Re: [boost] [filesystem.v3] v1.49 Path constructor issue in VS Debug configuration

Subject: Re: [boost] [filesystem.v3] v1.49 Path constructor issue in VS Debug configuration
From: Fraser Hutchison (fraser.hutchison_at_[hidden])
Date: 2012-03-11 10:28:09

On 11/03/2012 12:17, Lars Viklund wrote:
> On Sun, Mar 11, 2012 at 10:46:36AM +0000, Fraser Hutchison wrote:
>> Hello,
>>
>> The changes made to filesystem/v3/src/path.cpp between v1.48 and v1.49
>> seem to be causing a crash if I initialise a path as e.g. a global
>> variable or as a static class member, but only for a Debug build.  (I'm
>> using VS10 on 64-bit Win 7).  The following program exhibits this
>> behaviour:
>>
>> |#include "boost/filesystem/v3/path.hpp"
>> boost::filesystem::path p("p");
>> int main() { return 0; }
>> |
>> In Debug, I can see that the third path constructor calls |codecvt()|,
>> which calls |wchar_t_codecvt_facet()|, which returns |codecvt_facet|.
>> However, |codecvt_facet| hasn't been initialised at this point.  This
>> appears to only be the case for Debug builds though.
> I've reproduced this after release but not had the time to jot together
> a bug report yet, just some ranting on IRC.
>
>> I'm unclear as to whether this is an MSVC bug though.  I thought that
>> the Standard required |codecvt_facet| to have been initialised before
>> |wchar_t_codecvt_facet()| since it comes earlier in the same translation
>> unit.
>>
>> Could someone please clarify whether I should file a bug report with
>> boost please?
> Please do, and you could always make a Connect bug against MSVC. It
> should be possible to make a non-Boost standalone test case with use of
> codecvt() in the dynamic initialization phase. Poking STL (the person)
> might also be beneficial, as this probably falls under the standard
> library.
>

Looks like it's already in: https://svn.boost.org/trac/boost/ticket/6638 
- I've just added a comment to that bug.

If I get a chance, I'll submit a Connect bug also.

Thanks,

Fraser.