gsd-pi 2.76.0 → 2.77.0-dev.538325aea

package/src/resources/extensions/gsd/tests/exec-sandbox.test.ts

@@ -0,0 +1,210 @@
+import { test } from 'node:test';
+import assert from 'node:assert/strict';
+import { mkdtempSync, readFileSync, rmSync } from 'node:fs';
+import { tmpdir } from 'node:os';
+import { join } from 'node:path';
+
+import { EXEC_DEFAULTS, runExecSandbox, type ExecSandboxOptions } from '../exec-sandbox.ts';
+import { buildExecOptions, executeGsdExec } from '../tools/exec-tool.ts';
+import { isContextModeEnabled } from '../preferences-types.ts';
+
+function freshBase(): string {
+  return mkdtempSync(join(tmpdir(), 'gsd-exec-test-'));
+}
+
+function cleanup(dir: string): void {
+  rmSync(dir, { recursive: true, force: true });
+}
+
+function baseOpts(base: string, overrides: Partial<ExecSandboxOptions> = {}): ExecSandboxOptions {
+  return {
+    baseDir: base,
+    clamp_timeout_ms: EXEC_DEFAULTS.clampTimeoutMs,
+    default_timeout_ms: 10_000,
+    stdout_cap_bytes: 1_024,
+    stderr_cap_bytes: 1_024,
+    digest_chars: 120,
+    env_allowlist: EXEC_DEFAULTS.envAllowlist,
+    ...overrides,
+  };
+}
+
+test('runExecSandbox: captures stdout, persists artifacts, returns digest', async () => {
+  const base = freshBase();
+  try {
+    const result = await runExecSandbox(
+      { runtime: 'bash', script: 'echo hello world' },
+      baseOpts(base),
+    );
+    assert.equal(result.exit_code, 0);
+    assert.equal(result.timed_out, false);
+    assert.ok(result.digest.includes('hello world'), `digest should contain stdout: ${result.digest}`);
+    assert.ok(result.stdout_path.startsWith(join(base, '.gsd', 'exec')), 'stdout path under .gsd/exec');
+    assert.equal(readFileSync(result.stdout_path, 'utf-8').trim(), 'hello world');
+    const meta = JSON.parse(readFileSync(result.meta_path, 'utf-8')) as Record<string, unknown>;
+    assert.equal(meta.runtime, 'bash');
+    assert.equal(meta.exit_code, 0);
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('runExecSandbox: enforces stdout cap and marks truncation', async () => {
+  const base = freshBase();
+  try {
+    const result = await runExecSandbox(
+      // Emit far more than the cap so truncation triggers.
+      { runtime: 'bash', script: 'head -c 8000 /dev/urandom | base64' },
+      baseOpts(base, { stdout_cap_bytes: 256 }),
+    );
+    assert.equal(result.stdout_truncated, true, 'should mark stdout truncated');
+    assert.ok(result.stdout_bytes <= 256, `stdout_bytes within cap (got ${result.stdout_bytes})`);
+    const stdout = readFileSync(result.stdout_path, 'utf-8');
+    assert.ok(stdout.endsWith('[truncated: stdout cap reached]\n'), 'truncation marker appended');
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('runExecSandbox: enforces timeout and surfaces timed_out', async () => {
+  const base = freshBase();
+  try {
+    const started = Date.now();
+    const result = await runExecSandbox(
+      { runtime: 'bash', script: 'sleep 10' },
+      baseOpts(base, { default_timeout_ms: 150, clamp_timeout_ms: 150 }),
+    );
+    const elapsed = Date.now() - started;
+    assert.equal(result.timed_out, true);
+    assert.ok(elapsed < 5_000, `should return well before 10s (took ${elapsed}ms)`);
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('runExecSandbox: forwards only allowlisted env vars', async () => {
+  const base = freshBase();
+  try {
+    const result = await runExecSandbox(
+      { runtime: 'bash', script: 'echo PATH=$PATH SECRET=$GSD_TEST_SECRET' },
+      baseOpts(base, {
+        env_allowlist: [],
+        env: { PATH: '/usr/bin:/bin', HOME: '/tmp', GSD_TEST_SECRET: 'should-be-blocked' },
+      }),
+    );
+    const stdout = readFileSync(result.stdout_path, 'utf-8');
+    assert.ok(stdout.includes('PATH=/usr/bin:/bin'), 'PATH forwarded');
+    assert.ok(!stdout.includes('should-be-blocked'), 'non-allowlisted var blocked');
