verify_directory should respect umask

[sqlalchemy-bot]

Migrated issue, originally created by Anonymous

Currently util.verify_directory is hardcoding the mode of 0750 in its call to os.makedirs. The following patch checks the umask and passes mode along accordingly:

Index: lib/mako/util.py
===================================================================
--- lib/mako/util.py    (revision 439)
+++ lib/mako/util.py    (working copy)
@@ -17,6 +17,7 @@
     from StringIO import StringIO
 
 import codecs, re, weakref, os, time
+import subprocess
 
 try:
     import threading
@@ -38,7 +39,10 @@
     while not os.path.exists(dir):
         try:
             tries += 1
-            os.makedirs(dir, 0750)
+            umask, _ = subprocess.Popen(['sh', '-c', 'umask'], stdout=subprocess.PIPE).communicate()
+            umask = int(umask, 8)
+            mode = int(oct(0777 ^ umask), 8)
+            os.makedirs(dir, mode)
         except:
             if tries > 5:
                 raise

[sqlalchemy-bot]

Anonymous wrote:

looks like this code was lifted from beaker.util. filed a [http://pylonshq.com/project/pylonshq/ticket/567 corresponding ticket] in the pylons trac.

[sqlalchemy-bot]

Anonymous wrote:

note: the line

mode = int(oct(0777 ^ umask), 8)

could of course be simplified to

mode = 0777 ^ umask

if you're into the whole brevity thing.

[sqlalchemy-bot]

Anonymous wrote:

alternatively it may be preferable to effect more permissions-sensitive behavior via configuration options mako.umask or some such.

[sqlalchemy-bot]

Anonymous wrote:

discussed with zzzeek in #pylons, decided this is really a [http://bugs.python.org/issue5220 python bug]. a workaround like the above (though preferably a more platform independent one) can be used if desired until this is fixed in python.

[sqlalchemy-bot]

Anonymous wrote:

Replying to [comment:4 guest]:

decided this is really a [http://bugs.python.org/issue5220 python bug]
scratch that. it seems like we were misunderstanding the mode parameter (see response in the python ticket). otherwise why pass 0750? it effectively throws out part of the umask.

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

I think we decided here to just use 0775 for the mask.

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

43b362d

[sqlalchemy-bot]

Changes by Michael Bayer (@zzzeek):

• changed status to closed