svn merge -r19937:19948 https://eris.llnl.gov/svn/slurm/branches/slurm-2.1

NEWS

@@ -168,6 +168,9 @@ documents those changes that are of interest to users and admins.
 
 * Changes in SLURM 2.1.7
 ========================
+ -- Modify srun, salloc and sbatch parsing for the --signal option to accept
+    either a signal name in addition to the previously supported signal
+    numbers (e.g. "--signal=USR2@200").
 
 * Changes in SLURM 2.1.6-2
 ==========================

doc/man/man1/salloc.1

@@ -757,8 +757,8 @@ When a job is within \fIsig_time\fR seconds of its end time,
 send it the signal \fIsig_num\fR.
 Due to the resolution of event handling by SLURM, the signal may
 be sent up to 60 seconds earlier than specified.
-Both \fIsig_time\fR and \fIsig_num\fR must have integer values
-between zero and 65535.
+\fIsig_num\fR may either be a signal number or name (e.g. "10" or "USR1").
+\fIsig_time\fR must have integer value between zero and 65535.
 By default, no signal is sent before the job's end time.
 If a \fIsig_num\fR is specified without any \fIsig_time\fR,
 the default time will be 60 seconds.

doc/man/man1/sbatch.1

@@ -841,8 +841,8 @@ When a job is within \fIsig_time\fR seconds of its end time,
 send it the signal \fIsig_num\fR.
 Due to the resolution of event handling by SLURM, the signal may
 be sent up to 60 seconds earlier than specified.
-Both \fIsig_time\fR and \fIsig_num\fR must have integer values
-between zero and 65535.
+\fIsig_num\fR may either be a signal number or name (e.g. "10" or "USR1").
+\fIsig_time\fR must have integer value between zero and 65535.
 By default, no signal is sent before the job's end time.
 If a \fIsig_num\fR is specified without any \fIsig_time\fR,
 the default time will be 60 seconds.

doc/man/man1/srun.1

@@ -953,8 +953,8 @@ When a job is within \fIsig_time\fR seconds of its end time,
 send it the signal \fIsig_num\fR.
 Due to the resolution of event handling by SLURM, the signal may
 be sent up to 60 seconds earlier than specified.
-Both \fIsig_time\fR and \fIsig_num\fR must have integer values
-between zero and 65535.
+\fIsig_num\fR may either be a signal number or name (e.g. "10" or "USR1").
+\fIsig_time\fR must have integer value between zero and 65535.
 By default, no signal is sent before the job's end time.
 If a \fIsig_num\fR is specified without any \fIsig_time\fR,
 the default time will be 60 seconds.

src/common/proc_args.c

@@ -753,17 +753,20 @@ int get_signal_opts(char *optarg, uint16_t *warn_signal, uint16_t *warn_time)
 	if (optarg == NULL)
 		return -1;
 
-	num = strtol(optarg, &endptr, 10);
-	if ((num < 0) || (num > 0x0ffff))
+	endptr = strchr(optarg, '@');
+	if (endptr)
+		endptr[0] = '\0';
+	num = (uint16_t) sig_name2num(optarg);
+	if (endptr)
+		endptr[0] = '@';
+	if ((num < 1) || (num > 0x0ffff))
 		return -1;
 	*warn_signal = (uint16_t) num;
 
-	if (endptr[0] == '\0') {
+	if (!endptr) {
 		*warn_time = 60;
 		return 0;
 	}
-	if (endptr[0] != '@')
-		return -1;
 
