[poky][PATCH v4 0/2] gstreamer1.0: Add patches for Gstreamer 1.4.5

[poky][PATCH v4 0/2] gstreamer1.0: Add patches for Gstreamer 1.4.5

[Yuqing Zhu]

Fix sticky events haven't been sent out when active track reach EOS

Fix QoS/lateness checking if subclass implements prepare/prepare_list vfuncs


Yuqing Zhu (2):
  gstreamer1.0: Fix ticky events haven't been sent out when active track
    reach EOS
  gstreamer1.0: Fix QoS/lateness checking if subclass implements    
    prepare/prepare_list vfuncs

 ...x-QoS-lateness-checking-if-subclass-imple.patch |  70 +++++++++
 ...cky-events-haven-t-send-out-when-ac-1-4-1.patch | 167 +++++++++++++++++++++
 .../gstreamer/gstreamer1.0_1.4.5.bb                |   2 +
 3 files changed, 239 insertions(+)
 create mode 100644 meta/recipes-multimedia/gstreamer/gstreamer1.0/0002-basesink-Fix-QoS-lateness-checking-if-subclass-imple.patch
 create mode 100755 meta/recipes-multimedia/gstreamer/gstreamer1.0/inputselector-sticky-events-haven-t-send-out-when-ac-1-4-1.patch

-- 
1.9.1

[poky][PATCH v4 1/2] gstreamer1.0: Fix ticky events haven't been sent out when active track reach EOS

[Yuqing Zhu]

EOS event hasn't been sent to down-element. The resolution is block EOS event
of inactive pad, sending the event after the pad actived.

Signed-off-by: Yuqing Zhu <b54851@freescale.com>
---
 ...cky-events-haven-t-send-out-when-ac-1-4-1.patch | 167 +++++++++++++++++++++
 .../gstreamer/gstreamer1.0_1.4.5.bb                |   1 +
 2 files changed, 168 insertions(+)
 create mode 100755 meta/recipes-multimedia/gstreamer/gstreamer1.0/inputselector-sticky-events-haven-t-send-out-when-ac-1-4-1.patch

