[Tutorial] SketchUp Ruby C Extension

thomthom

Ah, so you initially used a Ruby version you built yourself?
I don't really know what makes the difference there. I just tried to find a minimum work solution - so I started out with a pre-built Ruby.

@jefftrull said:

I did the config.h hack for the one-click version, but for the VS2008 and VS2010 I was able to compile the extension with the same version of VC++ I used to build Ruby itself, so it wasn't necessary... I think?

Dunno - you might not need it if you built Ruby yourself. The version you had might not have that check in it. With the pre-built version I used I had to remove that check otherwise it would not compile.

Dan Rathbun

@jefftrull said:

That's an interesting question. I haven't changed anything in the Sketchup program dir, beyond copying the SX_HelloWorld.so into Plugins. What is the significance of those msvcrt* files?

It's THE Ruby interpreter that the Sketchup executable loads. (Copied from the pre-built v1.8.6-p287 binary package.)

The one that your full Ruby install (ruby.exe and rubyw.exe,) loads is in the "bin" subdir of the full install folder heirarchy.
see: Ruby Interpreter DLLs (Win32)

and making sure you saw this link list: [Info] C/C++ Ruby extensions & SketchUp plugins

jefftrull

Thanks again for the helpful information, Dan. I gathered from the "Ruby Interpreter DLLs" link that you can replace the interpreter used by Sketchup with a different one; indeed that seemed to work, but it made no difference in my ability to load the .so file. Still works OK in irb (when the compiler versions match), fails in Sketchup. The good news is it doesn't really matter - I have a working recipe now. The one-click binary is the only thing that's worked for me over all the combinations, and with that, either VS2008 or VS2010 works fine.

I'm off to turn this into a CMake recipe (this application is cross-platform); would anyone be interested in seeing that when it's done?

Cheers,
Jeff

Dan Rathbun

Be aware that Mac Sketchup still distro's with Ruby v1.8.5 initial release.

thomthom

@jefftrull said:

I'm off to turn this into a CMake recipe (this application is cross-platform); would anyone be interested in seeing that when it's done?

Yes! If you want to add it to the existing project just let me know. I think you can just make a pull request anyway (though I've never used this feature.)

I know that Dana tried to set up a CMake project - which AFAIK failed under SU. So a lean clean CMake bare bone recipe would be very much welcome!

jefftrull

Let's see how this works for people. Tested only under Windows, although I tried to make the analogous changes for OSX based on my reading of extconf.rb. This file is intended to live under "src" and will create targets for both extensions.

CMake control file for Sketchup Ruby C Extension

thomthom

@jefftrull said:

Let's see how this works for people. Tested only under Windows, although I tried to make the analogous changes for OSX based on my reading of extconf.rb. This file is intended to live under "src" and will create targets for both extensions.

Sorry, haven't had time to look at this yet. I've got quite a list of things to do for the time being...

chyn2000

Hey guys,

So, I got the ruby extension process to work, but I seem to be running into one little catch... multithreading.

FYI: I'm using Pelle's C with the options TBD suggested in his SUExt example. I'm on a Win 7 (64 bit) system (though hopefully that won't matter).

I'm trying to use _beginthread, but using TBD's settings, it seems I have an unresolved symbol '_beginthread'.

Just to let you know, I am kind of new to multithreading and DLLs. I have created threaded applications, but never a threaded dll.

Taking an example program that I have gotten to compile and run as a standalone app (pelle's C option multithreaded (lib)), I add the libs, includes and functions required to make it a ruby extension and it suddenly can't resolve _beginthread.

A couple of modifications to TBD's project settings and I get it to compile and even be recognized by the irb. However, it seems the second the thread tries to run, the whole irb closes.

The settings I modified to "get it to work" was:
Runtime library from multithread (dll) to multithread (lib)
and turn off "Omit default library in object files"

Any ideas?

