Initial commit of the blimp/ folder and target

blimp/client/compositor/blimp_compositor.cc

Index: blimp/client/compositor/blimp_compositor.cc
diff --git a/blimp/client/compositor/blimp_compositor.cc b/blimp/client/compositor/blimp_compositor.cc
new file mode 100644
index 0000000000000000000000000000000000000000..edb32bcaa7a254f4198177fac563aff591bf7c36
--- /dev/null
+++ b/blimp/client/compositor/blimp_compositor.cc
@@ -0,0 +1,154 @@
+// Copyright 2015 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "blimp/client/compositor/blimp_compositor.h"
+
+#include "base/bind_helpers.h"
+#include "base/command_line.h"
+#include "base/lazy_instance.h"
+#include "base/single_thread_task_runner.h"
+#include "base/thread_task_runner_handle.h"
+#include "base/threading/thread.h"
+#include "base/threading/thread_local.h"
+#include "base/threading/thread_restrictions.h"
+#include "blimp/client/compositor/blimp_context_provider.h"
+#include "blimp/client/compositor/blimp_output_surface.h"
+#include "blimp/client/compositor/blimp_task_graph_runner.h"
+#include "blimp/client/compositor/test/dummy_layer_driver.h"
+#include "blimp/common/compositor/blimp_layer_tree_settings.h"
+#include "cc/layers/layer.h"
+#include "cc/output/output_surface.h"
+#include "cc/trees/layer_tree_host.h"
+#include "ui/gl/gl_surface.h"
+
+namespace {
+
+base::LazyInstance<blimp::BlimpTaskGraphRunner> g_task_graph_runner =
+    LAZY_INSTANCE_INITIALIZER;
+
+}  // namespace
+
+// Testing Code:
+namespace {

[Wez, 2015/08/27 02:01:51]

Why does this need a separate anonymous namespace?

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

I wanted it separate so I could delete this chunk once we actually draw
something in the compositor.  This was just for testing purposes so something
actually displays (makes sure the scheduler/threads/pipelines are working
properly).

I'll move it into the other namespace though :).

[Wez, 2015/09/03 00:49:27]

OK, suggest commenting like:

TODO(dtrainor): Replace this with a real layer driver impl (see crbug.com/...).

:)

[David Trainor- moved to gerrit, 2015/09/03 06:33:21]

Done.

+base::LazyInstance<blimp::DummyLayerDriver> g_dummy_layer_driver =
+    LAZY_INSTANCE_INITIALIZER;
+}
+
+namespace blimp {
+
+BlimpCompositor::BlimpCompositor(float device_scale_factor)
+    : device_scale_factor_(device_scale_factor), weak_factory_(this) {}
+
+BlimpCompositor::~BlimpCompositor() {
+  // Make sure things get freed in the right order.

[Wez, 2015/08/27 02:01:51]

nit: This comment explains why we're doing it explicitly, but not the specific
constraints.

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

Done.

[Wez, 2015/09/03 00:49:27]

Acknowledged.

+  host_.reset();
+  settings_.reset();
+  if (compositor_thread_)
+    compositor_thread_->Stop();
+}
+
+void BlimpCompositor::SetVisible(bool visible) {
+  if (visible && !host_) {
+    if (!settings_) {
+      settings_.reset(new cc::LayerTreeSettings);
+      GenerateLayerTreeSettings(*settings_,
+                                *base::CommandLine::ForCurrentProcess());
+    }
+    // Create the LayerTreeHost

[Wez, 2015/08/27 02:01:51]

nit: Blank line before block comment?

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

Done.

+    cc::LayerTreeHost::InitParams params;
+    params.client = this;
+    params.shared_bitmap_manager = nullptr;
+    params.gpu_memory_buffer_manager = nullptr;

[Wez, 2015/08/27 02:01:51]

nit: These two = nullptrs are redundant.

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

Done.

+    params.task_graph_runner = g_task_graph_runner.Pointer();
+    params.main_task_runner = base::ThreadTaskRunnerHandle::Get();
+    params.settings = settings_.get();
+    // TODO(dtrainor): Swap this out with the remote client proxy when
+    // implemented.
+    host_ =
+        cc::LayerTreeHost::CreateThreaded(GetCompositorTaskRunner(), &params);
+
+    host_->SetVisible(true);
+    host_->SetLayerTreeHostClientReady();
+    host_->SetViewportSize(viewport_size_);
+    host_->SetDeviceScaleFactor(device_scale_factor_);
+
+    // Build the root Layer

[Wez, 2015/08/27 02:01:51]

nit: Punctuation - missing .

