Correctly recreate the database directory during tracker re-initialise

The database directory value was being fetched from the tracker config
and being appended to the tracker home directory, but the database
directory value in the config already has the tracker home prepended
which resulted in the database directory being recreated in a
nonsensical location.

This fix is a bit of a hack, but the relevant code is likely to be
removed in v1.6, so it shouldn't be too bad in the short term.

Author: jerrykan

CHANGES.txt

@@ -121,6 +121,8 @@ Fixed:
   application/octet-stream whenever 'html' is contained in the string
   (case insensitive). Thanks to Kay Hayen for reporting and helping
   debug this. (Ralf Schlatterbeck)
+- Correctly recreate the database directory when re-initialising a tracker
+  instance. (John Kristensen)
 
 Minor:
 - demo.py usage message improved: explains "nuke" now. (Bernhard Reiter)

roundup/init.py

@@ -175,12 +175,15 @@ def saveTemplateInfo(dir, info):
     finally:
         f.close()
 
-def write_select_db(instance_home, backend, dbdir = 'db'):
+def write_select_db(instance_home, backend, dbdir=None):
     ''' Write the file that selects the backend for the tracker
     '''
-    # dbdir may be a relative pathname, os.path.join does the right
-    # thing when the second component of a join is an absolute path
-    dbdir = os.path.join (instance_home, dbdir)
+    # dbdir is only supplied when AdminTool.do_initialise() invokes this
+    # function and the value is fetched from the tracker config which has
+    # already determined the correct path. This is bit of a hack, but it is
+    # likely this function will be removed in v1.6
+    if not dbdir:
+        dbdir = os.path.join(instance_home, 'db')
     if not os.path.exists(dbdir):
         os.makedirs(dbdir)
     f = open(os.path.join(dbdir, 'backend_name'), 'w')