Add patch skipping TestMachoIssue32233

[dagood]

• Tracked at darwin-amd64 test flaky TestMachoIssue32233: error: unknown type name 'NSUInteger' #2200

[gdams]

One concern I have with this is that the patches build in order CI job is getting longer and longer here. I wonder if we should add a [test]- prefix to patches that only affect tests and update the CI job to simply apply the patch and skip over the build?

[Copilot]

Pull request overview

Adds a downstream patch to skip a flaky Go linker DWARF test (TestMachoIssue32233) that is currently blocking PR validation, tracked in #2200.

Changes:

• Introduces a new patch file that injects an unconditional t.Skip(...) into TestMachoIssue32233.
• Documents the rationale and tracking issue in the patch header and skip message.

Comments suppressed due to low confidence (1)

patches/0015-Skip-consistently-failing-TestMachoIssue32233.patch:1

• This unconditionally skips the test on all builders, which can permanently hide real regressions even on platforms where it’s stable. Prefer making the skip conditional to the known failing conditions (e.g., specific GOOS/GOARCH/builder signal), and/or add a clear TODO with criteria for removal so it doesn’t become a long-lived blind spot.

From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001

[dagood]

One concern I have with this is that the patches build in order CI job is getting longer and longer here. I wonder if we should add a [test]- prefix to patches that only affect tests and update the CI job to simply apply the patch and skip over the build?

Didn't notice the time--still shorter than the AzDO-side builds so far in my experience.

I think we should update the workflow to build them in parallel, so we don't have to worry about number of patch files, and so we also don't lose the sequence-buildable guarantee for some subset of patches.