using saft-ctl from the command line causes large delayes in another application

[dietrichb]

Background.

An application using ECAs software action channel

• receives timing messages form the data master
• injects timing messages to the local ECA that are received by the lm32 channel.

This works fine.

However, starting saft-ctl at the command line causes messages in the lm32 channel to be late by typically 45ms.

Inspecting the software action queue (saft-ctl -s) does yield 'delayed' messages on the software action queue but no 'late' messages.
It is not possible to inspect the eCPU channel using saft-ecpu-ctl

Either the messages are received with a delay (!) of ~45 ms at the application, or the message injected by the application arrives late at the input of the ECA.

This only happens while saft-ctl starts (at the command line). Once saft-ctl has been started, everything works fine again.

This is unexpected as a similar issue has already been fixed [1].

The relevant frontend runs saftlib [2] on yocto [3]

[1] #119

[2] me@host:~# saft-ctl tr0 -i
saftlib source version : saftlib 3.1.3 (v3.1.3-dirty): Jul 25 2024 09:41:36
saftlib build info : built by jenkins on Jul 25 2024 09:41:36 with vmlb030 running Rocky Linux

[3] me@host:# uname -a
Linux host 5.19.17-rt10-yocto-preempt-rt #1 SMP PREEMPT_RT Mon Jan 2 02:58:07 UTC 2023 x86_64 GNU/Linux
met@host:# cat /etc/os-release
ID=ffos
NAME="ffos (Fair Frontend OS)"
VERSION="4.1.1 (langdale)"
VERSION_ID=4.1.1
PRETTY_NAME="ffos (Fair Frontend OS) 4.1.1 (langdale)"
DISTRO_CODENAME="langdale"