docs(migrate): match the Playwright mapping to what e2e does
Merged
OK
13d ago+46+46+01234567890123456789-33-33-01234567890123456789
What and why
docs/migrate/playwright.mdx claimed parity in rows that the ported Playwright tests in apps/testbed/tests (#1505) disprove or leave out. Each row below now matches main, checked against the source and the tests' fail/skip marks and "Not ported" headers.
- Runner and CLI rows (
workers,retries,trace,video, hooks,test.skip,test.setTimeout,test.use, CLI flags,-u) go fromsametorenamed/missingand name their limits. test.fail()is nowsame. The row notes that members of a serial group and setup tests can't be marked (keeps main's #1539 row).inputValue()on an element that isn't an input reads"", where Playwright throws (packages/e2e/src/locator/screen.ts:562).uncheck()on a checked radio,role=radioincluded, isNOT_ACTIONABLE(packages/web/src/actions.ts:59).- The 1 s hold of a negated matcher is capped by its timeout.
?in abrowser.routeglob matches one character, where Playwright reads it literally.route.fulfillbodies are strings only, and each header takes one value, so oneset-cookie. Aroute.continueheader can't be deleted (packages/web/src/route-options.ts).{ timeout: 0 }:INVALID_ARGUMENTonwaitForResponse,waitForDownload, andtoHaveScreenshot. Any other call times out at once.- An element inside a
visibility: hiddeniframe reads as visible. goto('data:')is nowby design. A new "What differs on purpose" bullet states the navigation rule:http,https, andabout:blankonly (packages/e2e/src/internal/urls.ts:137). That bullet replaces the undefined "navigation rule" reference.- No
about:blankpage before the firstgoto. browser.evaluateand locator reads run in the page's main world.
Rows describe how main behaves today. Open PRs that close gaps (clearing file inputs, binary bodies, header deletion, several set-cookie, hidden iframes, isolated world) update their own rows.
Verified
Ran it locally: yes
pnpm docs:check: 70 pages compile as MDX, and 1018 anchors resolve.- A fresh-context reviewer checked every changed row against
packages/*/srcand the testbed marks. It found two false claims, both now fixed:test.skipinafterEach, andtimeout: 0on everybrowsercall.
Checklist
- Changeset added (
pnpm changeset) for any change to a published package, or the change is testbed, docs, or CI only. - Docs updated in the same PR.