RFR (XS) 8210559: ClassLoaderData Symbols can leak

coleen.phillimore at oracle.com coleen.phillimore at oracle.com
Thu Sep 13 11:29:59 UTC 2018

On 9/12/18 8:03 PM, Ioi Lam wrote:
> Hi Coleen,
> The changes look good.
> I am wondering if we should add a WhiteBox function to test whether a 
> given Symbol is alive. That way, you can write a test case to ensure 
> that after the CLD is destroyed, its name will be removed from the 
> SymbolTable.

We can't really test dead symbols because looking up a symbol makes it 
alive, and if we save a reference to a dead symbol, it can be freed at 
any time.

I could make up a funny name for a class loader, and have a white box 
function watch its refcount, compare before with after unloading.
> I know tests cases are a hassle, but I think in the long run that 
> would be beneficial.

A white box function to check refcounts of Symbols might be useful for 
other tests, so I can try to do it.


> Thanks
> - Ioi
> On 9/12/18 4:32 PM, coleen.phillimore at oracle.com wrote:
>> Summary: unrefcount the symbol names when the CLD is destroyed
>> I was going to clean up the sometimes null value of _name and 
>> _name_and_id, and have the boot classloader data have a name, but 
>> there's a bootstrapping problem.  NULL CLD is created before the 
>> SymbolTable.  So this wasn't done.
>> Tested with mach5 hs-tier1-3, and runThese jck which does a lot of 
>> class unloading.
>> open webrev at http://cr.openjdk.java.net/~coleenp/8210559.01/webrev
>> bug link https://bugs.openjdk.java.net/browse/JDK-8210559
>> Thanks,
>> Coleen

More information about the hotspot-runtime-dev mailing list