[C++ PATCH for 3.4, committed] Fix PR9188

[Kriang Lerdsuwanakij]

Joseph S. Myers wrote

>Why haven't you added a testcase for this?  Patches need either testcases,
>or justification for their absence (especially in cases like this where it
>looks like a testcase ought to be easy, using dg-bogus "expected
>expected").
>
>  
>
Sorry I forgot to mention this in the patch.  I felt it's not useful to have
a testcase for this one.  The patch only changed an error message so
a testcase could only detect broken diagnostics for this particular case 
but doesn't find duplicate message in general.

--Kriang