joshuaboelter said:
Mike Wilson provided me with a concise example the reproduces the problem on a wider variety of systems. I uploaded slightly modified version of that source to the group files. I'm ready to blame it on the OS :) If anyone else has any insight, it's appreciated.
Sorry I've been so long in responding. I'm a tad swamped right now. I did look at your original code, and though there were a few things I thought might be tweaked, nothing was glaringly wrong. You may well have found an error in the library, but I need a lot more time to evaluate this. The boost::condition is the most complex implementation on Win32 of all of the Boost.Threads code. It's a very tricky thing to emulate this concept correctly, unfortunately. I don't think I'd be too quick to blame the problem on the OS, since the OS doesn't have conditions. It's much more likely that the problem is either in the test code, or in the Boost.Threads library. Give me a few days to evaluate this one thoroughly. Oh, and thanks for the other example as well. I haven't looked at it yet, but having two examples is likely to make it easier to track down the issue. William E. Kempf