What a pity

[James Linder]

> On 3 Jan 2022, at 7:32 am, Brad Hards via gstreamer-devel <gstreamer-devel at lists.freedesktop.org> wrote:
> 
> On Monday, 3 January 2022 10:16:31 AM AEDT James via gstreamer-devel wrote:
>> gstreamer seems very nice in concept. The fact that I've been trying for 3
>> months and can get no help is a big deterent.
> This email (especially the title) comes across as passive-aggressive, which 
> isn't pleasant. 
> 
> That said, perhaps the problem is that you aren't asking for support in the 
> right way. Possibly reading and applying
> http://www.catb.org/~esr/faqs/smart-questions.html would help.  
> 
> In your case, simplifying the problem (e.g. removing parts of the pipeline 
> that aren't required to reproduce the issue) might help. Saying what you've 
> tried to do to debug the issue might help. E.g. does it still stutter with 
> just the video parts, or just the pulsesrc parts.
> 
> If you still need support, and can't get any for free, perhaps you could 
> consider a commercial provider. I'm not associated with any of them, and 
> haven't tried any of them, but free software doesn't mean no-cost support.

This single thread generated more responses than all my queries over months. Thank you.

Since I have 30 or 40 ‘try’ scripts I’ll just sumarize:

Recoding pulsesrc is perfect.
Recording or streaming v4l2 is perfect.
A 2 core i3 runs 60% ish busy
A 4 core i5 at 30% ish
A 4 core i7 at 25% ish
RT prior is worse 
Gstreamer and threads (as pulseaudio is) at nice -11 is slightly worse. (normally 0! since I’m using m2 flash I think disc io is not relevant but the hlssink2 stores files in ram disk!)
Recording and streaming audiotestsrc and v4l2 is perfect.
But FILES are not broken, FRAMES are dropped when stutter happens.
htop and nmon show usage peak under 50% BUT the clock often gets boosted. Not showing the cpu being stretched !

Judicious use of queues makes little difference, this helps recording:

vt. ! h264parse ! queue max-size-buffers=0 max-size-bytes=0 max-size-time=1000000000 ! \
mux. \

I’m relutant to spend nearly $1000 on an 8 core rysen to prove that it does not work. Has anybody done that?

Peter off-list mail was a fumble, sorry.

> In your case, simplifying the problem (e.g. removing parts of the pipeline 
> that aren't required to reproduce the issue) might help. Saying what you've 
> tried to do to debug the issue might help. E.g. does it still stutter with 
> just the video parts, or just the pulsesrc parts.

I’ve tried taking the redunt tee’s out. Logic says that makes no difference and indeed it does not.
After that I can’t see how to simplify the pipeline.

I’ll spelunk trying to find commercial support.
Finally I have and do use a compiled pipeline, but gst-launch shows what you will get :-)
James