Please describe the problem:
After 3 days of debugging I finally managed to reduce this weird bug down to a small understandable test case.

Basically the Vala code generator sometimes generates code for methods in such a way that exceptions (g_error codes) are suppressed and forgotten.

That this is a codegen bug because evident when you look at the generated code for ForwardingMethod(), it looks like this:
static gint someclass_ForwardingMethod (GError** error) {
        GError * inner_error;
        inner_error = NULL;
        return someclass_MethodTwentyTwo (&inner_error);
}
Clearly, whatever is written to inner_error will be ignored and forgotten, instead I think it was intended that it should say &error directly and that the inner_error should never have been declared?

Feel free to use this test case as a regression test in the future.

Steps to reproduce:
1. download the attached test case
2. execute "make run"

Actual:
The assert(false) is reached.

Expected:
Since the method always throws an error, it should be impossible to reach the assert.

Actual results:


Expected results:


Does this happen every time?


Other information: