Keep visible Planning Mode controls bound to the current turn during asynchronous refreshes.
- preserve active question and plan-review workspaces during same-session hydration
- submit from live session state and retain dirty answers across response identity changes
- add desktop and mobile regression coverage for deferred hydration and stale Stop polling
- document the ownership race and add a patch changeset
Files changed:
.changeset/fn-9117-planning-turn-ownership.md | 7 +
.../suite-only-flakes-observed-register.md | 24 ++++
.../dashboard/app/components/PlanningModeModal.tsx | 141 ++++++++++++---------
.../PlanningModeModal.planning-flow.test.tsx | 132 +++++++++++++++++++
.../PlanningModeModal.ui-interactions.test.tsx | 20 +++
5 files changed, 265 insertions(+), 59 deletions(-)
Fusion-Task-Id: FN-9117
Fusion-Task-Lineage: 18d63f63-b532-49eb-a6d1-241e31c5b2c7
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Measured wall times: gate 14.0s, boot smoke 20.4s (-6.3s), changed-only
pnpm test 17.5s; quarantine ledger 0. Shard-timing snapshot rebuilt
from today's CI shard artifacts (run 31929730933) plus a locally
measured full dashboard suite so dashboard lane weighting keeps its
per-file data. The trim shows: the former top-6 core PG offenders are
gone from the slowest-20 (sqlite-migrator 2m29s serial -> 20.4s;
SettingsModal.general off the table entirely). Caveat: the dashboard
rows come from the pre-trim analysis measurement, so entries like
SettingsModal.scheduling-merge (30.2s, now ~13s) are pessimistic until
the next dashboard re-measure.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
One clean solo-lane sighting of the parametrized duplicate-response
reconciliation test (mobile row; the desktop row failed once earlier
under a contaminated concurrent run, recorded as context). Third
distinct flaky test in this suite plus one prior FN-8936 stabilization
— the AGENTS.md repeated-quarantine rule calls that a subsystem
product-race smell, so the product-path audit is filed as FN-9116
rather than stabilizing a fourth time.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Second root cause of the browser-back flake: the board is a lazy chunk
behind Suspense fallback={null} and waitForAppShell only proves the
header, so a synchronous getByText raced chunk resolution under
sharded-lane load (failure DOM showed a header-only page). The first
board lookup in both back-nav tests is now the async finder — the
correct wait mechanism, no widened timeouts. Complements the
once-queue purge fix; 3/3 solo runs green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Local e2e runs regenerated the fn-8602/fn-8607 baselines with small rendering
deltas and produced the FN-8806 task-title-stable-modal baselines that were
never checked in.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
check:changesets enforces a 120-char summary; this legacy entry was 133 and
failed the PR-check gate.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The FN-8923 drift guard was red on main: the fixture predated the FN-9047-9062
workspace wave and regeneration was blocked by 18 unclassified workspace
TaskStore methods. Classify the FN-9059 workspace coordination surface (lease
acquire/renew/release/reclaim/reconcile, land-intent record/resolve, fence-ref
recording, mergeWorkspaceWorktreeEntry, withValidWorkspaceLease as writers; the
inspect/list/find/validate reads as non-writers), regenerate the manifest, and
supply lifecycle verdicts for the 13 new call sites: lease/intent writes in the
land path are already-fenced by FN-9059 store-seam fence-token validation, the
land-failure note is benign-unfenced bookkeeping, and the remaining sites carry
their enclosing symbols' out-of-frontier verdicts. Also drops the deleted
diff-volume gate module from scannedModules.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
An SSE subscription with a nonexistent projectId (stale client tab, e2e fixture
page using projectId "fixture") surfaced the PG startup-factory construction
chain as a 500 on every poll, filling operator logs with alarming
"failed to construct TaskStore" errors. Map project-not-found to a clean 404,
matching the project routes' existing handling.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The gate blocked approved clean-room squashes on per-file shrinkage with no
override path ("AI merge diff-volume gate blocked the approved squash").
