2520 replies to this topic

[christophecvr]

Build over now errors:

NOTE: Preparing RunQueue
ERROR: Multiple .bb files are due to be built which each provide module-init-tools (/home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-core/busybox/busybox_1.23.2.bb /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-kernel/kmod/kmod_git.bb).
 This usually means one provides something the other doesn't and should.
NOTE: Executing SetScene Tasks
NOTE: Executing RunQueue Tasks
WARNING: enigma2-plugin-softcams-scam: the directory ${WORKDIR}/scam_v${PV}/ (/home/christophe/openpli40-gst1/openpli-oe-core/build/tmp/work/mips32el-oe-linux/enigma2-plugin-softcams-scam/3.53-r2/scam_v3.53/) pointed to by the S variable doesn't exist - please set S within the recipe to point to where the source has been unpacked to
WARNING: Failed to fetch URL file://strace-0001-sigaction-wrap-sa_restorer-in-ifdef-SA_RESTORER-cons.patch, attempting MIRRORS if available
ERROR: Fetcher failure: Unable to find file file://strace-0001-sigaction-wrap-sa_restorer-in-ifdef-SA_RESTORER-cons.patch anywhere. The paths that were searched were:
    /home/christophe/openpli40-gst1/openpli-oe-core/meta-openpli/recipes-devtools/strace/strace-4.10/openpli
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace-4.10/openpli
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace/openpli
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/files/openpli
    /home/christophe/openpli40-gst1/openpli-oe-core/meta-openpli/recipes-devtools/strace/strace-4.10/vuduo2
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace-4.10/vuduo2
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace/vuduo2
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/files/vuduo2
    /home/christophe/openpli40-gst1/openpli-oe-core/meta-openpli/recipes-devtools/strace/strace-4.10/mipsel
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace-4.10/mipsel
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace/mipsel
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/files/mipsel
    /home/christophe/openpli40-gst1/openpli-oe-core/meta-openpli/recipes-devtools/strace/strace-4.10/
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace-4.10/
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace/
    /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/files/
    /home/christophe/openpli40-gst1/sources
ERROR: Function failed: Fetcher failure for URL: 'file://strace-0001-sigaction-wrap-sa_restorer-in-ifdef-SA_RESTORER-cons.patch'. Unable to fetch URL from any source.
ERROR: Logfile of failure stored in: /home/christophe/openpli40-gst1/openpli-oe-core/build/tmp/work/mips32el-oe-linux/strace/4.10-r0/temp/log.do_fetch.19135
ERROR: Task 2259 (/home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace_4.10.bb, do_fetch) failed with exit code '1'
NOTE: Tasks Summary: Attempted 1512 tasks of which 0 didn't need to be rerun and 1 failed.
Waiting for 0 running tasks to finish:

Summary: 1 task failed:
  /home/christophe/openpli40-gst1/openpli-oe-core/openembedded-core/meta/recipes-devtools/strace/strace_4.10.bb, do_fetch
Summary: There were 5 WARNING messages shown.
Summary: There were 3 ERROR messages shown, returning a non-zero exit code.

My system ubuntu 14.04 lts x86_64

[MiLo]

Please read the last commit comment on that branch. It does not build yet.

[christophecvr]

Well sorry I misunderstood.

 

I understood that You requested us to try out the master-next-gst1 branch.

 

 

But currently I have my builds back with gst-1 and latest gstreamer HEAD.(vuduo2 and dm8000)

[mx3L]

 

agreed, @OpenPli if possible please apply above patches

 

Also from now on we should test playbin using the same flags in order to emulate servicemp3!.

gst-launch-1.0 playbin uri=http://megahdlive1-f.akamaihd.net/i/live_1@105260/master.m3u8 flags=0x47

Also I think GST_PLAY_FLAG_SOFT_VOLUME is not honored by gst-launch (it should not display a warning message)

 

Thanks @mx3L for the help

 

 

Edit1. Patch for not displaying warning message when GST_PLAY_FLAG_SOFT_VOLUME is not set also submited https://bugzilla.gno...gi?id=748809#c2

 

no problem , glad we could figure this out..

