Context
#19061 (#19056) settles an unhandled NEO_DEAD_PORT rejection only when the window it targeted departed: reason.code === 'NEO_DEAD_PORT' && isWindowDeparted(reason.windowId ?? reason.destination). That is the two-half predicate #18741 established. Four call sites catch the same rejection themselves, because each also has a second purpose: naming its subsystem, or settling Neo.isDestroyed. They decide "departed" in two different ways:
| site |
predicate |
container/Viewport.mjs:144 |
both halves (NEO_DEAD_PORT + isWindowDeparted(windowId)) |
draggable/DragZone.mjs:445 |
code only |
grid/Body.mjs:1321 |
code only (+ Neo.isDestroyed) |
table/Body.mjs:631 |
code only (+ Neo.isDestroyed) |
The Problem
The code-only sites swallow a dead port whose window never existed, which is a real misroute. #19067 is the latest proof that one exists: a message for a departed window was being delivered to a sibling. The central boundary would surface that error; these four sites silence it. It is the same rule written four times with two meanings. @neo-opus-ada agreed to the shared predicate on #19061's fork.
The Fix
Add one predicate, worker.Base#isDeparture(reason, windowId), that answers the two halves. onUnhandledRejection and the four catches call it in place of their inline tests. The code-only sites tighten to both halves; their Neo.isDestroyed handling and subsystem messages stay as they are.
Acceptance Criteria
Out of Scope
- Deleting any of the four catches. Each has a named second purpose (#19061's AC-4).
- The
DomEvent and Overflow sites, which check before the call and skip the work.
Related
- #19056 / #19061 — central settlement
- #18741 — the departure predicate
- #19067 — the misroute a code-only predicate hides
- #19011 — Viewport's both-halves precedent
Sweeps: live latest-open at 2026-09-23T00:29Z; open issues for isDeparture: none. Own assignments: none overlapping.
Origin Session ID: 19051047-2f38-4bf0-be51-efd6b957564d
Retrieval Hint: query_raw_memories("isDeparture shared predicate NEO_DEAD_PORT code-only catches Viewport DragZone grid Body table Body")
Context
#19061 (#19056) settles an unhandled
NEO_DEAD_PORTrejection only when the window it targeted departed:reason.code === 'NEO_DEAD_PORT' && isWindowDeparted(reason.windowId ?? reason.destination). That is the two-half predicate #18741 established. Four call sites catch the same rejection themselves, because each also has a second purpose: naming its subsystem, or settlingNeo.isDestroyed. They decide "departed" in two different ways:container/Viewport.mjs:144NEO_DEAD_PORT+isWindowDeparted(windowId))draggable/DragZone.mjs:445grid/Body.mjs:1321Neo.isDestroyed)table/Body.mjs:631Neo.isDestroyed)The Problem
The code-only sites swallow a dead port whose window never existed, which is a real misroute. #19067 is the latest proof that one exists: a message for a departed window was being delivered to a sibling. The central boundary would surface that error; these four sites silence it. It is the same rule written four times with two meanings. @neo-opus-ada agreed to the shared predicate on #19061's fork.
The Fix
Add one predicate,
worker.Base#isDeparture(reason, windowId), that answers the two halves.onUnhandledRejectionand the four catches call it in place of their inline tests. The code-only sites tighten to both halves; theirNeo.isDestroyedhandling and subsystem messages stay as they are.Acceptance Criteria
isDepartureexists onworker.Base, andonUnhandledRejectionuses it with no change in behaviour (the existingWindowDeparturearms stay green).console.error. One unit arm per site covers the newly visible case.NEO_DEAD_PORTcomparison remains insrcoutsideworker/Base.mjs.Out of Scope
DomEventandOverflowsites, which check before the call and skip the work.Related
Sweeps: live latest-open at 2026-09-23T00:29Z; open issues for
isDeparture: none. Own assignments: none overlapping.Origin Session ID: 19051047-2f38-4bf0-be51-efd6b957564d
Retrieval Hint:
query_raw_memories("isDeparture shared predicate NEO_DEAD_PORT code-only catches Viewport DragZone grid Body table Body")