platform/upstream/gst-plugins-good.git
7 years agoqtdemux: Use upstream's StreamFlags if there are
Seungha Yang [Thu, 1 Dec 2016 03:47:08 +0000 (12:47 +0900)]
qtdemux: Use upstream's StreamFlags if there are

When multiple demuxer's are used, upstream might want to indicate
default streams using GST_STREAM_FLAG_{SELECT, UNSELECT}

https://bugzilla.gnome.org/show_bug.cgi?id=775440

7 years agoqtmux: Timecode track fixes for STSD entry
Vivia Nikolaidou [Fri, 27 Jan 2017 14:14:16 +0000 (16:14 +0200)]
qtmux: Timecode track fixes for STSD entry

The n_frames field (frames per second) should follow the nominal frame
rate for drop-frame timecodes.

Also, the trak's timescale (and duration, accordingly) should follow the
STSD entry's timescale and frame duration (fps_n and fps_d accordingly),
not the other way around.

https://bugzilla.gnome.org/show_bug.cgi?id=777832

7 years agosouphttpsrc: retry request on early termination from the server
Arnaud Vrac [Thu, 19 Jan 2017 10:08:11 +0000 (11:08 +0100)]
souphttpsrc: retry request on early termination from the server

Fix a regression introduced by commit 183695c61a54f1 (refactor to use
Soup's sync API). The code previously attempted to reconnect when the
server closed the connection early, for example when the stream was put
in pause for some time.

Reintroduce this feature by checking if EOS is received before the
expected content size is downloaded. In this case, do the request
starting at the previous read position.

https://bugzilla.gnome.org/show_bug.cgi?id=776720

7 years agortspsrc: find_stream_by_channel should ignore unconfigured streams
Matt Staples [Tue, 10 Jan 2017 16:40:56 +0000 (09:40 -0700)]
rtspsrc: find_stream_by_channel should ignore unconfigured streams

https://bugzilla.gnome.org/show_bug.cgi?id=777101

7 years agoqtmux: Fix debug typo and remove misleading warning
Brendan Shanks [Wed, 25 Jan 2017 18:43:00 +0000 (18:43 +0000)]
qtmux: Fix debug typo and remove misleading warning

https://bugzilla.gnome.org/show_bug.cgi?id=777362

7 years agortp: Remove unused variable in example
Sebastian Dröge [Wed, 25 Jan 2017 18:56:24 +0000 (20:56 +0200)]
rtp: Remove unused variable in example

client-PCMA.c:84:22: warning: unused variable 'isrc' [-Wunused-variable]
  GObject *session, *isrc, *osrc;
                     ^

7 years agowavparse: check for not NULL before clearing adapter
Rahul Bedarkar [Mon, 9 Jan 2017 06:02:35 +0000 (11:32 +0530)]
wavparse: check for not NULL before clearing adapter

In case wavparse receives a manually injected FLUSH_STOP event
while operating in pull mode we get criticals because we'd try
to clear a NULL adapter.

https://bugzilla.gnome.org/show_bug.cgi?id=777123

7 years agomeson: Properly use ':' for defining keywords
Thibault Saunier [Tue, 24 Jan 2017 22:23:44 +0000 (19:23 -0300)]
meson: Properly use ':' for defining keywords

7 years agov4l2allocator: reference memory before the buffer is queued
Jean-Christophe Trotin [Tue, 17 Jan 2017 15:41:58 +0000 (16:41 +0100)]
v4l2allocator: reference memory before the buffer is queued

In gst_v4l2_allocator_qbuf(), the memory is referenced after the
buffer is queued. Once queued (VIDIOC_QBUF), the buffer might be handled
by the V4L2 driver (e.g. decoded) and dequeued (gst_v4l2_allocator_dqbuf),
through a different thread, before the memory is referenced (gst_memory_ref).
In this case, in gst_v4l2_allocator_dqbuf(), the memory is unreferenced
(gst_memory_unref) before having been referenced: the memory refcount
reaches 0, and the memory is freed.
So, to avoid this crossing case, in gst_v4l2_allocator_qbuf(), the
memory shall be referenced before the buffer is queued.

https://bugzilla.gnome.org/show_bug.cgi?id=777399

7 years agoqtmux: Only write 4 byte zero padding to the Video Sample Description in MOV
Sebastian Dröge [Tue, 24 Jan 2017 15:59:59 +0000 (17:59 +0200)]
qtmux: Only write 4 byte zero padding to the Video Sample Description in MOV

For MP4 this is not defined, and it actually breaks things for MSE in
Chrome if we do this. For MOV this is required by some broken software
but the official specification says it's optional:
https://developer.apple.com/library/content/documentation/QuickTime/QTFF/QTFFChap3/qtff3.html

https://bugzilla.gnome.org/show_bug.cgi?id=777540

7 years agortpstats: Keep number of nacks sent/received per source
Santiago Carot-Nemesio [Mon, 2 Jan 2017 12:42:04 +0000 (13:42 +0100)]
rtpstats: Keep number of nacks sent/received per source

Currently, the nack packets sent or received are kept at session level,
which makes it impossible to distinguish how many of these packages were
sent/received per ssrc when several sources are in the same session. This
patch is aligned with the https://www.w3.org/TR/webrtc-stats/#dom-rtcrtpstreamstats

https://bugzilla.gnome.org/show_bug.cgi?id=776714

7 years agortph265pay: Fix handling of config-interval
Jonas Holmberg [Thu, 8 Dec 2016 14:59:33 +0000 (15:59 +0100)]
rtph265pay: Fix handling of config-interval

Insert VPS/SPS/PPS before the first NAL unit containing an I-frame in an
access unit only. If an access unit consists of several such NAL units
(tiles) VPS/SPS/PPS should only be inserted before the first of them so
that parameters are only updated between frames.

