Files
ForgeMultipart/JAVA_MIGRATION.md
T

103 KiB
Raw Blame History

Java migration plan and audit findings

Status: living document
Initial audit: 2026-08-14
Audited revision: f10595d (master)

Decision summary

Converting ForgeMultipart to Java is realistic. It should be treated as a staged compatibility migration, not as an automatic source conversion or a ground-up rewrite.

The main difficulty is preserving behavior and compatibility, not translating Scala syntax. In particular, ForgeMultipart dynamically builds TileMultipart subclasses from combinations of part capabilities, and its published bytecode exposes Scala types and Scala compiler artifacts to other mods.

The current direction is:

  • Preserve observable behavior unless a divergence is an intentional, documented bug fix or performance fix.
  • Add clean Java APIs with equivalent features.
  • Keep deprecated compatibility bridges for the old Scala-facing API for at least the initial Java release.
  • Use the Hodgepodge/GTNH convention for any new Sponge Mixins: UniMixins plus explicit normal, early, and late registration where the target lifecycle requires it.
  • Allow GTNHLib as a dependency. Use fastutil selectively for measured primitive-collection or allocation problems, not as a blanket collection replacement.
  • Keep ForgeMultipart's runtime composite-tile generator separate from the Sponge Mixin decision until it is proven replaceable. They solve different problems.

No trustworthy tool will produce a maintainable Java port automatically. A decompiler can help recover control flow or compare generated bytecode, but its output should be used as a reference rather than committed as the implementation.

Current status

The binary and source-level consumer audits, low-coupling Phase 3 queue, audit-derived immediate compatibility gate, focused profile, first traversal optimization, complete redstone interaction helper unit, MicroRecipe port, and the first eight Java-trait ports are complete. TPartialOcclusionTile proved the no-field path, TSlottedTile proved state and lifecycle rewriting, TRedstoneTile removed the measured redstone allocation, TTileChangeTile exercised mutable state plus inherited access, TFluidHandlerTile preserved ordered Forge fluid distribution, and TIInventoryTile/JInventoryTile preserved AE2's direct rebuild linkage plus flattened sided inventory routing. TileMultipartClient and TRandomDisplayTickTile complete the built-in trait queue with their render-cache and random display-tick behavior intact. The first handler wave now also includes both mod entry points, both event handlers and the ForgeMicroblock registry packet handlers. All retain their exact runtime interfaces. See JAVA_MIGRATION_PROFILE.md.

That audit found and this branch has now repaired one existing regression: Schematica 1.12.6 reflects the original private MultiPartRegistry$ field codechicken$multipart$MultiPartRegistry$$typeMap and casts it to a Scala mutable map. The companion again exposes that exact private field as a live view of the canonical Java HashMap, and a test performs Schematica's reflective lookup. The remaining source-only member guards, TileMultipart ordering/move lifecycle, compact mixed-part NBT/description packet, generated Scala and Java traits, and a server-only pass-through interface are also frozen.

Phase 1 has not properly started: GTNHLib is not declared because no migration change currently needs it. Select a pack-aligned version only when a concrete port or measured data-structure change requires it. Phase 8 is answered and deferred.

The working rules established while converting, which apply to every remaining area:

  • Check both consumer audits before changing a compatibility surface. JAVA_MIGRATION_ABI_INVENTORY.md is the authority on binary linkage; JAVA_MIGRATION_CONSUMER_AUDIT.md is the authority on runtime behavior, data formats, reflection, and mixin field access.
  • Emit a Java default method only when no class in an implementor's superclass chain declares the same member. Otherwise leave it abstract and add explicit forwarders, because a superclass method always beats an interface default and the failure is silent.
  • Audit existing super call sites on every trait conversion, not just missing overrides. A class that already overrides a method and ends with super.x() can silently lose the trait's contribution while still compiling.
  • Introducing a trait default can change the declared members of concrete implementors that never carried their own override. Diff those classes too, not just the trait.

Scope and compatibility target

“Behave the same” means preserving the following unless a change is explicitly entered in a divergence log:

  • Multipart placement, removal, occlusion, collision, selection, ticking, redstone, rendering, and scheduled-tick behavior.
  • Client/server separation and Forge/FML lifecycle timing.
  • Part and tile registration order and lookup behavior.
  • Part identifiers, NBT keys and structures, saved-world loading, and conversion of existing multipart tiles.
  • Packet identifiers, field order, encoding, dispatch timing, and client update behavior.
  • Dynamically generated tile interfaces and capabilities expected by parts and integrations.
  • Public class names, method descriptors, static entry points, and reflective lookups that supported downstream mods use.
  • Behavior in both development and obfuscated production environments.

Allowed divergences are:

  • Confirmed bug fixes.
  • Removal of avoidable allocation, boxing, exception-based control flow, or other performance problems.
  • Better failure messages where success-path behavior is unchanged.

Each divergence should record the old behavior, new behavior, reason, compatibility impact, and how it was manually verified. Unintentional behavior changes are regressions even if the new implementation appears cleaner.

Java runtime policy

The main GTNH release targets modern Java across the Java 17–26 range. A Java 8 artifact must remain available for legacy launchers, with the same functional behavior even if it cannot match the modern runtime's performance.

  • Modern Java syntax and APIs may be used when the GTNH build's JVM Downgrader/desugaring path provides a verified Java 8 artifact.
  • Do not enable desugaring speculatively. Introduce it with the first source change that needs it, then verify both the modern and downgraded artifacts.
  • Java 8 startup, serialization, networking, registration, and gameplay semantics are compatibility requirements. Equal throughput or allocation behavior between Java 8 and modern Java is not required.
  • Validate routinely in the Java 8 deobfuscated Forge runner. Validate the packaged release in a representative modern-Java server before release, because the current GTNH deobfuscated runServer17/runServer21/runServer25 tasks do not provide a working dedicated-server path in this checkout.
  • Avoid modern runtime APIs whose downgrader stubs are unavailable or whose semantics materially differ on Java 8, especially in compatibility-critical and hot code.

Audit findings

Codebase size

At the audited revision:

Area Scala files Approximate Scala lines
Multipart core 38 4,076
Multipart ASM 8 1,933
Microblocks 28 3,013
Total 74 9,022

There are also 19 Java source files with approximately 1,148 lines. The existing README explicitly describes the API as Java-friendly, and the bundled Minecraft multipart implementation already serves as a Java consumer of it.

The current Scala source compiles successfully with Scala 2.11.5. This gives the migration a working reference implementation rather than requiring recovery of a broken codebase.

What Scala currently costs

The compiled mod contains 635 class files and is approximately 1.64 MiB before its Scala runtime dependency. Compiler-generated artifacts include:

  • 236 anonfun closure classes, approximately 448 KiB.
  • 39 trait helper $class files, approximately 98 KiB.
  • 121 object/companion-related class files, approximately 415 KiB.

These figures support a maintainability and allocation argument for Java, but not the claim that Scala wraps most methods in try/catch.

A scan of 5,126 Scala methods found 52 methods with exception tables and 53 total entries. The important groups were:

  • 33 catch-all/finally handlers, mostly compiler machinery for synchronization and lazy initialization.
  • 9 scala.runtime.NonLocalReturnControl handlers caused by return from inside closures.
  • 11 ordinary exception handlers covering Exception, ClassCastException, IllegalStateException, and IOException.

The 222 methods emitted from the Java sources had no exception-table entries in the same scan.

Exception-table metadata does not impose a meaningful steady-state cost merely by existing. The nine non-local-return sites are different: when the early return is taken, Scala implements it by throwing and catching NonLocalReturnControl. Those sites are worth removing, but they do not justify assuming exception overhead across the whole mod.

More credible performance risks

The source and bytecode audit identified these allocation patterns; the named ports have now resolved them:

  • The earlier Java TileMultipart.updateEntity() traversed through operate(Function1), whose parts() helper materialized an ArrayList, backing array, Scala iterator and Java-conversion wrappers per call.
  • The former Scala TRedstoneTile queries created closures and scala.runtime.IntRef boxes; the Java trait now walks the normal immutable list without allocation while preserving an arbitrary-Seq fallback.
  • Several Scala 2.11 collection, range, and JavaConversions paths create closure objects, boxed primitives, iterators, or temporary collections.
  • The former TSlottedTile.canAddPart and MicroRecipe scans used exception-based non-local returns; both are now ordinary Java loops.
  • Rendering code in TileMultipart already uses cached arrays and while loops in important paths. It should be preserved as evidence of deliberate optimization, not mechanically rewritten into streams or higher-level collection pipelines.

The focused Forge/JFR baseline confirmed both risks. With eight parts, updateEntity and operate allocated about 184 bytes per call, while a three-query redstone iteration allocated 80.5 bytes. The completed ports reduce the normal paths to effectively zero measured allocation; these remain focused measurements, not whole-pack TPS. See JAVA_MIGRATION_PROFILE.md.

Runtime code generation is architectural, not incidental

MultipartGenerator selects a tile class from the interfaces required by the parts in a block. ASMMixinFactory caches generated constructors by a trait bit set, so class generation is primarily a registration/first-combination cost rather than a per-tick cost.

ASMMixinCompiler already supports two inputs:

  • registerScalaTrait, which reads Scala signature data and understands Scala trait output.
  • registerJavaTrait, which transforms a concrete Java mixin class into the interface/static-helper form used by the generator.

The Java path rejects Java interfaces, inner classes, abstract classes, static initializers, and constructors with arguments. Those restrictions must be accounted for when converting built-in tile traits, but the path proves that Java implementations can participate in the current composition system.

MultipartGenerator.registerPassThroughInterface also explicitly supports interfaces that generated tiles must expose. This is part of the compatibility surface and must not be lost in a static subclass rewrite.

The least risky plan is therefore to keep the current composition model during the first Java migration. Port or redesign its ASM implementation only after the rest of the behavior is stable.

Public ABI exposes Scala

Several public descriptors contain Scala runtime types. Examples include:

  • TileMultipart.partList exposing scala.collection.Seq through generated accessors.
  • TileMultipart.operate accepting scala.Function1.
  • Occlusion methods accepting Scala Seq values.
  • Tile conversion helpers returning scala.Tuple2.
  • MultiPartRegistry overloads accepting Scala Function2 or Seq, and returning Seq.
  • Scala object singletons (MODULE$), companion/static forwarders, and trait $class helpers.

Changing these descriptors directly to java.util or java.util.function types can cause NoSuchMethodError, NoClassDefFoundError, or reflective integration failures in already compiled mods. Source compatibility alone is not sufficient.

This means “all source is Java” and “the Scala runtime is gone” are separate milestones. Deprecated bridges may still require Scala library types even after the implementation is Java.

API migration design

Create a Java-first API with the same capabilities, then make legacy entry points delegate into it. Delegation should have one direction: deprecated Scala bridge to canonical Java implementation. Avoid two independently maintained implementations.

