized in a differen C extension module.

All subclassable types must use _PyType_GetModuleByDef() to safely access module and module state struct. It's easy to miss the problem. I missed it when I ported some of my code to heap types and multiphase init.

I propose to either a guard to PyType_GetModule() to prevent misuse of the function or to add a variant of the function that prevents misuse.

I propose to either a guard to PyType_GetModule() to prevent misuse of the function or to add a variant of the function that prevents misuse.
+1. Looks like check Py_TPFLAGS_BASETYPE in PyType_GetModule() is an easy way:)

I don't understand this issue. Why would Py_TPFLAGS_BASETYPE make the result of PyType_GetModule() wrong? Is the note in the documentation of PyType_GetModule() not clear?
A type and its subtypes may be defined in different modules; why would it be wrong to ask which module the superclass was defined in?

Note that PyType_GetModule() is needed if users want to implement something like _PyType_GetModuleByDef().

I missed it when I ported some of my code to heap types and multiphase init.

I guess PyType_GetModule() is not good for that use case.
If your code is open, could you add a link?

I will close this issue.
The docs have a description of what PyType_GetModule does, and I'll be working on better API that avoids this potential surprise.