[issue18619] atexit leaks callbacks in subinterpreters

Antoine Pitrou added the comment: The patch was simple enough that I decided to push it as-is :) Don't hesitate to make comments if you want, though. -- resolution: -> fixed stage: needs patch -> committed/rejected status: open -> closed ___ Python t

[issue18619] atexit leaks callbacks in subinterpreters

Roundup Robot added the comment: New changeset 71b63a32b1e3 by Antoine Pitrou in branch 'default': Issue #18619: Fix atexit leaking callbacks registered from sub-interpreters, and make it GC-aware. http://hg.python.org/cpython/rev/71b63a32b1e3 -- nosy: +python-dev _

[issue18619] atexit leaks callbacks in subinterpreters

Antoine Pitrou added the comment: Here is a patch. It also enables GC on the atexit module. -- keywords: +patch Added file: http://bugs.python.org/file31113/atexit_leaks.patch ___ Python tracker ___

[issue18619] atexit leaks callbacks in subinterpreters

New submission from Antoine Pitrou: atexit uses PEP 3121 module initialization, and by this scheme each subinterpreter gets a separate module state. However, atexit's m_free doesn't actually decref the callbacks: it only frees the callbacks array, thinking the callbacks were already decref'ed