Created attachment 106631 [details]
Trace of 'Prague' (system #1)

Created attachment 106632 [details]
Trace of 'Watermelons' (system #1)

Created attachment 106633 [details]
Trace of 'Prague' (system #2)

Created attachment 106634 [details]
Trace of 'Watermelons' (system #2)

Created attachment 106635 [details]
'pactl list' (system #1)

Created attachment 106636 [details]
'pactl list' (system #2)

The .pps format is something that LibreOffice would open, and play sound from. Is there a reason you believe this problem is related to PulseAudio?

(In reply to comment #7)
> The .pps format is something that LibreOffice would open, and play sound
> from. Is there a reason you believe this problem is related to PulseAudio?

Hi Arun,

1.  As you can see from the traces (also, if you believe my words :) LibreOffice plays sound through the active PulseAudio.

2.  As you can see from the traces (also, if you believe my words :) LibreOffice
plays BOTH files on Ubuntu (PA v4.0 and 64-bit system).  That leads me to believe that

2.1.  there's a difference (also, if you believe my words :) between v4.0 and v5.0 for this particular sound situation.

2.2.  this may be caused by PA (any version) different behavior on 64 vs 32 bits
as well.

As an aside, I can imagine that the two files differ as to how they "create" their respective sound (through LO, granted, but played by PA in the final analysis), while, no matter what, on Ubuntu (v4.0 on 64 bits) PulseAudio takes the "differences" in stride (as it should).

Thank you for you interest,
-- Alex

In the bad watermelons trace, there is no sign of LibreOffice even trying to connect to PulseAudio. So I'd throw this ball over to the LibreOffice ballpark to figure out the difference between "prague.pps" and "watermelons.pps".

(In reply to comment #9)
> In the bad watermelons trace, there is no sign of LibreOffice even trying to
> connect to PulseAudio.

Hi David,

That's odd.
Let me drill a little deeper.
Let's stay strictly with Ubuntu and watermelons.pps for the time being.

PRELIMINARY COMMENTS
I've been using Ubuntu 14.04 LTS downloaded/installed from the Ubuntu site
on a "virgin" partitiion (i.e., never had, worked with any other Ubuntus before).
In pristine condition, no modifications, no additions, etc.
Just keeping it up to date as per Ubuntu official Updates.
Just for my 64-bits "exposure".
-------------------------------------------------------------------------------
NEW TESTS (as stated in the OP, with PA not running by default, "autospawn=no")

Test #1.
PA not started, not running.
Attach. 'psax1.txt' (ps ax)
Start LibreOffice Impress ("LOI", from now on).
Open 'watermelons.pps' ("W", from now on).
NO SOUND (quiet as a mouse)
Attach. 'psax2.txt'
Stop LOI (after approx. 20 sec.).
Attach. 'psax22.txt'

Test #2.
Start PA (pulseaudio --start --log-level=4 --log-time=true \
 --log-target=newfile:palog1.txt)

The trace 'palog1.txt' starts.
Attach. psax3.txt
Start LOI
Open W
Attach. 'psax4.txt'
SOUND (from 'watermelons.pps' playing).  Good.
Stop LOI (after approx. 10 sec.).
Attach. 'psax5.txt'
Kill PA. (the trace, 'palog1.txt' closes)
Attach. 'palog1.txt'
Attach. 'psax6.txt'

FINAL COMMENTS
Please note again that WITHOUT PA running 'watermelons.pps' is SILENT.
WITH PA running 'watermelons.pps' SOUNDS.

I cannot explain how and why.
Can you?

Thanks,
-- Alex

Created attachment 106864 [details]
ps ax

Created attachment 106865 [details]
ps ax

Created attachment 106866 [details]
ps ax

Created attachment 106867 [details]
ps ax

Created attachment 106868 [details]
ps ax

Created attachment 106869 [details]
ps ax

Created attachment 106870 [details]
ps ax

Created attachment 106871 [details]
PulseAudio debug trace

Is the point of the test in comment 10 that it tries to prove that on Ubuntu both files play through PulseAudio? That was visible already in the initial logs.

The interesting question is why LibreOffice doesn't even connect to PulseAudio on the BLFS system when trying to play the "watermelons" file, but it connects fine with the "prague" file.

