1117501 - [Bluetooth][FFOS7715 v2.1] When getting an incoming call, music comes out from speaker for several seconds before ringtone when user is using BT headset to listen to music

Status: RESOLVED WONTFIX

(Firefox OS Graveyard :: AudioChannel, defect)

Description

[ffos_st]

Description:
When getting an incoming call, music comes out from speaker for several seconds before ringtone when user is using BT headset to listen to music

Device:
SPRD 7715ga

Build indentifier:
20150104175050

Steps to reproduce:
1)Let your phone connected to wifi
2)Use BT headset to listen to music
3)Use another phone to call this test mobile phone

Actual Result:
Music comes out from speaker for several seconds before ringtone

Expected Result:
Music can be paused normally

Repro frequency:
10/10, 100%

Comment 1

[jingmei.zhang]

Attachment: log.log

Hi Wayne,

We have merged the patch in bug983539,mute other stream when in ringtone mode,but the issue still exist.

can you help to check the issue?

log is in attachment.it is about 12:44 when the issue occur.

Comment 2

[Wayne Chen [:xwaynec]]

Hi Jingmei,

I can't reproduce this problem by using Mozilla build at the following version.

Build ID               20150107100706
Gaia Revision          b04a8cb7b2482e0a44e6702b48c42283a00b5b1e
Gaia Date              2015-01-05 20:06:25
Gecko Revision         f6a4b637c4f59581ad64e4dfec3689442e98e268
Gecko Version          34.0
Device Name            scx15_sp7715ga
Firmware(Release)      4.4.2
Firmware(Incremental)  eng.waynechen.20150107.095743
Firmware Date          Wed Jan  7 09:57:52 CST 2015

I'll check log later.

Sincerely,
Wayne

Comment 3

[jingmei.zhang]

Hi Wayne,

we can reproduce the issue in flame at the following version:

OS version:2.1.0.0-prerelease
Platform version:34.0
Build Identifier:20150104161204

I get the above information from flame.and this is a 5/5 issue in our locale flame.

Thanks!

Comment 4

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

Hi Alastor, 
Could you help to check this?

Comment 5

[jingmei.zhang]

update version information:

Gaia-Rev        73be51f998031f06db0cd660c0e388fa621c9f4c
Gecko-Rev       https://hg.mozilla.org/releases/mozilla-b2g34_v2_1/rev/05dd053f1d90
Build-ID        20150104161204
Version         34.0
Device-Name     flame
FW-Release      4.4.2
FW-Incremental  39
FW-Date         Thu Oct 16 18:19:14 CST 2014
Bootloader      L1TC00011880

this is a 5/5 issue in above version.

Thanks!

Comment 6

[jingmei.zhang]

Hi Wayne,

Is there anything wrong in the log?

I have check the issue and find something improper:

01-07 16:54:30.667   123   838 E AudioPolicyManagerBase: JINGMEI mute stream:3 in setPhoneState  ---》state in ringtone and mute music
01-07 16:54:30.667   123   838 E AudioPolicyManagerBase: JINGMEI mute stream:3 In setStreamMute  ---》call setStreamMute mute music
01-07 16:54:30.667   123   838 E AudioPolicyManagerBase: checkAndSetVolume stream:3,volume:0.000000 ---》set music volume to 0

>it is strange device 2 is selected??

01-07 16:54:31.057   123   830 V AudioPolicyManagerBase: getNewDevice() selected device 2 --》device 2 is selected
01-07 16:54:31.057   123   830 E AudioPolicyManagerBase: checkAndSetVolume stream:3,volume:0.801217 ---》music volume is reset to normal

Is the patch exclude some corner cases?

Wayne,would you please help to give some suggestions?

