RFR(S): 8240204: Optimize package handling for archived classes

Yumin Qi yumin.qi at oracle.com
Fri Apr 17 17:54:34 UTC 2020

Hi, Harold

   Thanks for the review.

On 4/17/20 8:30 AM, Harold Seigel wrote:
> Hi Yumin,
> The changes look good.  Just one question.
> InstanceKlass::set_classpath_index() returns a value. That value is 
> only checked in one of the places from where it's called.  Is it worth 
> checking the return value in the other places and asserting if it is 
> false?
Let me investigate this issue first. If so, we should put the assert 
inside set_classpath_index and change return type to 'void'.

> Thanks, Harold
> On 4/17/2020 1:56 AM, Yumin Qi wrote:
>> Hi,
>>   Please review:
>>   bug: 8240204: https://bugs.openjdk.java.net/browse/JDK-8240204
>>   webrev: http://cr.openjdk.java.net/~minqi/8240204/webrev-01/
>>   Summary: Move ClassLoader::add_package to InstanceKlass since what 
>> it does is calling pkg->set_classpath_index. Also a minor 
>> optimization for InstanceKlass::set_package, for shared class, avoid 
>> call check_prohibited_package since CDS does not archive prohibited 
>> classes.
>>   Tests: hs-tier1-4
>>   Performance data: javac and zprint showed a little improvement.
>>   Thanks
>>   Yumin

More information about the hotspot-runtime-dev mailing list