On Sat, Aug 3, 2013 at 8:42 AM, Vicente J. Botet Escriba < vicente.botet@wanadoo.fr> wrote:
there were an issue fixed on 1.54 https://svn.boost.org/trac/** boost/ticket/8136 https://svn.boost.org/trac/boost/ticket/8136.
Please, could you tell me if this fix is related?
It might be related to this fix mixed with another fix from another dependency. I use Ogre3D with TBB, where Ogre3D will look for TBB using a CMake script. There was a bug I reported[1] recently to the Ogre3D team because in specific conditions (TBB from sources+32bit) where the CMake script was adding add_definitions(-D_WIN32_WINNT=0x0501) which hid some windows functions use by Ogre code. I fixed it by adding a condition so that it does it only with MinGW (I'm using exclusively Visual Studio 2012 U3 right now). Now, I don't know if it's related, but this change looks like it might have an impact on the fix you're talking about. As I was saying, I have a friend who can reproduce the bug systematically (windows Vista, core 2 duo) so maybe finding a computer with these characteristics might make the issue find it easily. When I reimplemented the loop using condition variables - which was actually better because now the thread wait as long as the time necessary to spawn the next task and is woken up when a new task is pushed in the task priority queue - solves the problem for my friend, so there might be a strong difference in waiting code there that might be related? I'll try to make a repro-case with my friend tomorrow if it helps. [1] http://www.ogre3d.org/forums/viewtopic.php?f=4&t=78443 Joel Lamotte