3 GSTREAMER 1.16 RELEASE NOTES
6 GStreamer 1.16 has not been released yet. It is scheduled for release in
9 1.15.x is the unstable development version that is being developed in
10 the git master branch and which will eventually result in 1.16.
12 1.16 will be backwards-compatible to the stable 1.14, 1.12, 1.10, 1.8,
13 1.6, 1.4, 1.2 and 1.0 release series.
15 See https://gstreamer.freedesktop.org/releases/1.16/ for the latest
16 version of this document.
18 _Last updated: Monday 14 January 2019, 13:00 UTC (log)_
23 The GStreamer team is proud to announce a new major feature release in
24 the stable 1.x API series of your favourite cross-platform multimedia
27 As always, this release is again packed with new features, bug fixes and
33 - GStreamer WebRTC stack gained support for data channels for
34 peer-to-peer communication based on SCTP, BUNDLE support, as well as
35 support for multiple TURN servers.
37 - AV1 video codec support for Matroska and QuickTime/MP4 containers
38 and more configuration options and supported input formats for the
41 - Support for Closed Captions and other Ancillary Data in video
43 - Spport for planar (non-interleaved) raw audio
45 - GstVideoAggregator, compositor and OpenGL mixer elements are now in
48 - New alternate fields interlace mode where each buffer carries a
51 - WebM and Matroska ContentEncryption support in the Matroska demuxer
53 - new WebKit WPE-based web browser source element
55 - Video4Linux: HEVC encoding and decoding, JPEG encoding, and improved
58 - Hardware-accelerated Nvidia video decoder gained support for VP8/VP9
59 decoding, whilst the encoder gained support for H.265/HEVC encoding.
61 - Many improvements to the Intel Media SDK based hardware-accelerated
62 video decoder and encoder plugin (msdk): dmabuf import/export for
63 zero-copy integration with other components; VP9 decoding; 10-bit
64 HEVC encoding; video post-processing (vpp) support including
65 deinterlacing; and the video decoder now handles dynamic resolution
68 - The ASS/SSA subtitle overlay renderer can now handle multiple
69 subtitles that overlap in time and will show them on screen
72 - The Meson build is now feature-complete (*) and it is now the
73 recommended build system on all platforms. The Autotools build is
74 scheduled to be removed in the next cycle.
76 - The GStreamer Rust bindings and Rust plugins module are now
77 officially part of upstream GStreamer.
79 - Many performance improvements
82 Major new features and changes
86 - GstAggregator has a new "min-upstream-latency" property that forces
87 a minimum aggregate latency for the input branches of an aggregator.
88 This is useful for dynamic pipelines where branches with a higher
89 latency might be added later after the pipeline is already up and
90 running and where a change in the latency would be disruptive. This
91 only applies to the case where at least one of the input branches is
92 live though, it won’t force the aggregator into live mode in the
93 absence of any live inputs.
95 - GstBaseSink gained a "processing-deadline" property and
96 setter/getter API to configure a processing deadline for live
97 pipelines. The processing deadline is the acceptable amount of time
98 to process the media in a live pipeline before it reaches the sink.
99 This is on top of the systemic latency that is normally reported by
100 the latency query. This defaults to 20ms and should make pipelines
101 such as “v4lsrc ! xvimagesink” not claim that all frames are late in
102 the QoS events. Ideally, this should replace max_lateness for most
105 - RTCP Extended Reports (XR) parsing according to RFC 3611:
106 Loss/Duplicate RLE, Packet Receipt Times, Receiver Reference Time,
107 Delay since the last Receiver (DLRR), Statistics Summary, and VoIP
110 - a new mode for interlaced video was added where each buffer carries
111 a single field of interlaced video, with buffer flags indicating
112 whether the field is the top field or bottom field. Top and bottom
113 fields are expected to alternate in this mode. Caps for this
114 interlace mode must also carry a format:Interlaced caps feature to
115 ensure backwards compatibility.
117 - The video library has gained support for three new raw pixel
120 - Y410: packed 4:4:4 YUV, 10 bits per channel
121 - Y210: packed 4:2:2 YUV, 10 bits per channel
122 - NV12_10LE40: fully-packed 10-bit variant of NV12_10LE32,
123 i.e. without the padding bits
125 - GstRTPSourceMeta is a new meta that can be used to transport
126 information about the origin of depayloaded or decoded RTP buffers,
127 e.g. when mixing audio from multiple sources into a single stream. A
128 new "source-info" property on the RTP depayloader base class
129 determines whether depayloaders should put this meta on outgoing
130 buffers. Similarly, the same property on RTP payloaders determines
131 whether they should use the information from this meta to construct
132 the CSRCs list on outgoing RTP buffers.
134 - gst_sdp_message_from_text() is a convenience constructor to parse
135 SDPs from a string which is particularly useful for language
138 Support for Planar (Non-Interleaved) Raw Audio
140 Raw audio samples are usually passed around in interleaved form in
141 GStreamer, which means that if there are multiple audio channels the
142 samples for each channel are interleaved in memory, e.g.
143 |LEFT|RIGHT|LEFT|RIGHT|LEFT|RIGHT| for stereo audio. A non-interleaved
144 or planar arrangement in memory would look like
145 |LEFT|LEFT|LEFT|RIGHT|RIGHT|RIGHT| instead, possibly with
146 |LEFT|LEFT|LEFT| and |RIGHT|RIGHT|RIGHT| residing in separate memory
147 chunks or separated by some padding.
149 GStreamer has always had signalling for non-interleaved audio, but it
150 was never actually properly implemented in any elements. audioconvert
151 would advertise support for it, but wasn’t actually able to handle it.
153 With this release we now have full support for non-interleaved audio as
154 well, which means more efficient integration with external APIs that
155 handle audio this way, but also more efficient processing of certain
156 operations like interleaving multiple 1-channel streams into a
157 multi-channel stream which can be done without memory copies now.
