This commit improves the JMS health indicator to identify a broken broker
that uses failover. An attempt to start the connection is a good way to
make sure that it is effectively available.
Closes gh-6818
This commit adds Spring Batch configuration validation that disables
database initializer in case custom table prefix is configured with
default schema.
See gh-6651
This commit adds Spring Session JDBC configuration validation that
disables database initializer in case custom table name is configured
with default schema.
See gh-6649
Update our `ResourceHttpRequestHandler` subclass so that the following
warning is no longer displayed:
"Locations list is empty. No resources will be served unless a
custom ResourceResolver is configured as an alternative to
PathResourceResolver."
Fixes gh-6791
This commit adds a new `spring.thymeleaf.check-template` property which
is only used for Thymeleaf 3.0+.
Since thymeleaf/thymeleaf#419, the Thymeleaf template resolver
implementations can implement the `setCheckExistence` method - this
enables the template existence verification at **resolution** time,
which means the resolver can return null as a `TemplateResolution`
and let other template resolvers in the chain try.
This new property is set to `true` by default and can be disabled if the
application only has a single resolver and the template existence check
is considered as a performance penalty with the configured resolver.
Fixes gh-6500
This commit adds the `spring.http.multipart.resolve-lazily`
property, in order to configure whether the multipart request
should be resolved lazily, as the parameters are accessed.
When extending `WebMvcConfigurerAdapter`, developers should always
declare those as `@Configuration` classes rather than `@Bean`s, which
can lead to dependency issues as described in #6853.
Fixes gh-6853
Add a `FailureAnalyzer` that handles the case where the context does
not start because no candidate bean was found for an `InjectionPoint`.
The implementation inspects the auto-configuration report for beans
that are candidate and output the condition(s) that lead to such beans
to be discarded on startup. If a whole auto-configuration class is
disabled (or excluded), its beans are inspected and candidates are
extracted in a similar way.
This works for both injection by type and by name.
Closes gh-6612
This works, and feels like the right thing to do, since there is no
guarantee that extensions won't in turn use ServiceLoader for things
that we haven't yet anticipated.
Fixes gh-6829. Cc @wilkinsona in case he has an opinion.
Post-processing of mocked beans causes a number of problems:
- The mock may be proxied for asynchronous processing which can cause
problems when configuring expectations on a mock (gh-6573)
- The mock may be proxied so that its return values can be cached or
so that its methods can be transactional. This causes problems with
verification of the expected calls to a mock (gh-6573, gh-5837)
- If the mock is created from a class that uses field injection, the
container will attempt to inject values into its fields. This causes
problems if the mock is being created to avoid the use of one of
those dependencies (gh-6663)
- Proxying a mocked bean can lead to a JDK proxy being created
(if proxyTargetClass=false) as the mock implements a Mockito
interface. This can then cause injection failures as the types don’t
match (gh-6405, gh-6665)
All of these problems can be avoided if a mocked bean is not
post-processed. Avoiding post-processing prevents proxies from being
created and autowiring from being performed. This commit avoids
post-processing by registering mocked beans as singletons as well as
via a bean definition. The latter is still used by the context for type
matching purposes.
Closes gh-6573, gh-6663, gh-6664
This commit adds `JdbcTemplateAutoConfiguration` to the list of auto-
configurations that are applied with `DataJpaTest`. This effectively
allows to inject a `JdbcTemplate` in any `@DataJpaTest` test.
Closes gh-6802