pyca
refs #11199 attempt to test on windows arm64
this is now blocked until https://github.com/actions/partner-runner-images/issues/77
It's the current blocker. Hacking around that limitation (which we will not merge) produces a segfault, but I have no intent of debugging that until Microsoft resolves the toolchain limitation.
Good news, we now have Rust in the windows image. Bad news is that we're experiencing segfaults.
https://github.com/openssl/openssl/issues/26239 indicates there was at least one known issue.
Unfortunately, https://github.com/actions/partner-runner-images/blob/main/images/arm-windows-11-image.md doesn't document what software versions are in what image versions.
Looks like it is indeed outdated. Visual Studio 2022 Developer Command Prompt v17.13.2
So now we wait some indeterminate amount of time for arm to update their images again. I have serious qualms with depending on these partner images given how poorly they're executing on this.
Great to hear. It would be extremely helpful if things like this were communicated/visible in the official repo.
@alex Sure! I’ll make an effort to improve communication. The new image should already be available.
@alex they asked to try once again and they'll reach out to GH team to see if there's any other issues with the runners. Meanwhile, here is a PR for changes: https://github.com/actions/partner-runner-images/pull/116/files
That package information is a major improvement, but ideally arm would move to doing GH releases that contain that data + information about what has changed much like the GH-owned images do (e.g., https://github.com/actions/runner-images/releases/tag/macos-14-arm64%2F20250723.1691). Do you know if that's the intent?
As is, we have a situation where Arm is providing images that aren't consistent with GH's own, doesn't accurately list information about the software manifest on them, makes it nearly impossible to tell what has changed, appears to have regressions that are suggestive of automation issues, and updates them more slowly than GH does. All of this makes us reluctant to depend on them.
@reaperhulk the update frequency was slow initially due to a few internal certification issues, but these were resolved and shouldn't impact future releases. Currently, a different team from Arm manages the images, as they have more expertise in this area (Arm) compared to the regular engineering team.
Thanks for providing this feedback, I'll ensure everything is taken into consideration, and the process will be improved to meet everyone's expectations.
@alex Rust is now installed in the latest image update. Latest versions and tools here - https://github.com/actions/partner-runner-images/blob/main/images/arm-windows-11-image.md. Please let us know if you face any issues.
Hey @alex , are there any other blockers keeping this from moving forward? Also, is there an ETA for getting support landed?
We’re going to merge this, but if we see regressions or stability issues we’ll revert. This project has a strict policy for all CI jobs to be passing to merge any code, so issues like we’ve seen the past several months would have blocked us entirely.
We aren’t planning to ship a wheel until our next release, which does not have a date.
I think this needs a test coverage, because openssl is known to crash on arm64 unless debug binary is used and cryptography uses openssl. https://github.com/openssl/openssl/issues/27030
We are running tests. They were crashing, but the issue we saw was resolved with a recent MSV release.
Oh, I see the test, and I see that https://github.com/sfackler/rust-openssl/pull/2407 doesn't crash anymore on CI (even though it does crash locally for me 🤔). Thank you!
If you want to see if our artifacts work for you locally: https://github.com/pyca/infra/actions/runs/16662850185
It's possible to go into CI and manually download a wheel if you want, but they're not published anywhere, no.
On Tue, Aug 5, 2025 at 5:47 AM Gleb Khmyznikov @.***> wrote:
khmyznikov left a comment (pyca/cryptography#12770) https://github.com/pyca/cryptography/pull/12770#issuecomment-3154398591
@reaperhulk https://github.com/reaperhulk is there any nightly builds available broadly?
— Reply to this email directly, view it on GitHub https://github.com/pyca/cryptography/pull/12770#issuecomment-3154398591, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAAGBAKQBXJPUDOU64A7PT3MB4T5AVCNFSM6AAAAAB3FRQSNCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTCNJUGM4TQNJZGE . You are receiving this because you were mentioned.Message ID: @.***>
-- All that is necessary for evil to succeed is for good people to do nothing.
@alex I guess it's not possible to do via CI? I'm asking because we could start the contribution to other projects using the links to nightly builds as temporally solution to prepare the PRs, as we did for scikit-learn.
For everyone who need to test their GH Actions:
steps:
- name: Download cryptography wheel
uses: robinraju/[email protected]
with:
repository: "khmyznikov/PyEnv-WoA-State"
tag: "cryptography-46.0.0"
fileName: "*.whl"
out-file-path: "wheels"
- name: Install from wheel
run: |
python -m pip install --find-links=wheels cryptography
# Verify installation
python -c "import cryptography; print(f'Installed cryptography version: {cryptography.__version__}')"
I've created temp release while we are waiting for official wheels.
https://pypi.org/project/cryptography/#files still doesn't show the arm64 wheel for 45.0.7, any idea what went wrong?
Seems it's just because this didn't enter 45.0.x branch, do you think it makes sense to do so?
We aren’t planning to backport windows arm64 wheels, they will appear in the next major release (46.0.0)
