Wouter Deconinck

This issue appears when (because) /usr/local already exists, is a directory, and isn't empty. In that case, spack won't remove it. Exception [IsADirectoryError](https://docs.python.org/3.10/library/exceptions.html#IsADirectoryError) is raised in https://github.com/spack/spack/blob/7c1a3094536730be698809daf2c3df267dd8a91b/lib/spack/llnl/util/filesystem.py#L226 See [os.replace](https://docs.python.org/3.10/library/os.html#os.replace) for...

FYI, https://github.com/spack/spack/pull/41875. We have at least one other case where we use a variant `use_spack_interpreter` to get around a cyclic dependency that way.

FYI @robertu94 @disheng222, for your consideration.

I see. I understand the need to test and benchmark on real world or at least representative data sets. It seems the smallest dataset on sdrbench is still quite large...

Ultimately this comes down to `setup-env.sh` not detecting the shell correctly under qemu in `_spack_determine_shell` (procfs returns the qemu emulator): https://github.com/spack/spack/blob/9bf6e05d027086f8363094aff7a091b25cdc8a87/share/spack/setup-env.sh#L231-L246 Then later it fails to `export -f _spack_shell_wrapper`, in...

For qemu in docker, the following change to `_spack_determine_shell` does the job: ``` # Determine which shell is being used _spack_determine_shell() { if [ -f "/proc/$$/exe" ]; then # If...

Rebased and conflicts resolved (file removed).

> I would push for the new workflow be an extra step inside of the existing style checks rather than its own standalone CI. Suggestion: add to https://github.com/spack/spack/blob/develop/.github/workflows/valid-style.yml after `validate`,...

Maintainers: @albestro @eschnett @msimberg @teonnik

Maybe @sly2j has some ideas of how this should be approached.