Re: feature request for setting _pack_ to a higher value

Axel Seibert wrote, On 03/21/2010 09:56 PM:
> Dear list!
>
> My quest continues... During this weekend, I learned a lot about C 
> data structures, more than I ever wanted to. After all, that's why I'm 
> using Python.

Yeah. ctypes can't hide the details and the simplistic complexity of C 
and the CPU, so ctypes is not a magic silver bullet.

If you couldn't do it in C then you probably can't do it in ctypes. The 
advantage of ctypes is that you don't have to actually do it in C and 
you can avoid a lot of trivial and error-prone C code ;-) And if you can 
do it with ctypes then you probably could have done it in other ways as 
well - but ctypes is often more convenient.

> It's surprising to learn that the alignment under Linux (32bit) is 
> different from Windows (32bit) and Solaris (32bit).
>
> Basically, now I have a feature request: The library I'm wrapping is 
> available under different plattforms (Windows, Linux, Solaris, AIX and 
> HP/UX). As this library contains the code to talk to a server across 
> the network, it probably tries to keep alignment across all plattforms 
> the same. At least, under Linux, I had to change the alignment of many 
> structs.

AFAIK C structs are - for this and other reasons - very seldom 
transferred directly over the network. This library might however do it 
anyway.

> When I tried to do this with the class variable _pack_, I could not 
> get the desired result; up to a value of 4 I saw a difference, but not 
> beyond. Then I found the following statement from Thomas:
>
> > To force an alignment that is larger than the native alignment, you
> > must use padding.
>
> But that's not really an option. The supporting header files define 
> 293 struct's; in 44 cases, I had to change the alignment from 4 to 8 
> -- but ctypes does not allow for this.
>

Can you give an (or several) exact example of a problematic C 
declaration and what you would have hoped worked with ctypes and how you 
have to workaround?

A reference to the Thomas quote might help the discussion too.

/Mads

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev