Le Jeudi 03 Novembre 2005 10:44, Manu Abraham a ?crit?: > Patrick Boettcher wrote: > > As Ralph pointed out, there are applications out there, which are not > > working with dvrX and demuxX (X >= 1). So in that case it's maybe > > better to register a dvb_adapter per stream-input to be more > > compatible with applications. > > It would be better to fix the broken applications, than to fix the > applications that are broken in software. IMHO, this is a very bad > workaround. I agree. For example, it would take about 10 minutes to fix Kaffeine. So, just do it the more convenient/consistent way, and apps will follow. > I remember somebody suggesting me to incorporate for a bug in an > application's CI stack for CA_PMT length in the driver since the > application was doing it wrong. > > IMHO, driver/kernelspace is not the place to fix userspace application > bugs/design. > > I would say that the application needs to be fixed in that case , > inorder to use the proper frontend/dvr/demux. I had been working a bit > on a dual DVB-S card on the same approach. (But still struggling a bit > on the tuner part a bit for now, since the same approach is used on both > card types). ie, > > /dev/dvb/adapterX > /dev/dvb/adapterX/frontend0 > /dev/dvb/adapterX/demux0 > /dev/dvb/adapterX/ca0 > /dev/dvb/adapterX/dvr0 > > /dev/dvb/adapterX/frontend1 > /dev/dvb/adapterX/demux1 > /dev/dvb/adapterX/ca1 > /dev/dvb/adapterX/dvr1 AFAIU, an application could stop probing as soon as there is no X+1 (as for frontendN+1). Can you confirm ? -- Christophe Thommeret
