Closed astrophys closed 1 year ago
I @astrophys ,
Not an author of CNVkit, but I think 2nd "atlas-related" error you got is not directly related to CNVkit itself => But rather your set up, or your PATH, or your Python installation or others => A bit hard to solve from here (as far as I am concerned, maybe someone could)
I advice you to get back to conda
way of installing CNVkit (should be simpler and more predictable)
=> Error 1st you got looks like you tried to install CNVkit inside base
env on your conda install ? Like conda install cnvkit
?
=> Generally speaking, this is something to be avoided with conda
(can cause conflicts between versions = "were found to be incompatible with the existing python installation in your environment")
Better to create a dedicated env: conda create -n cnvkit
Then install CNVkit package within this env: conda install -n cnvkit -c bioconda cnvkit
=> Tell me if this worked for you ?
Hope this helped. Have a nice day. Felix.
Looks like CentOS 8 is past end of life, so dependency resolution will get trickier as time goes on.
Another option you can try besides conda is the prebuilt Docker image: https://hub.docker.com/r/etal/cnvkit/
For HPC usage, you may also find conda or podman can avoid problems in the global environment.
I am trying to install this software on CentOS 8.1 using pip from within a virtual environment. I had previously tried installing it using
conda
but after 3h, it failed withfollowed by about 10k lines of output. So I tried something similar. I created a virtual environment, activated the environment and loaded some modules. We use lmod modules (which are written in lua) to manage software. The module files basically modify the user's environment (e.g.
PATH
andLD_LIBRARY_PATH
) to find the executables and runtime libraries. So I ran :Amongst the many errors, I see
However I have those libraries installed via
yum
and they are in the/usr/lib64/atlas
directory :Also, I have my lapack libraries within
LD_LIBRARY_PATH
, i.e.Questions:
/usr/lib64/atlas
) and the lapack libraries are in myLD_LIBRARY_PATH
pip
installation. I've done several installs ofscipy
via virtual environments withpip
and I did not have to install alapack
system library. Why must I install the system level libraries in this case when in previous instances I did not? Perhaps it is a difference between compiling from source vs. downloading precompiled executables?Thank you for your time.