Closed nfraprado closed 4 weeks ago
Is it works fine and improve things? If it is tested, i can merge it.
@nfraprado I just caught an instance of input overrun
on corsola
as well, causing a false regression being reported in the watchdog-reset test: https://lava.collabora.dev/scheduler/job/14402618#L2265 (KernelCI node: https://staging.kernelci.org:9000/viewer?node_id=6670b75ed0b7694b3998a7ea)
It'd be great to have the delay added for corsola
too in this PR, so we can confirm that it fixes the watchdog-reset issue. Alternatively, I can open a separate PR to tackle this.
@laura-nao good catch! I've added corsola to the PR now.
Merging, as it seems improve results.
Add test_character_delay to the Spherion and Tomato platforms to workaround the fact that they're sometimes unable to process serial input fast enough, resulting in mangled commands and consequently flaky test results, as described in https://github.com/kernelci/kernelci-project/issues/366.
The right place to do this change would be in the device-type template as described in LAVA's documentation [1]. This overriding in KernelCI is meant only as a temporary workaround to verify whether this fixes the issue. If it does, then we'll do it in LAVA upstream instead.
[1] https://docs.lavasoftware.org/lava/debugging.html#differences-in-input-speeds Signed-off-by: Nícolas F. R. A. Prado nfraprado@collabora.com