[IPython-dev] nbconvert on Scientific Linux 5

[Aaron Meurer]

Oh, I missed that you are already using Anaconda. In that case, you
already have conda.

By the way, here's the conda recipe I used for pandoc:
https://github.com/conda/conda-recipes/tree/master/pandoc. It's kind
of hacky (i.e., it installs things into the local cabal, because I
didn't feel like building up a true conda package stack for Haskell
packages just for the sake of pandoc). It depends on gmp 4, libffi,
and zlib, all of which are also in that same repo.

The notes.md unfortunately doesn't do justice to the pain I went
through trying to get things to work. For instance, it matters if you
have profiling enabled in the haskell-platform, and you have to do it
from the start, because you can't build against dependencies unless
they also have it enabled. At one point, I ended up manually
recompiling about 100 recursive dependencies of pandoc. And I think in
the end it didn't even work for some reason...

Aaron Meurer

On Fri, Jun 27, 2014 at 6:46 PM, Aaron Meurer <asmeurer at gmail.com> wrote:
> Good luck. I tried for a while to get pandoc compiled on CentOS 5
> (which I guess is basically the same), and I never got it to work.
>
> I was able to make it work on Debian 6. The important thing is the
> version of libc that you have on the system. If you want to try it,
> you can get it with conda. Install Miniconda
> (http://conda.pydata.org/miniconda.html) if you don't already have it,
> and
>
> conda install -c asmeurer pandoc
>
> And try running pandoc. If you get errors about the glibc version,
> then it's too old.
>
> It would be great if nbconvert could completely remove its dependency
> on pandoc. It's one of the hardest things to get compiled for a "full
> ipython stack" (arguably harder than qt even, at least if you care
> about RHEL 5 flavors of Linux).
>
> Aaron Meurer
>
> On Fri, Jun 27, 2014 at 5:01 PM, Jon Wilson <jsw at fnal.gov> wrote:
>> Hi,
>> Because the scientific computing world moves quite slowly, many of the
>> systems I use are still running Scientific Linux 5 (based on RHEL 5).
>> This presents a problem for nbconvert, with its Haskell dependency.
>>
>> Most of these systems don't have any Haskell packages installed, and I
>> don't have root in order to install them myself.  On one system (not
>> SL5), I was able to download the GHC binary and haskell-platform and
>> install them in userland.  But SL5 is still using GCC 2.5, so this
>> approach fails.
>>
>> Is there any alternative to pandoc, or has anyone else encountered this
>> problem and found a workaround?  I don't think that I have the time to
>> set up GHC 6.8 and bootstrap my way up to 7.6.3+ as suggested here:
>>
>> http://stackoverflow.com/questions/8426472/compiling-ghc-7-2-on-linux-with-libc-version-2-7
>>
>> In the meantime, my workaround is to copy notebooks from the SL5 machine
>> to the Fedora machine with working pandoc, nbconvert them there, and
>> copy them back.  It's quite painful compared to the more usual method.
>>
>> Suggestions?
>> Regards,
>> Jon
>> _______________________________________________
>> IPython-dev mailing list
>> IPython-dev at scipy.org
>> http://mail.scipy.org/mailman/listinfo/ipython-dev