Fix Power Advantage 1.13.2 machine state and strengthen tests

.gitignore

@@ -59,6 +59,7 @@
 /test-rig/.gradle/
 /test-rig/.cache/
 /test-rig/build/
+/test-rig/logs/
 /test-rig/migration-probe/.gradle/
 /test-rig/migration-probe/build/
 /test-rig/migration/.venv/

CHANGELOG.txt

@@ -1,3 +1,14 @@
+Unreleased test-coverage repairs for 1.13.2
+
+- Keep steel frames non-opaque on dedicated servers as well as clients.
+
+- Keep machine timers, stored power and progress in the next world save, even when no inventory changes.
+- Charge energy sources only for power that another machine actually accepts.
+- Preserve item NBT and check every destination slot when moving conveyor items.
+- Drain waterlogged blocks without destroying their host, and do not discharge fluid through solid blocks.
+- Restore machine names, readable inventory layouts and bounded fluid gauges.
+- Add measured operating jobs, negative controls and exact saved-state restart checks. Keep visual acceptance separate from automated checks.
+
 Power Advantage 2.4.4.113021
 
 - Exclude empty directories so development jars match between existing workspaces and fresh CI checkouts.

README.md

@@ -50,6 +50,9 @@ The [client resource probe](test-rig/client-resources/README.md) checks native b
 
 ## Compatibility
 
+The [stronger regression-test backport](docs/1.13/TEST-COVERAGE-BACKPORT.md)
+records the local repairs, tested jar hashes and remaining manual checks.
+
 The implementation namespace is `zone.moddev.mc.poweradvantage`. The historical `cyano.poweradvantage.api` surface, including `FluidNetworkApi`, remains available to dependent mods. Registry names, fluids, tile IDs, NBT keys, OreSpawn provider IDs, and configuration meaning remain stable.
 
 PowerAdvantage owns the shared pre-flattening migration hook used by all three Advantage mods. See [the 1.13 migration guide](docs/1.13/MIGRATION.md) before opening an irreplaceable 1.10 or 1.12 world. The [Advantage Works rig](test-rig/README.md) provides repeatable machine, restart, world-generation, and migration checks.
@@ -73,3 +76,31 @@ normal chunk loading reconnects them without replacing pipes. The same protocol
 records acquisition-stack diagnostics and pipe-only bridge regressions.
 
 The historical 1.10 handbook remains on the [1.10.2 branch](https://github.com/MinecraftModDevelopmentMods/PowerAdvantage/blob/master-1.10.2/docs/1.10/README.md).
+
+## Generator Follow-Up
+
+The [generator follow-up](docs/1.13/TEST-COVERAGE-BACKPORT.md#generator-follow-up) records the older `pa113-candidate-g` inputs.
+Its 126 operating jobs and 126 cold checks, 39 GUI windows, 351 caption cases,
+355 rejected broken controls, eight frozen routes, five connected routes and
+paired 30-minute runs passed. This is a pre-final Steam-control checkpoint,
+not qualification of the newer SteamAdvantage jars or harness. Those still need
+the final sweep. Manual play-testing and release review remain separate.
+
+## Final T Candidate
+
+The [final-t checkpoint](docs/1.13/TEST-COVERAGE-BACKPORT.md#final-t-candidate)
+records profile `pa113-candidate-i` and its exact jar hashes. All 132 operating
+jobs and 132 cold-restart comparisons pass, including an independent audit of
+six Steam controls and their nine tile-hold records. GUI checks pass for 39
+windows, 351 caption cases and 355 rejected broken controls.
+Frozen migration passed on 8 routes and connected-source operation passed on
+all five declared routes.
+Fresh and migrated-copy 30-minute runs pass. Power and Electric retain their
+exact-hash release and reproducibility proof; final-t Steam passes its full suite
+and two independent export builds. All 18 mod/version release and reproducibility
+gates pass across the six maintained Minecraft lines. Both requested broad
+profiles and their independent audit pass; the reporter-only unit/replay
+qualification preserves the original native proof. Survival harvest checks pass
+before and after restart. The report's extra network/drill scopes are now
+qualified; manual acceptance remains separate. Earlier checkpoints keep their original scope;
+nothing has been published by this local test work.

build.gradle

@@ -1144,7 +1144,9 @@ tasks.register('prepareCombinedEclipseLaunch') {
     description = 'Creates an additional production-only Eclipse client launch with all three Advantage mods.'
     dependsOn tasks.named('verifyEclipseProductionClasspath')
     doLast {
-        def profile = new JsonSlurper().parse(file('test-rig/profiles/dev-core.json'))
+        File inputProfile = file(providers.gradleProperty('advantageDevelopmentProfile').getOrElse(
+                'test-rig/profiles/dev-core.json'))
+        def profile = new JsonSlurper().parse(inputProfile)
         advantageExtensionJars.each { String id, File jar ->
             def pin = profile.mods.find { it.id == id }
             if (!jar.isFile()) throw new GradleException("Build ${id} first, or override its *AdvantageDeobfJar property: ${jar}")

docs/1.13/CURRENT-QUALIFICATION.md

@@ -1,5 +1,10 @@
 # Current Candidate Qualification and Remaining Tests
 
+> The [stronger local regression candidate](TEST-COVERAGE-BACKPORT.md) has its
+> own jar hashes, measured results and limits. The archived results below apply
+> to the previous jars, not the modified sources. Do not combine their coverage
+> into a claim that every scenario has been rerun on the new candidate.
+
 ## Scope
 
 This is the 23 October 2026 rerun against the current, unpublished 1.13.2 jars.