Repository navigation
Specification conformance, 51degrees-java, section 15 - #175
Merged
Merged
Conversation
The Pipeline specification's Properties page, at 51Degrees/specifications dbaf099c, gives the text of each missing property message in the table at pipeline-specification/features/properties.md lines 123 to 131, and says at line 136: "An implementation SHOULD give the same text, and MAY add to each link the campaign parameters its own repository requires." with lists "joined with a comma and no space" (line 135). https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/pipeline-specification/features/properties.md#L136 The tests compare each message, with the campaign parameters taken off its links, to the table's text for the reasons Java gives: data file upgrade with the tiers known and not known, excluded, product and property not accessible with the resource key, and unknown. Specification conformance item PR-17.
The upgrade message now says "your license or data file", ends with the type name and a link to the pricing page, and has a second form for data tiers that are not known, which an empty list of tiers gave as "with the license/data". The two resource key messages now point to the configurator, rather than to the resource keys explainer, in the table's words. The properties in a message are joined with a comma and no space. Each link keeps the repository's campaign parameters. Answers pipeline-specification/features/properties.md line 136 at 51Degrees/specifications dbaf099c, "An implementation SHOULD give the same text, and MAY add to each link the campaign parameters its own repository requires." The text for CloudRequestFailed (line 130) comes with that reason, under the order of deciding the reason. Specification conformance item PR-17.
The Pipeline specification's Properties page, at 51Degrees/specifications dbaf099c, says in pipeline-specification/features/properties.md line 143: "An implementation SHOULD decide the reason in this order, using the first that applies." and gives the order in lines 146 to 177. https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/pipeline-specification/features/properties.md#L143 The tests cover each step Java misses. Step 1, a property of a Cloud Aspect Engine whose Cloud Request Engine's request for the Flow Data failed is missing with CloudRequestFailed, naming the first engine of the Element Data, through the shared Cloud Aspect Engine and through the Device Detection cloud engine, with no response and with an error response. Step 3.1, the data tier is compared for an On-premise Engine only, so a Cloud Engine or another engine whose metadata lists the property gives no DataFileUpgradeRequired, and a Cloud Engine's property that is not available is excluded. Step 3.2, a Cloud Engine whose metadata is not loaded gives CloudRequestFailed when its request for the property list failed, and Unknown otherwise. Step 3.3, an On-premise Engine whose metadata does not list the property, but whose Element Data has an inherited typed accessor for it, gives DataFileUpgradeRequired with no data tiers, and a Cloud Engine does not. Step 2, when every engine gives Unknown the message names the last engine. The tests read each reason by its name, so they build before the CloudRequestFailed reason exists. Specification conformance item PR-18.
MissingPropertyServiceDefault now decides the reason for each engine in the order of pipeline-specification/features/properties.md lines 146 to 177 at 51Degrees/specifications dbaf099c: 1. A Cloud Aspect Engine records in its Element Data that the Cloud Request Engine's request for the Flow Data failed, which the request engine's error in the Flow Data shows, and a property the Element Data does not hold is then missing with the new CloudRequestFailed reason, naming the first engine. The Device Detection cloud engine, which processes the response itself, does the same, and no longer fails reading a missing response. 2. Each engine is checked in turn, as before. 3. For one engine, the data tier is compared for an On-premise Engine only, so a Cloud Engine no longer gives DataFileUpgradeRequired for a property its metadata lists. A Cloud Engine whose metadata is not loaded gives CloudRequestFailed when its request for the property list failed, and Unknown otherwise, where it gave ProductNotAccessibleWithResourceKey. An On-premise Engine whose metadata does not list the property, but whose Element Data has a typed accessor for it, counting inherited ones, gives DataFileUpgradeRequired with no data tiers. CloudAspectEngine gains propertyListRequestFailed, false by default, MissingPropertyService gains getCloudRequestFailedReason, and AspectDataBase gains setCloudRequestFailed. MissingPropertyServiceTests' upgrade case now uses an On-premise Engine, and its two Cloud Engine cases say their metadata is loaded, as the order requires. An On-premise Engine whose metadata gives a property no data tiers still gives DataFileUpgradeRequired, as before, which the open rule question on data tiers that are not known leaves as it is. Answers pipeline-specification/features/properties.md line 143, "An implementation SHOULD decide the reason in this order, using the first that applies." Specification conformance item PR-18.
