Don't refresh if palette is within a hidden branch

[fcollonval]

Fixes #744

• Add new method Widget.isWithinHiddenWidget to test if a widget is hidden or within an hidden parent

I'm not happy with the name isWithinHiddenWidget nor with isAnyHidden. Happy to change it to something else; I thought about isAnyParentHidden

I would recommend merging #745 prior to this to ensure tests are passing on CI.

---

This PR has two side effects:

• commenting a left over skipped test from Forbid .only and .skip on CI #745 - the skipped test are marked as failing when run in debug mode but not otherwise it seems. I have no ideas why this is happening.
• I turn on the generation of source maps for tests; this makes debugging much more efficient.

[krassowski]

I would recommend merging #745 prior to this to ensure tests are passing on CI.

I just merged #745.

I'm not happy with the name isWithinHiddenWidget nor with isAnyHidden. Happy to change it to something else

Sounds good. Some food for thought: in Qt, isVisible() and isHidden() are not exact equivalents. isHidden == true means that isVisible == false, but isVisible == false does not mean that the widget isHidden (because it may not be visible due to the parent being hidden, or due to being detached). isHidden returns per-widget state ignoring parents, isVisible returns the actual visibility (ignoring it being obscured by other windows etc).

If an ancestor is not visible, the widget won't become visible until all its ancestors are shown

A hidden widget will only become visible when show() is called on it. It will not be automatically shown when the parent is shown.

To check visibility, use !isVisible() instead (notice the exclamation mark).

isHidden() implies !isVisible(), but a widget can be not visible and not hidden at the same time. This is the case for widgets that are children of widgets that are not visible.

https://doc.qt.io/qt-5/qwidget.html#isHidden

isAnyParentHidden

I think this name is fine - unless we want to mirror Qt way of doing things.

One advantage of doing it Qt way is that we could also encapsulate "is any parent not attached to the DOM" in the same logic. What do you think?

[fcollonval]

Thanks @krassowski for all the thoughts and ideas. And actually as PhosphorJS mimick Qt API, this is indeed what isVisible doc mentions...

lumino/packages/widgets/src/widget.ts

Lines 120 to 127 in 7251957

	/**
	* Test whether the widget is visible.
	*
	* #### Notes
	* A widget is visible when it is attached to the DOM, is not
	* explicitly hidden, and has no explicitly hidden ancestors.
	*/
	get isVisible(): boolean {

So in e40bd13, I removed the new API to instead change the behavior of isVisible and I deprecated the flag Widget.Flag.isVisible as this is not reliable.

I gave a shot at adding a default layout that I thought would solve the issue. But it did not and more importantly it was having much more side effects.