159 New API to support this has been added to the GStreamer Audio support
160 library: There is now a new GstAudioMeta which describes how data is
161 laid out inside the buffer, and buffers with non-interleaved audio must
162 always carry this meta. To access the non-interleaved audio samples you
163 must map such buffers with gst_audio_buffer_map() which works much like
164 gst_buffer_map() or gst_video_frame_map() in that it will populate a
165 little GstAudioBuffer helper structure passed to it with the number of
166 samples, the number of planes and pointers to the start of each plane in
167 memory. This function can also be used to map interleaved audio buffers
168 in which case there will be only one plane of interleaved samples.
170 Of course support for this has also been implemented in the various
171 audio helper and conversion APIs, base classes, and in elements such as
172 audioconvert, audioresample, audiotestsrc, audiorate.
174 Support for Closed Captions and Other Ancillary Data in Video
176 The video support library has gained support for detecting and
177 extracting Ancillary Data from videos as per the SMPTE S291M
178 specification, including:
180 - a VBI (Video Blanking Interval) parser that can detect and extract
181 Ancillary Data from Vertical Blanking Interval lines of component
182 signals. This is currently supported for videos in v210 and UYVY
185 - a new GstMeta for closed captions: GstVideoCaptionMeta. This
186 supports the two types of closed captions, CEA-608 and CEA-708,
187 along with the four different ways they can be transported (other
188 systems are a superset of those).
190 - a VBI (Video Blanking Interval) encoder for writing ancillary data
191 to the Vertical Blanking Interval lines of component signals.
193 The new closedcaption plugin in gst-plugins-bad then makes use of all
194 this new infrastructure and provides the following elements:
196 - cccombiner: a closed caption combiner that takes a closed captions
197 stream and another stream and adds the closed captions as
198 GstVideoCaptionMeta to the buffers of the other stream.
200 - ccextractor: a closed caption extractor which will take
201 GstVideoCaptionMeta from input buffers and output them as a separate
202 closed captions stream.
204 - ccconverter: a closed caption converter that can convert between
207 - line21decoder: extract line21 closed captions from SD video streams
209 - cc708overlay: decodes CEA 608/708 captions and overlays them on
212 Additionally, the following elements have also gained Closed Caption
215 - qtdemux and qtmux support CEA 608/708 Closed Caption tracks
217 - mpegvideoparse extracts Closed Captions from MPEG-2 video streams
219 - decklinkvideosink can output closed captions and decklinkvideosrc
220 can extract closed captions
222 - playbin and playbin3 learned how to autoplug CEA 608/708 CC overlay
225 The rsclosedcaption plugin in the Rust plugins collection includes a
226 MacCaption (MCC) file parser and encoder.
230 - overlaycomposition: New element that allows applications to draw
231 GstVideoOverlayCompositions on a stream. The element will emit the
232 "draw" signal for each video buffer, and the application then
233 generates an overlay for that frame (or not). This is much more
234 performant than e.g. cairooverlay for many use cases, e.g. because
235 pixel format conversions can be avoided or the blitting of the
236 overlay can be delegated to downstream elements (such as
237 gloverlaycompositor). It’s particularly useful for cases where only
238 a small section of the video frame should be drawn on.
240 - gloverlaycompositor: New OpenGL-based compositor element that
241 flattens any overlays from GstVideoOverlayCompositionMetas into the
244 - glalpha: New element that adds an alpha channel to a video stream.
245 The values of the alpha channel can either be set to a constant or
246 can be dynamically calculated via chroma keying. It is similar to
247 the existing alpha element but based on OpenGL. Calculations are
248 done in floating point so results may not be identical to the output
249 of the existing alpha element.
251 - rtpfunnel funnels together rtp-streams into a single session. Use
252 cases include multiplexing and bundle. webrtcbin uses it to
253 implement BUNDLE support.
255 - testsrcbin is a source element that provides an audio and/or video
256 stream and also announces them using the recently-introduced
257 GstStream API. This is useful for testing elements such as playbin3
258 or uridecodebin3 etc.
260 - New closed caption elements: cccombiner, ccextractor, ccconverter,
261 line21decoder and cc708overlay (see above)
263 - wpesrc: new source element acting as a Web Browser based on WebKit
266 - Two new OpenCV-based elements: cameracalibrate and cameraundistort
267 who can communicate to figure out distortion correction parameters
268 for a camera and correct for the distortion.
270 - new sctp plugin based on usrsctp with sctpenc and sctpdec elements
272 New element features and additions
274 - playbin3, playbin and playsink have gained a new "text-offset"
275 property to adjust the positioning of the selected subtitle stream
276 vis-a-vis the audio and video streams. This uses subtitleoverlay’s
277 new "subtitle-ts-offset" property. GstPlayer has gained matching API
278 for this, namely gst_player_get_text_video_offset().
280 - playbin3 buffering improvements: in network playback scenarios there
281 may be multiple inputs to decodebin3, and buffering will be done
282 before decodebin3 using queue2 or downloadbuffer elements inside
283 urisourcebin. Since this is before any parsers or demuxers there may
284 not be any bitrate information available for the various streams, so
285 it was difficult to configure the buffering there smartly within
286 global constraints. This was improved now: The queue2 elements
287 inside urisourcebin will now use the new bitrate query to figure out
288 a bitrate estimate for the stream if no bitrate was provided by
289 upstream, and urisourcebin will use the bitrates of the individual
290 queues to distribute the globally-set "buffer-size" budget in bytes
291 to the various queues. urisourcebin also gained "low-watermark" and
292 "high-watermark" properties which will be proxied to the internal
293 queues, as well as a read-only "statistics" property which allows
294 querying of the minimum/maximum/average byte and time levels of the
295 queues inside the urisourcebin in question.
