Change the way lazy background pages shutdown.

Change the way lazy background pages shutdown.

The browser process now keeps a generic count of activity from the lazy background page. This will be used for all types of activity, such as outstanding events, resource requests, and API calls.

BUG=81752
TEST=no

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=123821

[Yoyo Zhou, 2012-02-24 20:25:12]

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
File chrome/browser/extensions/extension_host.cc (right):

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
chrome/browser/extensions/extension_host.cc:407: // TODO(mpcomplete): is this
check really necessary?
Well, the comment in extension_host.h on extension_host_type_ is certainly
wrong...
It's excluding VIEW_TYPE_NOTIFICATION perhaps?

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
File chrome/browser/extensions/extension_process_manager.h (right):

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
chrome/browser/extensions/extension_process_manager.h:96: int
GetLazyKeepaliveCount(const Extension* extension);
These names are a little unwieldy. Is the "lazy" necessary? I would spell it
KeepAlive.

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
chrome/browser/extensions/extension_process_manager.h:97: int
AdjustLazyKeepaliveCount(const Extension* extension, int delta);
Seems like the conventions are usually to have two functions for +1/-1 (such as
AddRef/Release). Would that make sense here?

http://codereview.chromium.org/9447042/diff/13001/chrome/browser/extensions/e...
File chrome/browser/extensions/extension_event_router.cc (right):

http://codereview.chromium.org/9447042/diff/13001/chrome/browser/extensions/e...
chrome/browser/extensions/extension_event_router.cc:423:
pm->OnLazyBackgroundPageIdle(extension_id);
Do you still need this?

http://codereview.chromium.org/9447042/diff/13001/chrome/browser/extensions/e...
File chrome/browser/extensions/extension_process_manager.cc (right):

http://codereview.chromium.org/9447042/diff/13001/chrome/browser/extensions/e...
chrome/browser/extensions/extension_process_manager.cc:41: // TODO(mpcomplete):
Should we store this on ExtensionHost instead?
I don't have a strong opinion on this.

[Matt Perry, 2012-02-24 22:51:27]

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
File chrome/browser/extensions/extension_host.cc (right):

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
chrome/browser/extensions/extension_host.cc:407: // TODO(mpcomplete): is this
check really necessary?
On 2012/02/24 20:25:12, Yoyo Zhou wrote:
> Well, the comment in extension_host.h on extension_host_type_ is certainly
> wrong...

Heh. I fixed the comment.

> It's excluding VIEW_TYPE_NOTIFICATION perhaps?

Maybe. But the restriction here seems really silly. All the listeners to the
notification check that the host matches what they expect anyway. Still, I'm
afraid to remove it with this CL, so I'll leave it for now.

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
File chrome/browser/extensions/extension_process_manager.h (right):

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
chrome/browser/extensions/extension_process_manager.h:96: int
GetLazyKeepaliveCount(const Extension* extension);
On 2012/02/24 20:25:12, Yoyo Zhou wrote:
> These names are a little unwieldy.

True... Any suggestions?

> Is the "lazy" necessary?

It only applies to lazy background pages. I suppose it's not necessary, but I
think it adds clarity.

> I would spell it
> KeepAlive.

I was going for http://en.wikipedia.org/wiki/Keepalive

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
chrome/browser/extensions/extension_process_manager.h:97: int
AdjustLazyKeepaliveCount(const Extension* extension, int delta);
On 2012/02/24 20:25:12, Yoyo Zhou wrote:
> Seems like the conventions are usually to have two functions for +1/-1 (such
as
> AddRef/Release). Would that make sense here?

Yeah, I went back and forth on this.. I'll go with 2 functions.

http://codereview.chromium.org/9447042/diff/13001/chrome/browser/extensions/e...
File chrome/browser/extensions/extension_event_router.cc (right):

http://codereview.chromium.org/9447042/diff/13001/chrome/browser/extensions/e...
chrome/browser/extensions/extension_event_router.cc:423:
pm->OnLazyBackgroundPageIdle(extension_id);
On 2012/02/24 20:25:12, Yoyo Zhou wrote:
> Do you still need this?

Yes. It's possible for us to spin up the process, but for the extension to opt
not to register for events.

I've restructured the code to make it flow a bit more naturally though. I now
wrap the event dispatch in a keepalive addref/release.

[Yoyo Zhou, 2012-02-24 23:15:00]

LGTM

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
File chrome/browser/extensions/extension_process_manager.h (right):

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
chrome/browser/extensions/extension_process_manager.h:96: int
GetLazyKeepaliveCount(const Extension* extension);
On 2012/02/24 22:51:27, Matt Perry wrote:
> On 2012/02/24 20:25:12, Yoyo Zhou wrote:
> > These names are a little unwieldy.
> 
> True... Any suggestions?
> 
> > Is the "lazy" necessary?
> 
> It only applies to lazy background pages. I suppose it's not necessary, but I
> think it adds clarity.

Yeah. But "lazy" (referring to not being created until necessary, which just
happened to be the first part to be implemented) doesn't capture the whole
picture either. I guess we can think of better names for this later.

> > I would spell it
> > KeepAlive.
> 
> I was going for http://en.wikipedia.org/wiki/Keepalive

Sure. But it does seem more like reference counting to me. So, perhaps something
like "IncrementLazyReferenceCount". But we don't want to confuse it with memory
refcounting...

[Matt Perry, 2012-02-24 23:20:12]

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
File chrome/browser/extensions/extension_process_manager.h (right):

http://codereview.chromium.org/9447042/diff/7001/chrome/browser/extensions/ex...
chrome/browser/extensions/extension_process_manager.h:96: int
GetLazyKeepaliveCount(const Extension* extension);
On 2012/02/24 23:15:00, Yoyo Zhou wrote:
> On 2012/02/24 22:51:27, Matt Perry wrote:
> > On 2012/02/24 20:25:12, Yoyo Zhou wrote:
> > > These names are a little unwieldy.
> > 
> > True... Any suggestions?
> > 
> > > Is the "lazy" necessary?
> > 
> > It only applies to lazy background pages. I suppose it's not necessary, but
I
> > think it adds clarity.
> 
> Yeah. But "lazy" (referring to not being created until necessary, which just
> happened to be the first part to be implemented) doesn't capture the whole
> picture either. I guess we can think of better names for this later.

"On-demand" is my favorite succinct name for the feature. But the code already
has lazy everywhere, so I just went with that. (And lazy's nice and short!)

> 
> > > I would spell it
> > > KeepAlive.
> > 
> > I was going for http://en.wikipedia.org/wiki/Keepalive
> 
> Sure. But it does seem more like reference counting to me. So, perhaps
something
> like "IncrementLazyReferenceCount". But we don't want to confuse it with
memory
> refcounting...

Yeah, that doesn't convey much about what we're counting, IMO.