[prev in list] [next in list] [prev in thread] [next in thread] 

List:       boost
Subject:    Re: [boost] Boost Filesystem v3 - requires wstring - an obstacle to testing the serialization librar
From:       Edward Diener <eldiener () tropicsoft ! com>
Date:       2017-01-21 2:07:44
Message-ID: o5ufpe$dh6$1 () blaine ! gmane ! org
[Download RAW message or body]

On 11/18/2016 12:55 AM, Robert Ramey wrote:
> On 11/17/16 9:27 PM, Thomas Trummer wrote:
>>
>>> On 17 Nov 2016, at 20:45, John Maddock <jz.maddock@googlemail.com>
>>> wrote:
>>>
>>> I'm pretty sure I've posted these before, but here are the
>>> serialization build/linker errors I see locally:
>>
>> For completeness I'll also post the reason for this problem again:
>>
>> The problem is that "BOOST_SERIALIZATION_DECL" expands to
>> "__attribute__((__dllimport__))" in "singleton.hpp" when it is
>> included from files that don't define "BOOST_SERIALIZATION_SOURCE"
>> e.g. "binary_iarchive.cpp".
>
> How is that a problem?  Isn't that what's it's supposed to do?  Note
> that this isn't failing for any other windows platforms.
>
> After I posted that message, I found that it ony applied to mngw 3.4.
>
> Other versions of mngw (4.x) don't display errors because the test
> matrix truncates the error messages. Or the library fails to build and
> the test matrix doesn't display any error messages at all.
>
> The test matrix "platform"
> http://www.boost.org/development/tests/develop/igaztanaga-develop-gcc-6-1c++14.html
> doesn't reveal whether this is a build of a shared or static library. So
> it's impossible to see what's happening here.
>
> Don't get the impression that I don't appreciate the extra input, but I
> really need a definitive solution in the test system itself.

I have fixed this problem using mingw(-64)/gcc on Windows with PR 
https://github.com/boostorg/serialization/pull/51. There is still a 
problem with the wide character serialization library build with 
mingw(-64)/gcc on Windows revolving around visibility and exporting of 
codecvt_null<wchar_t> functionality. The latter problem does not occur 
with gcc on Linux.

>
> Robert Ramey
>
>
>
>
>>
>> Thomas



_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
[prev in list] [next in list] [prev in thread] [next in thread]