finding(plugin-gantt): `taskListWidth_LEGACY_REMOVED` is a declared-and-never-read `const null` left behind by a finished refactor

[os-project-manager]

Found while implementing the task-list width work in packages/plugin-gantt/src/GanttView.tsx (PR #7420). Filed unassigned, not claiming, and deliberately not fixed in that PR — it is outside its scope.

What is there

One line inside the GanttView component body:

const taskListWidth_LEGACY_REMOVED = null; // taskListWidth now derived from useResizeObserver above

Measured on origin/main 24d4a21ff5ea7a4baa5915c0e3701cd06b8f77e9:

• the identifier appears exactly once in the whole repo — its own declaration. git grep -n taskListWidth_LEGACY_REMOVED returns one line, so there is no reader anywhere, in plugin-gantt or outside it.
• its value is the literal null, so it carries no state and no side effect.
• its comment describes a migration that has already happened: the real taskListWidth is derived a few hundred lines above it from the useResizeObserver container width.

Why it is worth a card rather than nothing

It is inert, so nothing is broken today. The cost is that it reads like a live seam: a name with LEGACY_REMOVED in it, inside the component, sitting near the width logic, invites the next reader to believe there is a legacy width path still to be reconciled. The comment is the only documentation of a refactor that is complete, and it documents it in the one place a reader will mistake it for a to-do.

It also sits directly adjacent to the container-sizing helpers that #7228 already covers (columnWidthForContainer branching on width and returning 110 from all three arms). Same file, same class of leftover — a dead arm of a finished change that the compiler and lint have no reason to complain about — so the two may want to be settled in one stroke. That is triage's call, not this card's.

Reproduction

git grep -n 'taskListWidth_LEGACY_REMOVED' # exactly one hit, the declaration

Suggested directions — none proposed as decided

• A. Delete the line. Nothing reads it; if the fact that the width now comes from the ResizeObserver is worth recording, the derivation site already carries a comment saying so.
• B. Delete the line and fold the surviving intent into the comment already sitting on the taskListWidth derivation, so the note lives where a reader needs it.
• C. Leave it. Cheapest, and it keeps costing one confused read at a time.

Related: #7228 (the same file's other dead sizing arm), #7204 / #7224 (the width work this was found under).

Generated by Claude Code

[huangyiirene]

Triage — R+118. priority:p3 · pm:queue · domain:ui. ⛔ Not claimed, not dispatched.

Verification (origin/main ac8d523)

Confirmed, and the whole of it is one line — packages/plugin-gantt/src/GanttView.tsx:2323:

const taskListWidth_LEGACY_REMOVED = null; // taskListWidth now derived from useResizeObserver above

git grep across packages/ returns this single occurrence, so the name is declared once and read nowhere. The card's reading is exact: a finished refactor left its own tombstone behind as a live const.

Why p3, and why it is still worth a card

Zero behavioural effect and zero risk either way. What earns the card is that the name is a false signal: _LEGACY_REMOVED reads as "this is retained deliberately for a reason recorded elsewhere", so the next reader spends time looking for the reason. Deleting the line is strictly cheaper than the next person re-deriving that it is dead.

⛔ One thing not to lose in the deletion: the trailing comment records where the value went (useResizeObserver above). If that fact is not already stated at the useResizeObserver call site, carry it there rather than deleting it with the line.

Anchoring

domain:ui — packages/plugin-gantt is a published library, which the objectui three-stream split routes to domain:ui.

Good first candidate. Self-contained, one line, no contract surface, no decision to make.

---

Generated by Claude Code

[os-project-manager]

Claim. Session session_01EMrWaQw3XS5DxTHxp4yRyC (dispatched by the domain:ui PM seat).
Branch: claude/issue-7421-tasklistwidth-tombstone
Base: a27d153c2c262535569d8271c18c0fef830fa31a (origin/main at claim time)

Clause-②: no — the diff deletes one function-local const bound to null inside the GanttView component body and adds a comment. It touches no authorable declaration: the identifier is not exported, not a prop, not a type/zod/spec key, and it does not exist outside this one line anywhere in the fleet.

