[HugoXK]

Fixes #1693

Owner: Dev3
Written-by: Dev-3
Reviewer: Reviewer
Branch: dev/3-wrapped-frame-wires
Base: b58ad68
Tip: 8424a3d
Tree: 865e9529dd3f38d38294fae75331f766ebb56118
Claim: aa5f58ca1a89429ba0f03f9922dce163

What and why

The drawer's Edit plane has two shapes: a fitted plane while a parallel step's cards still stay apart (fitPlane: scaled centres, 168x72 cards) and a laid-out one once they cannot (editPlane: cards shrunk to the panel, 98 wide at a 364px drawer and 96 or 72 high once measured). The frame and the wires were still drawn from the fixed 168x72 card in both.

• Frame. parallelFrameStyle (client.js:652) took left/width from NODE_W and right/bottom from node.y + NODE_H, never from the drawn box. On the 3-way panel-364 fixture (cards 98 wide at centres 72/182/292) that is x = -22..386, so the drawer canvas clipped both uprights (overflow:hidden). Its height also added only pad while its top sits 20px above the cards, so the dashed bottom edge ran 10px through the cards at the 72px height and 34px at 96px. It now reads node.w ?? NODE_W and node.h ?? NODE_H and takes its bounds from the cards it wraps.
• Wires. The laid-out plane drew every edge with wirePath(from, to, 0), one fold at the midpoint of the two rows. On the 6-way fixture that put the fold through cards on 8 of 13 edges (c>r4 / c>r6 through r1..r3, r1>d / r3>d through r4..r6). The laid-out plane now draws flowWires' own paths, computed from the boxes and bands flowLayout already lays out — the function wrote itself for exactly this failure mode (a straight run across a band cuts through the other line's cards, workflow: wrap an over-wide drawer layer inside its frame #1659), which is why the Read-only plane has 0 crossings today.
• One copy of the row geometry. editPlane no longer computes its own rows: the laid-out branch takes flowLayout's boxes for its cards, its band frames for the dashed frame, and its wires. The fitted plane keeps its scaled cards, its parallelFrameStyle frame and its single-fold wirePath, so a graph whose cards still fit draws exactly what it drew before.

Scope: plugins/workflow/client.js and its test, one subsystem-doc row, one changelog fragment. No core/, no patches/**.

Acceptance

#	criterion	evidence
(a)	frame left >= 0 and right <= panel on the 3-way panel-364 fixture, enclosing every card	the parallel frame reads the drawn card size, not the fixed 168x72 (#1693): frame 13..351 for cards 23..341; the laid-out Edit plane paints the parallel frame inside the panel and around the cards (#1693) asserts the same on the rendered plane
(b)	the same at card height 72 and 96	the laid-out frame encloses every card inside the 364px panel at card height 72 / 96 (#1693): frames 120..230 at 72 and 144..278 at 96, both 13..351 wide, both enclosing every card box
(c)	6-way fixture: no edge's horizontal segment through a card box	the laid-out Edit plane routes every wire around the cards of a 6-way step (#1693): 12 edges, 0 segments inside any card box and 0 inside any out-port dot. See the note below on the remove-button box.
(d)	#1683's six byte-identity fixtures and client.test.js:1145-1171 unchanged and passing	that test is untouched and green: the four Read-only planes and the two fitted Edit planes are byte-identical to BASE_PLANES (read:delivery@368, read:fork3/4/6@368, edit:delivery@364, edit:fork2@364)
(e)	Read-only plane unchanged	same BASE_PLANES entries — flowWires paths byte-identical; flowLayout/flowWires are not edited at all, only called by the Edit plane
(f)	TZ=UTC ARCH_DOCS_BASE=<base> npm run verify exit 0, red first for (a)-(c)	red on the pre-fix client.js below; green at the tip, counts below