Comment 7

[Wayne Chen [:xwaynec]]

hi Jingmei,

Please try to delete AudioPolicyManagerSPRD.cpp::startOutput functions

Comment 8

[jingmei.zhang]

(In reply to Wayne Chen [:xwaynec] from comment #7)
> hi Jingmei,
> 
> Please try to delete AudioPolicyManagerSPRD.cpp::startOutput functions

Hi Wayne,

AudioPolicyManagerSPRD.cpp::startOutput is used for SPRD hardware,we can't delete it……

Comment 9

[Wayne Chen [:xwaynec]]

Hi Alastor,

From log.log, I found something need your help.

I think in step 2, the ring maybe need to use AUDIO_STREAM_RING = 2 to start ringtone. But from log, I'm not sure why gaia or gecko use AUDIO_STREAM_MUSIC to start a ringtone. 

1. start ring
12-31 12:44:46.069   123   900 V AudioPolicyManagerBase: setPhoneState() state 1

2. start music or ringtone? 
12-31 12:44:46.269   123   900 D AudioPolicyManagerSPRD: startOutput() output 2, stream 3, session 12

3. pick phone up
12-31 12:44:53.089   123   123 V AudioPolicyManagerBase: setPhoneState() state 2


Sincerely,
Wayne

Comment 10

[Alastor Wu [:alwu]]

Hi, Wanyne,
Sorry for my late reply, I guess the reason that may results from the wrong usage/force?

Because I see that from the log.log.
"12-31 12:44:46.079 123 900 V AudioPolicyManagerBase: checkAndSetVolume() cannot set stream 6 volume with force use = 0 for comm"

Because now I still not very familiar with this part, I will look for it more details.

Comment 11

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

Add back ni? for tracking.

Comment 12

[jingmei.zhang]

Hi Wayne,

AudioPolicyManagerSPRD: startOutput() might be the root_cause of the issue,but no matter in a normal or abnormal case,AudioPolicyManagerSPRD: startOutput() is called. devices changes is the main difference between the two cases:

in a normal case:
AudioPolicyManagerSPRD: startOutput() is_voip_set 0,stream 3,
AudioPolicyManagerBase: setOutputDevice() output 2 device 0002 delayMs 0
>AudioPolicyManagerBase: setOutputDevice() prevDevice 0002, device 0002
AudioPolicyManagerBase: setOutputDevice() changing device 0002


in a abnormal case:
AudioPolicyManagerSPRD: startOutput() is_voip_set 0,stream 3,
AudioPolicyManagerBase: setOutputDevice() output 2 device 0002 delayMs 0
>AudioPolicyManagerBase: setOutputDevice() prevDevice 0001, device 0002
AudioPolicyManagerBase: setOutputDevice() changing device 0002


------>we can see from the setOutputDevice that AudioPolicyManagerBase::checkDeviceMuteStrategies is called:
 in normal case checkDeviceMuteStrategies returned without set volume
 in abnormal case,maybe checkDeviceMuteStrategies call setStrategyMute and the volume is reset.

Comment 13

[Alastor Wu [:alwu]]

I found out when I were not listening the music, the ringtone also came from the speaker instead of BT headset.
Even after I answered the phone, the sound still came from the speaker.

Comment 14

[jingmei.zhang]

(In reply to Alastor Wu [:alwu] from comment #13)
> I found out when I were not listening the music, the ringtone also came from
> the speaker instead of BT headset.
> Even after I answered the phone, the sound still came from the speaker.
Hi Alastor,
Do you mean the BT is useless in the process?? But BT works well in our locale:

when we were not listening the music, the ringtone came from both the speaker & BT headset.

Comment 15

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

(In reply to Alastor Wu [:alwu] from comment #13)
> I found out when I were not listening the music, the ringtone also came from
> the speaker instead of BT headset.
> Even after I answered the phone, the sound still came from the speaker.
Hi Alastor, 
How about your android phone?

Comment 16

[Wayne Chen [:xwaynec]]

Hi Jingmei,

I think you don't need to inherit SPRD::startOutput because AudioPolicyManagerBase has one.
From comment 7, I just wanna sure the problem maybe is in SPRD::startOutput. 
Because I can't reproduce in mozilla build.

