start-command 0.42.0 → 0.44.0

package/test/fixtures/pause-golden.cjs

@@ -0,0 +1,142 @@
+/**
+ * Issue #263: runs one `pause-control` shared golden case
+ * (parity/fixtures/contracts.json) through the JavaScript `--pause` /
+ * `--unpause` control path. rust/tests/shared_golden.rs runs the same cases
+ * through the Rust path, so both runtimes are held to identical reports,
+ * errors, docker argv and persisted records.
+ *
+ * The in-memory docker follows the rules spelled out in
+ * `pause-control.docker` of the fixture file; keep it in step with
+ * `GoldenDocker` in rust/tests/shared_golden.rs.
+ */
+
+const TIMESTAMP =
+  /\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}(?:\.\d+)?(?:Z|[+-]\d{2}:\d{2})/g;
+
+/** Replace every ISO-8601 timestamp, which differs per run, with a marker. */
+function normalize(value) {
+  if (typeof value === 'string') {
+    return value.replace(TIMESTAMP, '<timestamp>');
+  }
+  if (Array.isArray(value)) {
+    return value.map(normalize);
+  }
+  if (value && typeof value === 'object') {
+    return Object.fromEntries(
+      Object.entries(value).map(([key, entry]) => [key, normalize(entry)])
+    );
+  }
+  return value;
+}
+
+function goldenDocker(initial, name) {
+  const docker = { state: initial, calls: [] };
+  const ok = (stdout = '') => ({
+    success: true,
+    stdout,
+    stderr: '',
+    status: 0,
+    error: null,
+  });
+  const fail = (stderr) => ({
+    success: false,
+    stdout: '',
+    stderr,
+    status: 1,
+    error: null,
+  });
+  docker.runner = (command, args, extra = {}) => {
+    docker.calls.push({ args: [...args], bounded: Boolean(extra.timeout) });
+    const verb = args[0];
+    if (docker.state === 'removed' && verb !== 'rm') {
+      return fail(`Error response from daemon: No such container: ${name}`);
+    }
+    const alive = docker.state === 'running' || docker.state === 'paused';
+    switch (verb) {
+      case 'inspect': {
+        const format = args[args.indexOf('-f') + 1];
+        if (format === '{{.State.Status}} {{.State.Paused}}') {
+          return ok(`${docker.state} ${docker.state === 'paused'}\n`);
+        }
+        if (format === '{{.State.Status}}') {
+          return ok(`${docker.state}\n`);
+        }
+        if (format === '{{.Id}} {{.State.Pid}}') {
+          return ok(`abc263def ${alive ? 4242 : 0}\n`);
+        }
+        return ok('');
+      }
+      case 'pause':
+        if (docker.state !== 'running') {
+          return fail(`container ${name} is not running`);
+        }
+        docker.state = 'paused';
+        return ok(`${name}\n`);
+      case 'unpause':
+        if (docker.state !== 'paused') {
+          return fail(`container ${name} is not paused`);
+        }
+        docker.state = 'running';
+        return ok(`${name}\n`);
+      default:
+        return ok('');
+    }
+  };
+  return docker;
+}
+
+/**
+ * Run one case and return its normalized observable result.
+ * @param {{ExecutionRecord: Function, pause: object}} modules - Code under test
+ * @param {object} store - An empty ExecutionStore
+ * @param {object} section - `pause-control` from the contracts file
+ * @param {object} fixture - One of `section.cases`
+ * @returns {object} The value `fixture.expected` must equal
+ */
+function runPauseCase({ ExecutionRecord, pause }, store, section, fixture) {
+  const base = section.record;
+  store.save(
+    ExecutionRecord.fromObject({
+      ...base,
+      status: fixture.status || base.status,
+      exitCode:
+        fixture.exitCode === undefined ? base.exitCode : fixture.exitCode,
+      options: { ...base.options, ...(fixture.options || {}) },
+    })
+  );
+  const docker = goldenDocker(fixture.docker, base.options.sessionName);
+  const deps = {
+    runner: docker.runner,
+    outputFormat: fixture.outputFormat,
+    reason: fixture.reason,
+    by: fixture.by,
+  };
+  const result =
+    fixture.action === 'pause'
+      ? pause.pauseExecution(store, fixture.identifier, deps)
+      : pause.unpauseExecution(store, fixture.identifier, deps);
+  const saved = store.get(base.uuid).toObject();
+  let output = result.output === undefined ? null : result.output;
+  if (output !== null && fixture.outputFormat === 'json') {
+    output = JSON.parse(output);
+  }
+  return normalize({
+    success: result.success,
+    output,
+    error: result.error === undefined ? null : result.error,
+    dockerCalls: docker.calls.map((call) => call.args),
+    unboundedDockerCalls: docker.calls
+      .filter((call) => !call.bounded)
+      .map((call) => call.args),
+    dockerState: docker.state,
+    record: {
+      status: saved.status,
+      exitCode: saved.exitCode === undefined ? null : saved.exitCode,
+      containerState:
+        saved.containerState === undefined ? null : saved.containerState,
+      options: saved.options,
+    },
+  });
+}
+
+module.exports = { normalize, goldenDocker, runPauseCase };

