Following on 3ebbfa2191 where the local
refCount was removed in favor of using the internal refCount of the
native data buffer, this commit ensures that LeakAwareDataBufferFactory
uses a PooledDataBufferFactory delegate by default.
There are also fixes for test issues with eager allocation uncovered by
these changes in StringDecoder and ResourceDecoder.
LeakAwareDataBuffer was keeping its own refCount rather than checking
through the delegate. This leads to false leak reports in a sequence
where an allocated buffer is retained and then sliced since it is not
aware of the changes to the refCount through the slice.
Use of Flux.just is problematic in that if the Flux is cancelled before
demand, the item may never be read, nor freed. Flux#just does not
even delegate cancellation signals.
Closes gh-22731
1. If the write Subscriber cancels with the item cached, release it.
2. If the write Publisher emits an error while the item is cached, when
the write Subscriber subscribes, release the cached item and emit the
error signal.
Closes gh-22720
The cancellation callback in asynchronousReadFileChannel must be called
before the first read I/O or otherwise if cancellation signals happens
immediately the onDispose callback may be missed.
The DefaultBufferFactory workaround however remains in place until an
expected additional fix arrives with Reactor Core 3.2.9.
See gh-22107
Following a response on
https://github.com/reactor/reactor-core/issues/1634 apparently
FluxSink is respecting doOnDiscard but there is a race condition
in DataBufferUtils with file reading.
This commit makes two changes:
1) Add more checks for onDispose to detect cancellation signals
as well as to deal with potentially concurrent such signal.
2) Do not close the channel through the Flux.using callback but
rather allow the current I/O callback to take place and only then
close the channel or else the buffer is left hanging.
Despite this tests still can fail due to a suspected issue in Reactor
itself with the doOnDiscard callback for FluxSink. That's tracked under
the same issue https://github.com/reactor/reactor-core/issues/1634
and for now the use of DefaultDataBufferFactory is enforced as a
workaround until the issue is resolved.
See gh-22107
