Re: [Qemu-devel] [PATCH] bswap: Don't rely on HOST_LONG_BITS

[Andreas Färber]

Am 30.01.2013 23:03, schrieb Andreas Färber:
> Am 30.01.2013 22:59, schrieb Richard Henderson:
>> On 01/30/2013 01:04 PM, Stefan Weil wrote:
>>>>   static inline unsigned long leul_to_cpu(unsigned long v)
>>>>   {
>>>> -    return le_bswap(v, HOST_LONG_BITS);
>>>> +    /* In order to break an include loop between here and
>>>> +       qemu-common.h, don't rely on HOST_LONG_BITS.  */
>>>> +#if ULONG_MAX == UINT32_MAX
>>>> +    return le_bswap(v, 32);
>>>> +#elif ULONG_MAX == UINT64_MAX
>>>> +    return le_bswap(v, 64);
>>>> +#else
>>>> +# error Unknown sizeof long
>>>> +#endif
>>>>   }
>>>>
>>>>   #undef le_bswap
>>>
>>> That would be wrong for 64 bit MinGW-w64 because
>>> HOST_LONG_BITS is _not_ the bit size of a long value.
>>>
>>> See qemu-common.h for the correct definition.
>>
>> I beg your pardon, but it *is* right.  HOST_LONG_BITS is set to the size
>> of a host pointer, not a host long.  Which suggests that my patch is the
>> *only* correct way to do this, at least for now.
> 
> IIUC on w64 ULONG_MAX == UINT32_MAX but 64 is desired for pointer.
> w64 is not C99-compliant.

Oh I see now: function is named ul and operates on unsigned long. ;)
So your patch carries an implicit bugfix that should be mentioned in the
commit message.

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg