Using this in a demuxer will cause deadlocks if there's
a pad with a pending pad-block downstream, no matter if
there is a queue between the pad or not. Queues pass
bufferalloc downstream from the same thread and only
act as a thread boundary for events and buffers.
Extra info can't hurt. Field names aren't necessarily consistent with
what's used elsewhere though (e.g. avidemux), but then neither are the
caps.
https://bugzilla.gnome.org/show_bug.cgi?id=623178
That is, if parse error occurs in state requiring to move to next cluster,
and doing so to the expected next position of cluster fails, then scan for a
next cluster from present position and resume from there.
Fixes#620790.
If some bits out of place in block(group) parsing, forego and move to next.
Also skip large blocks in pull mode, but need to give up in push mode.
Fixes#626463.
Improves #620790.
That is, in files that have no index (Cue), perform seek by scanning for
nearest cluster with timecode before requested position. Scanning is done
as a combination of interpolation and sequential scan.
Fixes#617368.
This allows us to skip delta units earlier and is a bit clearer in my
opinion. It also makes only video buffers ever be delta units, not
just for SimpleBlock as before.
When the keyframe bit of SimpleBlock Flags wasn't set, the buffer was being
marked with GST_BUFFER_FLAG_DELTA_UNIT, causing all buffers to be skipped
after a seek. This may be a problem with the Sorenson Squish encoder, but
arguably the keyframe bit should only be applied to video.
Fixes bug #620358.
In demuxer and muxer use the gst_util_uint64 scaling functions rather than
standard integer division. Add warnings (to be changed to debug) for debugging
the timestamp and duration.
Before, vp8dec had no option but to decode all frames even if some/all
of them would be late. With this change, performance when keyframes are
frequent is helped a great deal. On my Thinkpad X60s, decoding a 20 s
1080p sunflower encode with keyframes every 10 frames went from taking
42 s with 5 frames shown to 21 s with 15 frames shown (still slow
enough to count by hand). When keyframes are more sparse, you will
still be able to catch up eventually, but the results won't be as
noticable.
The original plan was to let WebM v1 be the same as Matroska v2 (with
extra constraints), but for simplicity it was decided to handle the
versions equally, such that e.g. SimpleBlock is only allowed in WebM v2.
Failure to do this for corrupt input can cause a subbuffer bigger
than the actual buffer to be created, quickly leading to segfault.
Test case:
bug_s222005751_r0.001____memcpy.webm
The comment says this cannot happen, but it did and I don't know
why. This is not the correct fix, needs investigation. Test case:
bug_s555010094_r0.0005:0.008____IA__g_assertion_message_expr.webm
Because GstMatroskaTrackContext *stream is set up in the first
SimpleBlock or Block, a rogue CodecState otherwise causes a segfault on
derefencing the NULL pointer. Test case:
bug_s5506167_r0.001____gst_matroska_demux_parse_blockgroup_or_simpleblock.webm
Changing it to the newest timestamp that was ever pushed will
increase the segment start in 500ms jumps, which could be just
after the next sparse stream buffer. E.g.
Video at 1.0s, sparse stream at 0.5s would jump the
sparse stream to 1.0s. Now a new sparse stream buffer could
appear that has a timestamp of 0.9s and this would be
dropped for no good reason because of bad luck.
As a side effect, avoid sending newsegment updates with start times
that go back and forth, which leads to bogus downstream running_time.
Also fixes seeking in bug #606744.
.. by initializing streams starting at 0, as that is basically
where we 'seek to' at the start and assume streams to start elsewhere.
Also enables newsegment update events for subtitle streams.
Matroska uses three-letter ISO 639-2B codes, but GST_TAG_LANGUAGE is
supposed to contain two-letter ISO 639-1 codes, so use new language
code mapping functions in -base to convert between those two as
needed.
Fixes#505823.
Send pending tags only from the streaming thread, just after we've sent
the newsegment event, not with e.g. flush-start. This not only does the
right thing, but also makes sure we're not trampling over variables set
up in the streaming thread from the seeking thread in case someone tries
to issue a seek just as the demuxer is parsing the headers.
Fixes#601617. Spotted by Ognyan Tonchev.
Don't leak buffers when resyncing to a keyframe.
Avoid leaking buffers when exiting the loop on error conditions.
Add some more debug info.
Fixes#585911
First of all a keyframe seek should be done to the
keyframe right before the requested position and not
to the keyframe that is nearest to the requested position.
Use per track index arrays and use our new binary search function
from core to speed up the search.
