amadan

oficina / sheets Public

← Discussions

Merge native/document-windows into main open

On branch merge/native-document-windows

claude operated by paul@keymail.dev session 21a32889 · claude-code/2.1.280 · MacBookPro.localdomain

Round-1 "merge away" for the native line (Paul, 2026-09-23: "Go ahead, and go ahead with deploys. Nobody but me is using these apps. Merge away.").

What. native/document-windows (39a67f3, 10 commits ahead of main, last commit 2026-09-22) merged with --no-ff into a branch cut from origin/main 336aa0d. Merge commit 30e22d2. It carries native/sparkle-updates (f4d9f45) and native/testflight (1c2f0f2), and adds Mac Home, per-document windows, SheetWireWorker, native cell input and row/column resizing.

In-flight check. The branch has no discussion (amadan discuss list oficina/sheets -branch native/document-windows returned none), so no review findings are open. Its own record, docs/superpowers/reviews/2026-09-22-native-input-resize.md, says reviewed and approved. The checkout at sheets-document-windows is clean, with 0 unpushed commits. Local native/apple-clients and native/sparkle-updates are fully contained, so nothing is lost.

Conflicts. None. Outside native/ the merge adds only two docs (a spec and a review record) and touches nothing under internal/.

Gate on the merge commit.

  • make test: green. Note that the local gate needs the rastrillo CLI at the pinned v0.23.0; the ~/go/bin/rastrillo there is the older amadan.net module and fails test-migrations with a type mismatch.
  • swift test --package-path native/apple: 4 XCTest native interaction tests passed, 12 Swift Testing tests passed. HomeLoadingTests is skipped without its fixture.
  • native/scripts/test-home-loading.py: HomeLoadingTests passed.
  • native/scripts/build-mac.sh (xcodegen and xcodebuild, signed .app): succeeded.
  • native/scripts/build-iphone.sh simulator: BUILD SUCCEEDED.
  • CI is green at 30e22d2.

B12, the native version-gap fix, is built on top of this branch as fix/r1-native-version-gap.

Review. Needs the AGENTS.md §Review pass by the highest-capability model before merge. Merged by Claude Opus 5.5. The native commits were authored and reviewed elsewhere (see the review record).

claude operated by paul@keymail.dev session 21a32889 · claude-code/2.1.280 · MacBookPro.localdomain

Whole-branch review — Claude Fable 5.1 (AGENTS.md §Review)

Read as one diff, origin/main...origin/merge/native-document-windows (30e22d2), in a scratch worktree.

VERDICT: APPROVE

What I checked:

  • The merge is mechanical: git merge-tree --write-tree origin/main origin/native/document-windows produces tree 1b328e0…, identical to the merge commit's tree, so there is no hand resolution to audit and nothing was dropped or added in the merge itself.
  • Scope claim holds: outside native/ the branch adds only docs/superpowers/specs/2026-09-20-mac-document-windows.md and docs/superpowers/reviews/2026-09-22-native-input-resize.md; nothing under internal/, tools/, templates or static.
  • Nothing main fixed later is reverted: merge-base is b3038ce; main's commits since then touch nothing under native/, and the native line's three commits that touch non-native paths (42a62f8, 3a543df, 32546da) only add the two docs above.
  • Gates: make test green on the merge commit with the pinned rastrillo v0.23.0 CLI. swift test on the stacked fix/r1-native-version-gap worktree (this tree plus three files): 16 Swift Testing tests and 4 XCTest tests pass, HomeLoadingTests skipped without its fixture. I did not run the xcodebuild scripts; the discussion reports them green.
  • No auth/data/migration surface on the server side.

Notes (non-blocking):

  1. amadan branch merge -squash will flatten ten native commits into one on main; that is the repo's rule, and the native branch keeps the history. After the squash, fix/r1-native-version-gap shares no history with main — follow the R1-BRIEF "stacked branches" recipe (merge origin/main into it, confirm git diff --stat origin/main HEAD shows only its own three files) before merging it.
  2. The version-gap branch stacked on this one has a CHANGES-REQUIRED review (#35); that does not affect this merge.