On Wed, 5 Feb 2025 05:00:14 GMT, David Holmes <dhol...@openjdk.org> wrote:
>> Cleanup SA JavaThread support. Details in first comment:
>>
>> Testing:
>> - Tier1
>> - Tier2 svc
>> - Tier3
>> - Tier5 svc
>> - Local testing of debuggee using graal java compiler threads. Verified that
>> the compiler threads shows up in jstack output if the property is set.
>
> src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/runtime/JavaThread.java
> line 36:
>
>> 34: import sun.jvm.hotspot.utilities.Observer;
>> 35:
>> 36: /** This class is no longer abstract. Platform dependent functionality
>> is now implmented
>
> Suggestion:
>
> /** Platform dependent functionality is now implemented
>
> "no longer" very quickly loses context if you don't know when it was
> abstract, or why.
Ok. I can probably just remove the comment. I guess I worded it the way I did
to help the reviewer understand (as opposed to a comment just disappearing).
> src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/runtime/Threads.java line
> 161:
>
>> 159: virtualConstructor.addMapping("JvmtiAgentThread",
>> JavaThread.class);
>> 160: virtualConstructor.addMapping("NotificationThread",
>> JavaThread.class);
>> 161: virtualConstructor.addMapping("AttachListenerThread",
>> JavaThread.class);
>
> It still seems a shame that we have to enumerate all the thread sybtypes.
In order the create a mirror instance, SA needs to map the hotspot type to the
SA type that implements the mirror. We have the same issue with CodeBlob
subtypes.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/23456#discussion_r1942292903
PR Review Comment: https://git.openjdk.org/jdk/pull/23456#discussion_r1942294672
