Bug? This code should be clearly rejected!

Clang Frontend
#1

Hi all,

why does this wrong code compile with clang++ 3.6 (g++ rejects it
correctly)?

class Abc
{
public:
    virtual void foo() const = 0;
    virtual ~Abc() {}
};
class Impl : public Abc {
public:
    void foo() const {}
};
class B
{
public:
    void bar(Abc o)
    {
        o.foo();
    }
};
int main()
{
    B b;
    Impl i;
    b.bar(i);
}

#2

Hi all,

why does this wrong code compile with clang++ 3.6 (g++ rejects it
correctly)?

Yep. Looks like a bug to me. I think we had the same bug for return types
of abstract classes too & that was fixed a few months ago. Seems we haven't
fully flushed out these issues...

#3

    Hi all,

    why does this wrong code compile with clang++ 3.6 (g++ rejects it
    correctly)?

Yep. Looks like a bug to me. I think we had the same bug for return
types of abstract classes too & that was fixed a few months ago. Seems
we haven't fully flushed out these issues...

Is this already filed as bug? Or fixed in upstream version?

#4

>
>
>
> Hi all,
>
> why does this wrong code compile with clang++ 3.6 (g++ rejects it
> correctly)?
>
>
> Yep. Looks like a bug to me. I think we had the same bug for return
> types of abstract classes too & that was fixed a few months ago. Seems
> we haven't fully flushed out these issues...

Is this already filed as bug? Or fixed in upstream version?

The abstract return type one I think was filed/fixed, but I can't find the
specific details.

The abstract parameter type you've described/shown here doesn't seem to be
filed or fixed. If you'd like to file a bug (llvm.org/bugs) that'd be great.

- David

#5

    >
    >
    >
    > Hi all,
    >
    > why does this wrong code compile with clang++ 3.6 (g++ rejects it
    > correctly)?
    >
    >
    > Yep. Looks like a bug to me. I think we had the same bug for return
    > types of abstract classes too & that was fixed a few months ago. Seems
    > we haven't fully flushed out these issues...

    Is this already filed as bug? Or fixed in upstream version?

The abstract return type one I think was filed/fixed, but I can't find
the specific details.

The abstract parameter type you've described/shown here doesn't seem to
be filed or fixed. If you'd like to file a bug (llvm.org/bugs
<http://llvm.org/bugs>) that'd be great.

Ok, just filed a new bug ...