add autobuild on delta

.github/workflows/autobuild.yml

@@ -0,0 +1,50 @@
+name: Autobuild
+
+on:
+ schedule:
+ # Run the build as part of a fixed nightly schedule
+ - cron: '15 06 * * *' # UTC 6:15am, corresponds to 00:15 CST or 01:15 CDT
+ push:
+ paths:
+ # Also run the build when this file gets modified as part of a PR
+ - '.github/workflows/autobuild.yml'
+ - '.github/workflows/delta-sbatch-slurm.sh'
+
+
+# Cancel in progress CI runs when a new run targeting the same PR or branch/tag is triggered.
+# https://stackoverflow.com/questions/66335225/how-to-cancel-previous-runs-in-the-pr-when-you-push-new-commitsupdate-the-curre
+concurrency:
+ group: ${{ github.workflow }}-${{ github.ref }}
+ cancel-in-progress: true
+
+jobs:
+ Delta:
+ timeout-minutes: 60
+
+ runs-on: delta
+ name: Delta mpi-linux-x86_64 # Could test various builds (e.g., MPI, UCX, ...) via a build matrix
+
+ steps:
+ - uses: actions/checkout@v3
+ - name: Host info
+ run: |
+ set -x
+ echo "Running autobuild on delta"
+ hostname
+ uname -a
+ lsb_release -a
+ pwd
+ - name: build
+ run: |
+ set -ex
+ export target="mpi-linux-x86_64"
+ .github/workflows/jobmonitor.sh .github/workflows/delta-sbatch-slurm.sh
+ - name: results
+ run: |
+ if grep '0' result.latest
+ then
+ echo "Success"
+ else
+ echo "Failure"
+ fi 
+ # should also https://github.com/marketplace/actions/send-email

.github/workflows/delta-sbatch-slurm.sh

@@ -0,0 +1,21 @@
+#!/bin/bash -l
+#SBATCH -N 2
+#SBATCH -n 64
+#SBATCH -o %j.output
+#SBATCH -e %j.output
+#SBATCH -t 1:00:00
+#SBATCH -J autobuild
+#SBATCH -p cpu
+#SBATCH -A mzu-delta-cpu
+#cd $indir
+set -x
+module load libfabric; module load cmake
+./build all-test $target --with-production --enable-error-checking -j64 -g
+#
+cd $target
+make -C tests test OPTS="$flags" TESTOPTS="$testopts" $maketestopts
+make -C examples test OPTS="$flags" TESTOPTS="$testopts" $maketestopts
+make -C benchmarks test OPTS="$flags" TESTOPTS="$testopts" $maketestopts
+# Save make exit status
+status=$?
+echo $status > ../$SLURM_JOBID.result

.github/workflows/jobmonitor.sh

@@ -0,0 +1,69 @@
+#!/bin/bash
+
+export script=$1;
+queue_qsub=sbatch
+queue_kill=scancel
+queue_stat="squeue -j"
+
+End() {
+ echo "autobuild> $queue_kill $jobid ..."
+ $queue_kill $jobid
+ exit $1
+}
+echo "Submitting batch job for> $target OPTS=$flags"
+echo " using the command> $queue_qsub $script"
+chmod 755 $script
+while [ -z "$jobid" ]
+do
+ $queue_qsub $script > .status.$$ 2>&1
+ if grep 'have no accounts' .status.$$ > /dev/null
+ then
+ echo "NO account for submitting batch job!"
+ rm -f .status.$$
+ exit 1
+ fi
+ jobid=`cat .status.$$ | tail -1 | awk '{print $4}'`
+ rm -f .status.$$
+done
+echo "Job enqueued under job ID $jobid"
+export output=$jobid.output
+export result=$jobid.result
+# kill job if interrupted
+trap 'End 1' 2 3
+retry=0
+# Wait for the job to complete, by checking its status
+while [ true ]
+do
+ $queue_stat $jobid > tmp.$$
+ linecount=`wc -l tmp.$$ | awk '{print $1}' `
+ exitstatus=$?
+ #if [[ $exitstatus != 0 || $linecount != 2 ]]
+ if [[ $linecount != 2 ]]
+ then
+ # The job is done-- print its output
+ rm tmp.$$
+ # When job hangs, result file does not exist
+ test -f $result && status=`cat $result` || status=1
+ echo "==================================== OUTPUT (STDOUT & STDERR) ========================================"
+ cat $output
+ echo "======================================================================================================"
+ if [[ "$status" != 0 ]];
+ then
+ #print script
+ echo "=============================================== SCRIPT ==============================================="
+ cat $script
+ echo "======================================================================================================"
+ echo "=============================================== RESULT ==============================================="
+ cat $result
+ echo "======================================================================================================"
+ fi
+ # mv result and output to result.latest
+ mv $result result.latest
+ mv $output output.latest
+ exit $status
+ fi
+# The job is still queued or running-- print status and wait
+ tail -1 tmp.$$
+ rm tmp.$$
+ sleep 60
+done