RFR: 8245264: Test runtime/cds/appcds/SignedJar.java fails

Calvin Cheung calvin.cheung at oracle.com
Wed May 27 21:32:08 UTC 2020


Hi Yumin,

The updated version looks good.

thanks,

Calvin

On 5/27/20 1:31 PM, Yumin Qi wrote:
> Hi,
>   After offline discussion (with Coleen, Ioi, Robbin Ehn  and Calvin), 
> the new version uses SymbolTable::_local_table to collect complete set 
> of symbols at safepoint (dumping shared archive) which can avoid the 
> invalid _shared_table with concurrent work on it.
>
>   bug: https://bugs.openjdk.java.net/browse/JDK-8245264
>   webrev: http://cr.openjdk.java.net/~minqi/2020/8245264/webrev-01/
>
>   Tests: mach5 tier1, tier2.
>
> Thanks
> Yumin
>
> On 5/22/20 9:11 AM, Yumin Qi wrote:
>> Hi, Please review:
>>
>>   bug: 8245264: Test runtime/cds/appcds/SignedJar.java fails
>>   Webrev: http://cr.openjdk.java.net/~minqi/2020/8245264/webrev-00/
>>
>> Summary: When dump shared archive in safepoint, there may exist 
>> current work for StringTable or SymbolTable, which led to the 
>> table(s) not walkable. This led symbols not sorted and binary search 
>> for symbol (method in this case) failed in runtime. This problem 
>> might exist long time ago but not triggered until the change of jar 
>> signing code (bug 8242151, 
>> https://bugs.openjdk.java.net/browse/JDK-8242151). 
>> <https://bugs.openjdk.java.net/browse/JDK-8242151>
>>    If table(s) work still in process, abort safepoint (exit from vm 
>> operation) to wait for them to finish then do dump work.
>>
>>    (Thanks Ioi, Calvin, Coleen and David H ...  for help to 
>> identifying problem and suggest solution!)
>>
>>    Tests: tier1, tier2.
>>
>>    Thanks
>>    Yumin
>>


More information about the hotspot-runtime-dev mailing list