[SirEdvin]

Summary

• Upgrade Broccolium, Tweakium and test-only Testiarium to exactly 2.0.0-rc.1 on Minecraft 1.20.1 Fabric/Forge and 1.21.1 Fabric/NeoForge, starting from verified multiversion merge 64720fcd02c6bc14922f0b99b55961255497ddb1.
• Select published Testiarium testMod/cctTestMod feature capabilities; retain Fabric's detached fixture-file path to avoid overlapping Loom variant-remap outputs and exclude the redundant transitive KFF/FML loader on the NeoForge test classpath.
• Align CC:Tweaked and Fabric dependencies with the RC graph, expand the version-specific CC metadata correctly, and enforce the qualified RC in jar-in-jar metadata/auditing.
• Keep production Kotlin/Java, saved-data formats, public Lua/TypeScript contracts, registry/upgrade IDs, config semantics and independent npm versions unchanged. Update the pocket conservation fixture for CC's public native registry lookup; strengthen persistence checks to compare complete item/fluid metadata and optionally preserve/load same-version snapshots across processes.

Local verification (source-identical candidate)

All Gradle commands use the repository wrapper, --no-daemon, explicit deadlines and complete saved logs.

• TypeScript fixture compilation: exit 0 (14s).
• Full six-leaf/four-node build plus test-mod classes and publication metadata: exit 0 (16s).
• Final aggregate native GameTests: exit 0 (85s), fresh 13 cases per node / 52 total, no failures, errors or skips. Includes all nine original Lua fixtures and the three occupied/partial/full/empty conservation regressions.
• Four native datagen tasks: exit 0 (48s); tracked resources unchanged. Removed only 56 newly generated byte-identical copies of shared assets after reconciliation.
• Separate JVM write/read phases: complete named/damaged oversized item, tagged/component-bearing fluid, energy, IDs, owners and decay metadata survive copied native saves on every node. Also generated snapshots with the old 1.4.3/1.4.6 libraries at the baseline (test-only fixture overlay) and read them with the RC in separate processes, same Minecraft version only. An intentionally altered expected item quantity fails exactly the required persistence case; the final positive matrix is green.
• New-empty-cache external wrapper consumer: 34 transitive configurations, 12 fixture feature selections, 90 own-artifact occurrences / 30 unique binaries all matching the independently verified public publisher receipt. Actual DigitalItems configuration graphs and artifact hashes recorded separately (24 compile/runtime configurations).
• Final four release JARs pass class/resource/loader-metadata/nesting/publication audits, no test-only dependency leakage; all eight nested library JARs equal the qualified public bytes, with exact Forge-family [2.0.0-rc.1] ranges.
• Documentation tests: 20/20; report-validator test: 1/1; documentation build: exit 0 (13s).

Relevant exact-head hosted build, GameTests and documentation checks must pass before handoff. Existing workflows already cover the literal digital-items/** branch head; no workflow, repository policy or Pages environment changes are included.

Boundaries / next gate

This is the implementation PR for card t_b245d69c; qualification/review/merge belongs to t_f3b1aa6d. Do not merge this PR as part of the implementation card. No 26.1.2 node, release/npm publication, user-world access or cross-Minecraft-version conversion. New interactive client/menu qualification and post-merge CI remain with the qualification gate; persistence evidence here is native saved-data restart coverage, not a real socket/client walkthrough.

[SirEdvin]

rc.1 qualification gate — reviewed head 3ce14136c6e585ccd1b8493e89d8487f1e0efff4

Reviewed the original 12-file upgrade against accepted multiversion base 64720fcd02c6bc14922f0b99b55961255497ddb1. Frozen source hashes, all 179 inherited archive payloads, public resolution, datagen, native save/reload and old-library-to-RC restart receipts reconcile. Production Kotlin/Java, stored-data layout, registry/upgrade IDs, config and typed API are unchanged. Final four release JARs equal the previously qualified public-RC binaries byte-for-byte.

Required node	Native cases	Real dedicated server + connected client	Basic/advanced block, turtle, pocket	Both menus: client move + authoritative server quantity	Copied full-world restart
Fabric 1.20.1	13/13	PASS	PASS	PASS	PASS
Forge 1.20.1	13/13	PASS	PASS	PASS	PASS
Fabric 1.21.1	13/13	PASS	PASS	PASS	PASS
NeoForge 1.21.1 (forge task path)	13/13	PASS	PASS	PASS	PASS

Qualification commit changes only the retained TypeScript fluid/energy fixture bodies. They now assert mismatched simulated extraction, invalid destination ID, incompatible/full insertion rejection, partial admission, physical + digital conservation, recovery, exhausted IDs and exact rejection errors. A deliberate expected1001/actual1000 calibration fails precisely those two native Lua cases; restored final matrix passes 52/52, no failures/errors/skips. This verifies consumer rejection/rollback-by-no-mutation semantics through the existing storage provider fixtures, not every third-party mod's transaction implementation.

Final local wrapper commands use explicit deadlines, --no-daemon and complete saved logs:

• TypeScript fixture compile: exit 0, 34s.
• Aggregate gameTest under xvfb-run -a -PminimalTestEnvironment: exit 0, 103s; fresh 13 cases per node.
• Full build testModClasses generatePomFileForMavenPublication generateMetadataFileForMavenPublication: exit 0, 11s; packaging audit passes.
• Runtime uses each actual :<loader>:<minecraft>:runServer / runClient with a 600s deadline and scratch init script. All 12 final native processes exited normally. Full commands/durations/logs, screenshots, copied disposable worlds and receipts are preserved by card t_f3b1aa6d.

Runtime was executed at b7d442d; the final head only changes Lua test bodies, and all production JAR bytes are identical. Existing method-inventory contract tests and item/fluid/energy representations pass. Basic/advanced left/right turtle, back pocket and one-based inventory slots work; old/RC effective slot validation and owner source paths are unchanged.

All 12 literal-head GitHub Actions checks must be successful before the authorized history-preserving merge; downloaded PR run 40221799638 independently passes 52 native cases and its four release JARs equal local bytes.

Boundaries: private-Xvfb/software-rendering correctness, not GPU/performance qualification; same-Minecraft-version disposable worlds only, no user-world access/cross-version conversion. The inherited Tweakium creative_filler inventory-model warning is identical in baseline/RC, with no DigitalItems model warnings. The existing Pages environment still permits only 1.20/1.21; a multiversion Pages deployment is not claimed or authorized to change. Documentation generation/check workflows remain required. No 26.x implementation, default-branch/policy edits, release/npm publication or history rewriting.

[SirEdvin]

Post-merge qualification verified at 48ca204: GameTests/build run40222441686 and Documentation checks run40222441667 succeed. Downloaded merge reports independently pass52/52 with no failures/errors/skips; four production JARs equal the tested local/head bytes. Separate Pages run40222441663 generates/assembles successfully but deploy fails the unchanged multiversion environment restriction; no live-site update claimed. Gate evidence, copied same-version worlds and registered hosted reports are preserved on t_f3b1aa6d. No26.x work or publication performed here.