Why @test signatures allowed to be mismatched?

[alexveden]

The test case signature states and test runner checks the errors of running the test case:

def TestFn = fn void!();

But this test func absolutely normally runs:

//fn void(!) <- why ! is not enforced
fn void test_sub() @test {
}

[lerno]

That is a bug.

[lerno]

Or wait, it was a feature :( So the question is whether to allow it or not. A void test function will be turned into a void! function implicitly. So the question is whether to keep that.

[alexveden]

Or wait, it was a feature :( So the question is whether to allow it or not. A void test function will be turned into a void! function implicitly. So the question is whether to keep that.

You mean in user code it's possible to return errors implicitly, like this?

fn void! test_sub() @test {
    some_call_with_error()!; // <<< proparages the error and the test fails?
}

[lerno]

Yes, but the code allows you to also use void but then the actual signature will be turned into void!

This is a little broken right now, but could be fixed. (It works but has bad side effects)