----- On Mar 31, 2021, at 2:56 PM, lttng-dev lttng-dev@lists.lttng.org wrote:

> the following tests fails on arm64
> - test_event_vpid_tracker ust 0 "${EVENT_NAME}"
> - test_event_vpid_track_untrack ust 0 "${EVENT_NAME}"
> - test_event_pid_tracker ust 0 "${EVENT_NAME}"
> - test_event_pid_track_untrack ust 0 "${EVENT_NAME}"
>
> Signed-off-by: Anders Wallin <wallinux@gmail.com>
> ---
> tests/regression/tools/tracker/test_event_tracker | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/tests/regression/tools/tracker/test_event_tracker
> b/tests/regression/tools/tracker/test_event_tracker
> index 711690af..649c7e61 100755
> --- a/tests/regression/tools/tracker/test_event_tracker
> +++ b/tests/regression/tools/tracker/test_event_tracker
> @@ -5,7 +5,7 @@
> #
> # SPDX-License-Identifier: GPL-2.0-only
>
> -TEST_DESC="LTTng - Event traker test"
> +TEST_DESC="LTTng - Event tracker test"
>
> CURDIR=$(dirname "$0")/
> TESTDIR="$CURDIR/../../.."
> @@ -42,6 +42,8 @@ function prepare_ust_app
>
>       $TESTAPP_BIN -i $NR_ITER -w $NR_USEC_WAIT -a "$AFTER_FIRST_PATH" -b
>       "$BEFORE_LAST_PATH" &
>       CHILD_PID=$!
> +     # voluntary context switch to start $TESTAPP_BIN
> +     sleep 0.1

No, we have been bitten again and again by test issues hidden by sleeps in the
test code. Using sleeps for synchronization is flaky.

I don't know if we documented it, but we as maintainers are strongly against
anything that looks like a delay-based approach to fixing a race in the tests.
This typically just bury the race under the carpet and it shows up only in
specific conditions on the CI workers.

We need to add proper rendez-vous based synchronization to the test if some
is missing.

Adding Jérémie in CC.

Thanks,

Mathieu


> }
>
> function trace_ust_app
> --
> 2.31.1
>
> _______________________________________________
> lttng-dev mailing list
> lttng-dev@lists.lttng.org
> https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

--
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com