Candidate mappings, with final names and mutability contracts to be decided from existing behavior:

Existing Scala-facing shape Java-first equivalent
Seq<TMultiPart> List<TMultiPart> or a purpose-specific read-only view
Function1<TMultiPart, Unit> Consumer<TMultiPart> or direct indexed traversal methods
Function2 factory callbacks A named Java functional interface with domain-specific arguments and return type
Tuple2<TileMultipart, Boolean> A small immutable Java result class with named accessors
Scala collection overloads in registries Collection, List, arrays, or Java Iterable, according to actual ordering and mutation semantics

Bridge rules:

  • Preserve old JVM descriptors, not merely old source names.
  • Mark source-visible legacy APIs deprecated and point their documentation to the Java replacement.
  • Preserve ordering, nullability, mutability, exception, and callback timing semantics.
  • Keep legacy singleton/static entry points when downstream bytecode may reference them.
  • Do not leak new Scala types into the Java API.
  • Do not remove Scala library dependencies until an ABI inventory shows that no retained bridge needs them.
  • Use javap plus an API/ABI comparison tool such as japicmp or Revapi against the reference dev jar at every public-API phase.

Before implementing bridges, inventory real consumers in the target modpack. Some Scala compiler artifacts may be technically public but unused; others may be load-bearing despite looking internal.

Mixin strategy

Follow the pattern currently used by Hodgepodge for fixed class transformations:

  • Set the GTNH build convention's usesMixins = true, which supplies UniMixins integration and declares UniMixins as a required dependency.
  • Use a normal/pre-init mixin configuration for transformations that do not require special early or optional-mod timing.
  • Use an IEarlyMixinLoader configuration only for targets that must be selected before ordinary mod loading.
  • Use an @LateMixin/ILateMixinLoader configuration for optional-mod targets whose presence is known later.
  • Centralize selection through the GTNHMixins builder pattern, including side, loaded-coremod, loaded-mod, and configuration gates.
  • Keep early and late lists minimal. A mixin should not be early or late merely because those mechanisms are available.
  • Prefer Sponge Mixins for fixed transformations of known classes. Retain raw ASM only where a mixin cannot express the required behavior safely.

The inspected Hodgepodge reference uses:

  • HodgepodgeCore implements IFMLLoadingPlugin, IEarlyMixinLoader and delegates early selection to IMixins.getEarlyMixins(...).
  • HodgepodgeLateMixins, annotated with @LateMixin, implementing ILateMixinLoader and delegating late selection to IMixins.getLateMixins(...).
  • Separate normal, early, and late JSON configurations.
  • Explicit dependencies on GTNHMixins and UniMixins.

Important distinction: Sponge/UniMixins modifies known classes during class loading. ForgeMultipart's current generator creates different tile classes for runtime combinations of part interfaces. The Hodgepodge registration pattern should be used for ordinary patches introduced by this port, but UniMixins is not assumed to replace the composite-tile generator.

GTNHLib and fastutil

GTNHLib is an acceptable dependency for the port. The inspected GTNHLib checkout declares fastutil 8.5.18 as compileOnlyApi and uses its dependency-loader/GTNHExtLib arrangement for external runtime libraries. Hodgepodge consumes GTNHLib as an API dependency and imports fastutil through that environment.

Integration requirements:

  • Select the GTNHLib version aligned with the target pack/release train rather than copying Hodgepodge's current version blindly.
  • Verify both compile and production runtime resolution in ForgeMultipart's actual build after adding the dependency.
  • Add the Forge mod dependency declaration with a compatible version range.
  • Use primitive fastutil collections where they remove boxing or object-key allocation in a demonstrated hot path.
  • Prefer JDK collections for small, cold, or API-facing collections unless fastutil provides a concrete benefit.
  • Do not expose fastutil implementation classes from the stable public API without a strong reason; ordinary Java interfaces keep consumers decoupled.

Candidate areas to measure include redstone query state, coordinate-keyed caches, scheduler structures, registry lookups, and render/tile scratch collections. A primitive map is only useful if the key can be represented without first allocating a wrapper such as BlockCoord.

GTNHLib's ASM utilities may also reduce low-level boilerplate when the runtime generator is eventually ported. Evaluate them against the generator's exact needs instead of changing ASM APIs during the earlier source migration.

Migration phases

Every implementation phase follows the same test-first gate:

  1. Write characterization tests against the current Scala implementation for the area about to change.
  2. Run them on the untouched implementation and retain any golden fixtures they produce.
  3. Commit the tests separately from the port so the baseline is reviewable.
  4. Replace the implementation while leaving the tests unchanged.
  5. Require the same tests to pass before moving to the next area.

If a characterization test captures a confirmed bug that the port intentionally fixes, record it in the divergence log and change that expectation in a separate, explicit bug-fix step. This prevents accidental changes from being presented as fixes.

Phase 0 — Freeze and measure the reference behavior

  • Produce and retain a reference dev jar from f10595d or the selected migration base.
  • Dump the public/protected JVM API and important generated tile class shapes.
  • Inventory downstream mods that compile against or reflect into ForgeMultipart. See JAVA_MIGRATION_ABI_INVENTORY.md and JAVA_MIGRATION_CONSUMER_AUDIT.md.
  • Record source-level consumers of part IDs, NBT, packets, registration order, reflection, and lifecycle behavior.
  • Freeze representative part IDs, NBT trees, packet bytes, registration order, and lifecycle behavior as tests.
  • Add a Java-8-compatible JUnit/Jupiter setup to the existing Gradle test task.
  • Add a separate Forge integration-test source set/runner for behavior that cannot execute in a plain JVM.
  • Write the first characterization suite against the untouched Scala implementation.
  • Create a remaining manual compatibility checklist for rendering, input, and other behavior that cannot be asserted reliably by the automated harness. See JAVA_MIGRATION_MANUAL_CHECKS.md.
  • Capture representative CPU and allocation profiles before optimization work. See JAVA_MIGRATION_PROFILE.md.
  • Start a divergence log.

Exit condition: there is a reproducible behavior and ABI baseline, the test layers run in CI or an equivalent repeatable command, and the initial suite passes against the Scala implementation.

Phase 1 — Add infrastructure only when first needed

  • Add GTNHLib with a pack-aligned version when the first Java migration or data-structure change uses it.
  • Enable UniMixins when the first fixed class transformation is introduced.
  • Copy the Hodgepodge loader/config structure selectively: normal by default, early/late only for demonstrated lifecycle needs.
  • Confirm client and dedicated-server startup after each dependency or loader change.

Exit condition: dependencies and loaders are present with no unintended gameplay behavior change.

Phase 2 — Establish the Java-first API and compatibility bridges

  • Inventory every public Scala descriptor and compiler-generated entry point used by downstream code.
  • Add ABI snapshots and compiled consumer fixtures before changing public descriptors.
  • Define Java collection, callback, and result types with documented behavior equivalent to the current API.
  • Implement the canonical behavior once in Java.
  • Keep deprecated Scala signatures and singleton/static entry points as thin adapters.
  • Compare the produced jar with the reference API and verify representative Java and Scala consumers can load.

Status: the two open items are open only because conversion is still in progress, not because the approach is undecided. Every converted area already implements its behavior once in Java with the Scala entry points delegating inwards, and every public boundary is diffed against the reference dev jar. The collection and callback decisions are made per area against the inventory rather than up front: readIDMap moved to java.util.List because nothing links against it, while getIdMap keeps its scala.Tuple2 array because extrautilities does.

Exit condition: new code can use a clean Java API while existing supported binaries still link through the deprecated surface.

Phase 3 — Convert low-coupling code

  • Characterize each related leaf group before converting it.
  • Convert the identified constants, value holders, simple utilities, registries, packet data helpers, and leaf classes.
  • Replace JavaConversions in converted files with explicit Java collections or controlled adapters.
  • Preserve iteration order and null behavior.
  • Keep mixed Scala/Java compilation working throughout the phase.

Status: complete. The low-risk queue is empty. All eight load-bearing $class helpers are Java with their bridges retained, as are both registries, the central part/tile types, scheduler/helper utilities, BlockMultipart, and MultipartRenderer. The exact completed list is maintained in JAVA_MIGRATION_HANDOFF.md.

Exit condition: leaf code is Java and no longer generates avoidable Scala closures or collection adapters.

Immediate compatibility gate — complete

Do this before another source conversion:

  • Restore Schematica's reflective MultiPartRegistry$ type-map view with its exact field name and Scala mutable-map shape, backed by the canonical Java registry state; add a regression test that performs Schematica's actual lookup.
  • Add exact member-level guards for GuideNH's BlockMicroMaterial.block/meta mixin fields, Et Futurum's mutable ButtonPart.metaSideMap/sideMetaMap, Iguana's ItemSaw.harvestLevel, and Galacticraft's name-only registerMaterial lookup.
  • Freeze TileMultipart order and lifecycle observations: parts/id NBT, partList/jPartList, partMap, slot rebinding, add/remove/replace callback order, moving the live tile, and onMoved.
  • Add compact mixed-part NBT and description-packet fixtures. The existing torch/button parts provide the needed distinct IDs, order, slots and payload without adding downstream mods or a test-only part registry.
  • Freeze one generated Scala trait, one Java trait, and representative pass-through interfaces before changing built-in tile traits or either generator.

Exit condition: every source-only constraint that can silently fail has an automated structural/behavioral guard, and the already-ported registry again supports Schematica.

Phase 4 — Convert measured hot paths

  • Characterize result ordering, callback timing, early exits, and edge cases before changing traversal code.
  • Rewrite TileMultipart update/operate traversal without per-call collection allocation.
  • Rewrite redstone queries without closure allocation or IntRef boxing.
  • Replace non-local returns in recipe scans with ordinary Java control flow.
  • Replace the remaining non-local return in the generated TSlottedTile slot scan.
  • Keep fastutil out of these paths because profiling did not justify another dependency or data structure.
  • Re-profile the same scenarios and record both improvements and regressions.

Status: complete for the measured targets. operate now preserves its captured-sequence mutation semantics without collection materialization, and the six-interface IRedstonePart/RedstoneInteractions$ unit is ported with its class list and descriptors intact. The comparison proved its allocation was owned by the generated TRedstoneTile, which is now Java and allocation-free on the normal immutable-list path. MicroRecipe is also Java: all five recipe forms and precedence are characterized, its published Scala façade remains, and its closure-backed scans and non-local returns are gone. TPartialOcclusionTile completed the simple no-field Java-trait pilot, and TSlottedTile completed the stateful field/lifecycle checkpoint while removing the remaining slot-scan closures and exception-backed return. The paired TRedstoneTile run removed 80.5 B per three-query iteration and improved throughput by 20.3% with an identical checksum.

Exit condition: identified steady-state Scala allocation sites are removed with compatible results.

Phase 5 — Convert built-in tile traits through the existing Java path

  • Characterize the direct class shape, generated interface, super dispatch, behavior, and class caching of a no-field pilot.
  • Port one simple built-in trait as a pilot using registerJavaTrait.
  • Verify generated field accessors, initialization, copying, lifecycle callbacks, super bridges, and class caching on the stateful TSlottedTile checkpoint.
  • Characterize ProjectRed/Extra Utilities redstone calls and port TRedstoneTile with an immediate paired profile.
  • Convert remaining traits in small related groups.
  • Document or deliberately relax the Java trait restrictions only when a real trait requires it.
  • Preserve pass-through interface behavior.
  • Keep the completed OpenComputers TSlottedTile.v_partMap mutation/rebinding case green and cover ProjectRed/ Extra Utilities behavior before TRedstoneTile.
  • Explicitly cover AE2's TIInventoryTile.rebuildSlotMap before the inventory trait.

