It took me quite a bit to figure out what this was referring to,
since the given issue number is wrong, and the original commit
message I found through git blame lists a different, also wrong
issue number... see https://bugs.python.org/issue27122#msg279449

(cherry picked from commit af88e7e)

Mr3x33 added a commit to Mr3x33/cpython that referenced this pull request

colesbury referenced this pull request in colesbury/nogil

Once the threads object queue is empty, other threads may explicitly
merge reference counts immediately. It's no longer safe for the original
thread to do RC operations because of races with accesses to ob_ref_local.

This ensures that the queue is empty when it's removed from the hashtable
mapping thread-id to queue. There are still other RC operations that can
happen during thread destruction, so the call to _Py_queue_destroy will
probably need to be moved.

See #50

colesbury referenced this pull request in colesbury/nogil

A partially initialized thread may have an entry in the interpreter
without a corresponding entry in the biased-refcounting hashtable.

See #50

jaraco pushed a commit that referenced this pull request

Now reads: `The PR then needs to be backported manually.`

jaraco added a commit to jaraco/cpython that referenced this pull request

* Added a config for dependabot.

* Update features list for dependabot.

Co-authored-by: KOLANICH <kolan_n@mail.ru>

mgmacias95 pushed a commit to mgmacias95/cpython that referenced this pull request

This was referenced

DinoV added a commit to DinoV/cpython that referenced this pull request

Flow in _PyInterpreterFrame and fix issue when global not defined