I'm Sorry! I thought the code was too big to attach.. anyway here it is
My component "TSession" is a descendent from the TWCliSocket.
WaitStatus is a public property, that states what "state" the class is
currently at.
David
------------------------
procedure TSession.MyDataAvailable(Sender: TObject; ErrCode: Word);
var
bytesReceived: integer;
response, strReceived: WideString;
binReceive: pointer; // the pointer that will receive the buffer and
passed back to the user
chrReceived: array[0..MAX_BUF-1] of WideChar;
i: integer;
tempPtr: pointer;
begin
with (Sender as TSession) do
{ note: for WaitSendFile, WaitSendChunk, we do not process them here.
they are to be done in ProcessCommand }
if(WaitStatus=wsGetChunk)or(WaitStatus=wsGetMsg)then
begin
{ new version; solves the "multiple packet" problem }
if(TempBuffer=nil) then
begin
{$IFDEF VERBOSE}
consoleOut('Preparing Temp Buffer...');
{$ENDIF}
GetMem(TempBuffer, TempSize); // first packet
if(not assigned(ReceiveStr)) then ReceiveStr :=
TMemoryStream.Create;
ReceiveStr.Clear;
end;
ZeroMemory(TempBuffer, TempSize);
bytesReceived := Self.Receive(TempBuffer, ReceiveSize);
{$IFDEF VERBOSE}
consoleOut('Received '+inttostr(bytesReceived)+' bytes,
Remaining '+inttostr(ReceiveSize)+' bytes');
{$ENDIF}
if(bytesReceived>0) then
begin
{ given that the transfer was a successfull one... }
dec(ReceiveSize, bytesReceived);
ReceiveStr.Write(TempBuffer^, bytesReceived);
{ we can test whether we have finally done. }
{
------------------------------------------------------------------------ }
if(ReceiveSize<=0) then
begin
{$IFDEF VERBOSE}
consoleOut('All packets finished.');
{$ENDIF}
{ we just finished the last packet }
FreeMem(TempBuffer);
TempBuffer := nil;
if(WaitStatus=wsGetChunk) then
begin
{ we can now notify the user about it }
ReceiveStr.Seek(0, soFromBeginning);
tempPtr := ReceiveStr.Memory;
if(WideCompareStr(ReceiveHash, '0')<>0) then
begin
{ verify checksum }
if( GetMD5(ReceiveStr.Memory,
ReceiveStr.Size) = ReceiveHash ) then
TriggerOnChunkDone(tempPtr,
ReceiveTag, Success) else
TriggerOnChunkDone(tempPtr,
ReceiveTag, Fail);
end else
TriggerOnChunkDone(tempPtr, ReceiveTag,
Success);
WaitStatus := wsNone;
end
else
if(WaitStatus=wsGetMsg) then
begin
{$IFDEF VERBOSE}
consoleOut('Redirecting Text Message...');
{$ENDIF}
{ now we've got the complete string. let's
put it in ReceiveLine and let
ProcessCommand() handle it }
//DumpStream;
ReceiveStr.position := 0;
GetMem(tempPtr, ReceiveStr.Size);
ZeroMemory(tempPtr, ReceiveStr.Size);
ReceiveStr.ReadBuffer(tempPtr^,
ReceiveStr.Size);
ReceiveLine := PWideChar(tempPtr);
WaitStatus := wsNone;
FreeMem(tempPtr);
ProcessCommand;
end;
ReceiveStr.Clear;
ReceiveStr.Free;
ReceiveStr := nil;
end;
{
------------------------------------------------------------------------- }
end else
begin
consoleOut('WARNING! Transfer went wrong! Error
'+inttostr(GetLastError));
{ receiveSize is -1 }
{ something terrible happened. no more file transfer
will occur! so we better dispose of what we have }
ReceiveStr.Clear;
ReceiveStr.Free;
ReceiveStr := nil;
WaitStatus := wsNone;
Self.Flush;
SendText('fail');
end;
end
else
begin
// we got a "normal length" response.
//Application.ProcessMessages;
ZeroMemory(@chrReceived, sizeof(chrReceived));
bytesReceived := Self.Receive(@chrReceived,
sizeof(chrReceived));
ReceiveLine := WideString(chrReceived);
bytesReceived := bytesReceived + 1;
consoleOut('(received '+inttostr(bytesReceived)+'
bytes): '+ ReceiveLine);
ProcessCommand;
end;
end; // end procedure
------------------------
Francois PIETTE wrote:
>>I asked about a method to recognize the "splitted packets" and to join
>>them upon arrival of the last one. So far my protocol works fine, and
>>
>>
>
>This is really the most common task when dealing with TCP transmission. You
>have such "packet reassembly" in almost all ICS components. TWSocket itself
>has a LineMode that does this reassembly for lines. Lines are not necessary
>text line. It is just a convenient name to specify any variable length data
>with end of data marker.
>
>
>
>>Please would anybody give me a clue what's going wrong here?
