If stdin has been closed, 0 is a valid socket descriptor.
Signed-off-by: Martin Storsjö <martin@martin.st>

Make sure we actually have an error code in ret, in case
getsockopt failed.
Signed-off-by: Martin Storsjö <martin@martin.st>

Also use ff_neterrno() instead of errno directly (which doesn't work
on windows), for getting the error code.
Signed-off-by: Martin Storsjö <martin@martin.st>

getnameinfo doesn't set errno on failure, it returns an error code,
which should be handled by gai_strerror instead of the normal
strerror.
Signed-off-by: Martin Storsjö <martin@martin.st>

Signed-off-by: Martin Storsjö <martin@martin.st>

This avoids warnings.
Signed-off-by: Martin Storsjö <martin@martin.st>

Rtmpt is effectively half duplex - the server can't return any
data unless we send a request (to which the server responds). If
we don't have any data to send currently, and the server didn't
return any data either, wait a little before doing the next request.

This avoids busy looping with idle posts with empty replies, while
waiting for more data from the server.
Signed-off-by: Martin Storsjö <martin@martin.st>

The check uses check_func_header, since this function is
conditionally available depending on the targeted MSVCRT
version.
Signed-off-by: Martin Storsjö <martin@martin.st>

Add a note that pkt->data and pkt->size must be initialized.
Signed-off-by: Martin Storsjö <martin@martin.st>

Signed-off-by: Martin Storsjö <martin@martin.st>

Signed-off-by: Martin Storsjö <martin@martin.st>

Signed-off-by: Martin Storsjö <martin@martin.st>

The packet size, signalled via block_align, has to be passed via
the container.
Signed-off-by: Martin Storsjö <martin@martin.st>

The library is 3-clause BSD licensed.
Signed-off-by: Martin Storsjö <martin@martin.st>

Add a new option 'rtmp_flush_interval' that allows specifying the
number of packets to write before sending it off as a HTTP request.

This is mostly relevant for RTMPT - for plain RTMP, it only controls
how often we check the socket for incoming packets, which shouldn't
affect the performance in any noticeable way.
Signed-off-by: Martin Storsjö <martin@martin.st>

This seems to be the correct mode to send, according to the
original RTSP RFC, and matches the method RECORD which is
sent later when starting to send data.

Darwin Streaming Server works fine with either of them.
Signed-off-by: Martin Storsjö <martin@martin.st>

Signed-off-by: Mans Rullgard <mans@mansr.com>

Signed-off-by: Justin Ruggles <justin.ruggles@gmail.com>

2x speedup of surround decoding on Cortex-A9.
Signed-off-by: Mans Rullgard <mans@mansr.com>

Signed-off-by: Mans Rullgard <mans@mansr.com>

This requires all NAL units to fit within single RTP packets. It
doesn't change the actual packetization for packets that fit, but
errors out and gives a helpful hint if the NAL units would have to
be split, and signals the right packetization mode in the SDP.
Signed-off-by: Martin Storsjö <martin@martin.st>

Only these three libraries were out of order, the rest was already
neatly sorted.
Signed-off-by: Martin Storsjö <martin@martin.st>

Signed-off-by: Mans Rullgard <mans@mansr.com>

This adds two protocols, but one of them is an internal implementation
detail just used as an abstraction layer/generalization in the code. The
RTMPT protocol implementation uses rtmphttp:// as an alternative to the
tcp:// protocol. This allows moving most of the lower level logic out
from the higher level generic rtmp code.
Signed-off-by: Martin Storsjö <martin@martin.st>