MSE: Add layout test for 'decode' error via an algorithm

LayoutTests/http/tests/media/media-source/mediasource-errors.html

Index: LayoutTests/http/tests/media/media-source/mediasource-errors.html
diff --git a/LayoutTests/http/tests/media/media-source/mediasource-errors.html b/LayoutTests/http/tests/media/media-source/mediasource-errors.html
index bb839522155b8a2d222ba16f977886464d8a5c76..fca9b770ee739bf1084c42e062cf8c153fd780d8 100644
--- a/LayoutTests/http/tests/media/media-source/mediasource-errors.html
+++ b/LayoutTests/http/tests/media/media-source/mediasource-errors.html
@@ -143,6 +143,28 @@
                     test.done();
                 });
             }, "Signaling 'network' error via endOfStream() after initialization segment has been appended and the HTMLMediaElement has reached HAVE_METADATA.");
+
+            ErrorTest(function(test, mediaElement, mediaSource, segmentInfo, sourceBuffer, mediaData)
+            {
+                assert_equals(mediaElement.readyState, HTMLMediaElement.HAVE_NOTHING);
+
+                var mediaSegment = MediaSourceUtil.extractSegmentData(mediaData, segmentInfo.media[0]);
+                var index = segmentInfo.init.size + (mediaSegment.length + 1) / 2;

[philipj_slow, 2014/11/20 09:02:02]

I don't understand the +1 here. The lowest index of the media segment ought to
be "segmentInfo.init.size" and the highest index "segmentInfo.init.size +
mediaSegment.length - 1", so to scale between those with a factor [0,1] (in this
case 0.5) the formula would be "segmentInfo.init.size + (mediaSegment.length -
1) / 2".

If there's some other intention to this formula, please document it :)

[jiajia.qin, 2014/11/20 10:12:23]

I just follow 'mediasource-append-buffer.html'. In that file's test case, use
'var harfIndex = (mediaSegment.length + 1) / 2'.
But I think no matter +1 or -1 won't affect this test. Here, I just random
select a point to corrupt the data.

[philipj_slow, 2014/11/20 10:45:45]

In that case, please use - 1 instead in both tests so that future readers don't
have to wonder if there's something special going on.

+                // Corrupt the media data from index of mediaData, so it can signal 'decode' error.
+                // Here use mediaSegment to replace the original mediaData[index, index + mediaSegment.length]

[philipj_slow, 2014/11/20 09:02:02]

If I'm reading this correctly, we'll the init segment, half of the media segment
and then the media segment again? That sounds corrupt enough to me :)

[jiajia.qin, 2014/11/20 10:12:24]

Yes, it is :)

+                mediaData.set(mediaSegment, index);
+
+                test.expectEvent(mediaElement, "error", "mediaElement error.");
+                sourceBuffer.appendBuffer(mediaData);
+
+                test.waitForExpectedEvents(function()
+                {
+                    assert_equals(mediaElement.readyState, HTMLMediaElement.HAVE_METADATA);
+                    assert_true(mediaElement.error != null);
+                    assert_equals(mediaElement.error.code, MediaError.MEDIA_ERR_DECODE);
+                    test.done();
+                });
+            }, "Signaling 'decode' error via segment parser loop algorithm after initialization segment and partial media segment has been appended.");
         </script>
     </body>
 </html>