Python-netcdf crashes in Feisty

Output from running test script:

*** glibc detected *** python: free(): invalid pointer: 0xb7de0de0 ***
======= Backtrace: =========
/lib/tls/i686/cmov/libc.so.6[0xb7eb27cd]
/lib/tls/i686/cmov/libc.so.6(cfree+0x90)[0xb7eb5e30]
/usr/lib/python2.5/site-packages/Scientific/linux2/Scientific_netcdf.so[0xb7d82efe]
python[0x8085269]
python(PyDict_SetItem+0x6e)[0x8086bae]
python(_PyModule_Clear+0x158)[0x8088808]
python(PyImport_Cleanup+0x153)[0x80dcc63]
python(Py_Finalize+0xbf)[0x80e8e5f]
python(Py_Main+0x4d1)[0x8058e41]
python(main+0x22)[0x8058862]
/lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xdc)[0xb7e60ebc]
python[0x80587b1]
======= Memory map: ========
08048000-0813f000 r-xp 00000000 08:07 655424 /usr/bin/python2.5
0813f000-08164000 rw-p 000f6000 08:07 655424 /usr/bin/python2.5
08164000-081dd000 rw-p 08164000 00:00 0 [heap]
b7c00000-b7c21000 rw-p b7c00000 00:00 0
b7c21000-b7d00000 ---p b7c21000 00:00 0
b7d23000-b7d2e000 r-xp 00000000 08:05 737554 /lib/libgcc_s.so.1
b7d2e000-b7d2f000 rw-p 0000a000 08:05 737554 /lib/libgcc_s.so.1
b7d2f000-b7d43000 r-xp 00000000 08:07 278720 /usr/lib/python2.5/site-packages/Numeric/_numpy.so
b7d43000-b7d45000 rw-p 00013000 08:07 278720 /usr/lib/python2.5/site-packages/Numeric/_numpy.so
b7d45000-b7d7b000 r-xp 00000000 08:07 132087 /usr/lib/libnetcdf.so.3.6.1
b7d7b000-b7d7c000 rw-p 00035000 08:07 132087 /usr/lib/libnetcdf.so.3.6.1
b7d7c000-b7d7d000 rw-p b7d7c000 00:00 0
b7d7d000-b7d85000 r-xp 00000000 08:07 278713 /usr/lib/python2.5/site-packages/Scientific/linux2/Scientific_netcdf.so
b7d85000-b7d86000 rw-p 00007000 08:07 278713 /usr/lib/python2.5/site-packages/Scientific/linux2/Scientific_netcdf.so
b7d86000-b7d8d000 r--s 00000000 08:07 131864 /usr/lib/gconv/gconv-modules.cache
b7d8d000-b7dc8000 r--p 00000000 08:07 180227 /usr/lib/locale/en_US.utf8/LC_CTYPE
b7dc8000-b7e4b000 rw-p b7dc8000 00:00 0
b7e4b000-b7f86000 r-xp 00000000 08:05 807011 /lib/tls/i686/cmov/libc-2.5.so
b7f86000-b7f87000 r--p 0013b000 08:05 807011 /lib/tls/i686/cmov/libc-2.5.so
b7f87000-b7f89000 rw-p 0013c000 08:05 807011 /lib/tls/i686/cmov/libc-2.5.so
b7f89000-b7f8c000 rw-p b7f89000 00:00 0
b7f8c000-b7fb1000 r-xp 00000000 08:05 807062 /lib/tls/i686/cmov/libm-2.5.so
b7fb1000-b7fb3000 rw-p 00024000 08:05 807062 /lib/tls/i686/cmov/libm-2.5.so
b7fb3000-b7fb5000 r-xp 00000000 08:05 807085 /lib/tls/i686/cmov/libutil-2.5.so
b7fb5000-b7fb7000 rw-p 00001000 08:05 807085 /lib/tls/i686/cmov/libutil-2.5.so
b7fb7000-b7fb8000 rw-p b7fb7000 00:00 0
b7fb8000-b7fba000 r-xp 00000000 08:05 807060 /lib/tls/i686/cmov/libdl-2.5.so
b7fba000-b7fbc000 rw-p 00001000 08:05 807060 /lib/tls/i686/cmov/libdl-2.5.so
b7fbc000-b7fcf000 r-xp 00000000 08:05 807080 /lib/tls/i686/cmov/libpthread-2.5.so
b7fcf000-b7fd1000 rw-p 00013000 08:05 807080 /lib/tls/i686/cmov/libpthread-2.5.so
b7fd1000-b7fd3000 rw-p b7fd1000 00:00 0
b7fe2000-b7fe4000 rw-p b7fe2000 00:00 0
b7fe4000-b7ffd000 r-xp 00000000 08:05 737590 /lib/ld-2.5.so
b7ffd000-b7fff000 rw-p 00019000 08:05 737590 /lib/ld-2.5.so
bfb2c000-bfb42000 rw-p bfb2c000 00:00 0 [stack]
ffffe000-fffff000 r-xp 00000000 ...

Python-Scientific NetCDF used PyMem_DEL instead of PyObject_DEL to deallocate memory of an object created with PyObject_NEW.
Python 2.4 used to catch that error and automatically correct it. Python 2.5 has removed this behaviour.

I ask you to test this patch and to let me know if it solved the problem.
Ciao

On 5/13/07, eolo999 <email address hidden> wrote:
> Python-Scientific NetCDF used PyMem_DEL instead of PyObject_DEL to deallocate memory of an object created with PyObject_NEW.
> Python 2.4 used to catch that error and automatically correct it. Python 2.5 has removed this behaviour.
>
> I ask you to test this patch and to let me know if it solved the problem.

Thank you! Certainly I will. Could you please point me at a good
description of how this is done with debian-style packages? (I know
how to prepare RPMs from sources, but this Ubuntu box is my first
serious attempt at maintaining a Debian-based distribution)

> Ciao

Best regards,

--
Tom Grydeland
  <Tom.Grydeland@(phys.uit.no|gmail.com)>

First Debdiff :)

Ok this hopefully is a valid debdiff.

Tom,

MOTU is the place to start here. Lots of work and lots of people willing to help you learn how to do it. See https://wiki.ubuntu.com/MOTU. Mostly it's done via the #ubuntu-motu IRC channel.

Uploaded for gutsy. Please keep an eye on it. Thank you for your contribution!

Tested (with good help from Scott K) on feisty, and fixes the above test cases!

Thanks for your great help!

Made a feisty-proposed version and tested that it corrects the problem. Uploaded and waiting for the archive admins to publish it for testing.

I've unsubscribed ubuntu-universe-sponsors, as it appears that no further sponsoring is required.

A testing version of python-netcdf has landed in feisty-proposed. To test,
please add:

deb http://archive.ubuntu.com/ubuntu/ feisty-proposed universe

to your sources.list(5). Then, update and upgrade (or install) python-netcdf.

Please provide feedback here. We need two reports of works for me before we can publish it to feisty-updates.

Running the following script should no longer crash:

#!/usr/bin/python
from Scientific.IO.NetCDF import *
fname = 'crash.nc'
ncfile = NetCDFFile(fname, 'w', 'Created now by me')

python-scientific (2.4.11-1build1.1) feisty-proposed; urgency=low

  * NetCDF file causing invalid pointer error (LP: #113803)
  * Changed maintainer to MOTU
  * Added python-central (>= 0.5) to the build-depends field in debian/control

 -- Scott Kitterman <email address hidden> Sat, 07 Jul 2007 00:41:51 -0400

The update in -proposed works.

The update also works for me wth the script at bug 123949. Thanks.

Copied to feisty-updates.