Files
rea/src/process/ArtifactAnalysisResources.test.ts
T
morluto 3fcba07962 fix(process): preserve file outcomes through exact cleanup recovery
Classify canonical physical file outcomes through the provider mapper before composing cleanup. Retain unknown capabilities as unresolved and replace runtime failure maps with exact leases, including mixed cleanup and partial retry.

Keep completed-read cancellation output and cleanup history. Closes #4865.

Validation: 179 unique focused tests passed across 13 selected files, 7 Linux-only skips; static checks, docs checks and 1551-file discovery passed. Authored Node producers and controlled real-descriptor faults do not qualify native providers.
2026-10-18 02:55:41 +08:00

901 lines
27 KiB
TypeScript

import { AnalysisCleanupLease } from "../artifacts/AnalysisCleanupLease.js";
import { ArtifactResourceRetryFailure } from "../artifacts/ArtifactResourceScope.js";
import { ArtifactResourceScope } from "../artifacts/ArtifactResourceScope.js";
import { withArtifactRegularFile } from "../artifacts/ArtifactFileOperation.js";
import { writeFile } from "node:fs/promises";
import { join } from "node:path";
import { createTestTempDirectory } from "../../tests/fixtures/temporaryDirectory.js";
import { OwnedCommandFailure } from "./OwnedCommand.js";
import { EventEmitter } from "node:events";
import { PassThrough } from "node:stream";
import { expect, it } from "vitest";
import { readStableArtifact } from "../artifacts/readStableArtifact.js";
import { ArtifactReaderFailure } from "../artifacts/ArtifactReader.js";
import {
AnalysisCancelledError,
AnalysisOutputError,
} from "../domain/analysisErrorCore.js";
import { createEvidence } from "../domain/evidence.js";
import { projectAnalysisError } from "../domain/analysisErrorProjection.js";
import { ProviderAdapterError } from "../domain/providerAdapterError.js";
import { ProviderCleanupError } from "../domain/providerCleanupError.js";
import { err, ok } from "../domain/result.js";
import { ArtifactAnalysisResources } from "./ArtifactAnalysisResources.js";
import { readLinuxResourceLimits } from "./LinuxResourceLimits.js";
import { ProviderProcessSupervisor } from "./ProviderProcess.js";
it("projects the specific reason and member evidence for retained cleanup", async () => {
const producer = Object.assign(new EventEmitter(), {
stdout: new PassThrough(),
stderr: new PassThrough(),
exitCode: null,
signalCode: null,
kill: () => true,
});
const owner = new ProviderProcessSupervisor({
process: producer,
ownsProcessLifetime: true,
ownership: {
runId: "rea-tshark-run-1",
leaderPid: 4242,
processGroupId: 4242,
},
cleanup: async () => ({
cleaned: false as const,
reason: "process ownership token could not be read for 1 live process",
failures: [
{
pid: 101,
reason: "environment-unreadable" as const,
diagnostic: "sysctl_errno_22",
},
],
unverified: [
{ pid: 900, diagnostic: "platform_binary_environment_withheld" },
],
}),
});
const resources = new ArtifactAnalysisResources(
"fixture-provider",
"fixture-operation",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
resources.retain({ path: "runtime-root", close: async () => {} }, owner);
const result = await resources.finish(
err(new ProviderAdapterError("fixture-provider", "fixture-operation")),
() => {
throw new Error("unexpected success observation");
},
);
owner.dispose();
if (result.ok) throw new Error("expected cleanup failure");
expect(projectAnalysisError(result.error)).toMatchObject({
code: "cleanup_incomplete",
details: {
resources: [
"rea-tshark-run-1",
"runtime-root",
"pid:4242",
"process-group:4242",
],
diagnostics: {
cleanup_error: {
details: {
diagnostics: {
reason:
"process ownership token could not be read for 1 live process",
failures: [
{
pid: 101,
reason: "environment-unreadable",
diagnostic: "sysctl_errno_22",
