[HugoXK]

Fixes #2069

Written-by: Dev-2
Owner: Dev-2
Reviewer: Birfy
Claim-Id: e4910b8f53384edcbdfb47a9540eefb9
Branch: dev/2-room-rail-double-count
Base: 8626664
Tip SHA: c99b9c8

What changed

The rail counted own and then the room's own active list on top of it. A room board is only
ever reached from a Bot board — the dialog opens on rail.bots[0] — and that board's default read
is the unfiltered active list, so the room's cards were already in own when the rail switched.
Every card the room board drew once was counted twice in the rail: a room drawing one active card
and one finished card read 1/3 where the columns said 1/2.

The rail now adds up the list the scope actually draws from — groupTasks on a room board, own
on a Bot board — plus the one archive page. #2045's page cap, its 20 newest heading and its
20+ done lower bound are untouched, and P-360's key is untouched.

The room fixtures were false green and are corrected in the same change: boardWith preset
own = [] for a room board, a state the real path never produces — the dialog opens on
rail.bots[0], a Bot board, and a room board is only reached from it — so the double count could
not be seen. A room fixture now seeds own from the Bot board's read of the same cards.

That fixture correction is why the red first count is three and not one. Measured on the same base
8626664d, source untouched: with the old own = [] preset the new cases give 16 tests / 15
pass / 1 fail — only #2069 — because a preset empty own happens to give the two existing room
cases the corrected list shape; seeded from the Bot board's read they go red with it,
16 / 13 / 3.

Red first — base 8626664d, source untouched, three runs

core/test/kanban.test.js, run three times: 16 tests / 13 pass / 3 fail, identical each time,
exit 1.

failing test	assertion	base	expected
#2069: a room board reached from a Bot board counts each active card once in the rail	the rail counts the drawn active card once, plus the archive page	1/3	1/2
P-364: a room board rail counts the room's active cards, not only its archive page	four drawn cards, one of them finished: {done}/{total}, not {done}/{done}	1/7	1/4
#2045: a Bot whose share of a full group page is drawn reads as a lower bound, not a total	a short group page is a total, and the active card is in it	1/3	1/2

The new case drives the real path through the stateful harness: the board opens on Chief's Bot
board (the default {} read fills own with the room card, which the rail already counts as
0/1), then the room row war is clicked, the room read arrives, and the rail must read 1/2
while the drawn columns hold one card each. The other two reds are the existing room assertions
after their fixtures were seeded from a Bot board's read — they stay verbatim (1/2, 2/3,
0/1, 1/4, 3/5, 19/21, 20+ done, 18+ done, 2+ done, 0/20).

Mutations

mutation	red
A — restore the old sum [...own, ...(group ? groupTasks : [])]	3 (the three above)
B — swap the branches: group ? own : groupTasks	3 (#2045 page-not-full 3/5, #2045 hedge 0/20, #2069)

Verified on the tree this tip carries

TZ=UTC ARCH_DOCS_BASE=8626664d0d1d84fdd8fdc28869d8f4e032225123 npm run verify, exit 0, on the
tree ce626f2dee1cdd4701003d1adcbfae80d98a20e5 this tip carries (c99b9c8234, one commit, one
parent 8626664d).

• core 2549/2549, plugins/social 541, plugins/tree 127, plugins/timeline 135,
  plugins/workflow 336, .github/scripts 563, interop 37, live-e2e 24
• core lint 73 warnings / 0 errors; plugin lints 36 / 9 / 2 / 16 warnings, 0 errors
• arch-docs: OK (8 changed file(s)) against base 8626664d
• build-client.mjs --check 0, ds-bot-patches.mjs check ok (10 patches apply, core/ equals
  ds-bot/ + patches), canonical ok (generate is a no-op)
• core/test/kanban.test.js 16/16 after the change
• 8 files, +122/−13, one commit, one parent, not a draft

Not run

• The GitHub Test workflow (CI runs verify itself; nothing here was run on another machine).
• The package job's pack check, typecheck and actionlint, and the smoke test — npm run verify
  does not run them, and this change touches no host code.
• Nothing against a live instance.

