RFA: Another installed-tree testing patch

[Benjamin Kosnik]

>> As I've noted in the comment, this is all an ugly hack; the right
>> answer is for the compiler to have a -print-include-dirs option that
>> we could pass to g++, and then just include that list for the C
>> compiler.  But, we don't have that.
> 
> Yes. We've been around about this before, remember?
> 
> http://gcc.gnu.org/ml/gcc/2006-10/msg00328.html
> 
>> Tested with a cross to arm-eabi.  OK to apply?

I just want to make it obvious that:

I would prefer that we fix this correctly now, and add the 
-print-includes-dirs option, instead of waiting for the next breakage in 
6 months, when somebody reports breakage with different include paths.

This is now RFE 31864, ie,

http://gcc.gnu.org/PR31864

It seems fair to allow you to do the expedient thing in the meantime.

best,
-benjamin