Prior to this commit, the `SimpleCommandLineArgsParser` would reject
"--" arguments as invalid. As reported by the community, the POSIX
utility conventions (Guideline 10) state that
> The first -- argument that is not an option-argument should be
> accepted as a delimiter indicating the end of options.
> Any following arguments should be treated as operands, even if they
> begin with the '-' character.
This commit updates `SimpleCommandLineArgsParser` to not reject "--"
arguments and instead to consider remaining arguments as non-optional.
See gh-31513
Prior to this commit, `ExchangeFilterFunction` could only get the
current observation from the reactor context. This is particularly
useful when such filters want to add KeyValues to the observation
context.
This commit makes this use case easier by adding the context of the
current observation as a request attribute. This also aligns the
behavior with other instrumentations.
Fixes gh-31609
Prior to this commit, the Jackson 2.x encoders, in case of encoding a
stream of data, would first release the `ByteArrayBuilder` and then the
`JsonGenerator`. This order is inconsistent with the single value
variant (see `o.s.h.codec.json.AbstractJackson2Encoder#encodeValue`) and
invalid since the `JsonGenerator` uses internally the
`ByteArrayBuilder`.
In case of a CSV Encoder, the codec can buffer data to write the column
names of the CSV file. Writing an empty Flux with this Encoder would not
fail but still log a NullPointerException ignored by the reactive
pipeline.
This commit fixes the order and avoid such issues at runtime.
Fixes gh-30493
Prior to this commit, the JMS listener observations would only cover the
actual listener invocation, but not the error handling phase. This means
that all errors (including the handled ones) are marked as errors in the
observation and that the observation is not current anymore when error
handling happens.
This commit aligns the behavior with the other Spring Framework
instrumentations and ensures that the error handling phase is fully part
of the observation recording.
Closes gh-31559
This commit documents that the "error" key in Micrometer Observations
should be preferred over the legacy "exception" one. Right now the
information is duplicated but we might remove the deprecated one in the
future.
Closes gh-31514
This commit surfaces the ETag generation feature for both
`ResourceHttpRequestHandler` and `ResourceWebHandler` on their
respective `ResourceHandlerRegistration` for easier configuration.
See gh-29031
Prior to this commit, the `ResourceHttpRequestHandler` would support
HTTP caching when serving resources, but only driving it through the
`Resource#lastModified()` information.
This commit introduces an ETag generator function that can be configured
on the `ResourceHttpRequestHandler` to dynamically generate an ETag
value for the Resource that is going to be served.
Closes gh-29031
This commit re-generates the protobuf Java classes with a recent version
of the protoc binary and adds JMH benchmarks that exercise the message
converter for both the reading and writing cases.
See gh-29496
Prior to this commit, the `ConcurrentReferenceHashMap#Segment` would
recalculate its element count during the restructure phase by
subtracting the number of polled elements from the queue to the current
count.
Later, the newly restructured `references` array would only contain
references that 1) are not in the set of elements to purge and
2) that hold a non-null value.
The issues with this approach are multiple. The newly calculated count
is only an estimate, as some situations can make it invalid, even
temporarily.
* since we initially collected all elements to be purged from the queue,
the GC might have collected new values. This means that we might
filter out more references that we initially intended to
* because the restructure operation re-creates new references for all
elements in the original array, we might later get references from the
queue that are not in the array anymore. This could lead to
"duplicate" removals for the same value
Because several methods in the Segment class have special no-op behavior
when `count == 0`, an invalid count can lead to keys appearing missing
when they are actually still present. In some scenarios, this can
decrease the performance of the cache since values need to be
recalculated.
This commit fixes this inconsistency count issue by first using an
estimate in order to decide whether the array needs a resize and then by
counting the actual numbers of elements inserted in the restructured
array as the new count.
Fixes gh-31373
Since the Spring WebFlux HTTP server instrumentation has been moved from
the `WebFilter` to the `HttpWebHandlerAdapter`, we need to apply similar
changes there.
See gh-31417
Prior to this commit, HTTP server observations for Spring WebFlux could
be recorded twice for a single request in some cases. The "COMPLETE" and
"CANCEL" signals would race in the reactive pipeline and would trigger
both the `doOnComplete()` and ` `doOnCancel()` operators, each calling
`observation.stop()` on the current observation.
This would in fact publish two different observations for the same
request.
This commit ensures that the instrumentation uses the `Mono#tap`
operator to guard against this case and only call `Observation#stop`
once for each request.
Fixes gh-31417
Prior to this commit, a cancelled exchange would always result in an
`"status":"UNKNOWN"` KeyValue. This only applied to reactive variants,
as cancelled exchanges are not currently detected for Servlet
implementations.
In some cases, exchanges can be cancelled by clients before they are
completed, but the response was actually received by the client. The
response status information has been set by the application and the
response has been committed. For those cases, we shouldn't assume an
"UNKNOWN" value.
