Bug 273245 – REGRESSION(274201@main): Error embedding AT-SPI socket: GDBus.Error:org.freedesktop.DBus.Error.NameHasNoOwner: org.freedesktop.DBus.Error.NameHasNoOwner

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKitGTK

[Michael Catanzaro]

With Epiphany Tech Preview using WebKitGTK 2.44.1:

 * Load about:overview in Epiphany
 * Load https://example.com

You'll see this warning:

(epiphany:2): Gtk-WARNING **: 07:46:40.187: Error embedding AT-SPI socket: GDBus.Error:org.freedesktop.DBus.Error.NameHasNoOwner: 
org.freedesktop.DBus.Error.NameHasNoOwner

Sometimes, but not always, simply opening a new tab is sufficient to trigger the warning to print. I have a huge spam of such warnings in my terminal.

Georges says it's a GTK bug, https://gitlab.gnome.org/GNOME/gtk/-/issues/6581

[Michael Catanzaro]

Pull request: https://github.com/WebKit/WebKit/pull/27740

[Michael Catanzaro]

New finding: the warning occurs only under flatpak, and it seems to block a11y from working (at least, orca cannot read my web views)

[Michael Catanzaro]

Notably, the name we're trying to own looks like this:

/org/a11y/webkit/accessible/f0d99090_bf5d_4e81_af25_9c4d1eff7583

Matthias suggests xdg-dbus-proxy might be blocking us from owning the name. We might only have permission to own names that match the app ID? (However, it's strange that we don't see a separate error message for failing to own the name!)

[Michael Catanzaro]

Matthias says no: there are no policies for ownership of object paths, only for bus names.

[Adrian Perez]

While discussing in the WebKitGTK chat room we realized that the issue seems
to pop only when running e.g. an Epiphany Flatpak. It should be easy enough
restricting connections to the a11y socket iff !WTF::isInsideFlatpak() instead
of a full revert. I am testing a patch in this direction and will submit an
alternative PR momentarily.

[Michael Catanzaro]

The actual bus name looks like ":1.261" but that seems expected.

[Adrian Perez]

Pull request: https://github.com/WebKit/WebKit/pull/27741

[Michael Catanzaro]

So poking around a bit more, I found AccessibilityAtspi::registerObject in AccessibilityAtspi.cpp is where the plug ID is actually coming from. And it's ignoring all D-Bus errors. ;)

[EWS]

Committed 278054@main (bfbc2e6fa7ad): <https://commits.webkit.org/278054@main>

Reviewed commits have been landed. Closing PR #27741 and removing active labels.

[Adrian Perez]

Reopening to do the actual fix later on =)

[Georges Basile Stavracas Neto]

Pull request: https://github.com/WebKit/WebKit/pull/32247

[EWS]

Committed 284894@main (3672c423722b): <https://commits.webkit.org/284894@main>

Reviewed commits have been landed. Closing PR #32247 and removing active labels.