2520 replies to this topic

[Taapat]

@Taapat, what problems did you see on this commit? https://cgit.freedes...03b2d0f424c0595

 

The problem is to build gst-plugins-good, because there is a changed GstClockTime.
Therefore, I prefer to use a commit before it.
If I'm not mistaken, this one also complained in opnepli-nex topic.

[MastaG]

 

@Taapat, what problems did you see on this commit? https://cgit.freedes...03b2d0f424c0595

 

The problem is to build gst-plugins-good, because there is a changed GstClockTime.
Therefore, I prefer to use a commit before it.
If I'm not mistaken, this one also complained in opnepli-nex topic.

 

 

I guess I have to do some more reading

Thank you.

[christophecvr]

 

@Taapat, what problems did you see on this commit? https://cgit.freedes...03b2d0f424c0595

 

The problem is to build gst-plugins-good, because there is a changed GstClockTime.
Therefore, I prefer to use a commit before it.
If I'm not mistaken, this one also complained in opnepli-nex topic.

 

Yes the problem is known I solved it temporarily by changing the parameter into inc recipe plugins good

 

from

TARGET_CFLAGS += "-Wno-error=maybe-uninitialized"

to

 

 

TARGET_CFLAGS += "-Wno-error"

TARGET_CFLAGS += "-Wno-error"

think it might work if You for the time being (i'm using my own recipe bundles) but by you bb append ad :

TARGET_CFLAGS += "-Wno-error"

Edited by christophecvr, 5 April 2016 - 14:53.

[mrvica]

here some more Matroskas for testing
https://sourceforge...._1.zip/download

[Taapat]

@christophecvr I test your solution, thank you!
I also do not like revert so much because it creates a problem with the update even if I never need to play the files, which problems solves these reverted fixes.
This is my first attempt to understand what implement this mkv regression.

[christophecvr]

@christophecvr I test your solution, thank you!
I also do not like revert so much because it creates a problem with the update even if I never need to play the files, which problems solves these reverted fixes.
This is my first attempt to understand what implement this mkv regression.

Well I found the cause quit quick just cause I worked for long to a descent dts downmix plugin for dm8000 which also support modern dts audio and produces a reasonable sound. (high vq streams) . And this on a time where we came from just one timestamp to the split dts audio time stamp and pts audio timestamp.(that was a real pis). And again it is.

 

Depending on the media it looked they are both needed and can be different. But from the moment the dts audiotrack is converted to raw or pretty common ac3 it is the pts timestamp we need not the dts.

[dazulrich]

VC1?

 

yes, by the looks of it. is this known to cause issues with mkv container? 

[mx3L]

Note, setting DTS as PTS will work only if PTS==DTS, i.e if video is not using b-frames. If video is using b-frames, picture reordering will be wrong and you will see problems in playback.

 

As stated in commit in bugzilla issue:

 

"matroskademux: V_MS/VFW/FOURCC streams have DTS instead of PTS" from bugzilla

 

According to mediainfo, this stream is exactly like this(see Codec ID) so only DTS timestamps are set.

General
Complete name                            : /home/marko/shared/problem.mkv
Format                                   : Matroska
Format version                           : Version 1
File size                                : 759 MiB
Duration                                 : 24mn 4s
Overall bit rate                         : 4 408 Kbps
Encoded date                             : UTC 2016-03-31 22:04:47
Writing library                          : WonderShare Matroska Muxer

Video
ID                                       : 1
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L5.0
Format settings, CABAC                   : No
Format settings, ReFrames                : 1 frame
Format settings, GOP                     : M=1, N=25
Codec ID                                 : V_MS/VFW/FOURCC / X264
Codec ID/Info                            : XiWave GNU GPL x264 MPEG-4

It means that before this patch it worked correctly only if video like this didn't have b-frames, since it set DTS as PTS.

 

If PTS timestamps are not set by demuxer they should be set by parser. Unfortunately this is not done in h264parse yet:

https://bugzilla.gno...g.cgi?id=659489

[christophecvr]

Well as test I just now modified the multibox sink. It will at all times play the video providing a timestamp is avbl. Wheater it's pts or dts . Looks to run nice whitout any regression by my on :

 

duo2

dm8000

 

For all media fixed , live , just streamed.

 

see just a simple change currently only for multibox-experimental.

 

https://github.com/c...8bf7b498f43c774

 

@MastaG

 

Gues you can try it out on next builds. Think it will be ok.

[MastaG]

@Chris, thank you my friend.

I was already building using your experimental branch since you have the sh4 stuff applied there

I'll restart it now.. it'll take some time (many receivers) and I'll also have to push the feed, so expect them thursday or friday for public testing.

[Taapat]

@mx3L but if we do not have PTS, the video playback is not possible.

Therefore, perhaps it is better to use DTS, if available.
I think it is better if there is a picture, even in some cases with the wrong reordering, than if there is no picture in any case.

[mx3L]

@Taapat

I think so, it's not a regression, since it was like that before 1.4.5.

 

@christophecvr

This fix should be correct, use_dts flag can be removed, since you're using mpeg4p2unpacker.

Edited by mx3L, 5 April 2016 - 16:36.

