Bug 279016 – [GTK4] Gamepad tests are timing out when run with run-webkit-tests

• Status: NEW
• Component: WebKitGTK

[Claudio Saavedra]

It seems that the permissions from the flatpak sandbox are not enough for the gamepad tests to work, so they are timing out in GTK4. I am going to skip them for now.

Extending a bit on the reasons: I am not entirely sure but it might be lack of input device access. Reading the flatpak documentation, it's mentioned that from 1.15.6, one can use 'input' to request input device acess, but that's a very recent version (https://docs.flatpak.org/en/latest/sandbox-permissions.html#device-access ). I am not sure how that should be handled in older versions.

[Michael Catanzaro]

Yes, Georges determined in https://gitlab.gnome.org/GNOME/epiphany/-/merge_requests/1428 that we should use --device=input and that it's not a sandbox escape.

[Philippe Normand]

Aren't these tests supposed to rely on some gamepad fake/mock infra?

I guess we're not plugging a real gamepad to each and every test bot :)

[Claudio Saavedra]

(In reply to Philippe Normand from comment #2)
> Aren't these tests supposed to rely on some gamepad fake/mock infra?
> 
> I guess we're not plugging a real gamepad to each and every test bot :)


I'm really not familiar with this area. But here is what I found out: if you run the tests directly with WebKitTestRunner (webkit-flatpak --gtk -c WebKitTestRunner gamepad/....) the tests run correctly and pass. If you use run-webkit-tests, they time out.

First difference between these scripts I noticed is that run-webkit-tests runs sandboxed, whereas webkit-flatpak doesn't seem to. So maybe you're right and it's not input related, but something in the sandbox is preventing the tests from running.

[Philippe Normand]

This has nothing to do with the sandbox.

The issue is that UIGamepadProvider::platformWebPageProxyForGamepadInput() returns nullptr when invoked from run-webkit-tests because the gtk_window_is_active(window) call there returns false.

This is a bug in the GTK4 ifdefs of that code.

[Claudio Saavedra]

Thanks for debugging this, Phil!

It seems that the problem is that there's no direct replacement for gtk_window_has_toplevel_focus(). gtk_window_is_active() is not equivalent because the window is not visible while the tests run?

I am not so familiar with this, and a quick test with gtk_window_get_focus() and gtk_widget_has_focus() didn't really solve the issue, so I defer to other people who might know GTK better.