[Edge] No more handle exceptions from Edge browser

[vogella]

In our RCP application we see no more handle exceptions during startup if we use Edge. This does not happen all the time, there seems to be a race condition or the error is reported incorrectly similar to https://bugs.eclipse.org/bugs/show_bug.cgi?id=576196.

If the user starts again the RCP application, the errors does not happen again, sometimes is takes 3-4 attempts to start the application. We are using latest version from https://developer.microsoft.com/de-de/microsoft-edge/webview2/#download-section, currently 104.0.1293.63 pointing to the download with -Dorg.eclipse.swt.browser.DefaultType=edge -Dorg.eclipse.swt.browser.EdgeDir=c:\PROGRA~1\WebView2 in the runtime arguments.

This happens in the Edge.createmethod, relevant stack trace:

org.eclipse.swt.SWTError: No more handles [0x8007139f]
at org.eclipse.swt.SWT.error(SWT.java:4944)
at org.eclipse.swt.browser.Edge.error(Edge.java:190)
at org.eclipse.swt.browser.Edge.create(Edge.java:343)
at org.eclipse.swt.browser.Browser.<init>(Browser.java:99)
at de.example.rcp.base.views.NavigationPanelView.createPartControl(NavigationPanelView.java:242)

Here is the code which triggers the exception:

		try {
			if (hiddenBrowser == null || hiddenBrowser.isDisposed()) {
				hiddenBrowser = new Browser(bar, SWT.EDGE);
				hiddenBrowser.setBounds(0, 0, 0, 0);
				hiddenBrowser.setVisible(false);
			}
		} catch (Exception  | Error  e) {
			LOG.warn("Unable to create hidden browser", e);
		}

callAndWait seems to spins the event queue, maybe this causes the other thread to sync with main thread and to dispose the composite?

static int callAndWait(long[] ppv, ToIntFunction<IUnknown> callable) {
	int[] phr = {COM.S_OK};
	IUnknown completion = newCallback((result, pv) -> {
		phr[0] = (int)result;
		if ((int)result == COM.S_OK) {
			ppv[0] = pv;
			new IUnknown(pv).AddRef();
		}
		return COM.S_OK;
	});
	ppv[0] = 0;
	phr[0] = callable.applyAsInt(completion);
	completion.Release();
	Display display = Display.getCurrent();
	while (phr[0] == COM.S_OK && ppv[0] == 0) {
		if (!display.readAndDispatch()) display.sleep();
	}
	return phr[0];
}

The incorrect behavior reported via https://bugs.eclipse.org/bugs/show_bug.cgi?id=576196 can be reproduced with the following snippet.

	public void createComposite(Composite parent) {
		parent.setLayout(new GridLayout(1, false));

		for (int i = 0; i <= 10; i++) {
			System.out.println("Iteration " + i);
			Composite c = new Composite(parent, SWT.NONE);
			Display.getCurrent().asyncExec(new Runnable() {

				@Override
				public void run() {
					c.dispose();
				}
			});
			Browser browser = null;
			if (!c.isDisposed()) {
				browser = new Browser(c, SWT.EDGE);
			}
			if (browser != null) {
				browser.dispose();
			}
		}

	}

Expected behavior

Correct error message or even better, just working.

Environment:

1. Select the platform(s) on which the behavior is seen:

• • All OS
• • Windows
• • Linux
• • macOS

2. Additional OS info (e.g. OS version, Linux Desktop, etc)

3. JRE/JDK version

Java 11

Version since

Still present in latest I-Buil

Workaround (or) Additional context

Start again the RCP application.

[vogella]

@nnemkin can you have a look? This affects the production environment of one of my clients.

[nnemkin]

I remember this one. The root cause is that the loop that fakes synchronous operation (callAndWait above) uses Display.readAndDispatch, which allows asyncExec actions to execute. This blows up when one of those actions does dispose(). I've tried switching the wait to low level OS primitives (GetMessage() etc.), but it broke some other things because asyncExec is also used internally by the Browser to queue event handlers.

I guess the proper fix is remove Display.readAndDispatch from the loop and also remove asyncExec from the Browser implementation. Unfortunately this means that the Browser will have to invent its own asyncExec equivalent to use with its own wait loop. I'll see what can be done.

[vogella]

Thanks @nnemkin

[Destrolaric]

I can confirm we have something similar since we added edge by default in our client. It happens randomly when opening a browser instance.

[Destrolaric]

@nnemkin Hi, any progress here?

[mandavaakhilkumar]

Hi @nnemkin @vogella any work arounds for this issue ?

[SyntevoAlex]

To my knowledge, @nnemkin is away. He hasn't contributed anything in last 8 months or so.