Re: [PATCH] libxl: unconst the event argument to the event_occurs hook.

[Jim Fehlig <jfehlig@suse.com>]

Ian Campbell wrote:
>>> diff --git a/tools/libxl/libxl.h b/tools/libxl/libxl.h
>>> index 25efa76..ef96bce 100644
>>> --- a/tools/libxl/libxl.h
>>> +++ b/tools/libxl/libxl.h
>>> @@ -273,9 +273,9 @@
>>>  #include <libxl_uuid.h>
>>>  #include <_libxl_list.h>
>>>  
>>> -/* API compatibility. Only 0x040200 is supported at this time. */
>>> +/* API compatibility. */
>>>  #ifdef LIBXL_API_VERSION
>>> -#if LIBXL_API_VERSION != 0x040200
>>> +#if LIBXL_API_VERSION != 0x040200 && LIBXL_API_VERSION != 0x040300
>>>  #error Unknown LIBXL_API_VERSION
>>>  #endif
>>>  #endif
>>>   
>>>       
>> Should this hunk be in a separate patch? It seems to be introducing a
>> new API version :).
>>     
>
> I suppose it could be, or we could argue that this patch is introducing
> the first user of this API version?
>   

Ok.

>   
>>> @@ -308,6 +308,16 @@
>>>   */
>>>  #define LIBXL_HAVE_DEVICE_BACKEND_DOMNAME 1
>>>  
>>> +/*
>>> + * LIBXL_HAVE_NONCONST_EVENT_OCCURS_EVENT_ARG
>>> + *
>>> + * This argument was erroneously "const" in the 4.2 release despite
>>> + * the requirement for the callback to free the event.
>>> + */
>>> +#if LIBXL_API_VERSION != 0x040200
>>>       
>
> Looking at this again today this happens to be true if LIBXL_API_VERSION
> is undefined but I wonder if we ought to explicitly define
> LIBXL_API_VERSION to 0xffffff if the user doesn't supply it?
>   

Ah, good point.  But setting LIBXL_API_VERSION to 0xffffffff if the app
doesn't supply it means the app, which e.g. built fine on 4.2, would no
longer compile on 4.3 right?

Regards,
Jim