use of missing ::abort

[viboes]

The following tester seems to don't have ::abort

Test output: VeecoFTC - thread - test_futures / msvc-9.0~wm5~stlport5.2
Rev 76100 / Thu, 22 Dec 2011 05:40:08 +0000
Report Time: Mon, 2 Jan 2012 07:46:13 +0000
Compile [2011-12-23 13:43:44 UTC]: fail
 
    call "%BOOST_ROOT%/TestTools/VSVars32-VC9-WM5.bat" x86 >nul
cl /Zm800 -nologo @"D:\BoostRT\results\boost\bin.v2\libs\thread\test\test_futures.test\msvc-9.0~wm5~stlport5.2\debug\stdlib-stlport-5.2~evc9~arm\test_futures.obj.rsp"
 
test_futures.cpp
D:\BoostRT\boost\boost/exception/detail/exception_ptr.hpp(460) : error C2039: 'abort' : is not a member of 'stlpd_std'
D:\BoostRT\boost\boost/exception/detail/exception_ptr.hpp(460) : error C3861: 'abort': identifier not found

Is there something that can be done to work on this tester?

[ulrich.eckhardt@…]

Same problem here, using Boost 1.48. The target platform is MS Windows CE 6 with STLport 5.2 and MSVC 8.0 (VS 2005). There is a workaround in a different place already that simply checks

  #if defined(UNDER_CE)
    // some CE platforms don't define ::abort(), let alone std::abort()
    exit(-1);
  #else
    abort();
  #endif

I have found other places that use abort() to e.g. mark code as unreachable. They could use similar workarounds. I'm just in the process of migrating to 1.48 but I should be able to produce a patch for this.

See also bug #5213.

[viboes]

Any new on this?

[Ulrich Eckhardt <ulrich.eckhardt@…>]

Sorry, I've been distracted by different things, but it's still on my TODO list...

[Emil Dotchevski]

Please check if trunk revision 77593 fixes the problem.

[Ulrich Eckhardt <ulrich.eckhardt@…>]

No, r77593 doesn't help. Neither CE5 nor CE6 have an abort() function anywhere.

BTW: If you commit with "(refs #6349)" in the commit message, TRAC automatically makes a note in the according bug ticket, so you don't have to add one yourself. All subscribers of the ticket also get an automatic notification.

[serggzz@…]

As I see this bug steel exists, ​https://svn.boost.org/trac/boost/search?q=std%3A%3Aabort&noquickjump=1&ticket=on

• Does anybody know why assert.hpp is so often included after config.hpp? I'm intended to use BOOST_USER_CONFIG to figure this problem out but now it's impossible. Examples are:

<boost/ptr_container/ptr_vector.hpp> => <boost/ptr_container/ptr_sequence_adapter.hpp> => <boost/ptr_container/detail/reversible_ptr_container.hpp> => <boost/ptr_container/detail/throw_exception.hpp> => #include <boost/assert.hpp> #include <boost/config.hpp>
<boost/scoped_ptr.hpp> => <boost/smart_ptr/scoped_ptr.hpp> => the first include is #include <boost/assert.hpp> I think that <boost/config.hpp> must be the first include every time.

• It's only my curiosity why there are no any guards in assert.hpp and it's intended?

[Emil Dotchevski]

assert.hpp probably follows the design of assert.h, which lacks include guards so that it can be included multiple times within the same source file, with different combination of _DEBUG/NDEBUG defined.

[David Deakins]

Hi Emil,

Since this issue with abort is disrupting a lot of test result reporting for Windows CE, do you mind if I apply the attached patch for this issue or would you like to do it yourself?

Thanks, -Dave

[David Deakins]

Committed the patch. Will let the tests cycle and confirm that the issue is resolved.