Skip to content

test: automate desktop protocol journey - #193

Merged
oratis merged 21 commits into
mainfrom
codex/desktop-preview-fixture
Aug 2, 2026
Merged

test: automate desktop protocol journey#193
oratis merged 21 commits into
mainfrom
codex/desktop-preview-fixture

Conversation

@oratis

@oratis oratis commented Aug 1, 2026

Copy link
Copy Markdown
Owner

Summary

  • turn preview-app.html into a protocol-aware Tauri fixture using the production renderer bridge
  • add Playwright coverage for shell geometry, session resume, fast-turn buffering, approval/tool/usage completion, and Files source/diff/history
  • keep completed turns mounted while refreshing the sidebar and prevent tool results from rewriting older resumed turns
  • add a dedicated desktop browser journey CI job with failure artifacts

Validation

  • pnpm install --frozen-lockfile
  • pnpm typecheck
  • pnpm lint (0 errors; one pre-existing warning)
  • pnpm format:check
  • pnpm docs:check
  • pnpm test (953 passed, 12 skipped)
  • pnpm build
  • pnpm --filter @deepcode/desktop test:e2e (3 passed)

Stack

Depends on #192.

@oratis
oratis changed the base branch from codex/desktop-protocol-runtime to main August 2, 2026 06:36
@oratis
oratis marked this pull request as ready for review August 2, 2026 06:41
@oratis

oratis commented Aug 2, 2026

Copy link
Copy Markdown
Owner Author

Review: approved ✅

Real browser-level coverage for the desktop client, using the production renderer bridge rather than a hand-written mock. That distinction is what makes the fixture worth maintaining — a mock would drift from the bridge and pass while the real thing broke.

What I checked

  • Coverage spans the flows that actually regressed historically: shell geometry, session resume, fast-turn buffering, approval/tool/usage completion, and Files source/diff/history.
  • "Keep completed turns mounted while refreshing the sidebar" and "prevent tool results from rewriting older resumed turns" are both fixes, not just tests. The second is a genuine correctness bug — a late tool result landing on a resumed historical turn would visibly rewrite history in the UI.
  • Dedicated CI job with failure artifacts. Browser-journey failures are near-impossible to debug from a log line alone; screenshots/traces on failure are what keep this from becoming a flaky job people skip.

Maintenance note: this is the stack's only end-to-end UI gate, so it is also the one most likely to go flaky and get muted. Worth watching its first few weeks — a muted journey test is worse than none, because it still looks like coverage.

Validation: CI green including the new desktop journey job.

@oratis
oratis merged commit 3dabe15 into main Aug 2, 2026
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant