Commit 5b1165 was an attempt to leave semicolon content in the URL path
while ignoring it for request mapping purposes. However, it becomes
quite difficult to manage and semicolon content should not always be
ignored (sometimes a semicolon is used as a separator of multiple items
in a path segment, rather than for matrix variables).
This change effectively reverts back to the original approach in 3.2
where a flag on AbstractHandlerMapping can be used to have semicolon
content removed or kept. If kept, path segments with matrix variables
must be represented with a path segment.
The main difference is that by default it is removed everywhere
including the MVC namespace and Java config.
Issue: SPR-10427, SPR-10234
Previously DeferredResult locked around the setting of the result
including handling up to the part of submitting a dispatch. This
can cause a deadlock if a timeout occurs at the same time since
the Tomcat timeout thread has its own locking that permits only
one thread to do timeout or dispatch processing.
The fix reduces the locking to cover only the attempt to set the
DeferredResult but not the dispatching.
Issue: SPR-10485
This change should be the final piece in the puzzle to let SmartInstantiationAwareBeanPostProcessor's predictBeanType predict a FactoryBean-produced type, effectively as a semantic alternative to its postProcessBeforeInstantiation-related role.
Issue: SPR-10517
(cherry picked from commit bc3e585)
Spring 3.2.2 introduced a change to avoid closing the response stream
in HttpMessageConverters (SPR-10095). However, the InputStream of
resources being written, for example as part of a multi-part request
should be closed. This change ensures that.
Issue: SPR-10460
If type conversion turns an empty request parameter value (i.e. "") to
null, we should treat it as a missing value. By default the
ConversionService doesn't change empty strings and therefore one must
explicitly convert them to null for example by registering a
StringTrimmerEditor with emptyAsNull=true.
Issue: SPR-10402
A recent change in FormTag to prepend the context and servlet paths if
not present, causes issues when used in portlet applications.
This change introduces a servletRelativeAction form tag attribute that
must be used for the context and servlet paths to be prepended.
Issue: SPR-10382
Previously ContentNegotiationManager continued with the next
ContentNegotiationStrategy only if the current one returned an empty
list. Now it also does that if the current ContentNegotiationStrategy
returns "*/*". Since the absence of an Accept header and "*/*" have
the same meaning, this allows a default content type to be used in
either case.
Issue: SPR-10513
@InitBinder and @ModelAttribute methods in @ControllerAdvice classes
are now invoked first, allowing any such methods in the @Controller
class to override them.
Issue: SPR-10419
HibernateJpaDialect's HibernateConnectionHandle does not call close() for Hibernate 4.x anymore, since on 4.2, the exposed Connection handle isn't a "borrowed connection" wrapper but rather the actual underlying Connection - with a close() call immediately returning the Connection to the pool and making the local handle invalid for further use within the transaction.
Issue: SPR-10395
Use LinkedHashMaps/Sets wherever exposed to users, and code tests defensively in terms of expected Map/Set ordering. Otherwise, there'll be runtime order differences between JDK 7 and JDK 8 due to internal HashMap/Set implementation differences.
Issue: SPR-9639
(cherry picked from commit 9c09a0a)
In particular, avoid accidental usage of ASM for core JDK types - which will fail in case of a new bytecode version in the JDK, even if the application itself has been compiled with an earlier bytecode target.
Issue: SPR-10292
(cherry picked from commit d3a4068)
