Re: [boost] [chrono][ratio][boost_1_52_0_beta1] warning in user code

Subject: Re: [boost] [chrono][ratio][boost_1_52_0_beta1] warning in user code
From: Jeff Flinn (Jeffrey.Flinn_at_[hidden])
Date: 2012-10-30 14:56:17

On 10/30/2012 11:56 AM, Vicente J. Botet Escriba wrote:
> Le 30/10/12 07:25, Andrey Semashev a écrit :
>> On Tue, Oct 30, 2012 at 3:15 AM, Vicente J. Botet Escriba
>> <vicente.botet_at_[hidden]> wrote:
>>> Le 29/10/12 20:16, Jeff Flinn a écrit :
>>>
>>>> I've just downloaded boost_1_52_0_beta1 and get the warning below in my
>>>> user code on VC11(I'm using thread which evidently brings in the above
>>>> libs). I didn't see mention of this in trac. I'm moving forward from
>>>> 1.47
>>>> and from VC8 so I'm not sure if this is an indication of some project
>>>> setting[s] of mine, or if it should be corrected in the lib?
>>>>
>>>> Thanks, Jeff
>>>>
>>>> 6>D:\boost_1_52_0_beta1\boost/ratio/detail/overflow_helpers.hpp(136):
>>>> warning C4293: '<<' : shift count negative or too big, undefined
>>>> behavior
>>> The concerned code is
>>>
>>>        static const boost::intmax_t nan =
>>>          (BOOST_RATIO_INTMAX_C(1) << (sizeof(boost::intmax_t) *
>>> CHAR_BIT -
>>> 1));
>>>
>>> I don't reach to see why the shift count could be negative or too big.
>> If I'm not mistaken, the shifted (positive) value exceeds the intmax_t
>> value range (which is signed) but fits into uintmax_t. E.g. for 64 bit
>> intmax_t it's 1 << 63 while intmax_t range is [-2^63; 2^63 - 1].
>>
>> _______________________________________________
>> Unsubscribe & other changes:
>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>
> The following program
>
> int main () {
>    std::intmax_t nan= (ULL << (sizeof(intmax_t) * 8 - 1)) ;
>    std::intmax_t max= nan-1 ;
>    std::intmax_t min= -max ;
>    std::cout << std::hex << "max="<< max <<std::endl ;
>    std::cout << std::hex << "min="<< min <<std::endl ;
>    std::cout << std::hex << "nan="<< nan <<std::endl ;
>    std::cout << std::dec << "max="<< max <<std::endl ;
>    std::cout << std::dec << "min="<< min <<std::endl ;
>    std::cout << std::dec << "nan="<< nan <<std::endl ;
> }
>
> prints
>
> max=7fffffffffffffff
> min=8000000000000001
> nan=8000000000000000
> max=9223372036854775807
> min=-9223372036854775807
> nan=-9223372036854775808
>
> This is what I'm expecting.

That's what I get on VS2012 also.

> Maybe to avoid the warning we could replace the nan declaration by
>
>    std::intmax_t nan= (ULL << (sizeof(intmax_t) * 8 - 1)) ;
> by
>    std::intmax_t nan= (std::intmax_t)(1ULL << (sizeof(intmax_t) * 8 - 1)) ;
>
> Note the constant 1ULL instead of 1LL. and of course that the output is
> the same :)
>
> Please could you try it and see if the warning disappears?

Yep, that clears up the warning. Thanks!

static const boost::intmax_t nan = (boost::intmax_t)(1ULL << 
(sizeof(boost::intmax_t) * CHAR_BIT - 1)) ;

The above is what I used locally.

Will this get into the boost 1.52.0 release?

Thanks again, Jeff