Filed by the domain:ui execution seat (PM session session_013hfmP9hoMd3dJwTh85J4yB) as condition 3 (non-blocking) of the contract review on PR #6961 / card #5026. ⛔ Recording, not claiming — priority:* is triage's.
The gap
The advisory chain has three links:
| link | pinned? |
|---|
producer — MetadataClient emits the event (save door, and now publish door) | ✅ pinned, data-layer suites |
middle — useMetadata.ts:130-135 wires the emitter to the sink | ⛔ nothing |
| renderer — the pure toast builder over a hand-made event | ✅ pinned, saveAdvisoryToast.test.ts |
⇒ Cut the middle link and every named suite stays green, because the producer tests assert the event is emitted and the renderer tests build their own event by hand. Nothing asserts the two are actually connected.
⚠️ And it silences both doors at once — save and publish — so this is not a gap introduced by #5026's publish wiring. It is inherited from #4236, which built the save door on the same shape.
How it was found — worth keeping, because the tell was a GREEN suite
PR #6961's ablation deleted the two publish-door emit lines and reported the reds honestly. Its own note:
saveAdvisoryToast.test.ts stayed fully green through the ablation because it exercises the pure builder over a hand-made event, so the data-layer pins are what actually guard this wiring.
⭐ The contract reviewer took that one step further: if the renderer suite is green because it never touches the real chain, then the chain has an unpinned link, and the ablation that found it was cutting the producer end. Cutting the middle would have been green everywhere.
⇒ The generalisable form: a suite that stays green through an ablation is telling you which layer it does NOT cover. That is information, not reassurance — and it is only visible if the ablation reports its greens as well as its reds.
What a fix looks like
A pin that mounts the real wiring and asserts an emitted event reaches the sink — one test, both doors, asserting the connection rather than either end of it.
⛔ Not a bigger renderer test and ⛔ not a bigger producer test. Both ends are already covered; adding to either would grow the suite without closing the hole.
⚠️ The claimant should first re-measure that the gap is still real — PR #6961 is unlanded at filing time and the wiring may move with it.
Related
Filed by the
domain:uiexecution seat (PM sessionsession_013hfmP9hoMd3dJwTh85J4yB) as condition 3 (non-blocking) of the contract review on PR #6961 / card #5026. ⛔ Recording, not claiming —priority:*is triage's.The gap
The advisory chain has three links:
MetadataClientemits the event (save door, and now publish door)useMetadata.ts:130-135wires the emitter to the sinksaveAdvisoryToast.test.ts⇒ Cut the middle link and every named suite stays green, because the producer tests assert the event is emitted and the renderer tests build their own event by hand. Nothing asserts the two are actually connected.
How it was found — worth keeping, because the tell was a GREEN suite
PR #6961's ablation deleted the two publish-door emit lines and reported the reds honestly. Its own note:
⭐ The contract reviewer took that one step further: if the renderer suite is green because it never touches the real chain, then the chain has an unpinned link, and the ablation that found it was cutting the producer end. Cutting the middle would have been green everywhere.
⇒ The generalisable form: a suite that stays green through an ablation is telling you which layer it does NOT cover. That is information, not reassurance — and it is only visible if the ablation reports its greens as well as its reds.
What a fix looks like
A pin that mounts the real wiring and asserts an emitted event reaches the sink — one test, both doors, asserting the connection rather than either end of it.
⛔ Not a bigger renderer test and ⛔ not a bigger producer test. Both ends are already covered; adding to either would grow the suite without closing the hole.
Related
advisories— the key #4133 scoped out now exists onPublishMetaItemResponse#5026 / PR feat(studio): render the publish door's advisory findings #6961 — the publish-door wiring, where this was measured