Fix numerous alignment problems, both horizontal and vertical, in drawing the...

Fix numerous alignment problems, both horizontal and vertical, in drawing the browser chrome.  This doesn't fix problems with the NTP background image offsets (see bug 51853).

This also makes sure the OTR avatar is reversed in RTL mode, and clipped as usefully as possible when there isn't enough room for the full image.

BUG=44157
TEST=Frame, toolbar, bookmark bar, and find bar all look lined up with default and custom themes, restored and maximized, LTR and RTL, side tabs and normal tabs, and in popup and app windows

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=56737

[Peter Kasting, 2010-08-18 01:01:10]

* I can't locally test the ChromeOS changes.  If you could do a Linux Views
build and test some of this stuff for me, I'd appreciate it.  I probably broke
something.

* Here are some notes on particular files to help understand the review:

opaque_browser_frame_view.cc: Made OTR avatar be drawn directly instead of as a
child ImageView in order to have more control over positioning when there isn't
enough space (the avatar has an odd number of vertical pixels and looks better
when the display zone is biased toward the image bottom instead of the top, like
ImageView does).  Had to add ugly boolean overrides to all kinds of functions in
order to be able to calculate the tabstrip vertical offset "as it is in restored
mode" regardless of the current mode, so the frame image could be aligned
properly in maximized mode.  A number of changes/simplifications to painting due
to the browser toolbar/tabstrip bound-getting functions returning the "apparent"
bounds of those objects instead of the true bounds.  Fixed
toolbar-corner-and-client-edge-images-overlap bug in popup windows.

browser_frame_win.cc: Removed code that tried to position the side tabs tabstrip
"below the title bar" as it didn't actually get the height right (things jammed
up against the caption buttons).  The replacement calculations in
glass_browser_frame_view.cc do the right thing.

browser_view.cc: Broke Aero Peek (which is off by default) content-offsetting
code since I removed the function it (incorrectly) used to calculate the content
offsets.  It's not clear to me what the right fix is here, I suppose I could
plumb some sort of pathway to the frame to get the content area rect... this is
all a big mess.

glass_browser_frame_view.cc: Similar changes to the opaque frame w.r.t. plumbing
ugly bool overrides everywhere and biasing the OTR avatar clipping rect down 1
px.  Simplified toolbar background painting code for vertical tabs since now the
toolbar and tabstrip "apparent bounds" returned by the BrowserView start at the
same coords.

normal_browser_frame_view.cc: Removed an Inset() call because that's effectively
done by the changes to return the "apparent bounds" of the toolbar.

[sky, 2010-08-18 16:13:49]

Chromeos compile fails with:

chrome/browser/chromeos/frame/normal_browser_frame_view.cc: In member function
'virtual gfx::Rect
chromeos::NormalBrowserFrameView::GetBoundsForTabStrip(BaseTabStr\
ip*) const':
chrome/browser/chromeos/frame/normal_browser_frame_view.cc:94: error: no
matching function for call to
'chromeos::NormalBrowserFrameView::GetHorizontalTabStripVert\
icalOffset() const'
./chrome/browser/chromeos/frame/normal_browser_frame_view.h:37: note: candidates
are: virtual int
chromeos::NormalBrowserFrameView::GetHorizontalTabStripVerticalOf\
fset(bool) const

[Peter Kasting, 2010-08-18 20:24:09]

Fixed the compile error.

[sky, 2010-08-18 22:13:09]

Assuming you get chromeos building right, LGTM.

http://codereview.chromium.org/3137019/diff/11002/6017
File chrome/browser/views/frame/browser_view.cc (right):

http://codereview.chromium.org/3137019/diff/11002/6017#newcode630
chrome/browser/views/frame/browser_view.cc:630: gfx::Point
window_point(point.Add(gfx::Point(MirroredX(), y())));
nit: indenting is off.

[Peter Kasting, 2010-08-19 01:41:34]

New snap up that tries to fix alignment on ChromeOS.  Can you test it?

[Peter Kasting, 2010-08-19 03:18:23]

Argh, looks like maybe I broke the opaque frame compile in my rush to get out
the door with this.  I'll look when I come in.

[Peter Kasting, 2010-08-19 17:33:27]

OK, try now.

[Peter Kasting, 2010-08-19 19:28:21]

New snap fixes the remaining 1 px offset in the ChromeOS frame, and also fixes a
DCHECK failure in the opaque frame when maximizing an app window.

I'm going to land this when the tree opens.

[sky, 2010-08-19 19:32:30]

OK