> Did you really mean to say that you are forcing a GC when atomic state drops > to 0 and a signal is pending? I don't get why that would be good. Not a GC, just a call to GC_gc, which will switch to the signal handler thread. A bit confusing, I admit.