No longer segfault on new versions. This fixes #4

CHANGES

@@ -17,6 +17,13 @@ Version 2.6
   of attributes.
 - scoped blocks not properly treat toplevel assignments and imports.
   Previously an import suddenly "disappeared" in a scoped block.
+- automatically detect newer Python interpreter versions before loading code
+  from bytecode caches to prevent segfaults on invalid opcodes.  The segfault
+  in earlier Jinja2 versions here was not a Jinja2 bug but a limitation in
+  the underlying Python interpreter.  If you notice Jinja2 segfaulting in
+  earlier versions after an upgrade of the Python interpreter you don't have
+  to upgrade, it's enough to flush the bytecode cache.  This just no longer
+  makes this necessary, Jinja2 will automatically detect these cases now.
 
 Version 2.5.5
 -------------

jinja2/bccache.py

@@ -15,6 +15,7 @@
     :license: BSD.
 """
 from os import path, listdir
+import sys
 import marshal
 import tempfile
 import cPickle as pickle
@@ -27,8 +28,16 @@ except ImportError:
 from jinja2.utils import open_if_exists
 
 
-bc_version = 1
-bc_magic = 'j2'.encode('ascii') + pickle.dumps(bc_version, 2)
+bc_version = 2
+
+# magic version used to only change with new jinja versions.  With 2.6
+# we change this to also take Python version changes into account.  The
+# reason for this is that Python tends to segfault if fed earlier bytecode
+# versions because someone thought it would be a good idea to reuse opcodes
+# or make Python incompatible with earlier versions.
+bc_magic = 'j2'.encode('ascii') + \
+    pickle.dumps(bc_version, 2) + \
+    pickle.dumps((sys.version_info[0] << 24) | sys.version_info[1])
 
 
 class Bucket(object):