The "refresh does not break interacting with open drop-downs" is flaky. It attempts to verify that a refresh of the "Actions" list page does not occur when the user is currently navigating the dropdowns on the page, as htmx would replace their dropdowns mid-action. It does this by clicking on the dropdowns, verifying the dropdown text is visible, simulating a refresh, and verifying that the dropdown text is still visible.
Example failure log:
```
1) [chromium] › tests/e2e/actions.test.e2e.ts:173:3 › workflow list dynamic refresh › refresh does not break interacting with open drop-downs
Error: Timed out 3000ms waiting for expect(locator).toBeVisible()
Locator: getByText('Waiting')
Expected: visible
Received: hidden
Call log:
- Expect "toBeVisible" with timeout 3000ms
- waiting for getByText('Waiting')
7 × locator resolved to <a class="item" href="?workflow=test-dispatch.yml&actor=0&status=5">↵⇆⇆⇆⇆⇆⇆⇆Waiting↵⇆⇆⇆⇆⇆⇆</a>
- unexpected value "hidden"
194 | await expect(page.getByText('Failure')).toBeVisible();
195 | await simulatePollingInterval(page);
> 196 | await expect(page.getByText('Waiting')).toBeVisible();
| ^
197 | await expect(page.getByText('Failure')).toBeVisible();
198 |
199 | // Actor dropdown
at /workspace/forgejo/forgejo/tests/e2e/actions.test.e2e.ts:196:45
```
The dropdown list stops refreshes if there are any `[aria-expanded=true]` elements on the page:
ab6ea6a743/templates/repo/actions/list.tmpl (L31-L33)
But fomantic doesn't set the `aria-expanded` attribute immediately on mouseup; it delays for up to 100ms:
f8a332c2e6/web_src/js/modules/fomantic/dropdown.js (L241-L246)
The easiest fix is to align the test's expectations with the underlying code, which is to verify that the `aria-expanded=true` element is present in the test before proceeding with the simulated refresh.
Before this change, this test would fail as frequently as 10-out-of-10 in local runs, but usually more around 70% (albeit closer to 0% in a playwright ui). After this change, I've executed the test with zero failures in 100 back-to-back runs with `npx playwright test ./tests/e2e/actions.test.e2e.ts --project chromium --repeat-each 100`.
Reviewed-on: https://codeberg.org/forgejo/forgejo/pulls/9059
Reviewed-by: Gusted <gusted@noreply.codeberg.org>
Co-authored-by: Mathieu Fenniak <mathieu@fenniak.net>
Co-committed-by: Mathieu Fenniak <mathieu@fenniak.net>
- Make the "Actions" list (for example, https://codeberg.org/forgejo/forgejo/actions) dynamically refresh using htmx and partial page reloading. This addresses a pet peeve of mine, I find it common to end up on this page and have workflows in-progress, but not be able to monitor the workflows to success or failure from the page as it currently doesn't do any data refreshing.
- There are a few major risks involves with this change.
- Increased server-side load & network utilization. In order to mitigate this risk, I have configured the refresh to occur every 30 seconds **only** when the Page Visibility API indicates that the web page is currently visible to the end-user. It is still reasonable to assume this change will increase server-side load though.
- UI interactions on the page, such as the "Actor" and "Status" dropdown and the workflow dispatch form, would be replaced from the server with non-expanded UI during the refresh. This problem is prevented by stopping the refresh while these UIs are in their expanded states.
- E2E tests added.
Reviewed-on: https://codeberg.org/forgejo/forgejo/pulls/7361
Reviewed-by: Gusted <gusted@noreply.codeberg.org>
Co-authored-by: Mathieu Fenniak <mathieu@fenniak.net>
Co-committed-by: Mathieu Fenniak <mathieu@fenniak.net>
As per https://codeberg.org/forgejo/forgejo/pulls/6400, the after hook runs for every test, resulting in duplicated screenshots.
Not all tests are supposed to generate screenshots, especially because they could be flaky (also see 206d4cfb7a ).
Additionally, the implicit behaviour might have caused confusion, so we now create screenshots explicitly, adding the statements from the tests that already generated screenshots.
Replaced manual login and context loading across tests with Playwright's `test.use` configuration for user authentication. This simplifies test setup, improves readability, and reduces repetition.
For #6362
Reviewed-on: https://codeberg.org/forgejo/forgejo/pulls/6400
Reviewed-by: Gusted <gusted@noreply.codeberg.org>
Co-authored-by: Julian Schlarb <julian.schlarb@denktmit.de>
Co-committed-by: Julian Schlarb <julian.schlarb@denktmit.de>
- create screenshots on tests that create a custom page
- skip screenshot on a page that uses the default page (it is added
automatically and already exists)
- replace a useless response wait and assert the correct page style
instead
Goals:
- speedup
- less flakiness
- best practices and more use
- documentation
config:
- sync ports in Makefile and playwright config
(otherwise, some tests fail locally because they assert the full URL including the (wrong) port)
- even more generous timeouts
- limit workers to one again (because I finally understand how
Playwright works)
- allow nested functions to group them together with the related test
all:
- deprecate waitForLoadState('networkidle')
- it is discouraged as per https://playwright.dev/docs/api/class-page#page-wait-for-load-state
- I could not find a usage that seems to require it actually (see
added documentation in README)
- adding an exception should be made explicitly
- it does not do what you might expect anyway in most cases
- only log in when necessary
webauthn:
- verify that login is possible after disabling key
- otherwise, the cleanup was not necessary after the previous refactor to create a fresh user each
issue-sidebar / WIP toggle:
- split into smaller chunks
- restore original state first
- add missed assertion to fix race condition (not waiting
before state was reached)
- explicitly toggle the state to detect mismatch earlier
issue-sidebar / labels:
- restore original state first
- better waiting for background request
