mirror of
https://github.com/nexu-io/open-design.git
synced 2026-10-27 09:51:12 +01:00
fix(web): stop replaying OD Next turns parked on the user (#8250)
* fix: preserve ChatPanel history and clarification state * test(daemon): align artifact replay assertions with durable history * fix(web): stop replaying OD Next turns parked on the user A clarification turn ends with its Run succeeded while the strategy task stays open (`clarification_required`, `plan_ready`). The recovery probe in `attachRecoverableRuns` only bailed out for terminal tasks, so it cleared the finished row and replayed the Run from event 0. A conversation refresh landing mid-replay adopted the persisted transcript, the rest of the replay was appended behind it, and the replay's terminal PUT stored the duplicate: the question form card rendered with raw form JSON under it. A succeeded Run whose task holds no active Run has nothing left to follow; the daemon claims any automatic successor in the same transaction as the verdict, which a probe would see as a different `activeRunId`. Seal such a Run without clearing or replaying it. Rows that still need transcript recovery keep replaying. Refs OPEND-3230
This commit is contained in:
lefarcen
@@ -79,6 +79,7 @@ import { requestAmrArtifactUpgrade } from '../runtime/amr-artifact-upgrade';
|
||||
import {
|
||||
resolveQuestionFormStrategyTaskExecutionId,
|
||||
strategySettledMessageFields,
|
||||
strategyTaskParkedOnSucceededRun,
|
||||
} from '../runtime/strategy-question-continuation';
|
||||
import {
|
||||
isTodoWriteToolName,
|
||||
@@ -6586,8 +6587,9 @@ export function ProjectView({
|
||||
recoverableGenericDisconnectFailed;
|
||||
// A predecessor can be persisted as physically succeeded immediately
|
||||
// before the logical task advances. Probe daemon task truth even when
|
||||
// this row otherwise looks terminal; completed task rows bail out
|
||||
// below without replaying their final Run again.
|
||||
// this row otherwise looks terminal; completed task rows, and rows whose
|
||||
// task is parked on the user, bail out below without replaying their
|
||||
// final Run again.
|
||||
const needsTaskProjectionProbe = Boolean(
|
||||
message.strategyTaskExecutionId
|
||||
&& message.runId
|
||||
@@ -6761,7 +6763,11 @@ export function ProjectView({
|
||||
needsTaskProjectionProbe
|
||||
&& !needsReplayForMessage
|
||||
&& !taskRunAdvanced
|
||||
&& (!status.strategyTask || status.strategyTask.terminal)
|
||||
&& (
|
||||
!status.strategyTask
|
||||
|| status.strategyTask.terminal
|
||||
|| strategyTaskParkedOnSucceededRun(status, runId)
|
||||
)
|
||||
) {
|
||||
const strategyTask = status.strategyTask;
|
||||
if (
|
||||
|
||||
@@ -79,3 +79,29 @@ export function strategySettledMessageFields(
|
||||
}
|
||||
return null;
|
||||
}
|
||||
|
||||
/**
|
||||
* True when a Run status probe proves the Run left its logical task parked on
|
||||
* the user: the Run itself succeeded, and the task is neither terminal nor
|
||||
* running on this Run (`clarification_required`, `plan_ready`).
|
||||
*
|
||||
* Such a Run has nothing left for a reattach to follow. The daemon task store
|
||||
* keeps an active Run only while the outcome is `running`, and it claims any
|
||||
* automatic successor in the same transaction as the Run's verdict — a probe
|
||||
* would then project that successor as a different `activeRunId`. The wire
|
||||
* projection names this same Run only because it falls back to the latest Run
|
||||
* when none is active. A row that already holds this Run's transcript gains
|
||||
* nothing from a replay from event 0; the replay only re-streams text that a
|
||||
* concurrent conversation refresh can double up (OPEND-3230).
|
||||
*/
|
||||
export function strategyTaskParkedOnSucceededRun(
|
||||
status: Pick<ChatRunStatusResponse, 'status' | 'strategyTask'>,
|
||||
runId: string,
|
||||
): boolean {
|
||||
const task = status.strategyTask;
|
||||
return status.status === 'succeeded'
|
||||
&& task !== undefined
|
||||
&& !task.terminal
|
||||
&& task.outcome !== 'running'
|
||||
&& task.activeRunId === runId;
|
||||
}
|
||||
|
||||
@@ -0,0 +1,620 @@
|
||||
// @vitest-environment jsdom
|
||||
//
|
||||
// OPEND-3230: an OD Next clarification turn sometimes rendered its question
|
||||
// form card followed by raw form JSON — the transcript T with a copy of its own
|
||||
// tail T[cut:] appended, cut somewhere inside the form.
|
||||
//
|
||||
// Mechanism: a clarification turn ends with the Run physically `succeeded`
|
||||
// while its logical task stays open (`clarification_required`, not terminal).
|
||||
// `attachRecoverableRuns` probed that task, saw "not terminal", and cleared the
|
||||
// finished row to replay the whole Run from event 0. A conversation refresh
|
||||
// landing mid-replay adopted the server's full transcript, the rest of the
|
||||
// replay was appended behind it, and the replay's terminal PUT persisted the
|
||||
// duplicated text.
|
||||
//
|
||||
// The invariant under test: a Run whose logical task holds no active Run (the
|
||||
// task is parked on the user) has nothing left to follow, so its row is never
|
||||
// cleared and replayed. Rows that still need recovery keep replaying.
