Move responsiblity of tab suspending from renderer to TabManager

Move responsiblity of tab suspending from renderer to TabManager

Now tab-suspending is determined and done in renderer side and browser side
doesn't know when this happens. As we are going to have a centralized memory
managing system, it would make sense if we had tab-suspending system in browser
side.

This CL moves this tab-suspending system from renderer side to TabManager in
browser side.

BUG=469049, 486135, 551995

[hajimehoshi, 2016-04-19 06:54:55]

Description was changed from

==========
Move responsiblity of suspending from renderer to TabManager

WIP: Do not commit this!

BUG=
==========

to

==========
Move responsiblity of tab suspending from renderer to TabManager

WIP: Do not commit this!

BUG=
==========

[hajimehoshi, 2016-04-19 07:01:06]

Description was changed from

==========
Move responsiblity of tab suspending from renderer to TabManager

WIP: Do not commit this!

BUG=
==========

to

==========
Move responsiblity of tab suspending from renderer to TabManager

WIP: Do not commit this!

BUG=469049, 486135
==========

[hajimehoshi, 2016-04-19 11:48:35]

Description was changed from

==========
Move responsiblity of tab suspending from renderer to TabManager

WIP: Do not commit this!

BUG=469049, 486135
==========

to

==========
Move responsiblity of tab suspending from renderer to TabManager

*---------*---------*---------*---------*---------*---------*---------*---------
Now tab-suspending is determined and done in renderer side and browser side
doesn't know when this happens. As we are going to have a centralized memory
managing system, it would make sense if we had tab-suspending system in browser
side.

This CL moves this tab-suspending system from renderer side to TabManager in
browser side.

BUG=469049, 486135
==========

[hajimehoshi, 2016-04-19 11:48:35]

hajimehoshi@chromium.org changed reviewers:
+ chrisha@chromium.org, georgesak@chromium.org, haraken@chromium.org

[hajimehoshi, 2016-04-19 11:48:47]

Description was changed from

==========
Move responsiblity of tab suspending from renderer to TabManager

*---------*---------*---------*---------*---------*---------*---------*---------
Now tab-suspending is determined and done in renderer side and browser side
doesn't know when this happens. As we are going to have a centralized memory
managing system, it would make sense if we had tab-suspending system in browser
side.

This CL moves this tab-suspending system from renderer side to TabManager in
browser side.

BUG=469049, 486135
==========

to

==========
Move responsiblity of tab suspending from renderer to TabManager

Now tab-suspending is determined and done in renderer side and browser side
doesn't know when this happens. As we are going to have a centralized memory
managing system, it would make sense if we had tab-suspending system in browser
side.

This CL moves this tab-suspending system from renderer side to TabManager in
browser side.

BUG=469049, 486135
==========

[hajimehoshi, 2016-04-19 11:55:35]

I think implementation is almost done and I need to check if this actually works
on actual Android machine. I still need to add comments and browser tests.

I implemented sending suspending-tabs-IPC in TabManager::UpdateTimerCallback
instead of having timers so that implementation is much simpler, but suspending
IPC is called much more often than necessary. Redundant suspending-tabs-IPCs are
ignored in the renderer side BTW. Is this a good trade off?

[Georges Khalil, 2016-04-19 13:55:19]

Can you add bug 551995 as well?

https://codereview.chromium.org/1895773002/diff/60001/chrome/browser/chrome_b...
File chrome/browser/chrome_browser_main.cc (right):

https://codereview.chromium.org/1895773002/diff/60001/chrome/browser/chrome_b...
chrome/browser/chrome_browser_main.cc:1254:
g_browser_process->GetTabManager()->Start();
This will also enable automatic discarding as it will subscribe to memory
pressure events. Which we don't want on Android.

Wrap the memory pressure stuff in Start with !defined(OS_ANDROID), that should
do it for now. Eventually, Start would need to be factored out into smaller
functions.

https://codereview.chromium.org/1895773002/diff/60001/chrome/browser/memory/t...
File chrome/browser/memory/tab_manager.cc (right):

https://codereview.chromium.org/1895773002/diff/60001/chrome/browser/memory/t...
chrome/browser/memory/tab_manager.cc:535: #if defined(OS_ANDROID)
Move this to an android delegate, similar to chromeos.

https://codereview.chromium.org/1895773002/diff/60001/chrome/browser/memory/t...
File chrome/browser/memory/tab_manager.h (right):

https://codereview.chromium.org/1895773002/diff/60001/chrome/browser/memory/t...
chrome/browser/memory/tab_manager.h:57: // On Chrome OS (via the delegate), the
kernel (via /proc/<pid>/oom_score_adj)
Add a mention here about what we do on Android.

[hajimehoshi, 2016-04-20 07:09:55]

Description was changed from

==========
Move responsiblity of tab suspending from renderer to TabManager

Now tab-suspending is determined and done in renderer side and browser side
doesn't know when this happens. As we are going to have a centralized memory
managing system, it would make sense if we had tab-suspending system in browser
side.

This CL moves this tab-suspending system from renderer side to TabManager in
browser side.

BUG=469049, 486135
==========

to

==========
Move responsiblity of tab suspending from renderer to TabManager

Now tab-suspending is determined and done in renderer side and browser side
doesn't know when this happens. As we are going to have a centralized memory
managing system, it would make sense if we had tab-suspending system in browser
side.

This CL moves this tab-suspending system from renderer side to TabManager in
browser side.

BUG=469049, 486135, 551995
==========

[hajimehoshi, 2016-04-20 08:56:33]

I'm sorry but I found that TabManager and related classes (e.g.
TabStripModelObserver which is the parent class of TabManager) don't exist on
Android, and it would be hard work to make them work on Android correctly since
the notion of tab-strips seems much different between mobiles and desktops. As
this is just refactoring for implementing purge + suspend, I'll revisit this
after implementing purge + suspend on desktops.

[Georges Khalil, 2016-04-20 15:28:35]

On 2016/04/20 08:56:33, hajimehoshi wrote:
> I'm sorry but I found that TabManager and related classes (e.g.
> TabStripModelObserver which is the parent class of TabManager) don't exist on
> Android, and it would be hard work to make them work on Android correctly
since
> the notion of tab-strips seems much different between mobiles and desktops. As
> this is just refactoring for implementing purge + suspend, I'll revisit this
> after implementing purge + suspend on desktops.

I figured it would not be a simple thing to enable this on Android.

We need to factor out desktop specific code to a helper class to purify
TabManager.

I'll discuss with chrisha@ on what the best course of action would be.