 

@MiLo

We were already commenting on gstreamer1.0 recipes in meta-openpli few pages back.

 

concerns:

- extra complexity when doing local changes in meta-local, we would do bbappend on top of bbappend.

- checking after every update of openembedded-core if our bbappends are still valid.

 

I guess it's not a big deal, In any case we already had bbappends besides oe-core recipes - we had to add gstreamer1.0 recipes from oe-core to meta-openpli, since I believe in january we have no openembedded-core submodule in openpli-oe-core. So maybe we can reuse them, if you choose bbappend way.

[Erik Slagter]

 

 

I recommend to make every flag specific to avoid relying on playbin defaults

 

 

Sure, better to use only what we require because any new default might cause new troubles into Enigma2 (servicemp3).

 

Here is the syncronazation of GstPlayFlags with latest GStreamer.

Subject: [PATCH] servicemp3: syncronize GstPlayFlags with GStreamer


diff --git a/lib/service/servicemp3.cpp b/lib/service/servicemp3.cpp
index a246ae5..e74b70d 100644
--- a/lib/service/servicemp3.cpp
+++ b/lib/service/servicemp3.cpp
@@ -38,17 +38,25 @@ typedef enum
     PROGRESSIVE_DOWNLOAD    = 0x00000002
 } eServiceMP3Flags;
 
+/*
+ * GstPlayFlags flags from playbin2. It is the policy of GStreamer to
+ * not publicly expose element-specific enums. That's why this
+ * GstPlayFlags enum has been copied here.
+ */
 typedef enum
 {
-    GST_PLAY_FLAG_VIDEO         = 0x00000001,
-    GST_PLAY_FLAG_AUDIO         = 0x00000002,
-    GST_PLAY_FLAG_TEXT          = 0x00000004,
-    GST_PLAY_FLAG_VIS           = 0x00000008,
-    GST_PLAY_FLAG_SOFT_VOLUME   = 0x00000010,
-    GST_PLAY_FLAG_NATIVE_AUDIO  = 0x00000020,
-    GST_PLAY_FLAG_NATIVE_VIDEO  = 0x00000040,
-    GST_PLAY_FLAG_DOWNLOAD      = 0x00000080,
-    GST_PLAY_FLAG_BUFFERING     = 0x00000100
+    GST_PLAY_FLAG_VIDEO         = (1 << 0),
+    GST_PLAY_FLAG_AUDIO         = (1 << 1),
+    GST_PLAY_FLAG_TEXT          = (1 << 2),
+    GST_PLAY_FLAG_VIS           = (1 << 3),
+    GST_PLAY_FLAG_SOFT_VOLUME   = (1 << 4),
+    GST_PLAY_FLAG_NATIVE_AUDIO  = (1 << 5),
+    GST_PLAY_FLAG_NATIVE_VIDEO  = (1 << 6),
+    GST_PLAY_FLAG_DOWNLOAD      = (1 << 7),
+    GST_PLAY_FLAG_BUFFERING     = (1 << 8),
+    GST_PLAY_FLAG_DEINTERLACE   = (1 << 9),
+    GST_PLAY_FLAG_SOFT_COLORBALANCE = (1 << 10),
+    GST_PLAY_FLAG_FORCE_FILTERS = (1 << 11),
 } GstPlayFlags;
 
 // eServiceFactoryMP3
--

Next we need to define "our" default flags. Bellow are the playbin default ones.

#define DEFAULT_FLAGS             GST_PLAY_FLAG_AUDIO | GST_PLAY_FLAG_VIDEO | GST_PLAY_FLAG_TEXT | \
                                  GST_PLAY_FLAG_SOFT_VOLUME | GST_PLAY_FLAG_DEINTERLACE | \
                                  GST_PLAY_FLAG_SOFT_COLORBALANCE

And here is "our" default flags.

