eliott/ladybird
1
0
Fork 0
mirror of https://github.com/LadybirdBrowser/ladybird synced 2026-04-26 17:55:07 +02:00
Code Issues Packages Projects Releases Wiki Activity

Tests/LibWeb: Test performance with relative timing rather than absolute

Browse Source 
The flaky test `input-value-does-not-force-layout.html` had an absolute
threshold of 5000ms before it would fail, which did not work that great
on CI. Since we are dealing with wildly different performance
characteristics on our CI runners (and ASAN enabled), make this test
perform relative timing comparisons instead.

Confirmed to still fail locally with the original fix removed.
This commit is contained in:
Jelle Raaijmakers
2026-03-12 16:41:06 +01:00
committed by Gregory Bertilson
parent 0019836fc1
commit cf050288fa
Notes: github-actions[bot] 2026-03-18 18:54:55 +00:00 
Author: https://github.com/gmta
Commit: https://github.com/LadybirdBrowser/ladybird/commit/cf050288fa7
Pull-request: https://github.com/LadybirdBrowser/ladybird/pull/8388
Reviewed-by: https://github.com/Zaggy1024 ✅
1 changed files with 13 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
Tests/LibWeb/Text/input/input-value-does-not-force-layout.html
View File

@@ -23,15 +23,24 @@
input.focus();
input.offsetWidth;
// Performance test: setting input.value must not synchronously trigger a full layout.
// Setting input.value with dirty layout must not be significantly slower than without it.
const values = ["a", "ab"];
const start = performance.now();
const baselineStart = performance.now();
for (let i = 0; i < 2000; i++)
app.children[i % 2000].style.padding = (4 + (i % 2)) + "px";
const baseline = performance.now() - baselineStart;
input.offsetWidth;
const testStart = performance.now();
for (let i = 0; i < 2000; i++) {
app.children[i % 2000].style.padding = (4 + (i % 2)) + "px";
input.value = values[i % 2];
}
const elapsed = performance.now() - start;
const test = performance.now() - testStart;
println(elapsed < 5000 ? "PASS" : `FAIL: took ${elapsed.toFixed(0)}ms`);
const ratio = test / Math.max(baseline, 1);
println(ratio < 5 ? "PASS" : `FAIL: ratio ${ratio.toFixed(1)}x (baseline=${baseline.toFixed(0)}ms, test=${test.toFixed(0)}ms)`);
});
</script>