Sincerely,
Wayne

Comment 17

[Alastor Wu [:alwu]]

I tested on three different phones, them all have the wrong audio routing problem.

[Test cases]
case 1 : listen music, then get a incoming call
case 2 : No doing anything, then get a incoming call

[Test phones]
1 : Flame-kk with the the gecko version 222539.
2 : Asus Zenfone with android 4.3
3 : HTC One 801e with android 4.4.3

[Test results]
1 : Sound came from the speaker on both test cases.
2 : Sound came from the speaker on both test cases.
3 : The same problem happened in the test case 1, but not always so.

Comment 18

[Alastor Wu [:alwu]]

In all test cases, phones have connected to BT headset.

Comment 19

[jingmei.zhang]

(In reply to Wayne Chen [:xwaynec] from comment #16)
Hi wayne,

> From comment 7, I just wanna sure the problem maybe is in SPRD::startOutput. 
I can not sure now,for this is not a 5/5 issue.and I will verify the issue in a later investigation.

> Because I can't reproduce in mozilla build.
I can reproduce the issue for 5/5 use the flame in our locale,and the build information is in comment5.
can you help to check the issue in above build?

> I think you don't need to inherit SPRD::startOutput because
> AudioPolicyManagerBase has one.
SPRD::startOutput is a necessary in SPRD.and we change nothing from 1.4 to 2.1.

Comment 20

[Alastor Wu [:alwu]]

I found that I got the wrong understanding of this question, so please ignore the comment13 & comment17.
Sorry about that.

Comment 21

[Alastor Wu [:alwu]]

Attachment: Temp patch for fixing this problem

Hi, Randy,

I think the root cause is that the timing issue of updating audio channel's visibility. 

There had a timing that both audio channels are visible on their state, although the music app was on the background, its visibility still was true. 

So I add a "dirty" condition to avoid this problem which was caused from the updating latency. 

I want to know if there have a better way to fix it? Or we need to wait for the completion of audio channel refactor? 

Thanks!

Comment 22

[jingmei.zhang]

Hi Alastor,

Issue still exist with the patch merged.

The issue has been workaround in bug983539,I am not sure if the workaround patch has been merged in flame.

In SPRD FFOS1.4,the issue is solved with the workaround patch.but in SPRD 2.1,issue occur with a low probability.

Comment 23

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

Hi jingmei, 
Does this issue still occur when you phase in bug 983539's patch on your 2.1 trunk?

Comment 24

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

(In reply to Alastor Wu [:alwu] from comment #21)
> Created attachment 8546611 [details] [diff] [review]
> Temp patch for fixing this problem
> 
> Hi, Randy,
> 
> I think the root cause is that the timing issue of updating audio channel's
> visibility. 
> 
> There had a timing that both audio channels are visible on their state,
> although the music app was on the background, its visibility still was true. 
> 
> So I add a "dirty" condition to avoid this problem which was caused from the
> updating latency. 
> 
> I want to know if there have a better way to fix it? Or we need to wait for
> the completion of audio channel refactor? 
> 
> Thanks!
To solve the visibility problem, I think the audiochannel refactor should fix it  (v3), but to meet the partner requirement, we should find some workaround to avoid those problem. 
BTW, 
As the patch in bug 983539, it would mute the other's stream volume when the ringtone comes.
Does the stream type is ring at that time?
==
public static final int STREAM_RING
Constant Value: 2 (0x00000002)

Comment 25

[Alastor Wu [:alwu]]

Hi, jingmei,
As you mentioned, so the problem is not occur with 100% probability ? 
I was confused about that the comment1 and comment5 says it always happened.
By the way, I could fixed this problem with my patch on the gecko version "222539:70de2960aa87".

Comment 26

[jingmei.zhang]

(In reply to Alastor Wu [:alwu] from comment #25)
> As you mentioned, so the problem is not occur with 100% probability ? 
> I was confused about that the comment1 and comment5 says it always happened.
Hi Alastor,
In flame,it is 100%.
But in SPRD device with the patch in bug 983539,this occur with a low probability.

