hi, I have been playing with 4.42 and find that when its starts up as part of the startup routing it brings up a window that says connection failed. This looks like a standard windows child window and is similar to the one that v4.13 used to bring up to tell you it could not clear the window sockets. Once this has appeared Boinc displays the disconnected message at the bottom right and sulks so that it does nothing. If you quit and restart it runs happily and connects. Anyone got any ideas how to sort this as I have a number of machines that run Boinc as a service and this would not make it easy!

Machine is an amd 2800 running XP SP2 broadband connection. I had no problems on 4.19 or 4.25 although I have had similar on 4.25 on other machines. broadband is via a router, 5 other machines going through it no problems. Also have zone alarm pro on machine but boinc has access rights.

Julian Rouse

hi, I have been playing with 4.42 and find that when its starts up as part of the startup routing it brings up a window that says connection failed. This looks like a standard windows child window and is similar to the one that v4.13 used to bring up to tell you it could not clear the window sockets. Once this has appeared Boinc displays the disconnected message at the bottom right and sulks so that it does nothing. If you quit and restart it runs happily and connects. Anyone got any ideas how to sort this as I have a number of machines that run Boinc as a service and this would not make it easy!

Machine is an amd 2800 running XP SP2 broadband connection. I had no problems on 4.19 or 4.25 although I have had similar on 4.25 on other machines. broadband is via a router, 5 other machines going through it no problems. Also have zone alarm pro on machine but boinc has access rights.

Julian Rouse

I have a similar problem Julian,
As far as I can tell, BOINC gets started before ZoneAlarm is fully loaded, and while ZA is loading it blocks everything until it has loaded the rules/exceptions.
I've not used 'Run as a Service' (yet) so I'm not too sure about this, but, the error message you're seeing is (AFAIK) from BOINCMgr, which is a normal application rather than a service.
I'm still experimenting with ways of delaying the start-up of BOINCMgr until the rest of the system (incl. ZA) is properly ready.

I think it may be only a matter of seconds (on my system) so perhaps the Devs could extend the time-out on connect (and retry a few times)?

---

I think it may be only a matter of seconds (on my system) so perhaps the Devs could extend the time-out on connect (and retry a few times)?

It is a timing problem between the BOINC Manager (GUI) and the BOINC Daemon. The two operate independently of each other. So, if the both start at one time, the GUI can get up and try to connect to the Daemon and "miss" as the Daemon is still initializing. Once they both are up and runing, all is well.

Maybe we should just slow down the computers so this does not happen. :)

---

I'm still experimenting with ways of delaying the start-up of BOINCMgr until the rest of the system (incl. ZA) is properly ready.

Last week (I explained that in the questions/problems thread about this issue), I used a batch file to delay the startup of Boincmanager for 1 minute after Windows startup. All harddisk activity had stopped already after 45 seconds... then 15 seconds later Boincmanager started, WITH the error message.
So this delay did not help for the PC I tried it on.

:-)

I think it may be only a matter of seconds (on my system) so perhaps the Devs could extend the time-out on connect (and retry a few times)?

It is a timing problem between the BOINC Manager (GUI) and the BOINC Daemon. The two operate independently of each other. So, if the both start at one time, the GUI can get up and try to connect to the Daemon and "miss" as the Daemon is still initializing. Once they both are up and runing, all is well.

Maybe we should just slow down the computers so this does not happen. :)

Well, I don't do C, but the CVS checkin log messages seem to indicte that Rom is adding some 5 second delays in the RPC code:

Quote:

- Bug Fix: After starting the CC wait for 5 seconds to give the CC time
to establish the listening socket for the next rpc request.

So, this issue may go away?

---