Expose config to set up the re-transform support explicitly

[wu-sheng]

• If this pull request closes/resolves/fixes an existing issue, replace the issue number. Closes [Bug] SkyWalking Java Agent performance degradation after upgrading to 9.0 skywalking#11460
• Update the CHANGES log.

I am waiting for @lujiajing1126 final report about this case in their environment.

[lujiajing1126]

For a real workload (Spring Boot application) with ~30 classes to be enhanced.

Local env

• OS: macOS sonoma 14.0
• JDK: 1.8.0_372
• Hardware spec: Apple M2 with 16GB memory

DescriptionStrategy	premain exec time*
POOL_FIRST	~6 secs
HYBRID	~3 secs

• specifically agentBuilder.installOn(instrumentation)

[wu-sheng]

@lujiajing1126 I updated the doc, please take a look.

[lujiajing1126]

LGTM

[kylixs]

This option might be better changed to enable or disable custom byte-buddy patch, for compatibility reasons.

[wu-sheng]

Compatible with which part?
I am waiting for your PR feedback. If it is good enough, we don't need this.

[kylixs]

Determine whether to use a custom AgentBuilder with patch or a normal one:

    private static AgentBuilder newAgentBuilder(SWDescriptionStrategy descriptionStrategy) {
        final ByteBuddy byteBuddy = new ByteBuddy()
                .with(TypeValidation.of(Config.Agent.IS_OPEN_DEBUGGING_CLASS))
                .with(new SWAuxiliaryTypeNamingStrategy(NAME_TRAIT))
                .with(new SWImplementationContextFactory(NAME_TRAIT));

        return new SWAgentBuilderDefault(byteBuddy, new SWNativeMethodStrategy(NAME_TRAIT))
                .with(descriptionStrategy);
    }

[wu-sheng]

I don't think we need to expose. This should not be aware ny users.
Anyone doesn't read codes, don't know what this is about.

[kylixs]

I mean the option is SW_ENABLE_RETRANSFORM_SUPPORT, disable byte-buddy patch if it is false. Users can understand this option.

[wu-sheng]

I know your point. But I doubt they will know what path or our customization mean.

[wu-sheng]

Like I documented, it should be a straight forward. If there are really some issues of your patch, that is a bug, nothing more

[kylixs]

I agree. I'm just wondering if there's a fallback mechanism that can be enabled if something goes wrong.

[wu-sheng]

Don't worry. 8.x is safe enough to fall back. 😊

[wu-sheng]

Close for now, as we seems to have a solution now.