Marvin Drees

Update: I'll tackle this PR again soon, finally got a little bit more time at my disposal

I'm currently updating the whole PR with a few better ideas around CI. I would keep running all tests on both CircleCI and GitHub actions for now as they would...

I added the CodeCov Orb to our CircleCI project as the bash uploader has been deprecated quite some time ago in favor of the orb. I'll make all the actions...

Will continue on this one this week and see if some of those failures are actually legit test failures that weren't caught before.

Note: This darft will only stay open as a reference to my follow-up PRs that implement an easier path to switching CIs. Those PRs will also be smaller and more...

> @MDr164 did you got a change to look at this? We added https://github.com/u-root/u-root/blob/main/cmds/exp/efivarfs/efivarfs.go quite some time ago which uses the efivarfs, maybe that does what you're looking for?

That's odd, I guess I'll have to check if the code regressed because we have been using it in a different project and that one still worked fine last time...

> Unfortunately requiring CGO is not ideal, indeed. Looks like it can also be used in pure-go mode but is then not as powerful as it uses psx from libcap....

I rebased the PR on master and added a build constraint so that the sandboxing will throw an error when attempted to be used outside of Linux. I'm not sure...

Any chance this could get a review (be it positive or negative)? This change allows Task to obtain the same amount of sandboxing as build systems like Bazel while maintaining...