kqueue(2) vs. cuse4bsd?

[Bernhard Fröhlich]

Am 18.07.2013 21:39 schrieb "Hans Petter Selasky" <hps at bitfrost.no>:
>
> On 07/18/13 20:34, Juergen Lock wrote:
>>
>> In article <
CAE-m3X0coVNak4fZ9hfC82izfgB1u0tX_JSBz89wb1d_qLhKtQ at mail.gmail.com> you
write:
>>>
>>> Hi guys.
>>
>> Hi!
>>>
>>>
>>> I have been working on tvheadend for some time and got it to the point
>>> where I think it starts to become useful for others too. Upstream was
>>> very helpful and merged a lot of patches already and also helped with
>>> the kqueue support.
>>>
>>> So what is tvheadend and why do you want it?
>>> - If you are using XBMC then you really want tvheadend as your DVB
>>> backend. This allows you to watch TV and record stuff within XBMC.
>>> - If you are using MythTV then you really want to use tvheadend because
>>> it's slim and light and still does everything you ever need. Without
all the
>>> Qt GUI and without a database and without a ton of plugins.
>>>
>>> The port itself is in a somewhat useable state but there is still quite
some
>>> work to bring it in a good shape. I still wanted to bring it to the
>>> tree to allow
>>> other people to jump in and give it a broader testing with other
devices and
>>> other environments.
>>>
>>> Known issues:
>>> - scanning does not work (at least not for my Dib7000 DVB-T device)
>>> - no libav support
>>> - no libdvbcsa support
>>> - no rc.d script
>>> - does not compile on FreeBSD 8.x
>>>
>>> Patches, feedback and help are always welcome!
>>>
>>> WWW: https://tvheadend.org/
>>
>>
>>   Ok I got it to build on 8.3 and found it couldn't scan because
>> kevent() in tvhpoll_add() in open_table() in src/dvb/dvb_input_filtered.c
>> always fails with ENODEV on the demux0 dvb node. :(  Is kqueue(2)
>> broken with cuse4bsd?  I've Cc'd hps, maybe he knows...
>>
>>   First the Makefile patch I used:

You're awesome! Thanks a lot for stepping into and debugging that.

>
> kevent() support is not implemented for cuse4bsd. The devfs callback is
NULL. Possibly could add that, but have to read up on kqueue first.

Oh what a bummer. It would definitely be good to add that callback but is
this everything that is needed or should we expect more problems?