Thanks for pointing this out! Maybe we can always print GNU attributes at the beginning (there are some test cases assume GNU attributes are printed at the end so they need to be updated.). Another idea is instead of hard code the print position of attribute we rely on source locations of the attributes. I think this would work for most cases but will run into situations for inherited attributes which Richard pointed out in previous reviews.
http://llvm-reviews.chandlerc.com/D395 _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
