Make backup before compressing (low memory servers)

Fix error : Call post hook plugins even in display only mode
Don't compute unordered hits (remove pasts if they are found after current)
Remove tags in stats diff
Don't do geolocalisation is visitor is not valid
Don't try to find search engine on robots
Update robot check rules
Add top_pages_diff plugin

iwla.py

@@ -252,12 +252,18 @@ class IWLA(object):
         if not os.path.exists(base):
             os.makedirs(base)
 
+        # Make a backup in case of something fails
+        if os.path.exists(filename):
+            shutil.copy(filename, filename + '.bak')
+            
         with open(filename + '.tmp', 'wb+') as f, self._openDB(filename, 'w') as fzip:
             cPickle.dump(obj, f)
-            os.fsync(f)
             f.seek(0)
             fzip.write(f.read())
+            os.fsync(fzip)
         os.remove(filename + '.tmp')
+        if os.path.exists(filename + '.bak'):
+            os.remove(filename + '.bak')
 
     def _deserialize(self, filename):
         if not os.path.exists(filename):
@@ -626,15 +632,13 @@ class IWLA(object):
 
         duplicated_stats['nb_visitors'] = stats['nb_visitors'] = len(self.valid_visitors.keys())
 
+        self._callPlugins(conf.POST_HOOK_DIRECTORY)
+
         if args.display_only:
             self._generateDisplay()
             return
 
-        self._callPlugins(conf.POST_HOOK_DIRECTORY)
-
         path = self.getDBFilename(cur_time)
-        if os.path.exists(path) and not self.dry_run:
-            os.remove(path)
 
         self.logger.info("==> Serialize to %s" % (path))
         self._serialize(self.current_analysis, path)
@@ -701,6 +705,8 @@ class IWLA(object):
                 self.logger.debug("Not in time")
                 return False
             self.analyse_started = True
+            if t < cur_time: # Don't accept past hits
+                return False
             if cur_time.tm_mon != t.tm_mon:
                 self._generateDayStats()
                 self._generateMonthStats()