_gdContributionsCalc() always uses DEFAULT_BOX_RADIUS

[cmb69]

Even though each filter is supposed to have its own radius, _gdContributionsCalc() always uses DEFAULT_BOX_RADIUS. That appears to be a bug.

[cmb69]

According to filter.c from GGemsIII this is certainly a bug.

Test Program

#include <gd.h>


int main()
{
    gdImagePtr src, dst;
    FILE *fp;

    src = gdImageCreateTrueColor(100, 100);
    gdImageFilledRectangle(src, 0,0, 99,99, gdTrueColorAlpha(255, 255, 255, gdAlphaOpaque));
    gdImageEllipse(src, 49,49, 40, 40, gdTrueColorAlpha(0, 0, 0, gdAlphaOpaque));

    gdImageSetInterpolationMethod(src, GD_CATMULLROM);
    dst = gdImageScale(src, 200, 200);

    fp = fopen("scale.png", "wb");
    gdImagePng(dst, fp);
    fclose(fp);

    gdImageDestroy(src);
    gdImageDestroy(dst);

    return 0;
}

Expected Result

Actual Result

[cmb69]

Better reproducer:

Test Program

#include <gd.h>


int main()
{
    gdImagePtr src, dst;
    FILE *fp;

    src = gdImageCreateTrueColor(11, 11);
    gdImageFilledRectangle(src, 0,0, 10,10, gdTrueColorAlpha(255, 255, 255, gdAlphaOpaque));
    gdImageEllipse(src, 5,5, 8, 8, gdTrueColorAlpha(0, 0, 0, gdAlphaOpaque));

    gdImageSetInterpolationMethod(src, GD_CATMULLROM);
    dst = gdImageScale(src, 132, 132);

    fp = fopen("scale-2.png", "wb");
    gdImagePng(dst, fp);
    fclose(fp);

    gdImageDestroy(src);
    gdImageDestroy(dst);

    return 0;
}

Expected Result

Actual Result

[pierrejoye]

Master fixes this and also does a little cleanup of the gd_interpolation.