The fromUri method of UriComponentsBuilder used uri.getXxx() methods,
which decode the URI parts causing URI parsing issues. The same method
now uses uri.getRawXxx().
Issue: SPR-9317
Prior to this change, request-scoped components having
@Resource-injected dependencies caused a memory leak in
DefaultListableBeanFactory#dependenciesForBeanMap.
Consider the following example:
@Component
@Scope(value="request", proxyMode=ScopedProxyMode.TARGET_CLASS)
public class MyComponent {
@Resource
private HttpServletRequest request;
// ...
}
The bean name for "MyComponent" will end up being
'scopedTarget.myComponent', which will become a key in
the #dependenciesForBeanMap structure.
On the first request, the injected HttpServletRequest bean will be a
proxy and will internally have a bean name of the form
"$Proxy10@1a3a2a52". This name will be added to the Set value associated
with the 'scopedTarget.myComponent' entry in #dependenciesForBeanMap.
On the second request, the process will repeat, but the injected
HttpServletRequest will be a different proxy instance, thus having a
different identity hex string, e.g. "$Proxy10@5eba06ff". This name will
also be added to the Set value associated with the
'scopedTarget.myComponent' entry in #dependenciesForBeanMap, and this
is the source of the leak: a new entry is added to the set on each
request but should be added only once.
This commit fixes the leak by introducing caching to
CommonAnnotationBeanPostProcessor#ResourceElement similar to that already
present in AutowiredAnnotationBeanPostProcessor#AutowiredFieldElement
and #AutowiredMethodElement. Essentially, each ResourceElement instance
now tracks whether it has been created, caches the ultimate value to be
injected and returns it eagerly if necessary. Besides solving the memory
leak, this has the side effect of avoiding unnecessary proxy creation.
This fix also explains clearly why injection into request-scoped
components using @Autowired never suffered this memory leak: because the
correct caching was already in place. Because @Resource is considerably
less-frequently used than @Autowired, and given that this particular
injection arrangement is relatively infrequent, it becomes
understandable how this bug has been present without being reported
since the introduction of @Resource support in Spring 2.5: developers
were unlikely to encounter it in the first place; and if they did, the
leak was minor enough (adding strings to a Set), that it could
potentially go unnoticed indefinitely depending on request volumes and
available memory.
Issue: SPR-9176
Prior to this commit, SpEL's OpPlus ('+' operator) would convert its
left and right operands to String directly via #toString calls.
This change ensures that operand values are delegated to any registered
converters, allowing for customization of the stringified output.
Note that the OpPlus constructor now throws IllegalArgumentException if
zero operands are supplied.
Issue: SPR-8308
Before this change there were numerous javadoc warnings being reported
while building Spring framework API.
This commit resolves most of the javadoc warnings, reducing the total
number from 265 to 103.
Issue: SPR-9113
Before this change javadoc in two classes had non-UTF-8 encoded
characters. This caused building Spring API to fail in Java 1.7.
Commit fixes this by replacing wrongly encoded characters with their
UTF-8 equivalents.
Issue: SPR-9097
When a controller returns a DeferredResult, the underlying async
request will eventually time out. Until now the default behavior was
to send a 503 (SERVICE_UNAVAILABLE). However, this is not desirable
in all cases. For example if waiting on an event, a timeout simply
means there is no new information to send.
To handle those cases a DeferredResult now accespts a timeout result
Object in its constructor. If the timeout occurs before the
DeferredResult is set, the timeout result provided to the constructor
is used instead.
Issue: SPR-8617
* Clarify semantics and behavior of AsyncWebRequest methods in most cases
making a best effort and not raising an exception if async processing
has completed for example due to a timeout. The startAsync() method is
still protected with various checks and will raise ISE under a number
of conditions.
* Return 503 (service unavailable) when requests time out.
* Logging improvements.
Issue: SPR-8517
From a programming model perspective, @RequestMapping methods now
support two new return value types:
* java.util.concurrent.Callable - used by Spring MVC to obtain the
return value asynchronously in a separate thread managed transparently
by Spring MVC on behalf of the application.
* org.springframework.web.context.request.async.DeferredResult - used
by the application to produce the return value asynchronously in a
separate thread of its own choosing.
The high-level idea is that whatever value a controller normally
returns, it can now provide it asynchronously, through a Callable or
through a DeferredResult, with all remaining processing --
@ResponseBody, view resolution, etc, working just the same but
completed outside the main request thread.
From an SPI perspective, there are several new types:
* AsyncExecutionChain - the central class for managing async request
processing through a sequence of Callable instances each representing
work required to complete request processing asynchronously.
* AsyncWebRequest - provides methods for starting, completing, and
configuring async request processing.
* StandardServletAsyncWebRequest - Servlet 3 based implementation.
* AsyncExecutionChainRunnable - the Runnable used for async request
execution.
All spring-web and spring-webmvc Filter implementations have been
updated to participate in async request execution.
The open-session-in-view Filter and interceptors implementations in
spring-orm will be updated in a separate pull request.
Issue: SPR-8517
Previously depending on 1.5.10 in certain locations, which caused
NoClassDefFoundErrors in EhCacheCacheTests and other tests due to
mismatches between slf4j -api and -log4j12 versions.
With the exception of one transitive dependency via Hibernate 3.3.1.GA,
all slf4j dependencies are now pegged at 1.6.1 (and all tests pass).