This commit assumes that committed responses have a response status set
by the application and that the observations should reflect that. From
now on, we only assume an "UNKNOWN" status if the response has not been
commited.
Fixes gh-31388
Prior to this commit, the bean definition properties code generator
would register hints for invoking the setter methods of registered
property values defined for the bean definition.
The internal algorithm is also reflecting on the Field to discover
annotations. Doing so actually calls `getDeclaredFields` to iterate on
the available fields. This is done recursively up the type hierarchy
until the field is found.
This commit registers the required reflection metadata.
Closes gh-31390
This commit removes the previously introduced reflection hints that were
working around known issues in GraalVM.
Spring Framework 6.1 will require recent maintenance versions of GraalVM
and should not contribute such hints anymore.
Closes gh-30394
Prior to this commit, the bean registration AOT contributions would
register introspection and invocation hints on both declared and public
methods for bean types. The bean introspection algorithm also looks at
default methods implemented by interfaces when collecting bean property
information.
This commit ensures that introspection hints are registered for all
implemented interfaces when registering beans.
Closes gh-31350
Prior to this commit, the `BeanRegistrationsAotContribution` would only
contribute introspection hints for declared methods. This does not cover
inherited public methods.
This commit adds the missing hint on public methods.
Fixes gh-31293
This commit configures the Foojay resolver plugin in the Gradle build in
order to auto-provision a required JDK distribution if it's not present
already on the host.
Currently the Spring Framework build requires both a JDK 17 and a JDK 21
distribution to build artifacts.
Closes gh-30474
Prior to this commit, `DisposableBeanAdapter` supported reactive bean
destroy methods by detected if `Publisher` is available on the
classpath. The AOT engine did not contribute a reflection hint for this
call.
This commit ensures that this reflection hint is registered in all
cases, even if there are no destroy methods detected on beans.
Fixes gh-31278
`PathMatchingResourcePatternResolver` is reflecting on
`org.eclipse.core.runtime.FileLocator` and invoking methods on it for
OSGi support. While this use case is highly unlikely in native images,
registering the reflection entry by itself should be enough.
Fixes gh-31271
This commit adds the relevant reflection hints required by
`KotlinDetector`; this class is trying to detect both `kotlin.Metadata`
and `kotlin.reflect.full.KClasses`.
Fixes gh- 31269
This commit adds reflection hints for `jakarta.inject.Provider` and
ensures that hints are always contributed even if jakarta classes are
not on the classpath.
Fixes gh-31259
Prior to this commit, `ClassUtils#forName` would always attempt to
resolve the given class name as a nested type. For example, searching
for `org.example.Spring` would try to resolve:
* `org.example.Spring`
* if not available, try `org.example$Spring` as well
Java classes usually start with uppercase letters, so this additional
lookup can be costly and not very useful.
This commit only attempts nested class lookups when the previous segment
starts with an uppercase. So `org.example.Spring.Issue` will look for
`org.example.Spring$Issue`, but `org.example.Spring` will not.
Closes gh-31258
Prior to this commit, the `RuntimeHintsPredicates` would assume that
registering introspection or invocation hints for "all declared methods"
on a type would also include "all public methods". This is not true, as
the Java reflection API itself behaves differently.
`getDeclaredMethods()` does not return a superset of `getMethods()`, as
the latter can return inherited methods, but not the former.
Same reasoning applies to fields.
This commit fixes the hints predicates to only match if the correct hint
has been registered.
Fixes gh-31224
This commit ensures that both `ObservationRegsitry` and
`ServerRequestObservationConvention` beans are automatically detected in
the application context if they are unique. This aligns with the
existing behavior for all other builder methods.
Closes gh-31205
This commit replaces the initial allocation size for the content caching
buffer by a `FastByteArrayOutputStream`. With this variant, allocations
are cheap and we don't need to apply heuristics anymore to guess the
best initial buffer size.
See gh-29775
Prior to this commit, `DispatcherServlet` would completely reset the
response (status, headers and body) before handling errors within Spring
MVC. This can cause unintended consequences when Servlet Filters added
response headers before the error happened. Such response headers might
be still required in case of error handling.
This commit changes the complete reset of the response to only resetting
the response buffer, if possible.
Closes gh-31154
See gh-31104
Prior to this commit, the `RestTemplateAdapter` would manually expand
templated URIs. This means that the `RestTemplate` instance itself would
never see the templated URIs and could not record it in the client
observations at runtime.
This commit ensures that when URI templates are available, the adapter
uses the correct `exchange` method variant.
Fixes gh-31144
This is a follow-up change related to gh-31104.
This change reverts the changes previously made in
`ExceptionHandlerExceptionResolver` and instead attempts to reset the
response directly in `DispatcherServlet` in order to cover all types or
exception handling.
