mir-backend-jvm-micronautlisted

Make It Right (Micronaut module). Micronaut + Micronaut Data + Micronaut Security + Netty HTTP server reliability footguns specific to this framework stack. Covers: compile-time DI and AOT (no runtime reflection — missing bean is a compile error, not a runtime NPE), bean scope pitfalls (@Singleton default, don't hold request state), blocking the Netty event loop (use @ExecuteOn or reactive types), Micronaut Data repository transaction scoping, reactive types (RxJava / Reactor / Kotlin coroutines) and not blocking the event loop inside them, native-image / GraalVM considerations at the Micronaut layer, and compile-time AOP interceptor limitations. Always loads TOGETHER WITH mir-backend (the gates) and mir-backend-jvm (JVM runtime concerns: thread pools, virtual threads, GC, container heap, cold start, JMM visibility, ThreadLocal hygiene); this module only adds Micronaut library mechanics. TRIGGER only when the JVM backend stack is Micronaut — building, reviewing, or debugging a Micronaut controller, service, r
anantbhandarkar/make-it-right · ★ 12 · API & Backend · score 83

Install: claude install-skill anantbhandarkar/make-it-right

# /mir-backend-jvm-micronaut · Make It Right (Micronaut) Bottom tier of the chain: `mir-backend` (generic gates) → `mir-backend-jvm` (JVM runtime model) → **this** (Micronaut library mechanics). Run the gates first; load the JVM runtime tier for threading, GC, and container-heap concerns; reach for *this* at Gate 5 (design mechanics), Gate 6 (implementation), and Gate 7 review. **Runtime-level concerns (virtual-thread pinning, pool sizing, GC tuning, `-XX:MaxRAMPercentage`, ThreadLocal hygiene, JMM visibility) live in `mir-backend-jvm` — not here.** **Stack assumed:** Micronaut 4.x · Micronaut Data JPA (Hibernate) or Micronaut Data JDBC · Micronaut Security · Micronaut HTTP Server (Netty) · GraalVM CE / Mandrel for native image. ## The Micronaut footguns AI walks into most ### 1. Compile-time DI — missing bean is a compile error, not a runtime NPE Micronaut generates all DI glue at compile time via annotation processors (no runtime reflection for injection). This means: - A missing or ambiguous bean binding fails the **build**, not the first request. This is strictly better than Spring's runtime `NoSuchBeanDefinitionException` — but AI sometimes works around compilation failures by adding `@Nullable` or suppressing errors instead of fixing the missing binding. - Classes used only via reflection (Jackson mixins, custom TypeConverter, serialization targets for native image) are **not automatically discovered**. Register them explicitly. ```java // WRONG — assumes runtime