BufferOverflowException when using callbacks

[iwagner]

I added the ["amd"] tag on some of my methods so I can use the auto-generated callback objects. This seems to work for a while, then I start getting java.nio.BufferOverflowException when I call ice_response() on the callback object. Any ideas? Am I pounding the server too hard and have too many active callbacks? At this point I'm lost. Thanks.

[bernard]

Hi Isaac,

Could you post the full stack trace, and platform/version information?

Thanks,
Bernard

[iwagner]

Hi Isaac,

Could you post the full stack trace, and platform/version information?

Thanks,
Bernard

I'm using Ice 3.4.2 for Java on Linux. Here's the stack trace:

java.nio.BufferOverflowException
        at java.nio.Buffer.nextPutIndex(Buffer.java:495)
        at java.nio.DirectByteBuffer.put(DirectByteBuffer.java:250)
        at IceInternal.BasicStream.writeBool(Unknown Source)
        at testproject.parser.types.ParserData.__write(ParserData.java:142)
        at testproject.parser.interfaces._AMD_ParserProcessor_unpackData.ice_response(_AMD_ParserProcessor_unpackData.java:39)
        at com.testproject.DataListener.notifySender(DataListener.java:83)
        at com.testproject.DataListener.onMessage(DataListener.java:47)
        at com.tibco.tibjms.TibjmsxSessionImp._submit(TibjmsxSessionImp.java:3821)
        at com.tibco.tibjms.TibjmsxSessionImp._dispatchAsyncMessage(TibjmsxSessionImp.java:2088)
        at com.tibco.tibjms.TibjmsxSessionImp$Dispatcher.run(TibjmsxSessionImp.java:3347)

[bernard]

Hi Isaac,

This does not look like a memory exhaustion issue.

Is there any possibility that your code is calling ice_response on the same callback object from two different threads? (Such a race condition is a possible explanation).

AMD callbacks are not thread safe, and you should not use an AMD callback after calling ice_response on it.

Best regards,
Bernard

[iwagner]

Is there any possibility that your code is calling ice_response on the same callback object from two different threads? (Such a race condition is a possible explanation).

I've been through the code several times and can't see any threading issues. But, I will look again and get some of my colleagues involved to see if they can spot a threading issue. Thanks.