Fix bug in srun when using --no-shell option if not running under debugger

 (was causing segmentation fault).
Verify that a connection remains before trying to send a return code.

NEWS

@@ -8,6 +8,8 @@ documents those changes that are of interest to users and admins.
  -- Create new allocation as needed for debugger in case old allocation 
     has been purged
  -- Add Blue Gene User Guide to html documents
+ -- Fix srun bug that could cause seg fault with --no-shell option if not 
+    running under a debugger
 
 * Changes in SLURM 0.4.0-pre2
 =============================

slurm.spec.in

@@ -115,13 +115,13 @@ DESTDIR="$RPM_BUILD_ROOT" make install
 install -D -m755 etc/init.d.slurm $RPM_BUILD_ROOT/etc/rc.d/init.d/slurm
 install -D -m644 etc/slurm.conf.example $RPM_BUILD_ROOT/etc/slurm/slurm.conf
 # Delete unpackaged files:
-rm -f $RPM_BUILD_ROOT/usr/lib/slurm/*.{a,la}
+rm -f $RPM_BUILD_ROOT/%{_libdir}/slurm/*.{a,la}
 
 # Build file lists for optional plugin packages
 for plugin in auth_munge auth_authd switch_elan; do
    LIST=./${plugin}.files
    touch $LIST
-   test -f $RPM_BUILD_ROOT/usr/lib/slurm/${plugin}.so &&
+   test -f $RPM_BUILD_ROOT/%{_libdir}/slurm/${plugin}.so &&
      echo %{_libdir}/slurm/${plugin}.so > $LIST
 done
 

src/common/slurm_protocol_api.c

@@ -930,6 +930,9 @@ int slurm_send_rc_msg(slurm_msg_t *msg, int rc)
         slurm_msg_t resp_msg;
         return_code_msg_t rc_msg;
 
+        if (msg->conn_fd < 0)
+                return (ENOTCONN);
+
         rc_msg.return_code = rc;
 
         resp_msg.address  = msg->address;

src/srun/launch.c

@@ -222,8 +222,12 @@ static void _join_attached_threads (int nthreads, thd_t *th)
 {
 	int i;
 	void *retval;
-	for (i = 0; i < nthreads; i++)
-		pthread_join (th[i].thread, &retval);
+	if (!opt.parallel_debug)
+		return;
+	for (i = 0; i < nthreads; i++) {
+		if (th[i].thread != (pthread_t) NULL)
+			pthread_join (th[i].thread, &retval);
+	}
 	return;
 }
 
@@ -294,6 +298,7 @@ static void _p_launch(slurm_msg_t *req, job_t *job)
 		if (job->ntask[i] == 0)	{	/* No tasks for this node */
 			debug("Node %s is unused",job->host[i]);
 			job->host_state[i] = SRUN_HOST_REPLIED;
+			thd[i].thread = (pthread_t) NULL;
 			continue;
 		}