Update StreamUtils.drain to use InputStream.transferTo with a null
OutputStream. This avoids allocating buffers for cases where the
supplied InputStream has an optimized transferTo method (e.g.,
ByteArrayInputStream and FileInputStream).
Additionally, update StreamUtils.emptyInput to simply call
InputStream.nullInputStream.
Closes gh-28961
This commit records `@PropertySource` declarations defined on
configuration classes so that these are contributed to the environment
of a context that is initialized by generated code.
Closes gh-28976
This commit moves MethodReference to an interface with a default
implementation that relies on a MethodSpec. Such an arrangement avoid
the need of specifying attributes of the method such as whether it is
static or not.
The resolution of the invocation block now takes an
ArgumentCodeGenerator rather than the raw arguments. Doing so gives
the opportunity to create more flexible signatures.
See gh-29005
This commit updates GeneratedMethod and its underlying infrastructure
to be able to produce a MethodReference. This simplifies the need when
such a reference needs to be created manually and reuses more of what
MethodReference has to offer.
See gh-29005
This commit renames registerResourceIfNecessary() to registerResource()
and throws an exception if the class path resource does not exist.
Closes gh-29083
This aligns the buffer size used in StreamUtils with the buffer sizes
used throughout the JDK (see InputStream, BufferedInputStream, Files).
Closes gh-28965
As of Java 18, the serial lint warning in javac has been expanded to
check for class fields that are not marked as `Serializable`.
See https://www.oracle.com/java/technologies/javase/18all-relnotes.html#JDK-8202056
In the Spring Framework codebase, this can happen with `Map`, `Set` or
`List` attributes which are often assigned with an unmodifiable
implementation variant. Such implementations are `Serializable` but
cannot be used as field types.
This commit ensures that the following changes are applied:
* fields are marked as transient if they can't be serialized
* classes are marked as `Serializable` if this was missing
* `@SuppressWarnings("serial")` is applied where relevant
Prior to this commit, the `RuntimeHintsAgent` and its testing
infrastructure would assume that calling `MyClass.class.getMethods()`
requires a reflection hint on the class for introspecting public/private
methods.
GraalVM does not require this, in fact this call only returns methods
that have reflection hints in the native image.
This commit refines the agent behavior for `Class.getMethods()`,
`Class.getDeclaredMethods()`, `Class.getFields()` and
`Class.getDeclaredFields()`. With this change, registering at least one
method/field for reflection is enough to match.
During the execution of Java tests, all methods and fields will be
provided, regardless of hints being registered or not. This could cause
false negatives where we're missing reflection hints on methods or
fields.
This risk is mitigated thanks to additional instrumentation on
`Method.getAnnotations()`, `Method.getParameterTypes()` and
`Method.invoke()`. If a method is found reflectively, chances are it
will be used for further reflection.
Closes gh-29091
Since getPath() returns a relative path if the resource was created
using the ClassPathResource(String,Class) constructor, there was
previously no way to consistently obtain the absolute path to the
resource within the class path.
This commit addresses this shortcoming by introducing a new
getAbsolutePath() for consistently obtaining the absolute path to the
resource within the class path.
See gh-29083
Closes gh-29094
This commit introduces a new registerResourceIfNecessary() method in
RuntimeHintsUtils that simplifies the registration of hints for
`classpath:` resources.
Closes gh-29083
Prior to this commit, if a ResourcePatternHint was created with a
resource pattern with a leading slash, the hint was registered and
eventually converted to configuration for the GraalVM native image
compiler.
However, such a resource pattern is invalid for GraalVM. Consequently,
the registered resources were not available within the compiled native
image.
This commit ensures that registered patterns are applicable in a native
image by preventing creation of a ResourcePatternHint with a pattern
with a leading slash.
Closes gh-29088
Prior to this commit, if the TypeReference supplied to
ResourceHintsPredicates.forResource(TypeReference,String) was for a
class declared in the default package (i.e., without a package), the
resolveAbsoluteResourceName() method incorrectly prepended two leading
slashes (//) to the absolute resource name, causing correct matches to
fail.
This commit fixes this by adding special handling for a TypeReference
without a package name. In addition, this commit introduces lenient
handling of resource names by consistently removing a leading slash in
ResourceHintsPredicates.forResource(*) methods. The latter aligns with
absolute resource path handling in other places in the framework, such
as ClassPathResource.
