backend: Make refreshing newly added tasks more robust

The mechanism for obtaining information about newly added task (also
called refreshing) was a bit fragile in situations when the only
reference to the new task object is an ID number.

In particular, if taskwarrior has some GC cleanup pending, and a new
task is added through tasklib, the GC will be performed while exporting
the information about the newly added task, thus changing the IDs and
rendering the ID reference invalid.

This effect is now mitigated in two ways:
1.) We use the new-uuid (available since 2.4.0) verbosity token, which
    makes "task add" command output a UUID for the newly added task.
2.) The GC mechanism is disabled during refreshing. This will only be
    useful if a manual ID reference was used to get a task by the user.

Author: tbabej

tasklib/backends.py

@@ -330,16 +330,19 @@ def save_task(self, task):
 
         args = [task['uuid'], 'modify'] if task.saved else ['add']
         args.extend(self._get_modified_task_fields_as_args(task))
-        output = self.execute_command(args)
+        output = self.execute_command(
+            args,
+            config_override={'verbose': 'new-uuid'}
+        )
 
         # Parse out the new ID, if the task is being added for the first time
         if not task.saved:
             id_lines = [l for l in output if l.startswith('Created task ')]
 
             # Complain loudly if it seems that more tasks were created
             # Should not happen.
-            # Expected output: Created task 1.
-            #                  Created task 1 (recurrence template).
+            # Expected output: Created task bd23f69a-a078-48a4-ac11-afba0643eca9.
+            #                  Created task bd23f69a-a078-48a4-ac11-afba0643eca9 (recurrence template).
             if len(id_lines) != 1 or len(id_lines[0].split(' ')) not in (3, 5):
                 raise TaskWarriorException(
                     'Unexpected output when creating '
@@ -349,11 +352,8 @@ def save_task(self, task):
             # Circumvent the ID storage, since ID is considered read-only
             identifier = id_lines[0].split(' ')[2].rstrip('.')
 
-            # Identifier can be either ID or UUID for completed tasks
-            try:
-                task._data['id'] = int(identifier)
-            except ValueError:
-                task._data['uuid'] = identifier
+            # Identifier is UUID, because we used new-uuid verbosity override
+            task._data['uuid'] = identifier
 
         # Refreshing is very important here, as not only modification time
         # is updated, but arbitrary attribute may have changed due hooks
@@ -385,7 +385,11 @@ def refresh_task(self, task, after_save=False):
         # of newly saved tasks. Any other place in the code is fine
         # with using UUID only.
         args = [task['uuid'] or task['id'], 'export']
-        output = self.execute_command(args)
+        output = self.execute_command(
+            args,
+            # Supress GC, which can change ID numbers (undesirable for refresh)
+            config_override={'gc': '0'}
+        )
 
         def valid(output):
             return len(output) == 1 and output[0].startswith('{')