Any reason mark this as wontfix?

There are two issues. Click a link will open a new tab, which is not supported. See issue 547231.

We should be able to scroll. The log shows the engine is alive. But there is no client log, can't tell.

[Bulk edit]

This Blimp bug is targeted to M-52, which ends today.  Please change the target milestone to M-53 (which ends June 30) if appropriate and if your team lead signs off.  Any items still targeted to M-52 on Monday will have their milestone tags removed.

Please do *not* remove the label Blimp-M52-Proj-Scope if it exists on this bug; that is a scope tracking label only.  Similarly, do not add Blimp-M53-Proj-Scope yet either; I'll take care of that.

Ping me with any questions, thanks!

Do we have any new information on why this site doesn't seem to work?

Haven't gotten around to it yet.

[Bulk edit]

Setting tracking label Blimp-M53-Proj-Scope.  This label is for scope tracking purposes only and should not be added / removed from any bugs, even if we add additional bugs to M-53 scope, or remove this bug from M-53 scope.

Tried debugging this a bit today. We are getting wrong values from Webkit. The bounds of the scrollable layer are incorrect, which is why the scroll is getting clamped on the client.

I need to see why we are getting different values from Webkit though.

news page is scrollable on nexus 5. But it's not scrollable on nexus 6P for blimp.

I added a log to print out the bounds set to the layer tree.
diff --git a/cc/layers/layer.cc b/cc/layers/layer.cc
index b02c456..606d7c9 100644
--- a/cc/layers/layer.cc
+++ b/cc/layers/layer.cc
@@ -11,6 +11,7 @@
 
 #include "base/atomic_sequence_num.h"
 #include "base/location.h"
+#include "base/logging.h"
 #include "base/metrics/histogram.h"
 #include "base/single_thread_task_runner.h"
 #include "base/time/time.h"
@@ -315,6 +316,10 @@ void Layer::SetBounds(const gfx::Size& size) {
   if (masks_to_bounds())
     SetSubtreePropertyChanged();
   SetNeedsCommit();
+
+  if (user_scrollable_vertical()) {
+    LOG(INFO) << "Bound is " + size.ToString();
+  }
 }

The logcat of opening news page on blimp and clank are attached.

One other thing i noticed:
if I manually change the user_agent on nexus 6p to "Linux; Android 6.0.1; Nexus 5 Build/MOB30F", which is the user_agent on nexus 5, then the news page is scrollable on nexus 6p.

Deleted: blimp_6p_logcat 5.8 KB

blimp_6p_logcat 5.8 KB View Download

Deleted: clank_6p_logcat 1.2 KB

clank_6p_logcat 1.2 KB View Download

not sure if it's a blink bug or a cc bug. + trchen@ for more information

Different user agent just means that we are getting a different version of the page, which is why you're not seeing it if you change the user agent.

When I tried debugging it, I traced it to getting a different value for the bounds of the layer from blink on this. Is there something that we might be missing because Clank is an android build but the Blimp engine is linux?

[Bulk edit]

Setting tracking label Blimp-M54-Proj-Scope.  This label is for scope tracking purposes only and should not be added / removed from any bugs, even if we add additional bugs to M-54 scope, or remove this bug from M-54 scope.

Turns out news.google.com serves Nexus 6p a special version of the page.

On normal mobile page, the layout overflow is propagated all the way up to the document. The resulting document will be ~6000px long and scrolled by the browser's native scrolling.

On the special version, both <div class="b-Jb-yb kc Vc" id=":4"> and <div class="b-Jb-N Ud-N" navid="FRONTPAGE" id=":5"> has overflow:hidden that clipped their contents. The page implemented its own touch event handler to simulate scrolling by apply transforms on the contents. As Blimp doesn't support touch event handling yet, the simulated scrolling won't work.

I don't know why news.google.com added the hack. It is bad for performance even on regular Chrome. We should let them know the hack should be removed if possible.

Thanks for helping investigate on this! i filed a bug b/30787958 and ask if they can remove the hack. 

marking this bug as won't fix