Conversion of any RGB pixel to signed CMYK conversion is broken

[mloskot]

RGB pixels based on unsigned integral channels are not correctly translated to CMYK space with full range of signed integral channels.

The original implementation did not handle properly signed CMYK pixels, so this issue is independent from #406.

C++ Minimal Working Example

#include <boost/gil.hpp>
namespace gil = boost::gil;
int main()
{
    // black
    {
        gil::rgb8_pixel_t src{0, 0, 0};
        gil::cmyk8s_pixel_t dst;
        gil::color_convert(src, dst); // dst={0, 0, 0, 0}
                             // expected dst={-128, -128, -128, 127}
    }
    // white
    {
        gil::rgb8_pixel_t src{255, 255, 255};
        gil::cmyk8s_pixel_t dst;
        gil::color_convert(src, dst); // dst={0, 0, 0, -128}
                             // expected dst={-128, -128, -128, -128}
    }
    // blue
    {
        gil::rgb8_pixel_t src{0, 0, 255};
        gil::cmyk8s_pixel_t dst;
        gil::color_convert(src, dst); // dst={127, 127, 0, -128}
                             // expected dst={127, 127, -128, -128}
        print(1, src, dst);
    }
    // yellow
    {
        gil::rgb8_pixel_t src{255, 255, 0};
        gil::cmyk8s_pixel_t dst;
        gil::color_convert(src, dst); // dst={0, 0, 127, -128}
                             // expected dst={-128, -128, 127, -128}
    }
}

References

• Confirmed by Fix invalid conversion from RGB8 to CMYK32 due to overflow #470 while testing bug in Invalid conversion from RGB8 to CMYK32 #406

Environment

• Version (Git ref or <boost/version.hpp>): <=1.72, current develop

[Vivekyadavgithub]

this is my first time . Can u give me some pointers to fix this bug

[mloskot]

Can u give me some pointers to fix this bug

I'm not sure what kind of pointers you may need.
The bug is described quite clearly.
The bug description is accompanied with minimal, reproducible example which clearly points where is the problem (actual output vs expected output).

The task list is quite obvious, for example, it could be:

1. Compile the program
2. Run it, observe the actual vs expected output.
3. Add test case(s) confirm the bug (it can be based on the minimal example provided).
4. Start debugging it and reading the source code of the implementation
5. Find out why the conversion does not output expected results.
6. Fix it.
7. Confirm the tests pass, add more test cases...
8. Submit PR and request review
9. Address reviewers' comments and requests
10. Get the PR merged
11. Go for a coffee or a pint of ale and celebrate the great achievement!

[mloskot]

Fixed by #522