select/listen (win xp sp1) bug, due to Windows login somehow.

Dave Korn
Mon Feb 28 22:53:00 GMT 2005

----Original Message----
>From: cygwin-owner On Behalf Of Bryan O'Donoghue 
>Sent: 28 February 2005 15:01

> b) You reboot the machine and _don't_ log in. If you simply leave the
> machine at the login prompt, this particular Windows Service, will
> always behave as expected.
> However, if you do start the login process and poll programatically/by
> hand the service to return data, you should notice as I have that
> roughly 1-2 seconds before the full login, to the Windows desktop, that
> the process stops accepting connections, 

  My bet is that you have personal firewall software installed, and that
time is the particular point in userinit at which it (or it's management
interface) gets cranked up, and that is what is somehow causeing the
breakage.  Am I close?

> At this point a "netstat -an" will show BugTester in TIME_WAIT for port
> 1984, however, if left long enough, it will disappear (from the netstat
> -an list, not from the process list).

  That's what TIME_WAIT sockets always do.  Those TIME_WAITS are leftovers
from the invocations while you were waiting for it to disappear.  The real
important question is whether the LISTEN socket is still open.

> /* Functions */
> static void mainThread(void*arg){

  I'd recommend you don't do that.  Send code as an attachment instead, to
prevent it getting line-wrapped.  Nobody wants to spend ages fixing text
strings that have now been broken in the middle and are causing errors, or
trailing ends of c++ style // comments that have been wrapped onto new
lines; if you really want someone to try your testcase, they need to be able
to have it build OOTB.

Can't think of a witty .sigline today....

Unsubscribe info:
Problem reports:

More information about the Cygwin mailing list