Infinite loop in RpcClient.callBlockingMethod about protobuf-rpc-pro

[GoogleCodeExporter]

Thanks for raising this. I've been on holiday for the last month, i hope to 
look into this in the next week. 

Original comment by [email protected] on 4 Aug 2013 at 10:47

[GoogleCodeExporter]

I think this is a synchronization problem which can be triggered with a race 
condition. I have made a fix for this with 3.0.5.

Original comment by [email protected] on 5 Aug 2013 at 4:53

• Changed state: Fixed

[GoogleCodeExporter]

Thanks! I will try it out and get back to you if it occurs again.

Original comment by [email protected] on 7 Aug 2013 at 8:35

[GoogleCodeExporter]

I just noticed a bug in the latest version. You are multiplying the ms timeout 
with 1000 in order to get the nanosecond timeout. However, you need to multiply 
by 1 000 000 for the conversion to be correct. Otherwise it will timeout too 
quickly.

Original comment by [email protected] on 8 Aug 2013 at 12:58

[GoogleCodeExporter]

Doh! Thanks for that - i'll fix immediately.

Original comment by [email protected] on 8 Aug 2013 at 5:42

[GoogleCodeExporter]

Thanks for the quick response! We just tried it and unfortunately it isn't 
solved yet. ClientRpcController.getTimeoutMs() returns an int and it therefore 
overflows when multiplied by 1 000 000, resulting in a negative timeout. I'd 
suggest changing it to a long. 

Original comment by [email protected] on 9 Aug 2013 at 9:39

[GoogleCodeExporter]

Double doh! Thanks again i was taking too many shortcuts. The fix is now in 
3.0.7.

regards,P.

Original comment by [email protected] on 9 Aug 2013 at 7:59

[GoogleCodeExporter]

I just found a bug related to this fix. The following code in RpcClient waits 
too long: 
if ( !callback.isDone() ) {
    callback.wait(timeToDeadline);
}
timeToDeadline contains the timeout in nanos but wait(long timeout) expects to 
receive the time in ms. 

Original comment by [email protected] on 22 Jan 2014 at 2:07

[GoogleCodeExporter]

you're right. Thanks!

Original comment by [email protected] on 22 Jan 2014 at 6:38

• Changed state: Started

[GoogleCodeExporter]

fixed with 3.2.2

Original comment by [email protected] on 22 Jan 2014 at 7:51

• Changed state: Fixed

[GoogleCodeExporter]

You are as quick as always! :)

Original comment by [email protected] on 22 Jan 2014 at 7:56

[GoogleCodeExporter]

Some of our clients have started throwing this exception: "Exception: Issue25: 
not fixed - callback after timeout handling not finished. Please re-open issue.
Value: com.googlecode.protobuf.pro.duplex.RpcClient Issue25: not fixed - 
callback after timeout handling not finished. Please re-open issue.". 
Unfortunately I don't know exactly how to reproduce. The only thing I know is 
that it started occurring when we started making a lot more concurrent calls.

Original comment by [email protected] on 14 Aug 2014 at 8:56

[GoogleCodeExporter]

Hi Carl, i think this "error log" is producable because of a race condition 
between the correct response handling and the timeout handling in two separate 
threads( and not fully synchronizing access to the BlockingRpcCallback ) . 

I'm pretty sure there is not really a problem and the logging lines:

if ( !callback.isDone() ) {
    log.error("Issue25: not fixed ..
}
can be safely removed. Or enclosed in a synchronize(callback){..} block

do you want a new revision to fix this - if yes, then i propose to just delete 
the log.

Original comment by [email protected] on 15 Aug 2014 at 6:52