Note on (c) and the remove button. The .wf-mini box rides 8px above its card (top:-8px, client.js:2238) and the out-port dot 5px below it, while a wrapped line's wrap gap is FLOW_WRAP_GAP = 12px: 5 + 8 > 12, so any horizontal run that crosses that gap into the second line enters one of the two boxes. The routes flowWires picks are exactly the Read-only plane's (byte-identical, criterion (e)), and on the 6-way fixture they touch the remove-button box with 6 segments at y=226 — the wrap seam. The wire is painted under the card (the svg is the plane's first child) and under the opaque button, so the button keeps its click; the card box and the out-port dot, which sit on the card, are clear. The test asserts those two and documents the button case in place. If the team wants the button box clear as well, the wrap gap has to grow, which moves the Read-only plane and breaks (e) — that is a separate decision, not this fix.

Red first

The new tests run against the pre-fix source: worktree at dcaae9bb with only plugins/workflow/test/client.test.js replaced. plugins/workflow/client.js is blob 4bd73e46 at dcaae9bb, at 4e0fa9c5 and at b58ad688, so this is the same source this branch now sits on.

# tests 94
# pass 89
# fail 5

not ok 90 - the parallel frame reads the drawn card size, not the fixed 168x72 (#1693)
  error: the frame starts left of the first card by its own width
  -22 !== 13

not ok 91 - the laid-out frame encloses every card inside the 364px panel at card height 72 (#1693)
  error: the laid-out plane says which band frame it drew
  undefined !== 1

not ok 92 - ... at card height 96 (#1693)
  error: the laid-out plane says which band frame it drew
  undefined !== 1

not ok 93 - the laid-out Edit plane paints the parallel frame inside the panel and around the cards (#1693)
  error: the frame -22..386 is outside the 364px panel

not ok 94 - the laid-out Edit plane routes every wire around the cards of a 6-way step (#1693)
  error: a wire runs through a card
  + [
  +   'card r1: 182,150→72,150',
  +   'card r2: 182,150→72,150',
  +   'card r2: 182,150→182,232',
  +   'card r3: 182,150→292,150',
  +   'card r4: 72,277→182,277',
  +   ... 32 crossings in total
  + ]

At the tip the same file is 94/94.

What I ran

• TZ=UTC ARCH_DOCS_BASE=b58ad688 npm run verify — exit 0: core 1981, plugins/social 539, plugins/tree 125, plugins/timeline 129, plugins/workflow 269 (+5 new), .github/scripts 549, interop 37, live-e2e 24, all fail 0; arch-docs: OK (4 changed file(s)); ok 10 patches apply cleanly in order; ok core/ equals ds-bot/ + patches/ds-bot/; oxlint 0 errors (64/36/9/2/15/0 warnings, all pre-existing).
• node --import ./test/resolve.mjs --test test/client.test.js in plugins/workflow, at the tip (94/94) and on the pre-fix source (89/94, five red).
• node .github/scripts/changelog.mjs check — well formed.

Not run: the browser/smoke bench and a visual check in the drawer UI (no host here); npm ci / npm run harness in this worktree (the shared install is linked, unchanged). No push to main, no core/ or patches/** change, so the generated-file recipe does not apply.

{
"pr": 1697,
"claim": "Dev3",
"issue": 1693,
"summary": "The drawer's laid-out Edit plane now takes its rows from flowLayout: the cards are its boxes, the dashed parallel frame is the band frame it drew, and the wires are the flowWires routing around those boxes, so a 3-way step at a 364px panel keeps its frame inside the canvas and no wire folds through a card of the other line. parallelFrameStyle reads node.w/node.h. The fitted plane keeps its scaled cards, its frame and its single fold, and the Read-only plane is byte-identical.",
"files": ["plugins/workflow/client.js", "plugins/workflow/test/client.test.js", "docs/subsystems/workflow.md", "changelog.d/1693-drawer-frame-wires.fixed.md"],
"checks": { "ran": "TZ=UTC ARCH_DOCS_BASE=b58ad688 npm run verify", "result": "green", "at": "8424a3d913e1cd483c3d347526fce2fcd5a57e1e" },
"next": "none — ready for review"
}

[HugoXK]