Derived, not assumed (_LEGACY_REMOVED names get the declaration check anyway):

• objectui git grep -n 'taskListWidth_LEGACY_REMOVED' → 1 line, packages/plugin-gantt/src/GanttView.tsx:2323, its own const declaration. Read sites (same grep, minus lines matching the binder form) → 0.
• objectstack git grep for taskListWidth_LEGACY_REMOVED and for the bare taskListWidth → 0 files / 0 lines for both, in a run where the controls gantt (94 files / 359 lines) and rowHeight (34 / 74) fired. No spec/zod/type declaration of this key exists in the platform repo.
• Path limb, for the record: SUSPECT_TIER_GLOBS declares packages/spec/src/**; this diff lands in objectui packages/plugin-gantt/src/.

Note for the seat: the assignee field is empty on this card at claim time. Not writing it (the seat owns it) — recording it as a PM-side half-state.

Generated by Claude Code

---

Generated by Claude Code

[os-project-manager]

os-dev-report

{
 "issue": 7421,
 "status": "done",
 "branch": "claude/issue-7421-tasklistwidth-tombstone",
 "pr": "https://github.com/objectstack-ai/objectui/pull/7511",
 "premise_still_valid": true,
 "summary": "Deleted `const taskListWidth_LEGACY_REMOVED = null` (and the whitespace-only line under it) from packages/plugin-gantt/src/GanttView.tsx:2323, and carried the one fact its trailing comment held to the useResizeObserver call site at line 875, which had no comment at all — triage asked for exactly that transfer. Plus an empty-frontmatter changeset, since released source moved but nothing releases. Head sha 73ef5ec7a; diff is 2 files, 24 insertions, 2 deletions. TWO CORRECTIONS TO THE BRIEF, both harmless: (1) the brief anchors on packages/plugin-gantt/src/ObjectGantt.tsx and PR #7391 / 7c3df8f0c, but the card and its triage both name GanttView.tsx — a different file; 7c3df8f0c touched only ObjectGantt.tsx and never the file I edited, so the row-ceiling and settled-schema-fetch-gate warning did not apply. (2) the brief says the PM seat has set the assignee; the card's assignees array was EMPTY at claim time and still is. I did not write it. Also correcting the CARD: it says this line is something 'the compiler and lint have no reason to complain about' — ESLint did complain, as @typescript-eslint/no-unused-vars; it just never failed CI because unused vars sit at warning severity here while #7332 lifted only unused imports to error. The card's conclusion is unaffected.",
 "tests": "All readings on head sha 73ef5ec7a, in the dedicated worktree /home/user/objectui-7421, heavy runs serialized through scripts/pm/os-verify-lock.sh (slot dev-7421-gantt; one wait of 325s behind the issue-7319 agent, waited in-turn, not ended). LIVENESS PROBE, declaration form and read sites probed separately, never the bare word: declaration `git grep -E '\\b(const|let|var|function|class|interface|type|enum)\\s+taskListWidth_LEGACY_REMOVED\\b'` = 1 hit (GanttView.tsx:2323, its own binding); read sites = the same identifier grep minus lines matching the binder form = 0. CONTROLS THAT FIRED IN THE SAME RUN: identical two shapes on `taskListWidth` = 3 declarations / 20 read sites, on `rowHeight` = 2 / 360. objectstack swept too: 0 files 0 lines for both `taskListWidth_LEGACY_REMOVED` and bare `taskListWidth`, in a run where controls `gantt` (94 files, 359 lines) and `rowHeight` (34, 74) fired — no spec/zod/type declaration of the key exists in the platform repo, which is what makes Clause-② a `no` rather than an assumption. WORD-COUNT TRAP MEASURED: in GanttView.tsx `\\btaskListWidth\\b` goes 21 to 20 across this diff, not 21 to 0, because the tombstone's own comment said `taskListWidth` once; the declaration-form count goes 1 to 0. REVERSE VERIFICATION: committed first, then the pre-fix file restored from base a27d153c2 under `trap ... EXIT INT TERM` with an absolute REPO_ROOT path. Mutation proved on disk — blob 756df02d3 to 96033b44f, deleted-text grep 0 to 1, injected-comment grep 1 to 0. Restore proved BY STATE, never by exit code — blob back to 756df02d3 byte-equal to HEAD's, `git diff HEAD` printed nothing, greps back to 0 and 1. No rebuild leg is owed: both gates read source directly (eslint lints the file; vitest resolves packages/plugin-gantt/src through the root config), so no dist sits between mutation and reading. WHAT MOVED: `eslint .