style: fix formatting in hooks-agent-flow.test.ts

- Ensure 'SystemRoot', 'COMSPEC', 'windir', and 'PATHEXT' are passed to node-pty on Windows to prevent 'posix_spawnp' failures.
- Clean up test directories in 'TestRig.setup' to ensure a fresh state for retries and prevent telemetry log accumulation (fixing the 1, 2, 3 failure pattern).
- Fix path normalization in 'Hook Disabling' test to ensure disabled hooks are correctly matched on Windows.

Part of https://github.com/google-gemini/gemini-cli/pull/18665

.github/workflows/chained_e2e.yml

@@ -275,7 +275,7 @@ jobs:
           UV_THREADPOOL_SIZE: '32'
           NODE_ENV: 'test'
         shell: 'pwsh'
-        run: 'npm run test:integration:sandbox:none'
+        run: 'npm run test:e2e'
 
   evals:
     name: 'Evals (ALWAYS_PASSING)'
@@ -315,19 +315,30 @@ jobs:
     needs:
       - 'e2e_linux'
       - 'e2e_mac'
+      - 'e2e_windows'
       - 'evals'
       - 'merge_queue_skipper'
     runs-on: 'gemini-cli-ubuntu-16-core'
     steps:
       - name: 'Check E2E test results'
         run: |
-          if [[ ${{ needs.e2e_linux.result }} != 'success' || \
-               ${{ needs.e2e_mac.result }} != 'success' || \
-               ${{ needs.evals.result }} != 'success' ]]; then
-            echo "One or more E2E jobs failed."
+          if [[ ${{ needs.e2e_linux.result }} != 'success' ]]; then
+            echo "Linux E2E job failed."
             exit 1
           fi
-          echo "All required E2E jobs passed!"
+          if [[ ${{ needs.e2e_mac.result }} != 'success' ]]; then
+            echo "macOS E2E job failed."
+            exit 1
+          fi
+          if [[ ${{ needs.e2e_windows.result }} != 'success' ]]; then
+            echo "Windows E2E job failed."
+            exit 1
+          fi
+          if [[ ${{ needs.evals.result }} != 'success' ]]; then
+            echo "Evals job failed."
+            exit 1
+          fi
+          echo "All E2E jobs passed!"
 
   set_workflow_status:
     runs-on: 'gemini-cli-ubuntu-16-core'

integration-tests/hooks-agent-flow.test.ts

@@ -5,7 +5,7 @@
  */
 
 import { describe, it, expect, beforeEach, afterEach } from 'vitest';
-import { TestRig } from './test-helper.js';
+import { TestRig, normalizePath } from './test-helper.js';
 import { join } from 'node:path';
 import { writeFileSync } from 'node:fs';
 
@@ -113,10 +113,9 @@ describe('Hooks Agent Flow', () => {
       }
       `;
 
-      const scriptPath = join(rig.testDir!, 'after_agent_verify.cjs');
-      writeFileSync(scriptPath, hookScript);
+      const scriptPath = rig.createScript('after_agent_verify.cjs', hookScript);
 
-      await rig.setup('should receive prompt and response in AfterAgent hook', {
+      rig.setup('should receive prompt and response in AfterAgent hook', {
         settings: {
           hooksConfig: {
             enabled: true,
@@ -127,7 +126,7 @@ describe('Hooks Agent Flow', () => {
                 hooks: [
                   {
                     type: 'command',
-                    command: `node "${scriptPath}"`,
+                    command: normalizePath(`node "${scriptPath}"`)!,
                     timeout: 5000,
                   },
                 ],
@@ -157,7 +156,7 @@ describe('Hooks Agent Flow', () => {
     });
 
     it('should process clearContext in AfterAgent hook output', async () => {
-      await rig.setup('should process clearContext in AfterAgent hook output', {
+      rig.setup('should process clearContext in AfterAgent hook output', {
         fakeResponsesPath: join(
           import.meta.dirname,
           'hooks-system.after-agent.responses',
@@ -171,18 +170,32 @@ describe('Hooks Agent Flow', () => {
         const input = JSON.parse(fs.readFileSync(0, 'utf-8'));
         const messageCount = input.llm_request?.contents?.length || 0;
         let counts = [];
-        try { counts = JSON.parse(fs.readFileSync('${messageCountFile}', 'utf-8')); } catch (e) {}
+        try { counts = JSON.parse(fs.readFileSync(${JSON.stringify(messageCountFile)}, 'utf-8')); } catch (e) {}
         counts.push(messageCount);
-        fs.writeFileSync('${messageCountFile}', JSON.stringify(counts));
+        fs.writeFileSync(${JSON.stringify(messageCountFile)}, JSON.stringify(counts));
         console.log(JSON.stringify({ decision: 'allow' }));
       `;
-      const beforeModelScriptPath = join(
-        rig.testDir!,
+      const beforeModelScriptPath = rig.createScript(
         'before_model_counter.cjs',
+        beforeModelScript,
       );
-      writeFileSync(beforeModelScriptPath, beforeModelScript);
 
-      await rig.setup('should process clearContext in AfterAgent hook output', {
+      const afterAgentScript = `
+        console.log(JSON.stringify({
+          decision: 'block',
+          reason: 'Security policy triggered',
+          hookSpecificOutput: {
+            hookEventName: 'AfterAgent',
+            clearContext: true
+          }
+        }));
+      `;
+      const afterAgentScriptPath = rig.createScript(
+        'after_agent_clear.cjs',
+        afterAgentScript,
+      );
+
+      rig.setup('should process clearContext in AfterAgent hook output', {
         settings: {
           hooks: {
             enabled: true,
@@ -191,7 +204,7 @@ describe('Hooks Agent Flow', () => {
                 hooks: [
                   {
                     type: 'command',
-                    command: `node "${beforeModelScriptPath}"`,
+                    command: normalizePath(`node "${beforeModelScriptPath}"`)!,
                     timeout: 5000,
                   },
                 ],
@@ -202,7 +215,7 @@ describe('Hooks Agent Flow', () => {
                 hooks: [
                   {
                     type: 'command',
-                    command: `node -e "console.log(JSON.stringify({decision: 'block', reason: 'Security policy triggered', hookSpecificOutput: {hookEventName: 'AfterAgent', clearContext: true}}))"`,
+                    command: normalizePath(`node "${afterAgentScriptPath}"`)!,
                     timeout: 5000,
                   },
                 ],
@@ -244,6 +257,22 @@ describe('Hooks Agent Flow', () => {
             import.meta.dirname,
             'hooks-agent-flow-multistep.responses',
           ),
+        },
+      );
+
+      // Create script files for hooks
+      const baPath = rig.createScript(
+        'ba_fired.cjs',
+        "console.log('BeforeAgent Fired');",
+      );
+      const aaPath = rig.createScript(
+        'aa_fired.cjs',
+        "console.log('AfterAgent Fired');",
+      );
+
+      await rig.setup(
+        'should fire BeforeAgent and AfterAgent exactly once per turn despite tool calls',
+        {
           settings: {
             hooksConfig: {
               enabled: true,
@@ -254,7 +283,7 @@ describe('Hooks Agent Flow', () => {
                   hooks: [
                     {
                       type: 'command',
-                      command: `node -e "console.log('BeforeAgent Fired')"`,
+                      command: normalizePath(`node "${baPath}"`)!,
                       timeout: 5000,
                     },
                   ],
@@ -265,7 +294,7 @@ describe('Hooks Agent Flow', () => {
                   hooks: [
                     {
                       type: 'command',
-                      command: `node -e "console.log('AfterAgent Fired')"`,
+                      command: normalizePath(`node "${aaPath}"`)!,
                       timeout: 5000,
                     },
                   ],

integration-tests/test-helper.ts

@@ -5,3 +5,4 @@
  */
 
 export * from '@google/gemini-cli-test-utils';
+export { normalizePath } from '@google/gemini-cli-test-utils';

packages/core/src/core/coreToolScheduler.ts

@@ -965,6 +965,7 @@ export class CoreToolScheduler {
 
       // The active tool is finished. Move it to the completed batch.
       const completedCall = activeCall as CompletedToolCall;
+
       this.completedToolCallsForBatch.push(completedCall);
       logToolCall(this.config, new ToolCallEvent(completedCall));
 

packages/core/src/hooks/hookRegistry.ts

@@ -93,7 +93,8 @@ export class HookRegistry {
   private getHookName(
     entry: HookRegistryEntry | { config: HookConfig },
   ): string {
-    return entry.config.name || entry.config.command || 'unknown-command';
+    const name = entry.config.name || entry.config.command || 'unknown-command';
+    return name.replace(/\\/g, '/');
   }
 
   /**

packages/core/src/hooks/hookRunner.ts

@@ -35,7 +35,6 @@ const DEFAULT_HOOK_TIMEOUT = 60000;
  * Exit code constants for hook execution
  */
 const EXIT_CODE_SUCCESS = 0;
-const EXIT_CODE_BLOCKING_ERROR = 2;
 const EXIT_CODE_NON_BLOCKING_ERROR = 1;
 