diff --git a/meta/recipes-multimedia/gstreamer/gstreamer1.0/inputselector-sticky-events-haven-t-send-out-when-ac-1-4-1.patch b/meta/recipes-multimedia/gstreamer/gstreamer1.0/inputselector-sticky-events-haven-t-send-out-when-ac-1-4-1.patch
new file mode 100755
index 0000000..f50ce6f
--- /dev/null
+++ b/meta/recipes-multimedia/gstreamer/gstreamer1.0/inputselector-sticky-events-haven-t-send-out-when-ac-1-4-1.patch
@@ -0,0 +1,167 @@
+From 83bed90c306ed3185d48febf6441177d638f7341 Mon Sep 17 00:00:00 2001
+From: Song Bing <b06498@freescale.com>
+Date: Wed, 24 Dec 2014 10:13:51 +0800
+Subject: [PATCH] inputselector: sticky events haven't send out when active
+ track reach EOS
+
+EOS event hasn't been send to down-element. The resolution is block EOS event
+of inactive pad, send the event after the pad actived.
+
+https://bugzilla.gnome.org/show_bug.cgi?id=740949
+
+Upstream-Status: Backport [1.5.1]
+
+Signed-off-by: Song Bing <b06498@freescale.com>
+---
+ plugins/elements/gstinputselector.c |   58 ++++++++++++++++++++++++++---------
+ plugins/elements/gstinputselector.h |    1 +
+ 2 files changed, 45 insertions(+), 14 deletions(-)
+
+diff --git a/plugins/elements/gstinputselector.c b/plugins/elements/gstinputselector.c
+index fb50802..4461f7c 100644
+--- a/plugins/elements/gstinputselector.c
++++ b/plugins/elements/gstinputselector.c
+@@ -440,6 +440,17 @@ gst_selector_pad_iterate_linked_pads (GstPad * pad, GstObject * parent)
+ }
+ 
+ static gboolean
++gst_input_selector_eos_wait (GstInputSelector * self, GstSelectorPad * pad)
++{
++  while (!self->eos && !self->flushing && !pad->flushing) {
++    /* we can be unlocked here when we are shutting down (flushing) or when we
++     * get unblocked */
++    GST_INPUT_SELECTOR_WAIT (self);
++  }
++  return self->flushing;
++}
++
++static gboolean
+ gst_selector_pad_event (GstPad * pad, GstObject * parent, GstEvent * event)
+ {
+   gboolean res = TRUE;
+@@ -486,6 +497,7 @@ gst_selector_pad_event (GstPad * pad, GstObject * parent, GstEvent * event)
+     case GST_EVENT_FLUSH_START:
+       /* Unblock the pad if it's waiting */
+       selpad->flushing = TRUE;
++      sel->eos = FALSE;
+       GST_INPUT_SELECTOR_BROADCAST (sel);
+       break;
+     case GST_EVENT_FLUSH_STOP:
+@@ -523,21 +535,12 @@ gst_selector_pad_event (GstPad * pad, GstObject * parent, GstEvent * event)
+     case GST_EVENT_EOS:
+       selpad->eos = TRUE;
+ 
+-      if (forward) {
+-        selpad->eos_sent = TRUE;
+-      } else {
+-        GstSelectorPad *active_selpad;
+-
+-        /* If the active sinkpad is in EOS state but EOS
+-         * was not sent downstream this means that the pad
+-         * got EOS before it was set as active pad and that
+-         * the previously active pad got EOS after it was
+-         * active
+-         */
+-        active_selpad = GST_SELECTOR_PAD (active_sinkpad);
+-        forward = (active_selpad->eos && !active_selpad->eos_sent);
+-        active_selpad->eos_sent = TRUE;
++      if (!forward) {
++        /* blocked until active the sind pad or flush */
++        gst_input_selector_eos_wait (sel, selpad);
++        forward = TRUE;
+       }
++      selpad->eos_sent = TRUE;
+       GST_DEBUG_OBJECT (pad, "received EOS");
+       break;
+     case GST_EVENT_GAP:{
+@@ -676,6 +679,12 @@ gst_input_selector_wait_running_time (GstInputSelector * sel,
+         gst_input_selector_activate_sinkpad (sel, GST_PAD_CAST (selpad));
+     active_selpad = GST_SELECTOR_PAD_CAST (active_sinkpad);
+ 
++    if (sel->eos) {
