Merge commit 'e35a2708' into swap-prep

Author: WilfordGrimley

CLAUDE.md

@@ -275,15 +275,21 @@ external reader's orientation to the whole fork, see
   artifact 2 (knowledge-inventory sweep) status — all 3 MISSING
   constants decided, items 1–2's fix merged (PR #343) and deployed
   2026-07-23T01:33Z — plus the #347-amended fire sequence, now
-  **COMPLETE end to end (2026-07-23) — gate FIRED**: Bug-B whole-DB
-  reparse dry-run, B(i) write, B(ii)+B(iii) retraction, the Bug-A
-  forced-escalation sample, the 4c pilot dry-run, the owner sample
-  audit, the pilot `--write` (130,210 votes, DB-verified exact match to
-  the dry-run's prediction), and `consensus_recompute --apply` (49,206
-  tag transitions, printing 3→4 resolved) all DONE, and #340 root-cause
-  footprint sizing. Bug-A's full re-scan remains deferred post-pilot as
-  the one tracked open item. Single source of
-  truth for this
+  **COMPLETE end to end, gate FIRED, true completion 2026-07-24**:
+  Bug-B whole-DB reparse dry-run, B(i) write, B(ii)+B(iii) retraction,
+  the Bug-A forced-escalation sample, the 4c pilot dry-run, the owner
+  sample audit, the pilot `--write` (130,210 votes, DB-verified exact
+  match to the dry-run's prediction), `consensus_recompute --apply`
+  (49,206 tag transitions), plus five further 2026-07-24 corrective/
+  completion passes (lexicon-gate retraction, marker reparse,
+  artist-credit fill, calculator re-pass, a second `consensus_recompute`
+  closer) all DONE and #340 root-cause footprint sizing. Live
+  resolved-printing count is **3** (a provisional 4 was corrected back
+  by the second closer); all 218,345 cards remain
+  `artist_vote_status=unresolved` (single-machine-vote-below-threshold
+  finding); review queue is **134,370** cards. Bug-A's full re-scan
+  remains deferred post-pilot as the one tracked open item. Single
+  source of truth for this
   gate's status — don't restate gate status/decisions elsewhere, link
   here.
 - [`docs/documentation-process.md`](docs/documentation-process.md) — docs/

docs/MANIFEST.md

@@ -43,9 +43,15 @@ The methodology and the systems it governs.
   2026-07-22, now closed history per the 2026-07-23 new-data basis
   ruling) and artifact 2 status (all 3 MISSING constants decided, fix
   for 2 merged AND deployed 2026-07-23T01:33Z), the #347-amended fire
-  sequence — now **COMPLETE end to end, gate FIRED (2026-07-23)**: Bug-B
-  whole-DB reparse dry-run, the pilot dry-run and `--write` (130,210
-  votes), and `consensus_recompute --apply` all DONE (Bug-A's full
+  sequence — now **COMPLETE end to end, gate FIRED, true completion
+  2026-07-24**: Bug-B whole-DB reparse dry-run, the pilot dry-run and
+  `--write` (130,210 votes), `consensus_recompute --apply`, plus five
+  further 2026-07-24 corrective/completion passes (lexicon-gate
+  retraction, marker reparse, artist-credit fill, calculator re-pass, a
+  second `consensus_recompute` closer) all DONE — live resolved-printing
+  count is **3** (corrected from a provisional 4), all 218,345 cards
+  remain `artist_vote_status=unresolved` (single-machine-vote-below-
+  threshold finding), review queue is **134,370** cards (Bug-A's full
   re-scan deferred post-pilot is the one tracked open item) — and the
   #340 root-cause footprint sizing. Single source of
   truth for this gate's status — `theory.md`, `identification-pipeline.md`,

