Backports for 0.23.x release#4924
Conversation
* allow pypy3.11 * run CI on pypy3.11 * change const declaration * change link name for PyExc_BaseExceptionGroup * PyObject_DelAttr* are inline functions on pypy3.11 * use nightly until official release * DOC: add a news fragment * conditionally compile 'use' statements * fix format * changes from review * pypy 3.11 released * fixes for PyPy * typo * exclude _PyInterpreterFrame on PyPy * more pypy fixes * more excluding _PyFrameEvalFunction on PyPy * fixes for PyPy struct differences * format * fix test --------- Co-authored-by: Icxolu <10486322+Icxolu@users.noreply.github.com>
|
Sigh, I hate hypothesis health checks. |
|
I think picking #4879 might resolve. 😓 |
* make FreeList explicitly a wrapper for *mut PyObject * fix multithreaded access to freelist pyclasses * add changelog entry * use a GILOnceCell to initialize the freelist * respond to code review * skip test on wasm --------- Co-authored-by: David Hewitt <mail@davidhewitt.dev>
Co-authored-by: Ariel Ben-Yehuda <arielby@amazon.com>
|
I think #4912 will fix the benchmarks build. |
| @@ -0,0 +1 @@ | |||
| * fixed spurious `test_double` failures. | |||
There was a problem hiding this comment.
For what it's worth, when editing the CHANGELOG later I would probably just omit this line as it's not really user facing.
|
Okay, this is green. Do you want me to wait for #4874, or file a separate backport PR for it later? |
|
Let's see if that PR looks likely to land today; if not then I propose we move forward to ship this. |
…4921) * add test that panics because __builtins__ isn't available * use a critical section to serialize adding __builtins__ to __globals__ * add release note * use safe APIs instead of PyDict_Contains
CodSpeed Performance ReportMerging #4924 will improve performances by 10.99%Comparing Summary
Benchmarks breakdown
|
|
Let's not wait for #4874 |
|
The benchmark measurement change is probably not material and just a result of inlining change or similar. |
|
Anything else I should do? In particular, should I do the release tasks in this PR, or will this one be merged, and then I should open another one for release tasks? |
|
I don't have a fixed process for that bit, maybe let's just merge this and do the release steps in a follow up. Thank you again and I'll aim to help ship this tomorrow evening UK time |
|
With apologies I was unwell yesterday, I may be able to do this tonight instead. |
|
No problem, I didn't really find time to look into it either. |
As discussed in #4689.
Changes included:
benchmarksci base image #4912