Previous incarnation of MockMvc Kotlin DSL tried to reuse directly
Java APIs like ModelResultMatchers or StatusResultMatchers, but
when using multiple matchers in DSL blocks like model { } or
status { }, only the last statement was taken in account which
was very confusing.
This refactoring provides dedicated Kotlin DSLs for matchers.
The main API breaking changes is that functions like isOk() need to be
invoked with the parenthesis, isOk is not supported anymore (on purpose).
Closes gh-24103
Prior to this commit, if a developer accidentally copied and pasted the
same @ContextConfiguration or @TestPropertySource declaration from a
test class to one of its subclasses or nested test classes, the Spring
TestContext Framework (TCF) would merge the inherited configuration
with the local configuration, resulting in different sets of
configuration metadata which in turn resulted in a different
ApplicationContext instance being loaded for the test classes. This
behavior led to unnecessary creation of identical application contexts
in the context cache for the TCF stored under different keys.
This commit ignores duplicate configuration metadata when generating
the ApplicationContext cache key (i.e., MergedContextConfiguration) in
the TCF. This is performed for the following annotations.
- @ContextConfiguration
- @ActiveProfiles (support already existed prior to this commit)
- @TestPropertySource
Specifically, if @ContextConfiguration or @TestPropertySource is
declared on a test class and its subclass or nested test class with the
exact same attributes, only one instance of the annotation will be used
to generate the cache key for the resulting ApplicationContext. The
exception to this rule is an "empty" annotation declaration. An empty
@ContextConfiguration or @TestPropertySource declaration signals that
Spring (or a third-party SmartContextLoader) should detect default
configuration specific to the annotated class. Thus, multiple empty
@ContextConfiguration or @TestPropertySource declarations within a test
class hierarchy are not considered to be duplicate configuration and
are therefore not ignored.
Since @TestPropertySource is a @Repeatable annotation, the same
duplicate configuration detection logic is applied for multiple
@TestPropertySource declarations on a single test class or test
interface.
In addition, this commit reinstates validation of the rules for
repeated @TestPropertySource annotations that was removed when support
for @NestedTestConfiguration was introduced.
Closes gh-25800
The migration from JUnit 4 assertions to AssertJ assertions resulted in
several unnecessary casts from int to long that actually cause
assertions to pass when they should otherwise fail.
This commit fixes all such bugs for the pattern `.isNotEqualTo((long)`.
Prior to this commit, two @ActiveProfiles declarations with the same
profiles but different order resulted in an identical, duplicate
ApplicationContext in the context cache in the Spring TestContext
Framework.
This commit uses a TreeSet to ensure that registered active profiles
are both unique and sorted, thereby avoiding cache misses for
semantically identical active profiles configuration on different test
classes.
Closes gh-25973
This commit makes TransactionInterceptor and TransactionAspectSupport
Coroutines aware, adapting Reactive transaction support to Coroutines.
Suspending functions returning a Flow are handled like Flux, for other
return types, they are handled like Mono.
Closes gh-23575
This commit introduces TestContextAnnotationUtils as a replacement for
MetaAnnotationUtils, with dedicated support for honoring the new
@NestedTestConfiguration annotation and related annotation search
semantics.
MetaAnnotationUtils has been reverted to its previous scope and is now
deprecated.
See gh-19930
The existing implementation was exposed to very poor performance when matching
with multiple delimiters against a large buffer with many delimiters. In that
case all matchers are invoked many times (as many as the number of delimiters)
even though some of them found no match at all on the first pass.
The revised implementation uses a single index and advances all matchers
together, checking one byte a time, and not letting any one of them search to
the end of the entire buffer on a single pass.
Closes gh-25915
Sébastien Deleuze
Juergen Hoeller
Sam Brannen
Stephane Nicoll
Rossen Stoyanchev
Riley Park
Robin Roos
Arjen Poutsma