|
||||
|
||||
import { act, cleanup, render, waitFor } from '@testing-library/react';
|
||||
import { afterEach, describe, expect, it, vi } from 'vitest';
|
||||
import { StrategyTaskProjectionV2Schema } from '@open-design/contracts';
|
||||
import type { ChatRunStatusResponse, StrategyTaskProjectionV2 } from '@open-design/contracts';
|
||||
import { ProjectView } from '../../src/components/ProjectView';
|
||||
import type { AgentEvent, ChatMessage } from '../../src/types';
|
||||
import clarificationFixture from '../fixtures/chat/odnext-clarification-form.json';
|
||||
|
||||
const listConversations = vi.fn();
|
||||
const listMessages = vi.fn();
|
||||
const fetchPreviewComments = vi.fn();
|
||||
const loadTabs = vi.fn();
|
||||
const fetchProjectFiles = vi.fn();
|
||||
const fetchProjectDesignSystemPackageAudit = vi.fn();
|
||||
const fetchLiveArtifacts = vi.fn();
|
||||
const fetchSkill = vi.fn();
|
||||
const fetchDesignSystem = vi.fn();
|
||||
const getTemplate = vi.fn();
|
||||
const fetchChatRunStatus = vi.fn();
|
||||
const listActiveChatRuns = vi.fn();
|
||||
const listProjectRuns = vi.fn();
|
||||
const reattachDaemonRun = vi.fn();
|
||||
const publishDaemonRunFinishedEvent = vi.fn();
|
||||
const streamViaDaemon = vi.fn();
|
||||
const saveMessage = vi.fn();
|
||||
const createConversation = vi.fn();
|
||||
const patchConversation = vi.fn();
|
||||
const patchProject = vi.fn();
|
||||
const saveTabs = vi.fn();
|
||||
|
||||
const chatPaneHarness = vi.hoisted(() => ({
|
||||
onSend: null as null | ((
|
||||
prompt: string,
|
||||
attachments: unknown[],
|
||||
commentAttachments?: unknown[],
|
||||
meta?: unknown,
|
||||
) => unknown),
|
||||
messages: [] as ChatMessage[],
|
||||
}));
|
||||
|
||||
vi.mock('../../src/i18n', () => ({
|
||||
useI18n: () => ({
|
||||
locale: 'en',
|
||||
setLocale: () => undefined,
|
||||
t: (value: string) => value,
|
||||
}),
|
||||
useT: () => ((value: string) => value),
|
||||
}));
|
||||
|
||||
vi.mock('../../src/providers/anthropic', () => ({
|
||||
streamMessage: vi.fn(),
|
||||
}));
|
||||
|
||||
vi.mock('../../src/providers/daemon', () => ({
|
||||
GENERIC_DAEMON_DISCONNECT_CODE: 'GENERIC_DAEMON_DISCONNECT',
|
||||
GENERIC_DAEMON_DISCONNECT_MESSAGE: 'daemon stream disconnected before run completed',
|
||||
fetchChatRunStatus: (...args: unknown[]) => fetchChatRunStatus(...args),
|
||||
fetchAmrWalletSnapshot: vi.fn().mockResolvedValue(null),
|
||||
listActiveChatRuns: (...args: unknown[]) => listActiveChatRuns(...args),
|
||||
listProjectRuns: (...args: unknown[]) => listProjectRuns(...args),
|
||||
publishDaemonRunFinishedEvent: (...args: unknown[]) => publishDaemonRunFinishedEvent(...args),
|
||||
reattachDaemonRun: (...args: unknown[]) => reattachDaemonRun(...args),
|
||||
streamViaDaemon: (...args: unknown[]) => streamViaDaemon(...args),
|
||||
}));
|
||||
|
||||
vi.mock('../../src/providers/registry', () => ({
|
||||
deletePreviewComment: vi.fn(),
|
||||
fetchPreviewComments: (...args: unknown[]) => fetchPreviewComments(...args),
|
||||
fetchDesignSystem: (...args: unknown[]) => fetchDesignSystem(...args),
|
||||
fetchProjectDesignSystemPackageAudit: (...args: unknown[]) =>
|
||||
fetchProjectDesignSystemPackageAudit(...args),
|
||||
fetchLiveArtifacts: (...args: unknown[]) => fetchLiveArtifacts(...args),
|
||||
fetchProjectFiles: (...args: unknown[]) => fetchProjectFiles(...args),
|
||||
fetchSkill: (...args: unknown[]) => fetchSkill(...args),
|
||||
patchPreviewCommentStatus: vi.fn(),
|
||||
upsertPreviewComment: vi.fn(),
|
||||
writeProjectTextFile: vi.fn(),
|
||||
}));
|
||||
|
||||
vi.mock('../../src/providers/project-events', () => ({
|
||||
useProjectFileEvents: vi.fn(),
|
||||
}));
|
||||
|
||||
vi.mock('../../src/router', () => ({
|
||||
navigate: vi.fn(),
|
||||
}));
|
||||
|
||||
vi.mock('../../src/state/projects', () => ({
|
||||
cacheTabsLocally: vi.fn((projectId: string, tabs: unknown) => ({ projectId, tabs })),
|
||||
createConversation: (...args: unknown[]) => createConversation(...args),
|
||||
deleteConversation: vi.fn(),
|
||||
getTemplate: (...args: unknown[]) => getTemplate(...args),
|
||||
listConversations: (...args: unknown[]) => listConversations(...args),
|
||||
listMessages: (...args: unknown[]) => listMessages(...args),
|
||||
loadTabs: (...args: unknown[]) => loadTabs(...args),
|
||||
patchConversation: (...args: unknown[]) => patchConversation(...args),
|
||||
patchProject: (...args: unknown[]) => patchProject(...args),
|
||||
persistTabsToDaemonNow: vi.fn(),
|
||||
saveMessage: (...args: unknown[]) => saveMessage(...args),
|
||||
saveTabs: (...args: unknown[]) => saveTabs(...args),
|
||||
}));
|
||||
|
||||
vi.mock('../../src/components/AppChromeHeader', () => ({
