Files
gstack/test/brain-cache-roundtrip.test.ts
T
ac20ef1e13 v1.91.16.0 fix: safe installs and upgrades, skills start everywhere, truthful completion (parallel wave) (#3014)
* fix(make-pdf,design): real TOC page numbers and anchors; design models move to gpt-5.5 + gpt-image-2 (#2903, #2807)

* ci: pin every action to a SHA, Scorecard, checksummed Bun, scoped Codex CLI in the CI image, macOS named-regressions job (#1948, #1706, #1997)

* fix(cso,browse): Windows/macOS lease identity, retention skips bad runs, truthful 'not assessed' reports, env.kv false positive, Windows browse spawn fixes (#2894, #2912, #2048, #2637, #2805)

* fix(outside-voice): one Codex model precedence for probe and dispatch, bounded timeouts with KILL escalation, probe cache identity, guarded mktemp, skills hidden in nested Codex (#2776, #2914, #2787, #2881, #2847)

* feat(setup): install registry, ./setup --status, staged activation, per-scope migrations, host tiers and conformance kit, OpenCode/Hermes fixes, Copilot CLI (experimental), disabled_skills, slimmer CLAUDE.md (#1925, #2879, #2347, #2338, #2626, #2629, #2825, #2015, #393, #1206, #2096)

Co-authored-by: Andrey Esipov <61357035+andrey-esipov@users.noreply.github.com>
Co-authored-by: Matthew Williams <5863886+ridermw@users.noreply.github.com>
Co-authored-by: Rocky Wu <92764750+lolisaigao1234@users.noreply.github.com>

* fix(preamble): skills start in worktree-isolated Claude Code sessions — one literal skill-start line, context-recovery helper, no eval in skill blocks, $N token tripwire, honest plan-mode precedence (#2763, #2896, #2851)

* fix(setup): per-install renders for renamed, vendored and custom-config installs, space-safe alias roots, per-host /cso launcher path (#1882, #2906)

* test: free tests never write the real HOME (home-write tripwire), no orphaned browse daemons, symlinked-TMPDIR (macOS) fixes, gbrain-on-PATH isolation (#2895, #2816, #2829, #2841, #2982, #2911, #2909, #2908, #2910)

* chore: integrate wave follow-ups — regenerate all hosts, README model docs, TODOS, touchfiles, duration seeds, Bun 1.3 telemetry test fix, replay fixtures pinned to their recorded bytes

* v1.91.14.0: release metadata

* test(browse): group-kill cleanup falls back to the leader when macOS returns EPERM for a live group

* ci: Codex login steps run under bash (the CI container's default sh rejects pipefail)

* fix: full-suite repairs — private state roots for 7 real-HOME writers, truthful home-guard attribution with --attribute-home, gstack-paths shim in ship-hook fixtures, Copilot in team-init probes, test-of-test detector, cookie review fixture pinned to the approved excerpt

* fix(windows): LF for Dockerfiles, normalized home-guard exclusions, path.relative fixture containment, ship push test spawn timeouts

* test(design-detect): widen the whole-scan budget test's timing margins so a loaded macOS runner can't time out the single-batch call

* fix(evals): qa protocol names defined for the judge, async PTY judge so overlapping sessions keep their cleanup deadline, bws setup cases use the standard JUDGE_MS tier

* test(pty): the non-blocking judge case is POSIX-only (its fake judge is a shebang script, like the PTY harness it covers)

* test(evals): ship-coverage-audit names absolute skill paths, uses Read, and skips the Step 7 artifact nothing consumes (removes the mechanism behind 3 of 4 historical timeouts)

* test: deflake terminal-agent concurrent startup (write the placeholder record before spawning, like the product) and qa-evidence resumed pipe (barrier on a full high-water mark instead of a sleep)

* fix(design): daemon log stream handles open errors instead of crashing the daemon; shutdown test awaits the stubbed exit

* test(evals): skillify checks the product's resolved state root (no GSTACK_HOME bypass), negative cases check every root; tpa shims live outside the agent's cwd and answer consistently

* test(evals): coverage-audit detector splits parallel columns at │ rails and [+] groups; coverage prompts name absolute skill paths; TODOs from the gate eval census

* fix(plan-eng-review): scope gate always sends the menu (no headless 'Scope pending' exit); PTY harness ends idle turn-ends and API-error panels promptly with a precise diagnostic

* test(evals): one position-bound coverage-read rule credits read-only && chains (labels, context ranges, git displays in either order) and still rejects forged output

* fix(ship): documentation section has one saved source inserted by cat (never retyped); spawned-marker evidence is the child output; refreshed codex/factory ship goldens

* test(docsync): the byte-exact insert positive is POSIX-only (the bash-command observer refuses backslash paths by design)

