Skip to content

A fast 2D and 3D game engine — TypeScript SDK, C++/WebAssembly core, visual editor. Ship one project to web, desktop, WeChat MiniGames, playable ads, and native Android / iOS.

docs(2d): record both renderer acceptance results

yhhe8cc4c4ff855
diff --git a/examples/platformer/VALIDATION.md b/examples/platformer/VALIDATION.md
index 6830d3673..30962c385 100644
--- a/examples/platformer/VALIDATION.md
+++ b/examples/platformer/VALIDATION.md
@@ -25,14 +25,30 @@ texture's shared 8-pixel border change.
Windowless regression: editor Rect drag 11 and texture-border geometry 19 tests;
SDK import settings 13, generated field metadata 1, scene format/loading/
-validation 84; pixel-predicate tests 4. No local desktop/Electron verifier ran.
+validation 84; pixel-predicate tests 4. Two existing post-process test fixtures
+were brought up to date with quality filtering and initialization state; their
+31 rendering/present tests also pass (163 local tests in total). All 63 example
+projects pass type checking. No local desktop/Electron verifier ran.
## Second backend
The built-in browser reports **no WebGPU adapter**. That is unavailable coverage,
-not a pass. The same full-corner predicate is registered as the PR-tier
-`sprite-nine-slice` scene on WebGL2 and WebGPU, so a CI machine with an adapter
-can run it. The roadmap must retain pending acceptance until this passes.
+not a local pass. The same full-corner predicate is registered as the PR-tier
+`sprite-nine-slice` scene on WebGL2 and WebGPU. Both passed in
+[branch CI](https://github.com/esengine/estella/actions/runs/37783264188):
+WebGL2 in 1.0s and WebGPU in 7.9s. These are whole verifier durations, not frame
+times. The complete renderer jobs passed on both backends. The run predates
+the post-process mock repair; engine code, fixture and pixel predicate are identical.
+
+The full build was not green: the old post-process mocks caused 14 SDK failures,
+and editor camera-preview CPU budgets / selected-only 3D gizmo checks also
+failed on master before this increment. These are distinct from the passing
+render jobs. Six stale API snapshots were synchronized to unblock static checks.
+
+The exported test copy temporarily gave the canvas `tabindex=0` so the browser
+locator could focus it for Space/ArrowRight. The engine and game code were
+unchanged; jump and movement rendered without console errors. The attribute
+was removed afterward so the retained package has the original export markup.
## Reproduce without desktop windows
diff --git a/tools/sprite-nine-slice-probe.mjs b/tools/sprite-nine-slice-probe.mjs
index 486d7a5a9..63499edaf 100644
--- a/tools/sprite-nine-slice-probe.mjs
+++ b/tools/sprite-nine-slice-probe.mjs
@@ -74,8 +74,7 @@ button.onclick = async () => {
} catch (error) { checks.push({ name: 'Probe error', pass: false, detail: error.stack }); }
const report = { backend, userAgent: navigator.userAgent, checks, passed: checks.filter(c => c.pass).length, failed: checks.filter(c => !c.pass).length };
result.textContent = JSON.stringify(report, null, 2);
- summary.textContent = `${backend}: ${report.passed} passed, ${report.failed} failed`;
+ summary.textContent = `${backend}: ${report.passed} passed, ${report.failed} failed. Reload to repeat.`;
const saved = await fetch('/report', { method: 'POST', body: JSON.stringify(report) });
if (!saved.ok) result.textContent += '\nReport could not be saved';
- button.disabled = false;
};