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('runExecSandbox: node runtime executes JS', async () => {
+  const base = freshBase();
+  try {
+    const result = await runExecSandbox(
+      { runtime: 'node', script: 'console.log("node-ok:" + (1+2))' },
+      baseOpts(base),
+    );
+    assert.equal(result.exit_code, 0);
+    assert.ok(result.digest.includes('node-ok:3'));
+  } finally {
+    cleanup(base);
+  }
+});
+
+// ── exec-tool executor ────────────────────────────────────────────────────
+
+test('executeGsdExec: runs by default when context_mode is unset', async () => {
+  const base = freshBase();
+  try {
+    const result = await executeGsdExec(
+      { runtime: 'bash', script: 'echo default-on-run' },
+      { baseDir: base, preferences: {} },
+    );
+    assert.ok(!result.isError, 'should succeed with no preferences');
+    assert.equal(result.details.operation, 'gsd_exec');
+    assert.equal(result.details.exit_code, 0);
+    assert.ok(result.content[0].text.includes('default-on-run'));
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('executeGsdExec: runs when preferences is null (fresh project)', async () => {
+  const base = freshBase();
+  try {
+    const result = await executeGsdExec(
+      { runtime: 'bash', script: 'echo null-prefs-run' },
+      { baseDir: base, preferences: null },
+    );
+    assert.ok(!result.isError, 'null preferences should not disable');
+    assert.ok(result.content[0].text.includes('null-prefs-run'));
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('executeGsdExec: blocked only when context_mode.enabled=false', async () => {
+  const base = freshBase();
+  try {
+    const result = await executeGsdExec(
+      { runtime: 'bash', script: 'echo should-not-run' },
+      { baseDir: base, preferences: { context_mode: { enabled: false } } },
+    );
+    assert.equal(result.isError, true);
+    assert.equal((result.details as { error?: string }).error, 'context_mode_disabled');
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('executeGsdExec: runs when enabled explicitly set to true', async () => {
+  const base = freshBase();
+  try {
+    const result = await executeGsdExec(
+      { runtime: 'bash', script: 'echo explicit-on' },
+      { baseDir: base, preferences: { context_mode: { enabled: true } } },
+    );
+    assert.ok(!result.isError);
+    assert.ok(result.content[0].text.includes('explicit-on'));
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('executeGsdExec: rejects empty script', async () => {
+  const base = freshBase();
+  try {
+    const result = await executeGsdExec(
+      { runtime: 'bash', script: '   ' },
+      { baseDir: base, preferences: { context_mode: { enabled: true } } },
+    );
+    assert.equal(result.isError, true);
+    assert.equal((result.details as { error?: string }).error, 'invalid_params');
+  } finally {
+    cleanup(base);
+  }
+});
+
+test('isContextModeEnabled: defaults to true; only explicit false disables', () => {
+  assert.equal(isContextModeEnabled(undefined), true, 'undefined prefs → on');
+  assert.equal(isContextModeEnabled(null), true, 'null prefs → on');
+  assert.equal(isContextModeEnabled({}), true, 'empty prefs → on');
+  assert.equal(isContextModeEnabled({ context_mode: {} }), true, 'empty block → on');
+  assert.equal(isContextModeEnabled({ context_mode: { enabled: true } }), true);
+  assert.equal(isContextModeEnabled({ context_mode: { enabled: false } }), false);
+});
+
+test('buildExecOptions: clamps out-of-range values to safe defaults', () => {
+  const opts = buildExecOptions('/tmp/base', {
+    enabled: true,
+    exec_timeout_ms: 999_999_999,
+    exec_stdout_cap_bytes: 1,
+    exec_digest_chars: -20,
+  });
+  assert.equal(opts.default_timeout_ms, EXEC_DEFAULTS.clampTimeoutMs, 'timeout clamped to upper bound');
+  assert.equal(opts.stdout_cap_bytes, 4_096, 'stdout cap clamped to floor');
+  assert.equal(opts.digest_chars, 0, 'digest chars clamped to floor');
+});

package/src/resources/extensions/gsd/tests/file-change-validator.test.ts

@@ -15,6 +15,64 @@ function git(cwd: string, ...args: string[]): string {
   }).trim();
 }
 
+test("validateFileChanges works on repos with a single commit (no HEAD~1)", (t) => {
+  const base = mkdtempSync(join(tmpdir(), "gsd-file-change-validator-"));
+  t.after(() => rmSync(base, { recursive: true, force: true }));
+
+  git(base, "init");
+  git(base, "config", "user.email", "test@example.com");
+  git(base, "config", "user.name", "Test User");
+
+  writeFileSync(join(base, "foo.ts"), "export const x = 1;\n");
+  git(base, "add", ".");
+  git(base, "commit", "-m", "initial");
+
+  // With only one commit, HEAD~1 doesn't exist — this must not throw
+  const audit = validateFileChanges(base, ["foo.ts"], []);
+
+  assert.ok(audit, "audit should be produced for single-commit repo");
+  assert.deepEqual(audit.unexpectedFiles, []);
+  assert.deepEqual(audit.missingFiles, []);
+});
+
+test("validateFileChanges excludes allowlisted files from unexpected-change warnings", (t) => {
+  const base = mkdtempSync(join(tmpdir(), "gsd-file-change-validator-"));
+  t.after(() => rmSync(base, { recursive: true, force: true }));
+
+  mkdirSync(join(base, "tracking", "history"), { recursive: true });
+  git(base, "init");
+  git(base, "config", "user.email", "test@example.com");
+  git(base, "config", "user.name", "Test User");
+
+  writeFileSync(join(base, "src.ts"), "initial\n");
+  writeFileSync(join(base, "tracking", "history", "2026-04-20-snapshot.md"), "initial\n");
+  git(base, "add", ".");
+  git(base, "commit", "-m", "initial");
+
+  writeFileSync(join(base, "src.ts"), "updated\n");
+  writeFileSync(join(base, "tracking", "history", "2026-04-20-snapshot.md"), "updated\n");
+  git(base, "add", ".");
+  git(base, "commit", "-m", "update");
+
+  // Without allowlist: tracking/history snapshot is unexpected
+  const auditWithout = validateFileChanges(base, ["src.ts"], []);
+  assert.ok(auditWithout, "audit should be produced");
+  assert.ok(
+    auditWithout.unexpectedFiles.includes("tracking/history/2026-04-20-snapshot.md"),
+    "snapshot should be unexpected without allowlist",
+  );
+
+  // With glob allowlist: snapshot is excluded
+  const auditWith = validateFileChanges(base, ["src.ts"], [], ["tracking/history/**"]);
+  assert.ok(auditWith, "audit should be produced with allowlist");
+  assert.deepEqual(auditWith.unexpectedFiles, [], "no unexpected files when snapshot is allowlisted");
+  assert.equal(
+    auditWith.violations.filter(v => v.severity === "warning").length,
+    0,
+    "no warnings when all unexpected files are allowlisted",
+  );
+});
+
 test("validateFileChanges ignores inline descriptions in expected output paths", (t) => {
   const base = mkdtempSync(join(tmpdir(), "gsd-file-change-validator-"));
   t.after(() => rmSync(base, { recursive: true, force: true }));

package/src/resources/extensions/gsd/tests/flat-rate-routing-guard.test.ts

@@ -39,19 +39,50 @@ describe("flat-rate provider routing guard (#3453)", () => {
   });
 
