Also note: this seems to be a regression in 70.x since downgrading to 69.0.3497.100 makes the problem disappear.

Switching this to Bug-Regression due to comment above.

What can we do to move this along? It's very concerning to have this out in the field since it means user's battery usage will likely be increased significantly and when looking at battery stats it will be whatever app that happens to use AdMob ads (which are many as far as I know) that will be blamed. I don't think those apps can do much besides switching off ads to work around this which is a step I'm sure neither apps nor AdMob prefers.

Reverting https://chromium-review.googlesource.com/c/chromium/src/+/1150961/ fixes this. I don't have a full explanation in my head yet. Scheduler state below.

needs_prepare_tiles_ is true (fine)
visible_ is true (a bit surprising)
can_draw_ is false (dunno what's setting it to false yet)

So we spin begin frame, but don't invalidate because can_draw_ is false. Questions:
* what's causing can_draw_ to be false?
* why is visible_ true when the app is in the background? this can happen on detached webviews, but need to confirm that
* why does invalidate actually fix this? I mean if webview is detached or otherwise not visible onscreen, what's responding to invalidate with draw?
* can we do better here? if we need to prepare tiles but can_draw_ is false, should we just prepare tiles with a delay or something?

Also noticed all the vsyncs are serialized on the UI thread. I wonder when we broke that :/

	
{state_machine: {major_state: {next_action: "Action::NONE",
                               begin_impl_frame_state: "BeginImplFrameState::INSIDE_BEGIN_FRAME",
                               begin_main_frame_state: "BeginMainFrameState::IDLE",
                               layer_tree_frame_sink_state: "LayerTreeFrameSinkState::ACTIVE",
                               forced_redraw_state: "ForcedRedrawOnTimeoutState::IDLE"},
                 minor_state: {commit_count: 2,
                               current_frame_number: 7857,
                               last_frame_number_submit_performed: -1,
                               last_frame_number_draw_performed: 9,
                               last_frame_number_begin_main_frame_sent: 9,
                               did_draw: true,
                               did_send_begin_main_frame_for_current_frame: false,
                               did_notify_begin_main_frame_not_sent: false,
                               wants_begin_main_frame_not_expected: false,
                               did_commit_during_frame: false,
                               did_invalidate_layer_tree_frame_sink: false,
                               did_perform_impl_side_invalidaion: false,
                               did_prepare_tiles: false,
                               consecutive_checkerboard_animations: 0,
                               pending_submit_frames: 0,
                               submit_frames_with_current_layer_tree_frame_sink: 0,
                               needs_redraw: false,
                               needs_prepare_tiles: true,
                               needs_begin_main_frame: false,
                               needs_one_begin_impl_frame: false,
                               visible: true,
                               begin_frame_source_paused: false,
                               can_draw: false,
                               resourceless_draw: false,
                               has_pending_tree: false,
                               pending_tree_is_ready_for_activation: false,
                               active_tree_needs_first_draw: false,
                               active_tree_is_ready_to_draw: true,
                               did_create_and_initialize_first_layer_tree_frame_sink: true,
                               tree_priority: "NEW_CONTENT_TAKES_PRIORITY",
                               scroll_handler_state: "SCROLL_DOES_NOT_AFFECT_SCROLL_HANDLER",
                               critical_begin_main_frame_to_activate_is_fast: false,
                               main_thread_missed_last_deadline: false,
                               skip_next_begin_main_frame_to_reduce_latency: false,
                               video_needs_begin_frames: false,
                               defer_commits: false,
                               last_commit_had_no_updates: false,
                               did_draw_in_last_frame: false,
                               did_submit_in_last_frame: false,
                               needs_impl_side_invalidation: false,
                               current_pending_tree_is_impl_side: false,
                               previous_pending_tree_was_impl_side: false}},

> what's causing can_draw_ to be false?

webview is 0x0

> why is visible_ true when the app is in the background? 

Yep, these are detached webviews except one. That one attached does become invisible when app is background.

> * why does invalidate actually fix this?

App is not responding to invalidates as I originally suspected. Before my CL, cc scheduler does call invalidate up with needs_draw=false if it's only required for prepare tiles. Then we have a shortcut for these that do not involve os or app code.


So fix is pretty easy. Keep invalidating if it's for needs_prepare_tiles

https://chromium-review.googlesource.com/c/chromium/src/+/1334550

> Also noticed all the vsyncs are serialized on the UI thread. I wonder when we broke that :/

Oh it's not broken. It's just that each webview is passed a different context, and we've only ever tried to optimize for webviews with the same context.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/43a7801fe8d646b7099d5e72a151c8eab5a3bffc

commit 43a7801fe8d646b7099d5e72a151c8eab5a3bffc
Author: Bo Liu <boliu@chromium.org>
Date: Wed Nov 14 02:33:06 2018

cc: Invalidate for needs_prepare_tiles

Problem is when CanDraw is false but needs_prepare_tiles is true
Scheduler will keep requesting begin frames but do nothing in a
begin frame since invalidate is blocked on CanDraw.

We already have an efficient shortcut that does not involve actually
producing a frame if invalidate is only requested for
needs_prepare_tiles. So the simple fix is simply allow invalidates for
this case.

Bug:  900557 
Change-Id: Ib3c3e213e2e2461d7aa1b8dd8d2e91ab64042feb
Reviewed-on: https://chromium-review.googlesource.com/c/1334550
Commit-Queue: Bo <boliu@chromium.org>
Reviewed-by: Sunny Sachanandani <sunnyps@chromium.org>
Cr-Commit-Position: refs/heads/master@{#607876}
[modify] https://crrev.com/43a7801fe8d646b7099d5e72a151c8eab5a3bffc/cc/scheduler/scheduler_state_machine.cc
[modify] https://crrev.com/43a7801fe8d646b7099d5e72a151c8eab5a3bffc/cc/scheduler/scheduler_unittest.cc

No issues on canary. Change only impacts webview

This bug requires manual review: M71 has already been promoted to the beta branch, so this requires manual review
Please contact the milestone owner if you have questions.
Owners: benmason@(Android), kariahda@(iOS), kbleicher@(ChromeOS), govind@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Approved for merge to 71, branch 3578.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/95e44621f06a1c3ffab2e3f860042da2fb254e48

commit 95e44621f06a1c3ffab2e3f860042da2fb254e48
Author: Bo Liu <boliu@chromium.org>
Date: Tue Nov 20 04:39:44 2018

[Merge M71] cc: Invalidate for needs_prepare_tiles

Problem is when CanDraw is false but needs_prepare_tiles is true
Scheduler will keep requesting begin frames but do nothing in a
begin frame since invalidate is blocked on CanDraw.

We already have an efficient shortcut that does not involve actually
producing a frame if invalidate is only requested for
needs_prepare_tiles. So the simple fix is simply allow invalidates for
this case.

(cherry picked from commit 43a7801fe8d646b7099d5e72a151c8eab5a3bffc)

Bug:  900557 
Change-Id: Ib3c3e213e2e2461d7aa1b8dd8d2e91ab64042feb
Reviewed-on: https://chromium-review.googlesource.com/c/1334550
Commit-Queue: Bo <boliu@chromium.org>
Reviewed-by: Sunny Sachanandani <sunnyps@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#607876}
Reviewed-on: https://chromium-review.googlesource.com/c/1343663
Reviewed-by: Bo <boliu@chromium.org>
Cr-Commit-Position: refs/branch-heads/3578@{#769}
Cr-Branched-From: 4226ddf99103e493d7afb23a4c7902ee496108b6-refs/heads/master@{#599034}
[modify] https://crrev.com/95e44621f06a1c3ffab2e3f860042da2fb254e48/cc/scheduler/scheduler_state_machine.cc
[modify] https://crrev.com/95e44621f06a1c3ffab2e3f860042da2fb254e48/cc/scheduler/scheduler_unittest.cc

The following revision refers to this bug: 
https://chromium.googlesource.com/chromium/src.git/+/95e44621f06a1c3ffab2e3f860042da2fb254e48

Commit: 95e44621f06a1c3ffab2e3f860042da2fb254e48
Author: boliu@chromium.org
Commiter: boliu@chromium.org
Date: 2018-11-20 04:39:44 +0000 UTC

[Merge M71] cc: Invalidate for needs_prepare_tiles

Problem is when CanDraw is false but needs_prepare_tiles is true
Scheduler will keep requesting begin frames but do nothing in a
begin frame since invalidate is blocked on CanDraw.

We already have an efficient shortcut that does not involve actually
producing a frame if invalidate is only requested for
needs_prepare_tiles. So the simple fix is simply allow invalidates for
this case.

(cherry picked from commit 43a7801fe8d646b7099d5e72a151c8eab5a3bffc)

Bug:  900557 
Change-Id: Ib3c3e213e2e2461d7aa1b8dd8d2e91ab64042feb
Reviewed-on: https://chromium-review.googlesource.com/c/1334550
Commit-Queue: Bo <boliu@chromium.org>
Reviewed-by: Sunny Sachanandani <sunnyps@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#607876}
Reviewed-on: https://chromium-review.googlesource.com/c/1343663
Reviewed-by: Bo <boliu@chromium.org>
Cr-Commit-Position: refs/branch-heads/3578@{#769}
Cr-Branched-From: 4226ddf99103e493d7afb23a4c7902ee496108b6-refs/heads/master@{#599034}