Re: [Testcase][Patch] http protocol

Rocco Caputo (troc@netrus.net)
Fri, 08 Jan 99 12:54:06 -0500


On Fri, 08 Jan 1999 08:48:31 -0800, Roy T. Fielding wrote:

>>1. At least one system still reads POST data by lines instead of by
>>Content-Length.  This leads to "it works in the browser but doesn't in
>>LWP" behavior.  The patch appends an "invisible" $CRLF (not accounted
>>for in Content-Length) to content if a trailing $CRLF isn't present.
>
>What braindead system would that be?

That would be the braindead system visited by the test case. :)

If the previous patch tastes bad, try this one instead.

--- http.pm.orig	Thu Nov 19 16:45:00 1998
+++ http.pm	Fri Jan  8 12:45:08 1999
@@ -140,8 +140,9 @@ sub request
 	die "write timeout" if $timeout && !$sel->can_write($timeout);
 	$n = $socket->syswrite($$cont_ref, length($$cont_ref));
 	die $! unless defined($n);
+	shutdown($socket, 1);
 	die "short write" unless $n == length($$cont_ref);
-	LWP::Debug::conns($buf);
+	LWP::Debug::conns($$cont_ref);
     }
     
     # read response line from server
__END__

-- Rocco Caputo / troc@netrus.net