[Pkg-privacy-maintainers] Bug#833821: txtorcon: Should age() use an old new timestamp?

[Petter Reinholdtsen]

Package: txtorcon
Version: 0.15.0-1
Tags: patch

I noticed this issue because the package fail to build reproducably, and
sphinx generate documentation where the now statement below is replaced
with a concrete timestamp.  See <URL: > for the details about this.

But while looking at the code, it occured to me that it probably is not
working as the author intended.  If I remember correctly (based on a
blog post from Chris Lamb[1]), the default function arguments in Python
are evaluated when the code is parsed (ie when the 'import' statements
are evaluated).  The effect of this is that the age() function is not
give a fresh new timestamp generated at call time, but the timestamp
generated when the file was imported.

 [1] <URL: https://chris-lamb.co.uk/posts/python-quirk-signatures-are-evaluated-import-time >

I suspect the author wanted fresh timestamp, and suggest to rewrite the
age() function like this to make sure age() without any now value
specified actually uses a fresh timestamp:

diff --git a/txtorcon/circuit.py b/txtorcon/circuit.py
index 5166fc1..cd9a3cf 100644
--- a/txtorcon/circuit.py
+++ b/txtorcon/circuit.py
@@ -157,7 +157,7 @@ class Circuit(object):
         d.addCallback(close_command_is_queued)
         return self._closing_deferred
 
-    def age(self, now=datetime.datetime.utcnow()):
+    def age(self, now=None):
         """
         Returns an integer which is the difference in seconds from
         'now' to when this circuit was created.
@@ -166,6 +166,8 @@ class Circuit(object):
         """
         if not self.time_created:
             return None
+        if now == None:
+            now = datetime.datetime.utcnow()
         return (now - self.time_created).seconds
 
     def _create_flags(self, kw):

-- 
Happy hacking
Petter Reinholdtsen