…ngines Step 1 of the order of deciding the reason, in pipeline-specification/features/properties.md lines 146 to 150 at 51Degrees/specifications dbaf099c: "When a Cloud Aspect Engine populated the Element Data, and the Cloud Request Engine has no response for this Flow Data because its request to the remote server failed, the reason is `CloudRequestFailed`." https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/pipeline-specification/features/properties.md#L146 The IP Intelligence and location cloud engines read the response themselves, as the Device Detection cloud engine does, so the tests give each a Flow Data whose Cloud Request Engine failed, with no response and with an error response, and read a property. Specification conformance item PR-18, under line 143, "An implementation SHOULD decide the reason in this order, using the first that applies."
…n engines The IP Intelligence and location cloud engines now record in their Element Data that the Cloud Request Engine's request for the Flow Data failed, as the shared Cloud Aspect Engine and the Device Detection cloud engine do, so a property they do not hold is missing with the CloudRequestFailed reason. Each also stops, with the shared engine's warning, where the failed request left no response, rather than failing to read it. Answers step 1 of pipeline-specification/features/properties.md lines 146 to 150 at 51Degrees/specifications dbaf099c, under line 143, "An implementation SHOULD decide the reason in this order, using the first that applies." Specification conformance item PR-18.
…y name The Device Detection on-premise page, at 51Degrees/specifications dbaf099c, says in device-detection-specification/pipeline-elements/device-detection-on-premise.md line 205: "- Every read SHOULD use the position." and Java is not among the implementations its exception lets read by name (lines 214 to 264). https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/device-detection-specification/pipeline-elements/device-detection-on-premise.md#L205 DeviceDataHashDefault reads a value by its position, but first asks the native results whether they hold the property, by its name, which finds the name's position in native code on every read. The tests read IsMobile and BrowserName from an engine built from the free Lite data file, with native results that hold every property, and check that the read goes by the engine's position and that nothing is asked by name. Specification conformance item DD-15.
DeviceDataHashDefault.propertyIsAvailable asked each native results whether it holds the property, by name, before every read, which finds the name's position in native code. The engine already maps each name it was built with to its position, which is what the native lookup answers, so a property with a position is now held whenever there are results, and only a name with no position is still asked of native code. A read then goes by position alone. Answers device-detection-specification/pipeline-elements/device-detection-on-premise.md line 205 at 51Degrees/specifications dbaf099c, "Every read SHOULD use the position." Specification conformance item DD-15.
…y name The IP Intelligence on-premise page, at 51Degrees/specifications dbaf099c, says in ip-intelligence-specification/pipeline-elements/ip-intelligence-on-premise.md line 411: "- Every read SHOULD use the position." and Java is not among the implementations its exception lets read by name (lines 420 to 453). https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/ip-intelligence-specification/pipeline-elements/ip-intelligence-on-premise.md#L411 IPIntelligenceDataHashDefault reads a value by its position, but first asks the native results whether they hold the property, by its name, which finds the name's position in native code on every read. The test reads RegisteredName from an engine built from the free Lite data file, with native results that hold every property, and checks that the read goes by the engine's position and that nothing is asked by name. Specification conformance item IP-23.
IPIntelligenceDataHashDefault.propertyIsAvailable asked each native results whether it holds the property, by name, before every read, which finds the name's position in native code. The engine already maps each name it was built with to its position, which is what the native lookup answers, so a property with a position is now held whenever there are results, and only a name with no position is still asked of native code. A read then goes by position alone. Answers ip-intelligence-specification/pipeline-elements/ip-intelligence-on-premise.md line 411 at 51Degrees/specifications dbaf099c, "Every read SHOULD use the position." Specification conformance item IP-23.
