[linux-dvb] Re: AVerMedia 771 problems [everything solved]

[Nathan Hand <nathanh@manu.com.au>]

My preference would be that the driver leaves the values alone. There is
nothing more infuriating than telling the driver FEC_1_2 and finding 
out later the driver thinks it's smarter than you and has been silently
"correcting" the values to FEC_3_4.

In those situations, how do you tell the driver "no, I really do want
FEC_1_2, do what I tell you"? If the driver is ignoring your values, in
the hope of being "useful", then what do you do when the driver gets it
wrong? Recompile the kernel? Annoying.

Do those sort of things in user space.


On Mon, 2004-06-28 at 20:18 +0200, Johannes Stezenbach wrote:
> Wolfram Joost wrote:
> > 
> > > This also works for me and the sleep problem is gone. But I still need
> > > other parts of the diff I sent here before. Patch attached.
> > 
> > I think the parameter validation-code makes the difference. Currently, the 
> > driver in the cvs rejects wrong parameters.
> > 
> > What's the philosophy for the dvb-subsystem? Tolerate and auto-correct wrong 
> > parameters or silently correct them(and live with applications which use 
> > wrong parameters)? 
> 
> I think that parameters which are outside the DVB specification
> should be rejected (EINVAL). However, a driver might want to
> handle gracefully slightly wrong tuning parameters (e.g. a wrong FEC
> value) -- the application should be able to get correct values
> via FE_GET_FRONTEND if it is interested.
> 
> Johannes
> 
> 
>