diff --git a/NEWS b/NEWS
index ea26b45cd81be71c93c972e3087792fe4145237f..af30aed7bb637260e1068ed28466759430e3b3c6 100644
--- a/NEWS
+++ b/NEWS
@@ -79,17 +79,17 @@ documents those changes that are of interest to users and admins.
* Changes in SLURM 1.2.19
=========================
-*** IMPORTANT CHANGE IN RPM BUILD BELOW ****
- -- slurm.spec file (used to build RPMs) updated in order to support Mock, a
- chroot build environment, see https://hosted.fedoraproject.org/projects/mock/
+*** NOTE IMPORTANT CHANGE IN RPM BUILD BELOW ****
+ -- slurm.spec file (used to build RPMs) was updated in order to support Mock, a
+ chroot build environment. See https://hosted.fedoraproject.org/projects/mock/
for more information. The following RPMs are no longer build by default:
- aix-federation, auth_none, authd, bluegene, sgijob. switch-elan. Change the
- RPMs built using the following options in ~/rpmmacros: "%_with_authd 1",
+ aix-federation, auth_none, authd, bluegene, sgijob, and switch-elan. Change
+ the RPMs built using the following options in ~/rpmmacros: "%_with_authd 1",
"%_without_munge 1", etc. See the slurm.spec file for more details.
-- Print warning if non-privileged user requests negative "--nice" value on
job submission (srun, salloc, and sbatch commands).
- -- In sched/wiki and sched/wiki2, add support for srun's
- --ntasks-per-node option.
+ -- In sched/wiki and sched/wiki2, add support for srun's --ntasks-per-node
+ option.
-- In select/bluegene with Groups defined for Images, fix possible memory
corruption. Other configurations are not affected.
-- BLUEGENE - Fix bug that prevented user specification of linux-image,
@@ -105,6 +105,10 @@ documents those changes that are of interest to users and admins.
-- If srun --output or --error specification contains a task number rather
than a file name, send stdout/err from specified task to srun's stdout/err
rather than to a file by the same name as the task's number.
+ -- For srun --multi-prog option, verify configuration file before attempting
+ to launch tasks, report clear explanation of any configuration file errors.
+ -- For sched/wiki2, add optional timeout option to srun's s --get-user-env
+ parameter, change default timeout from 3 to 8 seconds.
* Changes in SLURM 1.2.18
=========================
diff --git a/doc/man/man1/sbatch.1 b/doc/man/man1/sbatch.1
index bfacb42c641ca7242810a81d2f789274d2b86a78..61a35ec6fcd2c0f6babe1ef083fa1072150a2fbf 100644
--- a/doc/man/man1/sbatch.1
+++ b/doc/man/man1/sbatch.1
@@ -154,13 +154,14 @@ The order of the node names in the list is not important; the node names
will be sorted my SLURM.
.TP
-\fB\-\-get\-user\-env\fR
+\fB\-\-get\-user\-env\fR[=\fItimeout\fR]
This option will tell sbatch to retrieve the
login environment variables for the user specified in the \-\-uid option.
The environment variables are retrieved by running "su - <username> -c
/usr/bin/env" and parsing the output. Be aware that any environment
variables already set in sbatch's environment will take precedence over any
environment variables in the user's login environment.
+Optional timeout value is in seconds. Default value is 8 seconds.
NOTE: This option only works if the caller has an effective uid of "root".
This option was originally created for use by Moab.
diff --git a/doc/man/man1/srun.1 b/doc/man/man1/srun.1
index a5afedb093c238ac8aff8841b91ca02fa0282dc0..7576aefefb4bf86a9d7e76043c1d5bdb87daf4a3 100644
--- a/doc/man/man1/srun.1
+++ b/doc/man/man1/srun.1
@@ -1348,9 +1348,13 @@ space:
Task rank
One or more task ranks to use this configuration.
Multiple values may be comma separated.
-Ranges may be indicated with two numbers separated with a '\-'.
+Ranges may be indicated with two numbers separated with a '\-' with
+the smaller number first (e.g. "0\-4" and not "4\-0").
To indicate all tasks, specify a rank of '*' (in which case you probably
should not be using this option).
+If an attempt is made to initiate a task for which no executable
+program is defined, the following error message will be produced
+"No executable program specified for this task".
