Jextract has two different backends - one generates bytecode, while the other generates sources. The bytecode generation, based on dynamic constants, is more efficient, but it's typically harder to work with. IDEs have issues in parsing classfiles containing dynamic constants, and there's no path to add things like javadoc.

For these reasons, it would be best for jextract to find a way to have a source-based generation scheme that is lazier than the one we currently have.