1944812 - Android debug-isolated-process TV jobs blocked looping on "W AudioSystem: AudioFlinger not published, waiting..."

Status: RESOLVED DUPLICATE of bug 1709948

(Testing :: General, defect, P2)

Description

[Andreas Pehrson [:pehrsons]]

Encountered after a couple of recent landings, that TV jobs were failing with the below pattern in logcat (see bug 1944573 and bug 1944575), only on debug-isolated-process. I looked at some occurences of Bug 1861710 on debug-isolated-process and they showed the same symptoms too.

I see this on repeat in logcat just after the test started:

01-29 13:39:21.059  3558  3572 W AudioSystem: AudioFlinger not published, waiting...
01-29 13:39:21.560  3558  3572 I ServiceManager: Waiting for service media.audio_flinger...
01-29 13:39:22.561  3558  3572 I ServiceManager: Waiting for service media.audio_flinger...
01-29 13:39:23.562  3558  3572 I ServiceManager: Waiting for service media.audio_flinger...
01-29 13:39:24.563  3558  3572 I ServiceManager: Waiting for service media.audio_flinger...
01-29 13:39:25.564  3558  3572 I ServiceManager: Waiting for service media.audio_flinger...

I see the mochitest start but early on it seems to get blocked, and these lines appear. Unclear exactly what is blocked.

I found bug 713301 comment 24 that fixed an eerily similar issue back in gonk. But in the emulator log for this failure I see [ 3.882493] init: Starting service 'servicemanager'....

Could the ServiceManager process have gotten hosed? Not sure where this should go, let's start with Testing.

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

The severity field is not set for this bug.
:jmaher, could you have a look please?

For more information, please visit BugBot documentation.

Comment 2

[Joel Maher ( :jmaher ) (UTC -8)]

I am not sure what specifically to do here. Test verify repeats a test in 2 scenarios:

1. in the same browser session
2. as a solo test restarting the browser each time

if I could get reference to a logcat (or a few of them) I could determine which scenario- it is possible if this is scenario 2 where we restart the browser each time that the AudioFlinger is not loaded up or registered fully- in that case the test case would need to wait before starting.

Comment 3

[Andreas Pehrson [:pehrsons]]

The lastest job on bug 1861710 (from nov) is here: https://firefoxci.taskcluster-artifacts.net/c69f1rbgRECip6JD60ggiQ/0/public/test_info/logcat-emulator-5554.log
It shows the AudioFlinger issue right after the first test_start.

The reported job on bug 1944573 (jan 29) is https://firefoxci.taskcluster-artifacts.net/N6TCGzEFSdmuN3XHp3qx6w/0/public/test_info/logcat-emulator-5554.log
Same pattern there. I'll note the first "AudioFlinger not published" warning is ~50s after 01-29 12:56:18.775 1373 1385 I AudioFlinger: systemReady.

Comment 4

[Joel Maher ( :jmaher ) (UTC -8)]

thank you, I was looking in the test log, not the logcat.

So process of elimination says:

• not webrtc test specific - this passes on other android
• not test verify specific - this passes on other android and other platforms
• not android emulator specific - this passes on other android

this leave debug-isolated-process. The key here is to figure out what is different with debug isolated process and if there is something different with AudioFlinger or whatever depends on that.

I am not sure what bugzilla component to put this in- is there one for debug-isolated-process?

Comment 5

[Andreas Pehrson [:pehrsons]]

I asked around and dug around. This seems to be covered by bug 1709948.