Previously, BeanTypeRegistry did not correctly determine the type
that would be created by a factory bean if that factory bean was
returned from a bean method with arguments on a configuration class
found via component scanning.
The key difference is that bean definitions for bean methods on
configuration classes found via component scanning use ASM-based
metadata rather than reflection-based metadata. The ASM-based method
data does not provide direct access to the Method that will create the
bean. In this case, BeanTypeRegistry was falling back to looking for
a method with the matching name and no arguments. Therefore, if
the bean method had any arguments it would fail to find the method
and would, therefore, be unable to determine the type of bean
produced by the factory bean.
This commit updates BeanTypeRegistry to use logic that is very similar
to Spring Framework's ConstructorResolver's
resolveFactoryMethodIfPossible method to locate the method that will
produce the factory bean. It looks for a single method with
the required name with any number of arguments. If it finds multiple
methods with the required name and different arguments it returns
null, just as ConstructorResolver does.
Closes gh-6755
Due to the layout format change in 1.4, Spring Framework is no longer
able to compute a default persistence unit root URL. If a Spring Boot 1.4
application has JPA but does not have any entity, the application started
from a fat jar now fails with a quite cryptic exception.
This commit introduces `ApplicationInfo` as a general replacement for
the `ApplicationArguments` and `Banner` singleton beans that
`SpringApplication` registers on startup. `ApplicationInfo` also defines
the detected "main" `Class` that can be used to compute a last resort
URL that makes sense.
If such bean is available, `EntityManagerFactoryBuilder` now sets the
default persistence unit root location, preventing Spring Framework to
attempt to resolve an unknown location. Note that in our case the
persistence unit root location is actually useless: given the way the
persistence unit is created, nothing actually uses it but Hibernate, as a
compliant JPA provider, has to make sure this setting is set to a valid
URL nevertheless.
Closes gh-6635
Hikari can configure both an underlying Driver or DataSource. Spring Boot
only supports the former so any attempt to trigger the creation of the
latter will lead to an exception since the auto-configuration has set
driver-specific field.
This commit adds an analyzer that intercepts such faulty scenario and
indicate that users should configure Hikari with an underlying DataSource
themselves.
Closes gh-6568
This commit changes `FailureAnalyzers` so that the loaded
`FailureAnalyzer` instances are initialized before the context eventually
fails.
Before this commit, the `BeanFactoryAware` callback was processed at a
time where the context has already shutdown. Some implementations need
to access components that are no longer available.
Instead, we now initialize them before the context is refreshed.
Closes gh-6748
This commit fixes the prefix for the WebClient and WebDriver auto-config
so that it complies with the prefix set on `AutoconfigureWebMvc`
Closes gh-6727
Previously, Spring Data Cassandra excluded Netty from the Cassandra
Driver's dependencies. This was an error that we replicated in
our dependency management. The exclusion has been removed from
Spring Data Cassandra's pom. This commit makes the equivalent update
to our dependency management.
Fixes gh-6616
Closes gh-6721
Previously, configuring the build-info goal in a pom would result in
Eclipse reporting an error for the pom as it didn’t know if/when to
execute the build-info goal.
This commit adds lifecycle mapping metadata so that the goal is executed
on incremental builds. This ensures that the contents of the generated
file are kept up-to-date, reflecting the latest build time, etc.
Closes gh-6723
This commit marks the auto-configured `JdbcTemplate` as `Primary`. Spring
session creates its own so we could end up with two in a fully auto
configured context.
Closes gh-6717
Previously, Tomcat’s background processing was only enabled on the
context but access logging was configured on the engine. This means that
the access log valve’s background processing method was never called
and, therefore, that it wasn’t flushed periodically.
This commit moves the enablement of background processing up to the
engine, thereby ensuring that the access log is flushed periodically.
Background processing cascades down the container hierarchy so, after
this change, background processing will still be performed on the
context as well.
Closes gh-6646
* gh-6706:
Polish “Add constructor to TestRestTemplate that takes a RestTemplateBuilder”
Add constructor to TestRestTemplate that takes a RestTemplateBuilder
We make Liquibase aware of our custom Commons Logging-based logger by
adding its package to the Liquibase ServiceLocator’s packages to scan.
Previously, this was happening too late so Liquibase may have already
initialized and cached a particular logger.
This commit moves the registration of the extra package from the
Liquibase auto-configuration to the application listener that customises
Liquibase’s ServiceLocator. This ensures that the package is added
before Liquibase is used. Unfortunately, configuring Liquibase’s
ServiceLocator and its packages to scan causes it to try to perform
some logging, resulting in it caching the wrong type of logger. We
work around this problem by resetting Liquibase’s LogFactory once we’ve
finished setting everything up.
Closes gh-6713
As of Spring Boot 1.4, IDEs are supposed to provide auto-completion for
hints and enum values. This commit removes the hard-coded copy of
allowed values for enum-based properties. This was mostly the case
already and this commit fixes the few ones remaining.
Closes gh-6654
Previously, when access logging was enabled and the management server
was running on a separate port, both the main server and the management
server would write their access logs to the same file. Having two
separate containers writing to the same file could cause problems such
as causing log rotation to break.
This commit updates the actuator so that when the management server is
running on a separate port (and therefore using a separate container)
it prepends management_ to the access log prefix so that the main
server and the management server write their access logs to separate
files in the same directory.
Closes gh-6618
This commit restores the support of "spring.datasource.type". If this
property is set to an unsupported `DataSource` implementation, we now
properly create it again, rather than ignoring it because its value did
not match any value we support.
Since Spring Boot 1.4 now redirects `DataSource` implementation settings
to dedicated namespaces, the documentation has been updated to explicitly
mention how one can create such arrangement for an unsupported
implementation.
As a convenience, `DataSourceProperties` can initialize a
`DataSourceBuilder` based on its internal state, making it very easy for
anyone to create a `DataSource` from a `DataSourceProperties` instance.
Closes gh-6695
Andy Wilkinson
Stephane Nicoll
Caleb Cushing
Mark Paluch
Johnny Lim
Maciej Walkowiak