Skip to content
This repository was archived by the owner on May 30, 2024. It is now read-only.

Commit 03ca79e

Browse files
LaunchDarklyReleaseBoteli-darklyLaunchDarklyReleaseBotaengelbergantonmos
authored
prepare 6.0.3 release (#293)
* more model updates * use LDContext in Evaluator; add per-kind targets logic * rm redundant test * skip inapplicable contract tests * implement new context kind logic for clauses * implement new context kind logic for rollouts * skip inapplicable contract tests * add context kind logic for regular segments and big segments * allow segments to reference other segments * guard against prerequisite cycles * implement event context redaction logic * misc fixes * implement the rest of U2C event logic * misc fixes * fix benchmarks * add contextKinds to summary + misc fixes * events schema 4 * update client API to use contexts; enable v2 contract tests * misc fixes * javadoc & test app fixes * javadoc fix * javadoc fix * unskip some tests since the behavior was fixed in java-sdk-common * bad attribute in clause should be reported as an error * update TestData to be context-aware * misc fixes + javadoc * fix allFlagsState to not generate prereq eval events * factor context deduplication logic out of DefaultEventProcessor * fix JSON serialization test * refactor diagnostic events logic to abstract out server-side-specific info * separate inner events implementation from public interface, don't expose event types * fix benchmarks * fix benchmarks * rm unused * rm unused * separate Java-SDK-specific HTTP functionality from events logic * fix tests * don't close the underlying Writer when we stop using the JsonWriter * rm unused * remove references to Java 8 types and Guava from shared event/HTTP code * fix benchmarks * also try not to use java.time.Duration in the event tests * also avoid using lambdas in event test code * move shared code into internal packages before splitting it off * add "...ForAll" TestData methods to replace "...ForAllUsers" * replace all component factory interfaces with a single generic one * change HttpConfiguration and LoggingConfiguration to concrete classes * bump okhttp & okhttp-eventsource dependencies * update comment to clarify that level() doesn't apply to SLF4J * update readme to mention different logging examples in hello-java * switch to use snapshot build of java-logging, pending next release * level setting does not apply to SLF4J and JUL * use java-logging 1.1.0 release * make sure META-INF files are never mistaken for classes and relocated * update shared data store test logic to pass ClientContext with logger * enable external javadoc links for com.launchdarkly.logging types * use variable for dependency version * remove another Guava usage * make events test code independent of server-side data model * fix flaky big segment status polling tests * fix benchmark code * update benchmarks * javadoc formatting * javadocs * undo accidental reversion * remove some more stray references to java-server-sdk test helper code * more efficient equals() and hashCode() for Operator * method can be static * clarify comment * typo * remove event logic, migrate to using java-sdk-internal * adjust for API change * use constant * fix packaging tests * remove inapplicable benchmarks * add error message if makefile can't run * force result to be "not in experiment" if bucketing calculation failed due to context kind not found * use newer HTTP/TCP test helpers * Update Windows orb, fix Windows JDK install in CI (#372) * update snakeyaml for CVE-CVE-2022-25857 * latest snakeyaml is 1.31 * bump snakeyaml version for CVE-2022-38752 * disable Windows Java 11 build * minor test updates for java-sdk-internal API changes * fix packaging of com.launchdarkly.logging classes * rm debugging * reconsidered - let's include the logging classes in the jars * fix packaging test logic * correct documentation * remove secondary meta-attribute * rm obsolete references to UserAttribute * support passing LDUser instead of LDContext in all SDK methods (#379) * support passing LDUser instead of LDContext in all SDK methods * actually we should use default methods * rm unused * enable test capability for user type * update java-sdk-common dependency * use okhttp-eventsource 3.0.0-SNAPSHOT * use non-snapshot okhttp-eventsource * include nested segment references in dependency update checks * use java-sdk-internal 1.0.0 * remove SLF4J dependency, remove "all" jar, simplify build (#381) * don't generate evaluation events for invalid context * use synchronous EventSource * comments * revise implementation of special HTTP configurations test * make sure we set readTimeout last * use okhttp-eventsource 4.0.0 * use newer HTTP test helpers (#385) * use synchronous EventSource (5.x backport) * update Gradle to 7.6 + fix snapshot releases * fix #288 (#289) * add unit test for externally-contributed YAML security fix * backport YAML CVE fix from 6.x * prepare 5.10.5 release (#290) * update CI and Gradle to test with newer JDKs (#259) * update okhttp to 3.14.9 (fixes incompatibility with OpenJDK 8.0.252) * prepare 4.14.2 release (#205) * Releasing version 4.14.2 * update okhttp to 4.8.1 (fixes incompatibility with OpenJDK 8.0.252) * gitignore * Bump SnakeYAML from 1.19 to 1.26 to address CVE-2017-18640 * prepare 4.14.3 release (#209) * Releasing version 4.14.3 * comments * only log initialization message once in polling mode * [ch89935] Correct some logging call format strings (#264) Also adds debug logs for full exception information in a couple locations. * [ch90109] Remove outdated trackMetric comment from before service support. (#265) * Fix compatibility with Java 7. * Remove import that is no longer used. * add Java 7 build (#267) * prepare 4.14.4 release (#214) * Releasing version 4.14.4 * add and use getSocketFactory * alignment * add socketFactory to builder * test socket factory builder * preserve dummy CI config file when pushing to gh-pages (#271) * fix concatenation when base URI has a context path (#270) * fix shaded jar builds to exclude Jackson classes and not modify Jackson return types (#268) * add test httpClientCanUseCustomSocketFactory for DefaultFeatureRequestor * add httpClientCanUseCustomSocketFactory() test for DefaultEventSenderTest * add httpClientCanUseCustomSocketFactory() test to StreamProcessorTest * pass URI to in customSocketFactory event test * make test less ambiguous * copy rules to new FlagBuilder instances (#273) * Bump guava version (#274) * Removed the guides link * increment versions when loading file data, so FlagTracker will work (#275) * increment versions when loading file data, so FlagTracker will work * update doc comment about flag change events with file data * add ability to ignore duplicate keys in file data (#276) * add alias events (#278) * add alias events and function * update tests for new functionality * update javadoc strings * add validation of javadoc build to CI * update commons-codec to 1.15 (#279) * Add support for experiment rollouts * add tests and use seed for allocating user to partition * test serialization and add check for isExperiment * fix PollingProcessorTest test race condition + other test issues (#282) * use launchdarkly-java-sdk-common 1.1.0-alpha-expalloc.2 * Update src/test/java/com/launchdarkly/sdk/server/EvaluatorTest.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * Update src/test/java/com/launchdarkly/sdk/server/EvaluatorTest.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * Update src/test/java/com/launchdarkly/sdk/server/EvaluatorTest.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * Update src/test/java/com/launchdarkly/sdk/server/EvaluatorTest.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * changes per code review comments * Please enter the commit message for your changes. Lines starting * fix null pointer exception * address code review comments * address more comments * missed a ! for isUntracked() * fix default boolean for json * make untracked FALSE by default * refactoring of bucketing logic to remove the need for an extra result object (#283) * add comment to enum * various JSON fixes, update common-sdk (#284) * simlpify the logic and make it match node/.Net sdks * Update src/main/java/com/launchdarkly/sdk/server/EventFactory.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * add the same comment as the Node SDK * Remove outdated/meaningless doc comment. (#286) * protect against NPEs if flag/segment JSON contains a null value * use java-sdk-common 1.2.0 * fix Jackson-related build issues (again) (#288) * update to okhttp-eventsource patch for stream retry bug, improve tests (#289) * update to okhttp-eventsource patch for stream retry bug, improve test * add test for appropriate stream retry * add public builder for FeatureFlagsState (#290) * add public builder for FeatureFlagsState * javadoc fixes * clarify FileData doc comment to say you shouldn't use offline mode (#291) * improve validation of SDK key so we won't throw an exception that contains the key (#293) * fix javadoc link in FileData comment (#294) * fix PollingProcessor 401 behavior and use new HTTP test helpers (#292) * re-fix metadata to remove Jackson dependencies, also remove Class-Path from manifest (#295) * make FeatureFlagsState.Builder.build() public (#297) * clean up tests using java-test-helpers 1.1.0 (#296) * use Releaser v2 config + newer CI images (#298) * [ch123129] Fix `PollingDataSourceBuilder` example. (#299) * Updates docs URLs * always use US locale when parsing HTTP dates * use Gson 2.8.9 * don't try to send more diagnostic events after an unrecoverable HTTP error * ensure module-info file isn't copied into our jars during build * use Gradle 7 * update build for benchmarks * more Gradle 7 compatibility changes for benchmark job * test with Java 17 in CI (#307) * test with Java 17 in CI * also test in Java 17 for Windows * fix choco install command * do date comparisons as absolute times, regardless of time zone (#310) * fix suppression of nulls in JSON representations (#311) * fix suppression of nulls in JSON representations * distinguish between situations where we do or do not want to suppress nulls * fix identify/track null user key check, also don't create index event for alias * use latest java-sdk-common * fix setting of trackEvents/trackReason in allFlagsState data when there's an experiment * implement contract tests (#314) * Merge Big Segments feature branch for 5.7.0 release (#316) Includes Big Segments implementation and contract test support for the new behavior. * Fix for pom including SDK common library as a dependency. (#317) * use new logging API * update readme notes about logging * set base logger name for SDK per test * comment * javadoc fixes * revert accidental commit * Upload JUnit XML to CircleCI on failure (#320) Fix a bug in the CircleCI config that was only uploading JUnit XML on _success_, not failure. * Add application tag support (#319) * Enforce 64 character limit on application tag values (#323) * fix "wrong type" logic in evaluations when default value is null * Rename master to main in .ldrelease/config.yml (#325) * Simpler way of setting base URIs in Java (#322) Now supports the `ServiceEndpoints` config for setting custom URIs for endpoints in a single place * update logging info in readme * use 1.0.0 release of logging package * misc cleanup * remove unnecessary extra interfaces, just use default methods instead * make BigSegmentStoreWrapper.pollingDetectsStaleStatus test less timing-sensitive * make LDEndToEndClientTest.test____SpecialHttpConfigurations less timing-sensitive * make data source status tests less timing-sensitive * use streaming JSON parsing for incoming LD data * fix tests * rm unused * rm unused * use okhttp-eventsource 2.6.0 * update eventsource to 2.6.1 to fix pom/manifest problem * increase efficiency of summary event data structures (#335) * make reusable EvaluationDetail instances as part of flag preprocessing (#336) * make evaluator result object immutable and reuse instances * comment * avoid creating List iterators during evaluations * remove unnecessary copy * fix allFlagsState to not generate prereq eval events * add "...ForAll" TestData methods to replace "...ForAllUsers" * bump okhttp & okhttp-eventsource dependencies * update comment to clarify that level() doesn't apply to SLF4J * update readme to mention different logging examples in hello-java * switch to use snapshot build of java-logging, pending next