Re: [certi-dev] Deadlock with jcerti

[Jan-Patrick Osterloh]

Hi!

--- Quoted from Eric Noulard (Date: 08.10.2012 21:27): ---
> 2012/10/8 Jan-Patrick Osterloh <address@hidden>:
>> Hi all,
>>
>> I'm currently using jcerti in my environment, and I have troubles with
>> some "deadlock". During initialisation, my program hangs in
>> rtia.publishObjectClass(classHandle, attributes);
>> The method does not return and hangs. Any idea what that could be, or I
>> can debug this?
> I don't really what could be the cause of that.
>
> Does this *always* happend or does it sometimes work?
Currently it is reproducible, so it happens always, always at the same
point. Also I had one afternoon (Friday) where it worked again, but when
I tried yesterday, same problem occoured.

> Is your application using rtia object in a concurrent fashion?
> i.e. does several thread use the same rtia object ?
Well, my Federate is a Thread (or to be correct, it implements Runnable
and is started in a Thread).

> If this is the case may be there is a race, could you try to serialize
> the access to the object using some Java lock or monitor?
Threading was also my first idea, therefore I tried to make it
thread-safe by using a "java.util.concurrent.locks.ReentrantLock" as
Semaphore for each rtia call, but it makes no difference between using
the lock or not (but maybe I must use it more often). I even tried to
use it without beeing a thread, but it makes no difference.


How much threading is used in jcerti itself, i.e. how are the callbacks
handled? Are they called in an extra thread? I used the UAV example as a
start for implementing my Federate.

>
> Concerning the debugging you should be able to debug into the call
> but if it is racy may be debugging it would make the problem disappear...
>
> You may try to bump the logger trace level, that way we may
> have more clue were the deadlock occurs.
>
> I don't remember right now how to do it.
> May be something like defining the jcerti.logLevel to appropriate value is
> the way to go. If you don't find your way in the Java code I'll find out
> tomorrow.
>
I've seen somewhere that the debug level is read from a property file,
I'll check that.

JPO



-- 
Dipl. Inform. Jan-Patrick Osterloh
FuE Bereich Verkehr | R&D Division Transportation
Human Centered Design Group

OFFIS
FuE Bereich Verkehr | R&D Division Transport
Escherweg 2 - 26121 Oldenburg - Germany
Phone/Fax: +49 441 97 22-524/502
E-Mail: address@hidden
URL: http://www.offis.de

signature.asc
Description: OpenPGP digital signature