Subject: [PATCH] servicemp3: don't rely on playbin default GstPlayFlags

Latest changes on playbin introduce new default GstPLayFlags causing issues.
It's better to define servicemp3 defaults.

diff --git a/lib/service/servicemp3.cpp b/lib/service/servicemp3.cpp
index e74b70d..63384cf 100644
--- a/lib/service/servicemp3.cpp
+++ b/lib/service/servicemp3.cpp
@@ -59,6 +59,13 @@ typedef enum
     GST_PLAY_FLAG_FORCE_FILTERS = (1 << 11),
 } GstPlayFlags;
 
+/*
+ * avoid video conversion, let the (hw) sinks handle that GST_PLAY_FLAG_NATIVE_VIDEO
+ * volume control is done by hardware ~GST_PLAY_FLAG_SOFT_VOLUME
+ */
+#define DEFAULT_FLAGS    GST_PLAY_FLAG_AUDIO | GST_PLAY_FLAG_VIDEO | GST_PLAY_FLAG_TEXT | \
+            GST_PLAY_FLAG_NATIVE_VIDEO
+
 // eServiceFactoryMP3
 
 /*
@@ -539,12 +546,7 @@ eServiceMP3::eServiceMP3(eServiceReference ref):
 #endif
     if ( m_gst_playbin )
     {
-        guint flags;
-        g_object_get(G_OBJECT (m_gst_playbin), "flags", &flags, NULL);
-        /* avoid video conversion, let the (hardware) sinks handle that */
-        flags |= GST_PLAY_FLAG_NATIVE_VIDEO;
-        /* volume control is done by hardware */
-        flags &= ~GST_PLAY_FLAG_SOFT_VOLUME;
+        guint flags = DEFAULT_FLAGS;
         if ( m_sourceinfo.is_streaming )
         {
             g_signal_connect (G_OBJECT (m_gst_playbin), "notify::source", G_CALLBACK (playbinNotifySource), this);
--

Athoik can you change the #define (preprocessor semantics) into an enum or const int (compiler semantics)?

 

BTW I do endorse what you did with the existing enum, 1<<x is so much more readable than a hex string.

Edited by Erik Slagter, 4 May 2015 - 17:50.

[athoik]

Athoik can you change the #define (preprocessor semantics) into an enum or const int (compiler semantics)?
BTW I do endorse what you did with the existing enum, 1<<x is so much more readable than a hex string.

 
I guess this one is better (there are other #defines in servicemp3 though, eg HTTP_TIMEOUT 30).
 

Subject: [PATCH] servicemp3: don't rely on playbin default GstPlayFlags

Latest changes on playbin introduce new default GstPLayFlags causing issues.
It's better to use servicemp3 default flags.

diff --git a/lib/service/servicemp3.cpp b/lib/service/servicemp3.cpp
index e74b70d..3732d8d 100644
--- a/lib/service/servicemp3.cpp
+++ b/lib/service/servicemp3.cpp
@@ -539,12 +539,13 @@ eServiceMP3::eServiceMP3(eServiceReference ref):
 #endif
 	if ( m_gst_playbin )
 	{
-		guint flags;
-		g_object_get(G_OBJECT (m_gst_playbin), "flags", &flags, NULL);
-		/* avoid video conversion, let the (hardware) sinks handle that */
-		flags |= GST_PLAY_FLAG_NATIVE_VIDEO;
-		/* volume control is done by hardware */
-		flags &= ~GST_PLAY_FLAG_SOFT_VOLUME;
+		/*
+		 * avoid video conversion, let the dvbmediasink handle that using native video flag
+		 * volume control is done by hardware, do not use soft volume flag
+		 */
+		guint flags = GST_PLAY_FLAG_AUDIO | GST_PLAY_FLAG_VIDEO | \
+				GST_PLAY_FLAG_TEXT | GST_PLAY_FLAG_NATIVE_VIDEO;
+
 		if ( m_sourceinfo.is_streaming )
 		{
 			g_signal_connect (G_OBJECT (m_gst_playbin), "notify::source", G_CALLBACK (playbinNotifySource), this);
-- 