>>
>>
>
>How would you receive help when we don't know anything about your code !
>It is likely that the problem is at the receiving part.
>
>--
>Contribute to the SSL Effort. Visit http://www.overbyte.be/eng/ssl.html
>--
>[EMAIL PROTECTED]
>http://www.overbyte.be
>
>
>----- Original Message -----
>From: "Kei" <[EMAIL PROTECTED]>
>To: "ICS support mailing" <twsocket@elists.org>
>Sent: Tuesday, November 01, 2005 9:45 AM
>Subject: [twsocket] transmission stress test
>
>
>
>
>>Hi, this is David.
>>
>>I asked about a method to recognize the "splitted packets" and to join
>>them upon arrival of the last one. So far my protocol works fine, and
>>all the data is assembled in an orderly manner. For large texts or
>>binary data, the clients knows how to handle it as follows:
>>The "Redirecting Text MEssage" just means, the text buffer is going to
>>be further processed in a function called "HandleResponse()".
>>
>>sender side-----------------------------------------
>>Line 7: out> txt 1561
>>Line 8: (Written 17 bytes)
>>Line 9: (received 16 bytes): sendtxt
>>Line 10: out> sendtxt
>>Line 11: out> msg AliasName and DriverName a...(too long to display)
>>
>>recipient side---------------------------------------------
>>Line 10: (received 18 bytes): txt 1561
>>Line 11: out> txt 1561
>>Line 12: out> sendtxt
>>Line 13: (Written 15 bytes)
>>Line 14: Preparing Temp Buffer...
>>Line 15: Received 1460 bytes, Remaining 3123 bytes
>>Line 16: Received 1460 bytes, Remaining 1663 bytes
>>Line 17: Received 203 bytes, Remaining 203 bytes
>>Line 18: All packets finished.
>>Line 19: Redirecting Text Message...
>>Line 20: out> msg AliasName and DriverName a...(too long to display)
>>Line 21: (!!) Command: msg AliasName and DriverName are m...(too long
>>to display)
>>------------
>>
>>The "sendtxt" is a command through which the recipient tells the sender
>>"hey I'm ready to get the long buffer".
>>This mechanism works very fine if I enter the command one by one. But
>>today I wanted to know if the recipient can handle successive "long msg"
>>requests by one single sender. So I tried a "stress test" to send the
>>1561 widechar long text on an interval of 10ms, to the recipient. There
>>are totally 200 requests to be sent consecutively.
>>
>>Sometimes it works fine, but sometimes .. the following happens
>>
>>Sender side-----------------------------------------
>>Line 1720: out> txt 1561
>>Line 1721: (Written 17 bytes)
>>Line 1722: Warning! Session is busy!
>>Line 1723: (received 25 bytes): sendtxt<JUNK>
>>Line 1724: out> sendtxt<JUNK>
>>Line 1725: Warning! Session is busy!
>>Line 1726: Warning! Session is busy!
>>Line 1727: Warning! Session is busy!
>> :
>> :
>>
>><JUNK> = some strange characters
>>
>>Recipent side---------------------------------------
>>Line 4112: (received 18 bytes): txt 1561
>>Line 4113: out> txt 1561
>>Line 4114: out> sendtxt
>>Line 4115: (Written 15 bytes)
>>Line 4116: Preparing Temp Buffer...
>>Line 4117: Received -1 bytes, Remaining 3123 bytes
>>Line 4118: WARNING! Transfer went wrong! Error 0
>>Line 4119: out> fail
>>Line 4120: (Written 9 bytes)
>>
>>Although the recipient sent 15 bytes of command "sendtxt" (line 4114),
>>the sender receives 25 bytes, with 10 bytes of junk following it.
>>Supposingly, the recipient should say "fail" to the sender and the
>>sender shoud receive it. But in this case, since the "sender" is kept
>>asking for more send (triggered by Timer, where interval=10ms), so there
>>are many "Warning! Session is busy!" messages appearing (because the
>>previous send-long-text action is still pending). Although the "fail"
>>message was sent from the other party, it seems that the "Sender" never
>>received this message.
>>
>>Also, out of so many times of test I've done, the "junk" fter the
>>"sendtxt" is ALWAYS THE SAME junk. So.. I think there might be some
>>problem in my code... perhaps is it something to do with multithreading?
>>
>>This problem only occur when I repeatly do the test. As in above you can
>>see the line number gets very big because I haven't got a problem from
>>line0... till now.. but I got to test the protocol under stress..
>>because it's going to be some sort of "database server."
>>
>>Please would anybody give me a clue what's going wrong here?
>>
>>Thanks!
>>
>>David
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>--
>>To unsubscribe or change your settings for TWSocket mailing list
>>please goto http://www.elists.org/mailman/listinfo/twsocket
>>Visit our website at http://www.overbyte.be
>>
>>
>
>
>
--
To unsubscribe or change your settings for TWSocket mailing list
please goto http://www.elists.org/mailman/listinfo/twsocket
Visit our website at http://www.overbyte.be
