[Python-bugs-list] [ python-Bugs-416943 ] 2.1 segfaults on pydoc HTTP
noreply@sourceforge.net
noreply@sourceforge.net
Fri, 11 May 2001 14:28:58 -0700
Bugs item #416943, was updated on 2001-04-17 20:58
You can respond by visiting:
http://sourceforge.net/tracker/?func=detail&atid=105470&aid=416943&group_id=5470
Category: Python Interpreter Core
Group: None
Status: Open
Resolution: None
Priority: 7
Submitted By: Andy Dustman (adustman)
Assigned to: Ka-Ping Yee (ping)
Summary: 2.1 segfaults on pydoc HTTP
Initial Comment:
Platform info: 2.1 final, Red Hat 7.1 beta for i386.
Run pydoc -p xxx. Pull up info on a C extension module.
Now
recompile that module so that it changes in a
significant way.
Try to pull it up again and segfault occurs.
#0 0x0808db62 in PyString_FromString (str=0x0) at
Objects/stringobject.c:89
84 }
85
86 PyObject *
87 PyString_FromString(const char *str)
88 {
89 register size_t size = strlen(str);
90 register PyStringObject *op;
91 if (size > INT_MAX) {
92 PyErr_SetString(PyExc_OverflowError,
93 "string is too long for a Python string");
(gdb) frame 1
#1 0x0808d003 in PyObject_GetAttrString (v=0x8164ec0,
name=0x80c888c "__name__") at Objects/object.c:1002
1002 return (*v->ob_type->tp_getattr)(v, name);
(gdb) list
997 v->ob_type->tp_name,
998 name);
999 return NULL;
1000 }
1001 else {
1002 return (*v->ob_type->tp_getattr)(v, name);
1003 }
1004 }
1005
1006 int
(gdb) print (*v->ob_type->tp_getattr)
$1 = {PyObject *()} 0x808b3c8 <meth_getattr>
$6 = {ob_refcnt = 3, ob_type = 0x80bb920, ob_size = 0,
tp_name = 0x80aa94e "builtin_function_or_method",
tp_basicsize = 16,
tp_itemsize = 0, tp_dealloc = 0x808b390
<meth_dealloc>, tp_print = 0,
tp_getattr = 0x808b3c8 <meth_getattr>, tp_setattr =
0,
tp_compare = 0x808b550 <meth_compare>, tp_repr =
0x808b4ec <meth_repr>,
tp_as_number = 0x0, tp_as_sequence = 0x0,
tp_as_mapping = 0x0,
tp_hash = 0x808b5a4 <meth_hash>, tp_call = 0, tp_str
= 0, tp_getattro = 0,
tp_setattro = 0, tp_as_buffer = 0x0, tp_flags = 0,
tp_doc = 0x0,
tp_traverse = 0, tp_clear = 0, tp_richcompare = 0,
tp_weaklistoffset = 0}
(gdb) print ((PyCFunctionObject *)v)->m_ml
$7 = (PyMethodDef *) 0x401c7ecc
(gdb) print
$8 = 0x0
(gdb) list meth_getattr
67 free_list = m;
68 }
69
70 static PyObject *
71 meth_getattr(PyCFunctionObject *m, char *name)
72 {
73 if (strcmp(name, "__name__") == 0) {
74 return PyString_FromString(m->m_ml->ml_name);
75 }
76 if (strcmp(name, "__doc__") == 0) {
((PyCFunctionObject *)v)->m_ml->ml_name ==
m->m_ml->ml_name == NULL. From what I can tell, all
members of v->m_ml are NULL. Note that name is getting
smashed somewhere, probably after the call at
Objects/object.c:1002.
Feel free to contact me for further information if
necessary. Complete backtrace attached.
----------------------------------------------------------------------
>Comment By: Andy Dustman (adustman)
Date: 2001-05-11 14:28
Message:
Logged In: YES
user_id=71372
I know what the problem is now. I was reinstalling packages
using Distutils, and when Distutils copies files, it
overwrites any existing file. Needless to say, when it
overwrites a dynamic library, bad things happen, if
something is using it.
Suggested solution: Distutils should write to a temporary
file and then rename this to the final destination name. On
UNIX, the original file will be unlinked, but not deleted
until the last reader closes it. Not sure about other
platforms, or how hard filename restrictions will make this.
----------------------------------------------------------------------
Comment By: Andy Dustman (adustman)
Date: 2001-04-26 12:05
Message:
Logged In: YES
user_id=71372
...and just to clarify, I don't think this is a pydoc bug.
It seems to be related to Python internals. The module used
in this case is MySQLdb
(http://sourceforge.net/projects/mysql-python). It could be
related to that module, but I really don't think so.
----------------------------------------------------------------------
You can respond by visiting:
http://sourceforge.net/tracker/?func=detail&atid=105470&aid=416943&group_id=5470