Crash when selecting input source in a mask layer

[phweyland]

Wanted to reuse the mask of the previous local contrast, invert it and set up a new local contrast

When I tick show mask, no mask appears. When I untick show mask pf crashes.

I've run gdb, the behaviour is a bit different. Doesn't crash on untick show mask, but does when untick invert mask.

PhotoFlow-01.txt 20181105_Agde_069.jpg.pfi.txt

using today pf build. Let me know if you need anything else.

[phweyland]

This one is crashing when I open the pfi without gdb.

With gdb the file opens, but pf crashes when I uncheck show mask.

PhotoFlow-02.txt

20181105_Agde_069.jpg.pfi.txt

[aferrero2707]

There must be still a bug in the input source selection...

I will try to reproduce the crash. Meanwhile, could you please try inserting an explicit layer clone below the "invert" layer, and see if it works as expected?

Thanks!

[phweyland]

I've added the clone layer, set the "mask sky" as input source, and set on "mask sea" the previous layer as input source. When I check show mask nothing appears. But crashes when uncheck (I think it crashed on the second time I unchecked ....)

PhotoFlow-03.txt

[aferrero2707]

I could reproduce the problem with your sample .pfi, and I'm investigating...

I was however able to modify the .pfi with my suggestion, and make it behave as expected. To fix the initially broken stack, I did the following:

• opened the .pfi
• hidden the topmost "local contrast" layer
• while hidden, I went into the corresponding mask, and added a clone layer below the "Mask Sea" layer. The clone layer is cloning the "Mask Sky (blended)"
• set the input mode of the "Mask Sea" layer to "process previous layer"
• went back to the main layer stack, and showed the topmost "local contrast" layer

Now if you go back into the mask, it should be possible to toggle the mask visibility and see the correct result... Here is my final arrangement for the mask:

[phweyland]

Great ! with clone layer that works ! :-) Thank you However I had to delete the topmost contrast layer and build it following your method. Maybe there was something wrong in my pfi. When I opened it the clone layer did appear on top. I've moved it on the bottom, but each time I leave and come back to mask view, the clone layer was back on top... like this: Thanks !

[aferrero2707]

I'm currently fixing the problem with the input source selection for mask layers... hopefully your initial idea will be working properly in a short while. Will let you know here before closing the issue.

[phweyland]

Thanks so much. I don't know if that's related but opening this pfi is endless (seems to process several times). In any case it's much faster to disable the group and each component, then enable the group and components one by one (I can share the raw if needed). 20181105_Agde_069.jpg.pfi.txt

EDIT: no, probably not related ... if I disable the bilateral blurr it's much faster.

[aferrero2707]

@phweyland bilateral blur is definitely not optimized, and in fact I am replacing it with the "guided filter" from RawTherapee. According to the literature, it provides better transitions between smooth areas and sharp edges, and moreover is is significantly faster...

The new "local contrast" tool is actually based on the guided filter.

[aferrero2707]

@phweyland I have just committed a fix of your initial issue (crash when selecting an input source for the invert tool). Pre-compiled packages should be available in less than one hour... could you please check if the new code works correctly for you as well?

Thanks!

[phweyland]

Yes, I can now set a previous mask in "invert layer" different than a "clone layer" without trouble when I enable/disable show mask. Thanks!

[aferrero2707]

Closing issue, fixed with 94b681ba0cc4afa26a56eb57fc3b8bbed475456f.