Exit condition: built-in behavior is implemented in Java while the established runtime composition mechanism remains stable.

Status: complete. All built-in tile traits are Java. The client pair required narrowly extending registerJavaTrait for a registered Java trait extending another registered Java trait, explicit Java field accessors, and transient runtime-only fields that must not participate in generated copyFrom. The generated runtime interfaces and class cache remain reference-identical; pass-through-interface coverage remains green.

Phase 6 — Convert multipart core and microblocks

  • Add characterization coverage for each subsystem immediately before its conversion.
  • Convert the central tile, part, registry, placement, rendering, networking, scheduler, and microblock code in dependency order.
  • Treat NBT, packets, registration, client/server separation, and rendering state as compatibility-critical.
  • Keep recipe scan order and output semantics stable while replacing exception-based early exits.
  • Check source and binary compatibility at each public boundary.

Exit condition: normal runtime implementation is Java; any remaining Scala is isolated to compatibility or ASM support.

Phase 7 — Port the ASM subsystem last

  • Freeze structural and behavioral fixtures for generated classes before changing the compiler.
  • Translate the existing compiler/generator into readable Java while preserving emitted class shapes and cache keys.
  • Evaluate GTNHLib ASM helpers where they reduce boilerplate without altering semantics.
  • Keep Scala-trait registration while deprecated bridges or supported external Scala traits require it.
  • Allow abstract Java mixins whose base type is abstract, skipping their abstract members instead of rejecting the class. Required before any Java mixin over Microblock is possible. See Phase 10.
  • Implement @SideOnly member stripping on the Java mixin path, matching what listSideOnly does for Scala signatures. Required before a client-only Java mixin method is safe on a dedicated server. See Phase 10.
  • Run ProjectRed's LightMicroblock external Scala-trait fixture and pass-through-interface fixtures against the ported generator.
  • Decide separately whether a future generator design can avoid custom bytecode generation. Do not couple that experiment to the source-language port.

Exit condition: the composition subsystem is maintainable Java with behavior and generated ABI verified against the reference.

Phase 8 — Decide on Scala runtime removal

Status: decided and deferred. The inventory showed four shipping mods linking against trait $class helpers and companion singletons, and ProjRed registering its own Scala trait through MicroblockGenerator.registerTrait. Scala removal is therefore not achievable for the first Java release, and Java-maintainable source is the milestone instead. Revisit only in a release allowed to break that ABI.

Complete Scala removal is not the goal in its own right. scala-library stays in the pack regardless of what FMP does, because OpenComputers (843 Scala files) and ProjectRed (171) are Scala end to end and are not being ported. The value of removing Scala from FMP is FMP's own maintainability and hot-path allocation, not a pack-level dependency win. Phase 10 is what actually retires individual items on this list; each Scala shape removed here needs a corresponding released consumer version there.

  • Measure actual downstream use of Scala descriptors, Scala traits, object singletons, and $class helpers.
  • Publish and complete a deprecation window if compatibility policy permits removal.
  • Remove Scala bridges, Scala signature decoding, and the Scala compiler/runtime dependency only in a release allowed to break that ABI.
  • Confirm no Scala types remain in published descriptors or runtime loading paths.

Exit condition: Scala removal is either completed deliberately or explicitly deferred as the cost of binary compatibility.

Phase 9 — Deprecate the Scala-shaped API and mark the internal boundary

Status: planned. This phase does not redesign the API. A surface audit against all 28 consumer checkouts found the published API is already largely Java-shaped: TMultiPart is an ordinary abstract Java class with conventional lifecycle, render, NBT, and packet methods, and IPartFactory2, IPartConverter, IMicroMaterial, PartMap, RedstoneInteractions, and the occlusion interfaces need no replacement. The remaining problems are a short list of signatures that still carry Scala types, and a boundary problem that is larger than it looks.

The governing rule is: deprecate only where a replacement already exists or is added in the same change. A deprecation with nowhere to go is noise a consumer cannot act on, and these bridges are being retained deliberately for years. Adding the Java sibling and marking the Scala-shaped entry deprecated is one change, not two.

Sequencing rule: do this work in-file, during the remaining port, not as a later sweep. The microblock package and the ASM package are still Scala and will each be opened exactly once. Adding a sibling, a @Deprecated, and an internal-marker javadoc while already editing the file is free; a separate pass later means re-reading every ported file.

9.1 — Scala-typed signatures

Each row keeps its existing descriptor for binary compatibility and gains a Java-shaped sibling. Two rows already have their sibling and only need the annotation and javadoc.

Deprecate Java-shaped replacement Notes
TileMultipart.partList(): scala.collection.Seq jPartList(): java.util.List Sibling already exists
MultiPartRegistry.registerParts(IPartFactory2, scala.collection.Seq) registerParts(IPartFactory2, String...) Sibling already exists
MicroMaterialRegistry.getIdMap(): scala.Tuple2[] materialCount(): int plus existing materialName(int) and getMaterial(int) The array index is the material ID, so no pair or map type is needed
TileMultipart.operate(scala.Function1<TMultiPart, BoxedUnit>) forEachPart(java.util.function.Consumer<TMultiPart>) Preserve the existing skip-unbound-part behavior
TileMultipart.occlusionTest(scala.collection.Seq, TMultiPart) occlusionTest(Collection<TMultiPart>, TMultiPart)
TileMultipart.loadParts(scala.collection.Iterable) loadParts(Collection<TMultiPart>) Schematica and GuideNH reflect the Scala descriptor; see Phase 10
TileMultipart.partList_$eq(scala.collection.Seq) setPartList(List<TMultiPart>) GuideNH reflects the _$eq name; see Phase 10
TileMultipart.renderID() / renderID_$eq(int) getRenderID() / setRenderID(int)
TileMultipart.getOrConvertTile2(): scala.Tuple2<TileMultipart, Object> Small immutable result type with named accessors No consumer in the audited set calls this

Explicitly not renamed: TMultiPart.world(), x(), y(), z(), and tile(). These carry Scala accessor naming but no Scala type, so renaming is cosmetic churn across 27 consumers with no compatibility or performance payoff. tile() is additionally unsafe to rename toward getTile(), which already exists with a different return type; a same-name overload there invites a silent wrong-overload bind.

  • Add the six missing Java-shaped siblings and route each Scala-shaped entry through it in one direction only.
  • Mark all nine rows @Deprecated with javadoc naming the replacement.
  • Confirm every original descriptor still exists in the ABI fixture after the change.

9.2 — Mark the internal boundary

Scala's private[multipart] compiles to public. As a result the published surface currently advertises implementation hooks as though they were API, and a consumer author cannot distinguish addPart from addPart_do. This is a more practical API defect than the Scala types above, and the fix is javadoc only: no rename, no descriptor change, no deprecation, and no new dependency for a marker annotation.

Verified against all 28 consumer checkouts as having zero external callers:

TileMultipart.addPart_impl, addPart_do, remPart_impl, writeAddPart, partAdded, partRemoved, from, copyFrom, loadFrom, setValid, getOrConvertTile2, operate; MicroMaterialRegistry.setupIDMap, calcMaxCuttingStrength, loadIcons, writeIDMap, readIDMap.

Two similar-looking members are externally load-bearing and must not be marked internal:

Member Consumer Call site
TileMultipart.bindPart OpenComputers li/cil/oc/integration/fmp/PrintPart.scala:171
TileMultipart.internalPartChange ProjectRed mrtjp/projectred/integration/gatepartrs.scala:74
  • Add an internal-marker javadoc line to each zero-caller member listed above.
  • Leave bindPart and internalPartChange documented as supported API and add them to the consumer audit's cross-cutting map.
  • Do not add an annotations dependency for this; javadoc is sufficient and changes no descriptor.

Exit condition: every Scala-typed public entry has a documented Java-shaped replacement and a deprecation pointing at it, the implementation-hook members are documented as internal, and the ABI fixture is unchanged.

Phase 10 — Upstream consumer cleanup

Status: planned, unsequenced. This phase does not block any other phase, and parts of it can start immediately, including before the Java port merges. Its purpose is to stop consumer-side hacks from dictating FMP's internal design.

Several consumers reach FMP through private fields, Scala-mangled names, name-only reflection, or third-party Scala traits. Today each one forces FMP to preserve an internal shape it would otherwise be free to change. Because all 27 consumers are GTNewHorizons forks, every one of these is patchable upstream. Extra Utilities is the only exception; it is decompiled-only and is being replaced by UtilitiesInExcess, so it constrains FMP until it is retired from the pack rather than being fixed.

The standard pattern is a three-step ratchet, and only the first step is on FMP's critical path:

  1. FMP adds a public, supported equivalent. Additive, no ABI break, safe to land at any time.
  2. The consumer is patched to use it and released.
  3. FMP drops the private shape in a release allowed to break that ABI. Feeds Phase 8.

Cleanup targets

Consumer What FMP is currently forced to preserve Upstream fix Can start
ProjectRed ScalaSignature.scala (369 lines) and ByteCodecs.scala (211 lines) exist for exactly one external caller: MicroblockGenerator.registerTrait(classOf[LightMicroblock]) Rewrite the ~60-line LightMicroblock trait as a Java mixin registered through the registerJavaTrait path Blocked on FMP; see the prerequisites below
Schematica Private Scala-mangled field codechicken$multipart$MultiPartRegistry$$typeMap, cast to scala.collection.mutable.Map FMP exposes a supported registry lookup; Schematica uses it instead of the field After step 1
GuideNH Companion-only MultipartGenerator$.MODULE$.generateCompositeTile and MicroblockGenerator$.create; partList_$eq(scala.collection.Seq); mixin into private BlockMicroMaterial.block and .meta FMP exposes supported static entry points and setPartList(List); GuideNH targets those and public material accessors After step 1 and Phase 9.1
Et Futurum Requiem Mutable static int[] ButtonPart.metaSideMap and sideMetaMap must stay public and mutable FMP exposes a supported orientation-override API; Et Futurum uses it After step 1
IguanaTweaksTConstruct Private ItemSaw.harvestLevel field name and type FMP exposes a supported harvest-level setter; Iguana uses it After step 1
Galacticraft Selects the first public method named registerMaterial without checking its signature, so FMP cannot add any overload of that name Make the reflection check the parameter types Immediately; independent of the port
UtilitiesInExcess MicroMaterialRegistry.getIdMap(): scala.Tuple2[] Both call sites read only the name, so materialCount() from Phase 9.1 plus the existing materialName(int) covers them; no material object is needed After Phase 9.1, before it enters the pack
UtilitiesInExcess Not an FMP constraint, but a shipping hazard: the server factory reads NBT key material while MaterialBasedPart.save writes mat Fix on the UtilitiesInExcess side; FMP must not special-case it Immediately; no FMP change needed

Galacticraft's row is the only one that can be done upstream today with no FMP change at all. It is a small signature check that removes a standing constraint on FMP's ability to overload a public registry method, and the current failure mode is silent: the reflection sits inside an empty catch (Exception), so a wrong overload bind registers nothing and reports nothing.

