Patch: Remove non-mapped location support

[Manuel López-Ibáñez]

On 25/02/2008, Chris Lattner <clattner@apple.com> wrote:
>  It emits it on the marked line, which is *after the wrong plus*.  I
>  think GCC has a long way to go before it will have accurate location
>  info in general.  Relying on input_location at all seems very dangerous.

You are right. We should perhaps change the API of diagnostics to
force the caller to give a explicit location. After all, a lot of the
calls now incorporate "%H" at the beginning, which doesn't look great
either. If you want the input_location, just pass that.

I guess that implementing the clang ranges may be much more difficult
with our current diagnostics machinery (Gabriel? Tom? any idea about
this?). However, we should be able to do the caret.

>  Clang puts the caret on the correct operator (or other relevant token)
>  every time, because we were aware of these issues from the beginning.

It shows. Congratulations. But I still think we can fix this in GCC
incrementally.

Cheers,

Manuel.