e.g., instead of matlab/R2021a we get tinkercliffs-rome/matlab/R2021a. This is confusing in addition to being cumbersome, as any module a user sees on a "tinkercliffs-rome" node is only available to that node type. I think this could be fixed by one of:
- Removing
${sysdir} from the module path associated with --system
- Changing
MODULEPATH on the clusters
The first option is cleanest but would change the module names and so break scripts using the existing structure.