Hello David,

1) Just downloaded the new version of: gmic_gimp_1.6.1.0_beta_win32.exe
2) I have overwritten the old version of gmic through this new installer
3) I have opened a small jpeg image (around 90 Kb) with Gimp 2.8.14
4) The Lab color mode crashes as usual as soon as I move the chroma curve:
https://dl.dropboxusercontent.com/u/309 ... _lab_color
4) Lch color always crashes this same image (chroma curve as usual) but the log file is extremely small (I have waited quite some time but it is always that small in the end...)
https://dl.dropboxusercontent.com/u/309 ... _lch_color

My system is:
Windows 7 - 64 bit
CPU Intel I7, Ram 8 gb, GPU nvidia geeforce

With the 32 bit gmic plugin I am trying right now I suppose I can not use my 8 gb Ram (on my 64 bit system) :-)

Thanks Silvio,

Did you really select 'Debug mode (logfile)' for the 'Output messages' option ?
In the second case, it looks like you have set it to 'Very verbose (logfile)' instead.

Anyway, thanks, that helps !

Hello David,

Here is the link for the Lch debug mode (logfile):
https://dl.dropboxusercontent.com/u/309 ... debug_mode

Thanks Silvio.
Now I don't see anything strange anymore
It seems it just crashes, without any clues on why it does that.

Could you now try to :
1) Update your filters (press the update button below the filter tree, with the Internet checkbox set).
2) Try again the Interactive color curves, using Lab mode.

And tell me if that is still crashing.

Thanks a lot for all your experiments !

Hello David,

I have just updated the G'MIC plugin through Internet (as you suggested).
At present, everything works fine :-)
No crash whatsoever so far: I have tried both the lab and lch color modes over and over again.

In addition, on a second different computer, I have installed GIMP 2.8.14 and today's latest gmic 1.6.1.0_beta_32_bit.exe plugin.
This new computer runs Windows 8.1 (64 bit).
On this 8.1 Windows system everything works fine (lab - lch colors): no need to "update" on internet the plugin either.

Btw, Lukáš Tvrdý has just updated the version of G'MIC for Krita next stable version as well [1].
Therefore, It is extremely likely the G'MIC 1.6.1.0 version is going to be available for Krita 2.9 users as well :-)

Mosf of all, THANKS a lot indeed for working on G'MIC (it is an amazing tool to work with) :-)

[1] http://quickgit.kde.org/?p=calligra.git ... 85d4692c27

Thanks Silvio,
Unfortunately, I've just found a workaround, so the bug is still here .
It probably needs further investigations to make me fix it.
I'll try to find out what is happening, and will probably request some more tests in the future !

Hi Silvio,

I have been able to test the latest beta of the G'MIC plug-in on a very similar configuration as yours : Win7, 64bits, with the plugin compiled in 32 bits.
So far, I've not been able to make it crash.
I don't know what I can do now...

Hello David,

Just tried myself on my usual computer (Windows 7 - 64 bit) and the Curve Interactive Lab - Lch mode both work fine: hence, do not worry :-)

BTW, I always got a very different sort of crash but this one is with the Extract foreground tool :-)
Whenever I re-apply this filter (shift+ctr+f), G'MIC 1.6.1.0 always crashes as soon as I close the window with the image (I mean the window containing the image to extract with all my greens - reds dots applied).
I suppose this bug was already there with the previous versions of G'mic
Molly, yesterday, on this same forum suggested not to re-apply twice (by typing shift+ctr+f) the filter :-)

Aside from this crash, what is really surprising (weird) to me is that the log file (debug mode log file) is always huge (tonight, the last one, was around 150 Mb)... :-)

Hello Silvio,

I've tried the Extract Foreground tool also in my experiments on Win7, and no crash...
I'm a bit worried about all those crashes, and the fact I cannot reproduce them easily. It means that can happen for other users and I've no easy ways to find a fix.

Your logfile size is quite normal, it just grows and grows, and this is a good idea to delete it before doing any experiment that leads to a crash. It won't delete itself automatically.