The IP Intelligence required examples page, at 51Degrees/specifications dbaf099c, says in ip-intelligence-specification/required-examples.md line 247: "IP Intelligence examples SHOULD disable usage sharing for console examples but enable it for web examples, following the same pattern as Device Detection. Include comments explaining that production deployments should enable usage sharing to improve service quality." https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/ip-intelligence-specification/required-examples.md#L247 The Device Detection pattern it points to, in device-detection-specification/required-examples.md lines 201 to 211, keeps the Usage Sharing Element out of a console example's Pipeline, with a comment that in normal operation usage sharing should be enabled. The tests read each console example's source, as FindOutMoreCoverageTest does. Each builder that adds a Usage Sharing Element unless told not to, the 51Degrees Pipeline builder and the on-premise IP Intelligence builder, needs a setShareUsage(false), and each of those needs a comment before it saying to leave usage sharing on in production. Specification conformance item IP-35.
GettingStartedOnPrem, GettingStartedMixed and UpdateDataFile built their Pipelines with builders that add a Usage Sharing Element, so the console examples shared their evidence. Each now calls setShareUsage(false). The five console examples that already turned usage sharing off now say why, in the same words: the example's evidence is already known to 51Degrees, and in production usage sharing should be left on, which adds a Usage Sharing Element to the Pipeline and helps improve the service. The web examples are unchanged and still share usage. Answers ip-intelligence-specification/required-examples.md line 247 at 51Degrees/specifications dbaf099c, "IP Intelligence examples SHOULD disable usage sharing for console examples but enable it for web examples, following the same pattern as Device Detection. Include comments explaining that production deployments should enable usage sharing to improve service quality." Specification conformance item IP-35.
…mple The Device Detection required examples page, at 51Degrees/specifications dbaf099c, says in device-detection-specification/required-examples.md lines 338 to 344: "The data update service has a range of options available. Implementors SHOULD illustrate:" "- file system watcher - remote update polling - update on start-up - programmatic (non-automatic) update" https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/device-detection-specification/required-examples.md#L339 UpdateDataFile illustrates the first three, and no example illustrates a programmatic update. The tests run a ManualDataUpdate console example with the free Lite data file, and check that it detects the same User-Agents before and after it reloads the data file, and that, run with no settings, it says what the Enterprise data file adds, with the contact link, and ends with Find out more. Specification conformance item EX-11.
ManualDataUpdate builds an on-premise Device Detection Pipeline with automatic updates and the file system watcher off, detects two User-Agents, reloads the data file with the engine's refreshData, and detects them again, saying the data file's published date before and after. It runs with the free Lite data file and no licence key, shows the Hardware Vendor that only the Enterprise data file has, says that the Enterprise data file adds daily automatic updates with the tagged contact link, turns usage sharing off with the comment the console examples carry, and ends with Find out more. The examples readme lists it. Answers device-detection-specification/required-examples.md line 339 at 51Degrees/specifications dbaf099c, "Implementors SHOULD illustrate:", for "programmatic (non-automatic) update" (line 344). Specification conformance item EX-11.
The Pipeline specification's Exception handling page, at 51Degrees/specifications dbaf099c, says in pipeline-specification/features/exception-handling.md line 17: "In all cases, standard language features SHOULD be used to let IDEs and users know what exceptions/errors a particular function can throw." https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/pipeline-specification/features/exception-handling.md#L17 The FlowData interface declares no exception, yet process throws IllegalStateException for a Flow Data processed twice or a closed Pipeline and passes on the Pipeline's AggregateException, the getters throw before processing, and getAs and the typed getters throw PipelineDataException. The tests make each method throw and check that the FlowData interface method declares what it threw, by a type more specific than RuntimeException. Specification conformance item CO-24.
The FlowData interface now declares, in each method's throws clause and its Javadoc, the exceptions it throws: IllegalStateException from process for a Flow Data processed twice or a closed Pipeline, and the Pipeline's AggregateException when a Flow Element failed and process exceptions are not suppressed; IllegalStateException from get and getFromElement before processing; and PipelineDataException and ClassCastException from getAs and the typed getters. FlowDataDefault.get by data key now throws IllegalStateException before processing, as get by typed key does, rather than a plain RuntimeException, which IllegalStateException extends, so a caller catching RuntimeException still catches it. Answers pipeline-specification/features/exception-handling.md line 17 at 51Degrees/specifications dbaf099c, "In all cases, standard language features SHOULD be used to let IDEs and users know what exceptions/errors a particular function can throw." Specification conformance item CO-24.
…on jars The Device Detection packaging page, at 51Degrees/specifications dbaf099c, says in device-detection-specification/packaging-and-structure.md line 16: "Examples and tests SHOULD NOT be included in any published packages." https://github.com/51Degrees/specifications/blob/dbaf099c712e2bd3a2a80bfb4eaf284b079f05eb/device-detection-specification/packaging-and-structure.md#L16 The published device-detection.shared jar holds the test helpers fiftyone.devicedetection.shared.testhelpers.FileUtils and KeyUtils, as version 4.5.130 on Maven Central does. The test reads the main sources of the four published Device Detection modules, which are what their jars hold, for a test helper package or a test class. Specification conformance item DD-7.
…ed jar FileUtils and KeyUtils, in fiftyone.devicedetection.shared.testhelpers, move from device-detection.shared's main sources to its test sources, so the published jar no longer holds them and the tests of the other modules take them from the test jar they already depend on. The web on-premise getting started example now depends on that test jar for its test, as its siblings do. The examples used the two helpers in their own code, so the example helpers now hold what they used: DataFileHelper names the TAC, Enterprise and Lite data files and the evidence file and finds the first Hash data file, and KeyHelper reads a named key and says whether a value can be a key. MatchMetrics, PerformanceBenchmark, TacCloud and UpdateDataFile take these from the example helpers. Answers device-detection-specification/packaging-and-structure.md line 16 at 51Degrees/specifications dbaf099c, "Examples and tests SHOULD NOT be included in any published packages." The test jar device-detection.shared builds is still published, as each shared module's test jar is, which is left for a rule on how far the sentence reaches. Specification conformance item DD-7.
A patch release removes no public name. FileUtils and KeyUtils, in fiftyone.devicedetection.shared.testhelpers, are public classes that device-detection.shared has published up to 4.5.130, and the previous commit moved them into the test jar. The tests check that each is loaded from the module's main classes, not its test classes, and that each is deprecated. Follows up specification conformance item DD-7, device-detection-specification/packaging-and-structure.md line 16 at 51Degrees/specifications dbaf099c.
FileUtils and KeyUtils move back to device-detection.shared's main sources, as a patch release removes no public name, and are marked @deprecated, with Javadoc naming where example code now takes what they gave: the data file names from DataFileHelper and the key reading and checking from KeyHelper, both in device-detection-java-examples.shared. The build treats the use of a deprecated class as an error, so the tests take the same helpers from undeprecated copies in the test jar, TestFileUtils and TestKeyUtils, and TestHelpersStayPublishedTests finds the published classes by name. PublishedPackageTests now lets a published test helper stay only when it is deprecated. The test helpers can leave the published jar in a release that may remove a public name. Follows up specification conformance item DD-7, device-detection-specification/packaging-and-structure.md line 16 at 51Degrees/specifications dbaf099c.
This was referenced Nov 15, 2026
jwrosewell
marked this pull request as ready for review
November 16, 2026 04:43
jwrosewell
added a commit
that referenced
this pull request
Nov 16, 2026
rc already carries the missing Property messages and the order the reason is decided in, from #175, so the merge keeps rc's version of those six files and this draft's own implementation of them goes, along with its two test files for them, which rc covers with MissingPropertyMessageTests and MissingPropertyReasonTests.
jwrosewell
added a commit
that referenced
this pull request
Nov 16, 2026
rc already moved the example tests and the example data file helper off the packages' test helpers, in #175, so the merge keeps rc's version of those and of the TacCloud key check. The two engine files keep this draft's version, which adds the lock around every read of the native results and only takes the position shortcut with one set of results. The device detection cloud request failed test processes through the engine, as the IP Intelligence one does, now that the engine leaves processing to the shared logic.
jwrosewell
added a commit
that referenced
this pull request
Nov 16, 2026
rc moved the examples off the packages' test helpers in #175, so TacCloud and UpdateDataFile take the key helper from the examples' own shared module whilst keeping this draft's ending for an example run with no resource key.