Do not insert VPS/SPS/PPS before P-frames when config-interval is -1.

https://bugzilla.gnome.org/show_bug.cgi?id=775817

7 years agosouphttpsrc: report a useful error message when soup_session_send fails
Arnaud Vrac [Thu, 19 Jan 2017 11:29:44 +0000 (12:29 +0100)]
souphttpsrc: report a useful error message when soup_session_send fails

This helps to understand cases where libsoup doesn't set the message
status code after running soup_session_send.

https://bugzilla.gnome.org/show_bug.cgi?id=777222

7 years agosouphttpsrc: properly check that seek range was respected
Arnaud Vrac [Thu, 19 Jan 2017 10:05:00 +0000 (11:05 +0100)]
souphttpsrc: properly check that seek range was respected

This check must be done only when we are sure the request was
successfully sent. soup_session_send() might fail without setting the
status code. In this case status code is 0 so we would only catch the
error after the seek range check. In this case we would report an error
saying that the seek range was not respected, instead of reporting the
underlying error that triggered the soup_session_send() failure.

https://bugzilla.gnome.org/attachment.cgi?bugid=777222

7 years agogdkpixbufoverlay: add a positioning coefficient pair
Mark Nauwelaerts [Mon, 9 Jan 2017 20:04:51 +0000 (21:04 +0100)]
gdkpixbufoverlay: add a positioning coefficient pair

... so as to allow one clearly defined (absolute) positioning mode
that can cater for a variety of absolute but also relative positioning
with respect to edge or center.

7 years agogdkpixbufoverlay: update composition in _before_transform
Mark Nauwelaerts [Sat, 21 Jan 2017 19:48:22 +0000 (20:48 +0100)]
gdkpixbufoverlay: update composition in _before_transform

... since we need to determine passthrough mode for buffer preparation before
calling into _transform_ip.

7 years agogdkpixbufoverlay: handle setting NULL gdkpixbuf
Mark Nauwelaerts [Sat, 7 Jan 2017 19:11:13 +0000 (20:11 +0100)]
gdkpixbufoverlay: handle setting NULL gdkpixbuf

... which is a clearer way to clear any current overlay, other than
fiddling with alpha or positioning properties to make it virtually go away.

7 years agoavidemux: Stop reading a ncdt sub-tag if it goes behind the surrounding tag
Sebastian Dröge [Fri, 20 Jan 2017 15:16:10 +0000 (17:16 +0200)]
avidemux: Stop reading a ncdt sub-tag if it goes behind the surrounding tag

https://bugzilla.gnome.org/show_bug.cgi?id=777532

7 years agoavidemux: Fix various out of bounds reads when parsing ncdt tags
Sebastian Dröge [Fri, 20 Jan 2017 05:58:26 +0000 (07:58 +0200)]
avidemux: Fix various out of bounds reads when parsing ncdt tags

https://bugzilla.gnome.org/show_bug.cgi?id=777500

7 years agoqtdemux: Increment current stts index whenever we finished one stts entry
Sebastian Dröge [Thu, 19 Jan 2017 11:46:58 +0000 (13:46 +0200)]
qtdemux: Increment current stts index whenever we finished one stts entry

Otherwise we could read more chunks than there are available, doing an
out of bounds read and potentially crash.

https://bugzilla.gnome.org/show_bug.cgi?id=777469

7 years agoRevert "qtdemux: Increment current stts index in all code paths after reading one...
Sebastian Dröge [Thu, 19 Jan 2017 11:25:53 +0000 (13:25 +0200)]
Revert "qtdemux: Increment current stts index in all code paths after reading one chunk"

This reverts commit 99d5d7570d0b53dad3bc8eb653b1320ee422aace. It broke
playback of various valid files.