.TP
Executable
The name of the program to execute.
diff --git a/src/common/env.c b/src/common/env.c
index d0098eb1ff39c56b6cd9ecbff44ee550391b8cd9..fa945bb91d4fa9a80f332024a4707514eb9b4a1c 100644
--- a/src/common/env.c
+++ b/src/common/env.c
@@ -75,7 +75,7 @@ strong_alias(env_array_append_fmt, slurm_env_array_append_fmt);
strong_alias(env_array_overwrite, slurm_env_array_overwrite);
strong_alias(env_array_overwrite_fmt, slurm_env_array_overwrite_fmt);
-#define SU_WAIT_MSEC 3000 /* 3000 msec for /bin/su to return user
+#define SU_WAIT_MSEC 8000 /* 8000 msec for /bin/su to return user
* env vars for --get-user-env option */
/*
* Return pointer to `name' entry in environment if found, or
@@ -1239,12 +1239,13 @@ static void _strip_cr_nl(char *line)
* environment variables, as determined by calling (more-or-less)
* "/bin/su - <username> -c /usr/bin/env".
*
+ * timeout value is in seconds or zero for default (8 secs)
* On error, returns NULL.
*
* NOTE: The calling process must have an effective uid of root for
* this function to succeed.
*/
-char **env_array_user_default(const char *username)
+char **env_array_user_default(const char *username, int timeout)
{
FILE *su;
char line[BUFSIZ];
@@ -1283,13 +1284,7 @@ char **env_array_user_default(const char *username)
snprintf(cmdstr, sizeof(cmdstr),
"echo; echo; echo; echo %s; env; echo %s",
starttoken, stoptoken);
-#if 0
- /* execute .profile only */
- execl("/bin/su", "su", username, "-c", cmdstr, NULL);
-#else
- /* execute .login plus .profile */
execl("/bin/su", "su", "-", username, "-c", cmdstr, NULL);
-#endif
exit(1);
}
@@ -1307,7 +1302,10 @@ char **env_array_user_default(const char *username)
found = 0;
while (!found) {
gettimeofday(&now, NULL);
- timeleft = SU_WAIT_MSEC;
+ if (timeout > 0)
+ timeleft = timeout * 1000;
+ else
+ timeleft = SU_WAIT_MSEC;
timeleft -= (now.tv_sec - begin.tv_sec) * 1000;
timeleft -= (now.tv_usec - begin.tv_usec) / 1000;
if (timeleft <= 0)
@@ -1343,7 +1341,10 @@ char **env_array_user_default(const char *username)
found = 0;
while (!found) {
gettimeofday(&now, NULL);
- timeleft = SU_WAIT_MSEC;
+ if (timeout > 0)
+ timeleft = timeout * 1000;
+ else
+ timeleft = SU_WAIT_MSEC;
timeleft -= (now.tv_sec - begin.tv_sec) * 1000;
timeleft -= (now.tv_usec - begin.tv_usec) / 1000;
if (timeleft <= 0)
diff --git a/src/common/env.h b/src/common/env.h
index 4b535380c4b2645354613efeee80b83a214ab309..c62e0054593ef4cc89e366db8be68952b67f2a9d 100644
--- a/src/common/env.h
+++ b/src/common/env.h
@@ -244,11 +244,12 @@ void env_array_set_environment(char **env_array);
* environment variables, as determined by calling (more-or-less)
* "/bin/su - <username> -c /usr/bin/env".
*
+ * timeout value is in seconds or zero for default (8 secs)
* On error, returns NULL.
*
* NOTE: The calling process must have an effective uid of root for
* this function to succeed.
