On 10/12/2012 05:15 PM, Guido Günther wrote:
> This fixes problems on platforms where sizeof(long) != sizeof(long long)
> like ia32.
>
> ---
> Cheers,
> -- Guido
>
> python/generator.py | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/python/generator.py b/python/generator.py
> index a98a894..1ef76e1 100755
> --- a/python/generator.py
> +++ b/python/generator.py
> @@ -259,8 +259,8 @@ py_types = {
> 'double': ('d', None, "double",
"double"),
> 'unsigned int': ('i', None, "int",
"int"),
> 'unsigned long': ('l', None, "long",
"long"),
> - 'long long': ('l', None, "longlong", "long
long"),
> - 'unsigned long long': ('l', None, "longlong",
"long long"),
> + 'long long': ('L', None, "longlong", "long
long"),
> + 'unsigned long long': ('K', None, "longlong",
"long long"),
I had to look this up:
http://docs.python.org/c-api/arg.html
Why did you choose 'K' instead of 'L' for the unsigned version? Either
we should also be using 'I' for 'unsigned int' for similar lack of
overflow checking, or this should be 'L'. And since I think we want
overflow checking, I'd rather see 'unsigned long long' changed to
'L'.
ACK with that fix.
Hmm...unsigned seemt to be a better fit but 'L' is more consistent with
the current usage. Pushed with that change.
Thanks,
-- Guido
--
Eric Blake eblake(a)redhat.com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org