-
Notifications
You must be signed in to change notification settings - Fork 27.5k
[chore] Test pybind11 2.6.0 (RC 2?) #46367
Copy link
Copy link
Open
Labels
enhancementNot as big of a feature, but technically not a bug. Should be easy to fixNot as big of a feature, but technically not a bug. Should be easy to fixmodule: buildBuild system issuesBuild system issuesmodule: ciRelated to continuous integrationRelated to continuous integrationtriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module
Description
Metadata
Metadata
Assignees
Labels
enhancementNot as big of a feature, but technically not a bug. Should be easy to fixNot as big of a feature, but technically not a bug. Should be easy to fixmodule: buildBuild system issuesBuild system issuesmodule: ciRelated to continuous integrationRelated to continuous integrationtriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module
Type
Projects
Status
Untriaged Archival - really old stuff
There is a bug in CPython 3.9.0 (already fixed in 3.9 and master) that breaks pybind11 extensions slightly, causing them to exhibit UB (on macOS, usually segfaults on process exit). While this will be fixed in CPython 3.9.1, and has already been patched in Conda-forge's Python 3.9 build, it is likely to affect users. pybind11 is about to release 2.6.0 (currently on 2.6.0rc2) that will detect 3.9.0 and leak instead of segfaulting; can that be tested with PyTorch to make sure it is an easy upgrade and no last fixes or patches are needed? There is at least one long-standing issue requested by PyTorch fixed in 2.6, as well, though I don't remember which one, as well as better CUDA support and more.
cc @malfet @seemethere @walterddr @ezyang