Issue 199 - Label causes scope to collapse into parent

[yebblies]

Do what C does - introduce a scope.

https://d.puremagic.com/issues/show_bug.cgi?id=199

[monarchdodra]

The issue "seems" that it is now always creating a scope after a label. The failure can be reduced to:

void main()
{
    Lcontinue: size_t curallocsize = void;
    curallocsize = 1; //Error: undefined identifier curallocsize
}

Maybe this will spontaneously fix after if you solve 10199 first?

In any case, I can't wait for this to get fixed. It's been one of my oldest standing issues with the language :)

[yebblies]

Forgot to push after I fixed it!

[monarchdodra]

This is failing on test 4448, but I think the test is wrong:

int bug4448()
{
    int n=2;
    L1:{ switch(n)
    {
       case 5:
        return 7;
       default:
       n = 5;
       break L1;
    }
    int w = 7;
    }
    return 3;
}

L1 labels a block. Regardless of whether or not said block is scoped, it is not labeling the switch. It is albelling a block which (currently) is not creating a scope. The switch not being labelled, the break L1; is (I think) currently an accepts invalid either way.

This fix also fixes that. The test should be moved to fail_compilation.

[yebblies]

I agree.

[monarchdodra]

I'm surprised this compiles. What does it do? It's a switch that contains a labeled body? Does the grammar allow this?

It's the switch that should be labeled:

    int n=2;
    {
        L1: switch(n)
        {
            ...

[yebblies]

It doesn't compile. I've wrapped it in a do-while, the test case seemed to intentionally not label the switch.

[yebblies]

Although, it is legal for a switch to have a LabelStatement as its body.

[WalterBright]

Need corresponding doc pull.

[yebblies]

How do I document this? The old behavior was undocumented... "A label on a statement doesn't do anything weird to the scope" ???

[yebblies]

dlang/dlang.org#457

[yebblies]

@WalterBright ping

[yebblies]

Thanks!