297 - splitmuxsink has gained a couple of new features:
299 - new "async-finalize" mode: This mode is useful for muxers or
300 outputs that can take a long time to finalize a file. Instead of
301 blocking the whole upstream pipeline while the muxer is doing
302 its stuff, we can unlink it and spawn a new muxer + sink
303 combination to continue running normally. This requires us to
304 receive the muxer and sink (if needed) as factories via the new
305 "muxer-factory" and "sink-factory" properties, optionally
306 accompanied by their respective properties structures (set via
307 the new "muxer-properties" and "sink-properties" properties).
308 There are also new "muxer-added" and "sink-added" signals in
309 case custom code has to be called for them to configure them.
311 - "split-at-running-time" action signal: When called by the user,
312 this action signal ends the current file (and starts a new one)
313 as soon as the given running time is reached. If called multiple
314 times, running times are queued up and processed in the order
317 - "split-after" action signal to finish outputting the current GOP
318 to the current file and then start a new file as soon as the GOP
319 is finished and a new GOP is opened (unlike the existing
320 "split-now" which immediately finishes the current file and
321 writes the current GOP into the next newly-started file).
323 - "reset-muxer" property: when unset, the muxer is reset using
324 flush events instead of setting its state to NULL and back. This
325 means the muxer can keep state across resets, e.g. mpegtsmux
326 will keep the continuity counter continuous across segments as
327 required by hlssink2.
329 - qtdemux gained PIFF track encryption box support in addition to the
330 already-existing PIFF sample encryption support, and also allows
331 applications to select which encryption system to use via a
332 "drm-preferred-decryption-system-id" context in case there are
335 - qtmux: the "start-gap-threshold" property determines now whether an
336 edit list will be created to account for small gaps or offsets at
337 the beginning of a stream in case the start timestamps of tracks
338 don’t line up perfectly. Previously the threshold was hard-coded to
339 1% of the (video) frame duration, now it is 0 by default (so edit
340 list will be created even for small differences), but fully
343 - rtpjitterbuffer has improved end-of-stream handling
345 - rtpmp4vpay will be prefered over rtpmp4gpay for MPEG-4 video in
346 autoplugging scenarios now
348 - rtspsrc now allows applications to send RTSP SET_PARAMETER and
349 GET_PARAMETER requests using action signals.
351 - rtspsrc also has a small (100ms) configurable teardown delay by
352 default to try and make sure an RTSP TEARDOWN request gets sent out
353 when the source element shuts down. This will block the downward
354 PAUSED to READY state change for a short time, but can be unset
355 where it’s a problem. Some servers only allow a limited number of
356 concurren clients, so if no proper TEARDOWN is sent clients may have
357 problems connecting to the server for a while.
359 - souphttpsrc behaves better with low bitrate streams now. Before it
360 would increase the read block size too quickly which could lead to
361 it not reading any data from the socket for a very long time with
362 low bitrate streams that are output live downstream. This could lead
363 to servers kicking off the client.
365 - filesink: do internal buffering to avoid performance regression with
366 small writes since we bypass libc buffering by using writev()
368 - identity: add "eos-after" property and fix "error-after" property
369 when the element is reused
371 - input-selector: lets context queries pass through, so that
372 e.g. upstream OpenGL elements can use contexts and displays
373 advertised by downstream elements
375 - queue2: avoid ping-pong between 0% and 100% buffering messages if
376 upstream is pushing buffers larger than one of its limits, plus
377 performance optimisations
379 - opusdec: new "phase-inversion" property to control phase inversion.
380 When enabled, this will slightly increase stereo quality, but
381 produces a stream that when downmixed to mono will suffer audio
384 - The x265enc HEVC encoder also exposes a "key-int-max" property to
385 configure the maximum allowed GOP size now.
387 - decklinkvideosink has seen stability improvements for long-running
388 pipelines (potential crash due to overflow of leaked clock refcount)
389 and clock-slaving improvements when performing flushing seeks
390 (causing stalls in the output timeline), pausing and/or buffering.
392 - srtpdec, srtpenc: add support for MKIs which allow multiple keys to
393 be used with a single SRTP stream
395 - The srt Secure Reliable Transport plugin has integrated server and
396 client elements srt{client,server}{src,sink} into one (srtsrc and
397 srtsink), since SRT connection mode can be changed by uri
400 - h264parse and h265parse will handle SEI recovery point messages and
401 mark recovery points as keyframes as well (in addition to IDR
404 - webrtcbin: "add-turn-server" action signal to pass multiple ICE
405 relays (TURN servers).
407 - The removesilence element has received various new features and
408 properties, such as a
409 "threshold"1 property, detecting silence only after minimum silence time/buffers, a“silent”property to control bus message notifications as well as a“squash”`
412 - AOMedia AV1 decoder gained support for 10/12bit decoding whilst the
413 AV1 encoder supports more image formats and subsamplings now and
414 acquired support for rate control and profile related configuration.
416 - The Fraunhofer fdkaac plugin can now be built against the 2.0.0
417 version API and has improved multichannel support
419 - kmssink now supports unpadded 24-bit RGB and can configure mode
420 setting from video info, which enables display of multi-planar
421 formats such as I420 or NV12 with modesetting. It has also gained a
422 number of new properties: The "restore-crtc" property does what it
423 says on the tin and is enabled by default. "plane-properties" and
424 "connector-properties" can be used to pass custom properties to the
427 - waylandsink has a "fullscreen" property now.
429 Plugin and library moves
431 - The stereo element was moved from -bad into the existing audiofx
432 plugin in -good. If you get duplicate type registration warnings
433 when upgrading, check that you don’t have a stale gststereo plugin
434 lying about somewhere.
436 GstVideoAggregator, compositor, and OpenGL mixer elements moved from -bad to -base
438 GstVideoAggregator is a new base class for raw video mixers and muxers
439 and is based on [GstAggregator][aggregator]. It provides defined-latency
440 mixing of raw video inputs and ensures that the pipeline won’t stall
441 even if one of the input streams stops producing data.
