Re: [Boost-users] [posix_time] truncation warning from VC71

From: Jeff Garland (jeff_at_[hidden])
Date: 2006-01-17 20:46:42

On Tue, 17 Jan 2006 12:08:06 -0600, Dale Wilson wrote 
> Hi,  (possible duplicate, but I think the original was rejected 'cause of a 
> mail alias.  If not, please forgive the noise.) 
> 
> I'm using boost 1.32 posix_time in some code that is intended to be portable
> (MS VC7.1 / Solaris gcc 3.3.1 / RHLinux gcc 3.3.1). 
> 
> When I build with VC 7.1.3088 on Windows XP Pro SP2 get  warnings about 
>integer trunction (see below). 
> 
> I can duplicate it by compiling a cpp file containing a single line: 
> 
> #include "boost/date_time/posix_time/posix_time.hpp" 
> 
> and I can work around it by: 
> 
> #ifdef _MSC_VER 
> #pragma warning(push)          
> #pragma warning(disable:4244) // boost/posix_time has truncation warnings 
> #endif // _MSC_VER 
> #include "boost/date_time/posix_time/posix_time.hpp" //include all types
plus i/o 
> #ifdef _MSC_VER 
> #pragma warning(pop) 
> #endif // _MSC_VER 
> 
> but this might hide real problems (or this might *BE* a real problem -- I 
>haven't really looked into the difference between a 
>boost::date_time::counted_time_rep<config>::int_type. and a 
>boost::date_time::gregorian_calendar_base<ymd_type_,date_int_type_>::date_int_type

> 
> Suggestions?

We've looked at and addressed a number of these warnings in 1.33 and non of
them would ever create a run-time problem.  They tended to be places where the
number of months was extracted into a short, but b/c of context it was assured
that the value would be within the range of a short -- but the compiler
couldn't really deduce that and hence the warnings.  Bottom line,  I expect
your warning disable will be safe.

Jeff