One minor thing: at_check_newline is trace-specific (it doesn't merely check newlines; it does other stuff). So we might as well migrate the "test -n" business into at_check_newline, no? It'll make the code shorter and easier to read. And perhaps its name could be changed to "at_check_traceable" or something like that.I disagree because you can omit the test completely if the nontraceability is determined at m4 time.
Huhm, sorry, I was not thinking. I'll prepare a follow-up. Paolo
