It turns out that `JdkSerializationRedisSerializer` by default is
based on the default Java class loader which may lead into
`ClassCastException` downstream after deserialization
* Make all the `JdkSerializationRedisSerializer` usage (default)
in Redis components based on the BeanFactory `ClassLoader`
* Fix tests to call `setBeanClassLoader()`
* Fix Mark Fisher's name in the `MultipartFileReader` :-)
**Cherry-pick to 5.1.x, 5.0.x & 4.3.x after restoring diamonds**
According Sonar report we need to re-implement `equals()` & `hashCode()`
methods in the `IntegrationWebExchangeBindException` since it
extends a `WebExchangeBindException` with those methods
When RSocket client connects to the server there is no reason to wrap
a `ConnectionSetupPayload` into a `Message` since we are not going to
send it downstream
* Refactor `IntegrationRSocket.handleConnectionSetupPayload()` just
return a `Mono<DataBuffer>` for converted `ConnectionSetupPayload`
* Ask for a `destination` and `RSocketRequester` from the
`IntegrationRSocket` instead of message headers
Looks like there is some connection race condition in Netty on Linux,
so we may not be started on server side fully when we try to select a
port from the OS
* Move local port resolution and, therefore, RSocket server start into
the earlier phase before creating a client
The `http://example.org` is not white-listed in `nohttp` and it
doesn't resolve SSL handshake on HTTPS schema
* Add `@Deprecated` into `SimpleMessageListenerContainerSpec.txSize()`
* It looks like we don't need to manage Reactor dependencies in the
RSocket module: the `io.rsocket:rsocket-transport-netty` brings
everything we need, although it doesn't that with its own supported
versions
* Use non-HTTPS url for `example.org` in test config to avoid SSL
handshake
* Add Validation to HTTP Inbound
* Pull a validation functionality from `WebFluxInboundEndpoint` to its
super class `BaseHttpInboundEndpoint` making a validation available for
the `HttpRequestHandlingEndpointSupport` as well
* Do the same for the `validator()` option in DSL for the
`HttpInboundEndpointSupportSpec`
* Add `validator` XML attribute for both HTTP and WebFlux inbound
endpoint XSDs
* Test parsers for a new `validator` option
* Document validation in the `http.adoc`
* Apply some polishing in the `http.adoc`, as well as in the
`HttpRequestHandlingEndpointSupport` JavaDocs
* Clarify in `webflux.adoc` that validation is applied for the
`Publisher` items before the payload is finally built for the message to
send.
* Add WebFlux into the table of endpoints in the `endpoint-summary.adoc`
* * Remove unused imports
* Fix `SimpleMessageListenerContainerSpec` for deprecated `txSize` option
Fixes https://github.com/spring-projects/spring-integration/issues/2974
There is a race in that we could get a `SocketException` in `inputStream`.
Since this is between payloads, it needs to be thrown as a
`SoftEndOfStreamException`.
* * Add javadocs to SoftEndOfStreamException
* * SEOSE is now a RuntimeException (can't be an UncheckedIOException - no default CTOR
* Add validation support into WebFlux Inbound
See https://stackoverflow.com/questions/56729517/spring-integration-webflux-validate-input-json-body
* Introduce a `WebFluxInboundEndpoint.validator` option to validate
a request payload after conversion HTTP request into a `Publisher`
* Expose a `WebFluxInboundEndpointSpec.validator()` option
* Introduce a `IntegrationWebExchangeBindException` to extend a
`WebExchangeBindException` for Spring Integration use-case.
such an exception is used in the `ResponseStatusExceptionHandler`
extensions, like one in Spring Boot for proper response rendering
* Test and document the solution
* * Remove unused imports from IntegrationWebExchangeBindException
* Fix `SmartLifecycle.stop(Runnable)` usage
We always have to call `callback` in the `SmartLifecycle.stop(Runnable)`
implementation independently of component state
* Fix `StandardIntegrationFlow.stop(Runnable)` for a logic when we
don't have any `this.lifecycles`
* Remove those `stop(Runnable)` which are fully equivalent of the
`default` on in the `SmartLifecycle`
* Some other simple polishing for the affected classes, e.g.
`isSingleton()` is `default` with `true` in the `InitializingBean`
**Cherry-pick to 5.1.x**
* * Fix checkstyle violation
It turns out that `contentType` header (as well as `Accept`) may be
converted into `MimeType` upstream before the message is sent into HTTP
Outbound Channel Adapter meanwhile the logic is only to support String
and MediaType
* Change the `DefaultHttpHeaderMapper` to support more generic `MimeType`
for the `Content-Type` and `Accept` HTTP headers which can simply be
converted to the expected `MediaType`
**Cherry-pick to 5.1.x & 5.0.x**
* GH-2967: Fix ScatterGatherH for headers copy
Fixes https://github.com/spring-projects/spring-integration/issues/2967
The `ChannelInterceptor` is added into the `this.gatherChannel` on each
request message making a subsequent requests for scatter-gather as
halting on reply.
* Add an interceptor into an injected `this.gatherChannel` only once
during `ScatterGatherHandler` initialization
* Introduce `ORIGINAL_REPLY_CHANNEL` and `ORIGINAL_ERROR_CHANNEL`
headers to carry a request reply and error channels from headers
* Populate `REPLY_CHANNEL` and `ERROR_CHANNEL` headers back before
sending scattering replies into gatherer
* Transfer a `GATHER_RESULT_CHANNEL` header now directly from the scatter
message to make it available in the reply from the gatherer
* Add note about those headers in the `scatter-gather.adoc`
* Modify `ScatterGatherTests` to be sure that `ScatterGatherHandler`
works for several requests
**Cherry-pick to 5.1.x**
* * Fix language in doc
Looks like we can't have a custom `whitelist.lines` in the `/checkstyle`
dir.
* Move `whitelist.lines` with the `http://www.jcraft.com` entry into the
`/src/nohttp`
* Upgrade `nohttp` to `0.0.3`
* Revert `jsch` URL in the `sftp.adoc`
* Upgrade to Micormeter-1.1.5
JIRA: https://jira.spring.io/browse/INT-3073
Add `FluxAggregatorMessageHandlerTests`
* Add `FluxAggregatorMessageHandlerTests`
* Fix `FluxAggregatorMessageHandler` for the default `messageForWindowFlux`
to rely on the `publish().autoConnect()` to build a target `Flux` for
window and also copy headers from the first `Message` in the window.
Looks like `switchOnFirst()` doesn't work somehow with windows or I just
don't use it properly, although it isn't clear how to continue chain but
get the whole `Flux` as a single entry for message payload
* Add more tests for `FluxAggregatorMessageHandlerTests`
* Fix `AbstractMessageSplitter` to cast properly a primitive array
* Implement `Lifecycle` in the `FluxAggregatorMessageHandler`
* Add JavaDocs into the `FluxAggregatorMessageHandler`
Javadoc polishing
Resolves https://github.com/spring-projects/spring-integration/issues/2958
The previous commit added the raw source message to a common header
`IntegrationMessageHeaderAccessor.SOURCE_DATA` so that the spring-cloud-stream
can construct an error message containing the data when using functions.
However, we also need to add it to the common header when the `ErrorMessage`
is created by Spring Integration for the normal message channel binder.
This allows the DLQ processing to be agnostic as to who created the
error message.
Looks like latest upgrades for the CI server made virtual machines a bit
slower (less resources are allocated), so we don't have a reasonable
time for switching between threads.
* Move the plugin and its task into the root project configuration
* Add more image file extensions to exclude
* Fix `sftp.adoc` do not use HTTP url for the JSCH library project page;
use GitHub repo link instead
