Massimiliano Culpo

I have a high-level question. Failed pipeline jobs let you run the build in a docker image to inspect the failure. Within that Docker image there's an `install.sh` script which...

> We used to split it in two, but it complicated the ci module even more, as we had to make a copy of the spack.yaml before the first install,...

I also started seeing this sometimes ago, but didn't have time to look into it yet. I suspect it's the same as #31265 nowadays (though the failure for bootstrapping `isort`...

@spackbot run pipeline

A few notes for later reference: - @becker33 pointed out in a conversation that "_with the new error message system, we could actually reimplement the logic from the old concretizer...

I'll go ahead. @becker33 said he was fine with it on Slack, and if there are reasons to revisit this decision the PR is simple enough to revert later without...

Here's a system where it works (Ubuntu 20.04): ```console $ ls /home/culpo/.spack/bootstrap/store/linux-ubuntu20.04-x86_64/gcc-9.4.0/py-isort-5.9.3-5ycfo3cc5sjmb3occ4quqvfymjcvm564 bin lib ``` and here's Fedora where this fails: ```console [root@166c7158c565 spack]# ls /root/.spack/bootstrap/store/linux-fedora36-x86_64/gcc-12.1.1/py-isort-5.9.3-k2i4if2s27nwicednjvpsejds5utaqmg/ local [root@166c7158c565 spack]# ls...

We have to make this line: https://github.com/spack/spack/blob/c533612ab6e4d4b4c46d3b827fae3687302cd22a/lib/spack/spack/bootstrap.py#L207 more resilient to changes in the directory where executables are placed, since they might change across platforms. I can confirm that hacking that...

I think this PR started making Windows unit tests fail. Apparently the CI job was still returning 0 so everything is :heavy_check_mark:

For reference, here's a link to the latest tests run on this PR: - https://github.com/spack/spack/actions/runs/3136510560/jobs/5093623993