Open ysnt27 opened 5 years ago
I've got the same bug with python 3.7 and boost-1.69:
#0 dict_dealloc (mp=0x7fffeab5c328) at Objects/dictobject.c:1901
#1 0x00007fffe9a2aa2f in boost::python::api::object_base::~object_base (this=0x7fffea9b56a0 <PyExecuter::main_namespace>, __in_chrg=<optimized out>)
at /home/artelys/repository/artelys/COE_BUILD/_dep/install/include/boost/python/object_core.hpp:423
#2 0x00007fffe9a2a8fa in boost::python::api::object::~object (this=0x7fffea9b56a0 <PyExecuter::main_namespace>, __in_chrg=<optimized out>)
at /home/artelys/repository/artelys/COE_BUILD/_dep/install/include/boost/python/object_core.hpp:238
#3 0x00007ffff7c68ccc in __run_exit_handlers () from /lib64/libc.so.6
#4 0x00007ffff7c68e00 in exit () from /lib64/libc.so.6
#5 0x00000000004dc995 in Py_Exit (sts=<optimized out>, sts@entry=0) at Python/pylifecycle.c:2282
#6 0x00000000004e45a2 in handle_system_exit () at Python/pythonrun.c:636
#7 0x00000000004e46ee in handle_system_exit () at Python/pythonrun.c:646
#8 PyErr_PrintEx (set_sys_last_vars=set_sys_last_vars@entry=1) at Python/pythonrun.c:646
#9 0x00000000004e4a5a in PyErr_Print () at Python/pythonrun.c:542
#10 0x00000000004e54ac in PyRun_InteractiveLoopFlags (fp=fp@entry=0x7ffff7deea20 <_IO_2_1_stdin_>, filename_str=filename_str@entry=0x57e94e "<stdin>",
flags=flags@entry=0x7fffffffce0c) at Python/pythonrun.c:134
#11 0x00000000004e5bee in PyRun_AnyFileExFlags (fp=fp@entry=0x7ffff7deea20 <_IO_2_1_stdin_>, filename=0x57e94e "<stdin>", closeit=closeit@entry=0,
flags=flags@entry=0x7fffffffce0c) at Python/pythonrun.c:78
#12 0x0000000000430ae4 in pymain_run_file (p_cf=0x7fffffffce0c, filename=0x0, fp=0x7ffff7deea20 <_IO_2_1_stdin_>) at Modules/main.c:427
#13 pymain_run_filename (cf=0x7fffffffce0c, pymain=0x7fffffffcee0) at Modules/main.c:1627
#14 pymain_run_python (pymain=0x7fffffffcee0) at Modules/main.c:2876
#15 pymain_main (pymain=pymain@entry=0x7fffffffcee0) at Modules/main.c:3037
#16 0x0000000000430ce7 in _Py_UnixMain (argc=<optimized out>, argv=<optimized out>) at Modules/main.c:3072
#17 0x00007ffff7c52413 in __libc_start_main () from /lib64/libc.so.6
#18 0x000000000042bd6e in _start () at Objects/dictobject.c:818
Python is compiled from sources (3.7.2) on Fedora 29 and boost is compiled from sources too. I tested with boost from github repo and the gdb trace is the one above.
I couldn't reproduce the problem with python-3.7.4 and boost-1.71.0 anymore. But from source code below, I got another segmentation fault around list_base destructor.
#include <boost/python.hpp>
#include <boost/python/str.hpp>
using namespace boost::python;
class teststr3 {
public:
static list aaa_list;
static const str aaa1;
static const str aaa2;
teststr3()
{
aaa_list.append(aaa1);
aaa_list.append(aaa2);
}
str get(int i) {
if (i==1) {
return aaa1;
} else {
return aaa2;
}
}
};
list teststr3::aaa_list;
const str teststr3::aaa1("1 AAA BBB CCC");
const str teststr3::aaa2("2 AAA BBB CCC");
BOOST_PYTHON_MODULE(teststr3) {
using namespace boost::python;
class_<teststr3>("teststr3")
.def("get", &teststr3::get);
}
/*
g++ -g -DPIC -shared -fpic -o teststr3.so teststr3.cc -L/usr/local/boost-1_71_0-py374/lib -lboost_python37 -I/usr/local/python374/include/python3.7dm -I/usr/local/boost-1_71_0-py374/include -Wl,-rpath,/usr/local/boost-1_71_0-py374/lib
*/
% gdb /usr/local/python374/bin/python3.7dm
GNU gdb (GDB) Red Hat Enterprise Linux 7.6.1-115.el7
Copyright (C) 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-redhat-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /usr/local/python374/bin/python3.7dm...done.
(gdb) run
Starting program: /usr/local/python374/bin/python3.7dm
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
Python 3.7.4 (default, Oct 7 2019, 08:57:43)
[GCC 4.8.5 20150623 (Red Hat 4.8.5-39)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import teststr3
>>> exit()
Program received signal SIGSEGV, Segmentation fault.
