Metadata extraction fails on unbound prefixes (e.g. gcse:search) during DTS publish

[natechadwick-intsof]

Summary

Metadata extraction fails on published HTML with unbound prefixes (e.g. Google CSE):

SAXParseException: The prefix "gcse" for element "gcse:search" is not bound
PSMetadataDeliveryHandler / PSMetadataExtractorService (RDFa / Semargl)

File publish may still succeed; metadata for that item errors.

Proposed fix

Lenient handling: sanitize/strip unknown prefixes or catch parse errors, WARN with path, do not fail entire extraction hard for third-party tags. Unit test with <gcse:search> fixture.

Workaround

Fix/remove CSE markup on the page (e.g. error.html) and re-publish.

Related

• Main-line sister: Metadata extraction fails on unbound prefixes (e.g. gcse:search) during DTS publish percussioncms#2355

[vijaya-boddipudi]

QA results (2026-08-26)

• QA: Blocked
• Instance: http://localhost:9992 (observed 8.1.8 Build 202608X25)

Why blocked

Issue #4 is a DTS metadata extraction defect (unbound prefixes / <gcse:search> → SAXParseException), not an Explorer UI qa task. There is no numbered Explorer test plan on the issue. Fix already landed in merged PR #37 (unit fixtures + lenient sanitize/WARN).

Steps attempted (Admin)

Step	Result
1. Login Admin at /login	PASS — classic dashboard (Welcome, Admin)
2. Open requested /cm/app/explorer	FAIL — branded 404 “This page cannot be found” (SPA Explorer route not on this 8.1.x UI)
3. Open Publish (/?view=publish → testsite)	PASS — Publish UI loads; recent Full jobs Completed
4. Acceptance (gcse unbound prefix → no metadata ERROR; WARN; dcterms; unit test)	BLOCKED — needs DTS publish of page with <gcse:search> + metadata/server log review (or unit-test gate from PR #37), not Explorer chrome

Pass / fail vs acceptance

Cannot mark Pass or Fail on product acceptance from Explorer UI alone. Recommend either:

1. Close human UI retest as N/A (covered by PR fix(metadata): lenient RDFa extraction for unbound XML prefixes (#4) #37 unit evidence), or
2. Provide a page with CSE markup + DTS publish checklist and re-run against metadata logs.

[vijaya-boddipudi]

QA results (2026-08-26 retest)

• QA: Blocked
• Instance: http://localhost:9992
• Product: Version 8.1.8 Build 202608X26 (0) (About dialog)
• Screenshots: D:\release_build\tmp\qa-26-08\4\

Issue scope vs requested URL

Issue #4 is a DTS metadata extraction defect (SAXParseException unbound prefix e.g. <gcse:search> in PSMetadataExtractorService / PSMetadataDeliveryHandler). Acceptance criteria (from sister #2355 / merged PR #37):

1. HTML with unbound gcse:search does not fail metadata delivery with ERROR for that reason alone
2. WARN identifies page path + problem prefix
3. Normal dcterms / RDFa still extracted when present
4. Unit test with gcse fixture

There is no numbered Explorer UI test plan on #4. Fix already merged: #37

Steps attempted (Admin)

Step	Result
1. Login Admin at /login	PASS — classic dashboard (Welcome, Admin) — 01-login.png, 02-dashboard-admin.png
2. Confirm version (About)	PASS — 8.1.8 Build 202608X26 — 03-about-version.png
3. Open requested /cm/app/explorer	FAIL (env) — branded 404 “This page cannot be found” (SPA Explorer route not on this 8.1.x UI) — 06-explorer-404.png
4. Open Publish (/?view=publish → testsite)	PASS — Publish UI loads; recent Full jobs Completed (CDP text + 05-publish-testsite-logs.png)
5. Acceptance: gcse unbound prefix → no metadata ERROR; WARN; dcterms; unit test	BLOCKED — needs DTS publish of a page containing <gcse:search> + metadata/server log review (or rely on PR #37 unit evidence). Not verifiable via Explorer chrome

Pass / fail vs acceptance

Cannot mark Pass or Fail on product acceptance from the Explorer UI alone on this instance.

Recommendation: treat human UI retest as N/A (covered by PR #37 unit evidence), or supply a page with CSE markup + DTS publish checklist and re-run against metadata logs.

Roles provided (Editor / Contributor) not exercised — Admin sufficient to confirm blocker.

[vijaya-boddipudi]

Re-opening to re-validate