release * level setting does not apply to SLF4J and JUL * use java-logging 1.1.0 release * make sure META-INF files are never mistaken for classes and relocated * update shared data store test logic to pass ClientContext with logger * enable external javadoc links for com.launchdarkly.logging types * use variable for dependency version * fix flaky big segment status polling tests * Update Windows orb, fix Windows JDK install in CI (#372) * update snakeyaml for CVE-CVE-2022-25857 * latest snakeyaml is 1.31 * bump snakeyaml version for CVE-2022-38752 * disable Windows Java 11 build * fix packaging of com.launchdarkly.logging classes * rm debugging * reconsidered - let's include the logging classes in the jars * fix packaging test logic * correct documentation * use synchronous EventSource (5.x backport) * backport YAML CVE fix from 6.x Co-authored-by: LaunchDarklyCI <dev@launchdarkly.com> Co-authored-by: Eli Bishop <eli@launchdarkly.com> Co-authored-by: LaunchDarklyCI <LaunchDarklyCI@users.noreply.github.com> Co-authored-by: Gavin Whelan <gwhelan@launchdarkly.com> Co-authored-by: ssrm <ssrm@users.noreply.github.com> Co-authored-by: Harpo Roeder <hroeder@launchdarkly.com> Co-authored-by: Ben Woskow <48036130+bwoskow-ld@users.noreply.github.com> Co-authored-by: Elliot <35050275+Apache-HB@users.noreply.github.com> Co-authored-by: Robert J. Neal <rneal@launchdarkly.com> Co-authored-by: Robert J. Neal <robertjneal@users.noreply.github.com> Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> Co-authored-by: LaunchDarklyReleaseBot <launchdarklyreleasebot@launchdarkly.com> Co-authored-by: Ember Stevens <ember.stevens@launchdarkly.com> Co-authored-by: ember-stevens <79482775+ember-stevens@users.noreply.github.com> Co-authored-by: Alex Engelberg <alex.benjamin.engelberg@gmail.com> Co-authored-by: Alex Engelberg <aengelberg@launchdarkly.com> * Releasing version 5.10.5 * don't allow uncaught RuntimeException on stream thread and don't report errors on shutdown * prepare 5.10.6 release (#292) * prepare 4.14.2 release (#205) * Releasing version 4.14.2 * update okhttp to 4.8.1 (fixes incompatibility with OpenJDK 8.0.252) * gitignore * Bump SnakeYAML from 1.19 to 1.26 to address CVE-2017-18640 * prepare 4.14.3 release (#209) * Releasing version 4.14.3 * comments * only log initialization message once in polling mode * [ch89935] Correct some logging call format strings (#264) Also adds debug logs for full exception information in a couple locations. * [ch90109] Remove outdated trackMetric comment from before service support. (#265) * Fix compatibility with Java 7. * Remove import that is no longer used. * add Java 7 build (#267) * prepare 4.14.4 release (#214) * Releasing version 4.14.4 * add and use getSocketFactory * alignment * add socketFactory to builder * test socket factory builder * preserve dummy CI config file when pushing to gh-pages (#271) * fix concatenation when base URI has a context path (#270) * fix shaded jar builds to exclude Jackson classes and not modify Jackson return types (#268) * add test httpClientCanUseCustomSocketFactory for DefaultFeatureRequestor * add httpClientCanUseCustomSocketFactory() test for DefaultEventSenderTest * add httpClientCanUseCustomSocketFactory() test to StreamProcessorTest * pass URI to in customSocketFactory event test * make test less ambiguous * copy rules to new FlagBuilder instances (#273) * Bump guava version (#274) * Removed the guides link * increment versions when loading file data, so FlagTracker will work (#275) * increment versions when loading file data, so FlagTracker will work * update doc comment about flag change events with file data * add ability to ignore duplicate keys in file data (#276) * add alias events (#278) * add alias events and function * update tests for new functionality * update javadoc strings * add validation of javadoc build to CI * update commons-codec to 1.15 (#279) * Add support for experiment rollouts * add tests and use seed for allocating user to partition * test serialization and add check for isExperiment * fix PollingProcessorTest test race condition + other test issues (#282) * use launchdarkly-java-sdk-common 1.1.0-alpha-expalloc.2 * Update src/test/java/com/launchdarkly/sdk/server/EvaluatorTest.