ALSA reset

[golowanow]

Reset ALSA settings with alsactl init and alsactl restore at set_alsa() before it calls set_alsa_settings which runs HW model-specific configuration scripts.

This way a test case can expect that the set_alsa call will apply the DUT's default ALSA settings as the baseline configuration to avoid side-effects possible after the previous tests execution (either passed, or failed) as well as other unexpected changes.

[marc-hb]

Why not just:

Suggested change

	[[ "${rc}" -ne 0 ]] && dlogw "alsactl init error=${rc}"
	alsactl -d init ... || dlogw "alsactl init error=$?"

[golowanow]

dlogw clears $?, so it looks like

ls /foo || dloge "error=$?"

ls: cannot access '/foo': No such file or directory
2025-07-23 18:44:58 UTC [ERROR] error=0

vs.
ls /foo || rc=$? ; dloge "error=$rc"

ls: cannot access '/foo': No such file or directory
2025-07-23 18:43:36 UTC [ERROR] error=2

[marc-hb]

OMG, they are aliases! Could you try to submit this fix that converts aliases to proper functions? I tested this change in isolation and it fixes the issue but I don't have time to submit it to CI.

Don't take the second diff hunk, it's just a hack to test locally.

--- a/case-lib/logging_ctl.sh
+++ b/case-lib/logging_ctl.sh
@@ -44,14 +44,15 @@ _func_log_cmd()
 
     for key in "${!LOG_LIST[@]}";
     do
-        # dymaic alias the command with different content
-        cmd="alias $key='echo \$(date -u \"+%F %T %Z\")$ext_message ${LOG_LIST[$key]}'"
+        # shellcheck disable=SC2016
+        local date_cmd='$(date -u "+%F %T %Z")'
+        local cmd="$key() { printf '%s%s %s\n' \"$date_cmd\" '$ext_message ${LOG_LIST[$key]}' \"\$*\" ; }"
         eval "$cmd"
     done
 }
 
 # without setting up the LOG_ROOT keyword, now create the log directory for it
-_func_log_directory()
+_Xfunc_log_directory()
 {
     # LOG_ROOT is the top of a single test run. SOF_LOGS_BASE is higher
     # up and for all tests and all runs.

source case-lib/logging_ctl.sh 
( exit 5 ) || dlogw "hello $?"
2025-07-23 20:01:08 UTC [WARNING] hello 5

[golowanow]

OMG, they are aliases! Could you try to submit this fix that converts aliases to proper functions?

• logging_ctl.sh: change log aliases to functions #1286

[marc-hb]

#1286 just merged, now this could be fixed.

At least it won't be needed elsewhere now...

[golowanow]

SOFCI TEST

[golowanow]

SOFCI TEST

[golowanow]

dear reviewers, do you have comments or objections ?
next is #1270 adjusted, to follow

[kv2019i]

Looks good @golowanow !

[lyakh]

I suppose you could just rc=$? unconditionally after the systemctl call and remove the initialisation of rc

[golowanow]

right, its my intent to deal with rc explicitly and to tie command calls and rc set, so it is declared as a local variable at this function with init value and, despite any change in the following command calls, it will have the default value.

With command || rc=$? the script works the same way if it runs with bash -e - continue its execution.

[marc-hb]

You could probably do this but && has many pitfalls and is best avoided

systemctl --no-pager status alsa-restore.service > "${alsa_log}" 2>&1 && local rc=$?

@lyakh check the && pitfalls in #312

[lyakh]

if for whatever reason dlogw exits with an error, rc won't be reset. I'd just use if here for determinism. More of these below, I'd be more explicit with rc

[golowanow]

when dlogw fails (something really bad ? hopefully with some stderr output) the rc should not reset to zero as well.

[majunkier]

LGTM