[cpunion]

Fix the Wasm failures reported in llgo-compat/ci#120:

# escape, crypto/mldsa and go/types need larger fixed Fiber stacks:
llgo test -target=emscripten -goroutine-stack-size=1MB ./test/escape

Apply that budget only to these three packages on JS profiles. Recognize native-only test/cgo/c11 as not applicable, and include test/std/simd in the experimental suites already tested by native/W32-WASI jobs. Neither classification reports a pass.

Validation: the original ML-DSA stack overflow reproduces locally; all three affected packages pass on GoJS, Emscripten32 and Emscripten64 with the budget. Full acceptance-runner unit tests pass, including profile isolation and source accounting.

[fennoai]

Review summary

This is a small, well-scoped change to the dev/wasmstdlib test-orchestration harness: it raises the WebAssembly goroutine stack to 1MB for three stack-heavy js packages, excludes test/cgo/c11 from wasm source selection, and classifies test/std/simd as a separately-executed experimental SIMD suite. Tests are updated accordingly, and TestFullLargeStackCommandsAreTargeted is a well-designed addition that exercises the full six-profile matrix including a negative control (test/std/fmt).

I verified:

• The -goroutine-stack-size flag is a real build flag accepting values like 1MB (cmd/internal/flags/flags.go:231).
• The guard !p.Reference && p.GOOS == "js" correctly scopes the flag to J32-GoJS, J32-Emscripten, and J64-Emscripten only (not W32-WASI or reference profiles).
• The test/cgo/c11 and test/std/simd reason strings accurately match the packages' build constraints.
• Package tests pass (go test ./dev/wasmstdlib/).

No security, performance, or correctness issues were found. One minor comment-wording nit is noted inline.

Optional (non-blocking): the two adjacent blocks at full.go:93 and full.go:103 share the identical !p.Reference && p.GOOS == "js" guard and could be merged, though the current grouping-by-concern is defensible.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!