Re: Design hole in nptl (and more general in POSIX threads)

Jamie Lokier <[email protected]> writes:
> It is never really portable anyway.  You can't safely call
> pthread_kill from a signal handler either, so a thread can't halt
> itself inside a signal handler.  Also, the only thread-related
> async-signal-safe operation, sem_drop, is an optional part of pthreads :)

I'm not sure what you mean here.

The code I posted never calls pthread_kill from a signal handler.  And
it never has a thread send itself a signal.  The signal handler makes
only two function calls: one to sem_post, to tell the collector it is
waiting, and one to sigsuspend, to wait for the collector's go-ahead.
See handle_pause_signal.

(I'm not saying this untested, never-compiled code doesn't have
serious flaws.  But it's intended to have well-defined behavior under
POSIX, so if I've botched something, I really want to know.)

(Roland pointed out my brain-dead use of assert; that's fixed in the
version at http://svn.red-bean.com/repos/minor/jimb/wip/gc/threads.c).