Repository navigation
fix: bugs from the frontend review — wrong runs in bulk actions, lost ignore areas, stale answers - #438
Merged
Conversation
… ignore areas, stale answers Test details - the left arrow on the first run no longer jumps to the last one - A/X hotkeys use the current "go to next" setting and unsaved-areas check - a failed ignore-area save keeps the areas marked unsaved - an update to the open run (a comment, another reviewer) no longer wipes drawn, unsaved ignore areas - applying an area to the whole build reports a failure instead of hanging - keydown listeners on the canvas no longer pile up (Delete ran twice) Lists and loading - the run selection is cleared when another build opens; delete, download and clear-ignore-areas act only on selected runs that are shown - the zip names images by test and tags and never overwrites one - a late answer for a run, a build or another project's build list no longer replaces the current one; a deleted run/build link closes cleanly Smaller - a non-JSON error (proxy HTML page) still logs out on 401 and shows a message - login after an expired session returns to the full link, query included - a role the server rejects reverts in the user list - no trailing space in VRT_PROJECT on the profile page - a failed login right after registering shows the error Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…iew-improvements-54c2a1
…e effect flush Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
3 of 4 tasks
pashidlos
added a commit
that referenced
this pull request
Oct 24, 2026
- handleResponse: a successful response that is not JSON goes through the one rejection path instead of a second Promise.reject - applying an ignore area to the whole build is one flat promise chain - zip entry numbering uses a while loop over the name it tests
pashidlos
added a commit
that referenced
this pull request
Oct 24, 2026
… list and switcher (#442) * build(lint): run SonarCloud's rules locally with eslint-plugin-sonarjs SonarCloud only looks at a PR once it is open, so its findings arrived after review (#438 → #439). eslint-plugin-sonarjs 4.2.2 runs the same rules in `npm run lint`, the pre-commit hook and CI. The 21 findings in the current code: - window.open for "Open an issue in GitHub" now passes noopener,noreferrer, so the opened page gets no handle on the app's tab (test first) - three nested ternaries and two one-case switch reducers rewritten as plain ifs; a float constant compared with toBeCloseTo - todo-tag is a warning: the TODOs are debt to see, not a failed build - off for Playwright specs only: assertions-in-tests (specs assert through helpers and page objects) and void-use (`void fixture` asks for a fixture used only for its setup) * feat(projects): favourites, recently opened and search in the project list and switcher A long project list was one unsorted grid, and the switcher on the build page a plain select. Now: - a star on each card marks a favourite; favourites are kept per user by the API (GET /projects/favorites, PUT/DELETE /projects/:id/favorite) and listed first, under "Favourites" - the last five projects opened are listed next, under "Recently opened", kept in this browser per user - the rest follow by name; the order is the same in the list and switcher - a search field filters by name, id or main branch, keeps the query in the URL (?q=) and takes focus on "/" - the switcher on the build, variations and profile pages is an autocomplete with the same groups and search - Builds and Variations are router links: they used to reload the whole app An API without favourites (older backend) answers 404: the list then works without them instead of greeting every login with an error. The star goes back and the error shows if saving it fails. ProjectCard, ProjectGrid and ProjectSearchField keep the page under 300 lines. The e2e sign-in fixture answers no favourites by default; new specs cover starring, the order, search, "/", the switcher and navigation without a reload. screens.spec gains the grouped list, a search without matches and the open switcher; baselines regenerated in Docker. * test: raise the coverage floors to the reached level * fix(projects): keep the switcher where the select was; address SonarCloud on #442; wait before legacy screenshots - The project switcher had a margin above its label, which the select it replaced had not: the label, the name and everything under them sat 8px lower on the build, variations and profile pages. Only the sides and the bottom keep the inset now, and the layout matches master again. - SonarCloud: the two favourite mocks drop an async they never awaited, and the lint config's comment no longer reads as an open task. require-await (Sonar's S7503, outside eslint-plugin-sonarjs) runs as a warning, since older tests and mocks have 40 such functions. - The per-spec "renders" tests of six pages took their screenshot before the lazy route had rendered, so they flaked (on master too) and an update run wrote the empty page as the baseline. They wait for the page's heading and for the network to settle first. Baselines regenerated in Docker; those within the 100px tolerance of master are master's again.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What and why
A review of the frontend turned up bugs where a reviewer could lose work, act on runs they could not see, or see a run other than the one in the URL. Each was checked against the code first; each fix starts with a test that failed before it (38 new tests). Security, socket reconnect, cross-tab logout and performance were left out on purpose: none of them is a small change.
Losing work or acting on the wrong runs
Navigation and stale answers
Array.at(-1)); a run outside the list no longer moves with the arrows.Smaller
?buildId=&testId=).VRT_PROJECTon the profile page; a failed login right after registering shows the error.Also: the tag field list moved to
testRunTags.helper(shared by the list and the zip names); coverage floors raised to the reached level (lines 52 → 73).Not covered: SSO login still lands on the home page rather than the link the user asked for (the target would have to survive the redirect to the provider); a failed
fetchstill reaches callers as a rawTypeError.How it was checked
Merged with master after #437, so everything runs on React 19.
npm run lint(0 errors) andnpm run typechecknpm test: 602 pass, coverage 73% statements / 74% linesnpm run test:pw:docker: 285/298 in the full run on a host with load average ~60; the 13 failures were timeouts opening pages and the details dialog. Rerun of those four specs with 2 workers: 68/69; the remaining one (zoomToolbar"shows the screenshots at their own size", the?testId=dialog-open timeout build(deps): React 19, with react-joyride 3 and react-konva 19 #437 describes on master) then passed 10/10 with--repeat-each=10. No screenshot baselines changed.Screenshots
Nothing visible changed.
🤖 Generated with Claude Code