Re: [linux-dvb] Re: [video4linux-cvs] [hg:v4l-dvb] Add support for Opera S1- DVB-USB

[Manu Abraham <abraham.manu@gmail.com>]

hermann pitton wrote:
> Am Freitag, den 20.04.2007, 03:19 +0400 schrieb Manu Abraham:
>> hermann pitton wrote:
>>> Am Freitag, den 20.04.2007, 02:51 +0400 schrieb Manu Abraham:
>>>> Markus Rechberger wrote:
>>>>> On 4/20/07, Manu Abraham <abraham.manu@gmail.com> wrote:
>>>>>> hermann pitton wrote:
>>>>>>> Am Freitag, den 20.04.2007, 00:55 +0400 schrieb Manu Abraham:
>>>>>>>> Mauro Carvalho Chehab wrote:
>>>>>>>>> Em Qui, 2007-04-19 às 16:41 -0400, Michael Krufky escreveu:
>>>>>>>>>> Marco Gittler wrote:
>>>>>>>>>>> this patch has applied the hints from mkrufky (dvb_attach,
>>>>>>>>>>> firmware-naming)
>>>>>>>>>>> and also one working rewrite of the i2c addresses stuff to fit the
>>>>>>>>>>> kernel i2c reqs.
>>>>>>>>>>>
>>>>>>>>>>> Signed-off-by: Marco Gittler<g.marco@freenet.de>
>>>>>>>>>>> diff -r c8b73ec18b42 linux/drivers/media/dvb/dvb-usb/opera1.c
>>>>>>>>>>> --- a/linux/drivers/media/dvb/dvb-usb/opera1.c    Thu Apr 19
>>>>>> 12:04:50
>>>>>> 2007 -0300
>>>>>>>>>>> +++ b/linux/drivers/media/dvb/dvb-usb/opera1.c    Thu Apr 19
>>>>>> 20:38:01
>>>>>> 2007 +0200
>>>>>>>>>>> @@ -25,6 +25,13 @@
>>>>>>>>>>>  #define REG_20_SYMBOLRATE_BYTE1 0x20
>>>>>>>>>>>  #define REG_21_SYMBOLRATE_BYTE2 0x21
>>>>>>>>>>>
>>>>>>>>>>> +#define ADDR_C0_TUNER (0xc0>>1)
>>>>>>>>>>> +#define ADDR_D0_PLL (0xd0>>1)
>>>>>>>>>>>
>>>>>>>>>> I don't like these two #define's.  These i2c addresses need only be
>>>>>>>>>> specified once, in the config structs / frontendfoo_attach calls for
>>>>>> the
>>>>>>>>>> tuner / demod.
>>>>>>>>>>
>>>>>>>>>> Better to just put them in as constants like all of the other dvb
>>>>>> drivers.
>>>>>>>>> I prefer the way it is. We should really avoid having magic numbers
>>>>>>>>> inside the code. The alias here helps to know that 0x60 is tuner
>>>>>> addres
>>>>>>>>> and 0x68 the pll.
>>>>>>>> Following a project's coding styles and conventions is "respecting" a
>>>>>>>> project
>>>>>>>>
>>>>>>>> Manu
>>>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> the other natural place for this should be the LKML to get more _good_
>>>>>>> arguments, instead of hanging soon in some "respect" stuff again.
>>>>>> DVB drivers generally have device addresses such as tuner_addresses and
>>>>>> demod_adresses defined in a config struct least to prevent them from
>>>>>> being global, wherever the header is included, since the very same
>>>>>> device can have multiple addresses and so on, which are non-probable
>>>>>> since being behind a repeater which is switched by a demod (private) and
>>>>>> hence.
>>>>>>
>>>>>> Those are some of the reasons to follow a certain coding
>>>>>> style/conventions. They are _not_ for fun.
>>>>>>
>>>>> cat *priv.h says something else too...
>>>>> there are also many global register defines in DVB drivers, they just
>>>>> don't include the register value in the define name.
>>>> *_priv.h from what i understand means private .. i don't know what you
>>>> make out from that.
>>>>
>>>>
>>>> HTH,
>>>> Manu
>>> ;)
>>>
>>> That means that I had to post the actual headers to every single tester
>> If you use a private header as a public header, of course yes. But that
>> is not what private explicitly means.
>> It _is_ indeed wrong to use a private header as a public header _even_
>> for workarounds.
>>
>> HTH,
>> Manu
> 
> Forget it.
> 
> That is as wrong as older Fedora distros were shipping v4l2 apps like
> tvtime, but only providing v4l1 headers on the user level.
> 

I don't know about Fedora shipping v4l2 apps.  Forgive my ignorance. But
it is really hopeless to include a private header for a device into
another device. Anyway not talking about V4L1/2/n headers, but about DVB
device (demod/tuner) private headers being included publicly. Private
means private, i don't understand how the notion comes around that a
private header is a public header.

It is _not_ named private for no reason.


Manu