sandboxfs is a FUSE-based file system that exposes an arbitrary view of the
host’s file system under the mount point, and offers access controls that
differ from those of the host. You can think of sandboxfs as an advanced
version of bindfs (or
mount --bind or
depending on your system) in which you can combine and nest directories under
an arbitrary layout.
The primary use case for this project is to provide a better file system sandboxing technique for the Bazel build system. The goal here is to run each build action (think compiler invocation) in a sandbox so that its inputs and outputs are tightly controlled, and sandboxfs attempts to do this in a more efficient manner than the current symlinks-based implementation.
But what makes sandboxfs more exciting are its potential secondary use cases outside of Bazel. For example, my personal goal is to use sandboxfs within pkg_comp 2.x to simplify the logic in isolating pkgsrc builds and make the isolation more robust and accurate.
So how did this happen? This summer I had the pleasure of hosting Pallav Agarwal as an intern in the Bazel team working from our Google NYC office. I leveraged my experience last year with sourcachefs to design and lead the implementation of sandboxfs. With this and Pallav’s strong skills, we got to a feature-complete implementation by the beginning of August.
And today, after a bit of extra work to make the full codebase open-sourceable and collecting all necessary approvals, I am happy to announce that sandboxfs is public!
Enjoy and stay tuned for further integrations! sandboxfs is a cool thing on its own, but until it serves as the base for builds within Bazel and pkg_comp, it’s not of much use. Or is it? If you have other use cases, let me know!