Re: bttv, tvaudio and ir-kbd-i2c probing conflict

[Trent Piepho <xyzzy@speakeasy.org>]

On Mon, 16 Mar 2009, Jean Delvare wrote:
> Come on, just look at ir-kbd-i2c and tvaudio again, see how great are
> these drivers which have been "designed" on top of the legacy i2c
> binding model. Look at the bttv mess. Look at the zoran driver
> conversion done by Hans a few weeks ago, which killed what, 3000 lines
> of code? The old binding model was so bad that DVB doesn't even use it.

IIRC, the zoran patch removed more like 1000 lines.  But it also deleted
v4l1 support, highmem support, and bigphys_area support.  Maybe other
things, Hans doesn't decribe his patches, so there's really no way to know
what the zoran patch really did other than to weed through 10,000+ lines of
diff which is mostly but not entirely moving blocks of code from one space
to another and reindenting them.

If one includes the v4l1-compat module that is now providing v4l1 support
(though it doesn't work correctly for zoran), the driver and the compat
module are larger than the old driver was.  Of course one can now turn off
v4l1 support and get a smaller driver than before.  And the v4l1 compat
already existed and can be shared.  But I think it's more correct to say
the size reduction of the zoran driver was from removing features and not
from v4l2_subdev.  It seems like more of the the other subdev conversions
have overall added more code than they removed.