Previously, when DevTools' was used it would set the application
context's ResourceLoader and overwrite any custom ResourceLoader that
had been configured. On the rare occasion when the user had customized
the ResourceLoader this meant that the customization was lost and
certain resources would become unavailable.
This commit updates DevTools' ResourceLoader to delegate a custom
ResourceLoader if one has been configured. If one has not been
configured it delegates as before, i.e. to
WebApplicationContextResourceLoader for web applications and to
DefaultResourceLoader for all others apps.
Closes gh-8010
The intent of the test is to:
1. Make a change
2. Sleep for long enough for that change to be picked up
3. Make another change
4. Stop that watcher after 1 further scan has been performed
5. Assert that a further scan was performed by checking that two
separate sets of changes (step 1 and step 3) have been picked up
Previously, step 2 relied on simply sleeping for a period of time
longer than the polling interval. In reality, the polling interval
is only a minimum time between scans and the actual time between them
depends on thread scheduling, GC pauses, etc. This lead to the
test failing intermittently if the scan didn't happen in a timely
manner.
This commit removes the sleep and replaces it with a while loop that
waits for first change to be picked up. This ensures that the second
change will be detected separately from the first and that two
separate change sets should always be available once the watcher has
stopped.
See gh-7782
Update ClassLoaderFilesResourcePatternResolver to support servlet
resources when it's being used with a WebApplicationContext.
Prior to commit 918e122ddc a `ResourceLoader` was not added to the
`ApplicationContext`, meaning that servlet resources could be found by
virtue of the protected `getResourceByPath()` method. Following commit
918e122ddc, the context `ResourceLoader` is set, meaning that all calls
to `getResource` delegate to the `ResourceLoader` and the
`ApplicationContext` methods are not invoked. Since the devtools
`ResourceLoader` wasn't Servlet aware, servlet resources could not
be found.
Fixes gh-7752
Previously, remote DevTools only correctly supported modifying
existing classes. New classes that were added would be missed, and
deleted classes could cause a failure as they would be found by
component scanning but hidden by RestartClassLoader.
This commit introduces a DevTools-specific ResourcePatternResolver
that is installed as the application context's resource loader. This
custom resolver is aware of the files that have been added and
deleted and modifies the result returned from getResource and
getResources accordingly.
New intergration tests have been introduced to verify DevTools'
behaviour. The tests cover four scenarios:
- Adding a new controller
- Removing an existing controller
- Adding a request mapping to a controller
- Removing a request mapping from a controller
These four scenarios are tested with:
- DevTools updating a local application
- DevTools updating a remote application packaged in a jar file
- DevTools updating a remote application that's been exploded
Closes gh-7379
If the resource chain is used, such as by using the
`"spring.resources.chain.strategy.content.enabled"` property,
resource chain caching can prevent the developer from seeing
changes made to resources, so that caching should be disabled
when DevTools is enabled.
This commit sets the `"spring.resources.chain.cache"` property
to `true` when devtools is enabled.
Update `DevToolsPropertyDefaultsPostProcessor` so that property defaults
are only added at development time. Properties are now added only when
`Restarter` is initialize or remote devtools is enabled.
Fixes gh-7014
Prior to this commit any application configured to write
`META-INF/build-info.properties` could trigger unexpected application
restarts. The problem is particularly prevalent when using Eclipse M2E in
combination with Maven's `spring-boot-maven-plugin` `build-info` goal
and Gradle's `springBoot` `buildInfo()`.
Closes gh-7002
Where possible, code that previously synchronized on this (or on the
class in the case of static methods) has been updated to use an
internal monitor object instead. This allows the locking model that's
employed to be an implementation detail rather than part of the
class's API.
Classes that override a synchronized method continue to declare
the overriding method as synchronized. This ensures that locking
is consistent across the superclass and its subclass.
Closes gh-6262
Previously, if an exception was resolved by a `HandlerExceptionResolver`
nothing the log indicated a failure to process the query.
This commit adds a new property `spring.mvc.log-resolved-exception` that
enables warning logs for supported `HandlerExceptionResolver` instances.
When Devtools is enabled, this flag is enabled by default.
Closes gh-2176
The upgrade to Hibernate 5.2.0.Final has provide to be too
problematic to live with. It requires Java 8, is incompatible with
a number of other projects in the Hibernate ecosystem, and it's
unclear for how long it will be maintained. We'd previously used
Hibernate 5.1.0.Final but its maintenance is also unclear with
Hibernate 5.1.1.Final being more than 3 months overdue.
This commit drops back to Hibernate 5.0.9.Final. This has a few
advantages:
- It's Java 7 compatible
- It's had some time to mature and should be reasonably free of
regressions for those moving from 4.3.x
- It's used in both Wildfly and JBoss EAP so there's a fair chance
that it will continue to be maintained.
Closes gh-6198
This commit changes the default version of Tomcat to 8.5.3 while
also retaining support for Tomcat 8.0 and 7.0. The main difference
in 8.5 is that the ServerSocketFactory abstraction that allowed the
TrustStore and KeyStore to be configured programatically no longer
exists. This logic has been replaced with the use of a custom URL
protocol (springbootssl) that provides access to the key store and
trust store of an SslStoreProvider. In addition to working with 8.5,
this approach has the advantage of also working with 8.0 and 7.0.
Closes gh-6164
DevTools deliberately throws an uncaught exception on the main thread
as a safe way of causing it to stop processing. This exception is
caught and swallowed by an uncaught exception handler. Unfortunately,
this has the unwanted side-effect of causing the JVM to exit with 1
once all running threads are daemons.
Normally, this isn't a problem. Non-daemon threads, such as those
started by an embedded servlet container, will keep the JVM alive and
restarts of the application context will occur when the user makes to
their application. However, if the user adds DevTools to an
application that doesn't start any non-daemon threads, i.e. it starts,
runs, and then exits, it will exit with 1. This causes both
bootRun in Gradle and spring-boot:run in Maven to report that the
build has failed. While there's no benefit to using DevTools with an
application that behaves in this way, the side-effect of causing the
JVM to exit with 1 is unwanted.
This commit address the problem by updating the uncaught exception
handler to call System.exit(0) if the JVM is going to exit as a
result of the uncaught exception causing the main thread to die. In
other words, if the main thread was the only non-daemon thread, its
death as a result of the uncaught exception will now cause the JVM
to exit with 1 rather than 0. If there are other non-daemon threads
that will keep the JVM alive, the behaviour is unchanged.
Closes gh-5968
Previously, Spring Data Redis assumed that the class loader that loaded
its classes would also be able to load the application’s classes. This
assumption is faulty when there are multiple class loaders involved
such as when using dev tools or when Spring Data Redis is installed as
a shared library in a servlet container.
DATAREDIS-427 and DATAREDIS-501 removed this assumption by making the
default serialiser use the bean class loader. DevTools configures this
class loader to be the restart class loader which can load the
application’s classes. As a result of moving to Hopper SR2 snapshots,
these fixes are now available and we can remove our custom serialised.
Closes gh-5760
- Limit shared state between FileSystemWatcher and the watching thread
- Use a private monitor rather than synchronizing on this
- Use a Runnable implementation rather than subclassing Thread
- Synchronize consistently when reading and writing state
Closes gh-6039
Extract TemplateAvailabilityProvider caching logic to a new
TemplateAvailabilityProviders class and provide property support to
disable it. Also update DevToolsPropertyDefaultsPostProcessor to
automatically set the property.
Fixes gh-5989
The list of changes is written to on one thread and read from on
another. Without some form of sychronization this is not thread-safe.
This commit makes changes a synchronized list which should guarantee
that the reading thread can see the changes made by the writing thread.
It also removes a redundant call to clear the list of changes at the
start of waitsForPollingInterval.
See gh-6038
Spring Buildmaster
Andy Wilkinson
Phillip Webb
Craig Andrews
Johnny Lim
Stephane Nicoll
Martin Lippert