|
||||
AppChromeHeader: () => null,
|
||||
}));
|
||||
|
||||
vi.mock('../../src/components/AvatarMenu', () => ({
|
||||
AvatarMenu: () => null,
|
||||
}));
|
||||
|
||||
vi.mock('../../src/components/ChatPane', () => ({
|
||||
ChatPane: ({
|
||||
messages,
|
||||
onSend,
|
||||
}: {
|
||||
messages: ChatMessage[];
|
||||
onSend: typeof chatPaneHarness.onSend;
|
||||
}) => {
|
||||
chatPaneHarness.messages = messages;
|
||||
chatPaneHarness.onSend = onSend;
|
||||
return null;
|
||||
},
|
||||
}));
|
||||
|
||||
vi.mock('../../src/components/FileWorkspace', () => ({
|
||||
DESIGN_SYSTEM_TAB: '__design_system__',
|
||||
FileWorkspace: () => null,
|
||||
}));
|
||||
|
||||
vi.mock('../../src/components/Loading', () => ({
|
||||
CenteredLoader: () => null,
|
||||
}));
|
||||
|
||||
function renderProjectView() {
|
||||
const project = {
|
||||
id: 'project-1',
|
||||
name: 'Project',
|
||||
skillId: null,
|
||||
designSystemId: null,
|
||||
} as never;
|
||||
return render(
|
||||
<ProjectView
|
||||
project={project}
|
||||
initialProjectDetail={{ project, resolvedDir: null }}
|
||||
routeConversationId={null}
|
||||
routeFileName={null}
|
||||
config={
|
||||
{
|
||||
mode: 'daemon',
|
||||
agentId: 'agent-1',
|
||||
notifications: undefined,
|
||||
agentModels: {},
|
||||
} as never
|
||||
}
|
||||
agents={[{ id: 'agent-1', name: 'OpenCode', models: [] } as never]}
|
||||
skills={[]}
|
||||
designTemplates={[]}
|
||||
designSystems={[]}
|
||||
daemonLive
|
||||
onModeChange={() => {}}
|
||||
onAgentChange={() => {}}
|
||||
onAgentModelChange={() => {}}
|
||||
onRefreshAgents={() => {}}
|
||||
onOpenSettings={() => {}}
|
||||
onBack={() => {}}
|
||||
onClearPendingPrompt={() => {}}
|
||||
onTouchProject={() => {}}
|
||||
onProjectChange={() => {}}
|
||||
onProjectsRefresh={() => {}}
|
||||
/>,
|
||||
);
|
||||
}
|
||||
|
||||
function deferred<T>() {
|
||||
let resolve!: (value: T) => void;
|
||||
const promise = new Promise<T>((r) => { resolve = r; });
|
||||
return { promise, resolve };
|
||||
}
|
||||
|
||||
const sleep = (ms: number) => new Promise((resolve) => setTimeout(resolve, ms));
|
||||
|
||||
/** The synthetic clarification turn: narration followed by one question form. */
|
||||
const FORM_TURN_TEXT = clarificationFixture.events
|
||||
.filter((event) => event.kind === 'text')
|
||||
.map((event) => ('text' in event ? event.text : ''))
|
||||
.join('');
|
||||
|
||||
const RUN_ID = 'run-clarification';
|
||||
const TASK_EXECUTION_ID = 'task-clarification';
|
||||
const STRATEGY_IDENTITY = {
|
||||
id: 'od-next-strategy',
|
||||
version: '2.0.0',
|
||||
packageHash: 'e'.repeat(64),
|
||||
snapshotId: 'snapshot-clarification',
|
||||
} as const;
|
||||
|
||||
/** What `POST /api/runs` projects while the initial request is still running. */
|
||||
const RUNNING_REQUEST_PROJECTION: StrategyTaskProjectionV2 = {
|
||||
taskExecutionId: TASK_EXECUTION_ID,
|
||||
strategy: STRATEGY_IDENTITY,
|
||||
inputStage: 'request',
|
||||
outcome: 'running',
|
||||
route: null,
|
||||
executionMode: null,
|
||||
activeRunId: RUN_ID,
|
||||
terminal: false,
|
||||
};
|
||||
|
||||
/**
|
||||
* What the daemon projects once the Run has settled a task that now waits on
|
||||
* the user. Mirrors `projectStrategyTask`
|
||||
* (apps/daemon/src/strategies/od-next/automatic-simple-production.ts): the task
|
||||
* store keeps no active Run for a non-`running` outcome, and the wire
|
||||
* projection then falls back to the latest Run — this same Run.
|
||||
*/
|
||||
const PARKED_PROJECTIONS: Record<'clarification_required' | 'plan_ready', StrategyTaskProjectionV2> = {
|
||||
clarification_required: {
|
||||
taskExecutionId: TASK_EXECUTION_ID,
|
||||
strategy: STRATEGY_IDENTITY,
|
||||
inputStage: 'request',
|
||||
outcome: 'clarification_required',
|
||||
route: 'full_plan',
|
||||
executionMode: null,
|
||||
activeRunId: RUN_ID,
|
||||
terminal: false,
|
||||
},
|
||||
plan_ready: {
|
||||
taskExecutionId: TASK_EXECUTION_ID,
|
||||
strategy: STRATEGY_IDENTITY,
|
||||
inputStage: 'request',
|
||||
outcome: 'plan_ready',
|
||||
route: 'full_plan',
|
||||
executionMode: 'simple',
|
||||
activeRunId: RUN_ID,
|
||||
terminal: false,
|
||||
},
|
||||
};
|
||||
|
||||
function succeededStatus(strategyTask: StrategyTaskProjectionV2): ChatRunStatusResponse {
|
||||
const now = Date.now();
|
||||
return {
|
||||
id: RUN_ID,
|
||||
status: 'succeeded',
|
||||
createdAt: now - 30_000,
|
||||
updatedAt: now,
|
||||
exitCode: 0,
|
||||
signal: null,
|
||||
strategyTask,
|
||||
} as ChatRunStatusResponse;
|
||||
}
|
||||
|
||||
const textOf = (message: ChatMessage | undefined) =>
|
||||
(message?.events ?? [])
|
||||
