playbin cannot recover from network interruption while streaming a live HTTP stream

Submitted by Eric Toombs

Link to original bug (#773665)

Description

Created attachment 338770
starts a command upon receipt of HTTP GET and returns the result.

By live stream, here, I mean a stream whose content is being decided in real time.

When I try to stream such a stream with playbin, it dies after the first major network interruption (perhaps a 5 second outage). The buffer either gets stuck at 0% or goes from 0 to 100, then right back to zero again, with nothing being played during this time. I did the test using a gst-launch command like this one:

$ gst-launch-1.0 playbin uri=http://example.com/live/stream

I was testing a live flac stream (~100KB/s) over a wifi network. I simulated interruptions by toggling the rfkill for a few seconds. I tested the same stream with ffplay and with a gst-launch pipeline that looks like this:

$ gst-launch-1.0 souphttpsrc location=http://server:10000/ ! queue max-size-bytes=65536 max-size-buffers=0 max-size-time=0 leaky=downstream !
decodebin ! autoaudiosink sync=false

Both were able to recover without any problems. I am attaching the script I used as a simple HTTP server. I saw similar problems with pulseaudio-dlna's HTTP server. This is how I invoked my server:

$ todd bash -c 'pacat -r -d alsa_output.pci-0000_00_07.0.analog-stereo.monitor --latency=100000000 | flac --force-raw-format --endian=little --sign=signed --sample-rate=44100 --bps=16 --channels=2 -'

I hope somebody else can reproduce this. I've managed to do so on two different hosts, one armv7l and the other amd64, both Linux.

Attachment 338770, "starts a command upon receipt of HTTP GET and returns the result.":
todd

Version: 1.9.90