Re: [boost] [1.33.0][date_time] Yr 1900 issue on x64 Win

From: Jeff Garland (jeff_at_[hidden])
Date: 2005-12-06 21:50:28

On Tue, 6 Dec 2005 15:55:15 -0500, Caleb Epstein wrote
> On 12/6/05, Norian N <noriann_at_[hidden]> wrote:
> >
> >   I'm trying to use Boost 1.33 libs on x64 Windows platform. Have assert
> > break when try to use date below year 1900.
> >
> > _VALIDATE_RETURN( ( timeptr->tm_year >=0 ), EINVAL, FALSE)
> 
> This is an issue with the Microsoft runtime libraries and not the 
> fault of Boost.DateTime.
> 
> There is absolutely no good reason for this check (IMHO).  There is nothing
> in the strftime code (aside from these debugging assertions) that 
> would fail on years prior to 1900, yet these assertions exist.  I 
> really wish they would remove them.

I have to agree with this.  This is a result of recent MS changes that now
break code that works on many other platforms and previously worked on MS
platforms.  I suppose someone ought to submit formal bug reports to complain.
 Only problem is AFAIK there is no C/C++ specification that says whether
negative values are allowed, so perhaps we need a LWG bug report to change the
standard.  The whole reason we need to do this is so that we can send a tm
value into the standard facet for formatting.  This makes working around the
problem in a generic way very difficult. So unfortunately, you might need to
consider using to_string or writing your own printing function if you are
using this particular platform.  I suppose we are going to have to add more
portability notes as well....sigh....

Jeff