[sage-devel] Re: Unsafe use of symbolic links when saving data

2013-04-06 Thread leif
Simon King wrote: Hi! Suppose you write data to orig.sobj, and then you create a symbolic link copy.sobj to orig.sobj. If you now save data under the name copy.sobj, then the data in orig.sobj get overridden: sage: d = tmp_dir() sage: save(1, os.path.join(d, 'orig'))

[sage-devel] Re: Unsafe use of symbolic links when saving data

2013-04-06 Thread Nils Bruin
Is this a bug? $ mkdir a $ ln -s a b $ touch b/t $ ls a t What touch b/t did was open b for writing to put a link to the freshly allocated file underlying t. The fact that opening a symbolic link means opening the thing to which the link is pointing is essential to the nature of symbolic links. N

[sage-devel] Unsafe use of symbolic links when saving data

2013-04-06 Thread Simon King
Hi! Suppose you write data to orig.sobj, and then you create a symbolic link copy.sobj to orig.sobj. If you now save data under the name copy.sobj, then the data in orig.sobj get overridden: sage: d = tmp_dir() sage: save(1, os.path.join(d, 'orig')) sage: os.symlink(os.pat

[sage-devel] Re: Clean up linestyle

2013-04-06 Thread Tobias Weich
Am Donnerstag, 4. April 2013 11:38:41 UTC+2 schrieb P Purkayastha: > I suggest that you introduce a function in sage/plot/misc.py, say, > > def get_matplotlib_linestyle(linestyle, return_type): > if return_type == 'short': > # change linestyle from 'dotted', 'dashed', etc to ':'

Re: [sage-devel] test failures for 5.9.beta0

2013-04-06 Thread Stephen Montgomery-Smith
Just to tell people that the first and second errors listed below no longer occur with 5.9.beta3. As for the third error, I have a trac ticket that fixes it: http://trac.sagemath.org/sage_trac/ticket/14392 On 03/20/2013 10:09 PM, Stephen Montgomery-Smith wrote: > This is under FreeBSD. (There a

[sage-devel] ImportError: libffi.so.6: cannot open shared object file

2013-04-06 Thread Ralf Hemmecke
>> In particular, I ran into the problem (see below): >> >> ERROR:root:code for hash md5 was not found. > Well, I did "sage -i openssl" and then at least I could start sage. But then, I ran into the thing below and gave up. I used ftp://ftp.fu-berlin.de/unix/misc/sage/linux/32bit/sage-5.8-linu

Re: [sage-devel] md5 missing

2013-04-06 Thread Dan Drake
On Sat, 06 Apr 2013 at 01:17PM +0200, Ralf Hemmecke wrote: > Hello, > > I want to install sage on a system that share the sage installation > directory and are all Debian6 machines, but they might have different > packages installed. > > In particular, I ran into the problem (see below): > > E

[sage-devel] Re: Does cython have to be slow for attribute lookup in deep MRO?

2013-04-06 Thread Nils Bruin
Yay! with a little digging around I think I know what feature cython- derived classes are missing that pure python classes do have: It's the Py_TPFLAGS_VALID_VERSION_TAG feature, which gets used in the "method" cache. in _PyType_Lookup (python source: typeobject.c) Evidence obtained with the follo

[sage-devel] md5 missing

2013-04-06 Thread Ralf Hemmecke
Hello, I want to install sage on a system that share the sage installation directory and are all Debian6 machines, but they might have different packages installed. In particular, I ran into the problem (see below): ERROR:root:code for hash md5 was not found. Unfortunately, I haven't found a