443 As part of the move to stabilise the API there were some last-minute API
444 changes and clean-ups, but those should mostly affect internal elements.
445 Most notably, the "ignore-eos" pad property was renamed to
446 "repeat-after-eos" and the conversion code was moved to a
447 GstVideoAggregatorConvertPad subclass to avoid code duplication, make
448 things less awkward for subclasses like the OpenGL-based video mixer,
449 and make the API more consistent with the audio aggregator API.
451 It is used by the compositor element, which is a replacement for
452 ‘videomixer’ which did not handle live inputs very well. compositor
453 should behave much better in that respect and generally behave as one
454 would expected in most scenarios.
456 The compositor element has gained support for per-pad blending mode
457 operators (SOURCE, OVER, ADD) which determines what operator to use for
458 blending this pad over the previous ones. This can be used to implement
461 A number of OpenGL-based video mixer elements (glvideomixer, glmixerbin,
462 glvideomixerelement, glstereomix, glmosaic) which are built on top of
463 GstVideoAggregator have also been moved from -bad to -base now. These
464 elements have been merged into the existing OpenGL plugin, so if you get
465 duplicate type registration warnings when upgrading, check that you
466 don’t have a stale gstopenglmixers plugin lying about somewhere.
470 The following plugins have been removed from gst-plugins-bad:
472 - The experimental daala plugin has been removed, since it’s not so
473 useful now that all effort is focused on AV1 instead, and it had to
474 be enabled explicitly with --enable-experimental anyway.
476 - The spc plugin has been removed. It has been replaced by the gme
479 - The acmmp3dec and acmenc plugins for Windows have been removed. ACM
480 is an ancient legacy API and there was no point in keeping them
481 around for a licensed mp3 decoder now that mp3 patents have expired
482 and we have a decoder in -good. We also didn’t ship these in our
483 cerbero-built Windows packages, so it’s unlikely that they’ll be
487 Miscellaneous API additions
489 - GstBitwriter: new generic bit writer API to complement the existing
492 - gst_buffer_new_wrapped_bytes() creates a wrap buffer from a GBytes
494 - gst_caps_set_features_simple() sets a caps feature on all the
495 structures of a GstCaps
497 - New GST_QUERY_BITRATE query: This allows determining from downstream
498 what the expected bitrate of a stream may be which is useful in
499 queue2 for setting time based limits when upstream does not provide
500 timing information. tsdemux, qtdemux and matroskademux have basic
501 support for this query on their sink pads.
503 - elements: there is a new “Hardware” class specifier. Elements
504 interacting with hardware devices should specify this classifier in
505 their element factory class metadata. This is useful to advertise as
506 one might need to put such elements into READY state to test if the
507 hardware is present in the system for example.
509 - protection: Add a new definition for unspecified system protection
511 - take functions for various mini objects that didn’t have them yet:
512 gst_query_take(), gst_message_take(), gst_tag_list_take(),
513 gst_buffer_list_take(). Unlike the various _replace() functions
514 _take() does not increase the reference count but takes ownership of
515 the mini object passed.
517 - clear functions for various mini object types and GstObject which
518 unrefs the object or mini object (if non-NULL) and sets the variable
519 pointed to to NULL: gst_clear_structure(), gst_clear_tag_list(),
520 gst_clear_query(), gst_clear_message(), gst_clear_event(),
521 gst_clear_caps(), gst_clear_buffer_list(), gst_clear_buffer(),
522 gst_clear_mini_object(), gst_clear_object()
524 - miniobject: new API gst_mini_object_add_parent() and
525 gst_mini_object_remove_parent()to set parent pointers on mini objects to ensure correct writability: Every container of miniobjects now needs to store itself as parent in the child object, and remove itself again later. A mini object is then only writable if there is at most one parent, that parent is writable itself, and the reference count of the mini object is 1.GstBuffer(for memories),GstBufferList(for buffers),GstSample(for caps, buffer, bufferlist), andGstVideoOverlayComposition`
526 were updated accordingly. Without this it was possible to have
527 e.g. a buffer list with a refcount of 2 used in two places at once
528 that both modify the same buffer with refcount 1 at the same time
529 wrongly thinking it is writable even though it’s really not.
531 - poll: add API to watch for POLLPRI and stop treating POLLPRI as a
532 read. This is useful to wait for video4linux events which are
533 signalled via POLLPRI.
535 - sample: new API to update the contents of a GstSample and make it
536 writable: gst_sample_set_buffer(), gst_sample_set_caps(),
537 gst_sample_set_segment(), gst_sample_set_info(), plus
538 gst_sample_is_writable() and gst_sample_make_writable(). This makes
539 it possible to reuse a sample object and avoid unnecessary memory
540 allocations, for example in appsink.
542 - ClockIDs now keep a weak reference to underlying clock to avoid
543 crashes in basesink in corner cases where a clock goes away while
544 the ClockID is still in use, plus some new API
545 (gst_clock_id_get_clock(), gst_clock_id_uses_clock()) to check the
546 clock a ClockID is linked to.
548 - The GstCheck unit test library gained a
549 fail_unless_equals_clocktime() convenience macro as well as some new
550 GstHarness API for for proposing meta APIs from the allocation
551 query: gst_harness_add_propose_allocation_meta(). ASSERT_CRITICAL()
552 checks in unit tests are now skipped if GStreamer was compiled with
553 GST_DISABLE_GLIB_CHECKS.
555 - gst_audio_buffer_truncate() convenience function to truncate a raw
559 Miscellaneous performance and memory optimisations
561 As always there have been many performance and memory usage improvements
562 across all components and modules. Some of them (such as dmabuf
563 import/export) have already been mentioned elsewhere so won’t be
566 The following list is only a small snapshot of some of the more
567 interesting optimisations that haven’t been mentioned in other contexts
570 - The GstVideoEncoder and GstVideoDecoder base classes now release the
571 STREAM_LOCK when pushing out buffers, which means (multi-threaded)
572 encoders and decoders can now receive and continue to process input
573 buffers whilst waiting for downstream elements in the pipeline to
574 process the buffer that was pushed out. This increases throughput
575 and reduces processing latency, also and especially for
576 hardware-accelerated encoder/decoder elements.
578 - GstQueueArray has seen a few API additions
579 (gst_queue_array_peek_nth(), gst_queue_array_set_clear_func(),
580 gst_queue_array_clear()) so that it can be used in other places like
581 GstAdapter instead of a GList, which reduces allocations and
582 improves performance.
584 - appsink now reuses the sample object in pull_sample() if possible
586 - rtpsession only starts the RTCP thread when it’s actually needed now
588 - udpsrc uses a buffer pool now and the GstUdpSrc object structure was
589 optimised for better cache performance
593 - API was added to fine-tune the synchronisation offset between
597 Miscellaneous changes
599 - As a result of moving to different FFmpeg APIs, encoder and decoder
600 elements exposed by the GStreamer FFmpeg wrapper plugin (gst-libav)
601 may have seen possibly incompatible changes to property names and/or
602 types, and not all properties exposed might be functional. We are
603 still reviewing the new properties and aim to minimise breaking
604 changes at least for the most commonly-used properties, so please
605 report any issues you run into!
609 - The OpenGL mixer elements have been moved from -bad to
610 gst-plugins-base (see above)
612 - The Mesa GBM backend now supports headless mode
614 - gloverlaycompositor: New OpenGL-based compositor element that
615 flattens any overlays from GstVideoOverlayCompositionMetas into the
618 - glalpha: New element that adds an alpha channel to a video stream.
619 The values of the alpha channel can either be set to a constant or
620 can be dynamically calculated via chroma keying. It is similar to
621 the existing alpha element but based on OpenGL. Calculations are
622 done in floating point so results may not be identical to the output
623 of the existing alpha element.
625 - glupload: Implement direct dmabuf uploader, the idea being that some
626 GPUs (like the Vivante series) can actually perform the YUV->RGB
627 conversion internally, so no custom conversion shaders are needed.
628 To make use of this feature, we need an additional uploader that can
629 import DMABUF FDs and also directly pass the pixel format, relying
630 on the GPU to do the conversion.
633 Tracing framework and debugging improvements
635 - There is now a GDB PRETTY PRINTER FOR VARIOUS GSTREAMER TYPES: For
636 GstObject pointers the type and name is added, e.g.
637 0x5555557e4110 [GstDecodeBin|decodebin0]. For GstMiniObject pointers
638 the object type is added, e.g. 0x7fffe001fc50 [GstBuffer]. For
639 GstClockTime and GstClockTimeDiff the time is also printed in human
640 readable form, e.g. 150116219955 [+0:02:30.116219955].
642 - GDB EXTENSION WITH TWO CUSTOM GDB COMMANDS gst-dot AND gst-print:
644 - gst-dot creates dot files that a very close to what
645 GST_DEBUG_BIN_TO_DOT_FILE() produces, but object properties and
646 buffer contents such as codec-data in caps are not available.
648 - gst-print produces high-level information about a GStreamer
649 object. This is currently limited to pads for GstElements and
650 events for the pads. The output may look like this:
652 (gdb) gst-print pad.object.parent
653 GstMatroskaDemux (matroskademux0) {
654 SinkPad (sink, pull) {
656 SrcPad (video_0, push) {
659 stream-id: 0463ccb080d00b8689bf569a435c4ff84f9ff753545318ae2328ea0763fd0bec/001:1274058367
663 pixel-aspect-ratio: 1/1
665 streamheader: < 0x5555557c7d30 [GstBuffer], 0x5555557c7e40 [GstBuffer], 0x7fffe00141d0 [GstBuffer] >
669 container-format: Matroska
671 SrcPad (audio_0, push) {
674 stream-id: 0463ccb080d00b8689bf569a435c4ff84f9ff753545318ae2328ea0763fd0bec/002:1551204875
679 codec_data: 0x7fffe0014500 [GstBuffer]
688 container-format: Matroska
690 audio-codec: MPEG-4 AAC audio
695 - gst_structure_to_string() now serialises the actual value of
696 pointers when serialising GstStructures instead of claiming they’re
697 NULL. This makes debug logging in various places less confusing,
698 because it’s clear now that structure fields actually hold valid
699 objects. Such object pointer values will never be deserialised
705 - gst-inspect-1.0 has coloured output now and will automatically use a
706 pager if the output does not fit on a page. This only works in a
707 unix environment and if the output is not piped. If you don’t like
708 the colours you can disable them by setting the
709 GST_INSPECT_NO_COLORS=1 environment variable or passing the
710 --no-colors command line option.
713 GStreamer RTSP server
715 - Improved backlog handling when using TCP interleaved for data
716 transport. Before there was a fixed maximum size for backlog
717 messages, which was prone to deadlocks and made it difficult to
718 control memory usage with the watch backlog. The RTSP server now
719 limits queued TCP data messages to one per stream, moving queuing of
720 the data into the pipeline and leaving the RTSP connection
721 responsive to RTSP messages in both directions, preventing all those
724 - Initial ULP Forward Error Correction support in rtspclientsink and
725 for RECORD mode in the server.
727 - API to explicitly enable retransmission requests (RTX)
729 - Lots of multicast-related fixes
731 - rtsp-auth: Add support for parsing .htdigest files
736 - this section will be filled in in due course
739 GStreamer Editing Services and NLE
741 - this section will be filled in in due course
746 - this section will be filled in in due course
749 GStreamer Python Bindings
751 - add binding for gst_pad_set_caps()
753 - pygobject dependency requirement was bumped to >= 3.8
755 - new audiotestsrc, audioplot, and mixer plugin examples, and a
756 dynamic pipeline example
759 GStreamer C# Bindings
761 - bindings for the GstWebRTC library
764 GStreamer Rust Bindings
766 The GStreamer Rust bindings are now officially part of the GStreamer
767 project and are also maintained in the GStreamer GitLab.
