FP: returnStdMoveLocal of unique_ptr

[Dan Workman]

We have a factory function that creates a unique_ptr to some struct. The only way to return this is via a std::move but cppcheck warns us about copy elision optimization

struct Foo
{
  std::string msg{"I'm unique :("};
};

std::unique_ptr<Foo> FooFactory()
{
  auto foo = std::make_unique<Foo>();
  return std::move(foo);
}

int main()
{
  auto foo = FooFactory();
  std::cout << foo->msg;
}

Output:

Checking cppcheckWarnings.cpp ...
cppcheckWarnings.cpp:16:20: warning: Using std::move for returning object by-value from function will affect copy elision optimization. More: https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#Rf-return-move-local [returnStdMoveLocal]
  return std::move(foo);

[CHR]

Seems to work just fine without std::move:
https://godbolt.org/z/M5j9KKeKd

[Dan Workman]

Ah I've simplified the snippet down too much then, this is closer to what we had:

std::unique_ptr<Foo> GetFirstFoo(std::vector<std::unique_ptr<Foo>>& foos)
{
  return std::move(foos.front());
}

int main()
{
  std::vector<std::unique_ptr<Foo>> foos;
  foos.emplace_back(std::make_unique<Foo>());
  auto foo = GetFirstFoo(foos);
  std::cout << foo->msg;
}

However this is rather unpleasant code as it leaves an invalid pointer in the vector so one might argue it's poor code

[CHR]

Thanks for reporting, this PR should fix it: https://github.com/danmar/cppcheck/pull/4244