},
],
unverified: [
{
pid: 900,
diagnostic: "platform_binary_environment_withheld",
},
],
},
},
},
},
},
});
});
it("preserves the primary error and resolves cleanup status after retry", async () => {
let closes = 0;
const resources = new ArtifactAnalysisResources(
"fixture-provider",
"fixture-operation",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
resources.retain({
path: "runtime-root",
close: async () => {
closes++;
if (closes === 1) throw new Error("temporary close failure");
},
});
const primary = new AnalysisOutputError(
"fixture-operation",
"producer returned malformed output",
{
partialObservation: createEvidence(
undefined,
{
id: "fixture",
name: "Fixture",
version: null,
},
{
operation: "inspect_signal_decoders",
parameters: {},
result: { decoders: [] },
confidence: "observed",
},
),
},
);
const result = await resources.finish(err(primary), () => {
throw new Error("unexpected success observation");
});
if (result.ok) throw new Error("expected cleanup failure");
expect(result.error.cause).toBeInstanceOf(AggregateError);
expect((result.error.cause as AggregateError).errors).toEqual([
primary,
expect.any(ProviderCleanupError),
]);
expect(result.error.partialObservation).toEqual(primary.partialObservation);
expect(result.error.cleanupIncomplete).toBe(true);
expect(result.error.diagnostics).toMatchObject({
primary_error: { code: "unreadable_output" },
});
await resources.release();
expect(result.error.cleanupIncomplete).toBe(false);
expect(closes).toBe(2);
await resources.close();
});
it("resolves each returned cleanup error with its own retained root", async () => {
let closesA = 0;
let closesB = 0;
const resources = new ArtifactAnalysisResources(
"fixture-provider",
"fixture-operation",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
resources.retain({
path: "runtime-root-a",
close: async () => {
closesA++;
if (closesA === 1) throw new Error("root A close failure");
},
});
const first = await resources.finish(
err(new AnalysisOutputError("fixture-operation", "primary failure")),
() => {
throw new Error("unexpected success observation");
},
);
if (first.ok) throw new Error("expected root A cleanup failure");
resources.retain({
path: "runtime-root-b",
close: async () => {
closesB++;
if (closesB === 1) throw new Error("root B close failure");
},
});
const partial = createEvidence(
undefined,
{ id: "fixture", name: "Fixture", version: null },
{
operation: "fixture-operation",
parameters: {},
result: { state: "partial" },
confidence: "observed",
},
);
const second = await resources.finish(ok("completed"), () => partial);
if (second.ok) throw new Error("expected root B cleanup failure");
expect(second.error.partialObservation).toEqual(partial);
expect(first.error.cleanupIncomplete).toBe(false);
expect(second.error.cleanupIncomplete).toBe(true);
await resources.release();
expect(second.error.cleanupIncomplete).toBe(false);
await resources.close();
});
it("resolves an artifact cleanup error when its scope owner later closes", async () => {
let closes = 0;
const readArtifact: typeof readStableArtifact = async (
_path,
_maximumBytes,
_signal,
_fileSystem,
scope,
) => {
if (scope === undefined) throw new Error("expected scoped artifact read");
if (closes === 0) {
const owner = {
kind: "lease",
resource: "fixture descriptor",
lease: new AnalysisCleanupLease(async () => {
closes++;
if (closes === 1) throw new Error("descriptor remains open");
}, ["fixture descriptor"]),
} as const;
expect(await scope.release(owner)).toMatchObject({ kind: "failed" });
throw new ArtifactResourceRetryFailure(
[{ owner, lease: owner.lease }],
"Descriptor close failed",
{
cleanup: {
reason: "descriptor remains open",
resources: ["fixture descriptor"],
},
},
);
}
return { bytes: Buffer.alloc(0), sha256: "0".repeat(64) };
};
