Caching the ThreadContext breaks in multi-deployment environments

[kabir]

My use-case is that I first deploy application A in WildFly, then undeploy application A, and then deploy application B.

Before this fix application B will still use the ThreadContext and underlying ContextManager from application A. This breaks the optimisation @stuartwdouglas added to cache the TransactionManager in the JtaThreadContextProvider as it tries to read data from the bean manager from deployment A which has been cleared. Also, I believe we cannot cache in case deployments supply their own ThreadContext implementations.

Creating the ThreadContext on demand as I am doing here fixes the problem.

[codecov]

Codecov Report

Merging #335 into master will increase coverage by 0.17%.
The diff coverage is 100.00%.

@@             Coverage Diff              @@
##             master     #335      +/-   ##
============================================
+ Coverage     89.62%   89.80%   +0.17%     
- Complexity     2636     2642       +6     
============================================
  Files           354      354              
  Lines         10150    10161      +11     
  Branches       1250     1252       +2     
============================================
+ Hits           9097     9125      +28     
+ Misses          545      537       -8     
+ Partials        508      499       -9     

Impacted Files	Coverage Δ	Complexity Δ
...ny/context/ContextPropagationMultiInterceptor.java	100.00% <100.00%> (ø)	4.00 <3.00> (ø)
...tiny/context/ContextPropagationUniInterceptor.java	86.66% <100.00%> (ø)	4.00 <3.00> (ø)
.../io/smallrye/mutiny/helpers/queues/DrainUtils.java	82.22% <0.00%> (-4.45%)	13.00% <0.00%> (-1.00%)
...io/smallrye/mutiny/converters/uni/ToPublisher.java	87.09% <0.00%> (-3.23%)	3.00% <0.00%> (ø%)
...subscription/SwitchableSubscriptionSubscriber.java	95.08% <0.00%> (-2.46%)	45.00% <0.00%> (-1.00%)
...a/io/smallrye/mutiny/converters/multi/ToMaybe.java	100.00% <0.00%> (ø)	4.00% <0.00%> (ø%)
...java/io/smallrye/mutiny/helpers/Subscriptions.java	79.55% <0.00%> (ø)	48.00% <0.00%> (ø%)
...rye/mutiny/operators/multi/MultiRepeatUntilOp.java	100.00% <0.00%> (ø)	5.00% <0.00%> (ø%)
...e/mutiny/streams/stages/FindFirstStageFactory.java	100.00% <0.00%> (ø)	5.00% <0.00%> (ø%)
...mutiny/operators/multi/builders/ResourceMulti.java	87.00% <0.00%> (+0.13%)	5.00% <0.00%> (ø%)
... and 14 more

[cescoffier]

I discussed with @kabir offline, and I'm worried about the cost of such a change. We are discussing alternatives:

• specific interceptors for EAP
• adding a way to clear the static on termination

[cescoffier]

Closing this PR for now, we may reopen it if the alternatives do not work.