Re: [Boost-users] [posix_time][date_time][string_algo] Another I/O error?

From: Jeff Garland (jeff_at_[hidden])
Date: 2006-06-03 18:37:01

Bogdan M. wrote:
> Hi Jeff,
> Thanks for your prompt answer. What I meant was to surround "" with 
> something like:
> 
> string_type ("")
> 
> where string_type is a template-based type defined in the surrounding 
> time_facet class.
> 
> Nevertheless, I tested a little bit more the code and it turned out that the 
> crash is due to ss.str() in the same line (time_facet.hpp:308).
> Changing the code to :
> 
> 
>           boost::algorithm::replace_all(format,
>                                         string_type (ss.str()), // <<<<<<<< 
> use explicitly the standard string
>                                         "");
> 
> seems to do the job.
> 
> What worries me is that the original/official version of the code looks 
> correct. I checked with some folks using the
> same deployment(e.g., boost 1.33.1 & Visual Studio 2005) and they have the 
> same problem with the sample I posted.

VS 2005 was released after 1.33.1 although we were regression testing on 
beta versions.

> I am wondering whether we made a mistake while installing this particular 
> version of boost or whether there is indeed something
> unusual with the way the compiler interprets the code in time_facet.hpp.

It's coming back to me now -- stupid me for relying on my memory -- 
completely forgotten even though it was just a few months.  This problem 
was reported right after vc8 was released.  We concluded that this was a 
serious problem in vc8 standard library -- someone reported it to MS. 
The following mails from the list might be helpful.

http://lists.boost.org/Archives/boost/2006/02/101042.php
http://lists.boost.org/Archives/boost/2006/02/101109.php
http://lists.boost.org/Archives/boost/2006/02/101122.php

Long story short is that if you pickup a few files from the current CVS 
head (see last link) things should be mostly better.  Note, I'm NOT 
going to guarantee that it is 100% -- as I recall there were a couple 
code paths I could not easily fix. Please complain to your vendor for a 
bug fix -- the problem in their library is serious and needs to be fixed.

> Would it be possible for you to see whether the program I posted works in 
> your Visual Studio 2005 setup?

Nope -- I don't develop date-time on Windows.

Jeff