An error message if it is an error is definitely a preferred response instead of an ICE so that needs reported either way; but I do not see why it should be considered recursive so I feel the response from gfortran is misleading at best; but not quite convinced it is standard-conforming anyway. Coarrays x. Variables x. Should it not be an error message? Nice to know nagfor works; as two compiler failures made me question if the case was standard although I could not think of a reason it was not.
nest...