Hello David,

On my Computer this crash always occurs:
Gimp 2.8.14 - G'MIC 1.6.1.0
CPU Intel I7; 8 Gb RAM; GPU Nvidia geeforce 1 gb ram

In short, when I re-apply (ctr+shift+f) the filter extract foreground
I add some more points (greens - reds) : G'mic always crashes when I close the window
https://dl.dropboxusercontent.com/u/309 ... GROUND.png

My jpeg image (a bug with legs and antenna very similar to the background) is very difficult to extract.
No matter how many points I add: this is why I have tried to re-apply the filter (ctr+shift+F) :-)

EDIT:
Just tried my workflow (step by step with the same jpeg image) on Linux (Kubuntu 14.10 - Kde desktop)
No crash whatsoever...
I have re-applied the filter "extract foreground "(by pressing shift+ctr+f) over and over again and it always works

Maybe it is GKT stuff : I have read the GTK 2.X version has many problems related to Windows (e.g. NO full Tablet support for painters)
Unfortunately, nobody knows when Gimp will be fully ported to GTK 3 (not to mention the Gegl compatibility with G'MIC...) :-)

EDIT:
Just tried on a different computer with:
Windows 8.1 - 64 bit
Gimp 2.8.14 - G'MIC 1.6.1.0
CPU Intel I7; 8 gb RAM, GPU Nvidia graphic card 2 gb Ram
Same image as usual (the jpeg with a bug)

On Windows 8.1 it is much more difficult to crash G'MIC when I re-apply the filter (Shift+ctr+F) but in the end it crashes as well...

To recap:
- On Linux (Kubuntu 14.10) it works: no crash whasoever.
- On Windows 7 (64 bit) : it crashes when I re-apply (shift+ctr+F) the filter (always the first time I re-apply)
- On Windows 8.1 (64 bit) : it crashes only when I re-apply the filters several times in a row (it crashed on the 4th attempt)

Thanks silvio,

Pat David has also encounter the same kind of problem with the Lab interactive curves, and I really need to figure out how to fix this.
I would be very interested by those experiments :

1. What if try with the current stable version of the plug-in (1.6.0.4), available here : http://gmic.eu/
Does it crash too ? You said that it worked with 1.6.0.3, so I need to know if this crash occurs because of regressions I've introduced in 1.6.0.4 or 1.6.1.0 (it will be easier for me to know where I have to look).

2. Does it crash also with the plug-in executable provided by Samj (1.6.1.0_beta), available here : http://www.aljacom.com/~gmic/gmic-1610b ... 0150221.7z
I'd be interested by the test for the 32bits version, and the 64bits version.
I understand this is a lot of work, but as I can't reproduce the bug on my machine, I'm a bit blind and I need someone to guide me

3. Does it crash when using the CLI tool ? You can get it from here : http://gmic.eu/
This is the 32bits version. Here are the steps you can try :
- Uncompress the archive, it creates a new directory.
- Open a Windows console, go to the newly created directory (cd ..blabla...)
- Type 'gmic.exe -testimage2d 2048 -x_color_curves lab'.
Try setting the lab curves. In case of crash, re-run it with : 'gmic.exe -debug -testimage2d 2048 -x_color_curves lab' and tell me what it displays at the end.

I know I'm asking a lot, anyway. I'm trying to look more closely to the code on my side to see what could have happened.
Thanks by advance!

OK, Pat David was kind enough to make some tests for me, and it seems that the problem arises for machines that have 8 available (virtual) CPUs.
That could explain why I couldn't reproduce the bug on my machine. What I've done so far is to limit the number of CPUs used for the computation to 4.

So, could you please try to refresh your filters once again, and try the Lab color curves [interactive] filter, and tell me if that is working as expected (no crash should happen ideally).

Thanks !

Hi David,

It is a real *pleasure* to test G'MIC (really ! ) :-)

Just tested G'MIC 1.6.0.4 (through your first link) and it is not usable for me.
It crashes on both the lab and the lch mode (interactive curve).
This version is even worse than the 1.6.1.0 for me...

Here is the video for this crash:
https://dl.dropboxusercontent.com/u/309 ... _CRASH.avi

Here is the log report for this same crash (see above video):
https://dl.dropboxusercontent.com/u/309 ... _LAB_CRASH

Btw, today I have tried the extract foreground filter on Windows Xp - 32 bit (G'MIC 1.6.1.0) and it is terrible.
It crashes sometimes even at the very beginning, when I start putting dots (greens - reds) without re-applying the filter (by pressing shift+ctr+F).
Most of all, it always crashes when I re-apply the filter (Shift+ctr+F) but this usually also occurs for me on Windows 7 and Windows 8.1 (both computers have an Intel CPU I7 - multi-core).

On the contrary, on Linux - Kde everything works fine instead (which is a pity because I could send you a nice gdb backtrace...).
Pay attention, please, on Linux, actually I run a virtual machine (thorugh Vbox) since Windows 7 is the host (and Kubuntu 14.10 is the guest)...

