[linux-dvb] Re: dvb-kernel/stv0299 cvs >20040224

[Andrew de Quincey <adq_dvb@lidskialf.net>]

On Saturday 13 March 2004 03:28, Kenneth Aafløy wrote:
> On Friday 12 March 2004 22:49, you wrote:
> > On Friday 12 March 2004 21:27, Kenneth Aafløy wrote:
> > > On Friday 12 March 2004 22:16, you wrote:
> > > > On Friday 12 March 2004 20:42, Kenneth Aafløy wrote:
> > > > > Hi,
> > > > >
> > > > > i've discovered that in cvs dvb-kernel above 20040224 tuning with
> > > > > mythtv is broken, but szap still works. The only difference between
> > > > > the two I can see is that szap uses FE_READ_STATUS, and mythtv uses
> > > > > FE_GET_EVENT to read the frontend status. Now, mythtv does indeed
> > > > > receive those events, but never with a status of |LOCK. What am i
> > > > > missing here? Is this a frontend bug, or application bug?
> > > > >
> > > > > Hauppauge Nexus-S (STV0299: BSRU6, TDQB-S00x)
> > > >
> > > > Odd; dvbtune (which uses the event interface) still works OK..... I
> > > > wonder what mythtv is doing differently.
> > >
> > > Doing some reading of the diff between 20040224 and 20040225 i noticed
> > > some changes to accomodate for FE_CAN_INVERSION_AUTO. Now, all my
> > > channels had been pre-set with INVERSION_OFF (because the mythtv dvb
> > > stuff is kind enough to warn about unsupported features of the
> > > frontend), and when I change this to INVERSION_AUTO all works well. So
> > > i guess it's a frontend bug then..
> >
> > Aha! Its a actually a bugfix which causes this.
> >
> > Previously the stv0299 frontend driver was always doing a nasty hack for
> > inversion which I removed, because it broke tuning to low symbol rate
> > channels.
>
> And reinstated for 1.1.1, as in:
> http://linuxtv.org/cgi-bin/cvsweb.cgi/dvb-kernel/linux/drivers/media/dvb/fr
>ontends/stv0299.c?only_with_tag=LINUXTV-DVB-1_1_1
>
> So, it's intentionally broken, or just a flop?

It was reinstated because 1.1.1 didn't have auto-inversion in the driver core 
for frontends which do not support it. It does now.

> > Unfortunately the side effect is that if the inversion is set
> > wrongly from the client side, it will not work.
>
> Or are you actually arguing that my inversion setting is wrong to begin
> with?

Well, in all my tests, I've tried INVERSION_OFF, INVERSION_ON, and 
INVERSION_AUTO. INVERSION_OFF/INVERSION_ON are passed directly through to the 
driver, and work as expected for me. INVERSION_AUTO works fine as well. And 
I've had no other complaints about inversion.

Previously the code flipped the inversion on every tune step attempt, so you 
could have specified absolutely anything, and it would have worked on high 
symbol rate channels. This, as I said, broke low symbol rate channels (for me 
at least) as they're far more sensitive.

> One other point, you seem to ignore the already existing coding styles when
> tabs is evaluated, please correct, as the code becomes rather ugly.

Does it? It looks fine to me... which editor are you using?


-- 
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe linux-dvb" as subject.