Re: Fixing bug in Harris/Hessian examples

From: Mateusz Loskot (mateusz_at_[hidden])
Date: 2019-09-13 16:17:56

On Fri, 13 Sep 2019 at 07:07, Olzhas Zhumabek
<anonymous.from.applecity_at_[hidden]> wrote:
> >
> > This is a bit unclear to me.
> > Where is the dependency issue?
>
>
>  Tests of gradient computation will rely on numeric if kernel_2d is used as
> kernel type. And IIRC numeric relies on core passing, does it not? If not,
> I guess there is no issue, just a bit of weirdness.

Yes, currently relationship between the numeric extension and the core
is a bit weird, or some may say annoying or incorrect.

Would merge of the numeric into core solve the problem for you?

I'll try to nudge Stefan about this thread.

> If there is a functionality missing from our kernel types,
> > perhaps we should improve it there, instead of using image_view.
>
> Without the dependency question, there is no issue. In fact, I could move
> the existing functionality that uses image_view to kernel_2d, but then I
> will break Miral's code that relies on generation of those kernels.

If Miral is not around, I will try to fix any breakage myself.
So, don't worry about sticking to your task.

Best regards,
-- 
Mateusz Loskot, http://mateusz.loskot.net