On 13-4-2015 11:28, Alex Peshkoff wrote:
>
>
> On 11.04.2015 11:57, Mark Rotteveel wrote:
>> On 6-4-2015 12:40, Mark Rotteveel wrote:
>>> If I add a small pause (1 to 5 milliseconds) between steps 2 and 3, the
>>> problem doesn't occur. This seems to indicate it is a
>>> threading/concurrency issue with invoking a callback in parallel with
>>> the invocation of isc_que_events for a different event.
>>>
>>> At the Java side I added a lot of synchronization already (more than I
>>> believe should be necessary).
>> I have as of yet been unable to reproduce it when I strip away more of
>> my code, so for now it looks like the problem is on my side, or there is
>> a very peculiar issue related to timing/interleaving of calls to methods
>> and callbacks and stripping away code changes this timing.
>>
>> When I look at the code in why.cpp for isc_que_events and QueCallback I
>> get a bit suspicious:
>>
>> As far as I can tell between these statements (specifically the second
>> and the last):
>>
>> events = attachment->queEvents(&statusWrapper, callback, length,
>> eventsData);
>> callback->setEvents(events); // should be called in case of
>> NULL events too
>> if (status.getState() & Firebird::IStatus::STATE_ERRORS)
>> {
>> return status[1];
>> }
>>
>> *id = FB_API_HANDLE_TO_ULONG(events->getHandle());
>>
>> the callback could already be triggered and have deleted events before
>> the id is obtained. Shouldn't it call addRef on events before setting it
>> to callback, and always call release at the end of the method? Or
>> shouldn't the mutex that is checked in the callback be locked until
>> isc_que_events has ended instead - like now - until the events object is
>> set?
>
> Mark, can you try this patch?
This unfortunately gives the same error. I tried something similar
yesterday (I assigned the handle to a temp variable before
callback->setEvents(events) and assigned it to id at the end), but that
lead to a different error (an invalid memory access), probably because I
used a wrong type or something.
For the time being I will assume that it is my code that is doing
something wrong, I just have to find out what exactly that is.
Mark
--
Mark Rotteveel
------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/lists/listinfo/firebird-devel