*/
-char **env_array_user_default(const char *username);
+char **env_array_user_default(const char *username, int timeout);
#endif
diff --git a/src/plugins/sched/wiki2/start_job.c b/src/plugins/sched/wiki2/start_job.c
index a6afae77f0f69e1d757a3ffb4be13107daf00878..f7cdd46d189ad8b080c780beb3ac794e7a9ca0ee 100644
--- a/src/plugins/sched/wiki2/start_job.c
+++ b/src/plugins/sched/wiki2/start_job.c
@@ -243,21 +243,6 @@ static int _start_job(uint32_t jobid, int task_cnt, char *hostlist,
lock_slurmctld(job_write_lock);
if (job_ptr->job_id != jobid)
job_ptr = find_job_record(jobid);
- if (job_ptr && (job_ptr->job_id == jobid) && job_ptr->details &&
- (job_ptr->job_state == JOB_RUNNING)) {
- /* Restore required node list, even if job has started so
- * that it can be requeued with the original requirements */
- xfree(job_ptr->details->req_nodes);
- job_ptr->details->req_nodes = save_req_nodes;
- FREE_NULL_BITMAP(job_ptr->details->req_node_bitmap);
- job_ptr->details->req_node_bitmap = save_req_bitmap;
- FREE_NULL_BITMAP(job_ptr->details->exc_node_bitmap);
- } else {
- xfree(save_req_nodes);
- FREE_NULL_BITMAP(save_req_bitmap);
- if (job_ptr && (job_ptr->job_id == jobid) && job_ptr->details)
- FREE_NULL_BITMAP(job_ptr->details->exc_node_bitmap);
- }
if (job_ptr && (job_ptr->job_id == jobid)
&& (job_ptr->job_state != JOB_RUNNING)) {
@@ -283,22 +268,26 @@ static int _start_job(uint32_t jobid, int task_cnt, char *hostlist,
*err_msg = tmp_msg;
error("wiki: %s", tmp_msg);
- /* Restore remainder of job state. Note the required node
- * information is restored above (even if job starts). */
+ /* restore some of job state */
job_ptr->priority = 0;
job_ptr->num_procs = old_task_cnt;
- if (job_ptr->details) {
- /* Details get cleared on job abort; happens
- * if the request is sufficiently messed up.
- * This happens when Moab tries to start a
- * a job on invalid nodes (wrong partition). */
- xfree(job_ptr->details->req_nodes);
- FREE_NULL_BITMAP(job_ptr->details->
- req_node_bitmap);
- xfree(job_ptr->details->req_node_layout);
- }
rc = -1;
}
+
+ if (job_ptr && (job_ptr->job_id == jobid) && job_ptr->details) {
+ /* Restore required node list in case job requeued */
+ xfree(job_ptr->details->req_nodes);
+ job_ptr->details->req_nodes = save_req_nodes;
+ FREE_NULL_BITMAP(job_ptr->details->req_node_bitmap);
+ job_ptr->details->req_node_bitmap = save_req_bitmap;
+ FREE_NULL_BITMAP(job_ptr->details->exc_node_bitmap);
+ xfree(job_ptr->details->req_node_layout);
+ } else {
+ error("wiki: start_job(%u) job missing", jobid);
+ xfree(save_req_nodes);
+ FREE_NULL_BITMAP(save_req_bitmap);
+ }
+
unlock_slurmctld(job_write_lock);
schedule_node_save(); /* provides own locking */
schedule_job_save(); /* provides own locking */
diff --git a/src/sbatch/opt.c b/src/sbatch/opt.c
index cf5f12aa785e0a73d64328997011ea0e47afb7f0..a1ac569a40e85de3fc4c7be20a70e6d4ba3a4873 100644
--- a/src/sbatch/opt.c
+++ b/src/sbatch/opt.c
@@ -278,7 +278,7 @@ static void _opt_default()
opt.ifname = xstrdup("/dev/null");
opt.ofname = NULL;
opt.efname = NULL;
- opt.get_user_env = false;
+ opt.get_user_env = -1;
}
/*---[ env var processing ]-----------------------------------------------*/
@@ -498,7 +498,7 @@ static struct option long_options[] = {
{"reboot", no_argument, 0, LONG_OPT_REBOOT},
{"tasks-per-node",required_argument, 0, LONG_OPT_NTASKSPERNODE},
{"wrap", required_argument, 0, LONG_OPT_WRAP},
- {"get-user-env", no_argument, 0, LONG_OPT_GET_USER_ENV},
+ {"get-user-env", optional_argument, 0, LONG_OPT_GET_USER_ENV},
