[dpdk-dev] [PATCH v2] librte_eal/common: Fix cast from pointer to integer of different size

[Qiu, Michael]

On 3/4/2015 4:19 PM, Wodkowski, PawelX wrote:
> On 2015-03-04 02:58, Qiu, Michael wrote:
>> On 3/3/2015 9:38 PM, Wodkowski, PawelX wrote:
>>>> -----Original Message-----
>>>> From: Qiu, Michael
>>>> Sent: Tuesday, March 03, 2015 11:00 AM
>>>> To: Wodkowski, PawelX; dev at dpdk.org
>>>> Subject: Re: [dpdk-dev] [PATCH v2] librte_eal/common: Fix cast from pointer to
>>>> integer of different size
>>>>
>>>> On 3/3/2015 4:25 PM, Wodkowski, PawelX wrote:
>>>>> On 2015-03-03 03:20, Michael Qiu wrote:
>>>>>> /i686-native-linuxapp-gcc/include/rte_memcpy.h:592:23: error:
>>>>>> cast from pointer to integer of different size
>>>>>> [-Werror=pointer-to-int-cast]
>>>>>>
>>>>>>     dstofss = 16 - (int)((long long)(void *)dst & 0x0F) + 16;
>>>>>>
>>>>>> Type 'long long' is 64-bit in i686 platform while 'void *'
>>>>>> is 32-bit.
>>>>>>
>>>>>> Signed-off-by: Michael Qiu <michael.qiu at intel.com>
>>>>>> ---
>>>>>> v2 --> v1:
>>>>>> 	Remove unnecessary casting (void *)
>>>>>>
>>>>>>    lib/librte_eal/common/include/arch/x86/rte_memcpy.h | 4 ++--
>>>>>>    1 file changed, 2 insertions(+), 2 deletions(-)
>>>>>>
>>>>>> diff --git a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
>>>> b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
>>>>>> index 7b2d382..85a5f4d 100644
>>>>>> --- a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
>>>>>> +++ b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
>>>>>> @@ -589,12 +589,12 @@ COPY_BLOCK_64_BACK15:
>>>>>>    	 * unaligned copy functions require up to 15 bytes
>>>>>>    	 * backwards access.
>>>>>>    	 */
>>>>>> -	dstofss = 16 - (int)((long long)(void *)dst & 0x0F) + 16;
>>>>>> +	dstofss = 16 - (int)((long)dst & 0x0F) + 16;
>>>>>>    	n -= dstofss;
>>>>>>    	rte_mov32((uint8_t *)dst, (const uint8_t *)src);
>>>>>>    	src = (const uint8_t *)src + dstofss;
>>>>>>    	dst = (uint8_t *)dst + dstofss;
>>>>>> -	srcofs = (int)((long long)(const void *)src & 0x0F);
>>>>>> +	srcofs = (int)((long)src & 0x0F);
>>>>>>
>>>>>>    	/**
>>>>>>    	 * For aligned copy
>>>>>>
>>>>> I think you should use here size_t, (u)ptrdiff_t or (u)intptr_t as this
>>>> Yes, but 'long' is enough, does it limit anything, or has any issue with
>>>> multiple platforms?
>>>>
>>> Those types ((u)ptrdiff_t, (u)intptr_t) exists specially for
>>> pointer-to-int and int-to-pointer casts. Using integer primitives might
>>> produce further warnings/error in the future and need further patches
>>> fixing the same place.
>> OK, I will send out the v3 patch.
>>
>>> Also why make offset variables signed and different type? This introduce
>>> a lot of unnecessary explicit and implicit casts or type promotions.
>> But Is it suitable to make offset (u)ptrdiff_t or (u)intptr_t?
>>
> I think, as final result is offset, its type should be size_t (the same 
> type as type of offsetof() macro). This way you can use 
> uptrdiff_t/uintptr_t and does not need of any signed-unsigned casting.

size_t should be acceptable

Thanks,
Michael

>
>> Thanks,
>> Michael
>>
>>>>> will be more portable.
>>>>> Also type of dstofss and srcofs should be changed accordingly.
>>>> No, I think, it should be offset.
>>>>
>>>> Thanks,
>>>> Michael
>>
>