The VideoLAN Forums

I'd like to report some vlc crash case:
make two channels vlm conf file.
there are two vlc instances
vlc instance 1:
new channel1 broadcaster enable
setup channel1 input udp://@238.1.1.1:1234
setup channel1 output #transcode{vcode=DIV3,vb=2800,acode=mp4,ab=32,scale=1}:std{access=http,mux=ts,dst=0.0.0.0:8080}

vlc instance 2:
new channel2 broadcaster enable
setup channel2 input http://192.168.0.37:8080
setup channel2 output #transcode{vcode=h264,vb=800,acode=mp4,ab=32,scal=1}:std{access=http,mux=ts,dst=0.0.0.0:7000}

channel1 receive a udp stream and transcode to DIV3
channel2 transcode it to h264.
it is work, and I play 192.168.0.37 http stream very good. but when I disconnect the network cable of channel1 or kill instance 1 and keep this status 2 or 3 even 8 hours the crash happen. I can not find out any vlc instance in windows, and even whole machine stuck, I can not visit that machine with telnet or ftp.

BTW for two ET port config, I've to disable and reenable NIC of instance2 or disconnect and reconnect its network cable at beginning to make instance 1 work due to vlc DO NOT support udp multicast NIC selection.

is this a memory leak bug? and no way to solve multiple NIC selection?

is it known memory leak bug?
I can repeat this. and I also found the memory abnormal record. 1G memory have been eaten up.
Date : Time : free physical mem : used physical mem: free swap file: used swap file: % of mem using
07/01/2010 10:27:56 530.28 483.14 542.72 374.05 47
07/01/2010 11:05:57 101.72 911.70 112.74 804.03 89
07/01/2010 11:06:57 90.98 922.44 101.80 814.96 91
07/01/2010 11:07:57 82.32 931.10 92.03 824.74 91
07/01/2010 11:08:57 70.53 942.89 80.18 836.59 93
07/01/2010 11:09:58 58.43 954.99 68.20 848.57 94
07/01/2010 11:10:58 46.99 966.43 56.82 859.95 95
07/01/2010 11:11:58 34.35 979.07 44.39 872.38 96
07/01/2010 11:12:58 24.04 989.38 33.71 883.05 97
07/01/2010 11:13:58 12.16 1001.25 21.94 894.83 98
07/01/2010 11:14:58 3.49 1009.93 10.18 906.59 99
07/01/2010 11:15:58 673.84 339.57 688.22 228.55 33
07/01/2010 11:16:58 673.82 339.60 688.31 228.46 33
07/01/2010 11:17:58 674.17 339.25 688.25 228.52 33

It's more or less a known problem. When an input elementary stream stops emitting data, the multiplexer can never dequeue and multiplex data from other elementary streams, and starts buffering to no end.

thanks the reply.
is there any plan to solve it?
how about only one ES for this? (no other ES) will it happen no end buffer?
is it possible to set a configable "used mem" value to force dequeue and end buffer and release these resouce?

Nobody committed to fixing the problem.

Same problem viewtopic.php?f=13&t=73966.

I got same problem... Solution I use: restart VLC each hour using cron or windows scheduler.

I got same problem... Solution I use: restart VLC each hour using cron or windows scheduler.

What OS you using? And did you try version 1.1-git?

I have several 24/7 VLC based streaming servers on FreeBSD 7.2 8.0 and Fedora Core 12. As I assume this is OS independent problem. FreeBSD is most stable for this purposes (on Fedora Core VLC some times stops streaming without any errors and starts only on cron's request ... )
I use VLC compiled from ports (on FreeBSD) and RPM packages for FC, VLC's version is 1.0.1. I use Fire Wire captured live DV stream as source for 24/7 streaming. If I use some file as source this problem not appears even on FC...

Is there some new abilities in git version which can increase stability ?

... I thinking about possibility of VLC integration to inittab ....

VLC 1.1 should be able to be more source-issues resiliant.

I'am experiencing same problem with newest VLC 1.1.5 on Linux and Windows system which prevents using VLC in production environment..