8216541: CompiledICHolders of VM locked unloaded nmethods are released too late

[Tobias Hartmann]

Hi Erik,

very nice analysis, thanks a lot for investigating!

On 28.01.19 14:56, Erik Österlund wrote:
> http://cr.openjdk.java.net/~eosterlund/8216541/webrev.00/

Why did you remove the call to thread->set_scanned_compiled_method(NULL) in sweeper.cpp?

> The proposed change has survived 200 rounds of kitchensink, hs-tier1-3 and hs-precheckin-comp.

In the meanwhile, could you please run some more 100x iterations of kitchensink?

Thanks,
Tobias