RFR: JEP 359-Records: hotspot runtime and serviceability code

[Ioi Lam]

On 10/21/19 11:40 AM, Lois Foltan wrote:
> On 10/18/2019 2:44 PM, Vicente Romero wrote:
>> Hi,
>>
>> Please review the hotspot runtime and serviceability code for JEP 359 
>> (Records).
>>
>> Thanks in advance for the feedback,
>> Vicente
>>
>> PS, Thanks to Harold for the development
>>
>>
>> [1] 
>> http://cr.openjdk.java.net/~vromero/records.review/hotspot_runtime/webrev.00/
>
> Harold, Vicente,
>
> src/hotspot/share/oops/recordComponent.hpp
> - line #95 - determination of TBD comment needed before committing.

This code is correct and the TBD can be removed.

     void collect_statistics(KlassSizeStats *sz) const {
       // TBD is this right?
       if (_annotations != NULL) {
         sz->_annotations_bytes += sz->count(_annotations);
         sz->_ro_bytes += sz->count(_annotations);
       }
       if (_type_annotations != NULL) {
         sz->_annotations_bytes += sz->count(_type_annotations);
         sz->_ro_bytes += sz->count(_type_annotations);
       }
     }

You can test this by running a JVM with something like:

jcmd <pid> Wait GC.class_stats 
IK_inner_classes,IK_nest_members,IK_record_components


Note that this will print out "WARNING: Ran out of C-heap; undercounted 
368 total instances in data below". This is JDK-8232758, which you can 
safely ignore for now.

BTW, all the collect_statistics() functions are pretty complicated and 
it's unclear whether GC.class_stats is still useful (I implemented it 
but I don't use it anymore). I have filed JDK-8232759.

Thanks
- Ioi