fix: 收敛 stale thread 和 tool-call 错误归因

This commit is contained in:
Codex
2026-06-02 12:08:38 +08:00
parent 0092f55249
commit 40a274d52b
6 changed files with 55 additions and 26 deletions
+20 -14
View File
@@ -452,19 +452,8 @@ async function runCodexStdioTurnWithSession(options: CodexStdioTurnOptions, sess
emitEvent({ type: "backend_status", payload: { phase: "thread/resume:completed", threadId } });
} catch (error) {
const failure = normalizeFailure(error);
if (!isStaleThreadResumeFailure(failure)) throw error;
emitEvent({
type: "backend_status",
payload: {
phase: "thread/resume:stale-thread-fallback",
requestedThreadId: options.threadId,
failureKind: failure.failureKind,
message: failure.message,
fallback: "thread/start",
valuesPrinted: false,
},
});
threadId = await startThread("thread/start:after-stale-resume");
if (!isMissingRolloutThreadResumeFailure(failure)) throw error;
throw threadResumeFailure(options.threadId, failure);
}
} else {
threadId = await startThread();
@@ -920,12 +909,27 @@ function normalizeFailure(error: unknown): CodexStdioFailure {
return new CodexStdioFailure(classifyMessageFailureKind(message, "backend-protocol-error"), message, "codex-stdio");
}
function isStaleThreadResumeFailure(error: CodexStdioFailure): boolean {
function isMissingRolloutThreadResumeFailure(error: CodexStdioFailure): boolean {
if (error.phase !== "response:thread/resume") return false;
const text = `${error.message}\n${JSON.stringify(error.details)}`.toLowerCase();
return text.includes("no rollout found for thread id");
}
function threadResumeFailure(threadId: string, error: CodexStdioFailure): CodexStdioFailure {
return new CodexStdioFailure(
"thread-resume-failed",
`codex app-server thread/resume failed for existing thread: ${error.message}`,
"thread/resume",
{
requestedThreadId: threadId,
originalFailureKind: error.failureKind,
originalPhase: error.phase,
originalDetails: redactJson(error.details),
valuesPrinted: false,
},
);
}
function classifyCodexErrorRecord(error: JsonRecord, fallback: FailureKind): FailureKind {
const parts: string[] = [];
if (typeof error.message === "string") parts.push(error.message);
@@ -937,6 +941,8 @@ function classifyCodexErrorRecord(error: JsonRecord, fallback: FailureKind): Fai
function classifyMessageFailureKind(message: string, fallback: FailureKind): FailureKind {
const text = String(message || "").toLowerCase();
if (/invalid[_ -]?prompt/u.test(text) && /invalid function arguments json string|tool_call_id/u.test(text)) return "provider-invalid-tool-call";
if (/invalid function arguments json string/u.test(text)) return "provider-invalid-tool-call";
if (/rate.?limit|too many requests|\b429\b/u.test(text)) return "provider-rate-limited";
if (/\b401\b|\b403\b|unauthori[sz]ed|forbidden|invalid api key|authentication|auth failed|oauth|access token/u.test(text)) return "provider-auth-failed";
if (isProviderUnavailableMessage(text)) return "provider-unavailable";
+2
View File
@@ -12,9 +12,11 @@ export type FailureKind =
| "backend-spawn-failed"
| "backend-json-parse-error"
| "backend-response-invalid"
| "thread-resume-failed"
| "backend-timeout"
| "provider-auth-failed"
| "provider-rate-limited"
| "provider-invalid-tool-call"
| "provider-unavailable"
| "infra-failed"
| "cancelled";
+12 -8
View File
@@ -105,15 +105,18 @@ const selfTest: SelfTestCase = async (context) => {
env: { CODEX_HOME: context.codexHome, AGENTRUN_FAKE_CODEX_MODE: "resume-no-rollout" },
oneShot: true,
}) as JsonRecord;
assert.equal(staleThreadResult.terminalStatus, "completed", "stale thread resume should start a replacement thread and complete the turn");
assert.equal(staleThreadResult.terminalStatus, "failed", "stale thread resume must fail instead of creating a replacement thread");
assert.equal(staleThreadResult.failureKind, "thread-resume-failed", "stale thread resume failure kind");
const staleEnvelope = await client.get(`/api/v1/runs/${staleThread.runId}/commands/${staleThread.commandId}/result`) as JsonRecord;
assert.equal(staleEnvelope.terminalStatus, "completed");
assert.equal(staleEnvelope.completed, true);
assert.equal((staleEnvelope.sessionRef as JsonRecord).threadId, "thread_selftest_1");
assert.equal(staleEnvelope.terminalStatus, "failed");
assert.equal(staleEnvelope.failureKind, "thread-resume-failed");
assert.equal(staleEnvelope.completed, false);
assert.equal((staleEnvelope.sessionRef as JsonRecord).threadId, "thread_missing_rollout");
const staleEvents = await client.get(`/api/v1/runs/${staleThread.runId}/events?afterSeq=0&limit=100`) as { items?: Array<{ type: string; payload: unknown }> };
assert.equal(staleEvents.items?.some((event) => event.type === "backend_status" && eventPayload(event).phase === "thread/resume:stale-thread-fallback"), true, "stale resume fallback must be visible");
assert.equal(staleEvents.items?.some((event) => event.type === "backend_status" && eventPayload(event).phase === "thread/start:after-stale-resume:completed"), true, "stale resume must start a replacement thread");
const stalePhases = (staleEvents.items ?? []).filter((event) => event.type === "backend_status").map((event) => String(eventPayload(event).phase ?? ""));
assert.equal(stalePhases.some((phase) => phase.startsWith("thread/start")), false, "stale resume must not start a replacement thread");