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * Update src/test/java/com/launchdarkly/sdk/server/EvaluatorTest.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * Update src/test/java/com/launchdarkly/sdk/server/EvaluatorTest.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * Update src/test/java/com/launchdarkly/sdk/server/EvaluatorTest.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * changes per code review comments * Please enter the commit message for your changes. Lines starting * fix null pointer exception * address code review comments * address more comments * missed a ! for isUntracked() * fix default boolean for json * make untracked FALSE by default * refactoring of bucketing logic to remove the need for an extra result object (#283) * add comment to enum * various JSON fixes, update common-sdk (#284) * simlpify the logic and make it match node/.Net sdks * Update src/main/java/com/launchdarkly/sdk/server/EventFactory.java Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> * add the same comment as the Node SDK * Remove outdated/meaningless doc comment. (#286) * protect against NPEs if flag/segment JSON contains a null value * use java-sdk-common 1.2.0 * fix Jackson-related build issues (again) (#288) * update to okhttp-eventsource patch for stream retry bug, improve tests (#289) * update to okhttp-eventsource patch for stream retry bug, improve test * add test for appropriate stream retry * add public builder for FeatureFlagsState (#290) * add public builder for FeatureFlagsState * javadoc fixes * clarify FileData doc comment to say you shouldn't use offline mode (#291) * improve validation of SDK key so we won't throw an exception that contains the key (#293) * fix javadoc link in FileData comment (#294) * fix PollingProcessor 401 behavior and use new HTTP test helpers (#292) * re-fix metadata to remove Jackson dependencies, also remove Class-Path from manifest (#295) * make FeatureFlagsState.Builder.build() public (#297) * clean up tests using java-test-helpers 1.1.0 (#296) * use Releaser v2 config + newer CI images (#298) * [ch123129] Fix `PollingDataSourceBuilder` example. (#299) * Updates docs URLs * always use US locale when parsing HTTP dates * use Gson 2.8.9 * don't try to send more diagnostic events after an unrecoverable HTTP error * ensure module-info file isn't copied into our jars during build * use Gradle 7 * update build for benchmarks * more Gradle 7 compatibility changes for benchmark job * test with Java 17 in CI (#307) * test with Java 17 in CI * also test in Java 17 for Windows * fix choco install command * do date comparisons as absolute times, regardless of time zone (#310) * fix suppression of nulls in JSON representations (#311) * fix suppression of nulls in JSON representations * distinguish between situations where we do or do not want to suppress nulls * fix identify/track null user key check, also don't create index event for alias * use latest java-sdk-common * fix setting of trackEvents/trackReason in allFlagsState data when there's an experiment * implement contract tests (#314) * Merge Big Segments feature branch for 5.7.0 release (#316) Includes Big Segments implementation and contract test support for the new behavior. * Fix for pom including SDK common library as a dependency. (#317) * use new logging API * update readme notes about logging * set base logger name for SDK per test * comment * javadoc fixes * revert accidental commit * Upload JUnit XML to CircleCI on failure (#320) Fix a bug in the CircleCI config that was only uploading JUnit XML on _success_, not failure. * Add application tag support (#319) * Enforce 64 character limit on application tag values (#323) * fix "wrong type" logic in evaluations when default value is null * Rename master to main in .ldrelease/config.yml (#325) * Simpler way of setting base URIs in Java (#322) Now supports the `ServiceEndpoints` config for setting custom URIs for endpoints in a single place * update logging info in readme * use 1.0.0 release of logging package * misc cleanup * remove unnecessary extra