Patch: Remove non-mapped location support

[Tom Tromey]

>>>>> "Manuel" == Manuel López-Ibáñez <lopezibanez@gmail.com> writes:

Manuel> Well, that seems the way to get perfect caret location. But I think we
Manuel> don't need that, we need to have a way to enable caret diagnostics and
Manuel> then incrementally improve it.

You can run with -fshow-column.  This isn't as pretty as the caret but
if you step through errors displayed this way you will quickly see
problems.

Not that I disagree... if you want to write the caret code, I'm all
for it.

Manuel> I am thinking about forcing diagnostics to take a location parameter.
Manuel> This will make more explicit where is the location coming from. Also
Manuel> it will avoid "%H". Initially, most diagnostics will just pass
Manuel> input_location but we can change that incrementally.

I do like the layer of indirection provided by having the diagnostic
machinery know how to extract the location from a tree.  I may need
this for the incremental compiler, though I'm not certain yet.

Tom