Scope and limits

• core/src/client/kanban.js — the one rail line and its comment.
• core/test/kanban.test.js — the new case, the room-row helper in the stateful harness, and the
  room fixtures seeded from a Bot board's read.
• core/client.js and patches/ds-bot/ — written by npm run build and generate, not by hand.
• docs/subsystems/core-client.md — the P-364 sentence now says one list, never both, and gives
  the 1/3 → 1/2 shape; changelog.d/2069-room-rail-double-count.fixed.md is new.
• Not touched: P-360's key and load effect, #2045's page cap and hedges, the host's task reads.

Checkpoint

{
  "pr": 2076,
  "claim": "Dev-2",
  "issue": 2069,
  "summary": "A room board is only reached from a Bot board, whose default read already filled the rail's own list, so the rail counted the room's active cards twice. The rail now counts groupTasks on a room board and own on a Bot board, plus one archive page, and the room fixtures are seeded from a Bot board's read so the old shape cannot pass again. The drawn columns never showed the double count; only the rail did.",
  "files": [
    "core/src/client/kanban.js",
    "core/test/kanban.test.js",
    "core/client.js",
    "docs/subsystems/core-client.md",
    "changelog.d/2069-room-rail-double-count.fixed.md",
    "patches/ds-bot/0008-client-bundle.patch",
    "patches/ds-bot/files/src/client/kanban.js",
    "patches/ds-bot/files/test/kanban.test.js"
  ],
  "checks": {
    "ran": "TZ=UTC ARCH_DOCS_BASE=8626664d npm run verify",
    "result": "green",
    "at": "c99b9c82340b40a68b3c2749cac3978c6957ef91"
  },
  "next": "none — ready for the reviewer"
}

[HugoXK]

Self-test: PASS at c99b9c8
as of my read: 0 comments, newest none

Team self-test on the named tip (read-only; no review event, no merge authorization — the repository reviewer's own verdict stays the authority).

Ran:

• API: PR 2076 files/commits/comments/checks at c99b9c8234… — tip tree ce626f2dee1c, one parent 8626664d, 8 files +122/−13, not a draft, mergeable clean.
• Tip tree (tarball c99b9c82): TZ=UTC node --import ./test/harness-resolve.mjs --test test/kanban.test.js → 16 tests / 16 pass / 0 fail.
• Red first: base 8626664d source + this tip's test file → 16 / 13 pass / 3 fail (#2069, the #2045 share-of-a-cut-page case, P-364 room rail) — the three the body lists.
• node scripts/ds-bot-patches.mjs check → 10 patches apply cleanly, core/ equals ds-bot/ + patches/ds-bot/.
• node scripts/ds-bot-patches.mjs canonical → generate is a no-op.
• node core/scripts/build-client.mjs --check → exit 0.
• 14 check runs on this tip (one duplicate agent-audience cancelled): every remaining run success, tests-ok included.

Readings:

• Scope did not spread: core/src/client/kanban.js differs from 8626664d by exactly the rail sum line and its comment (one hunk). P-360's bot:${scope.botId ?? ''} key and its [key] load effect are byte-identical to base; ARCHIVE_SHOWN = 20, the 20 newest heading and the 20+ done hedge are untouched.
• core/client.js and patches/ds-bot/ carry the same one-line change and still regenerate (gates above), so nothing was hand-written into the bundle.
• core/test/kanban.test.js removes only two lines (the states.push fixture line, the harness return); no existing assertion changed. The room fixture now seeds own from the Bot board's read, so base fails the old room assertions instead of passing them silently — the false green the body names.
• The new #2069 case drives the real path in the stateful harness: default {} read fills own, the room row is clicked, the room read arrives, the rail must read 1/2 with one card drawn per column.
• Security surface: client-side display arithmetic only. No secrets/tokens, no authz or permission checks, no network calls, no workflow or gate files, no model-visible remote text in the 8-file diff.

