Huge memory leak sometime after starting a pipeline

[Sergei Vorobyov]

You should have rather taken a quick look at the program or just an
explanation how it works:

it iteratively changes location in the pipeline

filescrc location=... ! decodebin ! imagefreeze ! videoconvert ! vaapisink

and it leaks. It does not if you replace vaapisink with ximagesink (hence
the program itself does not leak, except in vaapisin). Is there anything
forbidden/suspicious in this pipeline and/or the program? I don't think so.
Do you?

This happens on all platforms: Intel, NVidia, AMD Radeon with video
acceleration. Hence vaapisink is to blame, since it's highly improbable
that 3 independent drivers leak similarly.

Do you guys understand any logical arguments? We are talking about a
program that definitely demonstrates leaks in vaapisink, and not about (any
number of) your particular programs which do not. Remember, in mathematics,
any number of examples that confirm a statement DO NOT prove it,
unfortunately (statement being "vaapisink does not leak"), but just one
that refutes it, DISPROVES it completely. Your program does not confirm
that there are no leaks, whereas mine definitely PROVES there are leaks in
vaapisink, fullstop. Explain to me what isn't still clear. In programming
terms: testing does not prove the absence of errors, just finds their
presence, sometimes. Bad luck...

I contacted the guys from gstreamer-vaapisink (which does not happen to
belong to gstreamer) maybe they understand somewhat better. It would be
stupid for me searching errors in vaapisink, I've never seen the source
code and I am pretty sure the developers will find the errors much faster
and more efficient.

---
Valgrind is a bad substitute for lack of imagination...

On Sun, Nov 9, 2014 at 6:07 PM, Sebastian Dröge <sebastian at centricular.com>
wrote:

> On So, 2014-11-09 at 15:12 +0100, Sergei Vorobyov wrote:
> > Hi, Nicolas,
> >
> > as I explained, it's 100% clear that the leak is in vaapisink, and it
> would
> > be more efficient if developers investigate it, given that I provided a
> > full minimal program that exhibits the leak on vaapisink and not on
> > ximagesink (on each of Intel, AMD, NVidia).
>
> The problem here is that it doesn't leak for others. For me using
> vaapisink after a videotestsrc or anything else really does not cause
> increased memory usage. So either you have to provide the information
> you're asked for or have to debug and fix it yourself.
>
> --
> Sebastian Dröge, Centricular Ltd · http://www.centricular.com
>
> _______________________________________________
> gstreamer-devel mailing list
> gstreamer-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/gstreamer-devel/attachments/20141109/46a2c958/attachment-0001.html>