Ensure that NativeWidgetWin::OnPaint calls ValidateRect; workaround for crbug.com/93530.

Ensure that NativeWidgetWin::OnPaint calls ValidateRect.
This is a workaround for crbug.com/93530; caused by crrev.com/84823.
Windows creates a WM_PAINT storm without validation; even on empty regions.

BUG=93530
TEST=Issue does not repro; no regressions.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=110239

[msw, 2011-11-15 19:36:38]

Sky, MAD: PTAL, thanks!

[MAD, 2011-11-15 19:39:51]

http://codereview.chromium.org/8568016/diff/1/views/widget/native_widget_win.cc
File views/widget/native_widget_win.cc (right):

http://codereview.chromium.org/8568016/diff/1/views/widget/native_widget_win....
views/widget/native_widget_win.cc:1778: // TODO(msw): Find a better solution for
crbug.com/93530; nix this workaround.
Shouldn't we only do this in an else statement?

[msw, 2011-11-15 20:12:19]

Comment addressed; please take another look.
I see our e-mail thread discussing the BeginPaint/EndPaint workaround, is that a
better temporary stop-gap?

http://codereview.chromium.org/8568016/diff/1/views/widget/native_widget_win.cc
File views/widget/native_widget_win.cc (right):

http://codereview.chromium.org/8568016/diff/1/views/widget/native_widget_win....
views/widget/native_widget_win.cc:1778: // TODO(msw): Find a better solution for
crbug.com/93530; nix this workaround.
On 2011/11/15 19:39:52, MAD wrote:
> Shouldn't we only do this in an else statement?

Done.

[sky, 2011-11-15 20:29:29]

LGTM you should merge this as far back and you can.

[MAD, 2011-11-15 21:23:27]

On 2011/11/15 20:29:29, sky wrote:
> LGTM you should merge this as far back and you can.

I was talking to Siggi about it and he was wondering if this would cover all
cases, since this is a case that we were not expecting to happen... Maybe in
that case, ValidateRect won't always do what we expect, whereas Begin/EndPaint
should... And we could add a DCHECK that the rect from the paintstruct is
actually empty...

Did you confirm that you could reliably repro without the fix?
So you could confirm that the fix always work?

Thanks!

[msw, 2011-11-15 21:54:57]

> Did you confirm that you could reliably repro without the fix?
Yes, I have can reliably repro 93530 without the fix.
> So you could confirm that the fix always work?
The fix has worked on all four of my repro attempts; Spy++ reports just a single
WM_PAINT posted to the app window.

I'm reading up online about the differences between ValidateRect and
BeginPaint/EndPaint. So far, MSDN and other sites indicate that ValidateRect is
correct for validating a window without doing any drawing and it's slightly more
efficient (doesn't need to get the DC or allocate the PAINTSTRUCT):

http://books.google.com/books?id=Lx8xmAwXrCgC&lpg=PA130&ots=PFOFMS9RqU&dq=Val...
To validate the area of a window that needs repainting and tell Windows that you
have restored the window, you could call BeginPaint()—EndPaint() after the call
to GetDC()—ReleaseDC(), but this would be inefficient. Instead, use the call
specifically designed for this, called ValidateRect().

http://www.changhai.org/articles/technology/programming/opengl2.php
once a WM_PAINT message is retrived from the message queue, Windows will send
another WM_PAINT message to the queue (and keep doing so) until the program
validates the invalid region (which basically tells Windows that the invalid
area is updated). To validate an area, one usually uses the ValidateRect
function.

[MAD, 2011-11-15 22:17:56]

lgtm

Excellent... Thanks for... hum... Validating... this... ;)

LGTM!

[msw, 2011-11-16 00:10:21]

Between the win trybot re-run and my local test passes, I think this is good to
land.

[commit-bot: I haz the power, 2011-11-16 00:13:30]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/msw@chromium.org/8568016/3002

[commit-bot: I haz the power, 2011-11-16 01:58:16]

Change committed as 110239