Unlike the previous change, we decided to continue even if the response
was already committed: exception handlers will have a chance to be
called, even if it means they'll have to operate on a garbled response.
This change will cause less disruption, in case existing exception
handlers were relying on this behavior.
See gh-31104
Prior to this commit, the Spring Framework build would publish several
zip artifacts:
* a "*-schema.zip" containing all the XSD schemas produced
* a "*-docs.zip" containing the API docs
* a "*-dist.zip" containing all of the above, plus module jars
Since the reference docs are now produced by Antora in a separate
process, the "*-docs.zip" does not contain the reference docs anymore.
But it still contains the API docs which are automatically fetched from
the artifact repository and published on the docs.spring.io website.
This commit intends to update the current arrangement and optimize the
build.
First, the "*-dist.zip" is not published anymore, since it cannot be
consumed anyway by the community: repo.spring.io does not distribute
release artifacts publicly, developers are expected to get them from
Maven Central. This arrangement is quite dated anyway and is not really
useful in current application build setups.
The generation of API docs is moved to a new "framework-api" module,
separating it from the reference docs module ("framework-docs") which
contains Java, Kotlin and Asciidoctor sources. This removes the custom
javadoc aggregation task and instead uses a dedicated Gradle plugin.
This change also adds a new `-PskipDocs` Gradle project property that
skips entirely the documentation tasks (javadoc, kdocs) as well as the
"distrbution" tasks managed in the framework-api module.
This allows developers to publish locally a SNAPSHOT of Spring Framework
without creating the entire documentation distribution. This is
particularly useful for local testing.
For example, `$ ./gradlew pTML -PskipDocs`.
Closes gh-31049
Prior to this commit, `@Async` and `@EventListener` annotated methods
would lose the the logging and observation contexts whenever their
execution was scheduled on a different Thread.
The Context Propagation library supports this use case and can propagate
context values in ThreadLocals, Reactor Context and more.
This commit introduces a new `TaskDecorator` implementation that
leverages the Context Propagation library. When configured on a
`TaskExecutor`, this allows to properly propagate context value through
the execution of the task.
This implementation is completely optional and requires the
"io.micrometer:context-propagation" library on the classpath. Enabling
this feature must be done consciously and sometimes selectively, as
context propagation introduces some overhead.
Closes gh-31130
This commit instruments the new `RestClient` HTTP client for
observability. Since this client is sharing its HTTP infrastructure with
`RestTemplate` and operates on the same request/response types, this
instrumentation reuses the Observation convention and context.
This choice makes sense since one can build a new `RestClient` instance
using a `RestTemplate` instance, effectively reusing the underlying
configuration.
Closes gh-31114
This commit ensures that the Observability instrumentation for
`@JmsListener` does not assume that micrometer-core is on the classpath
and guards against direct usage.
Closes gh-30335
This commit adds observability support for Jakarta JMS support in
spring-jms support. This feature leverages the `JmsInstrumentation`
infrastructure in `io.micrometer:micrometer-core` library.
This instruments the `JmsTemplate` and the `@JmsListener` support to
record observations:
* "jms.message.publish" when the `JmsTemplate` sends a message
* "jms.message.process" when a message is processed by a `@JmsListener`
annotated method
The observation `Convention` and `Context` implementations are shipped
with "micrometer-core".
Closes gh-30335
This commit changes the `synchronized` usage into a `ReentrantLock`, in
order to guard write operations with a construct that does not pin
virtual threads to the current platform thread on JDK21.
Closes gh-30996
Prior to this commit, the `SseEventBuilder` would be used to create SSE
events and write them to the connection using the `ResponseBodyEmitter`.
This would send each data item one by one, effectively writing and
flushing to the network for each. Since multiple data lines are prepared
by the `SseEventBuilder`, a typical write of an SSE event performs
multiple flushes operations.
This commit adds a method on `ResponseBodyEmitter` to perform batch
writes (given a `Set<DataWithMediaType>`) and only flush once all
elements of the set have been written.
This also applies in case of early writes, where now all buffered
elements are written then flushed altogether.
Fixes gh-30912
This commit documents the fact that the Servlet Filter based
observations for MVC applications is limited by the Servlet Filter
contract in the first place. All processing and logging that happens
outside of the scope of the filter is not observed.
Log statements from the catalina engine (in the case of Tomcat), or any
container-specific infrastructure, is not covered by the
instrumentation.
Closes gh-29398
This commit improves the documentation for the
`ShallowEtagHeaderFilter`, stating that it is only meant to support a
subset of conditional HTTP requests: GET requests with "If-None-Match"
headers. Other headers and state changing HTTP methods are not supported
here, as the filter only operates on the content of the response and has
no knowledge of the resource being served.
Closes gh-30517
Brian Clozel