[robclark]

That branch is pretty old, might be best to extract the gstdecoder bits and bring in to current checkout from trunk. I did this several months ago when I was looking into gst. Was running under OSX in fact

yeah, I did rebase on latest master. Entirely possible that I could have screwed something up. In theory, when we seek, I should send some events to gst to flush out the pipeline, but not sure yet how I can get notification in DVDVideoCodecGStreamer.

I sort of hope your full gst player is using gst demuxers as well, because some of the stuff like seeking and trick play would work better if you let gst do the demuxing too.

[tarzan23]

I just wanted to say thanks for the great work Rob, here and on the pandaboard mailing list. If you need help for testing let me know.

[davilla]

yeah, I did rebase on latest master. Entirely possible that I could have screwed something up. In theory, when we seek, I should send some events to gst to flush out the pipeline, but not sure yet how I can get notification in DVDVideoCodecGStreamer.

I sort of hope your full gst player is using gst demuxers as well, because some of the stuff like seeking and trick play would work better if you let gst do the demuxing too.

codec ::Reset, you will get that call when seeking. does not really to reset the codec but might in the case of a gross error.

gst demuxers, hehe. it's a hit or miss went trying trick play, it all depends if the gst pipeline that was setup can handle trick play.

[robclark]

I just wanted to say thanks for the great work Rob, here and on the pandaboard mailing list. If you need help for testing let me know.

fyi, it looks like pvr/sgx update was just pushed to PPA a few minutes ago. I don't think gst-plugins-base / gst-ducati updates I made are pushed yet, but the impatient can build from the git trees at http://gitorious.org/gstreamer-omap

We're seeing some slightly weird problems on a couple ppls boards around here, which oddly I don't see with exact same binaries. So could be that some debugging is needed just as much as some testing ;-)

[robclark]

codec ::Reset, you will get that call when seeking. does not really to reset the codec but might in the case of a gross error.

ahh, this is probably what I'm missing. I think I'd need to flush the gst pipeline, otherwise you'll still get old buffers popping out appsink. I'll give this a try tonight.

gst demuxers, hehe. it's a hit or miss went trying trick play, it all depends if the gst pipeline that was setup can handle trick play.

Yeah, depends somewhat on the demuxer.. I guess at least for the more mainstream container formats it should work ok. (Well, maybe it's possible for the decoder element to screw this up.. but I wouldn't have expected that.)

[robclark]

fyi, pushed another update to my github branch.. seeking seems to be working better now. Also cleaned up a bit some buffer refcnt'ing stuff so we don't get buffers freed/recycled while the renderer still has 'em.

[RockerC]

CuBox by Solid-Run could be a great ARM and OpenGL ES development platform for XBMC on Linux?

http://www.solid-run.com/products/cubox
http://www.linuxfordevices.com/c/a/News/SolidRun-CuBox/
http://www.youtube.com/watch?v=twEoMYEJls4

it is relatively inexpensive at €99 and you do not need to buy any additional hardware for it

They say to already have XBMC for Ubuntu Linux running on it
http://www.youtube.com/watch?v=twEoMYEJls4
http://www.solid-run.com/phpbb/viewtopic.php?f=2&t=5

http://www.solid-run.com/phpbb/viewtopic.php?f=11&t=4

Typically a distro that supports ARMv7 and VFPv3-d16 instruction sets would be best, in order to get best performance.
Best if using hardvfp, but currently we haven't ported any (Debian probably will be first).

Marvell Armada 510 / PXA510 SoC it uses seams little special

http://www.marvell.com/application-proce.../index.jsp
http://www.linuxfordevices.com/c/a/News/...-and-1000/

Because Marvell is not only an ARM processor licensee, but also an architecture licensee, it has the right to take ARM-designed cores and use it in SoCs of its own devising. This is precisely what the chipmaker has done with its newly announced Armada products, creating its own, unique Sheeva designs.

Marvell's Sheeva PJ1 architecture -- used in the new, single-core Armada 100 family and dual-core Armada 1000 family -- is based on the ARMv5 instruction set. The company's Sheeva PJ4 architecture -- used in the Armada 500 and 600 families -- is based on the ARMv7 instruction set, previously employed by processors including those using ARM's Cortex-A8 core.

Instead of employing ARM's Neon SIMD (single instruction multiple data) instruction set, the Armada products incorporate Wireless MMX2 technology, inherited from Intel when Marvell acquired the latter's XScale-based SoCs in 2006. The Armada SoCs additionally include Marvell's Qdeo video processing technologies, and the processors can also accelerate Adobe's Flash technology, the company says.

[RockerC]

CuBox by Solid-Run could be a great ARM and OpenGL ES development platform for XBMC on Linux?

http://www.solid-run.com/products/cubox
http://www.linuxfordevices.com/c/a/News/SolidRun-CuBox/
http://www.youtube.com/watch?v=twEoMYEJls4

it is relatively inexpensive at €99 and you do not need to buy any additional hardware for it

Fork by rabeeh from SolidRun now available on GitHub with GStreamer hardware decoding and Dove overlay video rendering
https://github.com/rabeeh/xbmc

http://www.solid-run.com/phpbb/viewtopic.php?f=2&t=5

I'v forked xbmc and created -

https://github.com/rabeeh/xbmc

Most of the code is up there now.

Notice that I build under scratchbox.
The way to build is -
1. populate an Ubuntu 10.04 rootfs under Scratchbox
2. ./bootstrap
3. ./configure --enable-gles --enable-dove --enable-dove-overlay --enable-gstreamer --disable-pulse

[quickie]

Could we use that code or adapt it with Raspberry Pi or Pandora Board?

[pfbach]

Hi guys,
I have to say thanks to those who have contributed so much already for making xbmc work with the pandaboard.

I am trying to build xbmc from robclarks github repo (gstreamer-eglimg branch) for my pandaboard rev A2, but I get the same message I have gotten a few other times I have tried to compile xmbc with openmax support (from various branches of xbmc).

My config settings:
./configure --enable-neon --disable-vdpau --disable-optical-drive --enable-gles --enable-gstreamer

The error messages from compile time:
OpenMaxVideo.cpp:81:30: error invalid use of incomplete type 'struct COpenMaxVideo'
DVDVideoCodec.h:41:7: error forward declaration of 'struct COpenMaxVideo'
....

I have installed
libomxil-bellagio0
libomxil-bellagio-dev
libomxil-bellagio-components-base

I am running latest Linaro 11.11 build.
I have also installed stuff from
https://launchpad.net/~tiomap-dev/+archive/release, but there seems to be dependency problems for gstreamer0.10-faac and ubuntu-omap4-extras in that (faac/libfaac) >=1.28 is not installed on linaro 11.11

I managed to compile a working version off the earlier gstreamer branch by topfs in june/july, but somehow I don't remember how to get around these error messages.

Any help and guidance appreciated.

[robclark]

Fwiw, I didn't have omx installed or enabled when I built, so you might try removing bellagio or at least disabling omx support when you configure

[pfbach]

Fwiw, I didn't have omx installed or enabled when I built, so you might try removing bellagio or at least disabling omx support when you configure

Ah, yeah you are right. That did it, but I thought one were supposed to use OpenMax with Pandaboard?

However, when I start xbmc it segfaults after warning messages:

(FEH.py:3674):Gtk-WARNING **:Unable to locate theme engine in module_path: "pixmap",
...

I guess the warning messages has nothing to do with the segfault itself so I will investigate this further. Might be because I don't have ubuntu-omap4-extras packages installed, due to conflicting faac/libfaac versions?

[robclark]

Ah, yeah you are right. That did it, but I thought one were supposed to use OpenMax with Pandaboard?

Nope, no omx in 11.10

However, when I start xbmc it segfaults after warning messages:

(FEH.py:3674):Gtk-WARNING **:Unable to locate theme engine in module_path: "pixmap",
...

I guess the warning messages has nothing to do with the segfault itself so I will investigate this further. Might be because I don't have ubuntu-omap4-extras packages installed, due to conflicting faac/libfaac versions?

I see the warning too, but not the segfault, so I think unrelated.. but you probably will need ubuntu-omap4-extras, as there might be some issues in some cases w/ sw decoders. But I think gst-ducati and gst-plugins-base you might still need to build from the git trees at http://gitorious.org/gstreamer-omap (ti branch in case of gst-plugins-base), since I think what is in the PPA hasn't been updated yet.

[pfbach]

I see the warning too, but not the segfault, so I think unrelated.. but you probably will need ubuntu-omap4-extras, as there might be some issues in some cases w/ sw decoders. But I think gst-ducati and gst-plugins-base you might still need to build from the git trees at http://gitorious.org/gstreamer-omap (ti branch in case of gst-plugins-base), since I think what is in the PPA hasn't been updated yet.

Ok, thanks.

I figured out the conflict issue. I needed to enable multiverse repo in apt, so now I have installed ubuntu-omap4-extras.

I will build gst-ducati and gst-plugins-base and see if that will help.

[pavo]

In the meantime new model of Raspbery PI has arrived, and it looks pretty good http://www.raspberrypi.org/archives/431