[sane-devel] imageClass bug

[Alexander Pevzner]

Hi Ralph,

On 11/14/20 8:54 PM, Ralph Little wrote:
> I will add a model flag to the machine definition to activate the
> behaviour for the problematic machines.
> Thankfully, the way that I have changed the pixma geometry calculations
> makes this very straighforward.

I finally came to approach when for WSD scanners I always clip image in 
software, because I know at least one model (my own Kyocera) where image 
clipping in firmware is totally broken, and I cannot test them all.

For eSCL I rely on firmware, because nobody complained so far. But 
firmware clips inaccurate (typically with +/- 1 pixel error), and 
because my sane_start() returns early, before actual image is available, 
while sane_get_parameters() expects accurate response, I have to adjust 
actual image to the size I've promised earlier.

I want to tell, it may make sense for you to use software clipping for 
any model, not only for known to be problematic.

-- 

	Wishes, Alexander Pevzner (pzz at apevzner.com)