mozilla-release/layout/style/test/test_bug1505254.html

<!DOCTYPE HTML>
<html>
<!--
https://bugzilla.mozilla.org/show_bug.cgi?id=1505254
-->
<head>
  <meta charset="utf-8">
  <title>Test for Bug 1505254</title>
  <script src="/tests/SimpleTest/SimpleTest.js"></script>
  <link rel="stylesheet" type="text/css" href="/tests/SimpleTest/test.css"/>
  <style>
    /* Note: this CSS/DOM structure is loosely based on WhatsApp Web. */
    #outerFlex {
      display: flex;
      height: 200px;
      border: 3px solid purple;
      overflow: hidden;
      position: relative;
    }
    #outerItem {
      flex: 0 0 60%;
      overflow: hidden;
      position: relative;
    }
    #abspos {
      position: absolute;
      display: flex;
      flex-direction: column;
      height: 100%;
      width: 100%;
    }
    #insideAbspos {
      position: relative;
      flex: 1 1 0;
      width: 100%;
      height: 100%;
    }
    #scroller {
      display: flex;
      flex-direction: column;
      position: absolute;
      top: 0;
      overflow-x: hidden;
      overflow-y: scroll;
      height: 100%;
      width: 100%;
    }
    #initiallyHidden {
      display:none;
    }

  </style>
</head>
<body>
<a target="_blank" href="https://bugzilla.mozilla.org/show_bug.cgi?id=1505254">Mozilla Bug 1505254</a>
<div id="display">
  <div id="content">
    <div id="outerFlex">
      <div id="outerItem">
        <div id="abspos">
          <div id="insideAbspos">
            <div>
              <div id="scroller">
                <div style="min-height: 600px">abc</div>
                <div id="initiallyHidden">def</div>
              </div>
            </div>
          </div>
          <div id="testNode"></div>
        </div>
      </div>
    </div>
  </div>
</div>
<pre id="test">
<script type="application/javascript">
"use strict";

/** Test for Bug 1505254 **/

/**
 * This test checks how many reflows are required when we make a change inside
 * of an abpsos element, which itself is inside of a flex item with cached
 * block-size measurements. This test is checking that this sort of change
 * doesn't invalidate those cached block-size measurements on the flex item
 * ancestor. (We're testing that indirectly by seeing how many frames are
 * reflowed.)
 */

const gUtils = SpecialPowers.getDOMWindowUtils(window);

// The elements that we will modify here:
const gInitiallyHidden = document.getElementById("initiallyHidden");
const gTestNode = document.getElementById("testNode");

// Helper function to undo our modifications:
function cleanup()
{
  gTestNode.textContent = "";
  gInitiallyHidden.style = "";
}

// Helper function to flush layout & return the global frame-reflow-count:
function getReflowCount()
{
  let unusedVal = document.getElementById("scroller").offsetHeight; // flush layout
  return gUtils.framesReflowed;
}

// This function adds some text in gTestNode and returns the number of frames
// that need to be reflowed as a result of that tweak:
function makeTweakAndCountReflows()
{
  let beforeCount = getReflowCount();
  gTestNode.textContent = "def";
  let afterCount = getReflowCount();

  let numReflows = afterCount - beforeCount;
  if (numReflows <= 0) {
    ok(false, "something's wrong -- we should've reflowed *something*");
  }
  return numReflows;
}

// ACTUAL TEST LOGIC STARTS HERE
// -----------------------------

// "Reference" measurement: see how many frames need to be reflowed
// in response to a tweak in gTestNode, before we've shown
// #initiallyHidden:
let numReferenceReflows = makeTweakAndCountReflows();
cleanup();

// "Test" measurement: see how many frames need to be reflowed
// in response to a tweak in gTestNode, after we've shown #initiallyHidden:
gInitiallyHidden.style.display = "block";
let numTestReflows = makeTweakAndCountReflows();
cleanup();

// Any difference between our measurements is an indication that we're reflowing
// frames in a non-"dirty" subtree. (The gTestNode tweak has no reason to cause
// #initiallyHidden to be dirty -- and therefore, the presence/absence of
// #initiallyHidden shouldn't affect the number of frames that get reflowed in
// response to the gTestNode tweak).
//
// More importantly: if our measurements differ by *more than 1*, then that
// indicates that we're reflowing the non-"dirty" #initiallyHidden subtree
// *multiple times* in response to the tweak in gTestNode. We definitely want
// to avoid that.
//
// XXXdholbert The "+1" in the is() comparison here is a bit magical, and it
// probably indicates that we are still doing some unnecessary work. But it's
// an improvement -- without the fix that's accompanying this test, we get +5
// reflows here instead of +1. Hopefully we can get rid of the +1 at some point.
is(numTestReflows, numReferenceReflows + 1,
   "Tweak should trigger roughly the same number of reflows regardless of " +
   "content in unmodified sibling");

</script>
</pre>
</body>
</html>