Upstream synchronization ledger, 2026-08-28

Scope and counting method

This hand-written ledger records every non-merge, patch-equivalent record returned by:

git log --right-only --cherry-pick --no-merges --format="%H%x09%s" origin/main...upstream/development

The refreshed command was run against origin/main at 83c91f6964cc1799fcc7e1d4fcd23f90e5e017f6``a5428c07d91830c109a37323fcee67ee4433a655 and upstream/development at b17e06dd0f0d9a45807eb39a51d223f52eb14da9, with merge base d9080117b1fd01193d3eee51ae243714468c8176. It returned 112 records in the isolated ledger checkout. Stable patch IDs identify 108 unique patch effects. Rows 094-097 are duplicate records paired exactly with rows 078, 080, 081, and 082 respectively. They remain listed so no source record is lost from review. The image-difference pair in rows 001-002 is not silently collapsed: row 001 is an intermediate rendering hint, and row 002 is the follow-up compositor correction.

Each row has a source SHA, exact subject, subsystem, disposition, and local evidence or intended lane. The refreshed disposition contract is deliberately closed: ported means the behavior is represented by a local implementation; already-equivalent means no code delta is needed because the current tree already has the behavior; superseded by stronger local behavior means the current implementation intentionally replaces the upstream behavior; inapplicable because of platform/scope records a platform, release-history, or workflow-policy boundary; reverted-history duplicate records duplicate or net-zero upstream history; and review required is a non-blocked, actionable disposition for behavior awaiting parent implementation review or focused verification. The ledger lane does not port product code, so review-required rows name the owning issue and local evidence.

The historical task/upstream-integration-20260828 query remains preserved in the earlier handoff record. It is not the refreshed baseline, and it is not used to decide whether an upstream record is accounted for.

Refreshed disposition counts

Re-verification, 2026-09-03

The counting baseline in this ledger was origin/main at 83c91f6964. That tip has since advanced to a5428c07d9, carrying nineteen integrated task lanes. The canonical query was re-run against the new tip and still returns 112 records, which is the expected result and worth stating plainly: this fork reimplements upstream behaviour rather than cherry-picking it, so --cherry-pick cannot match the patches. The record count does not move; only dispositions do.

Two rows were corrected. Rows 002 and 099 said their local adaptation was not in origin/main; both are now ancestors of it and the evidence says so.

No review required row has been flipped without per-row verification, and that is deliberate. Several areas the ledger lists as review-required now have real implementations on main, and a reviewer can work through them against these commits:

Naming the implementation is not the same as proving the upstream behaviour is matched, which is why the dispositions still read review required. Flipping them on subject similarity alone would make this ledger assert something nobody checked, and a ledger that does that is worse than a stale one.

The copilot-conflict-ui lane is not integrated. It and upstream row 015 (delete-vs-modify conflicts) describe two parallel designs of the same code paths; merging it mechanically would drop one of them.

Hard GUI boundary

The current GUI is the deliberately restored presentation. It remains Material Design 3 through shared controls, dialogs, tokens, focus behavior, accessibility, and state layers, but its reverted shell style is protected. This ledger authorizes compatible behavior and control-level repairs only. It does not authorize a shell redesign, a screen rebuild, a navigation rewrite, or import of the retired shell.

The history that establishes this boundary is:

These commits are the provenance for the ledger rule: upstream behavior may land in the existing GUI architecture, but upstream shell styling never becomes an implementation order.

Evidence key

The local evidence references below point at the current repository's corresponding architecture and define the intended implementation lane.

Complete record ledger

Refreshed source-link and issue index

The source SHA in every row is linked to the exact upstream commit. Stable patch IDs are computed from each non-merge record, yielding 108 unique patch effects across 112 source records. The evidence key links to the local implementation or evidence map above. Every row is tracked by the open parent issue #212; no row is marked complete merely because its upstream commit is merged upstream.

Count check for this index: 112 rows, 112 exact upstream SHA links, 108 unique stable patch IDs, and 112 issue-status cells. The index is hand-written review data copied from the refreshed command output; it is not a generated discovery list.

The ledger is checked against the exact source command in this checkout by check-ledger.ps1:

1. The source command returned 112 rows and 108 unique stable patch IDs.
2. The hand-written table contains rows 001 through 112 exactly once.
3. Every row has a 40-character source SHA, subject, subsystem, refreshed disposition, and local evidence or lane.
4. The source-link index contains 112 exact commit links, 108 unique stable patch IDs, and 112 open parent issue links.
5. Duplicate and net-zero history is explicitly identified in rows 033-034; the duplicate patch pairs are 078/094, 080/095, 081/096, and 082/097.
6. Platform/scope, stronger-local-behavior, already-equivalent, ported, and review-required dispositions are represented.
7. The checker was deliberately made red by removing row 112 from a temporary copy, then restored and returned green. It does not claim that review-required rows have been implemented or verified.