Bridge Town V1: native water-colour study

Previous goal turn: progress. It verified unattended Railway access, diagnosed the production gallery click-handler failure and published a focused correction through successful branch/main CI. The same Railway rollout is being observed; no duplicate deployment is started.

The original Bridge Town V1 (WOG-A-003714) and its exact named export (WOG-A-008110) remain intact. The current partial study, WOG-A-009275, corrects 1,237 selected front-river RGB samples at native 1254 × 1254 dimensions. This includes 302 opaque water samples. All source alpha, the complete town above row995, bright foam and every outside-mask RGBA byte remain exact. Weighted-luminance rounding differs by at most0.481. This is a colour correction in the manually bounded water outlet; it does not resample, blur, crop, remove silhouettes or introduce a repeating texture.

Independent PNG/Sharp comparisons cover both complete native images. Two native browser comparisons on light/dark backgrounds detect the same1,237 changed samples, no outside-mask display differences and exact stationary redraws. Source/output hashes were rechecked after verification. All browser/server resources are terminal, and no raw buffers or new screenshots were retained. The complete final artwork and native before/after comparison were directly inspected.

The current study reduces the selected purple islands while retaining the architecture, wave strokes, composition and fine detail. A grey/purple boundary is still visible. Broadening the purple predicate adds only44 candidates, which is insufficient evidence to repaint the remaining grey outline. No extra variant was generated. The approved native Gabled Court3363 was directly viewed as an architectural-detail benchmark with its exact source approval hash; its approval does not transfer to Bridge Town.

The renderer-role audit finds neither the original nor this study in the354 local approvals,322 release approvals or regional object descriptors. Existing canonical-place selection matches feature/profile/existing state and occupancy; a generic town placement or dry-ground fit does not establish this painted two-bank channel, river mouth, road/bridge approaches or wheel contact. Those facts, whole-artwork acceptance, physical scale, normal desktop/phone Region/Hex, full performance/lifecycle, source admission and actual-live appearance remain open. Painted market and wheel details create no gameplay production or crossing rule.

This is the current preferred partial water-colour study, not an approved final/game-ready source. Catalog/ledger registration and live Atlas publication are verified for commit74e7569. The authoritative catalog contains9,206 files and9,231 goal records,with5,634 pending qualifications,including the newly registered partial study. Registration reconciled against the verified correction release1d9d6b9 closeout. The four ground-role findings and30-frame full wetland trace are registered in the same batch.

Retention: both original copies and the current native study are kept. One206,122-byte source-only diagnostic crop was retired after complete pixel comparisons proved that both its light/dark source panes are already preserved in the retained native comparison. The retirement receipt keeps its exact hash and replacement mapping. No catalogued artwork or earlier blocked cleanup set was touched.

Evidence: repair water-colour-repair.json, native visual finding native-visual-review.json, native validation native-validation.json, renderer role renderer-role-audit.json, retention diagnostic-retirement.json.

Local registration and gallery preflight

An executed in-memory registration check validates eleven prospective metadata changes, preserving all existing approvals,checksum rows and nonparticipants. It assumes the preceding44 metadata rows have first been verified and closed; it does not clear those rows or modify any current catalog/ledger root. The projection is9,206 catalog files,9,231 ledger records and5,634 pending qualifications because the new partial study brings its own open record. Two inherited fringe labels are queued to change from smoke to fringe.

The prepared six-image gallery presents the preserved Bridge Town,desert/meadow and wetland-apron originals beside their studies, with no new artwork copies. All six source bytes match the existing catalog or the explicitly queued native receipt. Both desktop and phone pass complete loading,light/dark pointer and keyboard round trips and no-overflow checks under production CSP. No screenshot or raw buffer is persisted. The page explicitly labels every study as unqualified and the Bridge Town Atlas registration as pending.

Registered local batch

Eleven metadata records and one native study are registered against the verified preceding release. All existing checksum rows,approvals and nonparticipants are preserved. Counts are9,206 catalog files,9,231 goal records and5,634 pending qualifications; the new partial study adds one open record. The original32-pixel failure and all source-role/native-quality/performance/lifecycle/live gates remain open. The two inherited wetland labels now correctly say fringe. The comparison gallery is rebuilt after registration so its study link resolves to the new local Atlas record. Final desktop/phone Atlas and publication verification passed after registration.

Verified live release

Commit74e756906f650e5e9d8b78160734a647b23c8e86 deployed automatically as26e20968-e47b-4494-b139-bc146a2a2b80. Exact live metadata/inventory membership,46 selected native HTTP identities,22 desktop/phone Atlas views and four gallery layouts passed with zero browser errors. All eleven matching metadata rows are synchronized. The full library retains9,206 files,9,231 goal records and5,634 open qualifications. There are no new source approvals,game-world placements or full game-ready assets. The separately prepared local portrait-sampling candidate is excluded from this release.