 0001-servicemp3-don-t-rely-on-playbin-default-GstPlayFlag.patch   1.32KB   5 downloads

Also please commit "Subject: [PATCH] servicemp3: syncronize GstPlayFlags with GStreamer" 0001-servicemp3-syncronize-GstPlayFlags-with-GStreamer.patch

 

The idea with shift was copied directly from GStreamer

[Erik Slagter]

Applied.

[christophecvr]

Some info about latest status gst-1 .

 

Currently all runs fine provided that :

 

The latest git is used for the full gstreamer package. (master git current HEAD)

 

To build the gstreamer the only required patches are :

 

gstreamer1.0   no patch required

gstreamer1.0-libav no patch required

gstreamer1.0-plugins-bad no patch required

gstreamer1.0-plugins-base

SRC_URI += "file://video-converter-fix-compiler-warning-due-missing-static-declaration.patch \ #patch of me taken from furtther gstreamer then master evolution patch origin Sebastian Drooge.
            file://0001-riff-media-added-fourcc-to-all-mpeg4-video-caps.patch \ #version off athoik

gstreamer1.0-plugins-good no patch required

gstreamer1.0-plugins-ugly no patch required.

 

For some media the flag 0x47 must be added. Which is already aplied to latest pli4 enigma2.

 

With gst-launch-1.0  for testing purposes ad the flag at the end for the concerned media  flags=0x47

 

Al test performed by me

 

I used the very last gstreamer1.0-plugin-multibox-dvbmediasink. (the last version is required).

 

vuduo2  well very simple all media ok

 

dm8000

 

almost all media     except:

 

dash and smoothstream   But I think there is really nothing to do as dm8000 has to few resources.

 

Then dtsdownmix now much better. The first start +- 15 seconds.    The pause unpause 100 % ok   The change off language still takes up to 60 seconds.

However the sound is almost perfect

[MastaG]

Great work christophecvr!

 

So when will everything be pushed to gst-1 or master-next-gst1?

[Erik Slagter]

For something to be pushed, it first has to be comitted and for that we need a patch...

[MastaG]

Well maybe you can apply the recipes from Athoik to use the HEAD branch?

http://forums.openpl...ndpost&p=488868

(see also his note to exclude cairo, gdk-pixbuf and gudev )

 

And his fix from here:

http://forums.openpl...ndpost&p=488861

 

Then add the mpeg-dash and smooth streaming packages (for the receivers capable of playing these)

http://forums.openpl...ndpost&p=488864

 

Now that only leaves the sink, I'm using christophecvr's multibox sink and it works.

However I'm not sure if the default sink needs patches (I haven't tested it for the last two weeks).

 

Correct me if I'm wrong

Edited by MastaG, 7 May 2015 - 15:06.

[christophecvr]

Some findings which I discovered with tests.

 

After that any off the gstreamer components have been updated,replaced or extra components are added

 

The gstreamer .cache/gstreamer-1.0/registry.mipsel.bin must be deleted and the box needs a reboot.

 

with components I mean any gst package or plugin this includes the dvbmediasink and or enigma2 plugins which installs an extra gst plugin.

 

Question to pli team :

 

Why is the .cache directory set in / directory when init 3 is used (the default) and not in /home/root ?

 

When using init 4 and then launch enigma2 or enigma2.sh  it is set in /home/root  like all the rest.

[betacentauri]

It should be /home/root. Gstreamer 0.10 used that. Never saw before that / was used. So maybe a bug.

[christophecvr]

It should be /home/root. Gstreamer 0.10 used that. Never saw before that / was used. So maybe a bug.

It's only happening when You bout the box. Actually if You use init 3.

