hang instead of pthread_exit during interpreter shutdown #4874
Conversation
CodSpeed Performance ReportMerging #4874 will not alter performanceComparing Summary
|
arielb1
force-pushed
the
safe-exit
branch
4 times, most recently
from
80f138a to
704ef0b
Compare
|
Yay it passes tests. |
pyo3-ffi/src/pystate.rs
Outdated
| #[cfg(not(Py_3_14))] | ||
| pub unsafe extern "C" fn PyGILState_Ensure() -> PyGILState_STATE { | ||
| let guard = HangThread; | ||
| let ret: PyGILState_STATE = raw::PyGILState_Ensure(); | ||
| std::mem::forget(guard); | ||
| ret | ||
| } |
There was a problem hiding this comment.
Disclaimer: I'm really no expert here.
According to my understanding of the C-unwind this is still UB. Specifically here is mentioned that forced unwinding (like pthread_exit) can not rely on Drop and thus crossing such framed is considered UB.
Given that this was UB before as well, but this one works, maybe that's better 🤷 (We really need a Python API to do this properly)
There was a problem hiding this comment.
So bjorn claimed this way is unspecified behavior but should work. I think it's better then the C++ version.
There was a problem hiding this comment.
There was a problem hiding this comment.
Thanks, I think that makes sense to me. I guess we aren't guaranteed that drop will run, but in case it is not run, we are in the same situation as currently (presumably a crash). So given this (currently) works in practice it seems like a net win to me.
I think it's better then the C++ version.
Definitely agree here, I was not really a fan of adding a C++ toolchain dependency 👍
pytests/tests/test_othermod.py
Outdated
| @@ -3,7 +3,7 @@ | |||
|
|
|||
| from pyo3_pytests import othermod | |||
|
|
|||
| INTEGER32_ST = st.integers(min_value=(-(2**31)), max_value=(2**31 - 1)) | |||
| INTEGER32_ST = st.integers(min_value=(-(2**30)), max_value=(2**30 - 1)) | |||
There was a problem hiding this comment.
this change here is to avoid the test failing due to a high rate of assume failure (filter_too_much).
|
split the test_double fix to #4879 |
This mimics the Python 3.14 behavior and avoids crashes in Rust 1.84.
See python/cpython#87135 and rust-lang/rust#135929 and the related pyo3-log issue vorner/pyo3-log#30
By submitting these contributions you agree for them to be dual-licensed under PyO3's MIT OR Apache-2.0 license.