Fix the dangling javadoc references jmsfx#136 left behind

[ctgnz]

Blocks #144. The 2.0.0 tag's second run failed here.

What broke

Six {@link} references to getGraphicLocation, which #136 renamed to getGraphicKey. javadoc with doclint=all treats an unresolvable reference as an error, and the release profile is the only thing that runs javadoc — so verify never saw them, on any build, for the whole epic.

file
MainElement.java:13	#getGraphicLocation(StandardIdentity)
AmplifierListItem.java:27	#getGraphicLocation(StandardIdentity)
HqtfDummy.java:28, :37	#getGraphicLocation(StandardIdentity, SymbolSet)
Status.java:55, :65	#getGraphicLocation(StandardIdentity, SymbolSet)

Only the arg-taking method was renamed. The no-argument getGraphicLocation() still exists on AmplifierListItem, Dimension and SymbolSet, where it is the authoring directory rather than a cache key — so a blanket rename would have been wrong, and I checked each of the ~100 remaining uses is real code rather than a doc link.

MainElement's paragraph was stale in three ways rather than one, so it is rewritten instead of patched: it described #124 as pending and said the categories #123 covers "still do" fall back to the classpath. Both are done, and since #130 no generated library returns null there for anything it can draw — which is exactly what the contract now asserts.

On the verification

I reported earlier that "the release profile builds clean". That run was -Dmaven.javadoc.skip=true -Dgpg.skip=true — it skipped both of the things -Prelease adds, so it proved nothing, and this is what it would have caught.

Done properly this time: a full -Pstandard,release clean package with javadoc enabled. No unresolved references, BUILD SUCCESS, and all four javadoc and sources jars produced. Then -Pstandard verify for Spotless, which reflowed one paragraph where the shorter method name moved the wrap point — and the release build re-run after that.

The pattern worth fixing separately

This is the third thing to fail only at tag time: the missing secrets, the stale artifact lists, and now javadoc. The doclint config's own comment records the same class of bug — "it is what caught a {@link} to a type the file never imported, the release profile having never been run before".

-Prelease drifts because nothing exercises it between releases. A CI step running javadoc:javadoc against the release profile's doclint would turn a broken {@link} into a failed PR check instead of a burned tag. Not in this PR — it costs CI time and that is a call worth making on its own — but I would suggest it as a follow-up before 2.1.

🤖 Generated with Claude Code