 

If You for example used trough telnet init 4

 

remove the .cache out off / dir and then start enigma2  with enigma2 or enigma2.sh it is set in the right place /home/root  .(even if You not remove the .cache out off / a new .cache is set in /home/root)

 

Somewhere there is a wrong setting when using init 3 but I can't find which.

Edited by christophecvr, 10 May 2015 - 09:12.

[betacentauri]

But it's not happening with gstreamer0.10, or?
At best you need to find the gstreamer source code to be able to check what's going on.
I see no real difference in init 3 (which also starts enigma2.sh) and starting enigma2.sh from shell. Maybe a variable in environment is not set when you use init.

[MastaG]

Yes, I have it as well.

Maybe some ./configure option which you can pass?

[christophecvr]

But it's not happening with gstreamer0.10, or?
At best you need to find the gstreamer source code to be able to check what's going on.
I see no real difference in init 3 (which also starts enigma2.sh) and starting enigma2.sh from shell. Maybe a variable in environment is not set when you use init.

 At this time I do not have gstreamer-0.10 images anymore.

 

But As far I remembered it was same issue with gstreamer-0.10 . by pli4 last images.

 

It has nothing to do with gstreamer self.

 

Example If You remove all .cache  dir's out off / if it was there and out off /home/root   . After You used init 4 or init 5.

 

Start gstreamer with gst-launch-1.0  xxxxx   The .cache/gstreamer-1.0/xxxx   is created in /home/root  So ok.

 

Even more proof it is not really to gstreamer self , If I use init 3.  The .cache  is created in /  map.

 

My current updated bb to build dvbmediasink has now extra rules :

pkg_preinst_${PN}_prepend () {
	if [ -d "/.cache/gstreamer-1.0" ]
	then
		rm -rf "/.cache/gstreamer-1.0"
	fi
	if [ -d "/home/root/.cache/gstreamer-1.0" ]
	then
		rm -rf "/home/root/.cache/gstreamer-1.0"
	fi
}

This is a requirement I noticed if You change one off the gst components.  (every plugin or sink)

 

Ok now I did install the updated for example dvbmediasink trough telnet with :

 

opkg update ; opkg upgrade.

or just opkg install xxxx --force-reinstall  (if its a changed ipk with same versions numbers  that for developpenent)

 

It will remove the gstreamer .cache files  so ok now.

 

Note as long You do not reboot the stb the current loaded gstreamer will remain actif !!

That's why standard users must reboot after an upgrade. And very good practice.

 

However as developper/tester. You have other options.

 

If You wan't to proceed whitout rebooting box you can use after that a gstreamer package is upgraded or reinstalled and the gstreamer cache files have been removed.:

 

And You wan't go to init 4 but remain in the initial booted init 3. stay using enigma in normal mode is the goal then.

 

sync; echo 3 > /proc/sys/vm/drop_caches

followed by

gst-inspect-1.0

 

You will see that .cache is made in /home/root dir and that's right.

 

It is a problem with a shell variable enviroment when using init 3. Perhaps a sequence problem. I can't find which script does load enigma2 when using init 3.

 

If You then do init 4  and back init 3 again .cache is set in / dir instead off /home/root

[christophecvr]

as extra note :  It is not stb related this problem off wrong location for .cache map. Same issue on vuduo2 or dm8000.

 

It is a general pli4 problem.

Do not know if other images do have same issue. Like for example openatv if they have same issue it's maybe some error in openembedded.

[christophecvr]

O yes again extra note by

 

If You then do init 4  and back init 3 again .cache is set in / dir instead off /home/root

 

You have two gstreamer registry files and this leads to very space situations.

 

Note that after a enigma2 crash and autorestart. The .cache map is created in /home/root . very very space situations are there then. Mostly not solved anymore after a standard reboot since the both .cache maps stays there.

Edited by christophecvr, 10 May 2015 - 10:34.

[theparasol]

Isnt it better to launch the gstreamer engine only when needed and kill it after usage?

This will clear for sure some stupid things like the leftovers of previous gstreamer usage.