Proposal: reduce standard padding to 128

255 is Too Large

Currently we have the Fermi-standard padding_depth set to 255. This has turned out to be problematic, to wit, a full grand-union build of dunesw hits a command-line-too long error when gcc invokes cc1, due to the summed -I and -L arguments passed to the compiler first stage. So I claim this demonstrates 255 is too long.

What is the actual minimum?

Looking at the longest paths we currently install packages in:

Length  Example
    61  /cvmfs/fermilab.opensciencegrid.org/packages/common/opt/spack
    59  /cvmfs/larsoft.opensciencegrid.org/spack-packages/opt/spack
    58  /exp/experiment/app/users/long_user_name/spack_1/opt/spack
    42  /afs/cern.ch/asis/spack_packages/opt/spack (theoretical)

We see that somewhere around 64 would be an actual minimum "reasonable" value

Proposal

Since the minimum we could get away with is around 64, I propose we set the Fermi-standard padding_depth (as setup by make_spack, etc.) to double that, or 128. This would cut our argument-list-size for compiling in large environments (like dunesw) by something like a third, and still maintain the relocation flexibility we want.