Open bluetech opened 6 months ago
I have same failure on older versions as well. E.g. 7.4.3: https://github.com/mpmath/mpmath/actions/runs/8030300102?pr=759
This happens for all tested in the mpmath pypy versions (3.8-3.10). But most often on 3.10.
Another example: https://github.com/pytest-dev/pytest/issues/11168#issuecomment-1714358183
@cfbolz, I don't see a bugreport in pypy issues. Is there any chance this might be related to pytest-dev/pytest#11168 (an fixed in pypy)?
@skirpichev it's not impossible that it's the same problem, yes. You can try a nightly PyPy build (or wait for the next release but that is still a bit out).
@cfbolz, thanks for a suggestion. I did tests for nightly builds (on 3.8-3.10) and, after ~20 re-runs, I got a TypeError on pypy3.8, which is similar to this issue:
Summary page: https://github.com/mpmath/mpmath/actions/runs/8491432157?pr=772
Maybe this fix isn't merged yet to the 3.8 branch (as it was stated in your blog post)? If so, I'll continue testing.
Yes, the 3.8 variants are not maintained any more, you need to try 3.9 or 3.10. Thanks for trying this out!
Well, I did ~50 re-runs of mpmath's tests on pypy3.9 and pypy3.10 - without any failures. Given the previous statistics, probably there is a chance that this issue was fixed in PyPy's nightly builds. I'll switch regular tests for pypy3.9/10 to them.
Thanks for bugfix!
Well, I did ~50 re-runs of mpmath's tests on pypy3.9 and pypy3.10 - without any failures.
thank you for trying this, that's very valuable feedback!
Sometimes it says
TypeError: expected some sort of stmt, but got <_ast.Load object at 0x000000000365d040>
instead.
During the sprints, we were seeing 3–4 variants of different AST object reprs there. Hopefully, bumping PyPy in CI will fix this. Ronny suggested this, but nobody got to actually doing it...
FTR @LilyFoote linked https://www.pypy.org/posts/2024/03/fixing-bug-incremental-gc.html in https://github.com/pytest-dev/pytest/issues/11168#issuecomment-2023988486 but that reference never got cross-posted into this issue.
Action item for whoever gets to prepare the PR:
Resolves #11771
somewhere in the commit message or the PR descriptionWorking on this issue
Sometimes it says
TypeError: expected some sort of stmt, but got <_ast.Load object at 0x000000000365d040>
instead.
I even saw a case of NameError
, which seems to have the same root cause:
E NameError: name '@py_assert2' is not defined
(https://github.com/pytest-dev/pytest/actions/runs/9746885084/job/26898335196#step:6:97)
Example: https://github.com/pytest-dev/pytest/actions/runs/7408595776/job/20157160097
This doesn't always happen, maybe 50% of the time. Started in the last few days. I checked if the PyPy version changed during these days, but seems not (it's 3.8.16).
The error is:
Sometimes it says
TypeError: expected some sort of stmt, but got <_ast.Load object at 0x000000000365d040>
instead.