mirror of
https://github.com/QwenLM/qwen-code.git
synced 2026-08-15 03:34:59 +00:00
test(integration): harden acp-cron diagnostics and isolation (#8076)
Fail fast when the cron_create tool call is not served to the first user prompt instead of timing out opaquely 75s later, and dump the fake server request log on failure so a dispatch shift is quick to diagnose. Close the fake server even when test setup throws, drop the dead FAKE_SERVER_OPTIONS (container mode is skipped by IS_SANDBOX), and move QWEN_HOME out of the agent workspace cwd so workspace scans never see it.
This commit is contained in:
Qwen Code Bot
parent
fa4d7592aa
commit
87a02fb498
1 changed files with 88 additions and 63 deletions
|
|
@ -17,12 +17,12 @@
|
|||
*/
|
||||
|
||||
import { spawn } from 'node:child_process';
|
||||
import { mkdirSync, readFileSync, writeFileSync } from 'node:fs';
|
||||
import { join } from 'node:path';
|
||||
import { mkdirSync, readFileSync, rmSync, writeFileSync } from 'node:fs';
|
||||
import { basename, dirname, join } from 'node:path';
|
||||
import { createInterface } from 'node:readline';
|
||||
import { setTimeout as delay } from 'node:timers/promises';
|
||||
import { describe, it, expect } from 'vitest';
|
||||
import { TestRig, fakeServerHostOptions } from '../test-helper.js';
|
||||
import { TestRig } from '../test-helper.js';
|
||||
import {
|
||||
startFakeOpenAIServer,
|
||||
fakeToolCall,
|
||||
|
|
@ -35,8 +35,6 @@ const IS_SANDBOX =
|
|||
process.env['QWEN_SANDBOX'] &&
|
||||
process.env['QWEN_SANDBOX']!.toLowerCase() !== 'false';
|
||||
|
||||
const FAKE_SERVER_OPTIONS = fakeServerHostOptions();
|
||||
|
||||
type PendingRequest = {
|
||||
resolve: (value: unknown) => void;
|
||||
reject: (reason: Error) => void;
|
||||
|
|
@ -87,7 +85,12 @@ function setupAcpCronTest(rig: TestRig, fakeServer: FakeOpenAIServer) {
|
|||
})[] = [];
|
||||
const stderr: string[] = [];
|
||||
|
||||
const qwenHome = join(rig.testDir!, '.qwen-home');
|
||||
// Keep the global config dir outside the agent's workspace cwd so workspace
|
||||
// scans (memory discovery, file tooling) never see it.
|
||||
const qwenHome = join(
|
||||
dirname(rig.testDir!),
|
||||
`${basename(rig.testDir!)}-home`,
|
||||
);
|
||||
mkdirSync(qwenHome, { recursive: true });
|
||||
|
||||
const agent = spawn(
|
||||
|
|
@ -103,6 +106,7 @@ function setupAcpCronTest(rig: TestRig, fakeServer: FakeOpenAIServer) {
|
|||
OPENAI_BASE_URL: fakeServer.baseUrl,
|
||||
OPENAI_MODEL: 'fake-model',
|
||||
QWEN_MODEL: 'fake-model',
|
||||
// Defends against an ambient proxy intercepting the local fake server.
|
||||
NO_PROXY: '127.0.0.1,localhost',
|
||||
no_proxy: '127.0.0.1,localhost',
|
||||
},
|
||||
|
|
@ -281,6 +285,7 @@ function setupAcpCronTest(rig: TestRig, fakeServer: FakeOpenAIServer) {
|
|||
pending.forEach(({ timeout }) => clearTimeout(timeout));
|
||||
pending.clear();
|
||||
await waitForExit();
|
||||
rmSync(qwenHome, { recursive: true, force: true });
|
||||
};
|
||||
|
||||
return {
|
||||
|
|
@ -340,72 +345,92 @@ async function initSession(
|
|||
};
|
||||
}
|
||||
return { content: 'Done.' };
|
||||
}, FAKE_SERVER_OPTIONS);
|
||||
|
||||
const { sendRequest, cleanup, stderr, waitForSessionUpdate } =
|
||||
setupAcpCronTest(rig, fakeServer);
|
||||
});
|
||||
|
||||
try {
|
||||
const sessionId = await initSession(sendRequest, rig.testDir!);
|
||||
const { sendRequest, cleanup, stderr, waitForSessionUpdate } =
|
||||
setupAcpCronTest(rig, fakeServer);
|
||||
|
||||
// --- Part 1: Create a cron job that fires every minute ---
|
||||
const createResult = (await sendRequest('session/prompt', {
|
||||
sessionId,
|
||||
prompt: [
|
||||
{
|
||||
type: 'text',
|
||||
text: 'Call cron_create with cron expression "*/1 * * * *" and prompt "Say CRONFIRE7742 and nothing else" and recurring true. Confirm briefly.',
|
||||
},
|
||||
],
|
||||
})) as { stopReason: string };
|
||||
expect(createResult.stopReason).toBe('end_turn');
|
||||
try {
|
||||
const sessionId = await initSession(sendRequest, rig.testDir!);
|
||||
|
||||
const promptDoneAt = Date.now();
|
||||
// --- Part 1: Create a cron job that fires every minute ---
|
||||
const createResult = (await sendRequest('session/prompt', {
|
||||
sessionId,
|
||||
prompt: [
|
||||
{
|
||||
type: 'text',
|
||||
text: 'Call cron_create with cron expression "*/1 * * * *" and prompt "Say CRONFIRE7742 and nothing else" and recurring true. Confirm briefly.',
|
||||
},
|
||||
],
|
||||
})) as { stopReason: string };
|
||||
expect(createResult.stopReason).toBe('end_turn');
|
||||
|
||||
// --- Part 2: Session stays responsive while cron is pending ---
|
||||
const interactiveResult = (await sendRequest('session/prompt', {
|
||||
sessionId,
|
||||
prompt: [
|
||||
{
|
||||
type: 'text',
|
||||
text: 'Say INTERACTIVE8899 and nothing else.',
|
||||
},
|
||||
],
|
||||
})) as { stopReason: string };
|
||||
expect(interactiveResult.stopReason).toBe('end_turn');
|
||||
// Fail fast if the cron_create tool call was not served to the first
|
||||
// user prompt. An internal model call before the first prompt (title
|
||||
// generation, a classifier pass) would shift dispatch and otherwise
|
||||
// surface only as an opaque 75s timeout in Part 3.
|
||||
expect(
|
||||
JSON.stringify(fakeServer.requests[0]?.body['messages']),
|
||||
'requestIndex 0 was not the cron_create prompt — dispatch shifted',
|
||||
).toContain('CRONFIRE7742');
|
||||
|
||||
// --- Part 3: Wait for cron-fired notification (up to 75s) ---
|
||||
// The cron fires at the next minute boundary. The model response
|
||||
// should stream back as sessionUpdate notifications after the
|
||||
// originating prompt has already returned.
|
||||
const promptDoneAt = Date.now();
|
||||
|
||||
// 3a: Check for user_message_chunk echoing the cron prompt with _meta.source
|
||||
const cronUserMsg = await waitForSessionUpdate(
|
||||
(u) =>
|
||||
u.update?.sessionUpdate === 'user_message_chunk' &&
|
||||
(u.update?.content?.text ?? '').includes('CRONFIRE7742') &&
|
||||
u.receivedAt > promptDoneAt,
|
||||
'cron-fired user_message_chunk with CRONFIRE7742',
|
||||
75_000,
|
||||
);
|
||||
expect(cronUserMsg.update?._meta).toBeDefined();
|
||||
expect(cronUserMsg.update?._meta?.source).toBe('cron');
|
||||
// --- Part 2: Session stays responsive while cron is pending ---
|
||||
const interactiveResult = (await sendRequest('session/prompt', {
|
||||
sessionId,
|
||||
prompt: [
|
||||
{
|
||||
type: 'text',
|
||||
text: 'Say INTERACTIVE8899 and nothing else.',
|
||||
},
|
||||
],
|
||||
})) as { stopReason: string };
|
||||
expect(interactiveResult.stopReason).toBe('end_turn');
|
||||
|
||||
// 3b: Check for agent_message_chunk after the cron user message
|
||||
// (the model's response to the cron prompt)
|
||||
const cronAgentMsg = await waitForSessionUpdate(
|
||||
(u) =>
|
||||
u.update?.sessionUpdate === 'agent_message_chunk' &&
|
||||
u.receivedAt > cronUserMsg.receivedAt,
|
||||
'agent_message_chunk after cron fire',
|
||||
15_000, // should already be here by now
|
||||
);
|
||||
expect(cronAgentMsg.receivedAt).toBeGreaterThan(promptDoneAt);
|
||||
} catch (e) {
|
||||
if (stderr.length) console.error('Agent stderr:', stderr.join(''));
|
||||
throw e;
|
||||
// --- Part 3: Wait for cron-fired notification (up to 75s) ---
|
||||
// The cron fires at the next minute boundary. The model response
|
||||
// should stream back as sessionUpdate notifications after the
|
||||
// originating prompt has already returned.
|
||||
|
||||
// 3a: Check for user_message_chunk echoing the cron prompt with _meta.source
|
||||
const cronUserMsg = await waitForSessionUpdate(
|
||||
(u) =>
|
||||
u.update?.sessionUpdate === 'user_message_chunk' &&
|
||||
(u.update?.content?.text ?? '').includes('CRONFIRE7742') &&
|
||||
u.receivedAt > promptDoneAt,
|
||||
'cron-fired user_message_chunk with CRONFIRE7742',
|
||||
75_000,
|
||||
);
|
||||
expect(cronUserMsg.update?._meta).toBeDefined();
|
||||
expect(cronUserMsg.update?._meta?.source).toBe('cron');
|
||||
|
||||
// 3b: Check for agent_message_chunk after the cron user message
|
||||
// (the model's response to the cron prompt)
|
||||
const cronAgentMsg = await waitForSessionUpdate(
|
||||
(u) =>
|
||||
u.update?.sessionUpdate === 'agent_message_chunk' &&
|
||||
u.receivedAt > cronUserMsg.receivedAt,
|
||||
'agent_message_chunk after cron fire',
|
||||
15_000, // should already be here by now
|
||||
);
|
||||
expect(cronAgentMsg.receivedAt).toBeGreaterThan(promptDoneAt);
|
||||
} catch (e) {
|
||||
if (stderr.length) console.error('Agent stderr:', stderr.join(''));
|
||||
console.error(
|
||||
'Fake server requests:',
|
||||
fakeServer.requests.map((r, i) => ({
|
||||
index: i,
|
||||
model: r.body['model'],
|
||||
messages: JSON.stringify(r.body['messages']).slice(0, 300),
|
||||
})),
|
||||
);
|
||||
throw e;
|
||||
} finally {
|
||||
await cleanup();
|
||||
}
|
||||
} finally {
|
||||
await cleanup();
|
||||
await fakeServer.close();
|
||||
}
|
||||
},
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue