Context
Since #19534 merged (b4065ca961), WorkstationHeaderStatePopOutNL fails on dev (6acb770a79, re-run 2026-10-10 00:3xZ). The arm is "the popped-out pane resolves the workspace provider and rea…". Its console-error assertion reads one extra line:
App Worker: PopupWorkspace: the landed pane's repaint render failed Worker port disconnected before reply: neo-port-2
A full Workstation and dashboard run on that base also failed two more arms with the same line. Both depend on timing:
WorkstationPopOutPaneIdentityNL "component, provider, store and descendant identities survive pop-out AND …";
WorkstationReclaimEmbodimentNL "control: a vessel that connected and then closed takes the release route …". This one passed alone on dev.
The Problem
PopupWorkspace#repaintLandedPane awaits pane.promiseUpdate(). #19534's RA-1 fix made the catch report every rejection on a surviving pane as a failed render. A popup window that leaves rejects that render without failing anything, and the pane survives, because a window disconnect is a render-target signal, not a destruction. The departure has two shapes:
- In flight: the frame was sent, then the worker retired the window's port.
worker.Base#removePort rejects the pending promise with a PortDisconnectedError ("Worker port disconnected before reply", no code). This is the line in the failing arm.
- After the port is gone: a frame sent to the departed window gets the dead-port answer (
code: 'NEO_DEAD_PORT').
(Corrected 2026-10-10 on review of #19536: the first version named only the dead-port shape, which the failing line is not.)
The Architectural Reality
apps/workstation/view/PopupWorkspace.mjs repaintLandedPane: catch (error) { if (pane.isDestroyed) return; console.error(...) }.
- The engine already classifies both shapes:
err?.name === 'PortDisconnectedError' for a rejection in flight, in src/mixin/VdomLifecycle.mjs, src/worker/App.mjs and the global rejection handlers in src/Neo.mjs;
src/worker/Base.mjs isDeparture(reason, windowId) for a dead-port answer: true when code === 'NEO_DEAD_PORT' and the window departed recently, as src/component/Sparkline.mjs uses it.
The Fix
Report only a rejection that is neither departure shape:
error?.name === 'PortDisconnectedError' || Neo.currentWorker.isDeparture(error)
|| console.error('PopupWorkspace: the landed pane\'s repaint render failed', error)The layer still comes off afterwards, as it does for any surviving pane.
Acceptance Criteria
Out of Scope
The pop-out lifecycle itself, and the repaint mechanism of #19531.
Related
- #19531 / #19534: the repaint and its RA-1 fix, which introduced the report.
- #19533: the e2e run that surfaced it.
Live latest-open sweep: checked the latest 20 open issues at 2026-10-10T00:35Z, plus searches for "repaint landed pane", "HeaderStatePopOut" and "departure console.error PopupWorkspace". No equivalent found. The A2A sweep (30, all read-states) shows no claim on this scope. The MC rationale sweep found nothing on this regression. My open assignments don't overlap.
Origin Session ID: fcc49158-63cf-4317-a407-741c5e5e3d8c
Context
Since #19534 merged (
b4065ca961),WorkstationHeaderStatePopOutNLfails ondev(6acb770a79, re-run 2026-10-10 00:3xZ). The arm is "the popped-out pane resolves the workspace provider and rea…". Its console-error assertion reads one extra line:A full Workstation and dashboard run on that base also failed two more arms with the same line. Both depend on timing:
WorkstationPopOutPaneIdentityNL"component, provider, store and descendant identities survive pop-out AND …";WorkstationReclaimEmbodimentNL"control: a vessel that connected and then closed takes the release route …". This one passed alone ondev.The Problem
PopupWorkspace#repaintLandedPaneawaitspane.promiseUpdate(). #19534's RA-1 fix made the catch report every rejection on a surviving pane as a failed render. A popup window that leaves rejects that render without failing anything, and the pane survives, because a window disconnect is a render-target signal, not a destruction. The departure has two shapes:worker.Base#removePortrejects the pending promise with aPortDisconnectedError("Worker port disconnected before reply", no code). This is the line in the failing arm.code: 'NEO_DEAD_PORT').(Corrected 2026-10-10 on review of #19536: the first version named only the dead-port shape, which the failing line is not.)
The Architectural Reality
apps/workstation/view/PopupWorkspace.mjsrepaintLandedPane:catch (error) { if (pane.isDestroyed) return; console.error(...) }.err?.name === 'PortDisconnectedError'for a rejection in flight, insrc/mixin/VdomLifecycle.mjs,src/worker/App.mjsand the global rejection handlers insrc/Neo.mjs;src/worker/Base.mjsisDeparture(reason, windowId)for a dead-port answer: true whencode === 'NEO_DEAD_PORT'and the window departed recently, assrc/component/Sparkline.mjsuses it.The Fix
Report only a rejection that is neither departure shape:
error?.name === 'PortDisconnectedError' || Neo.currentWorker.isDeparture(error) || console.error('PopupWorkspace: the landed pane\'s repaint render failed', error)The layer still comes off afterwards, as it does for any surviving pane.
Acceptance Criteria
removePort,isDeparture): a departure in either shape reports nothing and still removes the layer. A failed render, and a dead port for a window the worker never saw leave, are still reported. A destroyed pane stays untouched.WorkstationHeaderStatePopOutNL"the popped-out pane resolves the workspace provider…" is green on the fix, alone and in the full Workstation folder run.WorkstationPopOutPaneIdentityNL"component, provider, store and descendant identities…" andWorkstationReclaimEmbodimentNL"control: a vessel that connected and then closed…" show no repaint line in their console reads.Out of Scope
The pop-out lifecycle itself, and the repaint mechanism of #19531.
Related
Live latest-open sweep: checked the latest 20 open issues at 2026-10-10T00:35Z, plus searches for "repaint landed pane", "HeaderStatePopOut" and "departure console.error PopupWorkspace". No equivalent found. The A2A sweep (30, all read-states) shows no claim on this scope. The MC rationale sweep found nothing on this regression. My open assignments don't overlap.
Origin Session ID: fcc49158-63cf-4317-a407-741c5e5e3d8c