Fix #1549, spring-plugins bug

[ylywyn]

ContextManager.stopSpan(); may not execute

Please answer these questions before submitting pull request

• Why submit this pull request?
  Bug fix

• Related issues

---

Bug fix

• Bug description.
• How to fix?

[ascrutae]

I think that this is not the root cause about #1549, I suspect that the problem is caused by that the plugin only stores the Request and does not store Response。because of the plugin had check if Request is empty when creating span in beforeMethod and the plugin check the Response is empty when finish the span in afterMethod.

[wu-sheng]

I think that this is not the root cause about #1549, I suspect that the problem is caused by that the plugin only stores the Request and does not store Response。because of the plugin had check if Request is empty when creating span in beforeMethod and the plugin check the Response is empty when finish the span in afterMethod.

Can you provide more detail or another fix for this? If plugin checked, where does the span come from? Any other plugin leak the span?

[wu-sheng]

@ylywyn Look like @ascrutae has a different thought, maybe this is not the right way to fix. But the bug is real.

@ascrutae Please provide the details and things which need @ylywyn to recheck for you.

[coveralls]

Coverage increased (+0.004%) to 24.686% when pulling d1c22b5 on ylywyn:master into dd942d1 on apache:master.

[ascrutae]

@ylywyn Because the Response and Request are not stored in RuntimeContext at the same time, please see InvokeForRequestInterceptor and GetBeanInterceptor.java, so I think that The Response interceptor is not in effect in some version. Could you give the demo to reproduce this issue. or you can check if the Response is stored in the RuntimeContext

[wu-sheng]

@ylywyn Need you to work with @ascrutae . His thoughts are reasonable.

[ylywyn]

yes, this is not the root cause about #1549. context has been removed before AbstractMethodInterceptor::afterMethod.

callstack like this:
6: InvokeForRequestInterceptor::afterMethod

5: AbstractMethodInterceptor::afterMethod context[] is null . ContextManager.stopSpan() can't be call
4: AbstractMethodInterceptor::beforeMethod

3: InvokeForRequestInterceptor::afterMethod context[SW_REQUEST] -> context[]
2: InvokeForRequestInterceptor::beforeMethod context[SW_REQUEST, SW_RESPONSE] -> context[SW_REQUEST, SW_RESPONSE]

1: InvokeForRequestInterceptor::beforeMethod context[SW_REQUEST] -> context[SW_REQUEST, SW_RESPONSE]

[ylywyn]

RUNTIME_CONTEXT management (put and remove) seems to have some problems.

[ascrutae]

I know it, @ylywyn just remove the clear RuntimeContext code in InvokeForRequestInterceptor#afterMethod . Don't worry about leak memory cause by RuntimeContext, it will clear in AbstractMethodInterceptor#afterMethod.

[ylywyn]

Moving ContextManager.stopSpan() to finally{} can solve my problem.
I am not a Java (Spring) developer, so I am not sure that this method AbstractMethodInterceptor::afterMethod will definitely be called under any conditions.

[ascrutae]

The 5.0.0-RC will release in this weekend. I hope that this issue will be fixed in this issue , And could you update this PR? @ylywyn

[wu-sheng]

The 5.0.0-RC will release in this weekend. I hope that this issue will be fixed in this issue , And could you update this PR? @ylywyn

@ascrutae Consider sending a separated PR to fix this? I will keep this PR open in next 24 hours, in case @ylywyn still has interested in the update.

[wu-sheng]

#1568 provided a better solution to fix this. @ylywyn still thanks for helping.