You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Upstream: Playwright's page tests in the testbed suite (#1505) #1060
Upstream commit: tester-army/e2e bdf769b (#1505), e2e-inc/e2e@bdf769b0 (8744 additions, 130 files: apps/testbed/app/playwright/** assets and the tests that use them, plus config). Per CLAUDE.md, port the tests and their testbed pages, as real-model or web tests where upstream uses a real browser. This is big: first list what the commit adds (assets, test files, it count), then port in slices (one PR per test file or group), each with Refs until the last one. Report which tests need features the port lacks (name each, and file an issue per missing feature).
What the commit adds. 130 files. apps/testbed/app/playwright/ holds Playwright's test assets (about 64 KB in total; the biggest file is a 6 KB png; chromium-linux.zip is a 325-byte stub), a LICENSE (Apache-2.0), a NOTICE, and routes.mjs (dynamic routes: /content, /echo, /status, /slow, /never, /redirect, /download, /respond, /cors-echo, /bounce, /count, /options, /upload, /binary, /download-stall, /sse, and a /ws upgrade). It adds 32 test files (about 496 it, tagged playwright, written against the browser fixture and the e2e public API). Smaller changes: a patternLabel step label for waitForURL, route, unroute, waitForResponse, and a translated(call, ...) timeout message (with a unit test in fixture-runtime.test.ts).
Size and licensing. No size problem and no big binary. The port is already Apache-2.0 and has a NOTICE. The assets need Playwright's LICENSE and NOTICE next to them. No decision needed. Plan: copy only the assets a slice uses, and add the license files in the first slice.
How the port hosts these. The port's testbed is tests/E2E.Agent.Tests/Testbed.cs (an HttpListener on a random port, linked into E2E.NUnit.Tests). Browser tests live in tests/E2E.NUnit.Tests/<UpstreamFile>/ (NUnit E2ETest, Browser, Screen). Upstream's fixed PREFIX (127.0.0.1:4271) and CROSS (localhost:4271) become the testbed's own port. Upstream's fail: 'reason' (an expected failure) has no port equivalent: a small .NET-only test helper keeps such a test with its upstream assertions and passes only while the body still fails.
route 40, request-continue 35, request-fulfill 16, request-fallback 6, wait-for-response 3, wait-for-navigation 9, network-idle 11 (need the routes.mjs routes and /ws)
120
E. Dialog and download
dialog 9, download 19
28
F. Screenshots
screenshot 20 (10 toHaveScreenshot goldens; check how the port stores goldens first)
20
Missing features. Upstream already left out every test whose feature e2e lacks (each file header has a "Not ported" list, for example: no force/trial/position options, no route(..., { times }), no fallback({ headers | url | method | postData }), no download.saveAs, no fullPage/clip screenshots, no element handles or Frame objects, no waitForEvent). The port mirrors e2e, so those stay unported and are named in each PR. Gaps specific to the port are not known until each file is ported. I check each slice and file an issue per gap (searching first). To verify first: the error texts of check/uncheck (group A), the fail: semantics, toHaveScreenshot goldens (group F), and the patternLabel / timeout-label source change (its own small PR with its unit test).
Slicing (one PR each, Refs #1060)
Slice 1 (starting now): the /content route and asset serving in the testbed, Playwright LICENSE/NOTICE, the fail helper, and check (13), wait-for-url (7), locator-list (1), selectors-text (2). 23 tests.
Rest of group A (one or two PRs: selectors; locator-*).
patternLabel and the timeout label (src change plus its unit test).
Group B in three PRs: fill + select-option + set-input-files + filechooser; click + click-scroll + elementhandle-scroll-into-view; keyboard.
Group C.
Group D in three PRs: route; request-continue + fallback + fulfill; wait-for-* + network-idle (with the routes.mjs routes and /ws).
All slices are merged. #1138 delivered the last one (upstream testbed edits from e2e-inc/e2e#1548). Known partials, as upstream: FrameHierarchy Partial_Should_send_framenavigated_when_navigating_on_anchor_URLs and Partial_Should_report_frame_from_inside_shadow_DOM need frame events and Frame objects, which the port does not have. The fixture-runtime test is tracked by #1120 (PR #1140).
