1. Open Roslyn.sln at the root of the Roslyn repo in VS2017.
2. Open source file <%ROOT%>srcCompilersCSharpTestSemanticSemanticsImplicitlyTypedLocalsTests.cs.
3. Add the following new unit test to the source file:

[Fact]
public void VarInferredTypeNotApparent()
{
  var source = @"
  class Class
  {
    void M(int x, string y)
    {
      var z = x + y;
    }
  }";

  CreateCompilationWithMscorlib(source).VerifyDiagnostics();
}

4. Build the test project and execute the unit test on a command-line console using the command line copied from the project's Debug property page, appending -method switch for the newly added unit test:

<%USERS_FOLDER%>.nugetpackagesxunit.runner.console2.2.0-beta4-build3444	oolsxunit.console.x86.exe "<%ROOT%>BinariesDebugUnitTestsCSharpCompilerSemanticTestRoslyn.Compilers.CSharp.Semantic.UnitTests.dll" -html "C:
oslynBinariesDebugUnitTestsCSharpCompilerSemanticTestxUnitResultsRoslyn.Compilers.CSharp.Semantic.UnitTests.html" -noshadow -method Microsoft.CodeAnalysis.CSharp.UnitTests.ImplicitlyTypedLocalTests.VarInferredTypeNotApparent

5. Verify the unit test fails with the missing CS0823 diagnostic:

Expected:
Actual:
  // (6,7): warning CS0823: Use an explicit type for declaration as the initializer type 'string' is not apparent due to conversions
  // var z = x + y;
  Diagnostic(ErrorCode.WRN_ImplicitlyTypedVariableNotRecommended, "z = x + y").WithArguments("string").WithLocation(6, 7)

Diff:
 ++> Diagnostic(ErrorCode.WRN_ImplicitlyTypedVariableNotRecommended, "z = x + y").WithArguments("string").WithLocation(6, 7)

6. Add the missing diagnostic as an argument to the VerifyDiagnostics invocation in our unit test:

7. Re-execute the unit test by repeating step 4 and verify that the test passes now.

8. Add another unit test to verify that the diagnostic does not fire for a case where the initializer binary expression has no implicit conversions:

[Fact]
public void VarInferredTypeApparent_NoDiagnostic()
{
  var source = @"
  class Class
  {
    void M(int x, string y)
    {
      var z = (string)(x + y);
    }
  }";

  CreateCompilationWithMscorlib(source).VerifyDiagnostics();
}

9. Execute the new unit test and verify that it passes.