Update the `AbstractJarWriter` so that it can directly build the layer
index as entries are written. Prior to this commit, a layer tracking
was handled by a decorator class which was broken because it didn't
override enough methods. Since `AbstractJarWriter` has quite a complex
API, it seems sensible to have it handle the layer index directly,
removing the need for a decorator entirely.
Fixes gh-23801
Prior to this commit, the how-to documentation would say that Spring
Boot does not support the h2c protocol. While it's not supported
out-of-the-box with a configuration property, this protocol can still be
configured using server customizers.
This commit documents, with code snippets, the server customizers one
should use to configure the h2c protocol in an application - for each
supported server.
Closes gh-21997
Prior to this commit, tests using the `@WebMvcTest` annotation would not
include the `HttpEncodingAutoConfiguration`. This means that, even if
configured, the encoding filter would not be configured in MVC tests,
resulting in an inconsistency with `@SpringBootTest` tests.
This commit ensures that the `HttpEncodingAutoConfiguration` is included
when `@WebMvcTest` is used.
Fixes gh-23749
This commit modifies the logic for finding properties files using
wildcard paths to ignore files if any part of the file path contains
a hidden directory. Hidden directories are common when Kubernetes
mounts config maps onto volumes in a pod, which was causing the same
properties files to be loaded multiple times.
Fixes gh-23160
This commit makes sure to defer registration of hibernate statistics
outside of the singleton lock as it can lead to deadlocks when the
EntityManagerFactory is initialized in deferred mode.
Closes gh-23740
Previously, the shutdown hook was only registered once refresh has
completed. If the JVM was shut down during refresh (or after refresh
and before the hook was registered) the hook wouldn't run and the
partially refreshed context would not be cleaned up.
This commit moves the registration of the shutdown hook to before
refresh processing begins. This ensures that the hook is available
to clean up the context if the JVM is shutdown while refresh is in
progress.
Fixes gh-23625
This commit modifies the documentation generated for the
Maven plugin to include links to javadoc when Spring
Boot types are mentioned. Some javadoc was also polished
to improve the generated docs.
Fixes gh-21555
This commit qualifies examples of configuring the CNB builder to clarify
that the examples apply to use of the default Paketo builder, and adds links
to the official Paketo docs for more details.
Fixes gh-19967
Previously, JettyEmbeddedErrorHandler would change the method of every
request that is handles to GET. This was being done to work around
Jetty's error handling only dealing with GET, POST, and HEAD requests
by default. It had the unwanted side-effect of causing an error
response to a HEAD request having a body as, from the error handling's
perspective, it was a GET request.
This commit updates JettyEmbeddedErrorHandler to only set the method
on a request for which error handling is being performed if the method
isn't already one that will be handled, leaving the method of GET,
POST, and HEAD requests unchanged.
Unfortunately, short of implementing an HTTP client, this change cannot
be tested as the Apache HttpClient, OkHttp, and the JDK's
HttpURLConnection all silently drop the body of a response to a HEAD
request, preventing a test from asserting that a body hasn't been sent.
Closes gh-23551
Previously the artifact's version was used. In an artifact's version,
SNAPSHOT is replaced with the timestamped version number of a specific
snapshot. As a result, it no longer matches the *:*:*SNAPSHOT pattern.
This commit replaces switches to using the artifact's base version.
This preserves the SNAPSHOT in the version number. For non-snapshot
artifacts, the version and base version are identical.
Fixes gh-23533
Previously, we configured embedded Tomcat in such a way that no TLD
scan patterns were configured. This differed from a standalone
Tomcat installation where 4 patterns are configured that take
precedence over some of the skip patterns. The missing scan patterns
resulted in the skip patterns preventing the discovery of Log4j2's
TLDs.
This commit updates TomcatServletWebServerFactory to configure the
same four scan patterns as standalone Tomcat configures by default.
Fixes gh-23302
Previously, the web environment configured on `@SpringBootTest` was not
part of the context cache key. As a result, two test classes that has
identical configuration other than one using a MOCK web environment and
the other using a DEFINED_PORT web environment would share a context
when they should not do so. Classes that use MOCK and RANDOM_PORT were
not affected as the use of RANDOM_PORT results in a property for the
port being added to the environment.
This commit adds a new ContextCustomizer, SpringBootTestWebEnvironment,
that is used to capture the `webEnvironment` from `@SpringBootTest`
and use it in its hashCode and equals implementations. This fixes the
problem as all context customizers are evaluated when determing the
equality of two context cache keys.
Fixes gh-23085
This commit rework the tip on locating templates when running the app
in the IDE. Using classpath* should not change anything as this won't
make a difference without a pattern in the path.