++      GST_DEBUG_OBJECT (sel, "Not waiting because inputselector reach EOS.");
++      GST_INPUT_SELECTOR_UNLOCK (sel);
++      return FALSE;
++    }
++
+     if (seg->format != GST_FORMAT_TIME) {
+       GST_DEBUG_OBJECT (selpad,
+           "Not waiting because we don't have a TIME segment");
+@@ -971,6 +980,12 @@ gst_selector_pad_chain (GstPad * pad, GstObject * parent, GstBuffer * buf)
+       GST_TIME_ARGS (GST_BUFFER_TIMESTAMP (buf)));
+ 
+   GST_INPUT_SELECTOR_LOCK (sel);
++  if (sel->eos) {
++    GST_DEBUG_OBJECT (pad, "inputselector eos.");
++    GST_INPUT_SELECTOR_UNLOCK (sel);
++    goto eos;
++  }
++
+   /* wait or check for flushing */
+   if (gst_input_selector_wait (sel, selpad)) {
+     GST_INPUT_SELECTOR_UNLOCK (sel);
+@@ -1151,6 +1166,13 @@ flushing:
+     res = GST_FLOW_FLUSHING;
+     goto done;
+   }
++eos:
++  {
++    GST_DEBUG_OBJECT (pad, "We are eos, discard buffer %p", buf);
++    gst_buffer_unref (buf);
++    res = GST_FLOW_EOS;
++    goto done;
++  }
+ }
+ 
+ static void gst_input_selector_dispose (GObject * object);
+@@ -1309,6 +1331,7 @@ gst_input_selector_init (GstInputSelector * sel)
+   g_mutex_init (&sel->lock);
+   g_cond_init (&sel->cond);
+   sel->blocked = FALSE;
++  sel->eos = FALSE;
+ 
+   /* lets give a change for downstream to do something on
+    * active-pad change before we start pushing new buffers */
+@@ -1377,6 +1400,11 @@ gst_input_selector_set_active_pad (GstInputSelector * self, GstPad * pad)
+   GST_DEBUG_OBJECT (self, "New active pad is %" GST_PTR_FORMAT,
+       self->active_sinkpad);
+ 
++  if (old != new && new->eos && !new->eos_sent) {
++    self->eos = TRUE;
++    GST_INPUT_SELECTOR_BROADCAST (self);
++  }
++
+   return TRUE;
+ }
+ 
+@@ -1771,6 +1799,7 @@ gst_input_selector_change_state (GstElement * element,
+   switch (transition) {
+     case GST_STATE_CHANGE_READY_TO_PAUSED:
+       GST_INPUT_SELECTOR_LOCK (self);
++      self->eos = FALSE;
+       self->blocked = FALSE;
+       self->flushing = FALSE;
+       GST_INPUT_SELECTOR_UNLOCK (self);
+@@ -1779,6 +1808,7 @@ gst_input_selector_change_state (GstElement * element,
+       /* first unlock before we call the parent state change function, which
+        * tries to acquire the stream lock when going to ready. */
+       GST_INPUT_SELECTOR_LOCK (self);
++      self->eos = TRUE;
+       self->blocked = FALSE;
+       self->flushing = TRUE;
+       GST_INPUT_SELECTOR_BROADCAST (self);
+diff --git a/plugins/elements/gstinputselector.h b/plugins/elements/gstinputselector.h
+index 96c680f..9bf924f 100644
+--- a/plugins/elements/gstinputselector.h
++++ b/plugins/elements/gstinputselector.h
+@@ -77,6 +77,7 @@ struct _GstInputSelector {
+   GMutex lock;
+   GCond cond;
+   gboolean blocked;
++  gboolean eos;
+   gboolean flushing;
+ };
+ 
+-- 
+1.7.9.5
+
diff --git a/meta/recipes-multimedia/gstreamer/gstreamer1.0_1.4.5.bb b/meta/recipes-multimedia/gstreamer/gstreamer1.0_1.4.5.bb
index 0a2ce86..902f79d 100644
--- a/meta/recipes-multimedia/gstreamer/gstreamer1.0_1.4.5.bb
+++ b/meta/recipes-multimedia/gstreamer/gstreamer1.0_1.4.5.bb
@@ -7,6 +7,7 @@ SRC_URI = " \
     http://gstreamer.freedesktop.org/src/gstreamer/gstreamer-${PV}.tar.xz \
     file://0001-Fix-crash-with-gst-inspect.patch \
     file://0001-gstinfo-Shorten-__FILE__-on-all-platforms.patch \
+    file://inputselector-sticky-events-haven-t-send-out-when-ac-1-4-1.patch \
 "
 SRC_URI[md5sum] = "88a9289c64a4950ebb4f544980234289"
 SRC_URI[sha256sum] = "40801aa7f979024526258a0e94707ba42b8ab6f7d2206e56adbc4433155cb0ae"
-- 
1.9.1