ProjectRed's row remains the highest payoff-to-effort item in the migration, because one roughly 60-line file in one consumer is the sole reason FMP carries runtime Scala-pickle decoding, and removing it deletes about 580 lines plus the Scala branch of ASMMixinCompiler. It cannot be done first, however. The Java mixin path does not currently accept what LightMicroblock needs, so two FMP prerequisites belong to Phase 7:

  1. Abstract Java mixins over an abstract base. registerJavaTrait rejects ACC_ABSTRACT classes (ASMMixinCompiler.scala), but Microblock is abstract (Microblock.scala:39) with abstract microClass, itemClassID, and render. A Java mixin over it must therefore be abstract as well, so the check has to admit abstract mixins and skip their abstract members rather than reject the class. The existing Java-mixin example, TPartialOcclusionTile, avoids this only because its base TileMultipart is concrete.
  2. @SideOnly stripping on the Java path. Side stripping is implemented only for Scala traits, through listSideOnly(sig: ScalaSignature) reading the Scala signature annotation table inside registerScalaTrait. LightMicroblock.renderDynamic is @SideOnly(Side.CLIENT) and the trait is applied on both sides, because ProjectRed's addTraits ignores its client argument. Without an equivalent annotation-driven strip in registerJavaTrait, a Java rewrite would merge a client-only method into server-side generated microblocks.

Both prerequisites are Phase 7 work and are worth doing regardless, since they are also what any future FMP-internal Java microblock mixin would need.

UtilitiesInExcess is the cheapest case in the table because it is not yet in the pack. Anything fixed before it ships never becomes a compatibility obligation at all.

Its NBT key mismatch is real but currently mild, and the detail matters for whoever fixes it. loadPart is always followed by part.load(tag) on every audited path, including BuildCraftCompat's schematic restore, so the factory's wrong material is corrected immediately afterwards. The visible symptom is therefore one logged Missing mapping for part with ID: per part per load, from materialID("") falling through to the missing-material placeholder, rather than lost material data. Extra Utilities writes mat as well, in PartMicroBlock, PartConnecting, and PartPipeJacket, so correcting the key also matches the legacy tags reached through UtilitiesInExcess's extrautils:* aliases; there is no legacy-conversion risk in the fix.

  • Land the two Phase 7 Java-mixin prerequisites, then the ProjectRed LightMicroblock Java rewrite, with a fixture proving the Java-trait path produces equivalent generated microblocks on both sides.
  • Land the Galacticraft reflection signature check.
  • Fix the UtilitiesInExcess mat/material key mismatch and its getIdMap() use before it enters the pack.
  • Add the supported public equivalents needed by Schematica, GuideNH, Et Futurum, and Iguana as additive API.
  • Patch those four consumers and record the released versions that no longer need the private shapes.
  • Only after a consumer's released version is in the pack, move its retained private shape onto the Phase 8 removal list.

Consumer migration remains opportunistic. Nothing here justifies a coordinated lockstep release of the pack's most load-bearing block system; the value is in removing FMP's design constraints, not in reaching zero deprecated call sites.

Exit condition: no consumer-side hack constrains an FMP internal that the port would otherwise be free to change, or each remaining constraint is recorded with the released consumer version required to lift it.

Risk order

Risk Work
Low Leaf utilities, value holders, straightforward packets, isolated Java API facades
Medium Registries, recipes, placement logic, render helpers, schedulers, hot-loop rewrites
High TileMultipart, networking/NBT compatibility, tile trait conversion, external API bridges
Very high Runtime composite generator, Scala trait compatibility, removal of Scala runtime

The work should move from low to high risk while keeping a buildable mixed-language repository. A one-shot conversion would combine source, ABI, lifecycle, data-format, and bytecode-generation failures and make regressions difficult to localize.

Characterization and compatibility test strategy

Extensive tests are both possible and recommended here. Their purpose is not to prove that the current Scala behavior is ideal; it is to make its observable behavior explicit before translation, so every later difference is intentional.

The repository now has a Java-8-compatible JUnit Jupiter suite and a small Forge integration runner. The current baseline is 153 plain-JVM tests and 70 Forge server tests. Minecraft 1.7.10 does not provide the modern GameTest framework, so game-dependent coverage continues through the existing narrow Forge harness rather than a second test framework.

Layer 1 — Plain JVM characterization tests

Run fast tests through the normal Gradle test task, using real implementation classes and small fakes only at world, network, or rendering boundaries. Good targets include:

  • Slot masks, part lookup, occlusion and collision decisions.
  • Placement grids, side transforms, microblock shape calculations, and recipe matching.
  • Registry ordering, duplicate handling, missing-entry behavior, and callback order.
  • Scheduler ordering and cancellation with a deterministic clock or tick source.
  • Redstone connection and power matrices for all six sides and relevant masks.
  • Packet codec logic that does not require a live channel.
  • NBT read/write logic that can use real NBTTagCompound values without starting Forge.

Finite input spaces such as sides, slot masks, and small recipe grids should be covered exhaustively. Larger combinations should use named regression cases plus deterministic property/fuzz tests with a recorded seed. Avoid tests that merely restate the implementation or assert interactions with a large graph of mocks.

Layer 2 — Golden compatibility fixtures

Store compact fixtures under test resources when behavior has a stable serialized or structural representation:

  • Canonicalized NBT trees including names, types, values, missing fields, and legacy data; do not depend on unspecified compound-key iteration order.
  • Exact packet bytes for representative updates and malformed/truncated input behavior.
  • Part identifiers, registration order, and conversion results.
  • Public/protected JVM descriptors and required Scala bridge entry points.
  • Generated tile superclass, interfaces, fields, methods, access flags, and cache identity. Compare structure and behavior rather than raw class bytes or constant-pool order.
  • Small precompiled Java and Scala consumer fixtures built once against the reference dev jar, then loaded against the migrated jar to catch binary linkage failures that source recompilation would hide.

Golden fixtures must be reviewed rather than regenerated automatically when they fail. Otherwise an accidental change can silently redefine the baseline.

Layer 3 — Real Forge integration tests

Use a dedicated test mod/source set launched in an actual Forge environment for behavior coupled to FML, the classloader, worlds, or networking. The harness should fail the Gradle task or process when an assertion fails. It should cover:

  • Client and dedicated-server startup, mod lifecycle order, mixin selection, and side-only class safety.
  • Registration and creation of representative dynamic tile combinations.
  • Add/remove part sequences, ticking, scheduled ticks, redstone neighbor updates, and chunk unload/reload.
  • Save with the reference implementation and load with the migrated implementation.
  • Server-to-client part description and incremental update packets.
  • Loading representative downstream binaries compiled against the reference API.

Keep this harness narrow: a tiny test mod and deterministic scenarios are preferable to a general automation framework. Run server cases routinely; reserve the slower client cases for changes that touch rendering, client packets, or side-specific initialization.

Layer 4 — Irreducible manual and performance checks

Some behavior remains poorly captured by assertions in a legacy client. Retain focused manual checks for visual correctness, selection boxes, particles, input interaction, resource reload, and third-party rendering compatibility. Screenshot comparisons can help, but GPU and renderer differences make them unsuitable as the only oracle.

Performance fixes require the same CPU/allocation scenario before and after the change. Functional tests can prove equivalent results, but they cannot prove reduced allocations or better tick time.

Coverage rule

Coverage percentage is not the goal. The required evidence for an area is:

  • Normal behavior, edge cases, invalid input, ordering, and state transitions are represented.
  • Compatibility-sensitive bytes, NBT, identifiers, descriptors, and generated interfaces are frozen where applicable.
  • At least one real Forge scenario covers code that depends on lifecycle or world state.
  • The tests demonstrably pass against the old Scala area before any Java replacement begins.
  • The same tests pass unchanged against the Java implementation, except for separately documented bug fixes.

Decisions

Resolved from the consumer audits:

  1. All 27 binaries in the target pack must keep working without recompilation; UtilitiesInExcess is an additional forward-compatibility target.
  2. Third-party Scala trait registration is public in practice: ProjectRed registers LightMicroblock.
  3. Java-maintainable source is the first-release milestone. Scala runtime removal is deferred because shipping mods link Scala descriptors, trait helpers, companions, and external trait generation.

Still open, and only to be decided when they become necessary:

  1. How long deprecated Scala bridges remain supported after the initial Java release. Phase 9 fixes the policy for adding deprecations (only alongside a replacement); the removal window is still open and is gated by Phase 10.
  2. Which known bugs are intentionally fixed during the port rather than after compatibility is established.
  3. Which GTNHLib release is the minimum supported version once code actually needs it.
  4. Whether runtime tile composition merits redesign after the Java port is stable.

Effort tracking

The old pre-audit calendar ranges are retired. They mixed the Java source port, ecosystem validation, performance work, and optional Scala-runtime removal into estimates that are no longer useful. Track the remaining work by completed compatibility gates instead:

  1. Finish the source-only reflection/member guards; the Schematica registry view is repaired.
  2. Freeze tile ordering, lifecycle, NBT, packet, and generator behavior with compact representative fixtures.
  3. Capture the first focused CPU/allocation baseline and port the complete redstone unit.
  4. Convert a representative generated tile-trait group, then the remaining trait groups.
  5. Convert the multipart/microblock core while keeping the frozen data and consumer contracts green.
  6. Update the ASM/runtime generator, then run packaged-pack and manual release validation.

Estimate calendar time again after gates 1 and 2. They expose the real cost of hidden consumer contracts and the generated-trait compatibility work. Scala-runtime removal remains a later project and is not part of this estimate.

