Implemented SA1121

[pdelvo]

Fixes #54. Unfortunatly the diagnostic itself was already implemented in #306. But this one contains a bunch of unit tests so it might be a good idea to use them

[sharwell]

This is currently blocked on #306. See #54 for additional information.

[sharwell]

Now that #306 is merged, can you rebase this and reduce it to a nice set of tests (plus enabling the diagnostic)? 👍

If you aren't comfortable with Git rebase, you can also do a pull from the master branch to bring in the updates.

[pdelvo]

This should bring in the tests. Im working on the code fix

[sharwell]

nameof(Object) for the right hand side here. 😉 The same for the items below.

[sharwell]

Can you add this test?

namespace Foo
{
  using Int32 = System.UInt32;
  class Bar
  {
    Int32 value = 3;
  }
}

A diagnostic should be reported, but the target built-in type is actually uint due to the misleading using alias directive.

[sharwell]

Also can you add this test?

namespace Foo
{
  using MyInt = System.Int32;
  class Bar
  {
    MyInt value = 3;
  }
}

[pdelvo]

What are the desired results here? Should using Int32 = System.UInt32; change to using Int32 = uint; or should the Int32 value = 3; change to uint value = 3; ? Or both? Currently it will report 2 diagnostics at MyInt value = 3; and the code fix im implementing is replacing it there

[sharwell]

1. Using alias directives are not allowed to use the language-specific keywords to refer to types. No diagnostic should be reported for the following line:

   using Int32 = System.Int32;

2. In the first example I posted (the misleading using alias directive), the using alias directive would not be altered by the code fix. The reference to the alias would be changed from Int32 to uint.

3. In the second example I posted, the using alias directive again would not change, and the reference to MyInt would be changed to int.

[pdelvo]

These new tests have some issues. The diagnostic analyzer is called twice for the same IdentifierToken. Do you know why this might happen?

[pdelvo]

I could not figure out why this happens. Maybe a roslyn bug?

[sharwell]

It's a bug somewhere in the unit test framework for this extension (modified from the one in the Roslyn templates). The diagnostics are only reported once for actual code, so I merged this and created an issue to find the bug in the unit test framework.