.filter((event): event is Extract<AgentEvent, { kind: 'text' }> => event.kind === 'text')
|
||||
.map((event) => event.text)
|
||||
.join('');
|
||||
|
||||
const assistantOf = () => chatPaneHarness.messages.find((message) => message.role === 'assistant');
|
||||
|
||||
/** Classifies a transcript against the one clean copy of the turn. */
|
||||
function transcriptShape(text: string): string {
|
||||
if (text === FORM_TURN_TEXT) return 'clean';
|
||||
if (text === '') return 'empty';
|
||||
if (text.length > FORM_TURN_TEXT.length && text.startsWith(FORM_TURN_TEXT)) {
|
||||
return FORM_TURN_TEXT.endsWith(text.slice(FORM_TURN_TEXT.length))
|
||||
? `duplicated-tail(${text.length - FORM_TURN_TEXT.length})`
|
||||
: `overlong(${text.length})`;
|
||||
}
|
||||
return FORM_TURN_TEXT.startsWith(text) ? `prefix(${text.length})` : `other(${text.length})`;
|
||||
}
|
||||
|
||||
function mockProjectShell() {
|
||||
listConversations.mockResolvedValue([{ id: 'conv-1', title: 'Conversation' }]);
|
||||
fetchPreviewComments.mockResolvedValue([]);
|
||||
loadTabs.mockResolvedValue({ tabs: [], activeTabId: null });
|
||||
fetchProjectFiles.mockResolvedValue([]);
|
||||
fetchLiveArtifacts.mockResolvedValue([]);
|
||||
fetchSkill.mockResolvedValue(null);
|
||||
fetchDesignSystem.mockResolvedValue(null);
|
||||
getTemplate.mockResolvedValue(null);
|
||||
listActiveChatRuns.mockResolvedValue([]);
|
||||
listProjectRuns.mockResolvedValue([]);
|
||||
}
|
||||
|
||||
/** Settles every status probe issued so far, then lets React commit what follows. */
|
||||
async function settleStatusProbes() {
|
||||
await act(async () => {
|
||||
await Promise.allSettled(fetchChatRunStatus.mock.results.map((result) => result.value));
|
||||
});
|
||||
await act(async () => {});
|
||||
}
|
||||
|
||||
const probedRunIds = () => fetchChatRunStatus.mock.calls.map((call) => call[0]);
|
||||
|
||||
/**
|
||||
* A reattach stream that stays open until ProjectView aborts it (unmount).
|
||||
* Terminal replays share a module-level concurrency gate, so a stream that
|
||||
* never settles would starve the next case's replay.
|
||||
*/
|
||||
function streamUntilAborted(options: { signal: AbortSignal }): Promise<void> {
|
||||
return new Promise<void>((resolve) => {
|
||||
if (options.signal.aborted) resolve();
|
||||
else options.signal.addEventListener('abort', () => resolve(), { once: true });
|
||||
});
|
||||
}
|
||||
|
||||
describe('OPEND-3230 a finished OD Next turn whose task waits on the user', () => {
|
||||
afterEach(() => {
|
||||
cleanup();
|
||||
vi.useRealTimers();
|
||||
vi.restoreAllMocks();
|
||||
vi.clearAllMocks();
|
||||
chatPaneHarness.onSend = null;
|
||||
chatPaneHarness.messages = [];
|
||||
window.sessionStorage.clear();
|
||||
});
|
||||
|
||||
it('keeps the fixture projections in the shape the daemon contract accepts', () => {
|
||||
expect(() => StrategyTaskProjectionV2Schema.parse(RUNNING_REQUEST_PROJECTION)).not.toThrow();
|
||||
for (const projection of Object.values(PARKED_PROJECTIONS)) {
|
||||
expect(() => StrategyTaskProjectionV2Schema.parse(projection)).not.toThrow();
|
||||
}
|
||||
});
|
||||
|
||||
it('does not clear and replay the clarification turn after its own live stream ends', async () => {
|
||||
mockProjectShell();
|
||||
saveMessage.mockResolvedValue(undefined);
|
||||
listMessages.mockResolvedValue([]);
|
||||
let sendOptions: any = null;
|
||||
const liveStream = deferred<void>();
|
||||
streamViaDaemon.mockImplementation(async (options: any) => {
|
||||
sendOptions = options;
|
||||
options.onRunCreated(RUN_ID, RUNNING_REQUEST_PROJECTION);
|
||||
options.onRunStatus?.('running');
|
||||
return liveStream.promise;
|
||||
});
|
||||
fetchChatRunStatus.mockResolvedValue(succeededStatus(PARKED_PROJECTIONS.clarification_required));
|
||||
reattachDaemonRun.mockImplementation(streamUntilAborted);
|
||||
|
||||
renderProjectView();
|
||||
await waitFor(() => expect(chatPaneHarness.onSend).toBeTruthy());
|
||||
void chatPaneHarness.onSend!('prompt', [], []);
|
||||
await waitFor(() => expect(sendOptions).not.toBeNull());
|
||||
sendOptions.handlers.onDelta(FORM_TURN_TEXT);
|
||||
sendOptions.handlers.onAgentEvent({ kind: 'text', text: FORM_TURN_TEXT });
|
||||
await waitFor(() => expect(textOf(assistantOf())).toBe(FORM_TURN_TEXT));
|
||||
const assistantId = assistantOf()!.id;
|
||||
|
||||
await act(async () => {
|
||||
sendOptions.onRunStatus('succeeded');
|
||||
sendOptions.handlers.onDone(FORM_TURN_TEXT);
|
||||
liveStream.resolve();
|
||||
await liveStream.promise;
|
||||
});
|
||||
// Recovery re-reads daemon task truth once the local finalizer releases
|
||||
// the Run; that probe is the decision point this case is about.
