fix(infra) - Make file system interactive test check only tool call (#11055)

Co-authored-by: gemini-cli-robot <gemini-cli-robot@google.com>

.github/workflows/e2e.yml

@@ -218,7 +218,7 @@ jobs:
           UV_THREADPOOL_SIZE: '32'
           NODE_ENV: 'test'
         shell: 'pwsh'
-        run: 'npm run test:integration:sandbox:none'
+        run: 'npm run deflake:test:integration:sandbox:none'
 
   e2e:
     name: 'E2E'

integration-tests/file-system-interactive.test.ts

@@ -7,7 +7,7 @@
 import { expect, describe, it, beforeEach, afterEach } from 'vitest';
 import { TestRig } from './test-helper.js';
 
-describe.skip('Interactive file system', () => {
+describe('Interactive file system', () => {
   let rig: TestRig;
 
   beforeEach(() => {
@@ -33,16 +33,16 @@ describe.skip('Interactive file system', () => {
     const readCall = await rig.waitForToolCall('read_file', 30000);
     expect(readCall, 'Expected to find a read_file tool call').toBe(true);
 
-    await run.expectText('1.0.0', 30000);
-
     // Step 2: Write the file
     const writePrompt = `now change the version to 1.0.1 in the file`;
     await run.type(writePrompt);
     await run.sendKeys('\r');
 
-    await rig.expectToolCallSuccess(['write_file', 'replace'], 30000);
-
-    const newFileContent = rig.readFile(fileName);
-    expect(newFileContent).toBe('1.0.1');
+    // Check tool calls made with right args
+    await rig.expectToolCallSuccess(
+      ['write_file', 'replace'],
+      30000,
+      (args) => args.includes('1.0.1') && args.includes(fileName),
+    );
   });
 });

integration-tests/test-helper.ts

@@ -628,7 +628,11 @@ export class TestRig {
     );
   }
 
-  async expectToolCallSuccess(toolNames: string[], timeout?: number) {
+  async expectToolCallSuccess(
+    toolNames: string[],
+    timeout?: number,
+    matchArgs?: (args: string) => boolean,
+  ) {
     // Use environment-specific timeout
     if (!timeout) {
       timeout = getDefaultTimeout();
@@ -642,7 +646,10 @@ export class TestRig {
         const toolLogs = this.readToolLogs();
         return toolNames.some((name) =>
           toolLogs.some(
-            (log) => log.toolRequest.name === name && log.toolRequest.success,
+            (log) =>
+              log.toolRequest.name === name &&
+              log.toolRequest.success &&
+              (matchArgs?.call(this, log.toolRequest.args) ?? true),
           ),
         );
       },

package.json

@@ -21,8 +21,8 @@
     "start:a2a-server": "CODER_AGENT_PORT=41242 npm run start --workspace @google/gemini-cli-a2a-server",
     "debug": "cross-env DEBUG=1 node --inspect-brk scripts/start.js",
     "deflake": "node scripts/deflake.js",
-    "deflake:test:integration:sandbox:none": "npm run deflake -- --command='npm run test:integration:sandbox:none -- --retry=0'",
-    "deflake:test:integration:sandbox:docker": "npm run deflake -- --command='npm run test:integration:sandbox:docker -- --retry=0'",
+    "deflake:test:integration:sandbox:none": "npm run deflake -- --command=\"npm run test:integration:sandbox:none -- --retry=0",
+    "deflake:test:integration:sandbox:docker": "npm run deflake -- --command=\"npm run test:integration:sandbox:docker -- --retry=0",
     "auth:npm": "npx google-artifactregistry-auth",
     "auth:docker": "gcloud auth configure-docker us-west1-docker.pkg.dev",
     "auth": "npm run auth:npm && npm run auth:docker",