[vishnuv688]

What & why

A Selenium mocha trace showed tests with no rows after a driver.wait(): no getText, no assert.equal, not even for the failing one. The trace's own stdout events had the real story, Wait timed out after 10054ms: example.com no longer serves an <h1>, so driver.wait(until.elementLocated(By.css('h1'))) timed out and nothing after it ran. Capture had not regressed, but the failure was invisible, because a failed wait() produced no row at all. Looking into it turned up three adapter bugs, fixed here alongside the example.

• Example: the mocha example and its README snippet read the h2 of the-internet's login page, which the other examples already drive.
• A failed driver.wait() gets a red row. The patched wait() returns a promise derived from selenium's that records the failure and rethrows it, re-wrapped in selenium's own WebElementPromise for an element wait, so an awaited failure rejects with the same error, an unawaited one still surfaces as unhandledRejection, and driver.wait(until.elementLocated(...)).getText() still chains. The failure turns the wait's own poll row into a wait row named after the condition (Waiting for element to be located By(css selector, h1)), mapped to the same trace method as WDIO's waitUntil, so it shows in the Actions list, the Errors tab and the trace.
• A wait's polls fold only into that wait's rows. Each wait() call carries its own scope, and a poll folds only into a row of the same scope. Before, the fold matched any earlier row by command and arguments, so a helper's wait could overwrite the user's own findElements row. Timer polls also never folded at all, because the patcher reported core's 'unknown:0' placeholder instead of no call source.
• The call site is right under source maps. Under tsx or vitest the adapters' own frames are named by their src/ paths, so every row's callSource pointed at core/src/stack.ts. Core now locates this monorepo's packages/ directory from the nearest pnpm-workspace.yaml and treats its src/ files as library code; an adapter installed from npm or bundled into a user's build finds no such workspace, so a user's own src/ stays user code. A malformed frame can no longer throw, and no user frame now means no call source rather than 'unknown:0', which fixes the same fold for Nightwatch.
• The first test keeps its rows under --require. The plugin loads before mocha publishes its globals, and the hooks were registered by a 100 ms poll, so a first test that started sooner lost every command. The package now exports mochaHooks, mocha's root hook plugin entry, which mocha reads before any spec; importing the plugin from a spec still registers through the globals.

Type of change

• Bugfix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Polish (an improvement to an existing feature)
• Breaking change (existing behavior changes for users)
• Documentation
• Internal (build, CI, dependencies, tooling)

Packages touched

• shared (types and contracts)
• core (framework-agnostic capture/reporting)
• elements (published element/snapshot API — @wdio/elements)
• service (WebdriverIO adapter)
• nightwatch-devtools (Nightwatch adapter)
• selenium-devtools (Selenium adapter)
• selenium-devtools-py (Selenium Python adapter)
• backend (server)
• app (UI)
• script (page-injected runtime)
• trace (Trace mode)

Notes for reviewers

• Why the call-site change is in core: getCallSourceFromStack is core's, and the service, Selenium and Nightwatch all inline it, so all three get a patch changeset. Nightwatch's change is test fixtures only. CommandLog.callSource was already optional, so the wire contract and the app are unchanged. The 'unknown:0' guards in trace-sources.ts and the backend's baseline attribution stay, because stored rows and zips from older adapters still carry the placeholder.
• RetryTracker: with folding scoped to the wait, its bare-signature fold (isRetryIgnoringSource) had no user left and is removed.
• Tests: each fix has regression tests that fail without it: a helper's wait cannot take over the user's row; a failed wait rejects with the same error, still raises unhandledRejection when unawaited, and keeps element chaining; source-mapped stacks, a host-ful file:// frame, the monorepo, an npm install and a user bundle; the first test starting before the globals poll. pnpm test 2840 passed, pnpm lint 0 errors, pnpm typecheck clean, and the dist leak grep is empty for selenium-devtools, nightwatch-devtools and service.
• Real browser: not run yet. To check: pnpm build && pnpm --filter @wdio/selenium-devtools example:mocha, then open the trace; the heading tests show Element.getText() and the assert.equal rows, and every row's call site points into example.js. Pointing a selector at something missing shows the red wait row.
• Docs: the website's Selenium page carries the same example.com / h1 snippet; a webdriverio docs PR updates it.

Screenshots / recordings

n/a (no UI change)

🤖 Generated with Claude Code

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no new blocking issue was found.

Summary

The PR makes failed Selenium waits visible, keeps separate calls on separate rows, fixes source locations, and registers Mocha hooks before the first test.

• A failed Selenium wait gets a named row in the trace.
• Command call sites point to user code under source maps.
• Mocha captures the first test when the plugin loads with --require.
• The Selenium example reads the login page’s heading.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[driver.wait] --> B[Choose wait scope]
  B --> C[Run condition]
  C --> D[Match each call to its row]
  D --> E[Update only that wait's rows]
  C --> F{Wait fails?}
  F -->|Yes| G[Replace latest poll row with failed wait]
  G --> H[Rethrow original error]

Loading

_{Reviews (4) · Last reviewed commit: "fix(selenium-devtools): tell nested wait..." · Reviewed by Greptile}

[vishnuv688]

The new conformance rules in #466 guard this fix. With every source file this PR changes reverted, Selenium fails in all three BiDi modes: failed-test-has-error-row (the timed-out wait left no failed action), call-site-names-the-scenario (a poll row named unknown:0), and tests-end-as-designed / one-row-per-browser-action (the first test lost its rows). The other adapters stay green. With this PR restored, the suite passes 15/15.