* reactor Mono/Flux enhancement : post review fixes (methods renamed to reactor)
* new span started on @ContinueSpan should be closed on completion
* #1027 tests for reactor Mono/Flux support enhancement
fixes gh-1027
without this change there's no easy way to work with message headers for Spring Integration
with this change we're allowing to register beans that will override the default behaviour
fixes gh-1032
* Added an option to filter out outbound requests
without this change there's no way to filter outbound requests basing on the path via a property.
with this change we're introducing that
fixes gh-1047
* close span when Exception is ClientAbortException
* Support custom to ignore exception in trace.
* explicit import statements
* Add test to verify that the span got reported. Fix typo.
Fixes gh-1038
* Adjusts expectations of data reporting of loopback spans
The latest version of Brave corrects a data merge problem where shared
spans (created via Tracer.joinSpan) would lose some data. Along the way,
this fixes some misuse of `Tracer.joinSpan` which should only be used
for RPC server spans.
See https://github.com/openzipkin/brave/pull/734 applied in Brave 5.1.3
* Makes the test more better
without this change we wrapped in the handleSend and handleReceive both the method wrapping the execution of the TraceLoadBalancerFeignClient and its delegate too. That way we had 2 pairs of CS & CR.
with this change we disable that functionality by wrapping only the delegate. The problem with such an approach was such that there were exceptions thrown before the delegate was executed. That's why we added a fallback mechanism in that case.
fixes gh-1007
without this change the kafka binder wasn't able to propagate tracing headers without explicitly adding on classpath the deprecated sleuth-stream module
with this change we're adding the stream support in core of Sleuth
fixes gh-1005
without this change the instrumentation of RibbonCommand lead to creation of an additional client span
with removal of the instrumentation of the RibbonCommand no longer do we create that span
fixes gh-997
This is done by implementing NO-OP QueueSubscription on the subscriber
and removing the guard against `Fuseable` publisher sources.
As `Fuseable` is a marker interface there should be no reason for code
to perform casts so for now the fact that the `lift` call in
ReactorSleuth doesn't actually produces a `Fuseable` publisher should
not be a problem.
fixes gh-994
