Text column edges

Feature plan: Step 1 · Step 2 · Step 3 · Step 4

Debug mode: enabled (debug log)

Effort study: railspike-effort / v3/pledge_fills; host ledger. Standing P2 enrollment. Parent: v3/pledge_withdraw.

Extended estimates: enabled

Status: Steps 1 (Option B) and 2 approved by the maintainer 2026-11-16. Step 3 written 2026-11-16, awaiting Approved Step 3.

Cycle estimate: 0.50

Step 3: Development Plan — Pledge fills

Completion Contract

  • Entry: "Mark delivered" (the pledger) or "Mark filled" (the author or an operator) on a pledge on the campaign thread page; or a typed reply to the pledge with filled or filled: 200.

  • End-to-end outcome: the pledge shows "Delivered by mara" or "Filled by ilyag" with the amount, linked to the reply. The block shows "N filled", plus "M confirmed" when part is pledger-reported. A bare fill reply adds no card.

  • Recovery: a failed press reports on the pledge and can be pressed again. Retracting a fill reply undoes it. Turning FORUM_PLEDGES_ENABLED off returns every line to plain text.

  • Deployment: no schema change, no rebuild, no new flag, no new dependency.

  • Release condition: with the flag on at zenmemes.com, Ilya pledges from his iPhone SE, marks it delivered, sees "filled" change, then retracts the fill. Not possible from this session; it stays pending in Step 4.

Findings from the checks Step 2 asked for

  • Operator cost is small. ApprovalPolicy::isActiveOperator() reads one approval seed file and one retirement file per key. The reader asks once per distinct fill author, through a check passed in by the thread page.

  • The buttons follow Withdraw. pledge_amount.php already compares the viewer from the identity hint with the pledge's author. "Mark filled" compares it with the opening post's author, or asks the same operator check. A baked page with no viewer shows no button. The read rule is the real guard.

  • Operator means operator now. A fill by a key that later stops being an operator stops counting. That matches Step 2's "an operator at the time the thread is shown".

Refinements to Step 2, within its scope

  • Fills are read before withdrawals, so order does not matter. Each pledge's fills add up in thread order, capped at its amount. A withdrawal takes effect only when the pledge is not fully filled. A withdrawn pledge counts in "pledged" as its filled amount. Retracting any reply gives the same result as if it never existed.

  • One button per pledge. The pledger sees "Mark delivered"; the author or an operator sees "Mark filled". Neither shows on a fully filled, withdrawn or uncounted pledge. An author's own pledge gets "Mark delivered".

Key Risks

  • High risk: totals disagree with the pledges shown. Impact: readers stop trusting the block. Early validation: Stage 1 tests fill then withdraw, withdraw then fill, over-fill, and retracting each reply. Mitigation: one reader computes pledged, filled and confirmed from the same pass.

  • Folding hides a reply that matters. Impact: a note, an answer or a reason disappears. Early validation: Stage 3 tests bare, noted, answered and uncounted fill replies. Mitigation: reuse the withdraw fold rule unchanged.

  • The second button breaks the first. Impact: Withdraw stops working. Early validation: Stage 4 runs the existing withdraw browser check before adding the fill press. Mitigation: extend the script by a role, not a rewrite.

Stage 1

Estimate: 0.14

  • Goal: fill replies are read and counted; withdrawal respects fills.

  • Dependencies: none.

  • Expected changes: new Support/FilledDirective with extract(string $body): ?string (text after filled, empty when bare, null when absent; quoted lines skipped) and isBare(). PledgeCampaign::fromPosts() gains an optional operator check, ?callable $isOperator, and a private fills() run before withdrawals(). Each pledge entry gains filled and a list of fills (reply ID, who, amount, counted amount). The result gains filled, confirmed and fills, keyed by reply ID: effective or a reason. Reasons: not a reply to a counted pledge, not the pledger, author or an operator, no name, unverified, not a number or unit, already filled. Withdrawal gains "already filled". "Pledged" counts a withdrawn pledge as its filled amount.

  • Verification approach: PledgeCampaignTest for pledger, author and operator fills; another key; partial and over-fill ("Counted 100 of 150"); two fills adding up; filled: 2 coffees and filled: lots; fill then withdraw and withdraw then fill; fully filled withdraw; retracting a fill; flag off. Smoke set alongside.

  • Estimate details: driver: the reasons table and its order; uncertainty: little.

  • Risks or open questions:

    • Impact: totals disagree. Early warning: the order tests. Mitigation: fills first, one pass.

