Chrome: Gui.html test failure

[Douglas Hays]

Run dijit/tests/layout/robot/Gui.html in Chrome 8/WinXP.
The first test fails consistently:

GROUP "API" has 1 test to run
failed with hint: 
	center inside embeddedBC{"x":310,"y":266,"w":454,"h":0}{"x":310,"y":257,"w":454,"h":8}			
FAILED test: initialConditions 4 ms

[bill]

Thanks. Hmm, I'm seeing something similar, on my PC the "Info" tab slightly overlaps the ContentPane above it, and the draggable splitter is also too high (compared to the bottom of that ContentPane). So there are panes actually overlapping, and it's not because of lack of viewport size either.

[bill]

This is an architectural problem with BorderContainer, nothing particular to do with chrome.

At the beginning of _layoutChildren(), it queries the height of the top and bottom panes:

topHeight = changedRegion == "top" ? changedRegionSize : dojo._getMarginSize(this._top).h;
...

Then much later on, it changes the width of the top and bottom panes:

if(sidebarLayout){
	topStyle.left = bottomStyle.left = leftWidth + leftSplitterThickness + pe.l + "px";
	topStyle.right = bottomStyle.right = rightWidth + rightSplitterThickness + pe.r + "px";
}else{
	topStyle.left = bottomStyle.left = pe.l + "px";
	topStyle.right = bottomStyle.right = pe.r + "px";
}

In this test case (depending on viewport size and font/font size), the top pane of the "bottomSplit" BorderContainer has a bunch of text that all fits into one line. However, after the width of the pane is changed (reduced by 10px, to account for the padding of the BorderContainer), that text overflows to two lines, thus increasing the height of the top pane.

The order of operations is incorrect.

This would be fixed by #11030, or perhaps there's a more minimal change to get it working.

[bill]

(In [23340]) Convert BorderContainer to use dijit.layout.layoutChildren() rather than custom layout code. Refs #11030, fixes #12078 !strict.

[bill]

(In [23346]) Get ExpandoPane working again after [23276] (refs #11816) and [23340] (refs #12078), !strict.

Reintroduced separate ContentPane._layoutChildren() method since ExpandoPane was referencing it and other classes might be theoretically referencing it too.

[bill]

(In [23380]) Fix [23340] regression where you can't see the splitters being dragged when liveSplitters=false. Refs #11030, fixes #12078 !strict.