   test("resolvePreferredModelConfig returns undefined for copilot start model", () => {
+    const originalCwd = process.cwd();
+    const originalGsdHome = process.env.GSD_HOME;
+    const tempProject = mkdtempSync(join(tmpdir(), "gsd-flat-rate-project-"));
+    const tempGsdHome = mkdtempSync(join(tmpdir(), "gsd-flat-rate-home-"));
+
     // When the user's start model is on a flat-rate provider,
     // resolvePreferredModelConfig should not synthesize a routing
     // config from tier_models — it should return undefined so the
     // user's selected model is preserved.
-    const result = resolvePreferredModelConfig("execute-task", {
-      provider: "github-copilot",
-      id: "claude-sonnet-4",
-    });
-
-    // Should be undefined (no routing config created for flat-rate)
-    // Note: this only tests the guard — if explicit per-unit config exists
-    // in preferences, that takes precedence regardless.
-    assert.equal(result, undefined, "Should not create routing config for copilot");
+    try {
+      mkdirSync(join(tempProject, ".gsd"), { recursive: true });
+      writeFileSync(
+        join(tempProject, ".gsd", "PREFERENCES.md"),
+        [
+          "---",
+          "dynamic_routing:",
+          "  enabled: true",
+          "  tier_models:",
+          "    light: gpt-4o-mini",
+          "    standard: claude-sonnet-4-6",
+          "    heavy: claude-opus-4-6",
+          "---",
+        ].join("\n"),
+        "utf-8",
+      );
+      process.env.GSD_HOME = tempGsdHome;
+      process.chdir(tempProject);
+
+      const result = resolvePreferredModelConfig("execute-task", {
+        provider: "github-copilot",
+        id: "claude-sonnet-4",
+      });
+
+      // Should be undefined (no routing config created for flat-rate)
+      // Note: this only tests the synthesis guard — explicit per-unit config
+      // still takes precedence when the user configured one.
+      assert.equal(result, undefined, "Should not create routing config for copilot");
+    } finally {
+      process.chdir(originalCwd);
+      if (originalGsdHome === undefined) delete process.env.GSD_HOME;
+      else process.env.GSD_HOME = originalGsdHome;
+      rmSync(tempProject, { recursive: true, force: true });
+      rmSync(tempGsdHome, { recursive: true, force: true });
+    }
   });
 });
 

package/src/resources/extensions/gsd/tests/freeform-decisions.test.ts

@@ -6,6 +6,7 @@ import * as fs from 'node:fs';
 import {
   openDatabase,
   closeDatabase,
+  _getAdapter,
 } from '../gsd-db.ts';
 import {
   parseDecisionsTable,
@@ -30,6 +31,13 @@ function cleanupDir(dir: string): void {
   } catch { /* swallow */ }
 }
 
+/** Query all decisions from the DB ordered by seq. */
+function queryAllDecisions(): Array<Record<string, unknown>> {
+  const adapter = _getAdapter();
+  if (!adapter) return [];
+  return adapter.prepare('SELECT * FROM decisions ORDER BY seq').all();
+}
+
 // ═══════════════════════════════════════════════════════════════════════════
 // Bug reproduction: freeform DECISIONS.md content destroyed (#2301)
 // ═══════════════════════════════════════════════════════════════════════════
@@ -87,6 +95,13 @@ describe('freeform-decisions', () => {
 
       assert.deepStrictEqual(result.id, 'D001', 'decision ID assigned correctly');
 
+      // ── Assert DB state ──
+      const dbRows = queryAllDecisions();
+      assert.equal(dbRows.length, 1, 'DB has exactly 1 decision after first save');
+      assert.equal(dbRows[0]['id'], 'D001', 'DB row has correct ID');
+      assert.equal(dbRows[0]['scope'], 'testing', 'DB row has correct scope');
+      assert.equal(dbRows[0]['decision'], 'Use Jest for unit tests', 'DB row has correct decision text');
+
       // Read back the file
       const afterContent = fs.readFileSync(mdPath, 'utf-8');
 
