v2.1.0

General

• The copyright statement on all source code files has been updated.
• A bug affecting file watches on non-macOS platforms has been identified and fixed. The issue was caused by an erroneous generic cast unintentionally converting the watched path from the internal path type to a type alias of the same name that was exported from another file.
• Log messages from the GraalVM Polyglot engine emitted when the Konf JavaScript interpreter is not running on a GraalVM JDK have now been suppressed.
• A bug affecting YAML writer output for strings that contain : has been identified and fixed.
• As promised in the release notes for version 2.0.2, the Java requirement has been reduced from JDK 21 to JDK 11.
  • The only part of the project that still requires JDK 21 is the Maven Central publishing plugin for Gradle.
  • All other build phases in Gradle, and the Konf runtime, are able to run on any version of the JDK 11+.

Build

• Build system upgrades:
  • Gradle: 8.6 -> 8.7
  • Gradle Resolver: 0.5.0 -> 0.8.0
  • Gradle Develocity: 3.0 -> 3.17.3
  • Kotlin: 1.9.22 -> 2.0.0
  • Kotlin All-Open: 1.9.22 -> 2.0.0
  • Dokka: 1.9.10 -> 1.9.20
  • Kover: 0.7.5 -> 0.8.0
  • Central Portal Publisher: 1.1.1 -> 1.2.3
• Dependency upgrades:
  • KotlinX Coroutines Core: 1.7.3 -> 1.8.1
  • Apache Commons Text: 1.11.0 -> 1.12.0
  • Jackson Core, Modules, and DataTypes: 2.16.1 -> 2.17.1
  • JGit: 6.8.0 -> 6.9.0
• Test dependency upgrades:
  • JUnit Jupiter Params extension: 5.10.2
  • SLF4J Simple: 2.0.12 -> 2.0.13

Tests

• All tests have been completely rewritten to use standard JUnit Jupiter with Kotlin test assertions instead of the unmaintained JetBrains Spek engine and Hamkrest assertions.
• All test classes have been reorganized to separate "helper" classes that are only used by tests from the actual tests.
• Line separators now use macOS/Linux style on all platforms to avoid test failures on Windows due to inconsistent behavior between the different configuration language parsers and writers.
• As a side effect of the JUnit conversion, test execution is now fully parallelized and takes significantly less time to complete than the previous Spek-based tests.
• Test coverage has been improved from around 50% to 90%, as Kover was missing a significant number of tested lines of code run by the Spek engine.
• Several minor and/or platform-specific bugs discovered during test conversion have been fixed.
• The coverage report no longer erroneously includes content from the benchmark and snippet source sets.
• All test utilities are now documented correctly.

Availability

• Maven Central
• JitPack Snapshots