Removed by operator decision: delete checkDiffVolume/DiffVolumeRegressionError,
the merge:diff-volume-blocked audit event, the runDiffVolumeGate call sites in
every legacy squash finalizer, the AI-merge pre-land check, and the
mergeDiffVolume* settings. File scope remains the pre-land guard; the
post-squash audit policy remains the shrinkage backstop.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The mobile-styles guard forbids raw px anywhere in the stylesheet,
comments included; the FN-8866 token-fix comment mentioned the named
steps' pixel values and tripped it. Reworded — CSS rules unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Root-cause fix for the order-sensitive 'closes board-opened main-panel
task detail on one browser back' flake: vi.clearAllMocks() clears calls
but never drops unconsumed mockResolvedValueOnce entries, and a plain
mockResolvedValue default does not purge them either — the once-queue
wins first. A prior test's unconsumed auth/settings/health/plugin-views
Once value poisoned the next test's first fetch, rendering an
auto-opened modal surface instead of the board. The global beforeEach
now mockReset()s every once-queue-prone API mock before re-applying its
default (fetchPluginDashboardViews gains an explicit empty default
since its factory impl dies with the reset). 4/4 consecutive full-file
runs green (was ~1-in-3 failing); no assertions changed, no quarantine
needed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Dashboard bare-run repair, mock-drift cluster. Six route suites'
inline @fusion/engine mocks predated FN-8902's
DEFAULT_MODEL_REGISTRY_REFRESH_TIMEOUT_MS import and now wrap the
canonical createEngineMock fallback proxy; App.test's useTasks mock
becomes an importOriginal spread so the real mergeTaskSnapshot reaches
components, its Todo/graph tests move to FN-8762's plugin-view
enablement invariants (with a mockReset fix for a stale
mockResolvedValueOnce queue the clearAllMocks reset never drops); the
Mailbox lucide mock learns FN-9014's Archive icon; and two source-scan
tests repoint files moved by the #2398 domain-folder refactor
(app/api/agents/agents.ts, engine healing/self-healing-constants.ts),
both verified via git log --follow. Verified 10 files / 252 tests green
under their standard lane projects; both dashboard typechecks clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Dashboard bare-run repair, route/store cluster. Real product regression
fixed: the FNXC:IntakeOwnership boundary replaced the 'Workflow ... not
found' message with a typed TaskIntakeOwnerResolutionError, so the
mission triage routes' message-pattern mapping stopped firing and an
unknown workflowId leaked as a 500 — both feature and slice triage
handlers now match the typed error structurally and return 404 (patch
changeset included). Everything else was stale fixtures behind the PG
cutover and recent seams: sse's mock stores learn getAsyncLayer, the
retry fixtures learn FN-8908's resetTerminalFailureAutoRecoveryBudget,
approve-plan tests materialize a real on-disk PROMPT.md per the
SpecLockApproval 409 contract (fingerprint assertion strengthened to
the always-hash contract), and the MCP settings route body pins the new
host-dependent fusionMemoryMcpAvailable field. Verified 7 files / 213
tests green under their assigned lane projects, src typecheck clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Dashboard bare-run repair, component-drift cluster. All nine suites
were stale expectations behind cited intentional commits — no UI
regressions found: FN-8826's always-on WIP lifecycle badge
(TaskCard.oversight now asserts no overseer element instead of an empty
header), FN-8762's Todo Lists plugin extraction (MobileNavBar,
MainContent.graph-popout, useAppSettings repurposed to the
pluginDashboardViews contract, with a regression pin that the removed
todosEnabled field never returns), FN-8796's clock-arbitrated
mergeTaskSnapshot, FN-8797's prompt-only planning refresh, FN-8702's
767.98px phone-sheet boundary, FN-8764's primary-role split, and
FN-8947's added touch-target selector. Every update carries an FNXC
comment citing the causing commit. Verified 9 files / 293 tests green,
tsconfig.app.json typecheck clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The browser-lane suite regenerates these PNGs when it runs locally; six
regenerated baselines were swept into 9673f15c11 from the shared index.
