Skip to content

Windows + (native) Linux ARM64 build #22

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
peterjc merged 2 commits into from
Sep 2, 2026
Merged

Windows + (native) Linux ARM64 build #22

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e90c515
Add Windows ARM64 build
ndabas Aug 31, 2026
be015f2
Move Linux aarch64 to native runner
ndabas Sep 1, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 4 additions & 11 deletions .github/workflows/cibuildwheel.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,16 +9,15 @@ jobs:
strategy:
matrix:
# macos-13 was an intel runner, macos-14 is apple silicon
os: [ubuntu-22.04, windows-2022, macos-14]
os: [ubuntu-22.04, ubuntu-22.04-arm, windows-2022, windows-11-arm, macos-14]
env:
# Seems to have trouble parsing our setup.py, thus:
CIBW_PROJECT_REQUIRES_PYTHON: ">=3.10"
# Install numpy with a wheel (do NOT compile it):
CIBW_BEFORE_BUILD: "pip install --only-binary :all: -r requirements.txt"
# There are/were no numpy wheels for these:
CIBW_SKIP: cp3*-manylinux_i686 *-musllinux_*
# configure cibuildwheel to build native arch & aarch64 too
CIBW_ARCHS_LINUX: auto aarch64
# There are/were no numpy wheels for these.
# Windows ARM64 CPython is only available from 3.11 onwards.
CIBW_SKIP: cp3*-manylinux_i686 *-musllinux_* cp310-win_arm64
# Quick test with C code using NumPy C API:
CIBW_TEST_COMMAND: "python {package}/Tests/test_Cluster.py"

Expand All @@ -28,12 +27,6 @@ jobs:
fetch-depth: 0 # Optional
submodules: true # Optional, use if you have submodules

- name: Set up QEMU
if: runner.os == 'Linux'
uses: docker/setup-qemu-action@v4
with:
platforms: all

- name: Checkout tag/commit
run: |
git submodule update --init biopython
Expand Down
5 changes: 3 additions & 2 deletions requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,5 +14,6 @@ numpy==1.14.5; python_version=='3.7'
numpy==1.17.3; python_version=='3.8'
numpy==1.21.0; python_version=='3.9'
numpy==1.22.3; python_version=='3.10'
numpy==1.23.3; python_version=='3.11'
numpy==1.26.1; python_version=='3.12'
# NumPy only has win_arm64 wheels from 2.3.0 onwards
numpy==1.23.3; python_version=='3.11' and (sys_platform!='win32' or platform_machine!='ARM64')
numpy==1.26.1; python_version=='3.12' and (sys_platform!='win32' or platform_machine!='ARM64')

@peterjc peterjc Sep 1, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What happen on Python 3.11 or 3.12 when (sys_platform!='win32' or platform_machine!='ARM64') is false? Does that just never declare a versioned numpy dependency...

@ndabas ndabas Sep 1, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, correct, that does not declare a version-pinned numpy dependency, so the unpinned numpy dependency declared in biopython itself takes precedence, and we get the latest version (in most circumstances.)

I thought that was the idea here -- that we test with the 1.x series of numpy where supported. For the existing 3.13+ builds on other platforms, we just get the unpinned numpy dependency as well.

I could pin a version like 2.3.0 for the Windows ARM64 cases here, but I'm not sure if that would really add anything apart from a cluttered requirements.txt.

(Just for clarity, the way I have written the conditional expression -- (!x || !y) is equivalent to !(x && y) by De Morgan's laws. In testing I found that there is no not (expression) operator that we can use.)

@peterjc peterjc Sep 2, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd have to go hunting for the docs, but until relatively recently (in the numpy v2 era) the advice for maximising compatibility was to build against the oldest supported numpy v1 - thus these pins.

@ndabas ndabas Sep 2, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, exactly -- that's the reason why I didn't add explicit pins for the Windows ARM64 case, because the earliest supported numpy wheels there are in the 2.x series anyway.