|
||||
await waitFor(() => expect(probedRunIds()).toContain(RUN_ID));
|
||||
await settleStatusProbes();
|
||||
|
||||
expect(reattachDaemonRun).not.toHaveBeenCalled();
|
||||
expect(textOf(assistantOf())).toBe(FORM_TURN_TEXT);
|
||||
expect(assistantOf()?.content).toBe(FORM_TURN_TEXT);
|
||||
const clearedWrites = saveMessage.mock.calls
|
||||
.map((call) => call[2] as ChatMessage)
|
||||
.filter((message) => message?.id === assistantId && message.content !== FORM_TURN_TEXT && message.runStatus === 'succeeded');
|
||||
expect(clearedWrites).toEqual([]);
|
||||
});
|
||||
|
||||
it.each(['clarification_required', 'plan_ready'] as const)(
|
||||
'does not clear and replay a hydrated succeeded row whose task is parked on the user (%s)',
|
||||
async (outcome) => {
|
||||
mockProjectShell();
|
||||
saveMessage.mockResolvedValue(undefined);
|
||||
const startedAt = Date.now() - 30_000;
|
||||
listMessages.mockResolvedValue([
|
||||
{ id: 'msg-user', role: 'user', content: 'prompt', createdAt: startedAt },
|
||||
{
|
||||
id: 'msg-clarification',
|
||||
role: 'assistant',
|
||||
agentId: 'agent-1',
|
||||
content: FORM_TURN_TEXT,
|
||||
events: clarificationFixture.events as AgentEvent[],
|
||||
createdAt: startedAt,
|
||||
startedAt,
|
||||
endedAt: startedAt + 20_000,
|
||||
runId: RUN_ID,
|
||||
runStatus: 'succeeded',
|
||||
lastRunEventId: '42',
|
||||
strategyTaskExecutionId: TASK_EXECUTION_ID,
|
||||
},
|
||||
] satisfies ChatMessage[]);
|
||||
fetchChatRunStatus.mockResolvedValue(succeededStatus(PARKED_PROJECTIONS[outcome]));
|
||||
reattachDaemonRun.mockImplementation(streamUntilAborted);
|
||||
|
||||
renderProjectView();
|
||||
await waitFor(() => expect(probedRunIds()).toContain(RUN_ID));
|
||||
await settleStatusProbes();
|
||||
|
||||
expect(reattachDaemonRun).not.toHaveBeenCalled();
|
||||
expect(assistantOf()?.content).toBe(FORM_TURN_TEXT);
|
||||
expect(textOf(assistantOf())).toBe(FORM_TURN_TEXT);
|
||||
expect(
|
||||
saveMessage.mock.calls
|
||||
.map((call) => call[2] as ChatMessage)
|
||||
.filter((message) => message?.id === 'msg-clarification'),
|
||||
).toEqual([]);
|
||||
},
|
||||
);
|
||||
|
||||
it('still replays a parked-task row whose persisted transcript never landed', async () => {
|
||||
mockProjectShell();
|
||||
saveMessage.mockResolvedValue(undefined);
|
||||
const startedAt = Date.now() - 30_000;
|
||||
listMessages.mockResolvedValue([
|
||||
{ id: 'msg-user', role: 'user', content: 'prompt', createdAt: startedAt },
|
||||
{
|
||||
id: 'msg-clarification',
|
||||
role: 'assistant',
|
||||
agentId: 'agent-1',
|
||||
content: '',
|
||||
events: [],
|
||||
createdAt: startedAt,
|
||||
startedAt,
|
||||
runId: RUN_ID,
|
||||
runStatus: 'succeeded',
|
||||
strategyTaskExecutionId: TASK_EXECUTION_ID,
|
||||
},
|
||||
] satisfies ChatMessage[]);
|
||||
fetchChatRunStatus.mockResolvedValue(succeededStatus(PARKED_PROJECTIONS.clarification_required));
|
||||
reattachDaemonRun.mockImplementation(streamUntilAborted);
|
||||
|
||||
renderProjectView();
|
||||
|
||||
await waitFor(() => expect(reattachDaemonRun).toHaveBeenCalledTimes(1));
|
||||
expect(reattachDaemonRun).toHaveBeenCalledWith(expect.objectContaining({
|
||||
runId: RUN_ID,
|
||||
initialLastEventId: null,
|
||||
}));
|
||||
});
|
||||
|
||||
// Real clock on purpose: this case is about how React commits interleave
|
||||
// with pending I/O (status probe, post-run refresh, message PUTs), which
|
||||
// fake timers would collapse into one act (docs/testing/test-efficiency.zh-CN.md).
|
||||
// The small sleeps inside the mocks stand in for network latency.
|
||||
it.each([3, 6, 12])(
|
||||
'keeps the rendered and persisted transcript clean when a post-run refresh overlaps recovery (replay pacing %sms)',
|
||||
async (replayPauseMs) => {
|
||||
mockProjectShell();
|
||||
const NON_TEXT = Array.from({ length: 36 }, (_, index) => ({ kind: 'status', label: `s${index}`, detail: 'x' }));
|
||||
const REPLAY_HEAD = Array.from({ length: 5 }, (_, index) => ({ kind: 'status', label: `r${index}`, detail: 'x' }));
|
||||
const TRAILING = [
|
||||
{ kind: 'usage', inputTokens: 1, outputTokens: 1 },
|
||||
{ kind: 'status', label: 'tail-a', detail: 'x' },
|
||||
{ kind: 'status', label: 'tail-b', detail: 'x' },
|
||||
];
|
||||
|
||||
// SIMPLIFIED DAEMON STORE (hand-written model, not the real route). It
|
||||
// keeps only the rules of PUT /api/projects/:id/conversations/:cid/messages/:mid
|
||||
// (apps/daemon/src/routes/project/conversations.ts) that decide this bug
|
||||
// for a terminal, daemon-known Run: a write with fewer events than stored
|
||||
// is refused (`shrinksEvents`); otherwise content is the strictly longer
|
||||
// of the two and events are the incoming ones. While the Run is live the
|
||||
// daemon owns the row, so client writes are ignored.