 /**
@@ -353,28 +352,25 @@ export class HookRunner {
 
         // Parse output
         let output: HookOutput | undefined;
-        if (exitCode === EXIT_CODE_SUCCESS && stdout.trim()) {
+
+        const textToParse = stdout.trim() || stderr.trim();
+        if (textToParse) {
           try {
-            let parsed = JSON.parse(stdout.trim());
+            let parsed = JSON.parse(textToParse);
             if (typeof parsed === 'string') {
-              // If the output is a string, parse it in case
-              // it's double-encoded JSON string.
               parsed = JSON.parse(parsed);
             }
-            if (parsed) {
+            if (parsed && typeof parsed === 'object') {
               // eslint-disable-next-line @typescript-eslint/no-unsafe-type-assertion
               output = parsed as HookOutput;
             }
           } catch {
             // Not JSON, convert plain text to structured output
-            output = this.convertPlainTextToHookOutput(stdout.trim(), exitCode);
+            output = this.convertPlainTextToHookOutput(
+              textToParse,
+              exitCode || EXIT_CODE_SUCCESS,
+            );
           }
-        } else if (exitCode !== EXIT_CODE_SUCCESS && stderr.trim()) {
-          // Convert error output to structured format
-          output = this.convertPlainTextToHookOutput(
-            stderr.trim(),
-            exitCode || EXIT_CODE_NON_BLOCKING_ERROR,
-          );
         }
 
         resolve({
@@ -435,18 +431,18 @@ export class HookRunner {
         decision: 'allow',
         systemMessage: text,
       };
-    } else if (exitCode === EXIT_CODE_BLOCKING_ERROR) {
-      // Blocking error
-      return {
-        decision: 'deny',
-        reason: text,
-      };
-    } else {
-      // Non-blocking error (EXIT_CODE_NON_BLOCKING_ERROR or any other code)
+    } else if (exitCode === EXIT_CODE_NON_BLOCKING_ERROR) {
+      // Non-blocking error (EXIT_CODE_NON_BLOCKING_ERROR = 1)
       return {
         decision: 'allow',
         systemMessage: `Warning: ${text}`,
       };
+    } else {
+      // All other non-zero exit codes (including 2) are blocking
+      return {
+        decision: 'deny',
+        reason: text,
+      };
     }
   }
 }

packages/core/src/policy/policy-engine.ts

@@ -360,8 +360,8 @@ export class PolicyEngine {
       );
 
       if (match) {
-        debugLogger.debug(
-          `[PolicyEngine.check] MATCHED rule: toolName=${rule.toolName}, decision=${rule.decision}, priority=${rule.priority}, argsPattern=${rule.argsPattern?.source || 'none'}`,
+        debugLogger.log(
+          `[PolicyEngine.check] MATCHED rule: toolName=${rule.toolName}, decision=${rule.decision}, priority=${rule.priority}, source=${rule.source}`,
         );
 
         if (isShellCommand && toolName) {
@@ -389,7 +389,7 @@ export class PolicyEngine {
 
     // Default if no rule matched
     if (decision === undefined) {
-      debugLogger.debug(
+      debugLogger.log(
         `[PolicyEngine.check] NO MATCH - using default decision: ${this.defaultDecision}`,
       );
       if (toolName && SHELL_TOOL_NAMES.includes(toolName)) {

packages/core/src/utils/getPty.ts

@@ -18,6 +18,9 @@ export interface PtyProcess {
 }
 
 export const getPty = async (): Promise<PtyImplementation> => {
+  if (process.env['GEMINI_PTY_INFO'] === 'child_process') {
+    return null;
+  }
   try {
     const lydell = '@lydell/node-pty';
     const module = await import(lydell);

packages/test-utils/src/test-rig.ts

@@ -345,6 +345,7 @@ export class TestRig {
   originalFakeResponsesPath?: string;
   private _interactiveRuns: InteractiveRun[] = [];
   private _spawnedProcesses: ChildProcess[] = [];
+  private _initialized = false;
 
   setup(
     testName: string,
@@ -359,6 +360,14 @@ export class TestRig {
       env['INTEGRATION_TEST_FILE_DIR'] || join(os.tmpdir(), 'gemini-cli-tests');
     this.testDir = join(testFileDir, sanitizedName);
     this.homeDir = join(testFileDir, sanitizedName + '-home');
+
+    if (!this._initialized) {
+      // Clean up existing directories from previous runs (e.g. retries)
+      this._cleanDir(this.testDir);
+      this._cleanDir(this.homeDir);
+      this._initialized = true;
+    }
+
     mkdirSync(this.testDir, { recursive: true });
     mkdirSync(this.homeDir, { recursive: true });
     if (options.fakeResponsesPath) {
@@ -373,6 +382,36 @@ export class TestRig {
     this._createSettingsFile(options.settings);
   }
 