{NULL, 0, 0, 0}
};
@@ -1205,7 +1205,10 @@ static void _set_options(int argc, char **argv)
/* handled in process_options_first_pass() */
break;
case LONG_OPT_GET_USER_ENV:
- opt.get_user_env = true;
+ if (optarg)
+ opt.get_user_env = strtol(optarg, NULL, 10);
+ else
+ opt.get_user_env = 0;
break;
default:
fatal("Unrecognized command line parameter %c",
diff --git a/src/sbatch/opt.h b/src/sbatch/opt.h
index 6d2009f38205a55db66e54ba2a6a4d0ab8731e3c..1c572e5f6e35a3fe4b691fd56aa012f56910be0b 100644
--- a/src/sbatch/opt.h
+++ b/src/sbatch/opt.h
@@ -133,7 +133,7 @@ typedef struct sbatch_options {
char *ifname; /* input file name */
char *ofname; /* output file name */
char *efname; /* error file name */
- bool get_user_env; /* --get-user-env */
+ int get_user_env; /* --get-user-env[=timeout] */
} opt_t;
extern opt_t opt;
diff --git a/src/sbatch/sbatch.c b/src/sbatch/sbatch.c
index c0fe4f113ef96145df6fbe819862ad8580954587..f608b538739f6f5d255fc16967960cc15fe88ed1 100644
--- a/src/sbatch/sbatch.c
+++ b/src/sbatch/sbatch.c
@@ -217,11 +217,12 @@ static int fill_job_desc_from_opts(job_desc_msg_t *desc)
desc->shared = opt.shared;
desc->environment = NULL;
- if (opt.get_user_env) {
+ if (opt.get_user_env >= 0) {
struct passwd *pw = NULL;
pw = getpwuid(opt.uid);
if (pw != NULL) {
- desc->environment = env_array_user_default(pw->pw_name);
+ desc->environment = env_array_user_default(pw->pw_name,
+ opt.get_user_env);
/* FIXME - should we abort if j->environment
* is NULL? */
}
diff --git a/src/slurmd/slurmstepd/multi_prog.c b/src/slurmd/slurmstepd/multi_prog.c
index 87e0861490eba5bfb4abfc541d1540d59b76ad34..6a1c2e8fd2efcb9e264582b133a05fef160947b3 100644
--- a/src/slurmd/slurmstepd/multi_prog.c
+++ b/src/slurmd/slurmstepd/multi_prog.c
@@ -151,13 +151,15 @@ multi_prog_get_argv(char *file_contents, char **prog_env, int task_rank,
char **prog_argv = NULL;
char *local_data = NULL;
+ prog_argv = (char **)xmalloc(sizeof(char *) * 128);
+
if (task_rank < 0) {
- *argc = 0;
- *argv = NULL;
+ error("Invalid task rank %d", task_rank);
+ *argc = 1;
+ *argv = prog_argv;
return -1;
}
- prog_argv = (char **)xmalloc(sizeof(char *) * 128);
local_data = xstrdup(file_contents);
line = strtok_r(local_data, "\n", &ptrptr);
@@ -165,7 +167,7 @@ multi_prog_get_argv(char *file_contents, char **prog_env, int task_rank,
if (line_num > 0)
line = strtok_r(NULL, "\n", &ptrptr);
if (line == NULL) {
- error("Could not identify executable program for this task");
+ error("No executable program specified for this task");
goto fail;
}
line_num ++;
@@ -272,9 +274,9 @@ multi_prog_get_argv(char *file_contents, char **prog_env, int task_rank,
error("Program for task rank %d not specified.", task_rank);
fail:
- xfree(prog_argv);
xfree(local_data);
- *argc = 0;
- *argv = NULL;
+ *argc = 1;
+ prog_argv[0] = NULL;
+ *argv = prog_argv;
return -1;
}
diff --git a/src/slurmd/slurmstepd/task.c b/src/slurmd/slurmstepd/task.c
index 2d10bb2a62c3da974fd058efca9475afb50e3fa6..94c021a3e729a136f81bb20c80893cd851c48c50 100644
--- a/src/slurmd/slurmstepd/task.c
+++ b/src/slurmd/slurmstepd/task.c
@@ -354,7 +354,7 @@ exec_task(slurmd_job_t *job, int i, int waitfd)
job->envtp->env = NULL;
xfree(job->envtp->task_count);
- if (job->multi_prog) {
+ if (job->multi_prog && task->argv[0]) {
/*
* Normally the client (srun/slauch) expands the command name
* to a fully qualified path, but in --multi-prog mode it
@@ -416,6 +416,10 @@ exec_task(slurmd_job_t *job, int i, int waitfd)
job->env[0] = (char *)NULL;
}
+ if (task->argv[0] == NULL) {
