Move the ViewHostMsg_ClearPredictorCache out of content into chrome as this functionality is chro...

Move the V8 benchmarking_extension.cc/.h files out of webkit/extensions into chrome as the functionality
exposed by this extension is chrome specific. The IPCs used by this extension have all been moved to chrome
to a new header file benchmarking_messages.h in chrome\common.

Added a new message filter object in chrome\browser which filters the benchmarking IPCs.

Most of the changes in this CL are centered around ensuring that the benchmarking stuff gets initialized
in the same order as before.

Continuing changes to get rid of the pattern of IPC messags spanning across content and chrome.

BUG=87335
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=101214

[jam, 2011-09-13 21:40:36]

it seems that benchmarking_extension.cc is really chrome specific. I don't quite
understand why you say you gave up trying to move it? if it needs to call
(generic) functions in content, that seems fine. but it seems that now we have
chrome specific code in content and below (i.e. benchmarking_extension.cc) and
then have to add functions going both way that are chrome specific to support
this

[jam, 2011-09-13 21:47:51]

I just looked at the ViewHostMsg_ClearCache, ViewHostMsg_ClearHostResolverCache
messages, and the other IPCs sent by the v8 extension through webkit_glue.
they're all only used by taht function, and that implementation can move to
ChromeRenderMessageFilter (or, better yet, a separate message filter). seems
like a good cleanup to move code that doesn't need to move in content out of it,
and will also reduce webkit_glue functions.

On 2011/09/13 21:40:36, John Abd-El-Malek wrote:
> it seems that benchmarking_extension.cc is really chrome specific. I don't
quite
> understand why you say you gave up trying to move it? if it needs to call
> (generic) functions in content, that seems fine. but it seems that now we have
> chrome specific code in content and below (i.e. benchmarking_extension.cc) and
> then have to add functions going both way that are chrome specific to support
> this

[ananta, 2011-09-14 03:10:27]

On 2011/09/13 21:47:51, John Abd-El-Malek wrote:
> I just looked at the ViewHostMsg_ClearCache,
ViewHostMsg_ClearHostResolverCache
> messages, and the other IPCs sent by the v8 extension through webkit_glue.
> they're all only used by taht function, and that implementation can move to
> ChromeRenderMessageFilter (or, better yet, a separate message filter). seems
> like a good cleanup to move code that doesn't need to move in content out of
it,
> and will also reduce webkit_glue functions.
> 
> On 2011/09/13 21:40:36, John Abd-El-Malek wrote:
> > it seems that benchmarking_extension.cc is really chrome specific. I don't
> quite
> > understand why you say you gave up trying to move it? if it needs to call
> > (generic) functions in content, that seems fine. but it seems that now we
have
> > chrome specific code in content and below (i.e. benchmarking_extension.cc)
and
> > then have to add functions going both way that are chrome specific to
support
> > this

Done. I moved the benchmarking messages out to a new header file
benchmarking_messages.h. Had to make some changes in the content chrome
initialization code path to ensure that the benchmarking code initializes in
similar order as before.

PTAL

[jam, 2011-09-14 20:45:40]

the changes look like a great cleanup.

i don't understand why the InitializeStatsTable code has to move around. when we
initialize the field trial can move around slightly, i.e. you can move it to
ChromeBrowserParts::PostEarlyInitialization. then you wouldn't need the
content_main/content_main_delegate/chrome_main.cc changes

http://codereview.chromium.org/7885013/diff/4064/chrome/browser/chrome_benchm...
File chrome/browser/chrome_benchmarking_message_filter.h (right):

http://codereview.chromium.org/7885013/diff/4064/chrome/browser/chrome_benchm...
chrome/browser/chrome_benchmarking_message_filter.h:9: #include
"chrome/browser/profiles/profile.h"
nit: can you just forward declare profile

http://codereview.chromium.org/7885013/diff/4064/chrome/browser/chrome_benchm...
chrome/browser/chrome_benchmarking_message_filter.h:22: Profile* profile,
nit: tabbing

[ananta, 2011-09-14 23:40:54]

Reverted the changes around InitializeStatsTable as per our discussion. Added a
DCHECK which validates that EnableBenchmarking and EnableStatsTable need to be
specified together.

http://codereview.chromium.org/7885013/diff/4064/chrome/browser/chrome_benchm...
File chrome/browser/chrome_benchmarking_message_filter.h (right):

http://codereview.chromium.org/7885013/diff/4064/chrome/browser/chrome_benchm...
chrome/browser/chrome_benchmarking_message_filter.h:9: #include
"chrome/browser/profiles/profile.h"
On 2011/09/14 20:45:40, John Abd-El-Malek wrote:
> nit: can you just forward declare profile

Done.

http://codereview.chromium.org/7885013/diff/4064/chrome/browser/chrome_benchm...
chrome/browser/chrome_benchmarking_message_filter.h:22: Profile* profile,
On 2011/09/14 20:45:40, John Abd-El-Malek wrote:
> nit: tabbing

Done.

[jam, 2011-09-14 23:47:35]

lgtm

http://codereview.chromium.org/7885013/diff/5032/chrome/browser/chrome_conten...
File chrome/browser/chrome_content_browser_client.cc (right):

http://codereview.chromium.org/7885013/diff/5032/chrome/browser/chrome_conten...
chrome/browser/chrome_content_browser_client.cc:376: if
(command_line->HasSwitch(switches::kEnableBenchmarking)) {
nit: this file doesn't use brackets for one line statements