Supports High DPI mode....

Supports High DPI mode.
When launching an application on the text size larger than 120 DPI, Windows runs the application on the DPI-virtualization mode to hide the text size from it. Unfortunately, this virtualization mode causes some problems when using a custom frame, such as we cannot click system buttons. To fix this issue, this change disables the DPI-virtualization mode on Vista or later.

BUG=1715, 36939
TEST=Run chrome.exe on the screen resolution higher than 120 DPI.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=52446

[Hironori Bono, 2010-07-07 08:02:04]

ping?

[Ben Goodger (Google), 2010-07-07 21:06:54]

http://codereview.chromium.org/2867031/diff/6001/7001
File app/win_util.cc (right):

http://codereview.chromium.org/2867031/diff/6001/7001#newcode588
app/win_util.cc:588: if (win_util::GetWinVersion() < win_util::WINVERSION_VISTA)
What if you're using the custom frame on Vista though?

This can happen in the following circumstances:

- a theme is active
- the user is in Vista Basic or Windows Classic display mode.

[Hironori Bono, 2010-07-09 07:58:42]

http://codereview.chromium.org/2867031/diff/6001/7001
File app/win_util.cc (right):

http://codereview.chromium.org/2867031/diff/6001/7001#newcode588
app/win_util.cc:588: if (win_util::GetWinVersion() < win_util::WINVERSION_VISTA)
On 2010/07/07 21:06:54, Ben Goodger wrote:
> What if you're using the custom frame on Vista though?
> This can happen in the following circumstances:
> - a theme is active
> - the user is in Vista Basic or Windows Classic display mode.

Even though it seems to be OK to call this function on the Classic mode, I have
changed this line with ShouldUseVistaFrame() because these problems happen only
on using the native frame.

[Ben Goodger (Google), 2010-07-09 17:12:14]

Well maybe you should leave it as you had it before... since you say calling
that function is harmless with the non-native frame, I think it's best if we
call it in both modes... since calling ShouldUseVistaFrame implies a bug to
me... since this code only executes once (at browser startup, for the entire
process), what if the display mode changes? That would imply any assumption as
to the presence or absence of the native frame would be invalid. In this case is
harmless but now I understand it better I think your original check was better.
LGTM for that.