* test(evals): plan-design-with-ui focus-title rule accepts the comma-less 'dimensions or focus' wording (replayed on the captured CI run, negative controls kept)

* fix(qa-callers): exploratory preflight says the scope/method Reads come after exploratory.md; design-review scan check credits an executed scan proven by one engine run over the changed files

---------

Co-authored-by: Andrey Esipov <61357035+andrey-esipov@users.noreply.github.com>
Co-authored-by: Matthew Williams <5863886+ridermw@users.noreply.github.com>
Co-authored-by: Rocky Wu <92764750+lolisaigao1234@users.noreply.github.com>
2026-10-03 07:54:09 -07:00

279 lines
13 KiB
TypeScript

/**
* brain-cache roundtrip integration tests (T2a / T19).
*
* Exercises the non-MCP-dependent parts of the cache layer:
* - Path resolution per scope (cross-project vs per-project)
* - Atomic _meta.json write/read
* - TTL staleness detection
* - Invalidate clears last_refresh
* - Schema-version mismatch triggers rebuild attempt (D4 A4)
* - Endpoint switch triggers rebuild attempt
*
* The brain-reachable refresh path (MCP fetch + compress) is tested
* separately in brain-cache-stale-but-usable.test.ts using a mocked
* spawnGbrain. T2a focuses on the cache-state machine.
*
* Uses tmp GSTACK_HOME per-test to avoid polluting the real ~/.gstack/.
* Gate-tier, free, ~50ms.
*/
import { describe, test, expect, beforeEach, afterEach } from 'bun:test';
import { mkdtempSync, existsSync, writeFileSync, readFileSync, rmSync, mkdirSync, readdirSync } from 'fs';
import { delimiter, join } from 'path';
import { tmpdir } from 'os';
import { spawnSync } from 'child_process';
let TMP_HOME: string;
let GBRAIN_STUB: string;
const ORIGINAL_HOME = process.env.GSTACK_HOME;
beforeEach(() => {
TMP_HOME = mkdtempSync(join(tmpdir(), 'gstack-cache-test-'));
GBRAIN_STUB = mkdtempSync(join(tmpdir(), 'gstack-unreachable-gbrain-'));
writeFileSync(join(GBRAIN_STUB, 'gbrain'), '#!/bin/sh\necho "gbrain unreachable (test stub)" >&2\nexit 1\n', { mode: 0o755 });
writeFileSync(join(GBRAIN_STUB, 'gbrain.cmd'), '@echo gbrain unreachable (test stub) 1>&2\r\n@exit /b 1\r\n');
const pathKey = Object.keys(process.env).find(name => name.toLowerCase() === 'path') ?? 'PATH';
process.env[pathKey] = `${GBRAIN_STUB}${delimiter}${process.env[pathKey] ?? ''}`;
process.env.GSTACK_HOME = TMP_HOME;
// Reload the cache module fresh per test so it picks up the new HOME.
delete require.cache[require.resolve('../bin/gstack-brain-cache')];
});
afterEach(() => {
if (ORIGINAL_HOME) process.env.GSTACK_HOME = ORIGINAL_HOME;
else delete process.env.GSTACK_HOME;
try { rmSync(TMP_HOME, { recursive: true, force: true }); } catch { /* best effort */ }
rmSync(GBRAIN_STUB, { recursive: true, force: true });
});
async function importCache(): Promise<typeof import('../bin/gstack-brain-cache')> {
return (await import('../bin/gstack-brain-cache')) as typeof import('../bin/gstack-brain-cache');
}
describe('brain-cache hermetic brain (#2829)', () => {
test('an installed gbrain is shadowed by the failing stub', () => {
const probe = spawnSync('gbrain', ['--version'], { encoding: 'utf8', shell: process.platform === 'win32', timeout: 10_000 });
expect(probe.status).toBe(1);
expect(probe.stderr).toContain('gbrain unreachable (test stub)');
});
});
describe('brain-cache paths', () => {
test('cross-project entity (user-profile) lives in ~/.gstack/brain-cache/', async () => {
const mod = await importCache();
const path = mod.entityPath('user-profile', null);
expect(path).toBe(join(TMP_HOME, 'brain-cache', 'user-profile.md'));
});
test('per-project entity (product) lives in ~/.gstack/projects/<slug>/brain-cache/', async () => {
const mod = await importCache();
const path = mod.entityPath('product', 'helsinki');
expect(path).toBe(join(TMP_HOME, 'projects', 'helsinki', 'brain-cache', 'product.md'));
});
test('throws on unknown entity', async () => {
const mod = await importCache();
expect(() => mod.entityPath('not-an-entity', null)).toThrow();
});
test('per-project entity without slug throws', async () => {
const mod = await importCache();
expect(() => mod.entityPath('product', null)).toThrow();
});
});
describe('brain-cache meta lifecycle', () => {
test('cmdMeta on empty cache returns valid fresh meta', async () => {
