Publish jmsfx-battleorder, and correct what CLAUDE.md gets wrong

[ctgnz]

Release preparation for 2.0.0, whose milestone is now clear — 20 issues closed, #82 among them. Part of #144.

jmsfx-battleorder publishes

It was excluded from Central "until it can draw" — it carried no SVG fragments of its own, so it compiled and registered itself as an IconLibrary that rendered nothing, which is not a thing to put somewhere a release cannot be withdrawn from.

The exclusion named its own trigger: comes off this list once jmsfx#82 … settles how a library gets its fragments. That is settled. Fragments are injected into the generated code, so this library draws jmsfx-standard's through the tree it extends (#133), with no second IconLibrary on the classpath. It reports 0 classpath fallbacks across 17,513 load calls, and ServiceBranchFrameTest renders the actual branch colours.

So leaving it out would now be by default rather than by decision — which is exactly what that comment existed to prevent. The comment is rewritten to say why it publishes, rather than left stating a condition that has expired.

Five artifacts go to Central: jmsfx-parent, jmsfx-core, jmsfx-standard, jmsfx-historical, jmsfx-battleorder. Creator, editor, server and the generator stay unpublished — applications and build tooling, not things anyone resolves.

CLAUDE.md had four claims that stopped being true during the epic

This matters more than most stale prose, because it is the file that steers work here.

claim	reality
ModelComparator is "bound to verify"	Not since #137 — every extension is an overlay composed onto its base, so the superset property holds by construction and there is nothing for a build-time check to catch
jmsfx-historical "still carries a complete model"	#137 reduced it to its additions
overlayModel: true distinguishes an overlay, "which only jmsfx-battleorder is"	The field no longer exists. isOverlay() is baseLibrary != null — once historical became an overlay too, there was no distinction left to carry
historical keeps "314 additions"	356, with the extinct status and the bicycle drawing

Two facts added that later work needs: model elements nest their scalars under details: with the flow style on the record rather than the class, and the fallback count is asserted on every build rather than something to go and measure.

The last two hallux mentions

Neither was in code or build config — the rename to jmsfx-historical was clean, and no module, package, artifact or config still says hallux. But .gitattributes described keeping fragments "byte-comparable against hallux's copies", and those copies go away once historical is on Central, so the comment now says what the setting actually buys.

Verification

BUILD SUCCESS on -Pstandard, -Phistorical and -Pbattleorder, and the release profile builds clean with signing and javadoc skipped.

No version bump here — that is the next step on #144, and it needs -DprocessAllModules=true or twelve of the thirteen poms stay on 2.0.0-SNAPSHOT.

🤖 Generated with Claude Code