Anthony Williams wrote:
That code is there to support move semantics for thread. It compiles OK on Borland 5.8.2.
This should be a direct constructor call. Any ideas for a Borland 5.6.4 workaround?
Yep, with this patch it builds OK, and the regex test passes again: Index: C:/data/boost/boost/trunk/libs/thread/src/win32/thread.cpp =================================================================== --- C:/data/boost/boost/trunk/libs/thread/src/win32/thread.cpp (revision 40407) +++ C:/data/boost/boost/trunk/libs/thread/src/win32/thread.cpp (working copy) @@ -184,12 +184,14 @@ thread::operator boost::move_t<thread>() { - return boost::move_t<thread>(*this); + boost::move_t<thread> m(*this); + return m; } boost::move_t<thread> thread::move() { - return boost::move_t<thread>(*this); + boost::move_t<thread> m(*this); + return m; } void thread::swap(thread& x) Running through the thread regression tests I still see quite a few failures though: some are dependencies on the string library that are failing, the timeout failures look more worrying though. Logfile attached. HTH, John.