7 years agoqtdemux: Increment current stts index in all code paths after reading one chunk
Sebastian Dröge [Thu, 19 Jan 2017 05:52:33 +0000 (07:52 +0200)]
qtdemux: Increment current stts index in all code paths after reading one chunk

Otherwise we could read more chunks than there are available, doing an
out of bounds read and potentially crash.

https://bugzilla.gnome.org/show_bug.cgi?id=777469

7 years agosouphttpsrc: Initialize return variable
Edward Hervey [Thu, 19 Jan 2017 07:37:37 +0000 (08:37 +0100)]
souphttpsrc: Initialize return variable

In the normal use-case we would end up with ret being unitialized
causing havoc.

https://bugzilla.gnome.org/show_bug.cgi?id=777222

7 years agoqtdemux: avoid XMP tag parsing fatal error.
David Warman [Fri, 13 Jan 2017 12:27:40 +0000 (12:27 +0000)]
qtdemux: avoid XMP tag parsing fatal error.

qtdemux_handle_xmp_taglist() requires a writable taglist,
but qtdemux->tag_list can become non-writable, specifically
after sending global tags (qtdemux.c:958), which adds a
second reference.  Ensure the list is made writable before
calling (make_writable will copy the list if necessary).

https://bugzilla.gnome.org/show_bug.cgi?id=766177

7 years agoqtdemux: rework taglist handling
Thiago Santos [Tue, 31 May 2016 16:17:45 +0000 (13:17 -0300)]
qtdemux: rework taglist handling

Keep taglist around during element existance to avoid having to
create it at different places before usage. Makes code simpler to handle.

https://bugzilla.gnome.org/show_bug.cgi?id=766177

7 years agosouphttpsrc: make flow return values handling clearer
Arnaud Vrac [Mon, 16 Jan 2017 10:58:02 +0000 (11:58 +0100)]
souphttpsrc: make flow return values handling clearer

The flow return values was stored in the element before because the
result had to be set from callbacks. This is not the case anymore, we
can return the flow result directly from functions, making the code
easier to understand.

https://bugzilla.gnome.org/show_bug.cgi?id=777222

7 years agosouphttpsrc: properly track redirections
Arnaud Vrac [Fri, 13 Jan 2017 15:40:43 +0000 (16:40 +0100)]
souphttpsrc: properly track redirections

The current code configures libsoup to handle redirections
transparently, without informing the caller, thus preventing the element
to record the redirect code and location uri.

Fix this by always setting the SOUP_MESSAGE_NO_REDIRECT, preventing
libsoup from handling the redirection. When we receive a redirection
request and libsoup can safely handle it, return a custom error which
triggers a retry with the new URI.

https://bugzilla.gnome.org/show_bug.cgi?id=777222

7 years agoqtmux: add 4444 and 4444xq variants to video/x-prores pad template caps
Aurélien Zanelli [Tue, 17 Jan 2017 09:53:39 +0000 (10:53 +0100)]
qtmux: add 4444 and 4444xq variants to video/x-prores pad template caps

They are handled since commit 7b565475bf551c53b8eed46f7086f3b372f1f6c4
(qt: Add support for ProRes 4444 XQ).

https://bugzilla.gnome.org/show_bug.cgi?id=777377

7 years agomatroska: Quiet a WARN when parsing push mode
Jan Schmidt [Mon, 16 Jan 2017 23:48:57 +0000 (10:48 +1100)]
matroska: Quiet a WARN when parsing push mode

This warning was noisy when returning EOS, which is
just used to indicate more data is needed from upstream.

7 years agoqtmux: Don't write Sync Sample Atom for ProRes
Georg Lippitsch [Mon, 16 Jan 2017 13:50:22 +0000 (14:50 +0100)]
qtmux: Don't write Sync Sample Atom for ProRes

https://bugzilla.gnome.org/show_bug.cgi?id=777331

7 years agov4l2: Remove usage and definition of LOG_CAPS macro
Enrico Jorns [Wed, 28 Jan 2015 07:58:26 +0000 (08:58 +0100)]
v4l2: Remove usage and definition of LOG_CAPS macro

Unlike former definitions of LOG_CAPS, the current implementation simply
expands to GST_DEBUG_OBJECT. The LOG_CAPS macro is rarely used and most
uses duplicate already existing GST_DEBUG_OBJECT lines. Therefore, the
caps are often printed twice which unnecessarily clutters the debug log.

Replace LOG_CAPS calls with GST_DEBUG_OBJECT, remove LOG_CAPS calls, and
delete the definition of LOG_CAPS.

https://bugzilla.gnome.org/show_bug.cgi?id=776899

7 years agov4l2bufferpool: remove duplicated line of code
Jean-Christophe Trotin [Mon, 16 Jan 2017 14:40:43 +0000 (15:40 +0100)]
v4l2bufferpool: remove duplicated line of code

https://bugzilla.gnome.org/show_bug.cgi?id=777330

7 years agov4l2allocator: fix memory type in allocator probe
Jean-Christophe Trotin [Mon, 16 Jan 2017 14:17:15 +0000 (15:17 +0100)]
v4l2allocator: fix memory type in allocator probe

The buffer memory type provided to the VIDIOC_CREATE_BUFS ioctl shall
be set with the value ("memory") given as input parameter of the
gst_v4l2_allocator_probe() function.

https://bugzilla.gnome.org/show_bug.cgi?id=777327

7 years agoflacenc: fix other icon counter check
Tim-Philipp Müller [Sat, 14 Jan 2017 15:27:19 +0000 (15:27 +0000)]
flacenc: fix other icon counter check

It's never going to be 0 if we first increment and then check.

7 years agoqtdemux: boldly assume that first 'covr' image is the front cover
Tim-Philipp Müller [Sat, 14 Jan 2017 15:16:53 +0000 (15:16 +0000)]
qtdemux: boldly assume that first 'covr' image is the front cover

7 years agoqtdemux: extract cover art images into GST_TAG_IMAGE not PREVIEW_IMAGE
Tim-Philipp Müller [Sat, 14 Jan 2017 15:09:07 +0000 (15:09 +0000)]
qtdemux: extract cover art images into GST_TAG_IMAGE not PREVIEW_IMAGE

These are usually much bigger than icon size and required by
iTunes to be certain fairly large sizes. In qtmux it is also
the IMAGE tags which we write out as 'covr' atoms.

7 years agoflacenc: also set PICTURE tag width and height if available
Tim-Philipp Müller [Sat, 14 Jan 2017 15:05:36 +0000 (15:05 +0000)]
flacenc: also set PICTURE tag width and height if available

7 years agoflacenc: fix encoder init error with some GST_TAG_PREVIEW_IMAGEs
Tim-Philipp Müller [Sat, 14 Jan 2017 14:58:52 +0000 (14:58 +0000)]
flacenc: fix encoder init error with some GST_TAG_PREVIEW_IMAGEs

The encoder fails to initialise when we try to set GST_TAG_PREVIEW_IMAGEs
sent to use by qtdemux from iTunes-generated m4a files. We should
not just blindly translate the PREVIEW tag to file icon image types,
but check if the specific conditions required are met (i.e. image
type 1 must be a 32x32 PNG icon, and what we're getting is 500x500).

https://bugzilla.gnome.org/show_bug.cgi?id=776962

7 years agomeson: bump version
Tim-Philipp Müller [Fri, 13 Jan 2017 12:39:00 +0000 (12:39 +0000)]
meson: bump version

7 years agosplitmuxsink: Add option for timecode-based split
Vivia Nikolaidou [Thu, 22 Dec 2016 15:40:40 +0000 (17:40 +0200)]
splitmuxsink: Add option for timecode-based split

If this option is given, it will calculate the next split point based on
timecode difference.

https://bugzilla.gnome.org/show_bug.cgi?id=774209

7 years agoqtmux: Don't reset request pad numbering across uses
Jan Schmidt [Thu, 12 Jan 2017 13:01:06 +0000 (00:01 +1100)]
qtmux: Don't reset request pad numbering across uses

When reset, don't restart request pad numberings, as
request pads can survive across state changes. Only
restart at 0 if all request pads are handed back first.

https://bugzilla.gnome.org/show_bug.cgi?id=777174

7 years agortxqueue: Expose basic statistics as properties.
Mathieu Duponchelle [Wed, 11 Jan 2017 17:52:28 +0000 (18:52 +0100)]
rtxqueue: Expose basic statistics as properties.

Statistics about the total number of retransmission requests
and the actual number of retransmitted packets can be helpful
at application-level.

https://bugzilla.gnome.org/show_bug.cgi?id=777182

7 years agoqtmux: simplify video/x-h264 caps handling
Aurélien Zanelli [Thu, 12 Jan 2017 16:45:35 +0000 (17:45 +0100)]
qtmux: simplify video/x-h264 caps handling

'stream-format' and 'alignment' are defined in pad template caps so
there is no need to check them again here. Also remove bitrate parsing from
caps as bitrate in caps doesn't make sense but from tags, which is
actually the case.

https://bugzilla.gnome.org/show_bug.cgi?id=777181

7 years agoqtmux: add basic HEVC/H.265 muxing support
Aurélien Zanelli [Thu, 8 Dec 2016 16:02:22 +0000 (17:02 +0100)]
qtmux: add basic HEVC/H.265 muxing support

https://bugzilla.gnome.org/show_bug.cgi?id=736752

7 years agoqtmux: Calculate clean aperture size
Georg Lippitsch [Wed, 11 Jan 2017 17:29:05 +0000 (18:29 +0100)]
qtmux: Calculate clean aperture size

Calculate clean aperture dimensions by first guessing
display aspect ratio based on pixel aspect ratio and
frame size.

https://bugzilla.gnome.org/show_bug.cgi?id=777100

7 years agoqtmux: Write tapt atom for MOV files if PAR not 1/1
Sebastian Dröge [Tue, 10 Jan 2017 16:19:55 +0000 (18:19 +0200)]
qtmux: Write tapt atom for MOV files if PAR not 1/1

Needed for QuickTime 7 to properly play files.

Also write the clap atom for MOV files always, not only when ProRes is
used as a video codec. It's mandatory for MOV.

https://bugzilla.gnome.org/show_bug.cgi?id=777100

7 years agoBack to development
Sebastian Dröge [Thu, 12 Jan 2017 14:32:45 +0000 (16:32 +0200)]
Back to development

7 years agoRelease 1.11.1
Sebastian Dröge [Thu, 12 Jan 2017 13:31:02 +0000 (15:31 +0200)]
Release 1.11.1

7 years agoUpdate .po files
Sebastian Dröge [Thu, 12 Jan 2017 12:38:55 +0000 (14:38 +0200)]
Update .po files

7 years agopo: Update translations
Sebastian Dröge [Thu, 12 Jan 2017 12:36:22 +0000 (14:36 +0200)]
po: Update translations

7 years agoqtdemux: free seqh after calling qtdemux_parse_svq3_stsd_data()
Andre McCurdy [Thu, 12 Jan 2017 01:53:32 +0000 (17:53 -0800)]
qtdemux: free seqh after calling qtdemux_parse_svq3_stsd_data()

The seqh buffer allocated in qtdemux_parse_svq3_stsd_data() needs to
be freed by the caller after use.

https://bugzilla.gnome.org/show_bug.cgi?id=777157

Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
7 years agoisomp4: Don't spam debug log with knonw/padding atoms
Edward Hervey [Tue, 10 Jan 2017 15:01:35 +0000 (16:01 +0100)]
isomp4: Don't spam debug log with knonw/padding atoms

Only output WARNING messages for atoms we don't know how to handle
instead of for padding/known atoms we don't need to do any processing
on

https://bugzilla.gnome.org/show_bug.cgi?id=777095

7 years agoFix indentation
Tim-Philipp Müller [Mon, 9 Jan 2017 19:05:10 +0000 (19:05 +0000)]
Fix indentation

7 years agotests: rtpjitterbuffer: fix compiler warning due to c99-ism
Tim-Philipp Müller [Mon, 9 Jan 2017 19:04:04 +0000 (19:04 +0000)]
tests: rtpjitterbuffer: fix compiler warning due to c99-ism

rtpjitterbuffer.c:592:3: error: ‘for’ loop initial declarations are only allowed in C99 mode

7 years agoautodetect: bring the element state down after success
Matthew Waters [Fri, 11 Nov 2016 03:31:03 +0000 (14:31 +1100)]
autodetect: bring the element state down after success

Otherwise some messages that are emitted by the element on NULL->READY
will not reach the application.

https://bugzilla.gnome.org/show_bug.cgi?id=764947

7 years agoqtmux: Write tfdt atom into fragmented files.
Jan Schmidt [Sat, 7 Jan 2017 14:13:32 +0000 (01:13 +1100)]
qtmux: Write tfdt atom into fragmented files.

The DASH spec requires that tfdt atoms be present, so
write one out. ISO/IEC 23009-1:2014 6.3.4.2

https://bugzilla.gnome.org/show_bug.cgi?id=708221

7 years agoqtdemux: Don't reset output timestamps when no tfdt
Jan Schmidt [Sat, 7 Jan 2017 12:55:42 +0000 (23:55 +1100)]
qtdemux: Don't reset output timestamps when no tfdt

If a fragmented stream doesn't have a tfdt, don't
reset the output timestamps at each fragment boundary
by erroneously using the default value of 0. Introduced
by commit 69fc48

https://bugzilla.gnome.org/show_bug.cgi?id=754230

7 years agomeson: Install presets files
Thibault Saunier [Fri, 16 Dec 2016 19:51:48 +0000 (16:51 -0300)]
meson: Install presets files

7 years agoavidemux: fix some caps leaks
Garima Gaur [Tue, 3 Jan 2017 04:42:30 +0000 (10:12 +0530)]
avidemux: fix some caps leaks

https://bugzilla.gnome.org//show_bug.cgi?id=776789

7 years agosplitmuxsink: Return a bin with a "location" property as a sink
Vivia Nikolaidou [Thu, 22 Dec 2016 15:34:08 +0000 (17:34 +0200)]
splitmuxsink: Return a bin with a "location" property as a sink

Splitmuxsink might be called with a custom bin as a sink. If it has a
"location" property, it can be used.

7 years agosplitmux: Rewrite buffer collection and scheduling
Jan Schmidt [Fri, 18 Nov 2016 11:42:18 +0000 (22:42 +1100)]
splitmux: Rewrite buffer collection and scheduling

Majorly change the way that splitmuxsink collects
incoming data and sends it to the output, so that it
makes all decisions about when / where to split files
on the input side.

Use separate queues for each stream, so they can be
grown individually and kept as small as possible.

This removes raciness I observed where sometimes
some data would end up put in a different output file
over multiple runs with the same input.

Also fixes hangs with input queues getting full
and causing muxing to stall out.

7 years agosplitmuxsink: Add format-location-full signal
Jan Schmidt [Thu, 17 Nov 2016 12:40:27 +0000 (23:40 +1100)]
splitmuxsink: Add format-location-full signal

Add a new signal for formatting the filename, which receives
a GstSample containing the first buffer from the reference
stream that will be muxed into that file.

Useful for creating filenames that are based on the
running time or other attributes of the buffer.

To make it work, opening of files and setting filenames is
now deferred until there is some data to write to it,
which also requires some changes to how async state changes
and gap events are handled.

7 years agoqtdemux: Always snap to the start of the keyframe
Jan Schmidt [Fri, 30 Dec 2016 14:54:01 +0000 (01:54 +1100)]
qtdemux: Always snap to the start of the keyframe

When performing a key-unit seek, always snap to the start ts
of the keyframe buffer we landed on so that the keyframe is
entirely within the resulting outgoing segment. That seems
the most sensible result, since the user requested snapping
to the keyframe position.

7 years agoqtdemux: Omit cslg_shift when snapping seeks
Jan Schmidt [Fri, 30 Dec 2016 14:48:04 +0000 (01:48 +1100)]
qtdemux: Omit cslg_shift when snapping seeks

Segments times and seek requests are stored and handled
in raw 'PTS' time, without the cslg_shift - which only applies
to outgoing samples. Omit the cslg_shift portion when
extracting PTS to compare for internal seek snaps.

If the cslg_shift is included, then keyframe+snap-before seeks
generate a segment start/stop time that already includes the
cslg_shift, and it's then added a 2nd time, causing the
first buffer(s) to have timestamps that are out of segment.

7 years agoqtmux: Remove bogus check in atom_stsc_add_new_entry()
Jan Schmidt [Fri, 30 Dec 2016 11:31:38 +0000 (22:31 +1100)]
qtmux: Remove bogus check in atom_stsc_add_new_entry()

Remove an old check from atom_stsc_add_new_entry() that
extends the last entry in the STSC if the samples per chunk
matches, as the new interleave merging logic requires that
the final entry by updateable. There's already code
below which simply merges the final entry into the previous
one when needed, so rely on that instead.

Fixes asserts like:

ERROR:atoms.c:2940:atom_stsc_update_entry: assertion failed:
(atom_array_index (&stsc->entries, len - 1).first_chunk == first_chunk)

7 years agoqtdemux: Fix key_time in gst_qtdemux_adjust_seek()
Seungha Yang [Sun, 24 Apr 2016 12:38:51 +0000 (21:38 +0900)]
qtdemux: Fix key_time in gst_qtdemux_adjust_seek()

time in segment should be PTS based (not DTS).

https://bugzilla.gnome.org/show_bug.cgi?id=765498

7 years agosplitmuxsrc: Pass seek flags when activating.
Jan Schmidt [Wed, 28 Dec 2016 11:49:27 +0000 (22:49 +1100)]
splitmuxsrc: Pass seek flags when activating.

Pass all seek flags when activating a part
based on a seek, so that SNAP flags are preserved.

7 years agosplitmux: Fix a small race in the splitmuxsrc
Jan Schmidt [Fri, 25 Nov 2016 14:13:19 +0000 (01:13 +1100)]
splitmux: Fix a small race in the splitmuxsrc

Make sure the state of the parser is set to
collecting streams before chaining up to the
parent change_state() method, to close a
small window that can cause playback to
never commence.

7 years agocheck: Remove dead code
Edward Hervey [Mon, 2 Jan 2017 14:06:33 +0000 (15:06 +0100)]
check: Remove dead code

7 years agomultifilesink: refactor max_files handling a bit
Tim-Philipp Müller [Sat, 31 Dec 2016 09:52:25 +0000 (09:52 +0000)]
multifilesink: refactor max_files handling a bit

Use GQueue instead of a GSList so we don't have to traverse
the whole list to append something every time. And it also
keeps track of the number of items in it for us.

Add a function to add filenames to the list of old files and
use it in more places, so that memory doesn't build up in
other modes either if no max_files limit is specified.

https://bugzilla.gnome.org/show_bug.cgi?id=766991

7 years agomultifilesink: don't leak memory when no max-files limit is set
Ursula Maplehurst [Sun, 29 May 2016 16:21:47 +0000 (17:21 +0100)]
multifilesink: don't leak memory when no max-files limit is set

Technically we weren't leaking the memory, just storing it internally
and never using it until the element is freed. But we'd still use more
and more memory over time, so this is not good over longer periods
of time. Only keep track of files if there's actually a limit set,
so that we will prune the list from time to time.

https://bugzilla.gnome.org/show_bug.cgi?id=766991

7 years agomatroskademux: adjust segment stop for KEY_UNIT negative rate seeking
Mark Nauwelaerts [Thu, 29 Dec 2016 11:39:20 +0000 (12:39 +0100)]
matroskademux: adjust segment stop for KEY_UNIT negative rate seeking

7 years agoqtdemux: implement pull mode SNAP flag seeking
Mark Nauwelaerts [Thu, 29 Dec 2016 11:25:35 +0000 (12:25 +0100)]
qtdemux: implement pull mode SNAP flag seeking

7 years agoavidemux: tweak KEY_UNIT SNAP seek handling
Mark Nauwelaerts [Thu, 29 Dec 2016 10:26:33 +0000 (11:26 +0100)]
avidemux: tweak KEY_UNIT SNAP seek handling

Previously, seeking to position y where y is (strictly) within a keyframe
would seek to that keyframe both with SNAP_BEFORE and SNAP_AFTER,
where the latter is now adjusted to really snap to the next keyframe.

7 years agoavidemux: correctly perform pull mode KEY_UNIT seeking
Mark Nauwelaerts [Wed, 28 Dec 2016 12:23:11 +0000 (13:23 +0100)]
avidemux: correctly perform pull mode KEY_UNIT seeking

Rather amazingly (and equally unnoticed), keyunit seeking resulted in segments
where start != time (which is bogus for simple avi timeline).  So, properly
adjust the segment (start) rather than fiddling with segment time (only).

