Skip to content
Snippets Groups Projects
Commit adf19d52 authored by Moe Jette's avatar Moe Jette
Browse files

Remove rest of kill_srun calls, clean up many salloc_kills too

parent d3fb3f19
No related branches found
No related tags found
No related merge requests found
Showing
with 63 additions and 110 deletions
......@@ -58,7 +58,7 @@ make_bash_script $file_script "
# Set umask and confirm it is propagated
#
set matches 0
spawn ./$file_script
set srun_pid [spawn ./$file_script]
expect {
-re (0123|123) {
set matches 1
......@@ -66,7 +66,7 @@ expect {
}
timeout {
send_user "\nFAILURE: srun not responding\n"
kill_srun
slow_kill [expr 0 - $srun_pid]
exit 1
}
eof {
......
......@@ -62,7 +62,7 @@ for {set inx 1} {$inx < 4} {set inx [expr $inx * 2]} {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill $salloc_pid
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
}
eof {
......
......@@ -63,11 +63,10 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
......@@ -103,11 +102,10 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
......
......@@ -68,11 +68,10 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
......
......@@ -76,7 +76,7 @@ if {$job_id1 == 0} {
#
set match_acct 0
set match_state 0
spawn $salloc --dependency=$job_id1 $slaunch $scontrol show job $job_id1
set salloc_pid [spawn $salloc --dependency=$job_id1 $slaunch $scontrol show job $job_id1]
expect {
-re "Granted job allocation ($number)" {
set job_id2 $expect_out(1,string)
......@@ -91,10 +91,12 @@ expect {
exp_continue
}
timeout {
send_user "\nFAILURE: srun not responding\n"
kill_srun
send_user "\nFAILURE: salloc not responding\n"
if {$job_id2 != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -55,13 +55,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -88,13 +86,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -56,13 +56,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -91,13 +89,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -82,11 +82,9 @@ expect {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id_1 != 0} {
cancel_job $job_id_1
} else {
slow_kill $salloc_pid
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -60,13 +60,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -68,13 +68,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -108,13 +106,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -154,13 +150,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -214,13 +208,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -289,13 +281,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -97,11 +97,10 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
exit 1
}
eof {
......@@ -183,11 +182,10 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
exit 1
}
eof {
......@@ -246,13 +244,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -103,18 +103,16 @@ expect {
exp_continue
}
timeout {
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
if {$waiting == 1} {
send_user "\nWARNING: salloc still not responding, quiting\n"
exit 0
} else {
send_user "\nFAILURE: salloc not responding\n"
exit 1
}
exit 1
}
eof {
wait
......
......@@ -81,13 +81,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -141,13 +139,11 @@ expect {
}
timeout {
send_user "\nFAILURE: srun not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -78,13 +78,11 @@ expect {
}
timeout {
send_user "\nFAILURE: srun not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -87,9 +87,8 @@ expect {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill $salloc_pid
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -119,9 +119,8 @@ expect {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill $salloc_pid
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -145,9 +144,8 @@ expect {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill $salloc_pid
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -71,13 +71,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -63,9 +63,8 @@ expect {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill $salloc_pid
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -55,13 +55,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......@@ -93,13 +91,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
......@@ -70,13 +70,11 @@ expect {
}
timeout {
send_user "\nFAILURE: salloc not responding\n"
if {$job_id == 0} {
slow_kill $salloc_pid
} else {
if {$job_id != 0} {
cancel_job $job_id
}
slow_kill [expr 0 - $salloc_pid]
set exit_code 1
exp_continue
}
eof {
wait
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment