On Tue, 18 Aug 2026 07:22:29 GMT, Markus Grönlund <[email protected]> wrote:

> > The JBS description of the problem and solution no longer matches what is 
> > actually being done. I really don't understand the changes being made.
> 
> What do you mean "what is actually being done"? You asked for ClassDefine 
> events to be issued after a class is actually selected, later, instead of 
> trying to protect for errors.
> 
> That is what is being done.

Okay. The JBS description focuses on the problem with no description of the 
solution. The original problem statement was:

> [JDK-8365400](https://bugs.openjdk.org/browse/JDK-8365400) changed the way 
> jdk.ClassDefine events are emitted.
> Instead of being emitted as the last step of 
> SystemDictionary::define_instance_klass() {
> jdk.ClassDefine events are instead emitted earlier, as part of 
> KlassFactory::create_from_stream(), as part of the JFR hook located there,
> The reason for this change was to gain full access to the ClassFileParser to 
> determine a "source" attribute for the defined klass.

I expected to see a simple restoration of the event being " emitted as the last 
step of systemDictionary::define_instance_klass()" because you now have a way 
to get the source info you needed. But the actual changes seem more elaborate 
than that. But as I said I don't know this JFR code or how JFR events work in 
general. My main concern/interest was just checking for any changes to non-JFR 
parts of the code - which I have flagged.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/32218#issuecomment-5325302793

Reply via email to