See this image instead for Windows Xp (extract foreground - G'MIC 1.6.1.0 - Gimp 2.8.14):
https://dl.dropboxusercontent.com/u/309 ... _ERROR.jpg

EDIT
Ok, Just read your last post: therefore, I have updated the filter (1.6.0.4) through GIMP - Internet and now everthing works fine !
No crash whatsoever (lab and lch work OK) on Windows 7 - 64 bit.

EDIT
Just tried the console mode as well on Windows 7 - 64 bit (CPU Intel I7 as usual).
I have tried over and over again the command:
gmic.exe -testimage2d 2048 -x_color_curves lab
and the command:
gmic.exe -testimage2d 2048 -x_color_curves lch
Both lab and lch work fine on the console (the image never crashes and it has changed fine by the curves).

Thanks a lot Silvio !
That helps for sure !

So you say that the Lab interactive color curves works nicely on 1.6.0.3, but not on 1.6.0.4 ? That is indeed very interesting, as I know a little bit more where to search the regression (in the Changelog of 1.6.0.4 :p ).
Also the snapshot you've send is again a problem of missing memory. When the error message says that G'MIC was not able to allocate a certain amount of memory (in your case 178Mb), then this is an exception returned by the system memory allocation procedure (operator 'new[]' in C++), and we cannot do much about that problem.

It seems also that running the filter from the console always work. Could you please try to emulate exactly what the plug-in calls, i.e. by typing :

gmic.exe your_image.jpg -gimp_curves_interactive 6,0,0,0,0,0,100,100,-1,0,0,100,100,-1,0,0,100,100,-1,0,0,100,100,-1,0,0,100,100

in the console ? It will tell me if the problem comes only in the GIMP environment or not.

Last thought : I still don't know why, but it seems that limiting the number of processors to 4 solves the problem for the 'Lab interactive color curves' filter. This is very strange, as the code I use for multi-threading does not depend on the number of threads run, but this is probably something I need to explore.
How many cores do you have on your machine ?

Anyway, we are going ahead with all your tests, so thanks again !

here is the message from the console (no crash whatsoever):
https://dl.dropboxusercontent.com/u/309 ... essage.png

As noted, the 1.6.0.4 downloaded today through your first link crashed on Gimp 2.8.14 (Windows 7 - 64 bit).
But, when I have updated this plugin through Internet (G'MIC Gui button), later on, the lab - lch modes were working super fine indeed :-)

Here is the info regarding my CPU:
https://dl.dropboxusercontent.com/u/309 ... U_INFO.png

Hi Silvio,

Thanks.
I've been able to set up a Win7 virtualized machine on my PC today, and the good news is :
the Lab interactive color curves filter crashes !

Now, I will be able to inspect the problem locally on my machine, and that is already a giant step for me.
Samj has also helped a lot in giving me her compiling environment for Win7, so I think I'm ready to hunt the bug now.

I'll give you news of course

Some news, after testing :
The new compiling environment I'm using (from Samj) on my Win7 virtualized machine produces both 32bits and 64bits versions of the plug-in, and surprisingly, when I'm testing them with GIMP 2.8.14, the 'Colors / Curves [interactive]' filter with the Lab color space works !
Both for the 32bits and 64bits version of the plug-in!
I don't really have an explanation. The compiler version is much more recent (g++ 4.9 vs. 4.5), so the generated code is different. It's quite hard to understand what happened here.

Anyway, I'll now build the G'MIC binaries using the compiling environment from Samj instead. So, at least, this bug should be gone. And I'll be able to produce new binaries more often, both for Win32 and Win64 systems. I'll try to automate that in my release scripts this afternoon.

OK, so everything went smooth. If you are interested, you can try these new installers (or .zip files) to install latest beta (from today, 02/25/2015) :

http://gmic.eu/

Let me know if that works as expected.

Hi David,

Just tested the new version 1.6.1.0_beta_win_32.exe.
The curve interactive works fine (no crash whatsoever on Windows 7 - 64 bit).
Tested both the lab and lch mode (on top of rgb mode etc).
Great work indeed: thanks a bunch :-)

Btw, the extract foreground (interactive) tool does not work when I re-apply this same filter twice (by pressing shift+ctr+f).
It has never worked for me (no matter what Windows version I try).
Needless to say, in the future I am going to stick to what Molly suggested on this forum: never re-apply it (shift+ctr+f) twice :-)

Hi Silvio,

I didn't say I won't try to understand what happens with your second bug (foreground extraction filter)
I'll try to make it happen on my machine too in a first step.
Maybe you could detail the procedure you are doing to get the crash ?

Thanks!