769 The releases will generally not be synchronized with the releases of
770 other GStreamer parts due to dependencies on other projects.
772 Also unlike the other GStreamer libraries, the bindings will not commit
773 to full API stability but instead will follow the approach that is
774 generally taken by Rust projects, e.g.:
776 1) 0.12.X will be completely API compatible with all other 0.12.Y
778 2) 0.12.X+1 will contain bugfixes and compatible new feature additions.
779 3) 0.13.0 will _not_ be backwards compatible with 0.12.X but projects
780 will be able to stay at 0.12.X without any problems as long as they
781 don’t need newer features.
783 The current stable release is 0.12.2 and the next release series will be
784 0.13, probably around March 2019.
786 At this point the bindings cover most of GStreamer core (except for most
787 notably GstAllocator and GstMemory), and most parts of the app, audio,
788 base, check, editing-services, gl, net. pbutils, player, rtsp,
789 rtsp-server, sdp, video and webrtc libraries.
791 Also included is support for creating subclasses of the following types
792 and writing GStreamer plugins:
795 - gst::Bin and gst::Pipeline
796 - gst::URIHandler and gst::ChildProxy
797 - gst::Pad, gst::GhostPad
798 - gst_base::Aggregator and gst_base::AggregatorPad
799 - gst_base::BaseSrc and gst_base::BaseSink
800 - gst_base::BaseTransform
802 Changes to 0.12.X since 0.12.0
806 - PTP clock constructor actually creates a PTP instead of NTP clock
810 - Bindings for GStreamer Editing Services
811 - Bindings for GStreamer Check testing library
812 - Bindings for the encoding profile API (encodebin)
814 - VideoFrame, VideoInfo, AudioInfo, StructureRef implements Send and
816 - VideoFrame has a function to get the raw FFI pointer
817 - From impls from the Error/Success enums to the combined enums like
819 - Bin-to-dot file functions were added to the Bin trait
820 - gst_base::Adapter implements SendUnique now
821 - More complete bindings for the gst_video::VideoOverlay interface,
823 gst_video::is_video_overlay_prepare_window_handle_message()
827 - All references were updated from GitHub to freedesktop.org GitLab
828 - Fix various links in the README.md
829 - Link to the correct location for the documentation
830 - Remove GitLab badge as that only works with gitlab.com currently
832 Changes in git master for 0.13
836 - gst::tag::Album is the album tag now instead of artist sortname
840 - Subclassing infrastructure was moved directly into the bindings,
841 making the gst-plugin crate deprecated. This involves many API
842 changes but generally cleans up code and makes it more flexible.
843 Take a look at the gst-plugins-rs crate for various examples.
845 - Bindings for CapsFeatures and Meta
847 ParentBufferMeta,VideoMetaandVideoOverlayCompositionMeta`
848 - Bindings for VideoOverlayComposition and VideoOverlayRectangle
849 - Bindings for VideoTimeCode
851 - UniqueFlowCombiner and UniqueAdapter wrappers that make use of the
852 Rust compile-time mutability checks and expose more API in a safe
853 way, and as a side-effect implement Sync and Send now
855 - More complete bindings for Allocation Query
856 - pbutils functions for codec descriptions
857 - TagList::iter() for iterating over all tags while getting a single
858 value per tag. The old ::iter_tag_list() function was renamed to
859 ::iter_generic() and still provides access to each value for a tag
860 - Bus::iter() and Bus::iter_timed() iterators around the corresponding
863 - serde serialization of Value can also handle Buffer now
865 - Extensive comments to all examples with explanations
866 - Transmuxing example showing how to use typefind, multiqueue and
868 - basic-tutorial-12 was ported and added
872 - Rust 1.31 is the minimum supported Rust version now
873 - Update to latest gir code generator and glib bindings
875 - Functions returning e.g. gst::FlowReturn or other “combined” enums
876 were changed to return split enums like
877 Result<gst::FlowSuccess, gst::FlowError> to allow usage of the
878 standard Rust error handling.
880 - MiniObject subclasses are now newtype wrappers around the underlying
881 GstRc<FooRef> wrapper. This does not change the API in any breaking
882 way for the current usages, but allows MiniObjects to also be
883 implemented in other crates and makes sure rustdoc places the
884 documentation in the right places.
886 - BinExt extension trait was renamed to GstBinExt to prevent conflicts
887 with gtk::Bin if both are imported
889 - Buffer::from_slice() can’t possible return None
891 - Various clippy warnings
894 GStreamer Rust Plugins
896 Like the GStreamer Rust bindings, the Rust plugins are now officially
897 part of the GStreamer project and are also maintained in the GStreamer
900 In the 0.3.x versions this contained infrastructure for writing
901 GStreamer plugins in Rust, and a set of plugins.
903 In git master that infrastructure was moved to the GLib and GStreamer
904 bindings directly, together with many other improvements that were made
905 possible by this, so the gst-plugins-rs repository only contains
906 GStreamer elements now.
908 Elements included are:
910 - Tutorials plugin: identity, rgb2gray and sinesrc with extensive
913 - rsaudioecho, a port of the audiofx element
915 - rsfilesrc, rsfilesink
917 - rsflvdemux, a FLV demuxer. Not feature-equivalent with flvdemux yet
919 - threadshare plugin: ts-appsrc, ts-proxysrc/sink, ts-queue, ts-udpsrc
920 and ts-tcpclientsrc elements that use a fixed number of threads and
921 share them between instances. For more background about these
922 elements see Sebastian’s talk “When adding more threads adds more
923 problems - Thread-sharing between elements in GStreamer” at the
924 GStreamer Conference 2017.
926 - rshttpsrc, a HTTP source around the hyper/reqwest Rust libraries.
