[hugovk]

Homebrew 7 no longer provides binary "bottles" for macOS Intel:

https://brew.sh/2026/09/13/homebrew-7.0.0/#-macos-users

Which means LLVM has to build from source on this platform. This takes more than the 60-minute limit and times out.

PEP 836 has been withdrawn, and in any case only targetted tier 1, where macOS Intel is tier 2.

Let's remove macOS Intel from the JIT CI. If any future work wants to test it, let's figure it out then.

• Before: https://github.com/python/cpython/actions/runs/39121087906: two timed-out x86_64-apple-darwin/clang jobs
• After: https://github.com/python/cpython/actions/runs/39580151999?pr=160033: no x86_64-apple-darwin/clang jobs

• Issue: GitHub Action: "Tail calling interpreter" & JIT times out after 60 minutes on macOS x86_64 on building LLVM from source #159612

[StanFromIreland]

What about the tail-call workflow, it's also affected?

[ned-deily]

What about the tail-call workflow, it's also affected?

There should be no need to use a non-Apple clang for tail-call builds these days. The only need in our installer builds is for creating the JIT stencils.

[StanFromIreland]

So we can switch to the Xcode Clang in that job IIUC?

[ned-deily]

Unless there's some other gotcha, yes. Let's try it with just Xcode clang.

[vstinner]

LGTM. We have an Intel macOS buildbot: https://buildbot.python.org/#/workers/68, "x86-64 MacOS Intel NoGIL 3.x" and "x86-64 MacOS Intel ASAN NoGIL 3.x".

[itamaro]

considering darwin-intel is tier-2, we could just drop all CI coverage for that platform, and keep just the buildbots