Mistserver 1.2RC1 live errors and crash

137 views
Skip to first unread message

matt...@ymir.it

unread,
Sep 23, 2013, 5:23:34 AM9/23/13
to mists...@googlegroups.com
Hi.
I'm testing version 1.2RC1 (release tarball) for live streaming.
I've found out 2 problems:

1. the default configuration activated from scratch (all protocols enabled) results in an endless waiting on the player side. I can see the push from console, stream is "active" but the player either waits forever or ends with "unable to play" message.
Removing default configuration and activating only HTTP, RTMP, HTTPLIVE results in a working environment.

2. when pushing a stream, i can get a working live only when using ffmpeg. every other program I tried (Wirecast, Adobe FMLE works for a couple of seconds, and then on console I can see

terminate called after throwing an instance of 'std::out_of_range'
  what():  vector::_M_range_check

and the playing stops with player buffering. Restarting the push results in another couple of seconds of playback, with the same message and the crash.

Any advice?

Thanks in advance





Jaron Vietor

unread,
Sep 23, 2013, 9:38:19 AM9/23/13
to mists...@googlegroups.com
Hello,

Your first problem is because 1.2 wil attempt HTML5 if the client supports it, but we don't support HTML5 live streaming properly just yet. I'll make sure this gets sorted before the proper 1.2 release - until then what you did is a good workaround.

As to your second problem - that's very strange... What encoder settings are you using to stream? We need to reproduce the problem here to be able to fix this.

Regards,
Jaron Viëtor



--
You received this message because you are subscribed to the Google Groups "Mistserver.org" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mistserver+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

matt...@ymir.it

unread,
Sep 23, 2013, 10:43:21 AM9/23/13
to mists...@googlegroups.com


On Monday, September 23, 2013 3:38:19 PM UTC+2, Jaron Viëtor wrote:
Hello,

Your first problem is because 1.2 wil attempt HTML5 if the client supports it, but we don't support HTML5 live streaming properly just yet. I'll make sure this gets sorted before the proper 1.2 release - until then what you did is a good workaround.

By "HTML5 live" you mean? HLS, DASH? I don't know if any "stable" browser does actually support live streams, maybe flash (with html5 as fallback) is the safer solution for now?
 

As to your second problem - that's very strange... What encoder settings are you using to stream? We need to reproduce the problem here to be able to fix this.


I did try a lot of presets, starting from the very same settings as the ffmpeg run.

Flash push, H264 from 360p to 720p, 25/30fps, baseline/main, 300/1200kps, AAC stereo 44.1/48k 96/128k bitrate.

Every test leads to the same result. I will try a video only stream, maybe AAC encoding is the culprit?

If you can suggest me a base profile to test I will send my results.

BTW, I did try to compile from git but the build is somehow broken,
/usr/bin/MistConnHTTPProgressive: symbol lookup error: /usr/bin/MistConnHTTPProgressive: undefined symbol: _ZN4Util6Config19addConnectorOptionsEi

Thanks


 

Jaron Vietor

unread,
Oct 19, 2013, 6:24:14 AM10/19/13
to mists...@googlegroups.com
Hello,

Sorry for the delay in my response.

With HTML5 I mean progressive MP4 and OGG. The latest RC release has this bug fixed, now. It will now use Flash for live but HTML5 for VoD if enabled and support is available.

The settings you've posted should work. Make sure the keyframe interval is set to a value somewhere between 1-4 seconds, for best results (higher values should still work, but not perform as well). And here too - make sure to try the latest RC available on the website (or recompile from the development branch - it's been updated to match), as we've fixed a lot of bugs and more other improvements recently.

In regard to your error - that's because the HTTP Progressive connector was split into several separate binaries. It's save (and recommended) to delete the old MistConnHTTPProgressive executable, after which that message should disappear.

Regards,
Jaron
Reply all
Reply to author
Forward
0 new messages