/lib64/libc.so.6: version `GLIBC_2.14' not found

[Matt]

Hi,

I'm trying to install Iris using the Anaconda Python distribution on Scientific Linux 6.6. I've tried both the Conda install and the installation from source. Either way, I end up with the following error message:

error: /lib64/libc.so.6: version `GLIBC_2.14' not found (required by /opt/anaconda/lib/python2.7/lib-dynload/../../libudunits2.so.0)

When I do "strings /lib64/libc.so.6 | grep GLIBC" I see:

GLIBC_2.2.5

GLIBC_2.2.6

GLIBC_2.3

GLIBC_2.3.2

GLIBC_2.3.3

GLIBC_2.3.4

GLIBC_2.4

GLIBC_2.5

GLIBC_2.6

GLIBC_2.7

GLIBC_2.8

GLIBC_2.9

GLIBC_2.10

GLIBC_2.11

GLIBC_2.12

GLIBC_PRIVATE

So it looks like I don't have a high enough version of GLIBC on my system. Is there a way to compile Iris against an older version?

Thanks in advance!

Matt

[Matt]

Hi Guys,

Does anyone have a suggestion for this? It'd be great to get this up and running!

Cheers,

Matt

[Andrew Dawson]

Hi Matt

See this issue on GitHub: https://github.com/SciTools/iris/issues/1543

The packages for conda were built using a higher version of GLIBC then you have on your system. Unfortunately this means you won't be able to use the conda packages on SL6.6 at this time.

The real solution is to have the official packages built on a machine with an older version of GLIBC, RHEL6 or equivalent. Since this hasn't been done yet you can either build the problem packages yourself, or you could try building them as conda packages using the SciTools conda recipes: https://github.com/SciTools/conda-recipes-scitools.

Also note that it is not actually the Iris package that has this problem, but some of the built dependencies, in the case of your specific error is from udunits.

Sorry I can't be more help right now.

[Will Seviour]

Hi Matt

I've just had a very similar problem trying to get iris installed on an old system with SL5. As Andrew said, the issue is that the conda binaries were compiled with a newer version of glibc/glibcxx. It would be nice to have the official packages built with older libraries, but in the meantime, here's roughly what I did to get things working:

- Install iris from using conda.

- Download and compile proj.4 (http://trac.osgeo.org/proj/) and copy executables in place of those from the conda install (in anaconda/lib).

- Uninstall cartopy (conda uninstall cartopy) and re-install from source (on github). This is needed because of the _trace.cpp file in cartopy. 

- Copy libexpat.so from /usr/lib64 to anaconda/lib and make symbolic link to libexpat.so.1 (this might have been a problem specific to my system).

- Again for udunits (ftp://ftp.unidata.ucar.edu/pub/udunits/), download source, compile, make and copy executables to anaconda/lib.

I hope that makes some sense!

Will 

[Matt]

Hi Will and Andrew,

Thank you very much for your replies! It all seems to work on my system now.

For completeness, I also had to do the following to get Will's final step to work:

- yum install expat-devel (otherwise udunits2 builds fine but complains about an XML parser when it is called)

- build udunits-2.1.24 from source (Iris doesn't seem to like earlier versions, and I couldn't get the later version to build easily on my system for some reason)

Cheers,

Matt