Thanks for the info, I will look into this. I am still having trouble
testing this properly.
Simon
-----Original Message-----
From: Erik Svensson [mailto:erik_at_phlogiston.se]
Sent: March-07-08 4:09 PM
To: users_at_grizzly.dev.java.net
Subject: Re: [Q] Testing ProtocolFilter
And talking of releaseBuffer():
I notice something here.
157 public boolean releaseBuffer() {
162 if(!hasMoreBytesToParse())
163 {
168 this.byteBuffer.clear();
169 this.position = 0;
170 this.byteBufferLimit = 0;
171 }
172 else
173 {
174 byteBuffer.position(position);
175 }
176
177 return hasMoreBytesToParse();
178 }
(I've snipped the LOG lines for brevity)
hasMoreBytesToParse() means 'there are bytes in this buffer that I
haven't looked at and i don't know if it contains part of a message,
one complete message or more than one complete message'. At this
point this should not be true. Here you should return 'needs more
data' (if the buffer still contains bytes that can't be parsed to a
complete message).
Here's what I think is happening:
hasNextMessage() is called and we end up here:
111 else
112 {
117 isExpectingMoreData = true;
118 position =
bufferContentString.getBytes().length;
119 }
120 }
saving the position == byteBuffer.limit() and setting
isExpectingMoredata == true and returning false.Then releaseBuffer is
called. There we start (see at the top) by calling hasMorebytes():
20 public boolean hasMoreBytesToParse() {
21 boolean hasMoreBytesToParse = position <
byteBufferLimit;
28 return hasMoreBytesToParse;
29 }
Note line 21! We check that the position is less than the limit().
Since we've set position to limit this returns false.
which leads to that we inadvertently clear the buffer when we shouldn't.
if you change hasMoreBytesToParse to isExpectingMoreData() I think
this might work more in the lines of how you were thinking.
hope this helps
and for your question: I can't remember :-) But it shouldn't be
called more than once since if you do compact twice in the same
buffer the buffer will be overwritten in subsquent puts.
cheers
/Erik
On Mar 7, 2008, at 21:35 , Simon Trudeau wrote:
> Your logic makes sense. Those lines were an almost cut and paste from
> ProtocolParserTest.java line 130-133. I will remove those lines.
>
> By the way, when I parse a complete message, the releaseBuffer()
> method
> gets called twice... which is a bit weird. I am having a bit of
> trouble
> tracing this since it is multithreaded... Have you ever encountered
> something like this?
>
> Thanks,
>
> Simon
>
> -----Original Message-----
> From: Erik Svensson [mailto:erik_at_phlogiston.se]
> Sent: March-07-08 3:11 PM
> To: users_at_grizzly.dev.java.net
> Subject: Re: [Q] Testing ProtocolFilter
>
> To keep up the flowing commentary... :-)
>
> This is turning out to be something of a flow-of-consciousness thing!
>
> On lines 65 - 70:
>
> 65 ByteBuffer roDuplicateBuffer =
> byteBuffer.asReadOnlyBuffer();
> 66
> 67 if (byteBuffer.position() == 0){
> 68 isExpectingMoreData = true;
> 69 return false;
> 70 }
>
> I'm wondering a bit about this. If the incoming buffers position is
> '0' (before flip()) means that the buffer is empty
> (ie the eq to hasRemaining() == false). 'isExpectingMoreData' means
> that you need more data to fill out a complete message.
> I cannot see how you can determine that from just looking at the
> position of the byte buffer.
> I might be misunderstanding somehting from your program logic.
>
> cheers
> /Erik
>
> On Mar 7, 2008, at 20:51 , Simon Trudeau wrote:
>
>> You are most certainly right, a curly brace is missing on line 94.
>> Thanks,
>>
>> Simon
>>
>> -----Original Message-----
>> From: Erik Svensson [mailto:erik_at_phlogiston.se]
>> Sent: March-07-08 2:49 PM
>> To: users_at_grizzly.dev.java.net
>> Subject: Re: [Q] Testing ProtocolFilter
>>
>>>
>>
>> Simon,
>>
>> I found something a bit odd but that might be because you've edited
>> the code before posting it.
>> In the hasNextMessage() method there's this snippet of code:
>>
>> 86 }
>> 87 /*
>> 88 * Validates that the buffer contains at least 2 bytes
>> so we can proceed to input validation
>> 89 */
>> 90 if(bufferContent.length < 2)
>> 91 {
>> 92 isExpectingMoreData = true;
>> 93 return !isExpectingMoreData;
>> 94
>> 95 // 1.1)
>> 96 if(CR == bufferContent[0])
>> 97 {
>> 98 if(LF == bufferContent[1])
>> 99 {
>> 100 // 1.1.1)
>> 101
>> if(bufferContentString.contains(PASS.toString()))
>> 102 {
>> 103 if(LOG.isTraceEnabled())
>> 104 {
>> 105 LOG.trace("Found:
>> PASS");
>> 106 }
>> 107 isExpectingMoreData = false;
>> 108 message = PASS.toString();
>> 109 position = PASS.toBytes().length;
>> 110 }
>>
>> Note on line 93 the return without any closing curly bracket.
