Christian Knoke <[EMAIL PROTECTED]> writes:
> There have been several reports in the past about
> hanging wwwoffled processes (including mine) and I
> had assumed that these probs are fixed with 2.7f
>
> Unfortunately not. Here is an example with an abrupt
> abort of the pppd. All browsers have been stopped
> 5 minutes before the output, too:
There has been no work to fix this problem yet although I believe that
I do know what it is. I think that the problem is that the WWWOFFLE
process is waiting for the data to be written to the remote server.
The write function will not return until the data is written or the
socket is finally declared as dead.
The processes will eventually time out, but it will take until the TCP
timeout period which is a long time. I need to add the timeout to the
write functions in suitable places and then test it. It is the
testing that is the most difficult part because I need to simulate or
cause a broken connection.
As you can see from the log file the socket (read) timeout and connect
timeout seem to be working OK. A number of processes have connection
timeouts at around 14:02:00 and another group have read timeouts at
around 14:03:30. These would be the default timeouts of 30 and 120
seconds after the link went down at around 14:01:30 I would imagine.
--
Andrew.
----------------------------------------------------------------------
Andrew M. Bishop [EMAIL PROTECTED]
http://www.gedanken.demon.co.uk/
WWWOFFLE users page:
http://www.gedanken.demon.co.uk/wwwoffle/version-2.7/user.html
