Re: [darktable-dev] Gaussian Display Error

[Roman Lebedev]

On Sun, Oct 30, 2016 at 3:28 PM, David Vincent-Jones  wrote:
> Yes, this is an x-trans raw file 
I'm mostly asking, can anyone reproduce this with bayer raw?

> and yes NaNs are in console output as
> below:
>
> [dev_pixelpipe] module `lowpass' outputs NaNs! [preview]
> [dev_pixelpipe] module `lowpass' min: (0.00; -128.00; -128.00)
> max: (100.00; 128.00; 128.00) [preview]
> [dev_pixelpipe] module `output color profile' outputs NaNs! [preview]
> [dev_pixelpipe] module `output color profile' min: (0.00; 0.00;
> 0.00) max: (1.329883; 1.130938; 1.493263) [preview]
> [dev_pixelpipe] module `velvia' outputs NaNs! [preview]
> [dev_pixelpipe] module `velvia' min: (0.00; 0.00; 0.00) max:
> (1.00; 1.00; 1.00) [preview][dev_pixelpipe] module `lowpass'
> outputs NaNs! [preview]
> [dev_pixelpipe] module `lowpass' min: (0.00; -128.00; -128.00)
> max: (100.00; 128.00; 128.00) [preview]
> [dev_pixelpipe] module `output color profile' outputs NaNs! [preview]
> [dev_pixelpipe] module `output color profile' min: (0.00; 0.00;
> 0.00) max: (1.329883; 1.130938; 1.493263) [preview]
> [dev_pixelpipe] module `velvia' outputs NaNs! [preview]
> [dev_pixelpipe] module `velvia' min: (0.00; 0.00; 0.00) max:
> (1.00; 1.00; 1.00) [preview]
>
>
>
> On 10/30/2016 01:53 AM, Roman Lebedev wrote:
>>
>> Does all of you expirience the problem with X-Trans sensor?
>> Or with Bayer too?
>>
>> For x-trans: https://redmine.darktable.org/issues/11272
>> For bayer need reproducer.
>> Hint: you can run darktable as  `darktable -d nan`  and look for
>> "Module xyz outputs NaNs!" in the console output."
>>
>> Roman.
>>
>> On Sun, Oct 30, 2016 at 11:21 AM, David Vincent-Jones
>>  wrote:
>>>
>>> I opened dt (with gdb) using an image that showed the problem yesterday
>>> and
>>> it was still apparent at start up ... I switched several times between
>>> gaussian and bilateral and the problem was clearly indicated; the 2
>>> locations blacked-out and reapeared with the change.
>>>
>>> Nothing unusual in the gdb listing.
>>>
>>> Tried to adjust memory settings (8 gb on this machine) ... no change.
>>>
>>> 2.1.pre+git2000.f63e5ce-1.1 Opensuse 42.1
>>>
>>>
>>> On 10/29/2016 10:59 AM, Patrick Shanahan wrote:

 * David Vincent-Jones  [10-29-16 12:33]:
>
> I find that if gaussian curve is used either with shad/high or lowpass
> the
> display of the histogram is lost also the top-left overall image.
>
> I first noticed this with shad/high and found the situation was
> corrected
> when I switched to bilateral filter.

 seems also to happen with perspective correction on 19c3415, but not all
 the time, and is still active/usable to adjust exposure.  it also will
 reappear when accessing some other modules.

>>>
>>> ___
>>> darktable developer mailing list
>>> to unsubscribe send a mail to
>>> darktable-dev+unsubscr...@lists.darktable.org
>>>
>>
>> ___
>> darktable developer mailing list
>> to unsubscribe send a mail to
>> darktable-dev+unsubscr...@lists.darktable.org
>>
>
> ___
> darktable developer mailing list
> to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org
>
___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org

Re: [darktable-dev] Gaussian Display Error

[David Vincent-Jones]

Yes, this is an x-trans raw file  and yes NaNs are in console output 
as below:


[dev_pixelpipe] module `lowpass' outputs NaNs! [preview]
[dev_pixelpipe] module `lowpass' min: (0.00; -128.00; 
-128.00) max: (100.00; 128.00; 128.00) [preview]

[dev_pixelpipe] module `output color profile' outputs NaNs! [preview]
[dev_pixelpipe] module `output color profile' min: (0.00; 0.00; 
0.00) max: (1.329883; 1.130938; 1.493263) [preview]

[dev_pixelpipe] module `velvia' outputs NaNs! [preview]
[dev_pixelpipe] module `velvia' min: (0.00; 0.00; 0.00) max: 
(1.00; 1.00; 1.00) [preview][dev_pixelpipe] module `lowpass' 
outputs NaNs! [preview]
[dev_pixelpipe] module `lowpass' min: (0.00; -128.00; 
-128.00) max: (100.00; 128.00; 128.00) [preview]

