I maintain virt-manager, which will use libappindicator for its statusicon support, if libappindicator is installed. The appindicator is initialized like this:

            self.systray_icon = AppIndicator3.Indicator.new(
                                "virt-manager",
                                "virt-manager",
                                AppIndicator3.IndicatorCategory.OTHER)


The first "virt-manager" is the app name, the second "virt-manager" is the icon name.

When running this on gnome-shell, libappindicator needs to fall back to using a traditional StatusIcon. However it has some weird logic for picking the icon to use;

    if (icon_name != NULL) {
        if (g_file_test(icon_name, G_FILE_TEST_EXISTS)) {
            gtk_status_icon_set_from_file(icon, icon_name);
        } else {
            gchar *longname = append_panel_icon_suffix(icon_name);

            if (longname != NULL && gtk_icon_theme_has_icon (icon_theme, longname)) {
                gtk_status_icon_set_from_icon_name(icon, longname);
            } else {
                gtk_status_icon_set_from_icon_name(icon, icon_name);
            }

            g_free(longname);
        }
    }


The first thing it checks for is a file in CWD named $icon_name, and if so, tries to use that for the icon. So when I would try to test this when running it from virt-manager.git, which contains a python script named 'virt-manager', the statustray area would be empty (which is actually worse than a missing icon).

I don't know what the purpose of the file_test is, maybe a test suite depends on it or something, but it definitely causes weird behavior. I figure it should check the icon_theme first, and at least perform some error checking on the icon file loading, it may not even be an image.