Re: [Boost-bugs] [Boost C++ Libraries] #3436: Code miscompilation with MSVC 7.1

Subject: Re: [Boost-bugs] [Boost C++ Libraries] #3436: Code miscompilation with MSVC 7.1
From: Boost C++ Libraries (noreply_at_[hidden])
Date: 2009-11-23 06:45:49

#3436: Code miscompilation with MSVC 7.1
---------------------------+------------------------------------------------
  Reporter:  andysem       |       Owner:  rogeeff 
      Type:  Bugs          |      Status:  reopened
 Milestone:  Boost 1.41.0  |   Component:  test    
   Version:  Boost 1.40.0  |    Severity:  Problem 
Resolution:                |    Keywords:  msvc    
---------------------------+------------------------------------------------
Changes (by andysem):

  * status:  closed => reopened
  * resolution:  worksforme =>

Comment:

 > Works fine for me. In fact I believe shared library build is part of
 routine regression tests.

 I'm not sure Boost.Test is build as a shared library with my options in
 the regular testing process.

 > Are you sure it's not something specific to your setup? Do you have an
 example which reproduces this?

 I have described my setup in ticket #3435. I still have problems with 1.41
 Boost release (after the patch from #3435 is applied), although the test
 now simply ends with the "Boost.Test framework internal error: unknown
 reason" message instead of a crash. I don't have much time to dig deeper
 now, but I suspect that the root of the problem is still the same.

 Here is my test case:

 {{{
 #define BOOST_ALL_DYN_LINK
 #define BOOST_TEST_MAIN

 #include <boost/test/unit_test.hpp>
 #include <boost/test/test_tools.hpp>
 #include <boost/test/unit_test_suite.hpp>

 BOOST_AUTO_TEST_CASE(MyTest)
 {
     BOOST_CHECK(true);
 }
 }}}

 I compile it with the following commad line:

 {{{
 cl -Ox -I ../STLport/stlport -I . -EHsc -MD ./test.cpp
 }}}

 Where . is in the boost 1.41 directory. However, I believe, pretty much
 any test case
 is affected.

 > Also what to you mean by speedup startup? How long does it take for you?

 It's not something I measured or intentionally wanted to speedup. It was
 just a side effect of the proposed patch. We have several thousands of
 tests here, many of them are using Boost.Test. If each particular test
 starts a bit faster, it has an observable effect on our testing as a
 whole. But since without the patch none of the tests work, I don't have
 anything to compare with.

-- 
Ticket URL: <https://svn.boost.org/trac/boost/ticket/3436#comment:2>
Boost C++ Libraries <http://www.boost.org/>
Boost provides free peer-reviewed portable C++ source libraries.