Also ensure we get a proper EOS when seeking past the end of the
file, or (when the snap after flag is set) between the last keyframe
and the end of the file.
Two things were suboptimal from a performance point of view:
a) consider a large media object such as a video keyframe, which
may be split up into multiple fragments. We would assemble
the media object as follows:
buf = join (join (join (frag1, frag2), frag3), frag4)
which causes many unnecessary memcpy()s, and malloc/free,
which could easily add up to a multiple of the actual object
size. To avoid this, we allocate a buffer of the size needed
from the start and copy fragments into that directly.
b) for every fragment to join, we would create a sub-buffer
before joining it (which would discard the sub-buffer again),
leading to unnecessary miniobject create/free churn.
Conflicts:
gst/asfdemux/asfpacket.c
gst/asfdemux/asfpacket.h
We have already retrieved the stream for that stream number and
made sure it's not NULL, so no need to do it again here; neither
the number nor the streams changed since the last time.
Add private replacements for deprecated functions such as
g_mutex_new(), g_mutex_free(), g_cond_new() etc., mostly
to avoid the deprecation warnings. We can't change most of
these in 0.10 because they're part of our API and ABI.
Suppress deprecation warnings in selected files, mostly for
g_static_rec_mutex_*. StaticRecMutex is part of our API/ABI,
not much we can do here in 0.10.
Quite a few (broken?) files have a packet duration of 1ms, which is
most definitely wrong for either audio or video packets.
We therefore avoid using that value and instead use other metrics to
determine the buffer duration (like using the extended stream properties
average frame duration if present and valid).
Only error out when downstream returns:
* NOT_SUPPORTED
* ERROR
* NOT_NEGOTIATED
* NOT_LINKED
If we got _UNEXPECTED, we push an EOS downstream (since maybe only one
of the streams had gone EOS) and then stop the task silently.
In the case of WRONG_STATE we just need to stop silently
https://bugzilla.gnome.org/show_bug.cgi?id=600412
When on push mode and receiving an EOS event, asfdemux
should push all pending data because we might be dealing
with a broken file that has a preroll value higher
than its actual length.
Some (broken) streams don't have the extended stream properties in
the header, resulting in applying a duration of zero on outgoing
buffers.
Fixes#611473
Some files have payload with timestamps smaller than the preroll duration.
Instead of blindly substracting the preroll value (and ending up with
insanely high timestamps on the outgoing buffers), we make sure we
never go below 0.
Fixes#610432
We previously only aggregated flow returns after the while(push) loop,
which meant that in some cases we would end-up not properly aggregating
the flow returns.
This is based on the same flow aggregation algorithm as oggdemux.
Adds chained asfs handling to pull mode. It now checks if
there is a new asf header after the last packet (when it
is possible to know how many packets are) or it tries
checking if a processed packet that fails is an header
object.
Fixes#599718
Adds support for detecting and playing chained asfs
in push mode. asfdemux tries to detect a new asf start
by identifying the header object guid in a input buffer.
When it finds it, it resets its state, removing its pads
and creates new ones for the new file.
When receiving bogus data, we have to avoid subtracting a value
larger than 'size' from 'size' variable, resulting in a wrap
that would make 'size' a really large bogus value.
Fixes#599333
asf packets in rtp packets should come with their padding fields
set to 0 and the depayload must update them to the correct
value before pushing downstream
This also fixes a bug by which the first buffer (in a multi-packet mode)
passed to asf_demux_parse_packet() would have a GST_BUFFER_SIZE of the
full incoming buffer and not just of the single asf packet.
Fixes corrupted frames introduced by latest commit.
We now have a chance for packets to be collected before we send out the
newsegment. If we're not in accurate seeking (keyunit) it will set
the segment start/time to the keyframe's timestamp.
We now *always* seek to the keyframe just before our requested position.
When we encounter the first keyframe and we were not accurate (therefore doing
keyframe seeking), we update the segment start position to the keyframe timestamp.
This will still cause some timestamp jitter, but giving a hint as to the duration
rather than nothing seems to be a better idea.
Also, this allows some scenarios (like remuxing with asfmux) to estimate the total
duration using the accumulated packet duration (which will be correct).
The simple index entries also contain the number of packets one needs
to retrieve at a given position to get a full keyframe. We therefore
use that information to retrieve all those packets in one buffer when
working in pull-mode.
In gst_asf_demux_chain_headers, when 'goto wrong_type' was called
asfdemux tried to free a const pointer that had been cast to a
normal pointer variable.
We weren't taking the preroll into account previously, meaning that we
were always seeking preroll nanoseconds too early... resulting in a lot
of dropped packets (which are before the start time).
This brings quit a bit closer to as-fast-as-possible seeking in asf files.
Post global tags only after we've added our source pads, so that
tag events get sent downstream in addition to tag messages posted
on the bus. This makes sure tags can be picked up automatically
when transcoding, but also by tagreadbin/playbin2. Fixes#519721.
While we're at it, also add a container-format tag.
When we receive a DISCONT as input, don't clear our complete state but simply
mark a discont that will be put on the next buffer. The code will be able to
handle and throw away incomplete data.
Add some more debug info.
Remove an unused variable.
Don't overwrite the origin flow return by whatever flow we get
when trying to push the remaining internally queued payloads.
We want to do our eos logic, ie. send an EOS event or segment-done
message in any case. Makes things EOS properly when an EOS event
is forced upon the pipeline so that the source returns
FLOW_UNEXPECTED to a pulling asfdemux. Should fix#582056.
This also makes timestamps (more) consistent before and after a possible
seek, and moreover makes for reasonable position reporting in live stream
(whose payload timestamps should not be taken for granted).
* Improve newsegment handling, e.g. upstream might live in TIME.
* Only send newsegment if we have needed info.
* Avoid reading past end of data section.
The problem that happens is the following:
* A packet with multiple payloads comes in
* Those payloads get handled one by one
* The first payload contains the first audio payload with timestamp A
* The second payload contains the first video (key)frame with timestamp V (where V < A)
With the previous code, the following would happen:
* the first payload gets processed, then passed to queue_for_stream
* queue_for_stream detects it's the first valid timestamp received and stores
first_ts = A
* the second payload gets processed, then pass to queue_for_stream
* queue_for_stream detects the timestamp is lower than first_ts... and
discards it... resulting in losing the first keyframe of the video stream
We've been having this issue for *ages*... it's just that nobody noticed it
that much with playbin. But with playbin2's aggresive multiqueue handling, this
will result in multiqueue not being able to preroll (because the video decoder will
be dropping a ton of buffers before (maybe) receiving the next keyframe).
Tested with over 200 asf files, and they all play the first frame correctly now,
even the most braindead ones.
This might be caused by entering the if() line 1214 and then not having
any activated_streams.. resulting in reaching line 1267 without having
any valid flow value.
On win32, we're required to link to all the libraries used - including
ones only indirectly used by other libs. So, add gstaudio, gsttag, and
(for windows only) winsock.
Drop packets with an invalid replicated data length
instead of continuing with an invalid timestamp
and uninitialized payload metadata.
All other code assumes that the timestamps are valid.
