Hi Tony,

Thanks for offering your interpretation of the racket reference manual.  Below is my attempt to test whether or not the input port does indeed become closed when the TCP connection is broken.

The broken TCP connection is provided by combining curl with GNU coreutils timeout.

I try testing the input port for closedness two different ways. The default code tests with port-closed-evt. The commented out code tests by calling port-closed? in a loop.  Neither method succeeds in detecting the broken TCP connection.

Thanks for any help.

Regards,


Jeff Henrikson


#lang racket

;; Tony Garnock-Jones, https://groups.google.com/g/racket-users/c/H43jr8QuM-4 ;;   "... if the connection breaks, the ports will be closed as usual. . . ."

(define (handle-by-event in out k)
  ; Discard the HTTP request header (up to blank line):
  (regexp-match #rx"(\r\n|^)\r\n" in)
  (displayln "about to sleep")
  (flush-output (current-output-port))
  (let ((evt (sync/timeout 10 (port-closed-evt in))))
    (if evt
      (begin
        (displayln "got close evt\n")
        (k))
      (begin
        (displayln "done sleeping")
        (flush-output (current-output-port))
        (display "HTTP/1.0 200 Okay\r\n" out)
        (display "Server: k\r\nContent-Type: text/html\r\n\r\n" out)
        (display "<html><body>Hello, world!</body></html>" out)
        (k)))))

;; same as handle-by-event but using port-closed? instead of port-closed-evt
(define (handle-by-polling in out k)
  (define (iter i)
    (when (> i 0)
        (printf "closed=~a\n" (port-closed? in))
        (sleep 1)
        (iter (- i 1))))
  ; Discard the request header (up to blank line):
  (regexp-match #rx"(\r\n|^)\r\n" in)
  (displayln "about to sleep")
  (flush-output (current-output-port))
  (iter 10)
  (displayln "done sleeping")
  (flush-output (current-output-port))
  (display "HTTP/1.0 200 Okay\r\n" out)
  (display "Server: k\r\nContent-Type: text/html\r\n\r\n" out)
  (display "<html><body>Hello, world!</body></html>" out)
  (k))


(define (accept-and-handle listener)
  (define-values (in out) (tcp-accept listener))
  ;; alternatively: (handle-by-polling in out (lambda ()
  (handle-by-event in out (lambda ()
    (displayln "handle finished")
  ))
  (close-input-port in)
  (close-output-port out))

(define (serve port-no)
  (define listener (tcp-listen port-no 5 #t))
  (define (loop)
    (accept-and-handle listener)
    (loop))
  (loop))

(serve 8000)

#|
    Test with curl and GNU coreutils timeout:
        timeout --foreground 3s curl http://127.0.0.1:8000
    actual output (handle-by-event version):
        about to sleep
        done sleeping
        handle finished
    expected output  (handle-by-event version):
        about to sleep
        got close evt
        handle finished

    Tested on:
        Racket v8.0 cs
        on Ubuntu 20.04.4 LTS
|#


Hi Jeff,

You can use `tcp-listen` and `tcp-accept` [0] to accept connections. Once accepted, the connection appears as a matched pair of ports, one for input and one for output, and if the connection breaks, the ports will be closed as usual. In some circumstances, you will get an exception such as "connection reset" (see `exn:fail:network`). If you need to reliably distinguish closed-by-peer, all-ok from closed-by-weird-network-weather, your application protocol has to handle that, there's nothing TCP can do for you there. HTTP does this via content-length and/or chunking, for example.

Cheers,
  Tony

[0]: https://docs.racket-lang.org/reference/tcp.html

On 6/30/22 11:34 AM, Jeff Henrikson wrote:
Racket users,

How do I accept an inbound TCP connection so that once I am processing it, if the connection is broken by the client or network, my program promptly finds out?

Regards,


Jeff Henrikson


--
You received this message because you are subscribed to the Google Groups "Racket 
Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to racket-users+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/racket-users/e3445bfa-8d75-bbf5-1900-226a67e71599%40gmail.com.

Reply via email to