BUG DESCRIPTION:----------------I tried to build a gateway for some C function, but it did not work. The compilation process went through w/o any problem or compilation errors, but after loading I get some error that the number of input argument is wrong. I modified the code to trace the number of input argument that the gateway function actually sees, and it is some nonsensical very big value. I tried to do the same With the foo6 sample code given in manual page:Aide de Scilab >> API Scilab > Getting started with API_Scilaband I got the same issue.BTW, I tried to have a look at SCI+'contrib/toolbox_skeleton/sci_gateway/c/sci_foo.c' and the code is different from what is documented in the manual, it seems to use the legacy API and not the new one.
Designs
Child items 0
Show closed items
No child items are currently assigned. Use child items to break down this issue into smaller parts.
Linked items 0
Link issues together to show that they're related or that one is blocking others.
Learn more.
Probably I am not using the right compiler for MAC OS X --- i.e. there is some discrepancy between the compiler which you used for generating scilab and the one that comes by default with Mac OS X on my machine, and the link therefore does not work properly.
I understand that what you recommend is just the default compiler...
Maybe scilab ilib_build should run some test program with the detected compiler in order to configure properly the dynamic link. This configuration could then be cached and retrieved at each ilib_build calll after checking that a md5sum of the output of gcc --version + ld --version, etc... is the same signature as that of the tools that were used to generate the configuration (or maybe MACOSX file system allows directly to get the m5sum signatures of gcc, ld, etc... program files).
Anyway, whatever the source of this problem this is really a serious issue for me, and your help is most welcome !
Ok, making a gateway did not work at my office (on MAC OSX 10.12), so I tried it at my home (on MSWindows10) still with the same sci_foo6 example taken from the Manual.
I had MSYS2 MinGW 64 bits already installed on my machine, but for some reasons after installing the mingw ATOMS the compiler was not detected (indeed it was not in my MSWindows PATH variable, but only in the MSYS2 PATH as set by my MSYS2 ~/.profile file.
OK, I read from the mingw ATOMS (~/Scilab/scilab-6.0.0/atoms/x64/mingw/0.10.5/readme.txt) that there was some gcc to install from scilab site (from http://atoms.scilab.org/toolboxes/mingw/0.10.0/files/gcc-6.2.0-64.exe), I downloaded it, installed it. For some Strange reason it was installed Under ActiveState perl site directory --- probably perl had its own gcc, and the installer found that in the registry.
Anyway, I made the test after rebooting my machine, and it failed too, but differently, it could not even build up (see log attached).
I would be great if the mingw ATOMS had some indication how to set the compiler to MSYS2 mingw (that would save disc space, ...
I would be great if the mingw ATOMS had some indication how to set the compiler to MSYS2 mingw (that would save disc space, and make the updates simpler, as MSYS2 has some package manager funilly named pacman).
OK, I could run the jogl2 self test after the JogAmp maintainer told me to fall back to java 8, instead of java 9, as they have a yet unresolved pb with java 9.
So I did that and tried again to build Scilab from source, and I got this error message :
checking jogl2... no
configure: error: Could not find or use the Java package/jar jogl2 used by Scilab 3D rendering - Version 2.0 (looking for package javax.media.opengl.glu.GLUnurbs)
Actually, the configure script is looking for javax.media.opengl.glu.GLUnurbs, but this is under com.jogamp, and not under javax.media.
So, there is some problem of the configure script, maybe the Makefile.in is OK and I just need to re autoconf again... well, some guidance from you would be most welcome.
Just not to loose from site, what I just want to do is to build a f**king gateway on a MAC OS X 10.12
OK, I could run the jogl2 self test after the JogAmp maintainer told me to fall back to java 8, instead of java 9, as they have a yet unresolved
pb with java 9.
So I did that and tried again to build Scilab from source, and I got this error message :
checking jogl2... no
configure: error: Could not find or use the Java package/jar jogl2 used by Scilab 3D rendering - Version 2.0 (looking for package
javax.media.opengl.glu.GLUnurbs)
Actually, the configure script is looking for javax.media.opengl.glu.GLUnurbs, but this is under com.jogamp, and not under javax.media.
So, there is some problem of the configure script, maybe the Makefile.in is OK and I just need to re autoconf again... well, some guidance from
you would be most welcome.
Just not to loose from site, what I just want to do is to build a f**king gateway on a MAC OS X 10.12
Hello Vincent,
Thanks for all the comments, I hope it will help newcomers. About the JOGL2 issue, this seems related to a JOGL version mismatch the javax.media.opengl API has been moved to com.jogamp on jogl-2.3.0 ; a Scilab patch is provided on https://gitlab.com/scilab/legacy_codereview/-/wikis/changes/17530/ .
Hello Clement,
Thank you for the update.
Actually, when I first reported the issue I had not tried to build with all the instruction on the Wiki (https://wiki.scilab.org/Compiling%20Scilab%20under%20MacOSX). Sorry for that --- and thank you for the pointers which you have sent to me, notably the one to the wiki.
So, I got some of the thirdparties from your svn repo, and I also installed fink from sources (quite a long process...).
It now goes further to the extent that I complete the configure step (the jogl2 step is passed).
Unfortunately it fails at the java compilation.
Vincent.
PS: For your notice here are the lib I used, this is not related to the issue, just for this to be tracked down.
arpack-ng 3.4.0-1
gettext-tools 0.19.8.1-2
gettext-bin : no such package on fink, but anyway I already had gettext from brew, and its version is 0.19.8.1
pkgconfig 0.29.2-1
hdf5.10-shlibs
I attached a revised patch for the java compilation (#15388#attach_4632)
Concerning the issue with HDF5, it was just that I had v1.10, and scilab needs the previous version. So I installed it and I could compile completely. However, when I launch it I get this error message:
Scilab startup function detected that the function proposed to the engine is the wrong one. Usually, it comes from a linker problem in your distribution/OS.
Here, Scilab should have 'libscijvm' defined but gets 'libscijvm-disable' instead.
If you do not know what it means, please report a bug on http://bugzilla.scilab.org/. If you do, you probably know that you should change the link order in SCI/modules/Makefile.am