The shift operators (<< and >>) are not processed correctly by antora-prep.py - the are treated as xrefs and generate dead 'xref:' text in the generated HTML for the docs site. Probably just a slightly over-aggressive regexp matching things that look like xrefs but are not.