Improve use of parented_ptr

[daschuer]

This PR fixed some cases where a patented_ptr is constructed without a parent. This creates a pointer that is temporary without any owner.

I have moved the assertion to the constructor to avoid this situation in future.

The risk is that we with debug assertion enabled, Mixxx will crash in case I have missed a case. So a double check during review is required.

This is part of the pending discussion about the future of parented_ptr in: #13395

[Swiftb0y]

I have moved the assertion to the constructor to avoid this situation in future.

Wdyt about adding some compile time checks to make_parented? This code will fail if make_parented is called without any QObjects (if concepts don't work in all the compilers, you can also replace it with constexpr bool and static_assert). I know its a bit liberal in the sense that it can't know whether the passed object is actually the parent, but it can know that the code is definitely wrong if none of the parameters is a QObject.

template<typename... Args>
concept AnyQObject = (... || std::is_base_of_v<QObject, Args>);

template<typename T, typename... Args> 
parented_ptr<T> make_parented(Args...) requires (AnyQObject<Args...>)

[Swiftb0y]

whoops. sorry its

template<typename... Args>
constexpr bool AnyQObject = (... || std::is_base_of_v<QObject, std::remove_pointer_t<Args>>);

[Swiftb0y]

if you prefer I can also fix it locally and issue a PR to your branch once I got it working.

[Swiftb0y]

ah nice, the static check actually caught something

[daschuer]

All green finally. Thank you for your perfect idea with the compile time check.

[Swiftb0y]

LGTM. thank you.