And again it's posing problem. As advised by svuorela on irc, here is a
version that defers the trigger. This is probably even safer since
that's what is committed upstream for the AT_SPI_BUS_ADDRESS environment
variable case.
Samuel
--- a/src/platformsupport/linuxaccessibility/bridge.cpp
+++ b/src/platformsupport/linuxaccessibility/bridge.cpp
@@ -65,6 +65,16 @@ QSpiAccessibleBridge::QSpiAccessibleBrid
{
dbusConnection = new DBusConnection();
connect(dbusConnection, SIGNAL(enabledChanged(bool)), this,
SLOT(enabledChanged(bool)));
+ // Now that we have connected the signal, make sure we didn't miss a
change,
+ // e.g. when running as root or when AT_SPI_BUS_ADDRESS is set by hand.
+ // But do that only on next loop, once dbus is really settled.
+ QMetaObject::invokeMethod(
+ this,
+ [this] {
+ if (dbusConnection->isEnabled())
+ enabledChanged(true);
+ },
+ Qt::QueuedConnection);
}
void QSpiAccessibleBridge::enabledChanged(bool enabled)