  • Canonical components/API contracts touched: PledgeCampaign.php (extended); WithdrawDirective.php and PledgeAmount.php (pattern and parser reused); ApprovalPolicy::isActiveOperator() (used unchanged).

Stage 2

Estimate: 0.08

  • Goal: readers see fills on each pledge and in the block.

  • Dependencies: Stage 1.

  • Expected changes: ThreadAndPostPageController passes the operator check. pledge_amount.php lists each fill ("Delivered by mara" or "Filled by ilyag", amount when partial, "Counted 100 of 150" when capped), linked to its reply, and shows a fill reply's reason. pledge_totals.php adds "N filled" and, when lower, "M confirmed". One style rule in site.css if needed.

  • Verification approach: rendering tests in PledgeCampaignTest for both labels, the cap note, a reason, and the block. A baked page before and after a fill.

  • Estimate details: driver: label wording; uncertainty: the author's display name.

  • Risks or open questions:

    • Impact: a long fill list crowds the card. Early warning: three fills on one pledge. Mitigation: one line per fill, plain text.

  • Canonical components/API contracts touched: pledge_amount.php and pledge_totals.php (extended); no second badge component.

Stage 3

Estimate: 0.05

  • Goal: a bare fill reply that took effect leaves the conversation.

  • Dependencies: Stage 1.

  • Expected changes: WithdrawLeafFilter::apply() gains the fills map and folds a bare filled reply by the same rule. thread.php passes it. The class keeps its name; its comment names both lines.

  • Verification approach: WithdrawLeafFilterTest for bare, noted, answered and uncounted fill replies, and a withdraw reply answered by a fill reply.

  • Estimate details: driver: none; uncertainty: little.

  • Risks or open questions:

    • Impact: a fold hides a reason. Early warning: the uncounted test. Mitigation: only effective replies fold.

  • Canonical components/API contracts touched: WithdrawLeafFilter.php and thread.php (extended, not copied).

Stage 4

Estimate: 0.12

  • Goal: one press marks a pledge delivered or filled.

  • Dependencies: Stage 2.

  • Expected changes: pledge_amount.php draws a fill form beside Withdraw, with the body filled and the label from the refinement above. The controller tells the template whether the viewer is an operator. pledge_withdraw.js handles both forms by role, with the failure message taken from the form.

  • Verification approach: rendering tests for who sees which button and where none shows. tests/browser/pledge_campaign.mjs: the withdraw press still works; "Mark delivered" changes the label and the block with no new card; a forced failure shows the message and a second press works. If no browser runs in the session, the browser check stays pending.

  • Estimate details: driver: three viewer rules; uncertainty: the operator flag on baked and high-traffic pages.

  • Risks or open questions:

    • Impact: Withdraw breaks. Early warning: run its check first. Mitigation: role-based extension.

  • Canonical components/API contracts touched: pledge_withdraw.js (extended); Outbox prepare and finalize calls and /compose/reply (used unchanged).

Stage 5

Estimate: 0.06

  • Goal: checked in every theme and documented.

  • Dependencies: Stages 3 and 4.

  • Expected changes: the pledge campaign entry gains fills; the spec notes fills read before withdrawals; the checklist, ledger and debug log; the Step 4 closeout.

  • Verification approach: the browser check in each theme at 375 and 1280 pixels; smoke set with PledgeCampaignTest, WithdrawLeafFilterTest and ThreadActivityTest; git diff --check; review the whole diff. The full suite is not run.

  • Estimate details: driver: 17 themes; uncertainty: browser availability.

  • Risks or open questions:

    • Impact: a theme check cannot run. Early warning: no browser binary. Mitigation: report it unverified.

  • Canonical components/API contracts touched: concept inventory (extended).

Concepts touched: pledge campaign, changed; thread activity, post retraction and operator grant, used unchanged.

Branch

plan/pledge-fills was rebased onto origin/main at c98b5221 on 2026-11-16 (two unrelated PRs landed). The rebase is local; the remote branch still holds the old base. After Approved Step 3, Step 4 cuts feat/pledge-fills from a fresh origin/main.

Study checkpoint

  • Step 3 handoff: 0.50 reaffirmed, scope v1 with the two refinements above (neither widens scope). Stage work 0.14, 0.08, 0.05, 0.12, 0.06, not summed; Stage 5 owns shared verification and documentation. Baseline c98b5221, R1, P2; Claude single exposed rater, 2026-11-16 UTC. R1 hashes not re-verified.

— Claude · 2026-11-16 UTC · Model: Claude Opus 5.5