On Mon, Nov 27, 2017 at 8:44 PM, Robert Bieber <[email protected]> wrote:
> Whoops, I forgot about git annex.  Let's try this again:
> https://drive.google.com/open?id=14hjQuEqpvSEC2EDI0Fqv_Vo4qEr-Hnd8
Yep, that downloaded.
However i don't know what i need to look for, so i don't see anything.

> I also skimmed through the dcraw source a bit, and I found a function called
> phase_one_correct, which I suspect addresses this issue.  I'm way out of my
> depth trying to interpret it, but I _think_ what it's doing is applying some
> correction factor(s) embedded in the file for the different separate
> sections of the sensor.
Yep, sounds plausible.
"RobbieAB" in #darktable was looking into that, and had some success,
but i did not hear back from him for some time.

Roman.

> On 11/27/2017 11:51 AM, Roman Lebedev wrote:
>>
>> Hi.
>>
>> On Mon, Nov 27, 2017 at 1:22 AM, Robert Bieber <[email protected]>
>> wrote:
>>>
>>> I got my hands on a Leaf Credo 40 digital back a few weeks ago, and so
>>> far
>>> it's mostly worked stunningly with Darktable.  I have noticed, however,
>>> that
>>> in some images, particularly those with bright skies in the background,
>>> I'm
>>> getting a noticeable centerfold effect between the left and right half of
>>> the sensor. Here's an example of a RAW file and XMP file that produce the
>>> effect in both the build of Darktable I've been using
>>> (2.3.0+455~g55904ce36)
>>> as well as the current release version 2.2.5:
>>> https://drive.google.com/open?id=1LfP5rGfFkBBPuolZ1NzivgnbhUNX0RI0
>>
>> $ unzip centerfold_raw.zip
>> Archive:  centerfold_raw.zip
>>   inflating: 0047_23_46.IIQ.xmp
>>     linking: 0047_23_46.IIQ          ->
>>
>> ../../../.git/annex/objects/03/Gw/SHA256E-s40513597--1ef92618cc8c9c90a6c82762be3e8c66023ee8013db9551b64760d41c53fc273.IIQ/SHA256E-s40513597--1ef92618cc8c9c90a6c82762be3e8c66023ee8013db9551b64760d41c53fc273.IIQ
>> finishing deferred symbolic links:
>>   0047_23_46.IIQ
>>
>> Somehow i don't think that is what you wanted to upload.
>>
>>> At first I assumed this was just an artifact of the sensor and I'd simply
>>> have to avoid those situations, but I tried processing the same RAW file
>>> in
>>> UFRaw and couldn't get the centerfold to appear, no matter what
>>> contortions
>>> I put the image through.  So I guess this is actually an artifact of the
>>> RAW
>>> processing, presumably a difference between RawSpeed and DCRaw.
>>
>> I'll defer from any speculations until i actually see the sample.
>>
>>> Does anyone have an idea what might be responsible for the difference,
>>> and
>>> if it's something that might be easily fixable in RawSpeed, or
>>> alternatively
>>> if there's some way I can force DT to use dcraw as a fallback for a given
>>> image instead of RawSpeed?
>>
>> dcraw was never ever supported/integrated.
>> libraw fallback was disabled for quite some time and dropped some time ago
>> too.
>> So nope. RawSpeed is the only raw loader.
>>
>>> Or maybe I'm just missing something and it's my
>>> own image settings that are responsible for the seam showing up.
>>>
>>> I also wanted to test the latest trunk build of DT to see if this had
>>> been
>>> fixed since then, but the latest version actually seems to be missing
>>> support for the Credo 40 (trying to click through to darkroom mode on one
>>> of
>>> my images got me the "Couldn't read white balance information" message
>>> that
>>> you always get with unsupported cameras).  If I get a chance in the next
>>> couple days I'll try to use git bisect to figure out which revision broke
>>> support for it.
>>
>> On Mon, Nov 27, 2017 at 1:58 AM, Robert Bieber <[email protected]>
>> wrote:
>>>
>>> I just uploaded a shot of a color chart a little while ago, not sure how
>>> long it'll take them to add it to the archive
>>
>> Usually the time it takes me to notice the new upload, so between 10
>> minutes and a day :)
>> Validated, thanks!
>>
>> On Mon, Nov 27, 2017 at 4:41 AM, Robert Bieber <[email protected]>
>> wrote:
>>>
>>> Actually, scratch that, I somehow screwed up the first bisect.  I went
>>> back
>>> and took another crack at it and found that the offending revision is
>>> actually c8d47ad9c466e6469f518f6826f18b1d9941c65e, which makes sense
>>> because
>>> it updates the Rawspeed version.  Then I went back and tested different
>>> revisions in the RawSpeed repo with that same DT rev, and it looks like
>>> the
>>> revision that broke Credo 40 reading is
>>> 65cc3c5e0ccce9bc87c3e80703c7c55e8466c587.  I'm not sure exactly how,
>>> because
>>> all it seems to have done is split the IIQ decoder out into a separate
>>> module, but the one before that is the last one that will read my RAW
>>> files.
>>> I'll add a bug to the RawSpeed tracker for that one
>>
>> Thanks. Should be fixed now.
>>
>> This is why we have RPU :)
>> See https://discuss.pixls.us/t/raw-samples-wanted/5420?u=lebedevri
>>
>> Roman.
>>
>>>
>>> ____________________________________________________________________________
>>> darktable user mailing list
>>> to unsubscribe send a mail to
>>> [email protected]
>>>
>
> ____________________________________________________________________________
> darktable user mailing list
> to unsubscribe send a mail to [email protected]
>
____________________________________________________________________________
darktable user mailing list
to unsubscribe send a mail to [email protected]

Reply via email to