The undefined behavior sanitizer buildbot is flagging a bunch of issues in master (3.8) of late:

AssertionError: 'Objects/classobject.c:74:29: runtime erro[139 chars]re\n' != ''
- Objects/classobject.c:74:29: runtime error: null pointer passed as argument 2, which is declared to never be null
- /usr/include/string.h:43:28: note: nonnull attribute specified here

 (see https://buildbot.python.org/all/#/builders/135/builds/1937/steps/5/logs/stdio)

This appears to be coming from a relatively new classobject.c:method_vectorcall() function

method_vectorcall(PyObject *method, PyObject *const *args,
                  size_t nargsf, PyObject *kwnames)

Which looks like it is probably being called with NULL args value and thus winds up calling memcpy() with src=NULL.

This was introduced in https://github.com/python/cpython/commit/aacc77fbd77640a8f03638216fa09372cc21673d for the PEP 590 implementation.

That's entirely my fault. I was not aware of the fact that calling memcpy() with src=NULL and length 0 was undefined behaviour. I disagree that it should be undefined, but there is not much point in arguing against the C standard.

New changeset 1f9531764cc0f8dbca1d8f429d162dc28282f4b4 by Gregory P. Smith (Jeroen Demeyer) in branch 'master':
bpo-37138: fix undefined behaviour with memcpy() on NULL array (GH-13867)
https://github.com/python/cpython/commit/1f9531764cc0f8dbca1d8f429d162dc28282f4b4

New changeset 6e053079ac3fe50ffbe9128bcf766298168c31cb by Miss Islington (bot) in branch '3.8':
bpo-37138: fix undefined behaviour with memcpy() on NULL array (GH-13867)
https://github.com/python/cpython/commit/6e053079ac3fe50ffbe9128bcf766298168c31cb

Thanks Jeroen!