|
||||
let serverAssistant: ChatMessage | null = null;
|
||||
let serverUser: ChatMessage | null = null;
|
||||
let runTerminal = false;
|
||||
saveMessage.mockImplementation(async (_projectId: string, _conversationId: string, message: ChatMessage) => {
|
||||
await sleep(2);
|
||||
if (message.role === 'user') {
|
||||
serverUser = message;
|
||||
return;
|
||||
}
|
||||
if (!runTerminal || !serverAssistant) return;
|
||||
const incomingEvents = message.events ?? [];
|
||||
if (incomingEvents.length < (serverAssistant.events ?? []).length) return;
|
||||
const content = (message.content?.length ?? 0) > (serverAssistant.content?.length ?? 0)
|
||||
? message.content
|
||||
: serverAssistant.content;
|
||||
serverAssistant = { ...message, content, events: incomingEvents };
|
||||
});
|
||||
listMessages.mockImplementation(async () => {
|
||||
if (!runTerminal) return [];
|
||||
await sleep(4);
|
||||
return [serverUser, serverAssistant].filter(Boolean);
|
||||
});
|
||||
|
||||
let sendOptions: any = null;
|
||||
const liveStream = deferred<void>();
|
||||
streamViaDaemon.mockImplementation(async (options: any) => {
|
||||
sendOptions = options;
|
||||
options.onRunCreated(RUN_ID, RUNNING_REQUEST_PROJECTION);
|
||||
options.onRunStatus?.('running');
|
||||
return liveStream.promise;
|
||||
});
|
||||
fetchChatRunStatus.mockImplementation(async () => {
|
||||
await sleep(3);
|
||||
return succeededStatus(PARKED_PROJECTIONS.clarification_required);
|
||||
});
|
||||
// What a replay of this Run from event 0 would deliver, in paced chunks.
|
||||
reattachDaemonRun.mockImplementation(async (options: any) => {
|
||||
let eventId = 2450;
|
||||
options.onRunEventId?.(String(++eventId));
|
||||
for (const event of REPLAY_HEAD) {
|
||||
options.onRunEventId?.(String(++eventId));
|
||||
options.handlers.onAgentEvent(event);
|
||||
}
|
||||
let chunks = 0;
|
||||
for (let index = 0; index < FORM_TURN_TEXT.length; index += 3) {
|
||||
const chunk = FORM_TURN_TEXT.slice(index, index + 3);
|
||||
options.onRunEventId?.(String(++eventId));
|
||||
options.handlers.onDelta(chunk);
|
||||
options.handlers.onAgentEvent({ kind: 'text', text: chunk });
|
||||
if (++chunks % 8 === 0) await sleep(replayPauseMs);
|
||||
}
|
||||
for (const event of TRAILING) {
|
||||
options.onRunEventId?.(String(++eventId));
|
||||
options.handlers.onAgentEvent(event);
|
||||
}
|
||||
options.onRunEventId?.(String(++eventId));
|
||||
options.onRunStatus?.('succeeded');
|
||||
await options.handlers.onDone?.(FORM_TURN_TEXT);
|
||||
});
|
||||
|
||||
renderProjectView();
|
||||
await waitFor(() => expect(chatPaneHarness.onSend).toBeTruthy());
|
||||
void chatPaneHarness.onSend!('prompt', [], []);
|
||||
await waitFor(() => expect(sendOptions).not.toBeNull());
|
||||
for (const event of NON_TEXT) sendOptions.handlers.onAgentEvent(event);
|
||||
sendOptions.handlers.onDelta(FORM_TURN_TEXT);
|
||||
sendOptions.handlers.onAgentEvent({ kind: 'text', text: FORM_TURN_TEXT });
|
||||
for (const event of TRAILING) sendOptions.handlers.onAgentEvent(event);
|
||||
await waitFor(() => expect(textOf(assistantOf())).toBe(FORM_TURN_TEXT));
|
||||
const assistantId = assistantOf()!.id;
|
||||
// The daemon finalizes the row with the full transcript before the
|
||||
// client observes the terminal frame.
|
||||
serverAssistant = {
|
||||
...assistantOf()!,
|
||||
content: FORM_TURN_TEXT,
|
||||
events: [...NON_TEXT, { kind: 'text', text: FORM_TURN_TEXT }, ...TRAILING] as AgentEvent[],
|
||||
runStatus: 'succeeded',
|
||||
};
|
||||
runTerminal = true;
|
||||
const renderedShapes: string[] = [];
|
||||
const sampleRendered = setInterval(() => {
|
||||
const shape = transcriptShape(textOf(chatPaneHarness.messages.find((message) => message.id === assistantId)));
|
||||
if (renderedShapes.at(-1) !== shape) renderedShapes.push(shape);
|
||||
}, 1);
|
||||
const listCallsBeforeEnd = listMessages.mock.calls.length;
|
||||
|
||||
try {
|
||||
await act(async () => {
|
||||
sendOptions.onRunStatus('succeeded');
|
||||
sendOptions.handlers.onDone(FORM_TURN_TEXT);
|
||||
liveStream.resolve();
|
||||
await liveStream.promise;
|
||||
// STAGED TIMING: let the local finalizer release the Run, then land
|
||||
// one late frame cursor on the row. Any messages change in that
|
||||
// window re-runs recovery before the 150ms post-run refresh; in the
|
||||
// field that is whatever updates the row next.
|
||||
await sleep(30);
|
||||
sendOptions.onRunEventId?.('2451');
|
||||
});
|
||||
|
||||
await waitFor(() => expect(probedRunIds()).toContain(RUN_ID));
|
||||
await settleStatusProbes();
|
||||
// The post-run refresh scheduled by the live terminal status.
|
||||
await waitFor(() => expect(listMessages.mock.calls.length).toBeGreaterThan(listCallsBeforeEnd));
|
||||
await act(async () => {
|
||||
await Promise.allSettled(listMessages.mock.results.map((result) => result.value));
|
||||
});
|
||||
if (reattachDaemonRun.mock.calls.length > 0) {
|
||||
// Only reached when recovery replays the Run: wait for that replay,
|
||||
// its finalizing PUTs, and the refresh its terminal status schedules.