package/test/regression-243.js

@@ -82,6 +82,9 @@ const CONTAINERS = [
   '.State.OOMKilled': String(oom),
   '.State.StartedAt': startedAt || '0001-01-01T00:00:00Z',
   '.State.FinishedAt': finishedAt || '0001-01-01T00:00:00Z',
+  // Paused state for --status/--list (issue #263).
+  '.State.Status': running ? 'running' : 'exited',
+  '.State.Paused': 'false',
 }));
 
 /**
@@ -290,7 +293,7 @@ posix('issue #243: one docker inspect per query, same output', () => {
 });
 
 describe('issue #243: batched inspect answers', () => {
-  const row = (name, id) => `/${name} ${id} 7 true 0 false s f`;
+  const row = (name, id) => `/${name} ${id} 7 true 0 false s f running false`;
 
   it('matches lines by name or ID prefix, not by order', () => {
     const answers = parseBatchInspect(

package/test/regression-258-files.js

@@ -11,6 +11,7 @@ const {
   EXIT,
 } = require('../src/lib/formal-ai-policy');
 const files = require('../src/lib/formal-ai-files');
+const { readFrame, readFrameText } = files;
 const { runFormalAiTask } = require('../src/lib/formal-ai');
 const {
   cleanup,
@@ -144,11 +145,14 @@ describe('read (#258)', () => {
     const cwd = workspace();
     const ctx = context(cwd, 'read-only');
     const receipt = call('read', { path: './a.js' }, ctx);
+    // Formal AI binds a read as complete only in its numbered frame (#261).
     expect(receipt).toMatchObject({
       exitCode: 0,
-      result: TEXT,
+      text: TEXT,
+      result: readFrame(TEXT),
       bytes: Buffer.byteLength(TEXT),
     });
+    expect(readFrameText(receipt.result)).toBe(TEXT);
     expect(ctx.observed.has('a.js')).toBe(true);
     expect(ctx.contents.get('a.js')).toBe(receipt.sha256);
   });
@@ -198,10 +202,16 @@ describe('edit (#258)', () => {
     );
     expect(receipt).toMatchObject({
       exitCode: 0,
-      result: '',
       replacements: 1,
       verification: { verified: true },
     });
+    expect(JSON.parse(receipt.result)).toMatchObject({
+      ok: true,
+      path: 'a.js',
+      tool: 'edit',
+      replacements: 1,
+      verified: true,
+    });
     expect(fs.readFileSync(file, 'utf8')).toBe(
       TEXT.replace('const b = 2;', () => 'const b = $&3;')
     );
@@ -215,7 +225,7 @@ describe('edit (#258)', () => {
   it('keeps a byte order mark', () => {
     const cwd = workspace({ 'bom.txt': '\ufefffirst\nsecond\n' });
     const ctx = context(cwd);
-    expect(call('read', { path: 'bom.txt' }, ctx).result).toBe(
+    expect(call('read', { path: 'bom.txt' }, ctx).text).toBe(
       '\ufefffirst\nsecond\n'
     );
     const receipt = call(
@@ -420,10 +430,22 @@ describe('file tools in the task loop (#258)', () => {
       },
       { content: 'The edit failed.' },
     ]);
+    // The failure stays unresolved with its original error (#261).
     expect(result.evidence.outcome).toMatchObject({
-      status: 'completed',
+      status: 'incomplete',
       exitCode: 1,
-      remainingObligations: ['the last tool call failed'],
+      remainingObligations: [
+        '`edit a.js` failed: a.js: old_string was not found; nothing changed',
+      ],
+      failures: [
+        {
+          command: 'edit a.js',
+          tool: 'edit',
+          target: 'a.js',
+          exitCode: 1,
+          error: 'a.js: old_string was not found; nothing changed',
+        },
+      ],
     });
     expect(result.after).toEqual(result.before);
   });

