Follow-up from review of #963 (#963 (comment)).
The unit tests for the APS renderer registry hand-construct the bid Prebid "delivered", so they assert the assumption under test — that Prebid preserves meta sub-keys through bid normalization — rather than verify it against real Prebid. That is the same blind spot that let the original trustedServerRenderer regression ship: the mocked path was always green.
Extend crates/trusted-server-integration-tests/browser/tests/shared/aps-renderer.spec.ts (which drives pbjs.requestBids → interpretResponse → registry → Universal Creative against the vendored Prebid build) to:
- delete the custom top-level
trustedServerRenderer field from a bidAdjustment hook, then assert the registry still populates via the meta carrier;
- assert
meta.trustedServerRenderer is absent from pbjs.getBidResponsesForAdUnitCode(...) after the auction (the scrub now runs on bidAccepted, so no later-event consumer should observe the descriptor);
- keep the existing surviving-top-level-field scenario as-is so both carriers stay covered.
Follow-up from review of #963 (#963 (comment)).
The unit tests for the APS renderer registry hand-construct the bid Prebid "delivered", so they assert the assumption under test — that Prebid preserves
metasub-keys through bid normalization — rather than verify it against real Prebid. That is the same blind spot that let the originaltrustedServerRendererregression ship: the mocked path was always green.Extend
crates/trusted-server-integration-tests/browser/tests/shared/aps-renderer.spec.ts(which drivespbjs.requestBids→interpretResponse→ registry → Universal Creative against the vendored Prebid build) to:trustedServerRendererfield from abidAdjustmenthook, then assert the registry still populates via themetacarrier;meta.trustedServerRendereris absent frompbjs.getBidResponsesForAdUnitCode(...)after the auction (the scrub now runs onbidAccepted, so no later-event consumer should observe the descriptor);