+ error("No executable program specified for this task");
+ exit(2);
+ }
execve(task->argv[0], task->argv, job->env);
/*
diff --git a/src/srun/multi_prog.c b/src/srun/multi_prog.c
index 0ed471ccfddd245c6869d9cba5c308278d0b38a1..334c8bdb2fe0cc2c47fa8ed7c887d2504f6ffa3e 100644
--- a/src/srun/multi_prog.c
+++ b/src/srun/multi_prog.c
@@ -54,6 +54,7 @@
#include <sys/types.h>
#include <unistd.h>
+#include "src/common/bitstring.h"
#include "src/common/log.h"
#include "src/common/xassert.h"
#include "src/common/xmalloc.h"
@@ -132,7 +133,7 @@ _set_exec_names(char *ranks, char *exec_name, int ntasks)
if (ranks[0] == '*' && ranks[1] == '\0') {
low_num = 0;
- high_num = ntasks -1;
+ high_num = ntasks - 1;
_set_range(low_num, high_num, exec_name);
return;
}
@@ -161,7 +162,8 @@ _set_exec_names(char *ranks, char *exec_name, int ntasks)
high_num = MIN((ntasks-1), atoi(upper));
_set_range(low_num, high_num, exec_path);
} else {
- error ("Invalid task range specification (%s) ignored.", range);
+ error ("Invalid task range specification (%s) ignored.",
+ range);
}
}
}
@@ -188,8 +190,8 @@ mpir_set_multi_name(int ntasks, const char *config_fname)
while (fgets(line, sizeof(line), config_fd)) {
line_num ++;
if (strlen (line) >= (sizeof(line) - 1)) {
- error ("Line %d of configuration file too long",
- line_num);
+ error ("Line %d of configuration file %s too long",
+ line_num, config_fname);
fclose(config_fd);
return -1;
}
@@ -206,7 +208,8 @@ mpir_set_multi_name(int ntasks, const char *config_fname)
ranks = strtok_r(p, " \t\n", &ptrptr);
exec_name = strtok_r(NULL, " \t\n", &ptrptr);
if (!ranks || !exec_name) {
- error("Line %d is invalid", line_num);
+ error("Line %d of configuration file %s is invalid",
+ line_num, config_fname);
fclose(config_fd);
return -1;
}
@@ -264,3 +267,150 @@ mpir_dump_proctable()
i, tv->host_name, tv->pid, tv->executable_name);
}
}
+
+static int
+_update_task_mask(int low_num, int high_num, int ntasks, bitstr_t *task_mask)
+{
+ int i;
+
+ if (low_num > high_num) {
+ error("Invalid task range, %d-%d", low_num, high_num);
+ return -1;
+ }
+ if (low_num < 0) {
+ error("Invalid task id, %d < 0", low_num);
+ return -1;
+ }
+ if (high_num >= ntasks) {
+ error("Invalid task id, %d >= ntasks", high_num);
+ return -1;
+ }
+ for (i=low_num; i<=high_num; i++) {
+ if (bit_test(task_mask, i)) {
+ error("Duplicate record for task %d", i);
+ return -1;
+ }
+ bit_set(task_mask, i);
+ }
+ return 0;
+}
+
+static int
+_validate_ranks(char *ranks, int ntasks, bitstr_t *task_mask)
+{
+ char *range = NULL, *p = NULL;
+ char *ptrptr = NULL, *upper = NULL;
+ int low_num, high_num;
+
+ if (ranks[0] == '*' && ranks[1] == '\0') {
+ low_num = 0;
+ high_num = ntasks - 1;
+ return _update_task_mask(low_num, high_num, ntasks, task_mask);
+ }
+
+ for (range = strtok_r(ranks, ",", &ptrptr); range != NULL;
+ range = strtok_r(NULL, ",", &ptrptr)) {
+ p = range;
+ while (*p != '\0' && isdigit (*p))
+ p ++;
+
+ if (*p == '\0') { /* single rank */
+ low_num = atoi(range);
+ high_num = low_num;
+ } else if (*p == '-') { /* lower-upper */
+ upper = ++ p;
+ while (isdigit (*p))
+ p ++;
+ if (*p != '\0') {
+ error ("Invalid task range specification");
+ return -1;
+ }
+ low_num = atoi(range);
+ high_num = atoi(upper);
+ } else {
+ error ("Invalid task range specification (%s)",
+ range);
+ return -1;
+ }
+
+ if (_update_task_mask(low_num, high_num, ntasks, task_mask))
+ return -1;
+ }
+ return 0;
+}
+
+/*
+ * Verify that we have a valid executable program specified for each task
+ * when the --multi-prog option is used.