const mod = await importCache();
const meta = mod.cmdMeta('helsinki');
expect(meta.schema_version).toMatch(/^\d+\.\d+\.\d+$/);
expect(meta.endpoint_hash).toMatch(/^[a-f0-9]{1,8}$|^local$/);
expect(meta.last_refresh).toEqual({});
});
test('cmdInvalidate writes meta even if no prior refresh', async () => {
const mod = await importCache();
mod.cmdInvalidate('product', 'helsinki');
const meta = mod.cmdMeta('helsinki');
// last_refresh remains empty (we just delete an absent key — that's a no-op
// but the meta file is now written to disk).
expect(meta.last_refresh.product).toBeUndefined();
expect(existsSync(join(TMP_HOME, 'projects', 'helsinki', 'brain-cache', '_meta.json'))).toBe(true);
});
});
describe('brain-cache malformed _meta.json (#1879)', () => {
function seedMeta(content: string): void {
const cacheDir = join(TMP_HOME, 'projects', 'helsinki', 'brain-cache');
mkdirSync(cacheDir, { recursive: true });
writeFileSync(join(cacheDir, '_meta.json'), content);
}
test('cmdInvalidate does not throw when last_refresh is missing', async () => {
const mod = await importCache();
// Valid JSON object, but no last_refresh map — the original crash.
seedMeta(JSON.stringify({ schema_version: '0.0.1', endpoint_hash: 'x' }));
expect(() => mod.cmdInvalidate('product', 'helsinki')).not.toThrow();
});
test('cmdGet does not throw on null / array / primitive _meta.json', async () => {
const mod = await importCache();
for (const bad of ['null', '[]', '"a string"', '42']) {
seedMeta(bad);
expect(() => mod.cmdGet('product', 'helsinki')).not.toThrow();
}
});
test('missing schema_version is treated as a mismatch (forces rebuild, not trust)', async () => {
const mod = await importCache();
const cacheDir = join(TMP_HOME, 'projects', 'helsinki', 'brain-cache');
mkdirSync(cacheDir, { recursive: true });
writeFileSync(join(cacheDir, 'product.md'), '# stale-no-schema\n');
// No schema_version field — must NOT be trusted as a warm hit.
seedMeta(JSON.stringify({ endpoint_hash: mod.detectEndpointHash(), last_refresh: { product: Date.now() } }));
const result = mod.cmdGet('product', 'helsinki');
// Brain unreachable in test → rebuild path runs; must not be a trusted warm hit.
expect(['missing', 'cold-refreshed', 'stale-fallback']).toContain(result.state);
});
});
describe('brain-cache endpoint detection', () => {
test('detectEndpointHash returns "local" when no ~/.claude.json gbrain MCP', async () => {
// We don't write ~/.claude.json in the temp env, so this falls through to local.
const mod = await importCache();
// The user's real ~/.claude.json may have an MCP server; in that case the hash
// will be a real sha8. Either way, it's a stable string.
const hash = mod.detectEndpointHash();
expect(typeof hash).toBe('string');
expect(hash.length).toBeGreaterThan(0);
});
// #2499: project-scoped registrations (.projects["/path"].mcpServers.gbrain)
// were never read — two different project-scoped brains both hashed to
// 'local', so switching between them never invalidated the cache.
test('detectEndpointHash resolves a project-scoped gbrain URL for a cwd inside the project (#2499)', async () => {
const mod = await importCache();
const cj = join(TMP_HOME, 'claude.json');
writeFileSync(cj, JSON.stringify({
projects: {
'/w/repo': { mcpServers: { gbrain: { type: 'http', url: 'https://a.example/mcp' } } },
},
}));
const inside = mod.detectEndpointHash(cj, '/w/repo/src/deep');
expect(inside).not.toBe('local');
expect(inside).toHaveLength(8);
// Path-boundary check: /w/repo2 is NOT inside /w/repo.
expect(mod.detectEndpointHash(cj, '/w/repo2')).toBe('local');
});
test('detectEndpointHash prefers the nearest-ancestor project entry (#2499)', async () => {
const mod = await importCache();
const cj = join(TMP_HOME, 'claude.json');
writeFileSync(cj, JSON.stringify({
projects: {
'/w/repo': { mcpServers: { gbrain: { url: 'https://outer.example/mcp' } } },
'/w/repo/nested': { mcpServers: { gbrain: { url: 'https://inner.example/mcp' } } },
},
}));
const inner = mod.detectEndpointHash(cj, '/w/repo/nested/sub');
const outer = mod.detectEndpointHash(cj, '/w/repo/other');
expect(inner).not.toBe(outer); // two brains → two hashes (the docstring scenario)
expect(inner).not.toBe('local');
