Backport Wave 0: shared build infrastructure (26.1.2)

[chimericdream]

Wave 0 of the main → 26.1.2 backport. Plan: docs/backport-26.1.2/README.md.

This is the first of a stacked series — every later PR is based on this one, so merge in order:

26.1.2 <- shared-build (this PR) <- chimeric-lib <- 14 per-mod PRs

What's here

Commit
.gitattributes	Makes LF structural. Normalizes gradlew.bat (blob LF, worktree CRLF).
plan correction	See below.
root gradle.properties	parallel = true; the five *_compat floors centralized.
root build.gradle	chimeric-lib as an in-build project() dep; drops mavenLocal(), the GitHub Packages repo and GITHUB_TOKEN; wires the fabric test source set.
settings.gradle + tpl	Hoists chimeric-lib to the front of the include order.
gradle/mod-conventions.gradle	Every mod's build.gradle collapses to one apply from: line.
CI + docs	JDK 21 → 25 (the build is options.release = 25, so CI literally could not compile); docs/, CODE-REVIEW-PLAN.md, DEPENDENCY-PLAN.md; CLAUDE.md rewritten.

Wave 0 owns every *.gradle and gradle.properties in the repo, so no two later PRs collide on a build file.

Why the settings.gradle hoist

Loom resolves each mod's classpath in that mod's afterEvaluate; a consumer resolving chimeric-lib before it has configured fails with "project components has not been calculated yet". The hoist and evaluationDependsOnChildren() in chimeric-lib/build.gradle are both required — neither works alone.

Coordinates changed, filenames did not

archivesName: <name>-<platform>-26.1.2  ->  <name>-<platform>
version:      5.0.0-alpha.0             ->  26.1.2-5.0.0-alpha.0

The produced jar is byte-identically named, so scripts/create-modpacks.ts needs no change. Verified by running it.

Correction carried in this PR

The plan claimed ~4,000 payload files were CRLF churn. That was wrong — measured with git show <rev>:<path>, which applies eol conversion. Real census via git cat-file blob: 26.1.2 has 1 CRLF file (gradlew.bat), c5f2cc4d 12, main 9. Only 3 payload files mismatch, all in minekea and all already on its conflict list. README.md and minekea.md are corrected here.

Verification

• ./gradlew clean build — SUCCESSFUL, all 15 mods × both loaders.
• bun run build — 15 fabric + 15 neoforge jars, filenames unchanged.
• git diff 26.1.2..HEAD -- '*.java' — empty. Wave 0 changes no Java.

Not backported (deliberate)

docs/MC-26.2-NOTES.md (documents a port this branch hasn't done) and .claude/skills/mc-visual-smoke-test (its samples are 26.2 client APIs). docs/TESTING.md points at the latter with a note.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 238c7e8b-396d-4a2f-8439-129df4beb530

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch backport/26.1.2/shared-build

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chimericdream]

Stack index — merge in this order

This is a stacked series. Each PR's base is its parent, so each diff shows only its own work. GitHub retargets children automatically as parents merge.

#	PR	Base	Blocking?
1	#81 shared build infrastructure	26.1.2	yes — everything else is based on it
2	#82 chimeric-lib	#81	yes — 6 mods compile against its new API
3	#83 hopper-xtreme · #84 minekea	#82	no — the 14 below are mutually independent
	#85 sponj · #86 shulker-stuff · #87 houdini-block	#82
	#88 villager-tweaks · #89 enchantment-numbers-fix	#82
	#90 banner-tweaks · #91 archaeology-tweaks · #92 miniblock-merchants	#82
	#93 artificial-heart · #94 athenaeum · #95 beacon-conduit-tweaks · #96 flat-bedrock	#82

After #81 and #82 land, the 14 mod PRs can merge in any order — verified: no two touch the same file, and merging all 14 into a scratch integration branch produced zero conflicts.

Integration evidence

backport/26.1.2/integration is pushed as verification only — do not merge it. It is all 14 mod branches merged together, and the full acceptance suite was run against it:

Check	Result
./gradlew clean build	SUCCESSFUL — 15 mods × both loaders
bun run build	exit 0 — 15 fabric + 15 neoforge jars, filenames unchanged
GameTests	70 passed, 0 failed across 15 runs
chimeric-lib unit tests	43 passed, 0 failures/errors
26.2-only API leaks	none
Line endings	all LF except the intended gradlew.bat

What is not in this series

The 17 "update to 26.2" commits (e4817fa4..c5f2cc4d) — that is the 26.2 port, and backporting it would defeat the purpose. Mod versions stay on the 5.x line; root dependency versions stay at their 26.1.2 values.

Also skipped deliberately: docs/MC-26.2-NOTES.md and .claude/skills/mc-visual-smoke-test, both of which describe 26.2-only APIs.

Corrections found while executing

Two plan claims turned out to be wrong, both from inferring absence rather than checking. Both are corrected in the docs, in the PR that carries them:

1. CRLF churn (Backport Wave 0: shared build infrastructure (26.1.2) #81) — claimed ~4,000 files, actually 3.
2. Data-component binding is not 26.2-only (Backport Wave 1: chimeric-lib shared API, fixes and test harness (26.1.2) #82) — the bake is required on 26.1.2 too, which also flipped minekea's datagen instruction.

The underlying lesson, now recorded in the plan: git grep <symbol> 26.1.2 proves a symbol is used, never that it is absent. Use javap on the deobf jar for absence.