[RUM-18686] Electron OS and Version testing [windows]

[cdn34dd]

Motivation

The compatibility workflow currently covers Linux and macOS. This PR adds Windows coverage across the same Electron targets, reusing the existing E2E and integration suites.

Windows introduces differences in command execution, filesystem paths, dependency packaging, and container display availability. These changes let the suites run on shared Windows runners and add regression coverage for Windows packaging failures.

This PR is stacked on carlosnogueira/RUM-15522/electron-os-and-version-test, with the diff starting after 9b408f3e.

Changes

The Windows workflow has three steps:

1. Prepare the environment. Build a Windows Server 2022 container on the existing windows-v2:2022 runner. The image supplies Node, Yarn, Git, 7-Zip, and the Visual C++ runtime. No dedicated AMI or published image is required.
2. Prepare apps and run tests. Mount the checkout read-only, copy sources into the short container path C:\w, and run the existing compatibility preparation and test commands. Host dependencies and previous build artifacts are excluded.
3. Collect results and clean up. Save command logs, environment details, container state, target metadata, and available Playwright reports and traces. Remove only the container and image created by the job.

Windows adds seven jobs to the compatibility matrix, bringing the full Linux/macOS/Windows matrix to 21 jobs. Existing default-branch, manual, scheduled, filtering, and nightly failure policies continue to apply.

Supporting changes:

• Replace Unix-specific build commands and resolve Windows Yarn/npm entry points without shell interpolation.
• Fix WASM chunk naming for Windows paths and verify that CJS/ESM chunks remain lazy and included in the package allowlist.
• Adapt crash-test cleanup and display assertions for Windows containers.
• Retain two CI retries, print Windows test errors immediately, retain failure traces, and stop after three failures.
• Generate child-pipeline YAML from structured objects, with tests preserving command strings and scalar types.

The Windows payload regression checks that default-copy dd-trace paths would exceed MAX_PATH at a long temporary destination, while output built with copyRuntimeDependencies: false extracts completely using Windows PowerShell 5.1. Separate packaging assertions verify dependency placement in app.asar. This covers the extraction regression without running a full MSIX build or signing pipeline.

Review guide

The main commits separate portable commands, WASM packaging, test changes, container execution, and CI integration. Suggested reading order:

• Portable commands: scripts/lib/commandInvocation.ts, scripts/lib/command.ts, and the SDK/fixture build scripts.
• WASM packaging: rollup.config.mjs and scripts/check-sdk-chunks.spec.ts.
• Test coverage: e2e/playwright.config.ts, the crash scenarios, and e2e/integration/scenarios/windows-payload-extraction.scenario.ts.
• Container execution: ci/windows/run.ps1, scripts/run-windows-container-tests.ts, and scripts/lib/windowsContainer.ts.
• CI orchestration: scripts/lib/compatibilityCi.ts and e2e/compatibility/config.json.

Test instructions

On a Windows host with Docker configured for Windows containers, use a regular Git clone:

docker info --format '{{.OSType}}' # Must print windows

powershell -NoProfile -ExecutionPolicy Bypass `
  -File ci/windows/run.ps1 -Target electron-41

Process isolation is the default. Hosts that require and support Hyper-V isolation can add -Isolation hyperv.

To run the regular suites through the same container:

powershell -NoProfile -ExecutionPolicy Bypass `
  -File ci/windows/run.ps1 -Suite e2e

powershell -NoProfile -ExecutionPolicy Bypass `
  -File ci/windows/run.ps1 -Suite integration

To exercise the compatibility workflow in GitLab, manually start a pipeline with:

COMPATIBILITY_TESTS=true
DD_ELECTRON_COMPATIBILITY_ENVIRONMENTS=windows
DD_ELECTRON_COMPATIBILITY_TARGETS=electron-41

Leave the target filter empty to run all configured Electron versions on Windows.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-02T12:57:42.194541Z	cf9dd2c	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[cursor]

---

PR Review — Quality: 5.0 / 5 · Review load: Broad

At 5.0 / 5 this is ready to merge from a code-review perspective: I am approving. The change extends the Electron compatibility matrix to Windows via a Server 2022 container workflow, makes build/test scripts portable (including Yarn/npm invocation without a shell), fixes WASM chunk detection on Windows paths, and hardens crash and telemetry scenarios for headless Windows environments. The added unit and integration coverage matches the new surfaces. Review load is Broad because container orchestration, command bootstrap semantics, and packaging behavior land together—worth a second pair of eyes on Windows CI, even though I have no blocking findings.

Customer impact: Low — default SDK behavior is unchanged; the rollup path normalization fixes Windows packaging for the WASM chunk, and the rest is CI, fixtures, and test infrastructure.

Concern
Windows container CI (image, workspace copy, GitLab matrix jobs)	Risky
Portable command() / Yarn and npm JavaScript entry-point resolution on Windows	Risky
Rollup WASM manualChunks path normalization and package allowlist checks	Risky
Playwright reporting, crash predicates, and display-count assertions	Routine
YAML-based compatibility pipeline generation and filters	Routine

Quality — 5.0: The design is coherent end to end: host run.ps1 keeps Docker scope narrow, in-container preparation reuses the same yarn targets as Linux/macOS, and exclusions in copyWindowsContainerWorkspace mirror compatibility materialization. Command invocation and logged retries are well tested; the WASM regression test exercises real Rollup output; crash tests now filter on is_crash, which addresses realistic intake noise on Windows.

Review load — Broad: Three risky boundaries (CI containers, bootstrap commands, Windows packaging) ship in one stack, though commits separate them logically. A Windows-runner owner should sanity-check the first real matrix jobs; nothing here blocks on that alone.

Suggested split

Not required for merge. If you ever need to land pieces independently: (1) portable scripts + rollup WASM fix + chunk tests, (2) Playwright/scenario hardening, (3) Windows container + GitLab matrix enablement last (depends on 1–2). Commits e4284b57 / 048a83be / 9d852f08 / 1b989df7 / b674ee7d already follow that rough order.

  

_{Sent by Cursor Automation: electron-sdk reviews}

[bcaudan]

Great!