chore: modernize build tooling + fix MariaDB Lock Exception regression

[delebedev]

Context

Modernizes the build tooling and dependencies to current standards.

Changes

• Gradle 5.6 → 8.14.3
• Java 11 → 17
• Spring Boot 2.2.4 → 3.5.7
• Spring Framework 5.2.3 → 6.2.12
• Groovy 2.5 → 4.x
• Spock 1.3 → 2.4-M5
• Jakarta EE namespace migration
• Version 1.7.3 → 2.0.0 (breaking changes)

🔍 Deep Dive: MariaDB Lock Exception Fix

The Problem

After upgrading to Spring 6, the MariaDB integration test should only allow single side-effect for a given action started failing:

Expected: conflictErrorCount == 100
Actual:   conflictErrorCount == 0

The test verifies that concurrent requests for the same idempotent action properly detect lock conflicts and throw ConflictingActionException.

The Investigation 🕵️

Initial hypothesis was that MariaDB's FOR UPDATE NOWAIT semantics were broken. We tested:

• ❌ MariaDB 10.3 (original)
• ❌ MariaDB 10.6
• ❌ MariaDB 11.4

All versions showed the same behavior. NOWAIT was being accepted syntactically but the lock provider wasn't detecting conflicts.

Root Cause Found! 🎯

Spring Framework 6 changed the default SQL exception translator.

Spring Version	Default Translator	Behavior
Spring 5.x	SQLErrorCodeSQLExceptionTranslator	Maps vendor error codes (e.g., MySQL 1205) to Spring exceptions
Spring 6.x	SQLExceptionSubclassTranslator	Uses JDBC exception subclasses, ignores vendor codes

This meant MariaDB error 1205 (lock wait timeout) was being translated to UncategorizedSQLException instead of CannotAcquireLockException:

// Before fix (Spring 6 default):
CAUGHT: org.springframework.jdbc.UncategorizedSQLException
  SQL state [HY000]; error code [1205]; Lock wait timeout exceeded

// After fix:
CAUGHT: CannotAcquireLockException
  Lock wait timeout exceeded

The JdbcMariaDbLockProvider catches CannotAcquireLockException to detect lock conflicts - with the wrong exception type, it was throwing instead of returning Optional.empty().

The Fix

Explicitly configure SQLErrorCodeSQLExceptionTranslator in the lock provider:

public JdbcMariaDbLockProvider(JdbcTemplate jdbcTemplate) {
    // Spring 6 switched to SQLExceptionSubclassTranslator by default,
    // which doesn't use vendor-specific error codes. We need
    // SQLErrorCodeSQLExceptionTranslator to properly translate
    // MariaDB error 1205 (lock wait timeout) to CannotAcquireLockException.
    DataSource dataSource = jdbcTemplate.getDataSource();
    JdbcTemplate localJdbcTemplate = new JdbcTemplate(dataSource);
    if (dataSource != null) {
        localJdbcTemplate.setExceptionTranslator(
            new SQLErrorCodeSQLExceptionTranslator(dataSource));
    }
    this.namedParameterJdbcTemplate = new NamedParameterJdbcTemplate(localJdbcTemplate);
}

Migration Notes

Services using idempotence4j-mariadb with Spring Boot 3.x / Spring Framework 6.x should upgrade to version 2.0.0 to get the corrected exception translation behavior.

Impact of the bug: Wrong exception type (UncategorizedSQLException instead of ConflictingActionException). Lock acquisition and timeout behavior are NOT affected.

Test Plan

• All MariaDB integration tests pass
• All PostgreSQL integration tests pass
• Unit tests pass
• Build succeeds with Java 17

---

🤖 Investigated and fixed with Claude Code 🔬

[Copilot]

Pull request overview

Changes: Maintenance (1)

This PR modernizes the build tooling by upgrading Gradle to 8.14.3, Spring Boot to 3.5.7, and updating several key dependencies including Lombok, Groovy, and Spock. It also updates the Java target to 17 and adds JUnit Platform support for tests.

Changes:

• Upgraded Gradle wrapper from 7.3.3 to 8.14.3
• Updated Spring Boot from 2.2.4 to 3.5.7 and Spring JDBC from 4.3.18 to 6.2.12
• Migrated Groovy from 2.5.13 (Codehaus) to 4.0.29 (Apache) and Spock from 1.3 to 2.4-M5
• Updated Java compatibility from 1.9 to 17 and added JUnit Platform launcher for test execution
• Added docker-compose executable auto-detection for macOS Homebrew and Linux installations

Reviewed changes

Copilot reviewed 9 out of 11 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
libraries.gradle	Updated dependency versions for Lombok, Spring Boot, Spring JDBC, Groovy, and Spock; added groovy-json, groovy-sql, and junit_platform_launcher dependencies
build.gradle	Updated Gradle version, Java target to 17, Spring Boot plugin to 3.5.7, dependency-management plugin to 1.1.7, added useJUnitPlatform() for tests, and docker-compose executable auto-detection
idempotence4j-test/build.gradle	Added groovy-json and groovy-sql dependencies for Groovy test support
idempotence4j-postgres/build.gradle	Added groovy-sql dependency for integration tests
idempotence4j-mariadb/build.gradle	Added groovy-sql dependency for integration tests and fixed indentation
idempotence4j-benchmarks/build.gradle	Changed bootJar classifier property from deprecated 'classifier' to 'archiveClassifier'
gradlew	Updated Gradle wrapper script to the version shipped with Gradle 8.14.3 with improved POSIX compliance and argument handling
gradle/wrapper/gradle-wrapper.properties	Updated Gradle distribution URL from 7.3.3 to 8.14.3
gradle/wrapper/gradle-wrapper.jar	Updated Gradle wrapper JAR binary to match version 8.14.3

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[delebedev]

/request-review

[platon-github-app-production]

Unfortunately, the review request could not be sent to any teams:

• No teams were found to request a review. If a review has already started (someone has already approved or added at least one comment), you need to explicitly specify the team you want to request the review from: /request-review <team-1-name> <team-2-name>.

[mscott-tw]

Think it'd be simpler to make the constructor take a DataSource directly instead of a JdbcTemplate instance, the auto configuration bean method already takes in a data source instance so should be easy enough.

Technically it'd be a breaking change, though given JdbcMariaDbLockProvider is being created by auto config, it feels unlikely to cause real issues for users.

[delebedev]

I like this suggestion, breaking change could be justified too - it's 2.0.0 + this bugfix requires careful implementation

[brettyukich]

Here are the release notes for the SQLExceptionSubclassTranslator change

https://github.com/spring-projects/spring-framework/wiki/Spring-Framework-6.0-Release-Notes#data-access-and-transactions

[normanma-tw]

Since we're upgrading all the things and making breaking changes, how much trouble would it be to test this against both Spring Boot 3.5.x and Spring Boot 4? There were quite a few changes (also changes in Spring Framework 7) that are worth testing against.

[delebedev]

Since this doesn't use any of wise provided infra it's a bit harder but not impossible

[delebedev]

done!

[wise-github-bot-app]

The approval(s) from normanma-tw do(es)n't fullfill the approvers requirements because:

• The approver's cost centre, ENGPLA, maps to the PLATFORM business function. As the code that was changed is owned by ENGINEERING, this approval won't satisfy our separation of duties check. We'll need an additional approval from someone in ENGINEERING. This approval may still help satisfy other codeowner requirements.