Looks like this has been fixed already - I can confirm the leak with 1.37, but cannot reproduce it with 1.40: 1.37: ~$ valgrind --tool=memcheck --leak-check=full --show-reachable=yes ./a.out ... ==13406== 8 bytes in 1 blocks are still reachable in loss record 1 of 2 ==13406== at 0x43F6405: malloc (vg_replace_malloc.c:149) ==13406== by 0x4456BFE: boost::detail::get_once_per_thread_epoch() (in /usr/local-1.1.9/lib/libboost_thread-gcc41-mt-p-1_37.so.1.37.0) ==13406== by 0x4454CF6: void boost::call_once(boost::once_flag&, void (*)()) (in /usr/local-1.1.9/lib/libboost_thread-gcc41-mt-p-1_37.so.1.37.0) ==13406== by 0x44510F9: boost::detail::get_current_thread_data() (in /usr/local-1.1.9/lib/libboost_thread-gcc41-mt-p-1_37.so.1.37.0) ==13406== by 0x4452CD5: boost::thread::join() (in /usr/local-1.1.9/lib/libboost_thread-gcc41-mt-p-1_37.so.1.37.0) ==13406== by 0x804909A: main (boo_th.cpp:13) ==13406== 1.40: ... ==30689== ==30689== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 5 from 1) ==30689== malloc/free: in use at exit: 0 bytes in 0 blocks. ==30689== malloc/free: 5 allocs, 5 frees, 648 bytes allocated. ==30689== For counts of detected errors, rerun with: -v ==30689== All heap blocks were freed -- no leaks are possible. -- Nikolai

Hi all, Valgrind reports follows memory leak. ==24567== at 0x4A1FE1B: malloc (vg_replace_malloc.c:207) ==24567== by 0x6AC7FDD: boost::detail::get_once_per_thread_epoch() (in /app/local/pkgs/boost/lib/libboost_thread-gcc41-mt-1_39.so.1.39.0) ==24567== by 0x6AC4A17: void boost::call_once(boost::once_flag&, void (*)()) (in /app/local/pkgs/boost/lib/libboost_thread-gcc41-mt-1_39.so.1.39.0) ==24567== by 0x6AC1D96: boost::detail::get_current_thread_data() (in /app/local/pkgs/boost/lib/libboost_thread-gcc41-mt-1_39.so.1.39.0) ==24567== by 0x6AC36B4: boost::thread::join() (in /app/local/pkgs/boost/lib/libboost_thread-gcc41-mt-1_39.so.1.39.0)

Code is looks like:

thread_->join(); delete thread_;