homepage

In the 3.2.3 interpreter execute the following line:
 `None == datetime.timezone(datetime.timedelta())`
The interpreter will crash with a `Segmentation fault: 11`.

Reproducible also on Linux with Python 3.3.

Crashes Python 3.2.3 and Python 3.3.0rc2 on Windows 7 as well.

I cannot reproduce on a Mac with py3k tip.
Python 3.3.0rc2+ (default:19c74cadea95, Sep 19 2012, 14:39:07) 
[GCC 4.2.1 Compatible Apple Clang 3.1 (tags/Apple/clang-318.0.61)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import datetime
>>> None == datetime.timezone(datetime.timedelta())
False
Can anyone reproduce in a debug build and post a stack trace?

The segfault does not occur in a debug build. The stack trace
suggests that timezone_richcompare() accesses other->offset
of the None object:
(gdb) f 2
#2 0x000000000041d4e9 in do_richcompare (v=None, w=<datetime.timezone at remote 0x7ffff6688ab0>, 
 op=<value optimized out>) at Objects/object.c:563
563 res = (*f)(w, v, _Py_SwappedOp[op]);
(gdb) f 1
#1 timezone_richcompare (self=0x7ffff6688ab0, other=<value optimized out>, op=2)
 at /home/stefan/pydev/cpython-commit/Modules/_datetimemodule.c:3218
3218 return delta_richcompare(self->offset, other->offset, op);
(gdb) l
3213 timezone_richcompare(PyDateTime_TimeZone *self,
3214 PyDateTime_TimeZone *other, int op)
3215 {
3216 if (op != Py_EQ && op != Py_NE)
3217 Py_RETURN_NOTIMPLEMENTED;
3218 return delta_richcompare(self->offset, other->offset, op);
3219 }
3220
3221 static Py_hash_t
3222 timezone_hash(PyDateTime_TimeZone *self)
(gdb) f 0
#0 0x00007ffff5d850cf in delta_richcompare (self=0x7ffff6688ab0, other=<value optimized out>, op=2)
 at /home/stefan/pydev/cpython-commit/Modules/_datetimemodule.c:1823
1823 if (PyDelta_Check(other)) {

On linux it segfaults for me in the debug interpreter. On default tip.

I think the following simple patch should do the trick. I'll add some tests and commit. Should this get in 3.3.0? 
diff -r 19c74cadea95 Modules/_datetimemodule.c
--- a/Modules/_datetimemodule.c	Wed Sep 19 08:25:01 2012 +0300
+++ b/Modules/_datetimemodule.c	Wed Sep 19 21:42:51 2012 -0400
@@ -3215,6 +3215,11 @@
 {
 if (op != Py_EQ && op != Py_NE)
 Py_RETURN_NOTIMPLEMENTED;
+ if (Py_TYPE(other) != &PyDateTime_TimeZoneType)
+	if (op == Py_EQ)
+	 Py_RETURN_FALSE;
+	else
+	 Py_RETURN_TRUE;
 return delta_richcompare(self->offset, other->offset, op);
 }

Hopefully, there will be some more braces, though. :)

What about datetime subclasses?

> What about datetime subclasses?
Do you mean timezone subclasses? Timezone type is not subclassable, but we should probably support comparison with any tzinfo subclass. I'll add this logic, but arguably that would be a new feature.

Alexander, did you send a contributor agreement?. I don't see it in the tracker :-??

> Alexander, did you send a contributor agreement?
At least twice. :-)

Looks good. It would be nice to have this in 3.3.0. There are a couple
of blockers open, so perhaps this could go in, too.
Georg, are we going to have an rc3 anyway?

New changeset 9fba12ceb2fd by Alexander Belopolsky in branch '3.2':
Issue #15973: Fixed segmentation fault on timezone comparison to other types.
http://hg.python.org/cpython/rev/9fba12ceb2fd 

Leeaving the issue open in case it will go to 3.3.0. Reassigning to the RM.

Your change does not compile on Windows:
_datetimemodule.c
..\Modules\_datetimemodule.c(3247) : error C2065: 'Py_RETURN_NOTIMPLEMENTED' : undeclared identifier
http://buildbot.python.org/all/builders/AMD64%20Windows7%20SP1%203.2/builds/194/steps/compile/logs/stdio 

New changeset f17f67f0ec4b by Alexander Belopolsky in branch '3.2':
Issue #15973: fixed 3.2 backport.
http://hg.python.org/cpython/rev/f17f67f0ec4b 

Which commit(s) do I have to pick?

This one: 63cb0a642c84 

Done as ec77f8fb9958.

New changeset ec77f8fb9958 by Georg Brandl in branch 'default':
Closes #15973: fix a segmentation fault when comparing timezone objects.
http://hg.python.org/cpython/rev/ec77f8fb9958