WORLD OF GONZ WIKIOpen game ↗

Engineering, operations & design history

Actual visible-map and default-budget pressure qualification

Full repository source: docs/art-direction/display-cache-history-20261104/visible-pressure-v2/review.md

Technical documentation includes historical decisions, proposals and verification reports. For playing instructions and practical rules, open Gameplay.

Actual visible-map and default-budget pressure qualification

The existing grasslands Gabled Court town248797 is selected by the unchanged normal world and exact original source contract. This continuation records68 normal-client frames across desktop and phone DPR2 layouts:36 town-source frames and32 primary-material frames at four existing natural biomes. Thirty-four frames use the private place-planner candidate. Together with the preceding52 town observations,the retained client evidence totals120 frames;the Gabled Court's own-source scope is88,not120. No source alias,world command,geography change or admission is used.

PlannerCSS viewport widthMatched cold/warm displayed pixels differingPressure-return displayed pixels differingNative-cache evictions during pressure
control144081,51381,52194
candidate14400093
control39081,196080
candidate3900070

All four cold/warm and all four pressure-return source transforms match exactly. The active terrain is exact in all four matched cold/warm comparisons and both private pressure returns. The control-phone pressure surface has four raw pixel differences,maximum channel difference44,while its displayed map is exact. Those raw differences are retained without localization,cause or waiver;raw terrain-buffer stability remains open. The renderer swaps worldStatic and worldStaticNext:the same HTML canvas ID can refer to an old hidden surface. Four cold/warm cases observe this role swap. The earlier large raw-cache differences therefore did not establish a visible-terrain regression in these cases. The new visibleMap comparison captures the complete map panel after normal composition;no raw buffer or passing PNG copy is saved.

The private candidate matches both cold/warm and pressure-return visible map panels exactly on both layouts. Fourteen selected-source replays are native-exact. The control phone pressure return is also exact;the improvement is not claimed for a case that already passed. The exact original Gabled Court SHA-256 remains7ba2c669cf81289dce6d88c29ac7097693ac6babe7fe70992f49c52d6d5eb3fc. This is scoped fidelity evidence,not isolated attribution of every changed control pixel to one source or a whole-world certificate.

The pressure tour uses actual arctic,desert,forest and wetland sites,Region/Hex controls and the default256MiB native/32MiB display caches. Each primary material is independently source-hash bound and reported painted in eight frames:four Region and four Hex. Source invocation/coverage is not isolated final-pixel contribution,nonrepetition or artistic acceptance. Existing grain/palette/contact/state and all earlier gates remain unchanged. The default128/192MiB adaptive profiles remain untested in this normal-world continuation. A phone layout on this browser is not physical-phone certification.

Two method guards stopped incomplete attempts. Regional entry zoom depends on the viewport;the initial .3-only assumption was false. Direct focus from Hex can restore the prior Region camera even while the selected-cell ID changes;the corrected tour returns to Region before focusing each destination and verifies camera/source paint. The17 frames from those stopped attempts are retained separately and excluded from the68-frame completed run. Neither is an artwork-failure waiver.

The source observer releases its retained image reference before leaving the town,so it does not keep that original alive through the pressure route. Thirty-six descriptive navigation measurements exclude pixel readbacks from timed calls. Some CDP task counters reset across page navigations;negative cross-document deltas are not valid CPU costs. JavaScript heap samples exclude native/GPU/aggregate memory,and order/cache effects are not a controlled performance benchmark. Full cost,aggregate memory,crowded canonical contexts,lower adaptive budgets and lifecycle remain open. The private planner is not activated. The deployed module differs,and prior wetland32/max245 and larger Hex pressure failures remain unresolved.

All owned contexts and the server are terminal. Native artwork,public modules,approved source contracts and saves remain unchanged. Retention preserves every rendering input,comparison,hash and failure while removing duplicated game-overview payloads and duplicate progress case bodies. New passing screenshot copies and raw readback files:0.

Evidence:exact summary validation-summary.json,completed observations world-results.json,natural sites pressure-sites.json,retention retention.json.

Verified live Atlas publication

Commitedc632141c32298b84ef57304ccd6ffe9e3a8646 deployed automatically asf71a1d2d-cec2-4b74-a00d-7b4e56c8eb61. Exact live metadata and9206-member inventory,eight native HTTP artwork identities including all five source records and both hamlets,ten desktop/phone Atlas finding views,and six generated wiki pages passed. Five metadata rows are synchronized. No artist byte,source approval,public renderer or production-world placement changes. The four raw control-phone terrain pixels,prior Hex failures,lower-profile/full controlled cost,aggregate memory and lifecycle remain open. All5634 qualifications remain in scope.