[dev_pixelpipe] module `output color profile' outputs NaNs! [preview]
[dev_pixelpipe] module `output color profile' min: (0.00; 0.00; 
0.00) max: (1.329883; 1.130938; 1.493263) [preview]

[dev_pixelpipe] module `velvia' outputs NaNs! [preview]
[dev_pixelpipe] module `velvia' min: (0.00; 0.00; 0.00) max: 
(1.00; 1.00; 1.00) [preview]



On 10/30/2016 01:53 AM, Roman Lebedev wrote:

Does all of you expirience the problem with X-Trans sensor?
Or with Bayer too?

For x-trans: https://redmine.darktable.org/issues/11272
For bayer need reproducer.
Hint: you can run darktable as  `darktable -d nan`  and look for
"Module xyz outputs NaNs!" in the console output."

Roman.

On Sun, Oct 30, 2016 at 11:21 AM, David Vincent-Jones
 wrote:

I opened dt (with gdb) using an image that showed the problem yesterday and
it was still apparent at start up ... I switched several times between
gaussian and bilateral and the problem was clearly indicated; the 2
locations blacked-out and reapeared with the change.

Nothing unusual in the gdb listing.

Tried to adjust memory settings (8 gb on this machine) ... no change.

2.1.pre+git2000.f63e5ce-1.1 Opensuse 42.1


On 10/29/2016 10:59 AM, Patrick Shanahan wrote:

* David Vincent-Jones  [10-29-16 12:33]:

I find that if gaussian curve is used either with shad/high or lowpass
the
display of the histogram is lost also the top-left overall image.

I first noticed this with shad/high and found the situation was corrected
when I switched to bilateral filter.

seems also to happen with perspective correction on 19c3415, but not all
the time, and is still active/usable to adjust exposure.  it also will
reappear when accessing some other modules.


___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org


___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org



___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org

Re: [darktable-dev] Gaussian Display Error

[Roman Lebedev]

Does all of you expirience the problem with X-Trans sensor?
Or with Bayer too?

For x-trans: https://redmine.darktable.org/issues/11272
For bayer need reproducer.
Hint: you can run darktable as  `darktable -d nan`  and look for
"Module xyz outputs NaNs!" in the console output."

Roman.

On Sun, Oct 30, 2016 at 11:21 AM, David Vincent-Jones
 wrote:
> I opened dt (with gdb) using an image that showed the problem yesterday and
> it was still apparent at start up ... I switched several times between
> gaussian and bilateral and the problem was clearly indicated; the 2
> locations blacked-out and reapeared with the change.
>
> Nothing unusual in the gdb listing.
>
> Tried to adjust memory settings (8 gb on this machine) ... no change.
>
> 2.1.pre+git2000.f63e5ce-1.1 Opensuse 42.1
>
>
> On 10/29/2016 10:59 AM, Patrick Shanahan wrote:
>>
>> * David Vincent-Jones  [10-29-16 12:33]:
>>>
>>> I find that if gaussian curve is used either with shad/high or lowpass
>>> the
>>> display of the histogram is lost also the top-left overall image.
>>>
>>> I first noticed this with shad/high and found the situation was corrected
>>> when I switched to bilateral filter.
>>
>> seems also to happen with perspective correction on 19c3415, but not all
>> the time, and is still active/usable to adjust exposure.  it also will
>> reappear when accessing some other modules.
>>
>
> ___
> darktable developer mailing list
> to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org
>
___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org

Re: [darktable-dev] Gaussian Display Error

[David Vincent-Jones]

I opened dt (with gdb) using an image that showed the problem yesterday 
and it was still apparent at start up ... I switched several times 
between gaussian and bilateral and the problem was clearly indicated; 
the 2 locations blacked-out and reapeared with the change.


Nothing unusual in the gdb listing.

Tried to adjust memory settings (8 gb on this machine) ... no change.

2.1.pre+git2000.f63e5ce-1.1 Opensuse 42.1


On 10/29/2016 10:59 AM, Patrick Shanahan wrote:

* David Vincent-Jones  [10-29-16 12:33]:

I find that if gaussian curve is used either with shad/high or lowpass the
display of the histogram is lost also the top-left overall image.

I first noticed this with shad/high and found the situation was corrected
when I switched to bilateral filter.

seems also to happen with perspective correction on 19c3415, but not all
the time, and is still active/usable to adjust exposure.  it also will
reappear when accessing some other modules.



___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org