Part of #611. Difficulty: good first issue. Audit baseline: master 3950416e (2026-09-12).
Current evidence
The Windows lists in windows_system_abi_is_target_scoped (tests/codegen_regressions.rs) and std_net_compiles_for_every_supported_socket_abi (tests/codegen_regressions.rs) currently cover GNU Windows entries but omit the two registered MSVC triples.
Scope
Extend these existing assertions to x86_64-pc-windows-msvc and aarch64-pc-windows-msvc when their LLVM backend is enabled. Keep GNU-specific linker assertions scoped to GNU; reuse the current target-availability helpers/patterns. This is a small test-coverage task, not a request to implement the C ABI.
Completion criteria
The targeted regressions run for MSVC under the enabled X86/AArch64 backend configurations and retain the existing symbol/layout assertions. Disabled targets are not requested. Keep non-Windows checks and the Linux rejection assertion. Test selections and output show which triples ran. Broad feature-isolated cleanup remains #490.
Dependencies and boundaries
#498; may start independently of native SDK provisioning because the added checks emit IR.
Part of #611. Difficulty: good first issue. Audit baseline: master
3950416e(2026-09-12).Current evidence
The Windows lists in windows_system_abi_is_target_scoped (tests/codegen_regressions.rs) and std_net_compiles_for_every_supported_socket_abi (tests/codegen_regressions.rs) currently cover GNU Windows entries but omit the two registered MSVC triples.
Scope
Extend these existing assertions to x86_64-pc-windows-msvc and aarch64-pc-windows-msvc when their LLVM backend is enabled. Keep GNU-specific linker assertions scoped to GNU; reuse the current target-availability helpers/patterns. This is a small test-coverage task, not a request to implement the C ABI.
Completion criteria
The targeted regressions run for MSVC under the enabled X86/AArch64 backend configurations and retain the existing symbol/layout assertions. Disabled targets are not requested. Keep non-Windows checks and the Linux rejection assertion. Test selections and output show which triples ran. Broad feature-isolated cleanup remains #490.
Dependencies and boundaries
#498; may start independently of native SDK provisioning because the added checks emit IR.