[Linux] [MacOS] CTabFolder renders tabs with a delay

[basilevs]

CTabFolder does not render tabs in a timely manner on Linux and MaxOS x86_64. The problem is not observed on Windows and MacOS aarch64.

Test steps

Execute following test from #3295

@ParameterizedTest
@CsvSource(value = {"false,0", "false,1", "true,0", "true,1"})
public void test_tabsAreRendered(boolean nestComposite, int activateTab) {
	Composite parent = shell;
	Control[] children = shell.getChildren();
	for (Control child : children) {
		child.dispose();
	}
	if (nestComposite) {
		parent = new Composite(parent, SWT.NONE);
		parent.setLayout(new FillLayout());
	}
	ctabFolder = new CTabFolder(parent, SWT.NONE);
	setWidget(ctabFolder);
	shell.setSize(800, 400);
	ctabFolder.setBackground(RED);
	ctabFolder.setForeground(BLUE);
	ctabFolder.setSelectionForeground(BLUE);
	CTabItem tab = createTabItem(1);
	openShell(shell);
	processEvents();
	Rectangle bounds = tab.getBounds();
	assertTrue(hasPixel(ctabFolder, BLUE, bounds));
	tab = createTabItem(2);
	ctabFolder.setSelection(activateTab);
	SwtTestUtil.processEvents();
	bounds = tab.getBounds();
	assertTrue(hasPixel(ctabFolder, BLUE, bounds)); // reference 1
	tab.dispose();
	SwtTestUtil.processEvents();
	assertFalse(hasPixel(ctabFolder, BLUE, bounds)); // reference 2
}

private CTabItem createTabItem(int i) {
	CTabItem item = new CTabItem(ctabFolder, SWT.CLOSE);
	item.setText("█".repeat(10));
	Label content = new Label(ctabFolder, SWT.NONE);
	content.setText("Content " + i);
	item.setControl(content);
	return item;
}

The test creates two tabs, activates one of them and ensures that tab text is rendered. The test always fails for any parameter.

A characteristic screenshot captured from CI environment for GTK3, x86_64, x11:

Instead of expected two tabs, only one is shown. This happens for all parameters but the last one failing test on line "reference 1".

[basilevs]

I've added more delays and it seems tabs are rendered after a while.
On MacOS test now gives false positive due to hasPixel mechanism:

[basilevs]

@akurtakov could you please review if current implementation of org.eclipse.swt.tests.junit.SwtTestUtil.hasPixel(Control, Color, Rectangle) is correct?

/**
 * Check if widget contains the given color in given bounds. The effective
 * search range to find the color is the union of current widget bounds and
 * given rectangle.
 *
 * @param control       widget to check
 * @param expectedColor color to find
 * @param rect          the bounds where the color is searched in. Can overlap
 *                      the control bounds or <code>null</code> to check the
 *                      widgets full bounds.
 * @return <code>true</code> if the given color was found in search range of
 *         widget
 */
public static boolean hasPixel(Control control, Color expectedColor, Rectangle rect) {
	ImageGcDrawer noOpGcDrawer = (gc, height, width) -> {};
	GC gc = new GC(control);
	final Image image = new Image(control.getDisplay(), noOpGcDrawer, control.getSize().x, control.getSize().y);
	gc.copyArea(image, 0, 0);
	gc.dispose();
	boolean result = hasPixel(image, expectedColor, rect);
	image.dispose();
	return result;
}