On the other hand, it doesn’t hurt anything to run the tests twice does it? I’d feel better if this patch were *adding* test coverage as opposed to changing coverage On Wed, Dec 20, 2017 at 6:50 AM Gábor Horváth via Phabricator < revi...@reviews.llvm.org> wrote:
> xazax.hun added a comment. > > In https://reviews.llvm.org/D41444#960848, @a.sidorin wrote: > > > In https://reviews.llvm.org/D41444#960841, @xazax.hun wrote: > > > > > Is it possible that this will hide other problems? Wouldn't it be > better to run the tests twice once with this argument and once without it? > > > > > > I don't think so. In fact, without instantiation, we are not even able > to check semantic code correctness inside templates. So, we are solving > this problem as well. > > > E.g. the following code only compiles with `-fdelayed-template-parsing` > flag added: > > template<typename T> > struct Base { > int x; > }; > > > template<typename T> > struct Derived : Base<T> { > int f() { > return x; > } > }; > > But yeah, maybe it is not very likely that we hit such issues. > > > Repository: > rC Clang > > https://reviews.llvm.org/D41444 > > > >
_______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits