The urn:mpeg:dash:utc:http-head:2014 method of time synchronisation
uses an HTTP HEAD request to a specified URL and then parses the
Date: HTTP response header.
This commit adds support to dashdemux for this method of time
synchronisation by making a HEAD request and then parsing the Date:
response.
This commit adds support to gstfragment to return the HTTP headers
and to uridownloader to support HEAD requests. To avoid creating a
new API, the RANGE get function is re-used (abused?) with start=-1
and end=-1 to indicate a HEAD request.
https://bugzilla.gnome.org/show_bug.cgi?id=752413
This commit addresses the following items from the code review:
use a portable way to define NTP_TO_UNIX_EPOCH,
fix memory leak on error, and
add documentation to UTCTiming parse functions
Using LL is not portable, so the G_GUINT64_CONSTANT needs to be instead.
If an error occurs during DNS resolution, the GError was not being
released, causing a memory leak.
https://bugzilla.gnome.org/show_bug.cgi?id=752413
Unless the DASH client can compensate for the difference between its
clock and the clock used by the server, the client might request
fragments that either not yet on the server or fragments that have
already been expired from the server. This is an issue because these
requests can propagate all the way back to the origin
ISO/IEC 23009-1:2014/Amd 1 [PDAM1] defines a new UTCTiming element to allow
a DASH client to track the clock used by the server generating the
DASH stream. Multiple UTCTiming elements might be present, to indicate
support for multiple methods of UTC time gathering. Each element can
contain a white space separated list of URLs that can be contacted
to discover the UTC time from the server's perspective.
This commit provides parsing of UTCTiming elements, unit tests of this
parsing and a function to poll a time server. This function
supports the following methods:
urn:mpeg:dash:utc:ntp:2014
urn:mpeg:dash:utc:http-xsdate:2014
urn:mpeg:dash:utc:http-iso:2014
urn:mpeg:dash:utc:http-ntp:2014
The manifest update task is used to poll the clock time server,
to save having to create a new thread.
When choosing the starting fragment number and when waiting for a
fragment to become available, the difference between the server's idea
of UTC and the client's idea of UTC is taken into account. For example,
if the server's time is behind the client's idea of UTC, we wait for
longer before requesting a fragment
[PDAM1]: http://www.iso.org/iso/home/store/catalogue_tc/catalogue_detail.htm?csnumber=66068
dashdemux: support NTP time servers in UTCTiming elements
Use the gst_ntp_clock to support the use of an NTP server.
https://bugzilla.gnome.org/show_bug.cgi?id=752413
This reverts commit ff11a1a8a0.
It can't be assumed that all buffers in a buffer list have the same SSRC or
are RTP or RTCP only. It has to be checked for every single buffer, and one
basically has to do the processing that is done by the default chain_list
implementation.
The current code was ignoring the par/dar aspect when transforming
from window coordinates to stream coordinates resulting in incorrect
coordinates being sent upstream in the navigation events.
The payloader didn't copy anything so far, the depayloader copied every
possible meta. Let's make it consistent and just copy all metas without tags or
with only the audio tag.
https://bugzilla.gnome.org/show_bug.cgi?id=751774
Checking the vector is not empty and checking the vector size is greater
than zero are the same thing, this is a redundancy in the code. Only
checking the vector is not empty is sufficient, therefore removing the
other check.
https://bugzilla.gnome.org/show_bug.cgi?id=744763
Add a pivot vector for setting the origin of rotations and scales.
With the pivot point the rotation and scale operations can have
different origins. This adds the ability to rotate around different points.
Currently the default (0, 0) pivot point is possible,
a rotation around the center, and zooming into and out of the center.
With an pivot point this is optional.
I defined the following image coordinates for the pivot point:
(-1,1) ------------------------- (1,1)
| |
| |
| |
| (0,0) |
| |
| |
| |
(-1,-1) ------------------------- (1,-1)
Example:
Rotate the video at the bottom left corner
gst-launch-1.0 videotestsrc \
! gltransformation \
scale-x=0.5 \
scale-y=0.5 \
rotation-z=25.0 \
pivot-x=-1.0 \
pivot-y=-1.0 \
! glimagesink
The pivot-z option defines the pivot point in 3D space.
This only affects rotation, since we have no Z data to scale.
With this option a video can be rotated around a point in 3D space.
Example:
Rotate around point behind the video:
gst-launch-1.0 videotestsrc \
! gltransformation \
rotation-x=10.0 \
pivot-z=-4.0 \
! glimagesink
Since the profile gchar depends on DEFAULT_FACE_PROFILE, it should never be
NULL. Furthermore CascadeClassifier accepts any input, even
an empty one, but if the profile fails to load it returns an empty cascade.
Check for this instead, and inform the user if there was an Error.
The gst_dash_demux_stream_update_fragment_info function could call
gst_dash_demux_stream_update_headers_info function twice. The
gst_dash_demux_stream_update_headers_info function will set header_uri and
index_uri to some newly allocated strings. The values set by the first call of
gst_dash_demux_stream_update_headers_info will leak when the function is
called for a second time.
The solution is to call gst_adaptive_demux_stream_fragment_clear before the
second call of gst_dash_demux_stream_update_headers_info
https://bugzilla.gnome.org/show_bug.cgi?id=753188
Check if profile is NULL before dereferencing it with new. Also, new will
never return NULL; if allocation fails, a std::bad_alloc exception will be
thrown instead. Remove check for a NULL return.
CID #1315258
The ref_object and object parameters were the wrong way around.
For the typical use case where an application is setting a
GstControlBinding on the returned ghost pad:
1. our control binding would be removed when the new one was set
2. sync_values calls were not being forwarded from the internal
pad to the ghost pad.
If an application attempts to perform other control binding
operations (get_* family of functions) on the internal pad, they
will also be forwarded to the ghost pad where a possible
GstControlBinding will provide the necessary values.
Only copy the values from the parent if the current node doesn't
have that value, they were being copied from the parent and
then overwriten by the child node, leaking the parent's copy
With the switch of gstopencv.c to C++, all OpenCV elements are built with
g++. The template variable clashes with C++'s feature of the same name.
Rename template to templ to avoid any clash.
The cascade classifier changes its structure on new version of OpenCV.
The need to migrate to C++ to utilize the new load method of OpenCV which
allows to load the new classifiers.
https://bugzilla.gnome.org/show_bug.cgi?id=748377
This is used to proxy GstControlBinding to the pad on the
parent object. This avoid having to sync the values in the proxy pad,
this is too early if you have a queue between the pad and the actual
aggregation operation.
https://bugzilla.gnome.org/show_bug.cgi?id=734060
When the sink does not know the window size (e.g not created yet)
it will not add any param to the the composition meta. This is no
reason not to forward this meta API. Fixes issue where it could not
attach until we resize the window.
https://bugzilla.gnome.org/show_bug.cgi?id=745107
The coordinate are relative to the texture dimension and not
the window dimension now. There is no need to pass the window
dimension or to update the overlay if the dimension changes.
https://bugzilla.gnome.org/show_bug.cgi?id=745107
Adds an GST_VIDEO_OVERLAY_COMPOSITION_META_API_TYPE query to glupload
and glimagesink. Detects the query from the downstream elements, so
it is executed only when downstream supports the overlay API.
This makes pipelines with textoverlay ! glupload ! gldownload ! xvimagesink possible.
Uses allocation meta struct for passing the window size upstream.
https://bugzilla.gnome.org/show_bug.cgi?id=745107
Previously, PLC frames always had a length of 120ms, which caused audio
quality degradation and synchronization errors. Fix this by calculating an
appropriate length for the PLC frame.
The length must be a multiple of 2.5ms. Calculate a multiple of 2.5ms that
is nearest to the current PLC length. Any leftover PLC length that didn't
make it into this frame is accumulated for the next PLC frame.
https://bugzilla.gnome.org/show_bug.cgi?id=725167
If a ContentProtection element is present in an AdaptationSet element,
send Protection events on the source pad, so that qtdemux can use this
information to correctly generate its source caps for DASH CENC
encrypted streams.
This allows qtdemux to support CENC encrypted DASH streams where the
content protection specific information is carried in the MPD file
rather than in pssh boxes in the initialisation segments.
This commit adds a new function to the adaptivedemux base class to allow
a GstEvent to be queued for a stream. The queue of events are sent the
next time a buffer is pushed for that stream.
https://bugzilla.gnome.org/show_bug.cgi?id=705991
They require to get_proc_address some functions through the
platform specific {glX,egl}GetProcAddress rather than the default
GL library symbol lookup.
Document that "widget" property must be accessed from the
main thread (where GTK is running). This is the same for
state transition on these elements. It is very natural to
do so un GTK applications.
This is a "pseudo" base class. Basically it's a shared instance
and class structure and a shared set of function between the
two widget. It cannot have it's own type like normal base class
since the one instance will implement GtkGLArea while the other
implements GtkDrawingAreay. To workaround this, the parent instance
and class is a union of both.
https://bugzilla.gnome.org/show_bug.cgi?id=752441
I notice that if you stop the pipeline during a renegotiation
the upload may be NULL while an allocation query is being run.
In that scenario, returning FALSE to the allocation query is the
best thing.
Move back the default property at the same place they are in the
other sink. This helps when using a diff viewer to synchronized
this unfortunate copy paste.
https://bugzilla.gnome.org/show_bug.cgi?id=751104
In GTK dispose can be called before the last ref is reached. This
happens when you close the container window. The dispose will be
explicitly called, and destroyed notify will be fired. This patch
fixes this race by properly tracking the widget state.
In the sink, we now set the widget pointer to NULL, so the widget
will properly get created again if you set your pipeline to NULL
state after the widget was destroy, and set it back to PLAYING.
https://bugzilla.gnome.org/show_bug.cgi?id=751104
Very much in the same spirit as the Gtk GL sink
Two things are provided
1. A QQuickItem subclass that renders out RGBA filled GstGLMemory
buffers that is instantiated from qml.
2. A sink element that will push buffers into (1)
To use
1. Declare the GstGLVideoItem in qml with an appropriate
objectName property set.
2. Get the aforementioned GstGLVideoItem from qml using something like
QQmlApplicationEngine engine;
engine.load(QUrl(QStringLiteral("qrc:/main.qml")));
QObject *rootObject = engine.rootObjects().first();
QQuickItem *videoItem = rootObject->findChild<QQuickItem *> ("videoItem");
3. Set the videoItem on the sink
https://bugzilla.gnome.org/show_bug.cgi?id=752185
Checking for a parent is not enough, it must have a toplevel one.
If widget has no toplevel parent then add it in a GtkWindow, that
make it usable from gst-launch-1.0.
https://bugzilla.gnome.org/show_bug.cgi?id=751104
Moved gst_mpd_client_get_next_segment_availability_end_time and
gst_mpd_client_add_time_difference functions to be grouped with
functions from the same category.
https://bugzilla.gnome.org/show_bug.cgi?id=752027
Corrected the initialisation of mimeType in
gst_mpdparser_get_list_and_nb_of_audio_language: the variable is used
in a loop, so it must be set to NULL at the beginning of each iteration.
https://bugzilla.gnome.org/show_bug.cgi?id=751911
Before returning the next fragment duration value, the
gst_mpd_client_get_next_fragment_duration function tries to validate it.
But the condition was incorrect.
https://bugzilla.gnome.org/show_bug.cgi?id=751539
We're interested in the offset between the period start timestamp and the
actual media timestamp so that we can properly correct for it. The absolute
presentation offset to timestamp 0 is useless as the only thing we really
care about is the offset between the current fragment timestamp and the
media timestamp.
Otherwise we will look for segments after the period usually. The seek
timestamp is relative to the start of the first period and we have to
select a segment relative to the current period's start.
We didn't do this for fragments that are generated on demand from a template,
only for the other cases when they were all generated upfront. This caused
fragment timestamps to start from 0 again for each new period.
If not set, the timeShiftBufferDepth has a default value of -1.
The standard says that this should be interpreted as infinite.
The gst_mpd_client_check_time_position function incorrectly compares
timeShiftBufferDepth with 0 instead of -1 to determine if it was set.
https://bugzilla.gnome.org/show_bug.cgi?id=751500
The last parameter of gst_mpd_client_add_media_segment function is a
duration. But when called from gst_mpd_client_setup_representation, the
last argument was wrongly set to PeriodEnd
https://bugzilla.gnome.org/show_bug.cgi?id=751449
The period start information, calculated in gst_mpd_client_setup_media_presentation
function is stored in stream_period->start. The information read from
xml file and stored in stream_period->period->start is not changed.
If the xml file does not contain the period start information,
stream_period->period->start will be -1.
The function gst_mpd_client_get_next_segment_availability_end_time wants to
use period start time, but incorrectly uses stream_period->period->start
(value from xml file, which could be -1) instead of stream_period->start
(computed value)
https://bugzilla.gnome.org/show_bug.cgi?id=751465
According to ISO/IEC 23009-1:2014(E), chapter 5.3.2.1
"The Period extends until the PeriodStart of the next Period, or until
the end of the Media Presentation in the case of the last Period."