[poky][PATCH v4 2/2] gstreamer1.0: Fix QoS/lateness checking if subclass implements prepare/prepare_list vfuncs

[Yuqing Zhu]

In function gst_base_sink_chain_unlocked(), it should calculate jitter based
on current media clock, rather than just passing 0.
Or it will drop all the frames when rewind in slow speed, such as -2X.

Signed-off-by: Yuqing Zhu <b54851@freescale.com>
---
 ...x-QoS-lateness-checking-if-subclass-imple.patch | 70 ++++++++++++++++++++++
 .../gstreamer/gstreamer1.0_1.4.5.bb                |  1 +
 2 files changed, 71 insertions(+)
 create mode 100644 meta/recipes-multimedia/gstreamer/gstreamer1.0/0002-basesink-Fix-QoS-lateness-checking-if-subclass-imple.patch

diff --git a/meta/recipes-multimedia/gstreamer/gstreamer1.0/0002-basesink-Fix-QoS-lateness-checking-if-subclass-imple.patch b/meta/recipes-multimedia/gstreamer/gstreamer1.0/0002-basesink-Fix-QoS-lateness-checking-if-subclass-imple.patch
new file mode 100644
index 0000000..b6edda1
--- /dev/null
+++ b/meta/recipes-multimedia/gstreamer/gstreamer1.0/0002-basesink-Fix-QoS-lateness-checking-if-subclass-imple.patch
@@ -0,0 +1,70 @@
+From 6914566ed6a89c96973a578aa5ecd01ee68cdcfd Mon Sep 17 00:00:00 2001
+From: Jian <Jian.Li@freescale.com>
+Date: Thu, 14 May 2015 15:49:43 +0800
+Subject: [PATCH] basesink: Fix QoS/lateness checking if subclass implements
+ prepare/prepare_list vfuncs
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+In basesink functions gst_base_sink_chain_unlocked(), below code is used to
+checking if buffer is late before doing prepare call to save some effort:
+    if (syncable && do_sync)
+      late =
+          gst_base_sink_is_too_late (basesink, obj, rstart, rstop,
+          GST_CLOCK_EARLY, 0, FALSE);
+
+    if (G_UNLIKELY (late))
+      goto dropped;
+
+But this code has problem, it should calculate jitter based on current media
+clock, rather than just passing 0. I found it will drop all the frames when
+rewind in slow speed, such as -2X.
+
+https://bugzilla.gnome.org/show_bug.cgi?id=749258
+
+Upstream-Status: Backport [1.5.1]
+---
+ libs/gst/base/gstbasesink.c |   26 ++++++++++++++++++++++----
+ 1 file changed, 22 insertions(+), 4 deletions(-)
+
+diff --git a/libs/gst/base/gstbasesink.c b/libs/gst/base/gstbasesink.c
+index a505695..5fb2d6a 100644
+--- a/libs/gst/base/gstbasesink.c
++++ b/libs/gst/base/gstbasesink.c
+@@ -3369,10 +3369,28 @@ gst_base_sink_chain_unlocked (GstBaseSink * basesink, GstPad * pad,
+     if (G_UNLIKELY (stepped))
+       goto dropped;
+ 
+-    if (syncable && do_sync)
+-      late =
+-          gst_base_sink_is_too_late (basesink, obj, rstart, rstop,
+-          GST_CLOCK_EARLY, 0, FALSE);
++    if (syncable && do_sync) {
++      GstClock *clock;
++
++      GST_OBJECT_LOCK (basesink);
++      clock = GST_ELEMENT_CLOCK (basesink);
++      if (clock && GST_STATE (basesink) == GST_STATE_PLAYING) {
++        GstClockTime base_time;
++        GstClockTime stime;
++        GstClockTime now;
++
++        base_time = GST_ELEMENT_CAST (basesink)->base_time;
++        stime = base_time + gst_base_sink_adjust_time (basesink, rstart);
++        now = gst_clock_get_time (clock);
++        GST_OBJECT_UNLOCK (basesink);
++
++        late =
++            gst_base_sink_is_too_late (basesink, obj, rstart, rstop,
++            GST_CLOCK_EARLY, GST_CLOCK_DIFF (stime, now), FALSE);
++      } else {
++        GST_OBJECT_UNLOCK (basesink);
++      }
++    }
+ 
+     if (G_UNLIKELY (late))
+       goto dropped;
+-- 
+1.7.9.5
+
diff --git a/meta/recipes-multimedia/gstreamer/gstreamer1.0_1.4.5.bb b/meta/recipes-multimedia/gstreamer/gstreamer1.0_1.4.5.bb
index 902f79d..db58754 100644
--- a/meta/recipes-multimedia/gstreamer/gstreamer1.0_1.4.5.bb
+++ b/meta/recipes-multimedia/gstreamer/gstreamer1.0_1.4.5.bb
@@ -8,6 +8,7 @@ SRC_URI = " \
     file://0001-Fix-crash-with-gst-inspect.patch \
     file://0001-gstinfo-Shorten-__FILE__-on-all-platforms.patch \
     file://inputselector-sticky-events-haven-t-send-out-when-ac-1-4-1.patch \
+    file://0002-basesink-Fix-QoS-lateness-checking-if-subclass-imple.patch \
 "
 SRC_URI[md5sum] = "88a9289c64a4950ebb4f544980234289"
 SRC_URI[sha256sum] = "40801aa7f979024526258a0e94707ba42b8ab6f7d2206e56adbc4433155cb0ae"
-- 
1.9.1

