259 replies to this topic

[christophecvr]

 

Why an old gstreamer work good and not the new version ?

 

Because:

 

1. The GStreamer version you are using is not the latest one (there might be a bug already reported and fixed)

2. The GStreamer version might have a bug unidentified (so someone should do proper bug report)

3. The DVBmediasink has a bug unidentified (I think @MastaG is using multimediasink that has lot improvements and maybe new bugs)

 

In any case proper bug reporting is required (gstreamer + enigma2 logs).

 

You're right on al points , except : multimediasink that has lot improvements and maybe new bugs

 

 

Since (it could have been indeed a couple of version ago from the multibox sink) but now it's apart from the fact that all usseles gst-0.10 is removed so what 100 % equal to the original pli's mediasink gst-1.0.

 

The only difference is the avbl dtsdownmix which really works and produces acceptable downmix off all dts audio including high vq streams(which do only produce noise on the pli's dvdbmediasink downmix module).

 

And it runs in sync. The only remaining hack is the delay I needed to include in case of manual requested audio resync for the use off dts downmix.

Edited by christophecvr, 28 August 2015 - 18:10.

[athoik]

You're right on al points , except : multimediasink that has lot improvements and maybe new bugs
 
 
Since (it could have been indeed a couple of version ago from the multibox sink) but now it's apart from the fact that all usseles gst-0.10 is removed so what 100 % equal to the original pli's mediasink gst-1.0.
 
The only difference is the avbl dtsdownmix which really works and produces acceptable downmix off all dts audio including high vq streams(which do only produce noise on the pli's dvdbmediasink downmix module).
 
And it runs in sync. The only remaining hack is the delay I needed to include in case of manual requested audio resync for the use off dts downmix.

The changes for dtsdownmix might affect other things can't tell for sure.

Here are some notable changes.

You are still using framed on lpcm but we removed it https://github.com/O...3597674bf8fe37d

#define LPCMCAPS \
- "audio/x-private1-lpcm; "
+ "audio/x-private1-lpcm, " \
+ "framed =(boolean) true; "

Do we need dispose and reset? If yes please create a patch against OpenPLi dvbmediasink explaining what it fixes etc etc.

+static void gst_dvbaudiosink_dispose(GObject *obj)
+{
+ G_OBJECT_CLASS(parent_class)->dispose(obj);
+ GST_INFO("GstDVBAudioSink DISPOSED");
+}
+
+static void gst_dvbaudiosink_reset(GObject *obj)
+{
+ G_OBJECT_CLASS(parent_class)->finalize(obj);
+ GST_INFO("GstDVBAudioSink RESET");
+}
+

Do we need GST_EVENT_TAG and GST_EVENT_TOC?

+ case GST_EVENT_TAG:
...
+ case GST_EVENT_TOC:
...

It would be nice to use OpenPLi dvbmediasink and solve any remaining problems (except dtsdownmix on dm8000 that you already solve on your sink).

OFC if GStreamer has new bugs they should reported in order to get fixed! (with proper logs and samples eg sample url or sample video/audio)

Edited by athoik, 28 August 2015 - 20:00.

[MastaG]

The problems he has are probably not related to the sink. Since my early builds were using the default mediasink from OpenPLi gst-1.0 branch. He had the same issue on those as well.
I guess proper logging is required.

 

Also I find that Chris' multibox sink is working better for pausing and resuming. 

I'm going to be a bit busy in the coming weeks but I'll try to make new images based on master git.

Edited by MastaG, 28 August 2015 - 21:08.

[christophecvr]

@athoik ,

 

case event tag and event toc yes are needed (in developpement  to have debug info of them ) But it's shure the day we pass to a real stable pli image some of these procedures needs to be commented ( not removed since for developpement they are crucial)

I well do agree that for standard use they are not required (as long not an extra action must be made on it).

 

the : https://github.com/O...3597674bf8fe37d

Are not aplied to my media sink however no problem to do it they won't bring anything more or less. lpcm can't cope anymore with the more heavy requiremts in downmix.

 

Ach yes the disposed and reset are a remaining left from me . the goal was to try something to be rid ohh the increased pointrest by each media chnage.

No succes whit it

[athoik]

I guess proper logging is required.

Totally agree, proper logging can shed some light..

[sonci]

you are my hero, everything works,

there was even no audio for mp4 videos with the last Open Pli, and lot of freeze,

and the're saying its a driver problem and no power processor, but with this image everything works like in Open Pli3 maybe even better,

I'm testing with Vu solo optical out to receiver, mp4, mkv, avi, subtittles, dts dolby out is everything ok, and smooth, nothing hangs like in official image,

I hope my plugins will work with this image, so I'll keep this for some time..

 

ps. If I'm asking much, please put the latest links in the first post, it will be easier to find..

THANKS

Edited by sonci, 6 September 2015 - 19:46.

[daif01]

please how can add ( blind scan ), i have vu+duo2.

[sonci]

you are my hero, everything works,

there was even no audio for mp4 videos with the last Open Pli, and lot of freeze,

and the're saying its a driver problem and no power processor, but with this image everything works like in Open Pli3 maybe even better,

I'm testing with Vu solo optical out to receiver, mp4, mkv, avi, subtittles, dts dolby out is everything ok, and smooth, nothing hangs like in official image,

I hope my plugins will work with this image, so I'll keep this for some time..

