[PATCH libXi 1/5] Handle unknown device classes.

[Daniel Stone]

On Mon, Aug 22, 2011 at 03:35:18PM +1000, Peter Hutterer wrote:
> @@ -1564,13 +1576,13 @@ wireToDeviceChangedEvent(xXIDeviceChangedEvent *in, XGenericEventCookie *cookie)
>      out->deviceid = in->deviceid;
>      out->sourceid = in->sourceid;
>      out->reason = in->reason;
> -    out->num_classes = in->num_classes;
>  
>      out->classes = (XIAnyClassInfo**)&out[1];
>  
>      info.classes = out->classes;
>  
> -    copy_classes(&info, (xXIAnyInfo*)&in[1], in->num_classes);
> +    copy_classes(&info, (xXIAnyInfo*)&in[1], &nclasses);
> +    out->num_classes = nclasses;

I like this change.

> diff --git a/src/XIQueryDevice.c b/src/XIQueryDevice.c
> index 4e2f392..c22aee1 100644
> --- a/src/XIQueryDevice.c
> +++ b/src/XIQueryDevice.c
> @@ -87,7 +87,7 @@ XIQueryDevice(Display *dpy, int deviceid, int *ndevices_return)
>          ptr += ((wire->name_len + 3)/4) * 4;
>  
>          lib->classes = Xmalloc(size_classes((xXIAnyInfo*)ptr, lib->num_classes));
> -        ptr += copy_classes(lib, (xXIAnyInfo*)ptr, lib->num_classes);
> +        ptr += copy_classes(lib, (xXIAnyInfo*)ptr, &lib->num_classes);
>      }

... and think you should probably make the same one here, for clarity.
I noted that while reading XIQueryDevice to make sure it was safe and
only spotted the same change in wireToDeviceChangedEvent when I hit
reply to suggest it, so clearly if both of us think the same way, it
must be true, right? :)

Cheers,
Daniel