Files
fusion/packages
gsxdsm 890d588891 Phase B — Column.tsx 2→0 and taskActivity.ts 2→0 (U11's cluster to zero) (#2636)
**Claimed:** `Column.tsx`, `taskActivity.ts` — both to zero.

| file | before | after |
|---|---:|---:|
| `packages/dashboard/app/components/Column.tsx` | **2** | **0** |
| `packages/dashboard/app/utils/taskActivity.ts` | **2** | **0** |

## `Column.tsx` — two different fixes, because the two sites are
different problems

**The preserve-progress prompt** routed through
`isPreImplementationColumnRole`. This is the *same* question that helper
was written for — ListView asks it about a move target, Column asks it
about itself — and the degraded id sets are identical (`{todo,
triage}`), so the consolidation is exact.

I verified the sets matched **before** consolidating, because the
sibling case is not interchangeable: `isPreExecutionHoldColumn` in
`TaskContextMenu` drives the Plan affordance and its degraded set is
`{triage}` alone. Routing *that* through this helper added `plan` to
flagless `todo` cards, caught by an existing test. **Same shape,
identical trait path, non-interchangeable fallbacks.**

**The legacy-board arm** (`workflowMode || column === "triage"`) —
deleted, on the third attempt.

I deleted it twice before and reverted both times because four Column
tests render without `workflowMode`. That was the delete-only rule
working, but **my conclusion from it was wrong**: a behaviour change
means the branch was not dead *for those callers*, and the callers are
**fixtures, not production**. Board is Column's only consumer and passes
`workflowMode` at all three render sites. Defending an unreachable arm
so four tests keep passing preserves the tests, not the behaviour.

Two notes for anyone converting the remaining dashboard files:

- I did **not** default `workflowMode` to `true`, which was the tempting
one-liner. `isArchived`, `isHoldColumn` and `isWipProcessingColumn` all
switch on that same flag, so a global default would silently reinterpret
every other fixture in an 85-test file.
- **"Four tests break" was itself an underestimate.** Two more FN-770
fixtures surfaced only after the first two were fixed, because they
render their own explicit `column="triage"` block instead of using
`defaultProps`. The blast radius only became accurate by fixing it in
waves.

## `taskActivity.ts` — composed, not copied

The degraded arm now composes `utils/columnRoles`' predicates instead of
naming ids. **No local copy** — which is the failure mode #2625 hit from
the other direction.

Equivalent *by construction*:

| lane | composition | resolves to |
|---|---|---|
| intake | `isIntakeColumnRole(undefined, col)` | `triage` |
| hold | `isPreImplementationColumnRole(...)` **and not** intake |
`todo` |

reproducing `col === "triage" || (col === "todo" && isReplanning)`
exactly, since the shared pre-implementation set is `{todo, triage}` and
the shared intake id is `triage`.

Deliberately phrased as *"pre-implementation and not intake"* rather
than a second id list: if either shared set changes, this composition
follows it instead of silently disagreeing with the file next door. That
disagreement is precisely what bit the `TaskContextMenu` consolidation
above.

**I previously reported this site as blocked on `TaskCard.tsx` (U12's)**
— on the theory that the arm could only die once every caller supplied
resolved flags. Wrong framing: the arm doesn't need to become
*unreachable*, it needs to stop *naming ids*. Composing the shared
predicates does that without touching any caller.

## Verification

**1139 of 1141** green across `app/utils`, `Column` and `TaskCard`
suites. The two `TaskCard` failures are **pre-existing** — verified by
stashing this change and re-running, where they fail identically.
Dashboard app typecheck and lint clean.

Takes U11's cluster to zero except `TaskContextMenu.tsx`, whose
remaining site is covered in **#2626** and whose second site is a
documented non-consolidation.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-29 22:37:43 -07:00
..
2026-07-26 18:11:47 -07:00