[yt-users] Yt not installing right

[rsoares]

It is gfortran.

Also I've tried installing numpy-1.2.1 myself since I notice my version 
of python is 2.5 and attached the (succesful ?) install log from that.  
The old yt used python 2.5 but the new one is 2.6.  I had to install 
python-dev however to get my numpy to install.


Matthew Turk wrote:
> [install_log trimmed]
>
> So, it seem to me that it is in fact a semi-known issue, where numpy
> is using g77 to link and install.  Unfortunately, it doesn't seem to
> correctly use the "-shared" option.  This shows up in the lapack
> compilation & linkage step, which is where the install fails.
>
> We can try a couple things here, but unfortunately it might be a
> little bit experimental.  I will be using the environment variables
> you have set in the install_script.sh here for convenience; you'll
> need to insert the correct value of DEST_DIR.  A lead $ means that
> command should be typed in at the prompt.
>
> ##
>
> First off, try installing gfortran, if it is not already installed.
>
> $ sudo apt-get install gfortran
>
> Then, if that succeeds:
>
> $ cd $DEST_DIR/src/numpy-1.2.1
> $ rm -rf build dist
> $ $DEST_DIR/bin/python2.6 setup.py build_ext --fcompiler=gfortran
> $ $DEST_DIR/bin/python2.6 setup.py install
>
> If this works, and it might not,
>
> $ touch done
>
> and then re-run the install script.  This should pass over most of the
> installation and continue on where it left off.
>
> The second option, which I think is less likely to work:
>
> ##
>
> $ cd $DEST_DIR/src/numpy-1.2.1/
> $ $DEST_DIR/bin/python2.6 setup.py config_fc --f77flags="-shared" built_ext
> $ $DEST_DIR/bin/python2.6 setup.py install
>
> and if that works,
>
> $ touch done
>
> and then, again, re-run the install_script.sh.
>
> Let me know how those two options go; hopefully we can figure this
> out.  If not, I have another idea that I'm a bit reluctant to suggest,
> which is to have you "$DEST_DIR/bin/easy_install scons numscons" and
> then run "$DEST_DIR/bin/python2.6 setup.py scons install" from the
> numpy directory.  David Cournapeau has done some amazing work with
> scons, and it seems like it will ultimately be the build system for
> NumPy (and it's pretty great) but I am hesitant because it requires
> two additional dependencies right now.
>
> Let me know how it goes,
>
> -Matt
>
> On Thu, Jan 22, 2009 at 2:33 PM, rsoares <dlleuz at xmission.com> wrote:
>   
>> Hi Matt,
>>
>> I have GNU Fortran (GCC) 4.2.4 on Ubuntu and here is the full installation
>> log.
>> I can use Intel fortran compiler if you think it would help.
>>
>> Matthew Turk wrote:
>>     
>>> Hi!
>>>
>>> Sorry to hear you're having problems, but it's good to hear that
>>> you're trying out the newest version, from the trunk.  :)  I have a
>>> suspicion I know what's going on, but it would help out if you could
>>> send (even off-list, if you'd rather) the rest of the installation
>>> log.  What I believe is happening is that the numpy distutils are
>>> using the wrong fortran compiler to compile and link the fortran
>>> interface libraries.  This is a problem I ran into a while back, but
>>> my solution was probably not very portable.
>>>
>>> If we can figure out which fortran compiler you're using, we might be
>>> able to make some progress.  Are you running on Debian?  Can you tell
>>> me if you have any of g95, gfortran or g77 installed?
>>>
>>> Thanks!
>>>
>>> -Matt
>>>
>>> On Thu, Jan 22, 2009 at 1:00 PM, rsoares <dlleuz at xmission.com> wrote:
>>>
>>>       
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
>   
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: install.log.2
URL: <http://lists.spacepope.org/pipermail/yt-users-spacepope.org/attachments/20090122/666e9a95/attachment.ksh>