workflows/bootstrap.sh: Add diagnostics for Metal Toolchain

[Dave-Allured]

Description

Show Metal Toolchain status at start and end of CI bootstrap stage.
Needed because Metal Toolchain was removed from Xcode 26, and is now provided by Github runner for CI.

Type(s)

• enhancement

Tested on

CI only.

Verification

Have you

• followed our Commit Message Guidelines?
• squashed and minimized your commits?
• checked that there aren't other open pull requests for the same change?\
• tested basic functionality of all binary files CI process?

[herbygillot]

CC: @jmroot

[reneeotten]

does this actually install the Metal framework or only shows whether it exists? If the latter it's pretty useless since we know that it will not be there... or?

[Dave-Allured]

That is a fair question. Yes this change only shows whether Metal Toolchain exists, at two different times in the bootstrap stage.

I do not think that we actually "know that it will not be there". My early testing found that Metal Toolchain is installed at start of bootstrap on macOS 26-arm64 runner. This is consistent with a recent statement from Github: actions/runner-images#13014 (comment)

The purpose is to help debug cases of "missing Metal Toolchain" on CI Github runners, such as in the iTerm2 pull request that was merged a few minutes ago. Also see https://trac.macports.org/ticket/73180.

[jmroot]

So you're just trying to find out if the Metal toolchain is installed on the runner and whether the bootstrap script changes that? If so this doesn't need to be a PR; you can just push a whitespace change to any port to the branch and CI should run in your fork, giving you the answer.

[reneeotten]

I'd say we can just ignore failures due to the Metal framework on the CI...

[Dave-Allured]

just push a whitespace change

We need more fine grained information to find cause of recent flakiness in Metal on macOS 26 CI. This diagnostic is one small step in the right direction to sort it out.

just ignore failures

That would leave maintainers such as myself in an uncomfortable position, given that the general rule is that valid PR's must pass all CI.

[reneeotten]

just ignore failures

That would leave maintainers such as myself in an uncomfortable position, given that the general rule is that valid PR's must pass all CI.

Ideally yes, but I have certainly just ignored failures due to the Metal framework and merged PRs/