[FFe] Android media support over hybris for gst-plugins-bad1.0

Bug #1224665 reported by Ricardo Salveti
10
This bug affects 1 person
Affects Status Importance Assigned to Milestone
gst-plugins-bad1.0 (Ubuntu)
Fix Released
Undecided
Unassigned

Bug Description

Hey,

Jim Hodapp just finalized the work to get hardware accelerated video decode & rendering using the Android libstagefright (HAL that provides hw decode for Android). This allow us having hw accelerated video decode in Ubuntu Touch without depending on the Android MediaPlayer service (the new solution just depends on the libstagefright libraries instead).

You can check more details about the implementation and planning at https://blueprints.launchpad.net/ubuntu/+spec/client-1303-hw-video-decode-rendering-support

As the work was just completed, it's still not available in upstream, this FFe request would be to have it as a distro patch for 13.10 until Jim is able to start the upstreaming process for it.

The extra code would be enabled just for armhf and i386 (only tested at armhf though), and it doesn't modify any of the existing plugins in the package.

List of extra plugins/files:
/usr/lib/*/gstreamer-1.0/libgstandroidmedia.so
/usr/lib/*/gstreamer-1.0/libgstmirsink.so
/usr/lib/*/libgstmiralloc-1.0.so.0.104.0

And the debdiff (I can push myself once approved): http://people.canonical.com/~rsalveti/gst-plugins-bad1.0_1.1.4-2ubuntu2.debdiff

Revision history for this message
Iain Lane (laney) wrote :

Are the arch restrictions because it's known to only possibly work on these arches or just because it hasn't been tested there?

I heard a rumour that this isn't going to be the final destination for this code (because you don't want -bad on touch images which seems quite reasonable)—is that true? If so, what is the plan?

Revision history for this message
Oliver Grawert (ogra) wrote :

it is a plugin that makes use of android codecs which are only avalable for these two arches

there are plans for a packaging change to at least split out the plugin sparately in its own binary package so we can seed it alone.
i know ricardo started on it, but dont know where it stands yet

Revision history for this message
Ricardo Salveti (rsalveti) wrote :

Iain, we can only test the code with i386 and arm, because we don't have android devices (4.2.2) with other archs.

We can try to enable it for other archs, but that will be useless at this point.

Now for the package split, at the latest version available at https://launchpad.net/~rsalveti/+archive/ppa I separated the hybris plugin into another package, so we don't need to migrate the code in another gst plugin for 13.10.

We could create a completely new plugin for it as well, but all we need to care for 13.10 (touch related) is not having the entire -bad in the image.

Revision history for this message
Iain Lane (laney) wrote :
Revision history for this message
Iain Lane (laney) wrote :

The broad approach seems fine. I have some comments/questions

  1. I think this will be useless to most desktop users, so can we not have the dependency from -bad to -hybris? If we have the split then we might as well use it to the best effect for everybody.
  2. "Forwarded: no" makes me uneasy. Can we have an assurance that after 13.10 this will be properly proposed upstream?
  3. Please do a quick check with Debian (slomo) that the split is appropriate and how it will be done there once 2. is achieved so that we don't have any problems further down the line.

Revision history for this message
Ricardo Salveti (rsalveti) wrote : Re: [Bug 1224665] Re: [FFe] Android media support over hybris for gst-plugins-bad1.0

On Fri, Sep 20, 2013 at 11:08 AM, Iain Lane <email address hidden> wrote:
> The broad approach seems fine. I have some comments/questions
>
> 1. I think this will be useless to most desktop users, so can we not have the dependency from -bad to -hybris? If we have the split then we might as well use it to the best effect for everybody.

Yup, agreed, will change.

> 2. "Forwarded: no" makes me uneasy. Can we have an assurance that after 13.10 this will be properly proposed upstream?

This needs to go upstream, we also don't want to maintain such big
diff. Hopefully it will happen during the 14.04 cycle.

> 3. Please do a quick check with Debian (slomo) that the split is appropriate and how it will be done there once 2. is achieved so that we don't have any problems further down the line.

Jim, mind checking with slomo? Otherwise I can check next week.

Revision history for this message
Jim Hodapp (jhodapp) wrote :

>> 2. "Forwarded: no" makes me uneasy. Can we have an assurance that
after 13.10 this will be properly proposed upstream?

> This needs to go upstream, we also don't want to maintain such big
> diff. Hopefully it will happen during the 14.04 cycle.

This is a goal of mine, to get my new plugins included with upstream. I'm
targeting this for 14.04.

>> 3. Please do a quick check with Debian (slomo) that the split is
appropriate and how it will be done there once 2. is achieved so that we
don't have any problems further down the line.

> Jim, mind checking with slomo? Otherwise I can check next week.

I assume you're referring to the layout of how I approached modifying
gstamc? If so, I already checked with slomo and I implemented the split
just as he suggested.

Jim

On Fri, Sep 20, 2013 at 1:28 PM, Ricardo Salveti <email address hidden>wrote:

> On Fri, Sep 20, 2013 at 11:08 AM, Iain Lane <email address hidden>
> wrote:
> > The broad approach seems fine. I have some comments/questions
> >
> > 1. I think this will be useless to most desktop users, so can we not
> have the dependency from -bad to -hybris? If we have the split then we
> might as well use it to the best effect for everybody.
>
> Yup, agreed, will change.
>
> > 2. "Forwarded: no" makes me uneasy. Can we have an assurance that
> after 13.10 this will be properly proposed upstream?
>
> This needs to go upstream, we also don't want to maintain such big
> diff. Hopefully it will happen during the 14.04 cycle.
>
> > 3. Please do a quick check with Debian (slomo) that the split is
> appropriate and how it will be done there once 2. is achieved so that we
> don't have any problems further down the line.
>
> Jim, mind checking with slomo? Otherwise I can check next week.
>

Revision history for this message
Iain Lane (laney) wrote :

Thanks for the reassurances about forwarding upstream. I'm glad to see that you've checked the broad strokes of your approach too.

With 3 I was referring to the Debian packaging really, as it can sometimes be quite hard to leave mistakes and divergances behind. Might be a Ricardo thing (but I saw that Oli pinged him on #ubuntu-devel the other day; maybe worth trying again)

Revision history for this message
Oliver Grawert (ogra) wrote :

(from #ubuntu-devel)
<ogra> slomo, could you take a peek at https://bugs.launchpad.net/ubuntu/+source/gst-plugins-bad1.0/+bug/1224665 and leave a comment if you find the time ?
<ubottu> Ubuntu bug 1224665 in gst-plugins-bad1.0 (Ubuntu) "[FFe] Android media support over hybris for gst-plugins-bad1.0" [Undecided,New]
...
<slomo_> ogra: if this is proposed for upstream integration, sure

... seems he didnt find the time to leave a comment but he kind of nodded it off in case it is planned to go upstream ...

Revision history for this message
Iain Lane (laney) wrote :

OK, please let's go ahead. You'll need an AA to NEW the package.

Also make sure to re-test on the latest upstream from master, as we'll be moving to 1.2.0 soon.

Iain Lane (laney)
Changed in gst-plugins-bad1.0 (Ubuntu):
status: New → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.