Closes gh-23068
Previously LOG_LEVEL_PATTERN and LOG_DATEFORMAT_PATTERN were not
consumed as system properties in log4j2.xml and log4j2-file.xml. As a
result, the logging.pattern.level and logging.pattern.dateformat
configuration properties, which are translated into the
LOG_LEVEL_PATTERN and LOG_DATEFORMAT_PATTERN system properties
respectively had no effect.
This commit updates the log4j2.xml and log4j2-file.xml config files to
consume LOG_LEVEL_PATTERN and LOG_DATEFORMAT_PATTERN as system
properties. When the system property is not set, the configuation falls
back to the default values specified in the config files. Tests for
both log4j2.xml and log4j2-file.xml to verify the behaviour have also
bean added.
Fixes gh-22983
Previously, spring.flyway.init-sqls was only applied to Flyway's
configuration if Flyway was being configured to create the DataSource.
If Flyway was being configured to use an existing DataSource, init-sqls
was not applied. This is a hangover from when the init SQLs support was
introduced. At that time, Flyway only supported SQL to initialize the
connection when it was creating the DataSource. Flyway 5.2 added init
SQL support no matter how Flyway's DataSource was configured.
This commit updates FlywayAutoConfiguration to always apply the
init-sqls property to Flyway's configuration. The property's
documentation does not describe the current limitation so this change
should align the behaviour with what the documentation leads people to
expect.
Fixes gh-23392
Update `JarFileEntries` so that the interface is obtained rather than
the concrete implementation. This allows `JarEntry` values to be used
without causing a ClassCastException.
Closes gh-19041
Update the performance improvements to push certificate loading
and storage into the `JarFileEntries` class. This allows us to
keep certificates without needing to cache all entry data. We
now also keep certificates and code signers in a dedicated class
which is set whenever the full jar stream as been read, even if
the contained values are `null`. The logic that assumes META-INF
entries are not signed has been removed in favor of delegating to
the streamed entry results.
See gh-19041
Update Spring Boot nested JarFile support to improve the performance of
signed jars. Prior to this commit, `certificates` and `codeSigners`
were read by streaming the entire jar whenever the existing values
were `null`. Unfortunately, the contract for `getCertificates` and
get `getCodeSigners` states that `null` is a valid return value. This
meant that full jar streaming would occur whenever either method was
called on an entry that had no result. The problem was further
exacerbated by the fact that entries might not be cached.
See gh-19041
Update `ValidationBindHandler` so that pushed fields that reference
map keys can be used. This fixes a regression that was introduced in
commit 4483f417 when we switched to a `AbstractBindingResult` that no
longer required public getters/setters.
Closes gh-20350
Remove casting "timestamp" to `java.util.Date` in
`ErrorMvcAutoConfiguration` as the cast is not necessary and it
prevents other types (e.g. `java.time`) from being used.
See gh-23256
Prior to this commit, Actuator would sanitize properties values when
serializing them on the dedicated endpoint. Keys like "password" or
"secret" are entirely sanitized, but other keys like "uri" or "address"
are considered as URI types and only the password part of the user info
is sanitized.
This commit fixes the sanitization process where lists of such URI types
would not match the first entries of the list since they're starting
with `'['`. This commit improves the regexp matching process to sanitize
all URIs within a collection.
The documentation is also updated to better underline the processing
difference between complete sanitization and selective sanitization for
URIs.
Fixes gh-23037
Previous to this commit the connection idle timeout and heartbeat
interval were mapped to seconds whereas Cassandra expects ms for all
duration types.
This commit fixes the mapping and removes the default duration unit
since it should be considered ms like every other duration properties.
Closes gh-23249
Previously, PropertiesLauncher would close each archive that it
iterated over when creating its ClassLoader. This was not aligned
with JarLauncher's behaviour and left the ClassLoader with closed
archives. The close was introduced in [1] and became more apparent
following the change to fail operations on closed archives [2].
This commit updates Launcher to remove the close() that was added in
[1]. This aligns the behavior of PropertiesLauncher with JarLauncher
and ensures that the ClassLoader does not have entries backed by
closed archives on its classpath.
Fixes gh-23165
[1] ad72f86bdb
[2] ed7a5db174
Update `OriginTrackedYamlLoader` to remove node limits and recursive
parsing restrictions. SnakeYAML 1.26 introduced these options in order
to protect against the "billion laugh attacks" but since we consider
`application.yml` files to be trusted, we don't need these restrictions.
Fixes gh-23096
Extend `initializeSystem` to search the exception stack for a
FileNotFoundException before reporting the error. This allows
us to provide a similar stack trace to the one that used to be
thrown when we had the `ResourceUtils.getURL` check.
See gh-22946
Remove `ResourceUtils.getURL` checking from `LoggingApplicationListener`
so that logging systems can implement custom location support.
Prior to this commit, we checked in the listener if the specified config
location could be opened as a URL. This unfortunately prevents Log4J
extensions such as `log4j-spring-cloud-config-client` from implementing
configurable SSL and credentials support.
