The expected result has been out of sync with what glcpp produces for some time; glcpp's actual result seems to be correct and is very close to GCC's cpp. Updating this will make it easier to catch regressions in upcoming commits.