Closes gh-29086
This commit adds support generating graalvm `jni-config.json` file.
Configuration format for jni/reflection in graalvm is documented
to be exactly same so we can re-use facilities for reflection hints
which should be relatively clean for a user as also graalvm uses same
classes for both jni/reflection.
Closes gh-29007
This annotation and the related processor allows to
register reflection hints for data binding purpose
(class, fields, properties, record components for
the whole type hierarchy)
Closes gh-28979
This commit provides an alternative
java.beans.Introspector#findCustomizerClass implementation
via a GraalVM substitution that avoids to include thousands
of unused AWT classes in the native image.
Closes gh-29060
Since Spring no longer adds the SynthesizedAnnotation interface to the
JDK dynamic proxy used to synthesize an annotation, this commit
officially deprecates SynthesizedAnnotation and related methods in
RuntimeHintsUtils.
See gh-29041, gh-29054
Closes gh-29053
SynthesizedAnnotation was originally introduced as a convenience for
easily detecting if an annotation had been synthesized by Spring via a
simple `if (myAnnotation instanceof SynthesizedAnnotation)` check.
However, the introduction of SynthesizedAnnotation in the JDK dynamic
proxy for a synthesized annotation results in a separate proxy class
for each annotation synthesized by Spring, and this causes issues with
GraalVM native images since users and framework developers must always
ensure that the additional proxy classes are registered.
This commit completely removes the use of SynthesizedAnnotation in
synthesized annotation proxies. This change allows the proxy class for
an annotation to be reused for a synthesized annotation of the same
annotation type.
Consequently:
- Extra proxy classes are not generated on the JVM or in a native image.
- Extra proxy classes are not required to be registered for a native image.
Closes gh-29041
Add `MemberCategory` and `FieldMode` shortcuts for type registration.
Helper `builtWith` methods have also been extracted to the Hint types
to allow general reuse (for example with `registerTypes`).
See gh-29011
Add a `FieldMode` enum analogous to `ExecutableHint` and update
`FieldHint` to ensure that registration cannot downgrade `WRITE`
to `READ`.
Fixes gh-29055
Reduce the visibility of hint factory methods that return a `Builder`.
Since the `build()` method of the `Builder` is package-private the
factory methods can also be package-private.
Since SynthesizedAnnotation will be deprecated (and potentially
completely removed) in Spring Framework 6.0, this commit introduces
AnnotationUtils.isSynthesizedAnnotation(Annotation) in 5.3.x to allow
people to migrate away from relying on SynthesizedAnnotation.
Closes gh-29054
Prior to this commit, the `ConcurrentLruCache` implementation would not
perform well under certain conditions. As long as the cache capacity was
not reached, the cache would avoid maintaining an eviction queue
(reordering entries depending with least/most recently read). When the
cache capacity was reached, the LRU queue was updated for each
read/write operation. This decreased performance significantly under
contention when the capacity was reached.
This commit completely rewrites the internals of `ConcurrentLruCache`.
`ConcurrentLruCache` is now a specialized version of the
`ConcurrentLinkedHashMap` [1]. This change focuses on buferring read and
write operations, only processing them at certain times to avoid
contention.
When a cached entry is read, a read operation is queued and buffered
operations are drained if the buffer reached a fixed limit. When a new
cache entry is added or removed, a write operation is queued and
triggers a drain attempt. When the capacity is outgrown, the cache polls
items from the eviction queue, which maintains elements with the
least recently used ones first. Entries are removed until the capacity
is under control.
The behavior described here and the buffer sizes are optimized with the
number of available processors in mind. Work is localized as much as
possible on a per-thread basis to avoid contention on the eviction queue.
The new implementation has been tested with the JMH benchmark provided
here, comparing the former `COncurrentLruCache`, the new implementation
as well as the `ConcurrentLinkedHashMap` [1].
When testing with a cache reaching capacity, under contention, with a
10% cache miss, we're seeing a 40x improvement compared to the previous
implementation and performance on par with the reference.
See [2] for how to replicate the benchmark.
[1] https://github.com/ben-manes/concurrentlinkedhashmap
[2] https://github.com/spring-projects/spring-framework/wiki/Micro-Benchmarks
Closes gh-26320
Brian Clozel
Patrick Strawderman
Stephane Nicoll
Sam Brannen
Phillip Webb
Sébastien Deleuze
Janne Valkealahti