suspend/audio_before_suspend always fails: pa_context_connect() failed: Connection refused

[Leon Liao]

The suspend/audio_before_suspend test case always fails.
Looks there are some thing wrong in pulseaudio.
But, we run this test case on different machines and got the same result.

[I/O log]
Connection failure: Connection refused
pa_context_connect() failed: Connection refused
Traceback (most recent call last):
  File "/tmp/nest-j81t0lps.3361f09f54625130ca8353599d2d0b04ec843608119aba04070f3a0948f151e7/audio_settings.py", line 9, in <module>
    sys.exit(main())
  File "/usr/lib/python3/dist-packages/checkbox_support/scripts/audio_settings.py", line 368, in main
    store_audio_settings(settings_file)
  File "/usr/lib/python3/dist-packages/checkbox_support/scripts/audio_settings.py", line 218, in store_audio_settings
    pactl_status = check_output(["pactl", "info"],
  File "/usr/lib/python3.8/subprocess.py", line 411, in check_output
    return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
  File "/usr/lib/python3.8/subprocess.py", line 512, in run
    raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['pactl', 'info']' returned non-zero exit status 1.

[Yuan-Chen Cheng (ycheng-twn)]

per check, this failed on the I+N machine but not the UMA machine.

Per further investigation, the window on the screen does not have the desktop DBUS session. I think it can't access the pulseaudio session, too.

I suspect that's the root cause.

[Alex Tu (alextu)]

more information followed #1,
The `pactl info` failed on the terminal opened by checkbox.

To manually open a new terminal then `pactl info` works on the new terminal.
The `pactl info` failure could be caused by the terminal environment created by the checkbox.

[Jonathan Cave (jocave)]

I'm sure I encountered this problem while adding ODM Cert testing to the CDTS snap. I think it required some additional environment setup, I'll see if I can dig out the info on what I did.

[Sylvain Pineau (sylvain-pineau)]

did you run checkbox as root?

[Sylvain Pineau (sylvain-pineau)]

Please redo the same test using checkbox remote

[Jonathan Cave (jocave)]

This does indeed depend on how checkbox was run and the environment setup. I believe this was the change I made to ensure it works in checkbox snaps for classic systems: https://code.launchpad.net/~jocave/plainbox-provider-checkbox/+git/plainbox-provider-checkbox/+merge/388884

Correct setup of PULSE_RUNTIME_PATH is key.

[Lukas Waymann (meribold)]

I've also encountered this during SRU testing. E.g.:

http://10.101.50.238:8080/job/cert-stock-sru-focal-desktop-inspiron-7591-c2-26881/1/testReport/junit/com.canonical/certification__suspend_audio_before_suspend/com_canonical_certification__suspend_audio_before_suspend/

http://10.101.50.238:8080/job/cert-stock-sru-focal-desktop-precision-5550-c5/9/testReport/junit/com.canonical/certification__suspend_audio_before_suspend/com_canonical_certification__suspend_audio_before_suspend/

[Jonathan Cave (jocave)]

As SRU is run using Checkbox Remote and these are "normal user" jobs, I believe they will be susceptible to the problems surrounding setup of the environment for jobs as described in this card: https://trello.com/c/k15lSgfj

As a result, I expect that things like lack of setup of variables like PULSE_RUNTIME_PATH is to blame.

[Lukas Waymann (meribold)]

Here's another example: <http://10.101.50.238:8080/job/cert-stock-sru-focal-desktop-xps13-7390-2in1/26/testReport/junit/com.canonical/certification__suspend_audio_before_suspend/com_canonical_certification__suspend_audio_before_suspend/>.

[Sylvain Pineau (sylvain-pineau)]

Fixed after January stable deb release.

See <Cert_Jenkins_URL>/job/cert-stock-sru-focal-desktop-inspiron-7591-c2-26881/6/testReport/junit/com.canonical/certification__suspend_audio_before_suspend/

[jeremyszu (os369510)]

$ apt policy plainbox-provider-checkbox
plainbox-provider-checkbox:
  Installed: 0.57.0+git202103031045+pkg160~ubuntu20.04.1
  Candidate: 0.57.0+git202103031045+pkg160~ubuntu20.04.1
  Version table:
 *** 0.57.0+git202103031045+pkg160~ubuntu20.04.1 500
        500 http://ppa.launchpad.net/checkbox-dev/ppa/ubuntu focal/main amd64 Packages
        100 /var/lib/dpkg/status

Still can reproduce this issue.

[jeremyszu (os369510)]

@Sylvain,

Hi, I try it with version "0.57.0+git202103031045+pkg160~ubuntu20.04.1", still get the same error. Would you please help to take a look?

[Yao Wei (medicalwei)]

This issue seems to be that the terminal session cannot use pulseaudio session due to wrong $XDG_RUNTIME_DIR that happens after a reboot due to GPU switching.

p-p-c maintainers say that using checkbox remote would avoid the problem and they recommend us to migrate:
https://code.launchpad.net/~medicalwei/plainbox-provider-checkbox/+git/plainbox-provider-checkbox/+merge/405995

Tested that running the following test cases in order:
- com.canonical.certification::graphics/1_auto_switch_card_UHD_Graphics_630__Mobile_
- com.canonical.certification::suspend/audio_after_suspend_auto

This would fail on local and pass on remote.

Closing this as wontfix and we should use checkbox remote for testing.

Also note that Sylvain Pineau says they are killing pm_test.py. We might need to update our script that uses pm_test.py for poweroff and reboot.

[jeremyszu (os369510)]

According to comment#13, SWE needs to prevent to use local sanity since it's not longer support.
It's confirmed by AlexTu and Sylvain.
Cert team will retire local sanity and local sanity will not longer maintain.