Findings log

2026-08-14

  • Confirmed that the codebase is feasible to migrate incrementally to Java.
  • Confirmed that generalized Scala try/catch overhead is not present; only a small set of exception-based non-local returns is relevant.
  • Identified closure/boxing/collection allocations as stronger performance candidates.
  • Confirmed that the existing runtime compiler already has a Java trait path.
  • Confirmed that public bytecode exposes Scala types and compiler artifacts, so bridges and Scala runtime removal must be separate decisions.
  • Confirmed Hodgepodge's normal/early/late UniMixins plus GTNHMixins registration pattern as the reference for fixed mixins.
  • Confirmed that UniMixins is not a drop-in replacement for ForgeMultipart's runtime composite-tile generator.
  • Confirmed GTNHLib as a viable route to fastutil, subject to pack-aligned versioning and runtime-resolution verification.
  • Adopted characterization tests against the existing Scala implementation as a prerequisite for each converted area, backed by golden ABI/data fixtures and real Forge integration scenarios where plain JVM tests are insufficient.
  • Bootstrapped JUnit Jupiter and passed an eight-case IDWriter characterization pilot against the untouched Scala implementation, covering byte/short/int thresholds and exact encoded bytes.
  • The pilot found that negative IDWriter.setMax values select the byte carrier because Scala 2.11 treats 0xffffffff as signed -1; current registry callers supply non-negative collection sizes, so this is recorded as a dormant edge case and possible deliberate divergence rather than changed during characterization.
  • Added a dedicated functionalTest source set and runFunctionalTestServer task. The task starts a real Java 8 Forge dedicated server, loads a test-only FML mod, runs JUnit after FMLServerStartedEvent, writes legacy JUnit XML, and shuts the server down.
  • Passed the first two Forge integration checks against the untouched Scala implementation: complete server lifecycle/world startup with all three ForgeMultipart mod IDs loaded, and generation plus constructor-class caching of a TSlottedTile composite with its 27-slot map.
  • Hardened the Gradle wrapper to delete stale reports and fail when no fresh report is produced, no tests run, or any failure/error is reported. This is necessary because Forge 1.7.10 can terminate after a test-mod exception with process exit code 0.
  • Confirmed the Java 8 server lane on Azul OpenJDK 8u492. The current deobfuscated modern-server tasks are not usable as release evidence: Java 21 needs the legacy Security Manager switch and then Lwjgl3ifyRelauncherTweaker rejects the server launch; Java 25 rejects System.setSecurityManager before Forge loads. Modern-Java validation therefore remains a packaged-server gate rather than a custom test-runner workaround.
  • Added eight plain-JVM characterization cases for NormalOcclusionTest and NormallyOccludedPart: empty sets, separation on every axis, face contact, tolerance-sized overlap, real overlap, containment, identical boxes, multiple-box all-pairs behavior, normal-box delegation, partial-box delegation, and parts without occlusion interfaces.
  • Confirmed that normal occlusion returns true when parts may coexist, despite the source comment saying “true if the test fails.” Face contact and overlaps no larger than CodeChickenLib's 1e-5 intersection tolerance are accepted; larger overlaps reject placement. Preserve the behavior and correct the misleading documentation separately when this area is ported.
  • Added nine plain-JVM characterization cases for PartialOcclusionTest: its fixed 8³ grid and x-major indexing, part-ID encoding, half-voxel coordinate rounding, required visibility, complete-occlusion exemption, cross-part overlap, same-part box overlap, unfilled entries, and the public JPartialOcclusion overload.
  • Confirmed that every non-exempt part needs at least one exclusively owned voxel. Any second write to an occupied voxel permanently changes it to -1, including overlap between two boxes supplied by the same part; this can make that part fail the test. Preserve this exact dynamic-tile behavior during translation unless it is changed later as an explicitly documented bug fix.
  • Completed the first behavior-preserving production conversion by replacing IDWriter.scala with IDWriter.java. All eight existing encoding cases, all 25 plain-JVM tests, the clean build, and both Java 8 Forge server checks pass unchanged.
  • Preserved the four legacy Scala function accessor descriptors as deprecated binary bridges and added direct Java write(MCDataOutput, int) and read(MCDataInput) methods. Recompiled Scala registry callers use the new primitive methods because Scala property auto-application does not apply to accessors declared in Java.
  • Removed the six IDWriter$$anonfun$setMax$* compiler artifacts from the packaged jar. Their disappearance and replacement with Java anonymous helper classes is recorded in JAVA_MIGRATION_DIVERGENCES.md; the supported IDWriter descriptors remain link-compatible.
  • Added a frozen Scala 2.11.5 consumer compiled against the reference dev jar. Its constructor and inherited default method call JPartialOcclusion$class directly, so it detects the linkage failure that freshly recompiled tests would miss.
  • Replaced TPartialOcclusion.scala with Java implementations of PartialOcclusionTest and JPartialOcclusion. The interface keeps its name and method descriptors, while allowCompleteOcclusion() is now a Java default method and the deprecated $class helper remains for old Scala binaries.
  • Passed all ten partial-occlusion behavior/API cases, the frozen Scala binary consumer, all 27 plain-JVM tests, a clean build, and both Java 8 Forge server checks. The existing marker-interface registration remains unchanged and continues to drive runtime tile generation.
  • Completed the downstream ABI inventory by constant-pool scan of 240 mod jars in GTNH daily 2026-08-14+678, recorded in JAVA_MIGRATION_ABI_INVENTORY.md with the scanner in tools/AbiScan.java and the frozen baseline in src/test/fixtures/abi/. GitHub code search was rejected as an oracle because it indexes default branches only and cannot see reflection strings or closed-source consumers.
  • Found 27 consumer jars referencing 35 inherited types, 255 exact member descriptors, 76 other types, and 20 reflective string constants.
  • Answered open decision 2: third-party Scala traits are registered externally. ProjRed passes its own LightMicroblock Scala trait to MicroblockGenerator.registerTrait, so registerScalaTrait and ScalaSignature decoding must survive Phase 7.
  • Answered open decision 4: Scala runtime removal is not achievable for the first Java release. ProjRed, OpenComputers, ProjectBlue, and ForgeRelocationFMP link against 16 static methods on 8 trait $class helpers plus 17 companion MODULE$ singletons.
  • Found only five referenced descriptors containing Scala types, and TileMultipart.jPartList() already outweighs partList() in downstream use. The Scala bridge surface is much smaller than the initial audit assumed.
  • Found zero downstream references to IDWriter, so its four retained Scala function accessors are not load-bearing and can be dropped.
  • Removed both speculative bridges the inventory proved dead: the four IDWriter Scala function accessors and the whole JPartialOcclusion$class helper, along with the ReferenceScalaPartialOcclusion fixture that only existed to verify the helper. JPartialOcclusion itself and both of its method descriptors are unchanged.
  • IDWriter now selects a carrier width instead of storing Scala closures, removing the per-call Integer boxing and Function1/Function2 allocation that the first port had preserved. The nine encoding cases still pass unchanged.
  • Established the working rule for the remaining phases: check JAVA_MIGRATION_ABI_INVENTORY.md before writing a bridge, rather than writing one reflexively for every converted file.
  • Ported TCuboidPart, JCuboidPart and TCuboidPart$class to Java, the first conversion of a trait whose $class helper is genuinely load-bearing. All reference descriptors are preserved, verified by diffing javap -s against the reference dev jar, and a frozen Scala 2.11.5 consumer whose forwarders call all four statics loads and runs against the port.
  • Found the first structural limit of the migration: Scala trait linearization cannot be reproduced by a Java interface, because a superclass method always beats an interface default on the JVM. Recompiled Scala consumers that mix TCuboidPart into a TMultiPart subclass now silently get TMultiPart's empty implementations. Binary compatibility is unaffected, and CuboidPartCharacterizationTest carries the regression guard, but this applies to every remaining trait that overrides TMultiPart members and should be assumed for TFacePart, TNormalOcclusion, TIconHitEffects and TItemMultiPart as well.
  • Recorded the consequence for consumers: Scala code that recompiles must extend JCuboidPart or declare the overrides itself. This belongs in the release notes for the first Java release, not only in the divergence log.
  • Found a build-order constraint: compileJava runs before compileScala, so ported Java that still references Scala types must live under src/main/scala for joint compilation, as minecraft/McBlockPart.java already does. Files can move to src/main/java once their Scala dependencies are ported.
  • Established the fixture recipe for the remaining load-bearing bridges: build a reference dev jar at the pre-port commit, compile the consumer with Scala 2.11.5 under Java 8 with -target:jvm-1.8 against that jar plus the project compile classpath, then store the class base64-encoded under src/test/resources/compat/.
  • Ported the whole TNormalOcclusion.scala unit: NormalOcclusionTest, NormalOcclusionTest$, JNormalOcclusion, TNormalOcclusion, TNormalOcclusion$class and NormallyOccludedPart. The singleton and both trait interfaces are descriptor-identical to the reference; the other three add only private members.
  • Found that Scala emits trait super accessors as ACC_SYNTHETIC, which javac hides completely. An abstract synthetic method cannot be implemented by any Java class or by Scala recompiled against a Java interface, so codechicken$multipart$TNormalOcclusion$$super$occlusionTest is emitted as a non-synthetic default returning true. Method resolution ignores the flag, so existing call sites are unaffected. Expect the same treatment wherever a remaining trait calls super.
  • Found the dangerous shape of the linearization problem. HollowMicroblock simply lacked an override, which is easy to spot. PostMicroblock already had one ending in super.occlusionTest(npart), which used to route through the trait and now reaches Microblock directly, silently dropping the box test while still compiling. Existing super call sites must be audited on every remaining trait conversion, not just missing overrides.
  • Recorded a second source-only divergence: NormalOcclusionTest was a Scala object, so NormalOcclusionTest(a, b) apply sugar no longer compiles and recompiled Scala callers must write NormalOcclusionTest.apply(a, b). Both static forwarders and the MODULE$ singleton keep their descriptors, so binaries are unaffected.
  • Kept scala.collection.Traversable in two descriptors, as the ABI inventory predicted. They are drained with JavaConversions.asJavaIterator(boxes.toIterator()), which is the one interop route callable from Java.
  • Removed two more closure artifacts, NormalOcclusionTest$$anonfun$apply$1 and its nested $$anonfun$apply$2, and corrected the scaladoc that claimed the test "returns true if the test fails".
  • Ported TFacePart, the first trait where Java default methods are actually safe. TMultiPart declares neither solid nor redstoneConductionMap, so there is no superclass method for a default to lose to. The interface is descriptor-identical to the reference and no in-repo Scala needed changing.
  • Adopted the resulting rule for the remaining traits: emit default methods only when no class in an implementor's superclass chain declares the same member, otherwise leave the member abstract and add explicit forwarders as TCuboidPart and TNormalOcclusion required.
  • Made TFacePart inheritable from Java as a side effect. Both members were abstract on the reference interface, so minecraft/McSidedMetaPart had to declare them even to accept the defaults. That class stays descriptor-identical, so the change is purely additive.
  • Ported the whole TIconHitEffects.scala unit: IconHitEffects, JIconHitEffects, JIconHitEffects$class, TIconHitEffects and TIconHitEffects$class. Both interfaces are descriptor-identical to the reference and the other three add only a private constructor.
  • Confirmed the default-versus-abstract rule discriminates correctly within a single file. JIconHitEffects.getBreakingIcon became a default because TMultiPart does not declare it, while TIconHitEffects.addHitEffects and addDestroyEffects stayed abstract because it does. MicroblockClient needed explicit forwarders only for the latter pair.
  • Removed IconHitEffects$ and its MODULE$ field. No jar in the pack references it, in bytecode or by name, so it is dead under the inventory rule. This is the first public class the migration has dropped rather than a method, and it contrasts with NormalOcclusionTest$, which is referenced and was kept.
  • Preserved evaluation order inside addDestroyEffects rather than hoisting the tile read, because the characterization pins which work happens before a null tile fails.
  • Established that particle appearance cannot be asserted headless. Icon selection, side order and bounds scaling are covered by tests; the visual result stays on the manual checklist.
  • Ported ItemMultiPart.scala into JItemMultiPart, TItemMultiPart and TItemMultiPart$class. The interface is descriptor-identical to the reference, JItemMultiPart gains only additive statics, and the bridge keeps all three public statics for ProjRed.
  • Generalised the default-versus-abstract rule. TItemMultiPart extends Item rather than TMultiPart, and the same split applied: getHitDepth became a default because Item does not declare it, while onItemUse stayed abstract because Item does. What matters is the implementor's actual superclass, not TMultiPart specifically.
  • Confirmed that placement logic is testable headless. onItemUse short-circuits as soon as newPart returns null, before any world, player or stack access, so the depth branch, the neighbour offset and the shared mutated BlockCoord can all be asserted with null arguments.
  • Added the first functional coverage of runtime class generation interacting with a ported type. Microblock classes are built by ASM at run time, so a plain-JVM test cannot show that a generated class resolves a Java interface default; the Forge server suite now generates a FaceMicroblock and asserts it inherits redstoneConductionMap, and generates a HollowMicroblock and asserts its own 0x10 and false still win.

