Bug 1408537: Slightly increase the epsilon in MathML reftest "font-inflation-1.html", so that it passes on local linux reftest runs. r?jfkthame
The prior epsilon (2px) seems like an arbitrarily chosen value, which must have
been large enough to allow for small sizing differences on our continuous
integration platforms. But it's not quite large enough to handle sizing
differences on my local linux machine (I'm guessing due to font differences).
Note that this doesn't reduce the sensitivity of the test. In the test's
anticipated failure mode (when font inflation fails to apply on some element),
the sizing differences are an order of magnitude larger than this amended
epsilon.
MozReview-Commit-ID: Iz7jqRvFh8w
--- a/layout/reftests/mathml/font-inflation-1.html
+++ b/layout/reftests/mathml/font-inflation-1.html
@@ -9,17 +9,17 @@
src: url(../fonts/math/dtls-1.otf);
}
p, math {
font-family: serif;
}
</style>
<script type="text/javascript">
function almostEqual(aX, aY) {
- var epsilon = 2;
+ var epsilon = 2.6;
return Math.abs(aX - aY) < epsilon;
}
function verifySize(aElement, aReference) {
/* Verify if the size of the element matches the reference, and
otherwise paint the element in red. */
if (!almostEqual(aElement.getBoundingClientRect().height,
aReference.getBoundingClientRect().height) ||