Home > Net >  Is this gcc 12.1 const problem a bug or feature: Attempts to call non-const function with const obje
Is this gcc 12.1 const problem a bug or feature: Attempts to call non-const function with const obje

Time:05-18

We're seeing C code that compiles successfully in gcc 11.3 and Visual Studio 2022 have issues with gcc 12.1. The code is in godbolt: https://godbolt.org/z/6PYEcsd1h (Thanks to @NathanPierson for simplifying it some.)

Basically, a template class is deciding to try to call a non-const base class function in a const function, even though a const overload is available. This appears to be some sort of compiler bug, but it could be some wierd new C rule I don't understand. Anybody know if this represents a compiler bug?

struct BaseClass
{
    // Commenting this non-const function out will also fix the compilation.
    int* baseDevice() { return nullptr; }
    const int* baseDevice() const { return nullptr; }
};

template <class ObjectClass>
struct DerivedClass : BaseClass
{

};

template <class ObjectClass>
struct TopClass : DerivedClass<ObjectClass>
{
  public:
    virtual int failsToCompile() const
    {
      // This should choose to call the const function, but it tries to call the non-const version.
      if (BaseClass::baseDevice())
         return 4;

      return 1;
    }
};

int main()
{
    TopClass<int> x;
}
<source>: In instantiation of 'int TopClass<ObjectClass>::failsToCompile() const [with ObjectClass = ConcreteObject]':
<source>:27:17:   required from here
<source>:30:32: error: passing 'const TopClass<ConcreteObject>' as 'this' argument discards qualifiers [-fpermissive]
   30 |       if (BaseClass::baseDevice())
      |           ~~~~~~~~~~~~~~~~~~~~~^~
<source>:14:15: note:   in call to 'MyDevice* BaseClass::baseDevice()'
   14 |     MyDevice* baseDevice() { return nullptr; }
      |               ^~~~~~~~~~
ASM generation compiler returned: 1

CodePudding user response:

Is this gcc 12.1 const problem a bug or feature

It's a bug. I filed a bug report and the issue has already been verified coming from this commit.

The ticket has been assigned and the resolution has a targeted milestone of version 12.2 - so we can hope for a quick fix.

  • Related