7 years agoavidemux: restore considering of pull mode KEY_UNIT seeking
Mark Nauwelaerts [Wed, 28 Dec 2016 12:04:54 +0000 (13:04 +0100)]
avidemux: restore considering of pull mode KEY_UNIT seeking

... by using the original seek event's flags rather than the corresponding
segment flags, which do not have such counterpart flags (and
do no longer have them covertly sneaking in nowadays).

7 years agomatroskamux: only drop actual streamheader buffers with xiph codecs
Nicola Murino [Fri, 8 May 2015 10:44:01 +0000 (12:44 +0200)]
matroskamux: only drop actual streamheader buffers with xiph codecs

With Xiph codecs the stream header buffers are both in the caps and are
usually also at the beginning of each input stream, but it's perfectly
possible that the input stream does not have the stream header buffers
inline in the data. Matroskamux would drop the first N buffers assuming
they're stream headers, but this meant it would drop actual payload data
when the stream didn't contain the stream headers inline. Fix this by
only dropping leading buffers if they're flagged as stream headers. This
fixes issues with streams that are being tapped into after streaming
has started.

https://bugzilla.gnome.org/show_bug.cgi?id=749098

7 years agomatroskamux: adjust unit test to modified behaviour
Nicola Murino [Wed, 21 Dec 2016 16:43:58 +0000 (17:43 +0100)]
matroskamux: adjust unit test to modified behaviour

Now matroskamux mark all packets of audio-only streams as keyframes so
in test_block_group after pushing the test audio data 4 buffers are produced
and not more 2. The last buffer is the original data and must match with what
pushed. The remaining ones are matroskamux headers

https://bugzilla.gnome.org/show_bug.cgi?id=754696

7 years agomatroskamux: mark all packets of audio-only streams as keyframes
Nicola Murino [Sun, 29 May 2016 23:15:31 +0000 (01:15 +0200)]
matroskamux: mark all packets of audio-only streams as keyframes

This helps with streaming audio-only streams via multifdsink,
tcpserversink and such.

https://bugzilla.gnome.org/show_bug.cgi?id=754696

7 years agomatroskamux: add G722 audio support
Nicola Murino [Sat, 28 Mar 2015 17:15:36 +0000 (18:15 +0100)]
matroskamux: add G722 audio support

https://bugzilla.gnome.org/show_bug.cgi?id=746574

7 years agoupdsrc: Add to join multiple multicast interfaces
Wonchul Lee [Tue, 13 Dec 2016 02:11:07 +0000 (11:11 +0900)]
updsrc: Add to join multiple multicast interfaces

https://bugzilla.gnome.org/show_bug.cgi?id=776030

7 years agortpklvdepay: add the SPARSE flag to the outgoing stream-start event
Tim-Philipp Müller [Wed, 25 Mar 2015 13:51:30 +0000 (13:51 +0000)]
rtpklvdepay: add the SPARSE flag to the outgoing stream-start event

7 years agortpmanager: place content before Since-version API marker
Reynaldo H. Verdejo Pinochet [Wed, 14 Dec 2016 22:37:45 +0000 (14:37 -0800)]
rtpmanager: place content before Since-version API marker

Avoids confusing the parser

7 years agoshout2: fix 404 in package origin
Reynaldo H. Verdejo Pinochet [Wed, 14 Dec 2016 22:16:53 +0000 (14:16 -0800)]
shout2: fix 404 in package origin

7 years agoqtdemux: Check if we have enough data available when parsing edit lists
Sebastian Dröge [Wed, 14 Dec 2016 19:45:15 +0000 (21:45 +0200)]
qtdemux: Check if we have enough data available when parsing edit lists

Also consume the data entry by entry to get complicated indexing out of
the code.

https://bugzilla.gnome.org/show_bug.cgi?id=776107

7 years agov4l2object: Don't check size in a non-list value
Víctor Manuel Jáquez Leal [Wed, 14 Dec 2016 18:15:03 +0000 (19:15 +0100)]
v4l2object: Don't check size in a non-list value

After commit 1ea9735a I see these error while using the webcam
integrated in my laptop:

GStreamer-CRITICAL **: gst_value_list_get_size: assertion 'GST_VALUE_HOLDS_LIST (value)' failed

The issue is gst_v4l2src_value_simplify() was doing its job of
generating a single value, rather than the original list. That why,
when getting the list size, a critical warning was raised.

