[mparrett]

Follow-up to #1080, whose review asked for a test of the _lgStorage shim.

TestWorkerStorageShim cuts workerCode out of the embedded lg-host-core.js and runs it under node with postMessage stubbed. It sends the worker an init carrying a two-key storage snapshot and a wasmExecJS that throws, so the worker stops after building _lgStorage and before any wasm loads. The test then checks:

• length and key(i) order, including after an add, an overwrite, and a remove;
• key past the end and getItem of a missing or removed key read null;
• setItem stores String(v);
• the storage-set and storage-remove messages relayed to the page, in order.

Since the test runs the shipped source, a change to the shim needs no matching change to a fixture. It skips when node is not on PATH.

Verification

• go test ./pkg/rt/wasm/ passes.
• With the keys = null reset on add removed from the shim, the test fails on the key order after the add.
• GOOS=js GOARCH=wasm, linux, and plan9 builds pass.

🤖 Generated with Claude Code

[mparrett]

Self-check: No actionable findings.

The test exercises the shipped worker source rather than a duplicate shim, initializes the storage snapshot through the real init handler, and verifies key order/cache invalidation, missing-key nulls, value string coercion, and the relayed write/remove messages. Source extraction has explicit failure checks, and the Node subprocess is time-bounded. The PR changes no production behavior.

Validation at 90468cd: TestWorkerStorageShim ran under Node and passed (not skipped). I also temporarily removed the add-path key-cache invalidation in lg-host-core.js; the test failed on the after-add and after-overwrite key lists as expected. Restored the original source and confirmed no diff remained. Existing CI is green.