See gh-22946
In some cases, a call to the Docker image load API will fail but
return a 200 OK response status code and an empty response. This
commit detects that the response from this call is empty and
treats this condition as an error instead of a silent failure.
Fixes gh-23130
This commit changes the path used to invoke lifecycle binaries in
CNB builders from `/lifecycle` to `/cnb/lifecycle` to conform to
the CNB spec. This will ensure the build plugin image-building
goals and tasks are compatible with future versions of builders
that may not support both paths.
See gh-23009
Update `BuildImageMojo` so that commons HTTP wire logging is configured
to ERROR. Without this update, running `mvn -X` will produce a great
deal of logging since by default HTTP Client logs all bytes transfered
at DEBUG and Maven will enable DEBUG for all logs.
Closes gh-22674
Update `Repackager` to ensure that `getLayout` is called before we
backup the source file. This restores earlier behavior that some
custom `ModuleFactory` implementations were relying on.
Closes gh-22995
This commit adds a check to the `layertools extract` command to
ensure that the jar file being processed is readable and has a
valid directory.
Fixes gh-22993
Create a new `JarFileWrapper` class so that we can wrap and existing
`JarFile` and offer a version that can be safely closed.
Prior to this commit, we provided wrapper functionality in the `JarFile`
class itself. Unfortunately, because we override `close` and also create
a lot of wrappers this caused memory issues when running on Java 11.
With Java 11 `java.util.zip.ZipFile` class uses `FinalizableResource`
for any implementation that overrides `close()`. This means that any
wrapper classes will not be garbage collected until the JVM finalizer
thread runs.
Closes gh-22991
Update metrics auto-configurations so that they are auto-configured
after `CompositeMeterRegistryAutoConfiguration` in order to ensure
the `MeterRegistry` bean has been defined.
Prior to this commit, metrics auto-configurations that depended on a
`MeterRegistry` has `@AutoConfigureAfter(MetricsAutoConfiguration.class)`
which is not sufficient since `MetricsAutoConfiguration` does not export
a `MeterRegistry`.
See gh-21134
With the upgrade to the new Couchbase SDK and the related changes in
Spring Data Couchbase, CacheManagerCustomizer can no longer be used to
customize the Couchbase cache manager as it is an immutable class.
This commit introduces a dedicated callback for the
CouchbaseCacheManagerBuilder that is used by the auto-configuration and
update the documentation to refer to it with a sample usage.
Closes gh-22573
Previously, UrlPathHelper's full path optimization was enabled when
there was a dispatcher servlet mapped to /. The UrlPathHelper is used
across Spring MVC and if there are multiple dispatcher servlets they
all share the sample UrlPathHelper. This meant that any additional
dispatcher servlets mapping to locations other than / would not be able
to map requests correctly as the UrlPathHelper would use the full path,
ignoring the url mapping of the dispatcher servlet.
This commit updates the MVC auto-configuration so that use of the full
path is only enabled if there's a single dispatcher servlet
registration.
Fixes gh-22682
Prior to this commit, DebugAgentEnvironmentPostProcessor throws a
RuntimeException with a generic error message if its initialization
fails. The causing exception is discarded, which makes error analysis
unnecessarily difficult.
This commit attaches the cause to the newly thrown RuntimeException.
See gh-22777
Previously, the Maven plugin integration tests used a settings.xml file
that defined https://repo.spring.io/snapshot as a repository. This
allowed them to resolve snapshots of the plugin's Spring Framework
dependencies but it had the unfortunate side-effect of also allowing
them to resolve snapshots of other Spring Boot modules from Artifactory
rather than using those currently being built.
This commit replaces the repositories in settings.xml with a Gradle
task that resolves the necessary dependencies and populates a local
repository with the dependencies' jars and pom files. This is achieved
using a ComponentMetadataRule that creates a custom variant of each
dependency that includes its pom file, inspired by the example in
gradle/gradle/#11449. A configuration that extends the
runtimeClasspath configuration and select the custom variant via its
attribute is then used to resolve the jars and pom files of the runtime
classpath such that they can then be used to populate the local
repository.
Closes gh-22828
Prior to this commit, the HTTP/2 server auto-configuration for Jetty
would require Conscrypt as a hard dependency.
This commit updates the auto-configuration for more flexibility and now
allows the following deployments:
* JDK9+ with the JDK ALPN implementation
* JDK8u252+ with the backported ALPN implementation
* Conscrypt with no JDK requirement
The auto-configuration now improves detection and guides developers in
case there is a missing `jetty-alpn-*-server` dependency.
The reference docs in the HOWTO section has been updated accordingly.
Closes gh-22188
Previously, the endpoints' responses could occasionally exceed
WebClient's in-memory buffer limt, for example if the threads endpoint
was reporting a large number of threads or the threads had large
stacks.