const resources = new ArtifactAnalysisResources(
"fixture-provider",
"fixture-operation",
readArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
const first = await resources.run(async () =>
ok(await resources.read("fixture", 1)),
);
if (first.ok) throw new Error("expected descriptor cleanup failure");
expect(first.error.cleanupIncomplete).toBe(true);
const retried = await resources.run(async () =>
ok(await resources.read("fixture", 1)),
);
expect(retried.ok).toBe(true);
expect(first.error.cleanupIncomplete).toBe(false);
expect(closes).toBe(2);
await resources.close();
});
it("keeps a runtime root until its descriptor cleanup owner is released", async () => {
let descriptorCloses = 0;
let rootCloses = 0;
const readArtifact: typeof readStableArtifact = async (
_path,
_maximumBytes,
_signal,
_fileSystem,
scope,
) => {
if (scope === undefined) throw new Error("expected scoped artifact read");
if (descriptorCloses === 0) {
const owner = {
kind: "lease",
resource: "fixture descriptor",
lease: new AnalysisCleanupLease(async () => {
descriptorCloses++;
if (descriptorCloses === 1)
throw new Error("descriptor remains open");
}, ["fixture descriptor"]),
} as const;
expect(await scope.release(owner)).toMatchObject({ kind: "failed" });
throw new ArtifactResourceRetryFailure(
[{ owner, lease: owner.lease }],
"Descriptor close failed",
{
cleanup: {
reason: "descriptor remains open",
resources: ["fixture descriptor"],
},
},
);
}
return { bytes: Buffer.alloc(0), sha256: "0".repeat(64) };
};
const resources = new ArtifactAnalysisResources(
"fixture-provider",
"fixture-operation",
readArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
const first = await resources.run(async () =>
ok(await resources.read("fixture", 1)),
);
if (first.ok) throw new Error("expected descriptor cleanup failure");
resources.retain({
path: "runtime-root",
close: async () => {
rootCloses++;
},
});
const outer = await resources.finish(first, () => {
throw new Error("unexpected success observation");
});
if (outer.ok) throw new Error("expected combined cleanup failure");
expect(rootCloses).toBe(0);
expect(outer.error.cleanupResources).toEqual([
"fixture descriptor",
"runtime-root",
]);
expect(outer.error.cause).toBeInstanceOf(AggregateError);
expect((outer.error.cause as AggregateError).errors[0]).toBe(first.error);
expect(first.error.cleanupIncomplete).toBe(true);
const retried = await resources.run(async () =>
ok(await resources.read("fixture", 1)),
);
expect(retried.ok).toBe(true);
expect(first.error.cleanupIncomplete).toBe(false);
await resources.release();
expect(rootCloses).toBe(1);
expect(outer.error.cleanupIncomplete).toBe(false);
await resources.close();
});
it.each([
null,
"opaque reason",
new ProviderCleanupError("foreign", ["foreign-resource"], {
reason: "caller cleanup",
}),
])(
"preserves caller cancellation at resource admission: %s",
async (reason) => {
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
let calls = 0;
try {
const result = await resources.run(async () => {
calls++;
return ok("unexpected");
}, AbortSignal.abort(reason));
expect(result.ok).toBe(false);
if (result.ok) throw new Error("Expected cancelled admission");
expect(result.error.cause).toBe(reason);
expect(result.error.cleanupResources).toEqual([]);
expect(calls).toBe(0);
if (reason instanceof ProviderCleanupError)
expect(reason.cleanupIncomplete).toBe(true);
} finally {
await resources.close();
}
},
);
it.each([
null,
"opaque reason",
new ProviderCleanupError("foreign", ["foreign-resource"], {
reason: "caller cleanup",
}),
])(
"projects the command owner's exact cancellation independently of physical causes: %s",
async (reason) => {
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
const physical = new Error("physical collection interruption");
const failure = new OwnedCommandFailure(