|
||||
await act(async () => {
|
||||
await Promise.allSettled(reattachDaemonRun.mock.results.map((result) => result.value));
|
||||
});
|
||||
const listCallsAfterReplay = listMessages.mock.calls.length;
|
||||
await waitFor(
|
||||
() => expect(listMessages.mock.calls.length).toBeGreaterThan(listCallsAfterReplay),
|
||||
{ timeout: 3_000 },
|
||||
);
|
||||
await act(async () => {
|
||||
await Promise.allSettled(listMessages.mock.results.map((result) => result.value));
|
||||
await Promise.allSettled(saveMessage.mock.results.map((result) => result.value));
|
||||
});
|
||||
}
|
||||
} finally {
|
||||
clearInterval(sampleRendered);
|
||||
}
|
||||
|
||||
const renderedFinal = transcriptShape(textOf(chatPaneHarness.messages.find((message) => message.id === assistantId)));
|
||||
const persistedFinal = transcriptShape(textOf(serverAssistant!));
|
||||
expect({ renderedFinal, persistedFinal, renderedShapes }).toEqual({
|
||||
renderedFinal: 'clean',
|
||||
persistedFinal: 'clean',
|
||||
renderedShapes: ['clean'],
|
||||
});
|
||||
expect(reattachDaemonRun).not.toHaveBeenCalled();
|
||||
},
|
||||
20_000,
|
||||
);
|
||||
});
|
||||
@@ -1008,6 +1008,99 @@ describe('a Home auto-send identifies its caller before the project scope resolv
|
||||
activeStream.resolve();
|
||||
});
|
||||
|
||||
// OPEND-3230 guard: the raw form tail observed in the field was cut INSIDE
|
||||
// the question form, after the client had already rendered the text before
|
||||
// the cut. The live-stream ownership on Home authority refresh must hold for
|
||||
// that cut too, not only for a cut in the prose before the form.
|
||||
it.each(['buffer fallback', 'same-batch terminal'] as const)(
|
||||
'does not append a flushed partial question form behind the authoritative transcript on Home authority refresh (%s)', async (settlement) => {
|
||||
const activeStream = deferred<void>();
|
||||
let runOptions: Parameters<typeof streamViaDaemon>[0] | undefined;
|
||||
mockedStreamViaDaemon.mockImplementation((options) => {
|
||||
runOptions = options;
|
||||
return activeStream.promise;
|
||||
});
|
||||
const transcriptRead = vi.spyOn(transcriptLoader, 'loadConversationTranscript');
|
||||
const view = renderProjectView();
|
||||
await waitFor(() => expect(runOptions).toBeDefined());
|
||||
await act(async () => {
|
||||
runOptions?.onRunCreated?.('run-clarification-fixture');
|
||||
runOptions?.onRunStatus?.('running');
|
||||
});
|
||||
const live = chatPaneSpy.mock.calls.at(-1)?.[0].messages as ChatMessage[];
|
||||
const assistant = live.find((message) => message.role === 'assistant')!;
|
||||
expect(assistant.runId).toBe('run-clarification-fixture');
|
||||
vi.useFakeTimers();
|
||||
const authorityReload = deferred<ChatMessage[]>();
|
||||
mockedListMessages.mockReturnValueOnce(authorityReload.promise);
|
||||
workspaceScopeMocks.projectScope = {
|
||||
loading: false,
|
||||
scope: {
|
||||
kind: 'team', projectId: PROJECT_ID, workspaceId: TEAM_WORKSPACE, visibility: 'team',
|
||||
context: {
|
||||
...CALLER_CONTEXT, role: 'admin',
|
||||
permissions: buildWorkspacePermissions({ role: 'admin', lifecycleState: 'active' }),
|
||||
} as WorkspaceCollabContext & { workspaceType: 'team' },
|
||||
},
|
||||
};
|
||||
await act(async () => {
|
||||
view.rerender(projectViewElement());
|
||||
await vi.advanceTimersByTimeAsync(0);
|
||||
});
|
||||
expect(mockedListMessages).toHaveBeenCalledTimes(2);
|
||||
const full = proseFixture.events.filter((event) => event.kind === 'text')
|
||||
.map((event) => 'text' in event ? event.text : '').join('');
|
||||
const prefix = full.slice(0, full.indexOf('"title":"Choose'));
|
||||
expect(prefix).toContain('<question-form>');
|
||||
await act(async () => {
|
||||
runOptions?.handlers.onDelta(prefix);
|
||||
runOptions?.handlers.onAgentEvent({ kind: 'text', text: prefix });
|
||||
});
|
||||
// Unlike the prose cut above, this prefix is already on screen before the
|
||||
// authoritative GET lands.
|
||||
await act(async () => { await vi.advanceTimersByTimeAsync(250); });
|
||||
expect((chatPaneSpy.mock.calls.at(-1)?.[0].messages as ChatMessage[])
|
||||
.find((message) => message.id === assistant.id)?.content).toBe(prefix);
|
||||
const pendingTranscript = transcriptRead.mock.results.at(-1)?.value as Promise<ChatMessage[]>;
|
||||
expect(pendingTranscript).toBeInstanceOf(Promise);
|
||||
try {
|
||||
await act(async () => {
|
||||
authorityReload.resolve(live.map((message) => message.id === assistant.id ? {
|
||||
...message, content: full, events: [{ kind: 'text', text: full }],
|
||||
} : message));
|
||||
await pendingTranscript;
|
||||
if (settlement === 'same-batch terminal') {
|
||||
const suffix = full.slice(prefix.length);
|
||||
runOptions?.handlers.onDelta(suffix);
|
||||
runOptions?.handlers.onAgentEvent({ kind: 'text', text: suffix });
|
||||
runOptions?.onRunStatus?.('succeeded');
|
||||
runOptions?.handlers.onDone(full);
|
||||
} else {
|
||||
await vi.advanceTimersByTimeAsync(0);
|
||||
}
|
||||
});
|
||||
if (settlement === 'buffer fallback') {
|
||||
await act(async () => { await vi.advanceTimersByTimeAsync(250); });
|
||||
await act(async () => {
|
||||
const suffix = full.slice(prefix.length);
|
||||
runOptions?.handlers.onDelta(suffix);
|
||||
runOptions?.handlers.onAgentEvent({ kind: 'text', text: suffix });
|
||||
await vi.advanceTimersByTimeAsync(250);
|
||||
});
|
||||
}
|
||||
const actual = (chatPaneSpy.mock.calls.at(-1)?.[0].messages as ChatMessage[])
|
||||
.find((message) => message.id === assistant.id)!;
|
||||
expect(actual.content).toBe(full);
|
||||
const text = actual.events?.filter((event) => event.kind === 'text')
|
||||
.map((event) => event.text).join('');
|
||||
expect(text).toBe(full);
|
||||
} finally {
|
||||
activeStream.resolve();
|
||||
