Bug 70672 – Make relayout-nested-positioned-elements-crash.html more reliable

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Benjamin Poulain]

The test relayout-nested-positioned-elements-crash.html is flaky when executing WebCore asynchronously. The reason is the test written in a weird way.

In the follow code, the code of runTest() _cannot_ be executed in order for the test to succeed. In the test result, the PASS is followed by a whitespace. That whitespace is what the TextIterator output for the input element, because it still has a renderer, because runTest() was never executed.

If you execute WebCore asynchronously from the test runner, the function runTest() may or may not be run, causing the test to be flaky.

<p>This tests that we don't cause an assertion failure on relayout of nested positioned elements. This test PASSED if we don't cause an assertion failure.</p>
<div style="position:absolute">
    <span style="position:relative">
        PASS
        <div style="position:absolute">
            <input id="hideMe"/>
        </div>
    </span>
</div>
<script>
window.setTimeout(runTest, 0); // For some reason we need the setTimeout() for this test to work.
function runTest()
{
    document.getElementById("hideMe").style.display = "none";
}

Assuming the test is correct, we should rewrite it in a way that is more reliable.

[Benjamin Poulain]

Created attachment 112079 [details]
Patch

[Kenneth Rohde Christiansen]

Comment on attachment 112079 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=112079&action=review

I see this as an improvement, but I have a few comments.

> LayoutTests/ChangeLog:9
> +        Change the test to use the Javascript test framework.
> +        Ensure the element has the correct style in the final state.

Any comment on why or when it was unreliable?

> LayoutTests/fast/block/positioning/relayout-nested-positioned-elements-crash.html:17
> +description("This tests that we don't cause an assertion failure on relayout of nested positioned elements. This test pass if we don't cause an assertion failure.");

and the element has the correct style in the final state. ?

[Kenneth Rohde Christiansen]

The description is comment 1 is pretty good and could be summarized in the ChangeLog.

[Benjamin Poulain]

Committed r98197: <http://trac.webkit.org/changeset/98197>