"cancelled",
"cancelled command",
null,
null,
{ cause: physical, cancellation: { reason } },
);
try {
const error = resources.commandFailure(
failure,
{ command_timeout_ms: 100, output_bytes: 1024 },
undefined,
);
expect(error).toMatchObject({ _tag: "AnalysisCancelledError" });
expect(error.cause).toBe(reason);
expect(error.cleanupResources).toEqual([]);
expect(failure.cause).toBe(physical);
if (reason instanceof ProviderCleanupError)
expect(reason.cleanupIncomplete).toBe(true);
} finally {
await resources.close();
}
},
);
it.each([false, true])(
"preserves artifact cancellation and independent failed cleanup (failed close: %s)",
async (failedClose) => {
const reason = { request: "artifact cancellation" };
const root = await createTestTempDirectory("rea-resource-cancel-");
const path = join(root, "descriptor");
await writeFile(path, "authored bytes");
let mayClose = false;
const readArtifact: typeof readStableArtifact = async (
_path,
_limit,
_signal,
_fs,
scope,
) => {
const primary = new ArtifactReaderFailure("cancelled", "read cancelled", {
cause: reason,
});
if (!failedClose) throw primary;
if (scope === undefined) throw new Error("Expected retained file scope");
return withArtifactRegularFile(path, scope, _signal, async (file) => {
const close = file.close.bind(file);
file.close = async () => {
if (!mayClose) throw new Error("injected descriptor close failure");
await close();
};
throw primary;
});
};
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
try {
const result = await resources.run(async () =>
ok(await resources.read("fixture", 1)),
);
if (result.ok) throw new Error("Expected cancelled artifact read");
const primary = failedClose ? result.error.cleanupPrimary : result.error;
expect(primary).toMatchObject({ _tag: "AnalysisCancelledError" });
expect(primary?.cause).toBe(reason);
expect(result.error.cleanupIncomplete).toBe(failedClose);
mayClose = true;
await resources.close();
expect(result.error.cleanupIncomplete).toBe(false);
expect(primary?.cause).toBe(reason);
} finally {
mayClose = true;
await resources.close();
}
},
);
it("retains a physical command failure when the owner also observed late cancellation", async () => {
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
const reason = { request: "late cancellation" };
const failure = new OwnedCommandFailure(
"process",
"physical failure",
null,
null,
{ cancellation: { reason } },
);
try {
const error = resources.commandFailure(
failure,
{ command_timeout_ms: 100, output_bytes: 1024 },
undefined,
);
expect(error).toMatchObject({ _tag: "AnalysisOutputError" });
expect(error.cause).toBe(failure);
} finally {
await resources.close();
}
});
it("preserves typed caller cancellation through failed workspace cleanup and retry", async () => {
let denied = true;
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
const reason = { request: "cancelled before workspace cleanup" };
const primary = new AnalysisCancelledError("inspect_signal_decoders", {
cause: reason,
});
resources.retain({
path: "workspace",
close: async () => {
if (denied) throw new Error("workspace close denied");
},
});
try {
const result = await resources.finish(err(primary), () => {
throw new Error("Must not produce success Evidence");
});
if (result.ok) throw new Error("Expected retained workspace");
expect(result.error.cleanupPrimary).toBe(primary);
expect(result.error.cleanupPrimary?.cause).toBe(reason);
await expect(resources.close()).rejects.toMatchObject({
cleanupIncomplete: true,
});
denied = false;
await resources.close();
expect(result.error.cleanupIncomplete).toBe(false);
expect(result.error.cleanupPrimary?.cause).toBe(reason);
} finally {
denied = false;
await resources.close();
}
});
it("observes cancellation after pending descriptor cleanup before admitting replacement work", async () => {
