Exit VR Shell on pause (And fix a bunch of minor bugs).

chrome/browser/ui/webui/vr_shell/vr_shell_ui_message_handler.cc

Index: chrome/browser/ui/webui/vr_shell/vr_shell_ui_message_handler.cc
diff --git a/chrome/browser/ui/webui/vr_shell/vr_shell_ui_message_handler.cc b/chrome/browser/ui/webui/vr_shell/vr_shell_ui_message_handler.cc
index 0f8a6325821e161b084a3cb9df87059be753d00a..610e52f2b4f4125730c0ee7246934a61f36a374a 100644
--- a/chrome/browser/ui/webui/vr_shell/vr_shell_ui_message_handler.cc
+++ b/chrome/browser/ui/webui/vr_shell/vr_shell_ui_message_handler.cc
@@ -42,7 +42,8 @@ void VrShellUIMessageHandler::HandleDomLoaded(const base::ListValue* args) {
 }
 
 void VrShellUIMessageHandler::OnJavascriptAllowed() {
-  CHECK(vr_shell_);
+  if (!vr_shell_)

[cjgrant, 2016/10/27 14:33:17]

Why proceed here with no UI?

[mthiesse, 2016/10/27 14:37:03]

This can only happen if vr shell either died, or doesn't exist. If vr shell
died, this webui page will very soon be dead too (there's a destruction race).
If vr shell doesn't exist (the user loaded chrome://vr-shell-ui manually), we
should just silently do nothing.

I ran into this when exiting VR shortly after entering it; we definitely should
crash in that case.

[bshe, 2016/10/27 14:37:11]

I have logged a bug about this here:
https://bugs.chromium.org/p/chromium/issues/detail?id=659707

The root issue is that we didn't try to destroy ui_content before vr_shell is
destroyed. We should fix that I think. And then I think we dont need vr_shell a
weak ptr here.

[mthiesse, 2016/10/27 14:45:55]

Added comment and added bug# to CL.

+    return;
   vr_shell_->GetUiInterface()->SetUiCommandHandler(this);
   vr_shell_->OnDomContentsLoaded();
 }