2
0
mirror of https://gitlab.isc.org/isc-projects/kea synced 2025-08-22 01:49:48 +00:00
kea/tools/tests_in_valgrind.sh

85 lines
2.7 KiB
Bash
Raw Normal View History

2011-03-02 15:16:31 +01:00
#!/bin/sh
################################################################################
# This script runs all tests in valgrind. Configure and compile kea the way
# you want it to be tested (you should use --with-gtest or --with-gtest-source,
# however, or you get no tests). Then run this script from the top build
# directory.
#
# Note that the test isn't what you would call "production quality" (it is
# expected to be used by the kea developers, not end user) and might break,
# some ways of breaking it are known.
#
# There are two variables that modify it's behaviour.
# * VALGRIND_FLAGS are the flag passed to valgrind. There are some, hopefully
# reasonable defaults which you can overwrite. Note that the variable is
# used unmodified inside a sed pattern with # as a modifier, which can
# easily break it. There was no motivation to fix this.
# * VALGRIND_FILE is the file to store the output into. Default is valgrind.log
################################################################################
# Change directory to Kea's top build directory.
script_path=$(cd "$(dirname "${0}")" && pwd)
cd "${script_path}/.."
# First, make sure the tests are up to date
make
if test $? = 2; then
printf 'Did you run configure?\n'
exit 1
fi
# Exit with error if commands exit with non-zero and if undefined variables are
# used.
set -eu
# Some configuration
# TODO Escape for sed, this might break
LOGFILE="${VALGRIND_FILE:-$(pwd)/valgrind.log}"
2011-03-02 13:20:33 +01:00
FLAGS="${VALGRIND_FLAGS:---leak-check=full --track-fds=yes}"
FLAGS="$FLAGS --log-file=$LOGFILE.%p"
FOUND_ANY=false
FAILED=
# Find all the tests (yes, doing it by a name is a nasty hack)
# Since the while runs in a subprocess, we need to get the assignments out, done by the eval
find . -type f -name '*_unittests' -print | grep -Fv '.libs/' | while read -r testname ; do
sed -e 's#exec "#exec valgrind '"$FLAGS"' "#' "$testname" > "$testname.valgrind"
chmod +x "$testname.valgrind"
echo "$testname" >>"$LOGFILE"
echo "===============" >>"$LOGFILE"
OLDDIR=$(pwd)
cd "$(dirname "$testname")"
"./$(basename "$testname").valgrind" >&2 &
PID="$!"
set +e
wait "$PID"
CODE="$?"
set -e
2011-03-02 15:16:31 +01:00
cd "$OLDDIR"
if [ "$CODE" != 0 ] ; then
printf 'FAILED="%s
%s"' "${FAILED}" "${testname}"
fi
NAME="$LOGFILE.$PID"
rm "$testname.valgrind"
# Remove the ones from death tests
if [ -e "${NAME}" ]; then
grep "==$PID==" "$NAME" >>"$LOGFILE"
rm "$NAME"
fi
echo 'FOUND_ANY=true'
done
if test -n "$FAILED"; then
echo "These tests failed:" >&2
echo "$FAILED" >&2
fi
if ! $FOUND_ANY ; then
2011-03-02 13:26:59 +01:00
echo "No test was found. It is possible you configured without --with-gtest or you run it from wrong directory" >&2
exit 1
fi