[boost] Re: Re: [serialization] MSVC8 warning flood (widened size_t)

From: Bo Persson (bop_at_[hidden])
Date: 2005-01-17 14:31:26

"Victor A. Wagner Jr." <vawjr_at_[hidden]> skrev i meddelandet 
news:6.2.0.14.2.20050117084004.041befd8_at_mail.rudbek.com...
> At Monday 2005-01-17 02:19, you wrote:
>>Sorry, wrong example. Should have been
>>
>>cout << strlen("");
>>
>>or
>>
>>cout << vector<int>().size();
>>
>>which both give me warnings with /Wp64
>>
>>warning C4267: 'argument' : conversion from 'size_t' to 'unsigned 
>>int',
>>possible loss of data
>
>
> IMO this falls directly in Dinkumware's lap.  ALL of the code was 
> written by them (well the headers).
> It's fairly clear that there should be no restrictions on outputting 
> the return from .size()

And there isn't.

The compiler chooses one of the gazillion overloads of operator<< for a 
32 bit value, and then informs us that it wouldn't be the proper one to 
use for a 64 bit value. Duh!

The feature is just broken. Why not use the proper 64 bit compiler for 
the 64 bit tests?

Bo Persson