assert.equal(staleEvents.items?.some((event) => event.type === "backend_status" && eventPayload(event).phase === "thread/resume:completed"), false, "stale resume must not be reported as a successful resume");
assert.equal(staleEvents.items?.some((event) => event.type === "error" && eventPayload(event).failureKind === "thread-resume-failed"), true, "stale resume failure must be visible as an error event");
assertNoSecretLeak({ staleThreadResult, staleEnvelope, staleEvents });
const live = await createRunWithCommand(client, context, "hello live events", "selftest-live-tool-events", 15_000);
@@ -173,6 +176,7 @@ const selfTest: SelfTestCase = async (context) => {
await runFailureCase({ client, managerUrl: server.baseUrl, context, mode: "missing-turn-result", expectedStatus: "failed", expectedFailureKind: "backend-response-invalid" });
await runFailureCase({ client, managerUrl: server.baseUrl, context, mode: "provider-401-rpc-error", expectedStatus: "failed", expectedFailureKind: "provider-auth-failed" });
await runFailureCase({ client, managerUrl: server.baseUrl, context, mode: "provider-429-terminal", expectedStatus: "failed", expectedFailureKind: "provider-rate-limited" });
await runFailureCase({ client, managerUrl: server.baseUrl, context, mode: "provider-invalid-tool-call", expectedStatus: "failed", expectedFailureKind: "provider-invalid-tool-call" });
await runFailureCase({ client, managerUrl: server.baseUrl, context, mode: "provider-503-rpc-error", expectedStatus: "failed", expectedFailureKind: "provider-unavailable" });
await runFailureCase({ client, managerUrl: server.baseUrl, context, mode: "provider-503-terminal", expectedStatus: "failed", expectedFailureKind: "provider-unavailable" });
await runFailureCase({ client, managerUrl: server.baseUrl, context, mode: "provider-503-retry-event", expectedStatus: "failed", expectedFailureKind: "provider-unavailable", expectRetryError: true });
@@ -181,7 +185,7 @@ const selfTest: SelfTestCase = async (context) => {
await runSecretFailureCase({ client, managerUrl: server.baseUrl, context });
await runSpawnFailureCase({ client, managerUrl: server.baseUrl, context });
return { name: "codex-stdio", tests: ["runner-lease-heartbeat", "codex-stdio-fake-turn", "codex-stdio-projected-writable-home", "codex-stdio-deepseek-profile-fake-turn", "codex-stdio-minimax-m3-profile-fake-turn", "codex-stdio-deepseek-missing-secret-no-fallback", "codex-stdio-minimax-m3-missing-secret-no-fallback", "codex-stdio-config-model-authoritative", "codex-stdio-explicit-model-forwarded", "codex-stdio-final-agent-message-only", "codex-stdio-stale-thread-fallback", "codex-stdio-live-tool-events", "codex-stdio-noisy-reasoning-suppression", "codex-stdio-missing-turn-result", "codex-stdio-provider-auth-failed", "codex-stdio-provider-rate-limited", "codex-stdio-provider-503-rpc-error", "codex-stdio-provider-503-terminal", "codex-stdio-provider-503-retry-event", "codex-stdio-invalid-json", "codex-stdio-timeout", "codex-stdio-secret-unavailable", "codex-stdio-spawn-failure"] };
return { name: "codex-stdio", tests: ["runner-lease-heartbeat", "codex-stdio-fake-turn", "codex-stdio-projected-writable-home", "codex-stdio-deepseek-profile-fake-turn", "codex-stdio-minimax-m3-profile-fake-turn", "codex-stdio-deepseek-missing-secret-no-fallback", "codex-stdio-minimax-m3-missing-secret-no-fallback", "codex-stdio-config-model-authoritative", "codex-stdio-explicit-model-forwarded", "codex-stdio-final-agent-message-only", "codex-stdio-stale-thread-resume-fails", "codex-stdio-live-tool-events", "codex-stdio-noisy-reasoning-suppression", "codex-stdio-missing-turn-result", "codex-stdio-provider-auth-failed", "codex-stdio-provider-rate-limited", "codex-stdio-provider-invalid-tool-call", "codex-stdio-provider-503-rpc-error", "codex-stdio-provider-503-terminal", "codex-stdio-provider-503-retry-event", "codex-stdio-invalid-json", "codex-stdio-timeout", "codex-stdio-secret-unavailable", "codex-stdio-spawn-failure"] };
} finally {
await new Promise<void>((resolve) => server.server.close(() => resolve()));
}
@@ -264,7 +268,7 @@ async function createStaleThreadRun(client: ManagerClient, context: SelfTestCont
},
traceSink: null,
}) as { id: string };
const command = await client.post(`/api/v1/runs/${run.id}/commands`, { type: "turn", payload: { prompt: "hello stale thread" }, idempotencyKey: "selftest-stale-thread-fallback" }) as { id: string };
const command = await client.post(`/api/v1/runs/${run.id}/commands`, { type: "turn", payload: { prompt: "hello stale thread" }, idempotencyKey: "selftest-stale-thread-resume-fails" }) as { id: string };
return { runId: run.id, commandId: command.id };
}
+15
View File
@@ -100,6 +100,21 @@ for await (const line of rl) {
respond(message.id, { turn });
continue;
}
if (mode === "provider-invalid-tool-call") {
turnCounter += 1;
const turn = {
id: `turn_selftest_${turnCounter}`,
status: "failed",
error: {
message: "invalid params, invalid function arguments json string, tool_call_id: call_function_selftest_2 (2013)",
code: "invalid_prompt",
},
};
notify("turn/started", { turn: { id: turn.id, status: "running" } });
notify("turn/completed", { turn });
respond(message.id, { turn });
continue;
}
if (mode === "provider-503-retry-event") {
turnCounter += 1;
const turn = {