The common case of the pointer having increased by one packet (which results
in no change to the modulus) can be detected with a 64-bit subtraction,
which is far cheaper than a division on many platforms.
Before After
Mean StdDev Mean StdDev Change
Divisions 248.3 8.8 51.5 7.4 +381.7%
Overall 2773.2 25.6 2372.5 43.1 +16.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
When a stream contains a single program, there's no point in doing a
PID -> program lookup. Normally the one and only program isn't disabled,
so no packets should be discarded.
Before After
Mean StdDev Mean StdDev Change
discard_pid() 73.8 9.4 20.2 1.5 +264.8%
Overall 2300.8 28.0 2253.1 20.6 +2.1%
Signed-off-by: Martin Storsjö <martin@martin.st>
This was being performed to ensure that a complete packet was held in
contiguous memory, prior to parsing the packet. However, the source buffer
is typically large enough that the packet was already contiguous, so it is
beneficial to return the packet by reference in most cases.
Before After
Mean StdDev Mean StdDev Change
memcpy 720.7 32.7 649.8 25.1 +10.9%
Overall 2372.7 46.1 2291.7 21.8 +3.5%
Signed-off-by: Martin Storsjö <martin@martin.st>
As long as there is enough contiguous data in the avio buffer,
just return a pointer to it instead of copying it to the caller
provided buffer.
Signed-off-by: Martin Storsjö <martin@martin.st>
A separate rtcp port can already be set when opening the rtp
protocol normally, but when doing port setup as in RTSP (where
we first need to open the local ports and pass them to the peer,
and only then receive the remote peer port numbers), we didn't
check the same url parameter as in the normal open routine.
Signed-off-by: Martin Storsjö <martin@martin.st>
I doubt that anyone ever would try to send a 1 byte packet
via the RTP protocol, but check just in case - it shouldn't
crash at least.
Signed-off-by: Martin Storsjö <martin@martin.st>
This allows the chained demuxer (or more precisely, the lavf
utility code) to better fill in timestamps on packets from
these, especially for cases where one stream is a raw ADTS
stream.
Signed-off-by: Martin Storsjö <martin@martin.st>
Add support for domain names, for multiple source addresses,
for exclusions, and for session level specification of addresses.
Signed-off-by: Martin Storsjö <martin@martin.st>
This allows us to explicitly fail if the caller tried to set
both inclusions and exclusions at the same time.
Signed-off-by: Martin Storsjö <martin@martin.st>
Previously this only allowed literal IP addresses. When these
are conveyed in a SDP file as in RFC4570, host names are allowed
as well.
Signed-off-by: Martin Storsjö <martin@martin.st>
Also parse segment durations as floating point, which is allowed
since HLS version 3.
This is based on a patch by Zhang Rui.
Signed-off-by: Martin Storsjö <martin@martin.st>
When first_timestamp was stored as-is, its actual time base
wasn't known later in the seek function.
Additionally, the logic (from 795d9594cf) for scaling it
based on stream_index is flawed - stream_index in the seek
function only specifies which stream the seek timestamp refers
to, but obviously doesn't say anything about which stream
first_timestamp belongs to.
In the cases where stream_index was >= 0 and all streams had the
same time base, this didn't matter in practice.
Seeking taking first_timestamp into account is problematic
when one variant is mpegts (with real timestamps) and one variant
is raw ADTS (with timestamps only being accumulated packet
duration), where the variants start at totally different timestamps.
Signed-off-by: Martin Storsjö <martin@martin.st>
Without the information, an application may choose audio from one
variant and video from another variant, which leads to fetching two
variants from the network. This enables av_find_best_stream() to find
matching audio and video streams, so that only one variant is fetched.
Signed-off-by: Martin Storsjö <martin@martin.st>
Also adjust the streams timestamps according to their start
timestamp when comparing. This helps getting correctly interleaved
packets if one stream lacks timestamps (such as a plain ADTS
stream when the other variants are full mpegts) when the others
have timestamps that don't start from zero.
This probably doesn't work properly if such a stream is
temporarily disabled (via the discard flags) and then reenabled,
and such streams are hard to correctly sync against the other
streams as well - but this works better than before at least.
The segment number restriction makes sure all variants advance
roughly at the same pace as well.
Signed-off-by: Martin Storsjö <martin@martin.st>
If passing the end of one segment while initializing the
chained demuxer, the parent demuxer's streams aren't set up
yet, so we can't recheck the discard flags.
Signed-off-by: Martin Storsjö <martin@martin.st>
This serves as a safeguard; normally we want to use the dts
comparison to interleave packets from all active variants. If that
dts comparison for some reason doesn't work as intended, make sure
that all packets in all variants for a certain sequence number have
been returned before moving on to the next one.
Signed-off-by: Martin Storsjö <martin@martin.st>
Incomplete crypted files would lead to a read after buffer boundary
otherwise.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Derived from VLC's http module.
Original authors:
Antoine Cellerier <dionoea@videolan.org>
Sébastien Escudier <sebastien-devel@celeos.eu>
Rémi Duraffort <ivoire@videolan.org>
Rémi Denis-Courmont <remi@remlab.net>
Francois Cartegnie <fcvlcdev@free.fr>
Normally, http servers shouldn't send this to us since we
don't advertise it with an Accept-Encoding header, but some
servers still do it anyway.
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes sure that values that are left-shifted by this constant
end up casted to 64 bit before shifting, avoiding overflow if the
value ends up larger than 2 GB.
Signed-off-by: Martin Storsjö <martin@martin.st>