0x000000000044602a in list_dealloc (op=op@entry=0x7ffff7f4b360)
at ../Objects/listobject.c:316
316 Py_TRASHCAN_SAFE_BEGIN(op)
Missing separate debuginfos, use: debuginfo-install libgcc-4.8.5-39.el7.x86_64 libstdc++-4.8.5-39.el7.x86_64
(gdb) where
#0 0x000000000044602a in list_dealloc (op=op@entry=0x7ffff7f4b360)
at ../Objects/listobject.c:316
#1 0x0000000000466452 in _Py_Dealloc (op=0x7ffff7f4b360)
at ../Objects/object.c:1971
#2 0x00007ffff05151f3 in boost::python::api::object_base::~object_base (
this=0x7ffff071ed00 <teststr3::aaa_list>, __in_chrg=<optimized out>)
at /usr/local/boost-1_71_0-py374/include/boost/python/object_core.hpp:423
#3 0x00007ffff0515088 in boost::python::api::object::~object (
this=0x7ffff071ed00 <teststr3::aaa_list>, __in_chrg=<optimized out>)
at /usr/local/boost-1_71_0-py374/include/boost/python/object_core.hpp:238
#4 0x00007ffff0515290 in boost::python::detail::list_base::~list_base (
this=0x7ffff071ed00 <teststr3::aaa_list>, __in_chrg=<optimized out>)
at /usr/local/boost-1_71_0-py374/include/boost/python/list.hpp:18
#5 0x00007ffff051536a in boost::python::list::~list (
this=0x7ffff071ed00 <teststr3::aaa_list>, __in_chrg=<optimized out>)
at /usr/local/boost-1_71_0-py374/include/boost/python/list.hpp:57
#6 0x00007ffff6eeac99 in __run_exit_handlers (status=status@entry=0,
listp=0x7ffff72786c8 <__exit_funcs>,
run_list_atexit=run_list_atexit@entry=true) at exit.c:77
#7 0x00007ffff6eeace7 in __GI_exit (status=status@entry=0) at exit.c:99
#8 0x0000000000517904 in Py_Exit (sts=sts@entry=0)
at ../Python/pylifecycle.c:2292
#9 0x000000000051fb14 in handle_system_exit () at ../Python/pythonrun.c:636
---Type <return> to continue, or q <return> to quit---
#10 0x0000000000520da8 in PyErr_PrintEx (
set_sys_last_vars=set_sys_last_vars@entry=1) at ../Python/pythonrun.c:646
#11 0x000000000052117f in PyErr_Print () at ../Python/pythonrun.c:542
#12 0x0000000000521d3a in PyRun_InteractiveLoopFlags (
fp=fp@entry=0x7ffff7279640 <_IO_2_1_stdin_>,
filename_str=filename_str@entry=0x5e55b2 "<stdin>",
flags=flags@entry=0x7fffffffe2d0) at ../Python/pythonrun.c:134
#13 0x00000000005223f8 in PyRun_AnyFileExFlags (
fp=fp@entry=0x7ffff7279640 <_IO_2_1_stdin_>, filename=0x5e55b2 "<stdin>",
closeit=closeit@entry=0, flags=flags@entry=0x7fffffffe2d0)
at ../Python/pythonrun.c:78
#14 0x0000000000422023 in pymain_run_file (fp=0x7ffff7279640 <_IO_2_1_stdin_>,
filename=0x0, p_cf=p_cf@entry=0x7fffffffe2d0) at ../Modules/main.c:427
#15 0x00000000004220e5 in pymain_run_filename (
pymain=pymain@entry=0x7fffffffe300, cf=cf@entry=0x7fffffffe2d0)
at ../Modules/main.c:1606
#16 0x00000000004221d8 in pymain_run_python (pymain=pymain@entry=0x7fffffffe300)
at ../Modules/main.c:2867
#17 0x000000000042550f in pymain_main (pymain=pymain@entry=0x7fffffffe300)
at ../Modules/main.c:3028
#18 0x0000000000425589 in _Py_UnixMain (argc=<optimized out>,
argv=<optimized out>) at ../Modules/main.c:3063
#19 0x000000000041f116 in main (argc=<optimized out>, argv=<optimized out>)
---Type <return> to continue, or q <return> to quit---
at ../Programs/python.c:15
(gdb)
With python-3.7.5(debug build) and GDB, I tried step execution.
Python-3.7.5/Objects/listobject.c
static void
list_dealloc(PyListObject *op)
{
Py_ssize_t i;
PyObject_GC_UnTrack(op);
Py_TRASHCAN_SAFE_BEGIN(op) // <--- SIGSEGV here!!
The macro definition is
#define Py_TRASHCAN_SAFE_BEGIN(op) \
do { \
PyThreadState *_tstate = PyThreadState_GET(); \
if (_tstate->trash_delete_nesting < PyTrash_UNWIND_LEVEL) { \
++_tstate->trash_delete_nesting;
/* The body of the deallocator is here. */
_tstate
was NULL.
So, _tstate->trash_delete_nesting
resulted in segmentation fault.
A patch on python which has worked for me on Python-3.6.7 (works also for other versions, but you need to adapt it):
$ diff dictobject.c-3.6.7 /tmp/Python-3.6.7/Objects/dictobject.c
2009c2009
< Py_ssize_t i, n, refcnt = mp->ob_base.ob_refcnt;
---
> Py_ssize_t i, n;
2013,2015c2013
< if (refcnt != 0) {
< Py_TRASHCAN_SAFE_BEGIN(mp)
< }
---
> Py_TRASHCAN_SAFE_BEGIN(mp)
2033,2035c2031
< if (refcnt != 0) {
< Py_TRASHCAN_SAFE_END(mp)
< }
---
> Py_TRASHCAN_SAFE_END(mp)
This is a workaround and not a fix ...
If possible, I need a workaround at the side of boost.python, not Python.
Though, Python's issue, https://bugs.python.org/issue17703 seems related.
I opened an issue at Python Bug Tracker, https://bugs.python.org/issue38609
On Python 3.6, when boost::python::str constructor is used at global scope, I receive "Bus error" at end of python process.
Python 3.4 and 3.5 worked fine, no errors.
My environment
x86_64
boost-1_69_0
rh-python36-python-3.6.3-3.el7.x86_64
and AnacondaSource code
Compile
Run
Prepare test python script.
with gdb