2026-08-15

  • Ported the six marker interfaces: TSlottedPart, IRandomDisplayTick, INeighborTileChange, TRandomUpdateTick, ISidedHollowConnect and IMicroMaterialRender. All six are member- and descriptor-identical to the reference, and the seven implementors and mixin tiles were diffed as well and are unchanged member for member.
  • Confirmed that five of the six were already pure abstract interfaces in bytecode with no $class helper, so their conversion had no bytecode consequence at all. Exactly one class left the jar: TRandomUpdateTick$class.
  • Applied the default-versus-abstract rule to TRandomUpdateTick.onWorldJoin and got the failing answer: TMultiPart declares onWorldJoin, so a default would be shadowed by the superclass and the auto-registration would silently never run. It stays abstract, and implementors declare it and call TickScheduler.loadRandomTick themselves, which RedstoneTorchPart — the only implementor in this codebase — already did.
  • Applied the bridge rule to TRandomUpdateTick$class and got a clean negative: no jar in the pack references TRandomUpdateTick in any form, and the frozen baseline contains exactly the eight $class helpers the inventory lists. No bridge was written, following the IDWriter precedent. The accepted cost is a recompiled-Scala-consumer break with no shipping consumer.
  • Found that IMicroMaterialRender is implemented for every part solely by TMultiPart's Scala-style world/x/y/z/getRenderBounds accessors. Renaming any of them to a bean accessor would silently unimplement the interface, so the names are now pinned by a characterization test rather than left to reviewer attention.
  • Established that for interfaces carrying no implementation, the characterization is the shape: every member abstract and public, no superinterface, and the exact member set. That is enough to catch the two realistic failure modes, a stray default and a renamed accessor.
  • Noted that IRedstonePart.scala is misleadingly named and is not a marker-trait file. It carries six traits plus RedstoneInteractions, whose MODULE$ is load-bearing, and was moved out of the low-risk group.
  • Noted a documentation gap, not addressed here: this log was last appended for the TItemMultiPart port. The eight ports between it and this entry — TEdgePart, Saw, both registries, TileMultipart, TMultiPart, TickScheduler and BlockMultipart — are recorded in JAVA_MIGRATION_DIVERGENCES.md but never made it here, so the "Current status" section above should be read with that in mind.
  • Ported MultipartHelper, MultipartHelper$ and MultipartHelper$IPartTileConverter to Java. All three are public-member- and descriptor-identical to the reference and the full emitted class list is unchanged, so this port neither added nor removed a class.
  • Kept MultipartHelper$ on the strength of a reflective string constant alone. It is not among the 17 MODULE$ singletons read from bytecode and no jar links against it, but guidenh names it, and the inventory warns that removing a companion breaks such consumers invisibly. This is the opposite call to IconHitEffects$, which was dropped because nothing referenced it in bytecode or by name, and it establishes that the MODULE$ list alone is not sufficient grounds to delete a companion.
  • Found a reusable characterization technique: a class that cannot class-initialize headless is a probe for control flow. MultipartSaveLoad reflects into TileEntity's static maps through ObfMapping and always throws under a plain JVM, so createTileFromNBT returning null rather than raising proves the id guard short-circuits before the loadingWorld assignment. Assert LinkageError rather than the exact type, because the first attempt raises ExceptionInInitializerError and later ones raise NoClassDefFoundError.
  • Confirmed that erasure does not weaken IPartTileConverter.convert. The cast to T compiles away, but the ClassCastException a mismatched tile produces comes from the synthetic bridge on the subclass overriding convertMulti(T), which is the same mechanism Scala's asInstanceOf[T] relied on. It is now pinned by a test, because the code reads as though the check was lost.
  • Added five Forge server tests, doubling that suite, because most of this class cannot run headless. They cover tile construction through the ASM generator, the NBT round trip through the save/load hooks, registerTileConverter appending to a Scala MutableList from Java, and sendDescPacket against a loaded chunk.
  • Dropped the two commented-out blocks the reference carried, the PlayerInstance.playersInChunk reflection and the multi-tile sendDescPackets, rather than reproducing dead Scala as dead Java. The reason they existed, a missing forge access transformer, is now in the class javadoc.
  • Read guidenh's actual source at 6137525 rather than inferring from its constant pool, and recorded the exact reflective member list in JAVA_MIGRATION_ABI_INVENTORY.md. The scan could see the 20 names; only the source shows which members are looked up on them, and none of it is visible to the ABI diff.
  • Found that MultipartGenerator$.MODULE$ is load-bearing through reflection. generateCompositeTile is private[multipart], so no static forwarder exists and guidenh's static attempt always misses, leaving the companion as the only route. Phase 6/7 must keep it.
  • Found that MicroblockGenerator$.create is matched by exact parameter types, with MicroblockClass's fully qualified name string-compared. Widening a parameter or renaming the class breaks the lookup while every call site still links.
  • Verified against the already-ported TileMultipart that partList_$eq(scala.collection.Seq), loadParts, notifyTileChange and markRender all survived. partList_$eq is a Scala var setter with no Java-facing equivalent and is reflectively load-bearing, so dropping it for a list mutator would have broken guidenh invisibly. It was kept.
  • Confirmed the MultipartHelper port satisfies guidenh directly: it resolves createTileFromNBT as a static on the plain class and only falls back to MultipartHelper$.MODULE$ if that misses, so the retained companion is belt-and-braces rather than the primary path.
  • Ported TileCache to Java, the first conversion free to change shape rather than preserve it. It has zero downstream references in bytecode and does not appear in guidenh's reflective names, so map() became a java.util.Map, add/remove became void, and apply returns a nullable FlaggedTile instead of a scala.Option. No internal caller used any of the discarded return values.
  • Dropped TileCache$, TileCache$FlaggedTile$ and the whole case-class machinery on FlaggedTile. It is only ever a map value, never compared, hashed, printed or destructured. This is the IconHitEffects$ case rather than the MultipartHelper$ one: nothing names it in bytecode or by string.
  • Pinned an unreachable branch before removing it. The reference's findTile ends in case _ => null, but Some(FlaggedTile(t, rem)) matches whatever the flag holds, so a tile flagged removed is returned like any other and only the warning is suppressed. The functional test asserts that before the port so the Java version cannot quietly start returning null.
  • Confirmed the adapter approach works for ports that change collection types. The characterization test named scala.Option in exactly one helper, so the port's test diff was two lines and every behavioral assertion carried over untouched.
  • Ported PacketScheduler, IScheduledPacketPart and TScheduledPacketPart. IScheduledPacketPart is descriptor-identical, schedulePacket is unchanged, and TScheduledPacketPart.read stays abstract because TMultiPart declares it.
  • Found the case the default-versus-abstract rule splits within a single interface. read cannot be a default because TMultiPart declares it, but writeScheduled and readScheduled can, because TMultiPart declares neither. Making the latter two defaults restores what a Scala implementor already got from the trait for free; only Java implementors ever had to write them out.
  • Added TScheduledPacketPart.readMask as a Java 8 interface static, carrying the dispatch TScheduledPacketPart$class.read used to hold. Where a trait's central member cannot become a default, an interface static plus a documented one-line forwarder is the replacement, and it is the pattern to reuse.
  • Widened sendScheduled from private[multipart] to public. Its only caller is in codechicken.multipart.handler, and Java has no scope spanning a package and its siblings, so package-private would not compile. Expect this wherever private[multipart] crosses into the handler package.
  • Recorded that the whole subsystem has no implementors at all, in this codebase or in the pack. It is a third-party extension point, which is why every behavioral test had to construct its own part.
  • Confirmed the Forge server suite can drive the packet path without players connected. MultipartSPH.getTileStream only needs a non-remote world and a position, so a part bound to a tile with setWorldObj is enough to exercise getWriteStream.
  • Ported ControlKeyModifer and ControlKeyHandler. isControlDown(EntityPlayer) and isClientPressing() keep their descriptors; map() became a java.util.Map, and the implicit-conversion pair playerControlValue/ControlKeyValue was dropped because it only ever existed to give Scala player.isControlDown sugar and the reference already shipped the explicit static "for Java users".
  • Kept the misspelling in ControlKeyModifer. It is the published type name; only the file name, which said ControlKeyModifier.scala, was corrected. Renaming the type is an API change with no forcing reason and should be a deliberate decision rather than a side effect of a port.
  • Converted the first Scala object that extends a class rather than being a plain singleton. ControlKeyHandler is now an ordinary KeyBinding subclass with an INSTANCE constant, which drops the twelve static forwarders Scala generated for inherited KeyBinding members along with the companion. The registration sites pass ControlKeyHandler.INSTANCE.
  • Found that a Forge FakePlayer makes the server half of player-dependent code testable. FakePlayerFactory.getMinecraft(worldServer) yields a player whose world is not remote, which is exactly the branch the server takes, and it needs no connection.
  • Recorded that the client half remains untestable in both harnesses. isClientPressing is only ever set from a client tick, so the key binding, the press/release packet and the client branch of isControlDown are all manual checks.
  • Removed both package objects rather than porting them. A Scala package object compiles to a class named package, which is a Java keyword, so no Java source can declare or even name it; the characterization had to reach both through Class.forName. Each held one member, a logger alias for its proxy, so the indirection was inlined at its eight call sites instead of being given an invented Java home.
  • Proved the substitution was an identity before making it. The functional test asserted that each alias returned the very instance its proxy holds, so replacing logger with MultipartProxy.logger / MicroblockProxy.logger cannot change behavior. ConfigContent and MultipartGenerator are descriptor-identical afterwards.
  • Noted the first case where the migration's answer is deletion. Zero downstream references plus an impossible target name means there is nothing to preserve, and inventing a MultipartLogger type to hold a one-line forwarder would add API rather than remove it.
  • Emptied the low-risk queue. Everything remaining is the medium group or later, so the next piece of work should be chosen deliberately rather than taken off the top of a list.
  • Ported MultipartRenderer in the TickScheduler$ shape, with the companion holding the implementation and the plain class holding static forwarders. Forge registers the singleton itself and guidenh resolves the static first, so both halves are load-bearing for different reasons and the emitted class list is unchanged.
  • Applied the reflective-name rule for the second time, and this is the case it was written for: MultipartRenderer has no bytecode consumer at all, only guidenh string constants, and both the class and its companion had to survive.
  • Found a Scala-to-Java trap that is not silent but will recur: uniform access makes renderer.hasOverrideBlockTexture read the same whether it is a field or a no-arg method. It is a method, and the port did not compile until it grew parentheses. Every remaining renderer conversion will meet this.
  • Dropped the three static forwarders Scala generated for inherited TileEntitySpecialRenderer members, the same category as the twelve KeyBinding forwarders dropped from ControlKeyHandler.
  • Dropped an unused ThreadSafeISBRH import. It annotated nothing and appears nowhere else; if the intent was to mark this ISBRH thread-safe for Angelica then the annotation was never applied in the reference either, which is worth checking independently of the migration.
  • Recorded that this port adds nothing to the automated suites and everything to the manual checklist. Static and dynamic rendering, the breaking overlay and the render id all need a client.