The logs on Ubuntu show that LibreOffice is having some trouble when trying to play the "prague" file: first GStreamer's probe stream fails (probably due to a PulseAudio bug that was fixed in 5.0), which causes LibreOffice to disconnect. Then LibreOffice connects again through the alsa plugin, but there's no trace of stream creation and LibreOffice disconnects again. Then LibreOffice connects for the third time, now using GStreamer again. The third attempt works. With the "watermelons" file it works on the first attempt. This is an interesting difference between the two files, but doesn't help much with figuring out why BLFS works differently (in that case "prague" works already on the first attempt, and as said before, "watermelons" results in zero connection attempts).

(In reply to comment #19)
> Is the point of the test in comment 10 that it tries to prove that on Ubuntu
> both files play through PulseAudio? That was visible already in the initial
> logs.

Hi Tanu,

This is really all water over the dam now, but just to justify that there was
some method to my madness, I felt David's Comment #9 was a little too general,
somehow implying that LO _never_ tries to use PA.
So my point in Comment 10 was to reiterate what was implied in the original "bug" submission that _same_ version LO, on a 32-bit system with PA 5.0 shows
inconsistencies with Ubuntu "perfect" situation.
BTW, why is Ubuntu's PA sticking with 1.4 and not upgrading to 5.0 (released way back in March)?

One point, maybe lost in the translation, is that on my BLFS 32-bit system
where practically everything else (including PA) is compiled from sources,
the LO is a "canned" binary so that it "exactly matches" (in version and
spirit (?)) the one on Ubuntu in an attempt to compare apples with apples.

Be that as it may, I see two courses of action now (with your permission):

1. Compiling LA (which is normally a matter of course for me ~ 80 min.), any
version you suggests (they are now at 4.3.2) and/or recompiling PA.
In that, any suggestions as to configure options will be highly appreciated
(like which GStreamer, 0.10.x or 1.x, etc.).
Test again on the BLFS _32-bit_ system.  As an aside, my system (still) works pretty smoothly on the current 3.16.3 kernel.

2. If it annoys you so much, drop the whole friggin' thing which, come to think of it, doesn't amount to a hill of beans in this crazy, cold and cruel world of today (and of mine).

Thank you for your comments and in general,
-- Alex

 > Please note again that WITHOUT PA running 'watermelons.pps' is SILENT.
 > WITH PA running 'watermelons.pps' SOUNDS.

So when you're NOT running PA, the result is silence. This can't be a problem with PA.

So far, I think everything points to that this is not a problem with PA, hence this bug tracker is not the right place to resolve your issue, as the people here are not LibreOffice experts.

(In reply to comment #21)
> So when you're NOT running PA, the result is silence. This can't be a
> problem with PA.
> 
> So far, I think everything points to that this is not a problem with PA,
> hence this bug tracker is not the right place to resolve your issue, as the
> people here are not LibreOffice experts.

Hi David,

I'm missing your logic.
Fact,

- On 64-bit Ubuntu system:
LO (v4.2.6.3) does NOT generate sound
 UNLESS PA (v4.0) is running.

- On 32-bit BLFS system:
LO (same v4.2.6.3) does NOT generate sound
 WHEN PA (v5.0) is running.
_As an aside_ (just to make the logic tighter),
LO (same v4.2.6.3) does NOT generate sound
 WHEN PA (v5.0) is NOT running.
 
This Bug submission is titled
"PA _v5.0_ does not play "pps" file (LibreOffice)"
on a _32-bit_ system and _same_ LO (v4.2.6.3).

------------------------------------------------------------
My _Ubuntu_ references have been just that, references for
comparison (and hopefully help in resolving this issue) only.

The three variables I see present/active for the Bug submission,
 32 bits (vs 64 bits)
 PA (v5.0 vs v4.0)
 _Possible_ LO different behavior (on 32 bits vs 64 bits)
 
may be in play, one, two or all three in combination.

Why you have so readily chosen to "logically" blame only
one of them, LibreOffice, is beyond me.

while you have shown no enthusiasm in answering my question (Comment #20):

Why has Ubuntu's PA been sticking with v1.4 [since March]?
(especially when Ubuntu makes a big point of using the latest
(and the greatest) versions for their packages.)

As I was very clear in my last Comment (#20), if you don't
want to handle this submission, feel free.  Fine with me.
Just say so.  Don't give a nonsensical explanation.
As I said, the aggravation is not worth for me.

