Files
rea/tests/boundary/mcp/reverseEngineeringPartialMcp.test.ts
morluto 60149282fc test: strengthen integration oracles and verification (#8295)
* test: strengthen integration oracles and verification

* test: bound lifecycle gates and strengthen workflow checks

* test: verify complete paged and extracted results

* test: preserve bytecode and cancellation evidence

* test: strengthen retained evidence and fixture cleanup

* test: require cross-projection bridge identities

* test: retain fixture owners through cleanup failures

* test: verify graph examples and project selectors

* test: tighten provider and lifecycle regressions

* test: consolidate hex coverage and bound cutter cleanup

* test: bound subprocess output waits and strengthen graph checks

* test: check mcp recovery and preserve cleanup owners

* test: verify structured mcp argument failures

* test: preserve artifact cleanup and gdb limit evidence

* test: preserve evidence identities and bound provider gates

* test: retain browser and comparison cleanup owners

* test: bound startup and strengthen fixture oracles

* test: bound lifecycle gates and verify recovered evidence

* test: verify public outcomes and lifecycle ownership

* test: assert precise evidence and release lifecycle gates

* test: verify selected facts and bound cleanup waits

* test: check retained evidence and decoded records

* test: verify final reports and retain cleanup ownership

* test: bound reviewed startup gates and retain cleanup

* test: verify PE errors and preserve MCP cleanup

* test: preserve MCP fixture cleanup ownership

* test: release provider gates before late startup
2026-11-01 00:15:28 +08:00

572 lines
19 KiB
TypeScript

import pino from "pino";
import type { CallToolResult } from "@modelcontextprotocol/server";
import { parseMcpToolError } from "../../fixtures/mcpToolError.js";
import { mkdtemp, rm, symlink, writeFile } from "node:fs/promises";
import { tmpdir } from "node:os";
import { dirname, join } from "node:path";
import { expect, it, vi } from "vitest";
import { TestReverseEngineeringService as ReverseEngineeringService } from "../../support/reverseEngineeringService.js";
import {
AnalysisCancelledError,
AnalysisInputError,
} from "../../../src/domain/analysisErrorCore.js";
import { projectAnalysisError } from "../../../src/domain/analysisErrorProjection.js";
import type { FridaDeviceList } from "../../../src/domain/fridaDiscovery.js";
import {
createBinarySessionTargets,
createCacheProvider,
createTestBinarySession,
} from "../../fixtures/binarySession.js";
import type { AnalysisOperationPort } from "../../../src/application/AnalysisProvider.js";
import { EvidenceIntegrityError } from "../../../src/domain/evidenceErrors.js";
import { err, ok } from "../../../src/domain/result.js";
import type { WithAdmittedAnalysis } from "../../../src/server/analysisAdmission.js";
import { createToolResultDelivery } from "../../../src/server/toolResult.js";
import { registerReverseEngineeringTools } from "../../../src/server/registerReverseEngineeringTools.js";
import {
createReverseEngineeringMcpHarness,
invokeWithLiveSignal,
} from "../../support/reverseEngineeringMcpHarness.js";
const createTimedOutRizinService = (): ReverseEngineeringService =>
new ReverseEngineeringService({
run: async () => {
throw Object.assign(new Error("timed out"), {
stdout: "partial provider output",
stderr: "",
code: "ETIMEDOUT",
signal: null,
killed: true,
});
},
});
it("keeps a non-Evidence partial with a caller-selected evidence_id inline in a reverse tool error", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const partial = {
devices: [],
cleanup_error: null,
evidence_id: "caller-selected",
};
const observation: FridaDeviceList = partial;
const reason = { request: "cancel inspection" };
const failure = new AnalysisCancelledError("inspect_with_objdump", {
cause: reason,
partialObservation: observation,
});
const service = new ReverseEngineeringService();
vi.spyOn(service, "inspectWithObjdump").mockResolvedValue(err(failure));
const recordEvidence = vi.fn(() => ok("added" as const));
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
recordEvidence,
);
const result = (await invokeWithLiveSignal(handler("inspect_with_objdump"), {
path: "package.json",
operation: "file_headers",
})) as CallToolResult;
const projected = parseMcpToolError(result);
expect(projected.error).toMatchObject({
code: "cancelled",
message: projectAnalysisError(failure).message,