 	num = strtol(endptr+1, &endptr, 10);
 	if ((num < 0) || (num > 0x0ffff))
@@ -773,3 +776,46 @@ int get_signal_opts(char *optarg, uint16_t *warn_signal, uint16_t *warn_time)
 		return 0;
 	return -1;
 }
+
+/* Convert a signal name to it's numeric equivalent.
+ * Return -1 on failure */
+int sig_name2num(char *signal_name)
+{
+	char *sig_name[] = {"HUP", "INT", "QUIT", "KILL", "TERM",
+			    "USR1", "USR2", "CONT", NULL};
+	int sig_num[] = {SIGHUP, SIGINT, SIGQUIT, SIGKILL, SIGTERM,
+			       SIGUSR1, SIGUSR2, SIGCONT};
+	char *ptr;
+	long tmp;
+	int sig;
+	int i;
+
+	tmp = strtol(signal_name, &ptr, 10);
+	if (ptr != signal_name) { /* found a number */
+		if (xstring_is_whitespace(ptr))
+			sig = (int)tmp;
+		else
+			return 0;
+	} else {
+		ptr = (char *)signal_name;
+		while (isspace(*ptr))
+			ptr++;
+		if (strncasecmp(ptr, "SIG", 3) == 0)
+			ptr += 3;
+		for (i = 0; ; i++) {
+			if (sig_name[i] == NULL)
+				return 0;
+			if (strncasecmp(ptr, sig_name[i],
+					strlen(sig_name[i])) == 0) {
+				/* found the signal name */
+				if (!xstring_is_whitespace(ptr + 
+							   strlen(sig_name[i])))
+					return 0;
+				sig = sig_num[i];
+				break;
+			}
+		}
+	}
+
+	return sig;
+}

src/common/proc_args.h

@@ -132,4 +132,8 @@ char *print_geometry(const uint16_t *geometry);
  * RET 0 on success, -1 on failure */
 int get_signal_opts(char *optarg, uint16_t *warn_signal, uint16_t *warn_time);
 
+/* Convert a signal name to it's numeric equivalent.
+ * Return 0 on failure */
+int sig_name2num(char *signal_name);
+
 #endif /* !_PROC_ARGS_H */

src/salloc/opt.c

@@ -189,7 +189,6 @@ static bool _opt_verify(void);
 
 static void  _proc_get_user_env(char *optarg);
 static void _process_env_var(env_vars_t *e, const char *val);
-static int _parse_signal(const char *signal_name);
 static void  _usage(void);
 
 /*---[ end forward declarations of static functions ]---------------------*/
@@ -741,9 +740,11 @@ void set_options(const int argc, char **argv)
 			break;
 		case 'K': /* argument is optional */
 			if (optarg) {
-				opt.kill_command_signal =_parse_signal(optarg);
-				if (opt.kill_command_signal == 0)
+				opt.kill_command_signal = sig_name2num(optarg);
+				if (opt.kill_command_signal == 0) {
+					error("Invalid signal name %s", optarg);
 					exit(error_exit);
+				}
 			}
 			opt.kill_command_signal_set = true;
 			break;
@@ -1581,61 +1582,6 @@ static char *print_constraints()
 	return buf;
 }
 
-/*
- * Takes a string containing the number or name of a signal and returns
- * the signal number.  The signal name is case insensitive, and may be of
- * the form "SIGHUP" or just "HUP".
- *
- * Allowed signal names are HUP, INT, QUIT, KILL, TERM, USR1, USR2, and CONT.
- */
-static int _parse_signal(const char *signal_name)
-{
-	char *sig_name[] = {"HUP", "INT", "QUIT", "KILL", "TERM",
-			    "USR1", "USR2", "CONT", NULL};
-	int sig_num[] = {SIGHUP, SIGINT, SIGQUIT, SIGKILL, SIGTERM,
-			       SIGUSR1, SIGUSR2, SIGCONT};
-	char *ptr;
-	long tmp;
-	int sig;
-	int i;
-
-	tmp = strtol(signal_name, &ptr, 10);
-	if (ptr != signal_name) { /* found a number */
-		if (xstring_is_whitespace(ptr)) {
-			sig = (int)tmp;
-		} else {
-			goto fail;
-		}
-	} else {
-		ptr = (char *)signal_name;
-		while (isspace(*ptr))
-			ptr++;
-		if (strncasecmp(ptr, "SIG", 3) == 0)
-			ptr += 3;
-		for (i = 0; ; i++) {
-			if (sig_name[i] == NULL) {
-				goto fail;
-			}
-			if (strncasecmp(ptr, sig_name[i],
-					strlen(sig_name[i])) == 0) {
-				/* found the signal name */
-				if (!xstring_is_whitespace(
-					    ptr + strlen(sig_name[i]))) {
-					goto fail;
-				}
-				sig = sig_num[i];
-				break;
-			}
-		}
-	}
-
-	return sig;
-
-fail:
-	error("\"%s\" is not a valid signal", signal_name);
-	return 0;
-}
-
 #define tf_(b) (b == true) ? "true" : "false"
 
 static void _opt_list()