On Mon, Dec 25, 2017 at 6:29 PM wm4 <nfx...@googlemail.com> wrote: > On Tue, 26 Dec 2017 02:14:11 +0000 > Aman Gupta <ffm...@tmm1.net> wrote: > > > On Mon, Dec 25, 2017 at 1:03 PM Michael Niedermayer > <mich...@niedermayer.cc> > > wrote: > > > > > On Mon, Dec 25, 2017 at 06:08:11PM +0000, Aman Gupta wrote: > > > > On Mon, Dec 25, 2017 at 9:58 AM Michael Niedermayer > > > <mich...@niedermayer.cc> > > > > wrote: > > > > > > > > > On Fri, Dec 22, 2017 at 11:00:34PM +0000, Aman Gupta wrote: > > > > > > ffmpeg | branch: master | Aman Gupta <a...@tmm1.net> | Wed Oct > 4 > > > > > 14:52:52 2017 -0700| [b7d6c0cd48dac7869b9e6803e2d47d05a4fa373b] | > > > > > committer: Aman Gupta > > > > > > > > > > > > avformat/hls: add http_persistent option > > > > > > > > > > > > This teaches the HLS demuxer to use the HTTP protocols > > > > > > multiple_requests=1 option, to take advantage of "Connection: > > > > > > Keep-Alive" when downloading playlists and segments from the HLS > > > server. > > > > > > > > > > > > With the new option, you can avoid TCP connection and TLS > negotiation > > > > > > overhead, which is particularly beneficial when streaming via a > > > > > > high-latency internet connection. > > > > > > > > > > > > Similar to the http_persistent option recently implemented in > > > hlsenc.c > > > > > > > > > > > > Signed-off-by: Aman Gupta <a...@tmm1.net> > > > > > > Signed-off-by: Anssi Hannula <anssi.hann...@iki.fi> > > > > > > > > > > > > > > > > > > > > > > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=b7d6c0cd48dac7869b9e6803e2d47d05a4fa373b > > > > > > --- > > > > > > > > > > > > doc/demuxers.texi | 4 ++++ > > > > > > libavformat/hls.c | 72 > > > > > +++++++++++++++++++++++++++++++++++++++++++++++++++---- > > > > > > 2 files changed, 72 insertions(+), 4 deletions(-) > > > > > > > > > > This causes ffmpeg to infinite loop with simplehttpd.py > > > > > > > > > > > > Is that the same as SimpleHTTPServer.py? > > > > > > > > Sounds like it doesn't support persistent connections, but also isn't > > > > returning "Connection: close" in the response. > > > > > > > > I can investigate if you tell me where to find that httpd server. > > > > > > Iam not sure where i got it from or its original source but it matches > the > > > source here: > > > > > > > > > > https://www.linuxjournal.com/content/tech-tip-really-simple-http-server-python > > > > > > Thanks. > > > > I was able to reproduce the issue. It appears to be because the simple > http > > server is quite too simple, and can only handle one request at a time. > The > > http_multiple option tries to make a second request before the first is > > completed, causing a deadlock. > > > > I guess we could disable http_multiple by default... > > > > Aman > > > > < > https://www.linuxjournal.com/content/tech-tip-really-simple-http-server-python > > > > So I hear this causes some infinite loop. Could this be made adaptive?
I guess we could detect if the server is http/1.1 with the first request and only make multiple connections if it is. Technically multiple connections should work with http/1.0 servers though. It's just this specific http server which can only handle one connection at a time. The "infinite loop" is because there is no timeout on the request socket, and the server never accepts the http request to send a response (because another connection is still active). > > What does the HTTP spec say about this, and what do other HTTP > implementations do here? The spec says nothing about multiple concurrent connections. Aman > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel > _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel