[serialization][test] Last few serialization regressions with VC6
Looking through the serialization regressions, there are a few spurious ones, then these errors http://tinyurl.com/y3qpqf seem to be caused by Boost.Test, I have two questions: 1) How do I run the serialization tests with bbv2 and an earlier version of Spirit? There's code in the bbv1 Jamfile, but not the v2 version? 2) Is there a particular reason why the problem template (equal_impl_frwd in test_tools.hpp) is marked as exported, when it contains only inline functions? I realise it shouldn't make any difference either which way, and if it's used by classes that do need to be exported then it suppresses a whole lot of warnings so I guess that's the reason ? :-) Thanks, John.
2) Is there a particular reason why the problem template (equal_impl_frwd in test_tools.hpp) is marked as exported, when it contains only inline functions? I realise it shouldn't make any difference either which way, and if it's used by classes that do need to be exported then it suppresses a whole lot of warnings so I guess that's the reason ? :-)
You probably right, but I would prefer not to touch this, since I don't have resources at the moment to test it. Gennadiy
Gennadiy Rozental wrote:
2) Is there a particular reason why the problem template (equal_impl_frwd in test_tools.hpp) is marked as exported, when it contains only inline functions? I realise it shouldn't make any difference either which way, and if it's used by classes that do need to be exported then it suppresses a whole lot of warnings so I guess that's the reason ? :-)
You probably right, but I would prefer not to touch this, since I don't have resources at the moment to test it.
Gennadiy, I've now tested this change with VC 6, 7, 7.1, 8 and Intel 9.1 and there are no regressions in Boost.Test. I've also confirmed that it fixes the serialisation failures. The change I made was simply: Index: boost/test/test_tools.hpp =================================================================== RCS file: /cvsroot/boost/boost/boost/test/test_tools.hpp,v retrieving revision 1.60.2.1 diff -r1.60.2.1 test_tools.hpp 485c485 < struct BOOST_TEST_DECL equal_impl_frwd { ---
struct /*BOOST_TEST_DECL*/ equal_impl_frwd {
Is this acceptable to commit? Thanks, John.
Gennadiy Rozental wrote:
2) Is there a particular reason why the problem template (equal_impl_frwd in test_tools.hpp) is marked as exported, when it contains only inline functions? I realize it shouldn't make any difference either which way, and if it's used by classes that do need to be exported then it suppresses a whole lot of warnings so I guess that's the reason ? :-)
You probably right, but I would prefer not to touch this, since I don't have resources at the moment to test it.
Gennadiy, I've now tested this change with VC 6, 7, 7.1, 8 and Intel 9.1 and there are no regressions in Boost.Test. I've also confirmed that it fixes the serialization failures.
The change I made was simply:
Index: boost/test/test_tools.hpp =================================================================== RCS file: /cvsroot/boost/boost/boost/test/test_tools.hpp,v retrieving revision 1.60.2.1 diff -r1.60.2.1 test_tools.hpp 485c485 < struct BOOST_TEST_DECL equal_impl_frwd { ---
struct /*BOOST_TEST_DECL*/ equal_impl_frwd {
Is this acceptable to commit?
Thanks, John.
Go ahead. But I wouldn't be able to fix issues, if any. Gennadiy
participants (3)
-
Gennadiy Rozental -
gennadiy.rozental＠thomson.com
-
John Maddock