Ok. So the fact that "%$(EXEEXT_FOR_BUILD)" becomes just "%" when EXEEXT_FOR_BUILD is empty and then matches literaly any target isn't really a problem then. A little bit surprising but I can see how it would work if any host code is built by sub-makefiles.
Anyway, there's nothing here that would work worse with rntcl than any other compiler as long as it isn't used as CC_FOR_BUILD (which it never should be). :-)