Activity
kblok commentedon Nov 4, 2026
Plan for #1060 (upstream
bdf769b, e2e-inc/e2e#1505)What the commit adds. 130 files.
apps/testbed/app/playwright/holds Playwright's test assets (about 64 KB in total; the biggest file is a 6 KB png;chromium-linux.zipis a 325-byte stub), aLICENSE(Apache-2.0), aNOTICE, androutes.mjs(dynamic routes:/content,/echo,/status,/slow,/never,/redirect,/download,/respond,/cors-echo,/bounce,/count,/options,/upload,/binary,/download-stall,/sse, and a/wsupgrade). It adds 32 test files (about 496it, taggedplaywright, written against thebrowserfixture and the e2e public API). Smaller changes: apatternLabelstep label forwaitForURL,route,unroute,waitForResponse, and atranslated(call, ...)timeout message (with a unit test infixture-runtime.test.ts).Size and licensing. No size problem and no big binary. The port is already Apache-2.0 and has a
NOTICE. The assets need Playwright'sLICENSEandNOTICEnext to them. No decision needed. Plan: copy only the assets a slice uses, and add the license files in the first slice.How the port hosts these. The port's testbed is
tests/E2E.Agent.Tests/Testbed.cs(anHttpListeneron a random port, linked intoE2E.NUnit.Tests). Browser tests live intests/E2E.NUnit.Tests/<UpstreamFile>/(NUnitE2ETest,Browser,Screen). Upstream's fixedPREFIX(127.0.0.1:4271) andCROSS(localhost:4271) become the testbed's own port. Upstream'sfail: 'reason'(an expected failure) has no port equivalent: a small .NET-only test helper keeps such a test with its upstream assertions and passes only while the body still fails.Upstream test files (
itcount)it/contentand a few assetsroutes.mjsroutes and/ws)toHaveScreenshotgoldens; check how the port stores goldens first)Missing features. Upstream already left out every test whose feature e2e lacks (each file header has a "Not ported" list, for example: no
force/trial/positionoptions, noroute(..., { times }), nofallback({ headers | url | method | postData }), nodownload.saveAs, nofullPage/clipscreenshots, no element handles or Frame objects, nowaitForEvent). The port mirrors e2e, so those stay unported and are named in each PR. Gaps specific to the port are not known until each file is ported. I check each slice and file an issue per gap (searching first). To verify first: the error texts ofcheck/uncheck(group A), thefail:semantics,toHaveScreenshotgoldens (group F), and thepatternLabel/ timeout-label source change (its own small PR with its unit test).Slicing (one PR each,
Refs #1060)/contentroute and asset serving in the testbed, PlaywrightLICENSE/NOTICE, thefailhelper, andcheck(13),wait-for-url(7),locator-list(1),selectors-text(2). 23 tests.patternLabeland the timeout label (src change plus its unit test).routes.mjsroutes and/ws).Link the routes into the xUnit test project and pin the known differe…
Harden the upload and respond routes and write large test files in 1M…
Serve the testbed on localhost too and mark the two partial frame-hie…
Fail AllAsync as soon as the action fails (Refs #1060)
Assert the dialog handler ran and quote the download filename (Refs #…
Share the failure and respond helpers and answer bad /respond input w…
Read the network-idle timer from its task (Refs #1060)
Add the mouse helper script the input pages load (Refs #1060)
Answer a bad /respond status with 400 and clean the /download filenam…
Label a string waitForURL pattern through PatternLabel (Refs #1060)
Share the After and Record helpers and guard the held-response close …
Drop the unused Asset helper (Refs #1060)
Tighten the frame test assertions, skip CROSS when localhost cannot b…
12 remaining items
Port Playwright's fill, select-option, set-input-files and filechoose…
Quote a glob in a pattern step label, as upstream's patternLabel (Refs …
Merge main into issue-1060-input-pages-2 (Refs #1060)
Port Playwright's frame page tests (Refs #1060) (#1070)
Merge main into issue-1060-input-pages-3 (Refs #1060)
Run the nested-frame role tests as normal tests (Refs #1060)
Drop the known-difference markers of five Playwright tests (#1091)
Merge main into issue-1060-input-pages-3 (Refs #1060)
Port Playwright's route, request-fallback and request-fulfill tests (…
Leave out the digit assets: loaded, they move grid.html one pixel dow…
Leave the page screenshot port to the screenshot slice of #1060 (Refs #…
Port Playwright's dialog and download tests (Refs #1060) (#1078)
Port Playwright's request-continue, wait-for and network-idle tests (…
kblok commentedon Nov 5, 2026
All slices are merged. #1138 delivered the last one (upstream testbed edits from e2e-inc/e2e#1548). Known partials, as upstream: FrameHierarchy Partial_Should_send_framenavigated_when_navigating_on_anchor_URLs and Partial_Should_report_frame_from_inside_shadow_DOM need frame events and Frame objects, which the port does not have. The fixture-runtime test is tracked by #1120 (PR #1140).