Prior to this commit, if a non-annotated [1] class was registered
directly with an ApplicationContext -- for example, via
AnnotatedBeanDefinitionReader, AnnotationConfigApplicationContext, or
@ContextConfiguration -- it was not considered a configuration class in
'lite' mode unless @Bean methods were declared locally. In other words,
if the registered class didn't declare local @Bean methods but rather
extended a class that declared @Bean methods, then the registered class
was not parsed as a @Configuration class in 'lite' mode, and the @Bean
methods were ignored.
Whereas, a non-annotated class registered via @Import is always
considered to be a configuration class candidate.
To address this discrepancy between @Import'ed classes and classes
registered directly with an ApplicationContext, this commit treats any
class registered via AnnotatedBeanDefinitionReader as a @Configuration
class candidate with @Bean 'lite' mode semantics.
[1] In this context, "non-annotated" means a class not annotated (or
meta-annotated) with @Component, @ComponentScan, @Import, or
@ImportResource.
Closes gh-30449
This commit removes the JBoss Application Server external javadoc link,
which covers too many packages and can thus problematically take
precedence for said packages and lead to dead links.
It is replaced by Hibernate javadocs for `org.hibernate.*` packages,
JakartaEE and jsr305 for annotations. We lose a limited number of links
around org.xnio classes and Hibernate-and-JTA-related classes, as a
trade-off.
See gh-30455
See gh-30428
This commit improves how the build deals with javadoc invalid references
in two ways.
Link/see references that are temporarily invalid during javadoc
generation of individual modules are better masked by using the option
`Xdoclint:syntax` instead of `Xdoclint:none` (warnings were still
visible in some cases, e.g. when individually building the javadoc for
a specific module).
Global javadoc-building task `api` now combines `syntax` and `reference`
`Xdoclint` groups, allowing to raise truly invalid references even when
all the modules have been aggregated.
This commit also fixes the 20+ errors which appeared following the later
change in doclet configuration.
Closes gh-30428
In order to restore native support with Hibernate 6.2, this
change updates the PersistenceUnitInfoDescriptor instance
created by SpringHibernateJpaPersistenceProvider in order to
skip the class transformer for native images.
Closes gh-30492
Bean post processors that use InjectionMetadata checks if a property
value for the element it is about to inject is set and skip it, so
that the property value is used. Previously, the AOT contribution for
the same behavior did not check if a matching property value is set
and therefore override the user-defined value.
This commit introduces an additional method that filters the injected
element list so that only the elements that should be processed are
defined.
Closes gh-30476
Prior to this commit, test AOT processing failed when using the GraalVM
tracing agent and GraalVM Native Build Tools (NBT) plugins for Maven
and Gradle.
The reason is that the AOT support in the TestContext framework (TCF)
relied on AotDetector.useGeneratedArtifacts() which delegates
internally to NativeDetector.inNativeImage() which does not
differentiate between values stored in the
"org.graalvm.nativeimage.imagecode" JVM system property.
This commit addresses this issue by introducing a TestAotDetector
utility that is specific to the TCF. This detector considers the
current runtime to be in "AOT runtime mode" if the "spring.aot.enabled"
Spring property is set to "true" or the GraalVM
"org.graalvm.nativeimage.imagecode" JVM system property is set to any
non-empty value other than "agent".
Closes gh-30281
Restores proper event type propagation to parent context.
Selectively applies payload type to given payload object.
Also reuses cached type for regular ApplicationEvent now.
Closes gh-30360
This commit fixes the check by avoiding a fallback to eventType's
hasUnresolvableGenerics(). This could previously lead to checking a
generic event type `A<T>` against a listener which accepts unrelated
`B` and return `true` despite the inconsistency.
Note that this wouldn't necessarily surface to the user because there is
a `catch (ClassCastException e)` down the line, which was primarily put
in place to deal with lambda-based listeners but happens to catch an
exception thrown due to the bad result of `supportsEventType`.
The `supportsEventType` now matches generic `PayloadApplicationEvent`
types with a raw counterpart, using the above fallback only in that case
and otherwise ultimately returning `false`.
Closes gh-30399
This commit adapts the generated code so that each injection point has
a dedicated namespace in the form of a private method. That prevents
the same variable to be reused twice which lead to a compilation failure
previously.
Closes gh-30437
When ReactorNetty2StompBrokerRelayIntegrationTests fail, typically there
are multiple exceptions "Connection refused: /127.0.0.1:61613" that
appear after we've conneted, sent CONNECT, and expecting CONNECTED, but
that does not come within the 10 second timeout.
61613 is the default port for STOMP. However, in all integration tests
we start ActiveMQ with port 0 which results in a random port. Moreover,
the stacktrace is for Netty 4 (not 5), and the eventloop thread id's
are different than the one where the connection to the correct, random
port was established.
The suspicion is that these are log messages from
MessageBrokerConfigurationTests which focuses on testing configuration
but nevertheless as a bean starts and attempts to connect to the default
port and fails. Perhaps those attempts to connect on the default port
somehow affect the ActiveMQ server, and it stops responding.
This change adds a no-op TcpClient in MessageBrokerConfigurationTests
to avoid unnecessary attempts to connect that are not needed.
See gh-29287
The current test were not catching the issue because they request 1
via StepVerifier, wait for it, and then cancel. In the case of
StringDecoder it means all chunks are used up to produce that first
String and so the cancel doesn't catch any cached chunks.
Closes gh-30299
This commit augment the AOT section with best practices we have
experienced while working on the AOT engine. In particular, it describes
how a FactoryBean with an unresolved generic should be handled.
Closes gh-30434
This commit reviews BeanInstanceSupplier to reuse more code from
ConstructorResolver. Previously, the autowired argument resolution was
partially duplicated and this commit introduces a new common path via
RegisteredBean#resolveAutowiredArgument.
Closes gh-30401
This commit adds a workaround for oracle/graal#6529
triggered by b3748243199fe809f177bb7922c69f7eb37ab6d2.
When the GraalVM fix will have reached a wide enough
audience, it should be removed via gh-30394.
Closes gh-30407
Sam Brannen
greg.lee
Sébastien Deleuze
Stephane Nicoll
Johnny Lim
Simon Baslé
Jan Mewes
Anyuruf
rstoyanchev
Spring Builds
Juergen Hoeller
Rob Winch
Arjen Poutsma
Brian Clozel