David Cournapeau schrieb:
>  Hi,
>  
>      I recently developed small wrappers around some C libraries, which 
>  worked find under python2.4 + ctypes 1.0.1, but failed with python 2.5.
>      After investigation, the problem is that for some reason, 
>  c_longlong, which was 8 bytes under python2.4, is 4 under python2.5. 
>  Also, importing c_int64 fails (not found error) under python2.5 (I 
>  replaced c_longlong with c_int64 since, the latter being a much easier 
>  to find error). All this with the dapper ubuntu package.
>      The error is seemed to be because ubuntu python2.5 packages use a 
>  packaged libffi4 instead of the libffi furnished by ctypes, and as such, 
>  python2.5 is configured with the option --with-system-ffi.

I can reproduce this bug :-(.

>      My questions are:
>         - is this intended for c_longlong to be only 4 bytes in some 
>  cases ? (in ISO C99, my understanding was that long long was guaranteed 
>  to be at least 8 bytes)

Of course not.  c_longlong should have the same size as the struct module's
'q' specifier.

>         - is the lack of c_int64 of bug of ctypes, libffi, both ?
>         - I try to understand where the problem lies, but I don't have 
>  much knowledge on python C Api, so I am a bit lost in the code of ctypes

I've seen that you already answered that yourself in the bug you
submitted to ubuntu - looks like a libffi bug.

I will add some testcases to ctypes so that this problem will at
least be seen in the ctypes unittests.

Thanks,
Thomas


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
ctypes-users mailing list
ctypes-users@[...].net
https://lists.sourceforge.net/lists/listinfo/ctypes-users