Hi Niklas, On Sunday, 20 May 2018 13:47:53 EEST Niklas Söderlund wrote: > On 2018-05-19 23:34:25 +0300, Laurent Pinchart wrote: > > It can be useful to capture kernel log messages in test log files for > > diagnostic purpose. Add a simple mechanism to do so by capturing the > > full kernel log at the end of the test. The kernel log is cleared first > > before starting the test to avoid capturing unrelated messages. > > > > Signed-off-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx> > > --- > > > > scripts/vsp-lib.sh | 5 +++++ > > 1 file changed, 5 insertions(+) > > > > diff --git a/scripts/vsp-lib.sh b/scripts/vsp-lib.sh > > index 0f3992a7827e..e672686a377e 100755 > > --- a/scripts/vsp-lib.sh > > +++ b/scripts/vsp-lib.sh > > @@ -1075,6 +1075,9 @@ test_init() { > > > > echo "Using device $mdev ($dev)" | ./logger.sh config >> $logfile > > > > vsp_runner=./vsp-runner.sh > > > > + > > + # Clear the kernel log > > + dmesg -c > /dev/null > > I agree it's useful to capture the kernel log messages during a test, I > did the same thing in vin-tests. What I did learn was that when > something do go wrong you kind of want the whole kernel log from boot > and truncating it could be annoying. Just the other day I got feed up > with this in vin-tests and found a different way, as it's still fresh in > my memory maybe it could be useful for you too, or not :-) > > marker=$(dmesg | tail -n 1 | sed 's/^\[\([^]]*\)\].*/\1/g') > > > } > > > > test_start() { > > > > @@ -1086,6 +1089,8 @@ test_complete() { > > > > echo "Done: $1" | ./logger.sh >> $logfile > > echo $1 >&2 > > > > + dmesg -c | ./logger.sh kernel >> $logfile > > + > > dmesg | sed "1,/$marker/d" | ./logger.sh kernel >> $logfile That's a good idea. I gave it a try and it worked, I'll send a v2. > > rm -f ${frames_dir}frame-*.bin > > rm -f ${frames_dir}histo-*.bin > > rm -f ${frames_dir}rpf.*.bin -- Regards, Laurent Pinchart