If an address_space attribute is defined in a macro, print the macro instead when diagnosing a warning or error for incompatible pointers with different address_spaces.
Originally started from http://lists.llvm.org/pipermail/cfe-dev/2018-August/058702.html where the original proposal was a new pragma, but instead just changing the way address_space gets printed in a diagnostic.
We now allow this for all attributes (not just address_space), and for multiple attributes declared in the same macro.
This comment is out of date. "[...] a type that was qualified by a qualifier written as a macro invocation" maybe?