Re: [poky][PATCH v4 2/2] gstreamer1.0: Fix QoS/lateness checking if subclass implements prepare/prepare_list vfuncs

[Otavio Salvador]

On Fri, Aug 21, 2015 at 11:29 AM, Yuqing Zhu <b54851@freescale.com> wrote:
> In function gst_base_sink_chain_unlocked(), it should calculate jitter based
> on current media clock, rather than just passing 0.
> Or it will drop all the frames when rewind in slow speed, such as -2X.
>
> Signed-off-by: Yuqing Zhu <b54851@freescale.com>

Acked-by: Otavio Salvador <otavio@ossystems.com.br>

+Carlos for his feedback on this one as well.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: [poky][PATCH v4 0/2] gstreamer1.0: Add patches for Gstreamer 1.4.5

[Carlos Rafael Giani]

These are backports, so I do not see a problem with them. However, keep 
in mind that GStreamer 1.6 will be out in a few days, so I recommend 
hold off any patches that are not backports. I have recipes prepared for 
1.5.90 (= 1.6 release candidate) that I will then send to the mailing 
list. New patches should be applied on top of that, not 1.4.5.


Am 2015-08-21 um 16:29 schrieb Yuqing Zhu:
> Fix sticky events haven't been sent out when active track reach EOS
>
> Fix QoS/lateness checking if subclass implements prepare/prepare_list vfuncs
>
>
> Yuqing Zhu (2):
>    gstreamer1.0: Fix ticky events haven't been sent out when active track
>      reach EOS
>    gstreamer1.0: Fix QoS/lateness checking if subclass implements
>      prepare/prepare_list vfuncs
>
>   ...x-QoS-lateness-checking-if-subclass-imple.patch |  70 +++++++++
>   ...cky-events-haven-t-send-out-when-ac-1-4-1.patch | 167 +++++++++++++++++++++
>   .../gstreamer/gstreamer1.0_1.4.5.bb                |   2 +
>   3 files changed, 239 insertions(+)
>   create mode 100644 meta/recipes-multimedia/gstreamer/gstreamer1.0/0002-basesink-Fix-QoS-lateness-checking-if-subclass-imple.patch
>   create mode 100755 meta/recipes-multimedia/gstreamer/gstreamer1.0/inputselector-sticky-events-haven-t-send-out-when-ac-1-4-1.patch
>

Re: [poky][PATCH v4 0/2] gstreamer1.0: Add patches for Gstreamer 1.4.5

[Otavio Salvador]

On Fri, Aug 21, 2015 at 6:57 PM, Carlos Rafael Giani
<dv@pseudoterminal.org> wrote:
> These are backports, so I do not see a problem with them. However, keep in
> mind that GStreamer 1.6 will be out in a few days, so I recommend hold off
> any patches that are not backports. I have recipes prepared for 1.5.90 (=
> 1.6 release candidate) that I will then send to the mailing list. New
> patches should be applied on top of that, not 1.4.5.

