[pulseaudio-discuss] Hpw to start the pa server

Wed Apr 14 12:47:14 PDT 2010

'Twas brillig, and Gene Heskett at 14/04/10 17:30 did gyre and gimble:
> On Wednesday 14 April 2010, Colin Guthrie wrote:
>> 'Twas brillig, and Gene Heskett at 14/04/10 16:38 did gyre and gimble:
>>> On Wednesday 14 April 2010, Colin Guthrie wrote:
>>>> 'Twas brillig, and Gene Heskett at 13/04/10 01:38 did gyre and gimble:
>>>>> On Monday 12 April 2010, Colin Guthrie wrote:
>>>>>> 'Twas brillig, and Gene Heskett at 12/04/10 20:29 did gyre and gimble:
>>>>>>> draksound, re-enabled, was on before by other means, and enabled user
>>>>>>> switching.  No sound yet.
>>>>>>> pacmd.ls.out attached.
>>>>>>
>>>>>> Hmm, what is strange here is that it shows no streams at all - e.g. no
>>>>>> sink inputs. To back this up, all the sinks are in a suspended state.
>>>>>>
>>>>>> The volumes are all incredibly high - well over the 100% mark. How
>>>>>> were these volumes set? Most tools only allow volumes up to 150%.
>>>>>
>>>>> I _think_ it was paprefs that allowed me to go as high as 400%.
>>>>
>>>> Ahh no, I think it was likely paman... it's evil :p Generally speaking
>>>> paman and padevchooser are obsolete... I should probably not ship them
>>>> really but some people do like them despite their evilness :p
>>>
>>> paman runs here, padevchooser doesn't output anything, shows a .4%
>>> memory usage, and responds to a ctrl+c to quit.
>>
>> It's an applet that sits in the system tray and shows popups and a
>> submenu and stores it's settings in a strange way that conflict with
>> normal usage. Don't use it :p
> So thats why the system tray cleaned itself up when I killed the last 
> padevchooser.
> 
>>>> Can you do the following for me:
>>>>
>>>> 1) Enable PA in draksound and reboot.
>>>> 2) Login.
>>>> 3) ps aux | grep pulseaudio
>>>
>>> [gene at coyote gene]$ ps aux | grep pulseaudio
>>> root     17772  0.0  0.0 206580  2264 ?        S<sl Apr12   0:00
>>> /usr/bin/pulseaudio --start --log-target=syslog gene     22513  0.0  0.1
>>> 221408  4904 ?        S<sl Apr13   0:13 /usr/bin/pulseaudio --start
>>> --log-target=syslog gene     28277  0.0  0.0   7372   948 pts/6    R+  
>>> 10:28   0:00 grep --color pulseaudio
>>
>> OK, the version running as root is probably cocking things up here. Can
>> you find out why it was started and kill it if possible? You'll want to
>> do this before logging in as your own user.
> 
> Without a reboot, the root session has gone away, didn't change anything 
> though.
>>
>>>> 4) paplay /usr/share/sounds/ia_ora-startup.wav
>>>
>>> Silence, does take a while to get the prompt back
>>
>> I suspect the silence is due to the fact that you are now running your
>> own PA daemon, but the root users own PA daemon is also running, hogging
>> the sound card and not letting your user access it at the same time.
>> Running PA as root is a generally bad idea so try to avoid it at all costs.
> 
> See above, no root session now exists.
> 
>>>> 5) (in a separate terminal, leave running and retry until it starts
>>>> properly): pulseaudio -k; pulseaudio -vvvvv
>>>
>>> All I can get, tried 30-40 times, is
>>> [gene at coyote gene]$ pulseaudio -k; pulseaudio -vvvvv
>>> I: main.c: setrlimit(RLIMIT_NICE, (31, 31)) failed: Operation not
>>> permitted
>>>
>>> E: pid.c: Daemon already running.
>>> E: main.c: pa_pid_file_create() failed.
>>>
>>> However, executing the above line, does report a connection lost for the
>>> line below if its been run. Expected...
>>
>> Hmm, OK, it seems to autospawn far too quickly for you (machine quicker
>> than mine :D).
> 
> 2.1Ghz quad core phenom, 4Gb of dram.
> 
>>>> 6) paplay /usr/share/sounds/ia_ora-startup.wav
>>>
>>> [gene at coyote gene]$ time paplay /usr/share/sounds/ia_ora-startup.wav
>>> 0.03user 0.00system 0:07.75elapsed 0%CPU (0avgtext+0avgdata
>>> 0maxresident)k 0inputs+0outputs (0major+869minor)pagefaults 0swaps
>>>
>>>> 7) cat ~/.pulse/client.conf
>>>
>>> [gene at coyote gene]$ cat ./pulse/client.conf
>>> cat: ./pulse/client.conf: No such file or directory
>>
>> You've got a typo. I said "cat ~/.pulse/client.conf" type exactly that.
> Copy/paste from your line ;)
> [gene at coyote gene]$ cat ~/.pulse/client.conf
> cat: /home/gene/.pulse/client.conf: No such file or directory
> both times. ;)
> 
>>>> 8) cat /etc/pulse/client.conf
>>>
>>> [gene at coyote gene]$ cat /etc/pulse/client.conf
>>
>> Cool. That's what I expect it to be. Ticked off the list :)
>>
>>>> 9) xprop -root | grep PULSE
>>>
>>> [gene at coyote gene]$  xprop -root | grep PULSE
>>> [gene at coyote gene]$
>>
>> Thanks. With a normal clean startup process this should contain
>> something but running pulseaudio -k or padevchooser can mess it up.
>> Being empty is good tho' and should still work fine.
>>
>>>> 10) env | grep PULSE
>>>
>>> [gene at coyote gene]$ env | grep PULSE
>>> [gene at coyote gene]$
>>
>> As above. I didn't expect anything to be here, but worth double checking.
>>
>>> However,
>>> [root at coyote Daily]# env |grep pulse
>>> CANBERRA_DRIVER=pulse
>>> [root at coyote Daily]#
>>
>> Yeah that's fine (this is actually one of the reasons that PA is
>> autospawned so quickly above :D)
>>
>>>> Just to keep things simple, I'd load up paprefs and untick the box to
>>>> create a combined output (it's the last tab IIRC). Then the above sink
>>>> wont load which keeps our setup cleaner.
>>>
>>> Already did, that is the condition for all of the above.
>>>
>>> And I just tried the "pulseaudio -k;pulseaudio -vvvvv" about 50 more
>>> times. Same instantly respawned result every time.  Where do I disable
>>> the auto-respawn?
>>
>> OK, the best bet here is to:
>> cp /etc/pulse/client.conf ~/.pulse/client.conf
>>
>> then edit the latter file and change so "autospawn = no"
>>
>> This will allow easier debug :)
>>
>>> I just fired up mcc, went to the screen for audio and ran everything
>>> there, getting the expected results except for the last one:
>>> [gene at coyote ~]$ /sbin/fuser -v /dev/dsp
>>> [gene at coyote ~]$
>>> but at this point I have NDI what that means.
>>
>> That's fine, /dev/dsp is a legacy device node and not much should have
>> it open anyway. 99% of apps use either alsa or pulse directly.
>>
>>> Also, doing that while paplay is running is equally uninformative.
>>> But it seems to me we aren't using, or do not have, a tool that will
>>> trace the paths being used.
>>
>> OK, so ultimately I think the next round of debug relates to that root
>> process.
>>
>> In my previous list of numbered steps can you add:
>>
>> -1) Ensure no root PA daemon is started prior to testing. e.g. do a
>> fresh reboot, login and then do ps aux | grep pulseaudio. If there is a
>> root process running, just stop right there and let me know. THis is the
>> problem and *something* is running on your machine as root that is
>> trying to produce sound... this is a bad thing and needs to be solved.
>>
>> 0) If you just have your own user's PA running, then run the steps
>> again. You can skip #8 tho' :)
>>
>> 1) as before
>> 2)    "
>> ...)
>>
>>
>> All the best
>>
>> Col
>>
> After turning off autospawn, the screen scrape can be seen at
> <http://gene.homelinux.net:85/gene/pulseaudio-vvvvv.out>
> Lots of 'trailing' white space.  Do we have a tool that will trim that?

You can pipe it through "trim", but it's really the terminal that puts
the spaces in. If your term supports a save backlog feature then it
should save without the spaces. Or you can just run:

pulseaudio -k; pulseaudio -vvvv 2>&1 | tee pulse.log

Which will save it to pulse.log.

I forgot to ask, but can you now give "pacmd ls" output now that the
server is running?

Oh and I'd recommend you actually clear out the files in ~/.pulse/. Your
volumes are set really really high and I reckon that when things kick in
and start working, you'll likely get deafened!

Cheers.

Col

-- 

Colin Guthrie
gmane(at)colin.guthr.ie
http://colin.guthr.ie/

Day Job:
  Tribalogic Limited [http://www.tribalogic.net/]
Open Source:
  Mandriva Linux Contributor [http://www.mandriva.com/]
  PulseAudio Hacker [http://www.pulseaudio.org/]
  Trac Hacker [http://trac.edgewall.org/]