PATCH: private ivars
Adrian Prantl
aprantl at apple.com
Mon Feb 18 16:40:03 PST 2013
On Feb 18, 2013, at 4:28 PM, David Blaikie <dblaikie at gmail.com> wrote:
> On Mon, Feb 18, 2013 at 4:24 PM, Adrian Prantl <aprantl at apple.com> wrote:
> On Feb 15, 2013, at 3:59 PM, Dmitri Gribenko <gribozavr at gmail.com> wrote:
> >
> > + ObjCInterfaceDecl* decl = cast<ObjCInterfaceType>(Ty)->getDecl();
> > + if (decl)
> >
> > cast<> can not return null (it will either succeed or assert in
> > +Asserts mode, or just produce a wrong value in -Asserts). Use
> > dyn_cast (that returns null on failure) or drop the check -- whatever
> > is appropriate. And there's a dyn_cast idiom:
> >
> > if (Foo *F = dyn_cast<Foo>(Blah))
> > ... use F...
>
> I'm actually not checking the return value from cast<>() but the the result of getDecl() [which just happens to also return an ObjCInterfaceDecl]. Do you think I should use a temporary for the result of the cast to make it clearer?
>
> Then you're missing test cases, because dyn_cast doesn't propagate null. It should fail if you pass it null.
>
> You probably want cast_or_null.
At least in my particular case I'm not so sure about this. The code fragment we are talking about looks like this:
switch (Ty->getTypeClass()) {
...
case Type::ObjCInterface: {
ObjCInterfaceDecl* Decl = cast<ObjCInterfaceType>(Ty)->getDecl();
if (Decl)
...
}
At this point Ty should always be nonzero and since we switch over the TypeClass the cast<> should always be safe, right?
-- adrian
More information about the cfe-commits
mailing list