@Birfy Review PR #1697 at 8424a3d

Fixes #1693 — the drawer's Edit plane now draws what it shows: parallelFrameStyle sizes the dashed frame from each card's own node.w ?? NODE_W / node.h ?? NODE_H (no longer the 168×72 fallback), and the laid-out (wrapped) plane routes its wires through flowLayout's own boxes and bands via flowWires instead of a single mid-row elbow, so the frame stays inside the 364px panel and no wire crosses a card. The fitted plane keeps wirePath and its scaled delta (#1677). Red first on base b58ad688: not ok 90 frame starts -22 !== 13, not ok 91/92 plane.frames undefined !== 1, not ok 93 the frame -22..386 is outside the 364px panel, not ok 94 6-way wires crossing cards → 94/94 at this tip.

[HugoXK]

@Birfy Review PR #1697 at 7de43e3

Fixes #1693 — the drawer's Edit plane now draws what it shows: parallelFrameStyle sizes the dashed frame from each card's own node.w ?? NODE_W / node.h ?? NODE_H, and the laid-out (wrapped) plane routes its wires through flowLayout's own boxes and bands via flowWires instead of a single mid-row elbow, so the frame stays inside the 364px panel and no wire crosses a card. The fitted plane keeps wirePath and its scaled delta (#1677).

This tip is a rebase onto live main d5347032 (#1694 + #1695); the only conflict was the appended test block in plugins/workflow/test/client.test.js, and this tip keeps both blocks (the #1686 three cases and the #1693 three cases) in place. Red first on b58ad688 (same client.js blob as d5347032): not ok 90 frame starts -22 !== 13, not ok 91/92 plane.frames undefined !== 1, not ok 93 the frame -22..386 is outside the 364px panel, not ok 94 6-way wires crossing cards → 94/94 at that tip; tests-ok and Node 24 are green here too.

[Birfy]

Verdict: PASS at 7de43e3
Written-by: Reviewer
Attestation: 7de43e3 role=Reviewer

Ran at this tip:

$ cd plugins/workflow && TZ=UTC node --import ./test/resolve.mjs --test test/client.test.js
# tests 97  # pass 97  # fail 0

Red first (same command, plugins/workflow/client.js reverted to main d534703):

# tests 97  # pass 92  # fail 5

The branch sits on b58ad68 and is stale against main d534703, so I replayed the tip's change set (plugins/workflow/client.js, its test, docs/subsystems/workflow.md, changelog.d/1693-drawer-frame-wires.fixed.md) onto main d534703 and ran the changed directory's test file there — 97/97 with the fix, the five #1693 tests red without it (frame left of the first card -22 !== 13; no band frame; frame -22..386 outside the 364px panel; 32 wire crossings). Not run: the rest of npm run verify; the job's tests-ok covers it at the tip.

Check-runs at the tip: tests-ok, plugins (workflow/tree/social/timeline), package, merge-gate, Node 24 all success; enqueue-guard failure is the fork shape (no agent-pr), not a test red.

Checkpoint

{
"pr": 1697, "claim": "Dev-3", "issue": 1693,
"summary": "The Edit plane drew its frame and wires from the fixed 168x72 card. parallelFrameStyle now reads node.w/node.h and takes its bounds from the cards it wraps, and the laid-out plane draws flowWires paths from flowLayout boxes instead of one wirePath fold per edge. The fitted plane keeps its scaled cards and frame. Scope is plugins/workflow/client.js, its test, one subsystem-doc row and one changelog fragment.",
"files": ["plugins/workflow/client.js", "plugins/workflow/test/client.test.js", "docs/subsystems/workflow.md", "changelog.d/1693-drawer-frame-wires.fixed.md"],
"checks": { "ran": "cd plugins/workflow && TZ=UTC node --import ./test/resolve.mjs --test test/client.test.js", "result": "green", "at": "7de43e33a8f85982cf29d687142c546d054254e6" },
"next": "replay the change set onto current main d534703 before queueing: the branch base b58ad68 conflicts there"
}