This was referenced Nov 16, 2026
Draft
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
The Java packages differed from seven SHOULD sentences of the specification.
DataFileUpgradeRequiredfrom data tiers that apply only to a data file, a Cloud Engine whose metadata had not loaded gaveProductNotAccessibleWithResourceKey, and there was noCloudRequestFailedreason for a failed cloud request.FlowDatadeclared none of the exceptions its methods throw.device-detection.sharedjar held the test utility classesFileUtilsandKeyUtils.What the specification says
All at 51Degrees/specifications dbaf099c.
properties.md lines 133 to 137, with the table of messages at lines 123 to 131
properties.md lines 143 to 144, with the steps at lines 146 to 175
device-detection-on-premise.md line 205 and ip-intelligence-on-premise.md line 411
ip-intelligence-specification/required-examples.md line 247
device-detection-specification/required-examples.md lines 338 to 344
exception-handling.md lines 17 to 19
packaging-and-structure.md line 16
What changed
Links are to the head, cce971b.
Constantsgives the upgrade message with the pricing link, a second form for data tiers that are not known, the two resource key messages pointing at the configurator, and the newCloudRequestFailedmessage. Lists are joined with a comma and no space.MissingPropertyServiceDefaultfollows the steps. A Cloud Aspect Engine records in its Element Data that the Cloud Request Engine's request for this Flow Data failed (CloudAspectEngineBase.setCloudRequestFailed), and a missing Property then givesCloudRequestFailed, naming the first engine. The data tiers are compared for an On-premise Engine only. A Cloud Engine whose metadata has not loaded givesCloudRequestFailedwhen its request for the property list failed andUnknownotherwise. An On-premise Engine whose Element Data has a typed accessor for a Property its metadata does not list, inherited accessors counted, givesDataFileUpgradeRequiredwith no tiers. The IP Intelligence and location cloud engines, which read the response themselves, record a failed request too. When a failed request left no response, the Device Detection, IP Intelligence and location cloud engines now log the shared engine's warning and stop, where they failed reading it.DeviceDataHashDefault.propertyIsAvailableandIPIntelligenceDataHashDefault.propertyIsAvailableanswer from the engine's map of name to position, which holds what the native lookup would answer, and ask the native results by name only for a name with no position.GettingStartedOnPrem,GettingStartedMixedandUpdateDataFilecallsetShareUsage(false), and all eight console examples that turn usage sharing off say that in production it should be left on, as it adds a Usage Sharing Element that helps improve the service. The web examples are unchanged.ManualDataUpdateconsole example builds from the free Lite data file with automatic updates and the file watcher off, detects two User-Agents, reloads the data file withrefreshData, and detects them again. It then says that the Enterprise data file adds daily automatic updates, with a link to ask about it, and ends with Find out more. It is listed in the examples readme.FlowDatadeclares what it throws.FlowDatagainsthrowsclauses and@throwstags onprocess(IllegalStateException,AggregateException),getby key, Data Key and element (IllegalStateException), andgetAsand the typed getters (PipelineDataException,ClassCastException).FlowDataDefault.get(String)before processing now throwsIllegalStateException, where it threw a plainRuntimeException.FileUtilsandKeyUtilsstay indevice-detection.shared's main jar, as a patch release never removes a public name, and are marked@Deprecated, with Javadoc pointing atDataFileHelperandKeyHelperindevice-detection-java-examples.shared. The tests use new copies,TestFileUtilsandTestKeyUtils, in the test jar, and the examples take what they used fromDataFileHelperandKeyHelper.New API, all backward compatible:
