Prior to this commit, a BeanDefinitionReader registered via
@ImportResource was required to be public and have a public
constructor that accepts a single BeanDefinitionRegistry. However, the
public visibility requirements are not necessary, and the requirements
for the constructor's formal parameter list is not documented.
To address those issues, this commit removes the public visibility
restrictions and documents that a BeanDefinitionReader registered via
@ImportResource must declare a constructor that accepts a single
BeanDefinitionRegistry.
In addition, this commit includes the cause of the instantiation
failure in case the registered BeanDefinitionReader cannot be
instantiated.
Closes gh-34928
After the bootstrap phase (and with spring.locking.strict=true during the bootstrap phase), getSingletonFactoryBeanForTypeCheck always locks. In a background bootstrap thread, it never locks. Otherwise, it tries locking and explicitly resolves the bean class for subsequent type-based resolution (even for a component-scanned class) when it fails to acquire the lock. Furthermore, getObjectFromFactoryBean follows the same locking algorithm for post-processing.
Closes gh-34902
Spring Framework 6.2 introduced support for an escape character for
property placeholders (by default '\'). However, as of Spring Framework
6.2.6, there was no way to either escape the escape character or disable
escape character support.
For example, given a `username` property configured with the value of
`Jane.Smith` and a `DOMAIN\${username}` configuration string, property
placeholder replacement used to result in `DOMAIN\Jane.Smith` prior to
6.2 but now results in `DOMAIN${username}`. Similarly, an attempt to
escape the escape character via `DOMAIN\\${username}` results in
`DOMAIN\${username}`.
In theory, one should be able to disable use of an escape character
altogether, and that is currently possible by invoking
setEscapeCharacter(null) on AbstractPropertyResolver and
PlaceholderConfigurerSupport (the superclass of
PropertySourcesPlaceholderConfigurer).
However, in reality, there are two hurdles.
- As of 6.2.6, an invocation of setEscapeCharacter(null) on a
PropertySourcesPlaceholderConfigurer applied to its internal
top-level PropertySourcesPropertyResolver but not to any nested
PropertySourcesPropertyResolver, which means that the `null` escape
character could not be effectively applied.
- Users may not have an easy way to explicitly set the escape character
to `null` for a PropertyResolver or
PropertySourcesPlaceholderConfigurer. For example, Spring Boot
auto-configures a PropertySourcesPlaceholderConfigurer with the
default escape character enabled.
This first issue above has recently been addressed by gh-34861.
This commit therefore addresses the second issue as follows.
- To allow developers to easily revert to the pre-6.2 behavior without
changes to code or configuration strings, this commit introduces a
`spring.placeholder.escapeCharacter.default` property for use with
SpringProperties which globally sets the default escape character that
is automatically configured in AbstractPropertyResolver and
PlaceholderConfigurerSupport.
- Setting the property to an empty string sets the default escape
character to `null`, effectively disabling the default support for
escape characters.
spring.placeholder.escapeCharacter.default =
- Setting the property to any other character sets the default escape
character to that specific character.
spring.placeholder.escapeCharacter.default = ~
- Setting the property to a string containing more than one character
results in an exception.
- Developers are still able to configure an explicit escape character
in AbstractPropertyResolver and PlaceholderConfigurerSupport if they
choose to do so.
- Third-party components that wish to rely on the same feature can
invoke AbstractPropertyResolver.getDefaultEscapeCharacter() to obtain
the globally configured default escape character.
See gh-9628
See gh-34315
See gh-34861
Closes gh-34865
Commit 3295289e17 fixed a number issues with placeholder resolution in
PropertySourcesPlaceholderConfigurer. However, in doing so, it replaced
a raw PropertySource with a CompositePropertySource which implements
EnumerablePropertySource.
Consequently, all property sources registered in the Environment must
now implement EnumerablePropertySource (which is not an actual
requirement). Otherwise, invocations of getPropertyNames() on the
CompositePropertySource result in an IllegalStateException, and that is
a breaking change which resulted in numerous build failures within the
Spring portfolio.
To address that regression, this commit introduces a private
ConfigurableEnvironmentPropertySource in
PropertySourcesPlaceholderConfigurer which is a "raw" PropertySource
that delegates directly to the PropertySources in a
ConfigurableEnvironment.
This commit also extracts the raw PropertySource for direct Environment
delegation into a new FallbackEnvironmentPropertySource.
See gh-17385
Closes gh-34861
Currently, the placeholder resolution algorithm in
PropertySourcesPlaceholderConfigurer fails in several scenarios, and
the root cause for this category of failures has actually existed since
PropertySourcesPlaceholderConfigurer was introduced in Spring Framework
3.1.
Specifically, PropertySourcesPlaceholderConfigurer creates its own
PropertySourcesPropertyResolver that indirectly delegates to another
"nested" PropertySourcesPropertyResolver to interact with
PropertySources from the Environment, which results in double
placeholder parsing and resolution attempts, and that behavior leads to
a whole category of bugs.
For example, #27947 was addressed in Spring Framework 5.3.16, and due
to #34315 and #34326 we have recently realized that additional bugs
exist with placeholder resolution: nested placeholder resolution can
fail when escape characters are used, and it is currently impossible
to disable the escape character support for nested resolution.
To address this category of bugs, we no longer indirectly use or
directly create a "nested" PropertySourcesPropertyResolver in
PropertySourcesPlaceholderConfigurer. Instead, properties from property
sources from the Environment are now accessed directly without
duplicate/nested placeholder resolution.
See gh-27947
See gh-34326
See gh-34862
Closes gh-34861
In commit 05ebca8677, the `public` modifier was removed from the
SpringAtInjectTckTests class, which prevents it from being run as a
JUnit 3 test class.