Restore the committed baselines — no intentional visual re-baseline
happened.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Dashboard bare-run repair, census/token cluster. Real product bugs: the
Command Center activity panel (FN-8866) and structural-mail badge
(FN-8872) referenced undefined --space-* tokens, zeroing their
gaps/padding — mapped to the defined named scale; the settings search
index lagged FN-8829/FN-9021 additions and FN-8855's requiredChecks
entry had no scroll anchor (now a SettingsTextRow). Test-side: the
theme census learns FN-8730's intentional midnight theme, and the
Chromium touch-resize suite self-gates with describe.runIf per the
sibling browser-lane convention (CI/FUSION_BROWSER_SMOKE_REQUIRE still
fail loudly; all 62 tests still run where Chromium exists).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Gateway 502/503/504 non-JSON bodies (for example Traefik "no available server")
were dumped as content-type diagnostics into the Planning error banner.
## What
- Retry post-merge target pushes after recognized transient Git
transport failures.
- Use a bounded schedule: the initial attempt plus two retries after 2s
and 5s.
- Apply the same helper to the unified fast path and the shared push
path used by divergence recovery and the soft-deprecated merger.
- Keep retry sleeps and subsequent attempts cancellation-aware.
- Add a patch changeset and regression coverage, including a real bare
remote that rejects the first push.
## Why
Fusion already retries non-fast-forward races, and #1942 made terminal
push failures durable, but a temporary network failure still ended
post-merge delivery after one attempt. That can leave the local
integration branch ahead of the remote even though retrying seconds
later would succeed.
The retry is deliberately provider-neutral. It uses Git error
classification and normal Fusion logs only; it does not add Telegram,
OpenClaw, or any other notification-vendor dependency.
## Behavior and impact
- Retries only transient transport signatures such as connection resets,
DNS failures, unreachable networks, selected HTTP 429/5xx RPC failures,
and unexpected disconnects.
- Permission, authentication, configuration, and ref-rejection errors
keep their existing immediate handling.
- Exhausted retries remain non-fatal to the already-landed merge and
flow through the existing audit/task-log failure reporting.
- Existing non-fast-forward pull/rebase recovery is unchanged apart from
making its backoff cancellation-aware.
## Checks
- [x] `pnpm --filter @fusion/engine exec vitest run
src/__tests__/merger-ai-push-after-merge.test.ts
src/__tests__/merger-prompt-and-utils.test.ts --silent=passed-only
--reporter=dot` (48 tests)
- [x] `pnpm --filter @fusion/engine typecheck`
- [x] `pnpm lint` (0 errors; 2 pre-existing warnings)
- [x] `pnpm check:changesets --strict`
- [x] `pnpm check:fnxc-future-dates`
- [x] `pnpm test` (changed-test gate; static checks and 688 tests
passed)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Improved reliability of post-merge pushes by retrying temporary Git
transport failures.
* Added bounded backoff between retries to prevent excessive repeated
attempts.
* Push retries now stop promptly when an operation is canceled.
* Configuration, authentication, and ref-rejection errors continue to
fail immediately.
* Successful retries and canceled operations now report accurate
outcomes.
* **Tests**
* Added coverage for successful retries, cancellation, and non-retryable
failures.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: flexi767 <flexi767@users.noreply.github.com>
Co-authored-by: v <v@m5.speedport.ip>
Slow-test trim, DELETE + dashboard OPTIMIZE tiers. Deletions follow the
docs/testing.md taxonomy exactly — 40 tests removed across
TaskDetailModal.rendering (134→129), QuickEntryBox (300→280),
TerminalModal (247→240), ModelOnboardingModal (205→197), and
AgentDetailView.settings (27→25): byte-identical field-presence twins,
CSS-class permutation residue, and strict-subset interaction copies,
with every FN-tagged describe and user-noticeable behavior kept and
condensed matrices converted to it.each with all cases retained. The
real payoff is the FN-2707 latency pass: SettingsModal's real 500ms
auto-save debounce (and AgentDetailView's 700ms config debounce) now
flush under fake timers via a shared flushSettingsAutoSave harness
export — SettingsModal splits 102.3s→63.6s. Also fixed (not appeased) a
pre-existing race the speedup exposed: nav-button lookups now
findByRole after the settings promise resolves, file-wide. Verified:
1235 tests green across the touched set, tsconfig.app.json typecheck
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Cursor provider card's Enable action only flips useCursorCli in
settings and never registers fusion-plugin-cursor-runtime, so
getRuntimeById("cursor") missed and every cursor-cli selection hit the
runtime-routed fail-fast error even with an authenticated cursor-agent.
