Difference in Mac and Windows versions of setGradientFill/fillEllipse when both alpha channel and RGB gradient used

[vladg]

Image 1 is basic image generated on Windows (software renderer).

Image 2 is generated on Windows by the following lines following the base image generation:

gd = ColourGradient(Colour(0x3bffffff), innerFlatRect.getCentreX(), innerFlatRect.getCentreY(),
    Colour(0x00000000), innerFlatRect.getCentreX(), innerFlatRect.getBottom(), true);

g.setGradientFill(gd);
g.fillEllipse(innerFlatRect);

Image 3 is basic image generated on Mac (Core Graphics renderer)

Image 4 is generated on Mac by the same lines as above following the base image generation.

You may see the same lines lead to different image on Mac and Windows.

Image 5 is generated on Mac by instead of Colour(0x00000000), Colour(0x00ffffff) was used.

Also on Windows both Colour(0x00000000) and Colour(0x00ffffff) lead to the same result.

As you may see there’s a problem or difference in implementation when both alpha channel and RGB gradients are used at the same time.

JUCE 5.4.7

[ttg]

I’m not really sure but you’ve might stumbled on a bug?

Try adding the following line:

jassert (CGColorSpaceGetNumberOfComponents(colourSpace) == 4);

To:

static CGGradientRef createGradient (const ColourGradient& g, CGColorSpaceRef colourSpace)

In juce_mac_CoreGraphicsContext.mm

It’ll fail. (at least on my machine).
For testing purposes try setting CGColorSpaceRef to be from:

CGColorSpaceCreateWithName (kCGColorSpaceSRGB);

Would be interesting to know if it helped on your case.
(another approach less nice is to premultiply the Colours).

[vladg]

There’s no difference between CGColorSpaceCreateWithName (kCGColorSpaceSRGB) or rgbColourSpace, in both cases CGColorSpaceGetNumberOfComponents(colourSpace) returns 3 in createGradient().

[matkatmusic]

@ttg

jassert (CGColorSpaceGetNumberOfComponents(colourSpace) == 4);

This failed for me as well in OS X 10.13.6 on a 27" imac from 2011. So, a normal imac, not a retina display or anything.

@ttg

(another approach less nice is to premultiply the Colours).

Can you explain this? I would really like to get matching gradients between what I’ve sketched out in an SVG editor and what I see in JUCE, and have no idea how to match them without using JUCE_LIVE_CONSTANT for the R, G, and B values while using Digital Color Meter to see what value I’m actually getting.