package: lmbench
version: 3.0-a9-1
Bug#496427 introduced a regression in lat_proc benchmark program.
/usr/lib/lmbench/bin/i686-pc-linux-gnu/lat_proc tries to measure latency of
executing /var/tmp/lmbench/hello . However, since path of hello program in
/usr/lib/lmbench/bin/i686-pc-linux-gnu/lmbench was not updated from /tmp/hello
to /var/tmp/lmbench/hello accordingly, lat_proc program tries to execute
non-existent hello program.
if [ X$BENCHMARK_OS = XYES -o X$BENCHMARK_PROC = XYES ]; then
cp hello /tmp/hello
for i in fork exec shell
do lat_proc -P $SYNC_MAX $i
done
rm -f /tmp/hello
fi
As a result, latency reported by
/usr/lib/lmbench/bin/i686-pc-linux-gnu/lat_proc does not include latency of
executing /var/tmp/lmbench/hello .
I don't know whether Bug#496427 introduced more regressions or not.
I think we should carefully review any /tmp usage in
/usr/lib/lmbench/bin/i686-pc-linux-gnu/lmbench .
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
