[C++ Patch/RFC] Final part of c++/26099, __is_convertible_to

[Paolo Carlini]

Andrew Pinski wrote:

> Well for one, you should not be changing the generic diagnostic
> functions at all, it just makes the code harder to understand what is
> really going on.  I in some cases put a break point in error and
> warning, so now I will get an extra stop in the debugger which is
> wrong as we are not producing an error.
>
> You should add a new variable to perform_implicit_conversion that says
> we don't need diagnostic, and then change all the places where it
> would do an error/warning, not to do the warning and if it was an
> error, just return error_mark_node and check expr.  Right now your
> change look very very wrong as you have the generic diagnostic
> function says when an error occured when that is not exactly what you
> want.

You are funny.

Paolo.