927 Not feature-equivalent with souphttpsrc yet.
929 - togglerecord, an element that allows to start/stop recording at any
930 time and keeps all audio/video streams in sync.
932 - mccparse and mccenc, parsers and encoders for the MCC closed caption
935 Changes to 0.3.X since 0.3.0
937 - All references were updated from GitHub to freedesktop.org GitLab
938 - Fix various links in the README.md
939 - Link to the correct location for the documentation
941 Changes in git master for 0.4
943 - togglerecord: Switch to parking_lot crate for mutexes/condition
944 variables for lower overhead
945 - Merge threadshare plugin here
946 - New closedcaption plugin with mccparse and mccenc elements
947 - New identity element for the tutorials plugin
949 - Register plugins statically in tests instead of relying on the
950 plugin loader to find the shared library in a specific place
952 - Update to the latest API changes in the GLib and GStreamer bindings
953 - Update to the latest versions of all crates
956 Build and Dependencies
958 - The MESON BUILD SYSTEM BUILD IS NOW FEATURE-COMPLETE (*) and it is
959 now the recommended build system on all platforms and also used by
960 Cerbero to build GStreamer on all platforms. The Autotools build is
961 scheduled to be removed in the next cycle. Developers who currently
962 use gst-uninstalled should move to gst-build. The build option
963 naming has been cleaned up and made consistent and there are now
964 feature options to enable/disable plugins and various other features
965 on a case-by-case basis. (*) with the exception of plugin docs which
966 will be handled differently in future
968 - Symbol export in libraries is now controlled via explicit exports
969 using symbol visibility or export defines where supported, to ensure
970 consistency across all platforms. This also allows libraries to have
971 exports that vary based on detected platform features and configure
972 options as is the case with the GStreamer OpenGL integration library
973 for example. A few symbols that had been exported by accident in
974 earlier versions may no longer be exported. These symbols will not
975 have had declarations in any public header files then though and
976 would not have been usable.
978 - The GStreamer FFmpeg wrapper plugin (gst-libav) now depends on
979 FFmpeg 4.x and uses the new FFmpeg 4.x API and stopped relying on
980 ancient API that was removed with the FFmpeg 4.x release. This means
981 that it is no longer possible to build this module against an older
982 system-provided FFmpeg 3.x version. Use the internal FFmpeg 4.x copy
983 instead if you build using autotools, or use gst-libav 1.14.x
984 instead which targets the FFmpeg 3.x API and _should_ work fine in
985 combination with a newer GStreamer. It’s difficult for us to support
986 both old and new FFmpeg APIs at the same time, apologies for any
987 inconvenience caused.
989 - Hardware-accelerated Nvidia video encoder/decoder plugins nvdec and
990 nvenc can be built against CUDA Toolkit versions 9 and 10.0 now. The
991 dynlink interface has been dropped since it’s deprecated in 10.0.
993 - The (optional) OpenCV requirement has been bumped to >= 3.0.0 and
994 the plugin can also be built against OpenCV 4.x now.
996 - New sctp plugin based on usrsctp (for WebRTC data channels)
999 Platform-specific changes and improvements
1003 - The way that GIO modules are named has changed due to upstream GLib
1004 natively adding support for loading static GIO modules. This means
1005 that any GStreamer application using gnutls for SSL/TLS on the
1006 Android or iOS platforms (or any other setup using static libraries)
1007 will fail to link looking for the g_io_module_gnutls_load_static()
1008 function. The new function name is now
1009 g_io_gnutls_load(gpointer data). data can be NULL for a static
1010 library. Look at this commit for the necessary change in the
1015 - macOS binaries should be fully relocatable now
1017 - The way that GIO modules are named has changed due to upstream GLib
1018 natively adding support for loading static GIO modules. This means
1019 that any GStreamer application using gnutls for SSL/TLS on the
1020 Android or iOS platforms (or any other setup using static libraries)
1021 will fail to link looking for the g_io_module_gnutls_load_static()
1022 function. The new function name is now
1023 g_io_gnutls_load(gpointer data). data can be NULL for a static
1024 library. Look at this commit for the necessary change in the
1029 - The webrtcdsp element is shipped again as part of the Windows binary
1030 packages, the build system issue has been resolved.
1032 - ‘Inconsistent DLL linkage’ warnings when building with MSVC have
1035 - Hardware-accelerated Nvidia video encoder/decoder plugins nvdec and
1036 nvenc build on Windows now, also with MSVC and using Meson.