await act(async () => { await activeStream.promise; });
|
||||
transcriptRead.mockRestore();
|
||||
}
|
||||
});
|
||||
|
||||
it('keeps a terminal cold Home run attached when the empty refresh outlives its controller', async () => {
|
||||
const activeStream = deferred<void>();
|
||||
let runOptions: Parameters<typeof streamViaDaemon>[0] | undefined;
|
||||
|
||||
@@ -0,0 +1,134 @@
|
||||
{
|
||||
"runId": "fixture-odnext-clarification-run",
|
||||
"formId": "b2b-sales-deck",
|
||||
"events": [
|
||||
{
|
||||
"kind": "done_key",
|
||||
"key": "a7f3c91ed2b40561"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "我先确认几个关键点,再动手写提案。\n\n<question-form id="
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "\"b2b-sales-deck\" title=\"销售提案简报 — 30 秒"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "\">\n{\n \"description\": \"我会据此锁定叙事、证据链和 "
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "ROI 口径,再输出逐页文案。不填的项我会按推荐默认值处理。\",\n \"q"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "uestions\": [\n {\n \"id\": \"audie"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "nce\",\n \"label\": \"谁是这份提案的经济型买家与决策"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "目标?\",\n \"type\": \"radio\",\n \"o"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "ptions\": [\n \"设计组织负责人(VP Design"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " / Head of Design)— 以设计效率与一致性为决策点\",\n "
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " \"产品与工程负责人(CPO / VP Engineering"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": ")— 以交付速度与工程集成为决策点\",\n \"采购与 IT(P"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "rocurement / IT)— 以合规、采购与总拥有成本为决策点\",\n"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " \"CEO / 业务负责人 — 以营收与组织杠杆为决策点\"\n"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " ],\n \"required\": true,\n "
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " \"allowCustom\": true\n },\n {\n "
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " \"id\": \"evidence\",\n \"label\": "
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "\"证据与数据口径用哪种?是否有必须保留的真实数据?\",\n \"ty"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "pe\": \"radio\",\n \"options\": [\n "
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " \"真实公开信息 + 明确标注的占位数据(推荐,不编造任何指标)\","
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "\n \"我有内部真实数据(留存、用量、ROI 基线),随后提供"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "\",\n \"只用定性价值论述,不放任何数字\"\n ],"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "\n \"required\": true,\n \"allow"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "Custom\": true\n },\n {\n \"id\""
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": ": \"scope\",\n \"label\": \"篇幅与交付形态?\","
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "\n \"type\": \"radio\",\n \"option"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "s\": [\n \"12–15 页完整提案:痛点 / 价值 / "
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "ROI / 落地路线 / 决策请求(推荐)\",\n \"8–10"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " 页精简版:聚焦价值主张与决策请求\",\n \"18–20 页评"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "审完整版:含反方观点、安全与合规、实施细节\"\n ],\n "
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": " \"required\": true,\n \"allowCustom"
|
||||
},
|
||||
{
|
||||
"kind": "text",
|
||||
"text": "\": true\n }\n ]\n}\n</question-form>"
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -4,6 +4,7 @@ import {
|
||||
resolveQuestionFormStrategyTaskExecutionId,
|
||||
strategyBlockedMessageFields,
|
||||
strategySettledMessageFields,
|
||||
strategyTaskParkedOnSucceededRun,
|
||||
} from '../../src/runtime/strategy-question-continuation';
|
||||
|
||||
describe('question-form strategy continuation handle recovery', () => {
|
||||
@@ -128,3 +129,53 @@ describe('strategySettledMessageFields', () => {
|
||||
expect(strategySettledMessageFields(undefined)).toBeNull();
|
||||
});
|
||||
});
|
||||
|
||||
describe('strategyTaskParkedOnSucceededRun', () => {
|
||||
const parked = (overrides: Partial<StrategyTaskProjectionV2> = {}) => blockedProjection({
|
||||
outcome: 'clarification_required',
|
||||
terminal: false,
|
||||
blockedContext: undefined,
|
||||
...overrides,
|
||||
});
|
||||
|
||||
it.each(['clarification_required', 'plan_ready'] as const)(
|
||||
'is true when a succeeded Run left its task waiting on the user (%s)',
|
||||
(outcome) => {
|
||||
expect(strategyTaskParkedOnSucceededRun(
|
||||
{ status: 'succeeded', strategyTask: parked({ outcome }) },
|
||||
'run-1',
|
||||
)).toBe(true);
|
||||
},
|
||||
);
|
||||
|
||||
it('is false while the task still runs, even on the same Run', () => {
|
||||
expect(strategyTaskParkedOnSucceededRun(
|
||||
{ status: 'succeeded', strategyTask: parked({ outcome: 'running' }) },
|
||||
'run-1',
|
||||
)).toBe(false);
|
||||
});
|
||||
|
||||
it('is false when the task has moved on to another Run', () => {
|
||||
expect(strategyTaskParkedOnSucceededRun(
|
||||
{ status: 'succeeded', strategyTask: parked({ activeRunId: 'run-2' }) },
|
||||
'run-1',
|
||||
)).toBe(false);
|
||||
});
|
||||
|
||||
it('is false while the Run itself has not succeeded', () => {
|
||||
for (const status of ['queued', 'running', 'failed', 'canceled'] as const) {
|
||||
expect(strategyTaskParkedOnSucceededRun(
|
||||
{ status, strategyTask: parked() },
|
||||
'run-1',
|
||||
)).toBe(false);
|
||||
}
|
||||
});
|
||||
|
||||
it('leaves terminal tasks and task-less Runs to their own checks', () => {
|
||||
expect(strategyTaskParkedOnSucceededRun(
|
||||
{ status: 'succeeded', strategyTask: blockedProjection() },
|
||||
'run-1',
|
||||
)).toBe(false);
|
||||
expect(strategyTaskParkedOnSucceededRun({ status: 'succeeded' }, 'run-1')).toBe(false);
|
||||
});
|
||||
});
|
||||
|
||||
Reference in New Issue
Block a user