docs/README.md

@@ -7,7 +7,14 @@ doc is the single, un-fragmented home for the round shipped 2026-07-23
 (`SPEC-display-left-rail.md`, owner-approved; built into PR #352) — see
 [`docs/upstreaming/readiness-audit.md`](../upstreaming/readiness-audit.md)'s
 §10 for the presentation-only upstream-divergence ledger this round seeded
-(not duplicated here).
+(not duplicated here). A corrected fidelity round (also 2026-07-23,
+Yori) normalized every rail block-boundary divider to `#16202b` (O1,
+below). A follow-up machine-diff round (same day) fixed 63 further
+mismatches a computed-style diff against the corrected mockup caught —
+mostly Bootstrap body-default font-size fall-throughs, plus (per an
+explicit owner ruling) the `AutofillCollapse` header hex, the Source
+toggle's visual shape, and the Filters toggle's `.btn-sm` sizing — see
+the "Machine-diff fix round" section below.
 
 Companion design artifacts (spec + mockup) live at
 [`docs/proposals/mockups/proposal-h/display-left-rail-mockup.html`](../proposals/mockups/proposal-h/display-left-rail-mockup.html)
@@ -270,7 +277,7 @@ bulk/pin/save-defaults controls (above). The empty-state "Find this card
 ↗" link stays a link — pure navigation out to Scryfall, the rule's own
 explicit exception.
 
-## Density (mechanical, SPEC §2)
+## Density (mechanical, SPEC §D.1)
 
 Rail blocks now butt against each other separated by 1px borders —
 vertical rhythm comes from each block's own compact padding
@@ -281,6 +288,102 @@ behavior changed — see `proposal-h-display-layout-spec.md` §4.1 for the
 unchanged R2 shell (Offcanvas placement per tier, 380px inline width,
 etc.); this round only restyles/recomposes the rail's interior.
 
