personally, I feel that the target .o's should be renamed (perhaps to
".to") and not the local ones. Having local files named ".o" which are 
not for the local architecture violates the rule of least surprise.

.pm

Jonathan Stone writes:
> 
> >Hm.  I thought about `.go' meaning "generator object," but that's been done
> >on some hosts for `-g' objects.
> >
> >However, `.oo' hasn't been done, and this one sounds more interesting to
> >use.  Perhaps `.xo'?
> 
> .lo, for "local"-cpu object files in an obj tree where .o files are
> otherwise for the target CPU. I think it conflicts with existing
> targets in GNU Makefiles, though.
> 
> Does the stock FSF gcc or egcs already use an extension for
> native-host object files, or are still supposed to compile and run
> those (e.g., make float.h) on the target host?