+ *
+ * Return 0 on success, -1 otherwise
+ */
+extern int
+verify_multi_name(char *config_fname, int ntasks)
+{
+ FILE *config_fd;
+ char line[256];
+ char *ranks, *exec_name, *p, *ptrptr;
+ int line_num = 0, i, rc = 0;
+ bitstr_t *task_mask;
+
+ if (ntasks <= 0) {
+ error("Invalid task count %d", ntasks);
+ return -1;
+ }
+
+ config_fd = fopen(config_fname, "r");
+ if (config_fd == NULL) {
+ error("Unable to open configuration file %s", config_fname);
+ return -1;
+ }
+
+ task_mask = bit_alloc(ntasks);
+ while (fgets(line, sizeof(line), config_fd)) {
+ line_num ++;
+ if (strlen (line) >= (sizeof(line) - 1)) {
+ error ("Line %d of configuration file %s too long",
+ line_num, config_fname);
+ rc = -1;
+ goto fini;
+ }
+ p = line;
+ while (*p != '\0' && isspace (*p)) /* remove leading spaces */
+ p ++;
+
+ if (*p == '#') /* only whole-line comments handled */
+ continue;
+
+ if (*p == '\0') /* blank line ignored */
+ continue;
+
+ ranks = strtok_r(p, " \t\n", &ptrptr);
+ exec_name = strtok_r(NULL, " \t\n", &ptrptr);
+ if (!ranks || !exec_name) {
+ error("Line %d of configuration file %s invalid",
+ line_num, config_fname);
+ rc = -1;
+ goto fini;
+ }
+ if (_validate_ranks(ranks, ntasks, task_mask)) {
+ error("Line %d of configuration file %s invalid",
+ line_num, config_fname);
+ rc = -1;
+ goto fini;
+ }
+ }
+
+ for (i=0; i<ntasks; i++) {
+ if (!bit_test(task_mask, i)) {
+ error("Configuration file %s invalid, "
+ "no record for task id %d",
+ config_fname, i);
+ rc = -1;
+ goto fini;
+ }
+ }
+
+fini: fclose(config_fd);
+ bit_free(task_mask);
+ return rc;
+}
diff --git a/src/srun/multi_prog.h b/src/srun/multi_prog.h
index bdb0c7de1989e5a4b05d3f68a1440f895907451d..d837ed97e92986c51a7aed7b69cffc2427bb5fad 100644
--- a/src/srun/multi_prog.h
+++ b/src/srun/multi_prog.h
@@ -48,5 +48,13 @@ extern void mpir_cleanup(void);
extern void mpir_set_executable_names(const char *executable_name);
extern void mpir_dump_proctable(void);
+/*
+ * Verify that we have a valid executable program specified for each task
+ * when the --multi-prog option is used.
