[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#622152: marked as done ([britney2] Port to python2.6 (s/PyFoo_Check/PyFoo_CheckExact/))

Your message dated Sat, 26 Nov 2011 20:52:10 +0000
with message-id <1322340731.16684.25.camel@hathi.jungle.funky-badger.org>
and subject line Re: Bug#622152: [britney2] Port to python2.6 (s/PyFoo_Check/PyFoo_CheckExact/)
has caused the Debian Bug report #622152,
regarding [britney2] Port to python2.6 (s/PyFoo_Check/PyFoo_CheckExact/)
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner@bugs.debian.org
immediately.)


-- 
622152: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=622152
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems
--- Begin Message --- 
Package: release.debian.org
Severity: normal
Tags: patch
User: release.debian.org@packages.debian.org
Usertags: britney

Hi.

It seems that britney2 segfaults at every run when compiled for Python
2.6. After some quick investigation, the problem boils down to the
following:

Britney2 uses PyFoo_Check to check that some structure has the correct
type. That class of functions changed between Python 2.5 and 2.6. The
(relevant part of the) diff is as follows:

In Include/dictobject.h:

-#define PyDict_Check(op) PyObject_TypeCheck(op, &PyDict_Type)
+#define PyDict_Check(op) \
+                 PyType_FastSubclass(Py_TYPE(op), Py_TPFLAGS_DICT_SUBCLASS)
-#define PyDict_CheckExact(op) ((op)->ob_type == &PyDict_Type)
+#define PyDict_CheckExact(op) (Py_TYPE(op) == &PyDict_Type)

In Include/object.h

 #define PyObject_TypeCheck(ob, tp) \
-       ((ob)->ob_type == (tp) || PyType_IsSubtype((ob)->ob_type, (tp)))
+    (Py_TYPE(ob) == (tp) || PyType_IsSubtype(Py_TYPE(ob), (tp)))

 #define PyType_HasFeature(t,f)  (((t)->tp_flags & (f)) != 0)
+#define PyType_FastSubclass(t,f)  PyType_HasFeature(t,f)

I (personally) don't care much about (how) python (does it) (tbh)
because in our case, we know that those structures have exactly the
type we are testing against. So, we can use PyFoo_CheckExact variants
instead (btw, that was winning test even with Python 2.5).

Attached is a simple patch that performs the described substitution.
Using this patch, I've been able to run britney2 on my machines
wihtout seeing any problem.

Regards,

-- System Information:
Debian Release: 6.0
  APT prefers stable
  APT policy: (500, 'stable'), (1, 'experimental')
Architecture: amd64 (x86_64)

Kernel: Linux 2.6.32-5-amd64 (SMP w/2 CPU cores)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash

>From e022f5f39203486c875e06ed47c73fbaf9f5aeeb Mon Sep 17 00:00:00 2001
From: Mehdi Dogguy <mehdi@debian.org>
Date: Fri, 11 Mar 2011 18:37:00 +0100
Subject: [PATCH 1/2] Use Py{Dict,List}_CheckExact instead of Py{Dict,List}_Check

---
 lib/britney-py.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/lib/britney-py.c b/lib/britney-py.c
index 69de370..8037e32 100644
--- a/lib/britney-py.c
+++ b/lib/britney-py.c
@@ -1,4 +1,4 @@
-#include <python2.5/Python.h>
+#include <python2.6/Python.h>
 
 #include "dpkg.h"
 
@@ -269,7 +269,7 @@ static PyObject *dpkgpackages_add_binary(dpkgpackages *self, PyObject *args) {
 	(void)self; /* unused */
 
     if (!PyArg_ParseTuple(args, "sO", &pkg_name, &value) ||
-        !PyList_Check(value)) return NULL;
+        !PyList_CheckExact(value)) return NULL;
 
     /* initialize the new package */
     dpkg_package *pkg;
@@ -950,7 +950,7 @@ static PyObject *build_system(PyObject *self, PyObject *args) {
 	(void)self; /* unused */
 
     if (!PyArg_ParseTuple(args, "sO", &arch, &pkgs) ||
-        !PyDict_Check(pkgs)) return NULL;
+        !PyDict_CheckExact(pkgs)) return NULL;
 
     /* Fields and positions for the binary package:
        # VERSION = 0
-- 
1.7.2.5

--- End Message ---
--- Begin Message --- 
On Mon, 2011-11-21 at 19:20 +0000, Adam D. Barratt wrote:
> On Sun, 2011-11-20 at 12:51 +0100, Mehdi Dogguy wrote:
> > On 11/20/2011 12:17 AM, Adam D. Barratt wrote:
> > > On Thu, 2011-11-17 at 20:54 +0000, Adam D. Barratt wrote:
> > >> I've finally (!) got to looking at this but have been unable to
> > >> reproduce the problem.  It might be a squeeze versus wheezy issue or
> > >> some oddity which has resolved itself in the meantime, but after
> > >> s/python2.5/python2.6/ in britney-py.c and rebuilding the library on
> > >> franck, everything seems to just work[tm].
> [...]
> > maybe we should close the bug then? (it could be a temporary problem that
> > got fixed somewhere since then…)
> 
> I was planning on running a py2.6 britney live for a couple of days and
> seeing what happened first; what's the worst that could happen? :)

Well, it seems to have worked okay so far, so yeah let's close this.

Regards,

Adam

--- End Message ---
Reply to: