[linux-dvb] extra switch for tzap, problems with the scan utility: filter timeout pid appearing very often

Uwe Bugla uwe.bugla at gmx.de
Sat Jul 21 17:44:22 CEST 2007


Am Samstag, 21. Juli 2007 15:57:12 schrieben Sie:
> Le samedi 21 juillet 2007 00:30, Uwe Bugla a écrit :
> > Am Freitag, 20. Juli 2007 15:03:55 schrieb Christophe Thommeret:
> > > Le jeudi 19 juillet 2007 16:26, Uwe Bugla a écrit :
> > > > > > > (The main difference between dvbscan and kaffeine is that
> > > > > > > kaffeine filters one pid at a time (+ the nit pid in a second
> > > > > > > thread) while dvbscan creates up to 32 (iirc), so dvbscan is a
> > > > > > > bit faster, but maybe your device/driver doesn't like it.)
> > > > > >
> > > > > > 1. That sounds VERY interesting! so please WHERE in scan's source
> > > > > > code can I reduce the pid filtering to ONE PID and how. I do not
> > > > > > call myself a programmer, so can you please give me a hint where
> > > > > > I can find the appropriate sections in latest "scan.c" to patch
> > > > > > that thing? It indeed would save me a lot of pain if I only knew
> > > > > > where to look at and what to patch! :)
> > >
> > > try to decrease MAX_RUNNING in scan.c
> > >
> > > (btw, something may be wrong with your mail reader, as you see your
> > > response was double quoted.)
> >
> > Thank you for that hint, Christophe. :)
> >
> > Reducing "define MAX_RUNNING=1" instead of 27 in fact slows down the
> > whole process, but the real problem unfortunately stays.
> > In fact I have tried a couple of numbers: 1, 2, 10, 20......
> >
> > 7 walkthroughs - 7 different scan results. :(
> > WHERE the timeouts happen is highly "by chance". :(
> > There are channels missing in one walkthrough that appear in another and
> > so on.......
> >
> > The scan result you can never call "standardized" or equal, as the number
> > of dumped services is never at least close to equal. :(
> >
> > Kaffeine's scan result appears far more confidential than the one of
> > "scan".
> >
> > Any other hints from your part where I could poke around in scan.c to
> > resolve that issue???
>
> Unfortunately, no :(

Ok - that's what I expected, unfortunately. :(

Moreover I forgot to mention that if I start a DVB-Scan after certain 
walkthroughs the filter timeout pid problem appears right after trying the 
first transponder (= initial transponder), which is equal to a card driver 
oops. Consequence: The machine must be shutdown (i. e. NO soft reset helps!), 
then be restarted, and then "scan" got to be run again. That appears like 
a "plug and pray issue", highly Microsoft-compatible!

My card is a Pinnacle PCTVSAT PCI, i. e. a bt8xx-based DVB-S card, whose 
drivers are maintained (well, whatever that means in this very special 
example) by Manu Abraham.

I did a little research in the dvb-apps repository, and noticed that very long 
time ago Johannes Stezenbach pulled in a patch to lower down the maximum 
number of running pids from 32 to 27.
Whatever card he used as a reference or even proof that this would help to 
resolve the filter timeout pid issue remains a secret for me until today.

So the perspective is to wait until Mister Abraham has finished his cx878 
project (or should I call that a dead born child either?).

As four unusable kernels (at least for bt8xx-based DVB cards - 2.6.13 - 16)
showed in the past the ultimate Abraham-based top record of waitstates amounts 
to 9 months. I wonder if he can top that this time!

So I am forced to build up a deeply horrible list scan feature as part of my 
TCL-TK-project, because if I do the channel scan transponder by transponder 
(i. e. list-oriented), using parameter -c plus a loop I get reliable scan 
results.
That's highly time consuming, but that at least works.
This option I wanted to avoid - this is what we call bad fate!

>
> P.S.
> I've just added check for CA descriptors in PMT to correct CA flag in some
> cases.

In kaffeine I suppose - looking forward to kaffeine's next release. :(

Many thanks for your hints!

Best regards
Uwe



More information about the linux-dvb mailing list