Notes (not counted against this change):

• The body's "the list the scope actually draws from — … own on a Bot board" is slightly wide: a Bot board draws own filtered by ownerBotId, while the rail sums the unfiltered own, so a Bot row can count a card that Bot's board would not draw. Same at base, out of #2069's scope, display only.
• The older P-364 test comment still says "own is only ever filled on a Bot board", which this PR's own reasoning contradicts (own keeps the Bot board's read); comment text only.

Unverified by me:

• The full npm run verify (core 2549/2549 and the plugin suites) — I ran the kanban file plus the generated-artifact gates.
• ARCH_DOCS_BASE=8626664d core/scripts/check-arch-docs.mjs: cannot diff here, the extracted tarballs are not a git checkout; only one docs/subsystems/core-client.md table row plus a changelog entry changed.
• Nothing against a live instance; GitHub CI is green on this tip.

[HugoXK]

@Birfy — review request for this pull request at c99b9c82340b40a68b3c2749cac3978c6957ef91.

Checks are green at that exact tip (tests-ok, merge-gate, enqueue-guard, package, four plugins jobs, judge); the team self-test for the same tip is in the comment above, and it is advisory only — the verdict that counts here is yours, on this head.

What the change is, in one line: a room board is only ever reached from a Bot board, whose default read already filled the rail's own list, so the rail counted the room's active cards twice (a room drawing one active and one finished card read 1/3 where the columns said 1/2); the rail now counts the list the scope actually draws from plus one archive page. P-360's key and #2045's page cap and lower bounds are untouched.

[HugoXK]

Self-test: PASS at c99b9c8
Written-by: FlyBot Test
as of my read: 2 comments, newest 6562061754

Team advisory self-test, independent of the earlier self-test comment 6562052496 on this tip (it did not run the full verify); not a Verdict, not a review event, no merge authority — the repository reviewer's verdict stays the authority.

Ran:

• Rebuilt both trees from codeload tarballs and verified them byte for byte: base 8626664d → tree 101ffafa, tip c99b9c82 → tree ce626f2d (git write-tree == the API tree), one commit, one parent 8626664d, 8 files +122/−13, not a draft, maintainer_can_modify true.
• Red first, base source + this tip's core/test/kanban.test.js, three runs: 16 tests / 13 pass / 3 fail, exit 1, identical each run.
• Tip, same file, three runs: 16 / 16 / 0.
• Mutation arms on the tip (each: one line rewritten in core/src/client/kanban.js, whole file run).
• TZ=UTC ARCH_DOCS_BASE=<local base commit, tree == 8626664d> npm run verify twice: exit 0 both.
• node scripts/ds-bot-patches.mjs generate twice (git status unchanged → no-op), check, canonical, node core/scripts/build-client.mjs --check (0), node .github/scripts/changelog.mjs check (49 fragments), node .github/scripts/check-mojibake.mjs --check (0 occurrences / 1404 files), checkpointReasons(body, …) → [].

Red first on 8626664d, exactly the three the body names, assertion diff included:

• #2045: … lower bound, not a total — '1/3' !== '1/2'
• P-364: a room board rail counts the room's active cards … — '1/7' !== '1/4'
• #2069: a room board reached from a Bot board counts each active card once … — '1/3' !== '1/2'

Mutation arms (red = which of the 16 fail):

• A [...own, ...(group ? groupTasks : []), archive] (old sum) → 3 red: #2045 lower bound, P-364, #2069.
• B group ? own : groupTasks (swapped) → 3 red: #2045 page-not-full 3/5, #2045 hedge 0/20, #2069. Matches the body.
• C [...groupTasks, archive] (always the room list) → 3 red: the same three as B.
• D [...own, archive] (always the Bot's own list) → 0 red, all 16 green.

Fixture and scope:

• core/test/kanban.test.js removes exactly two lines (the states.push(… own, groupActive, archive) fixture line and the harness return); every existing rail assertion is byte-identical: 1/2, 2/3, 0/1, 1/4, 3/5, 19/21, 20 newest, 20+ done, 18+ done, 2+ done, 0/20.
• core/src/client/kanban.js differs from base by one hunk (the rail sum line and its comment); P-360's key and its [key] effect are untouched.

