Comment is at wrong place.

The naming is a bit tricky here. random_value has different meanings for different type. The behavior is basically determined by the needs of the Random Attribute node and might not apply to different contexts. Therefore it might be better to leave these functions in node_geo_attribute_random.cc.

That doesn't seem to make too much sense.. Not sure if the seed is randomized before already.

Random colors shouldn't be generated like this. Usually you don't want transparency to be random.
Just remove it for now.

I'm not too fond of adding these methods to colors. Not sure..
Generally these are not operations one should do with colors.

It might be best to just not support colors in this node yet.

This operation also isn't well defined, but I guess it's fine. There should be an assert that checks for division by zero.

I'm not sure about the usefulness of the and and or modes. They basically set/unset approximately half of the elements that aren't already set/unset. I'd expect that usually when this behavior is needed, the user would want to have more control over the percentage. This should not be something the Random Attribute node is concerned about.

Do you have a use case for the difference and average mode?

The xor mode does not seem to be very useful. In practice it should behave exactly the same as GEO_NODE_ATTRIBUTE_RANDOMIZE_REPLACE_CREATE, if I'm not mistaken.

Interesting, this was my first ASAN crash at compile time :D
The {0, NULL, 0, NULL, NULL} entry is missing.

Alright, I removed them for now. Maybe the mixer in attribute_math is a step in the right direction, because this problem will come up again.

Hmm, I can't assert here because it might actually be dividing by 0, oops... At this point I'm just going to remove the operations that require division from this patch. I don't know of a use case for them anyway.

Eventually we might want to add a attribute_math::safe_divide<T>

Interesting, worked for me somehow! Thanks.

Agh, I forgot the word "doesn't"