Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:33614
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain optilian.com from 195.68.251.133 cause and error)
Message-ID: <475491BD.9040309@optilian.com>
Date: Tue, 04 Dec 2007 00:31:09 +0100
User-Agent: Thunderbird 2.0.0.9 (Windows/20071031)
MIME-Version: 1.0
To: internals@lists.php.net
References: <47520C61.605@optilian.com>
In-Reply-To: <47520C61.605@optilian.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Subject: Re: [PATCH] implement stream_set_write_buffer for sockets
From: vincent@optilian.com (Vincent NEGRIER)

I made a wrong assumption in the first patch and although it worked, 
stream_set_write_buffer() did return an error.

This one is fixes the return value issue :

--- main/streams/xp_socket.c.orig       2007-12-01 16:56:29.000000000 +0100
+++ main/streams/xp_socket.c    2007-12-03 21:07:02.000000000 +0100
@@ -254,6 +254,7 @@
         int oldmode, flags;
         php_netstream_data_t *sock = 
(php_netstream_data_t*)stream->abstract;
         php_stream_xport_param *xparam;
+       size_t size;

         switch(option) {
                 case PHP_STREAM_OPTION_CHECK_LIVENESS:
@@ -388,6 +389,14 @@
                                         return 
PHP_STREAM_OPTION_RETURN_NOTIMPL;
                         }

+               case PHP_STREAM_OPTION_WRITE_BUFFER:
+                       if (ptrparam)
+                               size = *(size_t *)ptrparam;
+                       else
+                               size = CHUNK_SIZE;
+                       php_stream_set_chunk_size(stream, size);
+                       return PHP_STREAM_OPTION_RETURN_OK;
+
                 default:
                         return PHP_STREAM_OPTION_RETURN_NOTIMPL;
         }



> Hello,
> 
> Currently stream_set_write_buffer() only works with file streams. If 
> used on socket streams it always returns -1 and does nothing. This can 
> be quite problematic when using datagram sockets because any datagram 
> bigger than the default 8Kb gets chopped and more than one datagram get 
> sent, eventually messing up with the receiving side.
> 
> This small patch adds support for sockets in stream_set_write_buffer(), 
> I have tested it and it fixed my problem, but if it needs refinements 
> i'd be glad to try and help more.
> 
> diff against 5.2.5 is here: 
> http://rectophobia.com/~six/socket_write_buffer.diff
> 
> Regards,
> Vincent