Hello, As part of my work on thread destruction, I noticed that some servers, all diskfs translators, exit from the main thread after initialization. But POSIX mandates that exiting from the main thread makes the whole process exit too. My intention is to first fix the Hurd servers involved and then fix libpthread. Ideally, we could reuse the main thread as part of the thread pool, but this would mean changing the interface of diskfs_startup_diskfs (at least), and I didn't take the time to check the effects on tmpfs (which makes the main thread perform some more initialization after diskfs_startup_diskfs returns).
In the mean time, I was thinking adding simple blocking calls, replacing pthread_exit with poll. Is anybody uncomfortable with that hack ? -- Richard Braun