Results loading now takes place in one shot (file locate and metadata read). It makes weeding out the bad files more convenient and fixes the Cancel loading glitch where we end up with "ghost" results.

core/app.py

@@ -82,17 +82,7 @@ class DupeGuru(RegistrableApplication, Broadcaster):
     def _do_load(self, j):
         self.directories.load_from_file(op.join(self.appdata, 'last_directories.xml'))
         self.notify('directories_changed')
-        j = j.start_subjob([1, 9])
         self.results.load_from_xml(op.join(self.appdata, 'last_results.xml'), self._get_file, j)
-        files = flatten(g[:] for g in self.results.groups)
-        try:
-            for file in j.iter_with_progress(files, 'Reading metadata %d/%d'):
-                file._read_all_info(attrnames=self.data.METADATA_TO_READ)
-        except (OSError, IOError):
-            # If this error is raised, it means that a file was deleted while we were reading
-            # metadata. Proper handling of this rare occurrence is complex because there's no easy
-            # way to remove an arbitrary file from the Results. Thus, we simply clear them.
-            self.results.groups = []
     
     def _get_display_info(self, dupe, group, delta=False):
         if (dupe is None) or (group is None):
@@ -105,7 +95,12 @@ class DupeGuru(RegistrableApplication, Broadcaster):
     
     def _get_file(self, str_path):
         path = Path(str_path)
-        return fs.get_file(path, self.directories.fileclasses)    
+        f = fs.get_file(path, self.directories.fileclasses)    
+        try:
+            f._read_all_info(attrnames=self.data.METADATA_TO_READ)
+            return f
+        except EnvironmentError:
+            return None
     
     def _job_completed(self, jobid):
         # Must be called by subclasses when they detect that an async job is completed.