fix(ci): resolve failures from telemetry regression, unhandled rejections, and flaky tests

[aseemxs]

Problem

Multiple CI checks failing across macOS, Windows, and Linux (CodeBuild). All 5849+ unit tests pass — failures are from test assertion regressions, unhandled promise rejections detected by the run_and_report script, and pre-existing flaky tests.

Root Causes & Fixes

1. Lambda URI Handler telemetry — 3 test failures (all platforms)

Introduced by: #8598 (fix(lambda): add confirmation prompt before initiating console login)

What broke: #8598 changed handleLambdaUriError from throwing a cancelled ToolkitError to calling telemetry.record() + return. This broke telemetry because run() overwrites the span result with Succeeded when the function completes normally — the record() call gets clobbered.

Fix: Revert the cancellation path to throw ToolkitError.chain(e, 'User cancelled operation', { cancelled: true }). This lets run() correctly record Cancelled as the result. Tests updated to assert the throw instead of checking telemetry directly.

Files: uriHandlers.ts (functional), uriHandlers.test.ts (test)

2. AppBuilder Walkthrough test — 1 test failure (macOS minimum)

Introduced by: Interaction between the walkthrough test (#8236) and background security scan errors from startSecurityScan.ts. The test's onDidShowMessage handler asserted on every message, but an unrelated security scan error message fired during the test.

Fix: Filter the onDidShowMessage handler to only assert on the expected overwrite prompt message, ignoring unrelated ones. Added explicit assertion that the prompt was shown.

Files: walkthrough.test.ts (test only)

3. Unhandled promise rejections — Linux CodeBuild failures (all 3 variants)

Introduced by:

• SageMaker server throw when env var missing — feat(sagemaker): Merge SageMaker SSH Kiro integration #8589 (feat(sagemaker): Merge SageMaker SSH Kiro integration)
• CloudFormation StackActionCodeLensProvider calling sendRequest on stopped client — feat(cloudformation): Merge CloudFormation LSP integration with toolkit updates #8275 (feat(cloudformation): Merge CloudFormation LSP integration)
• CloudFormation deactivate() client.stop() rejecting — same PR feat(cloudformation): Merge CloudFormation LSP integration with toolkit updates #8275

What broke: The run_and_report script in CodeBuild fails the build if it detects "rejected promise not handled" in stdout. All 5849 tests pass, but these 5 unhandled rejections triggered the check.

Fix:

• SageMaker: console.error + return instead of throw for missing env var — avoids unhandled rejection while keeping extension host alive
• CloudFormation CodeLens: isRunning() guard + try/catch before sendRequest, errors logged at debug level
• CloudFormation deactivate: .catch() on client.stop()

Files: server.ts, stackActionCodeLensProvider.ts, extension.ts (all functional — adds resilience)

4. CloudFormation LSP client lifecycle rejections — Linux CodeBuild minimum

Introduced by: #8275 (feat(cloudformation): Merge CloudFormation LSP integration)

What broke: The vscode-languageclient LanguageClient creates unhandled promise rejections internally when the LSP server process fails to start during CI. These rejections (connection got disposed, Client is not running) happen inside the library's handleConnectionClosed/doInitialize and cannot be caught from our code.

Fix: Skip the CloudFormation LSP client activation when AWS_TOOLKIT_AUTOMATION === 'unit'. No unit tests depend on the LSP client — the CloudFormation unit tests cover template parsing, not the language server. E2E tests are unaffected (they use a different automation value).

Files: extension.ts (functional — conditional activation)

5. Pre-existing flaky test fixes

These are not caused by any recent PR but fail intermittently across CI:

SharedCredentialsProvider (handleInvalidConsoleCredentials — does not prompt reload for non-session errors):

• Previous test's messages leak through getTestWindow().shownMessages — the array isn't cleared between tests
• Fix: snapshot the count before the test and assert no new messages were added

ToolkitLogger (logs to a file — logs warn):

• Windows file I/O race: log file exists but write hasn't flushed yet, single read misses content
• Fix: retry loop that keeps reading until expected content appears (within existing 10s timeout)

editorContext + recommendationHandler (30s timeout on macOS insiders):

• Slow VS Code startup + async telemetry/mock setup exceeds default 30s Mocha timeout
• Fix: bump to 60s for these two tests

Files: sharedCredentialsProvider.test.ts, toolkitLogger.test.ts, editorContext.test.ts, recommendationHandler.test.ts (all test only)

Functional vs Test-Only Changes

File	Type	Change
uriHandlers.ts	Functional	Reverts cancellation to throw for correct telemetry
server.ts	Functional	console.error + return instead of throw for missing env var
stackActionCodeLensProvider.ts	Functional	isRunning() guard + debug logging on catch
extension.ts	Functional	.catch() on deactivate, skip LSP in unit tests
uriHandlers.test.ts	Test only	Updated assertions
walkthrough.test.ts	Test only	Filtered onDidShowMessage handler
sharedCredentialsProvider.test.ts	Test only	Snapshot message count instead of absolute check
toolkitLogger.test.ts	Test only	Retry loop for file content read
editorContext.test.ts	Test only	Timeout bump to 60s
recommendationHandler.test.ts	Test only	Timeout bump to 60s

Testing

• All GitHub Actions checks pass (macOS/Windows/Linux × stable/minimum/insiders × toolkit/amazonq)
• All 3 Linux CodeBuild variants pass (stable/minimum/insiders)
• CloudFormation LSP E2E tests pass (all 3 platforms)

[amazon-inspector-ohio]

⏳ I'm reviewing this pull request for security vulnerabilities and code quality issues. I'll provide an update when I'm done

[amazon-inspector-ohio]

✅ I finished the code review, and didn't find any security or code quality issues.

[amazon-inspector-ohio]

⏳ I'm reviewing this pull request for security vulnerabilities and code quality issues. I'll provide an update when I'm done

[amazon-inspector-ohio]

✅ I finished the code review, and didn't find any security or code quality issues.

[github-actions]

• This pull request implements a feat or fix, so it must include a changelog entry (unless the fix is for an unreleased feature). Review the changelog guidelines.
  • Note: beta or "experiment" features that have active users should announce fixes in the changelog.
  • If this is not a feature or fix, use an appropriate type from the title guidelines. For example, telemetry-only changes should use the telemetry type.

[amazon-inspector-ohio]

⏳ I'm reviewing this pull request for security vulnerabilities and code quality issues. I'll provide an update when I'm done

[amazon-inspector-ohio]

✅ I finished the code review, and didn't find any security or code quality issues.

[amazon-inspector-ohio]

⏳ I'm reviewing this pull request for security vulnerabilities and code quality issues. I'll provide an update when I'm done

[amazon-inspector-ohio]

✅ I finished the code review, and didn't find any security or code quality issues.

[amazon-inspector-ohio]

⏳ I'm reviewing this pull request for security vulnerabilities and code quality issues. I'll provide an update when I'm done

[amazon-inspector-ohio]

✅ I finished the code review, and didn't find any security or code quality issues.

[amazon-inspector-ohio]

⏳ I'm reviewing this pull request for security vulnerabilities and code quality issues. I'll provide an update when I'm done

[amazon-inspector-ohio]

✅ I finished the code review, and didn't find any security or code quality issues.