- The browse client half mirrors -native's rollback: a pair construction
that throws halfway (declared-but-occupied second hole) disposes the
earlier deferral before rethrowing, so no orphaned slot subscription
outlives the failed fiber (ds-review-bot).
- The TestWorkspaces browse methods gain the coverage CI's per-file gate
requires (defaults + stub overrides, recorded like their siblings).
A declared-but-occupied second hole registers synchronously, so the pair
construction can throw after the first deferral installed its subscription;
that orphan then fires against the failed fiber's inactive context. The
effect now disposes already-created deferrals before rethrowing
(ds-review-bot on the browse twin; same shape here).
master's new ctx.workspaces contract face (IWorkspaces + TestWorkspaces
double) predates this stack's listDirectory/createDirectory; widening the
face is the explicit act of letting the browse flow drive them, and the
double records/stubs them like its siblings.
The assembled fixture app now takes over the composer while fx-alpha's
resident approval is pending — the feature under test, not chrome noise —
so the session-title snapshot answers it (允许一次) before asserting the
model seat, mirroring what a user sees. The eleven keyless ui.expected.md
goldens re-record (DSH_SNAPSHOT=refresh, no key) for the Access placeholder
→ projection chip swap (Danger Full Access under the roster default) plus
the master-side chrome drift the goldens had accumulated (copy/branch/edit
buttons, tab rows) since their last refresh.
The runtime's new outward session contract exposes projections as a
ProjectionsFace (faceOf only); the ui-goal verb closure reads the current
value through faceOf('goal').getSnapshot() instead of the store-level get.
Catalogs regenerated after the merge.
- HMR/unmount bumps both request generations, so a listing or creation
settling after disposal neither updates dead state nor issues the
post-create relist.
- A selection whose preview listing fails clears back to the single pane:
the breadcrumb names the level, so the level must be the committing
target (reverses the earlier keep-marked behavior with its test).
The TUI keyless smoke exposed the wrong idle signal: /plan submitted while
the driver is between turns still saw status running (checkpointing), so
the selection queued and the immediate copy never appeared. The commit
gate is now the log's own open-turn fold — plan/mode commits at once
whenever no turn is open, and queues only inside one. Specs drive the
distinction through explicit turn/start / turn/end events instead of a
fake status; the keyless smoke pins the new idle copy on both /plan and
/plan off; regenerated cordis/service catalogs ride along.