Before this change, IDE settings generated via import-into-eclipse.sh
created a classpath dependency on slf4j-api version 1.6.1 and
slf4j-log4j12 version 1.5.10. As a result running tests inside the IDE
resulted in a NoSuchMethodException.
build.gradle sets the variable slf4jLog4jVersion to '1.5.10'. However,
the hibernate-validator dependency in spring-context pulls in
slf4j-api version 1.6.1. The change ensures the version specified in
the build script variable is used consistently. Whether it should be
1.5.10 or 1.6.1 is a separate concern.
- Spring remains compatible against AJ version 1.6.8, but is now
compiling and testing against 1.6.12
- Encountered what appears to be an AJ bug introduced in 1.6.10: an
assertion in org.aspectj.weaver.UnresolvedType causes a false
negative failure when encountering org.springframework.io.Resource
arrays, e.g. "[org.springframework.io.Resource@xxx". This problem
has been reported to the AJ team and in the meantime, the recommended
workaround is to disable assertions either completely, or at least
selectively with
-disableassertions:org.aspectj.weaver.UnresolvedType
Issue: SPR-7989, SPR-9272
- Fix deprecation warnings about dynamic properties; favor use of
"extra properties extensions" per [1]
- Certain such deprecations are still issued due to violations within
the docbook-reference plugin. These will be fixed in an upcoming
revision of the plugin, at which point spring-framework will upgrade
to depend on it and these warnings will be eliminated
[1] http://gradle.org/docs/current/dsl/org.gradle.api.plugins.ExtraPropertiesExtension.html
Issue: SPR-9327
The new @MVC support classes select a HandlerMethodArgumentResolver
and a HandlerMethodReturnValueHandler statically, i.e. based on
the signature of the method, which means that a controller method
can't declare a more general return type like Object but actually
return a more specific one, e.g. String vs RedirectView, and
expect the right handler to be used.
The fix ensures that a HandlerMethodReturnValueHandler is selected
based on the actual return value type, which is something that was
supported with the old @MVC support classes. One consequence
of the change is the selected HandlerMethodReturnValueHandler can
no longer be cached but that matches the behavior of the old
@MVC support classes.
Issues: SPR-9218
Previously 'Content-Disposition' was passed to Part.getHeader(String).
However the Javadoc for that method specifies the header should be
case insensitive. Note that the JavaDoc in tomcat-servlet-api doesn't
mention this. It can only be found in the official api JavaDoc:
http://download.oracle.com/otndocs/jcp/servlet-3.0-fr-oth-JSpec/
Issue: SPR-9149
Invalid Content-Type or Accept header values previously resulted in the
IllegalArgumentException getting propagated. After this change such
errors are detected and generally treated as a "no match", which
may for example result in a 406 in the case of the Accept header.
Issue: SPR-9148
A custom RequestCondition which can be provided by overriding methods
in RequestMappingHandlerMapping worked only for conditions that match
and did not return null (as it should have) for conditions that don't
match.
Issues: SPR-9134
Copy spring-*-3.1.xsd => spring-*-3.2.xsd; this commit introduces no
substantive changes, but rather prepares for them by creating a clean
baseline. All internal references to 3.1 schemas (e.g. spring-tool) have
also been updated.
Before this change JibxMarshaller did not use the configured encoding
when unmarshalling XML. This caused issues when content being
unmarshalled was not encoded using the default encoding.
This commit fixes the issue by passing configured encoding to JiBX so
it gets used when unmarshalling instead of the default encoding.
Issue: SPR-7865
Use the preferred SessionFactoryImplementor#getDialect call as
opposed to the previous Dialect#getDialect approach which required
explicitly setting the 'hibernate.dialect' property.
Issue: SPR-7396
This change fixes a mistake made during an earlier sync with the 3.1.x
branch in which the new QueryTimeoutException class was accidentally
placed in spring-transaction instead of spring-tx.
Issue: SPR-7680
Before this change if FreeMarker Spring form macro was bound to a path
which contains square brackets, those brackets would also appear in id
of generated tag, making the id invalid.
As of this fix all FreeMarker Spring form macros generate tag with id
that does not contain square brackets.
Issue: SPR-8732
Before this fix MultipartResolver javadoc mentioned that there is only
one concrete implementation included in Spring. This was true as of
Spring 2.5 but Spring 3.1 added another one, Servlet 3.0 Part API based
implementation.
This fix changes MultipartResolver javadoc so that the other one,
StandardServletMultipartResolver implementation, gets also listed.
Changed javadoc mentions also as of which version does Spring bundle
two MultipartResolver implementations.
Issue: SPR-5984
Before this change bean scopes chapter in Spring reference documentation
had a note which mentioned that creating a scoped proxy for singleton-
or prototype-scoped beans will throw BeanCreationException.
This is no longer the case and this change removes the mentioned note.
Issue: SPR-7940
Prior to this commit, MutablePropertySources#get(String) would throw
IndexArrayOutOfBoundsException if the named property source does not
actually exist. This is a violation of the PropertySource#get contract
as described in its Javadoc.
The implementation now correctly checks for the existence of the named
property source, returning null if non-existent and otherwise returning
the associated PropertySource.
Other changes
- Rename PropertySourcesTests => MutablePropertySourcesTests
- Polish MutablePropertySourcesTests for style, formatting
- Refactor MutablePropertySources for consistency
Issue: SPR-9179
Rossen Stoyanchev
Chris Beams
Dridi Boukelmoune
Ivo Smid
Stevo Slavic
Arjen Poutsma
Mike Krumlauf
Phani Kandula