<Swing Dev> [11] Review Request: 8198342 Test FileSystemViewListenerLeak.java is unstable

Jayathirth D V jayathirth.d.v at oracle.com
Wed Apr 25 06:27:27 UTC 2018


Hi Sergey,

I went through the changes done in JDK-8175968 and I can understand why you would have added creation of multiple CustomFileSystemView like stress test scenario.
But as you have mentioned in highly loaded systems there can sometimes be problem with CleanerFactory clearing all listeners on time under GC.

My observations:
1) Instead of decreasing the number of CustomFileSystemView to 1, can we add some wait after we generate OOM in test case and check for number of listeners later.
2) Also adding jtreg @summary tag would help to understand the basic functionality  of test case.

Thanks,
Jay

-----Original Message-----
From: Sergey Bylokhov 
Sent: Wednesday, April 25, 2018 3:43 AM
To: swing-dev at openjdk.java.net
Subject: Re: <Swing Dev> [11] Review Request: 8198342 Test FileSystemViewListenerLeak.java is unstable

Any volunteers to review? =)

On 12/03/2018 22:27, Sergey Bylokhov wrote:
> Hello.
> Please review the update of the test for jdk11.
> 
> Bug: https://bugs.openjdk.java.net/browse/JDK-8198342
> Webrev can be found at: 
> http://cr.openjdk.java.net/~serb/8198342/webrev.00
> 
> This test creates a number of FileSystemView which internally register 
> a listeners in UIManager, and the test assumes that these listeners 
> will be removed when FileSystemView will be collected by GC.(The 
> listener itself is removed by the Cleaner which uses a separate thread).
> 
> The problem is that in the high-loaded systems when the test is 
> executed in parallel with many other heavyweight tests the cleaner's 
> thread is not always able to remove the listeners in time and test fails.
> 
> In the fix the umber of FileSystemView in the test is decreased to 1, 
> I have checked that the test still fail before related 
> bug(JDK-8175968) was fixed.
> 
> 


--
Best regards, Sergey.



More information about the swing-dev mailing list