[Patch] PR14995 Too eagerly rejects operator function templates

Richard Smith metafoo at gmail.com
Thu Jan 30 13:56:30 PST 2014


LGTM, thanks. The new test coverage looks great. Do you need someone to
commit this for you?

On Thu Jan 30 2014 at 1:48:13 PM, Rahul Jain <1989.rahuljain at gmail.com>
wrote:

>
> Hi Richard,
>
> Thanks for the inputs.
>
> I have updated the patch as per your review comments.
> Please if you could help by reviewing the same!
>
>
> Index: lib/Sema/SemaDeclCXX.cpp
> ===================================================================
> --- lib/Sema/SemaDeclCXX.cpp (revision 200475)
> +++ lib/Sema/SemaDeclCXX.cpp (working copy)
> @@ -10909,11 +10909,10 @@
>    //   increment operator ++ for objects of that type.
>    if ((Op == OO_PlusPlus || Op == OO_MinusMinus) && NumParams == 2) {
>      ParmVarDecl *LastParam = FnDecl->getParamDecl(FnDecl->getNumParams()
> - 1);
> -    bool ParamIsInt = false;
> -    if (const BuiltinType *BT =
> LastParam->getType()->getAs<BuiltinType>())
> -      ParamIsInt = BT->getKind() == BuiltinType::Int;
> +    QualType ParamType = LastParam->getType();
>
> -    if (!ParamIsInt)
> +    if (!ParamType->isSpecificBuiltinType(BuiltinType::Int) &&
> +        !ParamType->isDependentType())
>        return Diag(LastParam->getLocation(),
>                    diag::err_operator_overload_post_incdec_must_be_int)
>          << LastParam->getType() << (Op == OO_MinusMinus);
> Index: test/SemaCXX/overloaded-operator.cpp
> ===================================================================
> --- test/SemaCXX/overloaded-operator.cpp (revision 200475)
> +++ test/SemaCXX/overloaded-operator.cpp (working copy)
> @@ -452,3 +452,58 @@
>      Result = 1; // expected-error {{no viable overloaded '='}} //
> expected-note {{type 'PointerUnion<int *, float *>' is incomplete}}
>    }
>  }
> +
> +namespace PR14995 {
> +  struct B {};
> +  template<typename ...T> void operator++(B, T...) {}
> +
> +  void f() {
> +    B b;
> +    b++;  // ok
> +    ++b;  // ok
> +  }
> +
> +  template<typename... T>
> +  struct C {
> +    void operator-- (T...) {}
> +  };
> +
> +  void g() {
> +    C<int> postfix;
> +    C<> prefix;
> +    postfix--;  // ok
> +    --prefix;  // ok
> +  }
> +
> +  struct D {};
> +  template<typename T> void operator++(D, T) {}
> +
> +  void h() {
> +    D d;
> +    d++;  // ok
> +    ++d; // expected-error{{cannot increment value of type 'PR14995::D'}}
> +  }
> +
> +  template<typename...T> struct E {
> +    void operator++(T...) {} // expected-error{{parameter of overloaded
> post-increment operator must have type 'int' (not 'char')}}
> +  };
> +
> +  E<char> e; // expected-note {{in instantiation of template class
> 'PR14995::E<char>' requested here}}
> +
> +  struct F {
> +    template<typename... T>
> +    int operator++ (T...) {}
> +  };
> +
> +  int k1 = F().operator++(0, 0);
> +  int k2 = F().operator++('0');
> +  // expected-error at -5 {{overloaded 'operator++' must be a unary or
> binary operator}}
> +  // expected-note at -3 {{in instantiation of function template
> specialization 'PR14995::F::operator++<int, int>' requested here}}
> +  // expected-error at -4 {{no matching member function for call to
> 'operator++'}}
> +  // expected-note at -8 {{candidate template ignored: substitution
> failure}}
> +  // expected-error at -9 {{parameter of overloaded post-increment operator
> must have type 'int' (not 'char')}}
> +  // expected-note at -6 {{in instantiation of function template
> specialization 'PR14995::F::operator++<char>' requested here}}
> +  // expected-error at -7 {{no matching member function for call to