1038 - The ksvideosrc camera capture plugin supports 16-bit grayscale video
1041 - The wasapisrc audio capture element implements loopback recording
1042 from another output device or sink
1044 - wasapisink recover from low buffer levels in shared mode and some
1045 exclusive mode fixes
1047 - dshowsrc now implements the GstDeviceMonitor interface
1052 Aleix Conchillo Flaqué, Alessandro Decina, Alexandru Băluț, Alex Ashley,
1053 Alexey Chernov, Alicia Boya García, Amit Pandya, Andoni Morales
1054 Alastruey, Andreas Frisch, Andre McCurdy, Andy Green, Anthony Violo,
1055 Antoine Jacoutot, Antonio Ospite, Arun Raghavan, Aurelien Jarno,
1056 Aurélien Zanelli, ayaka, Bananahemic, Bastian Köcher, Branko Subasic,
1057 Brendan Shanks, Carlos Rafael Giani, Christoph Reiter, Corentin Noël,
1058 Daeseok Youn, Daniel Drake, Daniel Klamt, Dardo D Kleiner, David Ing,
1059 David Svensson Fors, Devarsh Thakkar, Dimitrios Katsaros, Edward Hervey,
1060 Emilio Pozuelo Monfort, Enrique Ocaña González, Ezequiel Garcia, Fabien
1061 Dessenne, Fabrizio Gennari, Florent Thiéry, Francisco Velazquez,
1062 Freyr666, Garima Gaur, Gary Bisson, George Kiagiadakis, Georg Lippitsch,
1063 Georg Ottinger, Geunsik Lim, Göran Jönsson, Guillaume Desmottes, H1Gdev,
1064 Haihao Xiang, Haihua Hu, Harshad Khedkar, Havard Graff, He Junyan,
1065 Hoonhee Lee, Hosang Lee, Hyunjun Ko, Ingo Randolf, Iñigo Huguet, James
1066 Stevenson, Jan Alexander Steffens, Jan Schmidt, Jerome Laheurte, Jimmy
1067 Ohn, Joakim Johansson, Jochen Henneberg, Johan Bjäreholt, John-Mark
1068 Bell, John Nikolaides, Jonathan Karlsson, Jonny Lamb, Jordan Petridis,
1069 Josep Torra, Joshua M. Doe, Jos van Egmond, Juan Navarro, Jun Xie,
1070 Junyan He, Justin Kim, Kai Kang, Kim Tae Soo, Kirill Marinushkin, Kyrylo
1071 Polezhaiev, Lars Petter Endresen, Linus Svensson, Louis-Francis
1072 Ratté-Boulianne, Luis de Bethencourt, Luz Paz, Lyon Wang, Maciej Wolny,
1073 Marc-André Lureau, Marc Leeman, Marcos Kintschner, Marian Mihailescu,
1074 Marinus Schraal, Mark Nauwelaerts, Marouen Ghodhbane, Martin Kelly,
1075 Matej Knopp, Mathieu Duponchelle, Matteo Valdina, Matthew Waters,
1076 Matthias Fend, memeka, Michael Drake, Michael Gruner, Michael Olbrich,
1077 Michael Tretter, Miguel Paris, Mike Wey, Mikhail Fludkov, Naveen
1078 Cherukuri, Nicola Murino, Nicolas Dufresne, Niels De Graef, Nirbheek
1079 Chauhan, Norbert Wesp, Ognyan Tonchev, Olivier Crête, Omar Akkila,
1080 Patricia Muscalu, Patrick Radizi, Patrik Nilsson, Paul Kocialkowski, Per
1081 Forlin, Peter Körner, Peter Seiderer, Petr Kulhavy, Philippe Normand,
1082 Philippe Renon, Philipp Zabel, Pierre Labastie, Roland Jon, Roman
1083 Sivriver, Rosen Penev, Russel Winder, Sam Gigliotti, Sean-Der, Sebastian
1084 Dröge, Seungha Yang, Sjoerd Simons, Snir Sheriber, Song Bing, Soon,
1085 Thean Siew, Sreerenj Balachandran, Stefan Ringel, Stephane Cerveau,
1086 Stian Selnes, Suhas Nayak, Takeshi Sato, Thiago Santos, Thibault
1087 Saunier, Thomas Bluemel, Tianhao Liu, Tim-Philipp Müller, Tomasz
1088 Andrzejak, Tomislav Tustonić, U. Artie Eoff, Ulf Olsson, Varunkumar
1089 Allagadapa, Víctor Guzmán, Víctor Manuel Jáquez Leal, Vincenzo Bono,
1090 Vineeth T M, Vivia Nikolaidou, Wang Fei, wangzq, Whoopie, Wim Taymans,
1091 Wind Yuan, Wonchul Lee, Xabier Rodriguez Calvar, Xavier Claessens,
1092 Haihao Xiang, Yacine Bandou, Yeongjin Jeong, Yuji Kuwabara, Zeeshan Ali,
1094 … and many others who have contributed bug reports, translations, sent
1095 suggestions or helped testing.
1100 - this section will be filled in in due course
1102 More than XXX bugs have been fixed during the development of 1.16.
1104 This list does not include issues that have been cherry-picked into the
1105 stable 1.16 branch and fixed there as well, all fixes that ended up in
1106 the 1.16 branch are also included in 1.16.
1108 This list also does not include issues that have been fixed without a
1109 bug report in bugzilla, so the actual number of fixes is much higher.
1114 After the 1.16.0 release there will be several 1.16.x bug-fix releases
1115 which will contain bug fixes which have been deemed suitable for a
1116 stable branch, but no new features or intrusive changes will be added to
1117 a bug-fix release usually. The 1.16.x bug-fix releases will be made from
1118 the git 1.16 branch, which is a stable branch.
1122 1.16.0 is scheduled to be released around January/February 2019.
1127 - possibly breaking/incompatible changes to properties of wrapped
1128 FFmpeg decoders and encoders (see above).
1130 - The way that GIO modules are named has changed due to upstream GLib
1131 natively adding support for loading static GIO modules. This means
1132 that any GStreamer application using gnutls for SSL/TLS on the
1133 Android or iOS platforms (or any other setup using static libraries)
1134 will fail to link looking for the g_io_module_gnutls_load_static()
1135 function. The new function name is now
1136 g_io_gnutls_load(gpointer data). See Android/iOS sections above for
1142 Our next major feature release will be 1.18, and 1.17 will be the
1143 unstable development version leading up to the stable 1.18 release. The
1144 development of 1.17/1.18 will happen in the git master branch.
1146 The plan for the 1.18 development cycle is yet to be confirmed, but it
1147 is expected that feature freeze will be around July 2019 followed by
1148 several 1.17 pre-releases and the new 1.18 stable release in
1151 1.18 will be backwards-compatible to the stable 1.16, 1.14, 1.12, 1.10,
1152 1.8, 1.6, 1.4, 1.2 and 1.0 release series.
1154 ------------------------------------------------------------------------
1156 _These release notes have been prepared by Tim-Philipp Müller with_
1157 _contributions from Sebastian Dröge._
1159 _License: CC BY-SA 4.0_