-- Alex

(In reply to Alex from comment #22)
> (In reply to comment #21)
> > So when you're NOT running PA, the result is silence. This can't be a
> > problem with PA.
> > 
> > So far, I think everything points to that this is not a problem with PA,
> > hence this bug tracker is not the right place to resolve your issue, as the
> > people here are not LibreOffice experts.
> 
> Hi David,
> 
> I'm missing your logic.
> Fact,
> 
> - On 64-bit Ubuntu system:
> LO (v4.2.6.3) does NOT generate sound
>  UNLESS PA (v4.0) is running.
> 
> - On 32-bit BLFS system:
> LO (same v4.2.6.3) does NOT generate sound
>  WHEN PA (v5.0) is running.
> _As an aside_ (just to make the logic tighter),
> LO (same v4.2.6.3) does NOT generate sound
>  WHEN PA (v5.0) is NOT running.
>  
> This Bug submission is titled
> "PA _v5.0_ does not play "pps" file (LibreOffice)"
> on a _32-bit_ system and _same_ LO (v4.2.6.3).
> 
> ------------------------------------------------------------
> My _Ubuntu_ references have been just that, references for
> comparison (and hopefully help in resolving this issue) only.
> 
> The three variables I see present/active for the Bug submission,
>  32 bits (vs 64 bits)
>  PA (v5.0 vs v4.0)
>  _Possible_ LO different behavior (on 32 bits vs 64 bits)
>  
> may be in play, one, two or all three in combination.

You're missing the fact that all distributions (except possibly LFS?), add patches to their packages, mostly to fix bugs. E g, Ubuntu's LibreOffice contains ~50 patches on top of 4.2.6.3, as you can see here:

http://anonscm.debian.org/cgit/pkg-openoffice/libreoffice.git/tree/patches?h=ubuntu-trusty-4.2

> Why you have so readily chosen to "logically" blame only
> one of them, LibreOffice, is beyond me.

I did so because the relevant log showed that LibreOffice did not try to connect to PulseAudio. Which does not rule out PulseAudio completely, just saying that it is more likely that LibreOffice is at fault here. 

Also, "pulsesink probe", as I saw in another log, sounds very gstreamer-ish. If LibreOffice uses the gstreamer library to play back audio through Pulseaudio, then GStreamer could also be at fault here.

> while you have shown no enthusiasm in answering my question (Comment #20):
> 
> Why has Ubuntu's PA been sticking with v1.4 [since March]?

v1.4 of what?

> (especially when Ubuntu makes a big point of using the latest
> (and the greatest) versions for their packages.)

I'd say that Ubuntu's policy has shifted a bit in recent years towards more stability and less latest-and-greatest, but that is not really on topic here, either.

Hi David, 
> > Why has Ubuntu's PA been sticking with v1.4 [since March]?
> 
> v1.4 of what?
> 
Typo.  Should've been pretty obvious from the whole thread where Ubuntu was always mentioned as PA 4.0 (vs. BLFS 5.0).  If it was misleading, I apologize:
a typo is an error after all.  This is a fast world and you cannot expect someone to translate a sentence like

"Why has Ubuntu's PA been sticking with v1.4 [since March]?"
to 
"Why has Ubuntu's PA been sticking with v4.0 [since March]?"

That someone would've been required to
1. Figure out I had been talking about v4.0 on Ubuntu all along.
2. Realize that there was no PA 1.4 release in March (at least this year).
3. Realize that in March, PA released v4.0.
Expecting way too much these days.

Anyway, like what you said about LO, Ubuntu PA 4.0 may be full of patches, making it like a PA 4.0 on steroids :)
So what do I know ...

-- Alex

(In reply to Alex from comment #24)
> 3. Realize that in March, PA released v4.0.

Should be
3. Realize that in March, PA released v5.0.

Another typo!

About 4.0 in Ubuntu 14.04 and Ubuntu 14.10:

 * 5.0 was released too late to be included in Ubuntu 14.04, given that we want stability, and even more so in the LTS.

 * 5.0 was not released into Ubuntu 14.10 because
 1) I worked part time, and I didn't feel I had the time to fix the potential regressions that could come out of releasing a new version into 14.10.
 2) There were relatively few news anyway - 5.0 and 4.0 aren't that much different, really.

My plan is to move from 4.0 to 6.0 in Ubuntu 15.04, provided that 6.0 gets released reasonably soon.

This bug is going off-topic, and probably it's not PA's fault.

If you have logs from the LO/gstreamer side that shows that it is trying to connect to PA, but the PA side does not show any connection, then let us know, because then the fault could be in PA's client library. If not, let's close this bug.

Let's close it.
Thanks,
-- Alex