cpp.Pointer requires a defined type / Breaks externs

[PDeveloper]

Revision: 4a6b30a
Date: 13/4/2016 9:06:06 AM

In this revision, this was added:

hxcpp/include/cpp/Pointer.h

Line 150 in a66838a

enum { elementSize = sizeof(T) };

This causes issues when compiling libraries that contain forward declared types, specifically snowkit/linc_sdl#8
Removing that one line allowed me to compile my project. In SDL sources, SDL_Window is defined as typedef struct SDL_Window SDL_Window; in include/SDL_video.h
SDL_Window is given a concrete definition ONLY in internals that shouldn't be used from the public API.

I would argue that requiring the sizeof() of a type breaks pointers, because of cases like the above, which in my view aren't all too uncommon.

[PDeveloper]

Currently exploring alternatives to using ::cpp::Pointer in snowkit/linc_sdl, perhaps this would be a better solution!

[PDeveloper]

Changing to use a cpp.RawPointer fixes things! Sorry for the disturbance! (Hadn't noticed cpp.RawPointer before :) )

[Simn]

I'm still a bit curious what exactly changed here and what the difference between Pointer and RawPointer is. That might be worth documenting?

[PDeveloper]

As far as I know, a RawPointer is a straight up C++ pointer Type* type_p;

The change here is that SDL_Window wasn't a complete type from the headers, and ::cpp::Pointer<T> does a sizeof(T), causing it to fail (expectedly).

[hughsando]

I think I can remove the sizeof from the header. You are right that it is an unnecessary restriction.
The difference between raw and cpp.Pointer is that the latter wraps the pointer in a struct to allow it to be passed Dynamically.

[PDeveloper]

Yep, changing to a RawPointer causes more issues relating to Dynamic!

Should I pull request or can you do a small commit for this? :)