This means that a configured value for optional attribute period duration
should be ignored if the next period contains a start attribute or it is
the last period and the MPD contains a mediaPresentationDuration attribute.
https://bugzilla.gnome.org/show_bug.cgi?id=750797
Support video with multiview info in the caps, transform
it to mono anaglyph by default, but allow for configuring
other output modes and handoff to the app via
the draw signal.
https://bugzilla.gnome.org/show_bug.cgi?id=611157
Added some warning messages in gst_mpd_client_setup_streaming to help
debug situations when the function will return FALSE.
Renamed a wrongly spelled variable.
https://bugzilla.gnome.org/show_bug.cgi?id=751149
Corrected some comments in gstmpdparser.h file.
Moved gst_mpd_client_get_adaptation_sets function to be grouped with
other functions from AdaptationSet group
https://bugzilla.gnome.org/show_bug.cgi?id=751149
The gst_mpdparser_get_rep_idx_with_max_bandwidth function assumes
representations are ordered by bandwidth and incorrectly returns the
first one when wanting the one with minimum bandwidth.
Corrected gst_mpdparser_get_rep_idx_with_max_bandwidth function to get the
correct representation in case max_bandwidth parameter is 0.
https://bugzilla.gnome.org/show_bug.cgi?id=751153
Getting the current viewport and modifying it relatively will produce an
interesting feedback loop during widget resizing. Over a few frames we
will gradually move the viewport a bit until it converged again, adding
unnecessary additional borders at the top and left.
We now know that pool caching can cause renegotiation issues
when an element in the pipeline change from passthrough to not
passthrough. As it's not needed, don't cache existing pools.
https://bugzilla.gnome.org/show_bug.cgi?id=748344
Added a check for a_node->ns before accessing a_node->ns->href in
gst_mpdparser_get_xml_node_namespace. This could happen if the xml
is missing the default namespace.
https://bugzilla.gnome.org/show_bug.cgi?id=750866
If the presentationTimeOffset attribute of a DASH manifest contains
a value that is larger than 2^32, gstmpdparser incorrectly calculates
the stream's presentation time offset. This is due to two bugs:
1: Using gst_mpdparser_get_xml_prop_unsigned_integer rather than
gst_mpdparser_get_xml_prop_unsigned_integer_64 to parse the
attribute
2: gst_mpd_client_setup_representation multiplying the value by
GST_SECOND and then dividing by timescale
https://bugzilla.gnome.org/show_bug.cgi?id=750804
This patch allow going gst-inspect-1.0 on these elements removing
ugly crash that was previously occurring. The method consist of
making the widget creation as lazy as possible. This way we don't
endup doing gtk_init() before the application. We also ref_sink()
the widget, so we don't crash if the parent widget is discarded,
and cleanly error out with GL if the widget has no parent window,
because calling gtk_widget_realized() can only be done if the widget
has been parented to a window).
This reverts commit 4ca3a22b6b.
The connection-speed=0 is used as a special value in the property
of hlsdemux to mean 'automatic' selection, m3u8.c doesn't need
to know about that as it should be as simple as possible.
So this patch hides this automatic selection documented in hlsdemux
into m3u8 logic and I think the gets harder to understand the code.
It also makes the hlsdemux unit tests work again
https://bugzilla.gnome.org/show_bug.cgi?id=749328
This reverts commit 37011e5198.
This change was actually completely unnecessary, the streams in question are
marked as static and are not considered live anyway.
Otherwise we'll only get half of its bits printed on 32 bit architectures.
For this, promote the %d-style format strings to something that accepts
64 bit integers with G_GINT64_MODIFIER.
Using format strings from an untrusted source without validation is
calling for problems, and at least allows to remotely crash your application.
If not worse.
In live situations, it is not uncommon for the current fragment to end
up out of the (updated) play range (lowest/highest sequence). But the next
fragment to play *is* present in the play range.
When advancing, if we can't find the current GstM3U8MediaFile, don't abort
straight away. Instead, look if a GstM3U8MediaFile with the next sequence value
is present, and if so switch to it.
https://bugzilla.gnome.org/show_bug.cgi?id=750028
Previously when compiling GstGL with both GL and GLES2,
GL_RGBA8 was picked from GL/gl.h. But a clash may happen at
runtime when one is selecting GLES2.
gst_gl_internal_format_rgba allows to check at runtime
if it should use GL_RGBA or GL_RGBA8.
The functions to get the next fragment, next fragment timestamp and to advance
to the next fragment need to work differently when stream->segments is NULL.
Use logic similar to that introduced by commit 2105a310 to perform these
functions.
https://bugzilla.gnome.org/show_bug.cgi?id=749684
Previously the VPS unit was detected and all next packets where copied
into the header buffer assuming only SPS and PPS would follow. This is
not always true, also other types of NAL units follow the VPS unit and
where copied to the header buffer. Now the VPS/SPS/PPS are explicitely
detected and copied in the header buffer.
1. Set the sync point after the (possible) upload has occured
2. Wait in the correct GL context (the draw context)
Note: We don't add the GL sync meta to the input buffer as it's not
writable and a copy would be expensive.
Similar to the change with the same name for glimagesink
1. Set the sync point after the (possible) upload has occured
2. Wait in the correct GL context (the draw context)
Note: We don't add the GL sync meta to the input buffer as it's not
writable and a copy would be expensive.
The property level has a minimum value of 0. But when we set the level as 0,
it gets an assertion error. The function icvPyrSegmentation8uC3R returns false
if level is set as 0, since the minimum level cant be 0 and thus results in error.
Hence changing the minimum value to 1.
https://bugzilla.gnome.org/show_bug.cgi?id=749525
When all fragments have already been downloaded on a live stream
dashdemux would busy loop as the default implementation of
has_next_fragment would return TRUE. Implement it to correctly
signal if adaptivedemux should wait for the manifest update before
trying to get new fragments.
When updating the manifest the timestamps on it might have changed a little
due to rounding and timescale conversions. If the change makes the timestamp
of the current segment to go up it makes dashdemux reposition to the previous
one causing one extra unnecessary download.
So when repositioning add an extra 10 microseconds to cover for that rounding
issues and increase the chance of falling in the same segment.
Additionally, also improve the time used when the client is already after the
last segment. Instead of using the last segment starting timestamp use the
final timestamp to make it reposition to the next one and not to the one that
has already been downloaded.
These functions of directly getting and setting segment indexes
are no longer useful as now we need 2 indexes: repeat and segment
index.
The only operations needed are advance_segment, going back to the
first one or seeking for a timestamp.
Segments are now stored with their repeat counts instead of spanding
them to multiple segments. This caused advancing to the next segment
using a single index to have to iterate over the whole list every time.
This commit addresses this by storing both the segment index as well
as the repeat index and makes advancing to next segment just an
increment of the repeat or the segment index.
Use a single segment to represent it internally to avoid using too
much memory. This has the drawback of issuing a linear search to
find the correct segment to play but this can be fixed by using
binary searches or caching the current position and just looking
for the next one.
https://bugzilla.gnome.org/show_bug.cgi?id=748369
The custom code is wrong as it ignores the templates, which leads to
missing fields in the result. Instead, simply use the default get_caps
implementation which does it correctly (get the template, intersect
with filter and return).
https://bugzilla.gnome.org/show_bug.cgi?id=749237
Without this, we will fixate weird pixel-aspect-ratios like 1/2147483647. But
in the end, all the negotiation code in videoaggregator needs a big cleanup
and videoaggregator needs to get rid of the software-mixer specific things
everywhere.
Upstream might not give us a caps event (dtlssrtpdec) because it might be an
RTP/RTCP mixed stream, but we split the two streams anyway and should report
proper caps downstream if possible.
Fixes "sticky event misordering" warnings with dtlssrtpdec.
And provide home-made fallback for older GLib versions,
so that we can later find these and remove them when
we bump the GLib requirement (which is certainly going
to happen before 2.0).
https://bugzilla.gnome.org/show_bug.cgi?id=748495
It's better to just select some random variant playlist instead of stopping,
chances are that it's still continuing to work and we might just have to
select a different variant again later.
We should only refresh the currently selected variant playlist (if any,
otherwise the main playlist), not the main playlist. And only try to
refresh the main playlist if updating the variant playlist fails.
Some servers (Wowza) use the request of the main playlist to create a
"session", which is then part of the URI of the variant playlist and
also the fragments. Refreshing the main playlist would generate a new
session, and the server rate limits that usually. And after a few retries
the server just kicks us out.
Also as a side effect we now use the same downloader for all playlists, so
that we only have 2 instead of 3 connections to the server. And also
previously we just ignored the downloaded data from the main playlist that
the base class gave to us.
