RFR: 8356126: Duplication handling and optimization of CaptureCallState [v4]

[Chen Liang]

On Tue, 6 May 2025 06:36:57 GMT, Jaikiran Pai <jpai at openjdk.org> wrote:

>> Enum.valueOf -> Class.enumConstantDirectory -> Class.getEnumConstantsShared -> Method.invoke -> MethodHandleAccessorFactory.makeSpecializedTarget(isStatic = true) -> MethodHandles.dropArguments -> LambdaForm.editor -> bytecode generation and loading because this currently cannot be pregenerated by CDS archive.
>> 
>> If this class is broken, this would probably already surface at build time because this is used by jlink; otherwise it would have surfaced in Process tests. I don't think ensuring EIIE vs IAE is worth a test here.
>
> Thank you Chen for updating the issue type of bug.
> 
>> If this class is broken, this would probably already surface at build time because this is used by jlink; otherwise it would have surfaced in Process tests. I don't think ensuring EIIE vs IAE is worth a test here.
> 
> The current bug fix I think should just address the incorrect result from `captureCallState()`.
> 
> If the change to OperatingSystem class has practical improvements to the startup performance, then I think it's worth proposing. I suggest we do it in a separate and independent PR and the discussion and review for that change would have to take into account the existing comment in that class and see if it is no longer necessary.

Yep, rolled back OS changes.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/25025#discussion_r2075768065