> if it's your intent, > fdexpand() shouldn't do malloc() unless nfiles > 1024 and > tests in fdcopy() and fdfree() are wrong. NDHISLOTS() steps up at 1024 boundary, doesn't it? enami.