Hi Folks,

Here is a new stable release of the G'MIC image processing framework, numbered 1.6.1.0.
Binaries are already available for users of Linux32/64 and Windows32/64 systems.

What is major this time ? Well, first, I have done a lot of code cleaning, and this leads to a more stable, fast and less memory-eating G'MIC interpreter, so every interfaces build upon this interpreter benefits from these improvements. In particular, the G'MIC plug-in for GIMP has been improved a lot: it better handles selections, multiple layer previews, uses libcurl natively (so filter update is faster). I've also added few new filters, e.g. the Black & White / Engrave filter, which can really do some awesomeness
But, really, most of the work has been done for the improvement of the internal G'MIC engine. I hope you'll enjoy this version !

Here is the Changelog :
---------------------------

New features:
- [all] G'MIC can now use libcurl for a native support of downloading network files. Thanks Lukas, for the suggestion!
- [all] New command -split_colors which decomposes an image into several images, each containing only a non-transparent single color. The plug-in filter Layers / Colors to layers has been updated to use this new command.


- [gimp] Optimal preview size is automatically guessed at startup.
- [gimp] New filter Black & White / Engrave which can render a wide variety of black and white effects from an image (opt. with color background), like comics-like rendering.




Improvements:
- [all] Substituted strings @{ind,b}, @{ind,f} and @{ind,x} are not limited to 256 char anymore.
- [all] Several sessions of valgrind and gprof and various static code analyzers allowed to optimize and stabilize the code of the interpreter.
- [all] Binaries for Windows 32bits and 64bits are now compiled automatically by the build scripts. Installers for Win64 are also available now.
- [all] Suppressed warnings when compiling G'MIC with Visual Studio.
- [all] Removed the 'modified' tag of image properties. Was computed but never used.
- [all] Command '-solve' now works to solve linear system with the right-hand side is a matrix instead of a vector.
- [all] Math parser has now a predefined variable 'ic' equal to the median of the reference image.
- [all] Better handling of 'inf' and 'nan' for G'MIC command arguments.
- [gimp] Name of each layer is now passed to all G'MIC filters, and a filter can also output new layers with custom names.
- [gimp] Improvements on how layers are modified. Filters now better preserve the user-defined blending mode, opacity, position and names of input layers.
- [gimp] The plug-in better handles filters applied on selections.
- [gimp] When updating filters, the current filter parameters are kept unchanged when possible.
- [gimp] Better and customizable preview layout when previewing multiple output layers.

Bug fixes:
- [all] Better precision and more robustness for texture mapping in 3d objects.
- [all] Fixed possible buffer overflows on Windows (Microsoft's snprintf() and vsnprintf() do not act as standard C99 versions..).
- [all] Fixed small bug that prevents getting wrong values of the 'ia' variable when evaluating math expressions in a multi-threaded environment.
- [all] Reading video files with 'step_frame>1' with OpenCV could hang, because of a unreleased mutex. This has been fixed.
- [all] Fixed bug happening when trying to use libcurl for reading an URL containing a Windows-reserved character.
- [all] Fixed: wrong line number was displayed in case of an error happening just after a -endlocal directive.
- [zart] ZArt now compiles for both versions 4 and 5 of Qt.

Cheers,

David.

. Sadly I cannot get on the internet with my computer to download it tonight. Will try again tomorrow..... Something wrong with my wireless modem I think. Even the comnection with my phone is disabled.

Thanks David; no more beta in the name now.

Thank you David for your great job to improve GMIC.

Perhaps you could also have announced that 1.6.1.0 marks the point at which GMIP-G'MIC doesn't work for GIMP 2.6.x any more:



Kevin

ps. I'm not suggesting that you continue supporting 2.6.x indefinitely, just nice to know where the end-point is.

@paynekj : This is due to the API change between 2.6 and 2.8. The G'MIC plug-in should run well on GIMP 2.6, but you need to compile it with the gimptools for 2.6.
I personally use 2.8 on my machine.

unless there is a prerequisite on the image, the filter Saturation Equalizer (Colors, author Iain) abends when simply entered;
*** Error in ./Saturation_EQ_p/*repeat/*local/apply_parallel_overlap/_apply_parallel_overlap/ *** Command '-check': file '$3>=0 && isint(0) && 0>=0' does not exist.

@dinasset : you mean, when you select it from the plug-in, it does not work ?

yes, having an image opened in Gimp, selecting that filter, I get that error message

OK, it has been fixed.
Thanks for reporting !

thanks, David!

When i install i get this dialog for pre release 1.6.2.0? Is this correct?

Love the new Engrave filter though.

Yes, Rod, this is correct to get the 1.6.2.0 prerelease version.
1.6.1.0 is a quite stable basis to work with, so I've started a kind of rolling release system where I provide the latest features to the users.

Hi David, There is a minor glitch in Layers -> Colors to layers. The minimal area needs to be divided by 100 to be correct.

Hi David.

Good work, as always.
64bit gmic_gimp.exe depends on zlib1.dll and zlib1__.dll,
it also depends on libeay32.dll and libeay32__.dll.
But it just depends on libssh2-1__.dll.
Why is that so?

Greets Hatti

@Onkel Hatti : Well I don't know I use the compiling environment set by samj, maybe she can tell a bit more about the dependencies !

@preben : I'll check that, thanks !

Hi David,
I don't remember if already posted, but I entered just now the filter
Test/Deconvolve (by Jéjé)
and it abends immediately without touching any parameter
after Apply I get the message
*** Error in ./jeje_deconvolve/*if/(...)/*repeat/*local/_apply_channels_old/deconvolve_richardsonlucy/ *** Command '-check': expression '0 && 5>=1' is false.

thanks for the attention

Edit: maybe it has been replaced by Sharpen/Richardson-Lucy
in this case the one under Test could be removed IMO

Hi David,
today I made some experiments with G'MIC Details equalizer
it seems that when using not very little values, the resulting image is split horizontally into two halves, with a different luminosity.
I attach here:
a- the source image
b- the .xcf with 3 layers each one with different (increasing) parameters values
(note: all the 3 layers were produced -one at a time- applying the filter on the BG)
c- the .png showing the last layer outcome

something to do with parallel processing?
thanks for your support

a- source image


b- .xcf output (scaled down for size problems)
each layer keeps info on parameters values used


c- .png output (last outcome)



Edit: definitely something to do with parallel processing:
forcing "one thread" outcome is OK!

You need to increase the Spacial Overlap slider dinasset. 16 to 24 usually works for me. If you don't, then you get this bar line (and even more if you use more processors). If you don't see the control bar called Spacial Overlap, then that is a problem.