Gregory P. Smith <g...@krypto.org> added the comment:
I left some comments on the PR. We can at least make it detect musl libc and improve its default behavior there. FWIW CPython's recursionlimit and the C stack have long been a thorn in the side of many things. Even when Python's thread_pthread.h is changed to set a better default when running under musl libc, there are still plenty of ways to run into problems. ex: Extension modules can create threads on their own that then call back into CPython. So can C/C++ code that embeds a Python interpreter. Those choose their own stack sizes. When they choose low values, surprising crashes ensue for people working on the Python side... As a feature (beyond just this issue): It'd be ideal to actually be able to introspect the C stack space remaining and issue a RecursionError whenever it falls below a threshold. Rather than just blindly using a runtime adjustable number that any Python code can tweak to allow crashes via `sys.setrecursionlimit()`. Other languages like Golang use a dynamic stack and allocate more on SIGSEGV. [meta/off-topic: Why does anyone use small thread stack sizes? This is often done to save virtual address space in applications that expect to have thousands of threads. It matters a lot less on 64-bit address space. But there are reasons...] ---------- assignee: -> gregory.p.smith versions: +Python 3.10, Python 3.11, Python 3.9 -Python 2.7, Python 3.6 _______________________________________ Python tracker <rep...@bugs.python.org> <https://bugs.python.org/issue32307> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
