Please ensure you are following the instructions in https://harvardmed.atlassian.net/wiki/spaces/O2/pages/1594263516 when attempting to install your own Conda packages. Below is a list of known packages and situations the O2 user community has troubleshooted with Research Computing and their resolutions (this list will be updated as new instances are encountered).
GCC within Conda
When trying to troubleshoot building something within your conda environment, you may notice that the output of which gcc will still point to /usr/bin/gcc. You may attempt to remedy this by trying something like conda install gcc into your environment, only to discover that it either failed, or you are still unable to find your freshly “installed” GCC version. We illustrate an example solution here.
Firstly, the packages to conda install are actually gcc_linux-64 and gxx_linux-64. You will (probably) need both packages. You can specify versions if you’d like, such as gcc_linux-64=9; make sure that if you do so, you specify the same version for both. Once these install successfully, your environment’s bin directory will look something like this:
Note that these executables are long-form (and there are already a number of softlinks that have been created); there is no actual executable that is just named gcc. What you will need to do next is create (at least) two more softlinks (replace path/to/env with the actual path to your conda environment - by default, it will be in $HOME/.conda/envs, but if you specified --prefix or -p when creating the environment, it may be somewhere else):
In most cases, this should get you far enough. If you encounter build errors that are looking for something like gcc-ar, you will need to set up a similar softlink for that as well. In any case, once you have the above softlinks (at a minimum) created, your conda environment should be able to reliably use your installed gcc installation to build packages instead of inadvertently using the one located at /usr/bin.