[bugfix] cover native collection.xconf.xsd via OASIS catalog + jaxp

[duncdrum]

Description:

Adds NativeSchemaJaxpCatalogGapTest to close the integration gap left by #6528, and fixes the gap it found:

• Green control: validation:jaxv-report + XSD 1.1 language URI against schema/collection.xconf.xsd (valid minimal collection.xconf; empty one fails xs:assert).
• Previously missing path (now fixed): validation:jaxp-report via an OASIS catalog whose <uri> maps http://exist-db.org/collection-config/1.0 to an absolute file: URI of that same native XSD — mirrors the WEB-INF/catalog.xml → $EXIST_HOME/schema/ wiring tools are meant to use.

Root cause: Jaxp#newXsd11Validator wired an OASIS/system catalog's org.xmlresolver.Resolver in directly as the XSD 1.1 validator's LSResourceResolver. That resolver's resolveResource() requires a non-null systemId before it will consult the catalog at all, so a <uri> entry keyed purely by namespace was invisible whenever the instance carried no schemaLocation hint — even though the same catalog entry already worked for the default SAX pipeline, which falls back to the namespace as the systemId (see XercesXmlResolverAdapter#resolveEntity).

Fix: XercesXmlResolverAdapter#asLSResourceResolver applies that same namespace-as-systemId fallback. Code review of the first fix found validation:jaxv()'s SchemaFactory catalog wiring had the identical unfixed gap (same raw-Resolver pass-through, for xs:import/xs:include resolution during schema compilation), so it's fixed the same way and both call sites now share a single asLSResourceResolver(LSResourceResolver) dispatch instead of duplicating the instanceof check.

Existing coverage did not catch this: JaxpXsdCatalogTest XSD 1.1 cases use directory-search over /db/.../, which SearchResourceResolver already resolves purely by namespace; CollectionConfigSchema6000Test / Maven validate use raw SchemaFactory, not validation:jaxp.

What Changed

• exist-core/src/main/java/org/exist/resolver/XercesXmlResolverAdapter.java — new asLSResourceResolver(Resolver) wrapping an org.xmlresolver.Resolver as an LSResourceResolver with a namespace→systemId fallback, plus an asLSResourceResolver(LSResourceResolver) overload that dispatches on the concrete resolver type (only wrapping when it's a Resolver) so both call sites below share one implementation.
• exist-core/src/main/java/org/exist/xquery/functions/validation/Jaxp.java — newXsd11Validator now routes its catalog resolver through the new adapter before handing it to the XSD 1.1 Validator.
• exist-core/src/main/java/org/exist/xquery/functions/validation/Jaxv.java — same fix for the SchemaFactory#setResourceResolver catalog wiring.
• exist-core/src/test/java/org/exist/xquery/functions/validate/NativeSchemaJaxpCatalogGapTest.java — the regression test; updated comments now that the gap is closed, and escapes the generated catalog's file: URI attribute (Path#toUri() doesn't XML-escape characters like & a checkout path could contain).

Reference:

• Closes validation:jaxp + OASIS catalog does not apply native XSD 1.1 schemas from $EXIST_HOME/schema #6686
• Related: Ship versioned native XSD schemas with governance CI #6528 / [feature] add exist-db specific schema files to distribution output #6189, [feature] use native schemas provided by exist eXide#842

Type of tests:

JUnit (ExistXmldbEmbeddedServer) in exist-core.

mvn -pl exist-core -Dtest=NativeSchemaJaxpCatalogGapTest test

Test Plan

• NativeSchemaJaxpCatalogGapTest — 3/3 green (was 2/3, 1 expected-red)
• JaxpXsdCatalogTest — 12/12 green (no regression)
• Full validate/validation package (JaxpParseTest, JaxvTest, JaxpDtdCatalogTest, JaxpSchemaLocationSecurityTest, JaxpXsd11DetectionCacheTest, IsMissingElementDeclarationTest, Jing*/Parse* suites, etc.) — 77/77 green
• CI: all checks green except a pre-existing network flake (unparsed-text-from-url, unrelated fn:unparsed-text() GitHub-raw fetch)

Made with Cursor

[github-actions]

📊 XQTS result comparison

Comparison of this run against develop.

Metric	develop	this run	Change
➖ Passed	28,846 (90.67%)	28,846 (90.67%)	0 (0.00 pp)
➖ Failures	1,568	1,568	0
➖ Errors	134	134	0
➖ Skipped	1,267	1,267	0
🧪 Total tests	31,815	31,815	0

Relative to develop: 1 newly passing, 1 newly failing, 0 new errors, 0 newly skipped — counting only tests recorded in both runs whose outcome changed.

🔴 Newly failing tests (1)

• Constr-inscope-1 (was passing)

🟢 Newly passing tests (1)

• Constr-inscope-2 (was failing)

_{Runtime: 466.4s (+1.25s vs develop).}

[line-o]

LGTM <3