This commit disables WebClient's in-memory buffer size limit so that
the tests passing is not dependent on the size of the endpoints'
responses.
Closes gh-22743
Previously, waitsForQuietPeriod would iterate 10 times, touching a new
file and then sleeping for 100ms at it did so. With a quiet period of
200ms, this was intended to result in a single change set containing
10 files. However, the test would fail occasionally as multiple change
sets were detected. The test is multi-threaded and is, therefore, at
the mercy of the scheduler. If the thread that is iterating and
touching the files takes over 200ms to be scheduled – exceeding the
watcher's quiet period – the watcher may detect a change set while the
changes are still being made. Eliminating this possibilty would require
the test to participate in the watcher's synchronization, which would
require some changes to its implementation. Instead, this commit
aims to avoid the problem by sleeping for 1/10 of the time (10ms) and
expecting a single change set of 100 files. The hope is that the much
shorter sleep time will result in the file touching thread being
scheduled well within the 200ms quiet period.
Closes gh-22732
Prior to this commit, an entry in the environment map provided to the
build plugin image building goal or task that had a null value would
result in a failure with a message that was difficult to diagnose.
This commit treats env map entries with a null value as an empty
entry to prevent the failure and also make it easier to provide an
explicit empty entry in the Maven XML.
Fixes gh-22703
Prior to this commit, the livenessState and readinessState health
indicators would not be configured automatically and would be missing
from the "liveness" and "readiness" health groups, leading to 404s when
hitting the `/actuator/health/liveness` or `/actuator/health/readiness`.
This commit ensures that the health indicators beans have the proper
name and revisits the auto-configuration conditions to reflect that as
well.
Fixes gh-22562
This commit harmonizes dependency declarations for Jackson in the
actuator. Both Jackson and JSR 310 are back to optional in the core
actuator module and mandatory when using the auto-configuration.
Closes gh-22624
Prior to this change, TaskExecutorBuilder used seconds as its precision
to map the awaitTerminationPeriod value. This caused a loss of
millisecond information in the period.
This commit fixes the loss by converting the period to millisecond and
using setAwaitTerminationMillis to configure the executor.
See gh-22604
Previously, ignoring invalid fields would cause the failure for an
unknown field to be ignored, irrespective of the ignoreUnknownFields
attribute on `@ConfigurationProperties`.
This commit updates the NoUnboundElementsBindHandler to ensure that
any UnboundConfigurationPropertiesException is thrown rather than
being ignored when the handler has been wrapped by an
IgnoreErrorsBindHandler.
Fixes gh-22308
The fix for gh-20916 updated DefinitionsParser so that the
ResolvableType for each MockBean or SpyBean field included the
implementation class from which the field was found. Where the field
was declared with a variable generic signature that was made constant
by its implementation class, this allowed the correct concrete type to
be determined. It also had the unintended side-effect of preventing two
test classes with identical `@MockBean` and `@SpyBean` configuration
from sharing a context as the resolvable types for their mock and spy
bean fields would now be different.
This commit updates DefinitionsParser to only include the
implementation class in the ResolvableType if the field's generic type
is variable. For cases where it is not variable, this restores the
behaviour prior to the fix for gh-20916.
Fixes gh-22566
Previously, background pre-init was started in response to the
application starting event. This meant that it would be running while
the logging turbo filter was in place and was rejecting all logging.
As a result, any logging performed during pre-init would be lost.
This commit changes background pre-init so that it is now started
in response to the application environment prepared event. By this
point, the logging turbo filter has been removed as the logging levels
have been configured from the environment.
Closes gh-21897
Previously, only UndertowOptions was used as the source of options for
both server and socket options, but it only contains server options.
As a result, attempting to configure any socket options defined by
XNIO's Options class would fail.
This commit updates the property-based configuration of options to use
UndertowOptions as the source for server options and XNIO's Options as
the source for socket options.
Fixes gh-22502
This commit fixes the auto-configuration of the Couchbase caching
support by configuring the cache once Spring Data Couchbase has gotten
a chance to be processed.
Closes gh-22542
Andy Wilkinson
Stephane Nicoll
Phillip Webb
Brian Clozel
Scott Frederick
cdalexndr
Asha Somayajula
Andy Wilkinson
Thorasine
Madhura Bhave
Shraddha Yeole
ketaki-t
Rujun Chen
Pavel Gordon
dreis2211
Sebastiaan Fernandez
mathieufortin01
lock14
David Good
Ralph Goers
spencergibb
Thomas Kåsene
Ivo Smid
Martin Benda
Lopfest
Jean-Baptiste Nizet
Maciej Walkowiak
smlc
Johnny Lim
John Franco
Gabriele Bianchet-David
davidbilge
ksmilek
Tadaya Tsuyukubo
biergit