MSE, MS, and any blob URL: Ignore preload 'none' on resource fetching

third_party/WebKit/Source/core/html/HTMLMediaElement.cpp

Index: third_party/WebKit/Source/core/html/HTMLMediaElement.cpp
diff --git a/third_party/WebKit/Source/core/html/HTMLMediaElement.cpp b/third_party/WebKit/Source/core/html/HTMLMediaElement.cpp
index 2d84e10503d87820f00dd1cdd9fbc8d8c0b20027..8a7762fe3abf71522413f9c70e2f54e6f5c7db4e 100644
--- a/third_party/WebKit/Source/core/html/HTMLMediaElement.cpp
+++ b/third_party/WebKit/Source/core/html/HTMLMediaElement.cpp
@@ -1095,7 +1095,7 @@ void HTMLMediaElement::loadResource(const WebMediaPlayerSource& source, ContentT
     if (attemptLoad && canLoadResource) {
         ASSERT(!webMediaPlayer());
 
-        if (effectivePreloadType() == WebMediaPlayer::PreloadNone) {
+        if (!m_mediaSource && effectivePreloadType() == WebMediaPlayer::PreloadNone) {

[philipj_slow, 2016/04/14 12:39:27]

I would like it very much if this matched the behavior of
https://github.com/whatwg/html/pull/1037 even if not done in precisely the same
place. In other words, it'd be nice if MediaStream was also exempted here.

Can you use isObjectOrBlobURL perhaps? That doesn't quite match the spec because
it'll be true even for a bogus blob: URL, but maybe that's a better solution
anyway. If it's a bogus blob: URL the load will just fail quicker.

[wolenetz, 2016/04/21 21:37:30]

I've added isObjectOrBlobURL usage here, but adding a related MediaStream layout
test is interesting: it looks like that would best be done in the
web-platform-tests repo, and then imported into Blink the way other media-stream
imported tests are. I'll put together such a wpt test PR shortly and eventual
import of it should hopefully satisfy the mediastream portion of this CR
question.

I've added an invalid blob url non-MSE/non-MS test with preload=none into this
CL.

             WTF_LOG(Media, "HTMLMediaElement::loadResource(%p) : Delaying load because preload == 'none'", this);
             deferLoad();
         } else {