Am 11.07.2016 um 02:21 schrieb Emmanuel Lécharny:
As you can see, I have the same error, but I don't have any selector
being recreated.
Strange... I had this on Windows and Linux on different machines, as
well as on the systems on a user machine.
There is something fishy somwhere, but I can't tell. Your code is quite
big, and from what I saw, nothing jumped to my eyes.
Yes, it's not that tiny :-) But glad to hear that overall it seems to be
okay.
FTR, I would not have implemented the decoder the way you did. Do you
really need a DemuxingProtocolDecoder ? All in all, one single byte is
enough for you to determinate the message type, and that would ddeserve
a switch(type) in a simple cumulative protocol decoder. Teh way teh
Demux decoder works is not exactly beautiful : it loops on all the
registred decoders, returning OK when it can decode one message, so you
have to check all the decoder for *every* message. Not that it shouldn't
be fixed in MINA (that would deserver a JIRA for sure), but at some
point, simplicity is teh key.
I will have a look at it.
I will also create an jira issue for the described problem.
Now, the error is systematic. It seems like the buffer is not correctly
processed, at least in the test. I would suggest you chack that the
buffer is emptied after each read.
Thanks for the hint. I'll check this.
Sadly, it would require me more than a few hours to understand exactly
what your code is doing, so I can't point on any places to start. Maybe
ading a Logger filter would help.
Already did this, without success. Too much data in this test
environment for a quick analysis. Will have to repeat it...
best regards and thank you very much for having a look at it,
Alex