+## O1 — divider normalization (corrected fidelity round, 2026-07-23)
+
+The rail's own block-boundary hairlines were inconsistent: `.d14` (the
+confidence band) already used an explicit `#16202b`, while `.rail-head`,
+`.artist-line`, and the Sources accordion's outer wrapper used the plain
+Bootstrap `.border-bottom` utility — whose active `--bs-border-color` is
+genuinely ambiguous in this theme's compiled CSS (both `#495057` and
+`#ced4da` are present), risking a pale line on the dark rail depending on
+cascade order. The unified Frame+Treatment filter's own border (and its
+internal Frame↔Treatment divider) separately hardcoded the unthemed
+`rgba(0,0,0,.22)`. The Select Version wrapper had no boundary divider at
+all.
+
+**Normalized (owner-approved, corrected `SPEC-display-left-rail.md` §A/
+§D.1) — every one of the above now explicitly renders `#16202b`, 1px**:
+`.rail-head`, `.artist-line`, and `.sources` (all three now via
+`RailRoot`'s own styled-component rules in `DisplayPage.tsx`, replacing
+the Bootstrap utility classes they used to carry); the Select Version
+wrapper (gained a `select-version-wrapper` className plus a new
+`RailRoot` rule — it never had a boundary before); the unified filter
+`fieldset`'s own border and its internal `UnifiedFilterDivider` (both
+inline/styled-component literals in `SelectVersionResults.tsx`). The
+Sources list's own inner border was deliberately left at `rgba(0,0,0,.22)`
+— the spec's own binding table marks that one specifically as unchanged
+(`I`, not `I (border N)`), not part of O1's scope. Each source ROW's own
+bottom divider was ALSO meant to stay `rgba(0,0,0,.22)` per that same
+table, but O1's own named list (`.rail-head`/`.artist-line`/`.sources`
+only) missed that it was ALSO using the ambiguous Bootstrap utility and
+resolving to the wrong colour (`#ced4da`) — caught and fixed in the
+machine-diff round below, not this one.
+
+## Machine-diff fix round (2026-07-23)
+
+A throwaway Playwright/computed-style diff (session tmp dir, not
+committed) measured the corrected mockup against the O1-round build and
+found 63 property mismatches — mostly Bootstrap body-default (16px)
+font-size fall-throughs on bespoke rail classnames that had never had
+their own font-size rule at all (`.rail-head .slot` -> 14px, `.rail-head .name` -> 15px + `margin-top:1px`, `.artist-line` -> 13px (replacing the
+close-but-not-exact Bootstrap `small` utility), `.select-version-heading`
+-> 14px), plus the Sources filter input's font-size/padding (Bootstrap's
+stock `.form-control` -> the spec's `14px`/`6px 10px`, fixed as a
+component-scoped inline style, not a `.form-control` RailRoot selector),
+the source row's own border-bottom (the same ambiguous-Bootstrap-utility
+pattern O1 missed on this one element, now `rgba(0,0,0,.22)` explicitly),
+the tile corner tag (`0.5rem`/alpha `.9` -> the spec's `7px`/alpha
+`.92`), and the Ghost "+N" tile's padding (a real `<button>` needed an
+explicit `padding:0` reset against the browser's own UA-stylesheet
+default).
+
+Three further rows the diff also flagged were held back for an explicit
+owner ruling rather than applied speculatively — **all three are now
+resolved, same day, per that ruling** (the corrected mockup is the
+binding reference for all three, confirmed by the owner):
+
+- **`AutofillCollapse` header background hex** — `#4E5D6B` is REVERTED
+  from PR #400's own "correction" to `#4e5d6c` (`$secondary`). The owner
+  confirmed `#4E5D6B` is deliberate, distinct from the panel/seticon
+  token elsewhere in the rail, one hex digit apart by design. Do not
+  "fix" this back again — `AutofillCollapse.tsx`'s own comment at this
+  line carries the note.
+- **Source toggle shape** — restyled from the shared
+  `react-bootstrap-toggle` library's stock sliding single-label switch
+  into the corrected mockup's static two-cell segmented control (both
+  On/Off labels always visible). The library's own DOM already renders
+  both cells with the correct colour classes unconditionally; only its
+  `overflow:hidden`/sliding-`.toggle-group` CSS needed overriding, scoped
+  to a `rail-source-toggle` className `SourcesAccordion.tsx`'s Toggle
+  passes — every other `react-bootstrap-toggle` mount sitewide (roughly
+  ten other call sites: `FinishSettings`, `PDFGenerator`,
+  `SearchTypeSettings`, the filter Toggles, etc) is unaffected.
+- **`.btn-sm` sizing** — the Filters disclosure toggle (`CompactButton`,
+  `SelectVersionResults.tsx`) returns to the spec's real Bootstrap `sm`
+  metrics (`14px`/`4px 8px`), superseding the earlier "the buttons are
+  too big" shrink for that one control. `CompactButton` has exactly one
+  call site, so this is component-scoped by construction.
+  `CompactToggleButton` (Frame/Border segments) and `TreatmentChip`
+  (Treatment chips) are UNCHANGED — they bind to their own distinct,
+  still-in-force spec rows (`11px` each), never the generic `.btn-sm`
+  row. Every OTHER `.btn-sm`-class rail control (Sources bulk/save-
+  defaults, Slot Actions, Artist support, D14 `✗`) was already at
+  `14px`/`4px 8px` via plain `<Button size="sm">` with no override at
+  all — this ruling only ever affected the one Compact-wrapped control.
+
+**Deliberately left open this round**: "Demoted body" `13px` (§D.1) — a
+known gap the diff's own measurement surfaced (via Slot Actions' body
+font-size), not fixed because the mechanical repair (an additive
+`bodyFontSize?: string` prop on `AutofillCollapse.tsx`, mirroring the
+existing `headerPadding` precedent) would touch every demoted
+`RailSection` — Card Details/Attributes/Printing Tags/Print Options/
+Report, not just Slot Actions (the only one the diff actually measured)
+— and the blast radius across the other four wasn't verified. See
+`SPEC-display-left-rail.md` §D.1's own note under the "Demoted body" row
+for the full reasoning.
+
+`frontend/tests/DisplayLeftRailFidelity.spec.ts` asserts every fix above.
+
 ## File-level summary
 
 | File                                             | Change                                                                                                                                                                          |