package/test/regression-258.js

@@ -30,6 +30,7 @@ const {
   recognizeCapabilityGap,
 } = require('../src/lib/formal-ai-client');
 const { runFormalAiTask } = require('../src/lib/formal-ai');
+const { readFrame } = require('../src/lib/formal-ai-files');
 const { createFakeServer, edits } = require('./fixtures/formal-ai/replay.cjs');
 const {
   CLI,
@@ -273,11 +274,22 @@ describe('structured file tools against captured receipts (#258)', () => {
     'reads, edits and reads back, preserving all other bytes',
     async () => {
       const result = await run(LITERAL_EDIT, { effects: 'workspace-write' });
+      // start verified the edit byte for byte, but Formal AI's final answer
+      // is a gap for the clause "Preserve all other content.": that is never
+      // reported as done (#261).
       expect(result.outcome).toMatchObject({
-        status: 'completed',
-        exitCode: 0,
-        message: 'finished after 3 tool call(s); changed snippet.mjs',
-        remainingObligations: [],
+        status: 'incomplete',
+        exitCode: EXIT.UNSUPPORTED,
+        message:
+          'Formal AI reported obligation_gap (no_artifact_in_clause) for the clause "Preserve all other content."; nothing was verified as done by Formal AI; changed snippet.mjs',
+        remainingObligations: [
+          'Formal AI did not confirm the requested change',
+        ],
+        gapRecord: {
+          schema: 'obligation_gap',
+          reason: 'no_artifact_in_clause',
+          clause: 'Preserve all other content.',
+        },
       });
       expect(result.text).toBe(EDITED);
       const [read, edit, check] = result.tools;
@@ -304,9 +316,9 @@ describe('structured file tools against captured receipts (#258)', () => {
         command: 'sha256sum -- snippet.mjs',
         exitCode: 0,
       });
-      // Formal AI received the exact file text and the readback hash.
+      // Formal AI received the exact file text (framed) and the readback hash.
       const toolMessages = result.sent[3].filter((m) => m.role === 'tool');
-      expect(toolMessages[0].content).toBe(ORIGINAL);
+      expect(toolMessages[0].content).toBe(readFrame(ORIGINAL));
       expect(toolMessages[2].content).toContain(edit.sha256);
       expect(result.evidence.effects).toMatchObject({
         changed: ['snippet.mjs'],
@@ -398,7 +410,7 @@ describe('edit requests through the CLI (#258)', () => {
     const result = cli(LITERAL_EDIT, {
       START_FORMAL_AI_EFFECTS: 'workspace-write',
     });
-    expect(result.status).toBe(0);
+    expect(result.status).toBe(EXIT.UNSUPPORTED);
     expect(result.text).toBe(EDITED);
     expect(result.stderr).toContain('[formal-ai] edit snippet.mjs');
     expect(result.stderr).toContain(
@@ -411,8 +423,9 @@ describe('edit requests through the CLI (#258)', () => {
       'write',
     ]);
     expect(result.evidence.outcome).toMatchObject({
-      status: 'completed',
-      exitCode: 0,
+      status: 'incomplete',
+      exitCode: EXIT.UNSUPPORTED,
+      gapRecord: { schema: 'obligation_gap' },
     });
   });