let closes = 0;
let cleanupStarted: () => void = () => {};
const started = new Promise<void>((resolve) => {
cleanupStarted = resolve;
});
let releaseCleanup: () => void = () => {};
const held = new Promise<void>((resolve) => {
releaseCleanup = resolve;
});
const readArtifact: typeof readStableArtifact = async (
_path,
_limit,
_signal,
_fs,
scope,
) => {
if (scope === undefined) throw new Error("Expected file scope");
const retained = {
kind: "lease",
resource: "descriptor",
lease: new AnalysisCleanupLease(async () => {
closes++;
if (closes === 1) throw new Error("descriptor close denied");
cleanupStarted();
await held;
}, ["descriptor"]),
} as const;
await scope.release(retained);
throw new ArtifactResourceRetryFailure(
[{ owner: retained, lease: retained.lease }],
"descriptor cleanup failed",
{
cleanup: {
reason: "descriptor close denied",
resources: ["descriptor"],
},
},
);
};
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
let calls = 0;
try {
const first = await resources.run(async () =>
ok(await resources.read("fixture", 1)),
);
expect(first).toMatchObject({
ok: false,
error: { cleanupIncomplete: true },
});
const caller = new AbortController();
const reason = { request: "cancel while old descriptor closes" };
const pending = resources.run(async () => {
calls++;
return ok("Must not run");
}, caller.signal);
await started;
caller.abort(reason);
releaseCleanup();
const result = await pending;
if (result.ok) throw new Error("Expected cancelled replacement");
expect(result.error.cause).toBe(reason);
expect(calls).toBe(0);
if (!first.ok) expect(first.error.cleanupIncomplete).toBe(false);
} finally {
releaseCleanup();
await resources.close();
}
});
it.each([false, true])(
"keeps an existing physical primary across cleanup-time cancellation (failed close: %s)",
async (failsClose) => {
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
const caller = new AbortController();
const primary = new AnalysisOutputError(
"inspect_signal_decoders",
"physical producer report failure",
);
let denied = failsClose;
resources.retain({
path: "workspace",
close: async () => {
caller.abort({ request: "late cancellation" });
if (denied) throw new Error("workspace close denied");
},
});
try {
const result = await resources.finish(
err(primary),
() => {
throw new Error("Failed work must not produce success Evidence");
},
"inspect_signal_decoders",
caller.signal,
);
if (result.ok) throw new Error("Expected original physical failure");
expect(failsClose ? result.error.cleanupPrimary : result.error).toBe(
primary,
);
denied = false;
await resources.close();
expect(result.error.cleanupIncomplete).toBe(false);
expect(failsClose ? result.error.cleanupPrimary : result.error).toBe(
primary,
);
} finally {
denied = false;
await resources.close();
}
},
);
it("keeps cleanup with no reported capability unresolved instead of clearing it from an empty scope", async () => {
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
const source = new ArtifactReaderFailure("io", "unverified cleanup", {
cleanup: {
reason: "no cleanup capability reported",
resources: ["unverified-resource"],
},
});
const result = await resources.run(async () => {
throw source;
});
if (result.ok) throw new Error("Expected unresolved cleanup");
expect(result.error.cleanupIncomplete).toBe(true);
expect(result.error.cleanupResources).toEqual(["unverified-resource"]);
let calls = 0;
const blocked = await resources.run(async () => {
calls++;
return ok("Must not run");
});
expect(blocked).toMatchObject({
ok: false,
error: { cleanupIncomplete: true },
});
expect(calls).toBe(0);
await expect(resources.close()).rejects.toMatchObject({
cleanupIncomplete: true,
});
expect(result.error.cleanupIncomplete).toBe(true);
// No OS resource was supplied. An unresolved observation never grants cleanup authority.
