I posted this some time ago and received n replies. I'll post it again here, in the hopes that this is the right forum.

Post subject: JForexClient Memory LeakPosted: Wed 10 Nov, 2010, 03:04

Joined: Mon 01 Nov, 2010, 18:08
Posts: 2
I adjusted the start and end dates to back test a strategy using the JForexClient within Netbeans. After the system runs for several minutes, it began to slow down. To find the issue, I profiled the code and noticed the "surviving generations" profile increasing with every tick, which is the symptom of a memory leak.

I went through my code and continued the process until at one point I removed all of my code leaving the onTick method blank and ran the profile again only to find that the empty method is still leaking memory. This means the JForexClient must contain the memory leak, and it is only reveled during longer back testing.

As I can not go live without thorough back testing, please advise. I have screen shots, but don't see how to up load them in this interface.

Kind regards,
David Heath

I've not done the testing you've done to look for a leak, but I've suspected it. If I start doing some testing with historical data, it gets slower and slower with each identical test until I just restart the application and it goes back to being fast.

Yes, stoping the system and restarting it will release the memory, but doesn't allow for back testing over a long period of time. My trades are open when the system runs out of memory, so a restart literally breaks the trades making the test invalid. I'm concerned that Dukascopy hasn't at least replied to this . . . after two months or so.

I've tried using a 3rd party UI with their data feed, and of course it is problematic.

The regular tester and the optimizer actually function differently. I can REALLY REALLY break the regular tester and then with the same stuff the optimizer works fine!

I agree on the late response, but it's not really a serious issue in that if this was live, you wouldn't lose money because of it. But I've found things where the stop placement doesn't work properly and they don't reply to that... that can make someone lose real money!

As far as you run the test in the Netbeans, please provide us with the tested code which has caused the problem.

The leak occurs in JForex and the JForexClientLibrary. Are you asking for my strategy source code to see the leak or asking how I used Netbeans to find it?

To be honest, it's taken so long to get a response from anyone I've moved to another broker. I was curious, so I checked back in.

fixed in next version (2.13+)