Previously, actuator endpoints were registered with Jersey upon
injection of the ResourceConfig bean into a registrar class rather than
using a ResourceConfigCustomizer. This was done to fix a problem
when running the Actuator on a separate port where the main application
context's customizers were also applied to the management context,
breaking the singleton contract for those resources. This approach
meant that the registration could be performed at any point after the
ResourceConfig had been created. When Jersey's configured as a Filter
this resulted in the registration failing as the attempt was being made
after the Filter lifecyle callbacks which make the ResourceConfig
immutable.
This commit reworks the endpoint registration to be performed using a
ManagementContextResourceConfigCustomizer, a resource config customizer
that's only applied to the ResourceConfig that's used by the Actuator.
When there's a separate management context, this ResourceConfig is
created by the Actuator's auto-configuration and the management context
resource config customizers are applied to it during its creation. The
main application's customizers are not applied. When the actuator is
using the same context as the main application, this ResourceConfig is
created by the main application. In this case a
ResourceConfigCustomizer is defined that delegates to all
ManagementContextResourceConfigCustomizers, allowing them to register
the actuator endpoints with the main ResourceConfig.
Fixes gh-25262
Prior to this commit, every test started a new testcontainer. By splitting the
tests apart we can make use of static containers that are only instantiated once
per application (TomEE, Wildfly etc.)
See gh-25446
Previously, DataSource initialization was triggered via a
BeanPostProcessor or a schema created event from JPA. This caused
numerous problems with circular dependencies, bean lifecycle, etc and
added significant complexity.
This commit reworks DataSource initialization to remove the use of a
BeanPostProcessor entirely. In its place, DataSource initialization is
now driven by an InitializingBean with dependency relationships
between beans ensuring that initialization has been performed before
the DataSource is used. This aligns with the approach that's worked
well with Flyway and Liquibase.
More changes are planned to further simplify DataSource initialization.
The changes in this commit are a foundation for those changes. Any new
public API in this commit is highly likely to change before the next
GA.
Fixes gh-13042
Fixes gh-23736
Update `build.gradle` files to ensure that `junit-platform-launcher` is
a `testRuntimeOnly` dependency. This ensures that tests can be run from
Eclipse.
Closes gh-25074
Update jar `Handler` fallback logic to directly support Tomcat
'jar:war:file' URLs. This commit allows contents to be accessed without
the JDK needing to extracted the nested jar to the temporary folder.
Closes gh-24553
Update the jar `Handler` class to support a non-reflective fallback
mechanism when possible. The updated code attempts to capture a regular
jar URL before our handler is installed. It can then use that URL as
context when creating the a fallback URL. The JDK jar `Handler` will
be copied from the context URL to the fallback URL.
Without this commit, resolving new Tomcat URLs of the form
`jar:war:file:...` would result in an ugly "Illegal reflective access"
warning.
Fixes gh-18631
Stephane Nicoll
dreis2211
Andy Wilkinson
Phillip Webb
Madhura Bhave
Hatef Palizgar
izeye