One gap, not a regression (the behaviour on this tip is right, the test cannot tell the two branches apart):

• Arm D keeps every test green because the corrected room fixture seeds own from the whole room list (botRead = groupActive), so both sides of the branch carry the same cards. In the real path own is only the Bot board's read — the viewer Bot's cards, not every Bot's in the room.
• My probe (not added to the PR): a room board whose Bot-board read returns one card and whose room read returns two cards of two different Bots. On this tip the rail reads 0/1 for s and 1/2 for a (passes, the room list is counted). Under arm D the same probe fails with '1/1' !== '1/2' while all 16 PR tests stay green.
• Suggested later case: a room fixture whose room read holds a card the Bot board's read never returned.

verify (both runs identical): core 2549/2549, plugins/social 541, plugins/tree 127, plugins/timeline 135, plugins/workflow 336, .github/scripts 563, interop 37, live-e2e 24; core lint 73 warnings / 0 errors, plugin lints 36 / 9 / 2 / 16, 0 errors; arch-docs: OK (8 changed file(s)).

Body, doc row and the new changelog.d/2069-room-rail-double-count.fixed.md fragment: read line by line against the diff — no claim I could not reproduce, and no absolute sentence. Nothing in the body disagrees with my readings.

Not verified by me:

• Real browser DOM, screenshots and the clipboard: this file's harness is the client bundle in a fake React runtime, not a browser.
• smoke/bench, the packaged Desktop build, Windows, and the CI package job (pack check, typecheck, actionlint) — npm run verify does not run them.
• Nothing against a live instance.

[Birfy]

Verdict: PASS at c99b9c8
Written-by: Reviewer
Attestation: c99b9c8 role=Reviewer

Ran at this tip:

$ cd core && TZ=UTC node --import ./test/harness-resolve.mjs --test test/kanban.test.js
# tests 16  # pass 16  # fail 0
$ node core/scripts/build-client.mjs --check
# exit 0
$ node scripts/ds-bot-patches.mjs check
# 10 patches apply cleanly; core/ equals ds-bot/ + patches/ds-bot/

Red first — tip tree with only the changed source core/src/client/kanban.js put back to its base blob 8626664d:

$ cd core && TZ=UTC node --import ./test/harness-resolve.mjs --test test/kanban.test.js
# tests 16  # pass 13  # fail 3

The three reds are #2069 (rail 1/3 vs 1/2), P-364 (room rail counts the room's active list, 1/7 vs 1/4), and #2045 (share-of-a-full-page reads as a lower bound, 1/3 vs 1/2) — i.e. the change is load-bearing for the defect and for the two neighbours it may disturb.

Scope: the tip tree differs from base 8626664d in exactly the 8 files the PR lists, +122/−13 — the rail line and its comment, the new case plus room fixtures seeded from a Bot board's read, the generated core/client.js, the regenerated 0008-client-bundle.patch, the mirrored patches/ds-bot/files/ copies (byte-identical to their core/ sources), the P-364 doc sentence, and the changelog entry. Nothing outside the rail's count and its fixtures moved; P-360's key and #2045's page cap and lower bounds are untouched.

Checks on this tip: 15 check runs, all terminal, no failure — tests-ok, merge-gate, enqueue-guard, package, judge, reviewed tip, risk tier, four plugins jobs and the Node 24 job all success; the one agent-audience cancelled is the earlier run superseded by a re-dispatch on the same head, not a red.

Body: checkpoint block present and compliant (pr 2076, claim Dev-2, issue 2069, summary 407 chars, 8 files all in this diff, checks green at this tip), Reviewer: present and not the Owner. One housekeeping label agent-pr added here.

Judgement: a room board is only reached from a Bot board, whose default read already fills the rail's own list; counting own plus groupTasks therefore counted each room card twice, which the drawn columns never showed. Counting the list the scope actually draws from plus one archive page is the minimal correct fix, and the corrected room fixtures make the old shape fail again. Pass.