+  private _cleanDir(dir: string) {
+    if (fs.existsSync(dir)) {
+      for (let i = 0; i < 10; i++) {
+        try {
+          fs.rmSync(dir, { recursive: true, force: true });
+          return;
+        } catch (err) {
+          if (i === 9) {
+            console.error(
+              `Failed to clean directory ${dir} after 10 attempts:`,
+              err,
+            );
+            throw err;
+          }
+          const delay = Math.min(Math.pow(2, i) * 1000, 10000); // Max 10s delay
+          try {
+            const sharedBuffer = new Int32Array(new SharedArrayBuffer(4));
+            Atomics.wait(sharedBuffer, 0, 0, delay);
+          } catch {
+            // Fallback for environments where SharedArrayBuffer might be restricted
+            const start = Date.now();
+            while (Date.now() - start < delay) {
+              /* busy wait */
+            }
+          }
+        }
+      }
+    }
+  }
+
   private _createSettingsFile(overrideSettings?: Record<string, unknown>) {
     const projectGeminiDir = join(this.testDir!, GEMINI_DIR);
     mkdirSync(projectGeminiDir, { recursive: true });
@@ -474,6 +513,17 @@ export class TestRig {
     return { command, initialArgs };
   }
 
+  createScript(fileName: string, content: string) {
+    if (!this.testDir) {
+      throw new Error(
+        'TestRig.setup must be called before creating files or scripts',
+      );
+    }
+    const scriptPath = join(this.testDir, fileName);
+    writeFileSync(scriptPath, content);
+    return normalizePath(scriptPath);
+  }
+
   private _getCleanEnv(
     extraEnv?: Record<string, string | undefined>,
   ): Record<string, string | undefined> {
@@ -499,6 +549,7 @@ export class TestRig {
     return {
       ...cleanEnv,
       GEMINI_CLI_HOME: this.homeDir!,
+      GEMINI_PTY_INFO: 'child_process',
       ...extraEnv,
     };
   }
@@ -801,6 +852,13 @@ export class TestRig {
     // Kill any interactive runs that are still active
     for (const run of this._interactiveRuns) {
       try {
+        if (process.platform === 'win32') {
+          // @ts-ignore - access private ptyProcess
+          const pid = run.ptyProcess?.pid;
+          if (pid) {
+            execSync(`taskkill /F /T /PID ${pid}`, { stdio: 'ignore' });
+          }
+        }
         await run.kill();
       } catch (error) {
         if (env['VERBOSE'] === 'true') {
@@ -814,6 +872,9 @@ export class TestRig {
     for (const child of this._spawnedProcesses) {
       if (child.exitCode === null && child.signalCode === null) {
         try {
+          if (process.platform === 'win32' && child.pid) {
+            execSync(`taskkill /F /T /PID ${child.pid}`, { stdio: 'ignore' });
+          }
           child.kill('SIGKILL');
         } catch (error) {
           if (env['VERBOSE'] === 'true') {
@@ -836,21 +897,21 @@ export class TestRig {
     // Clean up test directory and home directory
     if (this.testDir && !env['KEEP_OUTPUT']) {
       try {
-        fs.rmSync(this.testDir, { recursive: true, force: true });
+        this._cleanDir(this.testDir);
       } catch (error) {
         // Ignore cleanup errors
-        if (env['VERBOSE'] === 'true') {
-          console.warn('Cleanup warning:', (error as Error).message);
+        if (env['VERBOSE'] === 'true' || env['CI'] === 'true') {
+          console.warn('Cleanup warning (testDir):', (error as Error).message);
         }
       }
     }
     if (this.homeDir && !env['KEEP_OUTPUT']) {
       try {
-        fs.rmSync(this.homeDir, { recursive: true, force: true });
+        this._cleanDir(this.homeDir);
       } catch (error) {
         // Ignore cleanup errors
-        if (env['VERBOSE'] === 'true') {
-          console.warn('Cleanup warning:', (error as Error).message);
+        if (env['VERBOSE'] === 'true' || env['CI'] === 'true') {
+          console.warn('Cleanup warning (homeDir):', (error as Error).message);
         }
       }
     }
@@ -1282,6 +1343,23 @@ export class TestRig {
 
     const envVars = this._getCleanEnv(options?.env);
 
+    // node-pty on windows often needs these to spawn correctly
+    if (process.platform === 'win32') {
+      const windowsCriticalVars = [
+        'SystemRoot',
+        'COMSPEC',
+        'windir',
+        'PATHEXT',
+        'TEMP',
+        'TMP',
+      ];
+      for (const v of windowsCriticalVars) {
+        if (process.env[v] && !envVars[v]) {
+          envVars[v] = process.env[v]!;
+        }
+      }
+    }
+
     const ptyOptions: pty.IPtyForkOptions = {
       name: 'xterm-color',
       cols: 80,
@@ -1364,3 +1442,11 @@ export class TestRig {
     throw new Error(`pollCommand timed out after ${timeout}ms`);
   }
 }
+
+/**
+ * Normalizes a path for cross-platform matching (replaces backslashes with forward slashes).
+ */
+export function normalizePath(p: string | undefined): string | undefined {
+  if (!p) return p;
+  return p.replace(/\\/g, '/');
+}