May be there is an influence in the blending with the underlying color of the fully transparent alpha image? (was it a fully transparent white or fully transparent black?)
could you try with this 
︎ image (you can’t see it but it’s there 
) ?
this is a transparent blue
No, that’s not the problem. I’ve been working with this setup for almost a year and the blue is red and red is blue behavior is about a week old. Very closely look at the sphere that’s drawn and you’ll see a gray line of lattitude. That’s the equator. Without blendMode(), you won’t see through an image and won’t see the equator. To show this, I took an earth.jpg image and load that instead of alphatest.png. I comment out the blendMode(SCREEN) and the blue ocean in earth.jpg is now red. The lat and lon are now red because I used GoToLoop’s workaround, but the equator is not visible, thus the image loses transparency without blendMode().
GoToLoop provided a link to Endianness and JDK 11. I ran java -version and javac -version. It says my jre is 1.8 and my jdk is the same, 1.8. I suppose the java.nio.ByteOrder is not unique to Java SE 11, or I couldn’t import it but I’m not sure “Alice”. 
Here’s a screenshot showing that blendMode(SCREEN) provides transparency. The land masses in the Northern Hemisphere would not be visible unless the earth.jpg is transparent, as provided by blendMode().
What did you change then?
PS: Nice images
I think you’ll have to ask Google and I’ll have to cry to them. It wasn’t till an update that this particular problem appeared. I’m interested in identifying and describing this problem so I can tell the “evil” developers what they broke. This wasn’t the only problem revealed by their update either. I have some processing programs that require reading data from an Arduino over USB and those devices are no longer visible.
it smells like an endianness issue, Intel and ARM platform with Linux used to not have the same endianness - may be they introduced a bug in the latest update and did not test as extensively on Intel as they did for ARM (seems ARM is the architecture for the future).
One thing I thought about trying is to reinstall 3.5.3 and see if something got corrupted in processing only. If I reinstall and the problem is gone, then the problem was in processing. Why or how would be a different question.
Yeah that’s likely not the issue but can’t hurt to double check
I noticed something in another program written in C++. Its a flight simulator called crrcsim that runs in linux.
I edited the config.xml file by taking out a section for video. When I ran the program, it also had the inverted red and blue. I put the video stuff back and it was normal again.