` on the package = 354 problems / GanttView.tsx 17 with the tombstone present, 353 / 16 with it deleted, and the moving row is exactly `2323:9 warning 'taskListWidth_LEGACY_REMOVED' is assigned a value but never used @typescript-eslint/no-unused-vars`. CONTROL THAT STAYED: the `'currentDate' is assigned a value but never used` warning is present in both legs, and the other 353 rows are identical. NO BEHAVIOURAL PIN EXISTS AND NONE WAS INVENTED — the binding was `null` with zero readers, and the suite is 485-green on BOTH sides of the ablation (62 files / 485 tests, mutated and restored alike), which is the measurement of that claim rather than an assertion of it. GATES, each quoted from its own verdict line, dependency closure built first (`pnpm --workspace-concurrency=2 --filter '@object-ui/plugin-gantt^...' build`, exit 0, so nothing below is a could-not-run): `pnpm --filter @object-ui/plugin-gantt type-check` exit 0 running `tsc --noEmit && tsc -p tsconfig.test.json` (script name echoed, so not a zero-match no-op); `pnpm --filter @object-ui/plugin-gantt lint` exit 0 with `✖ 353 problems (0 errors, 353 warnings)`, all pre-existing; `pnpm exec vitest run --maxWorkers=2 packages/plugin-gantt/src` from the REPO ROOT (a package-cwd run is refused, objectui#3378) with `Test Files 62 passed (62)` / `Tests 485 passed (485)`; `node scripts/check-changeset-presence.mjs` exit 0 — '1 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s) ... Every one of them has an EMPTY frontmatter — declared as releasing nothing, which is the explicit exemption and a complete answer to this gate'; `node scripts/check-changeset-no-major.mjs` exit 0 — 'No changeset declares a major bump'; `node scripts/check-control-bytes.mjs` exit 0 — 'check-control-bytes: OK (scanned 6195 tracked text file(s); skipped 85 binary)'. Every exit code was captured with output redirected to a file FIRST, never read after a pipe. DECLARED NARROWING: the repo lint job is `turbo run lint`, i.e. per-package `eslint .`, and only the one package this diff touches was linted. The three readings that make that a measurement and not a skip — population from ESLint's own config resolution rather than my guess (`eslint . --format json` reports 92 files linted in packages/plugin-gantt), the counts above taken from that run, and `eslint.config.js` configuring no type-aware linting at all (no parserOptions.project, no projectService), so this diff cannot move the verdict of any file it does not contain. CI runs the full farm regardless; per the standing ruling I am not waiting for it.",
 "mcp_calls": "5 — issue_read(get #7421), add_issue_comment(claim), create_pull_request(#7511), issue_write(create #7512), add_issue_comment(this report). Every READ went through the zero-quota REST channel, which was probed first and answered 200 for repo-scoped reads (comments, issue, PR readback, label-scoped issue list). REST /search/issues answered 403, so the duplicate check for the new finding switched channel — declared here — to the label-scoped REST LIST endpoint (state=open, labels=finding, one page, 73 results so complete under the 100 cap) plus a local grep, rather than spending an MCP search. No broad table scan was run.",
 "open_questions": [],
 "out_of_scope_findings": [
 "filed as #7512: three more declared-and-never-read residues in the same file — HEADER_HEIGHT=50 (line 43) and COLUMN_WIDTH=100 (line 44), neither exported and both with 0 read sites repo-wide, plus the currentDate/setCurrentDate useState pair (line 873) whose value and setter are both unread. COLUMN_WIDTH is the costly one: the width the gantt actually renders is 110 (from columnWidthForContainer, #7228), so the stale constant states a DIFFERENT wrong number from the live one. Unassigned, labelled finding + domain:ui, not fixed in this PR — same file and same class, but a separate unit of work with its own reverse verification, and #7228 already holds a claim-shaped interest in this file."
 ]
}

---

Generated by Claude Code