Mirror the FN-7761 Grok eager bootstrap in serve, dashboard, and daemon,
guarded by the same source-scan regression test.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Slow-test trim, OPTIMIZE tier (harness owner). Profiling showed 163ms
of the shared harness's 169ms per-test beforeEach was one TRUNCATE over
~110 tables (per-table catalog/fsync constants, row-independent). The
reset is now a single DO block: transaction-local
session_replication_role=replica, DELETE only from non-empty tables,
and sequence resets reproducing RESTART IDENTITY exactly — 169ms → 10ms
per test across every harness file, with the legacy TRUNCATE retained
as an automatic error fallback for non-superuser roles. sqlite-migrator
(the live SQLite→PG upgrade path; all 43 tests kept incl.
VAL-MIGRATE-001..006) now provisions targets from a golden-template
clone via new createBaselinedPgTestDatabase/createEmptyPgTestDatabase
exports — the dry-run test keeps a pristine empty DB so its
no-schemas-left-behind assertions stay meaningful — 148.8s → ~26s.
taskstore-lifecycle, data-layer, and taskstore-remaining migrate onto
the shared harness with byte-identical test bodies (187s → ~11s
combined); data-layer's close() test closes a private layer so the
shared pool survives. Seeding hoists were measured and deliberately
skipped: the harness fix already exceeded their projected savings and a
beforeAll seed would weaken the clean-DB-per-test contract.
Verified: 287 tests green across the touched files plus untouched
harness consumers (schema-applier 80/80, mission-store 62/62).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Slow-test trim, OPTIMIZE tier (mechanical sweep). These files each
created and baselined a fresh database per test (~3-4.6s/test); they now
use the shared golden-template harness (TRUNCATE per test) or, where a
private database is genuinely required (postgres-health drift/VACUUM
subjects, allocator-cross-project RLS connection sets), a per-test
template clone instead of raw DDL. Zero assertion lines changed
(verified), identical test counts, 110/110 green. Deliberately left
alone: startup-factory-integration and embedded-lifecycle (own-database
boot IS the subject) and schema-applier (real DDL is the subject).
Measured: ~195s serial → ~30s for these files.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Slow-test trim, CONDENSE tier. The three satellite-store files each
hand-rolled CREATE DATABASE + schema baseline per test (38 database
creations, ~75s serial); their describes now live verbatim — every it
name, assertion, and FNXC provenance comment preserved, 38/38 tests —
inside one shared-harness file with a single boot. The 3-test
custom-v1-workflow-dispatch characterization (13s alone) moved into the
adjacent no-selection-default-workflow-move harness file. Measured:
~92s serial → ~8s for the same coverage.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
## Summary
- replaces newly added future-dated FNXC metadata with the actual UTC
change time
- tightens the FNXC future-date baseline to zero known exceptions
## Test plan
- `pnpm check:fnxc-future-dates`
- `pnpm check:lifecycle-columns`
- `pnpm check:changesets`
- `pnpm --filter @runfusion/fusion exec vitest run
src/__tests__/staged-plugin-core-imports.test.ts --silent=passed-only
--reporter=dot`
- `pnpm --filter @fusion/engine typecheck`
- `pnpm lint`
Confirm the W33 spike as cold-start variance and preserve a repeatable diagnosis protocol.
- Record five sequential phase-timed samples and the 20.5-second median threshold.
- Link the controlled remeasurement protocol from the testing guide.
- Explain the timing snapshot handoff and why incomplete CI artifacts were not published.
- Preserve the no-appeasement requirement beside boot-smoke phase timing.