details: { operation: "inspect_with_objdump" },
});
expect(projected.error.details?.partial_observation).toEqual(partial);
expect(result.structuredContent).toEqual(projected);
expect(recordEvidence).not.toHaveBeenCalled();
expect(failure.cause).toBe(reason);
expect(failure.partialObservation).toBe(partial);
expect(result._meta).toEqual({
provenance: { untrusted_content: true, source: "analysis_target" },
});
});
it("records partial Evidence and references it in the original MCP error", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const service = createTimedOutRizinService();
const recordEvidence = vi.fn(() => ok("added" as const));
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
recordEvidence,
);
const invoke = handler("execute_rizin_command");
const response = (await invokeWithLiveSignal(invoke, {
path: "package.json",
command: "iI",
})) as {
readonly isError?: boolean;
readonly structuredContent?: {
readonly error?: {
readonly details?: Readonly<Record<string, unknown>>;
};
};
};
expect(response.isError).toBe(true);
expect(recordEvidence).toHaveBeenCalledTimes(1);
expect(response.structuredContent?.error?.details).toMatchObject({
partial_observation: {
kind: "retained-evidence",
evidence_id: expect.any(String),
},
});
});
it("admits Rizin mutation after invalidating before dispatch", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const events: string[] = [];
const service = new ReverseEngineeringService({
run: async () => {
events.push("dispatch");
return { stdout: "rizin output", stderr: "" };
},
});
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
undefined,
undefined,
() => events.push("invalidate"),
async (_path, _signal, operation) => {
events.push("admit");
return ok(await operation());
},
);
const invoke = handler("execute_rizin_command");
await invoke(
{ path: "package.json", command: "iI" },
{ mcpReq: { signal: new AbortController().signal } },
);
expect(events).toEqual(["admit", "invalidate", "dispatch"]);
});
it("invalidates the snapshot in the legacy Rizin admission path", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const invalidateSnapshot = vi.fn();
const service = new ReverseEngineeringService({
run: async () => ({ stdout: "rizin output", stderr: "" }),
});
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
undefined,
undefined,
invalidateSnapshot,
);
const invoke = handler("execute_rizin_command");
const response = (await invokeWithLiveSignal(invoke, {
path: "package.json",
command: "iI",
})) as { readonly isError?: boolean };
expect(response.isError).not.toBe(true);
expect(invalidateSnapshot).toHaveBeenCalledOnce();
});
it("skips snapshot invalidation for malformed Rizin input", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const invalidateSnapshot = vi.fn();
const executeRizinCommandValidated = vi.fn();
const service = {
executeRizinCommandValidated,
} as unknown as ReverseEngineeringService;
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
undefined,
undefined,
invalidateSnapshot,
);
const invoke = handler("execute_rizin_command");
const response = (await invokeWithLiveSignal(invoke, {
path: "package.json",
})) as CallToolResult;
expect(parseMcpToolError(response).error).toMatchObject({
code: "invalid_request",
details: {
issues: [
{ path: ["command"], reason: "missing_argument", expected: "string" },
],
},
});
expect(executeRizinCommandValidated).not.toHaveBeenCalled();
expect(invalidateSnapshot).not.toHaveBeenCalled();
});
it("delivers failed Rizin execution after external admission", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const service = new ReverseEngineeringService({
run: async () => {
throw new Error("Rizin command failed");
},
});
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
undefined,
undefined,
undefined,
async (_path, _signal, operation) => ok(await operation()),
);
const invoke = handler("execute_rizin_command");
const response = (await invokeWithLiveSignal(invoke, {
path: "package.json",
command: "iI",
})) as { readonly isError?: boolean };
expect(response.isError).toBe(true);
});
it("preserves the Rizin operation error when analysis retirement also fails", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const reason = { stage: "operation-input" };
const primary = new AnalysisInputError("execute_rizin_command", {
cause: reason,
});
const retirement = new AnalysisInputError("rebind");
const service = {
retryCleanup: async () => ok(null),
executeRizinCommandValidated: async () => err(primary),