In this case it is better to not merge those and wait for 1.5.90 to be
send. This allow for testing and also FSL checking for any other
pending fixes.


-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: [poky][PATCH v4 0/2] gstreamer1.0: Add patches for Gstreamer 1.4.5

[Carlos Rafael Giani]

Am 2015-08-22 um 00:01 schrieb Otavio Salvador:
> On Fri, Aug 21, 2015 at 6:57 PM, Carlos Rafael Giani
> <dv@pseudoterminal.org> wrote:
>> These are backports, so I do not see a problem with them. However, keep in
>> mind that GStreamer 1.6 will be out in a few days, so I recommend hold off
>> any patches that are not backports. I have recipes prepared for 1.5.90 (=
>> 1.6 release candidate) that I will then send to the mailing list. New
>> patches should be applied on top of that, not 1.4.5.
> In this case it is better to not merge those and wait for 1.5.90 to be
> send. This allow for testing and also FSL checking for any other
> pending fixes.
>
>

Yes. In fact, several of the present non-backport patches no longer 
apply. I had to delete them because fixing them was a nontrivial task, 
and should better be done by the original authors. I already sent a 
heads up about this to them.

Re: [poky][PATCH v4 0/2] gstreamer1.0: Add patches for Gstreamer 1.4.5

[Zhu Carol]

Hi Carlos,
   If I didn't misunderstand, you meant it's ok for you because they are backports. 
   So community will accept these two patches if no other problem. Am I right?
   The patches are still not being applied to master branch, I am waiting for your feedback.
 
B.R.
Yuqing Zhu


-----Original Message-----
From: openembedded-core-bounces@lists.openembedded.org [mailto:openembedded-core-bounces@lists.openembedded.org] On Behalf Of Carlos Rafael Giani
Sent: Saturday, August 22, 2015 5:58 AM
To: openembedded-core@lists.openembedded.org
Subject: Re: [OE-core] [poky][PATCH v4 0/2] gstreamer1.0: Add patches for Gstreamer 1.4.5

These are backports, so I do not see a problem with them. However, keep in mind that GStreamer 1.6 will be out in a few days, so I recommend hold off any patches that are not backports. I have recipes prepared for
1.5.90 (= 1.6 release candidate) that I will then send to the mailing list. New patches should be applied on top of that, not 1.4.5.


Am 2015-08-21 um 16:29 schrieb Yuqing Zhu:
> Fix sticky events haven't been sent out when active track reach EOS
>
> Fix QoS/lateness checking if subclass implements prepare/prepare_list 
> vfuncs
>
>
> Yuqing Zhu (2):
>    gstreamer1.0: Fix ticky events haven't been sent out when active track
>      reach EOS
>    gstreamer1.0: Fix QoS/lateness checking if subclass implements
>      prepare/prepare_list vfuncs
>
>   ...x-QoS-lateness-checking-if-subclass-imple.patch |  70 +++++++++
>   ...cky-events-haven-t-send-out-when-ac-1-4-1.patch | 167 +++++++++++++++++++++
>   .../gstreamer/gstreamer1.0_1.4.5.bb                |   2 +
>   3 files changed, 239 insertions(+)
>   create mode 100644 meta/recipes-multimedia/gstreamer/gstreamer1.0/0002-basesink-Fix-QoS-lateness-checking-if-subclass-imple.patch
>   create mode 100755 
> meta/recipes-multimedia/gstreamer/gstreamer1.0/inputselector-sticky-ev
> ents-haven-t-send-out-when-ac-1-4-1.patch
>

--
_______________________________________________
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.openembedded.org/mailman/listinfo/openembedded-core