[utility] [config] value_init tests on msvc-14.0, 14.1
Currently, value_init_workaround_test.cpp fails on both msvc-14.0 and 14.1, because in: #ifdef BOOST_DETAIL_VALUE_INIT_WORKAROUND_SUGGESTED // One or more failures are expected. return num_failures > 0 ? EXIT_SUCCESS : EXIT_FAILURE; #else // No failures are expected. return num_failures == 0 ? EXIT_SUCCESS : EXIT_FAILURE; #endif failures are supposedly expected (BOOST_DETAIL_VALUE_INIT_WORKAROUND_SUGGESTED is defined), but there are none. BOOST_DETAIL_VALUE_INIT_WORKAROUND_SUGGESTED is controlled by BOOST_NO_COMPLETE_VALUE_INITIALIZATION, which is defined on 14.0/14.1. The Config macro is defined because - presumably - msvc 14.0/14.1 fail the Config test for value initialization, boost_no_com_value_init.ipp. But this test is much more comprehensive than the value_init one. So to fix the value_init_workaround_test.cpp failure, we need to at minimum move one failing test case from the Config test to the value_init one.
But this test is much more comprehensive than the value_init one. So to fix the value_init_workaround_test.cpp failure, we need to at minimum move one failing test case from the Config test to the value_init one.
OK, done. This is what MSVC still gets wrong: https://github.com/boostorg/utility/commit/51ba9f1b4568dd0c22bef3c9820da48f9...
Hi Peter, Do you know if all of these failures have been reported to Microsoft? Thanks, --Beman On Sun, Sep 24, 2017 at 5:50 AM, Peter Dimov via Boost < boost@lists.boost.org> wrote:
But this test is much more comprehensive than the value_init one. So to
fix the value_init_workaround_test.cpp failure, we need to at minimum move one failing test case from the Config test to the value_init one.
OK, done. This is what MSVC still gets wrong:
https://github.com/boostorg/utility/commit/51ba9f1b4568dd0c2 2bef3c9820da48f9d6a3731
_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman /listinfo.cgi/boost
Beman Dawes wrote:
Hi Peter,
Do you know if all of these failures have been reported to Microsoft?
The comments in boost/config/compiler/visualc.hpp say, among other things, // Reported again by John Maddock in 2015 for VC14: // https://connect.microsoft.com/VisualStudio/feedback/details/1582233/c-subobj... and that bug report seems to contain all the problematic cases, although the only one that 2015/2017 still get wrong is the member data pointer one. ... On Sun, Sep 24, 2017 at 5:50 AM, Peter Dimov via Boost < boost@lists.boost.org> wrote:
But this test is much more comprehensive than the value_init one. So to
fix the value_init_workaround_test.cpp failure, we need to at minimum move one failing test case from the Config test to the value_init one.
OK, done. This is what MSVC still gets wrong:
https://github.com/boostorg/utility/commit/51ba9f1b4568dd0c2 2bef3c9820da48f9d6a3731
_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman /listinfo.cgi/boost
_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
participants (2)
-
Beman Dawes
-
Peter Dimov