gh-140550: Initial implementation of PEP 793 – PyModExport - #140556
Conversation
Sorry, something went wrong.
This uses a "def-like" structure: a PyModuleDef* that's not a valid Python object.
encukou
left a comment
There was a problem hiding this comment.
Thank you for the initial review!
Yes, but there's a lot of tests. And I've also renamed things that changed semantics, so unrelated changes/backports will conflict or fail to build, rather than merge cleanly at the Git level but be subtly broken.
Sorry, something went wrong.
| import unittest | ||
| import types |
There was a problem hiding this comment.
I couldn't find this in PEP 8.
Sorry, something went wrong.
|
!buildbot AMD64.FreeBSD.Refleaks |
Sorry, something went wrong.
|
🤖 New build scheduled with the buildbot fleet by @encukou for commit 01d52ba 🤖 Results will be shown at: https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F140556%2Fmerge The command will test the builders whose names match following regular expression: AMD64.FreeBSD.Refleaks The builders matched are:
|
Sorry, something went wrong.
Yes. There's a lot to add since the PEP calls for soft-deprecation of PyInit_; see the issue. |
Sorry, something went wrong.
Co-authored-by: Kumar Aditya <kumaraditya@python.org>
vstinner
left a comment
There was a problem hiding this comment.
LGTM. Just please move _PyModule_GetGCHooks() to pycore_moduleobject.h.
Sorry, something went wrong.
| static inline PyModuleDef *_PyModule_GetDefOrNull(PyObject *arg) { | ||
| PyModuleObject *mod = _PyModule_CAST(arg); | ||
| if (mod->md_token_is_def) { | ||
| return (PyModuleDef *)((PyModuleObject *)mod)->md_token; |
There was a problem hiding this comment.
| return (PyModuleDef *)((PyModuleObject *)mod)->md_token; | |
| return (PyModuleDef *)mod->md_token; |
Sorry, something went wrong.
|
|
||
| PyMODINIT_FUNC | ||
| FUNC_NAME(MODULE_NAME)(void) | ||
| INITFUNC_NAME(MODULE_NAME)(void) |
There was a problem hiding this comment.
Is it really needed to define a PyInit function if a modexport function is defined with slots? Developers may reuse this code as an example.
Sorry, something went wrong.
There was a problem hiding this comment.
Thanks for the catch! I think I had some platform-specific issues with the current version of setuptools. I added this to the TODO list in the issue; I'll get to it in a later PR.
Sorry, something went wrong.
|
Thank you for the review! |
Sorry, something went wrong.
|
Congrats :-) |
Sorry, something went wrong.
|
make check-limited-abi does now fail on the main branch: ./python -E ./Tools/build/generate-build-details.py `cat pybuilddir.txt`/build-details.json Checked 115 modules (37 built-in, 77 shared, 1 n/a on linux-x86_64, 0 disabled, 0 missing, 0 failed on import) ./python ./Tools/build/stable_abi.py --all File /home/vstinner/python/main/Doc/data/stable_abi.dat differs from expected! --- /home/vstinner/python/main/Doc/data/stable_abi.dat +++ <expected> @@ -979,14 +979,6 @@ type,Py_buffer,3.11,,full-abi type,Py_intptr_t,3.2,, macro,Py_mod_abi,3.15,, -macro,Py_mod_doc,3.15,, -macro,Py_mod_methods,3.15,, -macro,Py_mod_name,3.15,, -macro,Py_mod_state_clear,3.15,, -macro,Py_mod_state_free,3.15,, -macro,Py_mod_state_size,3.15,, -macro,Py_mod_state_traverse,3.15,, -macro,Py_mod_token,3.15,, macro,Py_mp_ass_subscript,3.2,, macro,Py_mp_length,3.2,, macro,Py_mp_subscript,3.2,, I created #141056 to fix the issue. |
Sorry, something went wrong.
|
Thank you! |
Sorry, something went wrong.
…honGH-140556) Co-authored-by: Victor Stinner <vstinner@python.org> Co-authored-by: Kumar Aditya <kumaraditya@python.org>
This adds the initial implementation of PEP-793.
See the issue for follow-up tasks.