Original bug ID: 2792
Reporter: administrator
Status: closed
Resolution: fixed
Priority: normal
Severity: minor
Category: ~DO NOT USE (was: OCaml general)

Bug description

Hi,

I have written a small example that demonstrates the earlier mentioned bug
in "flush" that leads to sudden death after a "Connection reset by peer":

file: client.ml

open Unix
open Pervasives

let _ =
sleep 1; (* Guarantee that the server is ready for action )
let s = socket PF_INET SOCK_STREAM 0 in
let sockaddr = ADDR_INET (inet_addr_any, 9876) in
connect s sockaddr;
let ic = in_channel_of_descr s in
let oc = out_channel_of_descr s in
sleep 1; ( Guarantee that the server process has terminated )
(try
( Try to read from dead server process. The server will respond
with an RST segment, which triggers an ECONNRESET-error. )
print_endline (input_line ic); flush stdout;
with exc ->
print_endline ("caught exception: " ^ Printexc.to_string exc);
flush stdout);
print_endline "before output"; flush stdout;
output_string oc "x"; ( Put dummy data into buffer )
flush oc; ( XXX: Flush buffer: we crash here )
print_endline "after output"; flush stdout ( Should not be reached *)

file: server.ml

open Unix

let _ =
let s = socket PF_INET SOCK_STREAM 0 in
setsockopt s SO_REUSEADDR true;
let sockaddr = ADDR_INET (inet_addr_any, 9876) in
bind s sockaddr;
listen s 1;
(* Wait for connection, but do not accept *)
select [s] [] [] 1000.0

Just compile the two examples, start the server, then the client.
It may take more than one invocation.

Best regards,
Markus

--
Markus Mottl http://www.oefai.at/~markus markus@oefai.at