make pthread_exit run dtors for last thread, wait to decrement thread count
diff --git a/src/thread/pthread_create.c b/src/thread/pthread_create.c
index b6c475d..7c43695 100644
--- a/src/thread/pthread_create.c
+++ b/src/thread/pthread_create.c
@@ -10,9 +10,6 @@
 	self = pthread_self();
 	if (self->cancel) self->result = PTHREAD_CANCELLED;
 
-	if (!a_fetch_add(&libc.threads_minus_1, -1))
-		exit(0);
-
 	LOCK(&self->exitlock);
 
 	not_finished = self->tsd_used;
@@ -28,6 +25,9 @@
 		}
 	}
 
+	if (!a_fetch_add(&libc.threads_minus_1, -1))
+		exit(0);
+
 	if (self->detached && self->map_base)
 		__unmapself(self->map_base, self->map_size);