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

Done.

+    scoped_refptr<cc::Layer> root(cc::Layer::Create(cc::LayerSettings()));
+    host_->SetRootLayer(root);
+
+    // For testing, set the dummy Layer.
+    g_dummy_layer_driver.Pointer()->SetParentLayer(root);
+
+  } else if (!visible && host_) {
+    // Destroy the LayerTreeHost

[Wez, 2015/08/27 02:01:51]

nit: Punctuation

Also, this comment seems redundant - it'd be more helpful to explain to the
reader why that's all that we need to do to become invisible?

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

Done.

+    host_.reset();
+  }
+}
+
+void BlimpCompositor::SetSize(const gfx::Size& size) {
+  viewport_size_ = size;
+  if (host_)
+    host_->SetViewportSize(viewport_size_);
+}
+
+void BlimpCompositor::Layout() {}
+
+void BlimpCompositor::RequestNewOutputSurface() {
+  gfx::AcceleratedWidget widget = GetWindow();
+  DCHECK(widget);
+
+  scoped_refptr<BlimpContextProvider> context_provider =
+      BlimpContextProvider::Create(widget);
+
+  scoped_ptr<cc::OutputSurface> output_surface(
+      new BlimpOutputSurface(context_provider));
+
+  host_->SetOutputSurface(output_surface.Pass());

[Wez, 2015/08/27 02:01:51]

nit: Do you need the output_surface temporary?

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

Done.

+}
+
+void BlimpCompositor::DidInitializeOutputSurface() {}
+
+void BlimpCompositor::DidFailToInitializeOutputSurface() {}
+
+void BlimpCompositor::DidCommit() {}
+
+void BlimpCompositor::DidCompleteSwapBuffers() {}
+
+void BlimpCompositor::ScheduleComposite() {}
+
+void BlimpCompositor::ScheduleAnimation() {}
+
+void BlimpCompositor::DidPostSwapBuffers() {}
+
+void BlimpCompositor::DidAbortSwapBuffers() {}
+
+void BlimpCompositor::GenerateLayerTreeSettings(cc::LayerTreeSettings& settings,
+                                                const base::CommandLine& cmd) {
+  PopulateCommonLayerTreeSettings(settings, cmd);
+}
+
+scoped_refptr<base::SingleThreadTaskRunner>
+BlimpCompositor::GetCompositorTaskRunner() {
+  if (!compositor_thread_) {

[Wez, 2015/08/27 02:01:51]

nit: Suggest restructing this as an early-return if it's already set.

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

Done.

+    base::Thread::Options thread_options;
+#if defined(OS_ANDROID)
+    thread_options.priority = base::ThreadPriority::DISPLAY;
+#endif
+    compositor_thread_.reset(new base::Thread("Compositor"));
+    compositor_thread_->StartWithOptions(thread_options);
+
+    scoped_refptr<base::SingleThreadTaskRunner> task_runner =
+        compositor_thread_->task_runner();
+    task_runner->PostTask(
+        FROM_HERE,
+        base::Bind(base::IgnoreResult(&base::ThreadRestrictions::SetIOAllowed),
+                   false));

[Wez, 2015/08/27 02:01:51]

What's all this about? Surely you can just create the thread with a normal
messgae-loop and IO won't be allowed on it?

[David Trainor- moved to gerrit, 2015/08/28 01:23:46]

I pulled this piece from render_thread_impl.cc.  Looking at
thread_restrictions.cc, it looks like the lazy ThreadLocalBoolean is called
g_io_disallowed (which defaults to false, which means IO is allowed).  Yay
double negatives?  I think I have to do this to turn off IO on this thread.

[Wez, 2015/09/03 00:49:27]

Ha! We have a poster about avoiding negatives in conditionals up on the wall
nearby right now. :P

Anyway... Ewwww, gross, but seemingly necessary. Do we also want to disallow
waiting?  Looks like Chromoting does that on its network thread, but I don't see
Chrome doing that.

[David Trainor- moved to gerrit, 2015/09/03 06:33:21]

Yes!  I want a poster like that.  I had to ask Tommy if I was reading the logic
correctly.  I'm not sure about waiting.  I'll add a TODO and investigate when I
get back from vacation if that's okay!

+
+    return task_runner;
+  } else {
+    return compositor_thread_->task_runner();
+  }
+}
+
+}  // namespace blimp