Hello Mirek,

We can easily implement RGBA constructor in C++11 branch... This standard changes a little bit POD rules. Now we can define constructor, but firstly we need to tell compiler that default constructor exists. I enclose demonstrative code:

#include <Core/Core.h>

#include <type_traits>

using namespace Upp;

namespace UppCpp11 {
struct RGBA {
	RGBA() = default;
	RGBA(byte r, byte g, byte b, byte a) {
		this->r = r;
		this->g = g;
		this->b = b;
		this->a = a;
	}
	
	byte b, g, r, a;
};

struct NotPodRGBA {
	NotPodRGBA(byte r, byte g, byte b, byte a) {
		this->r = r;
		this->g = g;
		this->b = b;
		this->a = a;
	}
	
	byte b, g, r, a;
};
}

CONSOLE_APP_MAIN
{
	UppCpp11::RGBA color(255, 255, 255, 255);
	Cout() << "Is RGBA POD? " << (std::is_pod<UppCpp11::RGBA>::value ? "True" : "False") << "!\n";
	Cout() << "Is NotPodRGBA POD? " << (std::is_pod<UppCpp11::NotPodRGBA>::value ? "True" : "False") << "!\n";
}

Result:

Is RGBA POD? True!
Is NotPodRGBA POD? False!

More information you can find on: http://stackoverflow.com/questions/4178175/what-are-aggregat es-and-pods-and-how-why-are-they-special/7189821#7189821 - second answer.

P.S.
Tested with g++ 4.8 using std=c++11 flag.

Sincerely,
Klugier

	RGBA(byte r, byte g, byte b, byte a) {
		ASSERT(r <= a && g <= a && b <= a);
		this->r = r;
		this->g = g;
		this->b = b;
		this->a = a;
	}

OK, but premultiplied alpha issue is still there... Should we add ASSERT

RGBA(byte r, byte g, byte b, byte a) {
    ASSERT(r <= a && g <= a && b <= a);
    this->r = r;
    this->g = g;
    this->b = b;
    this->a = a;
}

or perform conversion?

I don't know anything about "premultiplied alpha issue", but personally I think that implicit conversion would be safer for U++ users.

Moreover, I would like to noticed that it will be fine if we will have unit tests for RGBA. Each time the test failed we will know that something isn't ok in RGBA.

Google "premultiplied alpha". I guess that alone is a reason not to rush to have constructor....

I am not quite sure what you want to unittest about RGBA. But surely, it would be nice addition to have some RGBA *related* unittests, there is a couple of functions in Draw and in fact all Image processing is RGBA...

I mean if "premultiplied alpha" is Upp related issue it would be nice to have unittest for this. It will always remind us that something is wrong in this part of code. But if it isn't...

Have you googled it?