GIMP Version: 2.8.10
Operating System: Windows
GIMP Experience: Beginner Level

---

Hello, I'm trying to apply this filter and after running it I get a odd outcome. There are wide stripes crossing the picture. I don't know if it's only a flaw on my side or if someone has the same problem. It only happens in this filter not the other ones.
I add screens so that you can see the problem. Thank you so much in advance.

I can't replicate your issue. The fact that your result doesn't even remotely mirror the preview is very strange. Although at first I thought it could be a system graphics issue, I don't think that's it. Was the original edited beforehand (because it looks like a mask)? Have you tried a different image with the same settings?

This looks like another multi-processing problem ..I'm running on a 2-core laptop and I only get half the image processed.
(I presume Issabella is running on a 4-core machine.)

Furthermore, it's NOT a regression in the latest GMIC (1.6.1.0 beta) ..the same thing happens with GMIC 1.6.0.3.

"Channel" modes "All" , "Blue/red chrominances", "ab-components" and "Hue" seem okay, most of the rest fail.

I think Ronounours should see this.

Yippie.

I'm seeing it. Just happen to have a test dog left over from another post.



Same settings as Issabella's.

Like jonait2's theory.... Into the hunt.

Garry

Update:
Tried it again on my laptop with multiprocessing disabled (at start-up) and Spotify works okay in all Channel modes.
So it does look very much like a multi-processing problem.

Interesting.
How is it possible to disable multiprocessing using the G'MIC plug-in for Gimp?
Is there an option to enable/disable it on the plug-in?

Thank you all for your interest on this problem. The odd thing is that it happens only applying Spotify. Grosood, your outcome is just as mine.
Odin, the pic hasn't been edited before, it's a original one. And Yes I tested spotify on other different ones and I get the same outcome as the one I dropped here.

Please, if you have any idea to solve it, let me know.

Thanks Isabella, very interesting indeed.

The problem is that this filter apparently doesn't use any explicit parallel processing. It does in fact run in parallel, but using OpenMP built-in functions, not explicit threading. This might sound as a detail, but to me, this means that the bug may come from the OpenMP pieces of code of G'MIC instead of the 'spotify' command itself. In brief, this is probably not easy to fix with a simple filter update
So far, on Linux, I haven't been able to reproduce the bug. I've used the same parameters as you, the same kind of image dimensions, but without success.
So, maybe the bug is happening also only on Windows ?

Any additional information is welcome, to better understand the problem.
Thanks for your report !

More infos :
I've been able to reproduce the bug on Windows too. Looks really like a problem with the way OpenMP parallelizes the loops in the G'MIC code
What I've noticed is that if you get the bug, then undo the action (CTRL+Z), then run the filter again (CTRL+F), sometimes it will work.
On my virtualized machine, I'd say it works 2/3 of the time, and I get the bug 1/3 of the time.
That is bad news because :
- It probably also affects other filters.
- I don't know how to solve this, apart from removing OpenMP support when compiling G'MIC on Windows (which would slow down almost all available filters).

Maybe there are some compiling options I need to give specifically to get rid of this OpenMP problem, I'll check.
But as far as I understand the bug, it looks like another Windows-specific problem to deal with

Thank you so much Ronounours, I really appreciate your interest in it. I'm sure you will study it and perhaps you find a solution.

Sorry, Ronounours, I'm running on Ubuntu 14.04 LTS at the moment ..so its not Windows-specific ..happens on Linux too!

Hello David,

Yep. Ditto here. The Defective Dog was rendered on a Linux box, running Gentoo. Currently the G'MIC 1.6.1.0 beta.

Linux harriet 3.17.8-gentoo-r1 #1 SMP Sat Feb 14 05:04:14 EST 2015 x86_64 Intel(R) Pentium(R) 4 CPU 2.80GHz GenuineIntel

I do git pulls against the repository. This is the most recent log entry:
commit 16b8bd119d5fae8893196b5a6158647aa2e1d251
Author: ronounours <[email protected]>
Date: Wed Feb 25 10:31:59 2015 +0100

This may be a Hail Mary pass, but the filter still uses the old -apply_channels. Perhaps migrating it to the new? Won't help those < 1.6.1.0, but might narrow the search field.

I have some time to poke on this today. Will report anything interesting here and to the mailing list. Right now, must walk The Little Girl (a corgi. Read the G'MIC tutorials for full documentation on her. She's in there. )

Garry
EDIT:
Build with standard compiler flags, just as the makefile comes from the repository: "make all"

Ha thanks, that helps to know the problem also occurs on Linux too.
Maybe I was not able to reproduce it (at home) because I've only 4 processors ?
I'll try again at work, and see what happens.

OK, good news everyone,
Jérome Boulanger (the author of the 'Spotify' filter) have figured out what was causing the bug, and found a workaround to fix it.
This means you can now just refresh your filters and the bug should not be here anymore.
For me, it also means that there is a real bug to fix in my G'MIC code (there is still a bug, even if the work-around makes it invisible).

So thanks Issabella for the report ! I know there is still something to improve in the G'MIC code

Thanks Ronounours.

Thank you Ronounours. Of course, should the day come when there is nothing to improve in G'MIC, you'd be devastated. You'd have to take up some other line of work like sword swallowing or walking tightropes over Niagra Falls or something.

David after refreshing i get 408 filters (was 410). Did you move 2 of them elsewhere?

@Rod : I've actually removed the filter 'Do nothing' + one I moved to Testing (I don't remember which).

@Garry : On the contrary, I think that the day I'll feel that G'MIC is a finished project, I won't be stressed anymore. I try to converge to this state as fast as I can.

Anyway, I've found the bug (was in my CImg library), and fixed it. Thanks again for all your testing and reports !
I'll post some new updated G'MIC binaries today with the fix included.

Thank you so much David. I'm so happy to have Spotify filter working so good.

I have been able to check the filter Spotify and it works great! Thanks David.