[meld] undo: Ensure that each buffer has a starting checkpoint (#159)

[Kai Willadsen <kaiw src gnome org>]

commit 09ecc4188626d48ed25f03287103d33e2082a375
Author: Kai Willadsen <kai willadsen gmail com>
Date:   Sun Feb 11 08:36:40 2018 +1000

    undo: Ensure that each buffer has a starting checkpoint (#159)
    
    With this change, clearing the undo sequence will still ensure that each
    buffer starts with an undo checkpoint.
    
    Previously, this was *almost* always the case, but on reload of an
    individual file we would clear out the checkpoints for all files, and
    only re-checkpoint the file we reloaded. This caused several weird
    issues, chief among which was that you cause a file to show as being
    changed (along with the action sensitivity issues accompanying this).

 meld/undo.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
---
diff --git a/meld/undo.py b/meld/undo.py
index 691a7b17..7f224938 100644
--- a/meld/undo.py
+++ b/meld/undo.py
@@ -100,7 +100,10 @@ class UndoSequence(GObject.GObject):
             self.emit('can-redo', 0)
         self.actions = []
         self.next_redo = 0
-        self.checkpoints = {}
+        self.checkpoints = {
+            # Each buffer's checkpoint starts at zero and has no end
+            ref(): [0, None] for ref in self.buffer_refs
+        }
         self.group = None
         self.busy = False