Reactive routes - fix race condition in generated handlers

[mkouba]

• follows up on Reactive routes - improve @Route method invocation performance #8536

[mkouba]

FYI @gsmet @michalszynkiewicz

[gsmet]

@michalszynkiewicz I let you check that one as you discovered the issue?

[michalszynkiewicz]

I know nothing about the code, but it looks that it moves code that was in initialize method to the constructor.
I added this commit to the PR I have consistently failing on the bug that is being fixed, let's see if it helps.

If I read correctly, it gets the arc container via static method in the constructor. Will the container always be ready at this point @mkouba ?

[mkouba]

I know nothing about the code, but it looks that it moves code that was in initialize method to the constructor.

Yes, it does.

If I read correctly, it gets the arc container via static method in the constructor. Will the container always be ready at this point @mkouba ?

Yes, it should. That's why I added the BeanContainerBuildItem parameter to the addAdditionalRoutes() method.

[mkouba]

I added this commit to the PR I have consistently failing on the bug that is being fixed, let's see if it helps.

+1

[stuartwdouglas]

I am not actually sure what race this is actually fixing, and there does not seem to be any info anywhere. The linked PR mentions that there is a bug but gives no info.

[mkouba]

@stuartwdouglas I'm sorry. So while thinking of it it was rather a JMM visibility problem where a non-volatile field was set in one thread but probably not visible on the other thread.

[michalszynkiewicz]

fyi, #8576 with this commit passes the checks