[ http://issues.apache.org/jira/brows...243
1107 ]

Jim Gallacher commented on MODPYTHON-187:
-----------------------------------------

The problem seems to be in the tableobject.c function table_subscript().

subprocess_env.get['SCRIPT_FILENAME'] is being evaluated in the post rea
d request phase. It would seem that req.add_common_vars adds a key for SCRIP
T_FILENAME to the table, but since the filename has not been translated yet 
its value is NULL.

table_subscript() retrieves the value with the following bit of code:

while (i--)
if (elts[i].key) {
if (apr_strnatcasecmp(elts[i].key, k) == 0) {
PyObject *v = PyString_FromString(elts[i].val);
PyList_Insert(list, 0, v);
Py_DECREF(v);
}
}

The problem is with the PyString_FromString. According to the Python docs, t
he parameter passed must not be NULL and it will not be checked. I assume th
is is the origin of the segfault.

Something like the following seems to fix the problem.

The following code snippet seems to fix the problem, and is attached as MP18
7-2006-08-28-jgallacher-1.diff. Am I correct in doing the PY_INCREF(v)??

while (i--)
if (elts[i].key) {
if (apr_strnatcasecmp(elts[i].key, k) == 0) {
-                PyObject *v = PyString_FromString(elts[i].val);
+                PyObject *v = NULL;
+                if (elts[i].val != NULL) {
+                    v = PyString_FromString(elts[i].val);
+                } else {
+                    v = Py_None;
+                    Py_INCREF(v);
+                }
PyList_Insert(list, 0, v);
Py_DECREF(v);





> Hang on subscripted access to request.subprocess_env.
> -----------------------------------------------------
>
>                 Key: MODPYTHON-187
>                 URL: http://issues.apache.org/jira/browse/MODPYTHON-187
>             Project: mod_python
>          Issue Type: Bug
>          Components: core
>    Affects Versions: 3.2.10
>         Environment: Apache:     2.0.59
> ModPython:  3.2.10
> Python:     2.4
> OS:         Windows Server 2003
>            Reporter: Alan Kennedy
>
> When subscripted access is used to access variable 'SCRIPT_FILENAME' in th
e request.subprocess_env table/mapping, the code hangs.
> The following snippet illustrates the problem.
> # -=-=-=-=-=-=
> def postreadrequesthandler(request):
>   request.add_common_vars()
>   value = request.subprocess_env['SCRIPT_FILENAME']    # This hangs
> # value = request.subprocess_env.get('SCRIPT_FILENAME')# This works
>   return apache.OK
> # -=-=-=-=-=-=
> The strange thing is that the .get() access works fine: only the subscript
 hangs?
> If anyone is wondering about a use-case, I don't actually have one. I was just ite
rating over the contents of the request.subprocess_env using a for loop (code below)
, and found that the code hung when accessing 'SCRIPT_FILENAME', and not for any oth
er
variable.
> # -=-=-=
> request.add_common_vars()
> d = {}
> for sek in request.subprocess_env.keys():
>   d[sek] = request.subprocess_env[sek]
> # -=-=-=

[ Post a follow-up to this message ]