@@ -125,6 +140,13 @@ describe('freeform-decisions', () => {
 
       assert.deepStrictEqual(result2.id, 'D002', 'second decision ID assigned correctly');
 
+      // ── Assert DB state after second save ──
+      const dbRows2 = queryAllDecisions();
+      assert.equal(dbRows2.length, 2, 'DB has exactly 2 decisions after second save');
+      assert.equal(dbRows2[0]['id'], 'D001', 'first DB row still D001');
+      assert.equal(dbRows2[1]['id'], 'D002', 'second DB row is D002');
+      assert.equal(dbRows2[1]['scope'], 'ci', 'second DB row has correct scope');
+
       const afterContent2 = fs.readFileSync(mdPath, 'utf-8');
 
       assert.ok(
@@ -182,6 +204,12 @@ describe('freeform-decisions', () => {
       // But the new decision should be there.
       assert.deepStrictEqual(result.id, 'D001', 'gets D001 since DB was empty');
 
+      // ── Assert DB state ──
+      const dbRows = queryAllDecisions();
+      assert.equal(dbRows.length, 1, 'DB has exactly 1 decision');
+      assert.equal(dbRows[0]['id'], 'D001', 'DB row has correct ID');
+      assert.equal(dbRows[0]['decision'], 'Use Vitest', 'DB row has correct decision text');
+
       const afterContent = fs.readFileSync(mdPath, 'utf-8');
       // Table-format file gets fully regenerated — this is the normal path
       assert.ok(
@@ -218,6 +246,13 @@ describe('freeform-decisions', () => {
       assert.deepStrictEqual(result.id, 'D001', 'first decision gets D001');
       assert.ok(fs.existsSync(mdPath), 'DECISIONS.md created');
 
+      // ── Assert DB state ──
+      const dbRows = queryAllDecisions();
+      assert.equal(dbRows.length, 1, 'DB has exactly 1 decision');
+      assert.equal(dbRows[0]['id'], 'D001', 'DB row ID is D001');
+      assert.equal(dbRows[0]['scope'], 'arch', 'DB row scope is arch');
+      assert.equal(dbRows[0]['decision'], 'Brand new decision', 'DB row decision text matches');
+
       const content = fs.readFileSync(mdPath, 'utf-8');
       assert.ok(content.includes('# Decisions Register'), 'new file has header');
       assert.ok(content.includes('Brand new decision'), 'new file has decision');
@@ -227,6 +262,33 @@ describe('freeform-decisions', () => {
     }
   });
 
+  test('parallel saveDecisionToDb calls assign unique IDs', async () => {
+    const tmpDir = makeTmpDir();
+    const dbPath = path.join(tmpDir, '.gsd', 'gsd.db');
+    openDatabase(dbPath);
+
+    try {
+      const [r1, r2, r3] = await Promise.all([
+        saveDecisionToDb({ scope: 'a', decision: 'Decision A', choice: 'A', rationale: 'A' }, tmpDir),
+        saveDecisionToDb({ scope: 'b', decision: 'Decision B', choice: 'B', rationale: 'B' }, tmpDir),
+        saveDecisionToDb({ scope: 'c', decision: 'Decision C', choice: 'C', rationale: 'C' }, tmpDir),
+      ]);
+
+      const ids = new Set([r1.id, r2.id, r3.id]);
+      assert.strictEqual(ids.size, 3, `expected 3 unique IDs but got: ${[r1.id, r2.id, r3.id]}`);
+
+      // Verify all 3 decisions exist in the markdown file
+      const mdPath = path.join(tmpDir, '.gsd', 'DECISIONS.md');
+      const content = fs.readFileSync(mdPath, 'utf-8');
+      assert.ok(content.includes('Decision A'), 'Decision A in file');
+      assert.ok(content.includes('Decision B'), 'Decision B in file');
+      assert.ok(content.includes('Decision C'), 'Decision C in file');
+    } finally {
+      closeDatabase();
+      cleanupDir(tmpDir);
+    }
+  });
+
   // ═══════════════════════════════════════════════════════════════════════════
 
 });