>> However, this shouldn't compile
>> as the code below is unreachable. Am I correct in assuming that there
>> should be
>> a closing bracket on line 94?
>> (the indention is a bit off due to being cut'n'pasted into my mail
>> client).
>>
>> cheers
>> Erik
>>
>>
>>
>>
>>
>>> Thanks!
>>>
>>> Simon
>>>
>>> -----Original Message-----
>>> From: Erik Svensson [mailto:erik_at_phlogiston.se]
>>> Sent: March-07-08 1:35 PM
>>> To: users_at_grizzly.dev.java.net
>>> Subject: Re: [Q] Testing ProtocolFilter
>>>
>>>
>>> On Mar 7, 2008, at 17:03 , Simon Trudeau wrote:
>>>
>>>> Have you filled up a ticket about the compacting issue?
>>>
>>> That I haven't.
>>>
>>>> Have you posted the code for your protocol parser? Here's mine,
>>>> maybe
>>>> you can give me some tips! :.)
>>>
>>> I haven't done that either. We're looking at using grizzly inhouse
>>> and it was in that context I wrote
>>> the protocol parser. I had thoughts about writing a tutorial/blog
>>> post about it but there aren't all that much time right now.
>>> At work I have more pressing concerns. I
>>>
>>> I attach my protocol parser to this mail so you can look at it. It's
>>> not tested with 1.7.2, though.
>>>
>>> Basically it uses a class XFlowMessage that contains the parser to
>>> parse the byte buffer into xflow messages.
>>>
>>>> I will add your buffer compacting tip!
>>>
>>> I'll look at your code and see if I can come up with something. Just
>>> the thing the wife wants me to do on a friday evening :-)
>>>
>>> cheers
>>>
>>> /Erik
>>>
>>>
>>>
>>>>
>>>> Simon
>>>>
>>>> -----Original Message-----
>>>> From: Erik Svensson [mailto:erik.svensson_at_six.se]
>>>> Sent: March-07-08 10:56 AM
>>>> To: users_at_grizzly.dev.java.net
>>>> Subject: Re: [Q] Testing ProtocolFilter
>>>>
>>>> On 3/7/08 4:40 PM, "Simon Trudeau" <strudeau_at_bluetreewireless.com>
>>>> wrote:
>>>>
>>>>> I would like to know, how can I test my protocol filter. I would
>>>>> like
>>>> to
>>>>> simulate, using unit tests, a truncated message.
>>>>>
>>>>>
>>>>>
>>>>> I try invoking the connectorHandler's send() method twice (part1
>>>>> and
>>>>> part2 of my message) with a Thread.sleep in between but the
>>>>> protocol
>>>>> parsers doesn't think tokens are missing, it sees the two
>>>>> different
>>>>> operations as two different message. Does this has something with
>>>> TCP/IP
>>>>> where the stack is configured to "know" the packet received has
>>>>> been
>>>>> truncated or do I need to investigate my ProtocolFilter further.
>>>>> How
>>>>> should I test this?
>>>>>
>>>>>
>>>>>
>>>>> client.connect(new
>>>>> InetSocketAddress(InetAddress.getLocalHost(), serverPort));
>>>>>
>>>>> ByteBuffer outputByteBuffer1 = ByteBuffer.wrap(new
>>>>> String("\r\nPA").getBytes());
>>>>>
>>>>> ByteBuffer outputByteBuffer2 = ByteBuffer.wrap(new
>>>>> String("SS\r\n\0").getBytes());
>>>>>
>>>>> client.send(outputByteBuffer1);
>>>>>
>>>>> Thread.sleep(25);
>>>>>
>>>>> client.send(outputByteBuffer2);
>>>>>
>>>>
>>>> I haven't looked at your previously posted code but I've written a
>>>> ProtocolParser. tcp knows nothing about messages nor does it know
>>>> anything
>>>> about the semantics of your message.
>>>> Your ProtocolParser needs to be able to determine if it has
>>>> received a
>>>> complete message and then return true when the 'I need more data'-
>>>> method
>>>> is
>>>> called.
>>>>
>>>> I also found that I had to do a compact() on my incoming bytebuffer
>>>> when
>>>> the
>>>> ReleaseBuffer() is called.
>>>> Otherwise, if you can't read a complete message the bytebuffer will
>>>> just
>>>> fill up until there's no more space and you don't get any more
>>>> messages.
>>>>
>>>>
>>>> -------------------------------------------------------------------
>>>> -
>
>>>> -
>>>> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
>>>> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>>>>
>>>> <MyProtocolParser.java>--------------------------------------------
>>>> -
>
>>>> -
>>
>>>> -
>>>
>>>> ----------------------
>>>> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
>>>> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>>>
>>>
>>> --------------------------------------------------------------------
>>> -
>>> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
>>> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
>> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
>> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
For additional commands, e-mail: users-help_at_grizzly.dev.java.net