feat: add summary to Action

[henryiii]

This prints out a summary in the GitHub action step.

Some logging commands were missing for Pyodide.

You can see the general idea, but I plan to flesh it out soon.

[joerick]

This is pretty cool.

I couldn't help myself - I had a play around with formatting for this, what do you think?

• Display build options (can't dim, so only those that differ from the defaults)
• Show the wheel name first, feels more primary to me
• Make the table more scannable - stop Github from wrapping the wheel name in the table by making an HTML table using <td nowrap> and <samp> elements

---

🎡 cibuildwheel

Build options

platform: macos
architectures: arm64, universal2, x86_64
build_selector: 
  build_config: *
  skip_config: 
  requires_python: None
  enable: ['cpython-experimental-riscv64', 'cpython-freethreading', 'cpython-prerelease']
output_dir: /Users/runner/work/cibuildwheel/cibuildwheel/wheelhouse
package_dir: /Users/runner/work/cibuildwheel/cibuildwheel/sample_proj
build_frontend: build[uv]

17 wheels created

    ^{in 1 minute 2 seconds}

Wheel	Size	Build identifier	Time
spam-0.1.0-cp38-cp38-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	8.2 kB	cp38-manylinux_x86_64	2s
spam-0.1.0-cp39-cp39-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	7.7 kB	cp39-manylinux_x86_64	1s
spam-0.1.0-cp310-cp310-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	7.8 kB	cp310-manylinux_x86_64	1s
spam-0.1.0-cp311-cp311-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	7.8 kB	cp311-manylinux_x86_64	1s
spam-0.1.0-cp312-cp312-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	8.0 kB	cp312-manylinux_x86_64	1s
spam-0.1.0-cp313-cp313-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	8.0 kB	cp313-manylinux_x86_64	1s
spam-0.1.0-cp313-cp313t-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	8.1 kB	cp313t-manylinux_x86_64	1s
spam-0.1.0-cp314-cp314-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	8.1 kB	cp314-manylinux_x86_64	1s
spam-0.1.0-cp314-cp314t-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl	8.1 kB	cp314t-manylinux_x86_64	1s
spam-0.1.0-cp38-cp38-musllinux_1_2_x86_64.whl	7.9 kB	cp38-musllinux_x86_64	2s
spam-0.1.0-cp39-cp39-musllinux_1_2_x86_64.whl	7.8 kB	cp39-musllinux_x86_64	1s
spam-0.1.0-cp310-cp310-musllinux_1_2_x86_64.whl	8.0 kB	cp310-musllinux_x86_64	2s
spam-0.1.0-cp311-cp311-musllinux_1_2_x86_64.whl	8.0 kB	cp311-musllinux_x86_64	1s
spam-0.1.0-cp312-cp312-musllinux_1_2_x86_64.whl	8.1 kB	cp312-musllinux_x86_64	1s
spam-0.1.0-cp313-cp313-musllinux_1_2_x86_64.whl	8.2 kB	cp313-musllinux_x86_64	2s
spam-0.1.0-cp313-cp313t-musllinux_1_2_x86_64.whl	8.3 kB	cp313t-musllinux_x86_64	1s
spam-0.1.0-cp314-cp314-musllinux_1_2_x86_64.whl	8.2 kB	cp314-musllinux_x86_64	2s
spam-0.1.0-cp314-cp314t-musllinux_1_2_x86_64.whl	8.3 kB	cp314t-musllinux_x86_64	2s

---

I didn't implement it yet, just playing around with HTML/markdown.

[henryiii]

Those improvements sound good!

[joerick]

Would you like me to have a crack at implementing that new format? I might get some time towards the end of this week.

[henryiii]

Sure, I'm heading to SciPy then PyHEP.dev in Seattle. Likely won't have much time for a while.

[joerick]

Cool! Enjoy!

[joerick]

Here's where I landed-

---

🎡 cibuildwheel

Build options

output_dir: /private/var/folders/ld/k24nt7054698bctspqwrjq1r0000gn/T/tmpxz77ycv4/wheelhouse
package_dir: /private/var/folders/ld/k24nt7054698bctspqwrjq1r0000gn/T/tmpxz77ycv4

Wheel	Size	Build identifier	Time
spam-0.1.0-cp38-cp38-macosx_11_0_arm64.whl	2.8 kB	cp38-macosx_arm64	9 seconds
spam-0.1.0-cp39-cp39-macosx_11_0_arm64.whl	2.8 kB	cp39-macosx_arm64	3 seconds
spam-0.1.0-cp310-cp310-macosx_11_0_arm64.whl	2.8 kB	cp310-macosx_arm64	3 seconds
spam-0.1.0-cp311-cp311-macosx_11_0_arm64.whl	2.8 kB	cp311-macosx_arm64	4 seconds
spam-0.1.0-cp312-cp312-macosx_11_0_arm64.whl	2.8 kB	cp312-macosx_arm64	3 seconds
spam-0.1.0-cp313-cp313-macosx_11_0_arm64.whl	2.8 kB	cp313-macosx_arm64	3 seconds

^{6 wheels created in 26 seconds}

---

[henryiii]

You can use inspect.cleandoc instead of textwrap.dedent then you don't have to worry about leading/trailing whitespace. Up to you which you prefer.

Suggested change

	textwrap.dedent("""\
	inspect.cleandoc("""

[freakboy3742]

No complaints from the iOS side of things - and the rest of the implementation makes sense to me as well.

[ryanking13]

Looks good for Pyodide-side changes as well. The summary looks quite decent.

[agriyakhetarpal]

This looks nice, thanks, @henryiii and @joerick!

I had one comment about the summaries when being added to Markdown: https://github.com/pypa/cibuildwheel/actions/runs/16222307793/attempts/1#summary-45805767785

As you can see when expanding the "Build options" dropdown, there are remnants of ANSI colour codes in the output at the end of each option, like this: �[0m. I think these are coming from the fact that GitHub step summaries convert ANSI into raw UTF-8, where these codes become invisible bytes and show up in this weird manner. I suggest that we turn off ANSI colouring when writing Markdown output to the step summary, and I think there should be two ways to handle this:

• either we add another argument such as options.summary(..., colors_enabled=True) and then do c = Colors(enabled=colors_enabled) instead of c = log.colors;
• or we add another options.summary_markdown(...) variant which disables colours and call that one in _github_step_summary()

What do you think? I'm also happy to do so myself in a separate PR, not blocking this PR, or push a few commits here itself :D

[agriyakhetarpal]

Also, would it make sense to add a SHA-256 checksum column to the table in the future? It would be really nice to grab the hashes for the wheels directly from cibuildwheel's output and sanity check (manually) if they match, say, artifacts for a GitHub release or those uploaded to PyPI.

[agriyakhetarpal]

Great! My concerns and questions were resolved, and it looks much better. Thank you!