On 24 September 2013 16:48, Fuyao Zhao <[email protected]> wrote: > I can't find any thing about re declare function twice with different > arguments in the standard. So I think it's a non defined behavior. > Currently, what clang does is always pick the latest attribute it sees. > > my change is to fix the case that one function has multiple format_arg > attribute > > and your suggestion is essentially about check the behavior when people > define a function with different attribute (it doesn't have to be > format_arg, it could be any of the attributes) > > Unless you could confirm current clang behavior is desired (always use the > latest declaration), I don't think it's good to write a test case for > non-defined behavior. Even we really want to do it, I want to do it in a > separate patch because people then could revert it separately.
For attributes that are gcc extensions, we normally just follow gcc unless there is a good reason to be different. I agree that if we already do the same thing gcc does it would be best to add a test for it first. That way it is clear your patch is not changing that behavior. Cheers, Rafael _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