Files changed:
.../boot-smoke-w33-walltime-anomaly.md | 100 +++++++++++++++++++++
docs/testing.md | 2 +-
scripts/boot-smoke.mjs | 7 ++
3 files changed, 108 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-9105
Fusion-Task-Lineage: 6956b037-db0f-4560-bb04-136080a975b5
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Full-suite repair, engine harness cluster (~85 red suites). Two causes:
(1) graph dispatch now fails closed without options.agentStore
(FN-8764/FN-8821, intended) — the shared executor test harness now
provisions the workflow-routing agent-store fixture for bare
TaskExecutor constructions, with explicit opt-out for the two tests
asserting the fail-closed park; (2) the wave-18 'pure peel' (#3317)
rebuilt executor.ts from a stale base and silently deleted shipped
behaviors, restored here: FN-8864 agent-activity writers (task
started/handed-off, workflow gate pass/fail, gate principal attribution
via executor/workflow-gate-activity.ts), FN-8768 Plan Review group
recognition, convergence primer, and modified-file review scoping,
FN-6782's fire-time guard on transient resume-after-restart retries,
FN-8868 session usage telemetry boundaries, recommendation-route
withheld-tool guidance, and the per-instance worktree retry cap.
Stale expectations updated for intended changes (FN-8823 shared-member
hold, FN-9060 zero-acquire fail-closed, heartbeat tool inventory,
peeled-module seams, PG harness provisioning). Verified: 23 files /
711 tests green, engine typecheck clean, merge gate green.
Known follow-ups (not addressed here): step-session error routing may
still bypass FN-5866 non-continuable classification
(post-done-continuation-no-wedge red), scheduler mission-loop trigger
gap (mission-validation-trigger-gap red), and executeWorkflowStep lost
routed workflow-principal session identity threading (untested drop
from the same peel).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Full-suite repair, engine source-scan cluster. The package code
organization waves moved ~30 engine modules into subdirectories
(plugins/, execution/, scheduling/, healing/, worktree/, executor/
peels); the log-severity manifest, prompt carve-out, emit-surface,
failure-lane, and worktree-invariant scanners now read the moved
locations, verified per file via git log --follow. Two scans caught
real drift rather than moves: the lifecycle census had 12 unexamined
column guards (resolved with DELIBERATE-LITERAL markers for the mailbox
archived tab, the FN-9059 lease-owner terminality check, and the FN-9056
legacy done fallback — baseline re-recorded with zero absorbed debt),
and planning-claim gained a genuine second writer in self-healing's
FN-8998 transport-failure recovery, admitted to the allowlist with its
CAS-guarded justification. 9 files / 119 tests green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Full-suite repair, core cluster. Two real product regressions caught by
the red tests: (1) the wave-18 refactor dropped the task.title argument
FN-8840 added to isDuplicateRedirectOnlyPrompt inside
isTaskAwaitingPlanning, so a title-only DUPLICATE:<ID> redirect read as
executable Ready instead of awaiting planning; (2) FN-8998 scoped every
workflow-definition read by layer.projectId but left the INSERT on the
session-GUC default, so a JS-bound layer created workflow rows it could
never read back — the insert now stamps the bound projectId like the
FN-8997 workflowSteps insert. Test-lag fixes: analytics renamed-lanes
harnesses now bind projectId 'p1' to match FN-8957/FN-8998 scoping,
project-ownership-runtime-scope replaces a hardcoded expired approval
date, schema-applier learns migrations 0059/0060 (115 project tables,
baseline 0060), and builtin-workflow-settings-triage learns FN-8932's
memoryConsolidationEnabled with position-independent lookups.
Verified: 10 files / 128 tests green, core typecheck clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Full-suite repair, desktop cluster: FN-8926's Node-only
@fusion/core/mcp-builtin-servers subpath needed the same vitest alias
desktop's engine/dashboard siblings already carry — without it the broad
prefix alias mangled the path and the runtime-resolution test failed with
ENOTDIR. The release-workflow test pinned actions/setup-java to an exact
patch that Dependabot rotates (broke twice, #3303 then #3367); it now
asserts the v5 major while keeping the JDK-21 and Android wiring
assertions intact. Desktop suite: 29 files / 265 tests green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