interfaces, just use default methods instead * make BigSegmentStoreWrapper.pollingDetectsStaleStatus test less timing-sensitive * make LDEndToEndClientTest.test____SpecialHttpConfigurations less timing-sensitive * make data source status tests less timing-sensitive * use streaming JSON parsing for incoming LD data * fix tests * rm unused * rm unused * use okhttp-eventsource 2.6.0 * update eventsource to 2.6.1 to fix pom/manifest problem * increase efficiency of summary event data structures (#335) * make reusable EvaluationDetail instances as part of flag preprocessing (#336) * make evaluator result object immutable and reuse instances * comment * avoid creating List iterators during evaluations * remove unnecessary copy * fix allFlagsState to not generate prereq eval events * add "...ForAll" TestData methods to replace "...ForAllUsers" * bump okhttp & okhttp-eventsource dependencies * update comment to clarify that level() doesn't apply to SLF4J * update readme to mention different logging examples in hello-java * switch to use snapshot build of java-logging, pending next release * level setting does not apply to SLF4J and JUL * use java-logging 1.1.0 release * make sure META-INF files are never mistaken for classes and relocated * update shared data store test logic to pass ClientContext with logger * enable external javadoc links for com.launchdarkly.logging types * use variable for dependency version * fix flaky big segment status polling tests * Update Windows orb, fix Windows JDK install in CI (#372) * update snakeyaml for CVE-CVE-2022-25857 * latest snakeyaml is 1.31 * bump snakeyaml version for CVE-2022-38752 * disable Windows Java 11 build * fix packaging of com.launchdarkly.logging classes * rm debugging * reconsidered - let's include the logging classes in the jars * fix packaging test logic * correct documentation * use synchronous EventSource (5.x backport) * backport YAML CVE fix from 6.x * don't allow uncaught RuntimeException on stream thread and don't report errors on shutdown Co-authored-by: Eli Bishop <eli@launchdarkly.com> Co-authored-by: LaunchDarklyCI <LaunchDarklyCI@users.noreply.github.com> Co-authored-by: LaunchDarklyCI <dev@launchdarkly.com> Co-authored-by: Gavin Whelan <gwhelan@launchdarkly.com> Co-authored-by: ssrm <ssrm@users.noreply.github.com> Co-authored-by: Harpo Roeder <hroeder@launchdarkly.com> Co-authored-by: Ben Woskow <48036130+bwoskow-ld@users.noreply.github.com> Co-authored-by: Elliot <35050275+Apache-HB@users.noreply.github.com> Co-authored-by: Robert J. Neal <rneal@launchdarkly.com> Co-authored-by: Robert J. Neal <robertjneal@users.noreply.github.com> Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> Co-authored-by: LaunchDarklyReleaseBot <launchdarklyreleasebot@launchdarkly.com> Co-authored-by: Ember Stevens <ember.stevens@launchdarkly.com> Co-authored-by: ember-stevens <79482775+ember-stevens@users.noreply.github.com> Co-authored-by: Alex Engelberg <alex.benjamin.engelberg@gmail.com> Co-authored-by: Alex Engelberg <aengelberg@launchdarkly.com> * Releasing version 5.10.6 Co-authored-by: Eli Bishop <eli@launchdarkly.com> Co-authored-by: LaunchDarklyReleaseBot <launchdarklyreleasebot@launchdarkly.com> Co-authored-by: Alex Engelberg <aengelberg@launchdarkly.com> Co-authored-by: Anton Mostovoy <anton@mostovoy.net> Co-authored-by: LaunchDarklyCI <dev@launchdarkly.com> Co-authored-by: LaunchDarklyCI <LaunchDarklyCI@users.noreply.github.com> Co-authored-by: Gavin Whelan <gwhelan@launchdarkly.com> Co-authored-by: ssrm <ssrm@users.noreply.github.com> Co-authored-by: Harpo Roeder <hroeder@launchdarkly.com> Co-authored-by: Ben Woskow <48036130+bwoskow-ld@users.noreply.github.com> Co-authored-by: Elliot <35050275+Apache-HB@users.noreply.github.com> Co-authored-by: Robert J. Neal <rneal@launchdarkly.com> Co-authored-by: Robert J. Neal <robertjneal@users.noreply.github.com> Co-authored-by: Sam Stokes <sstokes@launchdarkly.com> Co-authored-by: Ember Stevens <ember.stevens@launchdarkly.com> Co-authored-by: ember-stevens <79482775+ember-stevens@users.noreply.github.com> Co-authored-by: Alex Engelberg <alex.benjamin.engelberg@gmail.com>
1 parent 7905504 commit 03ca79e

File tree

3 files changed

+66
-4
lines changed

3 files changed

+66
-4
lines changed

‎CHANGELOG.md‎

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2,6 +2,10 @@
22

33
All notable changes to the LaunchDarkly Java SDK will be documented in this file. This project adheres to [Semantic Versioning](http://semver.org).
44

5+
## [5.10.6] - 2023-01-06
6+
### Fixed:
7+
- Fixed unintended error behavior when the SDK is being shut down, if streaming is enabled. The symptom was that 1. the SDK could log a misleading message about a network error (in reality this was just the connection being deliberately closed) and 2. an uncaught exception could be thrown from the worker thread that managed that connection. The uncaught exception would be ignored in a default JVM configuration, but it could have more serious consequences in an application that had configured a default exception handler to be triggered by all uncaught exceptions.
8+
59
## [6.0.2] - 2023-01-04
610
### Fixed:
711
- Fixed vulnerability [CVE-2022-1471](https://nvd.nist.gov/vuln/detail/CVE-2022-1471) which could allow arbitrary code execution if using `FileDataSource` with a YAML file. (Thanks, [antonmos](https://github.com/launchdarkly/java-server-sdk/pull/289)!)

‎src/main/java/com/launchdarkly/sdk/server/StreamProcessor.java‎

Lines changed: 25 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -90,6 +90,7 @@ final class StreamProcessor implements DataSource {
9090
private final DataStoreStatusProvider.StatusListener statusListener;
9191
private volatile EventSource es;
9292
private final AtomicBoolean initialized = new AtomicBoolean(false);
93+
private final AtomicBoolean closed = new AtomicBoolean(false);
9394
private volatile long esStarted = 0;
9495
private volatile boolean lastStoreUpdateFailed = false;
9596
private final LDLogger logger;
@@ -182,11 +183,25 @@ public Future<Void> start() {
182183
// EventSource will start the stream connection either way, but if we called start(), it
183184
// would swallow any FaultEvents that happened during the initial conection attempt; we
184185
// want to know about those.
185-
for (StreamEvent event: es.anyEvents()) {
186-
if (!handleEvent(event, initFuture)) {
187-
// handleEvent returns false if we should fall through and end the thread
188-
break;
186+
try {
187+
for (StreamEvent event: es.anyEvents()) {
188+
if (!handleEvent(event, initFuture)) {
189+
// handleEvent returns false if we should fall through and end the thread
190+
break;
191+
}
192+
}
193+
} catch (Exception e) {
194+
// Any uncaught runtime exception at this point would be coming from es.anyEvents().
195+
// That's not expected-- all deliberate EventSource exceptions are checked exceptions.
196+
// So we have to assume something is wrong that we can't recover from at this point,
197+
// and just let the thread terminate. That's better than having the thread be killed
198+
// by an uncaught exception.
199+
if (closed.get()) {
200+
return; // ignore any exception that's just a side effect of stopping the EventSource
189201
}
202+
logger.error("Stream thread has ended due to unexpected exception: {}", LogValues.exceptionSummary(e));
203+
// deliberately log stacktrace at error level since this is an unusual circumstance
204+
logger.error(LogValues.exceptionTrace(e));
190205
}
191206
});
192207
thread.setName("LaunchDarkly-streaming");
@@ -205,6 +220,9 @@ private void recordStreamInit(boolean failed) {
205220

206221
@Override
207222
public void close() throws IOException {
223+
if (closed.getAndSet(true)) {
224+
return; // was already closed
225+
}
208226
logger.info("Closing LaunchDarkly StreamProcessor");
209227
if (statusListener != null) {
210228
dataSourceUpdates.getDataStoreStatusProvider().removeStatusListener(statusListener);
@@ -223,6 +241,9 @@ public boolean isInitialized() {
223241
// Handles a single StreamEvent and returns true if we should keep the stream alive,
224242
// or false if we should shut down permanently.
225243
private boolean handleEvent(StreamEvent event, CompletableFuture<Void> initFuture) {
244+
if (closed.get()) {
245+
return false;
246+
}
226247
logger.debug("Received StreamEvent: {}", event);
227248
if (event instanceof MessageEvent) {
228249
handleMessage((MessageEvent)event, initFuture);