Thanks,
Reg


// general Windows
#pragma comment(lib, "kernel32.lib")
#pragma comment(lib, "gdi32.lib")
#pragma comment(lib, "user32.lib")
#pragma comment(lib, "shell32.lib")
#pragma comment(lib, "advapi32.lib")

// Ruby
#define HAVE_ISINF 1  // isinf macro redefinition workaround
#pragma comment(lib, "msvcrt-ruby18.lib");

// DLL
#pragma comment(lib, "crt.lib");

#include <stdio.h>
#include <windows.h>
#include <process.h>         // needed for _beginthread()
 
// Ruby headers 
#include "ruby.h"

void  silly( void * );       // function prototype

// De fining a space for information and references about the module to be stored internally
VALUE KMRH = Qnil;

int j = -1;

// our function
VALUE method_display_model(VALUE self, VALUE input)
{
    // Our program's first thread starts in the main function.
    printf( "Now in the main function.\n" );

    // Let's now create our second thread and ask it to start
    // in the silly() function.
//    _beginthread( silly, 0, (void*)12 );  //Commented out to see if the program would stay alive and it does

    // From here on there are two separate threads executing
    // our one program.

    // This main thread can call the silly() function if it wants to.
	while(j != 0)
	{
    	silly( (void*)j );
		j--;
	}

    Sleep( 10000 );

	int i = NUM2INT(input);
	return INT2NUM(i + 1);
}

// The initialization method for this module
void Init_KMRH(void)
{
	KMRH = rb_define_class("KMRH", rb_cObject);
	rb_define_method(KMRH, "display", method_display_model, 1);
}


void  silly( void *arg )
{
	int k = 10;
	while(k-- != 0)
	{
//    	printf( "The silly() function was passed %d\n", (INT_PTR)arg );
    	printf( "The silly() function was passed %d\n", k );
	}
}

thomthom

That's way beyond me. I've not dealt with threading at all. But I'll be keeping an eye on this as it something I might want to do in the future.

chyn2000

So, I have to apologize for wasting your time...

First thing I need to do (yet again) is RTFM.

The answer to threading is incredibly simple:
_beginthread( silly, 0, (void*)12 );

becomes

rb_thread_create( silly, (void*)12);

found it on:
http://ruby-doc.org/docs/ProgrammingRuby/html/ext_ruby.html

So, if you will excuse me, I need to go beat myself half to death...

thomthom

You will not get any joy from using Ruby threads. They are "green" threads. It's just Ruby dividing up the tasks itself - not the system. So making a thread in Ruby will not prevent SU from freezing.

Parallelism is a Myth in Ruby - igvita.com

(www.igvita.com)

Dan Rathbun

And Daniel Berger's win32-thread package that wraps native threads is still Alpha and unstable (AFAIK.)

jefftrull

I'm on to the next challenge now... building for the 1.8.7-p334 version my colleagues prefer (better compatibility with certain gems). Unfortunately this one was built with with mingw, and my VS2010 build flow that worked for 1.8.6-p287 (mswin) is not successful here. I have the following results:

build against 1.8.7-p334 built from source with VS2010: compile succeeds, load fails with missing dll msvcrt-ruby18-vc100 (or something like that). If I supply the dll, Sketchup crashes.
build against 1.8.7-p334 one-click install (mingw): compile fails. Lots of warnings about win32.h and missing ruby API functions
hybrid build pointing includes to mswin build of 1.8.7-p334, library from mingw one-click: link error

I'm under the impression that because dll's use the C API it should be possible to load a VS2010-generated extension into a mingw-built Ruby. Is it a hopeless cause?

Thanks,
Jeff

jefftrull

Today I made a startling discovery: the plugin I built under VS2010 against the 1.8.6-p287 mswin Ruby will load and run without changes in the 1.8.7-p334 mingw build. So the Ruby version and compiler both mismatch, and yet everything works. I'm dumbfounded but delighted, and my previous question is now moot.

thomthom

thomthom

@thomthom said:

Btw, does anyone know how one can make extconf.rb and make/nmake output their files to a given folder? I'd like to make it so that each platform outputs its generated files to separate directories.

Anyone know how to achieve this?

Making it so that when I build and compile under Windows everything is placed in a "win32" folder and "osx" when using OSX?
Or rather, to clarify, I know how to do conditionals - just don't know how to control where the generated files are output.

Dan Rathbun

and you are compiling from the commandline, not from within the Visual Studio GUI interface, correct ?

thomthom

Currently using nmake that comes with MS VS C++.

And I'm using a extconf.rb with mkfm to generate the makefile.

thomthom

Is it possible to build an Ruby C Extension under Windows that doesn't require additional libraries? When using Visual C++ to build the extensions require the C++ runtime library.

tbd

yes. my old example did that. best is to check the DLL imports section to see what functions it requires