Add a RWH_TabSwitchPaintDuration histogram.

Add a RWH_TabSwitchPaintDuration histogram.

The RWH_TabSwitchPaintDuration histogram calculates
the amount of time it took after a particular view was
selected for it to be fully painted.

Previous versions of this patch included modifications
to the tab switching test.  Those are blocked on setting
up a new builder for accurate measurement and are now
removed from this CL.

BUG=4104
TEST=none

Commit: http://src.chromium.org/viewvc/chrome?view=rev&revision=31632
Revert: http://src.chromium.org/viewvc/chrome?view=rev&revision=31645
Commit: http://src.chromium.org/viewvc/chrome?view=rev&revision=31679
Revert: http://src.chromium.org/viewvc/chrome?view=rev&revision=31683

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=31747

[chase, 2009-11-10 02:05:10]

John, what do you think of this approach?  Tests locally are stalled while the
more general chrome.dll loading issue is looked into.  A previous test with the
same patch in an older checkout worked as expected, though.

[jam, 2009-11-10 20:06:55]

Why not do this in TabContents, that way don't need to repeat the code for each
platform?

[jam, 2009-11-10 20:07:12]

Why not do this in TabContents, that way don't need to repeat the code for  
each
platform?

http://codereview.chromium.org/378038

[jam, 2009-11-10 20:07:14]

Why not do this in TabContents, that way don't need to repeat the code for  
each
platform?

http://codereview.chromium.org/378038

[jam, 2009-11-10 20:29:16]

argh, never mind, you would need extra hooks anyways so this is fine since it
mirrors the whiteout duration stuff.  lgtm

[jam, 2009-11-10 20:29:35]

argh, never mind, you would need extra hooks anyways so this is fine since  
it
mirrors the whiteout duration stuff.  lgtm

http://codereview.chromium.org/378038

[chase, 2009-11-10 20:30:30]

On 2009/11/10 20:29:35, John Abd-El-Malek wrote:
> argh, never mind, you would need extra hooks anyways so this is fine since  
> it
> mirrors the whiteout duration stuff.  lgtm

np.  i will read up on histograms and rename histogram.  something like
'TabSwitchPaintDuration'.

[chase, 2009-11-10 20:31:02]

On 2009/11/10 20:29:35, John Abd-El-Malek wrote:
> argh, never mind, you would need extra hooks anyways so this is fine since
> it
> mirrors the whiteout duration stuff.  lgtm

np.  i will read up on histograms and rename histogram.  something like
'TabSwitchPaintDuration'.

http://codereview.chromium.org/378038

[chase, 2009-11-10 23:03:56]

Histogram name updated, added to wiki.  Also, the tab switching test now asserts
that it can locate the expected histogram.

[chase, 2009-11-10 23:04:13]

Histogram name updated, added to wiki.  Also, the tab switching test now  
asserts
that it can locate the expected histogram.

http://codereview.chromium.org/378038

[jam, 2009-11-10 23:11:22]

I think the RWHH_ for whiteout duration got an extra H by mistake, we shouldn't
do that as well here.

[jam, 2009-11-10 23:11:42]

I think the RWHH_ for whiteout duration got an extra H by mistake, we  
shouldn't
do that as well here.

http://codereview.chromium.org/378038

[chase, 2009-11-10 23:14:12]

On 2009/11/10 23:11:42, John Abd-El-Malek wrote:
> I think the RWHH_ for whiteout duration got an extra H by mistake, we  
> shouldn't
> do that as well here.

I found after talking with you earlier the extra H stands for HWND, see:

    // Now open the corresponding log file and collect average and std dev from
    // the histogram stats generated for
    // RenderWidgetHostHWND_TabSwitchPaintDuration

[chase, 2009-11-10 23:14:30]

On 2009/11/10 23:11:42, John Abd-El-Malek wrote:
> I think the RWHH_ for whiteout duration got an extra H by mistake, we
> shouldn't
> do that as well here.

I found after talking with you earlier the extra H stands for HWND, see:

     // Now open the corresponding log file and collect average and std dev  
from
     // the histogram stats generated for
     // RenderWidgetHostHWND_TabSwitchPaintDuration

http://codereview.chromium.org/378038

[chase, 2009-11-10 23:19:18]

On 2009/11/10 23:14:12, chase wrote:
> On 2009/11/10 23:11:42, John Abd-El-Malek wrote:
> > I think the RWHH_ for whiteout duration got an extra H by mistake, we  
> > shouldn't
> > do that as well here.
> 
> I found after talking with you earlier the extra H stands for HWND, see:
> 
>     // Now open the corresponding log file and collect average and std dev
from
>     // the histogram stats generated for
>     // RenderWidgetHostHWND_TabSwitchPaintDuration

That said if you prefer RWH, happy to make that change.

[chase, 2009-11-10 23:19:35]

On 2009/11/10 23:14:12, chase wrote:
> On 2009/11/10 23:11:42, John Abd-El-Malek wrote:
> > I think the RWHH_ for whiteout duration got an extra H by mistake, we
> > shouldn't
> > do that as well here.

> I found after talking with you earlier the extra H stands for HWND, see:

>      // Now open the corresponding log file and collect average and std dev
from
>      // the histogram stats generated for
>      // RenderWidgetHostHWND_TabSwitchPaintDuration

That said if you prefer RWH, happy to make that change.

http://codereview.chromium.org/378038

[chase, 2009-11-10 23:55:11]

Renamed to RWH_TabSwitchPaintDuration, comments and CL updated.

[chase, 2009-11-10 23:55:32]

Renamed to RWH_TabSwitchPaintDuration, comments and CL updated.

http://codereview.chromium.org/378038

[chase, 2009-11-11 01:07:39]

Got LGTM from John over IM.  (Why are most comments repeated twice in this
review?)

[chase, 2009-11-11 01:07:57]

Got LGTM from John over IM.  (Why are most comments repeated twice in this
review?)

http://codereview.chromium.org/378038