Description of problem: *** glibc detected *** /usr/bin/python: free(): invalid pointer: 0x08b84040 *** ======= Backtrace: ========= /lib/libc.so.6[0x95defd] /lib/libc.so.6(cfree+0x90) [0x961550] /usr/lib/python2.5/site-packages/_snackmodule. so[0xdd3846] /usr/lib/libpython2.5.so.1.0[0xcaec02] /usr/lib/libpython2.5.so.1.0[0xc896b1] /usr/lib/libpython2.5.so.1.0[0xcc4d5d] /usr/lib/libpython2.5.so.1. 0[0xca5032] /usr/lib/libpython2.5.so.1.0[0xcaec02] /usr/lib/libpython2.5.so.1. 0[0xc896b1] /usr/lib/libpython2.5.so.1.0(PyEval_EvalFrameEx+0x9e0) [0xcf83a0] /usr/lib/libpython2.5.so.1.0(PyEval_EvalFrameEx+0x63cf)[0xcfdd8f] /usr/lib/lAbgebrochen BTW, "Abgebrochen" means "Canceled". Version-Release number of selected component (if applicable): python25-2.5-0 How reproducible: Running system-config-date in TUI and quitting without any changes. Actual results: Invalid pointer. Expected results: No invalid pointer... ;-)
BTW, same at system-config-keyboard in TUI. IMHO either newt/snack or python is broken.
Can you run with _MALLOC_CHECK_=2 and get a full backtrace from gdb?
Created attachment 139757 [details] Full backtrace Is this what you expected to get by me?
I am providing the requested information for this bug *ticks*
this problem still happens in python-2.5-3.fc7. Starting system-config-display, *** glibc detected *** python2: free(): invalid pointer: 0x09c81b00 *** happens.
Created attachment 143485 [details] another backtrace file pyxf86config-0.3.31-4.fc7
Ping? Nobody interested in fixing this annoying bug?
Sorry, thought I had reassigned to the right component after looking into it.
Hum? Why is this problem re-asssigned to pyxf86config? Any TUI application (system-config-*) seems to be affected. IMHO the problem is somewhere between newt and python, isn't it?
Ping?
It appears that bug 220923 will be another instance, or closely related, to this problem (although that appears to start quite a bit later than a date on the original report from Robert Scheck). Presumably glibc got stricter in catching those things as I see also similar bombs which can be likely be traced back to bugs in fontconfig or Tk (or both playing together?).
This is being caused by snack allocating memory with PyObject_Malloc and then freeing it with PyMem_DEL. In python 2.5, these actually have a differences (so that the interpreter can return memory to the system). Fixed in newt-0.52.4-3.fc7
Yepp, your fix is confirmed to work in newt-0.52.4-3. Is there any reason why you left open this bug report?
Forgot to hit the closed button :)