Commit 84714fbae9 introduced usage of the
-Djava.locale.providers=COMPAT command-line argument for javac in order
to allow our JDK 20 builds to pass by using legacy locale data.
That was done to ensure that Date/Time formats using AM/PM produced a
standard space (" ") before the "AM" or "PM" instead of a narrow
non-breaking space (NNBSP "\u202F"), which was introduced in Java 20
due to adoption of Unicode Common Locale Data Repository (CLDR-14032).
This commit removes usage of the -Djava.locale.providers=COMPAT
command-line argument and updates all affected tests to:
- Use an NNBSP before "AM" or "PM" in input text when running on Java 20
or higher.
- Leniently match against any Unicode space character in formatted
values containing "AM" or "PM".
See https://jdk.java.net/20/release-notes#JDK-8284840
See https://unicode-org.atlassian.net/browse/CLDR-14032
See gh-30185
Closes gh-33144
Prior to this commit, the fix for gh-32730 disabled the involvment of
the osbervation filter for async dispatches. Instead of relying on ASYNC
dispatches to close the observation for async requests, this is now
using an async listener instead: async dispatches are not guaranteed to
happen once the async request is handled.
This change caused another side-effect: because async dispatches are not
considered anymore by this filter, the observation scope is not
reinstated for async dispatches. For example, `ResponseBodyAdvice`
implementations do not have the observation scope opened during their
execution.
This commit re-enables async dispatches for this filter, but ensures
that observations are not closed during such dispatches as this will be
done by the async listener.
Fixes gh-33091
Prior to this commit, the fix for gh-32575 introduced cases where the
client observation would be stopped twice.
This commit ensures that `RestClient` observations are stopped only once
when the response is closed, or before throwing an unhanlded exception.
Fixes gh-33068
This commit changes the use of HttpSessionRequiredException in
ModelFactory::initModel to an IllegalStateException, because the former
extends ServletException and cannot be used in WebFlux.
Closes gh-33043
Prior to this commit, the `ContentCachingRequestWrapper` could allocate
a `FastByteArrayOutputStream` block that was larger than the content
cache limit given as a consturctor argument. This was due to an
optimization applied in gh-31834 for allocating the right content cache
size when the request size is known.
Fixes gh-32987
Prior to this commit, the `ServerHttpObservationFilter` would support
async dispatches and would do the following:
1. start the observation
2. call the filter chain
3. if async has started, do nothing
4. if not in async mode, stop the observation
This behavior would effectively rely on Async implementations to
complete and dispatch the request back to the container for an async
dispatch. This is what Spring web frameworks do and guarantee.
Some implementations complete the async request but do not dispatch
back; as a result, observations could leak as they are never stopped.
This commit changes the support of async requests. The filter now
opts-out of async dispatches - the filter will not be called for those
anymore. Instead, if the application started async mode during the
initial container dispatch, the filter will register an AsyncListener to
be notified of the outcome of the async handling.
Fixes gh-32730
Commit d7970e4ab8 introduced support for JAXBElement in
Jaxb2XmlEncoder's encodeValue() method; however, canEncode() still
returned false for a JAXBElement element type.
This commit revises canEncode() so that it returns true for an element
type that is assignable from JAXBElement.
See gh-30552
See gh-32972
Closes gh-32977
Prior to this commit, #31870 added support for constraint annotations on
container elements for handler method argument validation. Supporting
this use case:
```
public void addNames(List<@NotEmpty String> names)
```
This commit does the same for `@Valid` annotation:
```
public void addPeople(List<@Valid Person> people)
```
Fixes gh-32964
Prior to this commit, HTTP requests sent with the
`HttpComponentsClientHttpRequestFactory` would not set a
"Content-Length" header for empty request bodies. Setting a request
entity is the expected behavior for unsafe HTTP methods, and this would
align the behavior with other HTTP clients.
Developers would often rely on `BufferingClientHttpRequestFactory` to
set this information on the request.
This commit ensures that a `NullEntity` is used for unsafe HTTP methods,
when no body has been set for the request. This result in a
"Content-Length:0" request header.
Fixes gh-32678
Prior to this commit, `RestClientException` thrown by status handlers
would not be registered as observation errors. This commit ensures that
such exceptions are first caught, registered in the observation and
rethrown as expected.
Closes gh-32575
Prior to this commit, the `RestClient` observations would be stopped as
soon as the exchange function was called. This means that all errors
related to response decoding or mapping would not be recorded by the
obsevations.
This commit extends the observation recording to the `ResponseSpec` DSL
calls as well as custom exchange functions.
Fixes gh-32575
Prior to this commit, the `BufferingClientHttpRequestFactory`, through
the `AbstractBufferingClientHttpRequest`, would set a "Content-Length"
header value, even if the buffered body was empty.
This behavior is invalid since no request body would be set by the
client code in the first place.
This commit ensures that this header is only set if a request body has
been buffered and is about to be written to the request.
Fixes gh-32650
Prior to this commit, `BufferingClientHttpRequestWrapper` would always
write to the actual client request body, even if the buffered content
was empty (empty byte array).
This would cause issues with specific client request factories,
especially the OkHttp variant, that would consider empty byte arrays as
non-empty body and would reject such cases for GET requests with an
"IllegalArgumentException: method GET must not have a request body".
This commit only writes to the request if the buffered content is not
empty.
Fixes gh-32612
The fix for #31254 resulted in an InvalidMimeTypeException being thrown
by MimeTypeUtils.sortBySpecificity() instead of an
IllegalArgumentException. However, InvalidMimeTypeException extends
IllegalArgumentException. Consequently, the change from
IllegalArgumentException to InvalidMimeTypeException did not result in
the desired effect in HeaderContentNegotiationStrategy.
HeaderContentNegotiationStrategy.resolveMediaTypes() still allows the
InvalidMimeTypeException to propagate as-is without wrapping it in an
HttpMediaTypeNotAcceptableException.
To address this issue, this commit catches InvalidMediaTypeException
and InvalidMimeTypeException in HeaderContentNegotiationStrategy and
wraps the exception in an HttpMediaTypeNotAcceptableException.
See gh-31254
See gh-31769
Closes gh-32483
The following adjustments are also made as a result:
- Use int to check if lock is held and unlock is needed, given that
for non-async requests we don't need to obtain a lock.
- Protect access methods getOutputStream and getWriter with the
same locking and state checks.
Closes gh-32340
ServletResponseHttpHeaders#get should be annotated with `@Nullable` and
return null instead of a singleton list containing null when there is no
content type header.
Closes gh-32362
The wrapped response prevents use after AsyncListener onError or completion
to ensure compliance with Servlet Spec 2.3.3.4.
The wrapped response is applied in RequestMappingHandlerAdapter.
The wrapped response raises AsyncRequestNotUsableException that is now
handled in DefaultHandlerExceptionResolver.
See gh-32340
The HttpHeaders.getAcceptLanguageAsLocales was incurring overhead from
using a Stream, as well as calling the fairly expensive
Locale.getDisplayName method.
Switch to using an ArrayList, and skipping over wildcard ranges to avoid
needing to check the display name.
Closes gh-32318
Based on feedback from several members of the community, we have
decided to revert the caching of the Content-Type header that was
introduced in ContentCachingResponseWrapper in 375e0e6827.
This commit therefore completely removes Content-Type caching in
ContentCachingResponseWrapper and updates the existing tests
accordingly.
To provide guards against future regressions in this area, this commit
also introduces explicit tests for the 6 ways to set the content length
in ContentCachingResponseWrapper and modifies a test in
ShallowEtagHeaderFilterTests to ensure that a Content-Type header set
directly on ContentCachingResponseWrapper is propagated to the
underlying response even if content caching is disabled for the
ShallowEtagHeaderFilter.
See gh-32039
Closes gh-32317
Prior to this commit, getHeaderNames() returned duplicates for the
Content-Type and Content-Length headers if they were set in the wrapped
response as well as in ContentCachingResponseWrapper.
This commit fixes that by returning a unique set from getHeaderNames().
In addition, this commit introduces a new test in
ContentCachingResponseWrapperTests to verify the expected behavior for
Content-Type and Content-Length headers that are set in the wrapped
response as well as in ContentCachingResponseWrapper.
See gh-32039
See gh-32317
Commit 375e0e6827 introduced a regression in
ContentCachingResponseWrapper (CCRW). Specifically, CCRW no longer
honors Content-Type and Content-Length headers that have been set on
the wrapped response and returns null for those header values if they
have not been set directly on the CCRW.
This commit fixes this regression as follows.
- The Content-Type and Content-Length headers set in the wrapped
response are honored in getContentType(), containsHeader(),
getHeader(), and getHeaders() unless those headers have been set
directly on the CCRW.
- In copyBodyToResponse(), the Content-Type in the wrapped response is
only overridden if the Content-Type has been set directly on the CCRW.
See gh-32039
Closes gh-32317
This commit ensures that, when creating a RestClient.Builder from a
RestTemplate, the UriBuilderFactory is only copied if it has been
changed from the default values.
Before this commit, the UriBuilderFactory was effectively alway copied,
resulting in not being able to use a baseUrl.
This commit also introduces a small memory optimization in
DefaultUriBuilderFactory, so that default environment variables are
created lazily.
Closes gh-32180
To improve consistency and avoid confusion regarding primitive types
and their wrapper types, this commit ensures that we always use class
literals for primitive types.
For example, instead of using the `Void.TYPE` constant, we now
consistently use `void.class`.
