Files
reverse-engineering-anything/tests/boundary/cli/cliToolCoverage.test.ts
T
morlutoandCursor Agent b81da90354 feat(native): page retained procedure references (#8588)
* feat(native): page retained feature traces

* fix(snapshot): retain linked source evidence for replay

* test(snapshot): verify linked foreign evidence stays scoped

* feat(native): page retained procedure references

* docs(mcp): clarify imported residual questions

* docs(mcp): guide trace reference classification

* fix(cli): report result sources when Evidence publication fails

A trace keeps linked sources on the result. Use those records for the
undelivered source closure as well as the bundle, so a refused output
does not report them as not captured.

* ci: rerun pull request checks

The merge of origin/main did not start the CI workflow for this head.

* fix(native): document shared procedure endpoints and the required selector

Describe kind shared and items.endpoints instead of inline procedure copies.
State that procedure-references requires a positional selector or --procedure.

* fix(native): count procedure-reference endpoints before keeping a row

Include the shared endpoint table in the page budget, with JSON punctuation,
and store a new procedure only when its row is accepted.

* fix(native): measure a procedure endpoint once before its comparison key

Size the candidate before building one canonical identity, and reuse that
key for the selected-procedure and shared-endpoint checks.

* fix(native): reject an oversized procedure endpoint before canonicalizing it

When the JSON size already exceeds every stored identity, skip the canonical
string and keep the row out of the page.

* fix(native): describe procedure-reference direction and coverage

Avoid execution wording in the tool description, and give the CLI direction
option the help text the catalog requires.

* fix(native): compare endpoint size with UTF-8 identity length

Keep one running maximum of stored identity bytes. A multibyte selected
procedure is recognized when its JSON no longer fits the page.

* test(cli): count procedure-references as a dedicated command

The MCP tool has a CLI command, so it is no longer an MCP-only exception.

* test(native): catch the string-length endpoint comparison

Use a 200,000-character multibyte selected procedure and a shorter ASCII
endpoint so the old JavaScript length check drops the selected row.

* ci: rerun after the binary-layout toolchain timeout

The required check failed because the toolchain install exceeded 15 minutes.
The same job completed on the previous head, and this commit does not change
the procedure-reference behavior.

* test: stabilize capture decoder exhaustion coverage

---------

Co-authored-by: Cursor Agent <cursoragent@cursor.com>
2026-11-08 02:16:24 +00:00

143 lines
4.6 KiB
TypeScript

import { describe, expect, it } from "vitest";
import { createCli } from "../../../src/cli.js";
import { createCliInventory } from "../../../scripts/lib/catalog-cli.mjs";
import { TOOL_CONTRACTS } from "../../../src/contracts/toolContracts.js";
import { CLI_COMMAND_NAMES } from "../../../src/cliCommandNames.js";
/** Canonical MCP operations reached by CLI commands with different names. */
const CLI_COMMAND_TOOL_ALIASES = Object.freeze({
"environment-build": ["export_environment_bundle"],
guidance: ["search_guidance", "read_guidance_section"],
analyze: ["binary_overview"],
inspect: ["binary_overview"],
decompile: ["procedure_pseudo_code"],
function: ["analyze_function"],
instructions: ["read_function_instructions"],
trace: ["trace_feature"],
search: ["search_strings", "search_procedures"],
"capture-process": ["capture_process_scenario"],
compare: ["compare_bundles"],
"list-objdump-architectures": ["inspect_with_objdump"],
"debug-with-gdb": ["gdb_console"],
"debug-with-rizin": ["rizin_debug_command"],
});
/** MCP operations without a dedicated CLI command or direct command alias. */
const MCP_TOOLS_WITHOUT_DEDICATED_CLI = Object.freeze([
// Runtime settings belong to the long-lived MCP server; CLI calls use launch/per-call controls.
"inspect_analysis_settings",
"update_analysis_settings",
// MCP session control selects a resource root; CLI resource commands take paths directly.
"open_bundle_resources",
"address_name",
"comment",
"current_address",
"current_procedure",
"current_document",
"goto_address",
"inline_comment",
"list_bookmarks",
"list_documents",
"list_names",
"list_procedures",
"list_segments",
"list_strings",
"next_address",
"prev_address",
"procedure_address",
"procedure_assembly",
"procedure_callees",
"procedure_callers",
"procedure_info",
"resolve_containing_procedure",
"set_address_name",
"set_addresses_names",
"set_bookmark",
"set_comment",
"set_inline_comment",
"unset_bookmark",
"get_objc_classes",
"get_objc_protocols",
"batch_decompile",
"get_call_graph",
"analyze_swift_types",
"find_xrefs_to_name",
"trace_call_path",
"open_binary",
"close_binary",
// Investigation reset acts on the long-lived MCP server's retained history.
"reset_investigation",
"binary_session",
"export_evidence_bundle",
"import_evidence_bundle",
"compare_functions",
"find_changed_behavior",
"build_call_path",
"correlate_static_and_runtime",
"verify_reconstruction",
"list_unknowns",
"record_unknown",
"update_unknown",
"verify_unknown_resolution",
"get_evidence_bundle",
"get_navigation_context",
"inspect_address_context",
"start_gdb_session",
"gdb_session_status",
"close_gdb_session",
"start_rizin_debug_session",
"rizin_debug_session_status",
"close_rizin_debug_session",
"start_frida_session",
"load_frida_script",
"resume_frida_session",
"unload_frida_script",
"frida_session_status",
"close_frida_session",
// Build computers are held by the long-lived MCP server between actions.
"start_build_computer",
"act_on_build_computer",
"close_build_computer",
"start_build_capture",
"mark_build_checkpoint",
"stop_build_capture",
"export_build_environment",
"check_build_task",
]);
describe("CLI canonical tool coverage", () => {
it("accounts for every MCP tool on the CLI or marks it intentionally MCP-only", () => {
const registeredCli = createCliInventory(createCli());
const registeredCommandNames = new Set([
...registeredCli.primary,
...registeredCli.aliases.map(({ name }) => name),
]);
const cliToolNames = new Set(
registeredCli.primary.map((command) => command.replaceAll("-", "_")),
);
for (const [command, names] of Object.entries(CLI_COMMAND_TOOL_ALIASES)) {
expect(registeredCommandNames.has(command)).toBe(true);
for (const name of names) cliToolNames.add(name);
}
for (const command of CLI_COMMAND_NAMES)
expect(registeredCommandNames.has(command)).toBe(true);
const intentionalMcpOnly = new Set<string>(MCP_TOOLS_WITHOUT_DEDICATED_CLI);
expect(
[...intentionalMcpOnly].filter((name) => cliToolNames.has(name)).sort(),
).toEqual([]);
const unaccounted = TOOL_CONTRACTS.map(({ name }) => name).filter(
(name) => !cliToolNames.has(name) && !intentionalMcpOnly.has(name),
);
expect(unaccounted).toEqual([]);
expect(new Set(MCP_TOOLS_WITHOUT_DEDICATED_CLI).size).toBe(
MCP_TOOLS_WITHOUT_DEDICATED_CLI.length,
);
expect(
MCP_TOOLS_WITHOUT_DEDICATED_CLI.every((name) =>
TOOL_CONTRACTS.some((contract) => contract.name === name),
),
).toBe(true);
});
});