Why isn't all memory freed when CPython exits?

Question

I want to be clear, I am not seeing the behavior described by this question. Instead my question is about the question itself:

The python 3 official FAQ says this verbatim:

Why isn't all memory freed when CPython exits?

And provides this answer:

Objects referenced from the global namespaces of Python modules are not always deallocated when Python exits. This may happen if there are circular references. There are also certain bits of memory that are allocated by the C library that are impossible to free (e.g. a tool like Purify will complain about these). Python is, however, aggressive about cleaning up memory on exit and does try to destroy every single object.

If you want to force Python to delete certain things on deallocation use the atexit module to run a function that will force those deletions.

This, assuming a managed memory operating system (Linux, Mac, Windows, GNU, BSD, Solaris...), sounds like total nonsense.

On a program exiting (be it Python or anything else) any memory it requested from the OS is freed (as the OS has control of the virtual page tables, etc, etc). The program doesn't have to de-allocate or de-construct anything (something programs used to have to do, as highlighted by the time someone's use of cp got bottlenecked by a hash table deconstruction), but I don't think any OS' Python 3 supports puts this requirement on programs.

Does this make sense in some context I'm not aware of? What is this referring to?

Answer 1

The FAQ just says that cpython itself does not actively deallocate all the memory it has acquired when it terminates

If you run cpython on a any normal server/desktop OS that releases all memory of a process when it exits, then there's no issue with memory leaks. The OS takes care of deallocating all memory when the process has exited.

The FAQ is more to inform you that cpython does not call free() or similar on all the memory it has allocated with malloc() or similar. This can have consequences if you run cpython on an OS that does not release all memory acquired by the process when the process exits (These operating systems exists, it is in particular the case with many embedded kernels). And if you run cpython under a memory profiler/leak detector, that detector might report the memory not free()'d as leaks.