Comment 27

[Alastor Wu [:alwu]]

Sorry, jingmei,
I tested my patch, it seems that problem wasn't always solved.
I would continue to check it again.

Comment 28

[jingmei.zhang]

(In reply to Randy Lin [:rlin] from comment #23)
> Hi jingmei, 
> Does this issue still occur when you phase in bug 983539's patch on your 2.1
> trunk?

Hi Randy,

Yes,with patch in bug 983539 merged,issue occur with a low probability.

I have discussed with the tester in SPRD,as probability is low,we are not see it as a blocker.

Comment 29

[Alastor Wu [:alwu]]

Attachment: Modify the visible state of audio channel

Hi, Randy,
I use this patch to implement that the audio channel is muted when the higher priority audio is coming. 

But after applying my patch, the problem would still exist, but happened not so often.

So I checked the "stream type" and "visible state" of the music. All of them is correct. It means that the audio track is created correctly by the stream's content.

Therefore, I guess the actual root causes may be like ... 
(1) Still exist the latency issue for pausing audio channels 
(2) The problems of AudioPolicyManager / AudioFliger.

Comment 30

[Alastor Wu [:alwu]]

Attachment: Modify the visible state of audio channel

Previous one is a wrong patch.

Comment 31

[Alastor Wu [:alwu]]

The word "correct" in comment29, it means that the audio agent has sent a pause to its media element. But the sound was still leaking from the speaker.

Comment 32

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

Hi Jingmei, 
Would like to test this? 
I think It looks fine for v2.2.
On v3, those workaround could be removed.

Comment 33

[jingmei.zhang]

Hi Alastor,

The patch in sprd doesn't work well when I remove the workaround in APMB.music will come out from speaker when the ringtone is run.

Comment 34

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

So this bug is testing base on removing the workaround in APMB?

Comment 35

[jingmei.zhang]

(In reply to Randy Lin [:rlin] from comment #34)
> So this bug is testing base on removing the workaround in APMB?
No,this bug is testing with the workaround in APMB.

for Alastor'patch,I test base on removing the workaround in APMB.

Comment 36

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

So base on the patch in bug 983539. 

+    if (state == AudioSystem::MODE_RINGTONE) {
+      for (int stream = 0; stream < AudioSystem::NUM_STREAM_TYPES; stream++) {
+        if (stream != AudioSystem::RING) {
+          setStreamMute(stream, true, mHardwareOutput);
+        }
+      }
The mHardwareOutput is on BT output or primary output? On this case.

Comment 37

[Wayne Chen [:xwaynec]]

Although bug 983539 only mute mHardwareoutput, but actually the code on sprd server is mute all mOutputs.
 
   if (state == AudioSystem::MODE_RINGTONE) {
      for (size_t i = 0; i < mOutputs.size(); i++) {
        for (int stream = 0; stream < AudioSystem::NUM_STREAM_TYPES; stream++) {
          if (stream != AudioSystem::RING) {
            setStreamMute(stream, true, mOutputs.keyAt(i));
          }
        }
      }
    } else if (oldState == AudioSystem::MODE_RINGTONE) {
      for (size_t i = 0; i < mOutputs.size(); i++) {
        for (int stream = 0; stream < AudioSystem::NUM_STREAM_TYPES; stream++) {
          if (stream != AudioSystem::RING) {
            setStreamMute(stream, false, mOutputs.keyAt(i));
          }
        }
      }
    }

Comment 38

[jingmei.zhang]

Hi Randy,

in setStreamMute:

in a normal case:
AudioPolicyManagerBase: setStreamMute() stream 3, mute 1, output 7, mMuteCount 0 device 0080

in a abnormal case:
AudioPolicyManagerBase: setStreamMute() stream 3, mute 1, output 2, mMuteCount 0 device 0001

Comment 39

[BMO Automation]

Firefox OS is not being worked on