‎src/test/java/com/launchdarkly/sdk/server/StreamProcessorTest.java‎

Lines changed: 37 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,8 @@
11
package com.launchdarkly.sdk.server;
22

33
import com.launchdarkly.eventsource.MessageEvent;
4+
import com.launchdarkly.logging.LDLogLevel;
5+
import com.launchdarkly.logging.LogCapture;
46
import com.launchdarkly.sdk.LDValue;
57
import com.launchdarkly.sdk.internal.events.DiagnosticStore;
68
import com.launchdarkly.sdk.server.DataModel.FeatureFlag;
@@ -55,6 +57,7 @@
5557
import static com.launchdarkly.sdk.server.TestComponents.dataSourceUpdates;
5658
import static com.launchdarkly.sdk.server.TestUtil.requireDataSourceStatus;
5759
import static com.launchdarkly.testhelpers.ConcurrentHelpers.assertFutureIsCompleted;
60+
import static com.launchdarkly.testhelpers.ConcurrentHelpers.assertNoMoreValues;
5861
import static org.hamcrest.MatcherAssert.assertThat;
5962
import static org.hamcrest.Matchers.allOf;
6063
import static org.hamcrest.Matchers.endsWith;
@@ -63,6 +66,7 @@
6366
import static org.hamcrest.Matchers.lessThanOrEqualTo;
6467
import static org.hamcrest.Matchers.notNullValue;
6568
import static org.hamcrest.Matchers.nullValue;
69+
import static org.hamcrest.Matchers.startsWith;
6670
import static org.junit.Assert.assertEquals;
6771
import static org.junit.Assert.assertFalse;
6872
import static org.junit.Assert.assertNotNull;
@@ -689,6 +693,39 @@ public void testSpecialHttpConfigurations() throws Exception {
689693
});
690694
}
691695

696+
@Test
697+
public void closingStreamProcessorDoesNotLogNetworkError() throws Exception {
698+
// This verifies that we're not generating misleading log output or status updates
699+
// due to simply seeing a broken connection when we have already decided to shut down.
700+
BlockingQueue<Status> statuses = new LinkedBlockingQueue<>();
701+
dataSourceUpdates.statusBroadcaster.register(statuses::add);
702+
703+
try (HttpServer server = HttpServer.start(streamResponse(EMPTY_DATA_EVENT))) {
704+
try (StreamProcessor sp = createStreamProcessor(null, server.getUri())) {
705+
sp.start();
706+
dataSourceUpdates.awaitInit();
707+
requireDataSourceStatus(statuses, State.VALID);
708+
709+
while (logCapture.awaitMessage(10) != null) {} // drain captured logs
710+
711+
sp.close();
712+
713+
requireDataSourceStatus(statuses, State.OFF); // should not see INTERRUPTED
714+
assertNoMoreValues(statuses, 100, TimeUnit.MILLISECONDS);
715+
716+
assertThat(logCapture.requireMessage(10).getText(), startsWith("Closing LaunchDarkly"));
717+
// There shouldn't be any other log output other than debugging
718+
for (;;) {
719+
LogCapture.Message message = logCapture.awaitMessage(10);
720+
if (message == null) {
721+
break;
722+
}
723+
assertThat(message.getLevel(), equalTo(LDLogLevel.DEBUG));
724+
}
725+
}
726+
}
727+
}
728+
692729
private void testUnrecoverableHttpError(int statusCode) throws Exception {
693730
Handler errorResp = Handlers.status(statusCode);
694731

0 commit comments

Comments
 (0)