[SciPy-dev] scipy cow and proc.py

[Pearu Peterson]

On Thu, 13 May 2004, Virgilio, Vincent wrote:

> > It turns out that proc.py is not portable even within linux systems.
> > For example, /proc/meminfo may have different headings when using
> > different kernels. As a result, mem_info, for one instance,
> > is failing with Linux 2.6 kernel. So, proc.py needs some work..

> Well I'm not surprised. I don't know how to reliably track the /proc
> filesystem in Linux. Which, anyway, leaves Windows out in the cold.

I think any differences in /proc are kernel specific, we probably need to 
support only 2.4 and 2.6 kernels.

> In fact, I don't know anything about how Python handles such portability
> issues. In Cygwin (Windows, of course), there is the glue layer in
> cygwin1.dll, which provides a canonical POSIX interface. Might there be
> something similar in the Python core release, which can be imitated and
> extended in the direction of proc.py?

I don't think that Cygwin will be a big problem, its /proc seems to be 
pretty similar to linux /proc (at least /proc/cpuinfo was so).
The problem is win32 port. I am not an expert on that but at least on 
Windows XP it should be possible to retrive process information.

> Eventually, I'll look closely at proc.py, as time permits (is that
> qualified enough?). It'll be my first exposure to such Pythonic things.

proc.py is quite simple module, and your patches are most welcome.

Thanks,
Pearu