AOT: loadXML DTD ID getElementById finds only first element — later IDs null + detached from tree (ext/dom/document.c)

[PurHur]

Category

stdlib · AOT / ext/dom

Problem

User-script AOT DOMDocument::loadXML() with a DTD ATTLIST … ID indexes only the first matching element (DomParseSimpleXmlIdsJitHelper::scanFirstAttributeValue). Later IDs return null from getElementById(). The first hit is also a detached shadow node (=== tree child is false; parentNode empty), not the live documentElement child. Zend/VM return the live tree node for every ID.

Repro	Zend 8.2+ / VM	AOT (master @ eb638b2)
DTD two <e id="a|b">; getElementById('a')	tree child, parentNode=r, text 1	detached text 1, parentNode empty
getElementById('b')	tree child text 2	null

Root cause: (1) ID scan is first-only per element name; (2) materializeIndexedElement creates orphan map entries before syncChildrenFromXml; (3) thin-AOT getElementById single-slot cache miss falls through to NestedJIT only, never PROP_ELEMENT_ID_MAP.

php-src reference

• php/php-src ext/dom/php_dom.c — dom_document_get_element_by_id
• libxml xmlAddID / DTD ID attribute type

PHP implementation target

• ext/dom/DomParseSimpleXmlIdsJitHelper.php — scan all DTD ID / xml:id elements (first-wins per id value)
• ext/dom/JitDomDocumentElement.php / JitDomLoadXMLUserScript.php — register live syncChildren nodes into the id map; drop orphan materializeIndexedElement pre-pass
• ext/dom/JitDomGetElementById.php — on element-cache miss, consult PROP_ELEMENT_ID_MAP before NestedJIT

Repro

./script/docker-exec.sh -- bash -lc 'php bin/compile.php -o /tmp/g.bin test/repro/issue_XXXX_dom_loadxml_multi_dtd_id_aot.php && /tmp/g.bin'

Done when

• AOT getElementById for every DTD ID matches Zend textContent
• Returned node === corresponding documentElement child (parentNode is the root)
• Existing dom_loadxml_getelementbyid_user_script.phpt still green
• ./script/aot-smoke.sh stays 8/8
• AOT fixture under test/fixtures/aot/cases/

claim: DOM/XML agent — implementing multi-ID loadXML getElementById + live tree registration.

[PurHur]

claim: advanced-stdlib agent — implementing multi-ID DTD loadXML getElementById + live tree registration (same specialty as DOM/XML lane).