Re: !EventConnect Problem
- From: "Paul G. Tobey [eMVP]" <p space tobey no spam AT no instrument no spam DOT com>
- Date: Mon, 25 Jun 2007 08:39:08 -0700
The fact that it's happening per interaction with the client seems very
strange. If it was happening once a day regardless of the number of
interactions, I could understand that, depending on the response times of,
say, the DHCP server on the network or something of that sort. I think that
you're going to have to contact MS and use one of the Platform Builder
support incidents (or buy more, if you've used yours already). A capture of
the network activity around the time of the failure would probably be useful
to the support engineer. You'll probably want to know whether the socket is
set for keep-alive, etc., also. You might also look for connections to
suspending that the device might be doing, etc. And, of course, make really
sure that it does change frequency with the frequency of interactions with a
client...
Paul T.
"Doug Allender" <doug.allender@xxxxxxxxx> wrote in message
news:C090DC0C-8D29-40CA-B05D-C781CB4F3D9F@xxxxxxxxxxxxxxxx
We can't seem to find any of the above causing us problems, but for what
its
worth I'll give you some more info, which may or may not be usefull.
We have our CE unit being polled by a PC to provide data via tcp/ip. with
a
poll time of 500mS we get a break in communication approx once a day, this
is
approx 2 minuts and accompanied by
0x83a2d400: ndisMResetCompleteStage2: Internal reset
0x83a2d400: *InsertTCB: SetIdleTimerReset
0x83a2d400: InsertTCB: setting fTCBTimerOn to 0
0x83a2d400: *InsertTCB: Restarting TCBTimer
... may get several copies of the TCB Timeouts.
with a poll every 5 seconds the 2 minute brakes were very infrequent
(approx every 8 days), but about once every 2 days we would get a break
of
approx 20minutes when the event connect messasges are generated over and
over.
we also see the event connect messages if we try to use the remote
performance tool... so? I'm not sure where I'm going from here. The socket
server we are using is multi-threaded, however this only seens to have
about
4 threads running to deal with the communications we have going on.
any thoughts
--
Snr. Des. Eng.
Northern Hi-Tec
"Paul G. Tobey [eMVP]" wrote:
The message is printed, best I can read, in the following situations:
1. The socket is not in a listening state. Presumably, this would occur
if
you bound a socket to the right TCP port number, but didn't call listen.
2. The address family of the incoming connection doesn't match that of
the
socket bound to the port (and listening).
3. There's a problem with the connection address, sockaddr. I'd have to
read a bunch more code to see what exactly this is testing for.
4. The incoming connection queue has no room for connections. I suppose
this might happen if you called listen( 0 ), or if you already had enough
pending connections to the listening socket that there wasn't room for
another potential client.
Paul T.
"Doug Allender" <doug.allender@xxxxxxxxx> wrote in message
news:2A9573EE-7565-433A-B7F9-43BEDBBF44A7@xxxxxxxxxxxxxxxx
We have a C++ application which uses a lot of TCP connections, and has
a
bug
we are currently trying to trace. We are using WinCE 4.2 on an arm
platform,
and recieve the following message, repeatedly when our problem occurs
"!EventConnect: connect not accepted IncConnQ = 0". Has anyone else
seen
this
message, or can anyone point me in the correct direction.
--
Snr. Des. Eng.
Northern Hi-Tec
.
- References:
- Re: !EventConnect Problem
- From: Paul G. Tobey [eMVP]
- Re: !EventConnect Problem
- From: Doug Allender
- Re: !EventConnect Problem
- Prev by Date: Re: Detecting any keyboard event independent of focus!! its urgent
- Next by Date: Re: Migrating from CE 4.0 to 4.2
- Previous by thread: Re: !EventConnect Problem
- Next by thread: AlphaBlend API
- Index(es):
Relevant Pages
|