});
it("joins mixed runtime cleanup and resolves each error only through its exact lease", async () => {
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
let denied = true;
let healthyCloses = 0;
const bad = {
path: "bad-root",
close: async () => {
if (denied) throw new Error("bad root denied");
},
};
const good = {
path: "good-root",
close: async () => {
healthyCloses++;
},
};
resources.retain(bad);
resources.retain(good);
try {
const failure: unknown = await resources
.release()
.catch((cause: unknown) => cause);
expect(failure).toMatchObject({
cleanupIncomplete: true,
cleanupResources: ["bad-root"],
});
expect(healthyCloses).toBe(1);
expect(resources.pendingPaths).toEqual(["bad-root"]);
denied = false;
await resources.close();
expect(failure).toMatchObject({
cleanupIncomplete: false,
cleanupResources: [],
});
expect(healthyCloses).toBe(1);
} finally {
denied = false;
await resources.close();
}
});
it("retains producer output when a completed read is cancelled during failed close", async () => {
const root = await createTestTempDirectory("rea-completed-read-cancel-");
const path = join(root, "descriptor");
await writeFile(path, "authored bytes");
const caller = new AbortController();
const reason = { request: "cancel during close" };
const scope = new ArtifactResourceScope();
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
let mayClose = false;
const output = {
stdout: "completed producer output",
stderr: "",
truncated: false,
exit_code: 0,
};
try {
const cause: unknown = await withArtifactRegularFile(
path,
scope,
caller.signal,
async (file) => {
const close = file.close.bind(file);
file.close = async () => {
caller.abort(reason);
if (!mayClose) throw new Error("controlled descriptor close failure");
await close();
};
return await file.readFile();
},
).catch((cause: unknown) => cause);
if (!(cause instanceof ArtifactReaderFailure))
throw new Error("Expected retained artifact failure");
const error = resources.cleanupFailure(
cause,
() => {
throw new Error("Completed read has no physical failure");
},
output,
);
const primary = error.cleanupPrimary;
expect(primary).toMatchObject({
_tag: "AnalysisCancelledError",
cause: reason,
capturedOutput: output,
});
expect(error.capturedOutput).toEqual(output);
expect(error.cleanupIncomplete).toBe(true);
await expect(scope.close()).rejects.toMatchObject({
reason: "unavailable",
});
mayClose = true;
await scope.close();
expect(error.cleanupIncomplete).toBe(false);
expect(error.cleanupPrimary).toBe(primary);
expect(error.capturedOutput).toEqual(output);
expect(projectAnalysisError(error)).toMatchObject(
projectAnalysisError(primary!),
);
} finally {
mayClose = true;
await scope.close();
await resources.close();
}
});
it("keeps aggregate cleanup live through partial recovery without reclosing healthy roots", async () => {
const resources = new ArtifactAnalysisResources(
"fixture",
"inspect_signal_decoders",
readStableArtifact,
readLinuxResourceLimits,
{ message: "busy", details: {} },
);
const denied = new Set(["first", "second"]);
const closes = new Map<string, number>();
for (const path of ["first", "healthy", "second"])
resources.retain({
path,
close: async () => {
closes.set(path, (closes.get(path) ?? 0) + 1);
if (denied.has(path)) throw new Error(`Close denied: ${path}`);
},
});
try {
const original: unknown = await resources
.release()
.catch((cause: unknown) => cause);
expect(original).toMatchObject({
cleanupIncomplete: true,
cleanupResources: ["first", "second"],
});
expect(closes.get("healthy")).toBe(1);
denied.delete("first");
const retry: unknown = await resources
.release()
.catch((cause: unknown) => cause);
expect(original).toMatchObject({
cleanupIncomplete: true,
cleanupResources: ["second"],
});
expect(retry).toMatchObject({
cleanupIncomplete: true,
cleanupResources: ["second"],
});
expect(resources.pendingPaths).toEqual(["second"]);
denied.clear();
await resources.close();
for (const error of [original, retry])
expect(error).toMatchObject({
cleanupIncomplete: false,
cleanupResources: [],
});
expect(closes).toEqual(
new Map([
["first", 2],
["healthy", 1],
["second", 3],
]),
);
} finally {
denied.clear();
await resources.close();
}
});