Add serialization round-trip tests for event types

[darklight3it]

Issue #, if available:

Related to #594

Description of changes:

Add a serialization round-trip test that catches data loss and non-idempotent serialization across the events library. Previously, EventLoaderTest covered 27 event types with field-by-field assertions after a single deserialization pass. The new round-trip tests cover 66 event types (61 passing + 5 known failures), bringing coverage from ~40% to ~97% of the 68 classes in aws-lambda-java-events.

How the test works

LambdaEventAssert.assertSerializationRoundTrip performs two consecutive round-trips (JSON → POJO → JSON → POJO → JSON) and compares the original JSON tree against the final output. Explicit nulls are stripped from the original since the serializer uses Include.NON_NULL. A JsonNodeUtils helper produces a structured diff listing every path that changed.

Two passes catch both silently dropped fields and unstable serialization (output that changes across round-trips).

The test is split into three parameterized classes: SerializationRoundTripTest (registered events), UnregisteredEventSerializationRoundTripTest (unregistered input events), and ResponseEventSerializationRoundTripTest (response types). Each has passingCases() for events that survive the round-trip cleanly, and knownFailureCases() for events expected to fail (wrapped in assertThrows). If a known failure suddenly passes, the test fails with a message to move it to passingCases().

Why this approach

Field-by-field assertions in EventLoaderTest only verify the fields the test author remembered to check. The round-trip approach is exhaustive: any field present in the JSON fixture that gets lost or mutated during serialization is caught automatically. This kind of broad tests are the ones needed when updating jackson.

Fixture corrections (false negatives in existing fixtures)

Several existing JSON fixtures contained fields that the event models silently drop during deserialization. These were false negatives — the fixtures looked complete but the data was being lost without any test catching it. Fixed by removing the unsupported fields from fixtures so the round-trip test has a clean baseline:

Event	Missing Field(s)	Root Cause
APIGatewayProxyRequestEvent	requestContext.identity.clientCert	RequestIdentity has no clientCert field
APIGatewayV2HTTPEvent	requestContext.authentication	RequestContext has no authentication field
CloudFrontEvent	request.querystring (as string)	Model uses Map<String, List<String>>, not String
ConnectEvent	contactData.MediaStreams	ContactData has no mediaStreams field
S3Event	s3.object.urlDecodedKey, s3.object.versionId	Fields exist in S3EventNotification but missing from S3ObjectEntity
ActiveMQEvent	destination.physicalname	Case mismatch: JSON "physicalname" vs getter getPhysicalName() — no @JsonProperty mapping

Events requiring separate round-trip fixtures

Some events needed dedicated _roundtrip.json fixtures because the original fixtures contain values that Jackson leniently coerces during deserialization (e.g., numbers to strings), making the output structurally different from the input:

Event	Fixture	Reason
DynamodbEvent	dynamo_event_roundtrip.json	approximateCreationDateTime epoch seconds → Date formatting
KinesisEvent	kinesis_event_roundtrip.json	approximateArrivalTimestamp epoch seconds → Date formatting
KafkaEvent	kafka_event_roundtrip.json	Numeric values in Map<String, String> coerced to strings
LexEvent	lex_event_roundtrip.json	slots typed as Map<String, String> — numeric "Nights" coerced
MSKFirehoseEvent	msk_firehose_event_roundtrip.json	Numeric metadata values coerced to strings
RabbitMQEvent	rabbitmq_event_roundtrip.json	expiration (String→int), timestamp (long→formatted string)

Known failures (3 events)

These five events fail the round-trip and are captured in knownFailureCases():

1. APIGatewayV2CustomAuthorizerEvent — getTime() parses "12/Mar/2020:19:03:58 +0000" into a Joda DateTime via a custom formatter. Jackson serializes it as ISO-8601, which the formatter can't parse back on the second round-trip. The time field is effectively mandatory (getTime() throws NPE if null), and the date format change is inherent to how the serialization works — not a bug, just a lossy transformation that prevents a clean round-trip.

2. ActiveMQEvent (bug) — Destination.physicalName (camelCase) vs JSON "physicalname" (lowercase). Since ACCEPT_CASE_INSENSITIVE_PROPERTIES is disabled in JacksonFactory, the field is silently dropped during deserialization. Fix: add a mixin with @JsonProperty("physicalname").

3. S3ObjectLambdaEvent (bug) — Lombok generates getXAmzRequestId() for field xAmzRequestId. With USE_STD_BEAN_NAMING, Jackson derives the property name as "XAmzRequestId" (capital X), so the original "xAmzRequestId" key is silently dropped during deserialization. Fix: add @JsonProperty("xAmzRequestId") on the field or getter.

Target (OCI, Managed Runtime, both):

Both

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 65.38%. Comparing base (09c4b4e) to head (27a1641).

Additional details and impacted files

@@            Coverage Diff            @@
##               main     #598   +/-   ##
=========================================
  Coverage     65.38%   65.38%           
  Complexity      211      211           
=========================================
  Files            34       34           
  Lines           988      988           
  Branches        142      142           
=========================================
  Hits            646      646           
  Misses          290      290           
  Partials         52       52           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.