[Collect Data] Windows pull request failure unrelated to the changes

[shahmishal]

This issue will be used to collect information about the Windows platform testing. Please add the information here if you see any failures on your pull request unrelated to the changes.

[Rajveer100]

Well, in my recent PR #63567 the smoke-tests passed for macOS and Linux platform, while failing for the windows platform, maybe this info might be helpful!

[shahmishal]

Well, in my recent PR #63567 the smoke-tests passed for macOS and Linux platform, while failing for the windows platform, maybe this info might be helpful!

Thanks @Rajveer100

Failed Tests (10):
  Swift(windows-x86_64) :: DebugInfo/alloc_stack_arg.swift
  Swift(windows-x86_64) :: DebugInfo/archetype.swift
  Swift(windows-x86_64) :: DebugInfo/inlined-generics.swift
  Swift(windows-x86_64) :: IRGen/closure.swift
  Swift(windows-x86_64) :: IRGen/generic_tuples.swift
  Swift(windows-x86_64) :: IRGen/opaque_result_type_debug.swift
  Swift(windows-x86_64) :: IRGen/struct_resilience.swift
  Swift(windows-x86_64) :: IRGen/synthesized_conformance.swift
  Swift(windows-x86_64) :: SILGen/protocol_with_superclass.swift
  Swift(windows-x86_64) :: SILGen/protocol_with_superclass_where_clause.swift

cc: @compnerd

[Rajveer100]

#63574

[shahmishal]

#63574

Test failures:

Failed Tests (4):
  Swift(windows-x86_64) :: DebugInfo/dynamic_layout.swift
  Swift(windows-x86_64) :: IRGen/big_types_corner_cases.swift
  Swift(windows-x86_64) :: IRGen/closure.swift
  Swift(windows-x86_64) :: SILGen/vtable_thunks_reabstraction.swift

[shahmishal]

@compnerd has a fix for the memory issue - #63635

[compnerd]

The original PR that introduced it actually did fail on Windows, we would've caught it then.

[ktoso]

This isn't about windows but about two PRs "racing" and thus ending up with broken main:

• [Distributed] Explicitly ban __owned and other specifiers we don't support #66119 added a test that checked a diagnostic
• And at the same time this commit was landed: 4720673

CI tested those independently and both passed.

Once they got merged into main, the changed diagnostic from (2) caused the new test in (1) to fail on a typo.

This was resolved in: #66138

Technically merge queues prevent such issues, but would add even more potential wait time to merge changes.

Some "middle ground" would be amazing to find but I don't think something automated exists for this yet.

[shahmishal]

This can be closed now. The swift repo now requires Windows testing before merging PR.