I tested 3.17-2 with sudo kill -PIPE and saw that it didn't die, whereas nis from sarge did die. Looking at the source, though, I think Mark Brown just applied my first patch. man 2 clone suggests that the signal handling thread will be ignoring SIGPIPE, inheriting that setting from the main thread:
Since Linux 2.5.35, flags must also include CLONE_SIGHAND if CLONE_THREAD is specified. So that's why it works. It does leave the code in the slightly misleading state of having a case for catching a SIGPIPE signal which will never be caught. No big deal. ------------------------------------- Martin's Outlook, BlueArc Engineering