Darren Cook wrote: > > It is failing on set_random_seed, which is the first command after > > getting the connection. Specifically it fails on my 3rd unit test - > > It worked (tested by doing my 3 tests 100 times) by adding a usleep(1); > after my fwrite() command and before doing the fgets() call. > > I also now explicitly kill the child pid when I disconnect, but as far > as I can tell that made no difference: ps still shows 300 [gnugo > <defunct>] zombies, but they all disappear when my program closes.
I assume you use wait() or waitpid(), not kill() here, right? Somehow, my `gtp-test' does neither, but I get no zombie processes anyway. > I guess my usleep() is a kludge and the real problem is how I connect or > disconnect. But as multiple connect/disconnects is probably only going > to happen in my unit tests it is good enough. Yes, it is. That usleep() ``fixes'' things indicates that pipes are asynchronous (non-blocking) and you should use select() function. Apparently, I use it too in my `gtp-test'; wrote it long ago, so I forgot about this. Another possible solution might using fcntl() to make the pipe IO blocking, but I don't know if this is possible. Paul _______________________________________________ gnugo-devel mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnugo-devel