To address that, this commit adds the missing `public` modifier as well
as a a code comment to help prevent this from happening again.
In addition, this commit updates spring-context.gradle to ensure that
the JUnit Vintage test engine is always applied. However, that Gradle
configuration is unfortunately ignored due to how our TestConventions
class has been implemented. Thus, that issue will have to be addressed
separately.
Closes gh-34800
Includes spring.locking.strict revision to differentiate between true, false, not set.
Includes checkFlag accessor on SpringProperties, also used in StatementCreatorUtils.
Closes gh-34729
See gh-34303
Historically, @Configuration classes that did not declare @Bean
methods were allowed to be abstract. However, the changes made in
76a6b9ea79 introduced a regression that prevents such classes from
being abstract, resulting in a BeanInstantiationException. This change
in behavior is caused by the fact that such a @Configuration class is
no longer replaced by a concrete subclass created dynamically by CGLIB.
This commit restores support for abstract @Configuration classes
without @Bean methods by modifying the "no enhancement required" check
in ConfigurationClassParser.
See gh-34486
Closes gh-34663
Prior to this commit, if a String 'value' attribute of an annotation
was annotated with @AliasFor and explicitly configured to alias an
attribute other than @Component.value, the value was still used as the
@Component name, but the warning message that was logged stated that
the 'value' attribute should be annotated with
@AliasFor(annotation=Component.class). However, it is not possible to
annotate an annotation attribute twice with @AliasFor.
To address that, this commit revises the logic in
AnnotationBeanNameGenerator so that it issues a log message similar to
the following in such scenarios.
WARN o.s.c.a.AnnotationBeanNameGenerator - Although the 'value'
attribute in @example.MyStereotype declares @AliasFor for an
attribute other than @Component's 'value' attribute, the value is
still used as the @Component name based on convention. As of Spring
Framework 7.0, such a 'value' attribute will no longer be used as the
@Component name.
See gh-34346
Closes gh-34317
Prior to this commit, `CronExpression` would support Quartz-style
expressions with "Nth occurence of a dayOfWeek" semantics by using the
`TemporalAdjusters.dayOfWeekInMonth` JDK support. This method will
return the Nth occurence starting with the month of the given temporal,
but in some cases will overflow to the next or previous month.
This behavior is not expected for our cron expression support.
This commit ensures that when an overflow happens (meaning, the
resulting date is not in the same month as the input temporal), we
should instead have another attempt at finding a valid month for this
expression.
Fixes gh-34360
This commit fixes a regression introduced by gh-24560, when adding
execution metadata support for scheduled tasks. The
`OutcomeTrackingRunnable` would delegate to the actual runnable but
could also hide whether it implements the `SchedulingAwareRunnable`
contract.
This commit ensures that `OutcomeTrackingRunnable` always implements
that contract and delegates to the runnable if possible, or return
default values otherwise.
Fixes gh-34058
This commit restores the user class in generated RootBeanDefinition
instances. Previously the CGLIB subclass was exposed. While this is
important in regular runtime as the configuration class parser operates
on the bean definition, this is not relevant for AOT as this information
is internal and captured in the instance supplier.
Closes gh-33960
This commit restores the signature of instance suppliers that are
exposing a CGLIB proxy. While calling the CGLIB proxy itself, and
making it available in BeanInstanceSupplier, is needed internally, such
type should not be exposed as it is an internal concern.
This was breaking InstanceSupplier.andThen as it expects the public
type of the bean to be exposed, not it's eventual CGLIB subclass.
Closes gh-33998
Prior to this commit, the equals() implementation in AdvisedSupport's
MethodCacheKey only considered methods to be equal based on an identity
comparison (`==`), which led to duplicate entries in the method cache
for the same logical method.
This is caused by the fact that AdvisedSupport's
getInterceptorsAndDynamicInterceptionAdvice() method is invoked at
various stages with different Method instances for the same method:
1) when creating the proxy
2) when invoking the method via the proxy
The reason the Method instances are different is due to the following.
- Methods such as Class#getDeclaredMethods() and
Class#getDeclaredMethod() always returns "child copies" of the
underlying Method instances -- which means that `equals()` should be
used instead of (or in addition to) `==` whenever the compared Method
instances can come from different sources.
With this commit, the equals() implementation in MethodCacheKey now
considers methods equal based on identity or logical equality, giving
preference to the quicker identity check.
See gh-32586
Closes gh-33915
Prior to this commit, the reactive Scheduled tasks would be wrapped as a
`SubscribingRunnable` which does not implement a custom `toString`. This
would result in task metadata using the default Java `toString`
representation for those.
This commit ensures that the bean class name and method name are used
for this `toString`.
Closes gh-34010
This commit introduces a test which verifies how many times the
matches() method of a StaticMethodMatcherPointcut is invoked during
ApplicationContext startup as well as during actual method invocations
via the advice chain, which also indirectly tests the behavior of the
equals() implementation in AdvisedSupport.MethodCacheKey.
In addition, this commit revises BeanFactoryTransactionTests to assert
that a transaction is started for the setAge() method.
See gh-33915
The log message for a NoClassDefFoundError is now a DEBUG level message
handled like a TypeNotPresentException and similar to the following.
DEBUG: Skipping validation constraint hint inference for class
org.example.CustomConstraint due to a NoClassDefFoundError for
com.example.MissingType
See gh-33949
Prior to this commit, AOT processing for bean validation failed with a
NoClassDefFoundError for constraints with missing dependencies.
With this commit, the processing no longer fails, and a warning is
logged instead.
See gh-33940
Closes gh-33949
Co-authored-by: Sam Brannen <sam.brannen@broadcom.com>