This patch takes advantage of the compiler optimizations to verify
first if the list was simplified, thus use it directly, otherwise,
if it is a list, verify its size.

https://bugzilla.gnome.org/show_bug.cgi?id=776106

7 years agotests/jitterbuffer: Major refactoring and cleanups
Havard Graff [Wed, 14 Dec 2016 09:39:12 +0000 (10:39 +0100)]
tests/jitterbuffer: Major refactoring and cleanups

* Changed PCMU->TEST for common macros
* Changed verify-functions (lost & rtx) into macros.
* Remove option to add marker-bit for test-buffers (not used anywhere)
* Add new push_test_buffer function that makes sure there are correlation
  between dts and the time on the clock. (classic test-mistake)
* Established a generic starting-point for tests with the
  construct_deterministic_initial_state function and use it where
  applicable, which removes lots of "boilerplate" everywhere.
* Add basic lost-event test
* Remove as much "magic constants" as possible.
* Remove 3 tests that no longer are testing anything that others don't,
  and was completely unmaintainable.
* Remove unnecessary use of the testclock
* Verify each test is testing what it actually says it does (and modify
  where it doesn't)

In general, make the tests much smaller, better, more maintainable and
readable.

https://bugzilla.gnome.org/show_bug.cgi?id=774409

7 years agoRemove generated .spec file
Tim-Philipp Müller [Wed, 14 Dec 2016 09:54:11 +0000 (09:54 +0000)]
Remove generated .spec file

Likely extremely bitrotten, and we should not ship this anyway.

7 years agoqtdemux: Check that the XiTh size is big enough
Sebastian Dröge [Wed, 14 Dec 2016 08:15:10 +0000 (10:15 +0200)]
qtdemux: Check that the XiTh size is big enough

https://bugzilla.gnome.org/show_bug.cgi?id=775794

7 years agoqtdemux: Check node length of video sample description
Heekyoung Seo [Fri, 9 Dec 2016 11:27:53 +0000 (20:27 +0900)]
qtdemux: Check node length of video sample description

Add check for node length of video sample description and its fields and
for the XiTh atom.

Also unify the code a bit.

https://bugzilla.gnome.org/show_bug.cgi?id=775794

7 years agoqtdemux: Enable xvid/mp2 codec support
Heekyoung Seo [Thu, 8 Dec 2016 09:50:52 +0000 (18:50 +0900)]
qtdemux: Enable xvid/mp2 codec support

Add support for xvid video and mp2 audio, add m2v1 fourcc.

https://bugzilla.gnome.org/show_bug.cgi?id=775794

7 years agogst: Don't declare variables inside the for loop header
Sebastian Dröge [Tue, 13 Dec 2016 20:32:46 +0000 (22:32 +0200)]
gst: Don't declare variables inside the for loop header

This is a C99 feature.

7 years agoscaletempo: Ensure to reinit buffers whenever they were not allocated yet
Sebastian Dröge [Sun, 11 Dec 2016 11:27:27 +0000 (13:27 +0200)]
scaletempo: Ensure to reinit buffers whenever they were not allocated yet

That is, whenever we go through start/stop we have to ensure that on the
next opportunity the buffers are reallocated again. Otherwise the
buffers might be NULL because the element was reused with the same
configuration as before (i.e. set_caps() wouldn't have reinited the
buffers).

https://bugzilla.gnome.org/show_bug.cgi?id=775898

7 years agodocs: design: remove, moved to gst-docs
Tim-Philipp Müller [Sat, 10 Dec 2016 12:52:18 +0000 (12:52 +0000)]
docs: design: remove, moved to gst-docs

7 years agomeson: Support building without Gst debug
Thibault Saunier [Fri, 9 Dec 2016 20:17:35 +0000 (17:17 -0300)]
meson: Support building without Gst debug

7 years agoflxdec: Only send SEGMENT events after CAPS
Sebastian Dröge [Fri, 9 Dec 2016 15:55:39 +0000 (17:55 +0200)]
flxdec: Only send SEGMENT events after CAPS

I.e., don't just forward the event but delay it if we don't have caps on
the srcpad yet.

7 years agoflxdec: Unref and unmap buffers in all code paths as needed
Sebastian Dröge [Fri, 9 Dec 2016 15:49:40 +0000 (17:49 +0200)]
flxdec: Unref and unmap buffers in all code paths as needed

https://bugzilla.gnome.org/show_bug.cgi?id=775888

7 years agov4l2object: Don't set empty interlace-mode list
Arun Raghavan [Tue, 6 Dec 2016 12:12:31 +0000 (17:42 +0530)]
v4l2object: Don't set empty interlace-mode list

If for some reason we fail to probe formats (all try_fmt calls fail, for
example), this is not a critical error, but we end up with an empty list
of interlace modes. This causes all subsequent negotiation to fail.

This patch fixes interlace-mode setting to be skipped if we failed to
detect any.

https://bugzilla.gnome.org/show_bug.cgi?id=775702

7 years agomonoscope: Unref allocation query after finished with it
Garima Gaur [Wed, 7 Dec 2016 11:52:22 +0000 (17:22 +0530)]
monoscope: Unref allocation query after finished with it

https://bugzilla.gnome.org/show_bug.cgi?id=775752

7 years agoflxdec: Allocate 0-initialized memory for the decoded frame
Sebastian Dröge [Tue, 6 Dec 2016 05:48:47 +0000 (07:48 +0200)]
flxdec: Allocate 0-initialized memory for the decoded frame

Otherwise we might leak arbitrary information from the uninitialized
memory if not every pixel is written.

https://scarybeastsecurity.blogspot.gr/2016/12/1days-0days-pocs-more-gstreamer-flic.html