Increase trash backend resilience of against stale NFS mounts
The stale NFS mounts can now easily block the whole backend. This is especially problem for the create_dir_monitor
and
create_file_monitor
methods that don't have asynchronous variants. This may cause hangs for nautilus, or file chooser dialog, but also for
the whole gnome-shell session (resp. desktop-icons extension). Let's run blocking methods on a thread pool and add a working thread for the rest of the problematic code. This should ensure that the create_dir_monitor
and create_file_monitor
methods respond immediately.
Related: https://bugzilla.redhat.com/show_bug.cgi?id=2152538