[cfe-dev] test case for stmt.label.

[sashan]

> 
> > Additionally I think that there's a bug in the clang error reporting since it
> > incorrectly identifies the location of the previous definition of the 'default'
> > label. If you compile the attached file with:
> > 
> > clang++ -fsyntax-only
> > 
> > You'll see these errors:
> > 
> > test/CXX/stmt.stmt/stmt.label/p1.cpp:22:5: error: multiple default labels in one switch
> >    default:; // expected-error{{multiple default labels in one switch}}
> >    ^
> > test/CXX/stmt.stmt/stmt.label/p1.cpp:23:5: note: previous case defined here
> >    default:; // expected-note{{previous case defined here}}
> >    ^
> > Note that clang thinks that the line location of the previous 'default' label is
> > after the first place the label 'default' is defined.  I think the note and
> > error messages should be swapped around.
> 
> I agree. The issue is that we store the list of case and default statements *backwards* in the AST, which I find rather unintuitive. It also means that we give warnings in the wrong order for code like this:
> 
> void f(char c) {
>   switch (c) {
>   case 1000: break;
>   case 1001: break;  
>   }
> }
> 
> Would you like to provide a patch that reverses the list of case/default statements, to fix both issues and make the AST cleaner?
> 
> 	- Doug

I did a fix for this by reversing the order that they are stored in the AST. I
modified SwitchStmt::addSwitchCase to do this, because that was the function
that added case and default statements. It resolved the issue where clang
would misdiagnose the line where the previous case/default statement was
defined. In other words it resolved the issue with this sort of code:

  switch (x)
  {
    default:;
    default:;
  }

But broke this code like this:

  switch (x)
  {
    default: //note the missing ;
    default:;
  }

Any ideas? What it looks like to me is that the token stream is fed backwards
into the parser from the first ; encountered but I'm not sure.

Thanks.