[christophecvr]

@Taapat

I think so, it's not a regression, since it was like that before 1.4.5.

 

@christophecvr

This fix should be correct, use_dts flag can be removed, since you're using mpeg4p2unpacker.

Yes the use_dts flag indeed should be removed. But before I just wan't to wait on more test results.

 

By me it's even not only mkv containers media improvements but strangely also a more smooth working from some live media like for example the tm_smill (rpc_tv)  and on top off it the :

#SERVICE 4097:0:0:0:0:0:0:0:0:0:http%3a//skyianywhere2-i.akamaihd.net/hls/live/200275/tg24/playlist.m3u8:Sky%20TG%2024TEST

Which played but before always for a couple off seconds ok and after that whit video/audio out off sink does now play ok.

 

I do not have a reasonable explanation for that but it is.

 

Note all my tests are done with :

 

debug lvl "0"

master-next

gstreamer version like on  my bundle posted in :

http://forums.openpl...ndpost&p=540796

[Taapat]

@christophecvr your changes also for me fix playback of this problematic mkv.

 

I'm not an expert in this question and therefore maybe write something wrong.
Trying my plugin mptgplay who plays the streams from mtg sites (tv3play etc ..), I also observe some problems with m3u8 playback from some sites.

 

As far I understand now gstreamer m3u8 playback start with a lower quality and then increase it to a better. I've noticed when Internet speed, or the site does not allow him to increase the quality, then sometimes after several seconds playback freezes.

The problems I had with links which are not intended for my region, so I'm not looking for cause of the problem in more detail.

[Beeker]

 

 

 

@Beeker why do you think that it's not necessary?
I am in my enigma a long time ago changed the dependence on gstreamer1.0-plugins-bad-hls: https://github.com/T...12bc420afddedbc
But I suppose there are some plugins, which can depending on the gstreamer1.0-plugins-bad-fragmented, so it is right that gstreamer1.0-plugins-bad-hls provides gstreamer1.0-plugins-bad-fragmented.

 

P.S.

I think that with this gstreamer issues we should move to gstreamer topoc .

 

Some plugins still depending on fragmented....Ok that could be.

 

But the problem is, if i use bbappend for gstreamer-plugins-bad,it's building fine.That's no problem.

 

But after a new git revision of plugins-bad,the box removes plugins-bad-hls, because it is obselete according to the box.

 

I can provide a log later this afternoon..Then you can see what happens, and make it clear.

Edited by Beeker, 9 April 2016 - 11:23.

[Taapat]

But in this case, it is not the plugins-bad-hls problem.

It is opkg error or image problem if it does not use the correct dependencies.

 

I also have a probem with plugins-bad-hls and gstreamer1.0-plugin-dvbmediasink packages update.
They both used provides, replaces, conflits, and it seems to me that they all do not work together properly.
But it is definitely opkg error or specific rules for the use these togheter, instead of the packet error.

Edited by Taapat, 9 April 2016 - 13:41.

[christophecvr]

I checked on the gstreamer website :

 

https://gstreamer.fr...on/plugins.html

 

But the plugins-bad-fragmented does not exist anymore.  it is now plugins-bad-hls and that plugin works 100 % perfect by me for hls since 1.9.0.1 . (I did not test 1.8.0) .

 

The links to the website needs well to be correct and maybe some links will not work from out some other countries but that has nothing to do with gstreamer self.

[Taapat]

Of course you can not think about others and use only package gstreamer1.0-plugins-bad-hls.
I respect the possibility that a plugins can depend on gstreamer1.0-plugins-bad-fragmented, so I left only provides: https://github.com/T...8664e43c36a28fc

This certainly is only opkg issue and have no influence on library work.

 

The same problem is also in your gstreamer1.0-plugin-dvbmediasink.bb.
Also, the gstreamer1.0-plugin-dvbmediasink package can not be normal update after the version changes.

 

If you think that with hls not problem then, see my answer in the Openpli-5 (still next master).
But please continue the discussion here.

[christophecvr]

p.s. renaming of plugin is not needed , the gst ver 1.9.0.1 just does not play those type off streams reason : ?? unknown yet.

 

