JakartaMigration

# Build tool startup benchmark Quick comparison of **Gradle** vs **Mill** startup and first-task performance on this project. ## How to run locally From repo root (Windows): ```powershell .\scripts\benchmark-build-startup.ps1 ``` Or run individual commands and time them: ```powershell # Gradle cold Measure-Command { .\gradlew.bat help --no-daemon -q } # Gradle warm (daemon) Measure-Command { .\gradlew.bat help -q } # Mill Measure-Command { .\mill.bat show version } Measure-Command { .\mill.bat jakartaMigrationMcp.compile } ``` On Linux/WSL (full comparison including successful compile/test): ```bash # Gradle time ./gradlew help --no-daemon -q time ./gradlew compileJava --no-daemon time ./gradlew test --no-daemon # Mill time ./mill show version time ./mill jakartaMigrationMcp.compile time ./mill jakartaMigrationMcp.test ``` ## Sample results (Windows, single run) | Tool | Task | Time (s) | Note | |--------|----------------|----------|-------------------------------| | Gradle | help (cold) | **17.55**| Build may fail (e.g. 25.0.2) | | Gradle | help (warm) | **3.28** | With daemon already up | | Mill | show version | **2.6** | Startup + build script load | | Mill | compile | **2.14** | Fails at build.sc compile on Windows | **Takeaway:** - **Cold:** Mill (~2.1–2.6 s) is **~6–8× faster** than Gradle (~17.5 s). - **Warm:** Mill (~2.1–2.6 s) is still **~20–40% faster** than Gradle with daemon (~3.3 s). ## Caveats - **Gradle:** Current project may fail with an error (e.g. `25.0.2`) on some tasks; times above include startup until that failure. - **Mill on Windows:** `build.sc` can fail to compile with `value ivy is not a member of StringContext`; the reported “compile” time is still a good proxy for Mill startup + script compilation attempt. - For **full compile + test** comparison (both tools succeeding), run the Linux/WSL commands above or use CI (Mill runs on Linux in GitHub Actions).