error: redefinition of 'struct Vector2/3/4'

[CodeSmith32]

Surely I'm doing something simple wrong.

I just started using raylib (seems like a really decent library!), but whenever I include both raylib.h and raymath.h I get redeclaration errors for some of the math types.

error: redefinition of 'struct Vector2'
...
error: redefinition of 'struct Vector3'
...
error: redefinition of 'struct Vector4'
...
error: redefinition of 'struct Matrix'
...

Is this to be expected? Both files do in fact declare all the math types, afaict, but only raymath.h appears to declare the math operators.

In general, how should I set up my project? Many of my files only need the math types and operators, whereas a few of my files need some of the other parts of raylib.h. But as soon as I include headers that pull in both, I get these issues.

Should I only include raylib.h? Do I need to define some special parameter?

TIA!

[CodeSmith32]

Ok, so for my header that includes two other headers that pull in raylib.h and raymath.h, I added manual includes for both, at the very top, in the order,

#include "raylib/raylib.h"
#include "raylib/raymath.h"

This appears to have worked. Is there a reason including raymath.h before raylib.h is not allowed? And, if this is an intentional decision, shouldn't there be a more explicit warning, instead of generating redefinition errors?

If this qualifies for a bug report, please let me know.

Thanks!

[OussamaTeyib]

This is intended behaviour.

raylib is a standalone library, and raymath was designed to work alongside it — though it can also be used standalone. For this reason, raymath checks whether each symbol is already defined before defining it. for exemple:

raylib/src/raymath.h

Lines 119 to 126 in b2ea5ea

	#if !defined(RL_VECTOR2_TYPE)
	// Vector2 type
	typedef struct Vector2 {
	float x;
	float y;
	} Vector2;
	#define RL_VECTOR2_TYPE
	#endif

raylib doesn't do the same, and as a standalone lib, it isn't required to. It instead sets defines after declaring its types, which raymath uses to detect them:

raylib/src/raylib.h

Lines 160 to 170 in b2ea5ea

	// NOTE: Set some defines with some data types declared by raylib
	// Other modules (raymath, rlgl) also require some of those types, so,
	// to be able to use those other modules as standalone (not depending on raylib)
	// this defines are useful for internal check and avoid type (re)definitions
	#define RL_COLOR_TYPE
	#define RL_RECTANGLE_TYPE
	#define RL_VECTOR2_TYPE
	#define RL_VECTOR3_TYPE
	#define RL_VECTOR4_TYPE
	#define RL_QUATERNION_TYPE
	#define RL_MATRIX_TYPE

So the rule is: always include raylib.h before raymath.h. That way the defines are set, and raymath skips the duplicate type definitions cleanly.

[CodeSmith32]

Hey @OussamaTeyib / @raysan5,

Got it. Makes sense!

Is there any reason why you wouldn't want to add a check + error if raymath.h / rlgl.h is improperly included first?
e.g., something simple, like,

#ifdef RAYMATH_H
#error "raymath.h must be included after raylib.h"
#endif

#ifdef RLGL_H
#error "rlgl.h must be included after raylib.h"
#endif

Thanks!

[raysan5]

Unneeded extra code pollution

[CodeSmith32]

@raysan5
Sure. This is at least documented somewhere, right? (I'm sure I just missed it..)

[raysan5]

not sure but feel free to add it in the FAQ, in the Wiki, it's open

[CodeSmith32]

Sounds good! I've edited this page, since it talked a bit about the external modules:

• https://github.com/raysan5/raylib/wiki/raylib-architecture

I kinda guessed that proper inclusion order - currently or in the future - is important for other modules besides just raymath..? But if it'll only ever apply to raymath, then my update could probably be tweaked..

Thanks @raysan5!