> 'operator++'}}
> +  // expected-note at -12 {{candidate template ignored: substitution
> failure}}
> +} // namespace PR14995
>
>
> Thanks,
> Rahul
>
>
> On Fri, Jan 31, 2014 at 12:35 AM, Richard Smith <metafoo at gmail.com> wrote:
>
> Please do add those other testcases. Please also add some tests which fail
> when we come to instantiate the operator++, such as:
>
> template<typename...T> struct A { void operator++(T...); };
> A<char> a;
>
> ... and ...
>
> struct A { template<typename...T> int operator++(T...); };
> int k1 = A().operator++(0, 0);
> int k2 = A().operator++('0');
>
>
> Also:
>
> -    if (!ParamIsInt)
> +    bool ParamIsDependent = false;
> +    if (LastParam->getType()->isDependentType())
> +      ParamIsDependent = true;
> +
> +    if (!ParamIsInt && !ParamIsDependent)
>
> This seems overly verbose. Please tidy this up; something like:
>
>   QualType ParamType = LastParam->getType();
>   if (!ParamType->isSpecificBuiltinType(BuiltinType::Int) &&
> !ParamType->isDependentType())
>
> ... would be better.
>
> On Thu Jan 30 2014 at 10:07:33 AM, rahul <1989.rahuljain at gmail.com> wrote:
>
> Hi Arthur,
>
> Yes it does fix the other two examples as well. I did check them too while
> working out the fix.
>
> I will add those too once I get a green signal for the fix :)
>
> Thanks,
> Rahul
>
>
> On 30-Jan-2014, at 11:29 pm, "Arthur O'Dwyer" <arthur.j.odwyer at gmail.com>
> wrote:
>
> Hi Rahul,
>
>   Does your patch fix the other two examples given in PR14995? Perhaps
> they should also be added as regression tests.
> (Just from eyeballing the code, I think your patch *does* fix them,
> including my example where the T is dependent on a class template instead
> of on a function template; but it might be nice to test explicitly.)
>
> LGTM.
>
> -Arthur
>
> On Thu, Jan 30, 2014 at 6:55 AM, Rahul Jain <rahul1.jain at samsung.com>wrote:
>
>
>
> Hi,
>
> This patch fixes PR14995 - Too eagerly rejects operator function
> templates.
>
> Please help review the same.
>
>
>
> Index: lib/Sema/SemaDeclCXX.cpp
> ===================================================================
> --- lib/Sema/SemaDeclCXX.cpp (revision 200465)
> +++ lib/Sema/SemaDeclCXX.cpp (working copy)
> @@ -10913,7 +10913,11 @@
>      if (const BuiltinType *BT =
> LastParam->getType()->getAs<BuiltinType>())
>        ParamIsInt = BT->getKind() == BuiltinType::Int;
>
> -    if (!ParamIsInt)
> +    bool ParamIsDependent = false;
> +    if (LastParam->getType()->isDependentType())
> +      ParamIsDependent = true;
> +
> +    if (!ParamIsInt && !ParamIsDependent)
>        return Diag(LastParam->getLocation(),
>                    diag::err_operator_overload_post_incdec_must_be_int)
>          << LastParam->getType() << (Op == OO_MinusMinus);
> Index: test/SemaCXX/overloaded-operator.cpp
> ===================================================================
> --- test/SemaCXX/overloaded-operator.cpp (revision 200465)
> +++ test/SemaCXX/overloaded-operator.cpp (working copy)
> @@ -452,3 +452,16 @@
>      Result = 1; // expected-error {{no viable overloaded '='}} //
> expected-note {{type 'PointerUnion<int *, float *>' is incomplete}}
>    }
>  }
> +
> +namespace PR14995 {
> +
> +  struct B {};
> +  template<typename ...T> void operator++(B, T...) {}
> +
> +  void f() {
> +    B b;
> +    b++;  // ok
> +    ++b;  // ok
> +  }
> +}
> +
>
>
>
>
>
> Thanks,
>
> Rahul
>
>
>
>   <201401302024474_BGFC2LL5.gif>
>
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140130/17f48696/attachment.html>


More information about the cfe-commits mailing list