} as unknown as ReverseEngineeringService;
const observedOperationResult = vi.fn();
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
undefined,
undefined,
undefined,
async (_path, _signal, operation, rebindFailure) => {
const result = await operation();
observedOperationResult(result);
return err(rebindFailure(result, retirement));
},
);
const invoke = handler("execute_rizin_command");
const response = (await invoke(
{ path: "package.json", command: "iI" },
{ mcpReq: { signal: new AbortController().signal } },
)) as {
readonly isError?: boolean;
readonly structuredContent?: {
readonly error?: {
readonly code?: string;
readonly details?: { readonly reason?: string };
};
};
};
expect(primary.cause).toBe(reason);
expect(primary.secondaryFailures).toEqual([retirement]);
expect(response.isError).toBe(true);
expect(observedOperationResult).toHaveBeenCalledWith(
expect.objectContaining({ ok: false }),
);
expect(response.structuredContent?.error).toMatchObject({
code: "invalid_request",
});
});
it.each(["success", "timeout"])(
"retains %s output and requires reopening after a symlink command",
async (outcome) => {
const [target] = await createBinarySessionTargets();
const alias = join(dirname(target), "active-target-alias.hop");
await symlink(target, alias);
const session = createTestBinarySession(createCacheProvider([]));
expect((await session.open(target)).ok).toBe(true);
expect(session.activeTarget()?.path).toBe(target);
const originalDigest = session.activeTarget()?.sha256;
expect(originalDigest).toEqual(expect.any(String));
const { server, handler } = createReverseEngineeringMcpHarness();
const service = new ReverseEngineeringService({
run: async () => {
await writeFile(alias, "changed by Rizin");
if (outcome === "timeout")
throw Object.assign(new Error("timed out"), {
stdout: "partial analysis output",
stderr: "timeout diagnostics",
code: "ETIMEDOUT",
signal: null,
killed: true,
});
return { stdout: "partial analysis output", stderr: "" };
},
});
const recordEvidence = vi.fn(() => ok("added" as const));
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
recordEvidence,
undefined,
() => session.invalidateSnapshotForExternalMutation(),
(path, signal, operation, rebindFailure) =>
session.withAdmittedExternalMutation(
path,
signal,
operation,
rebindFailure,
),
);
const invoke = handler("execute_rizin_command");
const response = (await invoke(
{ path: alias, command: "wx 00 @ 0" },
{ mcpReq: { signal: new AbortController().signal } },
)) as {
readonly isError?: boolean;
readonly structuredContent?: {
readonly error?: {
readonly code?: string;
readonly details?: Readonly<Record<string, unknown>>;
};
};
};
expect(response.isError ?? false).toBe(outcome === "timeout");
if (outcome === "timeout") {
expect(response.structuredContent?.error?.code).toBe("provider_timeout");
expect(response.structuredContent?.error?.details).toMatchObject({
captured_output: {
stdout: "partial analysis output",
stderr: "timeout diagnostics",
},
});
expect(response.structuredContent?.error?.details).toMatchObject({
partial_observation: {
kind: "retained-evidence",
evidence_id: expect.any(String),
},
});
}
expect(recordEvidence).toHaveBeenCalledTimes(1);
expect(session.activeTarget()).toBeUndefined();
expect((await session.execute("address_name", {})).ok).toBe(false);
expect((await session.open(target)).ok).toBe(true);
expect(session.activeTarget()?.sha256).toEqual(expect.any(String));
expect(session.activeTarget()?.sha256).not.toBe(originalDigest);
await session.close();
await rm(dirname(target), { recursive: true, force: true });
},
);
it("preserves a partial MCP error without claiming failed Evidence retention", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const service = createTimedOutRizinService();
const recordEvidence = vi.fn((_evidence: unknown) =>
err(new EvidenceIntegrityError("retention failed")),
);
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
recordEvidence,
);
const invoke = handler("execute_rizin_command");
const response = (await invokeWithLiveSignal(invoke, {
path: "package.json",
command: "iI",
})) as {
readonly isError?: boolean;
readonly structuredContent?: {
readonly error?: {
readonly details?: Readonly<Record<string, unknown>>;
};
};
};
expect(response.isError).toBe(true);
expect(response.structuredContent?.error?.details).toMatchObject({
partial_observation: expect.objectContaining({
operation: "execute_rizin_command",
}),
});
expect(
JSON.stringify(response.structuredContent?.error?.details),
