[Udpcast] Experiencing reproducible corruption when using udpcast-20090912
jice at wink.com
Wed Sep 16 06:58:32 CEST 2009
John Eisenman wrote:
> I'm new to udpcast and so may be doing something wrong. However, I find
> that with the latest two versions (udpcast-20090912 or
> udpcast-20090830), I experience repeated file corruption. I've had
> better, but not perfect, success with an old version (udpcast-20071228).
> In order to test, I made an ordered file, 500MB in size:
> cl6 tmp # ls -l test-52428800.dat
> -rw-r--r-- 1 jice jice 524288000 Sep 15 17:52 test-52428800.dat
> cl6 tmp # head -14 test-52428800.dat
> I find that every time I send and receive a file, it is the correct
> length. However, certain data is missing and other data is repeated.
> Here is an example of my command line to send the data:
> udp-sender --log udpcast.log --bw-period 1 --nokbd --min-receivers 1
> --min-wait 5 --portbase 9010 -f test-52428800.dat --pointopoint
> Here is my command line to receive data:
> udp-receiver -f test2 --nokbd --portbase 9010
> I have test result files that I can provide if they would be helpful.
> I am running under Gentoo on AMD64 processors. I've tried under two
> different kernels:
> # uname -a
> Linux cl7 2.6.20-gentoo-r8 #1 SMP PREEMPT Wed Aug 22 14:28:25 PDT 2007
> x86_64 AMD Opteron(tm) Processor 246 AuthenticAMD GNU/Linux
> # uname -a
> Linux m231 2.6.30-gentoo-r4pb #2 SMP Thu Aug 27 18:41:51 PDT 2009 x86_64
> Quad-Core AMD Opteron(tm) Processor 2374 HE AuthenticAMD GNU/Linux
By running in async mode (and setting an fec), I find that I am able to
produce a non-corrupt output file. Perhaps this provides a clue as to
the source of the problem.
More information about the Udpcast