log of it

<  3013.337> action ->  OkCancelActions ok
<  3013.338> [eDVBServicePlay] timeshift
<  3013.339> playing 4097:0:0:0:0:0:0:0:0:0:https%3a//manifest.googlevideo.com/api/manifest/hls_playlist/id/aE7JOds_GtY.2/itag/95/source/yt_live_broadcast/requiressl/yes/ratebypass/yes/live/1/cmbypass/yes/goi/160/sgoap/itag%253D140/sgovp/itag%253D136/hls_chunk_host/r7---sn-uxaxoxu-5oge.googlevideo.com/gcr/be/playlist_type/DVR/mm/32/mn/sn-uxaxoxu-5oge/ms/lv/mv/m/pl/14/dover/3/fexp/9405267,9405989,9406175,9408214,9416126,9416891,9418751,9420452,9422596,9423662,9425428,9426927,9427902,9428398,9429049,9429585,9430026,9431287,9433425,9433457,9433998/upn/Q08ymapO2ck/sver/3/mt/1460215160/ip/84.197.126.128/ipbits/0/expire/1460236775/sparams/ip,ipbits,expire,id,itag,source,requiressl,ratebypass,live,cmbypass,goi,sgoap,sgovp,hls_chunk_host,gcr,playlist_type,mm,mn,ms,mv,pl/signature/42ECEA4D5D4AE8841A1576D4F1BCDAC0F0352948.387D2AB66031C2838F333B0AA9B048CA11E8E35E/key/dg_yt0/playlist/index.m3u8:Трансляция прямого эфира NewsOne
<  3013.340> [eDVBCAService] free slot 0 demux 0 for service 1:0:1:10E0:418:1:C00000:0:0:0:
<  3013.340> [eDVBCAService] free service 1:0:1:10E0:418:1:C00000:0:0:0:
<  3013.353> [eTSMPEGDecoder] decoder state: play, vpid=ffffffff, apid=ffffffff
<  3013.353> [eDVBPCR0] DEMUX_STOP ok
<  3013.354> [eDVBPCR0] destroy
<  3013.354> [eDVBVideo0] DEMUX_STOP  ok
<  3013.355> [eDVBVideo0] VIDEO_STOP ok
<  3013.358> [eDVBVideo0] destroy
<  3013.358> [eDVBAudio0] AUDIO_STOP ok
<  3013.383> [eDVBAudio0] DEMUX_STOP ok
<  3013.383> [eDVBAudio0] AUDIO_CONTINUE ok
<  3013.389> [eDVBAudio0] destroy
<  3013.389> [eDVBText0] DEMUX_STOP ok
TuxTxt stopped service 28
<  3013.391> [eDVBText0] destroy
cleaning up
TuxTxt cache cleared
<  3013.396> [eDVBResourceManager] start release channel timer
<  3013.396> [eServiceMP3] construct!
<  3013.397> [eServiceMP3] playbin uri=https://manifest.googlevideo.com/api/manifest/hls_playlist/id/aE7JOds_GtY.2/itag/95/source/yt_live_broadcast/requiressl/yes/ratebypass/yes/live/1/cmbypass/yes/goi/160/sgoap/itag%3D140/sgovp/itag%3D136/hls_chunk_host/r7---sn-uxaxoxu-5oge.googlevideo.com/gcr/be/playlist_type/DVR/mm/32/mn/sn-uxaxoxu-5oge/ms/lv/mv/m/pl/14/dover/3/fexp/9405267,9405989,9406175,9408214,9416126,9416891,9418751,9420452,9422596,9423662,9425428,9426927,9427902,9428398,9429049,9429585,9430026,9431287,9433425,9433457,9433998/upn/Q08ymapO2ck/sver/3/mt/1460215160/ip/84.197.126.128/ipbits/0/expire/1460236775/sparams/ip,ipbits,expire,id,itag,source,requiressl,ratebypass,live,cmbypass,goi,sgoap,sgovp,hls_chunk_host,gcr,playlist_type,mm,mn,ms,mv,pl/signature/42ECEA4D5D4AE8841A1576D4F1BCDAC0F0352948.387D2AB66031C2838F333B0AA9B048CA11E8E35E/key/dg_yt0/playlist/index.m3u8
<  3013.414> [eServiceMP3] starting pipeline
<  3014.039> [eServiceMP3] state transition NULL -> READY
<  3014.066> resolved to PLAY
<  3014.069> resolved to PLAY
<  3014.071> resolved to PLAY
<  3014.073> new service started! trying to download cuts!
<  3014.074> RemovePopup, id = ZapError
<  3015.341> [DVBCAHandler] no more services
<  3016.396> [eDVBResourceManager] release cached channel (timer timeout)
<  3016.396> [eDVBLocalTimerHandler] remove channel 0xe91e68
<  3016.396> [eEPGCache] remove channel 0xe91e68
<  3016.397> [eDVBResourceManager] stop release channel timer
<  3019.398> [eDVBFrontend] close frontend 0
0:01:24.803236393  4903   0x713980 INFO            dvbvideosink gstdvbvideosink.c:2013:gst_dvbvideosink_change_state:<dvbvideosink0> GST_STATE_CHANGE_NULL_TO_READY

So it's stops that's it end off line.  same happens when running out off youtube plugin.

 

So first needs to be found the very last git version off gst where it was ok. If You say that it worked with 1.7.9 (the plugin fragmented was already removed since 1.7.2 so ... it should have worked with plugins-hls or the reports about used gstreamer version where wrong)

[christophecvr]

@MastaG

 

Concerning hls o yes again a problem

 

Reports did tell that a lot off hls streams did played well on one off You're published images with gst-1.7.9

 

repport from Taapat on toppic :

 

http://forums.openpl...ndpost&p=541695

 

however 1.7.9 already used gstreamers-plugins-bad-hls, and gstreamer1.0-plugins-bad-fragmented did not exist anymore (there was just an annoying warning about the try the install it while it did not exist )

 

Can You give the gstreamer (full gstreamer bundle recipes you used in a tar.gz bundle when You builde with gst-1.7.9 )

Edited by christophecvr, 9 April 2016 - 18:00.