[Mrtrix-discussion] Compiling and running MRtrix on Mac OS X

[Jon Clayden]

Dear all,

Just to follow up on the earlier thread quoted below (I wasn't on the mailing list when it was sent, so I can't directly continue it), I thought it might be useful to post my experiences of compiling MRtrix on Mac OS X.

I've had problems with Fink in the past, and I don't particularly like its approach of creating a whole software ecosystem separate from the rest of the system, so instead I used Homebrew <http://mxcl.github.com/homebrew/> to sort out the dependencies, which I find much cleaner. (I did - foolishly - attempt to build the dependencies by hand initially, but got stuck with pango and gave up.) So the steps were

1. Install Homebrew as per <https://github.com/mxcl/homebrew/wiki/installation>:

  $ ruby -e "$(curl -fsSLk https://gist.github.com/raw/323731/install_homebrew.rb)"

(Xcode and X11 from the OS X install DVD are also needed, if they're not already installed.)

2. Get and build the dependencies:

  $ brew install gtkglext gtkmm gsl

(The rest should be picked up automagically as dependencies of these three.)

3. Patch MRtrix. A sysconf/darwin.py file needs to be created, as Michael says (mine is attached - it only includes the "-mtune" change), and since I'm using the system OpenGL library rather than something provided by Fink, some .cpp/.h files need modifying. A patch using conditional compilation directives (to avoid breaking other platforms) is at the end of this e-mail. It should be safe to incorporate this into the main code-base.

4. Run ./build and proceed as normal.

This seems to result in working binaries. Naturally, most OS X systems won't have the run-time dependencies installed, so unless the whole lot gets bundled up together I wouldn't expect the binaries to be portable.

If it's not too much work, though, it might be helpful to provide an alternative to the "build" script which only builds the command-line applications. Presumably the dependencies are far fewer in that case, so it might represent a simple "way in" for OS X users...

All the best,
Jon

-------------- next part --------------
A non-text attachment was scrubbed...
Name: darwin.py
Type: text/x-python-script
Size: 1187 bytes
Desc: not available
Url : http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20110519/1ab365f1/darwin.bin
-------------- next part --------------


--
diff -r mrtrix-0.2.9/src/dwi/render_frame.cpp mrtrix-0.2.9-patched/src/dwi/render_frame.cpp
27a28,30
> #ifdef __APPLE__
> #include <OpenGL/glu.h>
> #else
28a32
> #endif
diff -r mrtrix-0.2.9/src/mrview/pane.cpp mrtrix-0.2.9-patched/src/mrview/pane.cpp
22a23,25
> #ifdef __APPLE__
> #include <OpenGL/glu.h>
> #else
23a27
> #endif
diff -r mrtrix-0.2.9/src/mrview/pane.h mrtrix-0.2.9-patched/src/mrview/pane.h
25a26,28
> #ifdef __APPLE__
> #include <OpenGL/glu.h>
> #else
26a30
> #endif
diff -r mrtrix-0.2.9/src/mrview/sidebar/tractography/roi_list.cpp mrtrix-0.2.9-patched/src/mrview/sidebar/tractography/roi_list.cpp
27a28,30
> #ifdef __APPLE__
> #include <OpenGL/glext.h>
> #else
28a32
> #endif
diff -r mrtrix-0.2.9/src/mrview/sidebar/tractography/track_list_item.cpp mrtrix-0.2.9-patched/src/mrview/sidebar/tractography/track_list_item.cpp
30a31,34
> #ifdef __APPLE__
> #include <OpenGL/gl.h>
> #include <OpenGL/glext.h>
> #else
32a37
> #endif
diff -r mrtrix-0.2.9/src/mrview/sidebar/tractography/track_list_item.h mrtrix-0.2.9-patched/src/mrview/sidebar/tractography/track_list_item.h
41a42,44
> #ifdef __APPLE__
> #include <OpenGL/gl.h>
> #else
42a46
> #endif
diff -r mrtrix-0.2.9/src/mrview/texture.h mrtrix-0.2.9-patched/src/mrview/texture.h
26a27,29
> #ifdef __APPLE__
> #include <OpenGL/gl.h>
> #else
27a31
> #endif
diff -r mrtrix-0.2.9/src/use_gl.h mrtrix-0.2.9-patched/src/use_gl.h
25a26,28
> #ifdef __APPLE__
> #include <OpenGL/gl.h>
> #else
26a30,31
> #endif
> 



> Hi Michael,
> 
> This is great news, and doesn't even sound too complicated! I'll add your
> instructions to the documentation and make sure they're available for the
> next release. In the meantime, I'd be interested to hear how things are
> going with the Mac version, and whether you've encountered any issues with
> it so far.
> 
> Thanks,
> 
> Donald.
> 
> 
> On 26 February 2011 13:19, Michael Zeineh <mmzeineh at gmail.com> wrote:
> 
>> Got it working! At least the basics ... mrview and mrconvert etc. Didn't test tractography yet ..
>> 
>> I tried copying from one machine (10.6.5) to another (10.6), but this didn?t work, so I don?t think it is worth putting binaries anywhere ...
>> 
>> Here is what I had to do
>> 
>> Install fink per webpage in /sw64
>> 
>> Download tarball
>> 
>> tar -xvzf fink-0.29.19.tar.gz
>> 
>> cd  fink-0.29.19
>> 
>> ./bootstrap sw64 (puts fink in /sw64)
>> 
>> fink install python glib gtk+ glibmm2.4 gtkmm2.4 gtkglext1 gsl
>> 
>> at some point it bottoms out, so just repeat it, and it works
>> 
>> copy linux.py Darwin.py
>> 
>> modify Darwin.py cpp flags to read ?mtune=native rather than ?march=native
>> 
>> modify darwin.py to add to the cpp flags: '-L/sw64/lib', '-I/sw64/include'
>> 
>> ./build
>> 
>> copy to /usr/local/mri
>> 
>> change permissions so all users can read
>> 
>> change path to include /usr/local/mri/bin
>> 
>> export DYLD_LIBRARY_PATH="/usr/local/mrtrix/lib"
>> 
>> Then things work!
>> 
>> One time, it was complaining regarding the freetype library, and this required copying the /usr/x11/bin/libfreetype.6.dylib to /sw64/lib/freetype219/lib, though it didn?t require it the second time.