 

ps. If I'm asking much, please put the latest links in the first post, it will be easier to find..

THANKS

Ok

Unfortunately I spoke too soon, randomly some mp4 videos won't play with audio, 

the same video sometimes play well, sometimes it just doesn't show the aac audio, I don't know if this has to do with the receiver (Vu solo),

I've checked with media player classic in pc, and these are well encoded videos with 48Khz sample rate,

but most of the files are ok,

[daif01]

I have vu+duo2. The image is great, but lack blind scan and backup suit. i tried but could not download them. Please can you add blind scan and backup suit . 

[Erik Slagter]

Download the satscan plugin.

[MastaG]

I'm getting lots of requests for new images and missing plugins.

Building all this requires lots of time and patience since for every small change I have to rebuild for each and every receiver.

And every small thing like a timeout when connecting to some github repository causes a build to fail.

 

Maybe it's best to just wait for gst-1 to be merged into master.

Gstreamer 1.5.1 is pretty stable on all of my receivers.

 

Or if someone is willing to host a feed server I can do a clean build and push the feeds.

[Erik Slagter]

Hehe, now you know what it's like

 

I can now more or safely announce, I think, that there will be at some point in the future, an "OpenPLi 5" release, which will be compiled from the latest OE (see master-next branch) and which will include the merged gst-1 branch. This allows us to prepare the transition thoroughly and do some testing before the release.

 

For the images to be able to compile with the latest OE (gcc/as), most older Linux kernels need to be patched. We have sent requests to all relevant manufacturers for such a patch. Only manufacturers' images that have such a patch in their BSP (of have sufficients old or new kernels) can be compiled successfully. For the time being that means that there will probably NO OpenPLi 5 images for DMM models, but who knows, maybe athoik can fix their BSP. This also means that some other manufacturer's models may not get a OpenPLi 5 image, or maybe not at first. That's not our decision, let's have that clear.

Edited by Erik Slagter, 15 September 2015 - 18:27.

[MastaG]

It's great to rebase on a newer snapshot of OE, especially from a security point-of-view.

However I think most kernels will compile fine with the latest gcc, they may need some small patches but I think this won't cause any problems for 3.2 and upwards.

It's the closed drivers however which will fail to load when the kernel has been built with a newer version of gcc.

in case of DMM, they'll need to provide new dvb drivers built with your version of gcc and I doubt they'll ever do that.

 

There are some dirty workarounds.

For example OpenATV is still building 2.6.18 for the old dm800 with gcc 4.9.

Then they patch the VERMAGIC in all closed source dvb modules to match gcc 4.9, something like sed 's/gcc-4\.6/gcc-4\.9/g'.

This is against the DMM license though.

I guess that only leaves force-loading them using modprobe -f.

Both are dirty workarounds imo

Edited by MastaG, 15 September 2015 - 20:02.

[athoik]

With some extra patches DMM kernels compiles with gcc 5. Although I did not test resulting image if it works yet.

One of the patches is just a workaround and someone should provide a proper patch.

The problem is the hard-float... kernels should be build only with soft-float...

More info https://git.kernel.o...4f2834b205b0a79

Ps. Here is the patch from vu for hard float: https://github.com/v...867d2200a66cdc9

Edited by athoik, 15 September 2015 - 20:26.

[christophecvr]

I followed a bit the discution here:

 

First I'm really interested in building with the last openembedded. With other words the https://github.com/O...ree/master-next

 

I'm not an expert in kernel building but did it for a couple off ubuntu revision on my desktop since they did had a couple off severe buggyy standard kernel builds for am64 (or more correct x86_64-linux-gnu) A couple off years ago , until they had back a non buggy base kernel.  In those days it was ubuntu revison if i'm not wrong 11.04 (or 12.04).

 

but ..

 

@athoik

 

You tell:

The problem is the hard-float... kernels should be build only with soft-float...

 

Normally if the targets device for which you build the linux kernel has a processor with fpu hard-float should be used . Why soft float ?

[athoik]

It's not me saying that...

https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/arch/mips/Makefile?id=6218cf4410cfce7bc7e89834e73525b124625d4c

[MIPS] Always pass -msoft-float.
    
Some people still haven't heared that fp in the kernel is forbidden.
    
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

So the workaround was simply to change soft-float to hard-float, to continue to other issues.

As I said before someone should create a proper patch (live VU+ did) for soft-float.

[Erik Slagter]

@MastaG, we are not ATV. Our goal is not to support as many receivers as possible (which has some consequences). That's for example we don't have the dm800 in our gamma (any more).

[christophecvr]

@athoik,

 

The message you refering to is from 2006.

 

At base it could be right or wrong .... I do not  see the relation from upgrade gcc to 5.xx now

[Beeker]

At base it could be right or wrong .... I do not  see the relation from upgrade gcc to 5.xx now

 

https://github.com/g...7ddde06c7d6e697

 

To revert this commit you can compile kernel DM8000 without any extra patch. but it fails on kernel-module for rt8712 usb wifi.

Edited by Beeker, 16 September 2015 - 18:02.

[athoik]

Hi,

Patches for DMM boxes applied at master-next, so currently you can build images.

Please note that samba4 is too fat to fit in boxes with small flash. We need to install only the required files to fee lot of memory (just like we did when using samba3).