).not.toContain("retained-evidence");
});
it("delivers pathless objdump capability Evidence and input errors through its named tool", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const service = new ReverseEngineeringService({
run: async () => ({ stdout: "objdump output", stderr: "" }),
});
const delivery = createToolResultDelivery(2 * 1024 * 1024);
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
delivery,
);
const context = { mcpReq: { signal: new AbortController().signal } };
const objdump = handler("inspect_with_objdump");
const rizin = handler("execute_rizin_command");
const success = (await objdump(
{ path: "package.json", operation: "file_headers" },
context,
)) as { readonly isError?: boolean };
const architectures = (await objdump(
{ operation: "architectures" },
context,
)) as {
readonly isError?: boolean;
};
const objdumpFailure = (await objdump({}, context)) as {
readonly isError?: boolean;
};
const failure = (await rizin({}, context)) as {
readonly isError?: boolean;
readonly structuredContent?: {
readonly error?: { readonly code?: string };
};
};
const cancelled = new AbortController();
cancelled.abort();
const cancelledContext = { mcpReq: { signal: cancelled.signal } };
const cancelledObjdump = (await objdump(
{ path: "package.json", operation: "file_headers" },
cancelledContext,
)) as { readonly isError?: boolean };
const cancelledRizin = (await rizin(
{ path: "package.json", command: "iI" },
cancelledContext,
)) as { readonly isError?: boolean };
expect(success.isError).not.toBe(true);
expect(architectures.isError).not.toBe(true);
expect(objdumpFailure.isError).toBe(true);
expect(failure.isError).toBe(true);
expect(cancelledObjdump.isError).toBe(true);
expect(cancelledRizin.isError).toBe(true);
});
it("does not treat a malformed partial observation as retained Evidence", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const invalidError = Object.assign(
new AnalysisInputError("execute_rizin_command"),
{ partialObservation: { evidence_id: 42 } },
);
const service = {
inspectWithObjdump: async () => err(invalidError),
executeRizinCommandValidated: async () => err(invalidError),
} as unknown as ReverseEngineeringService;
const recordEvidence = vi.fn(() => ok("added" as const));
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
recordEvidence,
);
const invoke = handler("execute_rizin_command");
const response = (await invoke(
{ path: "package.json", command: "iI" },
{ mcpReq: { signal: new AbortController().signal } },
)) as { readonly isError?: boolean };
expect(response.isError).toBe(true);
expect(recordEvidence).not.toHaveBeenCalled();
});
it("keeps provider execution and Evidence recording inside analysis admission", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const events: string[] = [];
const service = new ReverseEngineeringService({
run: async () => {
events.push("provider");
return { stdout: "objdump output", stderr: "" };
},
});
const recordEvidence = vi.fn(() => {
events.push("record");
return ok("added" as const);
});
const admission: WithAdmittedAnalysis = async (_name, _signal, operation) => {
events.push("admission-start");
const result = await operation({} as AnalysisOperationPort);
events.push("admission-end");
return ok(result);
};
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
recordEvidence,
admission,
);
const invoke = handler("inspect_with_objdump");
await invoke(
{ path: "package.json", operation: "file_headers" },
{ mcpReq: { signal: new AbortController().signal } },
);
expect(events).toEqual([
"admission-start",
"provider",
"record",
"admission-end",
]);
});
it("returns admission failures without invoking a reverse-engineering provider", async () => {
const { server, handler } = createReverseEngineeringMcpHarness();
const inspectWithObjdump = vi.fn();
const service = {
inspectWithObjdump,
executeRizinCommandValidated: vi.fn(),
} as unknown as ReverseEngineeringService;
const admission: WithAdmittedAnalysis = async () =>
err(new AnalysisInputError("inspect_with_objdump"));
registerReverseEngineeringTools(
server,
service,
pino({ enabled: false }),
createToolResultDelivery(2 * 1024 * 1024),
undefined,
admission,
);
const invoke = handler("inspect_with_objdump");
const response = (await invoke(
{ path: "package.json", operation: "file_headers" },
{ mcpReq: { signal: new AbortController().signal } },
)) as { readonly isError?: boolean };
expect(response.isError).toBe(true);
expect(inspectWithObjdump).not.toHaveBeenCalled();
});