workaround for spurious EOF on stdin when operating as subprocess of emacs

workaround for spurious EOF on stdin when operating as subprocess of emacs

[Thien-Thi Nguyen]

greetings,

below is a `(read-line port)' replacement that works around a bug in the
way the emacs lisp function `process-send-string' works.  the gist is
that emacs will send EOF after every 500 bytes or so, which means that
something like:

  (process-send-string PROC (make-string 512 ?A))

from emacs tends to show up (in guile) as #\A 508 times followed by EOF
followed by #\A four times.  this is generally undesirable.

note that this doesn't replace `read-line' per se, but the construct
`(read-line port)', which returns a string or EOF.

long-term, a better fix would involve examining the need for:

          /* If we sent just part of the string, put in an EOF
             to force it through, before we send the rest.  */
          if (len > 0)
            Fprocess_send_eof (proc);

in emacs/src/process.c (line 5664 in cvs revision 1.512) and change
emacs to flush via some out-of-band mechanism instead of via EOF.

thi


__________________________________________________________
(use-modules ((ice-9 rdelim) #:select (read-line)))

(define (get-line port)
  (define (next)
    (read-line port 'split))
  (let ((so-far ""))
    (let loop ((pair (next)))
      (let ((s (car pair))
            (d (cdr pair)))
        (cond ((and (eof-object? d)
                    (or (eof-object? s)
                        (and (string-null? s)
                             (string-null? so-far))))
               d)
              ((and (eof-object? d)
                    (string-null? s))
               so-far)
              ((eof-object? d)
               (set! so-far (string-append so-far s))
               (loop (next)))
              (else
               (string-append so-far s)))))))


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-user