+ *
+ * Return 0 on success, -1 otherwise
+ */
+extern int verify_multi_name(char *config_fname, int ntasks);
+
#endif
diff --git a/src/srun/opt.c b/src/srun/opt.c
index e114a3ab87b49b0f4d87bc65d56ec3333b76bf21..7979b5808c498e2b40d2d572a3e32d2823855726 100644
--- a/src/srun/opt.c
+++ b/src/srun/opt.c
@@ -84,6 +84,7 @@
#include "src/common/optz.h"
#include "src/api/pmi_server.h"
+#include "src/srun/multi_prog.h"
#include "src/srun/opt.h"
#include "src/srun/debugger.h"
#include "src/common/mpi.h"
@@ -697,7 +698,6 @@ static void _opt_default()
opt.msg_timeout = 15;
}
- opt.get_user_env = false;
opt.pty = false;
}
@@ -1009,7 +1009,7 @@ static void set_options(const int argc, char **argv)
{"mloader-image", required_argument, 0, LONG_OPT_MLOADER_IMAGE},
{"ramdisk-image", required_argument, 0, LONG_OPT_RAMDISK_IMAGE},
{"reboot", no_argument, 0, LONG_OPT_REBOOT},
- {"get-user-env", no_argument, 0, LONG_OPT_GET_USER_ENV},
+ {"get-user-env", optional_argument, 0, LONG_OPT_GET_USER_ENV},
{"pty", no_argument, 0, LONG_OPT_PTY},
{"checkpoint", required_argument, 0, LONG_OPT_CHECKPOINT},
{NULL, 0, 0, 0}
@@ -1494,7 +1494,7 @@ static void set_options(const int argc, char **argv)
opt.reboot = true;
break;
case LONG_OPT_GET_USER_ENV:
- opt.get_user_env = true;
+ error("--get-user-env is no longer supported in srun, use sbatch");
break;
case LONG_OPT_PTY:
#ifdef HAVE_PTY_H
@@ -1635,7 +1635,6 @@ static void _opt_args(int argc, char **argv)
exit(1);
}
_load_multi(&opt.argc, opt.argv);
-
}
else if (opt.argc > 0) {
char *fullpath;
@@ -1645,6 +1644,9 @@ static void _opt_args(int argc, char **argv)
opt.argv[0] = fullpath;
}
}
+
+ if (opt.multi_prog && verify_multi_name(opt.argv[0], opt.nprocs))
+ exit(1);
}
/*
diff --git a/src/srun/opt.h b/src/srun/opt.h
index bc574bb2c46e2f83f5ebef7d41df0b98a107551b..2e150aaf2e8d77735d6a8e1d4654de13e69d061f 100644
--- a/src/srun/opt.h
+++ b/src/srun/opt.h
@@ -198,7 +198,6 @@ typedef struct srun_options {
uint16_t mail_type; /* --mail-type */
char *mail_user; /* --mail-user */
char *ctrl_comm_ifhn; /* --ctrl-comm-ifhn */
- bool get_user_env; /* --get-user-env */
bool pty; /* --pty */
int argc; /* length of argv array */
char **argv; /* left over on command line */
diff --git a/testsuite/expect/test1.54 b/testsuite/expect/test1.54
index 573422264b63f05102e119f680c1a576078af566..21a153bfb34e04ff38dfb6b6de8d7135d0038195 100755
--- a/testsuite/expect/test1.54
+++ b/testsuite/expect/test1.54
@@ -96,6 +96,8 @@ if {$matches != 4} {
}
if {$exit_code != 0} {
exit $exit_code
+} else {
+ send_user "\nSo far, so good\n\n"
}
#
@@ -146,6 +148,37 @@ if {$matches != 4} {
set exit_code 1
}
+if {$exit_code != 0} {
+ exit $exit_code
+} else {
+ send_user "\nSo far, so good\n\n"
+}
+
+#
+# Submit a slurm job that asks for more tasks than specified in our
+# configuration file
+#
+set matches 0
+set srun_pid [spawn $srun -N1 -n5 --overcommit -l -t1 --multi-prog ./$file_in]
+expect {
+ -re "Configuration file .* invalid" {
+ send_user "\nNo worries. This error is expected\n"
+ incr matches
+ }
+ timeout {
+ send_user "\nFAILURE: srun not responding\n"
+ slow_kill $srun_pid
+ set exit_code 1
+ }
+ eof {
+ wait
+ }
+}
+if {$matches != 1} {
+ send_user "\nFAILURE: did not note lack of a executable for task 5.\n"
+ set exit_code 1
+}
+
if {$exit_code == 0} {
exec $bin_rm -f $file_in
send_user "\nSUCCESS\n"
diff --git a/testsuite/expect/test6.7 b/testsuite/expect/test6.7
index c3b363f8e33ec24f20dbd1a437dd54148c121091..61066ed411248a59e176af256c4c2e480ceb645d 100755
--- a/testsuite/expect/test6.7
+++ b/testsuite/expect/test6.7
@@ -75,7 +75,8 @@ if {[wait_for_job $job_id RUNNING] != 0} {
cancel_job $job_id
exit 1
}
-exec $bin_sleep 2
+# Allow time for the step to start
+exec $bin_sleep 4
#
# Test verbose scancel