MissingPropertyReason.CloudRequestFailed,CloudAspectEngine.propertyListRequestFailed()(a default method giving false),MissingPropertyService.getCloudRequestFailedReason(...)(a default method),AspectDataBase.setCloudRequestFailed(), and the protectedCloudAspectEngineBase.setCloudRequestFailed(...). No public name is removed or renamed.Before you upgrade
FlowDataDefault.get(String)before processing throwsIllegalStateException. It threw a plainRuntimeException.IllegalStateExceptionis a subclass ofRuntimeException, so a caller that catchesRuntimeExceptionsees no change.CloudRequestFailed. A Property missing because the cloud request failed now givesMissingPropertyReason.CloudRequestFailed, with the message "This is because an upstream cloud request failed, so no data was available to populate the property." Code that switches onMissingPropertyReasonshould handle it.DataFileUpgradeRequired, and one whose metadata has not loaded givesUnknownorCloudRequestFailedwhere it gaveProductNotAccessibleWithResourceKey. The messages are now the specification's text, so code that matches the old text needs the new one.FileUtilsandKeyUtilsare deprecated. They are test utility classes and will leave the published jar in a later release. UseDataFileHelperandKeyHelperfrom the examples instead.Tests
Each item has a test commit followed by its fix, and each test commit was run on a runner before its fix.
Red, run 43773859199 on f368fae and run 43775179085 on cce971b, each run's commit being that head with only a temporary workflow added. Each job checks out one test commit, builds the modules it tests with the modules they need on Java 17, runs only that commit's new tests, and fails.
MissingPropertyMessageTestsMissingPropertyReasonTests,CloudRequestFailedReasonTestsand Device Detection'sCloudRequestFailedTestsCloudRequestFailedTestsandGeoLocationCloudRequestFailedTestsPropertyNotAccessibleWithResourceKeyUsageSharingCoverageTestManualDataUpdateTest,ExamplesWithoutSettingsTestManualDataUpdateFlowDataThrowsTestsPublishedPackageTeststesthelpers/FileUtils.javaandKeyUtils.javain the published sourcesTestHelpersStayPublishedTestsThe two
ReadByPositionTests(5b7d01b for Device Detection and a22f3db for IP Intelligence) need the native engine and the free Lite data file, so they ran only in the full build.Green, the full build below, in which these tests ran on the head.
Build
Run 43775135770 built cce971b merged with
rc(e2472ee) on nine configurations, Ubuntu x64 with Java 8, 11, 17 and 21, Ubuntu arm64 with Java 17, and Windows with Java 8, 11, 17 and 21, and every job passed. Each of its tenSetting 'PullRequestSha'lines names cce971b, and its Configure job printed "Left out 6 macOS entries". Every test class in the red table passed in all nine jobs, and bothReadByPositionTests, which need the native engine and the free Lite data files, passed in each of them.macOS was left out of this draft's full build. The rc head is built with macOS before it merges into main.
Merging with #171 and #177
FlowData.java,DeviceDataHashDefault.javaandIPIntelligenceDataHashDefault.java, and a trial merge of the two heads conflicts in all three. Whichever of the two merges second must first mergerc, after the other has merged.IPIntelligenceCloudEngine.processEngine. Lock every read of IP Intelligence native results, type Latitude and Longitude as Float, and process cloud IP Intelligence with the shared engine logic #177 removes that method, so that the shared Cloud Aspect Engine logic processes cloud IP Intelligence, and this draft's fix (196b21f) edits it. Whichever of the two merges second should take Lock every read of IP Intelligence native results, type Latitude and Longitude as Float, and process cloud IP Intelligence with the shared engine logic #177's version, because the sharedprocessEngine, which this draft also changes, then does that work.What remains
device-detection.sharedjar still holdsFileUtilsandKeyUtils, deprecated, until a release may remove them.device-detection.shared,pipeline.common,pipeline.enginesandip-intelligence.sharedstill publish a-tests.jar, andpipeline.commonandip-intelligence.sharedkeep test utility classes in their main jars.Ready for review, head cce971b, build https://github.com/51Degrees/51degrees-java/actions/runs/43775135770