Replies: 3 comments 3 replies
-
Thanks for the request and I appreciate the pain. The problem is that without CI testing we cannot guarantee that outdated PyPy versions will not simply just crash. We prevent install on these versions for safely. Can you explain how exactly the current situation is a problem? As far as I can see, it will only affect users running PyPy 3.7 or 3.8. the PyPy team does not support those versions any more nor encourage users to use them, so I don't see why we should be making it easier for users to continue to try to use these out of support versions. |
Beta Was this translation helpful? Give feedback.
-
I understand that. Can we have a reduced (in terms of success definition) CI for them?
It is simple really. Doing a Each project (from trustme and dependencies) have a We cannot, using simple means, exclude PyPy 3.7, and PyPy 3.8 due to what is exposed.
That I understand also. But there is an issue that is often underrated or ignored by maintainers. A lot of companies are still bound to EOLs CPython or PyPy, and they are massively abandoned by us. Big corporations may have slow upgrade scheme to upgrade their interpreters.
CPython 3.7 and 3.8 are both EOL but statistics shows that they are still massively used. See https://pypistats.org/packages/__all__ I hope you will consider that request. |
Beta Was this translation helpful? Give feedback.
-
FWIW, we (pyca/cryptography) have been able to drop older
PyPy's without incident.
…On Wed, Dec 25, 2024 at 9:34 AM TAHRI Ahmed R. ***@***.***> wrote:
We would need to give up on freethreaded builds for that, and every bit of
improvements served in 0.23.x releases.
If the answer is no, we would have to consider dropping building for PyPy
altogether, which would be unfortunate. We can't keep python_require >=
3.7 when it cause crashes, especially when it is about a dependency of a
dependency.
—
Reply to this email directly, view it on GitHub
<#4799 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAAAGBDON5EP72L2Q74TWUD2HK663AVCNFSM6AAAAABTUK6EJGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCNRWGQYDAMQ>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
--
All that is necessary for evil to succeed is for good people to do nothing.
|
Beta Was this translation helpful? Give feedback.
-
Hello,
I am writing to ask for a revert of recent change introduced in 0.23.x releases.
The change I am referring to is located at #4582
I far as I can understand, no API / Bridge change occurred in the PR. No code simplification occurred, only the check for minimal version was patched.
While this change appear straightforward, it induce friction for integrators.
We want to publish new versions, but we are stuck/embarrassed due to the high level constraint (namely
python_requires >= 3.7
). There is no easy way to mark PyPy 3.7 and 3.8 excluded defacto so that pip or alike wouldn't pick the newer version.I also understand that the change was motivated due to CI flakyness.
My proposal is as follow:
This would be greatly appreciated.
Regards,
Beta Was this translation helpful? Give feedback.
All reactions