expect(outer).not.toBe('local');
});
test('detectEndpointHash prefers project-local scope over user scope (#2392 wave)', async () => {
// Empirically verified against claude 2.1.233 with hermetic fixtures:
// `claude mcp get gbrain` reports "Scope: Local config" when both scopes
// define the server — project-local WINS. The old pin here encoded the
// opposite (user-first) assumption, which mis-hashed endpoints whenever
// the two scopes disagreed.
const mod = await importCache();
const cj = join(TMP_HOME, 'claude.json');
writeFileSync(cj, JSON.stringify({
mcpServers: { gbrain: { url: 'https://user.example/mcp' } },
projects: {
'/w/repo': { mcpServers: { gbrain: { url: 'https://proj.example/mcp' } } },
},
}));
const conflictHash = mod.detectEndpointHash(cj, '/w/repo');
// Same file minus the USER entry → identical hash proves project scope won.
writeFileSync(cj, JSON.stringify({
projects: {
'/w/repo': { mcpServers: { gbrain: { url: 'https://proj.example/mcp' } } },
},
}));
expect(mod.detectEndpointHash(cj, '/w/repo')).toBe(conflictHash);
});
});
describe('brain-cache schema mismatch behavior', () => {
test('schema-version mismatch in meta triggers full-rebuild attempt on next get', async () => {
const mod = await importCache();
// Pre-seed meta with a different schema version, and a cache file that's
// recent enough to be "warm" by TTL but stale by schema version.
const cacheDir = join(TMP_HOME, 'projects', 'helsinki', 'brain-cache');
mkdirSync(cacheDir, { recursive: true });
writeFileSync(join(cacheDir, 'product.md'), '# stale-from-old-schema\n');
writeFileSync(join(cacheDir, '_meta.json'), JSON.stringify({
schema_version: '0.0.1',
endpoint_hash: mod.detectEndpointHash(),
last_refresh: { product: Date.now() },
last_attempt: {},
}));
const result = mod.cmdGet('product', 'helsinki');
// Brain is unreachable in this test (no gbrain mock), so refresh fails and
// the file gets deleted by the rebuild step. State should be 'missing' or
// 'stale-fallback' depending on whether the rebuild left a file behind.
expect(['missing', 'cold-refreshed', 'stale-fallback']).toContain(result.state);
}, 30000);
// ^ 30s: the schema-mismatch rebuild refreshes EVERY per-project entity,
// each spawning the real gbrain CLI (no mock here). With an unreachable
// brain each spawn runs to its own timeout, and under machine load the
// stack exceeds bun's 5s default — observed at 5.2-5.4s on a loaded box,
// identically on pre-fix binaries (load flake, not a code regression).
});
describe('brain-cache state machine', () => {
test('warm: pre-seeded fresh cache returns warm without touching brain', async () => {
const mod = await importCache();
const cacheDir = join(TMP_HOME, 'projects', 'helsinki', 'brain-cache');
mkdirSync(cacheDir, { recursive: true });
const productContent = '# Product: helsinki\n\nA test product.\n';
writeFileSync(join(cacheDir, 'product.md'), productContent);
writeFileSync(join(cacheDir, '_meta.json'), JSON.stringify({
schema_version: '1.0.0', // matches GSTACK_SCHEMA_PACK_VERSION
endpoint_hash: mod.detectEndpointHash(),
last_refresh: { product: Date.now() }, // fresh
last_attempt: {},
}));
const result = mod.cmdGet('product', 'helsinki');
expect(result.state).toBe('warm');
expect(readFileSync(result.path, 'utf-8')).toBe(productContent);
});
test('missing: no cache + no brain returns missing state', async () => {
const mod = await importCache();
const result = mod.cmdGet('brand', 'helsinki');
expect(result.state).toBe('missing');
});
test('stale-fallback: stale cache with unreachable brain returns stale-fallback', async () => {
const mod = await importCache();
const cacheDir = join(TMP_HOME, 'projects', 'helsinki', 'brain-cache');
mkdirSync(cacheDir, { recursive: true });
writeFileSync(join(cacheDir, 'product.md'), '# stale\n');
// Set last_refresh way in the past (> 1d TTL for product)
writeFileSync(join(cacheDir, '_meta.json'), JSON.stringify({
schema_version: '1.0.0',
endpoint_hash: mod.detectEndpointHash(),
last_refresh: { product: 0 }, // epoch start = very stale
last_attempt: {},
}));
const result = mod.cmdGet('product', 'helsinki');
// Brain unreachable → cold refresh fails → stale-but-usable fallback
expect(result.state).toBe('stale-fallback');
});
});