Your PR, reviewed twice at ce68347: Cody's process and ours, each by a fresh agent, then merged and re-checked. Nothing was posted to GitHub.
Preview:rancheri19854r, built from ce68347, backend the mf Rancher (v2.16 head). Baseline: dashboard-master at 55438bb, same backend. Filter namespace:i19854-group-header, then View > Group By > Description.
Where it stands: the issue is fixed, one regression to fix first
Everything the issue asks for in 2.16.0 works: descriptions show as typed on ConfigMaps and Projects/Namespaces, nothing asks for x, and the Namespace, Name and Age headers look as before.
CI: unticked checklist boxes, the code-freeze gate, a11y-test on a 503
Checklist only
CI notes
FYI
Pass A (Cody's process)
Pass B (our process)
Verdict
Comment, 1 comment filed
Ready after S1
Checked by both
The issue's case on ConfigMaps, master vs PR (markup shown as typed, no request for x). Namespace and Age groupings unchanged. The API Key regression, each with its own throwaway Linode/DigitalOcean credential. The new test failing without the fix.
Only one pass did
Before/after demo videos. A sourced acceptance table. Typed entities: <b> in a description now shows as typed where master decoded it to <b>. Screen reader view: master exposed an unnamed image, the PR only the text.
Projects/Namespaces, a list with its own heading slot. The Name grouping. A mutant of the month branch (equivalent). A sweep of every value that can reach the changed line and every model getter or formatter that returns HTML. Lint, type-check and the CI results. A coverage grid.
The PR escapes every View menu grouping's value, which assumes a column value is plain text. The Cloud Credentials list's API Key column isn't: its value is the model's publicData, which is already markup (the key escaped, plus … for DigitalOcean and Linode), and its CloudCredPublicData cell renders it as HTML. So the header escapes it a second time. Both passes reproduced it with a throwaway credential: master's header reads i19854pb… like the cell, the PR's reads i19854pb….
master: the header matches the cell.PR: the header shows the entity as text.
From the same cause, read from the code only: a full-mode key (AWS, Azure) holding & would show &. I re-checked Pass B's sweep: in shell, publicData is the only value handed to a column as markup. The two formatters that render HTML are CloudCredPublicData and ServiceTargets, and the second builds its markup inside the formatter, not in the value. The new test only covers a plain value.
2
No 2.17.0 issue yet for cutting long values short
Found by Pass A (Cody)Found by Pass B (ours)
Your call
Richard asked for the truncation to be spun out to 2.17.0, and the PR body says so, but nothing is filed. Fixes #19854 closes the issue on merge, and with it the only record of the 389px header. Both passes measured it unchanged on master and the PR.
3
CI: three red checks, none from the code
Found by Pass B (ours)
FYI
Description: three checklist boxes are unticked (reviewer assigned, Accessibility, Global Roles), and the gate wants all of them.
Branch Milestone: the milestone matches (v2.16.0). It fails on "the target branch 'master' is currently in code freeze", so it needs the maintainers' release-blocker exception.
a11y-test: Rancher answered 503 for 59 tries at startup. Infrastructure, so a rerun should clear it.
lint, type-check, unit-test, i18n, the build and all 18 e2e jobs pass.
Pass A: Cody's process
A fresh agent followed my-pr-full-review from codyrancher/ai-skills at c04c11a, mapped to our setup: the PR preview for his dev server, dashboard-master (rebuilt today at 55438bb) for his master build, the mf Rancher behind both, local files instead of his harness, our recorder for the demos.
Before this change: master.With this change: the PR.
Acceptance, from issue 19854 and its comment
Criterion (source)
Verdict
AC1
The header shows <b>bold</b> <i>italic</i> <img src=x> end as typed (issue, Expected result)
Met
AC2
No request for x (issue, step 6 and Actual result)
Met
AC3
Fixed in 2.16.0 (Richard's comment)
Met: milestone v2.16.0 on master
AC4
Namespace and month headers unchanged (implied)
Met
-
A long description cut short in the header (issue, Expected result)
Not in this PR: Richard moved it to 2.17.0; still 389px on both builds
Base review: no bug in the changed lines. The new test passes, and fails with master's ResourceTable.vue.
Simplify: already simple. Skipping the escape on the month label would be taste.
Side effects: finding 1. Extension group-row slots and Cypress page objects checked and dropped.
Component reuse, design fidelity: nothing in scope (no new component, no design).
Accessibility: no findings; the PR removes the unnamed image master exposed.
Verify, refine, condense, preexisting: finding 1 holds and is new with the PR (master's header is right).
Pass B: our process
A separate agent that didn't see Pass A. Verdict: ready after S1. No blockers, 1 suggestion (finding 1, evidence live), no nits, 1 question (finding 2), plus the CI notes (finding 3).
Live, PR vs master: ConfigMaps by Description (markup, & and quotes, the long one, none), Namespace, Name, Age by month; Projects/Namespaces by Description with a throwaway namespace (its own heading slot gives way, so the fix applies there too); Cloud Credentials by API Key.
Tests: 10 ResourceTable suites pass. Removing the escape from the column branch fails the new test; removing it from the month branch changes nothing, since a month label never holds markup.
Code: every caller with its own #group-by, every reader of group.ref, every kind of value that reaches the changed line, every model getter or formatter returning HTML.
Coverage grid: both changed files cleared on all 7 dimensions apart from finding 1.
Also seen
Shared state: Pass B's visit to Cluster Management moved admin's cluster preference to _ and back. At the end all 27 admin preferences matched the backup taken before the passes, and no table view was saved.
Projects/Namespaces opened with test already in its query box on both previews. Neither pass touched it.
Throwaways (two cloud credentials, a namespace, a ConfigMap) are deleted. The i19854-group-header ConfigMaps stay for testing while the PR is open.
What I'd do
It's your PR, so nothing to post as a review. Fix 1 in this PR before it's merged, since it's the one place the change makes a header worse than master. The cleanest fix is at the source: publicData returns plain text (the … character, no escaping) and its cell shows it as text. Then every column value is plain text and the header's escaping holds everywhere, with a test for that value. Then file the 2.17.0 issue for long values, tick Accessibility and Global Roles (both passes covered them, and the change doesn't depend on the role), ask Richard to review, and rerun a11y-test.