When built in an environment where many Gradle build workers are
available, :spring-webflux:test is regularly the only task running at
the end of the build. Therefore, the build's overall execution time
can be reduced by running its tests in parallel, spreading the tests'
execution across the available workers. The configured number of forks
is a maximum with Gradle reducing this as necessary for environments
with low numbers of cores where multiple workers will not improve
build performance.
Closes gh-23701
RSocket is still in its RC phase, but those artifacts are published on
maven central. This dependency is currently still depending on a
milestone dependency for Reactor adn this is causing issues with our
build.
This commits adds back the milestone repository until the RSocket POMs
and BOMs are fixed. Once fixed, the Framework build should use the
RSocket BOM for managing RSocket dependencies.
See gh-23698
The parallel forks configuration of Test tasks can have a big impact
on their execution time. This commit updates the build can
configuration to capture each Test task's maxParallelForks
configuration as custom values in the build scan. This will make this
data available when comparing scans which will be useful as we
experiment with different parallel fork settings to reduce the
build's overall execution time.
Closes gh-23691
Prior to this commit, the parameterized DataBufferFactory was never
actually used when setting up the WebClient for each test. This was due
to an oversight when migrating from JUnit 4 to JUnit Jupiter.
See: https://github.com/reactor/reactor-netty/issues/860
This commit fixes this by converting the existing @BeforeEach method to
a local setup method that is invoked from each
@ParameterizedDataBufferAllocatingTest method.
In addition, in order to avoid the 2 second "quiet period" that is
incurred when destroying the ReactorResourceFactory, this commit moves
the setup and destruction of the ReactorResourceFactory to new
@BeforeAll and @AfterAll methods.
The test instance lifecycle has also been switched to PER_CLASS to avoid
static state in the test class.
If an error looks like a "disconnected client" but the response is not
yet committed then it can't be an I/O error from writing to the server
response. It is most likely as a result of a remote call as part of
request handling.
Not setting the response to 500 in this case results in a 200 response
status despite the error. Even if it was an I/O error from the server
response, setting the status won't impact a failed response.
Closes gh-23319
Prior to this change, the Gradle API Diff task was failing due to a
missing class (a Guava class) during processing. This might be a
dependency that was previously brought by some other dependency.
This commit adds that dependency as a workaround, and improves the
Gradle API Diff task to only select relevant projects when running the
task (projects using the Java Plugin and the Maven Publish Plugin).
Take a slice of the metadata if not using composite metadata to allow
reading it multiple times. For composite metadata this is not an isuse
as it is ensured by the underlying RSocket Java API.
TransactionOperator.as(Mono) now no longer short-cuts via a Flux.next() but provides an implementation via Mono.usingWhen(…).
The short-cut previously issued a cancellation signal to the transactional Mono causing the transaction cleanup to happen without a handle for synchronization.
Using Mono.usingWhen(…) initiates transaction cleanup when the Mono completes eliminating the need for cancellation of the transactional Publisher.
This change does not fully fix gh-23304 but it softens its impact because TransactionalOperator.transactional(Mono) avoids cancellation.
This commit reverts b2704e1db6 and
configures the ReactorResourceFactory not to use global resources,
thereby allowing all tests in WebClientDataBufferAllocatingTests to
execute within approximately 2 seconds again on Mac OS.
