On Thu, 21 Sep 2023 06:21:13 GMT, Axel Boldt-Christmas <abold...@openjdk.org>
wrote:
>> `inUseList` will end up with the same value as `inUseListHead`. The reason
>> the old code worked is because `getAddressField` does not type check and
>> `reinterpret_cast<addres>(&ObjectSynchronizer::_in_use_list) ==
>> reinterpret_cast<addres>(&ObjectSynchronizer::_in_use_list._head)`
>>
>> Effectively I changed this to load it correctly (regardless of what
>> `offset_of(MonitorList, _head)` ends up being) and name the variables more
>> appropriately.
>>
>> C++ interpretation of what the java change does:
>> ```C++
>> // Old code
>> // Type type = db.lookupType("ObjectSynchronizer");
>> // inUseList = type.getAddressField("_in_use_list").getValue();
>> address inUseList =
>> *(reinterpret_cast<address*>(&ObjectSynchronizer::_in_use_list));
>>
>> // New code
>> // Type objectSynchronizerType = db.lookupType("ObjectSynchronizer");
>> // Type monitorListType = db.lookupType("MonitorList");
>> // Address monitorListAddr =
>> objectSynchronizerType.getField("_in_use_list").getStaticFieldAddress();
>> // inUseListHead =
>> monitorListType.getAddressField("_head").getAddress(monitorListAddr);
>> address monitorListAddr =
>> reinterpret_cast<address>(&ObjectSynchronizer::_in_use_list);
>> address inUseList = *(reinterpret_cast<address*>(monitorListAddr +
>> offset_of(MonitorList, _head)));
>
> Just to clarify what might cause confusion (at least it is what confused me
> at first when I read this code) is that `getAddress()`/
> `getAddressField(...).getValue()` does not return the address of the field.
> It returns the value of the field (loaded and) interpreted as an address.
I see now. _in_use_list is a MonitorList (not a pointer to one) and the first
field of a MonitorList is the _head field. So that means the address of the
_in_use_list field is also the address of the _head field.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/15782#discussion_r1333390292
