R packages error - unable to load shared object

Question

[Francis]

I use the conda update the R language and then I require packages but there is a error.

> unable to load shared object
> '/home/slee_pkuhpc/lustre1/test/software/miniconda2/lib/R/modules//internet.so':
> libssl.so.1.0.0: cannot open shared object file: No such file or
> directory

I Googled the question but I did not find a fix. I do not have the root authority and do not change something.

So any advice is helpful.

Thanks.

[Fabio Marroni]

HI Francis,

I think that the category tool is for announcing tools that you are making available, not for asking questions regarding tools (everybody is asking questions regarding a bioinformatics tool, more or less!)

Not having root permission complicates things. Looks like you have some problems with ssl libraries. Try to give a looke here: https://askubuntu.com/questions/339364/libssl-so-10-cannot-open-shared-object-file-no-such-file-or-directory. One idea could be to also include the ssl libraries in the conda environment you are building for R, but I am not very expert on this.

Another option could be to see if you manage to install R without root permission: try to see if these posts help:

https://unix.stackexchange.com/questions/149451/install-r-in-my-own-directory

https://community.rstudio.com/t/installing-other-version-of-r-without-root-access-and-working-on-rstudio-server/5275/5

[Fabio Marroni]

Ok, Devon already solved the problem... Thanks Devon, I might need this trick!!!

[Devon Ryan]

This is biting everyone at the moment, unfortunately :(

[Francis]

I am sorry about the category and thanks for your advice. I learn something from this links, thank you.

Answer 1

[Devon Ryan]

conda install openssl==1.0.2p

What you're observing happens when you use conda install inside an environment that has packages with pinned dependencies for which an updated version of the dependency is available. Since conda install seemingly ignores pinnings, it gleefully updates everything in the environment thereby randomly breaking things. The best practices are one of the following:

• Use conda create to make a new environment for software packages/combinations
• Use --no-update-deps so conda doesn't try to update everything if you MUST use conda install

[Francis]

Thanks for your answer. I am sorry that the reply is late.

Yes, it`s a good ideal to creat a new environment for analysis. I uninstalled miniconda and reinstalled again, it works now.

Thanks!