[RyuJit Wasm] fix off by one error for br_table

[AndyAyersMS]

The count of switch targets does not include the default.

[AndyAyersMS]

@dotnet/jit PTAL

We can now run examples with switches, like so:

    static int sw(int n)
    {
        switch (n)
        {
            case 1: case 3: return 10;
            case 2: case 4: return 20;
            case 5: case 7: return 30;
            case 6: case 8: return 40;
            default: return 0;
        }
    }

via @kg's harness

Loading test-module.wasm...
Compiling test-module.wasm...
Instantiating test-module.wasm...
OK!
exports=["wasm_test_Program__addTwoFloats","wasm_test_Program__addTwoInts","wasm_test_Program__derefBytePtr","wasm_test_Program__derefIntPtr","wasm_test_Program__increaseFloat","wasm_test_Program__increment","wasm_test_Program__select","wasm_test_Program__sum","wasm_test_Program__sw"]
running 'exports.wasm_test_Program__sw(/* sp */ 1024, 6, /* pep */ 0)...
result was 40

[Copilot]

Pull request overview

This PR fixes an off-by-one error in the RyuJIT WASM code generation for switch statements using the br_table instruction.

Changes:

• Fixed the count parameter passed to emitIns_I for INS_br_table to correctly exclude the default case from the count

[adamperlin]

Can you potentially add a comment here explaining that this loop will handle emitting all the targets, including the default, even though in the above count we excluded it because of Wasm's convention? We could also consider pulling emitting the default out of the loop just to be extra clear here, though I know we also have the assert above to ensure we do have a default case.

[AndyAyersMS]

Yes. Take another look.