2026-08-27

  • Reproduced Schematica 1.12.6's exact ReflectionHelper lookup of the private MultiPartRegistry$.codechicken$multipart$MultiPartRegistry$$typeMap field. The test failed against the existing Java port with UnableToFindFieldException, matching Schematica's silent integration-disable path.
  • Restored the exact private field name and scala.collection.mutable.Map descriptor as Scala's live wrapper over the canonical Java registry map. The wrapper adds no copied state or synchronization path.
  • Proved through the reflected Scala view that Schematica can resolve a factory and that the Java registry immediately sees the same entry. The focused registry suite and all 126 plain-JVM tests pass.
  • Added four source-consumer structural guards without adding downstream mods as test dependencies. They pin GuideNH's two BlockMicroMaterial mixin targets, Et Futurum's mutable static button-orientation arrays, and Iguana's reflected ItemSaw.harvestLevel field.
  • Reproduced Galacticraft's name-only method scan and froze its dangerous assumption: exactly one public MicroMaterialRegistry.registerMaterial method may exist, it must accept (IMicroMaterial, String), and BlockMicroMaterial(Block, int) must remain reflectively constructible. All 130 plain-JVM tests pass.
  • Accepted the dedicated-server EULA locally in ignored run/server/eula.txt, enabling the Forge characterization suite without adding a distributable acceptance file to the repository.
  • Froze tile NBT order with two different built-in parts: the outer tile id, ordered parts entries and per-part id/payload values survive reconstruction, and jPartList plus partMap reproduce the same order and slots.
  • Froze add/remove callback order, list publication, slot binding and detachment in a live Forge world. The plain-JVM replacement test also proves that a successful canReplacePart query does not mutate order or bindings.
  • Reproduced MatterManipulator's live-tile relocation shape. Removing the tile first calls onWorldSeparate for each part in list order; after reinsertion, onMoved calls each part's onMoved and default onWorldJoin, again in list order. The original generated tile, part instances, order, slot map and tile references survive.
  • Froze the complete logical chunk-description payload around the same torch/button pair: packet type, chunk coordinates, tile count, nibble-relative X/Z, full Y, part count, sorted runtime part IDs, part order and both meta bytes. CodeChickenLib's deferred Deflater transport step is deliberately outside this FMP fixture.
  • Froze representative generated-trait behavior. The existing TSlottedTile fixture pins built-in Scala trait class caching and field initialization; a ProjectRed-shaped external Scala microblock trait pins registration by name, trait initialization and method dispatch; and TPartialOcclusionTile pins Java-trait rewriting and override dispatch through the generated tile.
  • Froze a server-only pass-through interface with primitive/reference overloads. The generated tile forwards to one implementing part, rejects a second implementor, preserves and rebinds its delegate through copyFrom, and clears the generated implementation field on removal.
  • Added an opt-in Forge/JFR workload and captured the first focused CPU/allocation baseline. With eight parts, updateEntity and operate allocate 184.0 and 183.9 bytes per call; generated redstone's three-query iteration allocates 80.5 bytes. CPU and allocation sites point to TileMultipart.parts() collection copies and Scala redstone IntRef/closure traversal. Full methodology and rerun commands are in JAVA_MIGRATION_PROFILE.md.
  • The expanded baseline is 130 plain-JVM tests and 28 Forge server tests, all passing.

2026-08-28

  • Ported all six interfaces and RedstoneInteractions from IRedstonePart.scala as one unit. The emitted class list is unchanged, every public descriptor is unchanged, and RedstoneInteractions$.MODULE$ remains the implementation singleton behind the static façade.
  • Added six plain-JVM tests for interface shape, façade/companion members, maps, part-mask precedence, connector routing, and neighbor coordinate transforms, plus three Forge tests for initialized vanilla blocks, world power routing, the redstone-wire metadata fallback, and generated IRedstoneTile selection.
  • Re-profiled immediately before and after the port. Allocation stayed at roughly 4.02 GB / 80.5 B per generated three-query iteration, and JFR retained the same TRedstoneTile Scala closure/iterator sites. That cost belongs to the Phase 5 generated-trait conversion, not this helper unit.
  • Chose MicroRecipe.scala as the next independent Phase 4 unit because its nested collection scans compile early exits into NonLocalReturnControl; characterize all five recipe paths and precedence before porting it.
  • Characterized MicroRecipe's complete 17-method façade/companion ABI, Scala splitMap, all five recipe forms, class-specific gluing rules, saw ordering, exact material/tag lookup, and hollow-over-gluing precedence against the original Scala implementation.
  • Ported the recipe singleton to Java with ordinary loops. The published MicroRecipe static façade, MicroRecipe$.MODULE$, scala.Tuple3 return and immutable Scala map descriptors remain; the 12 private Scala closure classes and their NonLocalReturnControl exits are gone.
  • The recipe port passes all 141 plain-JVM and 37 Forge tests.
  • Characterized TPartialOcclusionTile against the untouched Scala implementation with four focused tests covering its exact direct class shape, partial/normal precedence, complete-occlusion behavior, and both short-circuit paths.
  • Ported TPartialOcclusionTile to direct Java source without changing the generator. Its compiled input still has one no-arg constructor, no fields, and the same two public descriptors; the Forge rewrite still exposes exactly the two behavior methods plus its super accessor, caches the generated class, and passes all 37 server tests.
  • The pilot raises the plain-JVM baseline to 145 tests. It deliberately proves no field or lifecycle behavior, so the next checkpoint is TSlottedTile, not a bulk conversion of the remaining traits.
  • Characterized TSlottedTile against untouched Scala with six focused Forge cases plus the existing live-world lifecycle fixture. The gate freezes its exact 13-method runtime interface, 27-slot per-instance initialization, getter/setter rebinding, copy identity, clear/remove/bind behavior, occupied-slot rejection, Scala value equality, generated field, super accessors, and class cache.
  • Ported TSlottedTile to a concrete Java mixin input with one public array field and ordinary loops. The runtime rewrite remains interface- and behavior-identical, while the four Scala range closures, $class artifact in the raw jar, and NonLocalReturnControl slot rejection disappear. All 145 plain-JVM and 43 Forge tests pass.
  • Found a Phase 5 source-build constraint: shipping binaries still see the same runtime interface, but a consumer recompiled directly against the untransformed dev jar sees the Java mixin input as a class. Consumers that name a generated trait therefore need a transformed compile stub or must avoid direct trait invocations; record this before claiming source-compatible downstream rebuilds.
  • Characterized TRedstoneTile against untouched Scala with three Forge cases covering its exact eight-method runtime interface, class caching, face/edge conduction, strong and weak maxima, mask filtering, arbitrary Seq input, neighbor masks, and vanilla-side translation. This directly covers ProjectRed's openConnections and Extra Utilities' one-argument weakPowerLevel calls.
  • Ported TRedstoneTile to Java with allocation-free immutable-list traversal and an iterator fallback for other published Seq values. A package-private shim isolates inherited TileMultipart methods and Minecraft coordinate fields from known Java-trait transformer limitations; no public redstone or generated-interface member changed.
  • The paired 50,000,000-iteration run fell from 4,023,855,000 bytes / 80.5 B per three-query iteration to zero measured allocation, while throughput rose 20.3% from 6,636,424 to 7,986,213 iterations/s. The checksum remained 3315999992, and all 145 plain-JVM plus 46 Forge tests pass.
  • Characterized mutable-Seq read behavior, direct ordered indexing, and BlockMultipart.getTile filtering, then extended the focused workload with getLightValue and getTile phases before changing production code.
  • Removed the Java-port-only ArrayList snapshots from all read paths and Java wrappers from internal block, renderer, and scheduler paths. The public jPartList() bridge is unchanged; add/remove retain intentional mutable snapshots before publishing a replacement immutable Seq.
  • In the paired run, getLightValue fell from 183.9 B to 0.0 B per call with 11.42x throughput, and getTile fell from 24.0 B to 0.0 B per call with 2.89x throughput. All 147 plain-JVM and 46 Forge tests pass.

2026-08-30

  • Characterized TileMultipartClient and TRandomDisplayTickTile together against untouched Scala. Six behavior cases freeze render-cache partitioning and bounds, lazy initialization, dynamic short-circuiting, base no-op display ticks, and ordered IRandomDisplayTick dispatch; a Forge smoke case freezes both exact runtime interfaces, inheritance, generated fields, and generated-class caching.
  • Ported both traits to Java with package-private access shims. registerJavaTrait now linearizes an already registered Java parent trait, recognizes explicit field accessors, and skips transient runtime caches when auto-generating copyFrom. Trait registration now precedes BlockMultipart construction so Forge defines the runtime interfaces before ordinary Java bytecode can preload the mixin inputs.
  • Added three no-op client dispatch hooks to TileMultipart, allowing in-repo Java callers to invoke generated overrides through the stable superclass rather than emit class opcodes against types Forge rewrites to interfaces. The runtime trait surfaces remain exact; the raw dev-jar source-build limitation remains documented.
  • All 153 plain-JVM tests and all 77 Java 8 Forge dedicated-server tests pass. Actual static/dynamic rendering and particle appearance remain client-only manual checks.
  • Characterized MultipartCompatiblity and MCPCCompatModule against untouched Scala. Three plain-JVM cases freeze both static facades, both MODULE$ companions, the private Scala Function4 field, default allow behavior and callback identity; two Forge cases freeze non-MCPC loading and the logged missing-hook fallback.
  • Ported both singletons to four Java types with unchanged public names and descriptors. The callback still propagates reflection and cast failures unchanged, while the two Scala anonymous-function artifacts become private named Java callback classes. No frozen binary or audited source consumer names those implementation classes.
  • All 156 plain-JVM tests and all 79 Java 8 Forge dedicated-server tests pass. This initialization-only hook is not a meaningful target for the focused allocation benchmark; successful MCPC integration remains environment-dependent.
  • Characterized MultipartMod against untouched Scala. Two plain-JVM cases freeze both annotated singleton types, all ten lifecycle methods and annotations, plus the MultipartPH.channel companion descriptor and identity; two Forge cases freeze FML's companion mod instance, completed initialization and server-stop cleanup.
  • Ported the singleton to an annotated Java facade/companion pair while retaining modLanguage = "scala". The one source use of the object as a value now names MultipartMod$.MODULE$ explicitly, leaving the compiled packet-handler field and accessor unchanged.
  • Found that the inherited MultipartProxy.postInit static forwarder carries @SideOnly(CLIENT) and is stripped on a dedicated server. The Java companion therefore invokes MultipartProxy$.MODULE$ directly, matching the reference Scala bytecode and allowing virtual resolution to reach the server implementation.
  • All 158 plain-JVM tests and all 81 Java 8 Forge dedicated-server tests pass. Mod lifecycle dispatch is startup-only, so it is not a meaningful focused throughput or allocation target.
  • Characterized MultipartEventHandler against untouched Scala. Two plain-JVM cases freeze both singleton types, all twelve public event methods, priorities and the client-only highlight boundary; three Forge cases freeze companion registration on both buses, chunk load/unload cleanup, queued watches and END-phase tick dispatch.
  • Ported the singleton to a Java facade/companion pair with unchanged public names, descriptors and annotations. The proxy now names MultipartEventHandler$.MODULE$ explicitly, preserving the exact object registered on both buses.
  • Server ticking still passes the configuration manager's live player list through Scala's Java-list buffer adapter; no copy or new traversal was introduced. All 160 plain-JVM and all 84 Java 8 Forge dedicated-server tests pass.
  • Characterized MicroblockMod against untouched Scala. Two plain-JVM cases freeze both annotated singleton types, all ten lifecycle/IMC methods, the mutable angelicaCompat accessors and shared identity; one Forge case freezes FML's companion mod instance and the completed microblock lifecycle.
  • Ported the singleton to an annotated Java facade/companion pair while retaining modLanguage = "scala". Lifecycle dispatch calls MicroblockProxy$.MODULE$ directly, matching the reference bytecode and avoiding side-only static forwarders that Forge strips on a dedicated server.
  • The internal client assignment now calls the preserved angelicaCompat_$eq method explicitly because recompiled Scala cannot apply property-assignment syntax to a Java-authored setter. The compiled accessor ABI is unchanged.
  • All 162 plain-JVM and all 85 Java 8 Forge dedicated-server tests pass. Startup and lifecycle dispatch are not a meaningful focused throughput or allocation target; MicroblockEventHandler.scala is the next target.
  • Reproduced the reported stale incremental @Mod(version) failure immediately after a commit. Gradle reran compileScala, but Zinc retained the joint-compiled Java classes that had inlined the previous Tags.VERSION.
  • Configured compileScala to force Zinc recompilation whenever Gradle schedules the task. The focused version test passes without cleaning both before and after a commit changes the generated version, so the assertion remains.
  • Characterized MicroblockEventHandler against untouched Scala. Two plain-JVM cases freeze both singleton types, all four public event methods, normal event metadata and both client-only boundaries. One Forge case freezes companion registration and complete method stripping on a dedicated server.
  • Ported the handler to a Java facade/companion pair and changed the one Scala registration to name MODULE$ explicitly. Texture-atlas filtering, highlight guards, matrix/render sequence and cancellation behavior are unchanged; their actual rendering remains on the client manual checklist.
  • All 164 plain-JVM and all 86 Java 8 Forge dedicated-server tests pass. This event-only adapter is not a meaningful focused performance target; microblock/handler/packethandlers.scala is next.
  • Characterized the ForgeMicroblock packet-handler unit against untouched Scala. Seven plain-JVM cases freeze the shared channel base, both facade/companion surfaces and exact packet interfaces, the integrated-server registry skip, ordered missing-material disconnect, unknown-type MatchError and no-op server callback. One Forge case freezes the handshake channel, type and complete material-ID payload.
  • Ported all five emitted packet-handler classes to Java and changed the two Scala proxy registrations to pass the companion singletons explicitly. Every callable public member and the emitted class list match the reference; the registry channel and wire format are unchanged.
  • All 171 plain-JVM tests pass. After the local EULA was accepted, all 87 Java 8 Forge dedicated-server tests pass, including the new registry handshake.
  • Characterized MultipartSaveLoad against untouched Scala. Three plain-JVM cases freeze the static facade, load-bearing companion, private fields and exact dummy class shape. Four Forge cases freeze ProjectRed-style binary linkage, both reflected vanilla maps, converter precedence/deletion and saved multipart reconstruction.
  • Ported the singleton to a Java facade/companion pair and kept the dummy as the static facade member MultipartSaveLoad.TileNBTContainer, which emits the literal MultipartSaveLoad$TileNBTContainer binary name. Nesting it under the companion would emit the wrong double dollar. Every callable public member matches the reference. Only the unreferenced compiler-generated $$anonfun$1 closure disappeared.
  • All 174 plain-JVM tests and all 91 Java 8 Forge dedicated-server tests pass. The next medium-risk target is MissingMicroMaterial.scala; its real icon/render paths remain client-manual work.
  • Characterized MissingMicroMaterial against untouched Scala. Three plain-JVM cases freeze the exact facade and companion surfaces, MODULE$, inert material values, interface defaults and all client-only boundaries. One Forge case freezes side stripping and the exact singleton registered under the missing-material name and ID.
  • Ported the singleton to a Java facade/companion pair and changed both Scala object-value uses to name MODULE$ explicitly. The placeholder key, stone item, sound, strength, resistance and missing-texture render pipeline are unchanged; only the actual client rendering remains manual.
  • All 177 plain-JVM and all 92 Java 8 Forge dedicated-server tests pass. This inert singleton is not a meaningful focused performance target; DefaultContent.scala is the next medium-risk target.
  • Characterized DefaultContent against untouched Scala. One plain-JVM case freezes its one-method static and companion surfaces. Two Forge cases freeze the five microblock factories and IDs, all 103 sorted built-in materials, their exact implementation types and the complete legacy-name remap table.
  • Ported the singleton to a Java facade/companion pair while continuing to use the existing BlockMicroMaterial$ overloads and Scala ranges. Registration contents, ordering and the historical meta-0-only log2/leaves2 overload behavior are unchanged.
  • All 178 plain-JVM and all 94 Java 8 Forge dedicated-server tests pass. Pre-init-only registration is not a meaningful focused performance target; GrassMicroMaterial.scala is the next material unit.
  • Characterized GrassMicroMaterial and TopMicroMaterial against untouched Scala. Five plain-JVM cases freeze both constructors, the grass overlay accessor used by UtilitiesInExcess, the top default-argument facade/companion, common-side boundaries and horizontal/side UV plus colour-pipeline routing. One Forge case freezes their registered blocks/meta and the exact surface left on a dedicated server.
  • Ported both classes and the default-argument companion directly to Java. Grass keeps its uncoloured base side pass, coloured top and height-adjusted side overlay; TopMicroMaterial keeps coloured horizontal faces and translated side UVs. All three emitted binary names and every callable public descriptor match the Scala reference.
  • A clean compile was required to evict the deleted Scala classes before the unchanged characterization could test the Java implementation. All 183 plain-JVM and all 95 Java 8 Forge dedicated-server tests pass. The render path retains the same per-side transformation work, so no separate performance claim is made; multipart/handler/proxies.scala is the next medium-risk target.

2026-08-31

  • Audited the published API surface against all 28 consumer checkouts to decide whether a cleaner API is warranted. It largely already is one: TMultiPart is an ordinary abstract Java class, and the factory, converter, material, PartMap, redstone and occlusion interfaces need no replacement. Added Phase 9 for the parts that are not.
  • Found nine remaining public entries carrying Scala types. Two already have a Java sibling (jPartList, the String... registerParts overload), so their deprecation is free. getIdMap() needs no pair or map replacement because the array index is the material ID; materialCount() plus the existing materialName(int) and getMaterial(int) covers every audited use.
  • Found a second and larger cleanliness problem: Scala's private[multipart] compiles to public, so seventeen implementation hooks are advertised as API. Verified all seventeen have zero external callers, and that the two similar-looking members that are load-bearing are bindPart (OpenComputers) and internalPartChange (ProjectRed). Fix is javadoc only, no descriptor change.
  • Confirmed that removing Scala from FMP has no pack-level dependency payoff: OpenComputers (843 Scala files) and ProjectRed (171) keep scala-library in the pack regardless. Recorded this under Phase 8 so the motivation is not restated later as a dependency argument.
  • Traced ScalaSignature.scala and ByteCodecs.scala to a single external caller, ProjectRed's ~60-line LightMicroblock trait. Rewriting it in Java deletes roughly 580 lines from FMP plus the Scala branch of ASMMixinCompiler, but it is not currently possible: registerJavaTrait rejects abstract classes while Microblock is abstract, and @SideOnly stripping exists only on the Scala signature path while LightMicroblock.renderDynamic is client-only and applied on both sides. Both gaps are recorded as Phase 7 prerequisites.
  • Added Phase 10 for upstream consumer cleanup. All 27 consumers are GTNewHorizons forks and therefore patchable; Extra Utilities is the sole exception and is constrained rather than fixed. The pattern is a three-step ratchet: FMP adds a supported equivalent (additive, safe now), the consumer is patched and released, then FMP drops the private shape in a release allowed to break that ABI.

2026-09-01

  • Characterized MultipartProxy against untouched Scala. Four plain-JVM cases freeze the four-type hierarchy, exact facade/companion surfaces, mutable singleton accessors, client-only boundaries and both chunk-index conversions. One Forge case freezes dedicated-server stripping, virtual fallback to the server implementation and initialized proxy state.
  • Ported the hierarchy and facade to four Java types without changing their emitted names or callable public descriptors. Common generated-tile registration now names MultipartProxy$.MODULE$ explicitly so it still reaches the inherited server method after Forge strips the client override and its static forwarder.
  • All 187 plain-JVM and all 96 Java 8 Forge dedicated-server tests pass. Startup registration and two bit-packing helpers are not a meaningful focused performance target; microblock/handler/proxies.scala is next.
  • Characterized MicroblockProxy against untouched Scala. Four plain-JVM cases freeze its four-type hierarchy, complete facade/companion ABI, eight mutable server fields, protected Scala saw list and exact lazy-renderer shape. Two Forge cases freeze side stripping, inherited server lifecycle resolution, item/ore/recipe registration and saw order.
  • Ported the hierarchy and facade to four Java types. The Scala MutableList remains for binary compatibility, the client renderer retains its field-only side annotation and lazy bitmap, and the saw-renderer closure becomes a direct iterator loop.
  • All 191 plain-JVM and all 98 Java 8 Forge dedicated-server tests pass. This startup-only proxy is not a meaningful focused performance target; multipart/handler/packethandlers.scala is the last Scala handler unit and is next.
  • Characterized the ForgeMultipart packet-handler unit against untouched Scala. Eleven plain-JVM cases freeze all six emitted retained types, exact facade/companion interfaces, channel identity, private prefixed state accessors, ordered registry and desync disconnects, control-key packets, coordinate streams and both update terminators.
  • Ported the five top-level types and nested byte stream directly to Java. The three Scala mutable collection descriptors remain where reflection already observes them, while direct loops replace the anonymous MultiMap and thirteen closure classes without changing watcher, batching, framing or cleanup behavior.
  • A clean build passes all 202 plain-JVM tests and the Java 8 Forge server passes all 98 tests. All six retained public surfaces match the reference by name and descriptor; only fourteen unreferenced compiler artifacts disappear. microblock/ItemMicroPart.scala is next.