E2E coverage: ReportContentEditor column-visibility toggles

[steilerDev]

[product-owner]

Component: E2E test coverage — ReportContentEditor column-visibility toggles
Priority: Should Have
Source: PR #1959, deliberately deferred at review time.

Problem

PR #1959 added column-visibility checkboxes above the summary table in ReportContentEditor (local state, no persistence). They shipped with no E2E coverage. The omission was deliberate: adding a test case changes Playwright's per-file test counts, which redistributes shard membership, and a worker-hash/shard-boundary change reddened an unrelated shard on PR #1961 the same day. On a branch gating the beta -> main promotion that risk was not worth taking, so the coverage was deferred here rather than rushed.

The feature is user-facing and has no automated guard: a renamed label, a removed checkbox, or an accidental switch to persisted preferences would all pass CI today.

Acceptance Criteria

• 1 e2e/tests/budget/reportWizardEditableContent.spec.ts (or a sibling spec) contains coverage that, on the summary tab: asserts every column checkbox is present and locatable by accessible name; unticks one column; asserts the corresponding <th> and every matching <td> are absent from the DOM; re-ticks it; asserts both return.
• 2 Coverage is complete against the component rather than a hand-picked subset: the test asserts the number of rendered column checkboxes equals the toggleable-column count the component defines, so adding a future column fails this test instead of silently going uncovered.
• 3 The toggles are asserted to be local state: toggling a checkbox issues no PATCH to /api/users/me/preferences (asserted by route interception or request logging). This pins the intended difference from DataTable's persisted column preferences and guards against someone adding persistence and inheriting the BUG: enabling two DataTable columns in a row silently hides the second one (column-preference save races its own re-sync) #1955 race.
• 4 Viewport scope is explicit: either the coverage runs at all three configured viewports, or a code comment in the spec states which viewports are excluded and why (e.g. the mobile card list exposes no column toggles). A silent desktop-only test does not satisfy this criterion.
• 5 Shard hygiene is handled explicitly on the implementing PR: the full E2E suite result is reported, and any shard that goes red is triaged as either (a) caused by this change or (b) a pre-existing latent hazard exposed by the shard redistribution — with the determination and its evidence recorded in a PR comment. Re-running until green does not satisfy this criterion.
• 6 If triage under criterion 5 identifies a latent cross-file hazard, it is linked to E2E hazard: cross-file shared-admin preference writes can silently flip a running test's state #1957 (or filed separately if out of that issue's scope) rather than worked around by pinning shard assignment.

Notes

• Adding tests here is expected to move shard boundaries. That is a known property of this suite, not a defect of this change — see E2E hazard: cross-file shared-admin preference writes can silently flip a running test's state #1957 for the latent cross-file isolation hazard that redistribution can expose.
• The toggles are presentation-only state in ReportContentEditor; nothing about them should reach the report content model or the generated PDF.

[steilerDev]

[product-owner]

Superseded by #1973 — closing (dated 2026-08-03).

This issue's premise has been reversed by a direct user requirement: "If i de-select a column in the preview it shouldn't render in the pdf". The column toggles are being wired through to the generated PDF, and the preview-only hint (sourceReports.editable.columnVisibilityHint) is being removed.

Two parts of this issue are now actively wrong rather than merely incomplete:

• The Notes state "nothing about them should reach the report content model or the generated PDF." That is the behaviour being deleted.
• AC1 asserts only DOM removal of the <th>/<td>. Under the new behaviour that assertion would pass while the PDF still contained every column — precisely the defect the user reported. See Report column visibility: wire the preview toggles through to the generated PDF #1973 AC 7.2.

Closing rather than rewriting, so the record of why the preview-only decision was reversed stays intact. Every still-valid ask here is carried forward into #1973 with attribution:

here	in #1973
AC1 (checkbox presence, DOM removal, re-tick)	7.1
AC2 (checkbox count == component's column count)	2.6
AC3 (no PATCH to preferences)	5.2 — rationale updated: the toggles now do reach the PDF, but still must not reach the server
AC4 (explicit viewport scope)	7.3
AC5 (shard hygiene, no re-run-until-green)	7.4
AC6 (link latent hazards to #1957)	7.5

Nothing is lost. Board status set to Wont-Do (superseded, not abandoned).