From my point of view, it seems more flexible to not catch exceptions in

Hi folks. Apologies if this has been discussed frequently before. I'm puzzled as to why the boost.thread thread_proxy function catches all exceptions and then quietly ends the thread. For example:- class MyThreadFunctor { public: void operator()() const { // Do some processing... throw very_bad_exception(); // ...never get to do some processing. } }; int main(int argc, char * argv[]) { MyThreadFunctor my_functor; boost::thread my_thread(my_functor); // Do lots of processing and then join to my_thread. // ... } In boost 1.32, the exception that is thrown in the created thread is caught in the library's thread_proxy function and then the thread exits quietly. Is there a compelling reason for this behaviour? the thread_proxy function. If I wanted to, I could still get the current behaviour by putting a try/catch-all block in the operator() function. I would have expected the normal C++ uncaught exception actions (i.e. calling terminate()) to occur if an exception is thrown out of the user provided thread code. I noticed this when (oh horror!) an error in my code threw an exception out of operator(). I could have identified where the problem was from the core file or Dr Watson log that would have been created if the exception had not been caught by thread_proxy(). Am I missing something obvious here? Thanks, Adrian.